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 mathematical regression expression in
230 23|0 220 210 13.5 0.09 0.09 230
220 22|0 230 230 15.5 0.10 0.19 201
021 021 021 021 15.5 0.10 0.29 211
201 201 201 201 15.5 0.10 0.39 132
301 3|01 321 321 (became infeasible) 301
221
2|21
201 101 12.5 0.08 0.47 132
301 30|1 301 301 16.5 0.11 0.58 301
211 21|1 211 211 21.0 0.14 0.72 021
132 132 132 132* 26.5 0.19 0.19 230
310 310 310 210 13.5 0.09 1.00 132
f
i
r
s
t

i
t
e
r
a
t
i
o
n
total fitness 150.0
t
h
i
r
d

i
t
e
r
a
t
i
o
n
s
e
c
o
n
d

i
t
e
r
a
t
i
o
n
p
o
p
u
l
a
t
i
o
n
n
e
w
p
o
p
.
c
u
m
.

p
r
o
b
.
s
e
l
.

p
r
o
b
.
f
i
t
n
e
s
s
m
u
t
a
t
i
o
n
c
r
o
s
s
o
v
e
r
230 230 230 16.0 0.08 0.08 221
201 201 201 15.5 0.08 0.16 132
211 21|1 212 27.5 0.14 0.30 230
132 13|2 131 20.5 0.10 0.40 212
301 301 301 16.5 0.08 0.48 201
132 132 132 26.5 0.14 0.62 132
301 3|01 001 7.5 0.04 0.66 301
021 0|21 321 23.5 0.12 0.78 221
230 230 230 15.5 0.08 0.86 212
132 132 132
220
201
212*
121
301
132
001
221
230
132 26.5 0.14 1.00 001
total fitness 195.5
221 22|1 222 222* 30.0 0.15 0.15 221
132 13|2 131 121 20.5 0.10 0.25 222
230 230 230 230 15.5 0.07 0.32 230
212 212 212 112 24.5 0.12 0.44 202
201 20|1 202 202 22.0 0.09 0.53 131
132 13|2 131 131 20.0 0.11 0.64 222
301 301 301 201 15.5 0.08 0.72 012
221 221 221 221 23.5 0.11 0.83 121
212 2|12 201 201 15.5
0.08 0.91 202
001 0|01 012 012 19.5 0.09 1.00 121
total fitness 206.5
E
0
2
0
8
2
0
e
Table 6.3. Several iterations
for solving the allocation
problem using genetic
algorithms.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 160
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 161
symbolic form that best produces the observed output
given the associated input. To perform this task GP uses
a physical symbol system divided into two sets. The first
set, the so-called terminal set, contains the symbols for
independent variables as well as parameter constants as
appropriate. The content of this set is determined only by
the nature of the problem to be solved. All the basic
operators used to form a function f() are in the second
set, called the functional set. For example, the second
set can contain the arithmetic operators (, , *, /) and
perhaps others such as log, sqrt, and sin as well,
depending on the perceived degree of complexity of the
regression.
As models have a well-defined grammar and are of
variable size and shape, the structures undergoing adap-
tation in genetic programming are often taken from a class
of parse trees. A parse tree is a node-link tree whose nodes
are procedures, functions, variables and constants. The
sub-trees of a node of a parse tree represent the arguments
of the procedure or function of that node. Variables and
functions requiring no arguments are called terminals.
They are the leaves in the parse tree. Functions that take
arguments are branches of a parse tree and are called
functions. All terminals in one parse tree compose the
terminal set, and all functions in that tree are in the
functional set.
One example of a parse tree for the expression
a (b/c) can be depicted as shown in Figure 6.11.
In Figure 6.11 the variables a, b and c are leaves in
the parse tree and hence belong to the terminal set. The
mathematical operations and / are functions having
two arguments and, hence, are members of the
functional set.
The advantage of encoding the induced GP model as
a parse tree is that mutations can readily be performed
by changing a nodes arguments or operator function.
Crossover can be achieved by replacing one or more
nodes from one individual with those from another with-
out introducing illegitimate syntax changes.
To produce new expressions (individuals) GP requires
that two parent expressions from the previous generation
be divided and recombined into two offspring expres-
sions. An example of this is illustrated in Figure 6.12.
Having two parent expressions presented as parse trees,
the crossover operation simply exchanges a branch of one
parent with a branch of the other.
The result of mutations and crossover operations is
the production of two new individuals or children
(Figure 6.12 lower). Each child inherits some character-
istics from its parents.
This process continues until the fitness of the entire
population increases and converges to find the near
optimal solution set. The benefit of using the parse tree
vehicle in GP can be seen in the crossover operation. This
presentation guarantees that every resulting expression is
grammatically and semantically correct, regardless of the
crossover or mutation site, assuming that it was
performed on the correct parents.
Mutation in GP corresponds to a random alteration of
the individual parse tree at the branch or node level.
There are several possible types of computational muta-
tions. Some examples are:
Branch-mutation. A complete sub-tree is replaced with
another, possibly arbitrary sub-tree.
Node-mutation. The value of a single node in the tree is
replaced by another, again in an arbitrary way.
Constant-mutation. Similar to node mutation, but here
the constant is selected and mutated using white noise.
Inversion-mutation. Inverts the order of operands in an
expression in a random way.
Edit-mutation. Edits the parse tree based on semantic
equivalence. This mutation does not alter the function
produced but just makes it shorter for example
elements like (x 0) will be replaced everywhere in
the expression with x only.
Mutation can affect both the parse tree structure and its
information content. With mutation, therefore, the search
explores a new domain. It also serves to free the search
+
a /
b
c
E
0
2
0
7
2
3
m
Figure 6.11. A parse tree of the expression a (b/c).
wrm_ch06.qxd 8/31/2005 11:53 AM Page 161
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
from the possibility of being trapped in local optima.
Mutation can be destructive, causing rapid degradation of
relatively fit solution sets if the probability of mutation in
the algorithm is set too high.
As in most evolutionary algorithms, the models that
produce the best fit to the data have the greatest oppor-
tunity to become parents and produce children, which
is called reproduction. The better models produce the
smallest errors, or differences between the calculated
output and the observed output.
The basic GP procedure is as follows:
1. Generate the initial population (of size N) of random
models (algebraic expressions). Several different
initialization procedures exist for GP. Details for
these methods can be seen in Babovic and Keijzer
(2000).
2. Evaluate each sample model in the population and
assign it a fitness value according to how well it solves
the problem.
3. Create a new population of size N of models by apply-
ing the following two primary operations:
a) copy the best M (M N) fitted individuals elite
to the new population (asexual reproduction);
162 Water Resources Systems Planning and Management
b) create new L (L N M) models by genetically
recombining randomly chosen parts of two existing
models (sexual reproduction).
4. Randomly apply a genetic operator mutation to the
offspring and determine the fitness of each model.
5. Repeat steps 3 and 4 until a predetermined stopping
criterion is reached.
The stopping criterion in Step 5 of the GP procedure is
usually that either (a) a certain number of generations has
been produced, or (b) a certain amount of wall-clock time
has passed.
Software programs have been written to implement
GP. For example GPKernel developed by Babovic and
Keijzer (2000) at the Danish Hydraulic Institute (DHI)
has been used in applications such as: rainfallrunoff
modelling (Babovic and Abbott, 1997; Drecourt, 1999;
Liong et al. 2000), sediment transport modelling, salt
intrusion in estuaries and roughness estimation for a
flow over a vegetation bed (Babovic and Abbott, 1997).
More details about GPKernel can be seen in Aguilera
(2000).
The challenge in applying genetic programming for
model development is not only getting a close fit between
parent 1 [(a+b/c)] parent 2 [(d-e)*(f+(g/h))]
child 1 [a+((d-e)/c)] child 2 [b*(f+(g ))]
h
+
a /
b c
*
-
+
d e
g
/ f
h
+
a /
c -
d e
*
b +
f
g
h
/
<
E
0
2
0
7
2
3
n
Figure 6.12. An illustration of a crossover
operation and mutation ( / *) operation for
genetic programming.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 162
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 163
observed and predicted outputs, given a set of input data,
but also of interpreting the model that is generated to
obtain additional understanding of the actual processes
taking place. This data mining is discussed in the next
section. There are also potential problems in creating a
dimensionally correct model if the input data are not
dimensionless. As a consequence, many applications using
GP seem to require some guidance based on a mix of both
physically based and data-based approaches.
5. Data Mining
Data mining is the process of finding new and potentially
useful knowledge from data usually data from large
databases. Data mining is also known as knowledge discov-
ery (KD). It is the non-trivial extraction of implicit, previ-
ously unknown, information from data. It uses machine
learning and statistical and visualization techniques to dis-
cover and present knowledge in a comprehensible form.
Data mining is a relatively new technology that has the
potential to help us learn more from our data. Data min-
ing tools can scour databases for hidden patterns, finding
predictive information that experts may miss because it
lies outside their expectations.
Data mining techniques can be implemented on exist-
ing software and hardware platforms to enhance the value
of existing information resources, and can be integrated
with new products and systems as they are brought
online. When implemented on high performance client/
server or parallel processing computers, data mining tools
can analyse massive databases to seek answers to
questions.
5.1. Data Mining Methods
Data mining is supported by three technologies (Babovic,
1998):
monitoring and data collection
multiprocessor computers
data mining algorithms
The core components of data mining technology have
been under development for decades in research areas
such as statistics, artificial intelligence and machine learn-
ing. The current level of development of these techniques,
coupled with high-performance relational database
engines and broad data-integration efforts, make these
technologies practical for current data environments.
Dynamic data access is critical for data navigation
applications, and the ability to store, manage and analyse
large databases is critical to data mining. Given databases
of sufficient size and quality, data mining technology can
provide:
Automated prediction of trends and behaviours. Data min-
ing automates the process of finding predictive infor-
mation in large databases. Questions that traditionally
required extensive hands-on analysis can now be
answered directly from the data and much more
quickly. A typical example of a predictive problem is
identifying segments of an ecosystem likely to respond
similarly to given events.
Automated discovery of previously unknown patterns.
Data mining tools sweep through databases and iden-
tify previously hidden patterns. When data mining
tools are implemented on high performance parallel
processing systems, they can analyse large databases
very quickly often in minutes. Faster processing
means that users can automatically experiment with
more models to understand complex data. Larger data-
bases, in turn, yield improved predictions.
Some of the techniques commonly used in data mining
are:
Artificial neural networks, which are nonlinear predic-
tive models that learn through training and resemble
biological neural networks in structure, as discussed
earlier.
Classification and regression trees (CARTs), which is
used for classification of a dataset. It provides a set of
rules that one can apply to a new (unclassified) dataset
to predict which records will have a given outcome. It
segments a dataset by creating two-way splits.
Chi square automatic interaction detection (CHAID),
which is a decision tree technique used for classifica-
tion of a dataset. It provides a set of rules that one can
apply to a new (unclassified) dataset to predict which
records will have a given outcome. It segments a
dataset by using chi square tests to create multi-way
splits. It requires more data preparation than CART.
Classification, which is the process of dividing a dataset
into mutually exclusive groups, such that the members of
wrm_ch06.qxd 8/31/2005 11:53 AM Page 163
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
each group are as close as possible to one another,
and different groups are as far as possible from one
another, where distance is measured with respect to
specific variable(s) one is trying to predict.
Clustering, which is similar to classification, except
that distance is measured with respect to all available
variables.
Data navigation, which is the process of viewing differ-
ent dimensions, slices and levels of detail of a multidi-
mensional database. See OLAP, defined below.
Data visualization, which is the visual interpretation of
complex relationships in multidimensional data.
Decision trees, which are tree-shaped structures that
represent sets of decisions. These decisions generate
rules for the classification of a dataset. Specific deci-
sion tree methods include classification and regression
trees (CARTs) and chi square automatic interaction
detection (CHAID).
Genetic algorithms and genetic programming, which search
techniques that use processes such as genetic combina-
tion, mutation and natural selection in a design based
on the concepts of evolution, as discussed earlier.
Nearest-neighbour method, which is a technique that
classifies each record in a dataset on the basis of a com-
bination of the classes of the k record(s) most similar
to it in a historical dataset. It is sometimes called the
k-nearest-neighbour technique.
Online analytical processing (OLAP), which refers to
array-oriented database applications that allow users
to view, navigate through, manipulate and analyse
multidimensional databases.
Rule induction, which is the method of extraction of useful
ifthen rules from data based on statistical significance.
Many of these technologies have been in use for more
than a decade in specialized analysis tools that work with
relatively small volumes of data. These capabilities are
now evolving to integrate directly with standard databases
and online application program platforms.
6. Conclusions
Most computer-based models used for water resources
planning and management are physical, mechanistic or
process-based models. Builders of such models attempt to
incorporate the important physical, biological, chemical,
164 Water Resources Systems Planning and Management
geomorphological, hydrological and other types of inter-
actions among all system components, as appropriate for
the problem being addressed and system being modelled.
This is done in order to be able to predict possible
economic, ecologic, environmental or social impacts
that might result from the implementation of some plan
or policy. These types of models almost always contain
parameters. These need values, and the values of the
parameters affect the accuracy of the impact predictions.
This chapter has introduced some data-based methods
of modelling. These have included two evolutionary
search approaches: genetic algorithms (GA) for estimating
the parameter values, and genetic programming for
finding models that replicate the real system. In some
situations, these biologically motivated search methods,
which are independent of the particular model being
calibrated, provide the most practical way for model
parameter calibrations to be accomplished. These same
search methods, sometimes coupled to physically based
simulation models, can be used to obtain the values of the
unknown decision-variables as well.
While physically based or process-based models are
appealing for those who wish to better understand these
natural processes, they clearly remain approximations of
reality. In some water resources problem applications, the
complexities of the real system are considerably greater
than the complexities of the models built to simulate them.
Hence, it should not be surprising that in some cases sta-
tistical or data-based models, which convert input variable
values to output variable values in ways that have nothing
to do with what happens in reality, may produce more
accurate results than physically-based models. This chap-
ter has briefly introduced two such types of data-based
model: genetic programming (GP) models and artificial
neural networks (ANN). When they work, they often pro-
duce results faster than their physical counterparts and as
accurately or more so, but only within the range of values
observed in the data used to build these models.
Data-driven modelling methods are increasingly
being developed and used to gain information from
data. They are especially useful when the data sets are
large, and where it becomes impractical for any human
to sort through it to obtain further insights into the
processes that produced the data. Many data-rich prob-
lems can be solved by using novel data-driven modelling
together with other techniques. Data mining methods are
wrm_ch06.qxd 8/31/2005 11:53 AM Page 164
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 165
also being increasingly used to gain greater understanding
and knowledge from large data sets. Some approaches to
data mining are listed in this chapter. Water resources
modellers are unlikely to be involved in the development
of such data mining techniques, so fortunately, as is the
case with GA, GP and ANN methods, many are available
on the Internet. Applications of such methods to ground-
water modelling, sedimentation processes along coasts
and in harbours, rainfall runoff prediction, reservoir
operation, data classification, and predicting surge water
levels for navigation represent only a small sample of what
can be found in the current literature. Some of this
literature is cited in the next section.
7. References
ABRAHART, R.J.; KNEALE, P.E. and SEE, L.M. (eds.).
2004. Neural networks for hydrological modeling. Leiden,
the Netherlands, AA Balkema.
AGUILERA, D.R. 2000. Genetic Programming with GPKER-
NEL. Unpublished, DHI, Copenhagen.
BABOVIC, V. 1996. Emergence, evolution, intelligence:
hydroinformatics.
BABOVIC, V. 1998. A data mining approach to time
series modelling and forecasting. In: V. Babovic and
L.C. Larsen (eds.), Proceedings of the Third International
Conference on Hydroinformatics (Hydroinformatics 98),
Rotterdam, Balkema, pp. 84756.
BABOVIC, V. and ABBOTT, M.B. 1997. The evolution of
equations from hydraulic data. Part II: applications.
Journal of Hydraulic Research, Vol. 35, No. 3, pp. 41127.
BABOVIC, V. and BOJKOV, V.H. 2001. D2K technical
report D2K TR 0401-1. Copenhagen, DHI, April.
BABOVIC, V. and KEIJZER, M. 2000. Genetic program-
ming as a model induction engine. Journal of
Hydroinformatics, Vol. 2, No. 1, pp. 35 60.
BANZHAF, W.; NORDIN, P.; KELLER, R.E. and
FRANNKONE, F.D. 1998. Genetic programming: an intro-
duction: on the automatic evolution of computer programs and
its applications. San Francisco, Calif., Morgan Kaufmann
Publishers, Inc. and dpunkt Verlag fur Digitale
Technologie GmbH.
BEALE, R. and JACKSON, T. 1990. Neural computing: an
introduction. Bristol, IOP.
CLAIR, T.A. and EHRMAN, J.M. 1998. Using neural net-
works to assess the influence of changing seasonal cli-
mates in modifying discharge, dissolved organic carbon,
and nitrogen export in eastern Canadian rivers. Water
Resources Research, Vol. 34, No. 3, pp. 44755.
CYBENKO, G. 1989. Approximation by superpositions of
a sigmoidal function. Mathematics of Control, Signals, and
Systems, Vol. 2, pp. 30314.
DRECOURT, J.-P. 1999. Application of neural networks
and genetic programming to rainfallrunoff modelling, D2K
Technical Report 0699-1. Copenhagen, DHI.
GOLDBERG, D.E. 1989. Genetic algorithms in search, opti-
mization, and machine learning. Reading, Addison-Westley.
HAGAN, M.T.; DEMUTH, H.B. and BEALE, M. 1996.
Neural network design. Boston, Mass., PWS.
HAYKIN, S. 1999. Neural networks: a comprehensive foun-
dation, 2nd edn. Upper Saddle River, N.J., Prentice-Hall.
HERTZ, J.; KROGH, A. and PALMER, R.G. 1991.
Introduction to the theory of neural computation. Reading,
Mass., Addison Wesley.
HORNIK, K.; STINCHCOMBE, M. and WHITE, H. 1989.
Multilayer feedforward networks are universal approxi-
mators. Neural Networks, No. 2, pp. 35966.
HSIEH, W.W. and TANG, B. 1998. Applying neural
network models to prediction and data analysis in
meteorology and oceanography. Bulletin of the American
Meteorological Society, Vol. 79, pp. 185570.
KOZA, J.R. 1992. Genetic programming: on the program-
ming of computers by means of natural selection (complex
adaptive systems). Cambridge, Mass., MIT Press.
KOZA, J.R.; BENNETT, F.H. III; ANDRE, D. and
KEANE, M.A. 1999. Genetic programming III: Darwinian
invention and problem solving. San Francisco, Calif.,
Morgan Kaufmann.
LANGE, N.T.G. 1998. Advantages of unit hydrograph
derivation by neural networks. In: V. Babovic and
L.C. Larsen (eds.), Hydroinformatics 98, Rotterdam,
Balkema, pp. 78389.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 165
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
LIONG, S.-Y.; KHU, S.T.; BABOVIC, V.; HAVNOE, K.;
CHAN, W.T. and PHOON, K.K. 2000. Construction of
non-linear models in water resources with evolutionary com-
putation. Singapore, National University of Singapore.
MINNS, A.W. 1996. Extended rainfallrunoff modelling
using artificial neural networks. In: A. Mller (ed.),
Hydroinformatics 96, Rotterdam, Balkema, pp. 20713.
MINNS, A.W. and HALL, M.J. 1996. Artificial neural
networks as rainfallrunoff models. Hydrological Sciences
Journal, Vol. 41, No. 3, pp. 399417.
MITCHELL, T.M. 1997. Machine learning. New York,
McGraw-Hill.
PRICE, R.K.; SAMEDOV, J.N. and SOLOMATINE, D.P.
1998. An artificial neural network model of a generalised
channel network. In: V. Babovic and L.C. Larsen (eds.),
Hydroinformatics 98, Rotterdam, Balkema, pp. 81318.
PRINCIPE, J.C.; EULIANO, N.R. and LEFEBVRE, W.C.
2000. Neural and adaptive systems: fundamentals through
simulation. New York, John Wiley.
PROAO, C.O.; VERWEY, A.; VAN DEN BOOGAARD,
H.F.P. and MINNS, A.W. 1998. Emulation of a sewerage
system computational model for the statistical processing
of large number of simulations. In: V. Babovic and
L.C. Larsen (eds.), Hydroinformatics 98, Vol. 2,
Rotterdam, Balkema, pp. 114552.
RECKNAGEL, F.; FRENCH, M.; HARKONEN P. and
YANUNAKA, K.I. 1997. Artificial neural network
approach for modelling and prediction of algal blooms.
Ecological Modelling, No. 96, pp. 1128.
SANCHEZ, L.; ARROYO, V.; GARCIA, J.; KOEV, K. and
REVILLA, J. 1998. Use of neural networks in design of
coastal sewage systems. Journal of Hydraulic Engineering,
Vol. 124, No. 5, pp. 45764.
SCARDI, M. 1996. Artificial neural networks as empirical
models for estimating phytoplankton production. Marine
Ecology Progress Series, No. 139, pp. 28999.
SEE, L. and OPENSHAW, S. 1999. Applying soft com-
puting approaches to river level forecasting. Hydrological
Sciences Journal, Vol. 44, No. 5, pp. 76378.
166 Water Resources Systems Planning and Management
SHEN, Y.; SOLOMATINE, D.P. and VAN DEN
BOOGAARD, H.F.P. 1998. Improving performance of
chlorophyl concentration time series simulation with
artificial neural networks. Annual Journal of Hydraulic
Engineering, JSCE, No. 42, pp. 751 6.
SMITH, M. 1993. Neural networks for statistical modelling.
New York, Van Nostrand Reinhold.
STELLING, G.S. 2000. A numerical method for inunda-
tion simulations. In: Y.N. Yoon, B.H. Jun, B.H. Seoh and
G.W. Choi (eds.), Proceedings of the 4th International
Conference on Hydro-Science and Engineering, Seoul,
2626 September 2000. Seoul, Korea Water Resources
Association.
VAN DEN BOOGAARD, H.F.P.; GAUTAM, D.K. and
MYNETT, A.E. 1998. Auto-regressive neural networks
for the modelling of time series. In: V. Babovic and
L.C. Larsen (eds.), Hydroinformatics 98, Rotterdam,
Balkema, pp. 7418.
VAN DEN BOOGAARD, H.F.P.; TEN BRUMMELHUIS,
P.G.J. and MYNETT, A.E. 2000. On-line data assimilation
in auto-regressive neural networks. Proceedings of the
Hydroinformatics 2000 Conference, The University of
Engineering, Iowa, USA, July 23 27, 2000. Iowa City,
University of Iowa.
VAN GENT, M.R.A. and VAN DEN BOOGAARD, H.F.P.
1998. Neural network modelling of forces on vertical
structures. In: Proceedings of the 27th International
Conference on Coastal Engineering, pp. 2096109. Reston,
Va., ASCE press.
WEN, C.-G. and LEE, C.-S. 1998. A neural network
approach to multiobjective optimization for water quality
management in a river basin. Water Resources Research,
Vol. 34, No. 3, pp. 42736.
WST, J.C. 1995. Current prediction for shipping
guidance. In: B. Kappen and S. Gielen (eds.). Neural
networks: artificial intelligence and industrial applications.
Proceedings of the 3rd Annual SNN Symposium
on Neural Networks, Nijmegen, The Netherlands,
14 15 September 1995. London: Springer-Verlag,
pp. 36673.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 166
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Data-Based Models 167
Additional References (Further Reading)
MCKINNEY, D.C. and LIN, M.D. 1994. Genetic algo-
rithm solution of groundwater management models.
Water Resources Research, Vol. 30, No. 6, pp. 1897906.
OLIVERA, R. and LOUCKS, D.P. 1997. Operating rules
for multi-reservoir operation. Water Resources Research,
Vol. 33, No. 4, pp. 83952.
SOLOMATINE, D.P. and AVILA TORRES, L.A. 1996.
Neural network application of a hydrodynamic model
in optimizing reservoir operation. In: A. Mller (ed.),
Hydroinformatics 96, Rotterdam, Balkema, pp. 20106.
wrm_ch06.qxd 8/31/2005 11:53 AM Page 167
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
7. Concepts in Probability, Statistics and Stochastic Modelling
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
wrm_ch07.qxd 8/31/2005 12:18 PM Page 168
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Uncertainty is always present when planning, developing,
managing and operating water resources systems. It arises
because many factors that affect the performance of water
resources systems are not and cannot be known with
certainty when a system is planned, designed, built,
managed and operated. The success and performance of
each component of a system often depends on future
meteorological, demographic, economic, social, technical,
and political conditions, all of which may influence
future benefits, costs, environmental impacts, and social
acceptability. Uncertainty also arises due to the stochastic
nature of meteorological processes such as evaporation,
rainfall and temperature. Similarly, future populations of
towns and cities, per capita water-usage rates, irrigation
patterns and priorities for water uses, all of which affect
water demand, are never known with certainty.
There are many ways to deal with uncertainty. One, and
perhaps the simplest, approach is to replace each uncertain
quantity either by its average (i.e., its mean or expected
value), its median, or by some critical (e.g., worst-case)
value, and then proceed with a deterministic approach. Use
of expected or median values of uncertain quantities may be
adequate if the uncertainty or variation in a quantity is rea-
sonably small and does not critically affect the performance
169
Concepts in Probability, Statistics
and Stochastic Modelling
Events that cannot be predicted precisely are often called random. Many if not most
of the inputs to, and processes that occur in, water resources systems are to some
extent random. Hence, so too are the outputs or predicted impacts, and even
peoples reactions to those outputs or impacts. To ignore this randomness or
uncertainty is to ignore reality. This chapter introduces some of the commonly used
tools for dealing with uncertainty in water resources planning and management.
Subsequent chapters illustrate how these tools are used in various types of
optimization, simulation and statistical models for impact prediction and evaluation.
7
of the system. If expected or median values of uncertain
parameters or variables are used in a deterministic model,
the planner can then assess the importance of uncertainty
by means of sensitivity analysis, as is discussed later in this
and the two subsequent chapters.
Replacement of uncertain quantities by either expected,
median or worst-case values can grossly affect the evalua-
tion of project performance when important parameters
are highly variable. To illustrate these issues, consider
the evaluation of the recreation potential of a reservoir.
Table 7.1 shows that the elevation of the water surface
varies over time depending on the inflow and demand for
water. The table indicates the pool levels and their associ-
ated probabilities as well as the expected use of the recre-
ation facility with different pool levels.
The average pool level L

is simply the sum of each

possible pool level times its probability, or
L

10(0.10) 20(0.25) 30(0.30)

40(0.25) 50(0.10) 30 (7.1)
This pool level corresponds to 100 visitor-days per day:
VD(L

) 100 visitor-days per day (7.2)

A worst-case analysis might select a pool level of ten as a
critical value, yielding an estimate of system performance
equal to 100 visitor-days per day:
VD(L
low
) VD(10) 25 visitor-days per day (7.3)
wrm_ch07.qxd 8/31/2005 12:18 PM Page 169
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Neither of these values is a good approximation of the
average visitation rate, that is
VD

0.10 VD(10) 0.25 VD(20) 0.30 VD(30)

0.25 VD(40) 0.10 VD(50)
0.10(25) 0.25(75) 0.30(100) 0.25(80)
0.10(70) (7.4)
78.25 visitor-days per day
Clearly, the average visitation rate, VD

78.25, the
visitation rate corresponding to the average pool level
VD(L

) 100, and the worst-case assessment VD(L

low
)
25, are very different.
Using only average values in a complex model can
produce a poor representation of both the average
performance and the possible performance range. When
important quantities are uncertain, a comprehensive
analysis requires an evaluation of both the expected
performance of a project and the risk and possible
magnitude of project failures in a physical, economic,
ecological and/or social sense.
This chapter reviews many of the methods of proba-
bility and statistics that are useful in water resources
planning and management. Section 2 is a condensed
summary of the important concepts and methods of
probability and statistics. These concepts are applied in
this and subsequent chapters of this book. Section 3
presents several probability distributions that are often
used to model or describe the distribution of uncertain
quantities. The section also discusses methods for fitting
these distributions using historical information, and
methods of assessing whether the distributions are
170 Water Resources Systems Planning and Management
adequate representations of the data. Sections 4, 5 and 6
expand upon the use of these mathematical models, and
discuss alternative parameter estimation methods.
Section 7 presents the basic ideas and concepts of
the stochastic processes or time series. These are used
to model streamflows, rainfall, temperature or other
phenomena whose values change with time. The section
contains a description of Markov chains, a special type of
stochastic process used in many stochastic optimization
and simulation models. Section 8 illustrates how synthetic
flows and other time-series inputs can be generated for
stochastic simulations. Stochastic simulation is introduced
with an example in Section 9.
Many topics receive only brief treatment in this intro-
ductory chapter. Additional information can be found in
applied statistical texts or book chapters such as Benjamin
and Cornell (1970), Haan (1977), Kite (1988), Stedinger
et al. (1993), Kottegoda and Rosso (1997), and Ayyub
and McCuen (2002).
2. Probability Concepts and
Methods
This section introduces the basic concepts and definitions
used in analyses involving probability and statistics. These
concepts are used throughout this chapter and later
chapters in the book.
2.1. Random Variables and Distributions
The basic concept in probability theory is that of the
random variable. By definition, the value of a random
variable cannot be predicted with certainty. It depends,
at least in part, on the outcome of a chance event.
Examples are: (1) the number of years until the flood
stage of a river washes away a small bridge; (2) the num-
ber of times during a reservoirs life that the level of the
pool will drop below a specified level; (3) the rainfall
depth next month; and (4) next years maximum flow at
a gauge site on an unregulated stream. The values of all
of these random events or variables are not knowable
before the event has occurred. Probability can be used
to describe the likelihood that these random variables
will equal specific values or be within a given range of
specific values.
10
20
30
40
50
0.10
0.25
0.30
0.25
0.10
25
75
100
80
70
possible
pool levels
probability
of each level
recreation potential
in visitor-days per day
for reservoir with
different pool levels
E
0
2
1
1
0
1
a
Table 7.1. Data for determining reservoir recreation potential.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 170
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 171
The first two examples illustrate discrete random
variables, random variables that take on values that are
discrete (such as positive integers). The second two exam-
ples illustrate continuous random variables. Continuous
random variables take on any values within a specified
range of values. A property of all continuous random
variables is that the probability that the value of any of
those random variables will equal some specific number
any specific number is always zero. For example, the
probability that the total rainfall depth in a month will be
exactly 5.0 cm is zero, while the probability that the total
rainfall will lie between 4 and 6 cm could be nonzero.
Some random variables are combinations of continuous
and discrete random variables.
Let X denote a random variable and x a possible value
of that random variable X. Random variables are generally
denoted by capital letters, and particular values they take
on by lowercase letters. For any real-valued random
variable X, its cumulative distribution function F
X
(x), often
denoted as just the cdf, equals the probability that the value
of X is less than or equal to a specific value or threshold x:
F
X
(x) Pr[X x] (7.5)
This cumulative distribution function F
X
(x) is a non-
decreasing function of x because
Pr[X x] Pr[X x ] for 0 (7.6)
In addition,
(7.7)
and
(7.8)
The first limit equals 1 because the probability that X
takes on some value less than infinity must be unity; the
second limit is zero because the probability that X takes
on no value must be zero.
The left half of Figure 7.1 illustrates the cumulative
distribution function (upper) and its derivative, the prob-
ability density function, f
X
(x), (lower) of a continuous
random variable X.
If X is a real-valued discrete random variable that takes
on specific values x
1
, x
2
, , then the probability mass
function p
X
(x
i
) is the probability X takes on the value x
i
.
p
X
(x
i
) Pr[X x
i
] (7.9)
lim ( )
x
X
F x

0
lim ( )
x
X
F x

1
The value of the cumulative distribution function F
X
(x)
for a discrete random variable is the sum of the probabil-
ities of all x
i
that are less than or equal to x.
(7.10)
The right half of Figure 7.1 illustrates the cumulative
distribution function (upper) and the probability mass
function p
X
(x
i
) (lower) of a discrete random variable.
The probability density function f
X
(x) (lower left plot in
Figure 7.1) for a continuous random variable X is the
analogue of the probability mass function (lower right
plot in Figure 7.1) of a discrete random variable X. The
probability density function, often called the pdf, is the
derivative of the cumulative distribution function so that:
(7.11)
It is necessary to have
(7.12)
Equation 7.12 indicates that the area under the probabil-
ity density function is 1. If a and b are any two constants,
the cumulative distribution function or the density
function may be used to determine the probability that X
is greater than a and less than or equal to b where
(7.13)
The area under a probability density function specifies the
relative frequency with which the value of a continuous
random variable falls within any specified range of values,
that is, any interval along the horizontal axis.
Life is seldomly so simple that only a single quantity is
uncertain. Thus, the joint probability distribution of two
or more random variables can also be defined. If X and Y
are two continuous real-valued random variables, their
joint cumulative distribution function is:
(7.14)
If two random variables are discrete, then
(7.15)
F x y p x y
XY XY i i
y y x x
i i
( , ) ( , )


F x y X x Y y
f u v u v
XY
XY
y x
( , ) Pr[ ]
( , )

and
d d


Pr[ ] ( ) ( ) ( ) a X b F b F a f x x
X X X
a
b
d

f x
X
( )

f x
F x
x
X
X
( )
( )

d
d
0
F x x
X X i
x x
i
( ) ( )

wrm_ch07.qxd 8/31/2005 12:18 PM Page 171

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
where the joint probability mass function is:
p
XY
(x
i
, y
i
) Pr[X x
i
and Y y
i
] (7.16)
If X and Y are two random variables, and the distribution
of X is not influenced by the value taken by Y, and vice
versa, then the two random variables are said to be inde-
pendent. For two independent random variables X and Y,
the joint probability that the random variable X will be
between values a and b and that the random variable Y
will be between values c and d is simply the product of
those separate probabilities.
Pr[a X b and c Y d]
Pr[a X b] Pr[c Y d] (7.17)
This applies for any values a, b, c, and d. As a result,
F
XY
(x, y) F
X
(x)F
Y
(y) (7.18)
which implies for continuous random variables that
f
XY
(x, y) f
X
(x)f
Y
(y) (7.19)
and for discrete random variables that
p
XY
(x, y) p
X
(x)p
Y
(y)
(7.20)
172 Water Resources Systems Planning and Management
Other useful concepts are those of the marginal and
conditional distributions. If X and Y are two random variables
whose joint cumulative distribution function F
XY
(x, y) has
been specified, then F
X
(x), the marginal cumulative distribu-
tion of X, is just the cumulative distribution of X ignoring Y.
The marginal cumulative distribution function of X equals
F
X
(x) Pr[X x] (7.21)
where the limit is equivalent to letting Y take on any
value. If X and Y are continuous random variables, the
marginal density of X can be computed from
(7.22)
The conditional cumulative distribution function is the
cumulative distribution function for X given that Y has
taken a particular value y. Thus the value of Y may have
been observed and one is interested in the resulting
conditional distribution for the so far unobserved value of
X. The conditional cumulative distribution function for
continuous random variables is given by
f x f x y y
X XY
( ) ( , )

lim ( , )
y
XY
F x y

1.0
0.0
x x
1.0
0.0
1.0
0.0
x x
1.0
0.0
a
possible values of a random variable X possible values of a random variable X
possible values of a random variable X possible values of a random variable X
b
E
0
2
0
5
2
7
a
F
X
(
x
)
F
X
(
x
)
F
X
(
x
)
F
X
(
x
)
Figure 7.1. Cumulative
distribution and probability
density or mass functions of
random variables: (a) continuous
distributions; (b) discrete
distributions.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 172
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 173
(7.23)
where the conditional density function is
(7.24)
For discrete random variables, the probability of observ-
ing X x, given that Y y equals
(7.25)
These results can be extended to more than two random
variables. Kottegoda and Rosso (1997) provide more detail.
2.2. Expectation
Knowledge of the probability density function of a
continuous random variable, or of the probability mass
function of a discrete random variable, allows one to
calculate the expected value of any function of the random
variable. Such an expectation may represent the average
rainfall depth, average temperature, average demand
shortfall or expected economic benefits from system
operation. If g is a real-valued function of a continuous
random variable X, the expected value of g(X) is:
(7.26)
whereas for a discrete random variable
(7.27)
The expectation operator, E[], has several important prop-
erties. In particular, the expectation of a linear function of
X is a linear function of the expectation of X. Thus, if a
and b are two non-random constants,
E[a bX] a bE[X] (7.28)
The expectation of a function of two random variables is
given by
(7.29)
E d d
E
[ ( , )] ( , ) ( , )
[ ( , )] ( , )
g X Y g x y f x y x y
g X Y g x y
XY
i i

or
pp x y
XY i i
j i
( , )

E[ ( )] ( ) ( ) g X g x p x
i X i
i

E d [ ( )] ( ) ( ) g X g x f x x
X

p x y
p x y
p y
X Y
XY
Y
|
( | )
( , )
( )

f x y
f x y
f y
X Y
XY
Y
|
( | )
( , )
( )

F x y X x Y y
f s y s
f y
X Y
XY
x
Y
|
( | ) Pr[ | ]
( , )
( )


d

If X and Y are independent, the expectation of the product

of a function g() of X and a function h() of Y is the
product of the expectations:
E[g(X) h(Y)] E[g(X)] E[h(Y)] (7.30)
This follows from substitution of Equations 7.19 and 7.20
into Equation 7.29.
2.3. Quantiles, Moments and Their Estimators
While the cumulative distribution function provides a
complete specification of the properties of a random
variable, it is useful to use simpler and more easily under-
stood measures of the central tendency and range of
values that a random variable may assume. Perhaps the
simplest approach to describing the distribution of a
random variable is to report the value of several quantiles.
The pth quantile of a random variable X is the smallest
value x
p
such that X has a probability p of assuming a
value equal to or less than x
p
:
Pr[X x
p
] p Pr[X x
p
] (7.31)
Equation 7.31 is written to insist if at some value x
p
, the
cumulative probability function jumps from less than p to
more than p, then that value x
p
will be defined as the pth
quantile even though F
X
(x
p
) p. If X is a continuous
random variable, then in the region where f
X
(x) 0,
the quantiles are uniquely defined and are obtained by
solution of
F
X
(x
p
) p (7.32)
Frequently reported quantiles are the median x
0.50
and
the lower and upper quartiles x
0.25
and x
0.75
. The median
describes the location or central tendency of the distribu-
tion of X because the random variable is, in the continu-
ous case, equally likely to be above as below that value.
The interquartile range [x
0.25
, x
0.75
] provides an easily
understood description of the range of values that the
random variable might assume. The pth quantile is also
the 100 p percentile.
In a given application particularly when safety is of
concern it may be appropriate to use other quantiles. In
floodplain management and the design of flood control
structures, the 100-year flood x
0.99
is a commonly selected
design value. In water quality management, a rivers
minimum seven-day-average low flow expected once in
ten years is commonly used in the United States as the
wrm_ch07.qxd 8/31/2005 12:18 PM Page 173
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
critical planning value: Here the one-in-ten year value is
the 10
th
percentile of the distribution of the annual min-
ima of the seven-day average flows.
The natural sample estimate of the median x
0.50
is
the median of the sample. In a sample of size n where
x
(1)
x
(2)

x
(n)
are the observations ordered by
magnitude, and for a non-negative integer k such that
n 2k (even) or n 2k 1 (odd), the sample estimate
of the median is
x
0.50
(7.33)
Sample estimates of other quantiles may be obtained by
using x
(i)
as an estimate of x
q
for q i/(n 1) and then
interpolating between observations to obtain x

p
for the
desired p. This only works for 1/(n 1) p n/(n1)
and can yield rather poor estimates of x
p
when (n1)p
is near either 1 or n. An alternative approach is to
fit a reasonable distribution function to the observa-
tions, as discussed in Section 3, and then estimate
x
p
using Equation 7.32, where F
X
(x) is the fitted
distribution.
Another simple and common approach to describing a
distributions centre, spread and shape is by reporting the
moments of a distribution. The first moment about the
origin is the mean of X and is given by
(7.34)
Moments other than the first are normally measured
about the mean. The second moment measured about the
mean is the variance, denoted Var(X) or
X
2
, where:
(7.35)
The standard deviation
X
is the square root of the vari-
ance. While the mean
X
is a measure of the central value
of X, the standard deviation
X
is a measure of the spread
of the distribution of X about
X
.
Another measure of the variability in X is the coefficient
of variation,
(7.36)
The coefficient of variation expresses the standard
deviation as a proportion of the mean. It is useful for
CV
X
X
X

X X
X X
2 2
Var( ) [( ) ] E

X X
X xf x x

E d [ ] ( )

x n k
x x n k
k k
( )
( ) ( )
k 1
1
2 1
1
2
2
for
for

1
]

174 Water Resources Systems Planning and Management

comparing the relative variability of the flow in rivers of
different sizes, or of rainfall variability in different regions
when the random variable is strictly positive.
The third moment about the mean, denoted
X
, meas-
ures the asymmetry, or skewness, of the distribution:

X
E[(X
X
)
3
] (7.37)
Typically, the dimensionless coefficient of skewness

X
is reported rather than the third moment
X
. The
coefficient of skewness is the third moment rescaled by
the cube of the standard deviation so as to be dimension-
less and hence unaffected by the scale of the random
variable:
(7.38)
Streamflows and other natural phenomena that are neces-
sarily non-negative often have distributions with positive
skew coefficients, reflecting the asymmetric shape of their
distributions.
When the distribution of a random variable is not
known, but a set of observations {x
1
, , x
n
} is available,
the moments of the unknown distribution of X can be
estimated based on the sample values using the following
equations. The sample estimate of the mean:
(7.39a)
The sample estimate of the variance:

X
2
(7.39b)
The sample estimate of skewness:

X
(7.39c)
The sample estimate of the coefficient of variation:
(7.39d)
The sample estimate of the coefficient of skewness:

X
/S
X
3
(7.39e)
The sample estimate of the mean and variance are often
denoted as x
_
and s
x
2
where the lower case letters are
used when referring to a specific sample. All of these
CV S X
X X
/

n
n n
X X
i
i
n
( )( )
( )
1 2
3
1

S
n
X X
X i
i
n
2 2
1
1
1 ( )
( )

X X n
i
i
n

/
1

X
X
X

wrm_ch07.qxd 8/31/2005 12:18 PM Page 174

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 175
sample estimators provide only estimates of actual or
true values. Unless the sample size n is very large, the
difference between the estimators and the true values of
may be large. In many ways, the
field of statistics is about the precision of estimators
of different quantities. One wants to know how well the
mean of twenty annual rainfall depths describes the true
expected annual rainfall depth, or how large the differ-
ence between the estimated 100-year flood and the true
100-year flood is likely to be.
As an example of the calculation of moments, consider
the flood data in Table 7.2. These data have the following
sample moments:
x
_
1549.2
s
X
813.5
CV
X
0.525

X
0.712
As one can see, the data are positively skewed and have a
relatively large coefficient of variance.
When discussing the accuracy of sample estimates, two
quantities are often considered, bias and variance. An
estimator

of a known or unknown quantity is a function

of the observed values of the random variable X, say in n
different time periods, X
1
, , X
n
, that will be available to
estimate the value of ;

may be written

[X
1
, X
2
, , X
n
]
to emphasize that

itself is a random variable. Its value

depends on the sample values of the random variable that
will be observed. An estimator

of a quantity is biased if
E[

] and unbiased if E[

]. The quantity {E[

]} is
generally called the bias of the estimator.
An unbiased estimator has the property that its
expected value equals the value of the quantity to be
estimated. The sample mean is an unbiased estimate of
the population mean
X
because
(7.40)
The estimator S
X
2
of the variance of X is an unbiased
estimator of the true variance
X
2
for independent obser-
vations (Benjamin and Cornell, 1970):
(7.41)
However, the corresponding estimator of the standard
deviation, S
X
, is in general a biased estimator of
X
because
E S
X X
2 2

1
]

E E E[ ] [ ] X
n
X
n
X
i
i
n
i
i
n
X


1 1
1 1

1
]
1


X X X X X
CV , , , ,
2
and
(7.42)
The second important statistic often used to assess the
accuracy of an estimator

is the variance of the estimator

Var

, which equals E{(

E[

])
2
}. For the mean of a set
of independent observations, the variance of the sample
mean is:
(7.43)
It is common to call the standard error of x

rather
than its standard deviation. The standard error of an
average is the most commonly reported measure of its
precision.
The bias measures the difference between the average
value of an estimator and the quantity to be estimated.

x
n /
Var(X
n
X
)

2
E[ ] S
X X

1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1946
1947
1948
1949
1950
date
E
0
2
1
1
0
1
b
410
1150
899
420
3100
2530
758
1220
1330
1410
3100
2470
929
586
450
1040
1470
1070
2050
1430
discharge
m /s
3
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
date
3070
2360
1050
1900
1130
674
683
1500
2600
3480
1430
809
1010
1510
1650
1880
1470
1920
2530
1490
discharge
m /s
3
Table 7.2. Annual maximum discharges on Magra River, Italy,
at Calamazza, 1930 70*.
* Value for 1945 is missing.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 175
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The variance measures the spread or width of the estima-
tors distribution. Both contribute to the amount by
which an estimator deviates from the quantity to be
estimated. These two errors are often combined into the
mean square error. Understanding that is fixed and the
estimator

is a random variable, the mean squared error

is the expected value of the squared distance (error)
between and its estimator

:
MSE(

) E[(

)
2
] E{[

E(

)] [E(

) ]}
2
[Bias]
2
Var(

) (7.44)
where [Bias] is E(

) .
Equation 7.44 shows that the MSE, equal to the expected
average squared deviation of the estimator

from the true

value of the parameter , can be computed as the bias
squared plus the variance of the estimator. MSE is a con-
venient measure of how closely

approximates because
it combines both bias and variance in a logical way.
Estimation of the coefficient of skewness
X
provides a
good example of the use of the MSE for evaluating the
total deviation of an estimate from the true population
value. The sample estimate

X
of
X
is often biased, has a
large variance, and its absolute value was shown by Kirby
(1974) to be bounded by the square root of the sample
size n:
|

X
|
(7.45)
The bounds do not depend on the true skew,
X
. However,
the bias and variance of

X
do depend on the sample size
and the actual distribution of X. Table 7.3 contains the
expected value and standard deviation of the estimated
coefficient of skewness

X
when X has either a normal
distribution, for which
X
0, or a gamma distribution
with
X
0.25, 0.50, 1.00, 2.00, or 3.00. These values are
adapted from Wallis et al. (1974 a,b) who employed
moment estimators slightly different than those in
Equation 7.39.
For the normal distribution, E[

] 0 and Var [

X
] 5/n.
In this case, the skewness estimator is unbiased but highly
variable. In all the other cases in Table 7.3, the skewness
estimator is biased.
To illustrate the magnitude of these errors, consider the
mean square error of the skew estimator

X
calculated from
a sample of size 50 when X has a gamma distribution with

X
0.50, a reasonable value for annual streamflows. The
n
176 Water Resources Systems Planning and Management
expected value of

X
is 0.45; its variance equals (0.37)
2
,
its standard deviation is squared. Using Equation 7.44, the
mean square error of

X
is:
MSE(

X
)
(7.46)
An unbiased estimate of
X
is simply (0.50/0.45)

X
. Here
the estimator provided by Equation 7.39e has been scaled
to eliminate bias. This unbiased estimator has a mean
squared error of:
(7.47)
The mean square error of the unbiased estimator of

X
is
larger than the mean square error of the biased estimate.
Unbiasing

X
results in a larger mean square error for
all the cases listed in Table 7.3 except for the normal
distribution for which
X
0, and the gamma distribu-
tion with
X
3.00.
As shown here for the skew coefficient, biased estima-
tors often have smaller mean square errors than unbiased
estimators. Because the mean square error measures the
total average deviation of an estimator from the quantity
being estimated, this result demonstrates that the strict or
unquestioning use of unbiased estimators is not advisable.
Additional information on the sampling distribution of
quantiles and moments is contained in Stedinger et al.
(1993).
2.4. L-Moments and Their Estimators
L-moments are another way to summarize the statistical
properties of hydrological data based on linear combina-
tions of the original observations (Hosking, 1990).
Recently, hydrologists have found that regionalization
methods (to be discussed in Section 5) using L-moments
are superior to methods using traditional moments
(Hosking and Wallis, 1997; Stedinger and Lu, 1995).
L-moments have also proved useful for construction of
goodness-of-fit tests (Hosking et al., 1985; Chowdhury
et al., 1991; Fill and Stedinger, 1995), measures of
regional homogeneity and distribution selection methods
(Vogel and Fennessey, 1993; Hosking and Wallis, 1997).
MSE
0 50
0 48
0 50 0 50
0 50
0 45
0 37
2
.
.
( . . )
.
.
( . )

_
,

_
,

1
]

11
2
0 169 0 17 . .


( . . ) ( . )
. .
0 45 0 50 0 37
0 0025 0 1369
2 2
0.139 0.14

wrm_ch07.qxd 8/31/2005 12:18 PM Page 176

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 177
The first L-moment designated as
1
is simply the
arithmetic mean:

1
E[X] (7.48)
Now let X
(i|n)
be the i
th
largest observation in a sample of
size n (i n corresponds to the largest). Then, for any
distribution, the second L-moment,
2
, is a description
of scale based upon the expected difference between two
randomly selected observations:

2
(1/2) E[X
(2|1)
X
(1|2)
] (7.49)
Similarly, L-moment measures of skewness and kurtosis use
three and four randomly selected observations, respectively.

3
(1/3) E[X
(3|3)
2X
(2|3)
X
(1|3)
] (7.50)

4
(1/4) E[X
(4|4)
3X
(3|4)
3X
(2|4)
X
(1|4)
] (7.51)
Sample L-moment estimates are often computed using
intermediate statistics called probability weighted moments
(PWMs). The r
th
probability weighted moment is defined as:

r
E{X[F(X)]
r
} (7.52)
where F(X) is the cumulative distribution function of X.
Recommended (Landwehr et al., 1979; Hosking and
Wallis, 1995) unbiased PWM estimators, b
r
, of
r
are
computed as:
(7.53)
b X
b
b
0
1
2
2
3
1
1
1
1
1 2
1 2

n n
n n n
j
j
n
j
( )
( )
( )( )
( )( )
( )
j X
j j

nn
j
X
( )
0
0.25
0.50
1.00
2.00
3.00
distribution
of X
E
0
2
1
1
0
1
c
=
=
=
=
=
=

X
normal
gamma
upper bound on skew
0.00
0.15
0.31
0.60
1.15
1.59
3.16
0.00
0.19
0.39
0.76
1.43
1.97
0.00
0.23
0.45
0.88
1.68
2.32
0.00
0.23
0.47
0.93
1.77
2.54
4.47 7.07 8.94
10 20 50 80
sample size
10 20 50 80
sample size

X
expected value of
0
0.25
0.50
1.00
2.00
3.00
distribution
of X
=
=
=
=
=
=
normal
gamma
0.69
0.69
0.69
0.70
0.72
0.74
0.51
0.52
0.53
0.57
0.68
0.76
0.34
0.35
0.37
0.44
0.62
0.77
0.26
0.28
0.31
0.38
0.57
0.77
X

standard deviation of

X
^
Table 7.3. Sampling properties of
coefficient of skewness estimator.
Source: Wallis et al. (1974b) who only divided by n in
the estimators of the moments, whereas in Equations
7.39b and 7.39c, we use the generally-adopted
coefficients of 1/(n 1) and n/(n 1)(n 2) for the
variance and skew.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 177
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
These are examples of the general formula for computing
estimators b
r
of

r
.
(7.54)
for r 1, , n 1.
L-moments are easily calculated in terms of PWMs
using:

1

0

2
2
1

0

3
6
2
6
1

0

4
20
3
30
2
12
1

0
(7.55)
Wang (1997) provides formulas for directly calculating
L-moment estimators of
r
. Measures of the coefficient
of variation, skewness and kurtosis of a distribution can be
computed with L-moments, as they can with traditional
product moments. Where skew primarily measures the
asymmetry of a distribution, the kurtosis is an additional
measure of the thickness of the extreme tails. Kurtosis is
b
n
j
r
X
n
r
r
j
r
r
j r
n
j
j r
n

1
1 1
1
1
1
1
1

( )

X
n
r
j ( )

1
178 Water Resources Systems Planning and Management
particularly useful for comparing symmetric distributions
that have a skewness coefficient of zero. Table 7.4 provides
definitions of the traditional coefficient of variation, coeffi-
cient of skewness and coefficient of kurtosis, as well as
the L-moment, L-coefficient of variation, L-coefficient of
skewness and L-coefficient of kurtosis.
The flood data in Table 7.2 can be used to provide an
example of L-moments. Equation 7.53 yields estimates of
the first three Probability Weighted Moments:
b
0
1,549.20
b
1
1003.89
b
2
759.02 (7.56)
Recall that b
0
is just the sample average x
_
. The sample
L-moments are easily calculated using the probability
weighted moments. One obtains:

1
b
0
1,549

2
2b
1
b
0
458

3
6b
2
6b
1
b
0
80 (7.55)
Thus, the sample estimates of the L-coefficient of
variation, t
2
, and L-coefficient of skewness, t
3
, are:
t
2
0.295
t
3
0.174 (7.58)
Table 7.4. Definitions of dimensionless
product-moment and L-moment ratios.
coefficient of
product-moment ratios
E
0
2
1
1
0
1
d
variation
skewness
kurtosis
L-coefficient of
variation
skewness
kurtosis
CV
L-CV,
L-skewness,
L-kurtosis,

X
X
X
*

2
3
4
/

E [
E [
(X -
(X -

X
)
3
] /
] /
X
)
4

X
X
3
4
X

X
2

3
4
/
/
/

2
2
L-moment ratios
name definition common symbol
*
Hosking and Wallis (1997) use instead of to represent the L-CV ratio
2

wrm_ch07.qxd 8/31/2005 12:18 PM Page 178

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 179
3. Distributions of Random Events
A frequent task in water resources planning is the
development of a model of some probabilistic or stochastic
phenomena such as streamflows, flood flows, rainfall, tem-
peratures, evaporation, sediment or nutrient loads, nitrate
or organic compound concentrations, or water demands.
This often requires one to fit a probability distribution
function to a set of observed values of the random variable.
Sometimes, ones immediate objective is to estimate a
particular quantile of the distribution, such as the 100-year
flood, 50-year six-hour-rainfall depth, or the minimum
seven-day-average expected once-in-ten-year flow. Then the
fitted distribution can supply an estimate of that quantity. In
a stochastic simulation, fitted distributions are used to
generate possible values of the random variable in question.
Rather than fitting a reasonable and smooth mathe-
matical distribution, one could use the empirical distribu-
tion represented by the data to describe the possible
values that a random variable may assume in the future
and their frequency. In practice, the true mathematical
form for the distribution that describes the events is not
known. Moreover, even if it was, its functional form may
have too many parameters to be of much practical use.
Thus, using the empirical distribution represented by the
data itself has substantial appeal.
Generally, the free parameters of the theoretical distri-
bution are selected (estimated) so as to make the fitted
distribution consistent with the available data. The goal is
to select a physically reasonable and simple distribution
to describe the frequency of the events of interest, to
estimate that distributions parameters, and ultimately to
obtain quantiles, performance indices and risk estimates
of satisfactory accuracy for the problem at hand. Use of a
theoretical distribution has several advantages over use of
the empirical distribution:
It presents a smooth interpretation of the empirical
distribution. As a result quantiles, performance indices
and other statistics computed using the fitted distribu-
tion should be more accurate than those computed
with the empirical distribution.
It provides a compact and easy-to-use representation
of the data.
It is likely to provide a more realistic description of
the range of values that the random variable may
assume and their likelihood. For example, by using
the empirical distribution, one implicitly assumes that
no values larger or smaller than the sample maximum
or minimum can occur. For many situations, this is
unreasonable.
Often one needs to estimate the likelihood of extreme
events that lie outside the range of the sample (either
in terms of x values or in terms of frequency). Such
extrapolation makes little sense with the empirical
distribution.
In many cases, one is not interested in the values of a
random variable X, but instead in derived values of
variables Y that are functions of X. This could be a
performance function for some system. If Y is the
performance function, interest might be primarily in
its mean value E[Y], or the probability some standard
is exceeded, Pr{Y standard}. For some theoretical
X-distributions, the resulting Y-distribution may
be available in closed form, thus making the analysis
rather simple. (The normal distribution works with
linear models, the lognormal distribution with product
models, and the gamma distribution with queuing
systems.)
This section provides a brief introduction to some
useful techniques for estimating the parameters of prob-
ability distribution functions and for determining if a
fitted distribution provides a reasonable or acceptable
model of the data. Sub-sections are also included on
families of distributions based on the normal, gamma
and generalized-extreme-value distributions. These three
families have found frequent use in water resources
planning (Kottegoda and Rosso, 1997).
3.1. Parameter Estimation
Given a set of observations to which a distribution is to
be fit, one first selects a distribution function to serve as
a model of the distribution of the data. The choice of a
distribution may be based on experience with data of
that type, some understanding of the mechanisms giving
rise to the data, and/or examination of the observations
themselves. One can then estimate the parameters of the
chosen distribution and determine if the fitted distribu-
tion provides an acceptable model of the data. A model is
generally judged to be unacceptable if it is unlikely that
wrm_ch07.qxd 8/31/2005 12:18 PM Page 179
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
one could have observed the available data were they
actually drawn from the fitted distribution.
In many cases, good estimates of a distributions
parameters are obtained by the maximum-likelihood-
estimation procedure. Give a set of n independent
observations {x
1
, , x
n
} of a continuous random variable
X, the joint probability density function for the observa-
tions is:
(7.59)
where is the vector of the distributions parameters.
The maximum likelihood estimator of is that
vector which maximizes Equation 7.59 and thereby
makes it as likely as possible to have observed the values
{x
1
, , x
n
}.
Considerable work has gone into studying the
properties of maximum likelihood parameter estimates.
Under rather general conditions, asymptotically the
estimated parameters are normally distributed, unbiased
and have the smallest possible variance of any asymptoti-
cally unbiased estimator (Bickel and Doksum, 1977).
These, of course, are asymptotic properties, valid for large
sample sizes n. Better estimation procedures, perhaps
yielding biased parameter estimates, may exist for small
sample sizes. Stedinger (1980) provides such an
example. Still, maximum likelihood procedures are
recommended with moderate and large samples, even
though the iterative solution of nonlinear equations is
often required.
An example of the maximum likelihood procedure
for which closed-form expressions for the parameter
estimates are obtained is provided by the lognormal
distribution. The probability density function of a lognor-
mally distributed random variable X is:
(7.60)
Here, the parameters and
2
are the mean and variance
of the logarithm of X, and not of X itself.
Maximizing the logarithm of the joint density for
{x
1
, , x
n
} is more convenient than maximizing the
joint probability density itself. Hence, the problem can
be expressed as the maximization of the log-likelihood
function
f x
x
x
X
( ) exp [ln( ) ]
1
2
1
2
2
2
2

f x x
f x f x f x
X X X X n
X X X n
n 1 2 3
1
1 2
, , , ,
( , , |
( | ( | ( |

)
) ) )
180 Water Resources Systems Planning and Management
(7.61)
The maximum can be obtained by equating to zero the
partial derivatives L/ and L/ whereby one obtains:
(7.62)
These equations yield the estimators

2
(7.63)
The second-order conditions for a maximum are met and
these values maximize Equation 7.59. It is useful to note
that if one defines a new random variable Y ln(X), then
the maximum likelihood estimators of the parameters
and
2
, which are the mean and variance of the Y
distribution, are the sample estimators of the mean and
variance of Y:

y
_

2
[(n 1)/n]S
Y
2
(7.64)
The correction [(n 1)/n] in this last equation is often
neglected.
The second commonly used parameter estimation
procedure is the method of moments. The method of
moments is often a quick and simple method for obtaining
parameter estimates for many distributions. For a distribu-
tion with m 1, 2 or 3 parameters, the first m moments of
the postulated distribution in Equations 7.34, 7.35 and
7.37 are equated to the estimates of those moments
calculated using Equations 7.39. The resulting nonlinear
equations are solved for the unknown parameters.

1
2
1
n
x
i
i
n
[ln( ) ]

1
1
n
x
i
i
n
ln( )

0
1
0
1
2
1
3
2
1

L
x
L n
x
i
i
n
i
i
n




[ln( ) ]
[ln( ) ]
L

ln [( , )]
ln ( , )
ln( ) ln(
f x
f x
x n
i
i
n
i
i
n
i
i
n
|
|

1
1
1
2

))
ln( )

1
2
2
2
1

[ ] x
i
i
n

wrm_ch07.qxd 8/31/2005 12:18 PM Page 180

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 181
For the lognormal distribution, the mean and variance
of X as a function of the parameters and are given by
(7.65)
Substituting x
_
for
X
and s
X
2
for
X
2
and solving for and

2
one obtains

(7.66)
The data in Table 7.2 provide an illustration of both
fitting methods. One can easily compute the sample mean
and variance of the logarithms of the flows to obtain

7.202

2
0.3164 (0.5625)
2
(7.67)
Alternatively, the sample mean and variance of the flows
themselves are
x
_
1549.2
s
X
2
661,800 (813.5)
2
(7.68)
Substituting those two values in Equation 7.66 yields

7.224

X
2
0.2435 (0.4935)
2
(7.69)
Method of moments and maximum likelihood are just
two of many possible estimation methods. Just as method
of moments equates sample estimators of moments to
population values and solves for a distributions parame-
ters, one can simply equate L-moment estimators to
population values and solve for the parameters of a
distribution. The resulting method of L-moments has
received considerable attention in the hydrological
literature (Landwehr et al., 1978; Hosking et al., 1985;
Hosking and Wallis, 1987; Hosking, 1990; Wang, 1997).
It has been shown to have significant advantages when
used as a basis for regionalization procedures that will be
discussed in Section 5 (Lettenmaier et al., 1987; Stedinger
and Lu, 1995; Hosking and Wallis, 1997).
Bayesian procedures provide another approach that
is related to maximum likelihood estimation. Bayesian

ln
/
ln
x
s x
x
X
1
1
2
2 2
2

_
,

ln / ( ) 1
2 2
s x
X


X
X


exp
exp( )[exp ( ) ]
1
2
2 1
2
2 2 2

_
,

inference employs the likelihood function to represent the

information in the data. That information is augmented
with a prior distribution that describes what is known
about constraints on the parameters and their likely
values beyond the information provided by the recorded
data available at a site. The likelihood function and the
prior probability density function are combined to obtain
the probability density function that describes the poste-
rior distribution of the parameters:
f

( | x
1
, x
2
, , x
n
)
f
X
(x
1
, x
2
, , x
n
| )() (7.70)
The symbol means proportional to and () is the
probability density function for the prior distribution
for (Kottegoda and Rosso, 1997). Thus, except for
a constant of proportionality, the probability density
function describing the posterior distribution of the
parameter vector is equal to the product of the likeli-
hood function f
X
(x
1
, x
2
, , x
n
| ) and the probability
density function for the prior distribution () for .
Advantages of the Bayesian approach are that it allows
the explicit modelling of uncertainty in parameters
(Stedinger, 1997; Kuczera, 1999) and provides a theoret-
ically consistent framework for integrating systematic
flow records with regional and other hydrological infor-
mation (Vicens et al., 1975; Stedinger, 1983; Kuczera,
1983). Martins and Stedinger (2000) illustrate how a
prior distribution can be used to enforce realistic
constraints upon a parameter as well as providing a
description of its likely values. In their case, use of a
prior of the shape parameter of a generalized extreme
value (GEV) distribution (discussed in Section 3.6)
allowed definition of generalized maximum likelihood
estimators that, over the -range of interest, performed
substantially better than maximum likelihood, moment,
and L-moment estimators.
While Bayesian methods have been available for
decades, the computational challenge posed by the
solution of Equation 7.70 has been an obstacle to their
use. Solutions to Equation 7.70 have been available for
special cases such as normal data, and binomial and
Poisson samples (Raiffa and Schlaifer, 1961; Benjamin
and Cornell, 1970; Zellner, 1971). However, a new and
very general set of Markov Chain Monte Carlo (MCMC)
procedures (discussed in Section 7.2) allows numerical
computation of the posterior distributions of parameters

wrm_ch07.qxd 8/31/2005 12:18 PM Page 181

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
for a very broad class of models (Gilks et al., 1996). As a
result, Bayesian methods are now becoming much more
popular and are the standard approach for many difficult
problems that are not easily addressed by traditional
methods (Gelman et al., 1995; Carlin and Louis, 2000).
The use of Monte Carlo Bayesian methods in flood
frequency analysis, rainfallrunoff modelling, and evalua-
tion of environmental pathogen concentrations are
illustrated by Wang (2001), Bates and Campbell (2001)
and Crainiceanu et al. (2002), respectively.
Finally, a simple method of fitting flood frequency
curves is to plot the ordered flood values on special
probability paper and then to draw a line through the
data (Gumbel, 1958). Even today, that simple method
is still attractive when some of the smallest values are
zero or unusually small, or have been censored as will
be discussed in Section 4 (Kroll and Stedinger, 1996).
Plotting the ranked annual maximum series against a
probability scale is always an excellent and recommended
way to see what the data look like and for determining
whether or not a fitted curve is consistent with the data
(Stedinger et al., 1993).
Statisticians and hydrologists have investigated which
of these methods most accurately estimates the parameters
themselves or the quantiles of the distribution. One also
needs to determine how accuracy should be measured.
Some studies have used average squared deviations, some
have used average absolute weighted deviations with
different weights on under and over-estimation, and some
have used the squared deviations of the log-quantile
estimator (Slack et al., 1975; Kroll and Stedinger, 1996).
In almost all cases, one is also interested in the bias of an
estimator, which is the average value of the estimator
minus the true value of the parameter or quantile being
estimated. Special estimators have been developed to
compute design events that on average are exceeded with
the specified probability and have the anticipated risk of
being exceeded (Beard, 1960, 1997; Rasmussen and
Rosbjerg, 1989, 1991a,b; Stedinger, 1997; Rosbjerg and
Madsen, 1998).
3.2. Model Adequacy
After estimating the parameters of a distribution, some
check of model adequacy should be made. Such checks
vary from simple comparisons of the observations with
182 Water Resources Systems Planning and Management
the fitted model (using graphs or tables) to rigorous
statistical tests. Some of the early and simplest methods
of parameter estimation were graphical techniques.
Although quantitative techniques are generally more
accurate and precise for parameter estimation, graphical
presentations are invaluable for comparing the fitted
distribution with the observations for the detection of
systematic or unexplained deviations between the two.
The observed data will plot as a straight line on probabil-
ity graph paper if the postulated distribution is the true
distribution of the observation. If probability graph paper
does not exist for the particular distribution of interest,
more general techniques can be used.
Let x
(i)
be the ith largest value in a set of observed
values {x
i
} so that x
(1)
x
(2)

x
(n)
. The random
variable X
(i)
provides a reasonable estimate of the pth
quantile x
p
of the true distribution of X for p i/(n 1).
In fact, when one considers the cumulative probability U
i
associated with the random variable X
(i)
, U
i
F
X
(X
(i)
), and
if the observations X
(i)
are independent, then the U
i
have
a beta distribution (Gumbel, 1958) with probability
density function:
(7.71)
This beta distribution has mean
(7.72a)
and variance
(7.72b)
A good graphical check of the adequacy of a fitted distribu-
tion G(x) is obtained by plotting the observations x
(i)
versus
G
1
[i/(n 1)] (Wilk and Gnanadesikan, 1968). Even if G(x)
equalled to an exact degree the true X-distribution F
X
[x], the
plotted points would not fall exactly on a 45 line through
the origin of the graph. This would only occur if F
X
[x
(i)
]
exactly equalled i/(n 1), and therefore each x
(i)
exactly
equalled F
X
1
[i/(n 1)].
An appreciation for how far an individual observation
x
(i)
can be expected to deviate from G
1
[i/(n 1)] can be
obtained by plotting G
1
[u
i
(0.75)
] and G
1
[u
i
(0.25)
], where
u
i
(0.75)
and u
i
(0.25)
are the upper and lower quartiles of the
distribution of U
i
obtained from integrating the probability
Var( )
( )
( ) ( )
U
i n i
n n
i



1
1 2
2
E[ ] U
i
n
i

1
f u
n
i n
u u u
U
i n i
i
( )
!
( ) ! ( ) !
( )



1 1
1 0 1
1
wrm_ch07.qxd 8/31/2005 12:18 PM Page 182
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 183
density function in Equation 7.71. The required incomplete
beta function is also available in many software packages,
including Microsoft Excel. Stedinger et al. (1993) show
that u
(1)
and (1 u
(n)
) fall between 0.052/n and 3(n 1)
with a probability of 90%, thus illustrating the great uncer-
tainty associated with the cumulative probability of the
smallest value and the exceedance probability of the
largest value in a sample.
Figures 7.2a and 7.2b illustrate the use of this
quantilequantile plotting technique by displaying the
results of fitting a normal and a lognormal distribution
to the annual maximum flows in Table 7.2 for the Magra
River, Italy, at Calamazza for the years 1930 70. The
observations of X
(i)
, given in Table 7.2, are plotted on
the vertical axis against the quantiles G
1
[i/(n 1)] on the
horizontal axis.
A probability plot is essentially a scatter plot of the
sorted observations X
(i)
versus some approximation of
their expected or anticipated value, represented by G
1
(p
i
),
where, as suggested, p
i
i/(n 1). The p
i
values are called
plotting positions. A common alternative to i/(n 1) is
(i 0.5)/n, which results from a probabilistic interpreta-
tion of the empirical distribution of the data. Many reason-
able plotting position formulas have been proposed based
upon the sense in which G
1
(p
i
) should approximate
X
(i)
. The Weibull formula i/(n 1) and the Hazen formula
(i 0.5)/n bracket most of the reasonable choices. Popular
formulas are summarized by Stedinger et al. (1993), who
also discuss the generation of probability plots for many
distributions commonly employed in hydrology.
Rigorous statistical tests are available for trying to
determine whether or not it is reasonable to assume that a
given set of observations could have been drawn from
a particular family of distributions. Although not the most
powerful of such tests, the KolmogorovSmirnov test
provides bounds within which every observation should lie
if the sample is actually drawn from the assumed
distribution. In particular, for G F
X
, the test specifies that
(7.73)
where C

is the critical value of the test at significance

level . Formulas for C

as a function of n are contained

in Table 7.5 for three cases: (1) when G is completely
Pr
( )
G
i
n
C X G
i
n
C i
i




1 1
1
1

_
,

_
,

1
]
1
0 1000 2000 3000 4000
E
0
2
0
5
2
7
d
quantiles of fitted normal distribution G
-1
[ i /(n+1)] m
3
/sec)
o
b
s
e
r
v
e
d

v
a
l
u
e
s

X
(
i
)
a
n
d

K
o
l
m
o
g
o
r
o
v
-
S
m
i
r
n
o
v

b
o
u
n
d
s

(
m
3
/
s
e
c
)
lower 90% confidence interval
for all points
upper 90% confidence
interval for all points
4000
3000
2000
1000
0
0 1000 2000 3000 4000
E
0
2
0
5
2
7
c
quantiles of fitted lognormal distribution G
-1
[i /(n+1)] (m
3
/sec)
o
b
s
e
r
v
e
d

v
a
l
u
e
s

X
(
i
)

a
n
d

K
o
l
m
o
g
o
r
o
v
-
S
m
i
r
n
o
v

b
o
u
n
d
s

(
m
3
/
s
e
c
)
4000
3000
2000
1000
0
lower 90% confidence interval
for all points
upper 90% confidence
interval for all points
Figure 7.2. Plots of annual maximum discharges of
Magra River, Italy, versus quantiles of fitted (a) normal and
(b) lognormal distributions.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 183
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
specified independent of the samples values; (2) when
G is the normal distribution and the mean and variance
are estimated from the sample with x
_
and s
X
2
; and (3)
when G is the exponential distribution and the scale
parameter is estimated as 1/(x
_
). Chowdhury et al. (1991)
provide critical values for the Gumbel and generalized
extreme value (GEV) distributions (Section 3.6) with
known shape parameter . For other distributions, the
values obtained from Table 7.5 may be used to construct
approximate simultaneous confidence intervals for
every X
(i)
.
Figures 7.2a and b contain 90% confidence intervals
for the plotted points constructed in this manner. For
the normal distribution, the critical value of C

equals
, where 0.819 corresponds
to 0.10. For n 40, one computes C

0.127. As
can be seen in Figure 7.2a, the annual maximum flows
are not consistent with the hypothesis that they were
drawn from a normal distribution; three of the observa-
tions lie outside the simultaneous 90% confidence inter-
vals for all the points. This demonstrates a statistically
significant lack of fit. The fitted normal distribution
underestimates the quantiles corresponding to small and
large probabilities while overestimating the quantiles in
an intermediate range. In Figure 7.2b, deviations between
the fitted lognormal distribution and the observations
can be attributed to the differences between F
X
(x
(i)
) and
i/(n 1). Generally, the points are all near the 45 line
through the origin, and no major systematic deviations
are apparent.
0 819 0 01 0 85 . /( . . / ) n n
184 Water Resources Systems Planning and Management
The KolmogorovSmirnov test conveniently provides
bounds within which every observation on a probability
plot should lie if the sample is actually drawn from the
assumed distribution, and thus is useful for visually
evaluating the adequacy of a fitted distribution. However,
it is not the most powerful test available for estimating
which distribution a set of observations is likely to have
been drawn from. For that purpose, several other more
analytical tests are available (Filliben, 1975; Hosking,
1990; Chowdhury et al., 1991; Kottegoda and Rosso,
1997).
The Probability Plot Correlation test is a popular and
powerful test of whether a sample has been drawn from a
postulated distribution, though it is often weaker than alter-
native tests at rejecting thin-tailed alternatives (Filliben,
1975; Fill and Stedinger, 1995). A test with greater power
has a greater probability of correctly determining that a
sample is not from the postulated distribution. The
Probability Plot Correlation Coefficient test employs the
correlation r between the ordered observations x
(i)
and the
corresponding fitted quantiles w
i
G
1
(p
i
), determined by
plotting positions p
i
for each x
(i)
. Values of r near 1.0 suggest
that the observations could have been drawn from the fitted
distribution: r measures the linearity of the probability plot
providing a quantitative assessment of fit. If x
_
denotes the
average value of the observations and w
_
denotes the average
value of the fitted quantiles, then
(7.74)
r
x x w w
x x w w
i i
i i

( ) ( )
( ) ( )
( )
( )
.
( )

1
]
1
2 2 0 5
Table 7.5. Critical values of
KolmogorovSmirnov statistic as a function
of sample size n (after Stephens, 1974).
F
x
completely specified:
E
0
2
1
1
0
1
e
C

(
n
+ 0.12 + 0.11 /
n
)
F
x
normal with mean and variance
F
x
exponential with scale parameter
C
C
estimated as x and s
b estimated as 1 / (x)

(
(
n
n
+ 0.01 + 0.85 /
+ 0.26 + 0.5 /
n
n
)
) (
2
x
+ 0.2 / ) n 0.926
0.775
1.138
significance level
0.990 1.094 1.190 1.308
0.819 0.895 0.995 1.035
1.224 1.358 1.480 1.628
0.150 0.100 0.050 0.025 0.010
values of C

are calculated as follows:
C for case 2 with = 0.10, = 0.819 / (
n
- 0.01 + 0.85 /
n
)
wrm_ch07.qxd 8/31/2005 12:18 PM Page 184
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 185
E
0
2
1
1
0
1
f
0.9180
0.9383
0.9503
0.9639
0.9715
0.9764
0.9799
0.9835
0.9870
0.99525
0.99824
0.9347
0.9506
0.9600
0.9707
0.9767
0.9807
0.9835
0.9865
0.9893
0.99602
0.99854
0.8804
0.9110
0.9290
0.9490
0.9597
0.9664
0.9710
0.9757
0.9812
0.99354
0.99755
10
15
20
30
40
50
60
75
100
300
1,000
0.10
significance level
0.01 0.05 n
E
0
2
1
1
0
1
g
0.9084
0.9390
0.9526
0.9594
0.9646
0.9685
0.9720
0.9747
0.9779
0.9902
0.99622
0.9260
0.9517
0.9622
0.9689
0.9729
0.9760
0.9787
0.9804
0.9831
0.9925
0.99708
0.8630
0.9060
0.9191
0.9286
0.9389
0.9467
0.9506
0.9525
0.9596
0.9819
0.99334
10
20
30
40
50
60
70
80
100
300
1,000
0.10
significance level
0.01 0.05 n
Table 7.6. Lower critical values of the probability plot
correlation test statistic for the normal distribution using
p
i
(i 3/8)/(n 1/4) (Vogel, 1987).
Table 7.7. Lower critical values of the probability plot
correlation test statistic for the Gumbel distribution using
p
i
(i 0.44)/(n 0.12) (Vogel, 1987).
Table 7.6 provides critical values for r for the normal
distribution, or the logarithms of lognormal variates,
based upon the Blom plotting position that has
p
i
(i 3/8)/(n 1/4). Values for the Gumbel distribu-
tion are reproduced in Table 7.7 for use with the
Gringorten plotting position p
i
(i 0.44)/(n 0.12).
The table also applies to logarithms of Weibull variates
(Stedinger et al., 1993). Other tables are available
for the GEV (Chowdhury et al., 1991), the Pearson
type 3 (Vogel and McMartin, 1991), and exponential
and other distributions (DAgostino and Stephens,
1986).
Recently developed L-moment ratios appear to provide
goodness-of-fit tests that are superior to both the
KolmogorovSmirnov and the Probability Plot Correlation
test (Hosking, 1990; Chowdhury et al., 1991; Fill and
Stedinger, 1995). For normal data, the L-skewness estima-
tor

3
(or t
3
) would have mean zero and Var

3
(0.1866
0.8/n)/n, allowing construction of a powerful test of
normality against skewed alternatives using the normally
distributed statistic
(7.75)
with a reject region |Z| z
/2
.
Chowdhury et al. (1991) derive the sampling variance
of the L-CV and L-skewness estimators

2
and

3
as a
function of for the GEV distribution. These allow
construction of a test of whether a particular data set is
consistent with a GEV distribution with a regionally
estimated value of , or a regional and a regional
coefficient of variation, CV. Fill and Stedinger (1995)
show that the

3
L-skewness estimator provides a test for
the Gumbel versus a general GEV distribution using the
normally distributed statistic
Z (

3
0.17)/ (7.76)
with a reject region |Z| z
/2
.
The literature is full of goodness-of-fit tests.
Experience indicates that among the better tests there
is often not a great deal of difference (DAgostino and
Stephens, 1986). Generation of a probability plot is
most often a good idea because it allows the modeller
to see what the data look like and where problems
occur. The KolmogorovSmirnov test helps the eye
( . . / ) / 0 2326 0 70 n n
Z t n n
3
0 1866 0 8 / ( . . / ) /
wrm_ch07.qxd 8/31/2005 12:18 PM Page 185
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
interpret a probability plot by adding bounds to a graph,
illustrating the magnitude of deviations from a straight
line that are consistent with expected variability. One
can also use quantiles of a beta distribution to illustrate
the possible error in individual plotting positions,
particularly at the extremes where that uncertainty is
largest. The probability plot correlation test is a popular
and powerful goodness-of-fit statistic. Goodness-of-fit
tests based upon sample estimators of the L-skewness

3
for the normal and Gumbel distribution provide
simple and useful tests that are not based on a proba-
bility plot.
3.3. Normal and Lognormal Distributions
The normal distribution and its logarithmic transforma-
tion, the lognormal distribution, are arguably the most
widely used distributions in science and engineering.
The probability density function of a normal random
variable is
(7.77)
where and
2
are equivalent to
X
and
X
2
, the mean
and variance of X. Interestingly, the maximum likelihood
estimators of and
2
are almost identical to the moment
estimates x
_
and s
X
2
.
The normal distribution is symmetric about its
mean
X
and admits values from to . Thus, it is
not always satisfactory for modelling physical phenomena
such as streamflows or pollutant concentrations, which
are necessarily non-negative and have skewed distribu-
tions. A frequently used model for skewed distributions is
the lognormal distribution. A random variable X has a
lognormal distribution if the natural logarithm of X, ln(X),
has a normal distribution. If X is lognormally distributed,
then by definition ln(X) is normally distributed, so that
the density function of X is
(7.78)
f x n x
x
x
x
X
( ) exp [ ( ) ]
( )
exp


1
2
1
2
1
2
1
2
2
2
2
2
2

1
d ln
d

[[ ( / )] ln x
2

f x x
X
( ) exp ( )

1
2
1
2
2
2
2

1
]
1
for X
186 Water Resources Systems Planning and Management
for x 0 and ln(). Here is the median of the
X-distribution.
A lognormal random variable takes on values in the
range [0, ]. The parameter determines the scale of
the X-distribution whereas
2
determines the shape of
the distribution. The mean and variance of the lognormal
distribution are given in Equation 7.65. Figure 7.3
illustrates the various shapes that the lognormal probability
density function can assume. It is highly skewed with
a thick right hand tail for 1, and approaches a
symmetric normal distribution as 0. The density func-
tion always has a value of zero at x 0. The coefficient of
variation and skew are:
CV
X
[exp(
2
) 1]
1/2
(7.79)

X
3CV
X
CV
X
3
(7.80)
The maximum likelihood estimates of and
2
are given
in Equation 7.63 and the moment estimates in Equation
7.66. For reasonable-sized samples, the maximum likeli-
hood estimates generally perform as well or better than
the moment estimates (Stedinger, 1980).
The data in Table 7.2 were used to calculate the
parameters of the lognormal distribution that would
describe these flood flows. The results are reported in
Equation 7.67. The two-parameter maximum likelihood
and method of moments estimators identify parameter
estimates for which the distribution skewness coefficients
Figure 7.3. Lognormal probability density functions with
various standard deviations .
2.5
2.0
1.5
1.0
0.5
0.0
0.0 0.5 1.0 .5 2.0 2.5 3.0 3.5
x
f
(
x
)
E
0
2
0
5
2
7
e

= 0.2
= 0.5
= 1.5
1
wrm_ch07.qxd 8/31/2005 12:18 PM Page 186
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 187
are 2.06 and 1.72, which is substantially greater than the
sample skew of 0.712.
A useful generalization of the two-parameter lognormal
distribution is the shifted lognormal or three-parameter log-
normal distribution obtained when ln(X ) is described
by a normal distribution, and X . Theoretically, should
be positive if, for physical reasons, X must be positive;
practically, negative values of can be allowed when the
resulting probability of negative values of X is sufficiently
small.
Unfortunately, maximum likelihood estimates of the
parameters ,
2
, and are poorly behaved because of irreg-
ularities in the likelihood function (Giesbrecht and
Kempthorne, 1976). The method of moments does fairly
well when the skew of the fitted distribution is reasonably
small. A method that does almost as well as the moment
method for low-skew distributions, and much better for
highly skewed distributions, estimates by:

(7.81)
provided that x
(1)
x
(n)
2x

0.50
0, where x
(1)
and x
(n)
are the smallest and largest observations and x

0.50
is the
sample median (Stedinger, 1980; Hoshi et al., 1984). If
x
(1)
x
(n)
2x

0.50
0, then the sample tends to be neg-
atively skewed and a three-parameter lognormal distribu-
tion with a lower bound cannot be fit with this method.
Good estimates of and
2
to go with

in Equation 7.81
are (Stedinger, 1980):

2
(7.82)
For the data in Table 7.2, Equations 7.81 and 7.82 yield
the hybrid moment-of-moments estimates of

7.606,

2
0.1339 (0.3659)
2
and

600.1 for the three-

parameter lognormal distribution.
This distribution has a coefficient of skewness of 1.19,
which is more consistent with the sample skewness
estimator than were the values obtained when a two-
parameter lognornal distribution was fit to the data.
Alternatively, one can estimate and
2
by the sample

ln 1
2
2
s
x
X
( )

1
]
1

ln
x
s x
X

1
2 2
/( )

1
]
1
1

x x x
x x x
n
n
( ) ( ) .
( ) ( ) .
1 0 50
2
1 0 50
2
mean and variance of ln(X

) which yields the

hybrid maximum likelihood estimates

7.605,

2
0.1407 (0.3751)
2
and again

600.1.
The two sets of estimates are surprisingly close in this
instance. In this second case, the fitted distribution has a
coefficient of skewness of 1.22.
Natural logarithms have been used here. One could
have just as well use base 10 common logarithms to
estimate the parameters; however, in that case the relation-
ships between the log-space parameters and the real-space
moments change slightly (Stedinger et al., 1993, Equation.
18.2.8).
3.4. Gamma Distributions
The gamma distribution has long been used to model
many natural phenomena, including daily, monthly and
annual streamflows as well as flood flows (Bobe and
Ashkar, 1991). For a gamma random variable X,
(7.83)
The gamma function, (), for integer is ( 1)!.
The parameter 0 determines the shape of the
distribution; is the scale parameter. Figure 7.4 illustrates
the different shapes that the probability density function for
a gamma variable can assume. As , the gamma
distribution approaches the symmetric normal distribu-
tion, whereas for 0 1, the distribution has a highly
asymmetric J-shaped probability density function whose
value goes to infinity as x approaches zero.
The gamma distribution arises naturally in many
problems in statistics and hydrology. It also has a very
reasonable shape for such non-negative random variables
as rainfall and streamflow. Unfortunately, its cumulative
distribution function is not available in closed form,
except for integer , though it is available in many
software packages including Microsoft Excel. The gamma
f x x x
CV
X
x
X
X
X X
( )
( )
( )

1
2
2
0
2
2 0
e
for

wrm_ch07.qxd 8/31/2005 12:18 PM Page 187

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
family includes a very special case: the exponential
distribution is obtained when 1.
The gamma distribution has several generalizations
(Bobe and Ashkar, 1991). If a constant is subtracted
from X so that (X ) has a gamma distribution, the
distribution of X is a three-parameter gamma. This is also
called a Pearson type 3 distribution, because the resulting
distribution belongs to the third type of distributions
suggested by the statistician Karl Pearson. Another
variation is the log Pearson type 3 distribution obtained
by fitting the logarithms of X with a Pearson type 3
distribution. The log Pearson distribution is discussed
further in the next section.
The method of moments may be used to estimate the
parameters of the gamma distribution. For the three-
parameter gamma distribution,

(7.84)
where x
_
, s
X
2
, and
X
are estimates of the mean, variance,
and coefficient of skewness of the distribution of X
(Bobe and Robitaille, 1977).

2
s
X X

4
2
( )
X
x
s
X
X
2

_
,

188 Water Resources Systems Planning and Management

For the two-parameter gamma distribution,

(7.85)
Again, the flood record in Table 7.2 can be used to
illustrate the different estimation procedures. Using
the first three sample moments, one would obtain for the
three-parameter gamma distribution the parameter
estimates

735.6

7.888

0.003452 1/427.2
Using only the sample mean and variance yields the
method of moment estimators of the parameters of the
two-parameter gamma distribution ( 0),

3.627

0.002341 1/427.2
The fitted two-parameter gamma distribution has a
coefficient of skewness of 1.05, whereas the fitted three-
parameter gamma reproduces the sample skew of 0.712.
As occurred with the three-parameter lognormal distribu-
tion, the estimated lower bound for the three-parameter
gamma distribution is negative (

735.6), resulting
in a three-parameter model that has a smaller skew
coefficient than was obtained with the corresponding
two-parameter model. The reciprocal of

is also
reported. While

has inverse x-units, 1/

is a natural
scale parameter that has the same units as x and thus can
be easier to interpret.
Studies by Thom (1958) and Matalas and Wallis
(1973) have shown that maximum likelihood parameter
estimates are superior to the moment estimates. For the
two-parameter gamma distribution, Greenwood and
Durand (1960) give approximate formulas for the
maximum likelihood estimates (also Haan, 1977).
However, the maximum likelihood estimators are often
not used in practice because they are very sensitive to the
smallest observations that sometimes suffer from
measurement error and other distortions.

x
s
X
2

( ) x
s
X
2
2
Figure 7.4. The gamma distribution function for various
values of the shape parameter .
2.5
2.0
1.5
1.0
0.5
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
x
f
(
x
)
E
0
2
0
5
2
7
f

= 0.50
= 1.00
= 8.00
= 2.00

wrm_ch07.qxd 8/31/2005 12:18 PM Page 188

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 189
When plotting the observed and fitted quantiles of a
gamma distribution, an approximation to the inverse of
the distribution function is often useful. For | | 3, the
WilsonHilferty transformation
(7.86)
gives the quantiles x
G
of the gamma distribution in terms
of x
N
, the quantiles of the standard-normal distribution.
Here , , and are the mean, standard deviation,
and coefficient of skewness of x
G
. Kirby (1972) and
Chowdhury and Stedinger (1991) discuss this and other
more complicated but more accurate approximations.
Fortunately the availability of excellent approximations
of the gamma cumulative distribution function and its
inverse in Microsoft Excel and other packages has
reduced the need for such simple approximations.
3.5. Log-Pearson Type 3 Distribution
The log-Pearson type 3 distribution (LP3) describes a
random variable whose logarithms have a Pearson type 3
distribution. This distribution has found wide use in
modelling flood frequencies and has been recommended
for that purpose (IACWD, 1982). Bobe (1975) and
Bobe and Ashkar (1991) discuss the unusual shapes that
this hybrid distribution may take allowing negative
values of . The LP3 distribution has a probability density
function given by
(7.87)
with 0, and either positive or negative. For 0,
values are restricted to the range 0 x exp(). For
0, values have a lower bound so that exp() X. Figure
7.5 illustrates the probability density function for the LP3
distribution as a function of the skew of the P3 distribu-
tion describing ln(X), with
lnX
0.3. The LP3 density
function for | | 2 can assume a wide range of shapes
with both positive and negative skews. For | | 2, the
log-space P3 distribution is equivalent to an exponential
distribution function, which decays exponentially as x
moves away from the lower bound ( 0) or upper
f x
x
x x
X
( )
| |
( )
[ (ln( ) )] exp{ (ln( ) )}

1

x
x
G
N

2
1
6 36
2
2
3

_
,

1
]
1
1
bound ( 0): as a result the LP3 distribution has a
similar shape. The space with 1 may be more
realistic for describing variables whose probability density
function becomes thinner as x takes on large values. For
0, the two-parameter lognormal distribution is obtained
as a special case.
The LP3 distribution has mean and variance
(7.88)
For 0 2, the variance is infinite.
These expressions are seldom used, but they do
reveal the character of the distribution. Figures 7.6 and
7.7 provide plots of the real-space coefficient of skew-
ness and coefficient of variation of a log-Pearson type 3
variate X as a function of the standard deviation
Y
and coefficient of skew
Y
of the log-transformation
Y ln(X). Thus the standard deviation
Y
and skew
Y
of Y are in log space. For
Y
0, the log-Pearson type 3
distribution reduces to the two-parameter lognormal
distribution discussed above, because in this case Y has
a normal distribution. For the lognormal distribution,
the standard deviation
Y
serves as the sole shape
parameter, and the coefficient of variation of X for small

Y
is just
Y
. Figure 7.7 shows that the situation is more

X
X

e
e
f
1
2 1
2 2
2

_
,

_
,

_
,

oor or 2 0 , .
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 4.0
x
E
0
2
0
5
2
7
g

=
=
=
=
3.5
=
+2
+1
- 1
0
- 2
f (x)
Figure 7.5. Log-Pearson type 3 probability density functions
for different values of coefficient of skewness .
wrm_ch07.qxd 8/31/2005 12:18 PM Page 189
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
complicated for the LP3 distribution. However, for
small
Y
, the coefficient of variation of X is approxi-
mately
Y
.
Again, the flood flow data in Table 7.2 can be used to
illustrate parameter estimation. Using natural logarithms,
one can estimate the log-space moments with the stan-
dard estimators in Equations 7.39 that yield:

7.202

0.5625

0.337
For the LP3 distribution, analysis generally focuses on
the distribution of the logarithms Y ln(X) of the
flows, which would have a Pearson type 3 distribution
with moments
Y
,
Y
and
Y
(IACWD, 1982; Bobe
and Ashkar, 1991). As a result, flood quantiles are calcu-
lated as
x
p
exp{
Y

Y
K
p
[
Y
]} (7.89)
where K
p
[
Y
] is a frequency factor corresponding to
cumulative probability p for skewness coefficient
Y
.
(K
p
[
Y
] corresponds to the quantiles of a three-parameter
gamma distribution with zero mean, unit variance, and
skewness coefficient
Y
.)
Since 1967 the recommended procedure for flood
frequency analysis by federal agencies in the United States
has used this distribution. Current guidelines in Bulletin
17B (IACWD, 1982) suggest that the skew
Y
be
190 Water Resources Systems Planning and Management
estimated by a weighted average of the at-site sample
skewness coefficient and a regional estimate of the
skewness coefficient. Bulletin 17B also includes tables of
frequency factors, a map of regional skewness estimators,
checks for low outliers, confidence interval formula, a
discussion of expected probability and a weighted-
moments estimator for historical data.
3.6. Gumbel and GEV Distributions
The annual maximum flood is the largest flood flow
during a year. One might expect that the distribution of
annual maximum flood flows would belong to the set of
extreme value distributions (Gumbel, 1958; Kottegoda
and Rosso, 1997). These are the distributions obtained in
the limit, as the sample size n becomes large, by taking the
largest of n independent random variables. The Extreme
Value (EV) type I distribution, or Gumbel distribution,
has often been used to describe flood flows. It has the
cumulative distribution function:
F
X
(x) exp{exp[ (x )/]} (7.90)
where is the location parameter. It has a mean and
variance of

X
0.5772

X
2

2

2
/6 1.645
2
(7.91)
Figure 7.6. Real-space coefficient of skewness
X
for LP3
distributed X as a function of log-space standard deviation

Y
and coefficient of skewness
Y
where Y ln(X).
12
10
8
6
4
2
0
-2
-1.6 -1.2 -0.8 -0.4 0.0 0.4 0.8 1.6
log-space coefficient of skewness
Y
E
0
2
0
5
2
7
h
1.2
real-space coefficient
of skewness
X
0.60
0.50
0.40
0.30
0.20
0.10
0.05 l
o
g
-
s
p
a
c
e

s
t
a
n
d
a
r
d
d
e
v
i
a
t
i
o
n

Y
Figure 7.7. Real-space coefficient of variation CV
X
for LP3
distributed X as a function of log-space standard deviation
Y
and coefficient of skewness
Y
where Y ln(X).
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
-1.6 -1.2 -0.8 -0.4 0.0 0.4 0.8 1.6
log-space coefficient of skewness
Y
E
0
2
0
5
2
7
j
1.2
0.60
0.50
0.40
0.30
0.20
0.10
0.05
real-space coefficient
of variation CV
X
log-space standard
deviation
Y
wrm_ch07.qxd 8/31/2005 12:18 PM Page 190
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 191
Its skewness coefficient has a fixed value equal to

X
1.1396.
The generalized extreme value (GEV) distribution is
a general mathematical expression that incorporates the
type I, II, and III extreme value (EV) distributions for
maxima (Gumbel, 1958; Hosking et al., 1985). In recent
years it has been used as a general model of extreme
events including flood flows, particularly in the context of
regionalization procedures (NERC, 1975; Stedinger and
Lu, 1995; Hosking and Wallis, 1997). The GEV distribu-
tion has the cumulative distribution function:
F
X
(x) exp{[1 (x)/]
1/
} for 0 (7.92)
From Equation 7.92, it is clear that for 0 (the typical
case for floods), x must exceed /, whereas for
0, x must be no greater than / (Hosking and
Wallis, 1987). The mean, variance, and skewness coeffi-
cient are (for 1/3):

X
(/) [1 (1)],

X
2
(/)
2
{(1 2) [(1 )]
2
} (7.93)

X
(Sign ){(1 3) 3(1 ) (1 2)
2[(1 )]
3
}/{(1 2) [(1 )]
2
}
3/2
where (1) is the classical gamma function. The
Gumbel distribution is obtained when 0. For ||
0.3, the general shape of the GEV distribution is similar
to the Gumbel distribution, though the right-hand tail is
thicker for 0, and thinner for 0, as shown in
Figures 7.8 and 7.9.
The parameters of the GEV distribution are easily
computed using L-moments and the relationships
(Hosking et al. (1985):
7.8590c 2.9554c
2

2
/[(1 )(1 2

)] (7.94)

1
(/)[(1) 1]
where
c 2
2
/(
3
3
2
) ln(2)/ln(3)
[2/(
3
3)] ln(2)/ln(3)
As one can see, the estimator of the shape parameter
will depend only upon the L-skewness estimator

3
. The
estimator of the scale parameter will then depend
on the estimate of and of
2
. Finally, one must also use
the sample mean
1
(Equation 7.48) to determine the
estimate of the location parameter .
Using the flood data in Table 7.2 and the sample
L-moments computed in Section 2, one obtains first
c0.000896 which yields

0.007036,

1,165.20
and

657.29.
The small value of the fitted parameter means that
the fitted distribution is essentially a Gumbel distribution.
Again, is a location parameter, not a lower bound, so
its value resembles a reasonable x value.
Madsen et al. (1997a) show that moment estimators
can provide more precise quantile estimators. Martins and
Figure 7.8. GEV density distributions for selected shape
parameter values.
0.20
0.15
0.10
0.05
0.00
0 10 15 20 25 30
x
f
(
x
)
E
0
2
0
5
2
7
k

=
=
=
-
-
0.3
0.1
0.1
5
Figure 7.9. Right-hand tails of GEV distributions shown in
Figure 7.8.
0.04
0.03
0.02
0.01
0.00
12 17 22 27
x
f
(
x
)
E
0
2
0
5
2
7
m

=
=
=
-
-
0.3
0.1
0.1
wrm_ch07.qxd 8/31/2005 12:18 PM Page 191
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Stedinger (2001b) found that with occasional uninforma-
tive samples, the MLE estimator of could be entirely
unrealistic resulting in absurd quantile estimators.
However the use of a realistic prior distribution on
yielded generalized maximum likelihood estimators
(GLME) that performed better than moment and
L-moment estimators over the range of of interest.
The generalized maximum likelihood estimators
(GMLE) are obtained by maximizing the log-likelihood
function, augmented by a prior density function on .
A prior distribution that reflects general world-wide
geophysical experience and physical realism is in the form
of a beta distribution:
() (p) (q) (0.5 )
p1
(0.5 )
q1
/(pq)
(7.95)
for 0.5 0.5 with p 6 and q 9. Moreover,
this prior assigns reasonable probabilities to the values of
within that range. For outside the range 0.4 to
0.2 the resulting GEV distributions do not have density
functions consistent with flood flows and rainfall (Martins
and Stedinger, 2000). Other estimators implicitly have
similar constraints. For example, L-moments restricts
to the range 1, and the method of moments
estimator employs the sample standard deviation so that
0.5. Use of the sample skew introduces the
constraint that 0.3. Then given a set of independ-
ent observations {x
1
, , x
n
} drawn for a GEV distribu-
tion, the generalized likelihood function is:
(7.96)
For feasible values of the parameters, y
i
is greater than 0
(Hosking et al., 1985). Numerical optimization of the
generalized likelihood function is often aided by the
additional constraint that min{y
1
, , y
n
} for some
small 0 so as to prohibit the search generating infea-
sible values of the parameters for which the likelihood
function is undefined. The constraint should not be bind-
ing at the final solution.
The data in Table 7.2 again provide a convenient data
set for illustrating parameter estimators. The L-moment
ln{ ( , , | , , )} ln( )
ln( ) ( )
L

x x n
y y
n
i i
1
1
1
1

_
,

1
]]
1

i
n
i i
y x

1
1
ln[ ( )]
[ ( / )( )]


with
192 Water Resources Systems Planning and Management
estimators were used to generate an initial solution.
Numerical optimization of the likelihood function in
Equation 7.96 yielded the maximum likelihood estima-
tors of the GEV parameters:

0.0359,

1165.4 and

620.2.
Similarly, use of the geophysical prior (Equation 7.95)
yielded the generalized maximum likelihood estimators

0.0823,

1150.8 and

611.4. Here the

record length of forty years is too short to define reliably
the shape parameter so that result of using the prior is
to pull slightly toward the mean of the prior. The other
two parameters adjust accordingly.
3.7. L-Moment Diagrams
Section 3 presented several families of distributions. The
L-moment diagram in Figure 7.10 illustrates the relation-
ships between the L-kurtosis (
4
) and L-skewness (
3
) for a
number of distributions often used in hydrology. It shows
that distributions with the same coefficient of skewness
still differ in the thickness of their tails. This thickness
is described by their kurtosis. Tail shapes are important if
an analysis is sensitive to the likelihood of extreme events.
The normal and Gumbel distributions have a fixed
shape and thus are presented by single points that fall on
the Pearson type 3 (P3) curve for 0, and the general-
ized extreme value (GEV) curve for 0, respectively.
The L-kurtosis/L-skewness relationships for the two-
parameter and three-parameter gamma or P3 distribu-
tions are identical, as they are for the two-parameter and
three-parameter lognormal distributions. This is because
the addition of a location parameter does not change
the range of fundamental shapes that can be generated.
However, for the same skewness coefficient, the lognor-
mal distribution has a larger kurtosis than the gamma or
P3 distribution, and thus assigns larger probabilities to
the largest events.
As the skewness of the lognormal and gamma distribu-
tions approaches zero, both distributions become normal
and their kurtosis/skewness relationships merge. For the
same L-skewness, the L-kurtosis of the GEV distribution is
generally larger than that of the lognormal distribution.
For positive yielding almost symmetric or even nega-
tively skewed GEV distributions, the GEV has a smaller
kurtosis than the three-parameter lognormal distribution.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 192
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 193
The latter can be negatively skewed when the log normal
location parameter is used as an upper bound.
Figure 7.10 also includes the three-parameter general-
ized Pareto distribution, whose cdf is:
F
X
(x) 1 [1 (x )/]
1/
(7.97)
(Hosking and Wallis, 1997). For 0 it corresponds
to the exponential distribution (gamma with 1).
This point is where the Pareto and P3 distribution
L-kurtosis/L-skewness lines cross. The Pareto distribution
becomes increasing more skewed for 0, which is the
range of interest in hydrology. The generalized Pareto
distribution with 0 is often used to describe peaks-
over-a-threshold and other variables whose probability
density function has its maximum at their lower bound.
In that range for a given L-skewness, the Pareto distribu-
tion always has a larger kurtosis than the gamma distri-
bution. In these cases the parameter for the gamma
distribution would need to be in the range 0 1, so
that both distributions would be J-shaped.
As shown in Figure 7.10, the GEV distribution has a
thicker right-hand tail than either the gamma/Pearson
type 3 distribution or the lognormal distribution.
4. Analysis of Censored Data
There are many instances in water resources planning
where one encounters censored data. A data set is
censored if the values of observations that are outside a
specified range of values are not specifically reported
(David, 1981). For example, in water quality investiga-
tions many constituents have concentrations that are
reported as T, where T is a reliable detection threshold
(MacBerthouex and Brown, 2002). Thus the concentra-
tion of the water quality variable of interest was too small
to be reliably measured. Likewise, low-flow observations
and rainfall depths can be rounded to or reported as zero.
Several approaches are available for analysis of censored
data sets, including probability plots and probability-plot
regression, conditional probability models and maximum
likelihood estimators (Haas and Scheff, 1990; Helsel,
1990; Kroll and Stedinger, 1996; MacBerthouex and
Brown, 2002).
Historical and physical paleoflood data provide
another example of censored data. Before the beginning
of a continuous measurement program on a stream or
river, the stages of unusually large floods can be
estimated on the basis of the memories of people who
have experienced these events and/or physical markings
in the watershed (Stedinger and Baker, 1987). Annual
maximum floods that were not unusual were not
recorded nor were they remembered. These missing data
are censored data. They cover periods between occa-
sional large floods that have been recorded or that have
left some evidence of their occurrence (Stedinger and
Cohn, 1986).
The discussion below addresses probability-plot
methods for use with censored data. Probability-plot
methods have a long history of use for this purpose
because they are relatively simple to use and to under-
stand. Moreover, recent research has shown that they
are relatively efficient when the majority of values are
observed, and unobserved values are known only to be
below (or above) some detection limit or perception
threshold that serves as a lower (or upper) bound. In such
cases, probability-plot regression estimators of moments
and quantiles are as accurate as maximum likelihood
estimators. They are almost as good as estimators
computed with complete samples (Helsel and Cohn,
1988; Kroll and Stedinger, 1996).
Perhaps the simplest method for dealing with censored
data is adoption of a conditional probability model. Such
models implicitly assume that the data are drawn from
one of two classes of observations: those below a single
threshold, and those above the threshold. This model is
Figure 7.10. Relationships between L-skewness and
L-kurtosis for various distributions.
0.8
0.6
0.4
0.2
0.0
-0.2 0.0 0.2 0.4 0.6 0.8
l-skewness
l
-
k
u
r
t
o
s
i
s
Normal
P3
LN
Gumbel
GEV
Pareto
E
0
2
0
5
2
7
b
wrm_ch07.qxd 8/31/2005 12:18 PM Page 193
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
appropriate for simple cases where censoring occurs
because small observations are recorded as zero, as often
happens with low-flow, low pollutant concentration,
and some flood records. The conditional probability
model introduces an extra parameter P
0
to describe
the probability that an observation is zero. If r of a
total of n observations were observed because they
exceeded the threshold, then P
0
is estimated as (n r)/n.
A continuous distribution G
X
(x) is derived for the
strictly positive non-zero values of X. Then the
parameters of the G distribution can be estimated
using any procedure appropriate for complete uncen-
sored samples. The unconditional cdf for any value x 0,
is then
F
X
(x) P
0
( 1 P
0
) G(x) (7.98)
This model completely decouples the value of P
0
from the
parameters that describe the G distribution.
Section 7.2 discusses probability plots and plotting
positions useful for graphical displays of data to allow
a visual examination of the empirical frequency curve.
Suppose that among n samples a detection limit is
exceeded by the observations r times. The natural estima-
tor of the exceedance probability P
0
of the perception
threshold is again (n r)/n. If the r values that exceeded
the threshold are indexed by i 1, , r, wherein x
(r)
is
the largest observation, reasonable plotting positions
within the interval [P
0
, 1] are:
p
i
P
0
(1 P
0
) [(i a)/(r 1 2a)] (7.99)
where a defines the plotting position that is used; a 0
is reasonable (Hirsch and Stedinger, 1987). Helsel and
Cohn (1988) show that reasonable choices for a generally
make little difference. Both papers discuss development
of plotting positions when there are different thresholds,
as occurs when the analytical precision of instrumenta-
tion changes over time. If there are many exceedances of
the threshold so that r (1 2a), p
i
is indistinguish-
able from
p
i
[i (n r) a]/(n 1 2a) (7.100)
where again, i 1, , r. These values correspond to the
plotting positions that would be assigned to the largest r
observations in a complete sample of n values.
The idea behind the probability-plot regression
estimators is to use the probability plot for the observed
data to define the parameters of the whole distribution.
And if a sample mean, sample variance or quantiles are
needed, then the distribution defined by the probability
plot is used to fill in the missing (censored) observations
so that standard estimators of the mean, standard
deviation and of quantiles can be employed. Such fill-in
procedures are efficient and relatively robust for fitting
a distribution and estimating various statistics with
censored water quality data when a modest number of the
smallest observations are censored (Helsel, 1990; Kroll
and Stedinger, 1996).
Unlike the conditional probability approach, here the
below threshold probability P
0
is linked with the selected
probability distribution for the above-threshold observa-
tions. The observations below the threshold are censored
but are in all other respects envisioned as coming from
the same distribution that is used to describe the observed
above-threshold values.
When water quality data are well described by a
lognormal distribution, available values ln[X
(1)
]

ln[X
(r)
] can be regressed upon F
1
[p
i
] F
1
[p
i
]
for i 1, , r, where the r largest observations in a
sample of size n are available. If regression yields constant
m and slope s corresponding to the first and second
population moments and , a good estimator of the pth
quantile is
x
p
exp[m sz
p
] (7.101)
wherein z
p
is the p
th
quantile of the standard normal
distribution. To estimate sample means and other
statistics one can fill in the missing observations with
x(j) exp{y(j)} for j 1, , (n r) (7.102)
where
y(j) m sF
1
{P
0
[(j a)/(n r 1 2a)]} (7.103)
Once a complete sample is constructed, standard estima-
tors of the sample mean and variance can be calculated,
as can medians and ranges. By filling in the missing small
observations, and then using complete-sample estimators
of statistics of interest, the procedure is relatively insensi-
tive to the assumption that the observations actually have
a lognormal distribution.
Maximum likelihood estimators are quite flexible, and
are more efficient than plotting-position methods when
the values of the observations are not recorded because
they are below or above the perception threshold (Kroll
and Stedinger, 1996). Maximum likelihood methods
194 Water Resources Systems Planning and Management
wrm_ch07.qxd 8/31/2005 12:18 PM Page 194
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 195
allow the observations to be represented by exact values,
ranges and various thresholds that either were or were
not exceeded at various times. This can be particularly
important with historical flood data sets because the
magnitudes of many historical floods are not recorded
precisely, and it may be known that a threshold was
never crossed or was crossed at most once or twice in a
long period (Stedinger and Cohn, 1986; Stedinger, 2000;
OConnell et al., 2002). Unfortunately, maximum likeli-
hood estimators for the LP3 distribution have proven
to be problematic. However, recently developed expected
moment estimators seem to do as well as maximum
likelihood estimators with the LP3 distribution (Cohn
et al., 1997, 2001; Griffs et al., 2004).
While often a computational challenge, maximum
likelihood estimators for complete samples, and samples
with some observations censored, pose no conceptual
challenge. One need only write the maximum likelihood
function for the data and proceed to seek the parameter
values that maximize that function. Thus if F(x | ) and
f(x | ) are the cumulative distribution and probability
density functions that should describe the data, and
is the vector of parameters of the distribution, then for
the case described above wherein x
1
, , x
r
are r of n
observations that exceeded a threshold T, the likelihood
function would be (Stedinger and Cohn, 1986):
L( | r, n, x
1
, , x
r
)
F(T | )
(nr)
f(x
1
| )f(x
2
| )

f(x
r
| ) (7.104)
Here, (n r) observations were below the threshold T,
and the probability an observation is below T is F(T | )
which then appears in Equation 7.104 to represent that
observation. In addition, the specific values of the r obser-
vations x
1
, , x
r
are available, where the probability an
observation is in a small interval of width around x
i
is
f(x
i
| ). Thus strictly speaking the likelihood function also
includes a term
r
. Here what is known of the magnitude
of all of the n observations is included in the likelihood
function in the appropriate way. If all that were known of
some observation was that it exceeded a threshold M, then
that value should be represented by a term [1 F(M | )]
in the likelihood function. Similarly, if all that were known
was that the value was between L and M, then a term
[F(M | ) F(L | )] should be included in the likelihood
function. Different thresholds can be used to describe
different observations corresponding to changes in the
quality of measurement procedures. Numerical methods
can be used to identify the parameter vector that maxi-
mizes the likelihood function for the data available.
5. Regionalization and Index-Flood
Method
Research has demonstrated the potential advantages
of index flood procedures (Lettenmaier et al., 1987;
Stedinger and Lu, 1995; Hosking and Wallis, 1997;
Madsen, and Rosbjerg, 1997a). The idea behind the
index-flood approach is to use the data from many
hydrologically similar basins to estimate a dimensionless
flood distribution (Wallis, 1980). Thus this method
substitutes space for time by using regional information
to compensate for having relatively short records at each
site. The concept underlying the index-flood method is
that the distributions of floods at different sites in a
region are the same except for a scale or index-flood
parameter that reflects the size, rainfall and runoff
characteristics of each watershed. Research is revealing
when this assumption may be reasonable. Often a more
sophisticated multi-scaling model is appropriate (Gupta
and Dawdy, 1995a; Robinson and Sivapalan, 1997).
Generally the mean is employed as the index flood.
The problem of estimating the p
th
quantile x
p
is then
reduced to estimation of the mean,
x
, for a site and the
ratio x
p
/
x
of the p
th
quantile to the mean. The mean can
often be estimated adequately with the record available at
a site, even if that record is short. The indicated ratio is
estimated using regional information. The British Flood
Studies Report (NERC, 1975) calls these normalized flood
distributions growth curves.
Key to the success of the index-flood approach is
identification of sets of basins that have similar coeffi-
cients of variation and skew. Basins can be grouped
geographically, as well as by physiographic characteristics
including drainage area and elevation. Regions need not
be geographically contiguous. Each site can potentially be
assigned its own unique region consisting of sites with
which it is particularly similar (Zrinji and Burn, 1994), or
regional regression equations can be derived to compute
normalized regional quantiles as a function of a sites
physiographic characteristics and other statistics (Fill and
Stedinger, 1998).
wrm_ch07.qxd 8/31/2005 12:18 PM Page 195
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Clearly the next step for regionalization procedures,
such as the index-flood method, is to move away from
estimates of regional parameters that do not depend upon
basin size and other physiographic parameters. Gupta et al.
(1994) argue that the basic premise of the index-flood
method that the coefficient of variation of floods is
relatively constant is inconsistent with the known
relationships between the coefficient of variation (CV)
and drainage area (see also Robinson and Sivapalan,
1997). Recently, Fill and Stedinger (1998) built such a
relationship into an index-flood procedure by using a
regression model to explain variations in the normalized
quantiles. Tasker and Stedinger (1986) illustrated how
one might relate log-space skew to physiographic basin
characteristics (see also Gupta and Dawdy, 1995b).
Madsen and Rosbjerg (1997b) did the same for a regional
model of for the GEV distribution. In both studies, only
a binary variable representing region was found useful in
explaining variations in these two shape parameters.
Once a regional model of alternative shape parameters
is derived, there may be some advantage to combining
such regional estimators with at-site estimators employing
an empirical Bayesian framework or some other weighting
schemes. For example, Bulletin 17B recommends weigh
at-site and regional skewness estimators, but almost cer-
tainly places too much weight on the at-site values
(Tasker and Stedinger, 1986). Examples of empirical
Bayesian procedures are provided by Kuczera (1982),
Madsen and Rosbjerg (1997b) and Fill and Stedinger
(1998). Madsen and Rosbjergs (1997b) computation of a
-model with a New Zealand data set demonstrates how
important it can be to do the regional analysis carefully,
taking into account the cross-correlation among concur-
rent flood records.
When one has relatively few data at a site, the index-
flood method is an effective strategy for deriving flood
frequency estimates. However, as the length of the
available record increases, it becomes increasingly advan-
tageous to also use the at-site data to estimate the coeffi-
cient of variation as well. Stedinger and Lu (1995) found
that the L-moment/GEV index-flood method did quite
well for humid regions (CV 0.5) when n 25, and for
semi-arid regions (CV 1.0) for n 60, if reasonable
care is taken in selecting the stations to be included
in a regional analysis. However, with longer records, it
became advantageous to use the at-site mean and L-CV
196 Water Resources Systems Planning and Management
with a regional estimator of the shape parameter for a
GEV distribution. In many cases this would be roughly
equivalent to fitting a Gumbel distribution corresponding
to a shape parameter 0. Gabriele and Arnell (1991)
develop the idea of having regions of different sizes for
different parameters. For realistic hydrological regions,
these and other studies illustrate the value of regionaliz-
ing estimators of the shape, and often the coefficient of
variation of a distribution.
6. Partial Duration Series
Two general approaches are available for modelling
flood and precipitation series (Langbein, 1949). An
annual maximum series considers only the largest event
in each year. A partial duration series (PDS) or peaks-
over-threshold (POT) approach includes all independent
peaks above a truncation or threshold level. An objection
to using annual maximum series is that it employs only
the largest event in each year, regardless of whether the
second-largest event in a year exceeds the largest events of
other years. Moreover, the largest annual flood flow in a
dry year in some arid or semi-arid regions may be zero, or
so small that calling them floods is misleading. When
considering rainfall series or pollutant discharge events,
one may be interested in modelling all events that occur
within a year that exceed some threshold of interest.
Use of a partial duration series framework avoids such
problems by considering all independent peaks that exceed
a specified threshold. Furthermore, one can estimate annual
exceedance probabilities from the analysis of partial
duration series. Arguments in favour of partial duration
series are that relatively long and reliable records are often
available, and if the arrival rate for peaks over the threshold
is large enough (1.65 events/year for the Poisson-arrival
with exponential-exceedance model), partial duration series
analyses should yield more accurate estimates of extreme
quantiles than the corresponding annual-maximum fre-
quency analyses (NERC, 1975; Rosbjerg, 1985). However,
when fitting a three-parameter distribution, there seems to
be little advantage from the use of a partial duration series
approach over an annual maximum approach. This is true
even when the partial duration series includes many more
peaks than the maximum series because both contain the
same largest events (Martins and Stedinger, 2001a).
wrm_ch07.qxd 8/31/2005 12:18 PM Page 196
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 197
A drawback of partial duration series analyses is that
one must have criteria to identify only independent
peaks (and not multiple peaks corresponding to the same
event). Thus, such analysis can be more complicated than
analyses using annual maxima. Partial duration models,
perhaps with parameters that vary by season, are often
used to estimate expected damages from hydrological
events when more than one damage-causing event can
occur in a season or within a year (North, 1980).
A model of a partial duration series has at least two
components: first, one must model the arrival rate of events
larger than the threshold level; second, one must model
the magnitudes of those events. For example, a Poisson
distribution has often been used to model the arrival of
events, and an exponential distribution to describe the
magnitudes of peaks that exceed the threshold.
There are several general relationships between the
probability distribution for annual maximum and the fre-
quency of events in a partial duration series. For a partial
duration series model, let be the average arrival rate of
flood peaks greater than the threshold x
0
and let G(x) be
the probability that flood peaks, when they occur, are less
than x x
0
, and thus those peaks fall in the range [x
0
, x].
The annual exceedance probability for a flood, denoted
1/T
a
, corresponding to an annual return period T
a
,
is related to the corresponding exceedance probability
q
e
[1 G(x)] for level x in the partial duration series by
1/T
a
1 exp{q
e
} 1 exp{1/T
p
} (7.105)
where T
p
1/(q
e
) is the average return period for level x
in the partial duration series.
Many different choices for G(x) may be reasonable. In
particular, the generalized Pareto distribution (GPD) is a
simple distribution useful for describing floods that exceed
a specified lower bound. The cumulative distribution
function for the generalized three-parameter Pareto
distribution is:
F
X
(x) 1 [1 (x )/]
1/
(7.106)
with mean and variance

X
/(1)

X
2

2
/[(1)
2
(12)] (7.107)
where for 0, x , whereas for 0, x
/ (Hosking and Wallis, 1987). A special case of
the GPD is the two-parameter exponential distribution
with 0. Method of moment estimators work relatively
well (Rosbjerg et al., 1992).
Use of a generalized Pareto distribution for G(x) with
a Poisson arrival model yields a GEV distribution for
the annual maximum series greater than x
0
(Smith, 1984;
Stedinger et al., 1993; Madsen et al., 1997a). The Poisson-
Pareto and Poisson-GPD models provide very reasonable
descriptions of flood risk (Rosbjerg et al., 1992). They
have the advantage that they focus on the distribution of
the larger flood events, and regional estimates of the GEV
distributions shape parameter from annual maximum
and partial duration series analyses can be used inter-
changeably.
Madsen and Rosbjerg (1997a) use a Poisson-GPD
model as the basis of a partial duration series index-flood
procedure. Madsen et al. (1997b) show that the estima-
tors are fairly efficient. They pooled information from
many sites to estimate the single shape parameter and
the arrival rate where the threshold was a specified
percentile of the daily flow duration curve at each site.
Then at-site information was used to estimate the mean
above-threshold flood. Alternatively, one could use the
at-site data to estimate the arrival rate as well.
7. Stochastic Processes and
Time Series
Many important random variables in water resources are
functions whose values change with time. Historical
records of rainfall or streamflow at a particular site are a
sequence of observations called a time series. In a time
series, the observations are ordered by time, and it is
generally the case that the observed value of the random
variable at one time influences ones assessment of the
distribution of the random variable at later times. This
means that the observations are not independent. Time
series are conceptualized as being a single observation
of a stochastic process, which is a generalization of the
concept of a random variable.
This section has three parts. The first presents the con-
cept of stationarity and the basic statistics generally used
to describe the properties of a stationary stochastic
process. The second presents the definition of a Markov
process and the Markov chain model. Markov chains are
wrm_ch07.qxd 8/31/2005 12:18 PM Page 197
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
a convenient model for describing many phenomena, and
are often used in synthetic flow and rainfall generation
and optimization models. The third part discusses the
sampling properties of statistics used to describe the
characteristics of many time series.
7.1. Describing Stochastic Processes
A random variable whose value changes through time
according to probabilistic laws is called a stochastic
process. An observed time series is considered to be one
realization of a stochastic process, just as a single
observation of a random variable is one possible value the
random variable may assume. In the development here, a
stochastic process is a sequence of random variables {X(t)}
ordered by a discrete time variable t 1, 2, 3,
The properties of a stochastic process must generally
be determined from a single time series or realization.
To do this, several assumptions are usually made. First,
one generally assumes that the process is stationary. This
means that the probability distribution of the process is
not changing over time. In addition, if a process is strictly
stationary, the joint distribution of the random variables
X(t
1
), , X(t
n
) is identical to the joint distribution of
X(t
1
t), , X(t
n
t) for any t; the joint distribution
depends only on the differences t
i
t
j
between the times
of occurrence of the events.
For a stationary stochastic process, one can write the
mean and variance as

X
E[X(t)] (7.108)
and

2
Var[X(t)] (7.109)
Both are independent of time t. The autocorrelations, the
correlation of X with itself, are given by

X
(k) Cov[X(t), X(t k)]/
X
2
(7.110)
for any positive integer time lag k. These are the
statistics most often used to describe stationary stochastic
processes.
When one has available only a single time series, it is
necessary to estimate the values of
X
,
X
2
, and
X
(k) from
values of the random variable that one has observed. The
mean and variance are generally estimated essentially as
they were in Equation 7.39.
198 Water Resources Systems Planning and Management

X
(7.111)

X
2
(7.112)
while the autocorrelations
X
(k) for any time lag k can be
estimated as ( Jenkins and Watts, 1968)

X
(k)
(7.113)
The sampling distribution of these estimators depends on
the correlation structure of the stochastic process giving
rise to the time series. In particular, when the observa-
tions are positively correlated, as is usually the case in
natural streamflows or annual benefits in a river basin
simulation, the variances of the estimated x
_
and

X
2
are
larger than would be the case if the observations were
independent. It is sometimes wise to take this inflation
into account. Section 7.3 discusses the sampling distribu-
tion of these statistics.
All of this analysis depends on the assumption of
stationarity, for only then do the quantities defined in
Equations 7.108 to 7.110 have the intended meaning.
Stochastic processes are not always stationary. Agricultural
and urban development, deforestation, climatic variability
and changes in regional resource management can alter
the distribution of rainfall, streamflows, pollutant concen-
trations, sediment loads and groundwater levels over time.
If a stochastic process is not essentially stationary over
the time span in question, then statistical techniques that
rely on the stationary assumption do not apply and the
problem generally becomes much more difficult.
7.2. Markov Processes and Markov Chains
A common assumption in many stochastic water
resources models is that the stochastic process X(t) is
a Markov process. A first-order Markov process has the
property that the dependence of future values of the
process on past values depends only on the current value
and not on previous values or observations. In symbols
for k 0,

r
x x x x
x x
k
t k t
t
T k
t
t
T
( )( )
( )
1
2
1

1
2
1
T
X X
t
t
T
( )

X
T
X
t
t
T
1
1

wrm_ch07.qxd 8/31/2005 12:18 PM Page 198

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 199
F
X
[X(t k) | X(t), X(t 1), X(t 2), ]
F
X
[X(t k) | X(t)] (7.114)
For Markov processes, the current value summarizes the
state of the processes. As a consequence, the current value
of the process is often referred to as the state. This makes
physical sense as well when one refers to the state or level
of an aquifer or reservoir.
A special kind of Markov process is one whose
state X(t) can take on only discrete values. Such a process
is called a Markov chain. Often in water resources
planning, continuous stochastic processes are approxi-
mated by Markov chains. This is done to facilitate the
construction of simple stochastic models. This section
presents the basic notation and properties of Markov
chains.
Consider a stream whose annual flow is to be
represented by a discrete random variable. Assume that
the distribution of streamflows is stationary. In the
following development, the continuous random variable
representing the annual streamflows (or some other
process) is approximated by a random variable Q
y
in
year y, which takes on only n discrete values q
i
(each
value representing a continuous range or interval of
possible streamflows) with unconditional probabilities
p
i
where
(7.115)
It is frequently the case that the value of Q
y1
is not
independent of Q
y
. A Markov chain can model such
dependence. This requires specification of the transition
probabilities p
ij
,
p
ij
Pr[Q
y1
q
j
| Q
y
q
i
] (7.116)
A transition probability is the conditional probability that
the next state is q
j
, given that the current state is q
i
. The
transition probabilities must satisfy
(7.117)
Figure 7.11 shows a possible set of transition probabilities
in a matrix. Each element p
ij
in the matrix is the probabil-
ity of a transition from streamflow q
i
in one year to
streamflow q
j
in the next. In this example, a low flow
tends to be followed by a low flow, rather than a high
flow, and vice versa.
p i
ij
j
n

1
1
for all

p
i
i
n

1
1

Figure 7.11. Matrix (above)

and histograms (below) of
streamflow transition
probabilities showing probability
of streamflow q
j
(represented by
index j) in year y1 given
streamflow q
i
(represented by
index i) in year y.
0.4 0.3 0.2 0.1
P
j
1
0.2 0.4 0.3 0.1
P
j
2
0.1 0.3 0.4 0.2
1 2 3 4
j
P
j
3
0.0 0.2 0.3 0.5
1 2 3 4
j
P
j
4
E
0
2
0
5
2
7
q
wrm_ch07.qxd 8/31/2005 12:18 PM Page 199
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Let P be the transition matrix whose elements are p
ij
.
For a Markov chain, the transition matrix contains all
the information necessary to describe the behaviour of
the process. Let p
i
y
be the probability that the streamflow
Q
y
is q
i
(in state i) in year y. Then the probability that
Q
y1
q
j
is the sum of the products of the probabilities
p
i
y
that Q
y
q
i
times the probability p
ij
that the next state
Q
y1
is q
j
given that Q
y
q
i
. In symbols, this relationship
is written:
(7.118)
Letting p
y
be the row vector of state resident probabilities
p
i
y
, , p
n
y
, this relationship may be written
p
(y1)
p
(y)
P (7.119)
To calculate the probabilities of each streamflow state in
year y 2, one can use p
(y1)
in Equation 7.119 to obtain
p
(y2)
p
(y1)
P or p
(y2)
p
y
P
2
Continuing in this manner, it is possible to compute
the probabilities of each possible streamflow state for
years y 1, y 2, y 3, , y k, as
p
(yk)
p
y
P
k
(7.120)
Returning to the four-state example in Figure 7.11,
assume that the flow in year y is in the interval repre-
sented by q
2
. Hence in year y the unconditional stream-
flow probabilities p
i
y
are (0, 1, 0, 0). Knowing each p
i
y
,
the probabilities p
j
y1
corresponding to each of the four
streamflow states can be determined. From Figure 7.11,
the probabilities p
j
y1
are 0.2, 0.4, 0.3 and 0.1 for
j 1, 2, 3 and 4, respectively. The probability vectors for
nine future years are listed in Table 7.8.
As time progresses, the probabilities generally reach
limiting values. These are the unconditional or steady-state
probabilities. The quantity p
i
has been defined as the
unconditional probability of q
i
. These are the steady-state
probabilities which p
(yk)
approaches for large k. It is clear
from Table 7.8 that as k becomes larger, Equation 7.118
becomes
(7.121)
or in vector notation, Equation 7.119 becomes
p pP (7.122)
p p p
j i i
j
i
n

p p p p p p p p p
j
y y
j
y
j n
y
nj i
y
ij
i
n

1
1 1 2 2
1
. . .

200 Water Resources Systems Planning and Management

where p is the row vector of unconditional probabilities
(p
1
, , p
n
). For the example in Table 7.8, the probability
vector p equals (0.156, 0.309, 0.316, 0.219).
The steady-state probabilities for any Markov
chain can be found by solving simultaneous Equation
7.122 for all but one of the states j together with the
constraint
(7.123)
Annual streamflows are seldom as highly correlated
as the flows in this example. However, monthly, weekly
and especially daily streamflows generally have high
serial correlations. Assuming that the unconditional
steady-state probability distributions for monthly
streamflows are stationary, a Markov chain can be
defined for each months streamflow. Since there are twelve
months in a year, there would be twelve transition
matrices, the elements of which could be denoted as
p
ij
t
. Each defines the probability of a streamflow q
j
t1
in month t 1, given a streamflow q
i
t
in month t. The
steady-state stationary probability vectors for each month
can be found by the procedure outlined above, except
that now all twelve matrices are used to calculate all
twelve steady-state probability vectors. However, once
the steady-state vector p is found for one month, the
others are easily computed using Equation 7.120 with t
replacing y.
p
i
i
n

1
1

Table 7.8. Successive streamflow probabilities based on

transition probabilities in Figure 7.11.
E
0
2
1
1
0
1
h
y
y + 1
y + 2
y + 3
y + 4
y + 5
y + 6
y + 7
y + 8
y + 9
0.000
0.200
0.190
0.173
0.163
0.159
0.157
0.156
0.156
0.156
1.000
0.400
0.330
0.316
0.312
0.310
0.309
0.309
0.309
0.309
0.000
0.300
0.310
0.312
0.314
0.315
0.316
0.316
0.316
0.316
0.000
0.100
0.170
0.199
0.211
0.216
0.218
0.219
0.219
0.219
year P
y
P
4
y
P
3
y
P
2
y
1
wrm_ch07.qxd 8/31/2005 12:18 PM Page 200
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 201
7.3. Properties of Time-Series Statistics
The statistics most frequently used to describe the
distribution of a continuous-state stationary stochastic
process are the sample mean, variance and various
autocorrelations. Statistical dependence among the obser-
vations, as is frequently found in time series, can have a
marked effect on the distribution of these statistics. This
part of Section 7 reviews the sampling properties of
these statistics when the observations are a realization
of a stochastic process.
The sample mean
(7.124)
when viewed as a random variable is an unbiased estimate
of the mean of the process
X
, because
(7.125)
However, correlation among the X
i
s, so that
X
(k) 0 for
k 0, affects the variance of the estimated mean X
_
.
(7.126)
The variance of X
_
, equal to
x
2
/n for independent obser-
vations, is inflated by the factor within the brackets. For

X
(k) 0, as is often the case, this factor is a nondecreas-
ing function of n, so that the variance of X
_
is inflated by a
factor whose importance does not decrease with increas-
ing sample size. This is an important observation, because
it means that the average of a correlated time series will be
less precise than the average of a sequence of independ-
ent random variables of the same length with the same
variance.
A common model of stochastic series has

X
(k) [
X
(1)]
k

k
(7.127)
This correlation structure arises from the autoregressive
Markov model discussed at length in Section 8. For this
correlation structure
Var( ) X X
n
X X
n
X
t X s X
s
n
t
n
X

E[(
E

) ]
( )( )
2
2
1 1
2
1

11 2 1
1
1

k
n
k
X
k
n

_
,

( )
E E[ ] [ ] X
n
X
i
i
n
X

1
1

X
n
X
i
i
n

1
1

(7.128)
Substitution of the sample estimates for
X
2
and
X
(1) in
the equation above often yields a more realistic estimate
of the variance of X
_
than does the estimate s
X
2
/n if the
correlation structure
X
(k)
k
is reasonable; otherwise,
Equation 7.126 may be employed. Table 7.9 illustrates
the effect of correlation among the X
t
values on the
standard error of their mean, equal to the square root of
the variance in Equation 7.126.
The properties of the estimate of the variance of X,

X
2
(7.129)
are also affected by correlation among the X
t
s. Here v
rather than s is used to denote the variance estimator,
because n is employed in the denominator rather than
n 1. The expected value of v
x
2
becomes
(7.130)
The bias in v
X
2
depends on terms involving
X
(1) through

X
(n 1). Fortunately, the bias in v
X
2
decreases with n
and is generally unimportant when compared to its
variance.
Correlation among the X
t
s also affects the variance of
v
X
2
. Assuming that X has a normal distribution (here the
variance of v
X
2
depends on the fourth moment of X), the
variance of v
X
2
for large n is approximately (Kendall and
Stuart, 1966, Sec. 48.1).
(7.131) Var( ) v
n
k
X
X
X
k
2
4
2
1
2 1 2

( )

E[ ] v
n n
k
n
k
X X X
k
n
2 2
1
1
1
1 2
1

_
,

( )

v
n
X X
X t
t
n
2 2
1
1
( )

Var( ) X
n n
n
X
n

2
2
1
2 1 1
1
[ ( ) ( )]
( )

Table 7.9. Standard error of X

_
when
x
0.25 and
X
(k)
k
.
E
0
2
1
1
0
1
j
25
50
100
0.050
0.035
0.025
0.067
0.048
0.034
0.096
0.069
0.050
sampl e
size n
correlation of consecutive observations
= 0.0 p = 0.3 p = 0.6 p
wrm_ch07.qxd 8/31/2005 12:18 PM Page 201
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
where for
X
(k)
k
, Equation 7.131 becomes
(7.132)
Like the variance of X
_
, the variance of v
X
2
is inflated by a
factor whose importance does not decrease with n. This
is illustrated by Table 7.10 which gives the standard
deviation of v
X
2
divided by the true variance
X
2
as a
function of n and when the observations have a normal
distribution and
X
(k)
k
. This would be the coefficient
of variation of v
X
2
were it not biased.
A fundamental problem of time-series analyses is
the estimation or description of the relationship among
the random variable values at different times. The
statistics used to describe this relationship are the
autocorrelations. Several estimates of the autocorrelations
have been suggested. A simple and satisfactory estimate
recommended by Jenkins and Watts (1968) is:

X
(k)
(7.133)
Here r
k
is the ratio of two sums where the numerator
contains n k terms and the denominator contains n
terms. The estimate r
k
is biased, but unbiased estimates
frequently have larger mean square errors ( Jenkins and
Watts, 1968). A comparison of the bias and variance of r
1
is provided by the case when the X
t
s are independent
normal variates. Then (Kendall and Stuart, 1966)
(7.134a)
and
(7.134b)
For n 25, the expected value of r
1
is 0.04 rather
than the true value of zero; its standard deviation is 0.19.
This results in a mean square error of (E[r
1
])
2
Var(r
1
)
0.0016 0.0353 0.0369. Clearly, the variance of
r
1
is the dominant term.
For X
t
values that are not independent, exact expres-
sions for the variance of r
k
generally are not available.
Var( ) r
n
n n n
1
2
2
2
1
1

( )
( )

E[ ] r
n
1
1

r
x x x x
x
k
t t k
t
n k
t
t
n
( )( )
( )
1
2
1

x
Var( ) v
n
X
X 2
4 2
2
2
1
1

_
,

202 Water Resources Systems Planning and Management

However, for normally distributed X
t
and large n (Kendall
and Stuart, 1966),
(7.135)
If
X
(k) is essentially zero for k q, then the simpler
expression (Box et al., 1994)
(7.136)
is valid for r
k
corresponding to k q. Thus for large n,
Var(r
k
) l/n and values of r
k
will frequently be outside
the range of 1.65/
,
even though
X
(k) may be zero.
If
X
(k)
k
, Equation 7.136 reduces to
(7.137)
In particular for r
l
, this gives
(7.138)
Approximate values of the standard deviation of r
l
for
different values of n and are given in Table 7.11.
The estimates of r
k
and r
kj
are highly correlated for
small j; this causes plots of r
k
versus k to exhibit slowly
varying cycles when the true values of
X
(k) may be zero.
This increases the difficulty of interpreting the sample
autocorrelations.
Var( ) r
n
1
2
1
1 ( )
Var( ) r
k

1
2
2 2
2
2
n
l l
l
k
k
k
( )( )

1
]
1
n
Var( ) r
n
l
k X
t
q

1
1 2
2
1

( )

1
]
1
Var( )
4
r
n
l l k l k
k l k l
k X X X
l
X X X

1
2
[ ( ) ( ) ( )
( ) ( ) ( )

2
2 2

X X
k l ( ) ( )]
E
0
2
1
1
0
1
k
25
50
100
0.28
0.20
0.14
0.31
0.22
0.15
0.41
0.29
0.21
sampl e
size n
correlation of consecutive observations
= 0.0 p = 0.3 p = 0.6 p
Table 7.10. Standard deviation of v
X
2
/
X
2
when observations
have a normal distribution and
X
(k)
k
.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 202
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 203
8. Synthetic Streamflow Generation
8.1. Introduction
This section is concerned primarily with ways of generat-
ing sample data such as streamflows, temperatures and
rainfall that are used in water resource systems simulation
studies (e.g., as introduced in the next section). The mod-
els and techniques discussed in this section can be used
to generate any number of quantities used as inputs to
simulation studies. For example Wilks (1998, 2002) dis-
cusses the generation of wet and dry days, rainfall depths
on wet days and associated daily temperatures. The dis-
cussion here is directed toward the generation of stream-
flows because of the historical development and frequent
use of these models in that context (Matalas and Wallis,
1976). In addition, they are relatively simple compared to
more complete daily weather generators and many other
applications. Generated streamflows have been called
synthetic to distinguish them from historical observations
(Fiering, 1967). The activity has been called stochastic
hydrological modelling. More detailed presentations can
be found in Marco et al. (1989) and Salas (1993).
River basin simulation studies can use many sets of
streamflow, rainfall, evaporation and/or temperature
sequences to evaluate the statistical properties of the
performance of alternative water resources systems. For
this purpose, synthetic flows and other generated quanti-
ties should resemble, statistically, those sequences that are
likely to be experienced during the planning period.
Figure 7.12 illustrates how synthetic streamflow, rainfall
and other stochastic sequences are used in conjunction
with projections of future demands and other economic
data to determine how different system designs and
operating policies might perform.
Use of only the historical flow or rainfall record in
water resource studies does not allow for the testing of
alternative designs and policies against the range of
sequences that are likely to occur in the future. We can be
very confident that the future historical sequence of flows
will not be the historical one, yet there is important infor-
mation in that historical record. That information is not
fully used if only the historical sequence is simulated.
Fitting continuous distributions to the set of historical
flows and then using those distributions to generate other
sequences of flows, all of which are statistically similar
and equally weighted, gives one a broader range of inputs
to simulation models. Testing designs and policies against
that broader range of flow sequences that could occur
more clearly identifies the variability and range of possi-
ble future performance indicator values. This in turn
should lead to the selection of more robust system designs
and policies.
The use of synthetic streamflows is particularly useful
for water resources systems having large amounts of
over-year storage. Use of only the historical hydrological
record in system simulation yields only one time history
of how the system would operate from year to year. In
water resources systems with relatively little storage, so
that reservoirs and/or groundwater aquifers refill almost
every year, synthetic hydrological sequences may not be
needed if historical sequences of a reasonable length are
E
0
2
1
1
0
1
m
25
50
100
0.20
0.14
0.10
0.19
0.13
0.095
0.16
0.11
0.080
sampl e
size n
correlation of consecutive observations
= 0. 0 p = 0. 3 p = 0. 6 p
Table 7.11. Approximate standard deviation of r
1
when
observations have a normal distribution and
X
(k)
k
.
Figure 7.12. Structure of a simulation study, indicating the
transformation of a synthetic streamflow sequence, future
demands and a system design and operating policy into
system performance statistics.
E
0
2
0
5
2
7
p
simulation model
of river basin system
synthetic streamflow
and other sequences
future demands
and economic data
system design
and operating policy
system
performance
wrm_ch07.qxd 8/31/2005 12:18 PM Page 203
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
available. In this second case, a twenty-five-year historical
record provides twenty-five descriptions of the possible
within-year operation of the system. This may be suffi-
cient for many studies.
Generally, use of stochastic sequences is thought to
improve the precision with which water resources system
performance indices can be estimated, and some studies
have shown this to be the case (Vogel and Shallcross,
1996; Vogel and Stedinger, 1988). In particular, if system
operation performance indices have thresholds and sharp
breaks, then the coarse descriptions provided by histori-
cal series are likely to provide relative inaccurate estimates
of the expected values of such statistics. For example,
suppose that shortages only invoke non-linear penalties
on average one year in twenty. Then in a sixty-year
simulation there is a 19% probability that the penalty will
be invoked at most once, and an 18% probability it will
be invoked five or more times. Thus the calculation of the
annual average value of the penalty would be highly unre-
liable unless some smoothing of the input distributions is
allowed, associated with a long simulation analysis.
On the other hand, if one is only interested in the
mean flow, or average benefits that are mostly a linear
function of flows, then use of stochastic sequences will
probably add little information to what is obtained simply
by simulating the historical record. After all, the fitted
models are ultimately based on the information provided
in the historical record, and their use does not produce
new information about the hydrology of the basin.
If in a general sense one has available n years of record,
the statistics of that record can be used to build a stochastic
model for generating thousands of years of flow. These
synthetic data can then be used to estimate more exactly
the system performance, assuming, of course, that the
flow-generating model accurately represents nature. But
the initial uncertainty in the model parameters resulting
from having only n years of record would still remain
(Schaake and Vicens, 1980).
An alternative is to run the historical record (if it is
sufficiently complete at every site and contains no gaps of
missing data) through the simulation model to generate n
years of output. That output series can be processed to
produce estimates of system performance. So the question
is the following: Is it better to generate multiple input
series based on uncertain parameter values and use those
to determine average system performance with great
204 Water Resources Systems Planning and Management
precision, or is it sufficient just to model the n-year output
series that results from simulation of the historical series?
The answer seems to depend upon how well behaved
the input and output series are. If the simulation model is
linear, it does not make much difference. If the simulation
model were highly non-linear, then modelling the input
series would appear to be advisable. Or if one is develop-
ing reservoir operating policies, there is a tendency to
make a policy sufficiently complex to deal very well with
the few droughts in the historical record, giving a false
sense of security and likely misrepresenting the probabil-
ity of system performance failures.
Another situation where stochastic data generating
models are useful is when one wants to understand the
impact on system performance estimates of the parameter
uncertainty stemming from short historical records. In
that case, parameter uncertainty can be incorporated into
streamflow generating models so that the generated
sequences reflect both the variability that one would
expect in flows over time as well as the uncertainty of
the parameter values of the models that describe that
variability (Valdes et al., 1977; Stedinger and Taylor,
1982a,b; Stedinger et al., 1985; Vogel and Stedinger,
1988).
If one decides to use a stochastic data generator, the
challenge is to use a model that appropriately describes
the important relationships, but does not attempt to
reproduce more relationships than are justified or can be
estimated with available data sets.
Two basic techniques are used for streamflow genera-
tion. If the streamflow population can be described by a
stationary stochastic process (a process whose parameters
do not change over time), and if a long historical stream-
flow record exists, then a stationary stochastic streamflow
model may be fit to the historical flows. This statistical
model can then generate synthetic sequences that
describe selected characteristics of the historical flows.
Several such models are discussed below.
The assumption of stationarity is not always plausible,
particularly in river basins that have experienced marked
changes in runoff characteristics due to changes in land
cover, land use, climate or the use of groundwater during
the period of flow record. Similarly, if the physical
characteristics of a basin change substantially in the
future, the historical streamflow record may not provide
reliable estimates of the distribution of future unregulated
wrm_ch07.qxd 8/31/2005 12:18 PM Page 204
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 205
flows. In the absence of the stationarity of streamflows or
a representative historical record, an alternative scheme is
to assume that precipitation is a stationary stochastic
process and to route either historical or synthetic precip-
itation sequences through an appropriate rainfallrunoff
model of the river basin.
8.2. Streamflow Generation Models
The first step in the construction of a statistical stream-
flow generating model is to extract from the historical
streamflow record the fundamental information about the
joint distribution of flows at different sites and at different
times. A streamflow model should ideally capture what
is judged to be the fundamental characteristics of the
joint distribution of the flows. The specification of what
characteristics are fundamental is of primary importance.
One may want to model as closely as possible the
true marginal distribution of seasonal flows and/or the
marginal distribution of annual flows. These describe
both how much water may be available at different times
and also how variable is that water supply. Also, model-
ling the joint distribution of flows at a single site in
different months, seasons and years may be appropriate.
The persistence of high flows and of low flows, often
described by their correlation, affects the reliability with
which a reservoir of a given size can provide a given
yield (Fiering, 1967; Lettenmaier and Burges, 1977a,
1977b; Thyer and Kuczera, 2000). For multi-component
reservoir systems, reproduction of the joint distribution of
flows at different sites and at different times will also be
important.
Sometimes, a streamflow model is said to resemble
statistically the historical flows if it produces flows with
the same mean, variance, skew coefficient, autocorrela-
tions and/or cross-correlations as were observed in the
historical series. This definition of statistical resemblance
is attractive because it is operational and only requires an
analyst to only find a model that can reproduce the
observed statistics. The drawback of this approach is that
it shifts the modelling emphasis away from trying to find
a good model of marginal distributions of the observed
flows and their joint distribution over time and over
space, given the available data, to just reproducing
arbitrarily selected statistics. Defining statistical resem-
blance in terms of moments may also be faulted for
specifying that the parameters of the fitted model should
be determined using the observed sample moments, or
their unbiased counterparts. Other parameter estimation
techniques, such as maximum likelihood estimators, are
often more efficient.
Definition of resemblance in terms of moments can
also lead to confusion over whether the population
parameters should equal the sample moments, or whether
the fitted model should generate flow sequences whose
sample moments equal the historical values. The two
concepts are different because of the biases (as discussed
in Section 7) in many of the estimators of variances and
correlations (Matalas and Wallis, 1976; Stedinger, 1980,
1981; Stedinger and Taylor, 1982a).
For any particular river basin study, one must
determine what streamflow characteristics need to be
modelled. The decision should depend on what character-
istics are important to the operation of the system being
studied, the available data, and how much time can be
spared to build and test a stochastic model. If time permits,
it is good practice to see if the simulation results are in fact
sensitive to the generation model and its parameter values
by using an alternative model and set of parameter values.
If the models results are sensitive to changes, then, as
always, one must exercise judgement in selecting the
appropriate model and parameter values to use.
This section presents a range of statistical models for
the generation of synthetic data. The necessary sophisti-
cation of a data-generating model depends on the
intended use of the generated data. Section 8.3 below
presents the simple autoregressive Markov model for
generating annual flow sequences. This model alone is
too simple for many practical studies, but is useful for
illustrating the fundamentals of the more complex models
that follow. It seems, therefore, worth some time explor-
ing the properties of this basic model.
Subsequent sections discuss how flows with any
marginal distribution can be produced, and present
models for generating sequences of flows that can
reproduce the persistence of historical flow sequences.
Other parts of this section present models for generating
concurrent flows at several sites and for generating
seasonal or monthly flows that preserve the characteristics
of annual flows. More detailed discussions for those wish-
ing to study synthetic streamflow models in greater depth
can be found in Marco et al. (1989) and Salas (1993).
wrm_ch07.qxd 8/31/2005 12:18 PM Page 205
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
8.3. A Simple Autoregressive Model
A simple model of annual streamflows is the autoregres-
sive Markov model. The historical annual flows q
y
are
thought of as particular values of a stationary stochastic
process Q
y
. The generation of annual streamflows and
other variables would be a simple matter if annual flows
were independently distributed. In general, this is not the
case and a generating model for many phenomena should
capture the relationship between values in different years
or in other time periods. A common and reasonable
assumption is that annual flows are the result of a first-
order Markov process (as discussed in Section 7.2).
Assume for now that annual streamflows are normally
distributed. In some areas the distribution of annual flows
is in fact nearly normal. Streamflow models that produce
non-normal streamflows are discussed as an extension of
this simple model.
The joint normal density function of two streamflows
Q
y
and Q
w
in years y and w having mean , variance
2
,
and year-to-year correlation between flows is
(7.139)
The joint normal distribution for two random variables
with the same mean and variance depend only on their
common mean , variance
2
, and the correlation
between the two (or equivalently the covariance
2
).
The sequential generation of synthetic streamflows
requires the conditional distribution of the flow in
one year given the value of the flows in previous years.
However, if the streamflows are a first-order (lag 1)
Markov process, then the distribution of the flow in year
y 1 depends entirely on the value of the flow in year y.
In addition, if the annual streamflows have a multivariate
normal distribution, then the conditional distribution of
Q
y1
is normal with mean and variance
E[Q
y1
Q
y
q
y
] (q
y
) (7.140)
Var(Q
y1
Q
y
q
y
)
2
(1
2
)
where q
y
is the value of the random variable Q
y
in year y.
This relationship is illustrated in Figure 7.13. Notice that
f q , q
q q q q
y w
y y w w
( )
exp
( )( )

1
2 1
2
2
2 2 0 5
2 2

( )
( ) ( )
.
22 2
1 ( )

1
]
1
206 Water Resources Systems Planning and Management
the larger the absolute value of the correlation between
the flows, the smaller the conditional variance of Q
y1
,
which in this case does not depend at all on the value q
y
.
Synthetic normally distributed streamflows that have
mean , variance
2
, and year-to-year correlation , are
produced by the model
(7.141)
where V
y
is a standard normal random variable, meaning
that it has zero mean, E[V
y
] 0, and unit variance,
. The random variable V
y
is added here to
provide the variability in Q
y1
that remains even after Q
y
is known. By construction, each V
y
is independent of past
flows Q
w
where w y, and V
y
is independent of V
w
for
w y. These restrictions imply that
E[V
w
V
y
] 0 for w y (7.142)
and
E[(Q
w
)V
y
] 0 for w y (7.143)
Clearly, Q
y1
will be normally distributed if both Q
y
and
V
y
are normally distributed because sums of independent
normally distributed random variables are normally
distributed.
E V
y
2
1

1
]

Q Q
y y y
V
+1
( ) 1
2
E
0
4
1
0
1
1
a
q
y
flow in year y
1 -
2

[ + (q
y
)]
Q
y
Q
y + 1
[Q
y + 1
|Q =q
y
] E
f
l
o
w

i
n

y
e
a
r

y
+
1
y
Figure 7.13. Conditional distribution of Q
y1
given Q
y
q
y
for
two normal random variables.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 206
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 207
It is a straightforward procedure to show that this
basic model indeed produces streamflows with the speci-
fied moments, as demonstrated below.
Using the fact that E[V
y
] 0, the conditional mean of
Q
y1
given that Q
y
equals q
y
is
(7.144)
Since E[V
y
2
] Var[V
y
] 1, the conditional variance of
Q
y1
is
(7.145)
Thus, this model produces flows with the correct condi-
tional mean and variance.
To compute the unconditional mean of Q
y1
one first
takes the expectation of both sides of Equation 7.141 to
obtain
(7.146)
where E[V
y
] 0. If the distribution of streamflows is
independent of time so that for all y, E[Q
y1
] E[Q
y
]
E[Q], it is clear that (1 ) E[Q] (1 ) or
E[Q] (7.147)
Alternatively, if Q
y
for y 1 has mean , then Equation
7.146 indicates that Q
2
will have mean . Thus repeated
application of the Equation 7.146 would demonstrate
that all Q
y
for y 1 have mean .
The unconditional variance of the annual flows can
be derived by squaring both sides of Equation 7.141 to
obtain
(7.148)
E[( ) ] E[{ ( )
E[( ) ]
1
2
2
Q Q
Q
y y y
y
V

1
2 2
2
} ]
2
E[( ) ] (1 )E
2 2
[ ]


1
2
2

Q
y y y
V V
E[ ] (E[ ] ) E[ ]
1
Q Q
y y y
V

1
2
Var E
E
[ | ] E[{ [ | ]} | ]
[{ (
1 1 1
2
Q Q Q
y y y y y y
y
q q q
q


))
[ ( )]}
[ ] (1
2
2 2


V q
V
y y
y


1
1
2
2
E )
2

E[
|
] E[ ( )
( )
+1 Qy y y y
y
q q V
q




1
2
]
Because V
y
is independent of Q
y
(Equation 7.143), the
second term on the right-hand side of Equation 7.148
vanishes. Hence the unconditional variance of Q satisfies
E[(Q
y1
)
2
]
2
E[(Q
y
)
2
]
2
(1
2
)
(7.149)
Assuming that Q
y1
and Q
y
have the same variance yields
E[(Q )
2
]
2
(7.150)
so that the unconditional variance is
2
, as required.
Again, if one does not want to assume that Q
y1
and Q
y
have the same variance, a recursive argument can be
adopted to demonstrate that if Q
1
has variance
2
, then
Q
y
for y 1 has variance
2
.
The covariance of consecutive flows is another
important issue. After all, the whole idea of building these
time-series models is to describe the year-to-year correla-
tion of the flows. Using Equation 7.141 one can show that
the covariance of consecutive flows must be
(7.151)
where E[(Q
y
)V
y
] 0 because V
y
and Q
y
are inde-
pendent (Equation 7.143).
Over a longer time scale, another property of this model
is that the covariance of flows in year y and y k is
E[(Q
yk
)(Q
y
)]
k

2
(7.152)
This equality can be proven by induction. It has already
been shown for k 0 and 1. If it is true for k j 1, then
(7.153)
where E[(Q
y
)V
yj1
] 0 for j 1. Hence Equation
7.152 is true for any value of k.
It is important to note that the results in Equations
7.144 to 7.152 do not depend on the assumption that the
random variables Q
y
and V
y
are normally distributed.
These relationships apply to all autoregressive Markov
processes of the form in Equation 7.141 regardless of the
E[( )( )] E{ ( )
(
1
1
Q Q Q
Q
y j y y
j
y
j
y
V

[
] 1
2

)}
E[( )]( )]
[ ]
1
2 2
Q Q
y y
j
j
j 1
E[( )( )] E{[ ( )
]( )}
E
1
Q Q Q
Q
y y y
y y
V

1
2

[[( ) ]
2 2
Q
y

wrm_ch07.qxd 8/31/2005 12:18 PM Page 207
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
distributions of Q
y
and V
y
. However, if the flow Q
y
in year
y 1 is normally distributed with mean and variance

2
, and if the V
y
are independent normally distributed
random variables with mean zero and unit variance, then
the generated Q
y
for y 1 will also be normally distrib-
uted with mean and variance
2
. The next section
considers how this and other models can be used to
generate streamflows that have other than a normal
distribution.
8.4. Reproducing the Marginal Distribution
Most models for generating stochastic processes deal
directly with normally distributed random variables.
Unfortunately, flows are not always adequately described
by the normal distribution. In fact, streamflows and many
other hydrological data cannot really be normally distrib-
uted because of the impossibility of negative values. In
general, distributions of hydrological data are positively
skewed, having a lower bound near zero and, for practi-
cal purposes, an unbounded right-hand tail. Thus they
look like the gamma or lognormal distribution illustrated
in Figures 7.3 and 7.4.
The asymmetry of a distribution is often measured by
its coefficient of skewness. In some streamflow models,
the skew of the random elements V
y
is adjusted so that
the models generate flows with the desired mean,
variance and skew coefficient. For the autoregressive
Markov model for annual flows
(7.154)
so that
(7.155)
By appropriate choice of the skew of V
y
, E[V
y
3
], the
desired skew coefficient of the annual flows can be
produced. This method has often been used to generate
flows that have approximately a gamma distribution by
using V
y
s with a gamma distribution and the required
skew. The resulting approximation is not always adequate
(Lettenmaier and Burges, 1977a).

Q
Q

E[( ) ] ( )
/ 3
3
2 3 2
3
3
1
1
E[ ] Vy
E[( ) ] E ( )
E[( ) ]
(1
1
3
3
3
3 2
[ ] Q Q
Q
y y
y

V
y
1
2

3
)) E
3/2
[ ] V
y
3
208 Water Resources Systems Planning and Management
The alternative and generally preferred method is to
generate normal random variables and then transform
these variates to streamflows with the desired marginal
distribution. Common choices for the distribution of
streamflows are the two-parameter and three-parameter
lognormal distributions or a gamma distribution. If Q
y
is
a lognormally distributed random variable, then
Q
y
exp(X
y
) (7.156)
where X
y
is a normal random variable. When the lower
bound is zero, Q
y
has a two-parameter lognormal
distribution. Equation 7.156 transforms the normal
variates X
y
into lognormally distributed streamflows. The
transformation is easily inverted to obtain
X
y
ln(Q
y
) for Q
y
(7.157)
where Q
y
must be greater than its lower bound .
The mean, variance, skewness of X
y
and Q
y
are related
by the formulas (Matalas, 1967)
(7.158)
If normal variates X
s
y
and
X
y
u
are used to generate lognor-
mally distributed streamflows Q
s
y
and
Q
y
u
at sites s and u,
then the lag-k correlation of the Q
y
s, denoted
Q
(k; s, u),
is determined by the lag-k correlation of the X variables,
denoted
X
(k; s, u), and their variances
x
2
(s) and
x
2
(u),
where
(7.159)
The correlations of the X
y
s
can be adjusted, at least in
theory, to produce the observed correlations among the
Q
y
s
variates. However, more efficient estimates of the true
correlation of the Q
y
s
values are generally obtained by
transforming the historical flows q
y
s
into their normal
equivalent and using the historical
correlations of these X
y
s
values as estimators of
X
(k; s, u)
(Stedinger, 1981).
x q
y
s
y
s
ln( )

Q
( ; , )
exp[ ( ; , ) ( ) ( )]
exp ( ) exp { [ ] } {
/
k s u
k s u s u
s
X X
X

X
1
1
2 1 2
[[ ] } ( )
/

X
u
2 1 2
1

Q
Q
Q
exp
exp 2 exp ( ) ( )


X
X
1
2
1
2
2 2 2
X
X X

[ ]

3 where exp
3
[
exp exp
exp
( ) ( )
( )
(
[ ]
/
3 3 2
1
2 2
2 3 2

X X
X

X
2 1/2
1] )
wrm_ch07.qxd 8/31/2005 12:18 PM Page 208
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 209
Some insight into the effect of this logarithmic
transformation can be gained by considering the resulting
model for annual flows at a single site. If the normal
variates follow the simple autoregressive Markov model
(7.160)
then the corresponding Q
y
follow the model (Matalas,
1967)
Q
y1
D
y
{exp[
x
(1
x
)]}(Q
y
)
x
(7.161)
where
(7.162)
The conditional mean and standard deviation of Q
y1
given that Q
y
q
y
now depends on (q
y

)
x
. Because
the conditional mean of Q
y1
is no longer a linear function
of q
y
, (as shown in Figure 7.14), the streamflows are said
to exhibit differential persistence: low flows are now more
likely to follow low flows than high flows are to follow
high flows. This is a property often attributed to real
streamflow distributions. Models can be constructed
to capture the relative persistence of wet and dry
periods (Matalas and Wallis, 1976; Salas, 1993; Thyer
and Kuczera, 2000). Many weather generators for
D V y
X
y exp[(1 ) ]
1/2

X
2
X X V
y X y y X

1
( ) 1
2
X
precipitation and temperature include such tendencies by
employing a Markov chain description of the occurrence
of wet and dry days (Wilks, 1998).
8.5. Multivariate Models
If long concurrent streamflow records can be constructed
at the several sites at which synthetic streamflows are
desired, then ideally a general multi-site streamflow model
could be employed. OConnell (1977), Ledolter (1978),
Salas et al. (1980) and Salas (1993) discuss multivariate
models and parameter estimation. Unfortunately, identifi-
cation of the most appropriate model structure is very
difficult for general multivariate models.
This section illustrates how the basic univariate
annual-flow model in Section 8.3 can be generalized to the
multivariate case. This exercise reveals how multivariate
models can be constructed to reproduce specified variances
and covariances of the flow vectors of interest, or some
transformation of those values. This multi-site generaliza-
tion of the annual AR(1) or autoregressive Markov model
follows the approach taken by Matalas and Wallis (1976).
This general approach can be further extended to generate
multi-site/multi-season modelling procedures, as is done in
Section 8.6, employing what have been called disaggrega-
tion models. However, while the size of the model matrices
and vectors increases, the model is fundamentally the same
from a mathematical viewpoint. Hence this section starts
with the simpler case of an annual flow model.
For simplicity of presentation and clarity, vector notation
is employed. Let Z
y
Z
y
1
, , Z
y
n

T
be the column vector of
transformed zero-mean annual flows at sites s 1, 2, , n,
so that
(7.163)
In addition, let V
y
V
y
1
, , V
y
n

T
be a column vector
of standard-normal random variables, where V
y
s
is
independent of V
w
r
for (r, w) (s, y) and independent
of past flows Z
w
r
where y w. The assumption that
the variables have zero mean implicitly suggests that the
mean value has already been subtracted from all the
variables. This makes the notation simpler and eliminates
the need to include a constant term in the models. With
all the random variables having zero mean, one can focus
on reproducing the variances and covariances of the
vectors included in a model.
E[ ] Z
y
s
0
E
0
4
1
0
1
1
b
flow in year y
f
l
o
w

i
n

y
e
a
r
y

+
1
Q
y
Q
y

+
1
2.5
2.0
1.5
1.0
0.5
0.0
0.0 0.5 1.0 1.5 2.0 2.5
[Q
y + 1
|q
y
] E
90% probability
range for Q
y+1
parameters

Q
Q
Q
= 1.00
= 0.25
= 0.50

X
X
X
= 0.030
= 0.246
= 0.508

Figure 7.14. Conditional mean of Q
y1
given Q
y
q
y
and 90%
probability range for the value of Q
y1
.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 209
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A sequence of synthetic flows can be generated by the
model
Z
y1
AZ
y
BV
y
(7.164)
where A and B are (n n) matrices whose elements are
chosen to reproduce the lag 0 and lag 1 cross-covariances
of the flows at each site. The lag 0 and lag 1 covariances
and cross-covariances can most economically be manipu-
lated by use of the two matrices S
0
and S
1
. The lag-zero
covariance matrix, denoted S
0
, is defined as
(7.165)
and has elements
(7.166)
The lag-one covariance matrix, denoted S
1
, is defined as
(7.167)
and has elements
(7.168)
The covariances do not depend on y because the stream-
flows are assumed to be stationary.
Matrix S
1
contains the lag 1 covariances and lag 1 cross-
covariances. S
0
is symmetric because the cross-covariance
S
0
(i, j) equals S
0
(j, i). In general, S
1
is not symmetric.
The variancecovariance equations that define the
values of A and B in terms of S
0
and S
1
are obtained by
manipulations of Equation 7.164. Multiplying both sides
of that equation by Z
y
T
and taking expectations yields
(7.169)
The second term on the right-hand side vanishes because
the components of Z
y
and V
y
are independent. Now the
first term in Equation 7.169, E[AZ
y
Z
y
T
], is a matrix whose
(i, j)th element equals
(7.170)
The matrix with these elements is the same as the matrix
AE Z
y
Z
y
T
.
Hence, A the matrix of constants can be pulled
through the expectation operator just as is done in the
E E a
y
k
k
ik y
j
k
n
ik y
k
y
j
n
a Z Z Z Z

1 1

1
]
1
[ ]
E[ E[ E[
T T T
Z Z Z Z V Z
y y y y y y

1
] ] ] A B
S ( ) E[ 1 i, j

Z Z
y y 1
i
j
]
S 1
T
1
E[ ] Zy
y
Z
S ( , ) E[
0
i j
y y
Z Z
i
j
]
S E[ ]
0
T
Z Z y
y
210 Water Resources Systems Planning and Management
scalar case where E[aZ
y
b] aE[Z
y
] b for fixed
constants a and b.
Substituting S
0
and S
1
for the appropriate expectations
in Equation 7.169 yields
S
1
AS
0
or A S
1
S
0
1
(7.171)
A relationship to determine the matrix B is obtained by
multiplying both sides of Equation 7.164 by its own
transpose (this is equivalent to squaring both sides of
the scalar equation a b) and taking expectations to
obtain
E[Z
y1
Z
T
y1
] E[AZ
y
Z
T
y
A
T
] E[AZ
y
V
T
y
B
T
]
E[BV
y
Z
y
A
T
] E[BV
y
V
T
y
B
T
] (7.172)
The second and third terms on the right-hand side of
Equation 7.172 vanish because the components of Z
y
and
V
y
are independent and have zero mean. E[V
y
V
y
T
] equals
the identity matrix because the components of V
y
are
independently distributed with unit variance. Thus
S
0
AS
0
A
T
BB
T
(7.173)
Solving for the B matrix, one finds that it should satisfy
the quadratic equation
BB
T
S
0
AS
0
A
T
S
0
S
1
S
0
1
S
1
T
(7.174)
The last equation results from substitution of the rela-
tionship for A given in Equation 7.171 and the fact that
S
0
is symmetric; hence, S
0
1
is symmetric.
It should not be too surprising that the elements of
B are not uniquely determined by Equation 7.174. The
components of the random vector V
y
may be combined in
many ways to produce the desired covariances as long as
B satisfies Equation 7.174. A lower triangular matrix that
satisfies Equation 7.174 can be calculated by Cholesky
decomposition (Young, 1968; Press et al., 1986).
Matalas and Wallis (1976) call Equation 7.164 the lag-
1 model. They do not call it the Markov model because
the streamflows at individual sites do not have the
covariances of an autoregressive Markov process given
in Equation 7.152. They suggest an alternative model
for what they call the Markov model. It has the same
structure as the lag-1 model except it does not preserve
the lag-1 cross-covariances. By relaxing this requirement,
they obtain a simpler model with fewer parameters
that generates flows that have covariances of an autore-
gressive Markov process at each site. In their Markov
model, the new A matrix is simply a diagonal matrix,
wrm_ch07.qxd 8/31/2005 12:18 PM Page 210
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 211
whose diagonal elements are the lag-1 correlations of
flows at each site:
A diag[(1; i, i)] (7.175)
where (1; i, i) is the lag-one correlation of flows at site i.
The corresponding B matrix depends on the new A
matrix and S
0
, where as before
BB
T
S
0
AS
0
A
T
(7.176)
The idea of fitting time-series models to each site
separately and then correlating the innovations in those
separate models to reproduce the cross-correlation
between the series is a very general and useful modelling
idea that has seen a number of applications with different
time-series models (Matalas and Wallis, 1976; Stedinger
et al., 1985; Camacho et al., 1985; Salas, 1993).
8.6. Multi-Season, Multi-Site Models
In most studies of surface water systems it is necessary to
consider the variations of flows within each year.
Streamflows in most areas have within-year variations,
exhibiting wet and dry periods. Similarly, water demands
for irrigation, municipal and industrial uses also vary, and
the variations in demand are generally out of phase with
the variation in within-year flows; more water is usually
desired when streamflows are low, and less is desired
when flows are high. This increases the stress on water
delivery systems and makes it all the more important that
time-series models of streamflows, precipitation and
other hydrological variables correctly reproduce the sea-
sonality of hydrological processes.
This section discusses two approaches to generating
within-year flows. The first approach is based on the disag-
gregation of annual flows produced by an annual flow
generator to seasonal flows. Thus the method allows for
reproduction of both the annual and seasonal characteris-
tics of streamflow series. The second approach generates
seasonal flows in a sequential manner, as was done for the
generation of annual flows. Thus the models are a direct
generalization of the annual flow models already discussed.
8.6.1. Disaggregation Models
The disaggregation model proposed by Valencia and
Schaake (1973) and extended by Mejia and Rousselle
(1976) and Tao and Delleur (1976) allows for the
generation of synthetic flows that reproduce statistics
both at the annual and at the seasonal level. Subsequent
improvements and variations are described by Stedinger
and Vogel (1984), Maheepala and Perera (1996),
Koutsoyiannis and Manetas (1996) and Tarboton et al.
(1998).
Disaggregation models can be used for either multi-
season single-site or multi-site streamflow generation.
They represent a very flexible modelling framework for
dealing with different time or spatial scales. Annual flows
for the several sites in question or the aggregate total
annual flow at several sites can be the input to the model
(Grygier and Stedinger, 1988). These must be generated
by another model, such as those discussed in the previous
sections. These annual flows or aggregated annual flows
are then disaggregated to seasonal values.
Let Z
y
Z
y
1
, , Z
y
N

T
be the column vector of N trans-
formed normally distributed annual or aggregate annual
flows for N separate sites or basins. Next, let X
y
X
1
1
y
, ,
X
1
T
y
,
X
1
2
y
, , X
2
T
y
, , X
1
n
y
, , X
n
T
y

T
be the column vector of
nT transformed normally distributed seasonal flows X
s
ty
for
season t, year y, and site s 1, , n.
Assuming that the annual and seasonal series, Z
y
s
and
X
s
ty
, have zero mean (after the appropriate transforma-
tion), the basic disaggregation model is
X
y
AZ
y
BV
y
(7.177)
where V
y
is a vector of nT independent standard normal
random variables, and A and B are, respectively, nT N
and nT nT matrices. One selects values of the elements
of A and B to reproduce the observed correlations among
the elements of X
y
and between the elements of X
y
and Z
y
.
Alternatively, one could attempt to reproduce the
observed correlations of the untransformed flows as
opposed to the transformed flows, although this is not
always possible (Hoshi et al., 1978) and often produces
poorer estimates of the actual correlations of the flows
(Stedinger, 1981).
The values of A and B are determined using the
matrices S
zz
E[Z
y
Z
y
T
], S
xx
E[X
y
X
y
T
], S
xz
E[X
y
Z
y
T
], and
S
zx
E[Z
y
X
y
T
] where S
zz
was called S
0
earlier. Clearly,
S
T
xz
S
zx
. If S
xz
is to be reproduced, then by multiplying
Equation 7.177 on the right by Z
y
T
and taking expectations,
one sees that A must satisfy
(7.178) E[ E[A
T
X Z Z Z
y y y y
T
] ]
wrm_ch07.qxd 8/31/2005 12:18 PM Page 211
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
or
S
xz
AS
zz
(7.179)
Solving for the coefficient matrix A one obtains
(7.180)
To obtain an equation that determines the required values
in the matrix B, one can multiply both sides of Equation
7.177 by their transpose and take expectations to obtain
S
xx
AS
zz
A
T
BB
T
(7.181)
Thus, to reproduce the covariance matrix S
xx
, the B matrix
must satisfy
BB
T
S
xx
AS
zz
A
T
(7.182)
Equations 7.180 and 7.182 for determining A and B are
completely analogous to Equations 7.171 and 7.174 for the
A and B matrices of the lag 1 models developed earlier.
However, for the disaggregation model as formulated, BB
T
,
and hence the matrix B, can actually be singular or nearly
so (Valencia and Schaake, 1973). This occurs because the
real seasonal flows sum to the observed annual flows. Thus
given the annual flow at a site and all but one (T 1) of the
seasonal flows, the value of the unspecified seasonal flow
can be determined by subtraction.
If the seasonal variables X
s
ty
correspond to non-linear
transformations of the actual flows Q
s
ty
, then BB
T
is gener-
ally sufficiently non-singular that a B matrix can be
obtained by Cholesky decomposition. On the other hand,
when the model is used to generate values of X
s
ty
to be
transformed into synthetic flows Q
s
ty
, the constraint that
these seasonal flows should sum to the given value of the
annual flow is lost. Thus the generated annual flows (equal
to the sums of the seasonal flows) will deviate from the
values that were to have been the annual flows. Some
distortion of the specified distribution of the annual flows
results. This small distortion can be ignored, or each years
seasonal flows can be scaled so that their sum equals the
specified value of the annual flow (Grygier and Stedinger,
1988). The latter approach eliminates the distortion in the
distribution of the generated annual flows by distorting the
distribution of the generated seasonal flows. Koutsoyiannis
and Manetas (1996) improve upon the simple scaling
algorithm by including a step that rejects candidate vectors
X
y
if the required adjustment is too large, and instead gen-
erates another vector X
y
. This reduces the distortion in the
monthly flows that results from the adjustment step.
A S S

xz zz
1
212 Water Resources Systems Planning and Management
The disaggregation model has substantial data require-
ments. When the dimension of Z
y
is n and the dimension
of the generated vector X
y
is m, the A matrix has mn
elements. The lower diagonal B matrix and the symmetric
S
xx
matrix, upon which it depends, each have m(m 1)/2
nonzero or non-redundant elements. For example, when
disaggregating two aggregate annual flow series to
monthly flows at five sites, n 2 and m 12 5 60;
thus, A has 120 elements while B and S
xx
each have 1,830
nonzero or non-redundant parameters. As the number of
sites included in the disaggregation increases, the size
of S
xx
and B increases rapidly. Very quickly the model
can become overly parameterized, and there will be
insufficient data to estimate all parameters (Grygier and
Stedinger, 1988).
In particular, one can think of Equation 7.177 as
a series of linear models generating each monthly flow
X
k
ty
for k 1, t 1, , 12; k 2, t 1, , 12 up to
k n, t 1, , 12 that reproduces the correlation of
each X
k
ty
with all n annual flows, Z
y
k
, and all previously
generated monthly flows. Then when one gets to the last
flow in the last month, the model will be attempting to
reproduce n (12n 1) 13n 1 annual to monthly
and cross-correlations. Because the model implicitly
includes a constant, this means one needs k* 13n
years of data to obtain a unique solution for this
critical equation. For n 3, k* 39. One could say
that with a record length of forty years, there would
be only one degree of freedom left in the residual
model error variance described by B. That would be
unsatisfactory.
When flows at many sites or in many seasons are
required, the size of the disaggregation model can be
reduced by disaggregation of the flows in stages. Such
condensed models do not explicitly reproduce every
season-to-season correlation (Lane, 1979; Stedinger and
Vogel, 1984; Gryier and Stedinger, 1988; Koutsoyiannis
and Manetas, 1996), Nor do they attempt to reproduce
the cross-correlations among all the flow variates at the
same site within a year (Lane, 1979; Stedinger et al.,
1985). Contemporaneous models, like the Markov
model developed earlier in Section 8.5, are models
developed for individual sites whose innovation vectors
V
y
have the needed cross-correlations to reproduce the
cross-correlations of the concurrent flows (Camacho
et al., 1985), as was done in Equation 7.176. Grygier and
wrm_ch07.qxd 8/31/2005 12:18 PM Page 212
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 213
Stedinger (1991) describe how this can be done for a
condensed disaggregation model without generating
inconsistencies.
8.6.2. Aggregation Models
One can start with annual or seasonal flows, and break
them down into flows in shorter periods representing
months or weeks. Alternatively one can start with a model
that describes the shortest time step flows. This latter
approach has been referred to as aggregation to distin-
guish it from disaggregation.
One method for generating multi-season flow
sequences is to convert the time series of seasonal flows
Q
ty
into a homogeneous sequence of normally distributed
zero-mean unit-variance random variables Z
ty
. These can
then be modelled by an extension of the annual flow gen-
erators that have already been discussed. This transfor-
mation can be accomplished by fitting a reasonable
marginal distribution to the flows in each season so as to
be able to convert the observed flows q
s
ty
into their trans-
formed counterparts Z
s
ty
, and vice versa. Particularly when
shorter streamflow records are available, these simple
approaches may yield a reasonable model of some streams
for some studies. However, they implicitly assume that
the standardized series is stationary, in the sense that the
season-to-season correlations of the flows do not depend
on the seasons in question. This assumption seems highly
questionable.
This theoretical difficulty with the standardized series
can be overcome by introducing a separate streamflow
model for each month. For example, the classic
ThomasFiering model (Thomas and Fiering, 1970) of
monthly flows may be written
(7.183)
where the Z
ty
s are standard normal random variables
corresponding to the streamflow in season t of year y,
t
is the season-to-season correlation of the standardized
flows, and V
ty
are independent standard normal random
variables. The problem with this model is that it often fails
to reproduce the correlation among non-consecutive
months during a year and thus misrepresents the risk
of multi-month and multi-year droughts (Hoshi et al.,
1978).
Z Z V
t y t ty t ty

1
2
1
,

For an aggregation approach to be attractive, it is
necessary to use a model with greater persisitence
than the ThomasFiering model. A general class of
time-series models that allow reproduction of different
correlation structures are the BoxJenkins Autoregressive-
Moving average models (Box et al., 1994). These models
are presented by the notation ARMA(p,q) for a model
which depends on p previous flows, and q extra
innovations V
ty
. For example, Equation 7.141 would be
called an AR(1) or AR(1,0) model. A simple ARMA(1,1)
model is
(7.184)
The correlations of this model have the values

1
(1
1

1
)(
1

1
)/(1
1
2
2
1

1
) (7.185)
for the first lag. For i 1
(7.186)
For values near and 0
1

1
, the autocorrelations

k
can decay much slower than those of the standard
AR(1) model.
The correlation function
k
of general ARMA(p,q)
model,
(7.187)
is a complex function that must satisfy a number of
conditions to ensure the resultant model is stationary and
invertible (Box et al., 1994).
ARMA(p,q) models have been extended to describe
seasonal flow series by having their coefficients depend
upon the season these are called periodic autoregressive-
moving average models, or PARMA. Salas and Obeysekera
(1992), Salas and Fernandez (1993), and Claps et al.,
(1993) discuss the conceptual basis of such stochastic
streamflow models. For example, Salas and Obeysekera
(1992) found that low-order PARMA models, such as
a PARMA(2,1), arise from reasonable conceptual repre-
sentations of persistence in rainfall, runoff and ground-
water recharge and release. Claps et al. (1993) observe
that the PARMA(2,2) model, which may be needed if
one wants to preserve year-to-year correlation, poses a
parameter estimation challenge that is almost unmanage-
able (see also Rasmussen et al., 1996). The PARMA (1,1)
Z Z V V
t i t i t
i
p
j t j
j
q

1 1 1
1
1
1



i
i

1
1
Z Z V V
t t t t

1 1 1 1

+
wrm_ch07.qxd 8/31/2005 12:18 PM Page 213
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
model is more practical and easy to extend to the multi-
variate case (Hirsch, 1979; Stedinger et al., 1985; Salas,
1993; Rasmussen et al., 1996). Experience has shown
that PARMA(1,1) models do a better job of reproducing
the correlation of seasonal flows beyond lag 1 than does a
ThomasFiering PAR(1,0) model (see for example,
Bartolini and Salas, 1993).
9. Stochastic Simulation
This section introduces stochastic simulation. Much more
detail on simulation is contained in later chapters. As
discussed in Chapter 3, simulation is a flexible and widely
used tool for the analysis of complex water resources
systems. Simulation is trial and error. One must define the
system being simulated, both its design and operating
policy, and then simulate it to see how it works. If the
purpose is to find the best design and operating policy,
many such alternatives must be simulated and their results
must be compared. When the number of alternatives to
simulate becomes too large for the time and money avail-
able for such analyses, some kind of preliminary screening,
perhaps using optimization models, may be justified. This
use of optimization for preliminary screening that is, for
eliminating alternatives prior to a more detailed simulation
is discussed in Chapters 3, 4 and later chapters.
As with optimization models, simulation models
may be deterministic or stochastic. One of the most useful
tools in water resources systems planning is stochastic
simulation. While optimization can be used to help
define reasonable design and operating policy alternatives
to be simulated, simulations can better reveal how each
such alternative will perform. Stochastic simulation of com-
plex water resources systems on digital computers provides
planners with a way to define the probability distributions
of multiple performance indices of those systems.
When simulating any system, the modeller designs
an experiment. Initial flow, storage and water quality
conditions must be specified if these are being simulated.
For example, reservoirs can start full, empty or at random
representative conditions. The modeller also determines
what data are to be collected on system performance and
operation, and how they are to be summarized. The
length of time the simulation is to be run must be
specified and, in the case of stochastic simulations, the
214 Water Resources Systems Planning and Management
number of runs to be made must also be determined.
These considerations are discussed in more detail by
Fishman (2001) and in other books on simulation. The
use of stochastic simulation and the analysis of the output
of such models are introduced here primarily in the
context of an example to illustrate what goes into a
stochastic simulation model and how one can deal with
the information that is generated.
9.1. Generating Random Variables
Included in any stochastic simulation model is some
provision for the generation of sequences of random
numbers that represent particular values of events such as
rainfall, streamflows or floods. To generate a sequence of
values for a random variable, the probability distribution
for the random variable must be specified. Historical data
and an understanding of the physical processes are used
to select appropriate distributions and to estimate their
parameters (as discussed in Section 7.2).
Most computers have algorithms for generating
random numbers uniformly distributed (equally likely)
between zero and one. This uniform distribution of
random numbers is defined by its cdf and pdf;
F
U
(u) 0 for u 0,
u for 0 u 1
and
1 if u 1 (7.188)
so that
f
U
(u) 1 if 0 u 1 and 0 otherwise (7.189)
These uniform random variables can then be transformed
into random variables with any desired distribution. If
F
Q
(q
t
) is the cumulative distribution function of a random
variable Q
t
in period t, then Q
t
can be generated using the
inverse of the distribution.
Q
t
F
Q
1
[U
t
] (7.190)
Here U
t
is the uniform random number used to generate
Q
t
. This is illustrated in Figure 7.15.
Analytical expressions for the inverse of many
distributions, such as the normal distribution, are
not known, so special algorithms are employed to
efficiently generate deviates with these distributions
(Fishman, 2001).
wrm_ch07.qxd 8/31/2005 12:18 PM Page 214
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 215
9.2. River Basin Simulation
An example will demonstrate the use of stochastic
simulation in the design and analysis of water resources
systems. Assume that farmers in a particular valley have
been plagued by frequent shortages of irrigation water.
They currently draw water from an unregulated river to
which they have water rights. A government agency has
proposed construction of a moderate-size dam on the
river upstream of the points where the farmers withdraw
water. The dam would be used to increase the quantity
and reliability of irrigation water available to the farmers
during the summer growing season.
After preliminary planning, a reservoir with an active
capacity of 4 10
7
m
3
has been proposed for a natural
dam site. It is anticipated that, because of the increased
reliability and availability of irrigation water, the quantity
of water desired will grow from an initial level of 3
10
7
m
3
/yr after construction of the dam to 4 10
7
m
3
/yr
within six years. After that, demand will grow more
slowly to 4.5 10
7
m
3
/yr the estimated maximum
reliable yield. The projected demand for summer irriga-
tion water is shown in Table 7.12.
A simulation study can evaluate how the system will
be expected to perform over a twenty-year planning
period. Table 7.13 contains statistics that describe the
hydrology at the dam site. The estimated moments are
computed from the forty-five-year historic record.
Using the techniques discussed in the previous
section, a ThomasFiering model is used to generate
twenty-five lognormally distributed synthetic streamflow
sequences. The statistical characteristics of the synthetic
flows are those listed in Table 7.14. Use of only the
forty-five-year historic flow sequence would not allow
examination of the systems performance over the large
range of streamflow sequences, which could occur during
the twenty-year planning period. Jointly, the synthetic
sequences should be a description of the range of inflows
that the system might experience. A larger number of
sequences could be generated.
particular value of Q
t
q
t
u
t
1
0
0
u
t
F =
Q
(q
t
)
F
Q
(q
t
)
E
0
2
0
5
2
7
n
u
-1
F
Q
( )
t
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
E
0
2
1
1
0
1
n
3.0
3.2
3.4
3.6
3.8
4.0
4.1
4.2
4.3
4.3
4.4
4.4
4.4
4.4
4.5
4.5
4.5
4.5
4.5
4.5
water
demand
year
( x 10
7
m
3
/yr)
Figure 7.15. The probability distribution of a random variable
can be inverted to produce values of the random variable.
Table 7.12. Projected water demand for irrigation water.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 215
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
9.3. The Simulation Model
The simulation model is composed primarily of continu-
ity constraints and the proposed operating policy.
The volume of water stored in the reservoir at the
beginning of seasons 1 (winter) and 2 (summer) in year
y are denoted by S
1y
and S
2y
. The reservoirs winter
operating policy is to store as much of the winters inflow
Q
1y
as possible. The winter release R
1y
is determined by
the rule
(7.191)
where K is the reservoir capacity of 4 10
7
m
3
and R
min
is 0.50 10
7
m
3
, the minimum release to be made if
possible. The volume of water in storage at the beginning
of the years summer season is
S
2y
S
1y
Q
1y
R
1y
(7.192)
The summer release policy is to meet each years
projected demand or target release D
y
, if possible, so
that
R
2y
S
2y
Q
2y
K if S
2y
Q
2y
D
y
K
D
y
if 0 S
2y
Q
2y
D
y
K
S
2y
Q
2y
otherwise (7.193)
This operating policy is illustrated in Figure 7.16.
The volume of water in storage at the beginning of the
next winter season is
S
1,y1
S
2y
Q
2y
R
2y
(7.194)
R
S K S R K
R K S R
S
y
y y y y
y y 1
1 1 1 1
1 1
1 1
0

Q Q
Q
Q
if
if
min
min min
y y y
otherwise

216 Water Resources Systems Planning and Management

9.4. Simulation of the Basin
The question to be addressed by this simulation study
is how well the reservoir will meet the farmers water
requirements. Three steps are involved in answering this
question. First, one must define the performance criteria
or indices to be used to describe the systems perform-
ance. The appropriate indices will, of course, depend on
the problem at hand and the specific concerns of the users
and managers of a water resources system. In this exam-
ple of a reservoir-irrigation system, several indices will be
used relating to the reliability with which target releases
are met and the severity of any shortages.
The second step is to simulate the proposed system to
evaluate the specified indices. For our reservoir-irrigation
system, the reservoirs operation was simulated twenty-five
E
0
2
1
1
0
1
p
standard deviation
mean flow
winter with following summer
correlation of flows:
summer with following winter
1.5
4.0
1.0
2.5
0.65
0.60
2.3
6.5
winter summer annual
x 10 m
7
3
x 10 m
7
3
Table 7.13. Characteristics of the river flow.
r
e
s
e
r
v
o
i
r
e
m
p
t
i
e
s
S 1
,
y
+
1
=

O
1
1
K reservoir capacity water available S
2y
+ Q
2y
E
0
2
1
0
2
9
c
r
e
s
e
r
v
o
i
r

r
e
l
e
a
s
e
R
2
y

=

S 2
y

+

Q
2
y
R
2
y

=

S 2
y

+

Q
2
y
-
K
r
e
s
e
r
v
o
i
r

f
i
l
l
s

S 1
,
y
+
1

=

K
Figure 7.16. Summer reservoir operating policy. The shaded
area denotes the feasible region of reservoir releases.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 216
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 217
times using the twenty-five synthetic streamflow sequences,
each twenty years in length. Each of the twenty simulated
years consisted of first a winter and then a summer season.
At the beginning of the first winter season, the reservoir was
taken to be empty (S
1y
0 for y 1) because construction
would just have been completed. The target release or
demand for water in each year is given in Table 7.13.
The third and final step, after simulating the system, is
to interpret the resulting information so as to gain an
understanding of how the system might perform both
with the proposed design and operating policy and with
modifications in either the systems design or its operating
policy. To see how this may be done, consider the opera-
tion of our example reservoir-irrigation system.
The reliability p
y
of the target release in year y is the
probability that the target release D
y
is met or exceeded in
that year:
p
y
Pr[R
2y
D
y
] (7.195)
The systems reliability is a function of the target release
D
y
, the hydrology of the river, the size of the reservoir and
the operating policy of the system. In this example, the
reliability also depends on the year in question. Figure
7.17 shows the total number of failures that occurred in
each year of the twenty-five simulations. In three of these,
the reservoir did not contain sufficient water after the
initial winter season to meet the demand the first sum-
mer. After year 1, few failures occur in years 2 through 9
because of the low demand. Surprisingly few failures
occur in years 10 and 13, when demand has reached its
peak; this is because the reservoir was normally full at the
beginning of this period as a result of lower demand in
the earlier years. Starting in years 14 and after, failures
occurred more frequently because of the higher demand
placed on the system. Thus one has a sense of how the
reliability of the target releases changes over time.
9.5. Interpreting Simulation Output
Table 7.14 contains several summary statistics of the
twenty-five simulations. Column 2 of the table contains
the average failure frequency in each simulation, which
equals the number of years the target release was not met
divided by twenty, the number of years simulated. At the
bottom of column 2 and the other columns are several
statistics that summarize the twenty-five values of the
different performance indices. The sample estimates of the
mean and variance of each index are given as one way of
summarizing the distribution of the observations. Another
approach is specification of the sample median, the
approximate inter-quartile range x
(6)
x
(20)
, and/or the
range x
(1)
x
(25)
of the observations, where x
(i)
is the ith
largest observation. Either set of statistics could be used to
describe the centre and spread of each indexs distribution.
Suppose that one is interested in the distribution of the
systems failure frequency or, equivalently, the reliability
with which the target can be met. Table 7.14 reports that
the mean failure rate for the twenty-five simulations is
0.084, implying that the average reliability over the
twenty-year period is 1 0.084 0.916, or 92%. The
median failure rate is 0.05, implying a median reliability
of 95%. These are both reasonable estimates of the centre
of the distribution of the failure frequency. Note that the
actual failure frequency ranged from 0 (seven times) to
0.30. Thus the systems reliability ranged from 100% to
as low as 70%, 75% and 80% in runs 17, 8, and 11,
respectively. Obviously, the farmers are interested not
only in knowing the mean failure frequency but also the
range of failure frequencies they are likely to experience.
If one knew the form of the distribution function
of the failure frequency, one could use the mean and
standard deviation of the observations to determine an
interval within which the observations would fall with
some pre-specified probability. For example, if the
observations are normally distributed, there is a 90%
probability that the index falls within the interval

x
1.65
x
. Thus, if the simulated failure rates are
normally distributed, then there is about a 90% probability
5
4
3
2
1
0
number of failures in each year
2 6 10 12 14 16 20 18
year of simulation
E021029m
4 8
Figure 7.17. Number of failures in each year of twenty-five
twenty-year simulations.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 217
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
218 Water Resources Systems Planning and Management
simulation
number, i
E
0
2
1
1
0
1
q
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
0.10
0.15
0.10
0.10
0.05
0.0
0.15
0.25
0.0
0.10
0.20
0.05
0.05
0.10
0.15
0.05
0.30
0.10
0.0
0.0
0.0
0.05
0.0
0.0
0.05
0.0
0.05
0.05
0.05
0.0
0.0
0.05
0.10
0.0
0.0
0.0
0.05
0.0
0.0
0.05
0.0
0.05
0.0
0.0
0.0
0.0
0.05
0.0
0.0
0.0
target
frequency
of failure to meet:
80% of
target
1.25
1.97
1.79
1.67
0.21
0.00
1.29
4.75
0.00
0.34
1.80
1.28
0.53
0.88
1.99
0.23
2.68
0.76
0.00
0.00
0.00
1.47
0.00
0.00
0.19
0.14
0.17
0.20
0.22
0.05
0.00
0.10
0.21
0.00
0.04
0.11
0.43
0.12
0.11
0.15
0.05
0.10
0.08
0.00
0.00
0.00
0.33
0.00
0.00
0.04
total
shortage
TS
x10 m
7
3
average
deficit,
AD
0.084 x mean 0.020 1.00 0.106
standard deviation of values;
0.081 s
x 0.029 1.13 0.110
0.05 median 0.00 0.76 0.10
approximate interquartile range;
0.0 - 0.15 0.0 - 0.05 0.0 - 1.79 0.0 - 0.17 x
(6)
-x
(20)
range;
0.0 - 0.30 0.0 - 0.10 0.0 - 4.75 0.0 - 0.43 x
( ) 1
-x
(25)
Table 7.14. Results of 25 twenty-year
simulations.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 218
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
that the actual failure rate observed in any simulation is
within the interval x
_
1.65s
x
. In our case this interval
would be [0.084 1.65(0.081), 0.084 1.65(0.081)]
[0.050, 0.218]. Clearly, the failure rate cannot be less
than zero, so this interval makes little sense in our example.
A more reasonable approach to describing the
distribution of a performance index whose probability dis-
tribution function is not known is to use the observations
themselves. If the observations are of a continuous random
variable, the interval x
(i)
x
(n1i)
provides a reasonable
estimate of an interval within which the random variable
falls with probability
(7.196)
In our example, the range x
(1)
x
(25)
of the twenty-five
observations is an estimate of an interval in which a
continuous random variable falls with probability
(25 1 2)/(25 1) 92%, while x
(6)
x
(20)
corre-
sponds to probability (25 1 2 6)/(25 1) 54%.
Table 7.14 reports that, for the failure frequency,
x
(1)
x
(25)
equals 00.30, while x
(6)
x
(20)
equals 00.15.
Reflection on how the failure frequencies are calculated
reminds us that the failure frequency can only take on
the discrete, non-negative values 0, 1/20, 2/20, ,
20/20. Thus, the random variable X cannot be less than
zero. Hence, if the lower endpoint of an interval is zero,
as is the case here, then 0 x
(k)
is an estimate of an inter-
val within which the random variable falls with a proba-
bility of at least k/(n 1). For k equal to 20 and 25, the
corresponding probabilities are 77% and 96%.
Often, the analysis of a simulated systems perform-
ance centres on the average value of performance indices,
such as the failure rate. It is important to know the
accuracy with which the mean value of an index approx-
imates the true mean. This is done by the construction of
confidence intervals. A confidence interval is an interval
that will contain the unknown value of a parameter with
a specified probability. Confidence intervals for a mean
are constructed using the t statistic,
(7.197)
which, for large n, has approximately a standard normal
distribution. Certainly, n 25 is not very large, but
t
x
s n
x
x

/
P
n i
n
i
n
n i
n

1
1 1
1 2
1
Concepts in Probability, Statistics and Stochastic Modelling 219
the approximation to a normal distribution may be
sufficiently good to obtain a rough estimate of how
close the average frequency of failure x
_
is likely to be to

x
. A 100(1 2)% confidence interval for
x
is, approx-
imately,
(7.198)
If 0.05, then using a normal distribution t

1.65
and Equation 7.118 becomes 0.057
x
0.11.
Hence, based on the simulation output, one can be
about 90% sure that the true mean failure frequency lies
between 5.7% and 11%. This corresponds to a reliability
of between 89% and 94%. By performing additional
simulations to increase the size of n, the width of this
confidence interval can be decreased. However, this
increase in accuracy may be an illusion because the
uncertainty in the parameters of the streamflow model
has not been incorporated into the analysis.
Failure frequency or system reliability describes only
one dimension of the systems performance. Table 7.14
contains additional information on the systems perform-
ance related to the severity of shortages. Column 3 lists
the frequencies with which the shortage exceeded 20% of
that years demand. This occurred in approximately 2%
of the years, or in 24% of the years in which a failure
occurred. Taking another point of view, failures in excess
of 20% of demand occurred in nine out of twenty-five, or
in 36% of the simulation runs.
Columns 4 and 5 of Table 7.14 contain two other
indices that pertain to the severity of the failures. The total
shortfall, TS, in Column 4 is calculated as the sum of the
positive differences between the demand and the release
in the summer season over the twenty-year period.
TS [D
2y
R
2y
]

where
[Q]

Q if Q 0; 0 otherwise (7.199)
The total shortfall equals the total amount by which the
target release is not met in years in which shortages
occur.
y

x t
s
n
x t
s
n
t t
x
x
x
x

or
0.084
0 081
25
0 084
0 081 .
.
.

_
,

225

_
,

wrm_ch07.qxd 8/31/2005 12:18 PM Page 219

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Related to the total shortfall is the average deficit. The
deficit is defined as the shortfall in any year divided by
the target release in that year. The average deficit, AD, is
(7.200)
where m is the number of failures (deficits) or nonzero
terms in the sum.
Both the total shortfall and the average deficit measure
the severity of shortages. The mean total shortfall TS
__
,
equal to 1.00 for the twenty-five simulation runs, is a dif-
ficult number to interpret. While no shortage occurred in
seven runs, the total shortage was 4.7 in run 8, in which
the shortfall in two different years exceeded 20% of the
target. The median of the total shortage values, equal to
0.76, is an easier number to interpret in that one knows
that half the time the total shortage was greater and half
the time less than this value.
The mean average deficit AD
__
is 0.106, or 11%.
However, this mean includes an average deficit of zero in
the seven runs in which no shortages occurred. The
average deficit in the eighteen runs in which shortages
occurred is (11%)(25/18) 15%. The average deficit in
individual simulations in which shortages occurred
ranges from 4% to 43%, with a median of 11.5%.
After examining the results reported in Table 7.14, the
farmers might determine that the probability of a shortage
exceeding 20% of a years target release is higher than
they would like. They can deal with more frequent minor
shortages, not exceeding 20% of the target, with little
economic hardship, particularly if they are warned at
the beginning of the growing season that less than the
targeted quantity of water will be delivered. Then they can
curtail their planting or plant crops requiring less water.
In an attempt to find out how better to meet the
farmers needs, the simulation program was re-run with
the same streamflow sequences and a new operating policy
in which only 80% of the growing seasons target release is
provided (if possible) if the reservoir is less than 80% full
at the end of the previous winter season. This gives the
farmers time to adjust their planting schedules and may
increase the quantity of water stored in the reservoir to be
used the following year if the drought persists.
As the simulation results with the new policy in
Table 7.15 demonstrate, this new operating policy appears
AD
m
D R
D
y y
y

1 2 2
2 1
20

1
]

y
220 Water Resources Systems Planning and Management
to have the expected effect on the systems operation. With
the new policy, only six severe shortages in excess of 20%
of demand occur in the twenty-five twenty-year simula-
tions, as opposed to ten such shortages with the original
policy. In addition, these severe shortages are all less severe
than the corresponding shortages that occur with the same
streamflow sequence when the original policy is followed.
The decrease in the severity of shortages is obtained at
a price. The overall failure frequency has increased from
8.4% to 14.2%. However, the latter value is misleading
because in fourteen of the twenty-five simulations, a
failure occurs in the first simulation year with the new
policy, whereas only three failures occur with the original
policy. Of course, these first-year failures occur because
the reservoir starts empty at the beginning of the first
winter and often does not fill that season. Ignoring these
first-year failures, the failure rates with the two policies
over the subsequent nineteen years are 8.2% and 12.0%.
Thus the frequency of failures in excess of 20% of
demand is decreased from 2.0% to 1.2% by increasing the
frequency of all failures after the first year from 8.2%
to 12.0%. Reliability decreases, but so does vulnerability.
If the farmers are willing to put up with more frequent
minor shortages, then it appears that they can reduce
their risk of experiencing shortages of greater severity.
The preceding discussion has ignored the statistical
issue of whether the differences between the indices
obtained in the two simulation experiments are of suffi-
cient statistical reliability to support the analysis. If care is
not taken, observed changes in a performance index from
one simulation experiment to another may be due to sam-
pling fluctuations rather than to modifications of the
water resource systems design or operating policy.
As an example, consider the change that occurred
in the frequency of shortages. Let X
1i
and X
2i
be the
simulated failure rates using the ith streamflow sequence
with the original and modified operating policies. The
random variables Y
i
X
1i
X
2i
for i equal 1 through 25
are independent of each other if the streamflow sequences
are generated independently, as they were.
One would like to confirm that the random variable Y
tends to be negative more often than it is positive, and
hence, that policy 2 indeed results in more failures overall.
A direct test of this theory is provided by the sign test. Of
the twenty-five paired simulation runs, y
i
0 in twenty-
one cases and y
i
0 in four cases. We can ignore the times
wrm_ch07.qxd 8/31/2005 12:18 PM Page 220
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 221
simulation
number, i
E
0
2
1
1
0
1
r
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
0.10
0.30
0.25
0.20
0.10
0.05
0.20
0.25
0.05
0.20
0.25
0.10
0.10
0.15
0.25
0.10
0.30
0.15
0.0
0.05
0.10
0.10
0.05
0.05
0.10
0.0
0.0
0.0
0.05
0.0
0.0
0.0
0.10
0.0
0.0
0.0
0.05
0.0
0.0
0.05
0.0
0.0
0.0
0.0
0.0
0.0
0.05
0.0
0.0
0.0
target
frequency
of failure to meet:
80% of
target
1.80
4.70
3.90
3.46
1.48
0.60
3.30
5.45
0.60
3.24
3.88
1.92
1.50
2.52
3.76
1.80
5.10
2.40
0.0
0.76
1.80
2.37
0.90
0.90
1.50
0.20
0.20
0.20
0.21
0.20
0.20
0.20
0.26
0.20
0.20
0.20
0.31
0.20
0.20
0.18
0.20
0.20
0.20
0.0
0.20
0.20
0.26
0.20
0.20
0.20
total
shortage
TS
x10 m
7
3
average
deficit,
AD
0.142 x mean 0.012 2.39 0.201
standard deviation of values;
0.087 s
x 0.026 1.50 0.050
0.10 median 0.00 1.92 0.20
approximate interquartile range;
0.05 - 0.25 0.0 - 0.0 0.90 - 3.76 0.20 - 0.20 x
(6)
-x
(20)
range;
0.0 - 0.30 0.0 - 0.10 0.0 - 5.45 0.0 - 0.31 x
( ) 1
-x
(25)
Table 7.15. Results of 25 twenty-year
simulations with modified operating policy to
avoid severe shortages.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 221
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
when y
i
0. Note that if y
i
0 and y
i
0 were equally
likely, then the probability of observing y
i
0 in all
twenty-one cases when y
i
0 is 2
21
or 5 10
7
. This is
exceptionally strong proof that the new policy has
increased the failure frequency.
A similar analysis can be made of the frequency with
which the release is less than 80% of the target. Failure
frequencies differ in the two policies in only four of the
twenty-five simulation runs. However, in all four cases
where they differ, the new policy resulted in fewer severe
failures. The probability of such a lopsided result, were it
equally likely that either policy would result in a lower
frequency of failures in excess of 20% of the target, is
2
4
0.0625. This is fairly strong evidence that the new
policy indeed decreases the frequency of severe failures.
Another approach to this problem is to ask if the
difference between the average failure rates x
_
1
and x
_
2
is
statistically significant; that is, can the difference between X
1
and X
2
be attributed to the fluctuations that occur in the
average of any finite set of random variables? In this example
the significance of the difference between the two means can
be tested using the random variable Y
i
defined as X
1i
X
2i
for i equal 1 through 25. The mean of the observed y
i
s is
(7.201)
and their variance is
(7.202)
Now, if the sample size n, equal to 25 here, is sufficiently
large, then t defined by
(7.203)
has approximately a standard normal distribution. The
closer the distribution of Y is to that of the normal
distribution, the faster the convergence of the distribution
of t is to the standard normal distribution with increasing
n. If X
1i
X
2i
is normally distributed, which is not the case
here, then each Y
i
has a normal distribution and t has
Students t-distribution.
If E[X
1i
] E[X
2i
], then
Y
equals zero, and upon sub-
stituting the observed values of y
_
and s
Y
2
into Equation
7.123, one obtains
t
y
s n
Y
Y

/
s x x y
i i
i
y
2
1 2
2
1
25
2
1
25
0 0400

( ) ( . )

y x x x x
i i
i


1
25
0 084 0 142 0 058
1 2
1
25
1 2
( )
. . .

222 Water Resources Systems Planning and Management

(7.204)
The probability of observing a value of t equal to 7.25
or smaller is less than 0.1% if n is sufficiently large that t
is normally distributed. Hence it appears very improbable
that
Y
equals zero.
This example provides an illustration of the advantage of
using the same streamflow sequences when simulating both
policies. Suppose that different streamflow sequences were
used in all the simulations. Then the expected value of Y
would not change, but its variance would be given by
(7.205)
where Cov(X
1
, X
2
) E[(X
1

1
)(X
2

2
)] and is the
covariance of the two random variables. The covariance
between X
1
and X
2
will be zero if they are independently
distributed, as they would be if different randomly gener-
ated streamflow sequences were used in each simulation.
Estimating
2
x
1
and
2
x
2
by their sample estimates, an
estimate of what the variance of Y would be if Cov(X
1
, X
2
)
were zero is

Y
(7.206)
The actual sample estimate
Y
equals 0.040; if independ-
ent streamflow sequences are used in all simulations,
Y
will take a value near 0.119 rather than 0.040 (Equation
7.202). A standard deviation of 0.119, with
y
0, yields
a value of the t test statistic
(7.207)
If t is normally distributed, the probability of observing a
value less than 2.44 is about 0.8%. This illustrates that use
of the same streamflow sequences in the simulation of both
policies allows one to better distinguish the differences in
the policies performance. By using the same streamflow
sequences, or other random inputs, one can construct a
simulation experiment in which variations in performance
caused by different random inputs are confused as little as
possible with the differences in performance caused by
changes in the systems design or operating policy.
t
y
Y

0 119 25
2 44
. /
.

x x
1 2
2 2 2 2 2
0 081 0 087 0 119 ( . ) ( . ) ( . )
Var( ) E[ ( )]
E[( ) ] 2E[( )
( )
2
2
Y X X
X X
X



1 2 1 2
1 1 1 1
2 2


]] E[( ) ]
Cov( )
2


X
X X
2 2
1 2

x x
1 2
2 2
2 ,
t

0 0580
0 0400 25
7 25
.
. /
.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 222
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 223
10. Conclusions
This chapter has introduced statistical concepts that
analysts use to describe the randomness or uncertainty
of their data. Most of the data used by water resources
systems analysts is uncertain. This uncertainty comes
from not understanding as well as we would like how
our water resources systems (including their ecosystems)
function as well as not being able to forecast, perfectly,
the future. It is that simple. We do not know the exact
amounts, qualities and their distributions over space and
time of either the supplies of water we manage or the water
demands we try to meet. We also do not know the bene-
fits and costs, however measured, of any actions we take
to manage both water supply and water demand.
The chapter began with an introduction to probability
concepts and methods for describing random variables
and parameters of their distributions. It then reviewed
some of the commonly used probability distributions and
how to determine the distributions of sample data, how
to work with censored and partial duration series data,
methods of regionalization, stochastic processes and
time-series analyses.
The chapter concluded with an introduction to a range
of univariate and multivariate stochastic models that
are used to generate stochastic streamflow, precipitation
depths, temperatures and evaporation. These methods are
used to generate temporal and spatial stochastic process
that serve as inputs to stochastic simulation models for
system design, for system operations studies, and for
evaluation of the reliability and precision of different
estimation algorithms. The final section of this chapter
provides an example of stochastic simulation, and the use
of statistical methods to summarize the results of such
simulations.
This is merely an introduction to some of the statisti-
cal tools available for use when dealing with uncertain
data. Many of the concepts introduced in this chapter will
be used in the chapters that follow on constructing and
implementing various types of optimization, simulation
and statistical models. The references cited in the
reference section provide additional and more detailed
information.
Although many of the methods presented in this and
in some of the following chapters can describe many of
the characteristics and consequences of uncertainty, it is
unclear as to whether or not society knows exactly what
to do with such information. Nevertheless, there seems to
be an increasing demand from stakeholders involved in
planning processes for information related to the uncer-
tainty associated with the impacts predicted by models.
The challenge is not only to quantify that uncertainty, but
also to communicate it in effective ways that inform, and
not confuse, the decision-making process.
11. References
AYYUB, B.M. and MCCUEN, R.H. 2002. Probability,
statistics, and reliability for engineers and scientists. Boca
Raton, Chapman and Hill, CRC Press.
BARTOLINI, P. and SALAS, J. 1993. Modelling of stream-
flow processes at different time scales. Water Resources
Research, Vol. 29, No. 8, pp. 257387.
BATES, B.C. and CAMPBELL, E.P. 2001. A Markov chain
Monte Carlo scheme for parameter estimation and
inference in conceptual rainfallrunoff modelling. Water
Resources Research, Vol. 37, No. 4, pp. 93747.
BEARD, L.R. 1960. Probability estimates based on small
normal-distribution samples. Journal of Geophysical
Research, Vol. 65, No. 7, pp. 21438.
BEARD, L.R. 1997. Estimating flood frequency and aver-
age annual damages. Journal of Water Resources Planning
and Management, Vol. 123, No. 2, pp. 848.
BENJAMIN, J.R. and CORNELL, C.A. 1970. Probability,
statistics and decisions for civil engineers. New York,
McGraw-Hill.
BICKEL, P.J. and DOKSUM, K.A. 1977. Mathematical
statistics: basic ideas and selected topics. San Francisco,
Holden-Day.
BOBE, B. 1975. The log Pearson type 3 distribution and
its applications in hydrology. Water Resources Research,
Vol. 14, No. 2, pp. 3659.
BOBE, B. and ASHKAR, F. 1991. The gamma distribution
and derived distributions applied in hydrology. Littleton
Colo., Water Resources Press.
BOBE, B. and ROBITAILLE, R. 1977. The use of
the Pearson type 3 distribution and log Pearson type 3
wrm_ch07.qxd 8/31/2005 12:18 PM Page 223
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
distribution revisited. Water Resources Research, Vol. 13,
No. 2, pp. 42743.
BOX, G.E.P.; JENKINS, G.M. and RISINSEL, G.C. 1994.
Times series analysis: forecasting and control, 3rd Edition.
New Jersey, Prentice-Hall.
CAMACHO, F.; MCLEOD, A.I. and HIPEL, K.W. 1985.
Contemporaneous autoregressive-moving average (CARMA)
modelling in water resources. Water Resources Bulletin,
Vol. 21, No. 4, pp. 70920.
CARLIN, B.P. and LOUIS, T.A. 2000. Bayes and empirical
Bayes methods for data analysis, 2nd Edition. New York,
Chapman and Hall, CRC.
CHOWDHURY, J.U. and STEDINGER, J.R. 1991.
Confidence intervals for design floods with estimated
skew coefficient. Journal of Hydraulic Engineering,
Vol. 117, No. 7, pp. 81131.
CHOWDHURY, J.U.; STEDINGER, J.R. and LU, L.H. 1991.
Goodness-of-fit tests for regional GEV flood distributions.
Water Resources Research, Vol. 27, No. 7, pp. 176576.
CLAPS, P. 1993. Conceptual basis for stochastic models
of monthly streamflows. In: J.B. Marco, R. Harboe and
J.D. Salas (eds). Stochastic hydrology and its use in water
resources systems simulation and optimization, Dordrecht,
Kluwer Academic, pp. 22935.
CLAPS, P.; ROSSI, F. and VITALE, C. 1993. Conceptual-
stochastic modelling of seasonal runoff using autoregressive
moving average models and different scales of aggregation.
Water Resources Research, Vol. 29, No. 8, pp. 254559.
COHN, T.A.; LANE, W.L. and BAIER, W.G. 1997. An
algorithm for computing moments-based flood quantile
estimates when historical flood information is available.
Water Resources Research, Vol. 33, No. 9, pp. 208996.
COHN, C.A.; LANE, W.L. and STEDINGER, J.R. 2001.
Confidence intervals for EMA flood quantile estimates.
Water Resources Research, Vol. 37, No. 6, pp. 16951706.
CRAINICEANU, C.M.; RUPPERT, D.; STEDINGER,
J.R. and BEHR, C.T. 2002. Improving MCMC mixing for a
GLMM describing pathogen concentrations in water supplies, in
case studies in Bayesian analysis. New York, Springer-Verlag.
DAGOSTINO, R.B. and STEPHENS, M.A. 1986.
Goodness-of-fit procedures. New York, Marcel Dekker.
224 Water Resources Systems Planning and Management
DAVID, H.A. 1981. Order statistics, 2nd edition. New
York, Wiley.
FIERING, M.B. 1967. Streamflow synthesis. Cambridge,
Mass., Harvard University Press.
FILL, H. and STEDINGER, J. 1995. L-moment and
PPCC goodness-of-fit tests for the Gumbel distribution
and effect of autocorrelation. Water Resources Research,
Vol. 31, No. 1, pp. 22529.
FILL, H. and STEDINGER, J. 1998. Using regional
regression within index flood procedures and an
empirical Bayesian estimator. Journal of Hydrology,
Vol. 210, Nos 14, pp. 12845.
FILLIBEN, J.J. 1975. The probability plot correlation
test for normality. Technometrics, Vol. 17, No. 1,
pp. 11117.
FISHMAN, G.S. 2001. Discrete-event simulation: modelling,
programming, and analysis. Berlin, Springer-Verlag.
GABRIELE, S. and ARNELL, N. 1991. A hierarchical
approach to regional flood frequency analysis. Water
Resources Research, Vol. 27, No. 6, pp. 12819.
GELMAN, A.; CARLIN, J.B.; STERN, H.S. and RUBIN, D.B.
1995. Bayesian data analysis. Boca Raton, Chapman and
Hall, CRC.
GILKS, W.R.; RICHARDSON, S. and SPIEGELHALTER,
D.J. (eds). 1996. Markov chain Monte Carlo in practice.
London and New York, Chapman and Hall.
GIESBRECHT, F. and KEMPTHORNE, O. 1976. Maximum
likelihood estimation in the three-parameter log normal
distribution. Journal of the Royal Statistical Society B, Vol. 38,
No. 3, pp. 25764.
GREENWOOD, J.A. and DURAND, D. 1960. Aids for
fitting the gamma distribution by maximum likelihood.
Technometrics, Vol. 2, No. 1, pp. 5565.
GRIFFS, V.W.; STEDINGER, J.R. and COHN, T.A. 2004.
LP3 quantile estimators with regional skew information
and low outlier adjustments. Water Resources Research,
Vol. 40, forthcoming.
GRYGIER, J.C. and STEDINGER, J.R. 1988. Condensed
disaggregation procedures and conservation corrections.
Water Resources Research, Vol. 24, No. 10, pp. 157484.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 224
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 225
GRYGIER, J.C. and STEDINGER, J.R. 1991. SPIGOT: a
synthetic flow generation software package, users manual and
technical description, version 2.6. Ithaca, N.Y., School of
Civil and Environmental Engineering, Cornell University.
GUMBEL, E.J. 1958. Statistics of extremes. New York,
Columbia University Press.
GUPTA, V.K. and DAWDY, D.R. 1995a. Physical inter-
pretation of regional variations in the scaling exponents of
flood quantiles. Hydrological Processes, Vol. 9, Nos. 34,
pp. 34761.
GUPTA, V.K. and DAWDY, D.R. 1995b. Multiscaling and
skew separation in regional floods. Water Resources
Research, Vol. 31, No. 11, pp. 276176.
GUPTA, V.K.; MESA, O.J. and DAWDY, D.R. 1994.
Multiscaling theory of flood peaks: regional quantile analy-
sis. Water Resources Research, Vol. 30, No. 12, pp. 340512.
HAAN, C.T. 1977. Statistical methods in hydrology. Ames,
Iowa, Iowa State University Press.
HAAS, C.N. and SCHEFF, P.A. 1990. Estimation of
averages in truncated samples. Environmental Science and
Technology, Vol. 24, No. 6, pp. 91219.
HELSEL, D.R. 1990. Less than obvious: statistical treat-
ment of data below the detection limit. Environ. Sci. and
Technol., Vol. 24, No. 12, pp. 176774.
HELSEL, D.R. and COHN, T.A. 1988. Estimation of
descriptive statistics for multiple censored water
quality data. Water Resources Research, Vol. 24, No. 12,
pp. 19972004.
HIRSCH, R.M. 1979. Synthetic hydrology and water
supply reliability. Water Resources Research, Vol. 15,
No. 6, pp. 160315.
HIRSCH, R.M. and STEDINGER, J.R. 1987. Plotting
positions for historical floods and their precision. Water
Resources Research, Vol. 23, No. 4, pp. 71527.
HOSHI, K.; BURGES, S.J. and YAMAOKA, I. 1978.
Reservoir design capacities for various seasonal operational
hydrology models. Proceedings of the Japanese Society of Civil
Engineers, No. 273, pp. 12134.
HOSHI, K.; STEDINGER, J.R. and BURGES, S. 1984.
Estimation of log normal quantiles: Monte Carlo results
and first-order approximations. Journal of Hydrology,
Vol. 71, Nos 12, pp. 130.
HOSKING, J.R.M. 1990. L-moments: analysis and
estimation of distributions using linear combinations of
order statistics. Journal of Royal Statistical Society, B,
Vol. 52, No. 2, pp. 10524.
HOSKING, J.R.M. and WALLIS, J.R. 1987. Parameter
and quantile estimation for the generalized Pareto distri-
bution. Technometrics, Vol. 29, No. 3, pp. 33949.
HOSKING, J.R.M. and WALLIS, J.R. 1995. A comparison
of unbiased and plotting-position estimators of
L-moments. Water Resources Research, Vol. 31, No. 8,
pp. 201925.
HOSKING, J.R.M. and WALLIS, J.R. 1997. Regional
frequency analysis: an approach based on L-moments.
Cambridge, Cambridge University Press.
HOSKING, J.R.M.; WALLIS, J.R. and WOOD, E.F. 1985.
Estimation of the generalized extreme-value distribution
by the method of probability weighted moments.
Technometrics, Vol. 27, No. 3, pp. 25161.
IACWD (Interagency Advisory Committee on Water
Data). 1982. Guidelines for determining flood flow frequency,
Bulletin 17B. Reston, Va., US Department of the
Interior, US Geological Survey, Office of Water Data
Coordination.
JENKINS, G.M. and WATTS, D.G. 1968. Spectral Analysis
and its Applications. San Francisco, Holden-Day.
KENDALL, M.G. and STUART, A. 1966. The advanced
theory of statistics, Vol. 3. New York, Hafner.
KIRBY, W. 1974. Algebraic boundness of sample statis-
tics. Water Resources Research, Vol. 10, No. 2, pp. 2202.
KIRBY, W. 1972. Computer oriented WilsonHilferty
transformation that preserves the first 3 moments and
lower bound of the Pearson Type 3 distribution. Water
Resources Research, Vol. 8, No. 5, pp. 12514.
KITE, G.W. 1988. Frequency and risk analysis in hydrology.
Littleton, Colo. Water Resources Publications.
KOTTEGODA, M. and ROSSO, R. 1997. Statistics, proba-
bility, and reliability for civil and environmental engineers.
New York, McGraw-Hill.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 225
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
KOUTSOYIANNIS, D. and MANETAS, A. 1996. Simple
disaggregation by accurate adjusting procedures. Water
Resources Research, Vol. 32, No. 7, pp. 210517.
KROLL, K. and STEDINGER, J.R. 1996. Estimation of
moments and quantiles with censored data. Water
Resources Research, Vol. 32, No. 4, pp. 100512.
KUCZERA, G. 1999. Comprehensive at-site flood
frequency analysis using Monte Carlo Bayesian inference.
Water Resources Research, Vol. 35, No. 5, pp. 15517.
KUCZERA, G. 1983. Effects of sampling uncertainty
and spatial correlation on an empirical Bayes procedure
for combining site and regional information. Journal of
Hydrology, Vol. 65, No. 4, pp. 37398.
KUCZERA, G. 1982. Combining site-specific and regional
information: an empirical Bayesian approach. Water
Resources Research, Vol. 18, No. 2, pp. 30614.
LANDWEHR, J.M.; MATALAS, N.C. and WALLIS, J.R.
1978. Some comparisons of flood statistics in real and log
space. Water Resources Research, Vol. 14, No. 5, pp. 90220.
LANDWEHR, J.M.; MATALAS, N.C. and WALLIS, J.R.
1979. Probability weighted moments compared with
some traditional techniques in estimating Gumbel
parameters and quantiles. Water Resources Research,
Vol. 15, No. 5, pp. 105564.
LANE, W. 1979. Applied stochastic techniques (users man-
ual). Denver, Colo., Bureau of Reclamation, Engineering
and Research Center, December.
LANGBEIN, W.B. 1949. Annual floods and the partial
duration flood series, EOS. Transactions of the American
Geophysical Union, Vol. 30, No. 6, pp. 87981.
LEDOLTER, J. 1978. The analysis of multivariate time
series applied to problems in hydrology. Journal of
Hydrology, Vol. 36, No. 34, pp. 32752.
LETTENMAIER, D.P. and BURGES, S.J. 1977a. Operational
assessment of hydrological models of long-term persistence.
Water Resources Research, Vol. 13, No. 1, pp. 11324.
LETTENMAIER, D.P. and BURGES, S.J. 1977b. An
operational approach to preserving skew in hydrological
models of long-term persistence. Water Resources
Research, Vol. 13, No. 2, pp. 28190.
226 Water Resources Systems Planning and Management
LETTENMAIER, D.P.; WALLIS, J.R. and WOOD, E.F.
1987. Effect of regional heterogeneity on flood frequency
estimation. Water Resources Research, Vol. 23, No. 2,
pp. 31324.
MACBERTHOUEX, P. and BROWN, L.C. 2002. Statistics
for environmental engineers, 2nd Edition. Boca Raton, Fla.,
Lewis, CRC Press.
MADSEN, H.; PEARSON, C.P.; RASMUSSEN, P.F. and
ROSBJERG, D. 1997a. Comparison of annual maximum
series and partial duration series methods for modelling
extreme hydrological events 1: at-site modelling. Water
Resources Research, Vol. 33, No. 4, pp. 74758.
MADSEN, H.; PEARSON, C.P. and ROSBJERG, D. 1997b.
Comparison of annual maximum series and partial duration
series methods for modelling extreme hydrological events 2:
regional modelling. Water Resources Research, Vol. 33,
No. 4, pp. 75970.
MADSEN, H. and ROSBJERG, D. 1997a. The partial
duration series method in regional index flood modelling.
Water Resources Research, Vol. 33, No. 4, pp. 73746.
MADSEN, H. and ROSBJERG, D. 1997b. Generalized
least squares and empirical Bayesian estimation in
regional partial duration series index-flood modelling.
Water Resources Research, Vol. 33, No. 4, pp. 77182.
MAHEEPALA, S. and PERERA, B.J.C. 1996. Monthly
hydrological data generation by disaggregation. Journal of
Hydrology, No. 178, 27791.
MARCO, J.B.; HARBOE, R. and SALAS, J.D. (eds). 1989.
Stochastic hydrology and its use in water resources
systems simulation and optimization. NATO ASI Series.
Dordrecht, Kluwer Academic.
MARTINS, E.S. and STEDINGER, J.R. 2000. Generalized
maximum likelihood GEV quantile estimators for
hydrological data. Water Resources Research, Vol. 36, No. 3,
pp. 73744.
MARTINS, E.S. and STEDINGER, J.R. 2001a. Historical
information in a GMLE-GEV framework with partial
duration and annual maximum series. Water Resources
Research, Vol. 37, No. 10, pp. 255157.
MARTINS, E.S. and STEDINGER, J.R. 2001b.
Generalized maximum likelihood Pareto-Poisson flood
wrm_ch07.qxd 8/31/2005 12:18 PM Page 226
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 227
risk analysis for partial duration series. Water Resources
Research, Vol. 37, No. 10, pp. 255967.
MATALAS, N.C. 1967. Mathematical assessment of
synthetic hydrology. Water Resources Research, Vol. 3,
No. 4, pp. 93745.
MATALAS, N.C. and WALLIS, J.R. 1973. Eureka! It fits a
Pearson type 3 distribution. Water Resources Research,
Vol. 9, No. 3, pp. 2819.
MATALAS, N.C. and WALLIS, J.R. 1976. Generation of
synthetic flow sequences. In: A.K. Biswas (ed.), Systems
approach to water management. New York, McGraw-Hill.
MEJIA, J.M. and ROUSSELLE, J. 1976. Disaggregation
models in hydrology revisited. Water Resources Research,
Vol. 12, No. 2, pp. 1856.
NERC (Natural Environmental Research Council)
1975. Flood studies report, Vol. 1: hydrological studies.
London.
NORTH, M. 1980. Time-dependent stochastic model of
floods. Journal of the Hydraulics Division, ASCE, Vol. 106,
No. HY5, pp. 64965.
OCONNELL, D.R.H.; OSTENAA, D.A.; LEVISH, D.R.
and KLINGER, R.E. 2002. Bayesian flood frequency
analysis with paleohydrological bound data. Water
Resources Research, Vol. 38, No. 5, pp. 16164.
OCONNELL, P.E. 1977. ARMA models in synthetic
hydrology. In: T.A. Ciriani, V. Maione and J.R. Wallis
(eds), Mathematical models for surface water hydrology.
New York, Wiley.
PRESS, W.H.; FLANNERY, B.P.; TEUKOLSKY, S.A. and
VETTERLING, W.T. 1986. Numerical recipes: the art
of scientific computing. Cambridge, UK, Cambridge
University Press.
RAIFFA, H. and SCHLAIFER, R. 1961. Applied statistical
decision theory. Cambridge, Mass., MIT Press.
RASMUSSEN, P.F. and ROSBJERG, D. 1989. Risk estima-
tion in partial duration series. Water Resources Research,
Vol. 25, No. 11, pp. 231930.
RASMUSSEN, P.F. and ROSBJERG, D. 1991a. Evaluation
of risk concepts in partial duration series. Stochastic
Hydrology and Hydraulics, Vol. 5, No. 1, pp. 116.
RASMUSSEN, P.F. and ROSBJERG, D. 1991b. Prediction
uncertainty in seasonal partial duration series. Water
Resources Research, Vol. 27, No. 11, pp. 287583.
RASMUSSEN, R.F.; SALAS, J.D.; FAGHERAZZI, L.;
RASSAM.; J.C. and BOBE, R. 1996. Estimation and
validation of contemporaneous PARMA models for
streamflow simulation. Water Resources Research, Vol. 32,
No. 10, pp. 315160.
ROBINSON, J.S. and SIVAPALAN, M. 1997. Temporal
scales and hydrological regimes: implications for flood
frequency scaling. Water Resources Research, Vol. 33,
No. 12, pp. 298199.
ROSBJERG, D. 1985. Estimation in Partial duration series
with independent and dependent peak values. Journal of
Hydrology, No. 76, pp. 18395.
ROSBJERG, D. and MADSEN, H. 1998. Design with
uncertain design values, In: H. Wheater and C. Kirby
(eds), Hydrology in a changing environment, New York,
Wiley. Vol. 3, pp. 15563.
ROSBJERG, D.; MADSEN, H. and RASMUSSEN, P.F.
1992. Prediction in partial duration series with general-
ized Pareto-distributed exceedances. Water Resources
Research, Vol. 28, No. 11, pp. 300110.
SALAS, J.D. 1993. Analysis and modelling of hydrological
time series. In: D. Maidment (ed.), Handbook of hydrology,
Chapter 17. New York, McGraw-Hill.
SALAS, J.D.; DELLEUR, J.W.; YEJEVICH, V. and
LANE, W.L. 1980. Applied modelling of hydrological time
series. Littleton, Colo., Water Resources Press Publications.
SALAS, J.D. and FERNANDEZ, B. 1993. Models for data
generation in hydrology: univariate techniques. In:
J.B. Marco, R. Harboe and J.D. Salas (eds), Stochastic
hydrology and its use in water resources systems simulation
and optimization, Dordrecht, Kluwer Academic. pp. 7695.
SALAS, J.D. and OBEYSEKERA, J.T.B. 1992. Conceptual
basis of seasonal streamflow time series. Journal of
Hydraulic Engineering, Vol. 118, No. 8, pp. 118694.
SCHAAKE, J.C. and VICENS, G.J. 1980. Design length
of water resource simulation experiments. Journal of
Water Resources Planning and Management, Vol. 106, No. 1,
pp. 33350.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 227
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
SLACK, J.R.; WALLIS, J.R. and MATALAS, N.C. 1975.
On the value of information in flood frequency analysis.
Water Resources Research, Vol. 11, No. 5, pp. 62948.
SMITH, R.L. 1984. Threshold methods for sample
extremes. In: J. Tiago de Oliveira (ed.), Statistical extremes
and applications, Dordrecht, D. Reidel. pp. 62138.
STEDINGER, J.R. 1980. Fitting log normal distributions
to hydrological data. Water Resources Research, Vol. 16,
No. 3, pp. 48190.
STEDINGER, J.R. 1981. Estimating correlations in
multivariate streamflow models. Water Resources Research,
Vol. 17, No. 1, pp. 20008.
STEDINGER, J.R. 1983. Estimating a regional flood
frequency distribution. Water Resources Research, Vol. 19,
No. 2, pp. 50310.
STEDINGER, J.R. 1997. Expected probability and annual
damage estimators. Journal of Water Resources Planning
and Management, Vol. 123, No. 2, pp. 12535. [With
discussion, Leo R. Beard, 1998. Journal of Water Resources
Planning and Management, Vol. 124, No. 6, pp. 36566.]
STEDINGER, J.R. 2000. Flood frequency analysis and
statistical estimation of flood risk. In: E.E. Wohl (ed.),
Inland flood hazards: human, riparian and aquatic communi-
ties, Chapter 12. Stanford, UK, Cambridge University Press.
STEDINGER, J.R. and BAKER, V.R. 1987. Surface water
hydrology: historical and paleoflood information. Reviews
of Geophysics, Vol. 25, No. 2, pp. 11924.
STEDINGER, J.R. and COHN, T.A. 1986. Flood
Frequency analysis with historical and paleoflood
information. Water Resources Research, Vol. 22, No. 5,
pp. 78593.
STEDINGER, J.R. and LU, L. 1995. Appraisal of regional
and index flood quantile estimators. Stochastic Hydrology
and Hydraulics, Vol. 9, No. 1, pp. 4975.
STEDINGER, J.R.; PEI, D. and COHN, T.A. 1985. A
disaggregation model for incorporating parameter
uncertainty into monthly reservoir simulations. Water
Resources Research, Vol. 21, No. 5, pp. 6657.
STEDINGER, J.R. and TAYLOR, M.R. 1982a. Synthetic
streamflow generation. Part 1: model verification and
228 Water Resources Systems Planning and Management
validation. Water Resources Research, Vol. 18, No. 4,
pp. 91924.
STEDINGER, J.R. and TAYLOR, M.R. 1982b. Synthetic
streamflow generation. Part 2: effect of parameter
uncertainty. Water Resources Research, Vol. 18, No. 4,
pp. 91924.
STEDINGER, J.R. and VOGEL, R. 1984. Disaggregation
procedures for the generation of serially correlated flow
vectors. Water Resources Research, Vol. 20, No. 1, pp. 4756.
STEDINGER, J.R.; VOGEL, R.M. and FOUFOULA-
GEORGIOU, E. 1993. Frequency analysis of extreme
events, In: D. Maidment (ed.), Handbook of hydrology,
Chapter 18. New York, McGraw-Hill.
STEPHENS, M. 1974. Statistics for Goodness of Fit, Journal
of the American Statistical Association, Vol. 69, pp. 73037.
TAO, P.C. and DELLEUR, J.W. 1976. Multistation, multi-
year synthesis of hydrological time series by disaggregation.
Water Resources Research, Vol. 12, No. 6, pp. 130311.
TARBOTON, D.G.; SHARMA, A. and LALL, U. 1998.
Disaggregation procedures for stochastic hydrology based
on nonparametric density estimation. Water Resources
Research, Vol. 34, No. 1, pp. 10719.
TASKER, G.D. and STEDINGER, J.R. 1986. Estimating
generalized skew with weighted least squares regression.
Journal of Water Resources Planning and Management,
Vol. 112, No. 2, pp. 22537.
THOM, H.C.S. 1958. A note on the gamma distribution.
Monthly Weather Review, Vol. 86, No. 4, 1958. pp. 11722.
THOMAS, H.A., JR. and FIERING, M.B. 1962.
Mathematical synthesis of streamflow sequences for
the analysis of river basins by simulation. In: A. Maass,
M.M. Hufschmidt, R. Dorfman, H.A. Thomas, Jr., S.A.
Marglin and G.M. Fair (eds), Design of water resources
systems. Cambridge, Mass., Harvard University Press.
THYER, M. and KUCZERA, G. 2000. Modelling
long-term persistence in hydroclimatic time series using
a hidden state Markov model. Water Resources Research,
Vol. 36, No. 11, pp. 330110.
VALDES, J.R.; RODRIGUEZ, I. and VICENS, G. 1977.
Bayesian generation of synthetic streamflows 2: the
wrm_ch07.qxd 8/31/2005 12:18 PM Page 228
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Concepts in Probability, Statistics and Stochastic Modelling 229
multivariate case. Water Resources Research, Vol. 13,
No. 2, pp. 29195.
VALENCIA, R. and SCHAAKE, J.C., Jr. 1973.
Disaggregation processes in stochastic hydrology. Water
Resources Research, Vol. 9, No. 3, pp. 5805.
VICENS, G.J.; RODRGUEZ-ITURBE, I. and SCHAAKE,
J.C., Jr. 1975. A Bayesian framework for the use of
regional information in hydrology. Water Resources
Research, Vol. 11, No. 3, pp. 40514.
VOGEL, R.M. 1987. The probability plot correlation
coefficient test for the normal, lognormal, and Gumbel
distributional hypotheses. Water Resources Research,
Vol. 22, No. 4, pp. 58790.
VOGEL, R.M. and FENNESSEY, N.M. 1993. L-moment
diagrams should replace product moment diagrams.
Water Resources Research, Vol. 29, No. 6, pp. 174552.
VOGEL, R.M. and MCMARTIN, D.E. 1991. Probability
plot goodness-of-fit and skewness estimation procedures
for the Pearson type III distribution. Water Resources
Research, Vol. 27, No. 12, pp. 314958.
VOGEL, R.M. and SHALLCROSS, A.L. 1996. The
moving blocks bootstrap versus parametric time series
models. Water Resources Research, Vol. 32, No. 6,
pp. 187582.
VOGEL, R.M. and STEDINGER, J.R. 1988. The value of
stochastic streamflow models in over-year reservoir
design applications. Water Resources Research, Vol. 24,
No. 9, pp. 148390.
WALLIS, J.R. 1980. Risk and uncertainties in the evalua-
tion of flood events for the design of hydraulic structures.
In: E. Guggino, G. Rossi and E. Todini (eds), Piene e Siccita,
Catania, Italy, Fondazione Politecnica del Mediterraneo.
pp. 336.
WALLIS, J.R.; MATALAS, N.C. and SLACK, J.R. 1974a.
Just a Moment! Water Resources Research, Vol. 10, No. 2,
pp. 21119.
WALLIS, J.R.; MATALAS, N.C. and SLACK, J.R. 1974b.
Just a moment! Appendix. Springfield, Va., National
Technical Information Service, PB-231 816.
WANG, Q. 2001. A. Bayesian joint probability approach
for flood record augmentation. Water Resources Research,
Vol. 37, No. 6, pp. 170712.
WANG, Q.J. 1997. LH moments for statistical analysis of
extreme events. Water Resources Research, Vol. 33, No. 12,
pp. 284148.
WILK, M.B. and GNANADESIKAN, R. 1968. Probability
plotting methods for the analysis of data. Biometrika,
Vol. 55, No. 1, pp. 117.
WILKS, D.S. 2002. Realizations of daily weather in fore-
cast seasonal climate. Journal of Hydrometeorology, No. 3,
pp. 195207.
WILKS, D.S. 1998. Multi-site generalization of a daily
stochastic precipitation generation model. Journal of
Hydrology, No. 210, pp. 17891.
YOUNG, G.K. 1968. Discussion of Mathematical assess-
ment of synthetic hydrology by N.C. Matalas and reply.
Water Resources Research, Vol. 4, No. 3, pp. 6813.
ZELLNER, A. 1971. An introduction to Bayesian
inference in econometrics, New York, Wiley.
ZRINJI, Z. and BURN, D.H. 1994. Flood Frequency
analysis for ungauged sites using a region of influence
approach. Journal of Hydrology, No. 13, pp. 121.
wrm_ch07.qxd 8/31/2005 12:18 PM Page 229
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
8. Modelling Uncertainty
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
wrm_ch08.qxd 8/31/2005 11:55 AM Page 230
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Water resources planners and managers work in an envi-
ronment of change and uncertainty. Water supplies are
always uncertain, if not in the short term at least in the
long term. Water demands and the multiple purposes
and services water provide are always changing, and
these changes cannot always be predicted. Many of the
values of parameters of models used to predict the multi-
ple hydrological, economic, environmental, ecological
and social impacts are also changing and uncertain.
Indeed models used to predict these impacts are, at least
in part, based on many imprecise assumptions. Planning
and managing, given this uncertainty, cannot be avoided.
To the extent that probabilities can be assigned to var-
ious uncertain inputs or parameter values, some of this
uncertainty can be incorporated into models. These mod-
els are called probabilistic or stochastic models. Most prob-
abilistic models provide a range of possible values for
each output variable along with their probabilities.
Stochastic models attempt to model the random processes
that occur over time, and provide alternative time series
of outputs along with their probabilities. In other cases,
sensitivity analyses (solving models under different
assumptions) can be carried out to estimate the impact of
any uncertainty on the decisions being considered. In
some situations, uncertainty may not significantly affect
the decisions that should be made. In other situations it
will. Sensitivity analyses can help estimate the extent to
231
Modelling Uncertainty
Decisionmakers are increasingly willing to consider the uncertainty associated
with model predictions of the impacts of their possible decisions. Information on
uncertainty does not make decisionmaking easier, but to ignore it is to ignore
reality. Incorporating what is known about the uncertainty into input parameters
and variables used in optimization and simulation models can help in quantifying
the uncertainty in the resulting model predictions the model output. This chapter
discusses and illustrates some approaches for doing this.
8
which we need to try to reduce that uncertainty. Model
sensitivity and uncertainty analysis is discussed in more
detail in Chapter 9.
This chapter introduces a number of approaches to
probabilistic optimization and simulation modelling.
Probabilistic models will be developed and applied to
some of the same water resources management problems
used to illustrate deterministic modelling in previous
chapters. They can be, and have been, applied to numer-
ous other water resources planning and management
problems as well. The purpose here, however, is simply to
illustrate some of the commonly used approaches to the
probabilistic modelling of water resources system design
and operating problems.
2. Generating Values From Known
Probability Distributions
Variables whose values cannot be predicted with certainty
are called random variables. Often, inputs to hydrological
simulation models are observed or synthetically generated
values of rainfall or streamflow. Other examples of such
random variables could be evaporation losses, point and
non-point source wastewater discharges, demands for
water, spot prices for energy that may impact the amount
of hydropower production, and so on. For random
processes that are stationary that is, the statistical attrib-
utes of the process are not changing and if there is
wrm_ch08.qxd 8/31/2005 11:55 AM Page 231
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
232 Water Resources Systems Planning and Management
lower and upper limits is equally likely. Using Equation 8.2,
together with a series of uniformly distributed (all equally
likely) values of p* over the range from 0 to 1 (that is, along
the vertical axis of Figure 8.2), one can generate a corre-
sponding series of variable values, r*, associated with any
distribution. These random variable values will have a
cumulative distribution as shown in Figure 8.2, and hence
a density distribution as shown in Figure 8.1, regardless of
the types or shapes of those distributions. The mean,
variance and other moments of the distributions will be
maintained.
The mean and variance of continuous distributions are:

r f
R
(r)dr E[R] (8.3)

(r E[R])
2
f
R
(r)dr Var[R] (8.4)
The mean and variance of discrete distributions having
possible values denoted by r
i
with probabilities p
i
are:
(8.5)
(8.6)
If a time series of T random variable values, r
t
, from the
same stationary random variable, R, exists, then the serial
or autocorrelations of r
t
and r
tk
in this time series for any
positive integer k can be estimated using:

R
(k)
(8.7)
[( [ ])( [ ])] ( [ ]) r E R r E R r E R
k
T k
t
t T

1
2
1

r E R p R
i
i
i
[ ] [ ]

2
Var
r p E R
i i
i
[ ]

r
(
)
r
f
R
r *
* p
E
0
2
0
1
1
0
a
Figure 8.1. Probability density distribution of a random
variable R. The probability that R is less than or equal r* is p*.
Figure 8.2. Cumulative distribution function of a random
variable R showing the probability of any observed
value of R being less than or equal to a given value r.
The probability of an observed value of R being less than or
equal to r* is p*.
no serial correlation in the spatial or temporal sequence
of observed values, then such random processes can be
characterized by single probability distributions. These
probability distributions are often based on past observa-
tions of the random variables. These observations or
measurements are used either to define the probability
distribution itself or to estimate parameter values of an
assumed type of distribution.
Let R be a random variable whose probability density
distribution, f
R
(r), is as shown in Figure 8.1. This distri-
bution indicates the probability or likelihood of an
observed value of the random variable R being between
any two values of r on the horizontal axis. For example,
the probability of an observed value of R being between 0
and r* is p*, the shaded area to the left of r*. The entire
shaded area of a probability density distribution, such as
shown in Figure 8.1, is 1.
Integrating this function over r converts the density
function to a cumulative distribution function, F
R
(r*),
ranging from 0 to 1, as illustrated in Figure 8.2.
(8.1)
Given any value of p* from 0 to 1, one can find its corre-
sponding variable value r* from the inverse of the cumu-
lative distribution function.
F
R
1
( p*) r* (8.2)
From the distribution shown in Figure 8.1, it is obvious
that the likelihood of different values of the random
variable varies; ones in the vicinity of r* are much
more likely to occur than are values at the tails of the
distribution. A uniform distribution is one that looks like a
rectangle; any value of the random variable between its
f r r R r F r
R R
r
( ) Pr( *) ( *)
*
d
0

wrm_ch08.qxd 8/31/2005 11:55 AM Page 232

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Uncertainty 233
The probability density and corresponding cumulative
probability distributions can be of any shape, not just
those named distributions commonly found in proba-
bility and statistics books.
The process of generating a time sequence t 1, 2,
of inputs, r
t
, from the probability distribution of a random
variable R where the lag 1 serial correlation,
R
(1) , is
to be preserved is a little more complex. The expected
value of the random variable R
t1
depends on the
observed value, r
t
, of the random variable R
t
, together
with the mean of the distribution, E[R], and the correla-
tion coefficient . If there is no correlation ( is 0), then
the expected value of R
t1
is the mean of the population,
E[R]. If there is perfect correlation ( is 1), then the
expected value of R
t1
is r
t
. In general, the expected value
of R
t1
given an observed value r
t
of R
t
is:
(8.8)
The variance of the random variable R
t1
depends on the
variance of the distribution, Var[R], and the lag 1 correla-
tion coefficient, .
(8.9)
If there is perfect correlation ( 1), then the process is
deterministic, there is no variance, and r
t1
r
t
. The
value for r
t1
is r
t
. If there is no correlation that is, serial
correlation does not exist ( 0) then the generated
value for r
t1
is its mean, E[R], plus some randomly gen-
erated deviation from a normal distribution having a mean
of 0 and a standard deviation of 1, denoted as N(0, 1). In
this case the value r
t1
is not dependent on r
t
.
When the serial correlation is more than 0 but less
than 1, then both the correlation and the standard devia-
tion (the square root of the variance) influence the value
of r
t1
. A sequence of random variable values from a
multivariate normal distribution that preserves the mean,
E[R]; overall variance, Var[R]; standard deviation , and
lag 1 correlation ; can be obtained from.
(8.10)
The term Z in Equation 8.10 is a random number gener-
ated from a normal distribution having a mean of 0 and
a variance of 1. The process involves selecting a
random number from a uniform distribution ranging
from 0 to 1, and using it in Equation 8.2 for an N(0, 1)
distribution to obtain a value of random number for
r E R r E R Z
t t

1
2 1 2
1 [ ] ( [ ]) ( )
/

Var Var [ | ] [ ]( ) R R r R
t t t

1
2
1
E R R r E R r E R
t t t t
[ | ] [ ] ( [ ])

1

use in Equation 8.10. This positive or negative number is
substituted for the term Z in Equation 8.10 to obtain a
value r
t1
. This is shown on the graph in Figure 8.3.
Simulation models that have random inputs, such as a
series of r
t
values, will generally produce random outputs.
After many simulations, the probability distributions of
each random output variable value can be defined. These
can be used to estimate reliabilities and other statistical
characteristics of those output distributions. This process
of generating multiple random inputs for multiple simu-
lations to obtain multiple random outputs is called Monte
Carlo simulation.
3. Monte Carlo Simulation
To illustrate Monte Carlo simulation, consider the water
allocation problem involving three firms, each of which
receives a benefit, B
i
(x
it
), from the amount of water, x
it
,
allocated to it in each period t. This situation is shown in
Figure 8.4. Monte Carlo simulation can be used to find
the probability distribution of the benefits to each firm
associated with the firms allocation policy.
Suppose the policy is to keep the first two units of flow
in the stream, to allocate the next three units to Firm 3,
and the next four units to Firms 1 and 2 equally. The
remaining flow is to be allocated to each of the three firms
equally up to the limits desired by each firm, namely 3.0,
2.33, and 8.0 respectively. Any excess flow will remain in
the stream. The plots in Figure 8.5 illustrate this policy.
Each allocation plot reflects the priorities given to the
three firms and the users further downstream.
E
0
2
0
1
1
0
c r 0
0
t
r
t
r
t
+
1
(r * E[ ] + R
t
-E[ ]) R
E[ ] R

1
*
N(0, Var[R] (1-
2
))
r * E[R
t
=
t+1
R
t
]
Figure 8.3. Diagram showing the calculation of a sequence of
values of the random variable R from a multivariate normal
distribution in a way that preserves the mean, variance and
correlation of the random variable.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 233
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A simulation model can be created. In each of a
series of discrete time periods t, the flows Q
t
are drawn
from a probability distribution, such as from Figure 8.2
using Equation 8.2. Once this flow is determined, each
successive allocation, x
it
, is computed. Once an allocation
is made it is subtracted from the streamflow and the
next allocation is made on the basis of that reduced
234 Water Resources Systems Planning and Management
streamflow, in accordance with the allocation policy
defined in Figures 8.5a d. After numerous time steps,
the probability distributions of the allocations to each of
the firms can be defined.
Figure 8.6 shows a flow chart for this simulation
model.
Q
x
x
x
E
0
2
0
1
1
0
d
t
Q
1t
1t
Q
2t
Q
3t
2t
3t
3
firm 3
B ( 8x -0.5x
2
x
3t
)=
3t 3t
1
firm 1
B ( 6x -x
2
x
1t
)=
1t 1t
2
firm 2
B ( 7x - .5x 1
2
x
2t
)=
2t 2t
Figure 8.4. Streamflow
allocations in each period t
result in benefits, B
i
(x
it
), to each
firm i. The flows, Q
it
, at each
diversion site i are the random
flows Q
t
less the upstream
withdrawals, if any.
E
0
2
0
1
1
0
e
12
10
8
6
4
2
0
0 2 4 6 8 10 12 14 16 26
a
l
l
o
c
a
t
i
o
n

a
t

d
i
v
e
r
s
i
o
n
s
i
t
e
:
x
t
1
1
streamflow Q
1t
18 20 22 24
Figure 8.5a. Water allocation policy for Firm 1 based on the
flow at its diversion site. This policy applies for each period t.
E
0
2
0
1
1
0
f
12
10
8
6
4
2
0
0 2 4 6 8 10 12 14 16 26
a
l
l
o
c
a
t
i
o
n

a
t

d
i
v
e
r
s
i
o
n
s
i
t
e
:
x
t
2
2
streamflow Q
2t
18 20 22 24
Figure 8.5b. Water allocation policy for Firm 2 based on
the flow at its diversion site for that firm. This policy
applies for each period t.
E
0
2
0
1
1
0
g
12
10
8
6
4
2
0
0 2 4 6 8 10 12 14 16 26
a
l
l
o
c
a
t
i
o
n

a
t

d
i
v
e
r
s
i
o
n
s
i
t
e
:
x
t
3
3
streamflow Q
3t
18 20 22 24
Figure 8.5c. Water allocation policy for Firm 3 based on
the flow at its diversion site. This policy applies for each
period t.
E
0
2
0
1
1
0
h
12
10
8
6
4
2
0
0 2 4 6 8 10 12 14 16 26
s
t
r
e
a
m
f
l
o
w
a
t
s
i
t
e

3
streamflow Q
3t
18 20 22 24
Figure 8.5d. Streamflow downstream of site 3 given the
streamflow Q
3t
at site 3 before the diversion. This applies for
each period t.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 234
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Uncertainty 235
E
0
2
0
1
1
0
j
stop
start
yes no
calculate and
plot probability
distributions
compute Q
t t = +1
set: t = 1
T
max
max
t =T
data
storage
t
(Eq.2) Q
1t
= Q
t
compute
(fig. 5a)
x
1t
compute
(fig. 5c)
x
3t
compute
(fig. 5b)
x
2t
Q
2t
= Q
1t
- x
1t
Q
3t
= Q
2t
- x
2t
Figure 8.6. Monte Carlo
simulation to determine
probability distributions of
allocations to each of three
water users, as illustrated in
Figure 8.4. The dashed lines
represent information (data)
flows.
Having defined the probability distribution of the
allocations, based on the allocation policy, one can
consider each of the allocations as random variables, X
1
,
X
2
, and X
3
for Firms 1, 2 and 3 respectively.
4. Chance Constrained Models
For models that include random variables, it may be
appropriate in some situations to consider constraints that
do not have to be satisfied all the time. Chance constraints
specify the probability of a constraint being satisfied, or
the fraction of the time a constraint has to apply.
Consider, for example, the allocation problem shown in
Figure 8.4. For planning purposes, the three firms may
want to set allocation targets, not expecting to have those
targets met 100% of the time. To ensure, for example, that
an allocation target, T
i
, of firm i will be met at least 90%
of the time, one could write the chance constraint:
Pr{T
i
X
i
} 0.90 i 1, 2 and 3 (8.11)
In this constraint, the allocation target T
i
is an unknown
decision-variable, and X
i
is a random variable whose
distribution has just been computed and is known.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 235
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
To include chance constraints in optimization
models, their deterministic equivalents must be defined.
The deterministic equivalents of the three chance con-
straints in Equation 8.11 are:
T
i
x
it
0.10
i 1, 2 and 3 (8.12)
where x
it
0.10
is the particular value of the random variable
X
i
that is equalled or exceeded 90% of the time. This
value is shown on the probability distribution for X
i
in
Figure 8.7.
To modify the allocation problem somewhat, assume
the benefit obtained by each firm is a function of its tar-
get allocation and that the same allocation target applies
in each time period t. The equipment and labour used in
the firm is presumably based on the target allocations.
Once the target is set, assume there are no benefits gained
by excess water allocations. If the benefits obtained are
to be based on the target allocations rather than
the actual allocations, then the optimization problem is
one of finding the values of the three targets that maximize
the total benefits obtained with a reliability of, say, at
least 90%.
Maximize6T
1
T
1
2
7T
2
1.5T
2
2
8T
3
0.5T
3
2

(8.13)
subject to:
Pr{T
1
T
2
T
3
[Q
t
min(Q
t
, 2)]} 0.90
for all periods t (8.14)
where Q
t
is the random streamflow variable upstream of
all diversion sites. If the same unconditional probability
236 Water Resources Systems Planning and Management
distribution of Q
t
applies for each period t, then only one
Equation 8.14 is needed.
Assuming the value of the streamflow, q
t
0.10
, that is
equalled or exceeded 90% of the time, is greater than 2
(the amount that must remain in the stream), the deter-
ministic equivalent of chance constraint Equation 8.14 is:
T
1
T
2
T
3
q
t
0.10
minq
t
0.10
, 2 (8.15)
The value of the flow that is equal to or exceeds 90%
of the time, q
t
0.10
, can be obtained from the cumulative
distribution of flows as illustrated in Figure 8.8.
Assume this 90% reliable flow is 8. The deterministic
equivalent of the chance constraint Equation 8.9 for all
periods t is simply T
1
T
2
T
3
6. The optimal
solution of the chance-constrained target allocation
model, Equations 8.8 and 8.9, is, as seen before, T
1
1,
T
2
1 and T
3
4. The next step would be to simulate
this problem to see what the actual reliabilities might be
for various sequences of flows q
t.
5. Markov Processes and Transition
Probabilities
Time-series correlations can be incorporated into
models using transition probabilities. To illustrate this
process, consider the observed flow sequence shown in
Table 8.1.
The estimated mean, variance and correlation coeffi-
cient of the observed flows shown in Table 8.1 can be
calculated using Equations 8.16, 8.17 and 8.18.
E
0
2
0
1
1
0
k
(
X
f
x
i
i
t
)
X
it
0. 0 1
0.90
X
it
Figure 8.7. Probability density distribution of the random
allocation X
i
to firm i. The particular allocation value x
it
0.10
has
a 90% chance of being equalled or exceeded, as indicated by
the shaded region.
E
0
2
0
1
1
0
m
q
t
0. 0 1
0. 0 1
1.00
Pr( q
t
Q ) = F
Q
(q
t
)
q
t
Figure 8.8. Example cumulative probability distribution
showing the particular value of the random variable, q
t
0.10
,
that is equalled or exceeded 90% of the time.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 236
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Table 8.1. Sequence of flows for thirty-one time periods t.
Modelling Uncertainty 237
E[Q] q
t
/31 3.155 (8.16)
Var[Q] (q
t
3.155)
2
/30 1.95 (8.17)
Lag-one correlation coefficient
(8.18)
The probability distribution of the flows in Table 8.1
can be approximated by a histogram. Histograms can be
created by subdividing the entire range of random vari-
able values, such as flows, into discrete intervals. For
example, let each interval be two units of flow. Counting
the number of flows in each interval and then dividing
those interval counts by the total number of counts results
in the histogram shown in Figure 8.9. In this case, just to
compare this with what will be calculated later, the first
flow, q
1
, is ignored.
Figure 8.9 shows a uniform unconditional probability
distribution of the flow being in any of the possible dis-
crete flow intervals. It does not show the possible
dependency of the probabilities of the random variable

( . )( . )
( . ) .
q q
q
t t
t
1
3 155 3 155
3 155 0 50
1
30
1
31
2

1
]
1
1
31

1
31

value, q
t1
, in period t 1 on the observed random vari-
able value, q
t
, in period t. It is possible that the probabil-
ity of being in a flow interval j in period t 1 depends on
the actual observed flow interval i in period t.
To see if the probability of being in any given interval
of flows is dependent on the past flow interval, one can
create a matrix. The rows of the matrix are the flow inter-
vals i in period t. The columns are the flow intervals j in
the following period t 1. Such a matrix is shown in
Table 8.2. The numbers in the matrix are based on the
flows in Table 8.1 and indicate the number of times a flow
in interval j followed a flow in interval i.
Given an observed flow in an interval i in period t, the
probabilities of being in one of the possible intervals j in
the next period t 1 must sum to 1. Thus, each number
in each row of the matrix in Table 8.2 can be divided by
the total number of flow transitions in that row (the sum
period
t
1
2
3
4
5
6
7
8
9
10
E
0
2
0
8
2
7
w
flow
Q
t
4.5
5.2
6.0
3.2
4.3
5.1
3.6
4.5
1.8
1.5
period
t
period
t
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
flow
Q
flow
Q
t t
1.8
2.5
2.3
1.8
1.2
1.9
2.5
4.1
4.7
5.6
1.8
1.2
2.5
1.9
3.2
2.5
3.5
2.7
1.5
4.1
4.8
E
0
2
0
1
1
0
n q
t
p
r
o
b
a
b
i
l
i
t
y
0 2 4 6 8 10
Figure 8.9. Histogram showing an equal 1/3 probability that
the values of the random variable Q
t
will be in any one of the
three two-flow unit intervals.
E
0
2
0
8
2
7
x
flow interval
in : t i
1
2
3
5
3
2
flow interval
in + t
4
4
2
1
3
6
1 2 3
j =
Table 8.2. Matrix showing the number of times a flow in
interval i in period t was followed by a flow in interval j in
period t 1.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 237
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
of the number of flows in the row) to obtain the proba-
bilities of being in each interval j in t 1 given a flow
in interval i in period t. In this case there are ten flows
that followed each flow interval i, hence by dividing
each number in each row of the matrix by 10 defines the
transition probabilities P
ij
.
P
ij
Pr{Q
t1
in interval j | Q
t
in interval i} (8.19)
These conditional or transition probabilities, shown in
Table 8.3, correspond to the number of transitions shown
in Table 8.2.
Table 8.3 is a matrix of transition probabilities. The
sum of the probabilities in each row equals 1. Matrices
of transition probabilities whose rows sum to 1 are also
called stochastic matrices or first-order Markov chains.
If each rows probabilities were the same, this would
indicate that the probability of observing any flow inter-
val in the future is independent of the value of previous
flows. Each row would have the same probabilities as the
unconditional distribution shown in Figure 8.9. In this
example the probabilities in each row differ, showing that
low flows are more likely to follow low flows, and high
flows are more likely to follow high flows. Thus the flows
in Table 8.1 are positively correlated, as indeed has
already determined from Equation 8.18.
Using the information in Table 8.3, one can compute
the probability of observing a flow in any interval at any
period on into the future given the present flow interval.
This can be done one period at a time. For example assume
the flow in the current time period t 1 is in interval
i 3. The probabilities, PQ
j,2
, of being in any of the three
238 Water Resources Systems Planning and Management
intervals in the following time period t 2 are the proba-
bilities shown in the third row of the matrix in Table 8.3.
The probabilities of being in an interval j in the follow-
ing time period t 3 is the sum over all intervals i of the
joint probabilities of being in interval i in period t 2 and
making a transition to interval j in period t 3.
Pr{Q
3
in interval j} PQ
j,3
Pr{Q
2
in interval i}
Pr{Q
3
in interval j | Q
2
in interval i} (8.20)
The last term in Equation 8.20 is the transition probabil-
ity, from Table 8.3, that in this example remains the same
for all time periods t. These transition probabilities,
Pr{Q
t1
in interval j | Q
t
in interval i} can be denoted as P
ij
.
Referring to Equation 8.19, Equation 8.20 can be
written in a general form as:
PQ
j,t1
PQ
it
P
ij
for all intervals j and periods t
(8.21)
This operation can be continued to any future time period.
Table 8.4 illustrates the results of such calculations for
i

E
0
2
0
9
1
1
b
1
2
3
flow interval in
t + : j
0.5
0.3
0.2
1 2 3
0.4
0.4
0.2
0.1
0.3
0.6
flow interval
in t: i
Table 8.3. Matrix showing the probabilities P
ij
of having a flow
in interval j in period t 1 given an observed flow in interval i
in period t.
E
0
2
0
8
2
7
y
t
i
m
e

p
e
r
i
o
d
t
1
2
3
4
5
6
7
8
flow interval i
0
0.2
0.28
0.312
0.325
0.330
0.332
0.333
1 2 3
probability PQ
it
0
0.2
0.28
0.312
0.325
0.330
0.332
0.333
1
0.6
0.44
0.376
0.350
0.340
0.336
0.334
Table 8.4. Probabilities of observing a flow in any flow interval i
in a future time period t given a current flow in interval i 3.
These probabilities are derived using the transition probabilities
P
ij
in Table 8.3 in Equation 8.21 and assuming the flow interval
observed in Period 1 is in Interval 3.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 238
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Uncertainty 239
up to six future periods, given a present period (t 1)
flow in interval i 3.
Note that as the future time period t increases, the
flow interval probabilities are converging to the uncondi-
tional probabilities in this example 1/3, 1/3, 1/3 as
shown in Figure 8.9. The predicted probability of
observing a future flow in any particular interval at some
time in the future becomes less and less dependent on
the current flow interval as the number of time periods
increases between the current period and that future time
period.
When these unconditional probabilities are reached,
PQ
it
will equal PQ
i,t1
for each flow interval i. To find
these unconditional probabilities directly, Equation 8.21
can be written as:
PQ
j
PQ
i
P
ij
for all intervals j (8.22)
Equation 8.22 (less one) along with Equation 8.23 can be
used to calculate all the unconditional probabilities PQ
i
directly.
PQ
i
1 (8.23)
Conditional or transition probabilities can be incorpo-
rated into stochastic optimization models of water
resources systems.
6. Stochastic Optimization
To illustrate the development and use of stochastic opti-
mization models, consider first the allocation of water to a
single user. Assume the flow in the stream where the diver-
sion takes place is not regulated and can be described by a
known probability distribution based on historical records.
Clearly, the user cannot divert more water than is available
in the stream. A deterministic model would include the con-
straint that the diversion x cannot exceed the available water
Q. But Q is a random variable. Some target value, q, of the
random variable Q will have to be selected, knowing that
there is some probability that in reality, or in a simulation
model, the actual flow may be less than the selected value q.
Hence, if the constraint x q is binding, the actual alloca-
tion may be less than the value of the allocation or diversion
variable x produced by the optimization model.
i

If the value of x affects one of the systems performance

indicators, such as the net benefits, B(x), to the user, a
more accurate estimate of the users net benefits will be
obtained from considering a range of possible allocations
x, depending on the range of possible values of the ran-
dom flow Q. One way to do this is to divide the known
probability distribution of flows q into discrete ranges, i
each range having a known probability PQ
i
. Designate a
discrete flow q
i
for each range. Associated with each spec-
ified flow q
i
is an unknown allocation x
i
. Now the single
deterministic constraint x q can be replaced with the
set of deterministic constraints x
i
q
i
, and the term B(x)
in the original objective function can be replaced by its
expected value,
i
PQ
i
B(x
i
).
Note, when dividing a continuous known probability
distribution into discrete ranges, the discrete flows q
i
,
selected to represent each range i having a given proba-
bility PQ
i
, should be selected so as to maintain at least the
mean and variance of that known distribution as defined
by Equations 8.5 and 8.6.
To illustrate this, consider a slightly more complex
example involving the allocation of water to consumers
upstream and downstream of a reservoir. Both the poli-
cies for allocating water to each user and the reservoir
release policy are to be determined. This example prob-
lem is shown in Figure 8.10.
If the allocation of water to each user is to be based
on a common objective, such as the minimization of
the total sum, over time, of squared deviations from
pre-specified target allocations, each allocation in each
time period will depend in part on the reservoir storage
volume.
allocation d
t
user D
allocation u
t
user U
flow Q
t
initial
storage S
t
reservoir
capacity K
release R
t
E
0
2
0
1
1
0
o
Figure 8.10. Example water resources system involving water
diversions from a river both upstream and downstream of a
reservoir of known capacity.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 239
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Consider first a deterministic model of the above prob-
lem, assuming known river flows Q
t
and upstream and
downstream user allocation targets UT
t
and DT
t
in each of
T within-year periods t in a year. Assume the objective is
to minimize the sum of squared deviations from actual
allocations, u
t
and d
t
, and their respective target alloca-
tions, UT
t
and DT
t
in each within-year period t.
Minimize {(UT
t
u
t
)
2
(DT
t
d
t
)
2
} (8.24)
The constraints include:
a) Continuity of storage involving initial storage volumes
S
t
, net inflows Q
t
u
t
, and releases R
t
. Assuming no losses:
S
t
Q
t
u
t
R
t
S
t1
for each period t,
T 1 1 (8.25)
b) Reservoir capacity limitations. Assuming a known
active storage capacity K:
S
t
K for each period t (8.26)
c) Allocation restrictions. For each period t:
u
t
Q
t
(8.27)
d
t
R
t
(8.28)
Equations 8.25 and 8.28 could be combined to eliminate
the release variable R
t
, since in this problem knowledge of
the total release in each period t is not required. In this
case, Equation 8.25 would become an inequality.
The solution of this model, Equations 8.24 8.28,
would depend on the known variables (the targets UT
t
and DT
t
, flows Q
t
and reservoir capacity K). It would
identify the particular upstream and downstream alloca-
tions and reservoir releases in each period t. It would not
provide a policy that defines what allocations and releases
to make for a range of different inflows and initial storage
volumes in each period t. A backward-moving dynamic
programming model can provide such a policy. This
policy will identify the allocations and releases to make
based on various initial storage volumes, S
t
, and flows, Q
t
,
as discussed in Chapter 4.
This deterministic discrete dynamic programming
allocation and reservoir operation model can be written for
different discrete values of S
t
from 0 S
t
capacity K as:
F
t
n
(S
t
, Q
t
) min(UT
t
u
t
)
2
(DT
t
d
t
)
2
F
t1
n1
(S
t1
, Q
t1
)
t
T

240 Water Resources Systems Planning and Management

The minimization is over all feasible u
t
, R
t
, d
t
:
u
t
Q
t
R
t
S
t
Q
t
u
t
R
t
S
t
Q
t
u
t
K
d
t
R
t
S
t1
S
t
Q
t
u
t
R
t
(8.29)
There are three variables to be determined at each stage or
time period t in the above dynamic programming model.
These three variables are the allocations u
t
and d
t
and the
reservoir release R
t
. Each decision involves three discrete
decision-variable values. The functions F
t
n
(S
t
, Q
t
) define
the minimum sum of squared deviations given an initial
storage volume S
t
and streamflow Q
t
in time period or
season t with n time periods remaining until the end of
reservoir operation.
One can reduce this three decision-variable model
to a single variable model by realizing that, for any fixed
discrete pair of initial and final storage volume states,
there can be a direct tradeoff between the upstream and
downstream allocations, given the particular streamflow
in each period t. Increasing the upstream allocation will
decrease the resulting reservoir inflow, and this in turn
will reduce the release by the same amount. This reduces
the amount of water available to allocate to the down-
stream use.
Hence, for this example problem involving these
upstream and downstream allocations, a local optimiza-
tion can be performed at each time step t for each combi-
nation of storage states S
t
and S
t1
. This optimization
finds the allocation decision-variables u
t
and d
t
that
minimize(UT
t
u
t
)
2
(DT
t
d
t
)
2
(8.30)
where
u
t
Q
t
(8.31)
d
t
S
t
Q
t
u
t
S
t1
(8.32)
This local optimization can be solved to identify the u
t
and d
t
allocations for each feasible combination of S
t
and
S
t1
in each period t.
Given these optimal allocations, the dynamic pro-
gramming model can be simplified to include only one
discrete decision-variable, either R
t
or S
t1
. If the decision
variable S
t1
is used in each period t, the releases R
t
in
wrm_ch08.qxd 8/31/2005 11:55 AM Page 240
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Uncertainty 241
those periods t do not need to be considered. Thus the
dynamic programming model expressed by Equations
8.29 can be written for all discrete storage volumes S
t
from 0 to K and for all discrete flows Q
t
as:
F
t
n
(S
t
, Q
t
) min(UT
t
u
t
(S
t
, S
t1
))
2
(DT
t
d
t
(S
t
, S
t1
))
2
F
t1
n1
(S
t1
, Q
t1
)
The minimization is over all feasible discrete values of
S
t1
,
S
t1
K (8.33)
where the functions u
t
(S
t
, S
t1
) and d
t
(S
t
, S
t1
) have been
determined using Equations 8.30 8.32.
As the total number of periods remaining, n, increases,
the solution of this dynamic programming model will
converge to a steady or stationary state. The best final
storage volume S
t1
given an initial storage volume S
t
will
probably differ for each within-year period or season t,
but for a given season t it will be the same in successive
years. In addition, for each storage volume S
t
, streamflow,
Q
t
, and within-year period t, the difference between
F
t
nT
(S
t
, Q
t
) and F
t
n
(S
t
, Q
t
) will be the same constant
regardless of the storage volume S
t
and period t. This
constant is the optimal, in this case minimum, annual
value of the objective function, Equation 8.24.
There could be additional limits imposed on storage
variables and release variables, such as for flood control
storage or minimum downstream flows, as might be
appropriate in specific situations.
The above deterministic dynamic programming model
(Equation. 8.33) can be converted to a stochastic model.
Stochastic models consider multiple discrete flows as well
as multiple discrete storage volumes, and their probabili-
ties, in each period t. A common way to do this is to
assume that the sequence of flows follow a first-order
Markov process. Such a process involves the use of tran-
sition or conditional probabilities of flows as defined by
Equation 8.20.
To develop these stochastic optimization models, it is
convenient to introduce some additional indices or sub-
scripts. Let the index k denote different initial storage
volume intervals. These discrete intervals divide the con-
tinuous range of storage volume values from 0 to the
active reservoir capacity K. Each S
kt
is a discrete storage
volume that represents the range of storage volumes in
interval k at the beginning of each period t.
Let the following letter l be the index denoting differ-
ent final storage volume intervals. Each S
l,t1
is a discrete
volume that represents the storage volume interval l at the
end of each period t or equivalently at the beginning
of period t 1. As previously defined, let the indices i
and j denote the different flow intervals, and each discrete
flow q
it
and q
j,t1
represent those flow intervals i and j in
periods t and t 1 respectively.
These subscripts and the volume or flow intervals they
represent are illustrated in Figure 8.11.
With this notation, it is now possible to develop
a stochastic dynamic programming model that will
identify the allocations and releases that are to be made
given both the initial storage volume, S
kt
, and the
flow, q
it
. It follows the same structure as the determin-
istic models defined by Equations 8.30 through 8.32,
and 8.33.
To identify the optimal allocations in each period t for
each pair of feasible initial and final storage volumes S
kt
and S
l,t1
, and inflows q
it,
one can solve Equations 8.34
through 8.36.
minimize (UT
t
u
kit
)
2
(DT
t
d
kilt
)
2
(8.34)
where
u
kit
q
it
k, i, t. (8.35)
d
kilt
S
kt
q
it
u
kit
S
l,t1
feasible k, i, l, t.
(8.36)
The solution to these equations for each feasible combi-
nation of intervals k, i, l, and period t defines the optimal
allocations that can be expressed as u
t
(k, i) and d
t
(k, i, l).
The stochastic version of Model 8.33, again expressed
in a form suitable for backward-moving discrete dynamic
programming, can be written for different discrete values
of S
kt
from 0 to K and for all q
it
as:
The minimization is over all feasible discrete values of
S
l,t1
S
l,t1
K
S
l,t1
S
kt
q
it
(8.37)
F S q UT u k t DT d k i l
P F
t
n
kt it t t t t
ij
t
( , ) min ( ( , )) ( ( , , ))

2 2
tt
n
t j t
j
S q

1
1
1 1 1
( , )
, ,

,

wrm_ch08.qxd 8/31/2005 11:55 AM Page 241

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Each P
ij
t
in the above recursive equation is the known
conditional or transition probability of a flow q
j,t1
within
interval j in period t 1 given a flow of q
it
within interval
i in period t.
P
ij
t
Pr{flow q
j,t1
within interval j in t 1 | flow of q
it
within interval i in t}
The sum over all flow intervals j of these conditional prob-
abilities times the F
t1
n1
(S
l,t1
, q
j,t1
) values is the expected
minimum sum of future squared deviations from alloca-
tion targets with n 1 periods remaining given an initial
storage volume of S
kt
and flow of q
it
and final storage
volume of S
l,t1
. The value F
t
n
(S
kt
, q
it
) is the expected
minimum sum of squared deviations from the allocation
targets with n periods remaining given an initial storage
volume of S
kt
and flow of q
it.
Stochastic models such as
these provide expected values of objective functions.
Another way to write the recursion equations of this
model, Equation 8.37, is by using just the indices k and
l to denote the discrete storage volume variables S
kt
and S
l,t1
and indices i and j to denote the discrete flow
variables q
it
and q
j,t1
:
242 Water Resources Systems Planning and Management
such that S
l,t1
K
S
l,t1
S
kt
q
it
(8.38)
The steady-state solution of this dynamic programming
model will identify the preferred final storage volume
S
l,t1
in period t given the particular discrete initial stor-
age volume S
kt
and flow q
it
. This optimal policy can be
expressed as a function that identifies the best interval l
given intervals k, i and period t.
l (k, i, t) (8.39)
All values of l given k, i and t, defined by Equation 8.39,
can be expressed in a matrix, one for each period t.
Knowing the best final storage volume interval l given
an initial storage volume interval k and flow interval i, the
F k i UT u k t DT d k i l
P F
t
n
l
t t t t
ij
t
t
( , ) min ( ( , )) ( ( , , ))

2 2
1
nn
j
l j
1
( , )

k
k
k
k
k
= 5
= 4
= 3
= 2
=
storage volume interval S
S
S
S
S
5
4
3
2
t
t
t
t
t 1
E
0
2
0
1
1
0
p
f
l
o
w

p
r
o
b
a
b
i
l
i
t
y
streamflow Q
1 2 3 4 5 6 7 8 9 t t t t t t t t t
flow interval
2 3 4 5 6 7 8 9
q q q q q q q q q
i = 1
P
r
{
Q
i
t
=
P
Q
i
t
}
P
r
{
Q
i
t
=
P
Q
i
t
}
Figure 8.11. Discretization of
streamflows and reservoir
storage volumes. The area
within each flow interval i below
the probability density
distribution curve is the
unconditional probability, PQ
it
,
associated with the discrete
flow q
it
.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 242
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Uncertainty 243
optimal downstream allocation, d
t
(k, i), can, like the
upstream allocation, be expressed in terms of only k and
i in each period t. Thus, knowing the initial storage
volume S
kt
and flow q
it
is sufficient to define the optimal
allocations u
t
(k, i) and d
t
(k, i), final storage volume S
l,t1
,
and hence the release R
t
(k, i).
S
kt
q
it
u
t
(k, i) R
t
(k, i) S
l,t1
k, i, t
where l (k, i, t) (8.40)
6.1. Probabilities of Decisions
Knowing the function l (k, i, t) permits a calculation
of the probabilities of the different discrete storage
volumes, allocations, and flows. Let
PS
kt
the unknown probability of an initial storage
volume S
kt
being within some interval k in period t;
PQ
it
the steady-state unconditional probability of flow
q
it
within interval i in period t; and
P
kit
the unknown probability of the upstream and
downstream allocations u
t
(k, i) and d
t
(k, i) and reservoir
release R
t
(k, i) in period t.
As previously defined,
P
ij
t
the known conditional or transition probability of a
flow within interval j in period t 1 given a flow within
interval i in period t.
These transition probabilities P
ij
t
can be displayed in
matrices, similar to Table 8.3, but as a separate matrix
(Markov chain) for each period t.
The joint probabilities of an initial storage interval k,
an inflow in the interval i, P
kit
in each period t must sat-
isfy two conditions. Just as the initial storage volume in
period t 1 is the same as the final storage volume in
period t, the probabilities of these same respective
discrete storage volumes must also be equal. Thus,
(8.41)
where the sums in the right hand side of Equation 8.41
are over only those combinations of k and i that result in
a final volume interval l. This relationship is defined by
Equation 8.39 (l (k, i, t)).
While Equation 8.41 must apply, it is not sufficient.
The joint probability of a final storage volume in interval
P P l t
j t
j
kit
i k
1 1 , ,
,

l in period t and an inflow j in period t 1 must equal the
joint probability of an initial storage volume in the same
interval l and an inflow in the same interval j in period
t 1. Multiplying the joint probability P
kit
times the con-
ditional probability P
ij
t
and then summing over all k and i
that results in a final storage interval l defines the former,
and the joint probability P
l,j,t1
defines the latter.
P
l,j,t1

P
kit
P
ij
t
l, j, t l (k, i, t) (8.42)
Once again the sums in Equation 8.42 are over all combi-
nations of k and i that result in the designated storage
volume interval l as defined by the policy (k, i, t).
Finally, the sum of all joint probabilities P
kit
in each
period t must equal 1.
P
kit
1 t (8.43)
Note the similarity of Equations 8.42 and 8.43 to the
Markov steady-state flow Equations 8.22 and 8.23. Instead
of only one flow interval index considered in Equations
8.22 and 8.23, Equations 8.42 and 8.43 include two
indices, one for storage volume intervals and the other for
flow intervals. In both cases, one of Equations 8.22 and
8.42 can be omitted in each period t since it is redundant
with that periods Equations 8.23 and 8.43 respectively.
The unconditional probabilities PS
kt
and PQ
it
can be
derived from the joint probabilities P
kit
.
PS
kt
P
kit
k, t (8.44)
PQ
it
P
kit
i, t (8.45)
Each of these unconditional joint or marginal probabili-
ties, when summed over all their volume and flow
indices, will equal 1. For example,
PS
kt
PQ
it
1 (8.46)
Note that these probabilities are determined only on the
basis of the relationships among flow and storage intervals
as defined by Equation 8.39, l (k, i, t) in each period
t, and the Markov chains defining the flow interval tran-
sition or conditional probabilities, P
ij
t
. It is not necessary to
know the actual discrete storage values representing those
intervals. Thus assuming any relationship among the stor-
age volume and flow interval indices, l (k, i, t) and a
i

k i

k i

wrm_ch08.qxd 8/31/2005 11:55 AM Page 243

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
knowledge of the flow interval transition probabilities P
ij
t
,
one can determine the joint probabilities P
kit
and their
marginal or unconditional probabilities PS
kt
. One does
not need to know what those storage intervals are to cal-
culate their probabilities.
Given the values of these joint probabilities P
kit
, the
deterministic model defined by Equations 8.24 to 8.28
can be converted to a stochastic model to identify the best
storage and allocation decision-variable values associated
with each storage interval k and flow interval i in each
period t.
(8.47)
The constraints include:
a) Continuity of storage involving initial storage volumes
S
kt
, net inflows q
it
u
kit
, and at least partial releases d
kit
.
Again assuming no losses:
S
kt
q
it
u
kit
d
kit
S
l,t1
k, i, t
l (k, i, t) (8.48)
b) Reservoir capacity limitations.
S
kit
K k, i, t (8.49)
c) Allocation restrictions.
u
kit
q
it
k, i, t (8.50)
More detail on these and other stochastic modelling
approaches can be found in Faber and Stedinger (2001);
Gablinger and Loucks (1970); Huang et al. (1991); Kim
and Palmer (1997); Loucks and Falkson (1970);
Stedinger et al. (1984); Su and Deininger (1974); Tejada-
Guibert et al. (1993 1995); and Yakowitz (1982).
6.2. A Numerical Example
A simple numerical example may help to illustrate how
these stochastic models can be developed without getting
buried in detail. Consider two within-year periods each
year. The random flows Q
t
in each period t are divided
into two intervals. These flow intervals are represented by
discrete flows of 1 and 3 volume units per second in the
first period, and 3 and 6 volume units per second in the
second period. Their transition probabilities are shown in
Table 8.5.
Minimize P UT u DT d
kit
T
t kit
i k
t kit

{( ) ( ) }
2 2
244 Water Resources Systems Planning and Management
Assuming equal within-year period durations, these
three discrete flow rates are equivalent to about 16, 47
and 95 million volume units per period.
Assume the active storage volume capacity K in the
reservoir equals 50 million volume units. This capacity can
be divided into different intervals of storage. For this sim-
ple example, assume three storage volume intervals repre-
sented by 10, 25 and 40 million volume units. Assume the
allocation targets remain the same in each period at both
the upstream and downstream sites. The upstream alloca-
tion target is approximately 2 volume units per second or
30 million volume units in each period. The downstream
allocation target is approximately 5 volume units per
second or 80 million volume units in each period.
With these data we can use Equations 8.34 8.36 to
determine the allocations that minimize the sum of
squared deviations from targets and what that sum is, for
all feasible combinations of initial and final storage vol-
umes and flows. Table 8.6 shows the results of these opti-
mizations. These results will be used in the dynamic
programming model to determine the best final storage
volumes given initial volumes and flows.
With the information in Tables 8.5 and 8.6, the
dynamic programming model, Equation 8.38 or as
expressed in Equation 8.51, can be solved to find the
optimal final storage volumes, given an initial storage
volume and flow. The iterations of the recursive equation,
sufficient to reach a steady state, are shown in Table 8.7.
such that S
l,t1
K
S
l,t1
S
kt
Q
it
(8.51)
F k i SD P F l j
t
n
l
kil ij
t
t
n
j
( , ) min ( , )

1
1

flow in = 2 t flow in = t
f
l
o
w

i
n
=

1
t
f
l
o
w

i
n
=

2
t
Q
i
Q
i
Q
j
Q
j
0.6 0.7 0.4 0.3
0.3 0.2 0.7 0.8
3 1 6 3
1 3
3 6
E
0
2
0
8
2
9
a
Table 8.5. Transition probabilities for two ranges of flows in
two within-year periods.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 244
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Uncertainty 245
initial
storage
flow final
storage
interval
indices
upstream
allocation
down-
stream
allocation
sum
squared
deviation
SD d
kil kil
u S
ki 1
k i l , , Q
i
S
k
10
10
10
10
10
10
10
10
25
25
25
25
25
25
25
25
25
4996.0
7141.0
1989.0
3204.0
4869.0
112.5
450.0
1012.5
3301.0
4996.0
7141.0
1152.0
1989.0
3204.0
0.0
112.5
450.0
16.0
1.0
47.0
32.0
17.0
72.5
65.0
75.5
31.0
16.0
1.0
56.0
47.0
32.0
80.0
72.5
65.0
0.0
0.0
0.0
0.0
0.0
22.5
15.0
7.5
0.0
0.0
0.0
6.0
0.0
0.0
30.0
22.5
15.0
1, 1, 1
1, 1, 2
1, 2, 1
1, 2, 2
1, 2, 3
1, 3, 1
1, 3, 2
1, 3, 3
2, 1, 1
2, 1, 2
2, 1, 3
2, 2, 1
2, 2, 2
2, 2, 3
2, 3, 1
2, 3, 2
2, 3, 3
10
25
10
25
40
10
25
40
10
25
40
10
25
40
10
25
40
16
16
47
47
47
95
95
95
16
16
16
47
47
47
95
95
95
40
40
40
40
40
40
40
40
40
2056.0
3301.0
4996.0
544.5
1152.0
1989.0
0.0
0.0
112.5
46.0
31.0
16.0
63.5
56.0
47.0
80.0
80.0
72.5
0.0
0.0
0.0
13.5
6.0
0.0
30.0
30.0
22.5
3, 1, 1
3, 1, 2
3, 1, 3
3, 2, 1
3, 2, 2
3, 2, 3
3, 3, 1
3, 3, 2
3, 3, 3
10
25
40
10
25
40
10
25
40
16
16
16
47
47
47
95
95
95
E
0
2
0
8
2
9
b
Table 8.6. Optimal allocations
associated with given initial storage,
S
k
, flow, Q
i
, and final storage, S
l
,
volumes. These allocations u
ki
and
d
kil
minimize the sum of squared
deviations, DS
kil
(30 u
ki
)
2
(80 d
kil
)
2
, from upstream and
downstream targets, 30 and 80
respectively, subject to u
ki
flow Q
i
,
and d
kil
release (S
k
Q
i
u
ki
S
l
).
the data in Tables 8.5 and 8.8. It is obvious that if the
policy from Table 8.9 is followed, the steady-state proba-
bilities of being in storage Interval 1 in Period 1 and in
Interval 3 in Period 2 are 0.
Multiplying these joint probabilities by the correspon-
ding SD
kit
values in the last column of Table 8.6 provides
the annual expected squared deviations, associated with
the selected discrete storage volumes and flows. This is
done in Table 8.11 for those combinations of k, i, and l
that are contained in the optimal solution as listed in
Table 8.9.
The sum of products of the last two columns in
Table 8.11 for each period t equals the expected squared
deviations in the period. For period t 1, the expected
This process can continue until a steady-state policy
is defined. Table 8.8 summarizes the next five iterations.
At this stage, the annual differences in the objective
values associated with a particular state and season
have come close to a common constant value.
While the differences between corresponding F
t
nT
and
F
t
n
have not yet reached a common constant value to the
nearest unit deviation (they range from, 3475.5 to 3497.1
for an average of 3485.7), the policy has converged to that
shown in Tables 8.8 and 8.9.
Given this operating policy, the probabilities of being
in any of these volume and flow intervals can be
determined by solving Equations 8.42 through 8.45.
Table 8.10 shows the results of these equations applied to
wrm_ch08.qxd 8/31/2005 11:55 AM Page 245
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
246 Water Resources Systems Planning and Management
E
0
2
0
8
2
9
d
storage
& flow optimal
1,1 6234.4 4996.0 + 0.6 (1989.0) +0.4 (112.5)
7141.0 + 0.6 (1152.0) +0.4 ( 0.0)
infeasible ----
l
l
l
= 1
= 2
= 3
SDkil + j Pij
t
Ft+1
n-1
(l, j) Ft
n
(k, i)
k, i
=
= 7832.2
= ---
6234.4
l
1
1,2
2,1
2,2
3,1
3,2
2664.5
4539.4
1827.5
3294.4
1220.0
1989.0 + 0.3 (1989.0) +0.7 (112.5)
3204.0 + 0.3 (1152.0) +0.7 ( 0.0)
4869.0 + 0.3 ( 544.5) +0.7 ( 0.0)
3301.0 + 0.6 (1989.0) +0.4 (112.5)
4996.0 + 0.6 (1152.0) +0.4 ( 0.0)
7141.0 + 0.6 ( 544.5) +0.4 ( 0.0)
1152.0 + 0.3 (1989.0) +0.7 (112.5)
1989.0 + 0.3 (1152.0) +0.7 ( 0.0)
3204.0 + 0.3 ( 544.5) +0.7 ( 0.0)
2056.0 + 0.6 (1989.0) +0.4 (112.5)
3301.0 + 0.6 (1152.0) +0.4 ( 0.0)
4996.0 + 0.6 ( 544.5) +0.4 ( 0.0)
544.5 + 0.3 (1989.0) +0.7 (112.5)
1152.0 + 0.3 (1152.0) +0.7 ( 0.0)
1989.0 + 0.3 ( 544.5) +0.7 ( 0.0)
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
=
= 3549.6
= 5032.35
2664.45
=
= 5687.2
= 7467.7
4539.4
=
= 2334.6
= 3367.35
1827.45
=
= 3992.2
= 5322.7
3294.4
=
= 1497.6
= 2152.35
1219.95
1
1
1
1
1
period = 1, = 2 t n
E
0
2
0
8
2
9
c
storage
& flow
period = 2, = 1 t n
optimal
1,2 1989.0 1989.0 + 0
3204.0 + 0
4869.0 + 0
l
l
l
= 1
= 2
= 3
SDkil + j Pij
t
Ft+1
n-1
(l, j) Ft
n
(k, i)
k, i l
1
1,3
2,2
2,3
3,2
3,3
112.5
1152.0
0.0
544.5
0.0
112.5 + 0
450.0 + 0
1012.0 + 0
1152.0 + 0
1989.0 + 0
3204.0 + 0
0.0 + 0
112.5 + 0
450.0 + 0
544.5 + 0
1152.0 + 0
1989.0 + 0
0.0 + 0
0.0 + 0
112.5 + 0
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
1
1
1
1
1,2
Table 8.7. First four iterations of dynamic
programming model, Equations 8.51, moving
backward in successive periods n, beginning in
season t 2 with n 1. The iterations stop when
the final storage policy given any initial storage
volume and flow repeats itself in two successive
years. Initially, with no more periods remaining,
F
1
0
(k, i) 0 for all k and i.
(contd.)
wrm_ch08.qxd 8/31/2005 11:55 AM Page 246
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Uncertainty 247
Table 8.7. Concluded.
E
0
2
0
8
2
9
e
storage
& flow optimal
1,2
6929.8
1989.0 + 0.7 (6234.4) +0.3 (2664.5)
3204.0 + 0.7 (4539.4) +0.3 (1827.5)
4869.0 + 0.7 (3294.4) +0.3 (1219.9)
l
l
l
= 1
= 2
= 3
SDkil + j Pij
t
Ft+1
n-1
(l, j) Ft
n
(k, i)
k, i
= 7152.4
=
= 7541.1
6929.8
l
2
1,3
2,2
2,3
3,2
3,3
112.5 + 0.2 (6234.4) +0.8 (2664.5)
450.0 + 0.2 (4539.4) +0.8 (1827.5)
1012.5 + 0.2 (3294.4) +0.8 (1219.9)
1152.0 + 0.7 (6234.4) +0.3 (2664.5)
1989.0 + 0.7 (4539.4) +0.3 (1827.5)
3204.0 + 0.7 (3294.4) +0.3 (1219.9)
0.0 + 0.2 (6234.4) +0.8 (2664.5)
112.5 + 0.2 (4539.4) +0.8 (1827.5)
450.0 + 0.2 (3294.4) +0.8 (1219.9)
544.5 + 0.7 (6234.4) +0.3 (2664.5)
1152.0 + 0.7 (4539.4) +0.3 (1827.5)
1989.0 + 0.7 (3294.4) +0.3 (1219.9)
0.0 + 0.2 (6234.4) +0.8 (2664.5)
0.0 + 0.2 (4539.4) +0.8 (1827.5)
112.5 + 0.2 (3294.4) +0.8 (1219.9)
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
= 3490.0
= 2819.8
= 2647.3
= 6315.4
=
= 5876.1
5714.8
= 3378.4
= 2482.3
= 2084.8
= 5707.9
= 4877.8
= 4661.1
= 3378.4
= 2369.8
= 1747.3
3
2
3
3
3
period = 2, = 3 t n
2647.3
5714.8
2084.8
4661.1
1747.3
E
0
2
0
8
2
9
f
storage
& flow optimal
1,1 10212.8 4996.0 + 0.6 (6929.8) +0.4 (2647.3)
7141.0 + 0.6 (5714.8) +0.4 (2084.8)
infeasible ---
l
l
l
= 1
= 2
= 3
SDkil + j Pij
t
Ft+1
n-1
(l, j) Ft
n
(k, i)
k, i
=
=
l
1
1,2
2,1
2,2
3,1
2,2
1989.0 + 0.3 (6929.8) +0.7 (2647.3)
3204.0 + 0.3 (5714.8) +0.7 (2084.8)
4869.0 + 0.3 (4661.1) +0.7 (1747.3)
3301.0 + 0.6 (6929.8) +0.4 (2647.3)
4996.0 + 0.6 (5714.8) +0.4 (2084.8)
7141.0 + 0.6 (4661.1) +0.4 (1747.3)
1152.0 + 0.3 (6929.8) +0.7 (2647.3)
1989.0 + 0.3 (5714.8) +0.7 (2084.8)
3204.0 + 0.3 (4661.1) +0.7 (1747.3)
2056.0 + 0.6 (6929.8) +0.4 (2647.3)
3301.0 + 0.6 (5714.8) +0.4 (2084.8)
4996.0 + 0.6 (4661.1) +0.4 (1747.3)
544.5 + 0.3 (6929.8) +0.7 (2647.3)
1152.0 + 0.3 (5714.8) +0.7 (2084.8)
1989.0 + 0.3 (4661.1) +0.7 (1747.3)
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
l
l
l
= 1
= 2
= 3
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
1
1
1
1
2
period = 1, = 4 t n
5921.1
8517.8
5084.1
7272.8
4325.8
10212.8
11403.8
5921.1
6377.8
7490.5
8517.8
9258.8
10636.6
5084.1
5162.8
5825.5
7272.8
7563.8
8491.6
4476.6
4610.5
4325.8
wrm_ch08.qxd 8/31/2005 11:55 AM Page 247
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
sum of squared deviations are 1893.3 and for t 2 they
are 1591.0. The total annual expected squared deviations
are 3484.3. This compares with the expected squared
deviations derived from the dynamic programming
model, after 9 iterations, ranging from 3475.5 to 3497.1
(as calculated from data in Table 8.8).
These upstream allocation policies can be displayed in
plots, as shown in Figure 8.12.
The policy for reservoir releases is a function not only of
the initial storage volumes, but also of the current inflow, in
other words, the total water available in the period.
Reservoir release rule curves such as shown in Figures 4.16
or 4.18 now must become two-dimensional. However, the
inflow for each period usually cannot be predicted with
certainty at the beginning of each period. In situations
where the release cannot be adjusted during the period as
the inflow becomes more predictable, the reservoir release
policy has to be expressed in a way that can be followed
without knowledge of the current inflow. One way to do
this is to compute the expected value of the release for each
discrete storage volume, and show it in a release rule. This
is done in Figure 8.13. The probability of each discrete
release associated with each discrete river flow is the proba-
bility of the flow itself. Thus, in Period 1 when the storage
volume is 40, the expected release is 46(0.41) 56(0.59)
52. These discrete expected releases can be used to define
a continuous range of releases for the continuous range of
storage volumes from 0 to full capacity, 50. Figure 8.13 also
248 Water Resources Systems Planning and Management
shows the hedging that might take place as the reservoir
storage volume decreases.
Another approach to defining the releases in each
period in a manner that is not dependent on knowledge
of the current inflow, even though the model used
assumes this, is to attempt to define either release targets
with constraints on final storage volumes, or final storage
targets with constraints on total releases. Obviously, such
policies will not guarantee constant releases throughout
each period. For example, consider the optimal policy
shown in Table 8.9. The releases (or final storage vol-
umes) in each period are dependent on the initial storage
and current inflow. However, this operating policy can be
expressed as:
If in period 1, the final storage target should be in
interval 1. Yet the total release cannot exceed the flow
in interval 2.
If in period 2 and the initial storage is in interval 1, the
release should be in interval 1.
If in period 2 and the initial storage is in interval 2, the
release should be in interval 2.
If in period 2 and the initial storage is in interval 3, the
release should equal the inflow.
This policy can be followed without any forecast of cur-
rent inflow. It will provide the releases and final storage
volumes that would be obtained with a perfect inflow
forecast at the beginning of each period.
E
0
2
0
8
2
9
g
storage
& flow F
t
n
(k, i)
k, i
1,1
1,2 10691.7 2
1,3 5927.7 3
= 2, = 5 t n
l*
13782.1 17279.2
F
t
n
(k, i) F
t
n
(k, i) F
t
n
(k, i) F
t
n
(k, i)
= 1, = 6 t n = 2, = 7 t n = 1, = 8 t n = 2, = 9 t n
l* l* l* l*
1 1
9345.9 14217.7 12821.4 17708.3
9381.5 12861.3
1 1 2 2
3 3
2,1
3,1
2,2
3,2
9476.7
8377.7
2
3
2,3
3,3
5365.2
5027.7
3
3
12087.1
10842.1
15584.2
14339.2
1 1
1 1
8508.9
7750.7
13002.7
11903.7
11984.4
11226.1
16493.2
15394.3
8819.0
8481.5
12298.7
11961.2
1 1
2 2
2 2
3 3
3
3
3
3
Table 8.8. Summary of objective
function values F
t
n
(k, i) and
optimal decisions for stages
n 5 to 9 periods remaining.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 248
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Uncertainty 249
period initial storage volume
and flow interval
final storage
volume interval
k t i l
1 1 1 1
1
1
1
1
1
1
2
2
3
3
2
1
2
1
2
1
1
1
1
2
2 1 2 2
2
2
2
2
2
1
2
2
3
3
3
2
3
2
3
3
2
3
3
3
E
0
2
0
8
2
9
h
Table 8.9. Optimal reservoir policy l (k, i, t) for the
example problem.
PQ (1, 1) = 0.4117647 PQ (2, 2) =0.4235294
PQ (2, 1) = 0.5882353 PQ (3, 2) = 0.5764706
PS (1, 1) = 0.0000000 PS (1, 2) = 0.5388235
PS (2, 1) = 0.4235294 PS (2, 2) = 0.4611765
PS (3, 1) = 0.5764706 PS (3, 2) = 0.0000000
P (1, 1, 1) = 0.0000000 P (1, 2, 2) = 0.2851765
P (1, 2, 1) = 0.0000000 P (1, 3, 2) = 0.2536471
P (2, 1, 1) = 0.2964706 P (2, 2, 2) = 0.1383529
P (2, 2, 1) = 0.1270588 P (2, 3, 2) = 0.3228235
P (3, 1, 1) = 0.1152941 P (3, 2, 2) = 0.0000000
P (3, 2, 1) = 0.4611765 P (3, 3, 2) = 0.0000000
unconditional probabilities
of flow intervals in the 2 time periods i t
PQ
it
unconditional probabilities
of storage intervals in the 2 time periods k t
PS
kt
joint probabilities P
intervals k and flow intervals in the 2 time periods t i
kit
of storage volume
E
0
2
0
8
2
9
j
Table 8.10. Probabilities of flow and storage volume
intervals associated with the policy as defined in
Table 8.9 for the example problem.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 249
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
250 Water Resources Systems Planning and Management
i
n
i
t
i
a
l
s
t
o
r
a
g
e
f
l
o
w
f
i
n
a
l
s
t
o
r
a
g
e
i
n
t
e
r
v
a
l
i
n
d
i
c
e
s
s
u
m

s
q
u
a
r
e
d
d
e
v
i
a
t
i
o
n
s
j
o
i
n
t
p
r
o
b
a
b
i
l
i
t
y
SD d
kit kit
u S
kit 1
k i l , , Q
i
S
k
E
0
2
0
8
2
9
k
10
10
25
25
40
40
10
10
25
25
40
40
16
47
16
47
16
47
47
95
47
95
47
95
10
10
10
10
10
25
25
25
25
40
40
40
1,1,1
1,2,1
2,1,1
2,2,1
3,1,1
3,2,2
1,2,2
1,3,2
2,2,2
2, 3,3
3,2,3
3,3,3
1
1
1
1
1
1
2
2
2
2
2
2
0.0
0.0
0.0
6.0
0.0
6.0
0.0
7.5
0.0
15.0
0.0
22.5
16.0
47.0
31.0
56.0
46.0
56.0
32.0
57.5
47.0
65.0
47.0
72.5
4996.0
1989.0
3301.0
1152.0
2056.0
1152.0
3204.0
1012.5
1989.0
450.0
1989.0
112.5
0.0
0.0
0.2964706
0.1270588
0.1152941
0.4611765
0.2851765
0.2536471
0.1383529
0.3228235
0.0
0.0
1.0
1.0
sum =
sum =
t
i
m
e

p
e
r
i
o
d
P
kit
o
p
t
i
m
a
l
a
l
l
o
c
a
t
i
o
n
Table 8.11. The optimal
operating policy and the
probability of each state and
decision.
river flow
river flow
100
100
25
20
15
10
5
0
25
20
15
10
5
0
u
p
s
t
r
e
a
m

u
s
e
r
a
l
l
o
c
a
t
i
o
n

i
n

p
e
r
i
o
d

2
u
p
s
t
r
e
a
m

u
s
e
r
a
l
l
o
c
a
t
i
o
n

i
n

p
e
r
i
o
d
1
90
90
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
0
0
22.5
15.0
7.5
s
t
o
r
a
g
e

v
o
l
u
m
e
i
n
d
e
p
e
n
d
e
n
t

o
f
s
t
o
r
a
g
e

v
o
l
u
m
e
E
0
2
0
1
1
0
q
Figure 8.12. Upstream user allocation policies. In Period 1
they are independent of the downstream initial storage
volumes. In Period 2 the operator would interpolate between
the three allocation functions given for the three discrete
initial reservoir storage volumes.
50
45
40
35
30
25
20
15
10
5
0
s
t
o
r
a
g
e

v
o
l
u
m
e
reservoir releases
period
E
0
2
0
1
1
0
r
10
20
30
42
44
46
48
50
52
54
period 2
4
53
55
57
59
61
62
16
28
40
63
64
49
55
40
0
60
65
52
0
Figure 8.13. Reservoir release rule showing an interpolated
release, increasing as storage volumes increase.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 250
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Modelling Uncertainty 251
Alternatively, in each period t one can solve the
model defined by Equation 8.37 to obtain the best deci-
sion for the current and a sequence of future periods,
taking into account all current information regarding
the objectives and possible inflow scenarios and their
probabilities. The actual release decision in the current
period can be the expected value of all these releases in
this current period. At the beginning of the next period,
the model is updated with respect to current initial
storage and inflow scenarios (as well as any changes in
objectives or other constraints) and solved again. This
process continues in real time. This approach is dis-
cussed in Tejada-Guibert et al. (1993) and is the cur-
rent approach for providing release advice to the Board
of Control that oversees the releases from Lake Ontario
that govern the water levels of the lake and the
St. Lawrence River.
These policies, and modifications of them, can be
simulated to determine improved release rules.
7. Conclusions
This chapter has introduced some approaches for includ-
ing risk in optimization and simulation models. The dis-
cussion began with ways to obtain values of random
variables whose probability distributions are known.
These values, for example streamflows or parameter val-
ues, can be inputs to simulation models. Monte Carlo
simulation involves the use of multiple simulations using
these random variable values to obtain the probability
distributions of outputs, including various system per-
formance indicators.
Two methods were reviewed for introducing
random variables along with their probabilities into opti-
mization models. One involves the use of chance con-
straints. These are constraints that must be met, as all
constraints must, but now with a certain probability. As in
any method there are limits to the use of chance con-
straints. These limitations were not discussed, but in cases
where chance constraints are applicable, and if their deter-
ministic equivalents can be defined, they are probably the
only method of introducing risk into otherwise determin-
istic models that do not add to the model size.
Alternatively, the range of random variable values
can be divided into discrete ranges. Each range can be
represented by a specific or discrete value of the
random variable. These discrete values and their proba-
bilities can become part of an optimization model. This
was demonstrated by means of transition probabilities
incorporated into both linear and dynamic programming
models.
The examples used in this chapter to illustrate the
development and application of stochastic optimization
and simulation models are relatively simple. These and
similar probabilistic and stochastic models have been
applied to numerous water resources planning and man-
agement problems. They can be a much more effective
screening tool than deterministic models based on the
mean or other selected values of random variables. But
sometimes they are not. Clearly if the system being
analysed is very complex, or just very big in terms of the
number of variables and constraints, the use of determin-
istic models for a preliminary screening of alternatives
prior to a more precise probabilistic screening is often
warranted.
8. References
FABER, B.A. and STEDINGER, J.R. 2001. Reservoir
optimization using sampling SDP with ensemble
streamflow prediction (ESP) forecasts. Journal of
Hydrology, Vol. 249, Nos. 14, pp. 11333.
GABLINGER, M. and LOUCKS, D.P. 1970. Markov
models for flow regulation. Journal of the Hydraulics
Division, ASCE, Vol. 96, No. HY1, pp. 16581.
HUANG, W.C.; HARBO, R. and BOGARDI, J.J. 1991.
Testing stochastic dynamic programming models
conditioned on observed or forecast inflows. Journal of Water
Resources Planning and Management, Vol. 117, No.1,
pp. 2836.
KIM, Y.O.; PALMER, R.N. 1997. Value of seasonal flow
forecasts in bayesian stochastic programming. Journal
of Water Resources Planning and Management, Vol. 123,
No. 6, pp. 32735.
LOUCKS, D.P. and FALKSON, L.M. 1970. Comparison
of some dynamic, linear, and policy iteration methods
for reservoir operation. Water Resources Bulletin, Vol. 6,
No. 3, pp. 384400.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 251
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
STEDINGER, J.R.; SULE, B.F. and LOUCKS, D.P. 1984.
Stochastic dynamic programming models for reservoir
operation optimization. Water Resources Research, Vol. 20,
No. 11, pp. 1499505.
SU, S.Y. and DEININGER, R.A. 1974. Modeling the
regulation of lake superior under uncertainty of future
water supplies. Water Resources Research, Vol. 10, No. 1,
pp. 1122.
TEJADA-GUIBERT, J.A.; JOHNSON, S.A. and STE-
DINGER, J.R. 1993. Comparison of two approaches for
implementing multi-reservoir operating policies derived
using dynamic programming. Water Resources Research,
Vol. 29, No. 12, pp. 396980.
TEJADA-GUIBERT, J.A.; JOHNSON, S.A. and STEDINGER,
J.R. 1995. The value of hydrologic information in stochastic
dynamic programming models of a multi-reservoir system.
Water Resources Research, Vol. 31, No. 10, pp. 25719.
YAKOWITZ, S. 1982. Dynamic programming
applications in water resources. Water Resources
Research, Vol. 18, No. 4, pp. 67396.
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.
BIRGE, J.R. and LOUVEAUX, F. 1997. Introduction to
stochastic programming. New York, Springer-Verlag.
DEGROOT, M.H. 1980. Optimal statistical decisions. New
York, McGraw-Hill.
DEMPSTER, M.A.H. (ed.). 1980. Stochastic programming.
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.
ERMOLIEV, Y. 1970. Methods of stochastic programming.
Moscow, Nauka. (In Russian.)
252 Water Resources Systems Planning and Management
ERMOLIEV, Y. and WETS, R., (eds.). 1988. Numerical
techniques for stochastic optimization. Berlin Springer-Verlag.
HEYMAN, D.P. and SOBEL, M.J. 1984. Stochastic models
in operations research. Vol. 2: Stochastic optimization. New
York, McGraw-Hill.
HILLIER, F.S. and LIEBERMAN, G.J. 1990. Introduction
to stochastic models in operations research. New York,
McGraw-Hill.
HOWARD, R.A. 1960. Dynamic programming and Markov
processes. Cambridge, Mass., MIT Press.
HUFSCHMIDT, M.M. and FIERING, M.B. 1966.
Simulation techniques for design of water-resource systems.
Cambridge, Mass., Harvard University Press.
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.
PRKOPA, A. 1995. Stochastic programming. Dordrecht,
Kluwer Academic.
RAIFFA, H. and SCHLAIFER, R. 1961. Applied statistical
decision theory. Cambridge, Mass., Harvard University Press.
REVELLE, C. 1999. Optimizing reservoir resources. New
York, Wiley.
ROSS, S.M. 1983. Introduction to stochastic dynamic
programming. New York, Academic Press.
SOMLYDY, L. and WETS, R.J.B. 1988. Stochastic
optimization models for lake eutrophication management.
Operations Research, Vol. 36, No. 5, pp. 66081.
WETS, R.J.B. 1990 Stochastic programming. In: G.L.
Nemhauser, A.H.G. Rinnoy Kan and M.J. Todd (eds.),
Optimization: handbooks in operations research and management
science, Vol. 1, pp. 573629. Amsterdam, North-Holland.
WURBS, R.A. 1996. Modeling and analysis of reservoir
system operations. Upper Saddle River, N.J., Prentice Hall.
wrm_ch08.qxd 8/31/2005 11:55 AM Page 252
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
9. Model Sensitivity and Uncertainty Analysis
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
wrm_ch09.qxd 8/31/2005 11:56 AM Page 254
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Models are the primary way we have to estimate the multi-
ple effects of alternative water resources system design and
operating policies. Models predict the values of various
system performance indicators. Their outputs are based on
model structure, hydrological and other time-series inputs,
and a host of parameters whose values describe the system
being simulated. Even if these assumptions and input data
reflect, or are at least representative of, conditions believed
to be true, we know they will be inaccurate. Our models
are always simplifications of the real systems we study.
Furthermore, we simply cannot forecast the future with
precision, so we know the model outputs of future condi-
tions are at best uncertain.
Some prediction uncertainties can be reduced by addi-
tional research and data collection and analysis. Before
undertaking expensive studies to gather and analyse
additional data, it is reasonable to ask what improvement in
estimates of system performance, or what reduction in the
uncertainty associated with those estimates, would result if
all data and model uncertainties could be reduced. Such
information helps determine how much one would be
willing to pay in order to reduce prediction uncertainty. If
prediction uncertainty is costly, it may pay to invest in addi-
tional data collection, more studies or better models all
255
Model Sensitivity
and Uncertainty Analysis
The usefulness of any model depends in part on the accuracy and reliability of its
output. Yet, because all models are imperfect abstractions of reality, and because
precise input data are rarely if ever available, all output values are subject to
imprecision. Input data errors and modelling uncertainties are not independent of
each other they can interact in various ways. The end result is imprecision and
uncertainty associated with model output. This chapter focuses on ways of
identifying, quantifying, and communicating the uncertainties in model outputs.
9
aimed at reducing that prediction uncertainty. If that uncer-
tainty has no, or only a very modest, impact on the likely
decision that is to be made, one should find other issues to
worry about.
If it appears that reducing prediction uncertainty is
worthwhile, then one should consider how best to do it.
If it involves obtaining additional information, then it is
clear that the value of this additional information, how-
ever measured, should exceed the cost of obtaining it. The
value of such information will be the increase in system
performance, or the reduction in its variance, that one can
expect from obtaining such information. If additional
information is to be obtained, it should be information
that reduces the uncertainties considered important, not
the unimportant ones.
This chapter reviews some methods for identifying
and communicating model prediction uncertainty. The
discussion begins with a review of the causes of risk
and uncertainty in model output. It then examines
ways of measuring or quantifying uncertainty and
model output sensitivity to model input imprecision,
concentrating on methods that seem most relevant or
practical for large-scale regional simulation modelling. It
builds on some of the statistical methods reviewed in
Chapter 7 and the modelling of risk and uncertainty in
Chapter 8.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 255
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
2. Issues, Concerns and
Terminology
Outcomes or events that cannot be predicted with
certainty are often called risky or uncertain. Some indi-
viduals draw a special and interesting distinction between
risk and uncertainty. In particular, the term risk is often
reserved to describe situations for which probabilities are
available to describe the likelihood of various events or
outcomes. If probabilities of various events or outcomes
cannot be quantified, or if the events themselves are
unpredictable, some would say the problem is then one of
uncertainty, and not of risk. In this chapter, what is not
certain is considered uncertain, and uncertainty is often
described by a probability distribution. When the ranges
of possible events are known and their probabilities are
measurable, risk is called objective risk. If the probabilities
are based solely on human judgement, the risk is called
subjective risk.
Such distinctions between objective and subjective
risk, and between risk and uncertainty, rarely serve any
useful purpose to those developing and using models.
Likewise, the distinctions are often unimportant to those
who should be aware of the risks or uncertainties associ-
ated with system performance indicator values.
Uncertainty in information is inherent in future-
oriented planning efforts. It stems from inadequate infor-
mation and incorrect assumptions, as well as from the
256 Water Resources Systems Planning and Management
variability of natural processes. Water managers often
need to identify both the uncertainty and the sensitivity
of, or changes in, system performance indicator values
due to any changes in possible input data and parameter
values from what were predicted. They need to reduce
this level of uncertainty to the extent practicable. Finally,
they need to communicate the residual uncertainties
clearly so that decisions can be made with this knowledge
and understanding.
Sensitivity analysis can be distinguished from uncer-
tainty analysis. Sensitivity analysis procedures explore
and quantify the impact of possible errors in input data
on predicted model outputs and system performance
indices. Simple sensitivity analysis procedures can be
used to illustrate either graphically or numerically the
consequences of alternative assumptions about the future.
Uncertainty analyses employing probabilistic descriptions
of model inputs can be used to derive probability distri-
butions of model outputs and system performance
indices. Figure 9.1 illustrates the impact of both input
data sensitivity and input data uncertainty on model
output uncertainty.
It is worthwhile to explore the transformation of
uncertainties in model inputs and parameters into uncer-
tainty in model outputs when conditions differ from those
reflected by the model inputs. Historical records of sys-
tem characteristics are typically used as a basis for model
inputs. Yet conditions in the future may change. There
Figure 9.1. Schematic diagram
showing relationship among
model input parameter
uncertainty and sensitivity to
model output variable
uncertainty (Lal, 1995).
input parameter
output variable
higher
lower
output
uncertainty
input
uncertainty
higher
sensitivity
lower
sensitivity
E
0
2
0
7
2
5
a
wrm_ch09.qxd 8/31/2005 11:56 AM Page 256
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 257
may be changes in the frequency and amounts of precipi-
tation, changes in land cover and topography, and
changes in the design and operation of control structures,
all resulting in changes of water stages and flows, and
their qualities, and consequently changes in the affected
ecosystems.
If asked how the system would operate with inputs
similar to those in the historical database, the model
should be able to interpolate within the available knowl-
edge base to provide a fairly precise estimate. Still, that
estimate will not be perfect. This is because our ability to
reproduce current and recent operations is not perfect,
though it should be fairly good. If asked to predict system
performance for situations very different from those in the
historical knowledge base, or when the historical data are
not considered representative of what might happen in
the future (due to climate change for example), such pre-
dictions become much less precise.
There are two reasons for this. First, our description of
the characteristics of those different situations or conditions
may be imprecise. Second, our knowledge base may not be
sufficient to calibrate model parameters in ways that would
enable us to reliably predict how the system will operate
under conditions unlike those that have been experienced
historically. The more conditions of interest differ from
those in the historical knowledge base, the less confidence
we have that the model is providing a reliable description
of systems operation. Figure 9.2 illustrates this issue.
Clearly, an uncertainty analysis needs to consider how
well a model can replicate current operations, and how
similar the target conditions or scenarios are to those
described in the historical record. The greater the
required extrapolation from what has been observed, the
greater will be the importance of parameter and model
uncertainties.
The relative and absolute importance of different
parameters will depend on the system performance indi-
cators of interest. Seepage rates may have a very large
local impact, but a small global effect. Changes in system-
wide evapotranspiration rates are likely to affect system-
wide flows. The precision of model projections and the
relative importance of errors in different parameters will
depend upon:
the precision with which the model can reproduce
observed conditions
the difference between the predicted conditions and
the historical experience included in the knowledge
base
system performance characteristics of interest.
Errors and approximations in input data measurement,
parameter values, model structure and model solution
algorithms are all sources of uncertainty. While there are
reasonable ways of quantifying and reducing these errors
and the resulting range of uncertainty of various system
performance indicator values, they are impossible to
eliminate. Decisions will still have to be made in the face
of a risky and uncertain future, and can be modified as
new data and knowledge are obtained in a process of
adaptive management.
There is also uncertainty with respect to human
behaviour and reactions related to particular outcomes and
their likelihoods, that is, to their risks and uncertainties. As
important as risks and uncertainties associated with human
reactions are to particular outcomes, they are not usually
part of the models themselves. Social uncertainty may often
be the most significant component of the total uncertainty
associated with just how a water resources system will
perform. For this reason we should seek designs and
operating policies that are flexible and adaptable.
When uncertainties associated with system operation
under a new operating regime are large, one should
anticipate the need to make changes and improvements as
experience is gained and new information accumulates.
When predictions are highly unreliable, responsible
Figure 9.2. The precision of model predictions is affected by
the difference between the conditions or scenarios of interest
and the conditions or scenarios for which the model was
calibrated.
m
o
d
e
l

o
u
t
p
u
t
y
conditions for
which prediction needed
E
0
2
0
1
1
0
y
model error
with calibration data
prediction error
for future conditions
calibration conditions
wrm_ch09.qxd 8/31/2005 11:56 AM Page 257
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
managers should favour actions that are robust (good
under a wide range of situations), gain information
through research and experimentation, monitor results to
provide feedback for the next decision, update assess-
ments and modify policies in the light of new informa-
tion, and avoid irreversible actions and commitments.
3. Variability and Uncertainty in
Model Output
Differences between model output and observed values
can result from either natural variability, such as is
caused by unpredictable rainfall, evapotranspiration,
water consumption and the like, and/or by both known
and unknown errors in the input data, the model param-
eters or the model itself. The later is sometimes called
knowledge uncertainty, but it is not always due to a lack
of knowledge. Models are always simplifications of
reality and, hence, imprecision can result. Sometimes
imprecision occurs because of a lack of knowledge, such
as just how a particular species will react to various envi-
ronmental and other habitat conditions. At other times,
known errors are introduced simply for practical reasons.
Imperfect representation of processes in a model con-
stitutes model structural uncertainty. Imperfect knowledge
of the values of parameters associated with these processes
constitutes model parameter uncertainty. Natural variability
includes both temporal variability and spatial variability, to
which model input values may be subject.
258 Water Resources Systems Planning and Management
Figure 9.3 illustrates these different types of uncer-
tainty. For example, the rainfall measured at a weather
station within a particular model grid cell may be used as
an input value for that cell, but the rainfall may actually
vary at different points within that cell and its mean value
will vary across the landscape. Knowledge uncertainty
can be reduced through further measurement and/or
research. Natural variability is a property of the natural
system, and is usually not reducible at the scale being
used. Decision uncertainty is simply an acknowledgement
that we cannot predict ahead of time just what decisions
individuals and organizations will make, or even just
what particular set of goals or objectives will be consid-
ered and the relative importance of each.
Rather than contrasting knowledge uncertainty
versus natural variability versus decision uncertainty, one
can classify uncertainty in another way based on specific
sources of uncertainty, such as those listed below, and
address ways of identifying and dealing with each source
of uncertainty.
Informational uncertainties
imprecision in specifying the boundary and
initial conditions that impact the output variable
values
imprecision in measuring observed output variable
values.
Model uncertainties
uncertain model structure and parameter values
variability of observed input and output values over
a region smaller than the spatial scale of the model
Figure 9.3. One way of
classifying types of uncertainty.
model
uncertainty
parameter
uncertainty
knowledge
uncertainty
temporal
variability
goals -
objectives
spatial
variability
values -
preferences
natural
variability
decision
uncertainty
uncertainty
E
0
2
0
7
2
5
b
wrm_ch09.qxd 8/31/2005 11:56 AM Page 258
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 259
variability of observed model input and output
values within a time smaller than the temporal scale
of the model. (e.g., rainfall and depths and flows
within a day)
errors in linking models of different spatial and
temporal scales.
Numerical errors
errors in the model solution algorithm.
3.1. Natural Variability
The main source of hydrological model output value
variability is the natural variability in hydrological and
meteorological input series. Periods of normal precipita-
tion and temperature can be interrupted by periods of
extended drought and intense meteorological events such
as hurricanes and tornadoes. There is no reason to believe
that such events will not continue to occur and become
even more frequent and extreme. Research has demon-
strated that climate has been variable in the past and there
are concerns about anthropogenic activities that may
increase that variability each year. Sensitivity analysis can
help assess the effect of errors in predictions if those pre-
dictions are based only on past records of historical time-
series data describing precipitation, temperature and
other exogenous forces across and on the border of the
regions being studied.
Time-series input data are often actual, or at least
based on, historical data. The time-series values typically
describe historical conditions including droughts and wet
periods. What is distinctive about natural uncertainty, as
opposed to errors and uncertainty due to modelling limi-
tations, is that natural variability in meteorological forces
cannot be reduced by improving the models structure,
increasing the resolution of the simulation or better
calibration of model parameters.
Errors result if meteorological values are not measured
or recorded accurately, or if mistakes are made in the
generation of computer data files. Furthermore, there is
no assurance that the statistical properties of historical
data will accurately represent those of future data. Actual
future precipitation and temperature scenarios will be
different from those in the past, and this difference in
many cases may have a larger effect than the uncertainty
due to incorrect parameter values. However, the effects of
uncertainties in the parameter values used in stochastic
generation models are often much more significant than
the effects of using different stochastic generation models
(Stedinger and Taylor, 1982).
While variability of model output is a direct result of
variability of model input (e.g. hydrological and meteoro-
logical data), the extent of the variability, and its lower
and upper limits, may also be affected by errors in the
inputs, the values of parameters, initial boundary condi-
tions, model structure, processes and solution algorithms.
Figure 9.4 illustrates the distinction between the
variability of a system performance indicator due to
input data variability, and the extended range of variability
due to the total uncertainty associated with any
combination of the causes listed in the previous section.
This extended range is what is of interest to water resources
planners and managers.
What can occur in practice is a time series of system
performance indicator values that can range anywhere
within or even outside the extended range, assuming the
confidence level of that extended range is less than 100%.
The confidence one can have that some future value of a
time series will be within a given range is dependent on two
factors. The first is the number of measurements used to
compute the confidence limits. The second is the assump-
tion that those measurements are representative (come
from the same statistical or stochastic process yielding)
future measurements, i.e. they come from the same statisti-
cal or stochastic process. Figure 9.5 illustrates this point.
s
y
s
t
e
m

p
e
r
f
o
r
m
a
n
c
e

i
n
d
i
c
a
t
o
r
time
a b
E
0
2
0
1
1
0
w
Figure 9.4. Time series of model output or system
performance showing variability over time. Range a results
from the natural variability of input data over time. The
extended range b results from the variability of natural input
data as well as from imprecision in input data measurement,
parameter value estimation, model structure and errors in
model solution algorithms. The extent of this range will
depend on the confidence level associated with that range.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 259
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Note that the time series may even contain values outside
the range b defined in Figure 9.4 if the confidence level of
that range is less than 100%. Confidence intervals associ-
ated with less than 100% certainty will not include every
possible value that might occur.
3.2. Knowledge Uncertainty
Referring to Figure 9.3, knowledge uncertainty includes
model structure and parameter value uncertainties. First
we consider parameter value uncertainty including
boundary condition uncertainty, and then model and
solution algorithm uncertainty.
3.2.1. Parameter Value Uncertainty
A possible source of uncertainty in model output results
from uncertain estimates of various model parameter
values. If the model calibration procedure were repeated
using different data sets, different parameter values would
result. Those values would yield different simulated sys-
tem behaviour and, thus, different predictions. We can call
this parameter uncertainty in the predictions because it is
caused by imprecise parameter values. If such parameter
value imprecision were eliminated, then the prediction
would always be the same and so the parameter value
uncertainty in the predictions would be zero. But this does
not mean that predictions would be perfectly accurate.
In addition to parameter value imprecision, uncertainty
in model output can result from imprecise specification
260 Water Resources Systems Planning and Management
of boundary conditions. These boundary conditions
may be either fixed or variable. However, because they
are not being computed on the basis of the state of the
system, their values can be uncertain. These uncertainties
can affect the model output, especially in the vicinity of the
boundary, in each time step of the simulation.
3.2.2. Model Structural and Computational Errors
Uncertainty in model output can also result from errors in
the model structure compared to the real system, and
approximations made by numerical methods employed in
the simulation. No matter how good our parameter value
estimates, our models are not perfect and there is a resid-
ual model error. Increasing model complexity in order to
more closely represent the complexity of the real system
may not only add to the cost of data collection, but may
also introduce even more parameters, and thus even more
potential sources of error in model output. It is not an
easy task to judge the appropriate level of model com-
plexity and to estimate the resulting levels of uncertainty
associated with various assumptions regarding model
structure and solution methods. Kuczera (1988) provides
an example of a conceptual hydrological modelling exer-
cise with daily time steps where model uncertainty dom-
inates parameter value uncertainty.
3.3. Decision Uncertainty
Uncertainty in model predictions can result from
unanticipated changes in what is being modelled. These
can include changes in nature, human goals, interests,
activities, demands and impacts. An example of this is the
deviation from standard or published operating policies
by operators of infrastructure such as canal gates, pumps
and reservoirs in the field, as compared to what is
specified in documents and incorporated into the water
systems models. Comparing field data with model data
for model calibration may yield incorrect calibrations if
operating policies actually implemented in the field differ
significantly from those built into the models. What do
operators do in times of stress?
What humans will want to achieve in the future may
not be the same as what they want today. Predictions of
peoples future desires are clearly sources of uncertainty.
s
y
s
t
e
m

p
e
r
f
o
r
m
a
n
c
e

i
n
d
i
c
a
t
o
r
time
b
E
0
2
0
1
1
0
x
Figure 9.5. Typical time series of model output or system
performance indicator values that are the result of input data
variability and possible imprecision in input data
measurement, parameter value estimation, model structure
and errors in model solution algorithms.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 260
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 261
A perfect example of this can be seen in the very flat
Greater Everglades region of south Florida in the United
States. Fifty years ago, folk wanted the swampy region
protected from floods and drained for agricultural and
urban development. Today many want just the opposite, at
least where there are no human settlements. They want to
return to a more natural hydrological system with more
wetlands and unobstructed flows, but now for ecological
restoration reasons, which were not a major concern
or much appreciated some half a century ago. Once the
mosquitoes return, and if the sea level continues to rise,
future populations who live there may want more flood
control and drainage again. Who knows? Complex and
changing social and economic processes influence human
activities and their demands for water resources and envi-
ronmental amenities over time. Some of these processes
reflect changes in local concerns, interests and activities,
but population migration and many economic activities
and social attitudes can also reflect changing national and
international trends.
Sensitivity scenarios that include human activities can
help define the effects of those activities within an area.
Careful attention should be given to the development of
these alternative scenarios so that they capture the real
forces or stresses that the system may face. The history
of systems studies are full of examples where the issues
studied were overwhelmed by much larger social forces
resulting from, for example, the relocation of major
economic activities, an oil embargo, changes in national
demand for natural resources, economic recession, an
act of terrorism or even war. One thing is certain: the
future will be different than the past, and no one knows
just how.
Surprises
Water resources managers may also want to consider how
vulnerable a system is to undesirable environmental
surprises. What havoc might an introduced species
like the zebra mussel invading the Great Lakes of
North America have in a particular watershed? Might some
introduced disease suddenly threaten key plant or animal
species? Might management plans have to be restructured
to address the survival of species such as salmon in the
Rhine River in Europe or in the Columbia River in North
America? Such uncertainties are hard to anticipate when
by their nature they will truly be surprises. But surprises
should be expected. Hence system flexibility and adapt-
ability should be sought to deal with changing manage-
ment demands, objectives and constraints.
4. Sensitivity and Uncertainty
Analyses
An uncertainty analysis is not the same as a sensitivity
analysis. An uncertainty analysis attempts to describe
the entire set of possible outcomes, together with their
associated probabilities of occurrence. A sensitivity analysis
attempts to determine the change in model output
values that results from modest changes in model input
values. A sensitivity analysis thus measures the change in
the model output in a localized region of the space of
inputs. However, one can often use the same set of model
runs for both uncertainty analyses and sensitivity analyses.
It is possible to carry out a sensitivity analysis of the model
around a current solution, and then use it as part of a first-
order uncertainty analysis.
This discussion begins by focusing on some methods
of uncertainty analysis, then reviews various ways of
performing and displaying sensitivity analyses.
4.1. Uncertainty Analyses
Recall that uncertainty involves the notion of random-
ness. If a value of a performance indicator or performance
measure, like the phosphorus concentration or the depth
of water at a particular location, varies, and this variation
over space and time cannot be predicted with certainty, it
is called a random variable. One cannot say with certainty
what the value of a random variable will be but only the
likelihood or probability that it will be within some spec-
ified range of values. The probabilities of observing par-
ticular ranges of values of a random variable are described
or defined by a probability distribution. There are many
types of distributions and each can be expressed in several
ways as presented in Chapter 7.
Suppose the random variable is denoted as X. As dis-
cussed in Chapter 7, if the observed values of this random
variable can only have discrete values, then the probability
wrm_ch09.qxd 8/31/2005 11:56 AM Page 261
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
distribution of X is easily described by a histogram, as
shown in Figure 9.6a. The sum of the probabilities for all
possible outcomes must equal 1. If the random variable is
a continuous variable that can assume any real value over a
range of values, the probability distribution of X can be
expressed as a continuous distribution, as shown in Figure
9.6b. The shaded area under the density function for the
continuous distribution, is 1. The area between two values
of the continuous random variable, such as between u and
v in Figure 9.6c, represents the probability that the
observed value x of the random variable value X will be
within that range of values.
The probability distributions shown in Figure 9.6b and
Figure 9.6c are called probability density functions (pdf )
and are denoted by f
X
(x). The subscript X on P
X
and f
X
represents the random variable, and the variable x is some
value of that random variable X.
Uncertainty analyses involve identifying characteristics
of various probability distributions of model input and
output variables, and subsequently functions of those
random output variables that are performance indicators
or measures. Often targets associated with these indica-
tors or measures are themselves uncertain.
A complete uncertainty analysis would involve a
comprehensive identification of all sources of uncer-
tainty that contribute to the joint probability distribu-
tions of each input or output variable. Assume such
analyses were performed for two alternative project
plans, A and B, and that the resulting probability density
distributions for a specified performance measure were
as shown in Figure 9.7. The figure also identifies the
costs of these two projects. The introduction of two per-
formance criteria cost and probability of exceeding a
performance measure target, e.g. a pollutant concentra-
tion standard introduces a conflict where a tradeoff
must be made.
262 Water Resources Systems Planning and Management
Figure 9.6. Probability
distributions for a discrete or
continuous random variable X.
The area under the distributions
(shaded areas in a and b) is 1,
and the shaded area in c is the
probability that the observed
value x of the random variable X
will be between u and v.
X
f ( ) X
X
u v
X X
f ( ) X
X
P ( ) X
X
a) c) b)
E
0
2
0
7
2
5
c
Figure 9.7. Tradeoffs involving cost and the probability that a
maximum desired target value will be exceeded. In this
illustration we want the lowest cost (B is best) and the lowest
probability of exceedance (A is best).
p
r
o
b
a
b
i
l
i
t
y
d
e
n
s
i
t
y

d
i
s
t
r
i
b
u
t
i
o
n
s
f
o
r

p
l
a
n
s
a
n
d
A
B
c
o
s
t
performance
measure value
probability maximum desired target is exceeded
plan A
plan B
E
0
2
0
7
2
5
d
maximum desired
target value
A
B
4.1.1. Model and Model Parameter Uncertainties
Consider a situation as shown in Figure 9.8, in which, for
a specific set of model inputs, the model outputs differ
from the observed values, and for those model inputs,
the observed values are always the same. Here, nothing
occurs randomly. The model parameter values or model
structure need to be changed. This is typically done in a
model calibration process.
Given specific inputs, the outputs of deterministic
models are always going to be the same each time those
inputs are simulated. If for specified inputs to any simu-
lation model the predicted output does not agree with
the observed value, as shown in Figure 9.8, this could
result from imprecision in the measurement of observed
data. It could also result from imprecision in the model
parameter values, the model structure or the algorithm
used to solve the model.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 262
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 263
Next, consider the same deterministic simulation
model, but now assume at least some of the inputs are
random (that is, not predictable), as may be case when
random outputs of one model are used as inputs into
another model. Random inputs will yield random outputs.
The model input and output values can be described by
probability distributions. If the uncertainty in the output is
due only to the uncertainty in the input, the situation is
similar to that shown in Figure 9.8. If the distribution of
performance measure output values does not fit or is not
identical to the distribution of observed performance meas-
ure values, then calibration of model parameter values or
modification of the model structure may be needed.
If a model calibration or identification exercise finds
the best values of the parameters to be outside reasonable
ranges of values based on scientific knowledge, then the
model structure or algorithm might be in error. Assuming
that the algorithms used to solve the models are correct
and that observed measurements of system performance
vary for the same model inputs, as shown in Figure 9.9, it
can be assumed that the model structure does not capture
all the processes that are taking place that affect the value
of the performance measures. This is often the case when
relatively simple and low-resolution models are used to
estimate the hydrological and ecological impacts of water
and land management policies. However, even large and
complex models can fail to include or adequately describe
important phenomena.
In the presence of informational uncertainties, there
may be considerable uncertainty about the values of the
Figure 9.8. A deterministic
system and a simulation model
of that system needing
calibration or modification in its
structure. There is no
randomness, only parameter
value or model structure errors
to be identified and corrected.
s
i
m
u
l
a
t
i
o
n

m
o
d
e
l
predicted performance
measure value
p
r
o
b
a
b
i
l
i
t
y
observed performance
measure value
p
r
o
b
a
b
i
l
i
t
y
model input model output
E
0
2
0
7
2
5
e
Figure 9.9. A deterministic
simulation model of a random or
stochastic system. To produce
the variability in the model
output that is observed in the real
system, even given the same input
values, the models parameter
values may need to vary over
distributions of values and/or the
model structure may need
modification along with additional
model inputs.
s
i
m
u
l
a
t
i
o
n

m
o
d
e
l
predicted performance
measure value
p
r
o
b
a
b
i
l
i
t
y
observed performance
measure value
p
r
o
b
a
b
i
l
i
t
y
model input model output
E
0
2
0
7
2
5
f
wrm_ch09.qxd 8/31/2005 11:56 AM Page 263
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
best parameters during calibration. This problem
becomes even more pronounced with increases in model
complexity.
An example:
Consider the prediction of a pollutant concentration at
some site downstream of a pollutant discharge site. Given
a streamflow Q (in units of 1000 m
3
/day), the distance
between the discharge site and the monitoring site, X(m),
the pollutant decay rate constant k (day
1
) and the pollu-
tant discharge W (Kg/day), we can use the following
simple model to predict the concentration of the pollutant
C (g/m
3
mg/l) at the downstream monitoring site:
C (W/Q) exp{k(X/U)}
In the above equation, assume the velocity U (m/day) is a
known function of the streamflow Q.
In this case, the observed value of the pollutant con-
centration C may differ from the computed value of C
even for the same inputs of W, Q, k, X and U.
Furthermore, this difference varies in different time peri-
ods. This apparent variability, as illustrated in Figure 9.9,
can be simulated using the same model but by assuming
a distribution of values for the decay rate constant k.
Alternatively, the model structure can be modified to
include the impact of streamflow temperature T on the
prediction of C.
C (W/Q) exp{k
T20
(X/U)}
Now there are two model parameters, the decay rate
constant k and the dimensionless temperature correction
factor , and an additional model input, the streamflow
temperature, T, in C. It could be that the variation in
streamflow temperature was the major cause of the first
equations uncertainty and that the assumed parameter
distribution of k was simply the result of the distribution
of streamflow temperatures on the term k
T20
.
If the output were still random given constant values
of all the inputs, then another source of uncertainty
exists. This uncertainty might be due to additional
random loadings of the pollutant, possibly from non-
point sources. Once again, the model could be modified
to include these additional loadings if they are knowable.
Assuming these additional loadings are not known, a
new random parameter could be added to the input
variable W or to the right hand side of the equations
above that would attempt to capture the impact on C of
264 Water Resources Systems Planning and Management
these additional loadings. A potential problem, however,
might be the likely correlation between those additional
loadings and the streamflow Q.
While adding model detail removed some uncertainty
in the above example, increasing model complexity will
not always eliminate or reduce uncertainty in model out-
put. Adding complexity is generally not a good idea when
the increased complexity is based on processes whose
parameters are difficult to measure, when the right equa-
tions are not known at the scale of application, or when
the amount of data for calibration is small compared to
the number of parameters.
Even if more detailed models requiring more input
data and more parameter values were to be developed, the
likelihood of capturing all the processes occurring in a
complex system is small. Hence, those involved will have
to make decisions while taking this uncertainty into
account. Imprecision will always exist due to a less than
complete understanding of the system and the hydro-
logical processes being modelled. A number of studies
have addressed model simplification, but only in some
simple cases have statisticians been able to identify just
how one might minimize modelling related errors in
model output values.
The problem of determining the optimal level of
modelling detail is particularly important when simulat-
ing the hydrological events at many sites over large areas.
Perhaps the best approach for these simulations is to
establish confidence levels for alternative sets of models
and then statistically compare simulation results. But even
this is not a trivial or cost-free task. Increases in the tem-
poral or spatial resolution typically require considerable
data collection and/or processing, model recalibrations,
and possibly the solution of stability problems resulting
from the numerical methods used in the models.
Obtaining and implementing alternative hydrological
simulation models will typically involve considerable
investment of money and time for data preparation and
model calibration.
What is needed is a way to predict the variability evi-
dent in the system shown in Figure 9.9. Instead of a fixed
output vector for each fixed input vector, a distribution of
outputs is needed for each performance measure based on
fixed inputs (Figure 9.9) or a distribution of inputs
(Figure 9.10.). Furthermore, the model output distribu-
tion for each performance measure should match as well
wrm_ch09.qxd 8/31/2005 11:56 AM Page 264
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 265
as possible the observed distribution of that performance
measure.
4.1.2. What Uncertainty Analysis Can Provide
An uncertainty analysis takes a set of randomly chosen
input values (which can include parameter values), passes
them through a model (or transfer function) to obtain the
distributions (or statistical measures of the distributions)
of the resulting outputs. As illustrated in Figure 9.11, the
output distributions can be used to
describe the range of potential outputs of the system at
some probability level
estimate the probability that the output will exceed a
specific threshold or performance measure target value.
Uncertainty analyses are often used to make general infer-
ences, such as the following:
estimating the mean and standard deviation of the
outputs
estimating the probability the performance measure
will exceed a specific threshold
assigning a reliability level to a function of the outputs,
for example, the range of function values that is likely
to occur with some probability
describing the likelihood of different potential outputs
of the system
estimating the relative impacts of input variable
uncertainties.
Implicit in any uncertainty analysis are the assumptions
that statistical distributions for the input values are correct
and that the model is a sufficiently realistic description of
the processes taking place in the system. Neither of these
assumptions is likely to be entirely correct.
4.2. Sensitivity Analyses
Sensitivity analysis aims to describe how much model
output values are affected by changes in model input
values. It is the investigation of the importance of impreci-
sion or uncertainty in model inputs in a decision-making
or modelling process. The exact character of a sensitivity
analysis depends upon the particular context and the
questions of concern. Sensitivity studies can provide a
Figure 9.10. Simulating variable
inputs to obtain probability
distributions of predicted
performance indices that match
the probability distributions of
observed performance values.
s
i
m
u
l
a
t
i
o
n

m
o
d
e
l
p
r
o
b
a
b
i
l
i
t
y
model input
E
0
2
0
7
2
5
g
p
r
o
b
a
b
i
l
i
t
y
model output
Figure 9.11. The distribution of
performance measures defines a range of
potential values and the likelihood that a
specified target value will be exceeded.
The shaded area under the density
function on the left represents the
probability that the target value will be
exceeded. This probability is shown in the
probability of exceedance plot on the right.
p
r
o
b
a
b
i
l
i
t
y

o
f
b
e
i
n
g

e
x
c
e
e
d
e
d
p
r
o
b
a
b
i
l
i
t
y
d
e
n
s
i
t
y

f
u
n
c
t
i
o
n
performance measure value performance measure value
target
value
target
value
range of values
1
0
E
0
2
0
7
2
5
h
wrm_ch09.qxd 8/31/2005 11:56 AM Page 265
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
general assessment of model precision when used to assess
system performance for alternative scenarios, as well as
detailed information addressing the relative significance of
errors in various parameters. As a result, sensitivity results
should be of interest to the general public, federal and
state management agencies, local water resources planners
and managers and model users and developers.
Clearly, upper level management and the public may be
interested in more general statements of model precision,
and should be provided such information along with model
predictions. On the other hand, detailed studies addressing
the significance and interactions among individual parame-
ters would likely be meaningful to model developers and
some model users. They can use such data to interpret
model results and to identify to where their efforts to
improve models and input values should be directed.
Initial sensitivity analysis studies could focus on two
products:
detailed results that guide research and assist model
development efforts
calculation of general descriptions of uncertainty associ-
ated with model predictions, so that policy decisions can
reflect both the modelling efforts best prediction of sys-
tem performance and the precision of such predictions.
In the first case, knowing the relative uncertainty in model
projections due to possible errors in different sets of
parameters and input data should assist in efforts to
improve the precision of model projections. This knowl-
edge should also contribute to a better understanding of
the relationships between model assumptions, parameters,
data and model predictions.
In the second case, knowing the relative precision
associated with model predictions should have a signifi-
cant effect on policy development. For example, the
analysis may show that, given data inadequacies, there
are very large error bands associated with some model
variables. When such large uncertainties exist, predic-
tions should be used with appropriate scepticism.
Incremental strategies should be explored along with
monitoring so that greater experience can accumulate to
resolve some of these uncertainties.
Sensitivity analysis features are available in many
linear and non-linear programming (optimization) pack-
ages. They identify the changes in the values of the
objective function and unknown decision-variables given
266 Water Resources Systems Planning and Management
a change in the model input values, and a change in
levels set for various constraints (Chapter 4). Thus, sensi-
tivity analysis can address the change in optimal system
performance associated with changes in various param-
eter values, and also how optimal decisions would
change with changes in resource constraint levels or tar-
get output requirements. This kind of sensitivity analysis
provides estimates of how much another unit of resource
would be worth, or what cost a proposed change in a
constraint places on the optimal solution. This informa-
tion is of value to those making design decisions.
Various techniques have been developed to determine
how sensitive model outputs are to changes in model
inputs. Most approaches examine the effects of changes in
a single parameter value or input variable assuming no
changes in all the other inputs. Sensitivity analyses can be
extended to examine the combined effects of multiple
sources of error, as well.
Changes in particular model input values can affect
model output values in different ways. It is generally true
that only a relatively few input variables dominate or
substantially influence the values of a particular output
variable or performance indicator at a particular location
and time. If the range of uncertainty of only some of the
output data is of interest, then undoubtedly only those
input data that significantly affect the values of those
output data need be included in the sensitivity analysis.
If input data estimates are based on repeated measure-
ments, a frequency distribution can be estimated that
characterizes natural variability. The shorter the record of
measurements, the greater will be the uncertainty regarding
the long-term statistical characteristics of that variability. If
obtaining a sufficient number of replicate measurements is
not possible, subjective estimates of input data ranges and
probability distributions are often made. Using a mixture of
subjective estimates and actual measurements does not
affect the application of various sensitivity analysis methods
that can use these sets or distributions of input values, but
it may affect the conclusions that can be drawn from the
results of these analyses.
It would be nice to have available accurate and easy-
to-use analytical methods for relating errors in input data
to errors in model outputs, and to errors in system
performance indicator values that are derived from
model output. Such analytical methods do not exist for
complex simulation models. However, methods based on
wrm_ch09.qxd 8/31/2005 11:56 AM Page 266
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 267
simplifying assumptions and approximations can be used
to yield useful sensitivity information. Some of these are
reviewed in the remainder of this chapter.
4.2.1. Sensitivity Coefficients
One measure of sensitivity is the sensitivity coefficient.
This is the derivative of a model output variable with
respect to an input variable or parameter. A number of
sensitivity analysis methods use these coefficients. First-
order and approximate first-order sensitivity analyses are
two such methods that will be discussed later. Analytical
methods are faced with considerable difficulties in:
obtaining the derivatives for many models
needing to assume mathematical (usually linear)
relationships when obtaining estimates of derivatives
by making small changes of input data values near
their nominal or most likely values
having large variances associated with most hydrological
process models.
These have motivated the replacement of analytical meth-
ods by numerical and statistical approaches to sensitivity
analysis.
Implicit in any sensitivity analysis are the assumptions
that statistical distributions for the input values are correct
and that the model is a sufficiently realistic description of
the processes taking place in the system. Neither of these
assumptions is likely to be entirely correct.
The importance of the assumption that the statistical
distributions for the input values are correct is easy to
check by using different distributions for the input
parameters. If the outputs vary significantly, then the
output is sensitive to the specification of the input distri-
butions, and hence they should be defined with care. A
relatively simple deterministic sensitivity analysis can be
of value here (Benaman, 2002). A sensitivity coefficient
can be used to measure the magnitude of change in an
output variable Q per unit change in the magnitude of an
input parameter value P from its base value P
0
. Let SI
PQ
be the sensitivity index for an output variable Q with
respect to a change P in the value of the input variable
P from its base value P
0
. Noting that the value of the out-
put Q(P) is a function of P, the sensitivity index could be
defined as
SI
PQ
[Q(P
0
P) Q(P
0
P)]/2P (9.1)
Other sensitivity indices could be defined (McCuen,
1973). Letting the index i represent a decrease and j rep-
resent an increase in the parameter value from its base
value P
0
, the sensitivity index SI
PQ
for parameter P and
output variable Q could be defined as
SI
PQ
{|(Q
0
Q
i
)/(P
0
P
i
)| |(Q
0
Q
j
)/
(P
0
P
j
)|}/2 (9.2)
or
SI
PQ
max{|(Q
0
Q
i
)/(P
0
P
i
)|, |(Q
0
Q
j
)/
(P
0
P
j
)|} (9.3)
A dimensionless expression of sensitivity is the elasticity
index, EI
PQ
, which measures the relative change in output
Q for a relative change in input P, and could be defined as
EI
PQ
[P
0
/Q(P
0
)]SI
PQ
(9.4)
4.2.2. A Simple Deterministic Sensitivity Analysis
Procedure
This deterministic sensitivity analysis approach is very
similar to those most often employed in the engineering
economics literature. It is based on the idea of varying one
uncertain parameter value, or set of parameter values, at a
time. The ideas are applied to a water quality example to
illustrate their use.
The output variable of interest can be any performance
measure or indicator. Thus, one does not know if more or
less of a given variable is better or worse. Perhaps too
much and/or too little is undesirable. The key idea is that,
whether employing physical measures or economic
metrics of performance, various parameters (or sets of
associated parameters) are assigned high and low values.
Such ranges may reflect either the differences between the
minimum and maximum values for each parameter, the
5
th
and 95
th
percentiles of a parameters distribution, or
points corresponding to some other criteria. The system
model is then run with the various alternatives, one at a
time, to evaluate the impact of those errors in various sets
of parameter values on the output variable.
Table 9.1 illustrates the character of the results that
one would obtain. Here Y
0
is the nominal value of the
model output when all parameters assume the estimated
best values, and Y
i,L
and Y
i,H
are the values obtained
by increasing or decreasing the values of the i
th
set of
parameters.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 267
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A simple water quality example is employed to illus-
trate this deterministic approach to sensitivity analysis.
The analysis techniques illustrated here are just as appli-
cable to complex models. The primary difference is that
more work would be required to evaluate the various
alternatives with a more complex model, and the model
responses might be more complicated.
The simple water quality model is provided by
Vollenweiders empirical relationship for the average phos-
phorus concentration in lakes (Vollenweider, 1976).
He found that the phosphorus concentration, P (mg/m
3
),
is a function of the annual phosphorus loading rate,
L (mg/m
2

a), the annual hydraulic loading, q (m/a, or

more exactly, m
3
/m
2

a), and the mean water depth, z (m):

P (L/q)/[1 (z/q)
0.5
] (9.5)
L/q and P have the same units; the denominator is an
empirical factor that compensates for nutrient recycling
and elimination within the aquatic lake environment.
268 Water Resources Systems Planning and Management
Data for Lake Ontario in North America would suggest
that reasonable values of the parameters are L 680 mg/m
3
,
q 10.6 m/a, and z 84 m, yielding P 16.8 mg/m
3
.
Values of phosphorus concentrations less than 10 mg/m
3
are
considered oligotrophic, whereas values greater than
20 mg/m
3
generally correspond to eutrophic conditions.
Reasonable ranges reflecting possible errors in the three
parameters yield the values in Table 9.2.
One may want to display these results so they can be
readily visualized and understood. A tornado diagram
(Eschenbach, 1992) would show the lower and upper
values of P obtained from variation of each parameter,
with the parameter with the widest limits displayed on
top, and the parameter with smallest limits on the
bottom. Tornado diagrams (Figure 9.12) are easy to
construct and can include a large number of parameters
without becoming crowded.
An alternative to tornado diagrams is a Pareto chart
showing the width of the uncertainty range associated
with each variable, ordered from largest to smallest. A
Pareto chart is illustrated in Figure 9.13.
Another visual presentation is a spider plot showing the
impact of uncertainty in each parameter on the variable in
question, all on the same graph (Eschenback, 1992;
DeGarmo et al., 1993). A spider plot, Figure 9.14, shows
the particular functional response of the output to each
parameter on a common scale, so one needs a common
metric to represent changes in all of the parameters. Here,
we use percentage change from the nominal or best values.
Spider plots are a little harder to construct than
tornado diagrams, and can generally include only four or
five variables without becoming crowded. However, they
provide a more complete view of the relationships
between each parameter and the performance measure. In
particular, a spider plot reveals non-linear relationships
Table 9.1. Sensitivity of model output Y to possible errors in
four parameter sets containing a single parameter or a group
of parameters that vary together.
low
value nominal
high
value
1
2
3
4
Y
2,L
Y
2,L
Y
3,L
Y
4,L
Y
0
Y
0
Y
0
Y
0
Y
1,H
Y
2,H
Y
3,H
Y
4,H
p
a
r
a
m
e
t
e
r

s
e
t
E020829m
z
L
q
mean depth (m)
P loading (mg/m
3
. a)
hydraulic loading (m/a)
81
500
8
87
900
13.5
parameter value
low high
17.0
12.4
20.0
16.6
22.3
14.4
phosphorus
concentration
P
low
P
high
E020829n
Table 9.2. Sensitivity of estimates of
phosphorus concentration (mg/m
3
) to
model parameter values. The two right-
most values in each row correspond to
the low and high values of the
parameter, respectively.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 268
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 269
Groups of factors
It is often the case that reasonable error scenarios would
have several parameters changing together. For this
reason, the alternatives have been called parameter sets.
For example, possible errors in water depth would be
accompanied by corresponding variations in aquatic vege-
tation and chemical parameters. Likewise, alternatives
related to changes in model structure might be accompa-
nied with variations in several parameters. In other cases,
there may be no causal relationship among possible errors
(such as model structure versus inflows at the boundary of
the modelled region), but they might still interact to affect
the precision of model predictions.
Combinations
If one or more non-grouped parameters interact in signif-
icant ways, then combinations of one or more errors
should be investigated. However, one immediately runs
into a combinatorial problem. If each of m parameters can
have 3 values (high, nominal, and low), then there are
3
m
combinations, as opposed to 2m 1 if each parameter
is varied separately. (For m 5, the differences are
3
5
243 versus 2(5) 1 11.) These numbers can be
reduced by considering instead only combinations of
extremes so that only 2
m
1 cases need be considered
(2
5
1 33), which is a more manageable number.
However, all of the parameters would be at one extreme or
the other, and such situations would be very unusual.
phosphorus concentration
E
0
2
0
1
1
0
z
0 5 10 15 20 25 30 35
L
Q
Z
Figure 9.12. A tornado diagram showing the range of the
output variables representing phosphorus concentrations for
high and low values of each of the parameter sets.
Parameters are sorted so that the largest range is on top,
and the smallest on the bottom, so the diagram looks like a
tornado.
E
0
2
0
1
1
4
a
p
h
o
s
p
h
o
r
u
s
u
n
c
e
r
t
a
i
n
t
y

r
a
n
g
e
L Q Z
10
8
6
4
2
0
9.9
5.6
0.4
Figure 9.13. A Pareto chart showing the range of the output
variable representing phosphorus concentrations resulting
from high and low values of each parameter set considered.
p
h
o
s
p
h
o
r
u
s

c
o
n
c
e
n
t
r
a
t
i
o
n
-80
30
25
20
15
10
5
0
-60 -40 -20 0 20 40 60 80
change in parameter (%)
Q
Z
L
E
0
2
0
1
1
4
b
Figure 9.14. A spider Plot illustrates the relationships
between model output describing phosphorus
concentrations and variations in each of the parameter
sets, expressed as a percentage deviation from their
nominal values.
and the relative sensitivity of the performance measure to
(percentage) changes in each variable.
In the spider plot, the linear relationship between P and
L and the gentle non-linear relationship between P and q is
illustrated. The range for z has been kept small, given the
limited uncertainty associated with that parameter.
4.2.3. Multiple Errors and Interactions
An important issue that should not be ignored is the
impact of simultaneous errors in more than one param-
eter. Probabilistic methods directly address the occurrence
of simultaneous errors, but the correct joint distribution
needs to be employed. With simple sensitivity analysis
procedures, errors in parameters are generally investigated
one at a time or in groups. The idea of considering pairs or
sets of parameters is discussed here.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 269
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Two factors at a time
A compromise is to consider all pairs of two parameters at a
time. There are m(m1)/2 possible pairs of m parameters.
Each parameter has a high and low value. Since there are
4 combinations of high and low values for each pair, there
are a total of 2m(m1) combinations. (For m5 there
are 40 combinations of two parameters each having two
values.)
The presentation of these results could be simplified by
displaying for each case only the maximum error, which
would result in m(m 1)/2 cases that might be displayed
in a Pareto diagram. This would allow identification of
those combinations of two parameters that might yield the
largest errors and thus are of most concern.
For the water quality example, if one plots the absolute
value of the error for all four combinations of high ()
and low () values for each pair of parameters, they
obtain Figure 9.15.
Considering only the worst error for each pair of vari-
ables yields Figure 9.16.
Here we see, as is no surprise, that the worst error results
from the most unfavourable combination of L and q values.
If both parameters have their most unfavourable values, the
predicted phosphorus concentration would be 27 mg/m
3
.
4.2.4. First-Order Sensitivity Analysis
The above deterministic analysis has trouble representing
reasonable combinations of errors in several parameter
sets. If the errors are independent, it is highly unlikely that
any two sets would actually be at their extreme ranges at
the same time. By defining probability distributions of
270 Water Resources Systems Planning and Management
the values of the various parameter sets, and specifying
their joint distributions, a probabilistic error analysis can
be conducted. In particular, for a given performance
indicator, one can use multivariate linear analyses to
evaluate the approximate impact on the performance
indices of uncertainty in various parameters. As shown
below, the impact depends upon the square of the sensi-
tivity coefficients (partial derivatives) and the variances
and covariances of the parameter sets.
For a performance indicator I F(Y), which is a func-
tion F() of model outputs Y, which are in turn a function
g(P) of input parameters P, one can use a multivariate
Taylor series approximation of F to obtain the expected
value and variance of the indicator:
E[I] F(based on mean values of input parameters)
(9.6) ( / ) [ / ] [ , ] 1 2
2



F P P P P
i j i j
j i
Cov
e
r
r
o
r

i
n

p
h
o
s
p
h
o
r
u
s
10
8
6
4
2
0
+L-q -L+q +L-z +L+z -L+z -L-z -q-z -q+z +q+z +L+q +q-z -L-q
combination of errors
E
0
2
0
1
1
4
c
Figure 9.15. Pareto diagram
showing errors in phosphorus
concentrations for all
combinations of pairs of input
parameters errors. A
indicates a high value, and a
indicates a low value for
indicated parameter. L is the
phosphorus loading rate, q is the
hydraulic loading, and z is the
mean lake depth.
E
0
2
0
1
1
4
d
e
r
r
o
r
i
n

p
h
o
s
p
h
o
r
u
s
+L-q +L-z -q-z
10
8
6
4
2
0
9.72
5.74
3.51
combination of parameters
Figure 9.16. Pareto diagram showing worst error combinations
for each pair of input parameters. A '' indicates a high value,
and a '' indicates a low value for indicated parameter.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 270
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 271
and
Var[I]
i

j
(F/P
i
)(F/P
j
) Cov[P
i
, P
j
] (9.7)
where (F/P
i
) are the partial derivatives of the function F
with respect to P
i
evaluated at the mean value of the input
parameters P
i
, and F
2
/P
i
P
j
are the second partial deriv-
atives. The covariance of two random input parameters P
i
and P
j
is the expected value of the product of differences
between the values and their means:
Cov[P
i
, P
j
] E[(P
i
E[P
i
])(P
j
E[P
j
])] (9.8)
If all the parameters are independent of each other, and
the second-order terms in the expression for the mean
E[I] are neglected, one obtains
E[I] F(based on mean values of input
parameters) (9.9)
and
Var[I] [F/P
i
]
2
Var[P
i
] (9.10)
(Benjamin and Cornell, 1970). Equation 9.6 for E[I]
shows that in the presence of substantial uncertainty, the
mean of the output from non-linear systems is not simply
the system output corresponding to the mean of the
parameters (Gaven and Burges, 1981). This is true for any
non-linear function.
Of interest in the analysis of uncertainty is the approx-
imation for the variance Var[I] of indicator I. In Equation
9.10 the contribution of P
i
uncertainty to the variance of
I equals Var[P
i
] times [F/P
i
]
2
, which are the squares of
the sensitivity coefficients for indicator I with respect to
each input parameter value P
i
.
An Example of First-Order Sensitivity Analysis
It may appear that first-order analysis is difficult because
the partial derivatives of the performance indicator I are
needed with respect to the various parameters. However,
reasonable approximations of these sensitivity coefficients
can be obtained from the simple sensitivity analysis
described in Table 9.3. In that table, three different
parameter sets, P
i
, are defined in which one parameter of
the set is at its high value, P
iH
, and one is at its low value,
P
iL
, to produce corresponding values (called high, I
iH
, and
low, I
iL
) of a system performance indicator I.
i

It is then necessary to estimate some representation of

the variances of the various parameters with some consistent
procedure. For a normal distribution, the distance between
the 5
th
and 95
th
percentiles is 1.645 standard deviations on
each side of the mean, or 2(1.645) 3.3 standard devia-
tions. Thus, if the high/low range is thought of as approxi-
mately a 595 percentile range for a normally distributed
variate, a reasonable approximation of the variance might be
Var[P
i
] {[P
iH
P
iL
]/3.3}
2
(9.11)
This is all that is needed. Use of these average sensitivity
coefficients is very reasonable for modelling the behaviour
of the system performance indicator I over the indicated
ranges.
As an illustration of the method of first-order uncer-
tainty analysis, consider the lake-quality problem
described earlier. The system performance indicator in
this case is the model output, the phosphorus concentra-
tion P, and the input parameters, now denoted as X L,
q and z. The standard deviation of each parameter is
assumed to be the specified range divided by 3.3. Average
sensitivity coefficients P/X were calculated. The results
are reported in Table 9.4.
Assuming the parameter errors are independent:
Var[P] 9.18 2.92 0.02 12.12 (9.12)
The square root of 12.12 is the standard deviation and
equals 3.48. This agrees well with a Monte Carlo analysis
reported below.
Note that 100*(9.18/12.12), or about 76% of the total
parameter error variance in the phosphorus concentration
P, is associated in the phosphorus loading rate L and the
remaining 24% is associated with the hydrological loading
q. Eliminating the uncertainty in z would have a negligible
impact on the overall model error. Likewise, reducing the
Table 9.3. Approximate parameter sensitivity coefficients.
value
1
2
3
p
a
r
a
m
e
t
e
r

s
e
t
E
0
2
0
8
2
9
p
low high
I
1L
I
1H
[I
1H
I
1L
]/[ P
1H
P
1L
]
I
2L
I
2H
[I
2H
I
2L
]/[ P
2H
P
2L
]
I
3L
I
3H
[I
3H
I
3L
]/[ P
3H
P
3L
]
sensitivity coefficient
wrm_ch09.qxd 8/31/2005 11:56 AM Page 271
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
272 Water Resources Systems Planning and Management
Table 9.4. Calculation of approximate
parameter sensitivity coefficients.
z m
L mg/m
2
. a
q m/a
variable
X units St Dev[X] Var[X] %
(P/ X)
2
P/ X
-0.074
0.025
-1.024
1.82
121.21
1.67
0.02
9.18
2.92
0.2
75.7
24.1
E
0
2
0
8
2
9
q
to exclude some parameter combinations. Such errors
can result in a distortion in the ranking of predominant
sources of uncertainty. However, in most cases very sim-
ilar results were obtained.
4.2.5. Fractional Factorial Design Method
An extension of first-order sensitivity analysis would be
a more complete examination of the response surface using
a careful statistical design. First, consider a complete facto-
rial design. Input data are divided into discrete levels.
The simplest case is two levels, which can be defined as a
nominal value, and a high (low) value. Simulation runs
are made for all combinations of parameter levels. For n
different inputs, this would require 2
n
simulation runs.
Hence, for a three-input variable or parameter problem,
eight runs would be required. If four discrete levels of
each input variable or parameter were allowed to provide a
more reasonable description of a continuous variable, the
three-input data problem would require 4
3
or 64 simula-
tion runs. Clearly this is not a useful tool for large regional
water resources simulation models.
A fractional factorial design involves simulating only a
fraction of what is required from a full factorial design
method. The loss of information prevents a complete
analysis of the impacts of each input variable or parameter
on the output.
To illustrate the fractional factorial design method,
consider the two-level with three-input variable or param-
eter problem. Table 9.5 below shows the eight simula-
tions required for a full factorial design method. The
and symbols show the upper and lower levels of each
input variable or parameter P
i
where i 1, 2, 3. If all
eight simulations were performed, seven possible effects
could be estimated. These are the individual effects of the
three inputs P
1
, P
2
, and P
3
; the three two-input variable or
error in q would at best have a modest impact on the total
error.
Due to these uncertainties, the estimated phosphorus
concentration has a standard deviation of 3.48. Assuming
the errors are normally distributed, and recalling that
1.645 standard deviations around the mean define a
595 percentile interval, the 595 percentile interval
would be about
16.8 1.645(3.48) mg/m
3
16.8 5.7 mg/m
3
11.1 to 22.5 mg/m
3
(9.13)
These error bars indicate that there is substantial uncer-
tainty associated with the phosphorus concentration P,
primarily due to uncertainty in the loading rate L.
The upper bound of 22.6 mg/m
3
is considerably less
than the 27 mg/m
3
that would be obtained if both L and
q had their most unfavourable values. In a probabilistic
analysis with independent errors, such a combination is
highly unlikely.
Warning on Accuracy
First-order uncertainty analysis is indeed an approximate
method based upon a linearization of the response
function represented by the full simulation model. It
may provide inaccurate estimates of the variance of
the response variable for non-linear systems with large
uncertainty in the parameters. In such cases, Monte Carlo
simulation (discussed below and in Chapters 7 and 8) or
the use of higher-order approximation may be required.
Beck (1987) cites studies that found that Monte Carlo and
first-order variances were not appreciably different, and a
few studies that found specific differences. Differences
are likely to arise when the distributions used for the
parameters are bimodal (or otherwise unusual), or some
rejection algorithm is used in the Monte Carlo analysis
wrm_ch09.qxd 8/31/2005 11:56 AM Page 272
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 273
parameter interactions, (P
1
)(P
2
), (P
1
)(P
3
), and (P
2
)(P
3
);
and the one three-input variable or parameter interaction,
(P
1
)(P
2
)(P
3
).
Consider an output variable Y, where Y
j
is the value of
Y in the j
th
simulation run. Then an estimate of the effect,
denoted (Y | P
i
), that input variable or parameter P
i
has
on the output variable Y is the average of the four separate
effects of varying P
i
:
For i 1:
(Y | P
1
) 0.25[(Y
2
Y
1
) (Y
4
Y
3
)
(Y
6
Y
5
) (Y
8
Y
7
)] (9.14)
Each difference in parentheses is the difference between a
run in which P
1
is at its upper level and a run in which P
1
is at its lower level, but the other two parameter values, P
2
and P
3
, are unchanged. If the effect is equal to 0, then,
in that case, P
1
has on average no impact on the output
variable Y.
Similarly the effects of P
2
and P
3
, on variable Y can be
estimated as:
(Y | P
2
) 0.25{(Y
3
Y
1
) (Y
4
Y
2
) (Y
7
Y
5
)
(Y
8
Y
6
)} (9.15)
and
(Y | P
3
) 0.25{(Y
5
Y
1
) (Y
6
Y
2
) (Y
7
Y
3
)
(Y
8
Y
4
)} (9.16)
Consider next the interaction effects between P
1
and P
2
.
This is estimated as the average of the difference between
the average P
1
effect at the upper level of P
2
and the
average P
1
effect at the lower level of P
2
. This is the same
as the difference between the average P
2
effect at the
upper level of P
1
and the average P
2
effect at the lower
level of P
1
:
(Y | P
1
, P
2
) (1/2){[(Y
8
Y
7
) (Y
4
Y
3
)]/2
[(Y
2
Y
1
) (Y
6
Y
5
)]/2}
(1/4){[(Y
8
Y
6
) (Y
4
Y
2
)]
[(Y
3
Y
1
) (Y
7
Y
5
)]} (9.17)
Similar equations can be derived to show the interaction
effects between P
1
and P
3
, and between P
2
and P
3
and the
interaction effects among all three inputs P
1
, P
2
, and P
3
.
Now assume only half of the simulation runs were
performed, perhaps runs 2, 3, 5 and 8 in this example. If
only outputs Y
2
, Y
3
, Y
5
, and Y
8
are available, for our
example:
(Y | P
3
) (Y | P
1
, P
2
)
0.5{(Y
8
Y
3
) (Y
2
Y
5
)} (9.18)
The separate effects of P
3
and of P
1
P
2
are not available
from the output. This is the loss in information resulting
from fractional instead of complete factorial design.
4.2.6. Monte Carlo Sampling Methods
The Monte Carlo method of performing sensitivity analy-
ses, illustrated in Figure 9.17, first selects a random set of
input data values drawn from their individual probability
distributions. These values are then used in the simula-
tion model to obtain some model output variable values.
This process is repeated many times, each time making
sure the model calibration is valid for the input data
values chosen. The end result is a probability distribution
of model output variables and system performance
indices that results from variations and possible values of
all of the input values.
Using a simple Monte Carlo analysis, values of all of
the parameter sets are selected randomly from distribu-
tions describing the individual and joint uncertainty in
each, and then the modelled system is simulated to obtain
estimates of the selected performance indices. This must
be done many times (often well over 100) to obtain a
statistical description of system performance variability.
The number of replications needed is generally not
dependent on the number of parameters whose errors are
to be analysed. One can include in the simulation the
Table 9.5. A three-input factorial design.
s
i
m
u
l
a
t
i
o
n

r
u
n
1
2
3
4
5
6
7
8
Y
Y
Y
Y
Y
Y
Y
Y
1
2
3
4
5
6
7
8
v
a
l
u
e

o
f

o
u
t
p
u
t

-

v
a
r
i
a
b
l
e
Y
-
+
-
+
-
+
-
+
-
-
+
+
-
-
+
+
-
-
-
-
+
+
+
+
v
a
l
u
e

o
f

i
n
p
u
t
P
1
P P
2 3
E
0
2
0
8
2
9
r
wrm_ch09.qxd 8/31/2005 11:56 AM Page 273
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
274 Water Resources Systems Planning and Management
probability of output data variable x
possible values of output x
possible values of output y
possible values of output z
probability of output data variable y
probability of output data variable z
0
0
0
0.0
0.0
0.0
expected or mean
value of input variable k
resulting random
value of input variable k
random number drawn from
uniform distribution
between 0 and 1
I)
II)
plot output distribution IV)
input data
output data
simulation hydrological simulation model
E
0
2
0
1
1
4
e
output data file
III)
r
e
p
e
a
t

s
i
m
u
l
a
t
i
o
n

f
o
r

m
a
n
y

s
a
m
p
l
e
s

o
f

s
e
l
e
c
t
e
d

i
n
p
u
t

d
a
t
a
possible values of input k 0
possible values of input k
0.0
0
probability that actual value S <k
probability of input k
0.5
1.0
0.0
other input data
Figure 9.17. Monte Carlo
sampling and simulation
procedure for finding
distributions of output variable
values based on distributions,
for specified reliability levels, of
input data values. This technique
can be applied to one or more
uncertain input variables at a
time. The output distributions
will reflect the combined effects
of this input uncertainty over the
specified ranges.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 274
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 275
uncertainty in parameters as well as natural variability.
This method can evaluate the impact of single or multiple
uncertain parameters.
A significant problem that arises in such simulations is
that some combinations of parameter values result in
unreasonable models. For example, model performance
with calibration data sets might be inconsistent with
available data sets. The calibration process places interest-
ing constraints on different sets of parameter values. Thus,
such Monte Carlo experiments often contain checks
that exclude combinations of parameter values that are
unreasonable. In these cases the generated results are
conditioned on this validity check.
Whenever sampling methods are used, one must
consider possible correlations among input data values.
Sampling methods can handle spatial and temporal
correlations that may exist among input data values, but
the existence of correlation requires defining appropriate
conditional distributions.
One major limitation of applying Monte Carlo meth-
ods to estimate ranges of risk and uncertainty for model
output variable values, and system performance indicator
values on the basis of these output variable values, is the
computing time required. To reduce the time needed to
perform sensitivity analyses using sampling methods,
some tricks and stratified sampling methods are available.
The discussion below illustrates the idea of a simple
modification (or trick) using a standardized Monte Carlo
analysis. The more general Latin Hypercube Sampling
procedure is also discussed.
Simple Monte Carlo Sampling
To illustrate the use of Monte Carlo sampling methods,
consider again Vollenweiders empirical relationship,
Equation 9.5, for the average phosphorus concentration
in lakes (Vollenweider, 1976). Two hundred values of
each parameter were generated independently from
normal distributions with the means and variances as
shown in Table 9.6.
The table contains the specified means and variances for
the generated values of L, q and z, and also the actual values
of the means and variances of the 200 generated values of
L, q, z and also of the 200 corresponding generated output
phosphorus concentrations, P. Figure 9.18 displays the
distribution of the generated values of P.
One can see that, given the estimated levels of uncer-
tainty, phosphorus levels could reasonably range from
below 10 to above 25. The probability of generating
a value greater than 20 mg/m
3
was 12.5%. The 595
percentile range was 11.1 to 23.4 mg/m
3
. In the figure,
the cumulative probability curve is rough because
only 200 values of the phosphorus concentration were
generated, but these are clearly enough to give a good
impression of the overall impact of the errors.
Sampling Uncertainty
In this example, the mean of the 200 generated values
of the phosphorus concentration, P, was 17.07. However,
a different set of random values would have generated a
Table 9.6. Monte Carlo analysis of lake
phosphorus levels.
parameter L q z P
mean standard
deviations
680.00
121.21
10.60
1.67
84.00
1.82

---
mean standard
deviations
674.18
130.25
10.41
1.73
84.06
1.82
17.07
3.61
specified means and standard deviations
generated means and standard deviations
E
0
2
0
8
2
9
s
wrm_ch09.qxd 8/31/2005 11:56 AM Page 275
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
different set of P values as well. Thus, it is appropriate
to estimate the standard error, SE, of this average. The
standard error equals the standard deviation of the P
values divided by the square root of the sample size n:
SE /(n)
0.5
3.61/(200)
0.5
0.25. (9.19)
From the central limit theorem of mathematical statistics,
the average of a large number of independent values
should have very nearly a normal distribution. Thus, 95%
of the time, the true mean of P should be in the interval
17.1 1.96(0.25), or 16.6 to 17.6 mg/m
3
. This level of
uncertainty reflects the observed variability of P and the
fact that only 200 values were generated.
Making Sense of the Results
A significant challenge with complex models is to deter-
mine from the Monte Carlo simulation which parameter
errors are important. Calculating the correlation between
each generated input parameter value and the output
variable value is one way of doing this. As Table 9.7 below
shows, on the basis of the magnitudes of the correlation
coefficients, errors in L were most important, and those in
q were second in importance.
One can also use regression to develop a linear model
defining variations in the output on the basis of errors
in the various parameters. The results are shown in the
Table 9.8. The fit is very good, and R
2
98%. If the
model for P had been linear, a R
2
value of 100% should
have resulted. All of the coefficients are significantly
different from zero.
Note that the correlation between P and z was positive
in Table 9.7, but the regression coefficient for z is nega-
tive. This occurred because there is a modest negative
correlation between the generated z and q values. Use of
partial correlation coefficients can also correct for such
spurious correlations among input parameters.
Finally, we display a plot, Figure 9.19, based on
this regression model, illustrating the reduction in the
variance of P that results from dropping each variable
276 Water Resources Systems Planning and Management
E
0
2
0
8
2
9
t
variable L q z P
L
q
z
P
1
0.079
0.130
0.851
1
-0.139
-0.434
1
0.144 1
Table 9.7. Correlation analysis of Monte Carlo results.
L
q
z
18.605
0.025
-1.068
-0.085
intercept
coefficient
standardized
error ratio t
1.790
0.000
10.39
85.36
0.022
0.021
-48.54
- 4.08
E
0
2
0
8
2
9
u
Table 9.8. Results of regression analysis on Monte Carlo
Results.
E
0
2
0
7
2
5
k
v
a
r
i
a
n
c
e

e
x
p
l
a
i
n
e
d
L Q Z
10
8
6
4
2
0
Figure 9.19. Reduction in the variance of P that is due to
dropping from the regression model each variable individually.
Clearly L has the biggest impact on the uncertainty in P, and z
the least.
Figure 9.18. Distribution of lake phosphorus concentrations
from Monte Carlo analysis.
c
u
m
u
l
a
t
i
v
e

p
r
o
b
a
b
i
l
i
t
y
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
phosphorus concentration
0 5 10 15 20 25 30
E
0
2
0
1
1
5
a
wrm_ch09.qxd 8/31/2005 11:56 AM Page 276
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 277
individually. Clearly, L has the biggest impact on the
uncertainty in P, and z the least.
Standardized Monte Carlo Analysis
Using a standardized Monte Carlo analysis, one could
adjust the generated values of L, q and z above so that the
generated samples actually have the desired mean and
variance. While making that correction, one can also
shuffle their values so that the correlations among the
generated values for the different parameters are near
zero, as is desired. This was done for the 200 generated
values to obtain the statistics shown in Table 9.9.
Repeating the correlation analysis from before (shown
in Table 9.10) now yields much clearer results that are
in agreement with the regression analysis. The correlation
between P and both q and z are now negative, as they
should be. Because the generated values of the three param-
eters have been adjusted to be uncorrelated, the signal from
one is not confused with the signal from another.
The mean phosphorus concentration changed very
little. It is now 17.0 instead of 17.1 mg/m
3
.
Generalized Likelihood Estimation
Beven (1993) and Binley and Beven (1991) suggest a
Generalized Likelihood Uncertainty Estimation (GLUE)
technique for assessment of parameter error uncertainty
using Monte Carlo simulation. It is described as a formal
methodology for some of the subjective elements of model
calibration (Beven, 1989, p. 47). The basic idea is to begin
by assigning reasonable ranges for the various parameters,
and then to draw parameter sets from those ranges using
a uniform or some similar (and flat) distribution. These
generated parameter sets are then used on a calibration
data set so that unreasonable combinations can be
rejected, while reasonable values are assigned a posterior
probability based upon a likelihood measure that may
reflect several dimensions and characteristics of model
performance.
Let L(P
i
) 0 be the value of the likelihood measure
assigned to the calibration sequence of the i
th
parameter
set. Then the model predictions generated with parameter
set/combination P
i
are assigned posterior probability,
p(P
i
), where
p(P
i
) L(P
i
) L(P
j
) (9.20)
These probabilities reflect the form of Bayes theorem,
which is well supported by probability theory (Devore,
1991). This procedure should capture reasonably well the
dependence or correlation among parameters, because rea-
sonable sequences will all be assigned larger probabilities,
whereas sequences that are unable to reproduce the system
response over the calibration period will be rejected or
assigned small probabilities.
However, in a rigorous probabilistic framework, the L
would be the likelihood function for the calibration series
for particular error distributions. (This could be checked
with available goodness-of-fit procedures; for example,
Kuczera, 1988.) When relatively ad hoc measures are
adopted for the likelihood measure with little statistical
validity, the p(P
i
) probabilities are best described as
pseudo probabilities or likelihood weights.
Another concern with this method is the potential
efficiency. If the parameter ranges are too wide, a large
j

E
0
2
0
8
2
9
v
parameter L q z P
680.00
121.21
680.00
121.21
10.60
1.67
10.60
1.67
84.00
1.82
84.00
1.82

---
17.03
3.44
specified means and standard deviations
generated means and standard deviations
mean
standard deviations
mean
standard deviations
E
0
2
0
8
2
9
w
variable L q z P
L
q
z
P
1.00
0.01
0.02
0.85
1.00
0.00
-0.50
1.00
-0.02 1.00
Table 9.10. Correlation analysis of standardized Monte Carlo
results.
Table 9.9. Standardized Monte Carlo analysis of lake
phosphorus levels.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 277
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
number of unreasonable or very unlikely parameter com-
binations will be generated. These will either be rejected
or else will have small probabilities and, thus, little effect
on the analysis. In this case, the associated processing
would be a waste of effort. A compromise is to use some
data to calibrate the model and to generate a prior or ini-
tial distribution for the parameters that is at least centred
in the best range (Beven, 1993, p. 48). Use of a different
calibration period to generate the p(P
i
) then allows
an updating of those initial probabilities to reflect the
information provided by the additional calibration period
with the adopted likelihood measures.
After the accepted sequences are used to generate sets
of predictions, the likelihood weights would be used in
the calculation of means, variances and quantiles. The
resulting conditional distribution of system output
reflects the initial probability distributions assigned to
parameters, the rejection criteria and the likelihood meas-
ure adopted to assign likelihood weights.
Latin Hypercube Sampling
For the simple Monte Carlo simulations described in
Chapters 7 and 8, with independent errors, a proba-
bility distribution is assumed for each input parameter
or variable. In each simulation run, values of all input
data are obtained from sampling those individual and
independent distributions. The value generated for an
input parameter or variable is usually independent of
what that value was in any previous run, or what other
input parameter or variable values are in the same run.
This simple sampling approach can result in a cluster-
ing of parameter values, and hence both redundancy
of information from repeated sampling in the same
regions of a distribution and lack of information
because there is no sampling in other regions of the
distributions.
A stratified sampling approach ensures more even
coverage of the range of input parameter or variable
values with the same number of simulation runs. This can
be accomplished by dividing the input parameter or vari-
able space into sections and sampling from each section
with the appropriate probability.
One such approach, Latin hypercube sampling (LHS),
divides each input distribution into sections of equal
278 Water Resources Systems Planning and Management
probability for the specified probability distribution, and
draws one observation randomly from each section.
Hence the ranges of input values within each section
actually occur with equal frequency in the experiment.
These values from each section for each distribution
are randomly assigned to those from other sections to
construct sets of input values for the simulation analysis.
Figure 9.20 shows the steps in constructing an LHS for
six simulations involving three inputs P
j
(P
1
, P
2
, and P
3
)
and six intervals of their respective normal, uniform and
triangular probability distributions.
5. Performance Indicator
Uncertainties
5.1. Performance Measure Target Uncertainty
Another possible source of uncertainty is the selection of
performance measure target values. For example, con-
sider a target value for a pollutant concentration based on
the effect of exceeding it in an ecosystem. Which target
value is best or correct? When this is not clear, there are
various ways of expressing the uncertainty associated
with any target value. One such method is the use of
fuzzy sets (Chapter 5). Use of grey numbers or intervals
instead of white or fixed target values is another. When
some uncertainty or disagreement exists over the selec-
tion of the best target value for a particular performance
measure, it seems to us the most direct and transparent
way to do this is to subjectively assign a distribution over
a range of possible target values. Then this subjective
probability distribution can be factored into the tradeoff
analysis, as outlined in Figure 9.21.
One of the challenges associated with defining and
including in an analysis the uncertainty associated with a
target or threshold value for a performance measure is
that of communicating just what the result of such an
analysis means. Referring to Figure 9.21, suppose the
target value represents some maximum limit of a pollu-
tant concentration, such as phosphorus, in the flow dur-
ing a given period of time at a given site or region, and it
is not certain just what that maximum limit should be.
Subjectively defining the distribution of that maximum
limit, and considering that uncertainty along with the
wrm_ch09.qxd 8/31/2005 11:56 AM Page 278
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 279
uncertainty (probability of exceedance function) of
pollutant concentrations the performance measure
one can attach a reliability to any probability of exceed-
ing the maximum desired concentration value.
The 95% probability of exceedance shown on Figure
9.21, say P
0.95
, should be interpreted as: we can be 95%
confident that the probability of the maximum desired
pollutant concentration being exceeded will be no greater
than P
0.95
. We can be only 5% confident that the proba-
bility of exceeding the desired maximum concentration
will be no greater than the lower P
0.05
value. Depending
on whether the middle line through the subjective distri-
bution of target values in Figure 9.21 represents the most
likely or median target value, the associated probability of
k =
k =
k =
1
1
1
2
2
2
3
3
3
4
4
4
5
5
5
6
6
6
normal distribution for P
uniform distribution for P
triangular distribution for P
1
2
3
model
run
randomly selected
interval k for input P
j
4
2
1
6
3
5
3
1
6
4
2
5
5
6
1
4
3
2
1
2
3
4
5
6
P P
2 3
P
1
I)
randomly assign each of 6 intervals k
for each input j to 6 model-runs
model
run
randomly selected values
P
jk
1
2
3
4
5
6
P P
2k 3k
P
1k
P P
23 35
P
14
for each simulation run
value of
output
variable
Y
P P
21 36
P
12
P P
26 31
P
11
P P
24 34
P
16
P P
22 33
P
13
P P
25 32
P
15
Y
1
Y
2
Y
3
Y
4
Y
5
Y
6
II)
randomly sample from each specified interval k
to obtain values P for use in each simulation run
jk
plot histogram of output
resulting from model-runs
III)
p
r
o
b
a
b
i
l
i
t
y
Y
2
Y Y Y Y Y
5 6 3 4 1
discrete range of output Y
E
0
2
0
1
1
5
d
k
k
Figure 9.20. Schematic
representation of a Latin
hypercube sampling procedure
for six simulation runs.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 279
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
exceedance is either the most likely, as indicated in Figure
9.21, or that for which we are only 50% confident.
Figure 9.22 attempts to show how to interpret the
reliabilities when the uncertain performance targets are:
minimum acceptable levels that are to be maximized,
maximum acceptable levels that are to be minimized, or
optimum levels.
An example of a minimum acceptable target level might
be the population of wading birds in an area. An example
of a maximum acceptable target level might be, again,
the phosphorus concentration of the flow in a specific
wetland or lake. An example of an optimum target level
might be the depth of water most suitable for selected
species of aquatic vegetation during a particular period of
the year.
For performance measure targets that are not
expressed as minimum or maximum limits but that are
the best values, referring to Figure 9.22, one can state
that there is a 90% reliability that the probability of
280 Water Resources Systems Planning and Management
achieving the desired target is no more than . The 90%
reliability level of not achieving the desired target is at
least . The probability of the performance measure
being too low is at least , and the probability of the per-
formance measure being too high is at least , again at the
90% reliability levels. As the reliability level decreases, the
bandwidth decreases, and the probability of not meeting
the target increases.
Now, clearly there is uncertainty associated with each of
these uncertainty estimations, and this raises the question
of how valuable is the quantification of the uncertainty of
each additional component of the plan in an evaluation
process. Will plan-evaluators and decision-makers benefit
from this additional information, and just how much addi-
tional uncertainty information is useful?
Now consider again the tradeoffs that need to be
made, as illustrated in Figure 9.7. Instead of considering
a single target value, as shown on the figure, assume
there is a 90% reliability or probability range associated
with that single performance measure target value. Also
Figure 9.21. Combining the
probability distribution of
performance measure values
with the probability distribution
of performance measure target
values to estimate the
confidence one has in the
probability of exceeding a
maximum desired target value.
p
r
o
b
a
b
i
l
i
t
y

d
e
n
s
i
t
y

d
i
s
t
r
i
b
u
t
i
o
n
s
p
r
o
b
a
b
i
l
i
t
y

o
f

e
x
c
e
e
d
a
n
c
e
performance measure values
performance measure values
90%
reliability band associated with a performance measure target value
performance measure distribution
0
1
0
90% reliability band associated with probability that
the maximum desired target value will be exceeded
for a maximum desired target:
95% confident of < this exceedance probability
most likely probability of exceeding target
5% confident of < this exceedance probability
E
0
2
0
7
2
5
m
performance measure target distribution
wrm_ch09.qxd 8/31/2005 11:56 AM Page 280
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 281
p
r
o
b
a
b
i
l
i
t
y
d
e
n
s
i
t
y

d
i
s
t
r
i
b
u
t
i
o
n
s
performance measure values
5% confident that the probability of meeting or
exceeding the desired target is at least +.
5% confident that the probability of not meeting
or exceeding the desired target is less than .
95% confident that the probability of meeting
or exceeding the desired target is at least .
95% confident that the probability of not meeting
or exceeding the desired target is less than +.

90% confidence band associated with a performance
measure target value
For a performance measure
that is to be maximized:
For a performance measure
that is to be maximized:
For a performance measure
that is to be minimized:
For a performance measure
that is to be minimized:
95% confident that the probability of exceeding
a desired target is no greater than +.
95% confident that the probability of not
exceeding the desired target is at least .
5% confident that the probability of exceeding
a desired target is no greater than .
5% confident that the probability of not
exceeding the desired target is at least +.
E
0
2
0
7
2
5
n
Figure 9.22. Interpreting the
results of combining
performance measure
probabilities with performance
measure target probabilities
depends on the type of
performance measure. The
letters , and represent
proportions of the probability
density function of performance
measure values. (Hence
probabilities 1.)
assume that the target is a maximum desired upper limit,
e.g. of some pollutant concentration.
In the case shown in Figure 9.23, the tradeoff is
clearly between cost and reliability. In this example, no
matter what reliability one chooses, Plan A is preferred
to Plan B with respect to reliability, but Plan B is
preferred to Plan A with respect to cost. The tradeoff is
only between these two performance indicators or
measures.
Consider, however, a third plan, as shown in Figure
9.24. This situation adds to the complexity of making
appropriate tradeoffs. Now there are three criteria: cost,
probability of exceedance and the reliabilities of those
probabilities. Add to this the fact that there will be multi-
ple performance measure targets, each expressed in terms
of their maximum probabilities of exceedance and the
reliability of those probabilities.
In Figure 9.24, in terms of cost the plans are ranked,
from best to worst, B, C and A. In terms of the 90 reliable
probability, they are ranked A, B and C, but at the 50
percent reliability level the ranking is A, C and B.
If the plan evaluation process has difficulty handling all
this, it may indicate the need to focus the uncertainty
analysis effort on just what is deemed important,
achievable and beneficial. Then, when the number of alter-
natives has been narrowed down to only a few that appear
to be the better ones, a more complete uncertainty analy-
sis can be performed. There is no need and no benefit in
performing sensitivity and uncertainty analyses on all
possible management alternatives. Rather one can focus on
those alternatives that look most promising, and then
carry out additional uncertainty and sensitivity analyses
only when important uncertain performance indicator
values demands more scrutiny.
5.2. Distinguishing Differences Between
Performance Indicator Distributions
Simulations of alternative water management infrastruc-
ture designs and operating policies require a comparison
of the simulation outputs the performance measures or
indicators associated with each alternative. A reasonable
wrm_ch09.qxd 8/31/2005 11:56 AM Page 281
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
question to ask is whether the observed differences are
statistically significant. Can one really tell whether one
alternative is better than another, or whether the observed
differences are explainable by random variations attribut-
able to variations in the inputs and the way the system
responds?
This is a common statistical issue that is addressed by
standard hypothesis tests (Devore, 1991; Benjamin and
Cornell, 1970). Selection of an appropriate test requires
that one first resolve what type of change one expects in
282 Water Resources Systems Planning and Management
the variables. To illustrate, consider the comparison of
two different operating policies. Let Y
1
denote the set of
output performance variable values with the first policy,
and Y
2
the set of output performance variable values of
the second policy. In many cases, one would expect one
policy to be better than the other. One measure might
be the difference in the mean of the variables; for exam-
ple, E[Y
1
] E[Y
2
]. Alternatively one could check the
difference in the median (50
th
percentile) of the two
distributions.
In addition, one could look for a change in the
variability or variance, or a shift in both the mean and
the variance. Changes described by a difference in the
mean or median often make the most sense, and many
statistical tests are available that are sensitive to such
changes. For such investigations parametric and non-
parametric tests for paired and unpaired data can be
employed.
Consider the differences between paired and unpaired
data. Suppose that the meteorological data for 1941 to
1990 is used to drive a simulation model generating data as
described in Table 9.11.
Here there is one sample, Y
1
(1) through Y
1
(50), for
Policy 1, and another sample, Y
2
(1) through Y
2
(50), for
Policy 2. However, the two sets of observations are not
independent. For example, if 1943 was a very dry year,
probability maximum performance target will be exceeded
E
0
2
0
7
2
5
p
c
o
s
t
5% confident
95% confident
plan A
plan B
plan C
Figure 9.24. Tradeoffs among cost, probabilities and the
reliability or confidence level of those probabilities. The
relative ranking of plans with respect to the probability of
exceeding the desired (maximum limit) target may depend
on the reliability or confidence associated with that
probability.
p
r
o
b
a
b
i
l
i
t
y
d
e
n
s
i
t
y

f
u
n
c
t
i
o
n
s
f
o
r

p
l
a
n
s
a
n
d
A
B
90% confidence interval for
performance measure target values
performance
measure value
probability maximum performance target will be exceeded
E
0
2
0
7
2
5
o
c
o
s
t
B A
5% confident 95% confident
plan A
plan B
Figure 9.23. Two plans showing ranges of
probabilities, depending on the reliability, that
an uncertain desired maximum (upper limit)
performance target value will be exceeded.
The 95% reliability levels are associated with
the higher probabilities of exceeding the
desired maximum target. The 5% reliability
levels are associated with the more desirable
lower probabilities of exceeding the desired
maximum target. Plan A with reduced
probabilities of exceeding the upper limit costs
more than Plan B.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 282
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 283
then we would expect both Y
1
(3) for Policy 1 in that year
and Y
2
(3) for Policy 2 to be unusually small. With such
paired data, one can use a paired hypothesis test to
check for differences. Paired tests are usually easier than
the corresponding unpaired tests that are appropriate in
other cases. For example, if one were checking for a
difference in average rainfall depth between the periods
1941 to 1960 and 1961 to 1990, then one would have
two sets of independent measurements for the two peri-
ods. With such data, one should use a two-sample
unpaired test.
Paired tests are generally based on the differences
between the two sets of output, Y
1
(i) Y
2
(i). These are
viewed as a single independent sample. The question is
then whether the differences are positive (say Y
1
tends to
be larger then Y
2
) or negative (Y
1
tends to be smaller), or
whether positive and negative differences are equally
likely (there is no difference between Y
1
and Y
2
).
Both parametric and non-parametric families of
statistical tests are available for paired data. The common
parametric test for paired data (a one-sample t test)
assumes that the mean of the differences
X(i) Y
1
(i) Y
2
(i) (9.21)
are normally distributed. The hypothesis of no difference
is rejected if the t statistic is sufficiently large, given the
sample size n.
Alternatively, one can employ a non-parametric test
and avoid the assumption that the differences X(i) are
normally distributed. In such a case, one can use the
Wilcoxon Signed Rank test. This non-parametric test
ranks the absolute values |X(i)| of the differences. If the
sum S of the ranks of the positive differences deviates
sufficiently from its expected value, n(n 1)/4 (were
there no difference between the two distributions), one
can conclude that there is a statistically significant
difference between the Y
1
(i) and Y
2
(i) series. Standard
statistical texts have tables of the distribution of the
sum S as a function of the sample size n, and provide a
good analytical approximation for n > 20 (for example,
Devore, 1991). Both the parametric t test and the non-
parametric Wilcoxon Signed Rank test require that the
differences between the simulated values for each year
be computed.
6. Communicating Model Output
Uncertainty
Spending money on reducing uncertainty would seem
preferable to spending it on ways of calculating and
describing it better. Yet attention to uncertainty commu-
nication is critically important if uncertainty analyses and
characterizations are to be of value in a decision-making
process. In spite of considerable efforts by those involved
in risk assessment and management, we know very little
about how to ensure effective risk communication to
gain the confidence of stakeholders, incorporate their
views and knowledge, and influence favourably the
acceptability of risk assessments and risk-management
decisions.
The best way to communicate concepts of uncertainty
may well depend on what the audiences already know
about risk and the various types of probability distributions
(e.g. density, cumulative, exceedance) based on objective
and subjective data, and the distinction between mean or
average values and the most likely values. Undoubtedly
graphical representations of these ways of describing
uncertainty considerably facilitate communication.
The National Research Council (NRC, 1994)
addressed the extensive uncertainty and variability
associated with estimating risk and concluded that risk
characterizations should not be reduced to a single
number or even to a range of numbers intended to
portray uncertainty. Instead, the report recommended
that managers and the interested public should be given
risk characterizations that are both qualitative and
quantitative, and both verbal and mathematical.
Table 9.11. Possible flow data from a fifty-year simulation.
1941
1942
1943
1944
1989
1990
Y
2
(50)
Y
2
(49)
Y
1
(50)
Y
1
(49)
Y
2
( 4)
Y
2
( 3)
Y
2
( 2)
Y
2
( 1)
Y
1
( 4)
Y
1
( 3)
Y
1
( 2)
Y
1
( 1)
E
0
2
0
8
2
9
x
wrm_ch09.qxd 8/31/2005 11:56 AM Page 283
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
In some cases, communicating qualitative information
about uncertainty to stakeholders and the public in
general may be more effective than providing quantitative
information. There are, of course, situations in which
quantitative uncertainty analyses are likely to provide
information that is useful in a decision-making process.
How else can tradeoffs such as those illustrated in
Figures 9.10 and 9.27 be identified? Quantitative
uncertainty analysis can often be used as the basis of
qualitative information about uncertainty, even if the
quantitative information is not what is communicated to
the public.
One should acknowledge to the public the widespread
confusion regarding the differences between variability
and uncertainty. Variability does not change through
further measurement or study, although better sampling
can improve our knowledge about it. Uncertainty reflects
gaps in information about scientifically observable phe-
nomena.
While it is important to communicate prediction
uncertainties and the reliabilities of those uncertainties, it
is equally important to clarify who or what is at risk,
the possible consequences, and the severity and irre-
versibility of an adverse effect should a target value,
for example, not be met. This qualitative information is
often critical to informed decision-making. Risk and
uncertainty communication is always complicated by the
reliability and amounts of available relevant information
as well as how that information is presented. Effective
communication between people receiving information
about who or what is at risk, or what might happen
and just how severe and irreversible an adverse effect
might be should a target value not be met, is just as
important as the level of uncertainty and the reliability
associated with such predictions. A two-way dialogue
between those receiving such information and those
providing it can help identify just what seems best for a
particular audience.
Risk and uncertainty communication is a two-
way activity that involves learning and teaching.
Communicators dealing with uncertainty should learn
about the concerns and values of their audience, their
relevant knowledge and their experience with uncertainty
issues. Stakeholders knowledge of the sources and
reasons for uncertainty needs to be incorporated into assess-
ment and management and communication decisions. By
284 Water Resources Systems Planning and Management
listening, communicators can craft risk messages that better
reflect the perspectives, technical knowledge and concerns
of the audience.
Effective communication should begin before impor-
tant decisions have been made. It can be facilitated in
communities by citizen advisory panels, which can give
planners and decision-makers a better understanding of
the questions and concerns of the community, and an
opportunity to test their effectiveness in communicating
concepts and specific issues regarding uncertainty.
One approach to make risk more meaningful is to
make risk comparisons. For example, a ten parts-per-
billion target for a particular pollutant concentration is
equivalent to ten seconds in over thirty-one years. If this
is an average daily concentration target that is to be
satisfied 99% of the time, then this is equivalent to an
expected violation of less than one day every three
months.
Many perceive the reduction of risk by an order of
magnitude as though it were a linear reduction. A better
way to illustrate orders of magnitude of risk reduction
is shown in Figure 9.25, in which a bar graph depicts
better than words that a reduction in risk from one in a
1,000 (10
3
) to one in 10,000 (10
4
) is a reduction of
90% and that a further reduction to one in 100,000 (10
5
)
is a reduction tenfold less than the first reduction of 90%.
The percent of the risk that is reduced by whatever
measures is a much easier concept to communicate than
reductions expressed in terms of estimated absolute risk
levels, such as 10
5
.
Risk comparisons can be helpful, but they should be
used cautiously and tested if possible. There are dangers
in comparing risks of diverse character, especially when
the intent of the comparison is seen as reducing a risk
(NRC, 1989). One difficulty in using risk comparisons is
that it is not always easy to find risks that are sufficiently
similar to make a comparison meaningful. How is one
to compare two alternatives that have two different costs
and two different risk levels, for example, as shown
in Figure 9.23? One way is to perform an indifference
analysis (Chapter 10), but that can lead to differing results
depending who performs it. Another way is to develop
utility functions using weights, where, for example
reduced phosphorus load by half is equivalent to a 25%
shorter hydroperiod in that area, but again each persons
utility or tradeoff may differ.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 284
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 285
At a minimum, graphical displays of uncertainty can
be helpful. Consider the common system performance
indicators that include:
time-series plots for continuous time-dependent indi-
cators (Figure 9.26 upper left),
probability exceedance distributions for continuous
indicators (Figure 9.26 upper right),
histograms for discrete event indicators (Figure 9.26
lower left),
overlays on maps for space-dependent discrete events
(Figure 9.26 lower right).
The first three graphs in Figure 9.26 could show, in
addition to the single curve or bar that represents the
most likely output, a range of outcomes associated with a
given confidence level. For overlays of information on
maps, different colours could represent the spatial extents
of events associated with different ranges of risk or
uncertainty. Figure 9.27, corresponding to Figure 9.26,
illustrates these approaches for displaying these ranges.
7. Conclusions
This chapter has provided an overview of uncertainty and
sensitivity analyses in the context of hydrological or water
resources systems simulation modelling. A broad range of
tools are available to explore, display and quantify the
sensitivity and uncertainty in predictions of key output
variables and system performance indices with respect to
imprecise and random model inputs and to assumptions
concerning model structure. They range from relatively
1000
800
600
400
200
0
10
-3
10
-4
10
-5
10
-6
E
0
2
0
7
2
5
r
risk level
e.g. probability of exceedance
Figure 9.25. Reducing risk (x-axis) by orders of magnitude is not
equivalent to linear reductions in some risk indicator (y-axis).
time
0
1
F(Y)
F
(
Y
)
F
(
Y
)
discrete events, times,
or alternatives
probability
exceeded
F(Y)
map F(Y)
E
0
2
0
7
2
5
s
Figure 9.26. Different types of displays
used to show model output Y or system
performance indicator values F(Y).
wrm_ch09.qxd 8/31/2005 11:56 AM Page 285
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
simple deterministic sensitivity analysis methods to more
involved first-order analyses and Monte Carlo sampling
methods.
Because of the complexity of many watersheds or river
basins, using Monte Carlo methods for uncertainty analy-
ses may be a very large and unattractive undertaking. It
is therefore often prudent to begin with the relatively
simple deterministic procedures. These, coupled with a
probabilistically based first-order uncertainty analysis
method, can help quantify the uncertainty in key output
variables and system performance indices, and the
relative contributions of uncertainty in different input
variables to the uncertainty in different output variables
and system performance indices. These relative contribu-
tions may differ depending upon which output variables
and indices are of interest.
A sensitivity analysis can provide a systematic assess-
ment of the impact of parameter value imprecision on
output variable values and performance indices, and of
the relative contribution of errors in different parameter
values to that output uncertainty. Once the key variables
are identified, it should be possible to determine the
extent to which parameter value uncertainty can be
reduced through field investigations, development of
better models and other efforts.
286 Water Resources Systems Planning and Management
Model calibration procedures can be applied to
individual catchments and subsystems, as well as to
composite systems. Automated calibration procedures
have several advantages, including the explicit use of
an appropriate statistical objective function, identification
of those parameters that best reproduce the calibration
data set with the given objective function, and the esti-
mations of the statistical precision of the estimated
parameters.
All of these tasks together can represent a formidable
effort. However, knowledge of the uncertainty asso-
ciated with model predictions can be as important to
management decisions and policy formulation as are
the predictions themselves.
No matter how much attention is given to quantifying
and reducing uncertainties in model outputs, uncertain-
ties will remain. Professionals who analyse risk, managers
and decision-makers who must manage risk, and the
public who must live with risk and uncertainty, have
different information needs and attitudes regarding risk
and uncertainty. It is clear that information needs differ
among those who model or use models, those who make
substantial investment or social decisions, and those who
are likely to be affected by those decisions. Meeting those
needs should result in more informed decision-making,
time
0
1
F(Y)
F
(
Y
)
F
(
Y
)
discrete events, times,
or alternatives
probability
exceeded
F(Y)
map F(Y)
E
0
2
0
7
2
5
t
Figure 9.27. Plots of ranges of possible
model output Y or system indicator values
F(Y) for different types of displays.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 286
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 287
but this comes at a cost that should be considered along
with the benefits of having this sensitivity and uncertainty
information.
8. References
BECK, M.B. 1987. Water quality modelling: a review of
the analysis of uncertainty. Water Resources Research,
Vol. 23, No. 8, pp. 13931442.
BENAMAN, J. 2002. A systematic approach to uncertainty
analysis for a distributed watershed model. Ph.D. Thesis,
School of Civil and Environmental Engineering, Cornell
University, Ithaca, N.Y.
BENJAMIN, J.R. and CORNELL, C.A. 1970. Probability,
statistics and decision for civil engineers. New York,
McGraw-Hill.
BEVEN, K. 1989. Changing ideas in hydrology: the case
of physically-based models. Journal of Hydrology. No. 105,
pp. 15772.
BEVEN, K. 1993. Prophecy, reality and uncertainty in
distributed hydrological modelling. Advances in Water
Resources, No. 16, 4151.
BINLEY, A.M. and BEVEN, K.J. 1991. Physically-based
modelling of catchment hydrology: a likelihood approach
to reducing predictive uncertainty. In: D.G. Farmer and
M.J. Rycroft (eds.), Computer modelling in the environmen-
tal sciences, Oxford, Clarendon, pp. 7588.
DEGARMO, E.P.; SULLIVAN, W.G. and BONTADELLI,
J.A. 1993. Engineering economy. New York, MacMillan.
DEVORE, J. 1991. Probability and statistics for engineering
and the sciences, 3rd Edition. London, Brooks/Cole.
ESCHENBACH, T.G. 1992. Spider plots versus tornado
diagrams for sensitivity analysis. Interfaces, No. 22,
pp. 406.
GAVEN, D.C. and BURGES, S.J. 1981. Approximate error
bounds for simulated hydrograph. Journal of Hydraulic
Engineering, Vol. 107, No. 11, pp. 151934.
KUCZERA, G. 1988. On the validity of first-order predic-
tion limits for conceptual hydrological models. Journal of
Hydrology, No. 103, pp. 22947.
LAL, W. 1995. Sensitivity and uncertainty analysis of a
regional model for the natural system of South Florida. West
Palm Beach, Fla., South Florida Water Management
District. Draft report, November.
MCCUEN, R. 1973. The role of sensitivity analysis in
hydrological modeling. Journal of Hydrology, Vol. 18,
No. 1, pp. 3753.
NRC (National Research Council). 1989. Improving risk
communication. Washington, D.C., National Academy Press.
NRC (National Research Council). 1994. Science and
judgment in risk assessment. Washington, D.C., National
Academy Press.
STEDINGER, J.R. and TAYLOR, M.R. 1982. Synthetic
streamflow generation, part II: Effect of parameter uncer-
tainty. Water Resources Research, Vol. 18, No. 4, pp. 91924.
VOLLENWEIDER, R.A. 1976. Advances in defining
critical loading levels for phosphorus in lake eutrophica-
tion, Memorie dellIstituto Italiano di Idrobiologia, No. 33,
pp. 5383.
Additional References (Further Reading)
ANDERSON, J.L. 1998. Embracing uncertainty: the interface
of Bayesian statistics and cognitive psychology. Conservation
Ecology (online), Vol. 2, No. 1, p. 2. Available from the
Internet. URL: http://www.consecol.org/vol2/iss1/art2
BERRY, D.A. 1996. Statistics: a Bayesian perspective.
Belmont, Calif., Duxbury.
CHAPRA, S.C. and RECKHOW, K.H. 1979. Expressing
the phosphorus loading concept in probabilistic terms.
Journal of the Fisheries Research Board of Canada, No. 36,
pp. 2259.
COLWELL, R.K. 1974. Predictability, consistency, and
contingency of periodic phenomena. Ecology, Vol. 55,
pp. 114853.
COVELLO, V.T. and MERKHOFER, M.W. 1993. Risk
assessment methods. London, Plenum.
COVELLO, V.T. 1987. Decision analysis and risk
management decision-making: issues and methods. Risk
Analysis, No. 7, pp. 1319.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 287
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
DEUTSCH, C.V. and JOURNEL, A.G. 1992. GS-LIB:
geostatistical software library and users guide. New York,
Oxford University Press.
DEVORE, J.L. and PECK, R. 1996. Statistics: the explo-
ration and analysis of data, 3rd edn. London, Brooks/Cole.
DILKS, D.W.; CANALE, R.P. and MEIER, P.G. 1992.
Development of Bayesian Monte Carlo techniques for water
quality model uncertainty. Ecological Modelling, Vol. 62,
pp. 14962.
DILKS, D.W. and JAMES, R.T. 2002. Application of Bayesian
Monte Carlo analysis to determine the uncertainty in the Lake
Okeechobee water quality model. Proceedings of Watershed
2002. Alexandria, Va., Water Environment Federation.
DOE (Department of Energy). 1998. Screening assessment
and requirements for a comprehensive assessment. DOE/RL-
96-16, Rev. 1. Richland, Washington, US Department of
Energy.
DUAN, Q.; SOROOSIAN, S. and IBBITT, R.P. 1988. A
maximum likelihood criterion for use with data collected
at unequal time intervals. Water Resources Research,
Vol. 24, No. 7, pp. 116373.
FEDRA, K.A. 1983. Monte Carlo approach to estimation
and prediction. In: M.B. Beck and G. Van Straten (eds.),
Uncertainty and forecasting of water quality, Berlin,
Springer-Verlag, pp. 25991.
FITZ, H.C.; VOINOV, A. and COSTANZA, R. 1995. The
Everglades landscape model: multiscale sensitivity analysis.
West Palm Beach, Fla., South Florida Water Management
District, Everglades Systems Research Division.
FONTAINE, T.A. and JACOMINO, V.M.F. 1997.
Sensitivity analysis of simulated contaminated sediment
transport. Journal of the American Water Resources
Association, Vol. 33, No. 2, pp. 31326.
FREY, H.C. and PATIL, S.R. 2002. Identification and
review of sensitivity analysis methods. Risk Analysis,
Vol. 22, No. 3, pp. 55378.
GARDNER, R.H.; ROJDER, B. and BERGSTROM, U.
1983. PRISM: a systematic method for determining the effect
of parameter uncertainties on model predictions. Technical
Report, Studsvik Energiteknik AB report NW-83-/555.
Sweden, Nykoping.
288 Water Resources Systems Planning and Management
GELMAN, A.; CARLIN, J.; STERN, H. and RUBIN, D.B.
1995. Bayesian data analysis. London, Chapman and Hall.
GLEICK, J. 1987. Chaos: making a new science. New York,
Penguin.
GUPTA, V.K. and SOROOSHIAN, S. 1985a. The auto-
matic calibration of conceptual watershed models using
derivative-based optimization algorithms. Water Resources
Research, Vol. 21, No. 4, pp. 47385.
GUPTA, V.K. and SOROOSHIAN, S. 1985b. The relationship
between data and the precision of parameter estimates of
hydrological models. Journal of Hydrology, No. 81, pp. 577.
HAIMES, Y.Y. 1998. Risk modelling, assessment and man-
agement. New York, Wiley.
HARLIN, J. and KUNG, C.-S. 1992. Parameter Uncertainty
in simulation of design of floods in Sweden. Journal of
Hydrology, No. 137, pp. 2009230.
HENDRICKSON, J.D.; SOROOSHIAN, S. and BRAZIL, L.E.
1988. Comparison of Newton-type and direct-search algo-
rithms for calibration of conceptual rainfallrunoff models.
Water Resources Research, Vol. 24, No. 5, pp. 691700.
HOLLING, C.S. 1978. Adaptive environmental assessment
and management. Chichester, UK, Wiley.
IBREKK, H. and MORGAN, M.G. 1987. Graphical
communication of uncertain quantities to nontechnical
people. Risk Analysis, Vol. 7, No. 4, pp. 51929.
ISAAKS, E.H. and SRIVASTAVA, R.M. 1989. An introduction
to applied geostatistics. New York, Oxford University Press.
JAFFE, P.R.; PANICONI, C. and WOOD, E.F. 1988.
Model calibration based on random environmental
fluctuations. Journal of Environmental Engineering, ASCE,
Vol. 114, No. 5, pp. 113645.
JAIN, S. and LALL, U. 2001. Floods in a changing
climate: does the past represent the future? Water
Resources Research, Vol. 37, No. 12, pp. 3193205 DEC.
JENSEN, F.V. 2001. Bayesian networks and decision graphs.
New York: Springer.
KANN, A. and WEYANT, J.P. 1999. A comparison of
approaches for performing uncertainty analysis in
integrated assessment models. Journal of Environmental
Management and Assessment, Vol. 5, No. 1, pp. 2946.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 288
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Model Sensitivity and Uncertainty Analysis 289
KELLY, E. and CAMPBELL, K. 2000. Separating variabil-
ity and uncertainty: making choices. Human and Ecological
Risk Assessment, An International Journal, Vol. 6, No. 1,
February, pp. 113.
KELLY, E. and ROY-HARRISON, W. 1998. A mathemat-
ical construct for ecological risk: a useful framework for
assessments. Human and Ecological Risk Assessment, An
International Journal, Vol. 4, No. 2, pp. 22941.
KELLY, E.J.; CAMPBELL, K. and HENRION, M. 1997. To
separate or not to separate that is the question: a
discourse on separating variability and uncertainty in
environmental risk assessments. Learned Discourses,
Society of Environmental Toxicology and Chemistry (SETAC)
News, November.
LAL, A.M.W. 2000. Numerical errors in groundwater and
overland flow models. Water Resources Research, Vol. 36,
No. 5, May, pp. 123748.
LAL, W.; OBEYSEKERA, J. and VAN ZEE, R. 1997.
Sensitivity and uncertainty analysis of a regional simula-
tion model for the natural system in South Florida.
In: Managing Water: Coping with Scarcity and Abundance.
Proceedings of the 27th Congress of International
Association for Hydraulic Research, San Francisco, Calif.,
August.
LEMONS, J. (ed.). 1996. Scientific uncertainty and environ-
mental problem solving. Cambridge, Mass. Blackwell.
LOPEZ, A. and LOUCKS, D.P. 1999. Uncertainty repre-
sentations in water quality prediction. Manuscript, Civil and
Environmental Engineering, Cornell University, Ithaca,
N.Y.
LOUCKS, D.P.; STEDINGER, J.R. and HAITH, D.A.
1981. Water resource systems planning and analysis.
Englewood Cliffs, N.J., Prentice-Hall.
LUDWIG, D.; HILBORN, R. and WALTERS, C. 1993.
Uncertainty, resource exploitation, and conservation:
lessons from history. Science, Vol. 260, pp. 1736.
MAJONI, H. and QUADE, E.S. 1980. Pitfalls of analysis.
New York, Wiley.
MCCARTHY, J.J.; CANZIANI, O.F.; LEARY, N.A.;
DOKKEN, D.J. and WHITE, K.S. (eds.). 2001. Climate
change 2001: impacts, adaptation, and vulnerability.
Contribution of Working Group II to the Third
Assessment Report of the Intergovernmental Panel on
Climate Change. Intergovernmental Panel on Climate
Change, http://www.ipcc.ch/ pub/tar/wg2/index.htm
MCCUEN, R.H. and SNYDER, W.M. 1983. Hydrological
Modelling: statistical methods and applications. Englewood
Cliffs, N.J., Prentice-Hall.
MEIXNER, T.; GUPTA, H.V.; BASTIDAS, L.A. and BALES,
R.C. 1999. Sensitivity analysis using mass flux and con-
centration. Hydrological Processes, Vol. 13, pp. 223344.
MISER, H.J. 1980. Operations research and systems
analysis. Science, No. 209, pp. 17482.
MORGAN, M.G. and HENRION, M. 1990. Uncertainty:
a guide to dealing with uncertainty in quantitative risk
and policy analysis. Cambridge, UK, Cambridge University
Press.
NEARING, M.A.; DEER-ASCOUGH, L. and LAFLEN, J.F.
1990. Sensitivity analysis of the WEPP hillslope profile
erosion model. Transactions of the American Society of
Agricultural Engineers, Vol. 33, No. 3, pp. 83949.
NRC (National Research Council). 1996. Committee on
risk characterization, understanding risk: informing decision
in a democratic society. In: P.S. Stern and H.V. Fineberg
(eds.), Washington, D.C., National Academy Press.
PHILLIPS, D.L. and MARKS, D.G. 1996. Spatial uncer-
tainty analysis: propagation of interpolation errors in
spatially distributed models. Ecological Modelling, No. 91,
pp. 21329.
PRESS, S.J. and TANUR, J.M. 2001. The subjectivity of
scientists and the Bayesian approach. New York: Wiley.
RECKHOW, K.H. 1994. Water quality simulation
modelling and uncertainty analysis for risk assessment
and decision-making. Ecological Modelling, No. 72,
pp. 120.
RECKHOW, K.H. 1999. Water quality prediction and
probability network models. Canadian Journal of Fisheries
and Aquatic Sciences, No. 56, pp. 11508.
RECKHOW, K.H. 2002. Applications of water models: pre-
diction uncertainty and decision-making. Presentation at
Model Uncertainty Workshop, January 2002. West Palm
Beach, Fla., South Florida Water Management District.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 289
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
SALTELLI, A.; CHAN, K. and SCOTT, E.M. (eds.). 2000.
Sensitivity analysis. Chichester, UK, Wiley.
SCHWEPPE, F.C. 1973. Uncertain dynamic systems.
Englewood Cliffs, N.J., Prentice-Hall.
SHAPIRO, H.T. 1990. The willingness to risk failure.
Science, Vol. 250, No. 4981, p. 609.
SIMON, I.I.A. 1988. Prediction and prescription in system
modelling. 15th Anniversary of IIASA, International
Institute for Applied Systems Analysis. Austria, Laxenburg.
SKLAR, F.H. and HUNSAKER, C.T. 2001. The Use and
uncertainties of spatial data for landscape models: an
overview with examples from the Florida Everglades. In:
C.T. Hunsaker, M.F. Goodchild, M.A. Friedl and T.J. Chase
(eds.), Spatial uncertainty in ecology, implications for remote
sensing and GIS applications, Chapter 2. New York, Springer.
SOROOSHIAN, S.; DUAN, Q. and GUPTA, V.K. 1993.
Calibration of rainfallrunoff models: application of
global optimization to the Sacramento soil moisture
accounting model. Water Resources Research, Vol. 29,
No. 4, pp. 118594.
SOROOSHIAN, S.; GUPTA, V.K. and FULTON, J.L. 1983.
Evaluation of maximum likelihood parameter estimation
techniques for conceptual rainfallrunoff models: influence
of calibration data variability and length on model credi-
bility. Water Resources Research, Vol. 19, No. 1, pp. 2519.
SOUTTER, M. and MUSY, A. 1999. Global sensitivity
analyses of three pesticide leaching models using a
Monte-Carlo approach. Journal of Environmental Quality,
No. 28, pp. 12907.
290 Water Resources Systems Planning and Management
SPEAR, R.C. and HORNBERGER, G.M. 1980. Eutrophica-
tion in Peel Inlet II. Identification of critical uncertainties
via generalized sensitivity analysis. Water Research, No. 14,
pp. 439.
STOKEY, E. and ZECKHAUSER, R. 1977. A primer for
policy analysis. New York, W.W. Norton.
SUTER, G.W. II. 1993. Ecological risk assessment. Boca
Raton, Fla., Lewis.
TATTARI, S. and BARLUND, I. 2001. The concept of
sensitivity in sediment yield modelling. Physics and
Chemistry of the Earth, Vol. 26, No. 1, pp. 2731.
VAN GRIENSVEN, A.; FRANCOS, A. and BAUWENS, W.
2002. Sensitivity analysis and auto-calibration of an inte-
gral dynamic model for river water quality. Water Science
and Technology, Vol. 45, No. 9, pp. 325332.
VAN HARN ADAMS, B.A. 1998. Parameter distributions
for uncertainty propagation in water quality modelling.
Doctoral Dissertation, Duke University. Durham, N.C.
VAN STRATEN, G. 1983. Maximum likelihood estima-
tion of parameters and uncertainty in phytoplankton
models. In: M.B. Beck and G. van Straten (eds.),
Uncertainty and forecasting of water quality, New York,
Springer-Verlag, pp. 15771.
VON WINTERFELDT, D. and EDWARDS, W. 1986.
Decision analysis and behavioural research. Cambridge, UK,
Cambridge University Press.
WARWICK, J.J. and CALE, W.G. 1987. Determining like-
lihood of obtaining a reliable model. Journal of Environmental
Engineering, ASCE, Vol. 113, No. 5, pp. 110219.
wrm_ch09.qxd 8/31/2005 11:56 AM Page 290
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
10. Performance Criteria
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
wrm_ch10.qxd 8/31/2005 11:57 AM Page 292
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Developing and managing water resources systems
involves making decisions. Decisions are made at various
levels of planning and management, even by those who
are simply recommending courses of action to higher
levels of an organization. The ability to make informed
decisions based on scientific as well as social or political
criteria is fundamental to the success of any water
resources planning and management organization.
Modelling and data management tools can con-
tribute to the information needed to make informed
decisions. These tools are quantitative, but they are
based on qualitative judgements about what informa-
tion is and is not important to include or consider. Even
the most objective or rational of decisions rests on
subjective choices that are influenced by what decision-
makers think is important, the objectives or criteria
that are to be achieved and for whom, and how much
they will care. One of the benefits of modelling is that
differences in assumptions, in judgements over what to
consider and what not to consider, or about just how
accurate certain information must be, can be evaluated
with respect to their influence on the decisions to be
made.
293
Performance Criteria
Water resources systems typically provide a variety of economic, environmental
and ecological services. They also serve a variety of purposes such as water
supply, flood protection, hydropower production, navigation, recreation, waste
reduction and transport. Performance criteria provide measures of just how well a
plan or management policy performs. There are a variety of criteria one can use to
judge and compare system performance. Some of these performance criteria may
conflict with one another. In these cases tradeoffs exist among conflicting criteria
and these tradeoffs must be considered when searching for the best compromise
decisions. This chapter presents ways of identifying and working with these
tradeoffs in the political process of selecting the best decision.
10
Decision-makers and those who influence them are
people, and peoples opinions and experiences and goals
differ. These differences force one to think in terms of
tradeoffs. Decisions in water resources management
inevitably involve making tradeoffs compromising
among competing opportunities, goals or objectives. One
of the tasks of water resources system planners or managers
involved in evaluating alternative designs and management
plans or policies is to identify the tradeoffs, if any, among
competing opportunities, goals or objectives. It is then up
to a largely political process involving all interested stake-
holders to find the best compromise decision.
Measures indicating just how well different management
plans and policies serve the interests of all stakeholders are
typically called system performance criteria. There are many
of these, some that can be quantified and some that cannot.
Lack of quantification does not make those criteria less
important than those that can be quantified. Those making
decisions consider all criteria, qualitative as well as quanti-
tative, as discussed in Chapter 2 among others.
If every system performance measure or objective
could be expressed in the same units, and if there were
only one decision-maker, then decision-making would
be relatively straight forward. Such is not the case when
dealing with the publics water resources.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 293
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The costbenefit framework, used for many decades in
water resources planning and management, converted the
different types of impacts into a single monetary metric.
Once that was done, the task was to find the plan or
policy that maximized the difference between the
benefits and costs. If the maximum difference between
benefits and costs was positive, then that was the best
plan or policy. But not all system performance criteria can
be easily expressed in monetary units. Even if monetary
units could be used for each objective, that in itself does
not address the distributional issues of who benefits and
who pays, and by how much. While all stakeholders may
agree that maximizing total net benefits is a desirable
objective, not everyone, if indeed anyone, will be likely to
agree on how best to distribute those net benefits.
Clearly, water resources planning and management
takes place in a multi-criteria environment. A key element
of most problems facing designers and managers is the
need to deal explicitly with multiple ecological, economic
and social impacts expressed in multiple metrics that may
result from management actions. Approaches that fail to
recognize and explicitly include ways of handling conflict
among multiple system performance measures and objec-
tives and among multiple stakeholders are not likely to be
very useful.
Successful decision-making involves creating a
consensus among multiple participants in the planning
and management process. These include stakeholders
individuals or interest groups who have an interest in the
outcome of any management plan or policy. The
relatively recent acknowledgement that stakeholders need
to be fully included in the decision-making processes
complicates the life of professional planners and man-
agers. However, important sources of information come
from discussion groups, public hearings, negotiations and
dispute-resolution processes. It is during the debates in
these meetings that information derived from the types of
modelling methods discussed in this book should be
presented and considered.
Eventually someone or some organization must make
a decision. Even if water resources professionals view
their job as one of providing technically competent
options or tradeoffs for someone else to decide upon, they
are still making decisions that define or limit the range of
those options or tradeoffs. The importance of making
informed, effective decisions applies to everyone.
294 Water Resources Systems Planning and Management
2. Informed Decision-Making
Informed decision-making involves both qualitative
thinking and analysis and quantitative modelling
(Hammond et al., 1999). Qualitative thinking and analysis
follows quantitative modelling. Qualitative analysis prior
to quantitative modelling is useful for identifying:
the real objectives of concern to all stakeholders
(which may be other than those expressed by them)
the likelihoods of events for which decisions are needed
the alternatives that address and meet each objective
the key tradeoffs among all interests and objectives
involved.
When the qualitative aspects of determining the objectives
and general alternative ways of meeting these objectives
are weak, no amount of quantitative modelling and
analysis will make up for this weakness. The identification
of objectives and general alternatives are the foundation
upon which water resources managers can develop
appropriate quantitative models. These models will pro-
vide additional insight and definition of alternatives and
their expected impacts. Models cannot identify new ideas
or so called out-of-the-box alternatives. Neither can they
identify the best criteria that should be considered in
specific cases. Only our minds can do this individually,
and then collectively.
For example, periodic municipal water supply
shortages in a river basin might be reduced by building
additional surface water reservoir storage capacity or by
increased groundwater pumping. Assuming the objective
in this case is to increase the reliability of some specified
level of water supply, models can be developed to identify
the tradeoffs between reservoir and pumping capacity
(and cost) and increased reliability in meeting the supply
target. However, these models will not consider and
evaluate completely different alternatives, such as import-
ing water by trucks, implementing water use restrictions,
or water reuse, in times of drought, unless of course those
options are included in the models. Someone has to
think of these general alternatives before models can
help identify the details: just how many trucks of what
capacity, or how best to implement water reuse and for
what uses.
Time needs to be taken to identify the relevant objectives
and general alternatives. Clearly some preliminary screening
wrm_ch10.qxd 8/31/2005 11:57 AM Page 294
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 295
of general alternatives should be carried out at the
qualitative level, but the alternatives that remain should
then be further analysed by means of quantitative modelling
methods.
Creativity in identifying possible objectives and
general alternatives is helped by addressing the following
questions:
What is an ideal decision?
What does each stakeholder think other stakeholders
ideal decisions would be?
What is to be avoided?
What makes a great alternative, even an infeasible one,
so great?
What makes a terrible alternative so terrible?
How would each individuals best alternative be justi-
fied to someone else?
When each manager and stakeholder has gone through
such thinking, the combined set of responses may be
more comprehensive and less limited by what others say
or believe. They can become a basis for group discussions
and consensus building.
General statements defining objectives can be con-
verted to ones that are short and include the words
maximize or minimize, e.g. minimize cost, maximize net
benefits, maximize reliability, maximize water quality,
maximize ecosystem biodiversity, minimize construction
time, or minimize the maximum deviations from some
target storage volume or target water allocation.
Economic, environmental, ecological and purely physical
objectives such as these can become the objective
functions that drive the solutions of optimization models,
as illustrated in many of the previous chapters. Social
objectives should also be considered. Examples might
include: maximize employment, maximize interagency
coordination, maximize stakeholder participation, and
minimize legal liability and the potential of future legal
action and costs.
Quantitative modelling is employed to identify more
precisely the design and operation of structural and
non-structural alternatives that best satisfy management
criteria, and to identify the impacts and tradeoffs among
various performance criteria. Once such analyses have
been performed, it is always wise to question whether or
not the results are reasonable. Are they as expected? If
not, why? If the results are surprising, are the analyses
providing new knowledge and understanding, or have
errors been made? How sensitive are the results to various
assumptions with respect to the input data and models
themselves (Chapter 9)?
Thus, the modelling process ends with some more
qualitative study. Models do not replace human judge-
ment. Humans, not quantitative models, are responsible
for water resources planning and management decisions
as well as the decision-making process that identifies
performance criteria and general alternatives.
3. Performance Criteria and
General Alternatives
There is a way to identify performance criteria that matter
most to stakeholders (Gregory and Keeney, 1994). One
can begin with very broad fundamental goals, such as
public health, national as well as individual security,
economic development, happiness and general well-being.
Almost anyone would include these as worthy objectives.
Just how these goals are to be met can be expressed by a
host of other more specific objectives or criteria.
Asking how any specified broad fundamental crite-
rion or objective can be achieved usually leads to more
specific system performance criteria or objectives and to
the means of improving these criteria: in other words, to
the general alternatives themselves. As one gets further
from the fundamental objective that most if not all will
agree to, there is a greater chance of stakeholder disagree-
ment. For example, we might answer the question How
can we achieve maximum public health? by suggesting
the maximization of surface and groundwater water qual-
ity. How? By minimizing wastewater discharges into sur-
face water bodies and groundwater aquifers. How? By
minimizing wastewater production, or by maximizing
removal rates at wastewater treatment facilities, or by
minimizing the concentrations of pollutants in runoff or
by increasing flow augmentation or by a combination of
flow augmentation and wastewater treatment. How? By
increasing reservoir storage capacity and subsequent
releases upstream and by upgrading wastewater treatment
to a tertiary level. And so on.
Each how question can have multiple answers. This
can lead to a tree of branches, each branch representing a
different and more specific performance criterion or an
wrm_ch10.qxd 8/31/2005 11:57 AM Page 295
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
alternative way to accomplish a higher-level objective.
In the example just illustrated, the answer to how to
improve water quality might include a combination of
water and wastewater treatment, flow augmentation,
improved wastewater collection systems and reduced
applications of fertilizers and pesticides on land to reduce
non-point source pollutant discharges. There are others.
If the lower-level objective of minimizing point source
discharges had been the first objective considered, many
of those other alternatives would not be identified. The
more fundamental the objective, the greater will be the
range of alternatives that might be considered.
If any of the alternatives identified for meeting some
objective are not considered desirable, this is a good sign
that there are other objectives that should be consid-
ered. For example, if flow augmentation is not desired,
it could mean that in addition to water quality consider-
ations, the regime of water flows or the existing uses of
the water are also being considered, and flow augmenta-
tion may detract from those other objectives. If a stake-
holder has trouble explaining why some alternative will
not work, then it is a possible sign that there are other
objectives and alternatives waiting to be identified and
evaluated. What are they? They need to be identified.
Consider them along with the others that have been
already defined.
More fundamental objectives can be identified by ask-
ing the question why? The answers will lead one back to
increasingly more fundamental objectives. A fundamental
objective is reached when the only answer to why is: it is
what everyone really wants or something similar.
Thinking hard about why will help clarify what is
considered most important.
3.1. Constraints On Decisions
Constraints limit alternatives. Constraints are often real.
There are some laws of physics and society that will limit
what can be done to satisfy any performance criterion in
many cases, but it is not the time to worry about them
during the process of identifying objectives and general
alternatives. Be creative, and do not linger on the status
quo, carrying on as usual or depending on a default (and
often politically risk-free) alternative. It may turn out these
default or risk-free alternatives are the best available, but
that can be determined later.
296 Water Resources Systems Planning and Management
When performing qualitative exercises to identify
objectives and general alternatives, enlarge the envelope
and think creatively. If water needs to flow uphill, assume
it can happen. If it is really worth it, engineers can do it.
Lawyers can change laws. Society can and will want to
adapt if it is really important. Consider for example the
water that is pumped uphill at all pumped storage hydro-
electric facilities, or the changing objectives over time
related to managing water and ecosystems in the Mekong,
Rhine and Senegal Rivers, or in the North American Great
Lakes and Florida Everglades regions.
There are other traps to avoid in the planning and man-
agement process. Do not become anchored to any initial
feasible or best-case or worst-case scenario. Be creative also
when identifying scenarios. Consider the whole spectrum
of possible events. Do not focus solely on extreme events
to the exclusion of the more likely events. While toxic
spills and floods bring headlines and suffering, they are
not the usual more routine events that one must plan for.
It is also tempting to consider past or sunk costs
when determining where additional investments should
be made. If past investments were a mistake, it is only our
egos that motivate us to justify those past investments by
spending more money on them instead of taking more
effective action.
Finally, objective targets should not be set too low.
The chances of finding good unconventional alternatives
is increased if targets are set high, even beyond reach.
High aspirations often force individuals to think in
entirely new ways. Politically it may end up that only
marginal changes to the status quo will be acceptable, but
it is not the time to worry about that when identifying
objectives and general alternatives.
3.2. Tradeoffs
Tradeoffs are inevitable when there are multiple objec-
tives, and multiple objectives are inevitable when there
are many stakeholders or participants in the planning and
management process. We all want clean water in our
aquifers, streams, rivers and lakes but, if it costs money
to achieve that desired quality, there are other activities
or projects involving education, health care, security,
or even other environmental restoration or pollution
prevention and reduction efforts, which compete for
those same, often limited, amounts of money.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 296
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 297
Tradeoffs arise because we all want more of good
things, and many of these good things are incompatible.
We cannot have cleaner water and at the same time
reduce our costs of wastewater treatment. Identifying
these tradeoffs is one of the tasks of qualitative as well as
quantitative analyses. Qualitatively we can identify what
the tradeoffs will involve. For example, cleaner streams
normally require increased costs. Just how many dinars,
dollars, euros, pounds, roubles or yen it will cost to
increase the minimum dissolved oxygen concentration in
a specific lake by 3, or 4, or 5 mg/l is the task of quanti-
tative analyses.
Finding a balance among all conflicting performance
criteria characterizes water management decision-making.
Understanding the technical information that identifies the
efficient or non-dominated tradeoffs is obviously essential.
Yet if the technical information fails to address the real
objectives or system performance measures of interest to
stakeholders, significant time and resources can be wasted
in the discussions that take place in stakeholder meetings,
as well as in the analyses that are performed in technical
studies.
Finding the best compromise among competing
decisions is a political or social process. The process is
helped by having available the tradeoffs among compet-
ing objectives. Models can identify these tradeoffs among
quantitative objectives but cannot go the next step, i.e.
identify the best compromise decision. Models can help
identify and evaluate alternatives, but they cannot take
the place of human judgements that are needed to make
the final decision.
Models can help identify non-dominated tradeoffs
among competing objectives or system performance
indicators. These are sometimes called efficient tradeoffs.
Efficient decisions are those that cannot be altered with-
out worsening one or more objective values. If one of
multiple conflicting objective values is to be improved,
some worsening of one or more other objective values will
likely be necessary. Dominated or so-called inferior deci-
sions are those in which it is possible to improve at least
one objective value without worsening any of the others.
Many will argue that these dominated decisions can
be eliminated from further consideration: why would
any rational individual prefer such decisions when
there are better ones available? However, quantitatively
non-dominated or efficient solutions can be considered
inferior and dominated with respect to one or more
qualitative objectives, or even other quantitative objec-
tives that were omitted from the analysis. Eliminating
inferior or dominated alternatives, as defined by a set of
performance criteria, from the political decision-making
process may not always be very helpful. One of them
could be viewed as the best by people who are consider-
ing other criteria that are either not included or, in their
opinions, not properly quantified in such analyses.
Tradeoffs must not only be identified and made among
conflicting outcome objectives, system performance indi-
cators, or impacts that help define what should be done.
Tradeoffs are also necessary among alternative processes
of decision-making. Some of the most important objec-
tives and toughest tradeoffs involve process decisions
that establish how a decision is going to be made.
What is, for example, the best use of time and financial
resources in performing a quantitative modelling study,
who should be involved in such a study, who should be
advising such a study, and how should stakeholders be
involved? Such questions often lie at the heart of water
and other resourceuse disputes and can significantly
influence the trust and cooperation among all who partic-
ipate in the process. They can also influence the willing-
ness of stakeholders to support any final decision or
selected management policy or plan.
4. Quantifying Performance
Criteria
So far this chapter has focused on the critical qualitative
aspects of identifying objectives and general alternatives.
The remainder of the chapter will focus on quantifying
various criteria and ways to use them to compare various
alternatives and identify the tradeoffs among them. What
is important here, however, is to realize that considerable
effort is worth spending on getting the general objectives
and alternatives right before spending any time on their
quantification. There are no quantitative aids for this, just
hard thinking, perhaps keeping in mind some of the
advice just presented.
Quantification of an objective is the adoption of some
quantitative (numerical) scale that provides an indicator
for how well the objective would be achieved. For example,
one of the objectives of a watershed conservation program
wrm_ch10.qxd 8/31/2005 11:57 AM Page 297
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
might be protection or preservation of wildlife. In order to
rank how various plans meet this objective, a numerical
criterion is needed, such as acres of preserved habitat or
populations of key wildlife species.
Quantification does not require that all objectives be
described in comparable units. The same project could
have a flood control objective quantified as the reduced
probability of flooding or the height of the protected
flood stage. It could have a regional development objec-
tive quantified as increased local income. Quantification
does not require that monetary costs and benefits be
assigned to all objectives.
In the following sub-sections various economic, envi-
ronmental, ecological and social criteria are discussed.
4.1. Economic Criteria
Water resources system development and management is
often motivated by economic criteria. Most of the devel-
opment of water resources systems involving reservoirs,
canals, hydropower facilities, groundwater-pumping
systems, locks and levees has been for economic reasons.
The benefits and costs of these systems can be expressed
in monetary units. The criteria have been either to
maximize the present value of the net benefits (total
benefits less the costs) or minimize the costs of providing
some purpose or service. To achieve the former involves
benefitcost analyses and to achieve the latter involves
cost-effectiveness analyses.
The rationale for benefitcost analysis is based on
two economic concepts: scarcity and substitution. The
former implies that the supplies of natural, synthetic
and human resources are limited. Hence people are
willing to pay for them. They should therefore be used
in a way that generates the greatest return, i.e. they
should be used efficiently. The concept of substitution,
on the other hand, indicates that individuals, social
groups and institutions are generally willing to trade
a certain amount of one objective value for more of
another.
Applied to water resources, maximization of net-
benefits requires the efficient and reliable allocation, over
both time and space, of water (in its two dimensions:
quantity and quality) to its many uses, including
hydropower, recreation, water supply, flood control,
navigation, irrigation, cooling, waste disposal and
298 Water Resources Systems Planning and Management
assimilation, and habitat enhancement. The following
example illustrates the maximization of net benefits from
a multiple-purpose reservoir.
Consider a reservoir that can be used for irrigation and
recreation. These two uses are not easily compatible.
Recreation benefits are greater when reservoir elevations
remain high throughout the summer recreation season,
just when the irrigation demand would normally cause a
drop in the reservoir storage level. Thus, the project has
two conflicting purposes: provision of irrigation water
and improvement of recreation opportunities.
Some basic principles of benefitcost analysis can be
illustrated with this example. Let X be the quantity of
irrigation water to be delivered to farmers each year and
Y the number of visitor-days of recreation use on the
reservoir. Possible levels of irrigation and recreation are
shown in Figure 10.1. The solid red line in Figure 10.1,
termed the production-possibility frontier or efficiency
frontier, is the boundary of the feasible combinations of
X and Y.
Any combination of X and Y within the shaded blue
area can be obtained by operation of the reservoir (that is,
regulating the amount of water released for irrigation
and other uses). Obviously, the more of both X and Y the
better. Thus attention generally focuses on the production-
possibility frontier which comprises those combinations
of X and Y that are technologically efficient in the sense
that more of either X or Y cannot be obtained without a
r
e
c
r
e
a
t
i
o
n
Y
irrigation X
efficient combinations of
X and Y on the boundary or
production possibility frontier
E
0
2
0
1
1
5
e
feasible combinations
of X and Y
Figure 10.1. A two-purpose planning problem showing the
feasible and efficient combinations of irrigation water X
and recreation visitor days Y.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 298
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 299
decrease in the other. The shape and location of the
production-possibility frontier is determined by the
quantity of available resources (water, reservoir storage
capacity, recreation facilities, etc.) and their ability to
generate both X and Y.
To continue the example, assume that a private entre-
preneur owns this two-purpose reservoir in a competitive
environment, i.e. there are a number of competing irriga-
tion water suppliers and reservoir recreation sites. Let the
unit market prices for irrigation water and recreation
opportunities be p
X
and p
Y
, respectively. Also, assume that
the entrepreneurs costs are fixed and independent of X
and Y. In this case the total income is
I p
X
X p
Y
Y (10.1)
Values of X and Y that result in fixed income levels I
1
I
2
I
3
are plotted in Figure 10.2. The value of X and Y that
maximizes the entrepreneurs income is indicated by the
points on the production-possibility frontier yielding an
income of I
3
. Incomes greater than I
3
are not possible.
Now assume the reservoir is owned and operated by a
public agency and that competitive conditions prevail.
The prices p
X
and p
Y
reflect the value of the irrigation
water and recreation opportunities to the users. The
aggregated value of the project is indicated by the users
willingness to pay for the irrigation and recreation outputs.
In this case, this willingness to pay is p
X
X p
Y
Y, which is
equivalent to the entrepreneurs income. Private opera-
tion of the reservoir to maximize income or government
operation to maximize user benefits should both, under
competitive conditions, produce the same result.
When applied to water resources planning, benefit
cost analysis assumes a similarity between decision-
making in the private and public sectors. It also assumes
that the income resulting from the project is a reasonable
surrogate for the projects social value.
4.1.1. Benefit and Cost Estimation
In benefitcost analysis, one may need to estimate the
monetary value of irrigation water, shoreline property,
land inundated by a lake, lake recreation, fishing oppor-
tunities, scenic vistas, hydropower production, navigation
or the loss of a wild river. The situations in which bene-
fits and costs may need to be estimated are sometimes
grouped into four categories, reflecting the way prices can
be determined. These situations are:
1. Market prices exist and are an accurate reflection of
marginal social values, i.e. marginal willingness to pay
for all individuals. This situation often occurs in the
presence of competitive market conditions. An exam-
ple would be agricultural commodities that are not
subsidized, that is, they do not have supported prices
(some do exist!).
2. Market prices exist but for various reasons do not
reflect marginal social values. Examples include
price-supported agricultural crops, labour that would
otherwise be unemployed, or inputs whose produc-
tion generates pollution, the economic and social cost
of which is not included in its price.
3. Market prices are essentially non-existent, but it is
possible to infer or determine what users or consumers
would pay if a market existed. An example is outdoor
recreation.
4. No real or simulated market-like process can easily be
conceived. This category may be relatively rare.
Although scenic amenities and historic sites are often
considered appropriate examples, both are sometimes
privately owned and managed to generate income.
For the first three categories, benefits and costs can be
measured as the aggregate net willingness to pay of those
affected by the project. Assume, for example, that alterna-
tive water resources projects X
1
, X
2
, are being
considered. Let B(X
j
) equal the amount beneficiaries of
r
e
c
r
e
a
t
i
o
n

i
n
c
o
m
e

Y
irrigation income X
E
0
2
0
1
1
5
f
3
2
1
maximum total
income allocation
I
I
I
Figure 10.2. Two-purpose project involving tradeoffs between
irrigation and recreation income and various constant total
income levels, I
1
I
2
I
3
.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 299
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
the plan X
j
are willing to pay rather than forego the
project. This represents the aggregate value of the project
to the beneficiaries. Let D(X
j
) equal the amount the non-
beneficiaries of plan X
j
are willing to pay to prevent it
from being implemented. This includes the social value of
the resources that will be unavailable to society if project
or plan X
j
is implemented. The aggregate net willingness
to pay W(X
j
) for plan X
j
is equal to the difference between
B(X
j
) and D(X
j
),
W(X
j
) B(X
j
) D(X
j
) (10.2)
Plans X
j
can be ranked according to the aggregate net
willingness to pay, W(X
j
). If, for example, W(X
j
) W(X
k
),
it is inferred that plan X
j
is preferable or superior to plan X
k
.
One rationale for the willingness-to-pay criterion is
that if B(X
j
) D(X
j
), the beneficiaries could compensate
the losers and everyone would benefit from the project.
However, this compensation rarely happens. There is
usually no mechanism established for such compensation
to be paid. While many resources required for the project,
such as land, labour and machinery, will be obtained and
paid for, those who lose favourite scenic sites or the
opportunity to use a wild river, or who must hear the
noise, breathe dirtier air or suffer a loss in the value of their
property because of the project are seldom compensated.
This compensation criterion also ignores the resultant
income redistribution, which should be considered
during the plan selection process. The compensation
criterion implies that the marginal social value of income
to all affected parties is the same. If a projects benefits
accrue primarily to affluent individuals and the costs are
borne by lower-income groups, B(X) may be larger than
D(X) simply because the beneficiaries can pay more than
the non-beneficiaries. It matters who benefits and who
pays, that is, who gets to eat the pie and how much of it,
as well as how big the pie is. Traditional benefitcost
analyses typically ignore this distribution issue.
In addition to these and other conceptual difficulties
related to the willingness-to-pay criterion, practical
measurement problems also exist. Many of the products of
water resources plans are public or collective goods. This
means that they are essentially indivisible, and once
provided to any one individual it is very difficult not to
provide them to others. Collective goods often also have
the property that consumption by one person does not
prohibit or infringe upon its consumption or use by others.
300 Water Resources Systems Planning and Management
Community flood protection is an example of a public
good. Once protection is provided for one individual, it is
simultaneously provided for many others. As a result, it is
not in an individuals self-interest to volunteer to help
pay for the project by contributing an amount equal to
his or her actual benefits if others are willing to pay for
the project. However, if others are going to pay for the
project, then individuals may exaggerate their own bene-
fits to ensure that the project is undertaken.
Determining what benefits should be attributed to a
project is not always simple and the required accounting
can become rather involved. In a benefitcost analysis,
economic conditions should first be projected for a base
case in which no project is implemented, and the benefits
and costs are estimated for that scenario. Then the benefits
and costs for each project are measured as the incremental
economic impacts that occur in the economy over these
baseline conditions. The appropriate method for benefit
and cost estimation depends on whether or not the market
prices reflecting true social values are available, or whether
such prices can be constructed.
Market Prices Equal Social Values
Consider the estimation of irrigation benefits in the
irrigationrecreation example discussed in the previous
section. Let X be the quantity of irrigation water sup-
plied by the project each year. If the prevailing market
price p
X
reflects the marginal social value of water, and
if that price is not affected by the projects operation, the
value of the water is just p
X
X. However, it often happens
that large water projects have a major impact on the
prices of the commodities they supply. In such cases,
the value of water from our example irrigation district
would not be based on prices before or after project
implementation. Rather, the total value of the water X to
the users is the total amount they would be willing to
pay for it.
Let Q(p) be the amount of water the consumers would
want to buy at a unit price p. For any price p, consumers
will continue to buy the water until the value of another
unit of water is less than or equal to the price p. The func-
tion Q(p) defines what is called the demand function. As
illustrated by Figure 10.3, the lower the unit price, the
more water individuals are willing to buy, and the greater
the demand will be.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 300
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 301
The willingness to pay a given unit price p is defined
by the area under the demand curve. As Figure 10.3
suggests, there are some who would be willing to pay a
higher price for a given amount of water than others. As
the unit price decreases, more individuals are willing to
buy more water. The total willingness to pay for a given
amount of water q is the area under the demand curve
from Q 0 to Q q:
willingness to pay for q Q
1
(q)dq. (10.3)
Consumers willingness to pay for a product is an impor-
tant concept in welfare economics.
Market Prices Not Equal to Social Values
Market prices frequently do not truly reflect the social value
of the various inputs of goods and services supplied by a
water resources project. In such cases it is necessary for the
planner to estimate the appropriate values of the quantities
in question. There are several procedures that can be used,
depending on the situation. A rather simple technique that
can reach absurd conclusions if incorrectly applied is to
equate the benefits of a service to the cost of supplying the
service by the least expensive alternative method. Thus the
benefits from hydroelectricity generation could be esti-
mated as the cost of generating that electricity by the least-
cost alternative method using solar, wind, geothermal,
coal-fired, natural gas or nuclear energy sources. Clearly,
this approach to benefit estimation is only valid if, were the
project not adopted, the service in question would in fact
0
q

be demanded at, and supplied by, the least-cost alternative

method. The pitfalls associated with this method of benefit
and cost estimation can be avoided if one clearly identifies
reasonable with- and without project scenarios.
In other situations, simulated or imagined markets can
be used to derive the demand function for a good or service
and to estimate the value of the amount of that good or
service generated or consumed by the project. The following
hypothetical example illustrates how this technique can be
used to estimate the value of outdoor recreation.
Assume a unique recreation area is to be developed,
which will serve two population centres. Centre A has a
population of 10,000, and the more distant Centre B has a
population or 30,000. From questionnaires it is estimated
that if access to the recreation area is free, then 20,000 visits
per year will be made from Centre A at an average round-
trip travel cost of $1. Similarly 30,000 visits per year will
come from Centre B, at an average round-trip cost of $2.
The benefits derived from the proposed recreation area
can be estimated from an imputed demand curve. First, as
illustrated in Figure 10.4, a graph of travel cost as a
function of the average number of visits per capita can be
constructed. Two points are available: an average of two
visits per capita (from Centre A) at a cost of $1/visit, and
an average of one visit per capita (from Centre B) at a cost
or $2 visit. These travel cost data are extrapolated to the
ordinate and abscissa.
Even assuming that there are no plans to charge
admission at the site, if users respond to an entrance fee
as they respond to travel costs, then it is possible to
u
n
i
t

p
r
i
c
e
p
E
0
2
0
1
1
5
g
Q
-1
( ) q =
q
willingness to pay for q*
quantity of water demanded q *
Figure 10.3. Demand function defining how much water Q will
be purchased for a specified unit price p. The shaded area
represents the willingness to pay for a quantity q.
E
0
2
0
1
1
5
h
t
o
t
a
l

c
o
s
t

$
/
v
i
s
i
t
visits/capita/year
3
2
0
1
0 1 2 3
Figure 10.4. Estimated relationship between travel cost and
visits per capita.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 301
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
estimate what the user response might be if an entrance
fee were to be charged. This information will provide a
demand curve for recreation at the site.
Consider first a $1 admission price to be added to the
travel cost for recreation. The total cost to users from
population Centre A would then be $2 per visit. From
Figure 10.4, at $2 per visit, one visit per capita is made;
hence, 10,000 visits per year can be expected from
Centre A. The resulting cost to users from Centre B is $3
per visit, and hence, from Figure 10.4 no visits would
be expected. Therefore, one point on the demand curve
(10,000 visits at $1) is obtained. Similarly, it can be
inferred that at a $2 admission price there will be no
visits from either centre. A final point, corresponding
to a zero admission price (no added costs) is just the
expected site attendance (20,000 30,000 50,000 visits).
The resulting demand curve is shown in Figure 10.5.
Recreation total willingness-to-pay benefits are equal to
the area under the demand curve, or $35,000. Assuming
no entrance fee is charged, this amounts to $0.70 per
visitor-day based on the expected 50,000 visits.
Obviously, this example is only illustrative. The cost of
travel time (which differs for each population centre) and
the availability of alternative sites must be included in a
more detailed analysis.
No Market Processes
In the absence of any market-like process (real or simu-
lated), it is difficult to associate specific monetary values
with benefits. The benefits associated with aesthetics and
302 Water Resources Systems Planning and Management
with many aspects of environmental and ecosystem
quality have long been considered difficult to quantify
in monetary terms. Although attempts (some by highly
respected economists) have been made to express
environmental benefits in monetary terms, the results
have had limited success. In most regions in the world,
water resources management guidelines, where they exist,
do not encourage the assignment of monetary values to
these criteria. Rather the approach is to establish environ-
mental and ecological requirements or regulations. These
constraints are to be met, perhaps while maximizing
other economic benefits. Legislative and administrative
processes rarely, if ever, determine the explicit benefits
derived from meeting these environmental and ecological
requirements or regulations.
To a certain extent, environmental (including aesthetic)
objectives, if quantified, can be incorporated into a multi-
objective decision-making process. However, this falls
short of the assignment of monetary benefits that is often
possible for the first three categories of benefits.
4.1.2. A Note Concerning Costs
To be consistent in the estimation of net benefits from
water resources projects, cost estimates should reflect
opportunity costs, the value of resources in their most
productive alternative uses. This principle is much easier
stated than implemented, and as a result true opportunity
costs are seldom included in a benefitcost analysis. For
example, if land must be purchased for a flood-control
project, is the purchase price (which would typically be
used in a benefitcost analysis) the lands opportunity
cost? Suppose that the land is currently a natural area
and its alternative use is as a nature preserve and camping
area. The lands purchase price may be low, but this price
is unlikely to reflect the lands true value to society.
Furthermore, assume that individuals who would
otherwise be unemployed are hired to work on the land.
The opportunity cost for such labour is the marginal value
of leisure forgone, since there is no alternative productive
use of those individuals. Yet the labourers must be paid
and their wages must be included in the budget for
the project.
The results of rigorous benefitcost analyses seldom
dictate which of competing water resources projects and
plans should be implemented. This is in part because of
E
0
2
0
1
1
5
j
t
o
t
a
l

c
o
s
t

$
/
v
i
s
i
t
recreation demand (1000 visits/year)
2
0
1
0 10 30 60 40 20 50
Figure 10.5. Demand function for recreation facility.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 302
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 303
the multi-objective nature of the decisions. One must
consider environmental impacts, income redistribution
effects and a host of other local, regional and national
goals, many of which may be non-quantifiable.
Other important considerations are the financial,
technical and political feasibilities of alternative plans.
Particularly important when a plan is undertaken by
government agencies is the relative political and legal
clout of those who support the plan and those who
oppose it. Still, a plans economic efficiency is an impor-
tant measure of its value to society and often serves as an
indicator of whether it should be considered at all.
4.1.3. Long and Short-Run Benefit Functions
When planning the capacities and target values associated
with water resources development projects, it is often
convenient to think of two types of benefit functions:
what economists call long-run benefits and short-run
benefits. In long-run planning, the capacities of proposed
facilities and the target allocations of flows or storage vol-
umes to alternative uses are unknown decision-variables.
The values of these variables are to be determined in a
way that achieves the most beneficial use of available
resources, even when the available resources vary in
magnitude and, hence, do not always equal the targets
over time. In short-run planning, the capacity of facilities
and the allocation targets of water promised to alternative
uses are specified, i.e. they are known. The problem is
one of managing or operating a given or proposed system
under varying supply conditions.
For example, if a water-using firm is interested in
building a factory along a river, of interest to those design-
ing the factory is the amount of water the factory can
expect to get. This in part may dictate its capacity, the
number of employees hired and so on. On the other
hand, if the factory already exists, the likely issue is how
to manage or operate the factory when the water supply
varies from the target levels that were (and perhaps still
are) expected.
Long-run benefits are those benefits obtained if all
target allocations are met. Short-run benefits are the
benefits one actually obtains by operating a system to best
use the available resources after capacities and target
values have been fixed by prior decisions. If the resources
available in the short-run are those that can meet all the
targets that were established when long-run decisions
were made, then estimated long-run benefits can be
achieved. Otherwise the benefits actually obtained may
differ from those expected. The goal is to determine the
values of the long-run decision-variables in a way that
maximizes the present value of all the short-run benefits
obtained given the varying water supply conditions.
The distinction between long-run and short-run
benefits can be illustrated by considering again a potential
water user at a particular site near a water body. Assume
that the long-run net benefits associated with various target
allocations of water to that use can be estimated. These
long-run net benefits are those that will be obtained if the
actual allocation Q equals the target allocation Q
T
. This
long-run net benefit function can be denoted as B(Q
T
).
Next assume that for various fixed values of the target Q
T
the actual net benefits derived from various allocations Q
can be estimated. These short-run benefit functions b(Q|Q
T
)
of allocations Q given a target allocation Q
T
are dependent
on the target Q
T
and, obviously, on the actual allocation Q.
The relationship between the long-run net benefits B(Q
T
)
and the short-run net benefits b(Q|Q
T
) for a particular target
Q
T
is illustrated in Figure 10.6.
The long-run function B(Q
T
) in Figure 10.6 reflects
the benefits that users receive when they have adjusted
their plans in anticipation of receiving an allocation equal
to the target Q
T
and actually receive it. The short-run
benefits function specifies the benefits users actually
b
e
n
e
f
i
t
allocation Q
E
0
2
0
1
1
5
k
Q
1
Q
T
Q
2
benefit function
long-run B Q (
( | b Q Q
T
)
T
) short-run
T
L(Q |Q
1
)
Figure 10.6. Long-run and short-run benefit functions,
together with the loss, L(Q
1
|Q
T
), associated with a deficit
allocation Q
1
and target allocation Q
T
.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 303
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
receive when a particular allocation is less (e.g. Q
1
) or
more (e.g. Q
2
) than the anticipated allocation, Q
T
, and
they cannot completely adjust their plans to the resulting
deficit or surplus.
Clearly the short-run benefits associated with any
allocation cannot be greater than the long-run benefits
obtainable had the firm planned or targeted for that
allocation. The short-run benefit function is always going
to be under, or tangent to, the long-run benefit function,
as shown in Figure 10.6. In other words the short-run
benefits b(Q | Q
T
) will never exceed the long-run benefits
B(Q) that could be obtained if the target Q
T
were equal to
the allocation Q. When the target Q
T
equals the allocation
Q, the values of both functions are equal.
Flipping the short run benefit function upside-down
along a horizontal axis running through the long-run
benefit function at the target Q
T
defines the short-run loss
function, as illustrated in Figure 10.7. The short-run loss
of any actual allocation Q equals the long-run benefit of
the target allocation Q
T
minus the short-run benefit of the
actual allocation, Q:
L(Q | Q
T
) B(Q
T
) b(Q | Q
T
) (10.4)
This function defines the losses that occur when the target
allocation cannot be met. When the actual allocation
equals the target allocation, the short-run loss is zero. It is
possible there might be short-run gains or benefits if there
is a surplus allocation over the target allocation, possibly
for a limited range of excess allocations.
The short-run benefit function, or its corresponding
loss function, usually depends on the value of the target
allocation. However, if the short-run losses associated
304 Water Resources Systems Planning and Management
with any deficit allocation (Q
T
Q
1
) or surplus allocation
(Q
2
Q
T
) are relatively constant over a reasonable range of
targets, it may not be necessary to define the loss as a
function of the target Q
T
. In this case the loss can be
defined as a function of the deficit D and/or as a function
of the surplus (excess) E. Both the deficit D or excess E
can be defined by the constraint
Q Q
T
D E (10.5)
Denote the loss function for a deficit allocation as L
D
(D)
and for a surplus allocation, L
E
(E). As indicated above, the
latter may be a negative loss a gain at least for some
range of E, as shown in Figure 10.7.
The costs of the capacity of many components or
multipurpose projects are not easily expressed as func-
tions of the targets associated with each use. For example,
the capacity, K, of a multipurpose reservoir is not usually
equal to, or even a function of, its recreation level target
or its active or flood storage capacities. The costs of its
total capacity, C(K), are best defined as functions of that
total capacity. If expressed as an annual cost, this would
include the annual amortized capital costs as well as the
annual operation, maintenance and repair costs.
Assuming that the benefit and loss functions reflect
annual benefits and losses, the annual net benefits, NB,
from all projects j is the sum of each projects long-run
benefits B
j
(T
j
) that are functions of their targets, T
j
, less
short-run losses L
j
(Q
j
| T
j
) and capacity costs C
j
(K
j
):
NB [B
j
(T
j
) L
j
(Q
j
| T
j
) C
j
(K
j
)] (10.6)
The monetary net benefits accrued by each group of water
users can also be determined so that the income redistri-
bution effects of a project can be evaluated.
The formulation of benefit functions as either long or
short-run is, of course, a simplification of reality. In
reality, planning takes place on many time scales, and for
each time scale one could construct a benefit function.
Consider the planning problems of farmers. In the very
long run, they decide whether or not to own farms, and if
so how big they are to be in a particular area. On a shorter
time scale, farmers allocate their resources to different
activities depending on what products they are producing
and on the processes used to produce them. Different
activities, of course, require capital investments in farm
machinery, storage facilities, pipes, pumps and the like,
some of which cannot easily be transferred to other uses.
j

l
o
s
s
allocation Q
E
0
2
0
1
1
5
m
Q
1
Q
T
Q
2
loss function
T
L(Q |Q
1
)
L Q Q ( |
T
) short-run
Figure 10.7. Short-run loss function, L(Q|Q
T
), associated with
an allocation Q and target allocation Q
T
.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 304
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 305
At least on an annual basis, most farmers re-appraise these
resource allocations in light of the projected market
prices of the produced commodities and the availability
and cost of water, energy, labour and other required
inputs. Farmers can then make marginal adjustments in
the amounts of land devoted to different crops, animals
and related activities within the bounds allowed by avail-
able resources, including capital and labour. At times dur-
ing any growing season some changes can be made in
response to changes in prices and the actual availability of
water.
If the farmers frequently find that insufficient water is
available in the short run to meet livestock and crop
requirements, then they will reassess and perhaps change
their long-run plans. They may shift to less water-
intensive activities, seek additional water from other
sources (such as deep wells), or sell their farms (and
possibly water rights) and engage in other activities. For
the purposes of modelling, however, this planning
hierarchy can generally be described by two levels,
denoted as long and short run. The appropriate decisions
included in each category will depend on the time scale of
a model.
These long-run and short-run benefit and loss func-
tions are applicable to some water users, but not all. They
may apply in situations where benefits or losses can be
attributed to particular allocations in each of the time
periods being modelled. They do not apply in situations
where the benefits or losses result only at the end of
a series of time periods, each involving an allocation
decision. Consider irrigation, for example. If each growing
season is divided into multiple periods, then the benefits
derived from each periods water allocation cannot be
defined independently of any other periods allocation, at
least very easily. The benefits from irrigation come only
when the crops are harvested, for example, in the last
period of each growing season. In this case some mecha-
nism is needed to determine the benefits obtained from
a series of interdependent allocations of water over a
growing season, as will be presented in the next chapter.
4.2. Environmental Criteria
Environmental criteria for water resources projects can
include water quantity and quality conditions. These are
usually expressed in terms of targets or constraints for
flows, depths, hydroperiods (duration of flooding), stor-
age volumes, flow or depth regimes, and water quality
concentrations that are considered desirable for aesthetic
or public health reasons or for various ecosystem habitats.
These constraints or targets could specify desired mini-
mum or maximum acceptable ranges, or rates of changes,
of these values, either for various times within each year
or over an n-year period.
Water quality constituent concentrations are usually
expressed in terms of some maximum or minimum
acceptable concentration, depending on the particular
constituents themselves and the intended uses of the
applicable water body. For example, phosphorus would
normally have a maximum permissible concentration,
while dissolved oxygen would normally have a minimum
acceptable concentration. These limiting concentrations
and their specified reliabilities are often based on
standards established by national or international organi-
zations. As standards, they are viewed as constraints.
These standards could also be considered as targets,
and the maximum or average adverse deviations from
these standards or targets could be a system performance
measure.
Some water quality criteria may be expressed in
qualitative terms. Qualitative quality criteria can provide
a fuzzy limit on the concentration of some constituent
in the water. Such criteria might be expressed as, for
example, the surface water shall be virtually free from
floating petroleum-derived oils and non-petroleum oils of
vegetable or animal origin. Fuzzy membership functions
can define what is considered virtually free, and these
can be included as objectives or constraints in models
(as discussed in Chapter 5).
Environmental performance criteria can vary depend-
ing on the specific sites and on the intended uses of water
at that site. They should be designed to assess, or define,
the risks of adverse impacts on the health of humans and
aquatic life from exposure to pollutants.
Environmental performance criteria of concern to
water resources planners and managers can also relate
to recreational and land use activities. These typically
address hydrological conditions, such as streamflows or
lake or reservoir storage volume elevations during speci-
fied times, or land use activities on watersheds. For exam-
ple, to increase the safety of sailors and individuals fishing
downstream of hydropower reservoirs, release rules may
wrm_ch10.qxd 8/31/2005 11:57 AM Page 305
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
have to be altered to reduce the rate of flow increase that
occurs during peak power production. Performance
criteria applicable to the adverse environmental impacts
of sediment loads such as those caused by logging or
construction activities, or to the impact of nutrient loads
in the runoff, perhaps from agricultural and urban areas
are other examples.
4.3. Ecological Criteria
Criteria that apply to aquatic ecosystems involve both
water quantity and quality and are often compatible with
environmental criteria. It is the time-varying regimes of
water quantities and qualities, not minimum or maximum
values, that benefit and have impacts upon ecosystems.
It is not possible to manage water and its constituent
concentrations in a way that maximizes the health or
well-being, however measured, of all living matter in an
ecosystem. When conditions favour one species that feeds
on another, it is hard to imagine how the latter can be as
healthy as the former. The conditions that favour one
species group may not favour another. Hence, variation in
habitat conditions is important for the sustainability of
resilient biodiverse ecosystems.
While ecosystem habitats exhibit more diversity when
hydrological conditions vary, as in nature, than when
they are relatively constant, hydrological variation is often
not desired by most human users of water resources sys-
tems. Reducing hydrological variation and increasing the
reliability of water resources systems has often been the
motivating factor in the design and operation of hydraulic
engineering works.
The states of ecologial habitats are in part functions of
how water is managed. One way to develop performance
indicators of ecosystems is to model the individual
species that they are composed of, or at least a subset of,
important indicator species. This is often very difficult.
Alternatively, one can define habitat suitability indices for
these important indicator species. This requires, first,
selecting the indicator species representative of each
particular ecosystem, second, identifying the hydrological
attributes that affect the well-being of those indicator
species during various stages of their life cycles, and third,
quantifying the functional relationships between the well-
being of those species and values of the applicable hydro-
logical attributes, usually on a scale from 0 to 1. A habitat
306 Water Resources Systems Planning and Management
suitability value of 1 is considered an ideal condition. A
value of 0 is considered to be very unfavourable.
Examples of hydrological attributes that affect eco-
systems could include flow depths, velocities, constituent
concentrations and temperatures, their durations or the
rate of change in any of those values over space and/or
time at particular times of the year. In wetlands, the
hydrological attributes could include the duration of
inundation (hydroperiod), time since last drawdown
below some threshold depth, the duration of time below
or above some threshold depth, and time rates of change
in depth. The applicable attributes themselves, or perhaps
just their functional relationships, can vary depending on
the time of year and on the stage of species development.
Figure 10.8 illustrates three proposed habitat suitabil-
ity indices for periphyton (algae) and fish in parts of the
Everglades region of southern Florida in the United
States. All are functions of hydrological attributes that
can be managed. Shown in this figure is the impact
of hydroperiod duration on the habitat of three different
species of periphyton located in different parts of the
Everglades, and the impact of the duration of the
hydroperiod as well as the number of years since the last
dry period on a species of fish.
There are other functions that would influence the
growth of periphyton and fish, such as the concentrations
of phosphorus or other nutrients in the water. These are
not shown. Figure 10.8 merely illustrates the construction
of habitat suitability indices.
There are situations where it may be much easier and
more realistic to define a range of some hydrological
attribute values as being ideal. Consider fish living in
streams or rivers for example. Fish desire a variety of
depths, depending on their feeding and spawning
activities. Ideal trout habitat, for example, requires both
deeper pools and shallower riffles in streams. There is no
ideal depth or velocity, or even a range of depths or veloc-
ities above or below some threshold value. In such cases,
it is possible to divide the hydrological attribute scale into
discrete ranges and identify the ideal fraction of the entire
stream or river reach or wetland that ideally would be
within each discrete range. This exercise will result in a
function resembling a probability density function. The
total area under the function is 1. (The first and last
segments of the function can represent anything less or
greater than some discrete value, where appropriate, and
wrm_ch10.qxd 8/31/2005 11:57 AM Page 306
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 307
1.0
0.5
0.0
epiphyton
1.0
0.5
0.0
floating mat
1.0
0.5
0.0
benthic mat
0
0
12
12
24
24
36
36
6
6
18
18
30
30
0 12 24 36 6 18 30
epiphyton
floating mat
benthic mat
s
i
s
i
s
i
hydroperiod (months)
E
0
2
0
1
1
5
n
fish density I)
applies to remnant Everglades
s
i
0 180 360
1.0
0.5
0.0
120 60 240 300
time (days)
(annual hydroperiod)
E020115o
fish density II)
applies to ridge and slough landscape
s
i
0 6
1.0
0.5
0.0
2 1 4
time (years)
(time since drawdown to "DRY" conditions)
3 5
Figure 10.8. Some proposed habitat suitability indices (SI) for three types of periphyton (algae) and a species of fish in portions
of the Everglades region in southern Florida of the United States.
if so the applicable segments are understood to cover
those ranges.) Such a function is shown in Figure 10.9.
Figure 10.9 happens to be a discrete distribution, but it
could have been a continuous one as well.
Any predicted distribution of attribute values resulting
from a simulation of a water management policy can
be compared to this ideal distribution, as is shown in
Figure 10.10. The fraction of overlapping areas of both
distributions is an indication of the suitability of that
habitat for that indicator species. In Figure 10.10, this is
the red shaded area under the blue curve, divided by the
total red shaded area.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 307
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Habitat suitability Fraction of area under the ideal
and simulated distributions of
attribute values (10.7)
To identify a representative set of indicator species of
any ecosystem, the hydrological attributes or stressors
that affect those indicator species, and finally the specific
functional relationships between the hydrological attributes
and the habitat suitability performance indicators, is not
a trivial task. The greater the number of experienced
individuals involved in such an exercise, the more difficult
308 Water Resources Systems Planning and Management
it may be to reach some agreement or consensus. This just
points to the complexity of ecosystems and the difficult
task of trying to simplify it in order to define habitat
suitability performance criteria. However, once identified,
these habitat suitability performance criteria can give water
resources planners and managers an admittedly incomplete
but at least relative indication of the ecosystem impacts of
alternative water management policies or practices.
The use of these habitat suitability functions along with
other performance criteria in optimization and simulation
models will be discussed later in this chapter.
4.4. Social Criteria
Social performance criteria are often not easily defined
as direct functions of hydrological attributes. Most social
objectives are only indirectly related to hydrological
attributes or other measures of water resources system
performance. Economic, environmental and ecological
impacts resulting from water management policies
affect people directly. One social performance criterion
that has been considered in some water resources
development projects, especially in developing regions,
has been employment. Where employment is considered
important, alternatives that provide more jobs may be
preferred to those that use more heavy machinery in place
of labour, for example.
Another social performance criterion is human settle-
ment displacement. The number of families that must
move from their homes because of, for example, flood-
plain restoration or reservoir construction is always of
concern. These impacts can be expressed as a function of
the extent of floodplain restoration or reservoir storage
capacity, respectively. Often the people most affected
are in the lower income groups, and this raises legitimate
issues of justice and equity. Human resettlement impacts
have both social and economic dimensions.
Social objectives are often the more fundamental
objectives discussed earlier in this chapter. Asking why
identifies them. Why improve water quality? Why prevent
flood damage? Why, or why not, build a reservoir? Why
restore a floodplain or wetland? Conversely, if social objec-
tives are first identified, by asking and then answering
how they can be achieved, this usually results in the
identification of economic, environmental and ecological
objectives more directly related to water management.
attribute value
E020115q
p
r
o
p
o
r
t
i
o
n
o
f

a
t
t
r
i
b
u
t
e

v
a
l
u
e
s
ideal distribution
predicted distribution
Figure 10.10. Predicted (simulated indicated in blue) and
ideal (indicated in red) distributions of attribute values for
some ecosystem indicator species. The fraction of the area
common to both distributions is a measure of habitat
suitability.
attribute value
E020115p
i
d
e
a
l

p
r
o
p
o
r
t
i
o
n
o
f

a
t
t
r
i
b
u
t
e

v
a
l
u
e
s
Figure 10.9. Ideal range of values of a hydrological attribute
for a particular component of an ecosystem.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 308
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 309
The extent of press coverage or of public interest
and participation in the planning and evaluation processes
can also be an indicator of social satisfaction with water
management. It is in times of social stress due to, for
example, floods, droughts, or disease caused by water-
borne bacteria, viruses and pollutants that social impacts
become important, that press coverage and public involve-
ment increases. (Public interest also increases when there
is a lot of money to be spent, but this is often a result
of substantial public interest as well.) It is a continuing
challenge to actively engage an often uninterested
public in water management planning at times when there
are no critical water management impacts being felt and
not a lot of money is being spent. Yet this is just the time
when planning for more stressful conditions should
take place.
5. Multi-Criteria Analyses
Given multiple performance criteria measured in multiple
ways, how can one determine the best decision, i.e. the
best way to develop and manage water? Just what is best,
or as some put it, rational? The answer to these questions
will often differ depending on who is being asked. There
is rarely an alternative that makes every interest group or
affected stakeholder the happiest. In such cases we can
identify the efficient tradeoffs among the objective values
each stakeholder would desire. In this section, some ways
of identifying efficient tradeoffs are reviewed. These
methods of multi-criteria or multi-objective analyses are
not designed to identify the best solution, but only to
provide information on the tradeoffs between given sets
of quantitative performance criteria. Again, any final
decision will be based on qualitative as well as this
quantitative information in a political process, not in a
computer.
Even if the same, say monetary, units of measure can
be used for each planning or management objective, it is
not always possible to identify the maximum net benefit
alternative. Consider for example a water resources
development project to be designed to maximize net
economic benefits. In the United States, this is some-
times designated as NED: national economic develop-
ment. That is one objective. A second objective is to
distribute the costs and benefits of the project in an
equitable way. Both objectives are measured in mone-
tary units. While everyone may agree that the biggest pie
(i.e. the maximum net benefits) should be obtained,
subject to various environmental and ecological con-
straints perhaps, not everyone is likely to agree as to
how that pie should be divided up among all the hungry
stakeholders. Again, issues of equity and justice involve
judgements, and the job of water resources planners and
managers and elected politicians, is to make good ones.
The result is often a solution that does not maximize net
economic benefits. Requiring all producers of waste-
water effluent to treat their wastes to the best practical
level before discharging the remainder, regardless of the
assimilative capacity of the receiving bodies of water, is
one example of this compromise between economic and
social criteria.
The irrigationrecreation example presented earlier in
this chapter illustrates some basic concepts in multi-
objective planning. As indicated in Figure 10.1, one of the
functions of multi-objective planning is the identification
of plans that are technologically efficient. These are plans
that define the production-possibility frontier. Feasible
plans that are not on this frontier are inferior in the sense
that it is always possible to identity alternatives that will
improve one or more objective values without making
others worse.
Although the identification of feasible and efficient
plans is seldom a trivial matter, it is conceptually straight-
forward. Deciding which of these efficient plans is the
best is quite another matter. Social welfare functions that
could provide a basis for selection are impossible to
construct, and the reduction of multiple objectives to a
single criterion (as in Figure 10.2) is seldom acceptable in
practice.
When the various objectives of a water resources
planning project cannot be combined into a single scalar
objective function, a single objective management model
may no longer be appropriate. Rather, a vector optimization
formulation of the problem may be applicable. Let the
vector X represent the set of unknown decision-variable
values that are to be determined, and Z
j
(X) be a perform-
ance criterion or objective that is to be maximized. Each
performance criterion or objective j is a function of these
unknown decision-variable values. Assuming that all J
objectives Z
j
(X) are to be maximized, the model can be
written
wrm_ch10.qxd 8/31/2005 11:57 AM Page 309
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
maximize [Z
1
(X), Z
2
(X), , Z
j
(X), , Z
J
(X)]
subject to:
g
i
(X) b
i
i 1, 2, , m (10.8)
The objective in Equation 10.8 is a vector consisting of
J separate objectives. The m constraints g
i
(X) b
i
define
the feasible region of solutions.
The vector optimization model is a concise way of
formulating a multi-objective problem, but it is not very
useful in solving it. In reality, a vector can be maximized
only if it can be reduced to a scalar. Thus, the multi objec-
tive planning problem defined by Model 10.8 cannot,
in general, be solved without additional information.
Various ways of solving this multi-objective model are
discussed in the following sub-sections.
The goal of multi-objective modelling is the generation
of a set of technologically feasible and efficient plans. An
efficient plan is one that is not dominated.
5.1. Dominance
A plan X dominates all others if it results in an equal or
superior value for all objectives, and at least one objective
value is strictly superior to those of each other plan. In
symbols, assuming that all objectives j are to be maxi-
mized, plan alternative i, X
i
, dominates if
Z
j
(X
i
) Z
j
(X
k
) for all objectives j and plans k (10.9)
and for each plan k i there is at least one objective j

such that
Z
j
(X
i
) Z
j
(X
k
) (10.10)
It is seldom that one plan dominates all others. If it
does, choose it! It is more often the case that different
plans will dominate all plans for different objectives.
However, if there exists two plans k and h such that
Z
j
(X
k
) Z
j
(X
h
) for all objectives j and for some objective
j

, Z
j
(X
k
) Z
j
(X
h
), then plan k dominates plan h and
plan X
h
can be dropped from further consideration,
assuming of course that all objectives are being consid-
ered. If some objectives are not, or cannot be, included
in the analysis, such inferior plans with respect to the
objectives that are included in the analysis should not be
rejected from eventual consideration. Dominance analy-
sis can only deal with the objectives being explicitly
considered.
310 Water Resources Systems Planning and Management
Dominance analysis requires that participants in the
planning and management process specify the objectives
that are to be maximized or minimized. It does not
require the assessment of the relative importance of each
objective. Non-inferior, efficient, or non-dominated solu-
tions are often called Pareto optimal because they satisfy
the conditions proposed by the nineteenth-century Italian
economist Vilfredo Pareto, namely that to improve the
value of any single objective, one should have to accept a
diminishment of at least one other objective.
Consider for example three alternatives A, B and C.
Assume, as shown in Figure 10.11, that plan C is inferior
to plan A with respect to objective Z
1
(X) and also inferior
to plan B with respect to objective Z
2
(X). Plan C might
still be considered the best with respect to both objectives
Z
1
(X) and Z
2
(X). While plan C could have been inferior
to both A and B, as is plan D in Figure 10.11, it should
not necessarily be eliminated from consideration just on
the basis of a pair-wise comparison. In fact, plan D, even
though inferior with respect to both objectives Z
1
(X) and
Z
2
(X), might be the preferred plan if another objective
were included.
Two common approaches for identifying non-
dominated plans that together identify the efficient trade-
offs among all the objectives Z
j
(X) in the Model 10.8
are the weighting and constraint methods. Both methods
Z
(
)
X
2
Z ( ) X
1
E
0
2
0
1
1
7
d
A
C
B
D
Figure 10.11. Four discrete plans along with a continuous
efficiency frontier associated with two objectives, Z
1
and Z
2
.
A pair-wise comparison of plans or objectives may not identify
all the non-dominated plans. All objectives should be
considered before declaring a plan inferior.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 310
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 311
require numerous solutions of a single-objective manage-
ment model to generate points on the objective functions
production-possibility frontier.
5.2. The Weighting Method
The weighting approach involves assigning a relative
weight to each objective to convert the objective vector (in
Equation 10.8) to a scalar. This scalar is the weighted sum
of the separate objective functions. The multi-objective
Model 10.8 becomes
maximize Z [w
1
Z
1
(X) w
2
Z
2
(X)

w
j
Z
j
(X)

w
J
Z
J
(X)]
subject to:
g
i
(X) b
i
i 1, 2, , m (10.11)
where the non-negative weights w
j
are specified constants.
The values of these weights w
j
are varied systematically,
and the model is solved for each combination of weight
values to generate a set of technically efficient (or non-
inferior) plans.
The foremost attribute of the weighting approach is
that the tradeoffs or marginal rate of substitution of one
objective for another at each identified point on the objec-
tive functions production-possibility frontier is explicitly
specified by the relative weights. The marginal rate of
substitution between any two objectives Z
j
and Z
k
, at a
specified constant value of X, is
[dZ
j
/dZ
k
] |
x

constant
w
k
/w
j
(10.12)
This applies when each of the objectives is continuously
differentiable at the point X in question. This is illustrated
for a two-objective maximization problem in Figure 10.12.
These relative weights can be varied over reasonable
ranges to generate a wide range of plans that reflect
different priorities. Alternatively, specific values of the
weights can be selected to reflect preconceived ideas of
the relative importance of each objective. It is clear that
the prior selection of weights requires value judgements.
If each objective value is divided by its maximum possible
value, then the weights can range from 0 to 1 and sum to
1, to reflect the relative importance given to each objective.
For many projects within developing countries, these
weights are often estimated by the agencies financing
the development projects. The weights specified by these
agencies can, and often do, differ from those implied by
national or regional policy. But regardless of who does it,
the estimation of appropriate weights requires a study of
the impacts on the economy, society and development
priorities involved.
Fortunately here we are not concerned with finding
the best set of weights, but merely using these weights
to identify the efficient tradeoffs among the conflicting
objectives. After a decision is made, the weights that
produce that solution might be considered the best, at
least under the circumstances and at the time when the
decision is made. They will probably not be the weights
that will apply in other places in other circumstances at
other times.
A principal disadvantage of the weighting approach is
that it cannot generate the complete set of efficient plans
unless the efficiency frontier is strictly concave (decreas-
ing slopes) for maximization, as it is in Figures 10.1 and
10.12. If the frontier, or any portion of it, is convex, then
only the endpoints of the convex region will be identified
using the weighting method when maximizing, as illus-
trated in Figure 10.13.
5.3. The Constraint Method
The constraint method for multi-objective planning can
produce the entire set of efficient plans for any shape of
efficiency frontier, including that shown in Figure 10.13,
Z
(
)
X
2
Z ( ) X
1
dZ
2
dZ
1
2
W
1
W
E
0
2
0
1
1
7
a
Figure 10.12. The efficiency frontier between two objectives,
Z
1
(X) and Z
2
(X), showing the reduction in one objective, say
Z
1
(X), as the relative weight, w
2
, associated with the other
objective, increases.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 311
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
assuming there are tradeoffs among the objectives. In this
method, one objective, say Z
k
(X), is maximized subject to
lower limits, L
j
, on the other objectives, j k. The solu-
tion of the model, corresponding to any set of feasible
lower limits L
j
, produces an efficient alternative if the
lower bounds on the other objective values are binding.
In its general form, the constraint model is
maximize Z
k
(X) (10.13)
subject to:
g
i
(X) b
i
i 1, 2, , m (10.14)
Z
j
(X) L
j
j k. (10.15)
Note that the dual variables associated with the right-
hand-side values L
j
are the marginal rates of substitution
or rate or change of Z
k
(X) per unit change in L
j
(or Z
j
(X)
if binding).
Figure 10.14 illustrates the constraint method for a
two-objective problem.
An efficiency frontier identifying the tradeoffs among
conflicting objectives can be defined by solving the model
many times for many values of the lower bounds. Just as
with the weighting method, this can be a big job if there
are many objectives. If there are more than two or three
objectives, the results cannot be plotted. Pair-wise tradeoffs
that can easily be plotted do not always clearly identify
non-dominated alternatives.
312 Water Resources Systems Planning and Management
The number of solutions to a weighting or constraint
method model can be reduced considerably if the
participants in the planning and management process can
identify the acceptable weights or lower limits. However,
this is not the language of decision-makers. Decision-
makers who count on the support of each interest group
are not content to assign weights that imply the relative
importance of those various stakeholder interests. In
addition, decision-makers should not be expected to
know what they may want until they know what they
can get, and at what cost (often politically as much as
economically). However, there are ways of modifying the
weighting or constraint methods to reduce the amount of
effort in identifying these tradeoffs as well as the amount
of information generated that is of no interest to those
making decisions. This can be done using interactive
methods that will be discussed shortly.
The weighting and constraint methods are among many
methods available for generating efficient or non-inferior
solutions (see, for example, Steuer, 1986). The use of
methods that generate many solutions, even just efficient
ones, assumes that once all the non-inferior alternatives
have been identified, the participants in the planning and
management process will be able to select the best
compromise alternative from among them. In some
situations this has worked. Undoubtedly, there will be
planning activities in the future where the use of these
non-inferior solution generation techniques alone will
Z
(
)
X
2
Z ( ) X
1
E
0
2
0
1
1
7
c
feasible combinations of objective values
L
2
Figure 10.14. The constraint method for identifying the
efficiency frontier by maximizing Z
1
(X) while constraining Z
2
(X)
to be no less than L
2
.
Z
(
)
X
2
Z ( ) X
1
E
0
2
0
1
1
7
b
range of efficient solutions that cannot
be identified using the weighting method
Figure 10.13. An efficiency frontier that cannot be completely
identified in its convex region using the weighting method
when objectives are being maximized. Similarly for concave
regions when objectives are being minimized.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 312
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 313
continue to be of value. However, in many other planning
situations, they will not be sufficient in themselves. Often,
the number of feasible non-inferior alternatives is simply
too large. Participants in the planning and management
process will not have the time or patience to examine and
evaluate each alternative plan. Participants may also need
help in identifying which alternatives they prefer. If they
are willing to work with the analyst, the analyst can help
them identify what alternatives they prefer without gener-
ating and comparing all the other non-inferior plans.
There are a number of methods available for assisting
in selecting the most desirable non-dominated plan. Some
of the more common ones are described next.
5.4. Satisficing
One method of further reducing the number of alterna-
tives is called satisficing. It requires that the participants
in the planning and management process specify a
minimum acceptable value for each objective. Those
alternatives that do not meet these minimum performance
values are eliminated from further consideration. Those
that remain can again be screened if the minimal
acceptable values of one or more objectives are increased.
When used in an iterative fashion, the number of
non-inferior alternatives can be reduced down to a single
best compromise plan or set of plans that everyone
considers equally desirable. This process is illustrated in
Figure 10.15.
Of course, sometimes the participants in the planning
process will be unwilling or unable to narrow down the
set of available non-inferior plans sufficiently with the
iterative satisficing method. Then it may be necessary to
examine in more detail the possible tradeoffs among the
competing alternatives.
5.5. Lexicography
Another simple approach is called lexicography. To use
this approach, the participants in the planning process
must rank the objectives in order of priority. This ranking
process takes place without considering their particular
values. Then, from among the non-inferior plans that
satisfy the minimum levels of each objective, the plan that
is the best with respect to the highest priority objective
will be the one selected as superior.
If there is more than one plan that has the same value
of the highest priority objective, then among this set of
preferred plans the one that achieves the highest value of
the second priority objective is selected. If here too there
are multiple such plans, the process can continue until
there is a unique plan selected.
This method assumes such a ranking of the objectives
is possible. Often the relative values of the objectives of
each alternative plan are of more importance to those
involved in the decision-making process. Consider, for
example, the problem of purchasing apples and oranges.
Assuming you like both types of fruit, which should you
buy if you have only enough money to buy one type? If
you know you already have lots of apples, but no oranges,
perhaps you would buy oranges, and vice versa. Hence,
the ranking of objectives can depend on the current state
and needs of those who will benefit from the plan.
5.6. Indifference Analysis
Another method of selecting the best plan is called indif-
ference analysis. To illustrate the possible application of
indifference analysis to plan selection, consider a simple
situation in which there are only two alternative plans
(A and B) and two planning objectives (1 and 2) being
considered. Let Z
1
A
and Z
2
A
be the values of the two respec-
tive objectives for plan A and Z
1
B
and Z
2
B
be the values of
Z
(
X
)
2
Z (X)
1
E
0
2
0
1
1
7
e
remaining range of alternatives
after second iteration
second iteration
first iteration
Figure 10.15. Two successive iterations of the satisficing
method in which minimum levels of each objective are set,
thereby eliminating those alternatives that do not meet these
minimum levels.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 313
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
the two respective objectives for plan B. Comparing both
plans when one objective is better than another for
each plan can be difficult. Indifference analysis can reduce
the problem to one of comparing the values of only one
objective.
Indifference analysis first requires the selection of an
arbitrary value for one of the objectives, say Z
2

, for
objective 2. It is usually a value within the range of the
values Z
2
A
and Z
2
B
, or in a more general case between the
maximum and minimum of all objective 2 values.
Next, a value of objective 1, say Z
1
, must be selected
such that the participants involved are indifferent
between the hypothetical plan that would have as its
objective values (Z
1
, Z
2

) and plan A that has as its

objective values (Z
1
A
, Z
2
A
). In other words, Z
1
must be
determined such that (Z
1
, Z
2

) is as desirable as or equiva-
lent to (Z
1
A
, Z
2
A
):
(Z
1
, Z
2

) (Z
1
A
, Z
2
A
) (10.16)
Next another value of the first objective, say Z
1
, must be
selected such that the participants are indifferent between
a hypothetical plan (Z
1
, Z
2

) and the objective values (Z

1
B
,
Z
2
B
) of plan B:
(Z
1
, Z
2

) (Z
1
B
, Z
2
B
) (10.17)
These comparisons yield hypothetical but equally desir-
able plans for each actual plan. These hypothetical plans
differ only in the value of objective 1 and, hence, they are
easily compared. If both objectives are to be maximized
and Z
1
is larger than Z
1
, then the first hypothetical plan
yielding Z
1
is preferred to the second hypothetical plan
yielding Z
1
. Since the two hypothetical plans are equiva-
lent to plans A and B, respectively, plan A must be pre-
ferred to plan B. Conversely, if Z
1
is larger than Z
1
, then
plan B is preferred to plan A.
This process can be extended to a larger number of
objectives and plans, all of which may be ranked by a
common objective. For example, assume that there are
three objectives Z
1
i
, Z
2
i
, Z
3
i
, and n alternative plans i. A
reference value Z
3

for objective 3 can be chosen and a

value Z
1
i
estimated for each alternative plan i such that one
is indifferent between (Z
l
i
, Z
2
i
, Z
3

) and (Z
l
i
, Z
2
i
, Z
3
i
). The
second objective value remains the same as in the
actual alternative in each of the hypothetical alternatives.
Thus, the focus is on the tradeoff between the values of
objectives 1 and 3. Assuming that each objective is to be
314 Water Resources Systems Planning and Management
maximized, if Z
3

is selected so that Z
3

Z
3
i
, then Z
l
i
will no
doubt be greater than Z
3
i
. Conversely, if Z
3

Z
3
i
, then Z
l
i
will
be less than Z
3
i
.
Next, a new hypothetical plan containing a reference
value Z
2

and Z
3

can be created and compared with (?, Z

2
i
,
Z
3

). The focus is on the tradeoff between the values of

objectives 1 and 2. A value of Z
1
i
must be selected such
that the participants are indifferent between (Z
1
i
, Z
2
i
, Z
3

)
and (Z
l
i
, Z
2

, Z
3

). Hence, for all plans i, the participants are

indifferent between two hypothetical plans and the actual
one. The last hypothetical plans differ only by the value of
the first objective. The plan that has the largest value for
objective 1 will be the best plan. This was achieved by
pair-wise comparisons only.
In the first step objective 2 remained constant and only
objectives 1 and 3 were compared to get
(Z
l
i
, Z
2
i
, Z
3
i
) (Z
l
i
, Z
2
i
, Z
3

) (10.18)
In the next step involving the hypothetical plans just
defined, objective 3 remained constant and only objec-
tives 1 and 2 were compared to get
(Z
l
i
, Z
2
i
, Z
3

) (Z
l
i
, Z
2

, Z
3

) (10.19)
Hence
(Z
l
i
, Z
2
i
, Z
3
i
) (Z
l
i
, Z
2
i
, Z
3

) (Z
l
i
, Z
2

, Z
3

) (10.20)
Having done this for all n plans, there are now n hypo-
thetical plans (Z
l
i
, Z
2

, Z
3

) that differ only in the value of

Z
l
i
. All n plans can be ranked just on the basis of the value
of this single objective.
Each of these plan selection techniques requires the
prior identification of discrete alternative plans.
5.7. Goal Attainment
The goal attainment method combines some of the advan-
tages of both the weighting and constraint plan genera-
tion methods already discussed. The participants in the
planning and management process specify a set of goals
or targets T
j
for each objective j and, if applicable, a
weight, w
j
, that reflects the relative importance of meeting
that goal compared to meeting other goals. If the partici-
pants are unable to specify these weights, the analyst must
select them and then later change them on the basis of
their reactions to the generated plans.
The goal attainment method identifies the plans that
minimize the maximum weighted deviation of any
wrm_ch10.qxd 8/31/2005 11:57 AM Page 314
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 315
objective value, Z
j
(X), from its specified target, T
j
. The
problem is to
minimize D (10.21)
subject to:
g
i
(X) b
i
i 1, 2, , m (10.22)
w
j
[T
j
Z
j
(X)] D j 1, 2, , J (10.23)
Constraints 10.22 contain the relationships among the
decision-variables in the vector X. They define the feasi-
ble region of decision-variable values.
This method of multi-criteria analysis can generate
efficient or non-inferior plans by adjusting the weights
and targets. It is illustrated for a two-objective problem in
Figure 10.16.
Unless T
j
Z
j
(X), some plans generated from goal
programming may be inferior with respect to the
objectives being considered.
5.8. Goal-Programming
Goal-programming methods also require specified target
values, along with relative losses or penalties associated
with deviations from these target values. The objective is
to find the plan that minimizes the sum of all such losses
or penalties. Assuming for this illustration that all such
losses can be expressed as functions of deviations from
target values, and again assuming each objective is to be
maximized, the general goal-programming problem is to
minimize v
j
D
j
w
j
E
j
(10.24)
subject to:
g
i
(X) b
i
i 1, 2, , m (10.25)
Z
j
(X) T
j

D
j
E
j
(10.26)
where the parameters v
j
and w
j
are the penalties (weights)
assigned to objective value deficits or excesses, as appro-
priate. The weights and the target values, T
j

, can be
changed to get alternative solutions, or tradeoffs, among
the different objectives.
5.9. Interactive Methods
Interactive methods allow participants in the planning
process to explore the range of possible decisions without
having to generate all of them, especially those of little
interest to the participants.
One such approach, called the step method, is an
iterative method requiring preference information from
the participants in the planning and management process
at each iteration. This information identifies constraints on
various objective values. The weighting method is used to
get an initial solution on the efficiency frontier. The
weights, w
j
, are calculated on the basis of the relative range
of values each objective j can assume, and on whether or
not the participants in the planning and management
process have indicated satisfaction regarding a particular
objective value obtained from a previous solution. If they
are satisfied with the value of, say, an objective Z
j
(X), they
must indicate how much of that value they would be will-
ing to give up to obtain unspecified increases in objectives
whose values they consider unsatisfactory. This defines a
lower bound on Z
j
(X). Then the weight w
j
for that objec-
tive is set to 0, and the weights of all remaining objectives
are re-calculated. The problem is again solved. The process
is repeated until some best compromise plan is identified.
This step method guides the participants in the
planning and management process among non-inferior
alternatives toward the plan or solution the participants
consider best without requiring an exhaustive generation
of all non-inferior alternatives. Even if the best compro-
mise solution is not identified or agreed upon, the
method provides a way for participants to learn what the
tradeoffs are in the region of solutions of interest to them.
However, the participants must be willing to indicate how
j

Z
(
)
X
2
Z ( ) X
1
E
0
2
0
1
1
7
f
ideal, but infeasible plan
T
1
T
2
Figure 10.16. The goal attainment method of generating
points on the efficiency frontier using different values of the
weights w
1
and w
2
for fixed objective target values T
1
and T
2
.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 315
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
much of some objective value can be reduced to obtain
some unknown improvement of other objective values.
This is not as easy as indicating how much more is desired
of any or all objectives whose values are unsatisfactory.
To overcome this objection to the step method, other
interactive methods have been developed. These begin
with an obviously inferior solution. Based on a series
of questions concerning how much more important it is
to obtain various improvements of each objective, the
methods proceed from that inferior solution to more
improved solutions. The end result is either a solution
everyone agrees is best, or an efficient one where no more
improvements can be made in one objective without
decreasing the value of another.
These iterative interactive approaches are illustrated in
Figure 10.17. To work, they require the participation of
the participants in the planning and management process.
5.10. Plan Simulation and Evaluation
The methods outlined above provide a brief introduction
to some of the simpler approaches available for plan
identification and selection. Details on these and other
potentially useful techniques can be found in many books,
some of which are devoted solely to this subject of
multi-objective planning (Cohon, 1978; Steuer, 1986).
Most have been described in an optimization framework.
This section describes ways of evaluating alternative
water management plans or policies on the basis of the
time series of performance criteria values derived from
simulation models.
Simulation models of water resources systems yield
time series of hydrological variable values. These values in
turn affect multiple system performance criteria, each
pertaining to a specific interest and measured in
its appropriate units. A process for evaluating alternative
water resources management plans or policies based on
these simulation model results includes the following
steps:
1. Identify system performance indicators that are
affected by one or more hydrological attributes whose
values will vary depending on the management policy
or plan being simulated. For example, navigation
benefits, measured in monetary units, might depend
on water depths and velocities. Hydropower produc-
316 Water Resources Systems Planning and Management
tion is affected by lake levels and discharges through
the power plant. Water quality might be expressed as
the average or maximum concentration over a fixed
time period at certain locations for various potential
pollutants, and will depend in part on the flows.
Ecological habitats may be affected by flows, water
depths, water quality, flooding frequency or duration,
and/or rates of changes in these attributes.
2. Define the functional relationships between these
performance indicators and the hydrological attrib-
utes. Figure 10.18 illustrates such functions. The units
on each axis may differ for each such function.
3. Simulate to obtain time series of hydrological attribute
values and map them into a time series of performance
Z
(
)
X
2
Z ( ) X
1
3
2
4
=
=
initial solution
third solution
unknown
efficiency frontier
first
Z
1
Z
1
Z
2
i
t
e
r
a
t
i
o
n
r
e
d
u
c
t
i
o
n
second
third
Z
(
)
X
2
Z ( ) X
1
E
0
2
0
1
1
7
g
successive iterations toward efficiency frontier
Figure 10.17. Two interactive iterative multiple criteria
approaches for identifying the tradeoffs of interest and
possibly the best decision.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 316
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 317
indicator values using the functional relationships
defined in Step 2. This step is illustrated in Figure 10.19.
4. Combine multiple time-series values for the same
performance criterion, as applicable, as shown in
Figure 10.20. This can be done using maximum or
minimum values, or arithmetic or geometric means, as
appropriate. For example, flow velocities, depths and
algal biomass concentrations may affect recreational
boating. The three sets of time series of recreational
boating benefits or suitability can be combined into
one time series, and statistics of this overall time series
can be compared to similar statistics of other perform-
ance indicators. This step gives the modeller an
opportunity to calibrate the resulting single system
performance indicator.
5. Develop and compare system performance exceedance
distributions, or divide the range of performance
values into colour-coded ranges and display on maps
or in scorecards, as illustrated in Figure 10.21.
p
e
r
f
o
r
m
a
n
c
e
i
n
d
i
c
a
t
o
r
1
hydrological attribute hydrological attribute
E
0
2
0
1
1
7
h
p
e
r
f
o
r
m
a
n
c
e
i
n
d
i
c
a
t
o
r
2
Figure 10.18. Performance indicators expressed as functions
of simulated hydrological attributes.
p
e
r
f
o
r
m
a
n
c
e
i
n
d
i
c
a
t
o
r
time
p
e
r
f
o
r
m
a
n
c
e
i
n
d
i
c
a
t
o
r
attribute
a
t
t
r
i
b
u
t
e
time
E
0
2
0
1
1
7
j
Figure 10.19. Mapping a time series of hydrological attribute
values into a corresponding time series of performance
indicator values.
Figure 10.20. Combining multiple time series of values of a
specified performance indictor into a single time series of
values of that performance indicator.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 317
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
318 Water Resources Systems Planning and Management
alternative strategy
p
e
r
f
o
r
m
a
n
c
e

i
n
d
i
c
a
t
o
r
s
A
3
10
-
0.2
B
100
E
0
2
0
1
1
7
o
B
5
12
0.8
C
45
C
65
+++
0.5
D
49
D
6
4
1.0
B
69
E
8
57
0.0
A
78
F
3
95
0.7
B
34
G
2
34
0.4
C
22
1
2
3
4
5
6
-- --- + --
Figure 10.21. Ways of summarizing and displaying time-series performance indicator data
involving exceedance distributions, and colour-coded maps and scorecards. Colour-coded
map displays on computers can be dynamic, showing changes over time. Green and red
coloured scorecard entries indicate best and worst plan or strategy, respectively, for
associated performance indicator (for scorecards see also Appendix E Section 6.4).
wrm_ch10.qxd 8/31/2005 11:57 AM Page 318
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 319
The area under each exceedance curve is the mean.
Different exceedance functions will result from different
water management policies, as illustrated in Figure 10.22.
One can establish thresholds to identify discrete zones
of performance indicator values, and assign a colour to
each zone. Measures of reliability, resilience and vulnera-
bility can then be calculated and displayed as well.
Scorecards can show the mean values of each indicator
for any set of sites. The best value for each indicator can
be coloured green; the worst value for each indicator
can be coloured red. The water management alternative
having the most green boxes will stand out and will prob-
ably be considered more seriously than the alternative
having the highest number of red (worst value) boxes.
This five-step process has been used in a study of
improved ways of managing lake levels and flows in Lake
Ontario and its discharges into the St. Lawrence River
in North America. Performance criteria were defined
for domestic and industrial water supplies, navigation
depths, shore bank erosion, recreational boating,
hydropower production, flooding, water quality and
ecological habitats. The performance measures for each of
these interests were identified and expressed as functions
of one or more hydrological variable values related to flow
and lake-level management. Models designed to simulate
alternative lake-level and flow-management policies were
used to generate sets of time series for each system
performance criterion. These in turn were combined,
summarized and compared.
The same five-step process has been implemented in
the Everglades restoration project in southern Florida.
The Everglades is a long, very wide and extremely flat
river of grass flowing generally south into, eventually, the
Atlantic Ocean and Gulf of Mexico. This ecosystem
restoration project has involved numerous local, state and
federal agencies. The project affects a large population
and agricultural industry that want secure and reliable
water supplies and flood protection. Its current estimated
cost over some three decades is about $8 billion. Hence,
it involves politics. But its goal is primarily focused on
restoring a unique ecosystem that is increasingly
degraded due to extensive alterations in its hydrology
over the past half-century.
The motto of the Everglades restoration project is to
get the water right. Those who manage the regions water
are attempting to restore the ecosystem by restoring the
hydrological regime the flows, depths, hydroperiods
and water qualities throughout the region to what they
think existed some fifty years ago. The trick is to accom-
plish this and still meet the water supply, flood protection
and land development needs of those who live in the
region. Clearly, achieving a hydrological condition that
existed before people began populating that region in sig-
nificant numbers will not be possible. Hence the ques-
tion: what if water managers are not able to get the water
right? What if they can only get the water right on 90%
of the area, or what if they can only get 90% of the water
right on all the area? What will be the likely impact on the
ecosystem? Are there opportunities for changing hydrol-
ogy to improve ecology? Where?
To address questions such as these, at least in a pre-
liminary way until more detailed ecological models
become available, the five-step approach outlined above
is being applied in the Everglades. It is being used to
extend simulated hydrological predictions to produce
relative values of ecological habitat suitability indicators
for selected indicator species, as illustrated in Figure 10.8,
and topographic characteristics.
This five-step procedure does not find an optimal
water management policy. It can however contribute
useful information to the political debate that must
take place in the search for that optimum. Each step of
the approach can and should include and involve the
various stakeholders and publics in the basin. These
individuals are sources of important inputs in this
evaluation process. Stakeholders who will be influenc-
ing or making water-management decisions need to
Figure 10.22. Three exceedance functions showing decrease
in mean performance value if Policy 1 is followed compared to
Policies 2 and 3.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 319
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
understand just how this multi-objective evaluation
process works if they are to accept and benefit from its
results. Stakeholder involvement in this process can
help lead to a common understanding (or shared
vision) of how their system works and of the tradeoffs
that exist among conflicting objectives. The extent to
which all stakeholders understand this evaluation
approach or procedure and how it is applied in their
basin will largely determine their ability to participate
effectively in the political process of selecting the best
water management policy or practice.
6. Statistical Summaries of
Performance Criteria
There are numerous ways of summarizing time-series
performance data that might result from simulation
analyses. Weighted arithmetic mean values or geometric
mean values are two ways of summarizing multiple time-
series data. The overall mean itself generally provides too
little information about a dynamic process. Multiple time-
series plots themselves are often hard to compare.
Another way to summarize and compare time-series
data is to calculate and compare the variance of the data.
Consider a time series of T values X
t
whose mean is X.
For example, suppose the time series consisted of 8, 5, 4,
9, 2, 1, 1, 3, 6, and 7. The mean of these 10 values is 4.6.
The variance is
(X
t
X)
2
/T [(8 4.6)
2
(5 4.6)
2

(7 4.6)
2
]/10 7.44 (10.27)
t
T

320 Water Resources Systems Planning and Management

A plot of these values and their mean is shown in
Figure 10.23.
The mean and variance for the time series shown in
Figure 10.23, however, are the same for its upside-down
image, as shown in Figure 10.24. They do not even
depend on the order of the time-series data.
Consider these two sets of time series shown again in
Figure 10.25, each having the same mean and variance.
Assume that any value equal to or less than the dashed
line (just above 2) is considered unsatisfactory. This value
is called a threshold value, dividing the time-series data
into satisfactory and unsatisfactory values.
It is clear from Figure 10.25 that the impact of these two
time series could differ. The original time series shown in a
red line remained in an unsatisfactory condition for a
longer time than did the rotated time series shown in blue.
However, its maximum extent of failure when it failed was
less than the rotated time series. These characteristics can
10
8
6
4
2
0
0 1 2 3 4 5 6 7 8 9 10
mean value
E
0
2
0
1
1
7
p
time
v
a
l
u
e
Figure 10.23. Plot of time-series data having a mean value
of 4.6 and a variance of 7.44.
10
8
6
4
2
0
E
0
2
0
1
1
7
q
0 1 2 3 4 5 6 7 8 9 10
mean value
time
v
a
l
u
e
Figure 10.24. A plot of two different time series having the
same mean and variance.
10
8
6
4
2
0
0 1 2 3 4 5 6 7 8 9 10
time
v
a
l
u
e
satisfactory
unsatisfactory
value:
threshold
E
0
2
0
1
1
7
r
Figure 10.25. Threshold value distinguishing values
considered satisfactory, and those considered unsatisfactory.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 320
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 321
be captured by the measures of reliability, resilience and
vulnerability (Hashimoto et al., 1982).
6.1. Reliability
The reliability of any time series can be defined as the
number of data in a satisfactory state divided by the total
number of data in the time series. Assuming satisfactory
values in the time series X
t
containing n values are those
equal to or greater than some threshold X
T
, then
Reliability[X] [number of time periods t such
that X
t
X
T
]/n (10.28)
The reliability of the original time series shown in red is
0.7. It failed three times in ten. The reliability of the
rotated time series shown in blue is also 0.7, failing three
times in ten.
Is a more reliable system better than a less reliable
system? Not necessarily. Reliability measures tell one
nothing about how quickly a system recovers and returns
to a satisfactory value, nor does it indicate how bad an
unsatisfactory value might be should one occur. It may
well be that a system that fails relatively often, but by
insignificant amounts and for short durations, will be
much preferable to one whose reliability is much higher
but where, when a failure does occur, it is likely to be
much more severe. Resilience and vulnerability measures
can quantify these system characteristics.
6.2. Resilience
Resilience can be expressed as the probability that if a
system is in an unsatisfactory state, the next state will be
satisfactory. It is the probability of having a satisfactory
value in time period t 1, given an unsatisfactory value
in any time period t. It can be calculated as
Resilience[X] [number of times a satisfactory value
follows an unsatisfactory value]/
[number of times an unsatisfactory
value occurred] (10.29)
Resilience is not defined if no unsatisfactory values occur
in the time series. For the original time series shown in
red, the resilience is 1/3, again assuming the value of 2 or
less is considered a failure. For the rotated time series
shown in blue the resilience is 2/2 1. We cannot judge
the resilience of the blue time series on the basis of the
last failure in period 10 because we do not have an
observation in period 11.
6.3. Vulnerability
Vulnerability is a measure of the extent of the differences
between the threshold value and the unsatisfactory time-
series values. Clearly, this is a probabilistic measure.
Some use expected values, some use maximum observed
values, and others may assign a probability of exceedance
to their vulnerability measures. Assuming an expected
value measure of vulnerability is to be used:
Vulnerability[X] [sum of positive values of (X
T
X
t
)]/
[number of times an unsatisfactory
value occurred] (10.30)
The expected vulnerability of the original red time series
is [(2 2) (2 1) (2 1)]/3 0.67. The expected
vulnerability of the time series shown by the rotated blue
line on Figure 10.25 is [(2 1.2) (2 0.2)]/2 1.3.
So, depending on whether a threshold value is consid-
ered a failure or not in this example, the reliability and
resilience of original (red) time series is equal or less than
the rotated (blue) time series. However, the expected
vulnerability of the original red time series is less than that
of the rotated blue time series. This shows the typical
tradeoffs one can define using these three measures of
system performance.
7. Conclusions
Many theoretical and practical approaches have been
proposed in the literature for identifying and quantifying
objectives and for considering multiple criteria or
objectives in water resources planning. The discussion
and techniques presented in this chapter serve merely as
an introduction to this subject. These tools, including
their modifications and extensions, are designed to
provide information that can be of value to the planning
and decision-making process.
Water resources systems planners and managers and
the numerous other participants typically involved in
decision-making face a challenge when they are required
to select one of many alternatives, each characterized
by different values among multiple performance criteria.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 321
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
It requires a balancing of the goals and values of the vari-
ous individuals and groups concerned with the project.
There is virtually no way in which the plan-selection step
can be a normative process or procedure; there can be no
standard set of criteria or methods which will identify the
preferred project. At best, an iterative procedure in which
those using tools similar to those described in this chapter
together with all the interested stakeholders may reach
some shared vision of what is best to do, at least until
conditions change or new knowledge or new goals or
new requirements emerge. This may be the only way to
identify a plan that is politically as well as technically,
socially, financially and institutionally feasible.
To many participants in the planning process, some of
these approaches for objective quantification and multi-
objective planning may seem theoretical or academic.
Many may be reluctant to learn quantitative policy
analysis techniques or to spend time answering seemingly
irrelevant questions that might lead eventually to a
compromise plan. Reluctance to engage in quantification
of tradeoffs among particular objectives of alternative
plans, and by implication alternative stakeholder groups,
may stem from the support decision-makers desire from
all of these conflicting interest groups. In such situations
it is obviously to their advantage not to be too explicit in
quantifying political values. They might prefer the
analyst to make these tradeoffs so that they just need not
discuss them. (We writers have participated in such
situations.) Planners, engineers or analysts are often very
willing to make these tradeoffs because they pertain to
subject areas in which they often consider themselves
experts. When political tradeoffs are at issue, no one is an
expert. No one has the optimal answer, but professionals
should be engaged in and informing and facilitating
the process of coming to an acceptable, and often
compromise, decision.
Through the further development and use of practical
analytical multi-objective planning techniques, analysts
can begin to interact with all participants in the planning
and management process and can enlighten any who
would argue that water resources policy evaluation and
analyses should not be political. Analysts, managers and
planners have to work in a political environment. They
need to understand the process of decision-making, what
information is most useful to that process, and how it can
best be presented. Knowledge of these facts in a particular
322 Water Resources Systems Planning and Management
planning situation might largely dictate the particular
approach to objective identification and quantification and
to plan selection that is most appropriate.
The method deemed most appropriate for a particular
situation will depend not only on the physical scale of
the situation itself but also on the decision-makers, the
decision-making process, and the responsibilities accepted
by the analysts, the participants and the decision-makers.
Finally, remember that the decisions being made at the
current time are only part of a sequence of decisions that
will be made on into the future. No one can predict with
certainty what future generations will consider as being
important or what they will want to do, but spending
some time trying to guess is not an idle exercise. It pays
to plan ahead, as best one can, and ask ourselves if the
decisions being considered today will be those we think
our descendants would have wanted us to make. This
kind of thinking gets us into issues of adaptive manage-
ment (Appendix B) and sustainability (ASCE, 1999).
8. References
ASCE (American Society of Civil Engineers). 1999.
Sustainability criteria for water resource systems. Reston, Va.,
ASCE Press, (also D.P. Loucks and J.S. Gladwell (eds.),
Cambridge, U K, Cambridge University Press, 2000).
COHON, J.L. 1978. Multiobjective programming and plan-
ning. New York, Academic Press.
GREGORY, R.S. and KEENEY, R.L. 1994. Creating policy
alternatives using stakeholder values. Management Science,
Vol. 40, pp. 103548.
HAMMOND, J.S.; KEENEY, R.L. and RAIFFA, H. 1999.
Smart choices: a practical guide to making better decisions.
Boston, Mass., Harvard Business School Press.
HASHIMOTO, T.; LOUCKS, D.P. and STEDINGER, J.R.
1982. Robustness of water resource systems. Water
Resources Research, Vol. 18, No. 1, pp. 2126.
HASHIMOTO, T.; STEDINGER, J.R. and LOUCKS, D.P.
1982. Reliability, resiliency and vulnerability criteria for
water resource system performance evaluation. Water
Resources Research, Vol. 18, No. 1, pp. 1420.
STEUER, R.E. 1986. Multiple criteria optimization: theory,
computation, and application. New York, Wiley.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 322
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Performance Criteria 323
Additional References (Further Reading)
CHEN, H.W. and CHANG, N.B. 1998. Water pollution
control in the river basin by fuzzy genetic algorithm-
based multiobjective programming modelling. Water
Science and Technology, Vol. 37, No. 8, pp. 5563.
CIENIAWSKI, S.E.; EHEART, J.W. and RANJITHAN, S.
1995. Using genetic algorithms to solve a multiobjective
groundwater monitoring problem. Water Resources
Research, Vol. 31, No. 2, February, pp. 399409.
ESCHENAUER, H.; KOSKI, J. and OSYCZKA, A. 1990.
Multicriteria design optimization. Berlin, Springer-Verlag.
GUPTA, H.V.; SORROSHIAN, S. and YAPO, P.O.
1998. Towards improved calibration of hydrological
models: multiple and non-commensurable measures
of information. Water Resources Research, Vol. 34, No. 4,
pp. 75163.
HAIMES, Y.Y.; HALL, W.A. and FREEDMAN, H.T. 1975.
Multiobjective optimization in water resources systems.
Amsterdam, Elsevier Scientific.
HALHAL, D.; WALTERS, G.A.; OUAZAR, D. and SAVIC,
D.A. 1997. Multiobjective improvement of water distri-
bution systems using a structure messy genetic algorithm
approach. Journal of Water Resources Planning and
Management ASCE, Vol. 123, No. 3, pp. 13746.
KEENEY, R.L. and RAIFFA, H. 1976. Decisions with mul-
tiple objectives: preferences and value tradeoffs. New York,
Wiley.
LOUCKS, D.P.; STEDINGER, J.R. 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.
REED, P.M.; MINSKER, B.S. and GOLDBERG, D.E. 2001.
A multiobjective approach to cost effective long-term
groundwater monitoring using an elitist nondominated
sorted genetic algorithm with historical data. Journal of
Hydroinformatics, Vol. 3, No. 2, April, pp. 7189.
RITZEL, B.J.; EHEART J.W. and RANJITHAN, S. 1994.
Using genetic algorithms to solve a multiple objective
groundwater pollution containment problem. Water
Resources Research, Vol. 30, No. 5, May, pp. 15891603.
SAWARAGI, Y.; NAKAYAMA, H. and TANINO, T. 1985.
Theory of multiobjective optimization. Orlando, Fla.,
Academic Press.
STATNIKOV, R.B. and MATUSOV, J.B. 1995. Multicriteria
optimization and engineering. New York, Chapman and Hall.
YEH, C.H. and LABADIE, J.W. 1997. Multiobjective
Watershed-level planning of storm-water detention sys-
tems. Journal of Water Resources Planning and Management,
Vol. 123, No. 6, November/December, pp. 336043.
ZELENY, M. 1982. Multiple criteria decision making.
New York, McGraw-Hill.
wrm_ch10.qxd 8/31/2005 11:57 AM Page 323
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
11. River Basin Planning Models
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 Resource 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
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 Optimzation 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
wrm_ch11.qxd 8/31/2005 11:57 AM Page 324
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
This chapter introduces some types of models commonly
used to assist those responsible for planning and manag-
ing various components of river systems. These compo-
nents include the watersheds that drain into the surface
water bodies and underlying aquifers of river systems.
They include the streams, rivers, lakes, reservoirs and
wetlands that can exist in river basins and that are affected
by water management policies and practices. First, each of
these components will be examined and modelled sepa-
rately. The management of any single component, how-
ever, can affect the performance of other components in a
river basin system. Hence, for the overall management of
river basin systems, a systems view is needed. Typically,
this systems view requires the modelling of multiple com-
ponents. These multi-component models can then be
325
River Basin Planning Models
Multipurpose river basin management typically involves the identification and use
of both structural and non-structural measures designed to increase the reliability
of municipal, industrial and agriculture water supplies when and where demanded,
to protect against floods, to improve water quality, to provide for commercial
navigation and recreation, and to produce hydropower, as appropriate for the
particular river basin. Structural measures may include diversion canals,
reservoirs, hydropower plants, levees, flood proofing, irrigation delivery and
drainage systems, navigation locks, recreational facilities, groundwater wells, and
water and wastewater treatment plants along with their distribution and collection
systems. Non-structural measures may include land use controls and zoning, flood
warning and evacuation measures, and economic incentives that affect human
behaviour with regard to water and watershed use. Planning the development and
management of water resources systems involves identifying just what, when and
where structural or non-structural measures are needed, the extent to which they
are needed, and their combined economic, environmental, ecological and social
impacts. This chapter introduces some modelling approaches for doing this. The
focus is on water quantity management. The following chapter reviews some
measures and models for water quality management.
11
used to analyse alternative designs and management
strategies for integrated multi-component systems.
River basin planning is a prerequisite for integrated
water resources management (IWRM). IWRM requires
the integration of the natural system components (surface
watergroundwater, quantityquality, land- and water
management, etc.) and the upstream and downstream
water-related demands or interests. Water resources
planning is increasingly done on a river basin scale. The
European Water Framework Directive, for example,
imposes the development of basin plans in Europe, forcing
riparian countries to work together on the development
and management of their river basins.
The discussion in this chapter is limited to water
quantity management. Clearly the regimes of flows, veloc-
ities, volumes and other properties of water quantity will
affect the quality of that water as well. Moreover, from the
wrm_ch11.qxd 8/31/2005 11:57 AM Page 325
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
326 Water Resources Systems Planning and Management
Surface-flow conditions that are of interest to those
managing water supplies, navigation, hydropower
production, recreation and water quality can be modelled
using mass balances in one-dimensional models with time
steps of up to a season in length. Flood-flow conditions
on the other hand require higher dimension models
and much smaller time steps. Groundwater flows can be
modelled with much longer time steps than surface
water flows but, if both surface and groundwater bodies
interact, some way has to be developed to make these
different time scales compatible.
Figure 11.1 attempts to identify the range of spatial
and temporal scales usually considered by those who
model these various components and processes. Much
depends, however, on just what impacts are to be identi-
fied and on which variables are considered unknown and
which are assumed known. For example, it is not uncom-
mon to model water supply systems that include reser-
voirs and diversions to water consumers on a monthly or
even seasonal time-step scale. However, if it is important
to capture the within-month changes in reservoir storage
for recreation or hydropower, smaller time steps may be
necessary during those months or seasons where the vari-
ation in storage and flows is important. If water quality
impacts are being estimated, time steps of a day or less
may be needed, depending on the detail or precision
desired. Some watershed processes, such as erosion and
evapotranspiration, are best modelled on an hourly or
minute-by-minute basis if sufficient data are available to
justify those short time steps, whereas other watershed
processes, such as changing ownership and land cover,
may be modelled using ten-year or longer time steps.
Much depends on the variability of both the meteorolog-
ical and hydrological inputs (air temperatures, radiation,
precipitation and streamflows) and the output targets
(demands). If there is not much variation during a partic-
ular season, it may not be necessary to define smaller time
periods within that season.
Figure 11.1 shows the ranges of spatial and temporal
scales usually considered when modelling the processes
taking place in the three major river basin components
watersheds, surface water bodies and groundwater
aquifers. There are exceptions. The boundaries of each
zone shown in Figure 11.1 are not as clearly defined as
the figure would suggest.
To provide some perspective on spatial extent of some
rivers and their watersheds, Table 11.1 lists a few of the
perspective of IWRM, quantity and quality aspects should
be considered jointly. However, unless water quantity
management strategies are based on requirements for
water quality, such as for the dilution of pollutants, water
quality measures do not normally affect water quantity.
For this reason it is common to separate analytical
approaches of water quantity management from those of
water quality management. However, when attempting to
predict the impacts of any management policy on both
water quantity and quality, both quantity and quality
models are needed.
This chapter begins with an introduction on the scales
of river basin processes and an overview of modelling
approaches. Section 2 describes the modelling of the
natural resources system (see also Figure 1.19 in Chapter 1).
This includes the estimation of unregulated surface and
groundwater runoff from watersheds, as well as the
prediction of streamflows at various sites of interest
throughout a basin. This section also includes several
methods for estimating reservoir storage requirements
for water supplies and flood storage. Reservoirs can
provide for water supply, flood control, recreation and
hydroelectric power generation. These use-functions
and their modelling are described in Section 3 on socio-
economic functions. The natural resource system and the
socio-economic functions are then, in Section 4, com-
bined into a multiple-purpose multi-objective planning
model for a river basin. The chapter concludes with an
introduction to some dynamic models for assisting in the
scheduling and time sequencing of multiple projects
within a river basin.
1.1. Scales of River Basin Processes
One of the challenges in constructing integrated models
of multiple river basin processes is the wide range of
spatial and temporal scales that characterize these
processes. Precipitation, which is the source of virtually
all freshwater in the hydrological cycle, is typically highly
variable and uneven in its distribution over time and
space. Similarly, the rates of evaporation and transpira-
tion vary considerably according to climatic and land-
cover conditions. The relative magnitudes of the fluxes
associated with individual components of the hydrologi-
cal cycle, such as evapotranspiration, may differ signifi-
cantly even at small spatial scales such as on a commercial
shopping mall, an agricultural field and a woodland.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 326
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 327
worlds major rivers with their watershed areas and
lengths.
1.2. Model Time Periods
When analysing and evaluating various water manage-
ment plans designed to distribute the natural unregulated
flows over time and space, it is usually sufficient to
consider average conditions within discrete time periods.
Commonly in river basin modelling, weekly, monthly or
seasonal flows are used as opposed to daily flows. The
shortest time-period duration usually considered in
models developed for planning analyses is one that is no
less than the time water takes to flow from the upper to
the lower end of the river basin. In this case stream and
river flows can be defined by simple mass-balance or
continuity equations. For shorter duration time periods,
some kind of flow routing may be required.
The actual length of each within-year period defined in
a model may vary from period to period. The appropriate
number and durations of modelled time periods will
depend, in part, on the variation of the supply of natural
flows and the variation of demands for water flows or
storage volumes.
Another important factor to consider in making a deci-
sion as to the number and duration of periods to include
in any model is the purpose for which the model is to be
used. Some analyses are concerned only with identifying
the desirable designs and operating policies of various
engineering projects for managing water resources at
some fixed time (a year, for instance) in the future.
Multiple years of hydrological records are simulated to
obtain an estimate of just how well a system might per-
form, at least in a statistical sense, in that future time
period. The within-year period durations can have an
impact on those performance indicator values, as well as
on the estimate of over-year as well as within-year storage
requirements that may be needed. These static analyses
are not concerned with investment project scheduling or
sequencing. Dynamic planning models include changing
economic, environmental and other objectives and design
and operating parameters over time. As a result, dynamic
Figure 11.1. Common spatial and
temporal scales of models of various river
basin processes.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 327
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
models generally span many more years than do static
models, but they may have fewer within-year periods.
1.3. Modelling Approaches For River Basin
Management
Chapter 3 presented a general overview of modelling
methods. The distinction made in that chapter between
simulation and optimization approaches applies also to
328 Water Resources Systems Planning and Management
river basin planning. The complexity involved in river
basin planning requires both approaches. In the design
phase of planning studies, optimization methods are often
useful for making a first selection among the countless
options for capacities and management measures.
However, applying optimization requires simplifications,
in particular with respect to spatial and temporal detail.
For this reason, simulation models should be applied to
check and refine the infrastructural designs and operating
policies as well as to develop detailed management
options. Reference is made to Chapter 3 (Section 3.4) for
a more detailed discussion of simulation and optimization
models.
Both optimization and simulation modelling approaches
require a description of the natural resource system and
related infrastructure, and a description of the socio-
economic functions in the river basin. These model
components will be introduced in the next two sections. In
Section 4 these two components are combined for more
comprehensive river basin analyses.
2. Modelling the Natural Resource
System and Related Infrastructure
The natural resource system (NRS) comprises all natural
physical, chemical and biological aspects and related
artificial elements of the water system. This chapter
focuses on the quantitative aspects; water quality and
ecology will be described in more detail in Chapter 12. The
hydrological surface water and groundwater system will be
described, as well as the function of reservoirs to augment
the supply in periods of low rainfall and river flows and
to contain flood flows in periods of high river flows.
2.1. Watershed Hydrological Models
Hydrological modelling is used to predict runoff from
land areas, infiltration into soils and percolation into
aquifers. Rainfallrunoff models are often used when
streamflow gauge data are not available or not reliable,
or cannot be made representative of natural flow condi-
tions (that is, what the flows in a stream or river would
be without upstream diversions or reservoirs that alter
the flows downstream). They are also used to provide
estimates of the impact that changing land uses and land
covers have on the temporal and spatial distribution of
runoff.
Amazon
Congo
La Plata
Ob Asia
Mississippi
Nile
Yenisei
Lena
Niger
Amur
Yangtze
Mackenzie
Gangas
Volga
Zambezi
St.Lawrence
Orinoco
Indus
Yukon
Danube
Mekong
Hwang Ho
Columbia
Kolyma
Sao-Francisco
Dnepr
Chutsyan
Indigirka
N.Dvina
Pechora
Godavari
Neva
Magdalena
Krishna
Fraser
Ogowe
Essequibo
Sanaga
Narmada
Rhine
Ebro
Atrato
San Juan
river length continent
km 10 km
3 2
E
0
2
0
1
1
8
b
area
S. America
Africa
S. America
Asia
N. America
Africa
Asia
Asia
Africa
Asia
Asia
N. America
Asia
Europe
Africa
N. America
S. America
Asia
N. America
Europe
Asia
Asia
N. America
Asia
S. America
Europe
Asia
Asia
Europe
Europe
Asia
Europe
S. America
Asia
N. America
Africa
S. America
Africa
Asia
Europe
Europe
S. America
S. America
6915.0
3680.0
3100.0
2990.0
2980.0
2870.0
2580.0
2490.0
2090.0
1855.0
1800.0
1790.0
1730.0
1380.0
1330.0
1030.0
1000.0
960.0
850.0
817.0
810.0
745.0
668.0
647.0
623.0
504.0
437.0
360.0
357.0
322.0
314.0
281.0
260.0
256.0
233.0
210.0
155.0
135.0
102.0
99.0
86.8
32.2
21.5
6280
4370
4700
3650
6420
6670
3490
4400
4160
2820
5520
5472
3000
3350
2660
3060
2740
3180
3000
2860
4500
4670
1950
2130
2800
2200
2130
1726
744
1810
1500
74
1530
1290
1370
850
970
860
1300
810
930
644
430
Table 11.1. Some of the worlds major rivers and their spatial
characteristics.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 328
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 329
2.1.1. Classification of Hydrological Models
Hydrological models are classified as either theoretical or
empirical models. A theoretical model is based on physical
principles. If all the governing physical processes are
described by mathematical functions, a model containing
those functions is a physically based model. However,
most existing hydrological models simplify the physics
and often include empirical components. For example, the
conservation of momentum equation or Mannings equa-
tion for predicting surface flow include empirical
hydraulic resistance terms. Darcys equation, used to pre-
dict subsurface flows, requires an empirical hydraulic con-
ductivity parameter value. Thus they are considered at
least partially, if not fully, empirical. Purely empirical or
statistical models (Chapter 6) omit the physics and are in
reality representations of the observed data.
Depending on the character of the results obtained,
hydrological models can also be classified as deterministic
or stochastic. If one or more of the variables in a mathe-
matical model are regarded as random variables whose
values can change unpredictably over time, then the
model is stochastic. If all the variables are considered to
be free from random variation, the model is deterministic.
Of course some deterministic models may include
stochastic processes that capture some of the spatial and
temporal variability of some of the sub-processes, such as
infiltration. Most hydrological models are deterministic in
spite of a host of random processes taking place in the
watersheds to which they are applied. The rather simple
modelling approach outlined below for estimating the
relative surface and groundwater runoff to surface water
bodies is an example of a deterministic model.
Hydrological models can also be classified as event-
based models or as continuous-time models. An event-based
model simulates a single runoff event, such as a single
storm, usually occurring over a period of time ranging
from about an hour to several days. The initial conditions
in the watershed for each event must be assumed or deter-
mined by other means and supplied as input data. The
accuracy of the model output may depend on the reliabil-
ity of these initial conditions. A continuous-time water-
shed model includes a sequence of time periods and for
each period determines the state of the watershed, whether
or not any events take place that will produce surface
runoff. The model keeps a continuous account of the
watershed surface and groundwater conditions. The effect
of any assumed initial conditions decreases rapidly as time
advances. Most continuous watershed models include
three water balances one for surface water, another for
unsaturated zone moisture content, and a third for
groundwater. An event model may omit one or both of the
subsurface components and also evapotranspiration when
those losses are small in relation to the surface runoff.
Finally, models that have been developed to simulate
hydrological processes can be classified as either lumped
or distributed or a mix of both. Lumped models assume
homogenous or average conditions over all or portions of
a watershed. Lumped models are not sensitive to the
actual locations of the varying features in the watershed.
Distributed models take into account the locations of
various watershed conditions such as land covers, soil
types and topography when estimating the total runoff.
Both types of models are useful. Distributed models
require more detailed data, but are needed if, for example,
one wishes to evaluate the impacts of riparian (tree-belt)
buffers along streambanks, or the effects of varying
topographic features within the watershed. Some models
are mixes of the two types of models, in other words,
quasi- or semi-distributed models made up of multiple
connected lumped models representing different parts of
watersheds.
2.1.2. Hydrological Processes: Surface Water
The manner and detail in which various hydrological
processes are modelled varies among different models
and their computer programs. Models that are developed
to better understand the physical and other natural
processes taking place on and under the surface of a
watershed are usually much more detailed than models
used for sizing drainage ditches or culverts or for evaluat-
ing alternative watershed land use management policies
or practices. Here only the basic processes will be
outlined. This will identify the minimum types of data
needed to model watershed runoff processes.
Figure 11.2 provides a schematic of the hydrological
cycle, identifying the processes that are either inputs,
such as precipitation and air temperature, and those that
are modelled, such as snowmelt or freezing, infiltration,
transpiration, percolation, evaporation, and surface and
groundwater runoff.
The surface water runoff from watershed land surfaces
is affected by many factors. Among them are the amounts
and intensity of precipitation and its rain and snow or
ice content, the land cover, soil type and slope, and the
wrm_ch11.qxd 8/31/2005 11:57 AM Page 329
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
infiltration, evaporation and transpiration that take place.
Infiltration, evaporation and transpiration rates are
dependent on the land cover, soil type, soil moisture
content and air temperature. Both the erosion that takes
place and the sediment in the runoff depend on the
intensity of precipitation, the land cover, the soil type and
the slope. The concentration of chemical constituents in
the runoff will depend on the accumulation of chemicals
on the land at the time of a storm, the extent they dissolve
in the surface water and the extent they are attached
to the soil particles that are contained in the runoff.
Here, we focus only on the runoff itself, as depicted in
Figure 11.2.
The groundwater contribution of runoff to surface
water bodies will depend on the relationship between the
groundwater and the surface water body. Groundwater
runoff can continue long after the last storm may have
ended.
Rainfall runoff processes are complex, and depend on
the particular physical and biological characteristics of
each watershed. One of the difficulties in developing
accurate predictive models of rainfallrunoff relationships
is the considerable amount of data one must collect to
understand just what processes are taking place at rather
330 Water Resources Systems Planning and Management
small spatial and temporal scales and then quantify these
processes. Very often such data are just not available, yet
we still need some way to predict, at least in some relative
sense, the quantities and qualities of surface runoff asso-
ciated with alternative land use policies or practices and
precipitation events.
Ideally, one would like to have continuous records of
net precipitation and air temperature for each watershed
or sub-watershed being modelled. Usually the hydrological
and meteorological data generally available are daily
averages. Thus most continuous-time rainfallrunoff
models use daily time steps, accepting that this does not
capture the intensities of shorter duration storms during
the day. A half-hour cloud burst from a thunderstorm
over a portion of a modelled watershed may generate
much more surface runoff (and sediment loads) than the
same amount of precipitation evenly distributed over the
entire watershed being modelled over a twenty-four-hour
period. The use of daily averages over larger areas will not
predict runoff accurately, but they are often the most
detailed data one can expect to obtain in practice. More
detailed rainfall data can sometimes be obtained from
radar images, but the processes of calibrating and
analysing such images are relatively expensive.
lake/stream
run-off
run-off
well
groundwater flow to lakes and streams
u
n
s
a
t
u
r
a
t
e
d

z
o
n
e
s
a
t
u
r
a
t
e
d

z
o
n
e e
v
a
p
o
r
a
t
i
o
n t
r
a
n
s
p
i
r
a
t
i
o
n
i
n
f
i
l
t
r
a
t
i
o
n
p
r
e
c
ip
i
t
a
t
i
o
n
w
t
e
a
r
ta
b
le
E
0
2
0
1
1
8
c
p
e
r
c
o
l
a
t
i
o
n
Figure 11.2. A schematic
diagram of the hydrological
cycle applicable to a watershed.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 330
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 331
Precipitation
Precipitation can be in liquid and/or frozen form,
depending of course on the location and time of year. In
this discussion the term rain will denote liquid precipita-
tion and snow will denote frozen precipitation (which
could include ice as well). The proportion of rain and
snow in precipitation will depend on the temperature of
the air through which the precipitation falls. The
proportion of rain and snow, if not directly measured,
usually has to be estimated from average daily surface air
temperature data, since these are the most detailed data
available. Clearly air temperatures will vary over a day, as
well as with altitude. The range of sub-freezing air
temperatures over time and at different altitudes can
affect the proportion of snow and ice in precipitation as
well as the fraction of water content in the snow and ice.
Estimates of these proportions, often based on average
daily surface temperatures at one or at best only a few
measuring stations, are sources of error when direct
measurements of snow depth (water equivalent) and rain
are not available.
The total net depth of precipitation, P
t
, in day t, equals
the sum of the net rain R
t
and snow S
t
(water equivalent)
depths that reach the ground. These are input data for
watershed runoff models.
Surface Water
The water available for surface runoff during any day is
the free water already on the ground surface plus the rain
and the snowmelt less any losses or reductions due to
infiltration and evaporation and any freezing of standing
surface water. The depth of snow that melts or the
amount of freezing of free-standing water on the grounds
surface will depend on the depth of snow available for
melting, or on the depth of free surface water available for
freezing, and on the air temperature.
Total snow depth accumulation, measured in water
content, is determined from a mass balance of snow in the
watershed. The initial snow depth, SD
t
, at the beginning
of day t plus the added depth of snow precipitation, S
t
,
and the depth of surface water that freezes on the surface,
WF
t
, less the snowmelt, SM
t
, equals the final snow depth,
SD
t1
, at the end of each day t:
SD
t1
SD
t
S
t
WF
t
SM
t
(11.1)
The actual daily snowmelt, SM
t
, will be the maximum
snowmelt associated with the average air temperature of
that day, SM
t
max
, or the actual amount of snow available
for melting, SD
t
S
t
:
SM
t
min{SM
t
max
, SD
t
S
t
} (11.2)
Similarly, for the freezing of surface water, SW
t
and
rainfall R
t
,
WF
t
min{WF
t
max
, SW
t
R
t
} (11.3)
One would not expect melting and freezing to occur on the
same day in a model based on average daily temperatures.
The net amount of water available for runoff, AW
t
,
before other losses are considered, equals the depth avail-
able at the beginning of the day, SW
t
, plus the rainfall
depth, R
t
, and snowmelt depth, SM
t
, less the freezing
depth, WF
t
:
AW
t
SW
t
R
t
SM
t
WF
t
(11.4)
This amount of water is reduced by that which infiltrates
into the unsaturated soil, if any, and by that which
evaporates.
Infiltration and Percolation
Infiltration may occur if the ground is not frozen or com-
pletely saturated. The ground is usually considered frozen
if snow is on top of it (that is, the accumulated snow
depth is greater than zero). It could also be frozen after
several days of air temperatures below 0, even if no snow
depth exists. In these cases infiltration is assumed to be
zero. If infiltration can occur, the amount will depend on
the soil cover, type and moisture content, as well as on
the available water to infiltrate if that amount is less than
the maximum infiltration rate over a day. Different soil
types have different effective porosities (fraction of total
soil volume occupied by pores that water can enter), and
different maximum rates of infiltration depending, in
part, on the soil moisture content.
The maximum soil moisture content of unsaturated
soil is reached when its available pore space is occupied
by water. This amount, expressed as depth, is equal to
the effective porosity of the soil times the depth of the
unsaturated zone, UD
t
. Expressing soil moisture content
variable, MC
t
, as a fraction of the soils maximum capacity,
Depth of soil moisture at beginning of day t,
MC
t
(effective porosity)(UD
t
) (11.5)
wrm_ch11.qxd 8/31/2005 11:57 AM Page 331
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The average depth of the unsaturated zone, UD
t
,
approximates the difference between the average water-
shed surface elevation, WE, and the average unconfined
groundwater table elevation, GE
t
:
UD
t
WE GE
t
(11.6)
The infiltration depth, I
t
, occurring when the ground is not
frozen and free of snow depth, will be the maximum rate,
I
max
, for the soil type, the soil moisture deficit, or the water
depth available for infiltration, AW
t
, whichever is less.
I
t
min{AW
t
, I
max
, [(effective porosity)(UD
t
)(1MC
t
)]}
if soil not frozen and SD
t
0
0 otherwise (11.7)
The depth of percolation of water, UP
t
, in the unsaturated
soil zone to the groundwater aquifer is assumed to equal
either the maximum daily percolation depth, P
max
, for the
soil type or the available moisture content down to some
minimum percentage, P
min
, of the maximum amount,
whichever is less. Often P
max
is assumed to equal the
maximum infiltration rate, I
max
, for lack of any better
assumption or measured data.
UP
t
min{P
max
, max[0, (MC
t
P
min
)
(effective porosity)(UD
t
)]} (11.8)
Transpiration and Evaporation
Transpiration depends on the moisture content in the
unsaturated soil, the land cover, the depth of plant roots
and the air temperature. Land cover and root depth may
vary over a year.
Plant transpiration, T
t
, reduces the soil moisture con-
tent of the unsaturated zone if the depth of that zone,
UD
t
, exceeds the average root zone depth, RZ
t
. If the
unsaturated zone depth is less than the average root
zone depth, then roots have access to the groundwater
and the transpiration will reduce the groundwater table.
In this case, the transpiration from groundwater, TG
t
, is
usually assumed equal to the maximum transpiration
rate, T
t
max
.
Some watershed rainfallrunoff models keep track of
the root depths, especially in agricultural croplands, and
their impact on soil moisture and transpiration.
Alternatively, one can assume transpiration will equal its
maximum rate, T
t
max
, for the given land cover and air
temperature until the soil moisture content fraction, MC
t
,
332 Water Resources Systems Planning and Management
goes below some minimal moisture content fraction for
transpiration, MCT
min
. Transpiration will cease when soil
moisture contents go below that minimum percentage.
This water is not available to the plants.
T
t
min{T
t
max
, max[0, (MC
t
MCT
min
)
(effective porosity)(UD
t
) UP
t
}]
and
TG
t
is 0 if UD
t
RZ
t
(11.9)
otherwise
TG
t
T
t
max
and T
t
is 0 if UD
t
RZ
t
(11.10)
The depth of evaporation of water on the land surface is a
function of the air temperature, relative humidity and
wind velocities. Maximum daily evaporation depths are
usually assumed to be fixed values that depend on
average meteorological conditions for various seasons of
the year when more precise data or functions defining
evaporation are not available or not used. The actual
evaporation each day will be the maximum evaporation
rate for that day, E
t
max
, or the total depth of free surface
water, AW
t
, less infiltration, I
t
, whichever is less.
E
t
min{AW
t
I
t
, E
t
max
} (11.11)
Surface Runoff
The depth of surface water available for runoff, SA
t
, in
each day t equals the depth of free surface water, AW
t
,
(which results from the initial amount, SW
t
, plus rain, R
t
,
and snowmelt, SM
t
, less the amount that freezes, WF
t
, as
defined in Equation vation. 11.4) less the infiltration I
t
,
and evaporation, E
t
:
SA
t
AW
t
I
t
E
t
(11.12)
The depth of surface runoff, SR
t
, is some fraction of
this amount, depending on the slope, the extent of
ponding, the surface area and the land cover of the
watershed.
Surface runoff can be estimated in a number of ways.
(See, for example, the curve number method of USDA,
1972, as presented in Chapter 13). For this discussion,
assume Mannings Equation applies. This is an equation
for determining the velocity of overland and channel flow
based on the hydraulic radius, HR
t
, (cross-sectional area
of flow divided by the wetted perimeter), land slope and
wrm_ch11.qxd 8/31/2005 11:57 AM Page 332
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 333
ground cover. For overland sheet flow, and for channel
flow of small depths, the hydraulic radius is essentially
the depth of available surface water, SA
t
. If SA
t
is in units
of metres, the velocity, V
t
, of surface runoff (metres/second)
based on Mannings equation, equals
V
t
(1/n)(SA
t
)
2/3
(Slope)
1/2
(11.13)
The parameter n represents an effective roughness coeffi-
cient that includes the effect of raindrop impact, drag over
the ground surface from obstacles such as litter, crop
ridges and rocks, and erosion and transportation of
sediments. This roughness coefficient will be the major
calibration parameter of the runoff model.
Watershed runoff will be a combination of sheet flow
and channel flow. Typical n values for sheet flow range
from 0.011 for smooth paved surfaces to 0.10.4 for grass
to 0.40.8 for underbrush in wooded areas. These n
values are for very shallow flow depths of about 2050
millimetres or so. For river channels the n values typically
range between 0.02 and 0.07 (USDA-SCS, 1972).
The maximum surface water flow, Q
t
max
, equals
the velocity times the cross-sectional area of the flow
perpendicular to the direction of flow. The cross-sectional
area is the depth of available water, SA
t
, times the average
width, W, of the watershed. If the units of width are
expressed in metres, the maximum flow in cubic metres
per second is
Q
t
max
(1/n)(SA
t
)
5/3
(Slope)
1/2
W (11.14)
The total quantity of surface water runoff, SR
t
, from
the watershed in day t will be this maximum flow rate
multiplied by the number of seconds in a day or the total
amount of water available for runoff, whichever is less.
SR
t
min{Q
t
max
(60)(60)(24),(SA
t
)(Watershed area)}
(11.15)
In reality, most watersheds are not uniform, flat, tilted
landscapes where sheet flow takes place. The water
available for runoff will either seep into the upper soil
layer and reappear down slope or flow into small chan-
nels. The above equations based on average watershed
conditions are thus a surrogate for what actually takes
place at much smaller space scales than are reasonable to
model.
2.1.3. Hydrological Processes: Groundwater
Groundwater Runoff to Streams and Rivers
To estimate the groundwater contribution of the total
runoff, it is necessary to model the surface waterground-
water interaction. As shown in Figure 11.3, groundwater
can move along flow paths of varying lengths from areas
of recharge to areas of discharge. The generalized flow
paths in Figure 11.3 start at the water table of the upper
unconfined aquifer and continue through the groundwa-
ter system, terminating at the surface water body. In
the uppermost, unconfined aquifer, flow paths near the
unconfined aquifer
water table
confining bed
confining bed
confined aquifer
confined aquifer
well
days
centuries
millennia
years
E
0
2
0
1
1
8
d
Figure 11.3. Groundwater
surface water interactions over a
range of time and space scales.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 333
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
stream can be tens to hundreds of metres in length and
have corresponding travel times ranging from days to
several years. The longest and deepest flow paths in
Figure 11.3 may be thousands of kilometres in length,
and travel times may range from decades to millennia. In
334 Water Resources Systems Planning and Management
general, shallow groundwater is more susceptible to
contamination from human sources and activities because
of its close proximity to the land surface.
Streams interact with groundwater in three basic
ways. They gain water from inflow of groundwater
through the streambed (gaining stream, Figure 11.5A),
they lose water to groundwater by outflow through the
streambed (losing stream, Figure 11.6A), or they do
both, gaining in some reaches and losing in others. For
groundwater to flow into a stream channel, the elevation
of the groundwater table in the vicinity of the stream
must be higher than that of the stream-water surface.
Conversely, for surface water to seep to groundwater, the
elevation of the water table in the vicinity of the stream
must be lower than that of the stream-water surface.
Contours of water-table elevation indicate gaining
streams by pointing in an upstream direction (Figure
11.5B), and indicate losing streams by pointing in a
downstream direction (Figure 11.6B) in the immediate
vicinity of the stream.
Losing streams can be connected to the groundwater
system by a continuous saturated zone (Figure 11.6A)
or can be isolated from the groundwater system by an
land surface
water table
surface water
groundwater flow path
E
0
2
0
1
1
8
e
Figure 11.4. Groundwater seepage into surface water usually
is greatest near shore. In flow diagrams such as that shown
here, the quantity of discharge is equal between any two flow
lines; therefore, the closer flow lines indicate greater
discharge per unit of bottom area.
unsaturated zone
water table
shallow aquifer
flow direction
70
60
50
40
30
20
water table countour
stream ground water flow line
E
0
2
0
1
2
1
a
Figure 11.5. Gaining streams receive water from
the groundwater system (A). This can be
determined from water-table contour maps
because the contour lines point in the upstream
direction where they cross the stream (B).
wrm_ch11.qxd 8/31/2005 11:57 AM Page 334
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 335
unsaturated zone. In the latter case, infiltration and
percolation processes apply. Pumping of shallow ground-
waters near disconnected streams does not affect the flow
of water in those streams near the pumped wells.
Given the complexity of the interactions between
surface water and groundwater at specific sites under
different conditions, it becomes difficult to model these
interactions without considerable site-specific data based
on many detailed observations and measurements.
As a first approximation, Darcys equation for laminar
(slowly moving) flow in saturated porous soils can be used
to estimate the flow either from the surface water to the
groundwater aquifer, or vice versa. This flow depends on
the hydraulic conductivity of the soil, K (having dimen-
sions of length per unit time, L/T), and the area through
which the flow occurs, A (L
2
). It also depends on the
groundwater flow head gradient, dH/dX (the change in
groundwater head H per change in distance X perpendi-
cular to the stream or river reach). Darcys equation states
that the groundwater flow is the product of KA(dH/dX).
Let (dH/dX)
L
and (dH/dX)
R
be the groundwater head
gradients in the direction of groundwater flow on the left
and right side of the stream or river reach, respectively.
Using Darcys equation, the net flow, Q
sg
, from surface
water stream or river reach to the unconfined groundwater
aquifer is the sum of the products of the hydraulic con-
ductivity K, the area A, and the groundwater head
gradient, dH/dX, for each side of the stream or river reach.
Q
sg
K
L
A
L
(dH/dX)
L
K
R
A
R
(dH/dX)
R
(11.16)
If Q
sg
is negative, the net flow is from the groundwater
aquifer to the surface water reach. Clearly, no more water
can flow from one water body to another than the volume
of water in the former. Models that use equations such as
Equation 11.16 need to check on this condition.
Groundwater Runoff to Lakes, Reservoirs and
Wetlands
Surface water in lakes, wetlands and reservoirs interacts
with groundwater rather the same way that streams and
rivers do, as illustrated in Figure 11.7. Although the basic
interactions are the same for these storage water bodies as
they are for streams, however, the specific interactions
differ in several ways.
unsaturated zone
water table
shallow aquifer
flow direction
E
0
2
0
1
2
1
b
stream
ground water flow line
water table countour
90
80
70
100
Figure 11.6. Losing streams lose water to the
groundwater system (A). This can be determined
from water-table contour maps because the
contour lines point in the downstream direction
where they cross the stream (B).
wrm_ch11.qxd 8/31/2005 11:57 AM Page 335
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The water levels of natural lakes are not controlled by
dams. Hence they generally do not change as rapidly as
the water levels of streams, and bank storage is therefore
of lesser importance in lakes than in streams.
Furthermore, lake sediments commonly have greater
volumes of organic deposits. These less permeable
organic deposits reduce the seepage and biogeochemical
exchanges of water and solutes that take place in lakes
compared with those which take place in streams.
Reservoirs are human-made lakes that are designed
primarily to control the flow and distribution of surface
water. Most reservoirs are constructed in stream valleys,
and therefore have some characteristics of both streams and
lakes. Like streams, reservoirs can have widely fluctuating
levels, bank storage can be significant, and they commonly
have a continuous flushing of water through them.
Wetlands that occupy depressions in the land surface
have interactions with groundwater similar to lakes and
streams. They can receive groundwater inflow, recharge
groundwater, or do both. Unlike streams and lakes,
336 Water Resources Systems Planning and Management
however, wetlands do not always occupy low points and
depressions in the landscape (Figure 11.8A). They can be
present on slopes (fens, for example) or even on drainage
divides (some types of bogs). Fens are wetlands that com-
monly receive most of their water from groundwater dis-
charge (Figure 11.8B). Bogs are wetlands that occupy
uplands (Figure 11.8D) or extensive flat areas, and they
receive much of their water from precipitation. While
riverine wetlands (Figure 11.8C) commonly receive
groundwater discharge, they are dependent primarily on
the stream for their water supply.
A major difference between lakes and wetlands, with
respect to their interaction with groundwater, is the ease
with which water moves through their beds. Lakes are
commonly shallow around their perimeter, where waves
can remove fine-grained sediments, permitting the
surface water and groundwater to interact freely.
Wetlands, on the other hand, typically contain fine-
grained and highly decomposed organic sediments near
the wetland edge, slowing the transfer of water and
solutes between groundwater and surface water.
Another difference in the interaction between ground-
water and surface water in wetlands compared with lakes
is the extent of rooted vegetation in wetlands. The fibrous
root mat in wetland soils is highly conductive to water
flow; therefore, water uptake by roots of emergent plants
results in significant interchange between surface water
and pore water of wetland sediments. Water exchanges in
this upper soil zone are usually not as restricted as the
exchanges between surface water and groundwater at the
base of the wetland sediments.
All these complex interactions between ground and
surface water systems can only be approximated by
watershed models, even the most detailed ones. Without
extensive data, it is unlikely the models will capture all
the interactions that take place over space and time.
2.1.4. Modelling Groundwater: Surface Water
Interactions
One of the simplest ways of estimating the groundwater
runoff from a watershed is to assume the runoff will equal
some fraction of the average groundwater head. This
fraction is called a recession constant.
If the aquifer lies solely under the watershed of
interest, then a recession rate can be defined for estimating
lake surface
E
0
2
0
1
2
1
g
Figure 11.7. Lakes can receive groundwater inflow (A), lose
water as seepage to groundwater (B), or both (C).
wrm_ch11.qxd 8/31/2005 11:57 AM Page 336
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 337
the extent of groundwater base flow depth when the aver-
age groundwater table elevation, GE
t
, is greater than the
average surface water body elevation, SE
t
. Letting the
dimensionless recession fraction be k, the groundwater base
flow contribution, GR
t
, in day t, to total surface runoff is
GR
t
Max{0, (GE
t
SE
t
}k(aquifer area)
(effective aquifer porosity) (11.17)
The total runoff in day t equals the sum of the ground-
water runoff, GR
t
, plus the surface water runoff SR
t
.
This method does not consider situations in which
water flows from the surface water bodies to the ground-
water aquifer.
Alternatively, the two-way interactions between the
ground and surface waters, perhaps involving multiple
watersheds and one or more underlying unconfined
area favorable for wetland formation
zone of low permeability zone of high permeability
area favorable
for wetland formation
seepage face
break in slope
direction of groundwater flow
wetland
wetland
stream
E
0
2
0
1
2
1
h
Figure 11.8. The source of water to
wetlands can be from groundwater
discharge where the (light-
coloured) land surface is underlain
by complex groundwater flow fields
(A), from groundwater discharge at
seepage faces and at breaks in
slope of the water table (B), from
streams (C), and from precipitation
in cases where wetlands have no
stream inflow and groundwater
gradients slope away from the
wetland (D).
wrm_ch11.qxd 8/31/2005 11:57 AM Page 337
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
aquifers, can be modelled using Darcys equation
(Equation 11.17).
If the groundwater aquifer is not being modelled in
detail, all one may have available from a lumped water-
shed model is the average groundwater table elevation or
head at the beginning or end of any time period. This
average head is usually based on the groundwater
volume. For an unconfined aquifer, the average head,
H
t
, at the beginning of day t, will equal the groundwater
volume, GWV
t
, divided by the product of the aquifer area
times its effective porosity. The head gradient is the
difference between the average groundwater elevation
(head), GE
t
, and the surface elevation, SE
t
, of the
surface water body, divided by some representative
distance X.
dH
t
/dX [GE
t
SE
t
]/X (11.18)
338 Water Resources Systems Planning and Management
Many lumped models of groundwater aquifers approxi-
mate the head gradient by dividing the elevation
difference between the average groundwater table
elevation, GE
t
, and the surface water body elevation, SE
t
,
by no more than a quarter of the distance from the
surface water body and the aquifer boundary. This is
shown in Figure 11.9 for water-gaining surface water
bodies. In this figure, the Wterm is equivalent to the X
terms use in Equations 11.18 and 11.19.
The approximate area, A
t
, through which the ground-
water flows is the average of the surface water body
elevation above the base of the aquifer, SE
t
, and the average
groundwater head elevation, GE
t
, times the length, L, of
the surface water body in its direction of flow. Assuming
groundwater flows from both sides of the river into the
river and the K and X (or W in Figure 11.9) terms are
the same for both sides of the river, Darcys equation for
watershed
main river
tributary stream
drainage divide
sub-watershed, draining into a river reach of length L
W = width of aquifer
river
baseflow
W
H
t
SE
t
GE
t
E
0
2
0
1
2
1
j
Figure 11.9. Cross-sectional
dimensions of a watershed
showing a surface water body
that is receiving water from the
groundwater aquifer.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 338
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 339
estimating the groundwater runoff, GR
t
from this portion
of the watershed is:
GR
t
2K((SE
t
GE
t
)/2)(L)[GE
t
SE
t
]/X
K(GE
t
2
SE
t
2
)(L)/X (11.19)
This equation can apply to any surface water body receiv-
ing runoff from groundwater. It can also apply to any
groundwater aquifer receiving water from a surface
water body. In this case, GR
t
would be negative. As for
Equation 11.16, no more water can flow from one water
body to another than the volume of water in the former.
Again, models that use equations such as Equation 11.19
need to check on this condition.
2.1.5. Streamflow Estimation
Water resources managers need to have estimates of
streamflows at each site where management decisions are
to be made. These streamflows can be based on the
results of rainfall runoff models or on measured histori-
cal flows at stream gauges. For modelling alternative
management policies, these streamflows should be those
that would have occurred under natural conditions.
These are called naturalized flows. These naturalized
flows are often computed from observed gauge flows
adjusted to take into account any upstream regulation
and diversions. Many gauge flow measurements reflect
actions taken upstream that alter the natural flows, such
as diversions and reservoir storage. Unless such upstream
water management and use policies are to continue,
these regulated flows should be converted to unregulated
or natural flows prior to their use in management
models.
Alternatively, naturalized flows can be estimated from
rainfallrunoff modelling.
RainfallRunoff Modelling
In rainfallrunoff models, mass balances must be main-
tained for snow depth, for free surface water depth, for
the moisture content of the unsaturated zone, and for the
groundwater volume, as applicable.
For surface water, the remaining free surface water will
be that water which is left at the end of period t. The free
water remaining on the watersheds surface at the end of
day t, SW
t1
, is what exists at the beginning of day t1.
It equals the surface water that exists at the beginning
of the day, SW
t
, plus the rainfall, R
t
, plus the snowmelt,
SM
t
, less the depth that freezes, WF
t
, the infiltration, I
t
,
the evaporation, E
t
, and the depth of surface runoff,
SR
t
/(watershed area):
SW
t1
SW
t
R
t
SM
t
WF
t
I
t
E
t
SR
t
/(watershed area)
SA
t
SR
t
/(watershed area) (11.20)
The snow depth, SD
t1
, at the end of each day t is:
SD
t1
SD
t
S
t
WF
t
SM
t
(11.21)
The final groundwater volume is equal to the initial
groundwater volume, GWV
t
, plus any additions and less
any reductions during the day. These additions and
reductions include the total net runoff to surface water
bodies, GR
t
, the percolation, UP
t
, the groundwater
transpiration, TG
t
, any groundwater abstractions or
artificial recharge, less any losses to, GWQ
out
t
, or plus any
gains from, GWQ
in
t
, deeper aquifers.
(11.22)
The average groundwater table elevation is computed
taking into account the percolation from the unsaturated
zone and its inflow from or outflow to the surface water
system. The groundwater head or elevation is the water-
shed elevation or its volume divided by its area times its
porosity, whichever is less:
GE
t1
min{WE, {GWV
t1
/[(aquifer area)
(effective porosity)]}} (11.23)
The unsaturated zone depth at the end of each day t is
UD
t1
WE GE
t1
(11.24)
The soil moisture fraction, MC
t1
, in the unsaturated soil
layer at the end of each day can be calculated as
MC
t1
[MC
t
(porosity)(UD
t
) I
t
T
t
UP
t
]/
(effective porosity)(UD
t1
) (11.25)
If the groundwater table is rising, the actual soil moisture
content will probably increase due to capillary action. If
the groundwater table is falling, the actual soil moisture
GWV GWV GR UP TG
t t t t t

1
watersheds
(watershed area)

( )

GW GW
t t
Q Q
in out
wrm_ch11.qxd 8/31/2005 11:57 AM Page 339
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
content is likely to be higher than estimated due to
the delayed percolation of the excess water from the once-
saturated zone. Here, we assume water table elevation
changes (that is, unsaturated zone depth changes) do not
by themselves change the percentage of soil moisture in
the unsaturated zone.
If the groundwater table equals the surface elevation,
then the unsaturated zone, UD, does not exist and no
infiltration and percolation will occur. In this case, the
unconfined saturated groundwater table is at the surface.
Streamflow Estimation Based on Flow Data
Assuming unregulated streamflow data are available at
gauge sites, these sites may not be where flow data
are needed. Thus, some way is needed to generate
corresponding streamflow data at sites where managers
need them, such as at potential diversion or reservoir
sites. Just how this can be done depends in part on the
durations of a models time periods.
Consider, for example, the simple river basin illus-
trated in Figure 11.10. The streamflows have been
recorded over a number of years at gauge sites 1 and 9.
Knowledge of the flows Q
t
s
in each period t at gauge sites
s 1 and 9 permits the estimation of flows at any other
site in the basin as well as the incremental flows between
those sites in each period t.
The method used to estimate flows at ungauged sites
will depend on the characteristics of the watershed of the
river basin. In humid regions where streamflows increase
in the downstream direction, and the spatial distribution
of average monthly or seasonal rainfall is more or less the
same from one part of the river basin to another, the
340 Water Resources Systems Planning and Management
runoff per unit land area is assumed constant over space.
In these situations, estimated flows, q
t
s
, at any site s are
usually based on the watershed areas, A
s
, contributing
flow to those sites, and the corresponding streamflows and
watershed areas above the nearest or most representative
gauge sites.
For each gauge site, the runoff per unit land area can
be calculated by dividing the gauge flow, Q
t
s
, by the
upstream drainage area, A
s
. This can be done for each
gauge site in the basin. Thus for any gauge site g, the
runoff per unit drainage area in month or season t is Q
t
g
divided by A
g
. This runoff per unit land area times the
drainage area upstream of any site s of interest will be the
estimated streamflow in that period at that site s. If there
are multiple gauge sites, as illustrated in Figure 11.10, the
estimated streamflow at some ungauged site s can be a
weighted linear combination of those unit area runoffs
times the area contributing to the flow at site s. The
non-negative weights, w
g
, that sum to 1 reflect the relative
significance of each gauge site with respect to site s. Their
values will be based on the judgements of those who are
familiar with the basins hydrology.
(11.26)
In all the models developed and discussed below, the
variable Q
t
s
will refer to the mean natural (unregulated)
flow (L
3
/T) at a site s in a period t.
The difference between the natural streamflows at any
two sites is called the incremental flow. Using Equation 11.26
to estimate streamflows will result in positive incremental
flows. The downstream flow will be greater than the
upstream flow. In arid regions, incremental flows may
not exist and hence, due to losses, the flows may be
decreasing in the downstream direction. In these cases,
there is a net loss in flow in the downstream direction.
This might be the case when a stream originates in a wet
area and flows into a region that receives little if any rain-
fall. In such arid areas the runoff into the stream, if any,
is less than the evapotranspiration and infiltration into the
ground along the stream channel.
For stream channels where there exist relatively uniform
conditions affecting water loss and where there are no
known sites where the stream abruptly enters or exits the
ground (as can occur in karst conditions), the average
Q Q
t
s
g t
g
g
g
s
w A A =


/
E
0
2
0
1
2
1
m
gauge site1
gauge site 9
Figure 11.10. River basin gauge sites where streamflows are
measured and recorded.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 340
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 341
streamflow for a particular period t at site s can be based on
the nearest or most representative gauge flow, Q
t
g
, and a loss
rate per unit length of the stream or river, L
gs
, between
gauge site g and an ungauged site s. If there are at least two
gauge sites along the portion of the stream or river that is
in the dry region, one can compute the loss of flow per unit
stream length, and apply this loss rate to various sites along
the stream or river. This loss rate per unit length, however,
may not be constant over the entire length between
the gauge stations, or even for all flow rates. Losses will
probably increase with increasing flows simply because
more water surface is exposed to evaporation and seepage.
In these cases one can define a loss rate per unit length of
stream or river as a function of the magnitude of flow.
In watersheds characterized by significant elevation
changes, and consequently varying rainfall and runoff dis-
tributions, other methods may be required for estimating
average streamflows at ungauged sites. The selection of
the most appropriate method to use, as well as the most
appropriate gauge sites to use, to estimate the streamflow,
Q
t
s
, at a particular site s can be a matter of judgement. The
best gauge site need not necessarily be the nearest gauge
to site s, but rather the site most similar with respect to
important hydrological variables.
2.1.6. Streamflow Routing
If the duration of a within-year period is less than the time
of flow throughout the stream or river system being
modelled, and the flows vary within the system, then some
type of streamflow routing must be used to keep track
of where the varying amounts of water are in each time
period. There are many proposed routing methods (as
described in any hydrology text or handbook, e.g.
Maidment, 1993). Many of these more traditional methods
can be approximated with sufficient accuracy by relatively
simple methods. Two such methods are described in the
following paragraphs.
The outflow, O
t
, from a reach of stream or river
during a time period t is a function of the amount of water
in that reach, i.e. its initial storage, S
t
, and its inflow, I
t
.
Because of bank storage, that outflow is often dependent
on whether the quantity of water in the reach is
increasing or decreasing. If bank inflows and outflows
are explicitly modelled, for example as described in
Sections 2.1.3 and 2.1.4, or if bank storage is not that
significant, then the outflow from a reach in any period t
can be expressed as a simple two-parameter power func-
tion of the form a(S
t
I
t
)
b
. Mass-balance equations,
which may take losses into account, update the initial
storage volumes in each succeeding time period. The
reach-dependent parameters a and b can be determined
through calibration procedures such as genetic algorithms
(Chapter 6), given a time series of reach inflows and
outflows. The resulting outflow function is typically
concave (the parameter b will be less than 1), and thus
the minimum value of S
t
I
t
must be at least 1. If, due
to evaporation or other losses, the reach volume drops
below this or any pre-selected higher amount, then the
outflow is assumed to be 0.
Alternatively, one can adopt a three- or four-parameter
routing approach that fits a wider range of conditions. Each
stream or river reach can be divided into a number of
segments. That number n is one of the parameters to be
determined. Each segment s can be modelled as a storage
unit, having an initial storage volume, S
st
, and an inflow, I
st
.
The three-parameter approach assumes the outflow, O
st
, is
a linear function of the initial storage volume and inflow:
O
st
S
st
I
st
(11.27a)
Equation 11.27a applies for all time periods t and for all
reach segments s in a particular reach. Different reaches
may have different values of the parameters n, and .
The calibrated values of and are non-negative and no
greater than 1. Again, a mass-balance equation updates
each segments initial storage volume in the following
time period. The outflow from each reach segment is the
inflow into the succeeding reach segment.
The four-parameter approach assumes that the
outflow, O
st
, is a non-linear function of the initial storage
volume and inflow:
O
st
, (S
st
I
st
)

(11.27b)
where the forth parameter is greater than 0 and no greater
than 1. In practice, is very close to 1. Again the values of
these parameters can be found using non-linear optimiza-
tion methods, such as genetic algorithms, together with a
time series of observed reach inflows and outflows.
Note the flexibility available when using the three- or
four-parameter routing approach. Even blocks of flow can
be routed a specified distance downstream over a speci-
fied time, regardless of the actual flow. This can be done
wrm_ch11.qxd 8/31/2005 11:57 AM Page 341
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
342 Water Resources Systems Planning and Management
each period t, reservoir storage volumes, S
t
, cannot exceed
their storage capacities, K.
S
t
K for each period t. (11.28b)
Equations 11.28a and 11.28b are the two fundamental
equations required when modelling water supply reservoirs.
They apply for each period t.
The primary purpose of all reservoirs is to provide a
means of regulating downstream surface water flows
over time and space. Other purposes may include storage
volume management for recreation and flood control,
and storage and release management for hydropower
production. Reservoirs are built to alter the natural
spatial and temporal distribution of the streamflows. The
capacity of a reservoir and its release (or operating) policy
determine the extent to which surface water flows can be
stored for later release.
The use of reservoirs for temporarily storing stream-
flows often results in a net loss of total streamflow due to
increased evaporation and seepage. Reservoirs also bring
with them changes in the ecology of a watershed and river
system. They may also displace humans and human
settlements. When considering new reservoirs, any
benefits derived from regulation of water supplies, from
floodwater storage, from hydroelectric power, and from
any navigational and recreational activities should be
compared to any ecological and social losses and costs.
The benefits of reservoirs can be substantial, but so may
the costs. Such comparisons of benefits and costs are
always challenging because of the difficulty of expressing
all such benefits and costs in a common metric. For a full
discussion on the complex issues and choices on reservoir
construction, refer to report of the World Commission on
Dams (WCD, 2000).
Reservoir storage capacity can be divided among three
major uses:
active storage used for downstream flow regulation and
for water supply, recreational development or
hydropower production
dead storage required for sediment collection
flood storage capacity reserved to reduce potential
downstream flood damage during flood events.
These separate storage capacities are illustrated in
Figure 11.11. The distribution of active and flood control
storage capacities may change over the year. For example
by setting and to 1, to 0, and the number of
segments n to the number of time periods it takes to travel
that distance. This may not be very realistic, but there
are some river basin reaches where managers believe this
particular routing applies.
Most river basin models offer a number of alternative
approaches to describe streamflow routing. Box 11.1
illustrates the methods included in the RIBASIM package
(WL | Delft Hydraulics, 2004). Which method is the best
to apply will depend on the required accuracy and the
available data.
2.2. Lakes and Reservoirs
Lakes and reservoirs are sites in a basin where surface water
storage needs to be modelled. Thus, variables defining the
water volumes at those sites must be defined. Let S
t
s
be the
initial storage volume of a lake or reservoir at site s in period
t. Omitting the site index s for the moment, the final storage
volume in period t, S
t1
, (which is the same as the initial
storage in the following period t 1) will equal the initial
volume S
t
plus the net surface and groundwater inflows, Q
t
,
less the release or discharge, R
t
, and evaporation and seep-
age losses, L
t
. All models of lakes and reservoirs include this
mass-balance equation for each period t being modelled:
S
t
Q
t
R
t
L
t
S
t1
(11.28a)
The release from a natural lake is a function of its sur-
rounding topography and its water surface elevation. It is
determined by nature, and unless it is made into a reser-
voir, its discharge or release is not controlled or managed.
The release from a reservoir is controllable, and is usually
a function of the reservoir storage volume and time of
year. Reservoirs also have fixed storage capacities, K. In
Box 11.1. Streamflow Routing in RIBASIM
Simulation Model
Streamflow routing in RIBASIM is simulated with link
storage nodes. The outflow from these nodes (actu-
ally river segments acting as reservoirs) into the next
node can be described by:
Manning formula
two-layered, multi-segmented Muskingum formula
Puls method
Laurenson non-linear lag and route method.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 342
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 343
there is no need for flood control storage in seasons that
are not going to experience floods.
Often these components of reservoir storage capacity
can be modelled separately and then added together to
determine total reservoir storage capacity. The next
several sections of this chapter address how these capaci-
ties may be determined.
2.2.1. Estimating Active Storage Capacity
Mass Diagram Analyses
Perhaps one of the earliest methods used to calculate the
active storage capacity required to meet a specified reser-
voir release, R
t
, in a sequence of periods t, was developed
by W. Rippl (1883). His mass diagram analysis is still used
today by many planners. It involves finding the maximum
positive cumulative difference between a sequence of pre-
specified (desired) reservoir releases R
t
and known
inflows Q
t
. One can visualize this as starting with a full
reservoir, and going through a sequence of simulations in
which the inflows and releases are added and subtracted
from that initial storage volume value. Doing this over
two cycles of the record of inflows will identify the maxi-
mum deficit volume associated with those inflows and
releases. This is the required reservoir storage. Having this
initial storage volume, the reservoir would always have
enough water to meet the desired releases. However, this
only works if the sum of all the desired releases does not
exceed the sum of all the inflows over the same sequence
of time periods. Reservoirs cannot make water. Indeed,
they can lose it from evaporation and seepage.
Equation 11.29 represents this process. The active
storage capacity, K
a
, will equal the maximum accumulated
storage deficit one can find over some interval of time
within two successive record periods, T.
(11.29)
where 1 i j 2T.
Equation 11.29 is the analytical equivalent of graphi-
cal procedures proposed by Rippl for finding the active
storage requirements. Two of these graphical procedures
are illustrated in Figures 11.12 and 11.13 for a nine-
period inflow record of 1, 3, 3, 5, 8, 6, 7, 2 and 1. Rippls
original approach, shown in Figure 11.12, involves
plotting the cumulative inflow
t
1
Q

versus time t.
Assuming a constant reservoir release, R
t
, in each period
t, a line with slope R
t
is placed so that it is tangent to the
cumulative inflow curve. To the right of these points of
tangency the release R
t
exceeds the inflow Q
t
. The maxi-
mum vertical distance between the cumulative inflow
curve and the release line of slope R
t
equals the maximum
water deficit, and hence the required active storage
capacity. Clearly, if the average of the releases R
t
is greater
than the mean inflow, a reservoir will not be able to meet
the demand no matter what its active storage capacity.
An alternative way to identify the required reservoir
storage capacity is to plot the cumulative non-negative
deviations,

t
(R

), and note the largest total devia-

tion, as shown in Figure 11.13.
These graphical approaches do not account for losses.
Furthermore, the method shown in Figure 11.12 is awk-
ward if the desired releases in each period t are not the
same. The equivalent method shown in Figure 11.13 is
called the sequent peak method. If the sum of the desired
releases does not exceed the sum of the inflows, calcula-
tions over at most two successive hydrological records of
flows may be needed to identify the largest cumulative
deficit inflow. After that the procedure will produce
repetitive results. It is much easier to consider changing as
well as constant release values when determining the
maximum deficit by using the sequent peak method.
Sequent Peak Analyses
The sequent peak procedure is illustrated in Table 11.2.
Let K
t
be the maximum total storage requirement needed
K
a
t i
j

maximum R
t t
Q ( )

E
0
2
0
1
2
1
n
dead storage capacity
active storage capacity
flood storage capacity
Figure 11.11. Total reservoir storage volume capacity
consisting of the sum of dead storage, active storage and flood
control storage capacities.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 343
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
for periods 1 up through period t. As before, let R
t
be the
required release in period t, and Q
t
be the inflow in that
period. Setting K
0
equal to 0, the procedure involves
calculating K
t
using Equation 11.30 (below) consecutively
for up to twice the total length of record. This assumes
that the record repeats itself to take care of the case when
the critical sequence of flows occurs at the end of the
streamflow record, as indeed it does in the example nine-
period record of 1, 3, 3, 5, 8, 6, 7, 2 and 1.
K
t
R
t
Q
t
K
t1
if positive,
0 otherwise (11.30)
The maximum of all K
t
is the required storage capacity
for the specified releases R
t
and inflows, Q
t
. Table 11.2
illustrates this sequent peak procedure for computing
344 Water Resources Systems Planning and Management
the active capacity K
a
, i.e. the maximum of all K
t
, required
to achieve a release R
t
3.5 in each period given the
series of nine streamflows. Note that this method does
not require all releases to be the same.
2.2.2. Reservoir StorageYield Functions
Reservoir storageyield functions define the minimum
active storage capacity required to ensure a given constant
release rate for a specified sequence of reservoir inflows.
Mass diagrams, sequent peak analyses and linear opti-
mization (Chapter 4) are three methods that can be used
to define these functions. Given the same sequence
of known inflows and specified releases, each method
will provide identical results. Using optimization models,
10
8
6
4
2
0
time t
0 2 4 6 8 10 12 14 16 18 20
E
0
2
0
1
2
1
q
10
t
= 4 for R
5
t
= 3 for R
c
u
m
u
l
a
t
i
v
e

p
o
s
i
t
i
v
e
d
e
v
i
a
t
i
o
n
s
Q
t
-
R
t
(
)
active storage
capacity requirement;
Figure 11.13. Alternative plot for identifying reservoir
active storage capacity requirements.
80
70
60
50
40
30
20
10
0
c
u
m
u
l
a
t
i
v
e

i
n
f
l
o
w

t
Q
time t
0 2 4 6 8 10 12 14 16 18 20
E
0
2
0
1
2
1
p
active storage
capacity requirement;
10
t
= 4 for R
5
t
= 3 for R
release rate;
t
R
= 3
= 4
)
(

Figure 11.12. The Rippl or mass diagram

method for identifying reservoir active
storage capacity requirements.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 344
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 345
minimum release constraints:
R
t
Y t 1, 2, , T (11.34)
for various values of the yield, Y.
Alternatively, one can maximize the constant release
yield, Y,
Maximize Y (11.35)
for various values of active storage capacity, K
a
, subject
to the same constraint Equations 11.32 through 11.34.
Constraints 11.32 and 11.34 can be combined to
reduce the model size by T constraints.
S
t
Q
t
Y S
t1
t 1, 2, , T; T 1 1 (11.36)
The solutions of these two linear programming models,
using the nine-period flow sequence referred to above
and solved for various values of yield or capacity, respec-
tively, are plotted in Figure 11.14. The results are the
same as could be found using the mass diagram or
sequent peak methods.
There is a probability that the storageyield function
just defined will fail. A record of only nine flows, for
example, is not very long and, hence, will not give one
much confidence that they will define the critical low-
flow period of the future. One rough way to estimate the
reliability of a storageyield function is to rearrange and
rank the inflows in order of their magnitudes. If there are
n ranked inflows, there will be n1 intervals separating
them. Assuming there is an equal probability that any
future flow could occur in any interval between these
ranked flows, there is a probability of 1/(n1) that a
future flow will be less than the lowest recorded flow. If
that record low flow occurs during a critical low flow
period, more storage may be required than indicated in
the function.
Hence, for a record of only nine flows that are
considered representative of the future, one can be only
about 90% confident that the resulting storageyield
function will apply in the future. One can be only 90%
sure of the predicted yield Y associated with any storage
volume K. A much more confident estimate of the
reliability of any derived storageyield function can be
obtained by using synthetic flows to supplement
any measured streamflow record, taking parameter
uncertainty into account (as discussed in Chapters 7, 8,
and 9). This will provide alternative sequences as well as
more intervals between ranked flows.
it is possible to obtain such functions from multiple
reservoirs and to account for losses based on storage
volume surface areas, as will be discussed later.
There are two ways of defining a linear optimization
(linear programming) model to estimate the active storage
capacity requirements. One approach is to minimize the
active storage capacity, K
a
, subject to minimum required
constant releases, Y, the yield. This minimum active
storage capacity is the maximum storage volume, S
t
,
required, given the sequence of known inflows Q
t
, and
the specified yield, Y, in each period t. The problem is to
find the storage volumes, S
t
, and releases, R
t
, that
Minimize K
a
(11.31)
subject to:
mass-balance constraints:
S
t
Q
t
R
t
S
t1
t 1, 2, , T; T 1 1 (11.32)
capacity constraints:
S
t
K
a
t 1, 2, , T (11.33)
E
0
2
0
1
2
1
r
time t R
1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
3.5
t
Q K
t t-1
K
t
= + - ( )
+
1.0
3.0
3.0
5.0
8.0
6.0
7.0
2.0
1.0
1.0
3.0
3.0
5.0
8.0
6.0
7.0
2.0
1.0
0.0
2.5
3.0
3.5
2.0
0.0
0.0
0.0
1.5
4.0
6.5
7.0
7.5
6.0
1.5
0.0
0.0
1.5
2.5
3.0
3.5
2.0
0.0
0.0
0.0
1.5
4.0
6.5
7.0
7.5
6.0
1.5
0.0
0.0
1.5
4.0
repetition begins
K
a
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Table 11.2. Illustration of the sequent peak procedure for
computing active storage requirements.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 345
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
While the mass diagram and sequent peak procedures
are relatively simple, they are not readily adaptable to
reservoirs where evaporation losses and/or lake level
regulation are important considerations, or to problems
involving more than one reservoir. Mathematical pro-
gramming (optimization) methods provide this capability.
These optimization methods are based on mass-balance
equations for routing flows through each reservoir. The
mass-balance or continuity equations explicitly define
storage volumes (and hence storage areas from which
evaporation occurs) at the beginning of each period t.
2.2.3. Evaporation Losses
Evaporation losses, L
t
, from lakes and reservoirs, if any,
take place on their surface areas. Hence, to compute these
losses, their surface areas must be estimated in each
period t. Storage surface areas are functions of the storage
volumes, S
t
. These functions are typically concave, as
shown in Figure 11.15.
In addition to the storage areavolume function, the
seasonal surface water evaporation loss depths, E
t
max
, must
be known. Multiplying the average surface area, A
t
, based
on the initial and final storage volumes, S
t
and S
t1
, by the
loss depth, E
t
max
, yields the volume of evaporation loss, L
t
,
in the period t. The linear approximation of that loss is
L
t
[a
0
a(S
t
S
t1
)/2]E
t
max
(11.37)
Letting
a
t
0.5aE
t
max
(11.38)
346 Water Resources Systems Planning and Management
the mass-balance equation for storage volumes that include
evaporation losses in each period t can be approximated as:
(1 a
t
)S
t
Q
t
R
t
a
0
E
t
max
(1 a
t
)S
t1
(11.39)
If Equation 11.39 is used in optimization models for iden-
tifying preliminary designs of a proposed reservoir, and if
the active storage capacity turns out to be essentially zero,
or just that required to provide for the fixed evaporation
loss, a
0
E
t
max
, then clearly any reservoir at the site is not
justified. These mass-balance equations, together with
any reservoir storage capacity constraints, should be
removed from the model before resolving it again. This
procedure is preferable to introducing 0,1 integer
0
0
E
0
2
0
1
2
1
t
storage volume St
s
u
r
f
a
c
e

a
r
e
a
A
t
dead storage capacity
a
0

active storage volume

Figure 11.15. Storage surface area as a function of reservoir
storage volume along with its linear approximation. The
parameter a is the assumed increase in surface area
associated with a unit increase in the storage volume.
4
3
2
1
0
y
i
e
l
d
Y
0 2 4 6 8 10 12 14
E
0
2
0
1
2
1
s
active storage capacity K
a
K
a
Y
0.0
0.5
2.0
3.5
5.0
10.0
1.0
1.5
2.0
2.5
3.0
4.0
Figure 11.14. Storageyield function for the
sequence of flows 1, 3, 3, 5, 8, 6, 7, 2 and 1.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 346
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 347
variables that will include the term a
0
E
t
max
in Equation
11.39 only if the active storage volume is greater than 0
(using methods discussed in Chapter 4).
An alternative way to estimate evaporation loss that does
not require a surface areastorage volume relationship, such
as the one shown in Figure 11.15, is to define the storage
elevationstorage volume function. Subtracting the evapora-
tion loss depth from the initial surface elevation associated
with the initial storage volume will result in an adjusted
storage elevation, which in turn defines the initial storage
volume after evaporation losses have been deducted. This
adjusted initial volume can be used in continuity equations
(Equation 11.32 or 11.36). This procedure assumes evapo-
ration is only a function of the initial storage volume in each
time period t. For relatively large volumes and short time
periods, such an assumption is usually satisfactory.
2.2.4. Over and Within-Year Reservoir Storage and
Yields
An alternative approach to modelling reservoirs is to sepa-
rate out over-year storage and within-year storage, and to
focus not on total reservoir releases, but on parts of the total
releases that can be assigned specific reliabilities. These
release components we call yields. To define these yields
and the corresponding reservoir release rules, we divide this
section into four parts. The first outlines a method for
estimating the reliabilities of various constant annual mini-
mum flows or yields. The second part discusses a modelling
approach for estimating the over-year and within-year active
storage capacities that are required to deliver specified
annual and within-year yields, each having pre-specified
reliabilities. The third and fourth parts expand this model-
ling approach to include multiple yields having different
reliabilities, evaporation losses and the construction of reser-
voir operation rule curves using these flow release yields.
It will be convenient to illustrate the yield models and
their solutions using a simple example of a single reser-
voir and two within-year periods per year. This example
will be sufficient to illustrate the method that can
be applied to larger systems having more within-year
periods. Table 11.3 lists the nine years of available
streamflow data for each within-year season at a potential
reservoir site. These streamflows are used to compare the
solutions of various yield models as well as to illustrate
the concept of yield reliability.
Reliability of Annual Yields
As discussed in Chapter 4, reservoirs can be operated so as
to increase the dependable flow downstream. The flow
that can always be depended upon at a particular site is
commonly referred to as the firm yield or safe yield.
These terms imply that this is what the reservoir will
always be able to provide. Of course, this may not be true.
If historical flows are used to determine this yield, then the
resulting yield might be better called an historical yield.
Historical and firm yield are often used synonymously.
A flow yield is 100% reliable only if the sequence of
flows in future years will never sum to a smaller amount
than those that have occurred in the historic record. Usually
one cannot guarantee this condition. Hence, associated with
any historic yield is the uncertainty (a probability) that it
might not always be available in the future. There are
various ways of estimating this probability.
Referring to the nine-year streamflow record listed
in Table 11.3, if no reservoir is built to increase the yields
downstream of the reservoir site, the historic firm yield
is the lowest flow on record, namely 1.0 that occurred
in year 5. The reliability of this annual yield is the
E
0
2
0
1
3
0
m
1
2
3
4
5
6
7
8
9
total
1.0
0.5
1.0
0.5
0.5
0.5
1.0
2.5
1.5
9.0
3.0
2.5
2.0
1.5
0.5
2.5
5.0
5.5
4.5
27.0
4.0
3.0
3.0
2.0
1.0
3.0
6.0
8.0
6.0
36.0
time
year y Q
1y
Q
2y
Q
y
annual within-year period
flow
Q
ty
1.0 3.0 4.0 average flow
Table 11.3. Recorded unregulated historical streamflows at a
reservoir site.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 347
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
probability that the streamflow in any year will be greater
than or equal to this value. In other words, it is the
probability that this flow will be equalled or exceeded.
The expected value of the exceedance probability of the
lowest flow in an n-year record is approximately n/(n1),
which for the n 9 year flow record is 9/(91), or 0.90.
This is based on the assumption that any future flow has
an equal probability of being in any of the intervals
formed by ordering the record of flows from the lowest to
the highest value.
Now, rank the n flows of record from the highest to
the lowest. Assign the rank m of 1 to the highest flow, and
n to the lowest flow. In general, the expected probability
p that any flow of rank m will be equalled or exceeded in
any year is approximately m/(n1). An annual yield hav-
ing a probability p of exceedance will be denoted as Y
p
.
For independent events, the expected number of years
until a flow of rank m is equalled or exceeded is the
reciprocal of its probability of exceedance p, namely
1/p (n1)/m. The recurrence time or expected time
until a failure (a flow less than that of rank m) is the recip-
rocal of the probability of failure in any year. Thus, the
expected recurrence time T
p
associated with a flow having
an expected probability p of exceedance is 1/(1p).
2.2.5. Estimation of Active Reservoir Storage
Capacities for Specified Yields
Over-Year Storage Capacity
A reservoir with active over-year storage capacity
provides a means of increasing the magnitude and/or
the reliabilities of various annual yields. For example, the
sequent peak algorithm defined by Equation 11.30
provides a means of estimating the reservoir storage
volume capacity required to meet various firm yields,
Y
0.9
, associated with the nine annual flows presented in
Table 11.3. The same yields can be obtained from a linear
optimization model that minimizes active over-year
storage capacity, K
a
o
,
Minimize K
a
o
(11.40)
required to satisfy the following storage continuity and
capacity constraint equations involving only annual
storage volumes, S
y
, inflows, Q
y
, yields, Y
p
, and excess
releases, R
y
. For each year y:
S
y
Q
y
Y
p
R
y
S
y1
(11.41)
S
y
K
a
o
(11.42)
348 Water Resources Systems Planning and Management
Once again, if the year y is the last year of record, then
year y 1 is assumed to equal 1. For annual yields of 3
and 4, the over-year storage requirements are 3 and 8
respectively, as can be determined just by examining the
right hand column of annual flows in Table 11.3.
The over-year model, Equations 11.40 through 11.42,
identifies only annual or over-year storage requirements
based on specified (known) annual flows, Q
y
, and speci-
fied constant annual yields, Y
p
. Within-year periods t
requiring constant yields y
pt
that sum to the annual yield
Y
p
may also be considered in the estimation of the
required over-year and within-year or total active storage
capacity, K
a
. Any distribution of the over-year yield
within the year that differs from the distribution of
the within-year inflows may require additional active
reservoir storage capacity. This additional capacity is
called the within-year storage capacity.
The sequent peak method, Equation 11.30, can be
used to obtain the total over-year and within-year active
storage capacity requirements for specified within-year
period yields, y
pt
. Alternatively a linear programming
model can be developed to obtain the same information
along with associated reservoir storage volumes. The
objective is to find the minimum total active storage
capacity, K
a
, subject to storage volume continuity and
capacity constraints for every within-year period of every
year. This model is defined as:
minimize K
a
(11.43)
subject to
S
ty
Q
ty
y
pt
R
ty
S
t1, y
t, y (11.44)
S
ty
K
a
t, y (11.45)
In Equation 11.44, if t is the final period T in year y, the
next period T 1 1 in year y 1, or year 1 if y is the
last year of record, n.
The difference in the active capacities resulting from
these two models, Equations 11.40 through 11.42, and
Equations 11.43 through 11.45, is the within-year storage
requirement, K
a
w
.
Table 11.4 shows some results from solving both of
the above models. The over-year storage capacity require-
ments, K
a
0
, are obtained from Equations 11.40 through
11.42. The combined over-year and within-year capaci-
ties, K
a
, are obtained from solving Equations 11.43
through 11.45. The difference between the over-year
storage capacity, K
a
0
, required to meet only the annual
wrm_ch11.qxd 8/31/2005 11:57 AM Page 348
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 349
yields and the total capacity, K
a
, required to meet each
specified within-year yield distribution of those annual
yields is the within-year active storage capacity K
a
w
.
Clearly, the number of continuity and reservoir capacity
constraints in the combined over-year and within-year
model (Equations 11.4311.45) can become very large
when the number of years n and within-year periods T are
large. Each reservoir site in the river system will require
2nT continuity and capacity constraints. Not all these
constraints are necessary, however. It is only a subset of
the sequence of flows within the total record of flows that
generally determines the required active storage capacity
K
a
of a reservoir. This is called the critical period. This crit-
ical period is often used in engineering studies to
estimate the historical yield of any particular reservoir or
system of reservoirs.
Even though the severity of future droughts is
unknown, many planners accept the traditional practice
of using the historical critical drought period for reservoir
design and operation studies. They assume events that
have happened in the past could happen again in the
future. In some parts of the world, notably those coun-
tries in the lower portions of the southern hemisphere,
historical records are continually proven to be unreliable
indicators of future hydrological conditions. In these
regions especially, synthetically generated flows based on
statistical methods are more acceptable as a basis for yield
estimation than seems to be the case elsewhere.
Over and Within-Year Storage Capacity
To begin the development of a smaller, but more approx-
imate, model, consider each combined over-year and
within-year storage reservoir to consist of two separate
reservoirs in series (Figure 11.16). The upper reservoir is
the over-year storage reservoir, whose capacity required
for an annual yield is determined by an over-year model,
such as Equations 11.40 through 11.42. The purpose of
the downstream within-year reservoir is to distribute as
desired in each within-year period t the annual yield pro-
duced by the upstream over-year reservoir. Within-year
storage capacity would not be needed if the distribution
E
0
2
0
1
3
0
n
3 4.0
3.5
4.5
5.5
3.0
3.0
3.0
3.0
1.0
0.5
1.5
2.5
3
2
1
0
0
1
2
3
within
-year
0. 9 K K K
a a a
t = 1 t = 2
w o
over
-year
total
required active storage
volume capacity
annual
yield
within-year yields
y
9.0
8.0
9.0
10.0
11.0
8.0
8.0
8.0
8.0
8.0
1.0
0.0
1.0
2.0
3.0
4
3
2
1
0
0
1
2
3
4
4
Table 11.4. Active storage requirements for various
within-year yields.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 349
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
of the average inflows into the upper over-year reservoir
exactly coincided with the desired distribution of within-
year yields. If they do not, within-year storage may be
required. The two separate reservoir capacities summed
together will be an approximation of the total active reser-
voir storage requirement needed to provide those desired
within-year period yields.
Assume the annual yield produced and released by
the over-year reservoir is distributed in each of the
within-year periods in the same ratio as the average
within-year inflows divided by the total average annual
inflow. Let the ratio of the average period t inflow divided
by the total annual inflow be
t
. The general within
year model is to find the minimum within-year storage
capacity, K
a
w
, subject to within-year storage volume
continuity and capacity constraints.
Minimize K
a
w
(11.46)
subject to
s
t

t
Y
p
y
pt
s
t1
t T1 1 (11.47)
s
t
K
a
w
t (11.48)
350 Water Resources Systems Planning and Management
Since the sum of
t
over all within-year periods t is 1, the
model guarantees that the sum of the unknown within-
year yields, y
pt
, equals the annual yield, Y
pt
.
The over-year model, Equations 11.40 through 11.42,
and within-year model, Equations 11.46 through 11.48,
can be combined into a single model for an n-year sequence
of flows:
Minimize K
a
(11.49)
subject to:
S
y
Q
y
Y
p
R
y
S
y1
y if y n, y1 1
(11.50)
S
y
K
a
o
y (11.51)
s
t

t
Y
p
y
pt
s
t1
t if t T, T1 1 (11.52)
s
t
K
a
w
t (11.53)
(11.54)
K
a
K
a
o
K
a
w
(11.55)
Constraint 11.54 is not required due to Equation 11.52,
but is included here to make it clear that the sum of within-
year yields will equal the over-year yield, and that such a
constraint will be required for each yield of reliability p if
multiple yields of different reliabilities are included in
the model. In addition, constraint Equation 11.53 can be
combined with Equation 11.55, saving a constraint. If this
is done, the combined model contains 2n 2T 1 con-
straints, compared to the more accurate model, Equations
11.43 through 11.45, that contains 2nT constraints.
If the fractions
t
are based on the ratios of the average
within-year inflow divided by average annual inflow in
the two within-year periods shown in Table 11.3, then
0.25 of the total annual yield flows into the fictitious
within-year reservoir in period t 1, and 0.75 of the total
annual yield flows into the reservoir in period t 2.
Suppose the two desired within-year yields are to be 3
and 0 for periods 1 and 2 respectively. The total annual
yield, Y
0.9
, is 3. Assuming the natural distribution of this
annual yield of 3 in period 1 is 0.25 Y
0.9
0.75, and in
period 2 it is 0.75 Y
0.9
2.25, the within-year storage
required to redistribute these yields of 0.75 and 2.25 to
become 3 and 0, respectively, is K
a
w
2.25. From
Table 11.4, we can see an annual yield of 3 requires an
over-year storage capacity of 3. Thus, the estimated total
storage capacity required to provide yields of 3 and 0 in
y Y
pt p
t

actual over-year
and within-year
reservoir
fictitious over-year
and within-year
reservoir
K
a
~
~
K
o
a
+ K
w
a
E
0
2
0
1
0
8
i
Figure 11.16. Approximating a combined over-year and
within-year reservoir as two separate reservoirs, one for
creating annual yields, the other for distributing them as
desired in the within-year periods.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 350
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 351
periods 1 and 2 is the over-year capacity of 3 plus the
within-year capacity of 2.25 equalling 5.25. This com-
pares with 3 plus 2.5 of actual within-year capacity
required for a total of 5.50, as indicated in Table 11.4.
There are ways to reduce the number of over-year
constraints without changing the solution of the over-year
model. Sequences of years whose annual inflow values
equal or exceed the desired annual yield can be combined
into one constraint. If the yield is an unknown variable,
then the mean annual inflow can be used since it is the
upper limit of the annual yield. For example in Table 11.3
note that the last three years and the first year have flows
equal or greater than 4, the mean annual inflow.
Thus,these four successive years can be combined into a
single continuity equation:
S
7
Q
7
Q
8
Q
9
Q
1
4Y
p
R
7
S
2
(11.56)
This saves a total of three over-year continuity constraints
and three over-year capacity constraints. Note that the
excess release, R
7
, represents the excess release in all four
periods. Furthermore, not all reservoir capacity constraint
Equation 11.51 are needed, since the initial storage
volumes in the years following low flows will probably be
less than the over-year capacity.
There are many ways to modify and extend this yield
model to include other objectives, fixed ratios of the
unknown annual yield for each within-year period,
and even multiple yields having different exceedance
probabilities p.
The number of over-year periods being modelled com-
pared to the number of years of flow records determines
the highest exceedance probability or reliability a yield
can have, e.g. 9/10 or 0.9 in the nine-year example used
here. If yields having lower reliabilities are desired, such
as a yield with a reliability of 0.80, then the yield variable
Y
P
(Y
0.90
) can be omitted from Equation 11.50 in the
critical year that determines the required over-year
capacity for a 0.90 reliable yield. (Since some outflow
might be expected, even if it is less than the 90% reliable
yield, the outflow could be forced to equal the inflow for
that year.) If a 0.70 reliable yield is desired, then the yield
variables in the two most critical years can be omitted
from Equation 11.50, and so on.
The number of years of yield failure determines the
estimated reliability of each yield. An annual yield that
fails in f out of n years has an estimated probability
(nf )/(n1) of being equalled or exceeded in any future
year. Once the desired reliability of a yield is known, the
modeller must select the appropriate failure years and
specify the permissible extent of failure in those f failure
years.
To consider different yield reliabilities p, let the
parameter
y
p
be a specified value between 0 and 1 that
indicates the extent of a failure in year y associated with
an annual yield having a reliability of p. When
y
p
is 1
there is no failure, and when it is less than 1 there is a fail-
ure, but a proportion of the yield Y
p
equal to
y
p
is
released. Its value is in part dependent on the conse-
quences of failure and on the ability to forecast when a
failure may occur and to adjust the reservoir operating
policy accordingly.
The over-year storage continuity constraints for n
years can now be written in a form appropriate for
identifying any single annual yield Y
p
having an
exceedance probability p:
S
y
Q
y

y
p
Y
p
R
y
S
y1
y
if y n, y1 1 (11.57)
When writing Equation 11.57, the failure year or years
should be selected from among those in which permitting
a failure decreases the required reservoir capacity K
a
. If a
failure year is selected that has an excess release, no reduc-
tion in the required active storage capacity will result and
the reliability of the yield may be higher than intended.
The critical year or years that determine the required
active storage volume capacity may be dependent on the
yield itself. Consider, for example, the seven-year
sequence of annual flows (4, 3, 3, 2, 8, 1, 7) whose mean
is 4. If a yield of 2 is desired in each of the seven years, the
critical year requiring reservoir capacity is year 6. If a yield
of 4 is desired (again assuming no losses), the critical years
are years 2 through 4. The streamflows and yields in these
critical years determine the required over-year storage
capacity. The failure years, if any, must be selected from
within the critical low flow periods for the desired yield.
When the magnitudes of the yields are unknown,
some trial-and-error solutions may be necessary to
ensure that any failure years are within the critical period
of years for the associated yields. To ensure a wider range
of applicable yield magnitudes, the year having the low-
est flow within the critical period should be selected as
the failure year if only one failure year is selected. Even
wrm_ch11.qxd 8/31/2005 11:57 AM Page 351
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
though the actual failure year may follow that year, the
resulting required reservoir storage volume capacity will
be the same.
Multiple Yields and Evaporation Losses
The yield models developed so far define only single
annual and within-year yields. Incremental secondary
yields having lower reliabilities can also be included in
the model. Referring to the nine-year streamflow record
in Table 11.3, assume that two yields are desired, one
90% reliable and the other 70% reliable. Let Y
0.9
and Y
0.7
represent those annual yields having reliabilities of 0.9
and 0.7, respectively. The incremental secondary yield
Y
0.7
represents the amount in addition to Y
0.9
that is only
70% reliable. Assume that the problem is one of estimat-
ing the appropriate values of Y
0.9
and Y
0.7
, their respective
within-year allocations y
pt
, and the total active reservoir
capacity K
a
that maximizes some function of these yield
and capacity variables.
In this case, the over-year and within-year continuity
constraints can be written
S
y
Q
y
Y
0.9

y
0.7
Y
0.7
R
y
S
y1
y
if y n, y1 1 (11.58)
s
t

t
(Y
0.9
Y
0.7
) y
0.9,t
y
0.7,t
s
t1
t
if t T, T1 1 (11.59)
Now an additional constraint is needed to ensure that all
within-year yields of a reliability p sum to the annual yield
of the same reliability. Selecting the 90%
reliable yield,
(11.60)
Evaporation losses must be based on an expected storage
volume in each period and year, since the actual storage
volumes are not identified using these yield models. The
approximate storage volume in any period t in year y can be
defined as the initial over-year volume S
y
, plus the estimated
average within-year volume (s
t
s
t1
)/2. Substituting this
storage volume into Equation 11.37 (see also Figure 11.15)
results in an estimated evaporation loss L
yt
:
L
yt
[a
0
a(S
y
(s
t
s
t1
)/2)]E
t
max
(11.61)
Summing L
yt
over all within-year periods t defines the
estimated annual evaporation loss, E
y
:
y Y
t
t
0 9 0 9 . , .

352 Water Resources Systems Planning and Management

(11.62)
This annual evaporation loss applies, of course, only
when there is a non-zero active storage capacity require-
ment. These annual evaporation losses can be included in
the over-year continuity constraints, such as Equation
11.58. If they are, the assumption is being made that their
within-year distribution will be defined by the fractions

t
. This may not be realistic. If it is not, an alternative
would be to include the average within-year period losses,
L
t
, in the within-year constraints.
The average within-year period loss, L
t
, can be defined
as the sum of each loss L
yt
defined by Equation 11.61 over
all years y divided by the total number of years, n:
(11.63)
This average within-year period loss, L
t
, can be added to
the within-years highest reliability yield, y
pt
, forcing
greater total annual yields of all reliabilities to meet
corresponding total within-year yield values. Hence,
combining Equations 11.60 and 11.61, for p equal to 0.9
in the example,
(11.64)
Since actual reservoir storage volumes in each period t
of each year y are not identified in this model, system
performance measures that are functions of those storage
volumes, such as hydroelectric energy or reservoir
recreation, are only approximate. Thus, as with any of
these screening models, any set of solutions should be
evaluated and further improved using more precise
simulation methods.
Simulation methods require reservoir operating rules.
The information provided by the solution of the yield
model can help in defining a reservoir operating policy for
such simulation studies.
Reservoir Operation Rules
Reservoir operation rules, as discussed in Chapters 4, 7
and 8, are guides for those responsible for reservoir
Y y a a S s s E n
p pt y t t t
y
n
t


[ ( ( )/ )] /
max
0 1
2

L a a S s s E n
t y t t t
y
n


[ ( ( )/ )] /
max
0 1
2

E a a S s s E
y y t t t
t


[ ( ( ) / )]
max
0 1
2

wrm_ch11.qxd 8/31/2005 11:57 AM Page 352

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 353
operation. They apply to reservoirs being operated in a
steady-state condition, i.e. not filling up immediately after
construction or being operated to meet a set of new and
temporary objectives. There are several types of rules, but
each indicates the desired or required reservoir release or
storage volume at any particular time of year. Some rules
identify storage volume targets (rule curves) that the oper-
ator is to maintain, if possible, and others identify storage
zones, each associated with a particular release policy.
This latter type of rule can be developed from the solution
of the yield model.
To construct an operation rule that identifies storage
zones, each having a specific release policy, the values of
the dead and flood storage capacities, K
D
and K
f
, are
needed together with the over-year storage capacity, K
a
o
,
and within-year storage volumes, s
t
, in each period t.
Since both K
a
o
and all s
t
derived from the yield model
are for all yields, Y
p
, being considered, it is necessary
to determine the over-year capacities and within-year
storage volumes required to provide each separate
within-year yield, y
pt
. Plotting the curves defined by the
respective over-year capacity plus the within-year storage
volume (K
a
o
s
t
) in each within-year period t will define
a zone of storage whose yield releases y
pt
from that zone
should have a reliability of at least p.
For example, assume again a nine-year flow record and
ten within-year periods. Of interest are the within-year
yields, y
0.9,t
and y
0.7,t
, having reliabilities of 0.9 and 0.7. The
first step is to compute the over-year storage capacity
requirement, K
a
o
, and the within-year storage volumes, s
t
,
for just the yields y
0.9,t
. The sum of these values, K
a
o
s
t
, in
each period t can be plotted as illustrated in Figure 11.17.
The sum of the over-year capacity and within-year
volume K
a
o
s
t
in each period t defines the zone of active
storage volumes for each period t required to supply the
within-year yields y
0.9,t
. If the storage volume is in the
shaded zone shown in Figure 11.17, only the yields y
0.9,t
should be released. The reliability of these yields, when
simulated, should be about 0.9. If at any time t the actual
reservoir storage volume is within this zone, then
reservoir releases should not exceed those required to
meet the yield y
0.9,t
if the reliability of this yield is to be
maintained.
The next step is to solve the yield model for both
yields Y
0.9
and Y
0.7
. The resulting sum of over-year storage
capacity and within-year storage volumes can be plotted
over the first zone, as shown in Figure 11.18.
If at any time t the actual storage volume is in the sec-
ond, lighter-shaded, zone in Figure 11.18, the release
should be the sum of both the most reliable yield, y
0.9,t
and the incremental secondary yield y
0.7,t
. If only these
releases are made, then the probability of being in that
zone, when simulated, should be about 0.7. If the actual
storage volume is greater than the total required over-year
storage capacity K
a
o
plus the within-year volume s
t
, the
non-shaded zone in Figure 11.18, then a release can
E
0
2
0
1
3
0
p
within-year period t
r
e
s
e
r
v
o
i
r

s
t
o
r
a
g
e

e
l
e
v
a
t
i
o
n
within-year volumes s
t
over-year capacity K
o
a
Figure 11.17. Reservoir release rule
showing the identification of the most
reliable release zone associated with the
within-year yields y
0.9,t
.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 353
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
354 Water Resources Systems Planning and Management
2.3. Wetlands and Swamps
Wetlands and swamps are important elements of the
water resources system. They provide regulating
functions with respect to both quantity (e.g. flood
attenuation) and quality (e.g. self-purification). Most
wetlands support valuable ecological systems. Appendix A
(Section 4) describes the natural processes involved
in wetlands.
From a quantitative point of view, wetlands are
comparable with (shallow) lakes. Lakes are usually
located on streams or rivers, while most wetlands are
separate elements connected to a stream or river only in
wet periods, either by flooding from the stream or river or
by draining to it. Some wetlands are part of river systems.
Examples are the Sudd Swamps in the Nile River, the
Pantanal in the Paraguay River in South America and
the marshlands in southern Iraq at the confluence of
the Euphrates and Tigris Rivers. In these cases wetlands
can be modelled as lakes or reservoirs, as described in the
previous section. If they are separate from the river
system, wetlands can be included as indicated in
Figure 11.20. The lateral flow Q
3
(t) can be bi-directional.
2.4. Water Quality and Ecology
River basin models focusing on water quantities are mostly
used to investigate whether sufficient water is available to
satisfy the various use functions (off-stream and
in-stream), and to identify measures to match supply and
demand. The core of most river basin models consists of
keeping track of the water balance of the whole river basin.
The analysis of water quality and ecology is mostly done
off-line, using another model for a specific part of the
system, e.g. a river stretch, reservoir or groundwater
system. These models will be described in Chapter 12.
Such an off-line approach makes sense. There is little
feedback from water quality to quantity (except in cases
where minimum flows are required to maintain a minimum
water quality level). Using separate water quality models for
parts of the system makes it possible to include more
temporal and spatial detail and to include more complex
water quality processes. Environmental flow requirements
can be included in river basin models by defining specific
flow regime demands (quantity, velocity, dynamics and the
like) at certain locations in the river basin.
be made to satisfy any down-stream demand. Converting
storage volume to elevation, this release policy is
summarized in Figure 11.19.
These yield models focus only on the active storage
capacity requirements. They can be a part of a model that
includes flood storage requirements as well (as discussed
later in this chapter). If the actual storage volume is
within the flood control zone in the flood season, releases
should be made to reduce the actual storage to a volume
no greater than the total capacity less the flood storage
capacity.
Once again, reservoir rules developed from simplified
models such as this yield model are only guides, and once
developed they should be simulated, evaluated and
refined prior to their actual adoption.
E
0
2
0
1
3
0 within-year period t
r
e
s
e
r
v
o
i
r

s
t
o
r
a
g
e

e
l
e
v
a
t
i
o
n
r
releases and their reliabilities
y
0.9, t
y
0.7, t
+
y
0.9, t
any
70%
90%
70% <
~
~
~
~
Figure 11.19. Reservoir release rule defined by the yield model.
E
0
2
0
1
3
0 within-year period t
r
e
s
e
r
v
o
i
r

s
t
o
r
a
g
e

e
l
e
v
a
t
i
o
n
q
K
o
a
over-year
capacity within-year volumes s
t
Figure 11.18. Reservoir release rule showing the
identification of the second most reliable release zone
associated with the total within-year yields y
0.9,t
y
0.7,t
.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 354
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 355
3. Modelling the Socio-Economic
Functions In a River Basin
3.1. Withdrawals and Diversions
Major demands for the withdrawal of water include those
for domestic or municipal uses, industrial uses (including
cooling water) and agricultural uses (including irrigation).
These uses generally require the withdrawal of water from
a river system or other surface or groundwater body. The
water withdrawn may be only partially consumed, and
that which is not consumed may be returned to the river
system, perhaps at a different site, at a later time period,
and containing different concentrations of constituents.
Water can also be allocated to instream uses that alter
the distribution of flows in time and space. Such uses
include:
reservoir storage, possibly for recreational use as well
as for water supply
flow augmentation, possibly for water quality control,
navigation or ecological benefits
hydroelectric power production.
The instream uses may complement or compete with
each other or with various off-stream municipal, industrial
and agricultural demands. One purpose of developing
management models of river basin systems is to help derive
policies that will best serve these multiple uses, or at least
identify the tradeoffs among the multiple purposes and
objectives.
The allocated flow q
t
s
to a particular use at site s in
period t must be no greater than the total flow available,
Q
t
s
, at that site and in that period:
q
t
s
Q
t
s
(11.65)
The quantity of water that any particular user expects to
receive in each particular period is termed the target
allocation. Given an annual (known or unknown) target
allocation T
s
at site s, some (usually known) fraction, f
t
s
, of
that annual target allocation will be expected in each
within-year period t. If the actual allocation, q
t
s
, is less
than the target allocation, f
t
s
T
s
, there will be a deficit, D
t
s
.
If the allocation is greater than the target allocation, there
will be an excess, E
t
s
. Hence, to define those unknown
variables, the following constraint equation can be written
for each applicable period t:
q
t
s
f
t
s
T
s
D
t
s
E
t
s
(11.66)
Even though allowed, one would not expect a solution to
contain non-zero values for both D
t
s
and E
t
s
.
S
1
(t )
S
2
(t)
S
2
(t- ) 1
Q
2
(t) q
2
(t)
R ( t) E(t)
A
2
(t)
A
2
(t- ) 1
river
swamp / wetland
r ( t)
A
1
(t- ) 1
A
1
(t)
q (t)
1
Q ( t)
1
Q ( t)
3
r ( t)
S
1
(t- ) 1 S river channel storage -
1
(t)
S swamp / wetland storage -
2
(t) S
2
(t- ) 1
A river channel surface area -
1
(t) A
1
(t- ) 1
A swamp / wetland surface area -
2
(t) A
2
(t- ) 1
q (t)
1
Q (t)
1
river channel inflow -
river channel outflow -
q
2
(t)
Q
2
(t)
Q ( t)
3
E evaporation (t)
R rainfall (t)
r s oil moisture recharge (t)
swamp / wetland inflow -
-
swamp / wetland outflow -
-
lateral flow -
-
S
1
(t- ) 1
E
0
4
0
7
1
2
a
Figure 11.20. Schematization of
the link between wetlands and a
river system.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 355
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Whether or not any deficit or excess allocation
should be allowed at any demand site s depends on the
quantity of water available and the losses or penalties
associated with deficit or excess allocations to that site.
At sites where the benefits derived in each period are
independent of the allocations in other periods, the
losses associated with deficits and the losses or benefits
associated with excesses can be defined in each period t
(Chapter 10). For example, the benefits derived from
the allocation of water for hydropower production in
period t will in some cases be essentially independent of
previous allocations.
For any use in which the benefits are dependent on a
sequence of allocations, such as at irrigation sites, the
benefits may be based on the annual (or growing season)
target water allocations T
s
and their within season distri-
butions, f
t
s
T
s
. In these cases one can define the benefits
from those water uses as functions of the unknown sea-
son or annual targets, T
s
, where the allocated flows q
t
s
must be no less than the specified fraction of that
unknown target;
q
t
s
f
t t
s
T
s
for all relevant t (11.67)
If, for any reason, an allocation q
t
s
must be zero, then
clearly from Equation 11.67 the annual or growing
season target allocation T
s
would be zero, and presumably
so would be the benefits associated with that target value.
Water stored in reservoirs can often be used to
augment downstream flows for instream uses such as
recreation, navigation and water quality control. During
natural low-flow periods in the dry season, it is not only
the increased volume but also the lower temperature of
the augmented flows that may provide the only means
of maintaining certain species of fish and other aquatic
life. Dilution of wastewater or runoff from non-point
sources may be another potential benefit from flow
augmentation. These and other factors related to water
quality management are discussed in greater detail in
Chapter 12.
The benefits derived from navigation on a potentially
navigable portion of a river system can usually be
expressed as a function of the stage or depth of water in
various periods. Assuming known stream or river flow-
stage relationships at various sites in the river, a possible
constraint might require at least a minimum acceptable
depth, and hence flow, for those sites.
356 Water Resources Systems Planning and Management
3.2. Domestic, Municipal and Industrial Water
Demand
Domestic, municipal and industrial (DMI) water demands
are typically based on projections of population and socio-
economic activities. The domestic demand is the house-
hold demand. The municipal demand covers the water
requirements of the commercial sector (shops, department
stores, hotels and so on) and for the services sector (hospi-
tals, offices, schools and others). In most river basins,
the DMI demand will be small compared to the demand for
irrigation. However, given the importance of this demand
for human health and economic developments, it is often
given first priority. The preferred source for drinking
water (domestic and municipal) is often groundwater
which requires only limited treatment. When sufficient
groundwater of good quality is not available, surface water
will be used, in most cases in combinati4on with advanced
treatment facilities.
About 80% to 90% of the water abstracted for DMI
will usually be returned to the system as wastewater.
About 95% of cooling water is also typically returned to
the basin. In river basin studies, the total demand has to
be taken into account, as this amount is actually with-
drawn from the system. The water not used will be
returned to the system, possibly at another location, in a
future time period and containing more pollutants
because of its use.
Predicting DMI demands can be difficult. A common
approach is to use the demand predictions of the public
water supply authorities in the basin. They base their pre-
dictions on estimates of population growth, growth in
demand per capita and so on, often using statistical
trends. As these authorities have a good knowledge of
what is going on in their area, their estimates are often the
most reasonable ones to use, although they often tend to
overestimate the demand.
In many cases these predictions are not available or
not sufficiently reliable, and it is then usually necessary
to collect basic information on population and socio-
economic activities and expected growth rates. Domestic
use can be estimated on the basis of water consumption
per capita and coverage rates, taking into account
differences in social strata. Municipal demand is often
assumed to be a percentage of the domestic water
demand (usually in the range of 1535%). Industrial
wrm_ch11.qxd 8/31/2005 11:57 AM Page 356
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 357
water demand is the most difficult to estimate, as it
depends on the type of industry and the production
processes being used. Not only is it difficult to get all the
data needed to make accurate predictions, but future
economic and technological developments are very
uncertain. In most cases estimates are made on the
basis of statistical projections or of water demand per
industrial employee (for which data are easier to get),
depending on the type of industry.
Once a total demand for DMI has been determined,
projections have to be made for the losses in the system.
Unaccounted for water losses depend on the condition
and maintenance of the distribution system and should be
taken into account. In some systems such losses can
exceed 40%.
3.3. Agricultural Water Demand
Most agricultural areas in the world depend on rain as the
primary source of water. During dry periods, additional
water may be supplied through irrigation systems. In
semi-arid and arid countries, irrigation is an absolute
requirement to enable agricultural production. In those
cases, agricultural water is often the dominant demand
category. The irrigation demand in Egypt accounts for
90% of the total net water demand, and in the western
United States for about 75% of the total.
Irrigation water demand depends on a large number of
factors, which can be grouped under four headings:
hydro-meteorological conditions
types of crops
soil types
irrigation practices and water use efficiency.
Crops transpire as they grow. At the same time, there is
evaporation from the soil surface. The combined quantity
of water used under conditions of optimum availability is
known as consumptive use or evapotranspiration (ET).
The potential evapotranspiration, E
p
, is estimated from
the following equation:
E
p
c
p
ET
0
(11.68)
in which:
E
p
potential crop evapotranspiration
(mm per time unit)
c
p
crop coefficient
ET
0
reference crop evapotranspiration
(mm per time unit)
The reference crop evapotranspiration, ET
0
, can be
computed using a standard FAO method (FAO, 1998)
based on sunshine, temperature, humidity and wind
speed. The crop coefficient, c
p
, is related to the type of
crop and the growth stage.
Based on crop water demand (E
p
), the crop irrigation
demand (C
d
E
p
effective rainfall), field irrigation
demand (F
d
C
d
field losses) and project irrigation demand
(P
d
F
d
distribution losses) can be calculated.
The calculation of irrigation demand is complex as
farmers use many different cropping patterns and cropping
calendars. Several computer programs (e.g. the FAOs
CROPWAT) are available to support that calculation, and
many river basin simulation packages (see Section 4.3)
have modules that estimate the irrigation demand.
3.4. Hydroelectric Power Production
Hydropower plants, Figure 11.21, convert the energy
from the flowing water into mechanical and then
electrical energy. These plants, containing turbines and
generators, are installed either in or adjacent to dams.
Pipelines (penstocks) carry water under pressure from the
reservoir to the powerhouse. Power transmission systems
transport energy from the powerhouse to where it is
needed.
The principal advantages of hydropower are the
absence of polluting emissions during operation, its
capability to respond quickly to changing utility load
demands, and its relatively low operating costs. Dis-
advantages can include high initial capital cost, potential
site-specific impacts and cumulative environmental ones.
The potential environmental impacts of hydropower
projects include altered flow regimes below storage
reservoirs, water quality degradation, mortality of fish
that pass through hydroelectric turbines, blockage of
fish migration and flooding of terrestrial ecosystems by
impoundments. Proper design and operation of hydro-
power projects can mitigate some of these impacts (see
also WCD, 2000). Hydroelectric projects can also provide
beneficial effects such as recreation in reservoirs or in
tailwaters below dams.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 357
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Hydropower technology can be grouped into two
types: conventional and pumped storage. Conventional
plants use the available water from a river, stream, canal
system or reservoir to produce electrical energy. In
conventional multipurpose reservoirs and run-of-river
systems, hydropower production is just one of many
competing purposes for which water resources may be
used. Competing water uses can include irrigation, flood
control, navigation, and municipal and industrial water
supply. Pumped storage plants pump the water, usually
through a reversible turbine, from a lower to an upper
reservoir. While pumped storage facilities are net energy
consumers, they are income producers. They are valued
by the utilities because they can be rapidly brought online
to operate in a peak power production mode when energy
demand and prices are highest. The pumping to replenish
the upper reservoir is performed during off-peak hours
when electricity costs are lowest. This process benefits the
utility by increasing the load factor and reducing the
cycling of its base-load units. In most cases, pumped
storage plants run a full cycle every twenty-four hours
(USDOE, 2002).
Run-of-river plants use the natural flow of the river and
produce relatively little change in the stream channel and
358 Water Resources Systems Planning and Management
streamflow, in contrast to a peaking plant that impounds
water when the energy is needed. A storage project
extensively impounds and stores water during high-flow
periods to augment the water available during low-flow
periods, allowing the flow releases and power production
to be more constant. Many projects combine the modes.
The power capacity of a hydropower plant is primarily
the function of the hydraulic head and the flow rate
through the turbines. The hydraulic head is the elevation
difference the water falls in passing through the plant or to
the tailwater, whichever elevation difference is less. Project
design may concentrate on either of these variables or
both, and on the hydropower plants installed capacity.
The production of hydroelectric energy during any
period at any particular reservoir site is dependent on the
installed plant capacity; the flow through the turbines; the
average effective productive storage head; the number of
hours in the period; the plant factor (the fraction of time
that energy is produced); and a constant for converting
the product of flow, head and plant efficiency to electrical
energy. The kilowatt-hours of energy, KWH
t
, produced in
period t is proportional to the product of the plant effi-
ciency, e, the productive storage head, H
t
, and the flow, q
t
,
through the turbines.
a dam - stores water
b penstock - carries water to the turbines
c transmission lines - conduct electricity, ultimately to homes and businesses
d generator - rotated by the turbines to generate electricity
e turbine - turned by the force of the water on their blades
Cross section of conventional hydropower facility that uses an impoundment dam
E
0
2
0
7
2
5
u
a
c
d
e
b
Figure 11.21. Hydropower plant
and system components.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 358
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 359
A cubic metre of water, weighing l0
3
kg, falling a
distance of one metre, acquires 9.81 10
3
joules
(Newton-metres) of kinetic energy. The energy generated
in one second equals the watts (joules per second) of
power produced. Hence, an average flow of q
t
cubic
metres per second falling a height of H
t
metres in period t
yields 9.81 10
3
q
t
H
t
watts or 9.81 q
t
H
t
kilowatts of
power. Multiplying by the number of hours in period t
yields the kilowatt-hours of energy produced from an
average flow rate of q
t
. The total kilowatt-hours of energy,
KWH
t
, produced in period t, assuming 100% efficiency in
conversion of potential to electrical energy, is
KWH
t
9.81q
t
H
t
(seconds in period t)/
(seconds per hour)
9.81q
t
H
t
(seconds in period t)/3600 (11.69)
Since the total flow, Q
T
t
through the turbines in period t
equals the average flow rate q
t
times the number of
seconds in the period, the total kilowatt-hours of energy
produced in period t, given a plant efficiency of e, equals
KWH
t
9.81Q
T
t
H
t
e/3600
0.002725Q
T
t
H
t
e (11.70)
The energy required for pumped storage, where instead of
producing energy the turbines are used to pump water up
to a higher level, is
KWH
t
0.002725Q
T
t
H
t
/e (11.71)
For Equations 11.70 and 11.71, Q
T
t
is in cubic metres
and H
t
is in metres. The storage head, H
t
, is the vertical
distance between the water surface elevation in the lake
or reservoir that is the source of the flow through the
turbines and the maximum of either the turbine elevation
or the downstream discharge elevation. In variable head
reservoirs, storage heads are functions of storage volumes.
In optimization models for capacity planning these heads
and the turbine flows are among the unknown variables.
The energy produced is proportional to the product of
these two unknown variables. This results in non-
separable functions in equations that must be written at
each hydroelectric site for each time period t.
A number of ways have been developed to convert
these non-separable energy production functions to
separable ones for use in linear optimization models for
estimating design and operating policy variable values.
These methods inevitably increase the model complexity
and the number of its variables and constraints. For a
preliminary screening of hydropower capacities prior to
a more detailed analysis (e.g. using simulation or other
non-linear or discrete dynamic programming methods)
one can:
1. Solve the model using both optimistic and pessimistic
assumed fixed head values.
2. Compare the actual derived heads with the assumed
ones and adjust the assumed heads.
3. Resolve the model.
4. Compare the capacity values.
From this iterative process one should be able to identify
the range of hydropower capacities that can then be
further refined using simulation.
Alternatively, assumed average heads, H
t
o
, and flows,
Q
t
o
, can be used in a linear approximation of the non-
separable product terms, Q
T
t
H
t
:
Q
T
t
H
t
H
t
o
Q
T
t
Q
t
o
H
t
Q
t
o
H
t
o
(11.72)
Again, the model may need to be solved several times in
order to identify reasonably accurate average flow and
head estimates in each period t.
The amount of electrical energy produced is limited by
the installed kilowatts of plant capacity, P, as well as
on the plant factor, p
t
. The plant factor is a measure of
hydroelectric power plant use. Its value depends on the
characteristics of the power system and the demand
pattern for hydroelectric energy. The plant factor is defined
as the average power load on the plant for the period
divided by the installed plant capacity. The plant factor
accounts for the variability in the demand for hydropower
during each period t. This factor is usually specified by
those responsible for energy production and distribution. It
may or may not vary for different time periods.
The total energy produced cannot exceed the product
of the plant factor, p
t
, the number of hours, h
t
, in the
period, and the plant capacity, P, measured in kilowatts:
KWH
t
Ph
t
p
t
(11.73)
3.5. Flood Risk Reduction
Two types of structural alternatives exist for flood risk reduc-
tion. One is flood storage capacity in reservoirs designed to
reduce downstream peak flood flows. The other is channel
enhancement and/or flood-proofing works that will contain
wrm_ch11.qxd 8/31/2005 11:57 AM Page 359
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
peak flood flows and reduce damage. This section introduces
methods of modelling both of these alternatives for inclusion
in either benefitcost or cost-effectiveness analyses. The latter
analyses apply to situations in which a significant portion of
the flood control benefits cannot be expressed in monetary
terms and the aim is to provide a specified level of flood
protection at minimum cost.
The discussion will first focus on the estimation of
flood storage capacity in a single reservoir upstream of a
potential flood damage site. This analysis will then be
expanded to include downstream channel capacity
improvements. Each of the modelling methods discussed
will be appropriate for inclusion in multipurpose river
basin planning (optimization) models having longer time
step durations than those required to predict flood peak
flows. Additional detail can be found in Appendix D.
3.5.1. Reservoir Flood Storage Capacity
A common approach for estimating reservoir flood stor-
age capacity is based on expected damage reduction
associated with various storage capacities in a particular
reservoir. A relationship between flood storage capacity
and expected flood damage reduction can be obtained.
This functional relationship can then be used within a
multipurpose reservoir model that considers other uses
of the reservoir and its stored water. This approach
avoids having to include short time steps applicable to
most floods in an overall multipurpose river basin model.
Consider a reservoir upstream of a potential flood
damage site along a river. The question is how much
flood storage capacity, if any, should it contain. For vari-
ous assumed capacities, simulation models can be used to
predict the impact on the downstream flood peaks. These
hydraulic simulation models must include flood routing
procedures from the reservoir to the downstream poten-
tial damage site and the flood control operating policy at
the reservoir. For various downstream flood peaks, eco-
nomic flood damages can be estimated. To calculate the
expected annual damages associated with any upstream
reservoir capacity, the probabilities of various damage
levels being exceeded in any year need to be calculated.
The expected annual flood damage at a potential
flood damage site can be estimated from an exceedance
probability distribution of peak flood flows at that poten-
tial damage site together with a flow or stage damage
360 Water Resources Systems Planning and Management
function. The peak flow exceedance distribution at any
potential damage site will be a function of the upstream
reservoir flood storage capacity K
f
and the reservoir oper-
ating policy. The computational process of determining
the probability of exceedance of particular flood damages
associated with a particular flood storage capacity and
operating policy is illustrated graphically in Figure 11.22.
The analysis requires three input functions that are shown
in quadrants (a), (b) and (c). The dashed-line rectangles
define point values on the three input functions in quad-
rants (a), (b) and (c) and the corresponding probabilities
of exceeding a given level of damages in the lower right
quadrant (d). The distribution in quadrant (d) is defined
by the intersections of these dashed-line rectangles. This
distribution defines the probability of equalling or
exceeding a specified damage in any given year. The
shaded area under the derived function is the annual
expected damage, E[FD].
The relationships between flood stage and damage, and
flood stage and peak flow, defined in quadrants (a) and (b)
of Figure 11.22, must be known. These do not depend on
the flood storage capacity in an upstream reservoir. The
information in quadrant (c) defines the exceedance
probabilities of each peak flow. Unlike the other two
functions, this distribution depends on the upstream flood
storage capacity and flood flow release policy. This peak
flow probability of exceedance distribution is determined
by simulating the annual floods entering the upstream
reservoir in the years of record.
The difference between the expected annual flood dam-
age without any upstream flood storage capacity and the
expected annual flood damage associated with a flood
storage capacity of K
f
is the expected annual flood
damage reduction. This is illustrated in Figure 11.23.
Knowing the expected annual damage reduction associated
with various flood storage capacities, K
f
, permits the defini-
tion of a flood damage reduction function, B
f
(K
f
).
If the reservoir is a single-purpose flood control reser-
voir, the eventual tradeoff is between the expected flood
reduction benefits, B
f
(K
f
), and the annual costs, C(K
f
), of
that upstream reservoir capacity. The particular reservoir
flood storage capacity that maximizes the net benefits,
B
f
(K
f
) C(K
f
), may be appropriate from a national eco-
nomic efficiency perspective but it may not be best from
a local perspective. Those occupying the potential damage
site may prefer a specified level of protection from that
wrm_ch11.qxd 8/31/2005 11:57 AM Page 360
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 361
exceedance probability
flood stage
E
0
2
0
1
2
1
v
flood flow damage $
high high
high
high
low
Figure 11.22. Calculation of
the expected annual flood
damage shown as the shaded
area in quadrant (D) derived
from the expected stage-
damage function (A), the
expected stage-flow relation
(B), and the expected
probability of exceeding an
annual peak flow (C).
exceedance probability
flood stage
K = 50
f
K = 0
f
expected
annual
flood
damage
reduction
benefits
E
0
2
0
1
2
1
w
flood flow damage $
Figure 11.23. Calculation of
expected annual flood
damage reduction benefits,
shown as the darkened
portion of quadrant (D),
associated with a specified
reservoir flood storage
capacity.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 361
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
reservoir storage capacity, rather than the protection that
maximizes expected annual net benefits, B
f
(K
f
) C(K
f
).
If the upstream reservoir is to serve multiple purposes,
say for water supply, hydropower and recreation as well
as for flood control, then the expected flood reduction
benefit function just derived could be a component in the
overall objective function for that reservoir. Alternatively,
the flood storage capacity, K
f
, can be based on a particu-
lar predefined level of flood protection, regardless of the
damage reduction. Again, both methods are discussed in
greater detail in Appendix D.
Total reservoir capacity K will equal the sum of dead
storage capacity K
d
, active storage capacity K
a
and flood
storage capacity K
f
, assuming they are the same in each
period t. In some cases they may vary over the year. If the
required active storage capacity can occupy the flood stor-
age zone when flood protection is not needed, then the
total reservoir capacity K will be the dead storage, K
d
, plus
the maximum of either the actual storage volume and
flood storage capacity in the flood season, or the actual
storage volume in non-flood season.
K K
d
S
t
K
f
for all periods t in flood season plus
the following period that represents the end of the flood
season (11.74)
K K
d
S
t
for all remaining periods t (11.75)
In the above equations, the dead storage capacity, K
d
,
is a known variable. It is included in the capacity Equations
11.74 and 11.75, assuming that the active storage capacity
is greater than zero. Clearly, if the active storage capacity
were zero, there would be no need for dead storage.
3.5.2. Channel Capacity
The unregulated natural peak flow of a particular design
flood at a potential flood damage site can be reduced by
upstream reservoir flood storage capacity, or it can be
contained within the channel at the potential damage site
by levees and other channel-capacity improvements. Of
course both options may be applicable at some sites.
What is needed is a function defining the reduction in
the peak stage of some design flood and the cost of
providing additional channel capacity, K
c
. This function
can then be included in a model along with the upstream
flood storage capacity, K
f
, to determine the best combi-
nation of alternative flood control measures that will
362 Water Resources Systems Planning and Management
protect the potential damage site from a predetermined
design flood.
For example, consider a single-purpose flood control
reservoir and channel capacity enlargement, either by
dredging or dykes, whichever is cheaper, as alternatives for
protecting from a design flood, Q
T
, having a return period
of T. Defining the functional relationship between K
f
and
peak flow reduction at the damage site, and defining K
c
as
the peak flow reduction due to channel capacity enlarge-
ment, then the least cost combination can be estimated
from solving the following optimization model:
minimize Cost
R
(K
f
) Cost
C
(K
c
) (11.76)
subject to
Q
T
f
T
(K
f
) K
c
(11.77)
Equations 11.76 and 11.77 assume that a decision will be
made to provide protection from a design flood Q
T
.
It is only a question of how to provide the required
protection.
Solving Equations 11.76 and 11.77 for peak flows Q
T
of various design floods of return period T will identify
the riskcost tradeoff. This tradeoff function might look
like what is shown in Figure 11.24.
3.6. Lake-Based Recreation
Recreation benefits derived from natural lakes as well as
reservoirs are usually dependent on their storage levels. If
docks, boathouses, shelters and other recreational facili-
ties are installed on the basis of some assumed (target)
E
0
2
0
1
2
1
x
c
o
s
t

o
f

f
l
o
o
d
p
r
o
t
e
c
t
i
o
n
flood return period T
0
0
Figure 11.24. Tradeoff between minimum cost of flood
protection and flood risk, as expressed by the expected return
period.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 362
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 363
lake level, and the lake levels deviate from the target
value, there can be reduced recreational benefits. These
storage targets and any deviations can be modelled in a
way similar to Equation 11.66. The actual storage volume
at the beginning of a recreation period t equals the target
storage volume less any deficit or plus any excess;
S
t
s
T
s
D
t
s
E
t
s
(11.78)
The recreational benefits in any recreational period t can
be defined on the basis of what they would be if the
target were met, less the average of any losses that may
occur from initial and final storage volume deviations, D
t
s
or E
t
s
, from the target storage volume in each period of the
recreation season (Chapter 10).
4. River Basin Analysis
4.1. Model Synthesis
Each of the model components discussed above can be
combined, as applicable, into a model of a river system.
One such river system together with some of its interested
stakeholders is shown in Figure 11.25.
One of the first tasks in modelling this basin is to iden-
tify the actual and potential system components and their
interdependencies. This is facilitated by drawing a
schematic of the system at the level of detail that will
address the issues being discussed and of concern to these
stakeholders. This schematic can be drawn over the basin
as in Figure 11.26. The schematic without the basin is
illustrated in Figure 11.27.
A site number must be assigned at each point of inter-
est. These sites are usually where some decision must be
made. Mass-balance and other constraints will need to be
defined at each of those sites.
Figure 11.25. A multipurpose river system whose
management is of concern to numerous stakeholders (with
permission from Engineering News Record, McGraw Hill).
Figure 11.26. A schematic representation of the basin
components and their interdependencies drawn over the map
image of the basin.
E
0
2
0
1
3
0
k
1
2
3
4
5
gauge
irrigation
urban area
levee
protection
pumped storage hydro-power
water supply
recreation flood storage
Figure 11.27. Schematic of river system showing components
of interest at designated sites.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 363
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
364 Water Resources Systems Planning and Management
As shown in the schematic in Figure 11.27, this river
has one streamflow gauge site, site 1, two reservoirs, sites 3
and 5, two diversions, sites 2 and 3, one hydropower plant,
site 5, and a levee desired at site 4 to help protect against
floods in the urban area. The reservoir at site 5 is a pumped
storage facility. The upstream reservoir at site 3 is used for
recreation, water supply, and flood control. The down-
stream reservoir is strictly for hydropower production.
Before developing a model of this river system, the
number of within-year time periods t to include in the
model and the length of each within-year time period
should be determined. If a river systems reservoirs are to
contain storage for the distribution of water among differ-
ent years, called over-year storage, then a number of peri-
ods encompassing multiple years of operation must be
included in the model. This will allow an evaluation of the
possible benefits of storing excess water in wet years for
release in dry years.
Many reservoir systems completely fill up almost every
year, and in such cases one is concerned only with the
within-year operation of the system. This is the problem
addressed here. To model the within-year operation of
the system, a year is divided into a number of within-year
periods. The number of periods and the duration of each
period will depend on the variations in the hydrology,
the demands and the particular objectives, as previously
discussed.
4.2. Modelling Approach Using Optimzation
Once the number and duration of the time steps to be
modelled have been identified, the variables and func-
tions used at each site must be named. It is convenient to
use notation that can be remembered when examining the
model solutions. The notation made up for this example
is shown in Table 11.5. Economic efficiency objective
function components are defined in Table 11.6.
The objective components listed in Table 11.6 are eco-
nomic efficiency objectives. The LD and LE loss functions
are associated with deficit and excess allocations, respec-
tively. These and the other functions (whose meaning
should be self evident) need to be known, of course. There
could, and no doubt should, be other objective components
defined as well, as discussed in Chapter 10. Nevertheless
these objective components serve the purpose here of illus-
trating how a model of this river system can be constructed.
The overall objective might be a weighted combination
of all objective components such as:
(11.79)
This objective function does not identify how much each
stakeholder group would benefit and how much they
would pay. Who benefits and who pays, and how much,
may be important. If it is known how much of each of the
Maximize w NB
s
s
s

design and operating variables and parameters

all sites and time periods : s t
natural streamflows, based on gage flows at site , (t) 1 Q
site 2
irrigation allocations, (t), for all periods
annual irrigation target allocation,
known fraction of annual irrigation target required for each period
irrigation diversion channel capacity,
X t
T
t,
X
site 3
active initial storage volume. S (t), in each period t
dead storage volume, K
reservoir release downstream, R (t), in each period t
recreation storage volume target, T
deficit and excess storage volumes, D (t) and E (t), for each period t
flood storage volume capacity, K
total reservoir storage capacity, K
urban diversions, X (t), in each period t
diversion capacity, X
site 4
channel flood flow capacity, Q
water supply target, T
known fraction of annual water supply target for each period
site 5
active initial storage volume, S (t), in each period t
dead storage volume, K
reservoir release through turbines, QO (t), in each period t
quantity of water pumped back into reservoir, QI (t), in each period t
energy produced, EP (t), in each period t
energy consumed, EC (t), in each period t
total storage capacity of reservoir, K
power plant/pump capacity, P
storage head function, h(S (t)), in each period t
average storage head, H (t), in each period t
E
0
2
0
9
1
2
a
s
2
2
2
3
3
d
3
3
3 3
3
f
3
3
3
4
4
5
D
5
5
5
5
5
5
5
5
5
2
t
t,
4
t
Table 11.5. Names associated with required variables and
functions at each site in Figure 11.27. The units of these
variables and parameters, however defined, must be
consistent.
wrm_ch11.qxd 12/7/2005 10:04 AM Page 364
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 365
net benefits derived from each site are to be allocated to
each stakeholder group i, then these allocated portions,
denoted as NB
i
s
, of the total net benefits, NB
s
, can be
included in the overall objective:
(11.80)
Using methods discussed in Chapter 10, solving the model
for various assumed values of these weights, w
s
or w
i
, can
help identify the tradeoffs between different conflicting
objectives, Equation 11.79, or conflicting stakeholder
interests, Equation 11.80.
The next step in model development is to define the
constraints applicable at each site. It is convenient to
begin at the most upstream sites and work downstream.
As additional variables or functions are needed, invent
notation for them. These constraints tie the decision-
variables together and identify the interdependencies
among system components. In this example, Equations
11.81 through 11.97 (see Box 11.2), together with objec-
tive Equation 11.79 or 11.80, define the general structure
of this river system model.
Before the model can be solved the actual functions
must be defined. Then they may have to be made piece-
wise linear if linear programming is to be the optimization
procedure used to solve the model. The process of
defining functions may add variables and constraints to
the model, as discussed in Chapters 4 and 10.
For T within-year periods t, this static model of a
single year includes between 14T 8 and 16T 5
Maximize w NB
i i
s
s i

constraints, depending on the number of periods in the
irrigation and recreation seasons. This number does not
include the additional constraints that will surely be
needed to define the functions in the objective function
components and constraints. Models of this size and
complexity, even though this is a rather simple river
system, are usually solved using linear programming
algorithms simply because other non-linear or dynamic
programming (optimization) methods are more difficult
to implement.
The model just developed is for a typical single year.
In some cases it may be more appropriate to incorporate
over-year as well as within-year mass-balance constraints,
and yields with their respective reliabilities, within this
modelling framework. This can be done as outlined in
Sections 2.2.4 and 2.2.5 of this chapter.
The information derived from optimization models of
river systems such as this one should not be considered as
a final answer. Rather, it is an indication of system design
and operation policies that should be further analysed
using more detailed analyses. Optimization models of the
type just developed serve as ways to eliminate inferior
alternatives from further consideration rather than as
ways of finding a solution that all stakeholders will accept
as the best.
4.3. Modelling Approach Using Simulation
A simulation approach basically uses the same water
balance equations and constraints as given in the text box.
Table 11.6. Economic efficiency (net benefit)
maximization objective components.
wrm_ch11.qxd 12/7/2005 10:04 AM Page 365
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
366 Water Resources Systems Planning and Management
In a simulation approach, the analyst will have to specify
the values of the decision variables (e.g. the diversion
policies). The best values of these decision-variables will
have to be determined by sensitivity analyses, i.e. by
changing these variables and seeing what the output will
be in terms of the objective functions. Hence, simulation
is a trial-and-error approach.
A simulation model of the simple system as given in
Figure 11.27 can easily be developed by means of a
spreadsheet. More complicated systems can be simulated
by using generic computer packages for river basin
planning such as RIBASIM (WL | Delft Hydraulics, 2004),
Box 11.2. General Structure of the River
System Model
At site 1
No constraints are needed at this gauge site.
At site 2
The diverted water, X
2
(t), cannot exceed the
streamflow, Q
2
(t), at that site.
Q
2
(t) X
2
(t) t in the irrigation season (11.81)
The diversion flow, X
2
(t), cannot exceed the diver-
sion channel capacity, X
2
.
X
2
X
2
(t) t (11.82)
The diversion flow, X
2
(t), must meet the irrigation
target,
t
2
T
2
X
2
(t)
t
2
T
2
t (11.83)
At site 3
Storage volume mass balances (continuity of stor-
age), assuming no losses.
S
3
(t1) S
3
(t) Q
3
(t) X
3
(t) R
3
(t) t,
T 1 1 (11.84)
Define storage deficits, D
3
(t), and excesses, E
3
(t),
relative to recreation target, T
3
.
S
3
(t) T
3
D
3
(t) E
3
(t) t in recreation season
plus following period (11.85)
Diverted water, X
3
(t), cannot exceed diversion
channel capacity, X
3
.
X
3
(t) X
3
t (11.86)
Reservoir storage capacity constraints involving
dead storage, K
D
3
, and flood storage, K
F
3
, capacities.
S
3
(t) K
3
K
D
3
K
F
3
t in flood season plus
following period
S
3
(t) K
3
K
D
3
for all other periods t (11.87)
At site 4
Define deficit diversion, D
4
(t), from site 3, associ-
ated with target, d
4
t
T
4
, if any.
X
3
(t)
4
t
T
4
D
4
(t) t (11.88)
Channel capacity, Q
4
, for peak flood flow, PQ
T
4
, of
specified return period T.
Q
4
PQ
T
4
(11.89)
At site 5
Continuity of pumped storage volumes, involving
inflows, QI
5
(t), and outflows, QO
5
(t), and assuming
no losses.
S
5
(t1) S
5
(t) QI
5
(t) QO
5
(t) t (11.90)
Active storage capacity involving dead storage,
K
D
5
.
S
5
(t) K
5
K
D
5
t (11.91)
Pumped inflows cannot exceed the amounts of
water available at the intake; this includes the
release from the upstream reservoir, R
3
(t), and
the incremental flow, Q
5
(t) Q
3
(t).
QI
5
(t) Q
5
(t) Q
3
(t) R
3
(t) t (11.92)
Define the energy produced, EP
5
(t), given the
average storage head, H(t), flow through the
turbines, QO
5
(t), and efficiency, e.
EP
5
(t) (const.)(H(t))(QO
5
(t))e t (11.93)
Define the energy consumed, EC
5
(t), from pumping
given the amount pumped, QI
5
(t).
EC
5
(t) (const.)(H(t))(QI
5
(t))/e t (11.94)
Energy production, EP
5
(t), and consumption,
EC
5
(t), constraints given power plant capacity, P
5
.
EP
5
(t) P
5
(hours of energy production in t)
t (11.95)
EC
5
(t) P
5
(hours of pumping in t) t (11.96)
Define the average storage head, H
5
(t), based on
storage head functions, h(S
5
(t)).
H
5
(t) (h(S
5
(t1) h(S
5
(t))/2 t (11.97)
wrm_ch11.qxd 8/31/2005 11:57 AM Page 366
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 367
MIKE BASIN (DHI, 2003), and WEAP-21 (SEI, 2001).
These river basin simulation packages support the
development of a model schematization consisting of a
network of nodes connected by links. The nodes repre-
sent reservoirs, dams, weirs, pumps, hydro-power
stations, water users, inflows, artificial and natural
bifurcations, intake structures, natural lakes and so on.
The links transport water between the different nodes.
Such a network represents the basins features that are
significant for the planning and management problem at
issue. The network can be adjusted to provide the level of
spatial and temporal detail required. The river basin is
represented as a network schematization superimposed
over a vector or raster map image.
Figure 11.28 shows the network schematization of
RIBASIM of the simple river basin system of Figure 11.27.
Within each time-step a water balance calculation is made
in two phases:
1. The target setting phase, in which the water demands
are determined (calculated from specific modules for
irrigation, DMI, and other uses, or directly specified in
volumes or discharges), resulting in targets for the
releases from surface water reservoirs, aquifers and
diversion flows at weirs and pumping stations.
2. The water allocation phase, in which the allocation to
the users takes place according to the targets, avail-
ability of supply and allocation rules (priorities).
Water allocation to users can be implemented in a variety
of ways. In its simplest format, water is allocated on a first
come, first serve principle along the natural flow direc-
tion. This allocation can be amended by rules which, for
WD_upstream inflow a
WD_upstream inflow b
WD_upstream inflow c
WD_upstream inflow e
WD_upstream inflow f
WD_upstream inflow d
gauge-1
diversion-1
irrigation-1
confluence-1
confluence-2
SW_reservoir-1
pumping
SW_reservoir-2
low flow
PWS
confluence-3
diversion-2
confluence-4
terminal downstream
terminal 4
E
0
4
0
7
2
0
a
Figure 11.28. RIBASIM representation the
river basin schematization of Figure 11.27.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 367
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
example, allocate priority to particular users or result in
allocations proportional to demands. On the basis of a set
of simulations in which the values of important decision-
variables are changed, thus defining a range of alter-
native development or management strategies, thusthe
performance of the basin is evaluated in terms of water
allocation, shortages, energy production, overall river basin
water balance, crop yields, production costs and so on.
4.4. Optimization and/or Simulation
For determining the best design and operating policy
variable values in river basin systems such as the one just
illustrated, or even more complex ones, the use of both
optimization and simulation models can be advantageous.
As discussed in Chapter 3 (Section 2.2), optimization is
often useful, not for finding the very best solution, but for
eliminating the worst alternatives from further considera-
tion. The remaining ones can then be analysed in greater
detail, using more detailed simulation models.
Simulation by itself begs the question: What to simulate?
Optimization by itself begs the question: Is the solution
really the best?
There is an art to using optimization as a preliminary
screening tool. Some of this art is illustrated in Chapter 4,
Section 5.3.
4.5. Project Scheduling
The river basin models discussed thus far in this chapter
deal with static planning situations. Project capacities,
targets and operating policies take on fixed values and
one examines snapshot steady-state solutions for a par-
ticular year in the future. These snapshots only allow for
fluctuations caused by hydrological variability. The non-
hydrological world is seldom static, however. Demands
and targets change in response to population growth,
investment in agriculture and industry, and shifting pri-
orities for water use. In addition, financial resources avail-
able for water resources investment are limited and may
vary from year to year.
Dynamic planning models can aid those responsible
for the long-run development and expansion of water
resources systems. Although static models can identify
target values and system configuration designs for a
particular period in the future, they are not well adapted
368 Water Resources Systems Planning and Management
to long-run capacity expansion planning over a ten,
twenty or thirty-year period. But static models may
identify projects for implementation in early years that in
later year simulations do not appear in the solutions. This
was seen in some examples presented in Chapter 4.
This is a common problem in capacity expansion
where each project has a fixed construction and imple-
mentation cost, as well as variable operating, repair and
maintenance cost components. If there are two mutually
exclusive competing projects, one may be preferred at a
site when the demand at that site is low, but the other
may be preferred if the demand is, as it is later projected
to be, much higher. Which of the two projects should be
selected now when the demand is low, given the uncer-
tainty of the projected increase over time, especially
assuming it makes no economic sense to destroy and
replace a project already built?
Whereas static models consider how a water resources
system operates under a single set of fixed conditions,
dynamic expansion models must consider the sequence
of changing conditions that might occur over the plan-
ning period. For this reason, dynamic expansion models
are potentially more complex and larger than are their
static counterparts. However, to keep the size and cost of
dynamic models within the limitations of most studies,
these models are generally restricted to very simple
descriptions of the economic and hydrological variables
of concern. Most models use deterministic hydrology and
are constrained either to stay within predetermined
investment budget constraints or to meet predeter-
mined future demand estimates.
Dynamic expansion models can be viewed as network
models for solution by linear or dynamic programming
methods. The challenge in river system capacity expansion
or project scheduling models is that each components
performance, or benefits, may depend on the design and
operating characteristics of other components in the
system. River basin project impacts tend to be dependent
on what else is happening in the basin, that is, what other
projects are present and how they are operated.
Consider a situation in which n fixed-scale discrete
projects may be built during the planning period. The sched-
uling problem is to determine which projects to build or
implement and in what order. The solution of this problem
generally requires a resolution of the timing problem. When
should each project be built or implemented, if at all?
wrm_ch11.qxd 8/31/2005 11:57 AM Page 368
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 369
For example, assume there are n 3 discrete projects
that might be beneficial to implement sometime over the
next twenty years. This twenty-year period consists of
four five-year construction periods y. The actual benefits
derived from any new project may depend on the projects
that already exist. Let S be the set of projects existing at
the beginning of any construction period. Finally, let
NB
y
(S) be the maximum present value of the total net
benefits derived in construction period y associated with
the projects in the set S. Here, benefits refers to any
composite of system performance measures.
These benefit values for various combinations of dis-
crete projects could be obtained from static river system
models, solved for all combinations of discrete projects
for conditions existing at the end of each of these four
five-year periods. It might be possible to do this for just
one or two of these four periods and apply applicable
discount rates for the other periods. These static models
can be similar to those discussed in the previous section
of this chapter. Now the challenge is to find the sequenc-
ing of these three projects over the periods y that meet
budget constraints and that maximize the total present
value of benefits.
This problem can be visualized as a network. As
shown in Figure 11.29, the nodes of this network repre-
sent the sets S of projects that exist at the beginning of the
construction period. For these sets S, we have the present
value of their benefits, NB
y
(S), in the next five years. The
links represent the project or projects implemented in
that construction period. Any set of new projects that
exceeds the construction funds available for that period is
not shown on the network. Those links are infeasible. For
the purposes of this example, assume it is not financially
feasible to add more than one project in any single con-
struction period. Let C
ky
be the present value of the cost
of implementing project k in construction period y.
The problem is to find the best (maximum benefits less
costs) paths through the network. Each link represents a
net benefit, NB
y
(S), over the next five-years obtained from
the set of projects, S, that exist less the cost of adding a
new project k.
Using linear programming, define a continuous non-
negative unknown decision-variable X
ij
for each link
between node i and node j. It will be an indicator of
whether a link is on the optimum path or not. If after
solving the model its value is 1, then the link connecting
nodes i and j represents the decision to make in that
construction period. Otherwise, its value is 0 indicating
the link is not on the optimal path. The sequence of links
having their X
ij
values equal to 1 will indicate the most
beneficial sequence of project implementations.
Let the net benefits associated with node i be designated
NB
i
(which equals the appropriate NB
y
(S) value), and C
ij
the cost, C
ky
, of the new project k associated with that link.
The objective is to maximize the present value of net bene-
fits less project implementation costs over all periods y.
(11.98)
Subject to:
Continuity at each node:
(11.99)
Sum of all decision-variable values on the links in any one
period y must be 1. For example, in period 1:
X
00
X
01
X
02
X
03
1 (11.100)
The sums in Equation 11.99 are over nodes h having links
to node i and over nodes j having links from node i.
The optimal path through this network can also be
solved using dynamic programming. (Refer to the
capacity expansion problem illustrated in Chapter 4). For
a backward-moving solution procedure, let
s subset of projects k not contained in the set S (s S).
$
y
the maximum project implementation funds
available in period y.
F
y
(S) the present value of the total benefits over the
remaining periods, y, y1, , 4.
F
Y
(S) 0 for all sets of projects S following the end of
the last period.
The recursive equations for each construction period,
beginning with the last period, can be written
(11.101)
Defining F
y
(S) as the present value of the total benefits of
all new projects in the set S implemented in all periods up
C
ky y
k
$

s
F NB C F
y y ky y
k
( ) ( )} S S S s S
S
s

maximum{ (
s
)

1
X X i
hi ij
j h
for each node in the network.

Maximise ( ) NB C X
i ij ij
j i

wrm_ch11.qxd 8/31/2005 11:57 AM Page 369
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
370 Water Resources Systems Planning and Management
0
0
1
2
0 0 0
2,3 2,3 2,3
1,3 1,3 1,3
1,2 1,2 1,2
3 3 3
2 2 2
1 1 1
1,2,3 1,2,3
E
0
2
0
1
3
0
s
period 1 period 2 period 3 period 4
S
y
NB
S, k
project
cost =
k
C
ky
period y
( ) S
y+1
NB ( ) S, k
3
K
Figure 11.29. Project scheduling
options. Numbers in nodes
represent existing projects.
Links represent new projects,
the difference between the
existing projects at both
connecting nodes.
sequencing of projects recognizing their interdependen-
cies. Of interest, again, is what to do in this first
construction period. The only reason for looking into the
future is to make sure, as best as one can, that the first
periods decisions are not myopic. Models like these
can be developed and solved again with more updated
estimates of future conditions when next needed.
Additional constraints and variables might be added to
these scheduling models to enforce requirements that
some projects precede others, or that if one project is built
another is infeasible. These additional restrictions usually
reduce the size of a network of feasible nodes and links,
as shown in Figure 11.29.
to and including period y, and the subset s of projects k
being considered in period y now belonging to the set S
of projects existing at the end of the period, the recursive
equations for a forward-moving solution procedure
beginning with the first period, can be written:
(11.102)
where F
0
(0) 0.
Like the linear programming model, the solutions
of these dynamic programming models identify the
C
ky y
k
$

s
F NB C F
y y ky y
k
( ) ( )} S S S s S
S
s

maximum{ (
s
)

1
wrm_ch11.qxd 8/31/2005 11:57 AM Page 370
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
River Basin Planning Models 371
Another issue that these dynamic models can address
is the sizing or capacity expansion problem. Frequently,
the scale or capacity of a reservoir, pipeline, pumping sta-
tion or irrigation project is variable and needs to be deter-
mined concurrently with the solution of the scheduling
and timing problems. To solve the sizing problem, the
costs and capacities in the scheduling model become
variables.
5. Conclusions
This chapter on river basin planning models has
introduced some ways of modelling river basin
components, separately and together within an integrated
model. Ignored during the development of these different
model types were the uncertainties associated with the
results of these models. As discussed in Chapters 7
through 9, these uncertainties may have a substantial
effect on the model solution and the decision taken.
Most of this chapter has been focused on the
development of simplified screening models, using simu-
lation as well as optimization methods, for identifying
what and where and when infrastructure projects should
be implemented, and of what capacity. The solution of
these screening models, and any associated sensitivity and
uncertainty analyses, should be of value prior to
committing to a more costly design modelling exercise.
Preliminary screening of river basin systems, especially
given multiple objectives, is a challenge to accomplish in
an efficient and effective manner. The modelling methods
and approaches discussed in this chapter serve as an
introduction to that art.
6. References
DHI (Danish Hydraulic Institute). 2003. MIKE BASIN: a
toolfor river basin planning and management. User Manual
for MIKE BASIN 2003. Horsholm, Denmark, Danish
Hydraulic Institute. Also http://www.dhisoftware.
com/mikebasin/index.htm (accessed 11 November
2004).
FAO. 1998. Crop evapotranspiration: guidelines for
computing crop water requirements. R.G. Allen, L.S. Pereira,
D. Raes and M. Smith. FAO Irrigation and Drainage
Paper 56. Rome, FAO.
MAIDMENT, D.R. 1993. Handbook of hydrology. New York,
McGraw-Hill.
RIPPL, W. 1883. The capacity of storage reservoirs
for water supply. Proceedings of the Institute of Civil
Engineers (UK), Vol. 71, pp. 2708.
SEI, WEAP. 2001. User guide for WEAP21. Boston Mass.,
Stockholm Environment Institute, Tellus Institute, July;
also http://www.weap21.org/index.sp (accessed 11
November 2004).
USDA (US Department of Agriculture). 1972. National
engineering handbook. Washington, D.C., Soil Conservation
Service, US Government Printing Office.
USDOE (US Departement of Energy). 2002. Hydropower program.
http://hydropower.inel.gov or http://hydropower.id.doe. gov/,
accessed April.
WCD (World Commission on Dams). 2000. Dams and
development: a new framework for decision-making. The report
of the World Commission on Dams. London, Earthscan.
WL | DELFT HYDRAULICS. 2004. RIBASIM Version 6.32:
Technical Reference Manual. Also http://www.wldelft.nl/
soft/ribasim/int/index.html (accessed 11 November
2004).
Additional References (Further Reading)
BASSON, 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.
BECKER, L. and YEH, W.W.G. 1974. Optimal timing,
sequencing, and sizing of multiple reservoir surface water
supply facilities. Water Resources Research, Vol. 10, No. 1,
pp. 5762.
BECKER, L. and YEH, W.W.G. 1974. Timing and sizing of
complex water resource systems. Journal of the Hydraulics
Division, ASCE, Vol. 100, No. HYIO,pp. 145770.
BEDIENT, P.B. and HUBER, W.C. 1992. Hydrology and flood-
plain analysis, 2nd edn. Reading, Mass., Addison Wesley.
CHIN, D.A. 2000. Water-resources engineering. Upper
Saddle River, N.J., Prentice Hall.
ERLENKOTTER, D. 1973a. Sequencing expansion proj-
ects. Operations Research, Vol. 21, No. 2, pp. 54253.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 371
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
372 Water Resources Systems Planning and Management
ERLENKOTTER, D. 1973b. Sequencing of interdepen-
dent hydroelectric projects. Water Resources Research,
Vol. 9, No. 1, pp. 217.
ERLENKOTTER, D. 1976. Coordinating scale and
sequencing decisions for water resources projects:
economic modelling for water policy evaluation.
NorthHolland/TIMS Studies in the Management Sciences,
Vol. 3, pp. 97112.
ERLENKOTTER, D. and ROGERS, J.S. 1977. Sequencing
competitive expansion projects. Operations Research,
Vol. 25, No. 6, pp. 93751.
HALL, W.A.; TAUXE, G.W. and YEH, W.W.G. 1969. An
alternative procedure for optimization of operations for
planning with multiple river, multiple purpose systems.
Water Resources Research, Vol. 5, No. 6, pp. 136772.
JACOBY, H.D. and LOUCKS, D.P. 1972. Combined use of
optimization and simulation models in river basin plan-
ning. Water Resources Research, Vol. 8, No. 6, pp. 140114.
JAMES, L.D. and LEE, R.R. 1971. Economics of water
resources planning. New York, McGraw-Hill.
LINSLEY, R.K.; FRANZINI, J.B.; FREYBERG, D.L. and
TCHOBANOGLOUS, G. 1992. Water-resources engi-
neering. New York, McGraw-Hill.
LOUCKS, D.P. 1976. Surface water quantity manage-
ment, In: A. K. Biswas (ed.), Systems approach to water
management, Chapter 5. New York, McGraw-Hill.
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.
MAJOR, D.C. and LENTON, R.L. 1979. Applied water
resources systems planning. Englewood Cliffs, N.J., Prentice Hall.
MAYS, L.W. 2005. Water resources engineering. New York,
Wiley.
MAYS, L.W. and TUNG, Y.K. 1992. Hydrosystems engi-
neering and management. New York, McGraw-Hill.
MORIN, T.L. 1973. Optimal sequencing of capacity
expansion projects. Journal of the Hydraulics Division,
ASCE, Vol. 99, No. HY9, pp. 160522.
MORIN, T.L. and ESOGBUE, A.M.O. 1974. A useful
theorem in the dynamic programming solution of
sequencing and scheduling problems occurring in capital
expenditure planning. Water Resources Research, Vol. 10,
No. 1, pp. 4950.
NATIONAL RESEARCH COUNCIL. 1999. New strategies
for Americas watersheds. Washington, D.C., National
Academy Press.
NATIONAL RESEARCH COUNCIL. 2000. Risk analysis
and uncertainty in flood damage reduction studies.
Washington, D.C., National Academy Press.
OLAOGHAIRE, D.T. and HIMMELBLAU, D.M. 1974.
Optimal expansion of a water resources system. New York,
Academic Press.
REVELLE, C. 1999. Optimizing reservoir resources.
New York, Wiley.
THOMAS, H.A. Jr. and BURDEN, R.P. 1963. Operations
research in water quality management. Cambridge, Mass.,
Harvard Water Resources Group, 1963.
USACE. 1991a. Hydrology and hydraulics workshop on river-
ine levee freeboard. Monticello, Minnesota, Report SP-24.
Hydrological Engineering Center, US Army Corps of
Engineers.
USACE. 1991b. Benefit determination involving existing
levees. Policy Guidance Letter No. 26. Washington, D.C.,
Headquarters, US Army Corps of Engineers.
USACE. 1999. Risk analysis in geotechnical engineering for
support of planning studies. ETL 1110-2-556. Washington,
D.C., Headquarters, US Army Corps of Engineers.
VIESSMAN, W. Jr. and WELTY, C. 1985. Water manage-
ment technology and institutions. New York, Harper and Row.
WURBS, R.A. and JAMES, W.P. 2002. Water resources
engineering. Upper Saddle River, N.J., Prentice Hall.
wrm_ch11.qxd 8/31/2005 11:57 AM Page 372
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
12. Water Quality Modelling and Prediction
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
wrm_ch12.qxd 8/31/2005 11:59 AM Page 374
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
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
wrm_ch12.qxd 8/31/2005 11:59 AM Page 375
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Water quality management is a critical component of
overall integrated water resources management. Most
users of water depend on adequate levels of water quality.
When these levels are not met, these water users must
either pay an additional cost for water treatment or incur
at least increased risks of damage or loss. As populations
and economies grow, more pollutants are generated. Many
of these are waterborne, and hence can end up in surface
and groundwater bodies. Increasingly, the major efforts
and costs involved in water management are devoted to
water quality protection and management. Conflicts
among various users of water are increasingly over issues
involving water quality as well as water quantity.
Natural water bodies are able to serve many uses,
including the transport and assimilation of waterborne
wastes. But as natural water bodies assimilate these
wastes, their quality changes. If the quality drops to
the extent that other beneficial uses are adversely affected,
the assimilative capacities of those water bodies have been
exceeded with respect to those affected uses. Water
quality management measures are actions taken to ensure
377
Water Quality Modelling and Prediction
The most fundamental human needs for water are for drinking, cooking and
personal hygiene. To meet these needs, the quality of the water used must pose no
risk to human health. The quality of the water in nature also affects the condition
of ecosystems that all living organisms depend on. At the same time, humans use
water bodies as convenient sinks for the disposal of domestic, industrial and
agricultural wastewaters. This of course degrades the quality of those water
bodies. Water resources management involves the monitoring and management of
water quality as much as the monitoring and management of water quantity.
Various models have been developed to assist in predicting the water quality
impacts of alternative land and water management policies and practices. This
chapter introduces some of the main principles of water quality modelling.
that the total pollutant loads discharged into receiving
water bodies do not exceed the ability of those water
bodies to assimilate those loads while maintaining the
levels of quality specified by quality standards set for
those waters.
What uses depend on water quality? One can identify
almost any use. All living organisms require water of
sufficient quantity and quality to survive, although differ-
ent aquatic species can tolerate different levels of water
quality. Regrettably, in most parts of the developed world
it is no longer safe to drink natural surface or ground
waters; they usually need to be treated before they become
fit for human consumption. Treatment is not a practical
option for recreational bathing, or for maintaining the
health of fish, shellfish and other organisms found in
natural aquatic ecosystems. Thus, standards specifying
minimum acceptable levels of quality are set for most
ambient waters. Various other uses have their own stan-
dards as well. Irrigation water must not be too saline or
contain toxic substances that can be absorbed by the
plants or destroy microorganisms in the soil. Water qual-
ity standards for industry can be very demanding, depend-
ing of course on the particular industrial processes.
12
wrm_ch12.qxd 8/31/2005 11:59 AM Page 377
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Pollutant loadings degrade water quality. High
domestic wasteloads can result in high concentrations of
bacteria, viruses and other organisms that affect human
health. High organic loadings can reduce dissolved
oxygen to levels that are fatal to parts of the aquatic
ecosystem and cause obnoxious odours. Nutrient
loadings from both urban and agricultural land runoff can
cause excessive algae growth, which in turn may degrade
the water aesthetically and recreationally, and upon
death result in low dissolved oxygen levels. Toxic heavy
metals and other micro-pollutants can accumulate in
the bodies of aquatic organisms, including fish, making
them unfit for human consumption even if they them-
selves survive.
Pollutant discharges originate from point and non-
point sources. A common approach to controlling point
source discharges, such as those from stormwater outfalls,
municipal wastewater treatment plants or industries, is to
impose standards specifying maximum allowable pollu-
tant loads or concentrations in their effluents. This is
often done in ways that are not economically efficient
or even environmentally effective. Effluent standards
typically do not take into account the particular assimila-
tive capacities of the receiving water body.
Non-point sources such as agricultural runoff or atmos-
pheric deposition are less easily controlled, and hence it is
difficult to apply effluent standards to non-point source
pollutants, and their loadings can be much more signifi-
cant than point source loadings. Management of non-point
water quality impacts requires a more ambient-focused
water quality management programme.
The goal of an ambient water quality management
programme is to establish appropriate standards for
water quality in water bodies receiving pollutant loads,
and then to ensure that these standards are met. Realistic
standard-setting takes into account the basins hydro-
logical, ecological and land use conditions, the potential
uses of the receiving water body, and the institutional
capacity to set and enforce water quality standards.
Ambient-based water quality prediction and manage-
ment involves considerable uncertainty. No one can
predict what pollutant loadings will occur in the future,
especially from area-wide non-point sources. In addition
to uncertainties inherent in measuring the attainment
of water quality standards, there are uncertainties in
models used to determine sources of pollution, to allocate
378 Water Resources Systems Planning and Management
pollutant loads, and to predict the effectiveness of actions
taken to meet water quality standards. The models
available to help managers predict water quality impacts
are relatively simple compared with the complexities of
actual water systems. These limitations and uncertainties
should be understood and addressed as water quality
management decisions are based on their outputs.
2. Establishing Ambient Water
Quality Standards
Identifying the intended uses of a water body whether
a lake, a section of a stream or an estuary is a first step
in setting water quality standards for that body. The most
restrictive of the specific desired uses of a water body is
termed a designated use. Barriers to achieving the desig-
nated use are the presence of pollutants, or hydrological
and geomorphic changes that affect the water quality.
The designated use dictates the appropriate type
of water quality standard. For example, a designated use
of human recreation should protect humans from
exposure to microbial pathogens while swimming,
wading or boating. Other uses include those designed to
protect humans and wildlife from consuming harmful
substances in water, in fish and in shellfish. Aquatic-life
uses include the protection and propagation of fish,
shellfish and wildlife resources.
Standards set upstream may affect the uses of water
downstream. For example, small headwater streams
may have aesthetic value but may not be able to support
extensive recreational uses. However, their condition
may affect the ability of a downstream area to achieve a
particular designated use such as fishable or swim-
mable. In this case, the designated use for the smaller
upstream water body may be defined in terms of
achieving the designated use of the larger downstream
water body.
In many areas, human activities have altered the land-
scape and aquatic ecosystems to the point where they
cannot be restored to their pre-disturbance condition. For
example, someones desire to establish a trout fish-farm in
downtown Paris, Phnom Penh, Prague or Pretoria may
not be attainable because of the development history of
these areas or the altered hydrological regimes of the rivers
flowing through them. Similarly, someone might wish to
wrm_ch12.qxd 8/31/2005 11:59 AM Page 378
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 379
designate an area near the outfall of a sewage treatment
plant for shellfish harvesting, but health considerations
would preclude any such use. Ambient water quality
standards must be realistic.
Designating the appropriate use for a water body is a
policy decision that can be informed by the use of water
quality prediction models of the type discussed in this
chapter. However, the final standard selection should
reflect a social consensus made while bearing in mind the
current condition of the watershed, its pre-disturbance
condition, the advantages derived from a certain desig-
nated use, and the costs of achieving that use.
2.1. Water-Use Criteria
The designated use is a qualitative description of
the desired condition of a water body. A criterion is a
measurable indicator surrogate for use attainment. The
criterion may be positioned at any point in the causal
chain of boxes shown in Figure 12.1.
In Box 1 of Figure 12.1 are measures of the pollutant
discharge from a treatment plant (such as biological
oxygen demand, ammonia (NH
3
), pathogens and sus-
pended sediments) or the amount of a pollutant entering
the edge of a stream from runoff. A criterion at this posi-
tion is referred to as an effluent standard. Criteria in Boxes
2 and 3 are possible measures of ambient water quality
conditions. Box 2 includes measures of a water quality
parameter such as dissolved oxygen (DO), pH, total phos-
phorus concentration, suspended sediment or tempera-
ture. Criteria closer to the designated use (e.g. Box 3)
include more combined or comprehensive measures
of the biological community as a whole, such as the
condition of the algal community (chlorophyll a) or
a measure of contaminant concentration in fish tissue.
Box 4 represents criteria that are associated with sources
of pollution other than pollutants. These criteria might
include measures such as flow timing and pattern
(a hydrological criterion), abundance of non-indigenous
taxa, or some quantification of channel modification
(e.g. a decrease in sinuosity) (NRC, 2001).
The more precise the statement of the designated use,
the more accurate the criterion will be as an indicator of
that use. For example, the criterion of fecal coliform count
may be a suitable criterion for water contact recreation.
The maximum allowable count itself may differ among
water bodies that have water contact as their designated
use, however.
Surrogate indicators are often selected for use as
criteria because they are easy to measure and in some
cases are politically appealing. Although a surrogate
indicator may have these appealing attributes, its useful-
ness can be limited unless it can be logically related to
a designated use.
As with setting designated uses, the connections among
water bodies and segments must be considered when
determining criteria. For example, where a segment of a
water body is designated as a mixing zone for a pollutant
discharge, the criterion adopted should assure that the
mixing zone use will not adversely affect the surrounding
water body uses. Similarly, the desired condition of a small
headwater stream may need to be specified in relation to
other water bodies downstream; thus, an ambient nutrient
criterion may be set in a small headwater stream to
ensure a designated use in a downstream estuary, even if
there are no local adverse impacts resulting from the nutri-
ents in the small headwater stream, as previously
discussed. Conversely, a high fecal coliform criterion
may be permitted upstream of a recreational area if the
fecal load dissipates before the flow reaches that area.
3. Water Quality Model Use
Monitoring data are the preferred form of information
for identifying impaired waters (Appendix B). Model
predictions might be used in addition to or instead of
4
pollutant load from each source
ambient pollutant concentration in waterbody
human health and biological condition
2
3
land use,
hydrological,
and
ecological
characteristics
of waterbody
appropriate designated use for the waterbody
E
0
2
0
7
3
0
a
Figure 12.1. Factors considered when determining designated
use and associated water quality standards.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 379
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
monitoring data for several reasons:
Modelling might be feasible in some situations where
monitoring is not.
Integrated monitoring and modelling systems could
provide better information than one or the other
alone for the same total cost. For example, regression
analyses that correlate pollutant concentration with
some more easily measurable factor (such as stream-
flow) could be used to extend monitoring data for
preliminary listing (of impared status) purposes.
Models can also be used in a Bayesian framework
to determine preliminary probability distributions of
impairment that can help direct monitoring efforts and
reduce the quantity of monitoring data needed for
making listing decisions at a given level of reliability
(see Chapter 7).
Modelling can be used to assess (predict) future water
quality situations resulting from different management
strategies. For example, assessing the improvement in
water quality after a new wastewater treatment plant is
built, or the effect of increased industrial growth and
effluent discharges.
A simple but useful modelling approach that may be
used in the absence of monitoring data is dilution calcula-
tions. In this approach the rate of pollutant loading from
point sources in a water body is divided by the streamflow
to give a set of estimated pollutant concentrations that may
be compared to the standard. Simple dilution calculations
assume conservative movement of pollutants. Thus, the
use of dilution calculations will tend to be conservative
and predict higher than actual concentrations for decaying
pollutants. Of course, one could include a best estimate of
the effects of decay processes in the dilution model.
Combined runoff and water quality prediction models
link stressors (sources of pollutants and pollution) to
responses. Stressors include human activities likely to
cause impairment, such as the presence of impervious
surfaces in a watershed, cultivation of fields close to the
stream, over-irrigation of crops with resulting polluted
return flows, the discharge of domestic and industrial
effluents into water bodies, installing dams and other
channelization works, introduction of non-indigenous
taxa and over-harvesting of fish. Indirect effects of humans
include land cover changes that alter the rates of
delivery of water, pollutants and sediment to water bodies.
380 Water Resources Systems Planning and Management
A review of direct and indirect effects of human
activities suggests five major types of environmental
stressors:
alterations in physical habitat
modifications in the seasonal flow of water
changes in the food base of the system
changes in interactions within the stream biota
release of contaminants (conventional pollutants)
(Karr, 1990; NRC, 1992, 2001).
Ideally, models designed to manage water quality should
consider all five types of alternative management
measures. A broad-based approach that considers these
five features provides a more integrative approach to
reduce the cause or causes of degradation (NRC, 1992).
Models that relate stressors to responses can be of
varying levels of complexity. Sometimes, they are sim-
ple qualitative conceptual representations of the
relationships among important variables and indicators
of those variables, such as the statement human activi-
ties in a watershed affect water quality, including the
condition of the river biota. More quantitative models
can be used to make predictions about the assimilative
capacity of a water body, the movement of a pollutant
from various point and non-point sources through a
watershed, or the effectiveness of certain best manage-
ment practices.
3.1. Model Selection Criteria
Water quality predictive models include both mathe-
matical expressions and expert scientific judgement. They
include process-based (mechanistic) models and
data-based (statistical) models. The models should link
management options to meaningful response variables
(such as pollutant sources and water quality standard
parameters). They should incorporate the entire chain
from stressors to responses. Process-based models should
be consistent with scientific theory. Model prediction
uncertainty should be reported. This provides decision-
makers with estimates of the risks of options. To do this
requires prediction error estimates (Chapter 9).
Water quality management models should be appro-
priate to the complexity of the situation and to the
available data. Simple water quality problems can be
addressed with simple models, while complex ones may
wrm_ch12.qxd 8/31/2005 11:59 AM Page 380
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 381
or may not require the use of more complex models.
Models requiring large amounts of monitoring data
should not be used in situations where such data are
unavailable. Models should be flexible enough to allow
updates and improvements as appropriate based on new
research and monitoring data.
Stakeholders need to accept the models proposed
for use in any water quality management study. Given
the increasing role of stakeholders in water management
decision processes, they need to understand and accept
the models being used, at least to the extent they wish to
do so. Finally, the cost of maintaining and updating the
model over time must be acceptable.
Although predictions are typically made with the aid
of mathematical models, there are certainly situations
where expert judgement can be just as good. Reliance
on professional judgement and simpler models is often
acceptable, especially when data are limited.
Highly detailed models require more time and
are more expensive to develop and apply. Effective and
efficient modelling for water quality management may
dictate the use of simpler models. Complex modelling
studies should be undertaken only if warranted by the
complexity of the management problem. More complex
modelling will not necessarily ensure that uncertainty is
reduced, and in fact added complexity can compound
problems of uncertainty analyses (Chapter 9).
Placing a priority on process description usually leads
to the development and use of complex mechanistic
models rather than simpler mechanistic or empirical
models. In some cases this may result in unnecessarily
costly analyses. In addition, physical, chemical and bio-
logical processes in terrestrial and aquatic environments
are far too complex to be fully represented in even
the most complicated models. For water quality manage-
ment, the primary purpose of modelling should be to
support decision-making. The inability to describe all
relevant processes completely contributes to the uncer-
tainty in the model predictions.
3.2. Model Chains
Many water quality management analyses require the use
of a sequence of models, one feeding data into another.
For example, consider the sequence or chain of models
required for the prediction of fish and shellfish survival as
a function of nutrient loadings into an estuary. Of interest
to the stakeholders are the conditions of the fish and
shellfish. One way to maintain healthy fish and shellfish
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 limits on 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.
The negative effects of excessive nutrients (e.g. nitro-
gen) in an estuary are shown in Figure 12.2. 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 lower levels of water. Fish
and shellfish may die or become weakened and more
vulnerable to disease.
A model consisting of a sequence of conditional
probabilities can be defined to predict the probability of
shellfish and fish abundance based on upstream nutrient
loadings into the estuary that might cause problems
for fish and shellfish populations. These conditional
probabilities can be judgemental, mechanistic and/or sta-
tistical. Each conditional probability can be a separate
sub-model. Assuming each sub-model can identify a
conditional probability distribution, the probability Pr{C
| N} of a specified amount of carbon, C, given some spec-
ified loading of a nutrient, say nitrogen, N, equals the
probability Pr{C | A} of that given amount of carbon,
given a concentration of algae biomass, A, times the prob-
ability Pr{A | N, R} of that concentration of algae biomass
given the nitrogen loading, N, and the river flow, R,
times the probability Pr{R} of the river flow, R. In other
words:
Pr{C | N} Pr{C | A}Pr{A | N, R}Pr{R} (12.1)
An empirical process-based model of the type to be
presented later in this chapter could be used to predict
the concentration of algae and the chlorophyll violations
on the basis of the river flow and nitrogen loadings. It
could similarly predict the production of carbon, on the
basis of algae biomass. A seasonal statistical regression
model might be used to predict the likelihood of algae
blooms based on algal biomass. A cross-system compari-
son may be made to predict sediment oxygen demand.
A relatively simple hydraulic model could be used to
wrm_ch12.qxd 8/31/2005 11:59 AM Page 381
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
predict the duration of stratification and the frequency
of hypoxia, given both the stratification duration and
sediment oxygen demand. Expert judgement and fish
survival models could be used to predict the shellfish
abundance and fishkill and fish health probabilities.
The biological endpoints shell-fish survival and
number of fishkills, are meaningful indicators to stake-
holders and can easily be related to designated water body
use. Models and even conditional probabilities assigned
to each link of the network in Figure 12.3 can reflect
a combination of simple mechanisms, statistical (regres-
sion) fitting and expert judgement.
Advances in the mechanistic modelling of aquatic
ecosystems have enabled us to include greater process
(especially trophic) detail and complexity, as well as to
perform dynamic simulations, although mechanistic
ecosystem models have not advanced to the point of
being able to predict community structure or biotic
integrity. In this chapter, only some of the simpler
mechanistic models will be introduced. More detail can
be found in books solely devoted to water quality
382 Water Resources Systems Planning and Management
modelling (Chapra, 1997; McCutcheon, 1989; Orlob,
1983; Schnoor, 1996; Thomann and Mueller, 1987) as
well as the current professional journal literature.
3.3. Model Data
Data availability and accuracy are sources of concern
in the development and use of models for water quality
management. The complexity of models used for water
quality management should be compatible with the
quantity and quality of available data. The use of complex
mechanistic models for water quality prediction in
situations with little useful water quality data does not
compensate for that lack of data. Model complexity
can give the impression of credibility, but this is usually
misleading.
It is often preferable to begin with simple models and
then, over time, add additional complexity as justified
by the collection and analysis of additional data. This
strategy makes efficient use of resources. It targets the
effort toward information and models that will reduce
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 12.2. The negative
impacts of excessive nutrients
in an estuary (NRC, 2001).
wrm_ch12.qxd 8/31/2005 11:59 AM Page 382
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 383
the uncertainty as the analysis proceeds. Models should
be selected (simple versus complex) in part on the basis
of the data available to support their use.
Water quality models of water bodies receiving
pollutant discharges require those pollutant loadings as
input data. These pollutant discharges can be from point
and non-point sources. Point source discharges are much
easier to measure, monitor and estimate than non-point
source inputs. Non-point discharge data often come
from rainfallrunoff models that attempt to predict the
quantity of runoff and its constituent concentrations. The
reliability of the predictions from these models is not very
good, especially if short time periods (e.g. each day or
week) are being simulated. Their average values over
longer time periods (e.g. a month or year) tend to be
more reliable. This is mainly because the short-term
inputs to those models, such as constituent loadings on
the land and the rainfall within an area, can vary over
space and time within the area and time period being
simulated, and are typically not known with any preci-
sion. Chapter 13 reviews some of these loading models
and their limitations.
4. Water Quality Model Processes
Water quality models can be applied to many different
types of water system, including streams, rivers, lakes,
reservoirs, estuaries, coastal waters and oceans. The models
describe the main water quality processes, and typically
require the hydrological and constituent inputs (the water
flows or volumes and the pollutant loadings). These
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 12.3. Cause and effect diagram for
estuary eutrophication due to excessive
nutrient loadings (Borsuk et al., 2004).
wrm_ch12.qxd 8/31/2005 11:59 AM Page 383
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
models include terms for dispersive and/or advective trans-
port depending on the hydrological and hydrodynamic
characteristics of the water body, and terms for the biolog-
ical, chemical and physical reactions among constituents.
Advective transport dominates in flowing rivers. Dispersion
is the predominant transport phenomenon in estuaries
subject to tidal action. Lake-water quality prediction is
complicated by the influence of random wind directions
and velocities that often affect surface mixing, currents and
stratification. For this and other reasons, obtaining reliable
quality predictions for lakes is often more difficult than for
streams, rivers and estuaries. In coastal waters and oceans,
large-scale flow patterns and tide are the most important
transport mechanisms.
The development and application of water quality
models is both a science and an art. Each model reflects
the creativity of its developer, the particular water quality
management problems and issues being addressed, the
available data for model parameter calibration and
verification, the time available for modelling and associ-
ated uncertainty, and other considerations. The fact that
most, if not all, water quality models cannot accurately
predict what actually happens does not detract from their
value. Even relatively simple models can help managers
understand the real world prototype and estimate at least
the relative, if not actual, change in water quality associ-
ated with given changes in the inputs resulting from
management policies or practices.
4.1. Mass-Balance Principles
The basic principle of water quality models is that of mass
balance. A water system can be divided into different
segments or volume elements, also called computational
cells. For each segment or cell, there must be a mass
balance for each water quality constituent over time. Most
water quality simulation models simulate quality over a
consecutive series of discrete time periods, t. Time is
divided into discrete intervals t and the flows are assumed
constant within each of those time period intervals. For
each segment and each time period, the mass balance of
a substance in a segment can be defined. Components of
the mass balance for a segment include: first, changes by
transport (Tr) into and out of the segment; second, changes
by physical or chemical processes (P) occurring within the
384 Water Resources Systems Planning and Management
segment; and third, changes by sources/discharges to or
from the segment (S).
(12.2)
The mass balance has the following components:
the mass in computational cell i at the beginning of a
time step t:
the mass in computational cell i at the end of a time
step t:
changes in computational cell i
by transport:
changes in computational cell i by physical,
(bio)chemical or biological processes:
changes in computational cell i by sources
(e.g. wasteloads, river discharges):
Changes by transport include both advective and disper-
sive transport. Advective transport is transport by flowing
water. Dispersive transport results from concentration
differences. Dispersion in the vertical direction is impor-
tant if the water column is stratified, and dispersion in the
horizontal direction can be in one or two dimensions.
Dispersion, as defined here, differs from the physical
concept of molecular diffusion as it stands for all
transport that is not advective.
Changes by processes include physical processes such
as re-aeration and settling, (bio)chemical processes
such as adsorption, transformation and denitrification,
and biological processes such as primary production and
predation on phytoplankton. Water quality processes
convert one substance to another.
Changes by sources include the addition of mass by
wasteloads and the extraction of mass by intakes. Mass
entering over the model boundaries can be considered
a source as well. The water flowing into or flowing out
of the modelled segment or volume element (the compu-
tational cell) is derived from a water quantity (possibly
hydrodynamic) model.

M
t
i
S

_
,

M
t
i
P

_
,

M
t
i
Tr

_
,

M
i
t t
M
i
t
M M t
M
t
t
M
t
t
M
t
i
t t
i
t
i
Tr
i
P
i
S

_
,

_
,

_
,

wrm_ch12.qxd 8/31/2005 11:59 AM Page 384

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 385
To model the transport of substances over space, a
water system is divided in small segments or volume
elements. The complete ensemble of all the segments or
elements is called the grid or schematization. Each compu-
tational cell is defined by its volume and its dimensions in
one, two or three directions (x, y, z) depending on
the nature of the schematization (1D, 2D or 3D). Note
that the cell dimensions x, y and z do not have to be
equal. The computational cell can have any rectangular
shape. A computational cell can share surface areas with
other cells, the atmosphere, or the bottom sediment or
coast line.
The following sections will look at the transport
processes in more detail, defining parameters or variables
and their units in terms of mass M, length L and time T.
4.1.1. Advective Transport
The advective transport, (M/T), of a constituent at a site
x
0
is the product of the average water velocity, (L/ T), at
that site, the surface or cross-sectional area, A (L
2
), through
which advection takes place at that site, and the average
concentration, (M/L
3
), of the constituent:
(12.3)
4.1.2. Dispersive Transport
The dispersive transport, (M/T), across a surface area
is assumed to be proportional to the concentration
gradient at site x
0
times the surface area A.
Letting (L
2
/T), be the dispersion or diffusion coeffi-
cient at site x
0
:
(12.4)
Dispersion is done according to Ficks diffusion law. The
minus sign originates from the fact that dispersion causes
net transport from higher to lower concentrations, and so
in the opposite direction of the concentration gradient.
The concentration gradient is the difference of concentra-
tions per unit length, over a very small distance across the
cross section:
T D A
C
x
x
D
x
x x
0
0
0

D
x
0
x x
C
x

0

T
x
D
0
T v A C
x
A
x x
0
0 0

C
x
0
v
x
0
T
x
A
0
(12.5)
Dispersion coefficients should be calibrated or be
obtained from calculations using turbulence models.
4.1.3. Mass Transport by Advection and Dispersion
If the advective and dispersive terms are added and the
terms at a second surface at site x
0
x are included,
a one dimensional equation results:
(12.6)
or equivalently:
(12.7)
where (L
3
/T) is the flow at site x
0
.
If the previous equation is divided by the volume and
the time interval t, then the following equation results in
one dimension:
(12.8)
Taking the asymptotic limit t 0 and x 0, the
advectiondiffusion equation for one dimension results:
(12.9)
The finite volume method for transport is a computa-
tional method of solving the advectiondiffusion
equation. The accuracy of the method will be related to
the size of x, A (A yz) and t.

_
,

C
t x
D
C
x x
vC ( )
i
t t
i
t
x x
x x
x
x
x x x x
C C
t
D
C
x
D
C
x
x
v C v

0
0
0
0
0 0 0

00

x x
C
x
Q
x
0
M
M
t
C C D A
C
x
D A
i
t t
i
t
x x x x x x x
x
x x x

Q Q
x
0 0 0 0 0 0
0
0

00
0

x
x x
C
x

_
,

M M t C C
D
C
x
D
C
x
i
t t
i
t
x x x x x x
x
x
x x
x

v v
0 0 0 0
0
0
0

00

x
_
,

C
x
C C
x
x
x
x x x x

lim
. .

0
0 5 0 5
wrm_ch12.qxd 8/31/2005 11:59 AM Page 385
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
By adding terms for transport in the y and z-direction,
a three-dimensional model is obtained. Taking the
asymptotic limit again will lead to a three-dimensional
advectiondiffusion equation
(12.10)
with dispersion coefficients D
j
defined for each direction.
If source terms S and f
R
are added as shown in the equa-
tion above, the so-called advectiondiffusion reaction
equation emerges. The additional terms represent:
Discharges or wasteloads (S): these source terms are
additional inflows of water or mass. As many source
terms as required may be added to Equation 12.10.
These could include small rivers, discharges of indus-
tries, sewage treatment plants, small wasteload outfalls
and so on.
Reaction terms or processes (f
R
).
Processes can be split into physical and other processes.
Examples of physical processes are:
settling of suspended particulate matter
water movement not affecting substances, like
evaporation
volatilization of the substance itself at the water
surface.
Examples of other processes are:
biochemical conversions like ammonia and oxygen
forming nitrite
growth of algae (primary production)
predation by other animals
chemical reactions.
These processes are described in more detail in the
remaining parts of this section.
The expression D(C/X) vC in Equation 12.9,
multiplied by the area A, is termed the total flux (M/T).
Flux due to dispersion, DA(C/X), is assumed to be
proportional to the concentration gradient over distance.
Constituents are transferred by dispersion from higher
concentration zones to lower ones. The coefficient of
dispersion D (L
2
/T) depends on the amplitude and
frequency of the tide, if applicable, as well as upon the

C
t
D
C
x
v
C
x
D
C
y
v
C
y
D
C
z
v
C
z
S f
x x y y
z z R


2
2
2
2

2
2
(CC, t)
386 Water Resources Systems Planning and Management
turbulence of the water body. It is common practice to
include in this dispersion parameter everything affecting
the distribution of C other than advection. The term vAC
is the advective flux caused by the movement of water
containing the constituent concentration C (M/L
3
) at a
velocity rate v (L/ T) across a cross-sectional area A (L
2
).
The relative importance of dispersion and advection
depends on the degree of detail with which the velocity
field is defined. A good spatial and temporal description
of the velocity field within which the constituent is being
distributed will reduce the importance of the dispersion
term. Less precise descriptions of the velocity field, such
as averaging across irregular cross sections or approxi-
mating transients by steady flows, may lead to a
dominance of the dispersion term.
Many of the reactions affecting the decrease or increase
of constituent concentrations are often represented
by first-order kinetics that assume the reaction rates are
proportional to the constituent concentration. While
higher-order kinetics may be more correct in certain
situations, predictions of constituent concentrations
based on first-order kinetics have often been found to be
acceptable for natural aquatic systems.
4.2. Steady-State Models
A steady state means no change over time. If we consider
a water body, for example a river, this means there are no
changes in the concentrations with time. In this case the
left hand side of Equation 12.9, C/t, equals 0. Assume
the only sink is the natural decay of the constituent
defined as kC where k, (T
1
), is the decay rate coefficient
or constant. Now Equation 12.9 becomes
0 D
2
C/X
2
vC/X kC (12.11)
Equation 12.11 can be integrated, since river reach
parameters A, D, k, v, and Q are assumed constant. For a
constant loading, W
C
(M/T) at site X 0, the concentra-
tion C at any distance X will equal
C(X) (W
C
/Qm) exp[(v/2D)(1 m)X] X 0
(W
C
/Qm) exp[(v/2D)(1 m)X] X 0 (12.12)
where
m (1 (4kD/v
2
))
1/2
(12.13)
wrm_ch12.qxd 8/31/2005 11:59 AM Page 386
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 387
Note from Equation 12.12 that the parameter m is always
equal to or greater than 1 and that the exponent of e is
always negative. Hence, as the distance X increases in
magnitude, either in the positive or negative direction, the
concentration C(X) will decrease if k 0. The maximum
concentration C occurs at X 0 and is W
C
/Qm.
C(0) W
C
/Qm (12.14)
These equations are plotted in Figure 12.4.
In flowing rivers not under the influence of tidal
actions the dispersion is usually small. Assuming the
dispersion coefficient D is 0, the parameter m defined by
Equation 12.13 is 1. Hence, when D 0, the maximum
concentration at X 0 is W
C
/Q.
C(0) W
C
/Q if D 0. (12.15)
Assuming D 0 and v, Q and k 0, Equation 12.12
becomes
(12.16)
The above equation for X 0 can be derived from
Equations 12.12 and 12.13 by noting that the term (1 m)
equals (1 m)(1 m)/(1 m) (1 m
2
)/2 when D is 0.
Thus when Dis 0 the expression v/2D)(1 m)X in Equation
12.12 becomes kX/v. The term X/v is sometimes denoted
as a single variable representing the time of flow: the time
flow Q takes to travel from site X 0 to some other down-
stream site for a distance of X.
As rivers approach the sea, the dispersion coefficient D
increases and the net downstream velocity v decreases.
Because the flow Q equals the cross-sectional area A times
the velocity v, Q Av, and since the parameter m can
be defined as (v
2
4kD)
1/2
/v, then as the velocity
v approaches 0, the term Qm Av(v
2
4kD)
1/2
/v
approaches 2A(kD)
1/2
. The exponent vX(1 m)/2D in
Equation 12.12 approaches X(k/D)
1/2
.
Hence for small velocities, Equation 12.12 becomes
(12.17)
Here, dispersion is much more important than advective
transport and the concentration profile approaches a
symmetric distribution, as shown in Figure 12.4, about
the point of discharge at X 0.
C X
W A kD X k D
W A kD X
C
C
X
( )
( / ( ) ) exp[ ( / ) ]
( / ( ) ) exp[
/ /
/

2
2
1 2 1 2
1 2
0
(( / ) ]
/
k D X
1 2
0
C X
X
W kX v X
C
( )
( / ) exp[ / ]

0 0
0 Q
Water quality management models are often used to
assess the effect of pollutant loadings on ambient waters
and to compare the results with specific water quality
standards. The above steady-state equations can be
used to construct such a model for estimating the
wastewater removal efficiencies required at each waste-
water discharge site that will result in an ambient
stream quality that meets the standards along a stream
or river.
Figure 12.5 shows a schematic of a river into which
wastewater containing constituent C is being discharged
at four sites. Assume that maximum allowable concentra-
tions of the constituent C are specified at each of those
discharge sites. To estimate the necessary reduction
in these discharges, the river must be divided into
approximately homogenous reaches. Each reach can be
characterized by constant values of the cross-sectional
area, A, dispersion coefficient, D, constituent decay rate
constant, k, and velocity, v, associated with some design
flow and temperature conditions. These parameter values
and the length, X, of each reach can differ; hence, the
subscript index i will be used to denote the particular
parameter values for the particular reach. These reaches
are shown in Figure 12.5.
c
o
n
c
e
n
t
r
a
t
i
o
n
C
(
X
)
distance from discharge point
advection only
advection and dispersion
dispersion only
waste load W
c
distance X X > 0 X = 0 X < 0
E
0
2
0
7
3
0
d
Figure 12.4. Constituent concentration distribution along
a river or estuary resulting from a constant discharge of that
constituent at a single point source in that river or estuary.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 387
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
In Figure 12.5 each variable C
i
represents the
constituent concentration at the beginning of reach i.
The flows Q represent the design flow conditions. For
each reach i the product (Q
i
m
i
) is represented by (Qm)
i
.
The downstream (forward) transfer coefficient, TF
i
,
equals the applicable part of Equation 12.12,
TF
i
exp[(v/2D)(1 m)X] (12.18)
as does the upstream (backward) transfer coefficient, TB
i
.
TB
i
exp[(v/2D)(1 m)X] (12.19)
The parameter m is defined by Equation 12.13.
Solving a model such as the one shown in Figure 12.5
does not mean that the least-cost wasteload allocation
plan will be implemented, but least cost solutions can
identify the additional costs of other imposed constraints,
for example, to ensure equity or extra safety. Models like
this can be used to identify the cost-quality tradeoffs
inherent in any water quality management programme.
Other than economic objectives can also be used to obtain
other tradeoffs.
The model in Figure 12.5 incorporates both advection
and dispersion. If upstream dispersion under design
streamflow conditions is not significant in some reaches,
then the upstream (backward) transfer coefficients, TB
i
,
for those reaches i will equal 0.
388 Water Resources Systems Planning and Management
4.3. Design Streamflows for Water Quality
In streams and rivers, the water quality may vary signifi-
cantly, depending on the water flow. If wasteload
discharges are fairly constant, a high flow serves to dilute
the waste concentration, while where there is a low
flow concentrations may become undesirably high. It is
therefore common practice to pick a low-flow condition
for judging whether or not ambient water quality
standards are being met. This can also be seen from
Equations 12.12, 12.14, 12.15, and 12.16. This often is
the basis for the assumption that the smaller (or more
critical) the design flow, the more likely it is that the
stream quality standards will be met. This is not always
the case, however.
Different regions of the world use different design
low-flow conditions. One example of such a design flow,
which is used in parts of North America, is the minimum
seven-day average flow expected once in ten years on
average. Each year the lowest seven-day average flow is
determined, as shown in Figure 12.6. The sum of each of
the 365 sequences of seven average daily flows is divided
by seven, and the minimum value is selected. This is the
minimum annual average seven-day flow.
These minimum seven-day average flows for each
year of record define a probability distribution whose
W
8
W
3
W
1
W
4

i
Cost
i
(R
i
)

C
i
C
i
max
i ; R
i
R
i
max
i
minimize
subject to:
mass balances:
quality standards and maximum removal
efficiencies:
C
1
= W
1
(1-R
1
)/(Qm)
1
+ C
2
TB
1
C
2
= [C
1
TF
1
(Qm)
1
+ C
3
TB
2
(Qm)
2
] / (Qm)
2
C
3
= [W
3
(1-R
3
) + C
2
TF
2
(Qm)
2
+ C
7
TB
3
(Qm)
3
] / (Qm)
3
C
4
= [W
4
(1-R
4
)/(Qm)
4
+ C
5
TB
4
C
5
= [C
4
TF
4
(Qm)
4
+ C
6
TB
5
(Qm)
5
] / (Qm)
5
C
6
= [C
5
TF
5
(Qm)
5
+ C
7
TB
6
(Qm)
6
] / (Qm)
6
C
7
= [C
3
TF
3
(Qm)
3
+ C
6
TF
6
(Qm)
6
+CB
8
TB
7
(Qm)
7
] / (Qm)
7
C
8
= [W
8
(1-R
8
) + C
7
TF
7
(Qm)
7
] / (Qm)
8
E
0
2
0
7
3
0
e
8
6
7
5
4
3
2
1
Figure 12.5. Optimization model
for finding constituent removal
efficiencies, Ri, at each
discharge site i that result in
meeting stream quality
standards, Ci
max
, at least total
cost.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 388
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 389
cumulative probabilities can be plotted. As illustrated
in Figure 12.7, the particular flow on the cumulative
distribution that has a 90% chance of being exceeded
is the design flow. It is the minimum annual average
seven-day flow expected once in ten years. This flow is
commonly called the 7Q10 flow. Analyses have shown
that this daily design flow is exceeded about 99% of the
time in regions where it is used (NRC, 2001). This means
that there is on average only a 1% chance that any daily
flow will be less than this 7Q10 flow.
Consider now any one of the river reaches shown in
Figure 12.5. Assume an initial amount of constituent
mass, M, exists at the beginning of the reach. As the reach
volume, Qt, increases due to the inflow of less polluted
water, the initial concentration, M/Qt, will decrease.
However, the flow velocity will increase, and thus the
time, t, it takes to transport the constituent mass to the
end of that reach will decrease. This means less time for
the decay of the constituent. Thus, wasteload allocations
that meet ambient water quality standards during low-
flow conditions may not meet them under higher-flow
conditions, which are observed much more frequently.
Figure 12.8 illustrates how this might happen. This does
not suggest low flows should not be considered when
allocating wasteloads, but rather that a simulation of
water quality concentrations over varying flow conditions
may show that higher-flow conditions at some sites
are even more critical and more frequent than are the
low-flow conditions.
Figure 12.8 shows that for a fixed mass of pollutant at
X 0, under low-flow conditions the more restrictive
(lower) maximum pollutant concentration standard in the
downstream portion of the river is met, but that same
standard is violated under more frequent higher-flow
conditions.
4.4. Temperature
Temperature affects almost all water quality processes
taking place in water bodies. For this reason, it may
be important to model temperature when it may vary
substantially over the period of interest, or when the
discharge of heat into water bodies is to be managed.
days
d
a
i
l
y

f
l
o
w
minimum annual average
7-day flow
minimum
7-day flow period
E
0
2
0
7
3
0
f
Figure 12.6. Portion of annual flow time series showing
low flows and the calculation of average seven and
fourteen-day flows.
minimum 7-days average annual flows q
Pr
(q < actual annual minimum 7-day average flow)
E
0
2
0
7
3
0
g
0.0
1.0
0.1
7Q 0 flow 1
Figure 12.7. Determining the minimum seven-day annual
average flow expected once in ten years, designated
7Q10, from the cumulative probability distribution of annual
minimum seven-day average flows.
distance downstream X
E
0
2
0
7
3
0
h
c
o
n
s
t
i
t
u
e
n
t

c
o
n
c
e
n
t
r
a
t
i
o
n maximum
allowable concentration
along the river (quality standard)
flow:
higher
lower
Figure 12.8. Increasing the streamflows decreases initial
concentrations but may increase downstream concentrations.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 389
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Temperature models are based on a heat balance in the
water body. A heat balance takes into account the sources
and sinks of heat. The main sources of heat in a water
body are short-wave solar radiation, long-wave atmos-
pheric radiation, conduction of heat from the atmosphere
to the water and direct heat inputs. The main sinks of heat
are long-wave radiation emitted by the water, evaporation
and conduction from the water to the atmosphere.
Unfortunately, a model with all the sources and sinks of
heat requires measurements of a number of variables and
coefficients that are not always readily available.
One temperature predictor is the simplified model that
assumes an equilibrium temperature T
e
(C) will be
reached under steady-state meteorological conditions.
The temperature mass balance in a volume segment
depends on the water density (g/cm
3
), the heat capacity
of water, c
p
(cal/g/C), and the water depth h (cm). The net
heat input, K
H
(T
e
T) (cal/cm
2
/day), is assumed to be
proportional to the difference of the actual temperature,
T, and the equilibrium temperature, T
e
(C).
dT/dt K
H
(T
e
T)/c
p
h (12.20)
The overall heat exchange coefficient, K
H
(cal/cm
2
/day/C),
is determined in units of Watts/m
2
/C (1 cal/cm
2
/day/C
0.4840 Watts/m
2
/C) from empirical relationships that
include wind velocity, dew point temperature and actual
temperature T (C) (Thomann and Mueller, 1987).
The equilibrium temperature, T
e
, is obtained from
another empirical relationship involving the overall heat
exchange coefficient, K
H
, the dew point temperature, T
d
,
and the short-wave solar radiation, H
s
(cal/cm
2
/day),
T
e
T
d
(H
s
/K
H
) (12.21)
This model simplifies the mathematical relationships of a
complete heat balance and requires less data.
4.5. Sources and Sinks
Sources and sinks of pollutants include wasteloads, and
the physical and biochemical processes that alter those
wasteloads. External inputs of each constituent would have
the form W/Qt or W/(A
X
X) where W(M/T) is the loading
rate of the constituent and Qt or A
X
X (L
3
) represents
the volume of water into which the mass of waste W is
discharged. Constituent growth and decay processes are
discussed in the remaining parts of this Section 4.
390 Water Resources Systems Planning and Management
4.6. First-Order Constituents
The first-order models are commonly used to predict
water quality constituent decay or growth. They can
represent constituent reactions such as decay or growth in
situations where the time rate of change (dC/dt) in the
concentration C of the constituent, say organic matter
that creates a biochemical oxygen demand (BOD), is
proportional to the concentration of either the same or
another constituent concentration. The temperature-
dependent proportionality constant k
c
(1/day) is called
a rate coefficient or constant. In general, if the rate of
change in some constituent concentration C
j
is propor-
tional to the concentration C
i
, of constituent i, then
dC
j
/dt a
ij
k
i

i
(T20)
C
i
(12.22)
where
i
is the temperature correction coefficient for k
i
at
20 C and T is the temperature in C. The parameter a
ij
is
the grams of C
j
produced (a
ij
0) or consumed (a
ij
0)
per gram C
i
. For the prediction of BOD concentration
over time, C
i
C
j
BOD and a
ij
a
BOD
1 in
Equation 12.22. Conservative substances, such as salt,
will have a decay rate constant k of 0. The concentration
of conservative substances depends only on the amount
of water, that is, dilution.
The typical values for the rate coefficients k
c
and
temperature coefficients
i
of some constituents C are
in Table 12.1. For bacteria, the first-order decay rate (k
B
)
can also be expressed in terms of the time to reach
90% mortality (t
90
, days). The relationship between these
coefficients is given by k
B
2.3/t
90
.
4.7. Dissolved Oxygen
Dissolved oxygen (DO) concentration is a common
indicator of the health of the aquatic ecosystem. DO was
originally modelled in the Ohio River (US) by Streeter
and Phelps (1925). Since then a number of modifica-
tions and extensions of the model have been made
relating to the number of sinks and sources of DO being
considered, and how processes involving the nitrogen
cycle and phytoplankton are being modelled, as
illustrated in Figure 12.9.
The sources of DO in a water body include re-aeration
from the atmosphere, photosynthetic oxygen production
from aquatic plants, denitrification and DO inputs. The
wrm_ch12.qxd 8/31/2005 11:59 AM Page 390
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 391
total coliform bacteria (freshwater)
total coliform bacteria (sediments)
total coliform bacteria (seawater)
fecal coliform bacteria (seawater)
BOD (no treatment)
BOD (activated sludge treatment)
carbofuran
DDT
PCB
pentachlorophenol
constituent
constituent
1.0-5.5
0.14-0.21
0.7-3.0
37-110
0.3-0.4
0.05-0.1
0.03
0.0-0.10
0.0-0.007
0.0-33.6
1/day
1/day
1/day
1/day
1/day
1/day
1/day
1/day
1/day
1/day
rate constant k

units
units
coliform bacteria (freshwater)
BOD
coliform bacteria (saltwater)
1.07
1.10
1.04

---
- a
- a
- a
- a
- a
- a
- b
- b
- b
- b
- b
- b
- a
a - Thomann and Mueller (1987) b - Schnoor (1996)
E
0
2
0
9
1
3
a
Table 12.1. Typical values of the first-order
decay rate, k, and the temperature correction
factor, , for some constituents.
oxygen
CBOD NBOD
organic matter input
sediment organic matter
dissolved oxygen
atmosphere
waterbody
bottom sediment
E
0
2
0
7
3
0
j
Figure 12.9. The dissolved
oxygen interactions in a water
body, showing the decay
(satisfaction) of carbonaceous,
nitrogenous and sediment
oxygen demands and water body
re-aeration or deaeration (if
supersaturation occurs at the
airwater interface).
wrm_ch12.qxd 8/31/2005 11:59 AM Page 391
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
sinks include oxidation of carbonaceous and nitrogenous
material, sediment oxygen demand and respiration by
aquatic plants.
O
2
/t loads transport re-aeration
net primary production denitrification
mineralization nitrification SOD
The rate of re-aeration is assumed to be proportional to
the difference between the saturation concentration, DO
sat
(mg/l), and the concentration of dissolved oxygen, DO
(mg/l). The proportionality coefficient is the re-aeration
rate k
r
(1/day), defined at temperature T 20 C , which
can be corrected for any temperature T with the
coefficient
r
(T20)
. The value of this temperature correc-
tion coefficient, , depends on the mixing condition of
the water body. Values generally range from 1.005 to
1.030. In practice, a value of 1.024 is often used (Thomann
and Mueller, 1987). The re-aeration rate constant is a
sensitive parameter. There have been numerous equations
developed to define this rate constant. Table 12.2 lists
some of them.
The saturation concentration, DO
sat
, of oxygen in
water is a function of the water temperature and salinity
(chloride concentration, Cl (g/m
3
)), and can be approxi-
mated by
DO
sat
{14.652 0.41022T (0.089392T)
2
(0.042685T)
3
}{1 (Cl/100,000)} (12.23a)
Elmore and Hayes (1960) derived an analytical expres-
sion for the DO saturation concentration, DO
sat
(mg/l), as
a function of temperature (T, C):
392 Water Resources Systems Planning and Management
DO
sat
14.652 0.41022T 0.007991T
2
0.000077774T
3
(12.23b)
Fitting a second-order polynomial curve to the data
presented in Chapra (1997) results in:
DO
sat
14.407 0.3369T 0.0035T
2
(12.23c)
as is shown in Figure 12.10.
Because photosynthesis occurs during daylight hours,
photosynthetic oxygen production follows a cyclic,
diurnal, pattern in water. During the day, oxygen
concentrations in water are high and can even become
supersaturated, i.e. concentrations exceeding the satura-
tion concentration. At night, the concentrations drop due
to respiration and other oxygen-consuming processes.
16
14
12
10
8
6
4
2
0
s
a
t
u
r
a
t
i
o
n
D
O

(
m
g
/
l
)
0 5 10 15 20 25 30 35 40
data
adjusted curve
E
0
2
0
7
3
0
k
temperature (
o
C )
Figure 12.10. Fitted curve to the saturation dissolved oxygen
concentration (mg/l) as a function of temperature (C).
E
0
2
0
9
1
3
b
water and wind velocity (m/s) water depth (m)
k
r
=
=
=
=
=
=
=
mass transport coefficient for reaeration (m/day) / (water depth)
5.026 (water velocity)
0.969
(water depth)
1.673
/ (Churchill, 1962)
3.95
(scale factor)
5.344
5.13
(water velocity)
(water velocity)
(water velocity)
(water velocity)
0.5
0.5
0.670
(water depth)
(water depth)
(water depth)
(water depth)
1.5
1.5
1.85
1.333
/
/
/
/
(O'Connor and Dobbiens, 1958)
(Owens, Edwards, Gibb, 1964)
(Langbien, Durum, 1967)
3.95
0.065 (wind velocity)
2
{
(van Pagee 1978, Delvigne 1980)
3.86 [(water velocity) (water depth)]
0.5
/ + } (water depth) /
u
n
i
t
s
Table 12.2. Some equations for
defining the re-aeration rate
constant, k
r
(day
1
).
wrm_ch12.qxd 8/31/2005 11:59 AM Page 392
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 393
One can distinguish between the biochemical oxygen
demand from carbonaceous organic matter (CBOD, mg/l)
in the water, and that from nitrogenous organic matter
(NBOD, mg/l) in the water. There is also the oxygen
demand from carbonaceous and nitrogenous organic
matter in the sediments (SOD, mg/l/day). These oxygen
demands are typically modelled as first-order decay
reactions with decay rate constants k
CBOD
(1/day) for
CBOD and k
NBOD
(1/day) for NBOD. These rate constants
vary with temperature, hence they are typically defined
for 20 C. The decay rates are corrected for temperatures
other than 20 C using temperature coefficients
CBOD
and
NBOD
, respectively.
The sediment oxygen demand SOD (mg/l/day) is usually
expressed as a zero-order reaction, that is, a constant
demand. One important feature in modelling NBOD is
ensuring the appropriate time lag between when it is
discharged into a water body and when the oxygen demand
is observed. This lag is in part a function of the level of
treatment in the wastewater treatment plant.
The dissolved oxygen (DO) model with CBOD, NBOD
and SOD is
dDO/dt k
CBOD

CBOD
(T20)
CBOD
k
NBOD

NBOD
(T20)
NBOD
k
r

r
(T20)
(DO
sat
DO) SOD (12.24)
dCBOD/dt k
CBOD

CBOD
(T20)
CBOD (12.25)
dNBOD/dt k
NBOD

NBOD
(T20)
NBOD (12.26)
The mean and range values for coefficients included in
these dissolved oxygen models are shown in Table 12.3.
4.8. Nutrients and Eutrophication
Eutrophication is the progressive process of nutrient
enrichment of water systems. The increase in nutrients
leads to an increase in the productivity of the water
system, which may result in an excessive increase in
the biomass of algae or other primary producers, such as
macrophytes or duck weed. When it is visible on the
surface of the water it is called an algae bloom. Excessive
algal biomass could affect the water quality, especially
if it causes anaerobic conditions and thus impairs the
drinking, recreational and ecological uses.
The eutrophication component of the model relates
the concentration of nutrients and the algal biomass.
Taking the example shown in Figure 12.11, consider
the growth of algae A (mg/l), depending on phosphate
phosphorus, P (mg/l), and nitrite/nitrate nitrogen, N
n
mg/l),
as the limiting nutrients. There could be other limiting
nutrients or other conditions as well, but here consider only
these two. If either of these two nutrients is absent, the
algae cannot grow, regardless of the abundance of the other
nutrient. The uptake of the more abundant nutrient will
not occur.
To account for this, algal growth is commonly
modelled as a MichaelisMenten multiplicative effect; in
other words, the nutrients have a synergistic effect. Model
parameters include a maximum algal growth rate
(1/day) times the fraction of a day, f
d
, that rate applies
(Figure 12.12), the half saturation constants K
P
and K
N
(mg/l) (shown as K
C
in Figure 12.13) for phosphate and
nitrate, respectively, and a combined algal respiration and
specific death rate constant e (1/day) that creates an
oxygen demand. The uptake of phosphate, ammonia and
nitrite/nitrate by algae is assumed to occur in proportion
to their contents in the algae biomass. Define these
proportions as a
P
, a
A
and a
N
, respectively.
In addition to the above parameters, one needs to
know the amounts of oxygen consumed in the oxidation
of organic phosphorus, P
o
, and the amounts of oxygen
produced by photosynthesis and consumed by respira-
tion. In the model below, some average values have
been assumed. Also assumed are constant temperature
correction factors for all processes pertaining to any
individual constituent. This reduces the number of
parameters needed, but is not necessarily realistic. Clearly
other processes as well as other parameters could be
added, but the purpose here is to illustrate how these
models are developed. Users of water quality simulation
programs will appreciate the many different assumptions
that can be made and the large amount of parameters
associated with most of them.
The source and sink terms of the relatively simple
eutrophication model shown in Figure 12.11 can be
written as follows:
For algae biomass:
dA/dt f
d

A
(T20)
[P/(P K
P
)][N
n
/(N
n
K
N
)]
A e
A
(T20)
A (12.27)
For organic phosphorus:
dP
o
/dt k
op

op
(T20)
P
o
(12.28)
wrm_ch12.qxd 8/31/2005 11:59 AM Page 393
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
For phosphate phosphorus:
dP/dt f
d

A
(T20)
[P/(P K
P
)]
[N
n
/(N
n
K
N
)]a
P
A (12.29)
For organic nitrogen:
dN
o
/dt k
on

on
(T20)
N
o
(12.30)
For ammonia-nitrogen:
dN
a
/dt f
d

A
(T20)
[P/(P K
P
)]
[N
n
/(N
n
K
N
)]a
A
A
k
on

on
(T20)
N
o
k
a

a
(T20)
N
a
(12.31)
394 Water Resources Systems Planning and Management
For nitrate-nitrogen:
dN
n
/dt f
d

A
(T20)
[P/(P K
p
)]
[N
n
/(N
n
K
N
)]a
N
A
k
a

a
(T20)
N
a
k
n

n
(T20)
N
n
(12.32)
For dissolved oxygen:
dDO/dt k
CBOD

CBOD
(T20)
CBOD
4.57k
a

a
(T20)
N
a
2k
op

op
(T20)
P
o
(1.5f
d
2e)
A
(T20)
A
k
r

r
(T20)
(DO
sat
DO) SOD (12.33)
E
0
2
0
9
1
3
c
k
k
k
k
k
k
k
parameter
0.1-0.4
0.4-1.5
1.5-4.0
4.0-10.0
0.35
0.20
0.075
1.04
1.047
1.04
1.024
1/day
1/day
1/day
1/day
1/day
1/day
1/day

---
value units
- a
- a
- a
- a
- b
- b
- b
- a
- a
- c
- c
r
r
r
r
CBOD
CBOD
CBOD
CBOD

slow, deep rivers

typical conditions
swift, deep rivers
swift, shallow rivers
,
,
,
,
,
,
,
untreated discharges
primary treatment
activated sludge
(0.20-0.50)
(0.10-0.30)
(0.05-0.10)
(1.02-1.09)
(1.005-1.030)
r
sediment oxygen demand
*
4
1.5
0.5
0.07
0.1-10.0
5-10
g O / m / day
value units
- c d
- c d
- c d
- c d
- a
- a
municipal sludge (outfall vicinity)
municipal sewage sludge
sandy bottom
mineral soils
natural to low pollution
moderate to heavy pollution
(2-10)
(1-2)
(0.2-1.0)
(0.05-0.1)
2
2
g O / m / day
2
2
g O / m / day
2
2
g O / m / day
2
2
g O / m / day
2
2
g O / m / day
2
2
a
b
- Schnoor (1996)
- Chapra (1997)
c
d
- Thomann and Mueller (1987)
- Bowie et al. (1985)
value has to be divided by the water height (m) *
Table 12.3. Typical values of
parameters used in the
dissolved oxygen models.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 394
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 395
Representative values of the coefficients for this model are
shown in Table 12.4.
Because the growth of phytoplankton cannot occur
without nutrients, the eutrophication modelling must be
coupled with that of nutrients. Nutrient modelling must
include all the different biochemical forms of the nutri-
ents, primarily nitrogen and phosphorus, as well as all the
interactions between the different forms. The sum of all
these interactions is referred to a nutrient cycling.
The nitrogen cycle includes ammonium (NH
4
-N) and
nitrate/nitrite (represented as NO
3
-N) as the main forms
of dissolved nitrogen in water. Furthermore, nitrogen is
present in algae, as well as detritus, resulting from algae
mortality, and suspended (non-detritus) organic nitrogen.
Nitrogen can also be present in different forms in
the bottom sediment. Two important reactions in the
nitrogen nutrient cycle are nitrification and denitrifica-
tion, which affect the flux of ammonium and nitrate in the
water column.
0 4 6 8 10 12 14 16 18 20 22 24 2
time of day (hrs)
g
r
o
w
t
h

r
a
t
e

c
o
n
s
t
a
n
t

maximum growth rate constant (1/day)

potential
production zone
effictive
daylight hours (EDH)
actual
production
zone
E
0
2
0
7
3
0
o
Figure 12.12. Calculation of the fraction, f
d
, of the maximum
growth rate constant, , to use in the algal growth equations.
The fraction f
d
is the ratio of actual production zone/potential
production zone.
limiting nutrient concentration C
g
r
o
w
t
h

r
a
t
e

c
o
n
s
t
a
n
t

maximum growth rate constant

E
0
2
0
7
3
0
p
0.5
half saturation concentration K
c
Figure 12.13. Defining the half saturation constant for a
MichaelisMenten model of algae. The actual growth rate
constant C/(C K
C
).
oxygen
CBOD
organic matter input
atmosphere
bottom sediment
E
0
2
0
7
3
0
n
CBOD
ammonia
dissolved oxygen
phytoplankton
detritus nitrate-N
CBOD phosphate-P
sediment organic matter
waterbody
Figure 12.11. The dissolved oxygen,
nitrogen and phosphorus cycles, and
phytoplankton interactions in a water
body, showing the decay (satisfaction) of
carbonaceous and sediment oxygen
demands, re-aeration or deaeration of
oxygen at the airwater interface,
ammonification of organic nitrogen in the
detritus, nitrification (oxidation) of
ammonium to nitrate-nitrogen and
oxidation of organic phosphorus in the
sediment or bottom layer to phosphate
phosphorus, phytoplankton production
from nitrate and phosphate consumption,
and phytoplankton respiration and death
contributing to the organic nitrogen and
phosphorus.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 395
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Nitrification is the conversion of ammonium to nitrite
and finally nitrate, requiring the presence of oxygen:
NH
4

2O
2
NO
3

2H
2
O2H

(12.34)
Denitrification is the process occurring during the break-
down (oxidation) of organic matter by which nitrate is
transformed to nitrogen gas, which is then usually lost
from the water system. Denitrification occurs in anaerobic
conditions:
NO
3

N
2
(12.35)
The phosphorus cycle is simpler than the nitrogen
cycle because there are fewer forms in which phosphorus
can be present. There is only one form of dissolved phos-
phorus, orthophosphorus (also called orthophosphate,
PO
4
-P). Like nitrogen, phosphorus also exists in algae,
in detritus and other organic material, as well as in the
bottom sediment. Unlike nitrogen, there can also be
inorganic phosphorus in the particulate phase.
Further details of the nutrient cycles are given in
Section 5.
4.9. Toxic Chemicals
Toxic chemicals, also referred to as micro-pollutants,
are substances that at low concentrations can impair the
reproduction and growth of organisms, including fish
396 Water Resources Systems Planning and Management
and human beings. These substances include heavy
metals, many synthetic organic compounds (organic
micro-pollutants) and radioactive substances.
4.9.1. Adsorbed and Dissolved Pollutants
An important characteristic of many of these substances
is their affinity with the surface areas of suspended or
bottom sediments. Many chemicals preferentially sorb
onto surfaces of particulate matter rather than remaining
dissolved in water. To model the transport and fate of
these substances, the adsorptiondesorption process,
estimations of the suspended sediment concentration,
resuspension from the bottom and settling are required.
Figure 12.14 depicts the adsorptiondesorption and
first-order decay processes for toxic chemicals and their
interaction in water and sediment. This applies to the
water and sediment phases in both the water body and
the bottom sediments.
The adsorptiondesorption model assumes (conve-
niently but not always precisely) that an equilibrium
exists between the dissolved (in water) and adsorbed
(on sediments) concentrations of a toxic constituent
such as a heavy metal or organic contaminant. This
equilibrium follows a linear relationship. The slope of that
linear relation is the partition coefficient K
p
(litres/kg).
This is shown in Figure 12.15.
E
0
2
0
9
1
3
e
parameter
g N/l
g NO /l
g N/l
g P/l
g P/l
g P/l

value units
- a
- c
- b
- a
- c
- b
- c
- c
- b
- c
- b
half saturation
half saturation
stoichoimetric ratio
stoichoimetric ratio
maximum algae growth rate
death algae rate
10-20
50-200
10
1-5
20-70
10
0.012-0.015
0.08-0.09
1.5
0.2-8
0.1
(1-20)
(1.0-2.0)
(0.05-0.025)
3
3
k
k
a
a
e

P
N
P
N
1/day
1/day
1/day
mg P/mg A
mg NO /mg A
a - Thomann and Mueller (1987)
b - Schnoor (1996)
c - Bowie et al. (1985)
Table 12.4. Typical values of
coefficients in the eutrophication
model.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 396
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 397
Each partition coefficient K
p
(mg/kgDW/mg/l water or
l/kg) is defined as the ratio of the particulate concentra-
tion C
p
of a micro-pollutant (mg/kgDW or mg/kgC)
divided by the dissolved concentration C
d
of a micro-
pollutant (mg/l water).
K
p
C
p
/ C
d
(12.36)
Representative values of partition coefficients K
p
are given
in Table 12.5.
The presence of a micro-pollutant in a water system is
described by the total concentration (sum of dissolved
and particulate concentrations), the total particulate
concentration and the total dissolved concentration for
each water and sediment compartment. The particulate
and dissolved concentrations are derived from the total
concentration and the respective fractions.
Because the fate of most micro-pollutants is largely
determined by adsorption to particulate matter, sus-
pended inorganic and organic matter (including phyto-
plankton) has to be included in the model in most cases.
It may be necessary to include dissolved organic matter
as well.
The adsorbed fractions in the water column are subject
to settling. The fractions in the sediment are subject to
resuspension. The adsorbed fractions in the sediment
can also be removed from the modelled part of the water
system by burial.
The rates of settling and resuspension of micro-
pollutants are proportional to the rates for particulate
matter. An additional process called bioturbation leads to
the redistribution of the micro-pollutant among sediment
layers. Bioturbation is caused by the physical activity of
organisms, and affects both the particulate and dissolved
phases but at different rates. Bioturbation is taken into
account by means of dispersion coefficients.
For modelling purposes, it is important to know how
much of a toxic chemical is present as a dissolved
constituent, as opposed to adsorbed. Assuming partition
coefficients apply to a particular toxic constituent, the
concentration, C
w
, of that constituent in the water body is
divided into a dissolved fraction (f
dw
) and an adsorbed
fraction (f
aw
);
C
w
(f
dw
f
aw
)C
w
(12.37)
In turn, the adsorbed fraction is composed the fractions
of a micro-pollutant adsorbed to inorganic particulate
bottom sediment
adsorbed
adsorption
decay
desorption
dissolved
E
0
2
0
7
3
0
v
waterbody
adsorbed
adsorption
decay
desorption
dissolved
settling / resuspension
Figure 12.14. Schematic of the adsorption/desorption and
decay processes of various toxic chemicals in water bodies
and bottom sediments.
c
o
n
c
e
n
t
r
a
t
i
o
n

i
n
s
o
l
i
d

p
h
a
s
e

m
g
/
k
g
(
d
r
y

w
e
i
g
h
t
)
concentration in water phase mg/l
K
p
(mg/kg / mg/l = l/kg)
E
0
2
0
7
3
0
w
Figure 12.15. Defining the partition coefficient K
p
(litres/kg)
as the slope of the fixed ratio between concentrations of a
constituent in the water and sediment phases of either a
water body or bottom sediments. Different constituents have
different partition coefficients, when they apply.
E
0
2
0
9
1
3
f
K
p
parameter
l/kg
l/kg
l/kg
l/kg
l/kg
l/kg
l/kg
l/kg
value units
arsenic
heavy metals (Cd,Cu,Cr,Zn)
benzo(a)pyrene
lead
PCB
plutonium-239
methoxychlor
napthalene
10
10 -10
10 -10
10 -10
10 -10
10 -10
10
10
4
4 6
4 5
5 6
5 6
4 5
4
3
Table 12.5. Typical values of partition coefficients in toxic
chemical model from Thomann and Mueller (1987).
wrm_ch12.qxd 8/31/2005 11:59 AM Page 397
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
matter, fim, dead particulate organic matter, fpoc, and
algae, falg. The total micro-pollutant concentration,
C
w
(mg/m
3
) is the sum of all these fractions:
C
w
(f
dw
fim fpoc falg)C
w
(12.38)
Considering the simple division into dissolved and
adsorbed fractions (f
dw
and f
aw
), these fractions depend
on the partition coefficient, K
p
, and on the suspended
sediment concentration, SS (mg/l). The proportions of
the total constituent concentration in the water body, C
w
,
dissolved in the water, DC
w
(mg/l), and adsorbed to the
suspended sediments, AC
w
(mg/l) are defined as
DC
w
f
dw
C
w
(12.39)
AC
w
f
aw
C
w
(12.40)
where the fractions
f
dw
1/(1 K
p
SS) (12.41)
f
aw
1 f
dw
K
p
SS/(1 K
p
SS) (12.42)
Similarly, in the bottom sediments, the dissolved concen-
tration DC
s
(mg/l) and adsorbed concentration AC
s
(mg/l)
are fractions, f
ds
and f
as
, of the total concentration C
s
(mg/l);
DC
s
f
ds
C
s
(12.43)
AC
s
f
as
C
s
(12.44)
These fractions are dependent on the sediment porosity,
, and density,
s
(kg/l):
f
ds
1/[
s
(1 )K
p
] (12.45)
f
as
1 f
ds
([
s
(1 )K
p
] 1)/
[
s
(1 )K
p
] (12.46)
First order decay occurs in the water and sediment phases
only in the dissolved fraction with decay rate constants k
w
and k
s
(1/day), respectively. Thus:
dC
w
/dt k
w

w
(T20)
f
dw
C
w
f
aw
C
wS
f
as
C
s
r (12.47)
dC
s
/dt k
s

s
(T20)
f
ds
C
s
f
aw
C
wS
f
as
C
s
r (12.48)
In the above two equations the parameter s represents
the mass of settling sediments (mg/day), r the mass of
resuspension sediments (mg/day), and the temperature
correction coefficient of the constituent at temperature
398 Water Resources Systems Planning and Management
T 20 C. If data are not available to distinguish
between the values of the decay rate constants k in water
and on sediments, they may be assumed to be the same.
Similarly, for the values of the temperature correction
coefficients . The settling and resuspension of suspended
solids can be determined each day from a sediment
model.
4.9.2. Heavy Metals
The behaviour of heavy metals in the environment
depends on their inherent chemical properties. Heavy
metals can be divided into different categories accord-
ing to their dissolved form and redox status. Some
metals, including copper, cadmium, lead, mercury,
nickel, tin and zinc, form free or complexed cations
when dissolved in water, e.g. Cu
2
or CuCl

. The
soluble complexes are formed with negatively charged
ions such as chlorine, oxygen or dissolved organic
compounds. These heavy metals also tend to form
poorly soluble sulphides under chemically reducing
conditions. These sulphides generally settle in bottom
sediments and are essentially unavailable ecologically.
Other metals such as arsenic and vanadium are present
as anions in dissolved form. The differences between
groups of metals have important consequences for the
partitioning of the metals among several dissolved and
particulate phases.
Metals are non-decaying substances. The fate of heavy
metals in a water system is determined primarily by
partitioning to water and particulate matter (including
phytoplankton), and by transport. The partitioning
divides the total amount of a pollutant into a dissolved
fraction and several adsorbed fractions (as described in
Equations 12.3912.42). The fractions of a metal that are
adsorbed onto particulate matter are influenced by all the
processes that affect particulate matter, such as settling
and resuspension.
Partitioning is described in general by sorption to
particulates, precipitation in minerals, and complexation
in solution. Complexation with inorganic and organic
ligands can be considered explicitly in connection with
the other processes. Sorption can be modelled as an
equilibrium process (equilibrium partitioning) or as the
resultant of slow adsorption and desorption reactions
(kinetic formulations). In the latter case, partitioning
wrm_ch12.qxd 8/31/2005 11:59 AM Page 398
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The fractions are functions of the partition coefficients
K
p
(for algae (m
3
/gC), for inorganic matter (m
3
/gDW)
and for dead particulate organic matter (m
3
/gC)), the
individual concentrations C (for algae biomass (gC/m
3
), for
inorganic matter (gDW/m
3
) and for dead particulate organic
matter (gC/m
3
)), and the porosity (m
3
water/m
3
bulk). In
surface water the value for porosity is 1.
fd /[ Kpalg Calg Kpim Cim
Kppoc Cpoc] (12.50)
fim (1 fd) Kpim Cim/
[Kpalg Calg Kpim Cim Kppoc Cpoc] (12.51)
fpoc (1 fd) Kppoc Cpoc/
[Kpalg Calg Kpim Cim Kppoc Cpoc] (12.52)
falg (1 fd fim fpoc) (12.53)
In terms of bulk measures, each partition coefficient K
p
(see Equation 12.36) also equals the porosity times the
bulk particulate concentration Cp (mg/m
3
bulk) divided
by the product of the dissolved (mg/l bulk) and particu-
late (mg/m
3
bulk) bulk concentrations, Cd Cs, all times
10
6
mg/kg.
K
p
10
6
Cp/(CdCs) (12.54)
Partitioning can be simulated based on the above equilib-
rium approach or according to slow sorption kinetics. For
the latter, the rate, dCp/dt, of adsorption or desorption
(mg/m
3
/day) depends on a first order kinetic constant
ksorp (day
1
) for adsorption and desorption times
the difference between equilibrium particulate concentra-
tion Cpe of a micro-pollutant (mg/m
3
bulk) and the
actual particulate concentration Cp (mg/m
3
bulk) of a
micro-pollutant.
dCp/dt ksorp(Cpe Cp) (12.55)
The kinetic constant for sorption is not temperature
dependent. All other kinetic constants for micro-pollutants
are temperature dependent.
Mass-balance equations are similar for all micro-
pollutants except for the loss processes.
Metals are conservative substances that can be
transformed into various species either through complexa-
tion, adsorption or precipitation. Organic micro-pollutants
are lost by volatilization, biodegradation, photolysis,
hydrolysis and overall degradation. Most of these processes
are usually modelled as first-order processes, with associ-
ated rate constants.
Water Quality Modelling and Prediction 399
is assumed to proceed at a finite rate proportional
to the difference between the actual state and the
equilibrium state.
To describe the fate of certain heavy metals in reducing
environments, such as sediment layers, the formation of
metal sulphides or hydroxides can be modelled. The
soluble metal concentration is determined on the basis of
the relevant solubility product. The excess metal is stored
in a precipitated metal fraction.
Sorption and precipitation affect the dissolved metal
concentration in different ways. Both the adsorbed
and dissolved fractions increase at increasing total
concentration as long as no solubility product is
exceeded. When it is, precipitation occurs.
4.9.3. Organic Micro-pollutants
Organic micro-pollutants are generally biocides (such as
pesticides or herbicides), solvents or combustion
products, and include substances such as hexachloro-
hexane, hexachlorobenzene, PCBs or polychloro-
biphenyls, benzo-a-pyrene and fluoranthene (PAHs or
polycyclic aromatic hydrocarbons), diuron and linuron,
atrazine and simazine, mevinfos and dichlorvos, and
dinoseb.
The short-term fate of organic micro-pollutants in a
water system is determined primarily by partitioning
to water and organic particulate matter (including
phytoplankton), and by transport. Additional processes
such as volatilization and degradation influence organic
micro-pollutant concentrations (this is in contrast to
heavy metals, which do not decay). Many toxic organic
compounds have decay (or daughter) products that
are equally, if not more, toxic than the original
compound. The rates of these processes are concentration
and temperature dependent.
Organic micro-pollutants are generally very poorly
soluble in water and prefer to adsorb to particulate
matter in the water, especially particulate organic matter
and algae. Therefore, the fractions of a micro-pollutant
adsorbed to inorganic matter, fim, dead particulate
organic matter, fpoc, the dissolved fraction of a micro-
pollutant, fd, and algae, falg, add up to the total
micro-pollutant concentration, C (mg/m
3
);
C (fd fim fpoc falg)C (12.49)
wrm_ch12.qxd 8/31/2005 11:59 AM Page 399
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Volatilization is formulated according to the double
film theory. The volatization rate dCd/dt (mg/m
3
/day)
of dissolved micro-pollutant concentrations, Cd (mg/m
3
water), in water depends on an overall transfer coefficient,
kvol (m/day), for volatilization and the depth of the water
column, H (m);
dCd/dt kvol Cd/H (12.56)
The numerator (kvol Cd) is the volatilization mass flux
(mg/m
2
/day).
This equation is only valid when the atmospheric
concentration is negligibly small, which is the normal
situation.
All other loss rates such as biodegradation, photolysis,
hydrolysis or overall degradation (mg/m
3
/day) are usually
modelled as
dC/dt kC (12.57)
where C is the total concentration of a micro-pollutant
(mg/m
3
), and k is a (pseudo) temperature-dependent
first-order kinetic rate constant for biodegradation,
photolysis, hydrolysis or overall degradation (day
1
).
4.9.4. Radioactive Substances
The fate of most radionuclides, such as isotopes of iodine
(
131
I) and cesium (
137
Cs), in water is determined prima-
rily by partitioning to water and particulate matter
(including phytoplankton), by transport and by decay.
Cesium (Ce

) adsorbs to particulate inorganic matter, to

dead particulate organic material, and to phytoplankton,
both reversibly and irreversibly. The irreversible fraction
increases over time as the reversible fraction gradually
transforms into the irreversible fraction. Radioactive
decay proceeds equally for all fractions. Precipitation of
cesium does not occur at low concentrations in natural
water systems.
Iodine is only present in soluble form as an anion
(IO
3

) and does not adsorb to particulate matter.

Consequently, with respect to transport, iodine is only
subject to advection and dispersion.
Concentrations of radionuclides, C
R
(mg/m
3
), are
essentially conservative in a chemical sense, but they
decay by falling apart in other nuclides and various types
of radiation. The radioactive decay rate (mg/m
3
/day) is
usually modelled as a first order process involving a
kinetic radioactive decay constant, kdec (day
1
). This
400 Water Resources Systems Planning and Management
kinetic constant is derived from the half-life time of the
radionuclide. The initial concentration may be expressed
as radioactivity, in order to simulate the activity instead of
the concentration. These state variables can be converted
into each other using:
Ac 10
3
N
A
kdecC
R
/[86400 Mw] (12.58)
where
Ac activity of the radionuclide (Bq/m
3
/s)
N
A
Avogadros number (6.02
*
10
23
mole)
Mw molecular weight of the radionuclide (g/mole).
4.10. Sediments
As discussed in the previous section, sediments play
an important role in the transport and fate of chemical
pollutants in water. Natural waters can contain a mixture
of particles ranging from gravel (2 mm to 20 mm) or
sand (0.07 mm to 2 mm) down to very small particles
classified as silt or clay (smaller than 0.07 mm). The very
fine fractions can be carried as colloidal suspension, for
which electrochemical forces play a predominant role.
Considering its large adsorbing capacities, the fine
fraction is characterized as cohesive sediment. Cohesive
sediment can include silt and clay particles as well as
particulate organic matter such as detritus and other
forms of organic carbon, diatoms and other algae. Since
flocculation and adsorbing capacities are of minor
importance for larger particles, they are classified as
non-cohesive sediment.
The behaviour of this fine-grained suspended matter
affects water quality in several ways. First, turbidity and
its effect on the underwater light climate is an important
environmental condition for algae growth. The presence
of suspended sediment increases the attenuation of
light in the water column, which leads to an inhibition
of photosynthetic activity, and hence a reduction in
primary production. Second, the fate of contaminants in
waters is closely related to suspended solids due to
their large adsorbing capacities. Like dissolved matter,
sediment is transported by advection and by turbulent
motion. In addition, the fate of the suspended cohesive
sediment is determined by settling and deposition, as well
as by bed processes of consolidation, bioturbation and
resuspension.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 400
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 401
4.10.1. Processes and Modelling Assumptions
From a water quality perspective, the three most important
sediment processes are sedimentation, resuspension or
erosion, and burial of sediment.
The modelling of sedimentation and erosion processes
originates in part from the PartheniadesKrone concept
(Partheniades, 1962; Krone, 1962). In this concept, the
bottom shear stress plays an essential role in defining
whether or not sedimentation of suspended particles or
erosion of bed material will occur. Sedimentation takes
place when the bottom shear stress drops below a critical
value. Resuspension occurs when the bottom shear stress
exceeds a critical value. The calculation of bed shear stress
is discussed in Section 4.10.5.
For purposes of modelling chemical concentrations
in the bottom sediment, one or more sediment layers
are usually defined. Each sediment layer is assumed
homogeneous (well mixed). The density of the layer can
vary, depending on the variable sediment layer composi-
tion. The porosity within a given layer is assumed constant
and user-defined. Water quality models do not generally
consider horizontal transport of bed sediments. This hori-
zontal transport would result in a change in the amount of
sediment present in the bottom, and hence a change in the
thickness of the layer of non-cohesive sediment.
4.10.2. Sedimentation
A characteristic feature of cohesive sediments is their
ability to form aggregates of flocs that settle to the bottom,
in a process called sedimentation. Whether a particle will
settle to the bottom depends on its size and density, and
the chemical conditions of the surrounding water system.
Various laboratory and field measurements show that the
suspended matter concentration strongly influences the
aggregation process and thereby the settling velocities of
the aggregates (Krone, 1962). Strong flocs are denser and
have larger settling velocities.
Sediment floc aggregation depends on the chemical
and physical properties of the sediment, and on salinity
and turbulence. At high sediment concentrations (several
g/l) the particles hinder each other, decreasing their
settling velocity. Turbulence affects the flocculation and
therefore the settling velocity in two opposing ways.
Increased turbulence will increase the collisions between
particles, resulting in larger flocs with higher settling
velocities, but increased turbulent shear stresses will
break up the flocs and decrease their settling velocities.
Sedimentation occurs when the bottom shear stress
drops below a critical value. Total shear stress is the sum of
the shear stresses induced by flows and wind waves.
Sedimentation rates can depend on salinity concentrations.
For any sediment particle j that settles out of the water
column, the rate of settling depends on the flow velocity
shear stress, (kg/m/sec
2
), at the bottom surfacewater
interface, the critical shear stress,
j
, for the substance j,
a zero order sedimentation rate, ZS
j
(g/m
2
/day), and the
sedimentation velocity, VS
j
(m/day), for the settling
substance, C
j
(g dry wt./m
3
/day), all divided by the depth,
H (m), of settling in the water column.
If the shear stress at the bottom-water interface is less
than the critical shear stress
j
for the substance j, settling
is assumed to take place. The rate of decrease of the
substance (g dry wt./m
3
/day) in the water column due to
settling is
dC
j
/dt [1 (/
j
)] (ZS
j
VS
j
C
j
)/H
j
(12.59)
The corresponding flux (g dry wt./m
2
of sediment layer
area/day) of substance j onto the sediment layer (accre-
tion) is
dC
j
sed
/dt (dC
j
/dt) H (12.60)
Sedimentation always results in an increase of the
substance settling in the upper sediment bed layer. The
quantity of sedimentation in one model time-step cannot
exceed the available amount of substance in the water
column.
No net resuspension is assumed to take place if
sedimentation occurs.
4.10.3. Resuspension
If the shear stress exceeds the critical shear stress
j
for
the type substance j, then no net settling takes place.
Instead, resuspension can take place (also called
erosion).
Erosion of cohesive bed material occurs when the bed
shear forces exceed the resistance of the bed sediment. The
resistance of the bed is characterized by a certain critical
erosive strength (bottom shear stress). This critical stress is
determined by several factors, such as the chemical
wrm_ch12.qxd 8/31/2005 11:59 AM Page 401
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
composition of the bed material, particle size distribution
and bioturbation. Erosion of sediment is induced by the
bed stress due to flow velocities, tidal and wind-induced
advective flows, and surface waves. Erosion is directly
proportional to the excess of the applied shear stress over
the critical erosive bottom shear stress. One formula for
erosion of homogeneous beds is based on Partheniades
(1962). The erosion/resuspension flux is limited by the
available amount of sediment on the sea bed. Typically a
one-layer homogeneous bed is assumed.
The rate of resuspension (g dry wt./m
2
of bottom layer
area/day) of substance j going into the water column
depends on these flow velocity shear stresses, a zero-order
resuspension flux ZR (g/m
2
/day) if any, a resuspension
rate constant kR
j
(day
1
), and the amount of substance
C
j
sed
in the top sediment layer (g/m
2
);
dC
j
/dt [(/
j
) 1](ZR
j
kR
j
C
j
sed
)/H
j
(12.61)
The corresponding flux (g dry wt./m
2
sediment layer
area/day) of substance C
j
sed
from the sediment layer area
(erosion) is
dC
j
sed
/dt (dC
j
/dt) H (12.62)
No sedimentation is assumed if resuspension occurs.
Resuspension results in a decrease of sediment in the
upper sediment bed layer. The sedimentation in one
model time step cannot exceed the available amount
of substance in the sediment layer. The values of ZR
j
and
strongly depend on the sediment properties and envi-
ronmental parameter values.
For water quality considerations, the chemical compo-
sition of the resuspending sediment is assumed to be
the same as that of the bottom sediment. Resuspension of
chemical component concentrations, N
j
(g/m
3
/day), in
a substance from the sediment layer is simply the
substance, C
sed
resuspension flux (g dry wt./m
2
/day)
(Equation 12.62) times the fraction, fr
j
, of component j in
the substance divided by the depth, H, of water column:
dN
j
/dt fr
j
(dC
sed
/dt)/H (12.63)
4.10.4. Burial
Burial is a term used to convey that bed sediment is
no longer available for resuspension because it has
been covered by much newer sediment, i.e. it has been
402 Water Resources Systems Planning and Management
buried. This concept is important especially with respect
to contaminated sediments, as it implies that buried
contaminated sediments are isolated from the overlying
water column and no longer pose a threat to the water
quality.
Burial is a sink for sediments that are otherwise
susceptible for resuspension. Assuming a fixed active
sediment layer thickness, burial occurs when this thick-
ness is exceeded. Establishing this requires knowledge
of the depth, h
j
(m/g dry wt/m
2
) of particle size
substance j per unit dry wt per square metre. Hence, if
the total depth of the active sediment layer,
j
C
j
sed
h
j
, is
greater than the assumed maximum depth, D
max
(m), of
active sediment layer, then burial will result. The
decrease in available sediment substance j will be in
proportion to its contribution, C
j
sed
h
j
, to the total depth,

j
C
j
sed
h
j
, of the sediment layer. The burial flux will
equal a burial rate constant k
B
(day
1
and usually
set equal to 1) times the amount C
j
sed
of substance
j times the depth ratio times the excess depth ratio,
(
j
C
j
sed
h
j
D
max
)/(
j
C
j
sed
h
j
).
(12.64)
4.10.5. Bed Shear Stress
The bed shear stress directly influences the sedimentation
and erosion rates. It depends on the flow (currents) and
the wind-generated (and sometimes human- generated)
surface waves. For sedimentationerosion processes,
it is usually assumed that the total bed shear stress,
(Newton/m
2
or kg/m/s
2
) due to flow,
flow
, and waves,

wave
, are additive:

flow

wave
(12.65)
The bed shear stress
flow
for depth-averaged flow
depends on the water density (1,000 kg/m
3
), the hori-
zontal flow velocity, U
h
(m/s), acceleration of gravity,
g (9.81 m/s
2
), and a Chezy coefficient, Cz (m
0.5
/s).

flow
1,000 (9.81)U
h
2
/Cz
2
(12.66)
d d
bur sed sed sed
sed
C t k C C h C h
C h D C
j B j j j j j
j
j j
j
/
/
max

jj j
j
h
sed

wrm_ch12.qxd 8/31/2005 11:59 AM Page 402

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 403
The Chezy coefficient, Cz (m
0.5
/s), is defined as a function
of water depth, H (m), and bottom roughness length
(Nikuradse equivalent roughness length) (m), Rough (m),
or Mannings roughness coefficient, n (s/m
1/3
):
Cz 18 log
10
(12 H/Rough) (12.67)
Cz (H)
1/6
/n (12.68)
Both Nikuradses roughness coefficient and Mannings
roughness coefficient may be changing due to bed load
movements. Here (and in most models) they are assumed
fixed.
Surface waves are caused by wind stress on the water
surface. The magnitude of the waves depends on the
wind conditions, wind duration, water depth and bottom
friction. Wave fields are commonly described by the
significant wave height, significant wave period and
wavelength. Waves induce a vertical circular movement
(orbital velocity) that decreases with depth. The waves
exert friction forces on the bed during propagation.
The magnitude of the bed shear stress,
wave
, due to
waves depends on a wave friction factor, f
w
, the density of
water, (1,000 kg/m
3
) and the effective orbital horizontal
velocity at the bed surface, U
o
.

wave
0.25 (1,000) f
w
U
o
2
(12.69)
The wave friction factor, f
w
, and the effective orbital
horizontal velocity at the bed surface, U
o
, are functions
of three wave parameters: the significant wave height,
H
s
(m), the mean wave period, T
m
(s), and mean wave
length, L
m
(m) (see also Figure 12.16). Also required is
the depth of water, H (m).
The effective horizontal bottom velocity due to waves
is defined as
U
o
H
s
/[T
m
sinh(2H/L
m
)] (12.70)
The friction or shear factor, f
w
, can be calculated in two
ways (Monbaliu, et al., 1999). One way is
f
w
0.16 [Rough/(U
o
T
m
/2)]
0.5
(12.71)
The other way uses a factor depending on a parameter A
defined as
A H
s
/[2 Rough sinh(2H/L
m
)] (12.72)
If A 1.47 then
f
w
exp{5.977 5.123H
0.194
} (12.73)
Otherwise
f
w
0.32 (12.74)
4.11. Lakes and Reservoirs
The water quality modelling principles discussed above
are applicable to all different types of water systems:
streams, rivers, lakes, estuaries and even coastal or
ocean waters. This section presents some of the unique
aspects of water quality modelling in lakes. The physi-
cal character and water quality of rivers draining
into lakes and reservoirs are governed in part by the
velocity and volume of river water. The characteristics
of the river water typically undergo significant changes
as the water enters the lake or reservoir, primarily
because its velocity reduces. Sediment and other
material that were carried in the faster-flowing water
settle out in the basin.
The structure of the biological communities also
changes from organisms suited to living in flowing
waters to those that thrive in standing or pooled waters.
There are greater opportunities for the growth of
algae (phytoplankton) and the development of
eutrophication.
Reservoirs typically receive larger inputs of water (as
well as soil and other materials carried in rivers) than
lakes do, and as a result, may receive larger pollutant
loads. However, because of greater water inflows, flushing
rates are more rapid than in lakes. Thus, although reser-
voirs may receive greater pollutant loads than lakes, they
have the potential to flush out the pollutants more
rapidly. Reservoirs may therefore exhibit fewer or less
severe negative water quality or biological impacts than
lakes for the same pollutant load.
The water quality of lakes and reservoirs is defined by
variables measured within the water basin. Although
there are many variables of limnological significance,
crest crest
length
time of passage = wave period
still-water line
trough
height
E
0
2
0
8
0
1
a
Figure 12.16. Wave dimensions of significant wave height,
period and length.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 403
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
water quality is typically characterized on the basis of
conditions such as:
water clarity or transparency (greater water clarity
usually indicates better water quality)
concentration of nutrients (lower concentrations
indicate better water quality)
quantity of algae (lower levels indicate better water
quality)
oxygen concentration (higher concentrations are
preferred for fisheries)
concentration of dissolved minerals (lower values
indicate better water quality)
acidity (a neutral pH of 7 is preferred).
Many waste chemical compounds from industry, some
with toxic or deleterious effects on humans and/or other
water-dependent organisms and products, are discharged
into lakes and reservoirs. Some can kill aquatic organisms
and damage irrigated crops. Inadequate water purification
resulting in the discharge of bacteria, viruses and other
404 Water Resources Systems Planning and Management
organisms into natural waters can be a primary cause of
waterborne disease. Although dangerous to human health
worldwide, such problems are particularly severe in
developing countries.
There can be major differences between deep and
shallow lakes or reservoirs. Deep lakes, particularly in
non-tropical regions, usually have poorer water quality in
their lower layers, due to stratification (see Section 4.11.3).
Shallow lakes do not exhibit this depth differentiation
in quality. Their more shallow, shoreline areas have
relatively poorer water quality because those sites are
where pollutant inputs are discharged and have a greater
potential for disturbance of bottom muds and the like.
The water quality of a natural lake usually improves as
one moves from the shoreline to the deeper central part.
In contrast, the deepest end of a reservoir is usually
immediately upstream of the dam. Water quality usually
improves along the length of a reservoir, from the shallow
inflow end to the deeper, lake-like end near the dam, as
shown in Figure 12.17.
riverine zone transitional zone lacustrine zone
narrow, channelized basin
relatively high flow rates
light-limited primary
productivity
high suspended solids
nutrient supply by
advection
low light availability at
depth
relatively high nutrient
levels
cell losses primarily
by sedimentation
organic matter sypply
primarily allochthonous
more eutrophic
broader, deeper basin
reduced flow rates
primary productivity
relatively high
reduced suspended solids
advective nutrient supply
reduced
light availability at
depth
cell losses by grazing
and sedimentation
intermediate
intermediate
broad, deep, lake basin -like
low flow rates
nutrient-limited primary
productivity
relatively clear
nutrient supply by
internal recycling
high light availability at
depth
relatively low nutrient
levels
cell losses primarily
by grazing
organic matter sypply
primarily autochthonous
more oligotrophic
E
0
2
0
8
0
9
a
Figure 12.17. Longitudinal
zonation of water quality and
other variables in reservoirs
(UNDP, 2000).
wrm_ch12.qxd 8/31/2005 11:59 AM Page 404
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 405
Reservoirs, particularly the deeper ones, are also
distinguished from lakes by the presence of a longitudinal
gradient in physical, chemical and biological water
quality characteristics from the upstream river end to the
downstream dam end. Because of this, reservoirs have
been characterized as comprising three major zones: an
upstream riverine zone, a downstream lake-like zone at
the dam end, and a transitional zone separating these two
(Figure 12.17). The relative size and volume of the three
zones can vary greatly in a given reservoir.
4.11.1. Downstream Characteristics
Constructing a dam can produce dramatic changes in the
downstream river channel below it. These are quite unlike
downstream changes from lakes. Because reservoirs act
as sediment and nutrient traps, the water at the dam end
of a reservoir is typically of higher quality than water
entering the reservoir. This higher-quality water subse-
quently flows into the downstream river channel below
the dam. This phenomenon is sometimes a problem,
in that the smaller the quantity of sediments and other
materials transported in the discharged water, the greater
the quantity that it can now pick up and transport.
Because it contains less sediment, the discharged water
can scour and erode the streambed and banks, picking up
new sediment as it continues downstream. This scouring
effect can have significant negative impacts on the flora,
fauna and biological community structure in the down-
stream river channel. The removal of sediments from a
river by reservoirs also has important biological effects,
particularly on floodplains.
Many reservoirs, especially those used for drinking
supplies, have water release or discharge structures located
at different vertical levels in their dams (Figure 12.18).
This allows for the withdrawal or discharge of water
from different layers within the reservoir, known
as selective withdrawal. Depending on the quality of
the water discharged, selective withdrawal can signifi-
cantly affect water quality within the reservoir itself, as
well as the chemical composition and temperature of
the downstream river. Being able to regulate both
quantities and qualities of the downstream hydrological
regimes makes it possible to affect both flora and
fauna, and possibly even the geomorphology of the
stream or river.
Constructing a reservoir may have significant social
and economic implications, including the potential for
stimulating urban and agricultural development adjacent
to, and below, the reservoir. These activities can have
both positive and negative impacts on downstream
water quality, depending on the nature and size of
development.
Agriculture is often the leading source of pollution
in lakes. Healthy lake ecosystems contain nutrients in
small quantities from natural sources, but extra inputs of
nutrients (primarily nitrogen and phosphorus) unbalance
lake ecosystems. When temperature and light conditions
are favourable, excessive nutrients stimulate population
explosions of undesirable algae and aquatic weeds. After
the algae die, they sink to the lake bottom, where bacte-
ria consume the available dissolved oxygen as they
decompose the algae. Fish may die and foul odours may
result if dissolved oxygen is depleted.
Heavy metals are another major cause of lake quality
impairment. Our knowledge of this is mainly due to
the widespread detection of heavy metals in fish tissue
samples. Since it is difficult to measure mercury in
ambient water, and since it bioaccumulates in fish tissue,
fish samples are commonly used to indicate the level of
contamination. Common sources of mercury involve
atmospheric transport from power-generating facilities
and other smoke-stack industries.
In addition to nutrient and metal siltation, enrich-
ment by organic wastes that deplete oxygen and by
noxious aquatic plants affect lakes and reservoirs. Often,
Figure 12.18. A multiple-outlet reservoir can be better used
to regulate the temperature and water quality downstream.
E
0
2
0
7
3
0
y
C
B
A
thermocline
anoxic layer
wrm_ch12.qxd 8/31/2005 11:59 AM Page 405
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
several pollutants and processes affect a single lake. For
example, a process such as removal of shoreline vegeta-
tion may accelerate erosion of sediment and nutrients
into a lake. Extreme acidity (low pH) resulting from acid
rain can eliminate fish in isolated lakes. Urban runoff
and storm sewers, municipal sewage treatment plants,
and hydrological modifications are also sources of lake
pollutants.
The prediction of water quality in surface water
impoundments is based on mass-balance relationships
similar to those used to predict water quality concentra-
tions in streams and estuaries. There are also significant
problems in predicting the water quality of lakes or reser-
voirs compared to those of river and estuarine systems.
One is the increased importance of wind-induced mixing
processes and thermal stratification. Another, for reser-
voirs, is the impact of various reservoir-operating policies.
4.11.2. Lake Quality Models
Perhaps the simplest way to begin modelling lakes is
to consider shallow well-mixed constant-volume lakes
subject to a constant pollutant loading. The flux of any
constituent concentration, C, in the lake equals the mass
input of the constituent less the mass output less losses
due to decay or sedimentation, if any, all divided by the
lake volume V (m
3
). Given a constant constituent input
rate W
C
(g/day) of a constituent having a net decay and
sedimentation rate constant k
C
(day
1
) into a lake having
a volume V (m
3
) and inflow and outflow rate of
Q (m
3
/day), then the rate of change in the concentration
C (g/m
3
/day) is
dC/dt (1/V) (W
C
QC k
C
CV) (12.75)
Integrating this equation yields a predictive expression of
the concentration C(t) of the constituent at the end of any
time period t based in part on what the concentration,
C(t1), was at the end of the previous time period, t1.
For a period duration of t days,
C(t) [W
C
/(Q k
C
V)]
[1 exp{t((Q/V) k
C
)}]
C(t1) exp{t((Q/V) k
C
)}
(12.76)
The equilibrium concentration, C
e
, can be obtained by
assuming each concentration is equal in Equation 12.76
or by setting the rate in Equation 12.75 to 0, or by setting
406 Water Resources Systems Planning and Management
C(t1) equal to 0 and letting t go to infinity in Equation
12.76. The net result is
C
e
W
C
/(Q k
C
V) (12.77)
The time, t

, since the introduction of a mass input W

C
that is required to reach a given fraction of the equilib-
rium concentration (i.e., C(t)/C
e
) is
t

V[ln(1 )]/(Q K
C
V) (12.78)
Similar equations can be developed to estimate the
concentrations and times associated with a decrease in
a pollutant concentration. For the perfectly mixed lake
having an initial constituent concentration C(0), say after
an accidental spill, and no further additions, the change
in concentration with respect to time is
dC/dt C (Q k
C
V)/V (12.79)
Integrating this equation, the concentration C(t) is
C(t) C(0) exp{ t((Q/V) K
C
) (12.80)
In this case, one can solve for the time t

required for the

constituent to reach a fraction (1 ) of the initial con-
centration C(0) (that is, C(t)/C(0) (1 )). The result is
Equation 12.78.
Equation 12.76 can be used to form an optimization
model for determining the wasteload inputs to this well-
mixed lake that meet water quality standards. Assuming
that the total of all natural wasteloads W
C
(t), inflows and
outflows Q(t), and the maximum allowable constituent
concentrations in the lake, C(t)
max
, may vary among
different within-year periods t, the minimum fraction, X,
of total waste removal required can be found by solving
the following linear optimization model:
Minimize X (12.81)
Subject to the following mass-balance and constituent
concentration constraints for each period t:
C(t) [W
C
(t)(1X)/(Q(t) k
C
V)]
[1 exp{t((Q(t)/V) k
C
)}]
C(t1) exp{t((Q(t)/V) k
C
)} (12.82)
C(t) C(t)
max
(12.83)
If each period t is a within-year period, and if the waste-
loadings and flows in each year are the same, then no
initial concentrations need be assumed and a steady state
solution can be found. This solution will indicate, for the
loadings W
C
(t), the fraction X of waste removal that meet
the quality standards, C(t)
max
, throughout the year.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 406
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 407
4.11.3. Stratified Impoundments
Many deep impoundments become stratified during
particular times of the year. Vertical temperature
gradients arise that imply vertical density gradients. The
depth-dependent density gradients in stratified lakes
effectively prevent complete vertical mixing. Particularly
in the summer season, lakes may exhibit two zones: an
upper volume of warm water called the epilimnion and
a lower colder volume called the hypolimnion. The transi-
tion zone or boundary between the two zones is called the
thermocline (Figure 12.18).
Because of lake stratification, many models divide the
depth of water into layers, each of which is assumed to be
fully mixed. To illustrate this approach without getting
into too much detail, consider a simple two-layer lake in
the summer that becomes a one-layer lake in the winter.
This is illustrated in Figure 12.19
Discharges of a mass W
C
of constituent C in a flow
Q
in
(t) into the lake in period t have concentrations of
W
C
/Q
in
(t). The concentration in the outflows from the
summer epilimnion is C
e
(t) for each period t in the sum-
mer season. The concentration of the outflows from the
winter lake as a whole is C(t) for each period t in the
winter season. The summer time rates of change in the epil-
imnion constituent concentrations C
e
(t) and hypolimnion
concentrations C
h
(t) depend on the mass inflow, W
C
(t), and
outflow, C
e
(t)Q
out
(t), the net vertical transfer across thermo-
cline, (v/D
T
)[C
h
(t)V
h
(t) C
e
(t)V
e
(t)], the settling on sediment
interface, sH
h
(t)C
h
(t), and the decay, kC
e
(t):
dC
e
(t)/dt (1/V
e
(t)){(W
C
(t) C
e
(t)Q
out
(t))
(v/D
T
)[C
h
(t)V
h
(t) C
e
(t)V
e
(t)]}
kC
e
(t)
(12.84)
dC
h
(t)/dt kC
h
(t) (v/D
T
)[C
h
(t) C
e
(t)V
e
(t)/V
h
(t)]
sH
h
(t)C
h
(t) (12.85)
In the above two equations, V
e
and V
h
(m
3
) are the time-
dependent volumes of the epilimnion and hypolimnion
respectively; k (day
1
) is the temperature-corrected decay
rate constant; v (m/day) is the net vertical exchange
velocity that includes effects of vertical dispersion,
erosion of hypolimnion, and other processes that transfer
materials across the thermocline of thickness D
T
(m); s is
the settling rate velocity (m/day); and H
h
(t) is the average
depth of the hypolimnion (m).
summer stratification winter circulation
mass output
C Q (t)
out
(t)
settling
E
0
2
0
7
3
0
z
mass input
W
c
(t)
Figure 12.19. Lake stratification during
summer and complete mixing during
winter season.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 407
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
In the winter season the lake is assumed to be fully
mixed. Thus for all periods t in the winter season the
initial concentration of a constituent is:
C(t) C
e
(t)V
e
(t) C
h
(t)V
h
(t)/[V
e
V
h
] (12.86)
dC(t)/dt (1/V(t)){(W
C
(t) C(t)Q
out
(t))
kC
e
(t) sH(t)C(t) (12.87)
At the beginning of the summer season, each epilimnion
and hypolimnion concentration will be the same:
C
e
(t) C(t) (12.88)
C
h
(t) C(t) (12.89)
5. An Algal Biomass Prediction
Model
An alternative approach to modelling the nutrient,
oxygen and algae parts of an ecological model has
been implemented in a simulation model developed by
WL | Delft Hydraulics called DELWAQ-BLOOM (Los,
1991; Los et al., 1992; Molen et al., 1994; Smits 2001;
WL | Delft Hydraulics, 2003). This model is used to
predict algae growth and mortality, oxygen concentra-
tions and nutrient dynamics.
The ecological model DELWAQ-BLOOM calculates
the advection and dispersion of constituents (state
variables) in the water column, and the water quality and
ecological processes affecting the concentrations of the
constituents. It is based on a three-dimensional version of
the governing Equation 12.10. The focus here will be on
the source and sink terms in that equation that define the
water quality and ecological processes, mostly related to
algae growth and mortality, mineralization of organic
matter, nutrient uptake and release, and oxygen produc-
tion and consumption.
For this discussion, consider three nutrient cycles
(nitrogen, N, phosphorus, P, and silica, Si) and five
different groups of algae, (phytoplankton (diatoms, flag-
ellates and dinoflagellates) or macroalgae (attached
or suspended Ulva)), suspended and bottom detritus,
oxygen and inorganic phosphorus particulate matter
in the bottom sediments. Different predefined sets
of species are available for marine and freshwater
ecosystems.
408 Water Resources Systems Planning and Management
The model processes relating these substances are all
interrelated. However, for clarity, the processes can
be grouped into nutrient cycling, algae modelling, and
oxygen-related processes.
5.1. Nutrient Cycling
The DELWAQ-BLOOM model assumes that algae
consume ammonia and nitrate in the water column. It
includes the uptake of inorganic nutrients by bottom
algae, algae-mortality-producing detritus and inorganic
nutrients, mineralization of detritus in the water col-
umn producing inorganic nutrients, and mineralization
of detritus in the bottom that also produces inorganic
nutrients. Optionally the model includes algae species
with the ability to take up atomic nitrogen (N-fixation)
and detritus (mixotrophic growth). The model accounts
for the settling of suspended detritus and algae and
inorganic adsorbed phosphorus, resuspension of
bottom detritus, release of inorganic bottom nutrients
to the water, burial of bottom detritus, nitrification
or denitrification depending on the dissolved oxygen
concentration, and adsorption/desorption of ortho-
phosphate.
5.2. Mineralization of Detritus
The oxidation or mineralization of the nutrients in detri-
tus (DetN, DetP, DetSi) and also of detritus carbon (DetC)
reduces detritus concentrations. The mineralization
process consumes oxygen and produces inorganic nutri-
ents (NH
4
, PO
4
, and SiO
2
). The fluxes, dC/dt, for these
four constituents C (mg/l or g/m
3
) are assumed to be
governed by first order processes whose temperature
corrected rate constants are k
C

C
(T20)
(1/day). Thus:
dC/dt k
C

C
(T20)
C (12.90)
This equation applies in the water column as well as
in the bottom sediments, but the mineralization rate
constants, k
C

C
(T20)
, may differ. This rate constant also
depends on the stochiometric composition of detritus,
relative to the requirements of the bacteria. The concen-
trations of these detritus constituents in the bottom are
usually expressed in grams per square metre of surface
area divided by the depth of the sediment layer.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 408
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 409
5.3. Settling of Detritus and Inorganic
Particulate Phosphorus
The rate of settling of nutrients in detritus and inorganic
particulate phosphorus out of the water column and on to
the bottom is assumed to be proportional to their water
column concentrations, C. Settling decreases the concen-
trations of these constituents in the water column.
dC/dt SR
C
(C)/H (12.91)
The parameter SR
C
is the settling velocity (m/day) of
constituent concentration C and H is the depth (m) of the
water column.
5.4. Resuspension of Detritus and Inorganic
Particulate Phosphorus
The rates at which nutrients in detritus and inorganic
particulate phosphorus are resuspended depend on the
flow velocities and resulting shear stresses at the bed
surfacewater column interface. Below a critical shear
stress no resuspension occurs. Resuspension increases the
masses of these constituents in the water column without
changing its volume, and hence increases their concentra-
tions in it. For C
B
representing the concentration (grams
of dry weight per cubic metre) of resuspended material in
the active bottom sediment layer, the flux of constituent
concentration in the water column is
dC/dt RR
C
C
B
/H (12.92)
where RR
C
(m/day) is the velocity of resuspension
(depending on the flow velocity) and H is the depth of the
water column.
5.5. The Nitrogen Cycle
The nitrogen cycle considers the water-column compo-
nents of ammonia (NH
4
-N), nitrite and nitrate (represented
together as NO
3
-N), algae (AlgN), suspended detritus
(DetN), and suspended (non-detritus) organic nitrogen
(OON). In the bottom, sediment bottom detritus (BDetN)
and bottom diatoms (BDiatN) are considered. Figure 12.20
shows this nitrogen cycle.
5.5.1. Nitrification and Denitrification
Two important reactions in the nitrogen nutrient cycle are
nitrification and denitrification. These reactions affect the
flux of ammonia and nitrate in the water column. Given
sufficient dissolved oxygen and temperature, nitrifying
bacteria in the water column transform ammonium to
mineralization
mineralization
mineralization
primary
production
autolysis
primary
production
mortality
nitrogen in
algae
nitrification
N
NH
denitrification
NO
3 4
2
gas
nitrogen in
bottom
sedimentation
nitrogen in
detritus
nitrogen in
organic waste
r
e
s
u
s
p
e
n
s
i
o
n
s
e
d
i
m
e
n
t
a
t
i
o
n
E
0
2
0
7
3
0
q
Figure 12.20. The nitrogen cycle
processes.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 409
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
nitrite and then nitrate. This can be considered as one
reaction,
NH
4

2O
2
NO
3

H
2
O2H

that occurs at a rate (gN/m

3
/day) of
NH
4
/dt k
NH
4

NH
4
(T20)
NH
4
(12.93)
Again, k
NH
4

NH
4
(T20)
is the temperature corrected rate
constant (1/day), and NH
4
is the concentration of
nitrogen in ammonium (gN/m
3
).
Bacterial activities decrease as temperatures decrease.
Bacterial activities also require oxygen. The nitrification
process stops if the dissolved oxygen level drops below
about 2 mg/l or if the temperature T is less than approxi-
mately 5 C.
For each gram of nitrogen in ammonium-nitrogen
NH
4
-N reduced by nitrification, a gram of nitrate-nitrogen
NO
3
-N is produced, consuming 2 moles (64 grams)
oxygen per mole (14 grams) of nitrogen (64/14 4.57
grams of oxygen per gram of nitrogen). Nitrification
occurs only in the water column.
In surface waters with a low dissolved-oxygen content,
nitrate can be transformed to free nitrogen by bacterial
activity as part of the process of mineralizing organic
material. This denitrification process can be written as:
organic matter 2NO
3
N
2
CO
2
H
2
O (12.94)
Nitrate is (directly) removed from the system by means of
denitrification. The reaction proceeds at a rate:
(12.95)
where NO
3
is the concentration of nitrate nitrogen
(gN/m
3
).
This process can occur both in the water column and
the sediment, but in both cases results in a loss of nitrate
from the water column. Algae also take up nitrate-
nitrogen. As with nitrification, denitrification decreases
with temperature. The reaction is assumed to stop below
about 5 C.
5.5.2. Inorganic Nitrogen
Ammonia is produced when algae die and the cells release
their contents into the surrounding water (in a process
called autolysis) and by the mineralization of organic
nitrogen in the water and bottom sediment. Ammonia is
converted to nitrate by nitrification. Algae use ammonia
dNO d NO
NO NO
( 20)
3 3
3 3
/ t k
T


410 Water Resources Systems Planning and Management
and nitrate for growth. Different algae prefer either NH
4
or
NO
3
nitrogen. Upon death they release part of their nitro-
gen contents as ammonia. The remaining nitrogen of dying
algae becomes suspended detritus and suspended other
organic nitrogen (OON). The latter degrades at a much
slower rate. Algae can also settle to the bottom. Some
macroalgae (Ulva) can be fixed to the bottom, unless wind
and water velocities are high enough to dislodge them.
Once in the bottom sediment, planktonic algae die and
release all their nitrogen contents, as ammonium into the
water column and to organic nitrogen in the sediment. In
contrast macroalgae and bottom diatoms are attached
and are subjected to the normal processes of growth,
mortality and respiration.
Suspended detritus and organic nitrogen are formed
upon the death of algae. Detritus is also produced by
excretion of phyto- and zooplankton and from resuspen-
sion of organic matter on and in the sediment. The
detritus concentration in the water column decreases
by bacterial decay, sedimentation and filtration by
zooplankton and benthic suspension feeders.
Bottom detritus is subject to the processes of miner-
alization, resuspension and burial. Mineralization of
bottom detritus is assumed to be slower than that of
suspended detritus. The ammonia produced from
mineralization is assumed to go directly to ammonia
in the water phase. Sedimentation from the water
column and mortality of algae in the bottom increase the
bottom pool of bottom detritus. The mineralization
rate depends on the composition of the detritus, i.e. is a
function of the nitrogen/carbon and phosphorus/carbon
ratios.
Nitrogen is removed from the system by means of
denitrification, a process that occurs under anoxic condi-
tions. Burial is a process that puts the material in a deep
sediment layer, and effectively removes it from the active
system. This is the only removal process for the other
nutrients (P and Si).
5.6. Phosphorus Cycle
The phosphorus cycle (Figure 12.21) is a simplified
version of the nitrogen cycle. There is only one dissolved
pool: orthophosphorus, and only one removal process:
burial. However, unlike nitrogen and silica, there is also
inorganic phosphorus in the particulate phase (AAP).
wrm_ch12.qxd 8/31/2005 11:59 AM Page 410
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 411
The phosphorus cycle in the water column includes
orthophosphate (PO
4
), algae (AlgP), suspended detritus
(DetP), suspended (non-detritus) organic phosphorus
(OOP), inorganic adsorbed (available) P (AAP), and
inorganic adsorbed (unavailable) P (UAP). In the bottom
sediment the cycle includes the bottom detritus (BDetP)
and the bottom inorganic adsorbed P (BAAP).
A reaction specific to the phosphorus cycle is
the adsorbtion/desorption of particulate inorganic phos-
phorus. Inorganic phosphorus can be present in the
aquatic environment in a dissolved form and adsorbed to
inorganic particles, such as calcium or iron. The transition
from one form into another is not a first order kinetic
process, yet in many models desorption of inorganic
phosphorus is assumed to be such.
5.7. Silica Cycle
The silica cycle is similar to the phosphorus cycle except
that there is no adsorption of silica to inorganic
suspended solids. Silica is only used by diatoms, so
uptake by algae depends on the presence of diatoms. The
silica cycle is shown in Figure 12.22.
The silica cycle in the water column includes dissolved
silica (Si), diatoms (Diat), suspended detritus (DetSi), and
suspended (non-detritus) organic silica (OOSi). In the
bottom sediment the cycle includes the bottom detritus
(BDetSi).
5.8. Summary of Nutrient Cycles
The nutrient cycles just described are based on the
assumption that nutrients can be recycled an infinite
number of times without any losses other than due to
transport, chemical adsorption, denitrification and burial.
This is an over-simplification of the organic part of the
mineralization
sedimentation
mineralization
primary
production
desorption
autolysis
mortality
phosphorus
in algae
PO
particulate
inorganic
phosphorus
4
phosphorus
in bottom
sedimentation
phosphorus
in detritus
phosphorus in
organic matter
r
e
s
u
s
p
e
n
s
i
o
n
s
e
d
i
m
e
n
t
a
t
i
o
n
E
0
2
0
7
3
0
r
mineralization
adsorption
Figure 12.21. The processes
involved in the phosphorus
cycle.
mineralization
primary
production
autolysis
mortality
silica
in diatoms
inorganic
dissolved silica
resuspension sedimentation
silica
in detritus
E
0
2
0
7
3
0
s
mineralization
silica
in the bottom
Figure 12.22. The processes involved in the silica cycle.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 411
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
nutrient cycles. The elementary composition of living
algae cells is a complicated function of their characteris-
tics as well as the environmental conditions. Upon dying,
the algae cell contents are released into the surrounding
water. As the cell breaks apart, many of the nutrients are
in a form that makes them instantly available for algae cell
growth (autolysis). The remaining material consists of
more or less degradable substances. Most of this material
is mineralized either in the water or at the bottom, but a
small portion degrades very slowly if at all. Most of this
material settles and is ultimately buried. Resuspension
delays but does not stop this process by which nutrients
are permanently removed from the water system.
For simplicity, all possible removal processes are
lumped into a single term, which is modelled as burial.
For example, if a nominal value of 0.0025 per day is used,
this means that 0.25% of the bottom amount is buried
each day.
The same formulation is used for all three nutrients.
Whether or not this is correct depends on the actual
removal process. If deactivation is mainly burial into
deeper layers of the sediment, there is no reason to
distinguish between different nutrients. Other processes
such as chemical binding, however, may deactivate
phosphorus, but not nitrogen or silica.
5.9. Algae Modelling
Algal processes include primary production, mortality
(producing detritus and inorganic nutrients) and grazing,
settling to become bottom algae and the resuspension of
bottom algae, the mortality of bottom algae to bottom
detritus, and the burial of bottom algae.
The basic behaviour of algae in surface water can be
illustrated by the two diagrams in Figures 12.23 and
12.24. These show the nutrient and carbon fluxes for
diatoms and other algae. Diatoms are distinguished from
other algae in that they need silicate to grow.
5.9.1. Algae Species Concentrations
The module BLOOM II computes phytoplankton within
DELWAQ-BLOOM. It is based upon the principle of
competition between different species. The basic variables
of this module are called types. A type represents the
412 Water Resources Systems Planning and Management
physiological state of a species. Usually a distinction
is made between three different types: an N-type for
nitrogen limitation, a P-type for phosphorus limitation
and an E-type for light energy limitation, whose nutrient
content is high (luxury uptake). The solution algorithm of
the model considers all potentially limiting factors and
first selects the one that is most likely to become limiting.
It then selects the best-adapted type for that limitation.
The suitability of a type (its fitness) is determined by the
ratio of its requirement for that particular resource and
its growth rate in equal proportion. This means that a
type can become dominant either because it needs a
inorganic carbon
NH
phosphate
silica
diatoms
4
and NO
3
carbon detritus
nitrogen detritus
phosphorus
detritus
silica detritus
net production mortality mineralization
E
0
2
0
7
3
0
t
Figure 12.23. Modelling of diatoms.
inorganic carbon
NH
phosphate
4
and NO
3
carbon detritus
nitrogen detritus
phosphorus
detritus
net production mortality mineralization
E
0
2
0
7
3
0
u
other
algae
Figure 12.24. Modelling of other algae besides diatoms.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 412
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 413
comparatively small amount of a limiting resource (it is
efficient) or because it grows rapidly (it is opportunistic).
Now the algorithm considers the next potentially
limiting factor and again selects the best-adapted phyto-
plankton type. This procedure is repeated until it is
impossible to select a new pair of a type and limiting
factor without violating (that is, over-exhausting) some
limiting factor. Thus the model seeks the optimum
solution consisting of n types and n limiting factors. To
that purpose, it uses linear programming. Typically,
BLOOM II considers between three and ten representative
algae species and six to fifteen types.
As a further refinement BLOOM II takes the existing
biomasses of all phytoplankton types into account. These
are the result of growth, loss processes and transport
during the previous time steps. Thus, the optimization
algorithm does not start from scratch.
As they represent different physiological stages of the
same species, the transition of one type to another is a rapid
process with a characteristic time step in the order of a day.
The equations of the model are formulated in a way that
allows such a rapid shift between types of the same species.
A transition between different species is a much slower
process as it depends on mortality and net growth rates.
It is interesting that the principle just described, by
which each phytoplankton type maximizes its own
benefit, effectively means that the total net production of
the phytoplankton community is maximized.
Denote each distinct species subtype (from now
on called type) by the index k. The BLOOM II module
identifies the optimum concentration of biomass, B
k
, of
each algae type k that can be supported in the aquatic
environment characterized by light conditions and nutri-
ent concentrations. Using the net growth rate constant Pn
k
,
it will
Maximize Pn
k
B
k
(12.96)
The sum of the biomass concentrations over all types k is
the total algae biomass concentration.
For each algae type, the requirements for nitrogen,
phosphorus and silica (used only by diatoms) are speci-
fied by coefficients n
ik
, the fraction of nutrient i per unit
biomass concentration of algae type k.
The total readily available concentration, C
i
(g/m
3
) of
each nutrient in the water column equals the amount in
k

the total living biomass of algae,

k
(n
ik
B
k
), plus the
amount incorporated in dead algae, d
i
, plus that dissolved
in the water, w
i
. These mass-balance constraints apply for
each nutrient i.
(n
ik
B
k
) d
i
w
i
C
i
(12.97)
The unknown concentration variables B
k
, d
i
and w
i
are
non-negative. All nutrient concentrations C
i
are the
modelled (optionally measured) total concentrations and
are assumed to remain constant throughout the time
period (typically a day).
5.9.2. Nutrient Recycling
A certain amount of each algae type k dies in each time
step, and this takes nutrients out of the live phytoplank-
ton pool. Most of it remains in the detritus and other
organic nutrient pools; a smaller fraction (in the order of
30%) is directly available to grow new algae because the
dead cells break apart (autolysis) and are dissolved in the
water column. Detritus may be removed to the bottom or
to the dissolved nutrient pools at rates in proportion to its
concentration. Needed to model this are the mortality
rate, M
k
(day
1
), of algae type k; the fraction, f
p
; of dead
phytoplankton cells that is not immediately released
when a cell dies; the remineralization rate constant, m
i
(day
1
), of dead phytoplankton cells; the fraction, n
ik
, of
nutrient i per unit biomass concentration of algae type k;
and the settling rate constant, s (day
1
), of dead phyto-
plankton cells.
The rate of change in the nutrient concentration of
the dead phytoplankton cells, dd
i
/dt, in the water column
equals the increase due to mortality less that which
remineralizes and that which settles to the bottom;
dd
i
/dt (f
p
M
k
n
ik
B
k
) m
i
d
i
sd
i
(12.98)
Both mortality and mineralization rate constants are
temperature dependent.
5.9.3. Energy Limitation
Algae absorb light for photosynthesis. Energy becomes
limiting through self-shading when the total light absorp-
tion of algae, called light extinction, exceeds the
maximum at which primary production is just balanced
k

wrm_ch12.qxd 8/31/2005 11:59 AM Page 413

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
by respiration and mortality. For each algae type k there
exists a specific extinction value K
k
max
(1/m) at which this
is the case. The light intensity can also be too high, which
means the total extinction is too low (photo-inhibition)
for growth. This specific extinction value is K
k
min
.
The ranges between K
k
min
and K
k
max
differ for different
algal types k. Letting K
k
(m
3
/m/gdry) represent the
specific light absorbing extinction constant for living
material of algae type k, the total extinction due to all
living algae is
KL (K
k
B
k
) (12.99)
Added to this must be the extinction caused by dead cells,
KD, and the background extinction of the water, KW
(1/m). As defined here, KW includes the extinction by
inorganic suspended matter and humic substances.
K
k
min
KL KD KW K
k
max
(12.100)
In an alternate formulation used in DELWAQ-BLOOM,
the total light extinction coefficient in the water column
(in m
1
) is the sum of background extinction (K
b
, water
and non-modelled substances) and extinction due to
detritus (K
det
), inorganic material (K
im
), dissolved organic
matter or yellow substances (K
dom
) and phytoplankton
(K
phyt
):
Extinction K
b
K
det
K
im
K
dom
K
phyt
(12.101)
The extinction from each component is calculated by
its concentration multiplied by its specific extinction
coefficient.
The ratio between the various extinction components
is highly site specific. In general it also varies seasonally.
The amount of dead cells not yet mineralized is, from
Equation 12.98,
k
(fpM
k
B
k
). Assuming some fraction e
d
(usually between 0.2 and 0.4) of the extinction rate of
live cells,
KD e
d
K
k
fpM
k
B
k
(12.102)
If the total extinction is not within the range for an algae
type k, then its concentration B
k
will be zero. To ensure
that B
k
is 0 if the total extinction is outside of its extinc-
tion range, a 0, 1 binary (integer) unknown variable Z
k
is needed for each algae type k. If Z
k
is 1, B
k
can be any
non-negative value; if it is 0, B
k
will be 0. This is modelled
by adding three linear constraints for each algae type k.
k

414 Water Resources Systems Planning and Management

KL KD KW K
k
max
KM (1 Z
k
) (12.103)
KL KD KW K
k
min
(Z
k
) (12.104)
B
k
BMZ
k
(12.105)
where KM and BM are any large numbers no less than the
largest possible value of the total extinction or biomass
concentration, respectively. Since the objective of maxi-
mizing the sum of all B
k
together with Equation 12.105 is
to set each Z
k
value equal to 1, it is only when the total
extinction is outside of the extinction range K
k
min
to K
k
max
that the Z
k
value will be forced to 0. Equation 12.105 then
forces the corresponding B
k
to 0.
5.9.4. Growth Limits
For all algae types k, the maximum potential biomass
concentration, B
k
max
(g dry/m
3
), at the end of the time
interval t (days) depends on the initial biomass concen-
tration, B
k
, (g dry/m
3
); the maximum gross production
rate, Pg
k
max
(day
1
); the respiration rate constant,
R
k
(day
1
); and a time and depth averaged production
efficiency factor, E
k
. Mortality is not included in this
computation. Using the net production rate constant,
Pn
k
( Pg
k
max
E
k
R
k
) (day
1
), for each algae type k:
B
k
max
B
k
0
exp{Pn
k
t} (12.106)
This condition is taken into account by the optimization
algorithm. For types with an initial biomass of zero, a
small base level is used instead in order to allow
previously absent species to start growing.
5.9.5. Mortality Limits
As in the case of growth, the mortality of each algae
species is also constrained to prevent a complete removal
within a single time step. The minimum biomass value of
a species is obtained when there is no production, but
only mortality. The minimum biomass, B
k
min
(g dry/m
3
), of
type k at the end of time interval t depends on the
initial biomass, B
k
0
(g dry/m
3
), of type k and the specific
mortality rate constant, M
k
(day
1
), of type k.
B
k
min
B
k
0
Z
k
exp{M
k
t} (12.107)
These maximum and minimum values are computed for
each individual algae type. However, the model sums each
of these maximum and minimum values over all subtypes
wrm_ch12.qxd 8/31/2005 11:59 AM Page 414
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 415
within each species and applies it to the total biomass of
the species. This way the maximum possible mortality
cannot be exceeded, but transitions between limit types
remain possible.
B
k
min
B
k
B
k
max
species j
(12.108)
As mortality is computed according to a negative
exponential function, the minimum biomass level is
always positive; in other words, a species can never
disappear completely. For computational purposes, the
minimum biomass of a species is set to zero if it drops
below some small threshold base level. It is conceivable
that the amount of biomass, which should minimally be
maintained according to Equation 12.108 exceeds the
amount permitted according to the available amount of
light energy (12.100). If this happens, Condition (12.100)
is dropped.
5.9.6. Oxygen-Related Processes
The oxygen concentration in the water column depends
in part on the biochemical and physical processes that
either produce or consume oxygen. For the algae
the model includes the production of oxygen (primary
production), and its consumption through respiration.
Oxygen is also consumed by the mineralization of
detritus and other organic material (in the water column
and bottom sediment), nitrification of ammonia to nitrate,
and exchange of oxygen with the atmosphere (e.g.
re-aeration).
The mineralization of carbon detritus in the water
column and bottom consumes oxygen at a molar ratio of
1:1, equivalent to a ratio of 32/12 g O
2
to 1 g C. The net
growth of algae produces oxygen at a molar ratio of 1:1,
equivalent to a ratio of 32/12 g O
2
to 1 g C. The miner-
alization of organic carbon in waste and carbon detritus
in the water column and bottom consumes oxygen
at a molar ratio of 1:1, equivalent to a ratio of 32/12 g
O
2
to 1 g C. The nitrification reaction consumes oxygen
at a molar ratio of 2:1, equivalent to a ratio of 64/14 g
O
2
to 1 g N. For all algae, oxygen is produced during
photosynthesis. The net growth of algae produces
oxygen at a molar ratio of 1:1, equivalent to a ratio of
32/12 g O
2
to 1 g C.
k of species j

k of species j

k of species j

Bottom Oxygen
The mineralization of organic material in the bottom
sediment consumes oxygen, which must be supplied
from the water column. The process consumes oxygen at
a molar ratio (oxygen to carbon) of 1:1, equivalent to a
ratio of 32/12 g O
2
to 1 g C.
Daily Oxygen Cycle
Because oxygen is produced by the photosynthesis of
algae during the daylight hours, there is a natural varia-
tion of oxygen concentrations over the twenty-four-hour
daynight cycle. Typically, oxygen concentrations are
lowest in the early morning as oxygen is consumed
during the night through the processes of algae respira-
tion and organic material mineralization. During the
daylight hours, oxygen is produced, and the highest
values (often supersaturated) are typically found in the
late afternoon. When biomasses are high, these variations
may be large enough to cause low oxygen conditions
during the night or in the early morning.
In the traditional BLOOM calculation, the water
quality processes are all calculated for a daily averaged
situation. This is reflected by the choice of the input
parameters for the light model: the daily averaged solar
radiation and the day length. Reducing the time step would
be the most straightforward way to include diurnal
variations. The drawback, however, is a considerable
increase in computation time. Thus an alternative approach
has been adopted in the model. The total daily rate of
primary production is computed first. Next, this produc-
tion is distributed over the day. The model takes into
account the day length, and oxygen production begins in
the first daylight hour. Oxygen production increases during
the morning, levels off at a (user-defined) maximum value
for a period in the middle of the day, and decreases during
the afternoon. There is no oxygen production during the
night-time. The hourly oxygen production is combined
with the daily averaged oxygen consumption processes
and the re-aeration to produce an hourly value of oxygen
concentration in the water.
Maintenance Respiration
Respiration in algae is a process in which organic carbon
is oxidized, using oxygen to produce energy. The process
occurs throughout the day and results in oxygen
wrm_ch12.qxd 8/31/2005 11:59 AM Page 415
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
consumption at a rate of 32 grams of oxygen per
12 grams of carbon: 32/12 g O
2
to 1 g C. Total respiration
is divided into growth and maintenance respiration.
Maintenance respiration is a first-order temperature
dependent process. Species growth respiration is
accounted for when calculating the net primary produc-
tion. Maintenance respiration, and thus the amount of
oxygen consumed, is governed by the temperature-
dependent respiration rate constant, k
k
res

k
res(T20)
(day
1
),
the temperature, T (C), of the water (higher respiration
at higher temperatures), and the concentration of algae
biomass, AlgC
k
. Each algae type can have a different
respiration rate. Hence, the rate (gC
2
/m/day) at which
carbon in algae is oxidized is
dAlgC
k
/dt k
k
res

k
res(T20)
AlgC
k
(12.109)
6. Simulation Methods
Most of those who will be using water quality models will
be using simulation models that are commonly available
from governmental agencies (e.g. USEPA), universities, or
private consulting and research institutions such as
the Danish Hydraulics Institute, Wallingford software or
WL | Delft Hydraulics (Ambrose et al., 1996; Brown and
Barnwell, 1987; Cerco and Cole, 1995; DeMarchi et al.,
1999; Ivanov et al., 1996; Reichert, 1994; USEPA, 2001;
WL | Delft Hydraulics, 2003).
These simulation models are typically based on
numerical methods that incorporate a combination of
plug flow and continuously stirred reactor approaches
to pollutant transport. Users must divide streams, rivers,
and lakes and reservoirs into a series of well-mixed
segments or volume elements. A hydrological or hydro-
dynamic model calculates the flow of water between all
of these. In each simulation time step, plug flow enters
these segments or volume elements from upstream
segments or elements. Flow also exits from them to
downstream segments or elements. During this time
the constituents can decay or grow, as appropriate,
depending on the conditions in those segments or
volume elements. At the end of each time-step, the
volumes and their constituents within each segment or
element are fully mixed. The length of each segment
or the volume in each element reflects the extent of
dispersion in the system.
416 Water Resources Systems Planning and Management
6.1. Numerical Accuracy
Water quality simulation models based on physical,
biological and chemical processes typically include time
rate of change terms such as dC/dt. While it is possible
to solve some of these differential equations analytically,
most water quality simulation models use numerical
methods. The purpose of this section is not to explain
how this can be done, but rather to point to some of the
restrictions placed on the modeller because of these
numerical methods.
Consider first the relationship between the stream,
river or lake segments and the duration of time steps, t.
The basic first-order decay flux, dC/dt (g/m
3
/day), for a
constituent concentration, C, that is dependent on a rate
constant, k (day
1
), is:
dC/dt kC (12.110)
The finite difference approximation of this equation can
be written
C(t t) C(t) C(t)kt (12.111)
or
C(t t) C(t)(1 kt) (12.112)
This equation can be used to illustrate the restriction
placed on the term kt. That term cannot exceed a value
of 1 or else C(t t) will be negative.
Figure 12.25 is a plot of various values of
C(t t)/C(t) versus kt. This plot is compared with
the analytical solution resulting from the integration of
Equation 12.110, namely:
C(t t) C(t) exp{ kt} (12.113)
Reducing the value of t will increase the accuracy of the
numerical solution. Hence, for whatever value of t, it
can be divided by a positive integer n to become 1/n
th
of
its original value. In this case the predicted concentration
C(t t) will equal
C(t t) C(t)(1 kt/n)
n
(12.114)
For example if kt 1, and n 2, the final concentra-
tion ratio will equal
C(t t)/C(t) (1 1/2)
2
0.25 (12.115)
Compare this 0.25 to 0.37, the exact solution, and to
0.0, the approximate solution when n is 1. Having n 2
brings a big improvement. If n 3, the concentration
wrm_ch12.qxd 8/31/2005 11:59 AM Page 416
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 417
ratio will be 0.30, an even greater improvement compared
to 0. However, no matter what value of n is selected,
the predicted concentration will always be less than the
actual value based on Equation 12.113, and hence the
error is cumulative. Whenever t n/k, the predicted
concentrations will alternate between positive and
negative values, either diverging, converging or just
repeating the cycle, depending on how much t exceeds
n/k. In any event, the predicted concentrations are not
very useful.
Letting m n/kt, Equation 12.114 can be
written as
C(t t) C(t)(1 1/m)
m (kt)
(12.116)
As n approaches infinity, so does the variable m, and
hence the expression (1 1/m)
m
becomes the natural
logarithm base e 2.718282. Thus, as n approaches
infinity, Equation 12.114 becomes Equation 12.113, the
exact solution to Equation 12.110.
6.2. Traditional Approach
Most water quality simulation models simulate quality
over a consecutive series of discrete time periods. Time is
divided into discrete intervals and the flows are assumed
constant within each of those time period intervals. Each
water body is divided into segments or volume elements,
and these are considered to be in steady-state conditions
within each simulation time period. Advection or plug
flow (i.e. no mixing or dispersion) is assumed during each
time period. At the end of each period mixing occurs
within each segment or volume element to obtain the
concentrations in the segment or volume element at
the beginning of the next time step.
This method is illustrated in Figure 12.26. The indices
i1, i and i1 refer to stream or river reach segments.
The indices t and t1 refer to two successive time peri-
ods, respectively. At the beginning of time period t, each
segment is completely mixed. During the time interval t
of period t, the water quality model predicts the concen-
trations, assuming plug flow in the direction of flow from
segment i toward segment i 1. The time interval t is
such that the flow from any segment i does not pass
through any following segment i 1. Hence, at the end of
each time period each segment has some of the original
water that was there at the beginning of the period,
and its end-of-period concentrations of constituents, plus
some of the immediately upstream segments water and
its end-of-period concentrations of constituents. These
two volumes of water and their respective constituent
concentrations are then mixed to achieve a constant
concentration within the entire segment. This is done for
all segments in each time step. Included in this plug flow
and then mixing process are the inputs to the reach from
point and non-point sources of constituents.
In Figure 12.26 a mass of waste enters reach i at a rate
of W
i
t
. The volume in each reach segment is denoted by V,
and the flows from one segment to the next are denoted
by Q. The drawing on the left represents a portion of a
stream or river divided into well-mixed segments. During
a period t, waste constituents enter segment i from the
immediate upstream reach i1 and from the point waste
source. In this illustration, the mass of each of these
wastes is assumed to decay during each time period,
independent of other wastes in the water. Depending on
the types of waste, the decay (or even growth) processes
that take place may be more complex than those assumed
in this illustration. At the end of each time period these
altered wastes are mixed together to create an average
concentration for the entire reach segment. This illustra-
tion applies for each reach segment i and for each time
period t.
The length, x
i
, of each completely mixed segment or
volume element depends on the extent of dispersion.
Reducing the length of each reach segment or size of each
volume element reduces the dispersion within the entire
stream or river. Reducing segment lengths, together with
1.0
0.5
0.0
0.5
1.0
C
(
)
t
+
t

C
(
t
)
k t
0.1 0.3 0.5 0.7 0.9 1.1
-
-
E
0
2
0
8
0
1
b
Figure 12.25. Plot of numerical approximation (red line)
based on Equation 12.112 compared to the true analytical
(blue line) value obtained from Equation 12.113.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 417
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
increasing flow velocities, also reduces the allowable
duration of each time period t. The duration of each
simulation time step t must be such that flow from any
segment or element enters only the adjacent down-
stream segment or element during that time step. Stated
formally, the restriction is:
t T
i
(12.117)
where T
i
is the residence time in reach segment or volume
element i. For a 1-dimensional stream or river system
consisting of a series of segments i of length x
i
, cross
section area A
i
and average flow Q
it
, the restriction is:
t min{x
i
A
i
/Q
it
; i, t} (12.118)
If time steps are chosen that violate this condition, then
numerical solutions will be in error. The restriction
defined by Equation 12.118 is often termed the
courant condition. It limits the maximum time-step
value. Since the flows being simulated are not always
known, this leads to the selection of very small time
steps, especially in water bodies having very little
dispersion. While smaller simulation time steps
increase the accuracy of the model output, they also
418 Water Resources Systems Planning and Management
increase the computational times. Thus the balance
between computational speed and numerical accuracy
restricts the model efficiency in the traditional
approach to simulating water quality.
6.3. Backtracking Approach
An alternative Lagrangian or backtracking approach to
water quality simulation eliminates the need to consider
the simulation time-step duration restriction indicated by
Equation 12.118 (Manson and Wallis, 2000; Yin, 2002).
The backtracking approach permits any simulation time-
step duration to be used along with any segmenting
scheme. Unlike the traditional approach, water can travel
through any number of successive segments or volume
elements in each simulation time step.
This approach differs from the traditional one in that,
instead of following the water in a segment or volume
element downstream, the system tracks back upstream to
find the source concentrations of the contaminants at
time t that will be in the control volume or segment i 1
at the beginning of time period t 1.
i-1
i
i+1
i
t
W
concentration C
C
i
i
t
t+1
in
segment having volume i
i
t
V
at beginning of time period t
concentration in segment
after mixing at end of time period
i
t
={( C
i-1
t
Q
t
i-1, i
+W
i
t
) t +
Q
t
i, i+1
C
i
t
( V
i
t +1
- t )} exp (-kt ) / V
concentration in segment
at end of time period before mixing
i
t
Q
t
i-1
[ C
i-1
t
exp (-k t ) upstream:
, i
+W
i
t
t
Q
t
i-1
C
i
t
exp (-k t ) downstream:
i,
V
i
t
t [
]
- ]
E
0
2
0
8
0
1
d
t
i
Figure 12.26. Water quality
modelling approach showing a
water system schematized into
computational cells.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 418
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 419
The backtracking process works from upstream to
downstream. It starts from the segment of interest, i, and
finds all the upstream sources of contaminants that
flow into segment i during time period t. The contami-
nants could come from segments in the same river reach
or storage site, or from upstream river reaches or storage
volume segments. They could also come from incremen-
tal flows into upstream segments. Flows between the
source site and the segment i1 transport the contami-
nants from their source sites to segment i during the time
interval t, as shown in Figure 12.27.
The simulation process for each segment and for
each time period involves three steps. To compute the
concentration of each constituent in segment i at the end
of time period t, as shown in Figure 12.27, the approach
first backtracks upstream to locate all the contaminant
particles at the beginning of period t that will be in the
segment i at the end of period t. This is achieved by
finding the most upstream and downstream positions of
all reach intervals that will be at the corresponding
boundaries of segment i at the end of time period t. To do
this requires computing the velocities through each of
the intermediate segments or volume elements. Second,
the changes in the amounts of the modelled quality
constituents, such as temperature, organics, nutrients and
toxics, are calculated, assuming plug flow during the
time interval, t, and using the appropriate differential
equations and numerical methods for solving them.
Finally, all the multiple incoming blocks of water
with their end-of-period constituent concentrations are
completely mixed in the segment i to obtain initial
concentrations in that segment for the next time step,
t1. This is done for each segment i in each time period
t, proceeding in the downstream direction.
If no dispersion is assumed, the backtracking process
can be simplified to consider only the end points of each
reach. Backtracking can take place to each end-of-reach
location whose time of travel to the point of interest is just
equal to or greater than t. Then, using interpolation
between end-of-period constituent concentrations at
those upstream sites, plus all loadings between those sites
and the downstream site of interest, the constituent
E
0
2
0
8
0
1
g
C
1
t
, V
1
t
C
2
t
, V
t
2
C
5
t
, V
t
5
incremental waste
load mass
segment i
C
3
t
, V
3
t
C
4
t
,V
t
4
C
ij
t+1
=C
j
t
exp{
ij
-k
j
( ij-20) T
t }
j =1,2,3,4,5
C
i
t+1
=
j
C
ij
t+1
V
j
t
/ V
i
t+1
after mixing:
Figure 12.27. The backtracking
approach for computing the
concentrations of constituents in
each reach segment or volume
element i during time-step
duration of t.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 419
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
concentrations at the end of the time period t at the
downstream ends of each reach can be computed. This
process, like the one involving fully mixed reach
segments, must take into account the possibility of
multiple paths from each pollutant source to the site of
interest, and the different values of rate constants,
temperatures and other water quality parameters in each
reach along those paths.
Figure 12.27 illustrates an example of backtracking
involving simple first-order decay processes. Assume
contaminants that end up in reach segment i at the
beginning of period t 1 come from J sources with initial
concentrations C
1
t
, C
2
t
, C
3
t
, , C
j
t
at the beginning of time
period t. Decay of mass from each source j during time
t in each segment or volume element is determined by
the following differential equation:
dC
j
t
/dt k
j

j
(T20)
C
j
t
(12.119)
The decay rate constant k
j
, temperature correction coeffi-
cient
j
and water temperature T are all temporally and
spatially varied variables. Their values depend on the
particular river reaches and storage volume sites through
which the water travels during the period t from sites j to
segment i.
Integrating Equation 12.119 yields:
C
j
t1
C
j
t
exp{ k
j

j
(T20)
t} (12.120)
Since t is the time it takes water having an initial
concentration C
j
t
to travel to reach i, the values C
j
t1
can be
denoted as C
ij
t1
:
C
ij
t1
C
j
t
exp{ k
ij

j
(Tij20)
t} (12.121)
In Equation 12.121 the values of the parameters are the
appropriate ones for the stream or river between the
source segments j and the destination segment i. These
concentrations times their respective volumes, V
j
t
, can
then be mixed together to define the initial concentra-
tion, C
i
t 1
, in segment i at the beginning of the next
time period t 1.
6.4. Model Uncertainty
There are two significant sources of uncertainty in water
quality management models. One stems from incomplete
knowledge or lack of sufficient data to estimate the
probabilities of various events that might happen.
420 Water Resources Systems Planning and Management
Sometimes it is difficult to even identify possible future
events. This type of uncertainty sometimes called
epistemic (Stewart, 2000) stems from our incomplete
conceptual understanding of the systems under study, by
models that are necessarily simplified representations
of the complexity of the natural and socio-economic
systems, as well as by limited data for testing hypotheses
and/or simulating the systems.
Limited conceptual understanding leads to parameter
uncertainty. For example, there is an ongoing debate
about the parameters that can best represent the fate
and transfer of pollutants through watersheds and water
bodies. Arguably, more complete data and more work on
model development can reduce this uncertainty. Thus, a
goal of water quality management should be to increase
the availability of data, improve their reliabilities and
advance our modelling capabilities.
However, even if it were possible to eliminate knowl-
edge uncertainty, complete certainty in support of water
quality management decisions will probably never be
achieved until we can predict the variability of natural
processes. This type of uncertainty arises in systems
characterized by randomness. Assuming past observa-
tions are indicative of what might happen in the future
and with the same frequency in other words, assuming
stationary stochastic processes we can estimate from
these past observations the possible future events or
outcomes that could occur and their probabilities. Even if
we think we can estimate how likely any possible type of
event may be in the future, we cannot predict precisely
when or to what extent that event will occur.
For ecosystems, we cannot be certain we know even
what events may occur in the future, let alone their
probabilities. Ecosystems are open systems in which it is
not possible to know in advance what all the possible
biological outcomes will be. Surprises are not only
possible, but likely; hence, neither type of uncertainty
knowledge uncertainty nor unpredictable variability or
randomness can be eliminated.
Thus, uncertainty is a reality of water quantity and
quality management. This must be recognized when
considering the results of water quality management
models that relate to actions taken to meet the desired
water quality criteria and designated uses of water bodies.
Chapter 9 suggests some ways of characterizing this
uncertainty.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 420
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 421
7. Conclusions: Implementing a
Water Quality Management Policy
This chapter has provided only a brief introduction to
some of the relationships contained in water quality
models. As can be said for the other chapters as well, it
summarizes a subject on which entire texts, and very
good ones, have been written (see, for example, Chapra,
1997; McCutcheon, 1989; Orlob, 1983; Schnoor, 1996;
Thomann and Mueller, 1987). Water quality modelling
and management demand skill and data. Skill comes with
experience. Sufficient expertise will not be gained by
working just with the material introduced in this chapter.
It serves only as an introduction to surface water quality
models, their assumptions and their limitations.
If accompanied by field data and uncertainty analysis,
many existing models can be used to assist those respon-
sible for developing water quality management plans in
an adaptive implementation or management framework.
Adaptive implementation or management will allow
for both model and data improvements over time. Such
approaches strive toward achieving water quality stan-
dards while relying on monitoring and experimentation
to reduce uncertainty. This is often the only way one can
proceed, given the complexity of the real world compared
to the predictive models and the data and time usually
available at the time a water quality analysis is needed.
Starting with simple analyses and iteratively expanding
data collection and modelling as the need arises is a
reasonable approach.
An adaptive management process begins with initial
actions that have reasonable chances of succeeding.
Future actions must be based on continued monitoring
of the water body to determine how it responds to the
actions taken. Plans for future regulation and public
spending should be subject to revision as stakeholders
learn more about how the system responds to actions
taken. Monitoring is an essential aspect of adaptive water
quality management and modelling (see Appendix B).
Regardless of what immediate actions are taken, there
may not be an immediate measurable response. For
example, there may be significant lags between the time
when actions are taken to reduce nutrient loads and
the resulting changes in nutrient concentrations. This is
especially likely if nutrients from past activities are tightly
bound to sediments or if nutrient-contaminated ground-
water has a long residence time before its release to
surface water. For many reasons, lags between actions
taken and responses must be expected. Water bodies
should be monitored to establish whether the trajectories
of the measured water quality criteria point toward
attainment of the designated use.
Wasteload allocations will inevitably be required if
quality standards are not being met. These involve costs.
Different allocations will have different total costs and
different distributions of those costs; hence they will have
different perceived levels of fairness. A minimum-cost
policy may result in a cost distribution that places most of
the burden on just some of the stakeholders. But until
such a policy is identified, one will not know this. An
alternative may be to reduce loads from all sources by the
same proportion. Such a policy has prevailed in the
United States over the past several decades. Even though
not very cost-effective from the point of view of water
quality management, the ease of administration and the
fulfilment of other objectives must have made such a
policy politically acceptable, even though expensive.
However, these types of wasteload allocations policies will
not in themselves be sufficient for many of the ecosystem
restoration efforts that are increasingly being made.
Restoration activities are motivated in part by the services
ecosystems provide for water quality management.
Our abilities to include ecosystem components within
water quantity and quality management models are at a
fairly elementary level. Given the uncertainty, especially
with respect to the prediction of how ecosystems will
respond to water management actions, together with the
need to take actions now, long before we can improve
these capabilities, the popular call is for adaptive manage-
ment. The trial and error aspects of adaptive management
based on monitoring and imperfect models may not satisfy
those who seek more definitive direction from water qual-
ity analysts and their predictive models. Stakeholders and
responsible agencies seeking assurances that the actions
taken will always work as predicted may be disappointed.
Even the best predictive capabilities of science cannot
ensure that an action that will lead to the attainment of
designated uses will be initially identified. Adaptive
management is the only reasonable option in most cases
for allowing water quality management programmes to
move forward in the face of considerable uncertainties.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 421
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
8. References
AMBROSE, R.B.; BARNWELL, T.O.; MCCUTCHEON,
S.C. and WILLIAMS, J.R. 1996. Computer models for
water quality analysis. In: L.W. Mays (ed.), Water
Resources Handbook. New York, McGraw-Hill. Chapter 14.
BORSUK, M.E.; STOW, C.A. and RECKHOW, K.H.
2004. A Bayesian network of eutrophication models for
synthesis, prediction, and uncertainty analysis. Ecological
Modelling, Vol. 173, pp. 21939.
BROWN, L.C. and BARNWELL, T.O. Jr. 1987. The
enhanced stream water-quality models QUAL2E and QUAL2E-
UNCAS: documentation and user manual. EPA-600/3-87/007.
Athens, Ga., EPA Environmental Research Laboratory.
CERCO, C.F. and COLE, T. 1995. Users guide to the
CE-QUAL-ICM three-dimensional eutrophication model,
release version 1.0. Technical Report EL-95-15. Vicksburg,
Miss., USACE Waterways Experiment Station.
CHAPRA, S. 1997. Surface water quality modelling.
New York, McGraw-Hill.
DEMARCHI, C.; IVANOV, P.; JOLMA, A.; MASLIEV, I.;
SMITH, M. and SOMLYDY, L. 1999. Innovative tools for
water quality management and policy analysis: DESERT
and STREAMPLAN. Water Science and Technology, Vol. 40,
No. 10, pp. 10310.
ELMORE, H.L. and HAYES, T.W. 1960. Solubility of
atmospheric oxygen in water. Journal of the Environmental
Engineering Division, ASCE, Vol. 86, No. SA4, pp. 4153.
ENGELUND, F. and HANSEN, E. 1967. A monograph
on sediment transport in alluvial streams. Copenhagen,
Teknisk forlag.
IVANOV, P.; MASLIEV, I.; DE MARCHI, C. and
SOMLYDY, L. 1996. DESERT: decision support system
for evaluating river basin strategies; users manual.
International Institute for Applied Systems Analysis.
Laxenburg, Austria, IISA.
KARR, J.R. 1990. Bioassessment and non-point source
pollution: an overview. In: Second National Symposium on
Water Quality Assessment. Washington, D.C., EPA Office
of Water; pp. 418.
KRONE, R.B. 1962. Flume studies of the transport of
sediment in estuarial shoaling processes. Berkeley, Calif.,
University of California, Hydraulic and sanitary engine-
ering laboratory.
422 Water Resources Systems Planning and Management
LOS, F.J. 1991. Mathematical simulation of algae blooms by
the model BLOOM II, Version 2, T68. Delft, the Netherlands,
WL | Delft Hydraulics Report.
LOS, F.J. et al. 1992. Process formulations DBS. WL | Delft
Hydraulics, model documentation T542. Delft, the
Netherlands, Delft Hydraulics, (In Dutch.)
MANSON, J.R. and WALLIS, S.G. 2000. A conservative
semi-lagrangian fate and transport model for fluvial
systems I. Theoretical Development. Journal of Water
Research, Vol. 34, No. 15, pp. 376977.
MCCUTCHEON, S.C. 1989. Water Quality Modelling,
Vol. 1. Boca Raton, Fla., CRC.
MOLEN, D.T. VAN DER; LOS, F.J.; VAN BALLEGOOI-
JEN, L. and VAN DER VAT, M.P. 1994. Mathematical
modelling as a tool for management in eutrophication
control of shallow lakes. Hydrobiologia, Vol. 2756,
pp. 47992.
MONBALIU, J.; HARGREAVES, J.C.; CARRETERO, J.C.;
GERRITSEN, H. and FLATHER, R.A. 1999. Wave model-
ling in the PROMISE project. Coastal Engineering, Vol. 37,
Nos. 34, pp. 379407.
NRC (National Research Council). 1992. Restoration of
aquatic ecosystems. Washington, DC, National Academy
Press.
NRC (National Research Council). 2001. Assessing the
TMDL approach to water quality management committee to
assess the scientific basis of the total maximum daily load
approach to water pollution reduction. Washington, D.C.,
Water Science and Technology Board, National Academy
Press.
ORLOB, G.T. (ed.). 1983. Mathematical modelling of
water quality: streams, lakes and reservoirs. Chichester,
UK, Wiley.
PARTHENIADES, E. 1962. A study of erosion and
deposition of cohesive soils in salt water. Berkeley, Calif.,
University of California.
REICHERT, P. 1994. AQUASIM: a tool for simulation
and data analysis of aquatic systems. Water Science and
Technology, Vol. 30, No. 2, pp. 2130.
SCHNOOR, J.L. 1996. Environmental modelling, fate and
transport of pollutants in water, air and soil. New York,
Wiley.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 422
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 423
SMITS, J. 2001. DELWAQ-BLOOM-switch. Delft, Delft
Hydraulics.
STEWART, T.R. 2000. Uncertainty, judgment, and error
in prediction. In: D. Sarewitz, R.A. Pielke Jr. and R. Byerly
Jr. (eds.), Prediction: science, decision-making, and the future
of nature. Washington, D.C., Island Press.
STREETER, H.W. and PHELPS, E.B. 1925. A study of the
pollution and natural purification of the Ohio River, III. Factors
concerned in the phenomena of oxidation and reaeration.
Washington, D.C., US Public Health Service.
THOMANN, R.V. and MUELLER, J.A. 1987. Principles
of surface water quality modelling and control. New York,
Harper and Row.
USEPA. 2001. BASINS Version 3.0 Users Manual. EPA-
823-B-01-001. Washington, D.C., EPA Office of Water
and Office of Science and Technology.
WL | DELFT HYDRAULICS. 2003. Delft3D-WAQ, user
manual; versatile water quality modelling in 1D, 2D, or 3D
systems including physical, (bio)chemical and biological
processes. Delft, the Netherlands, WL | Delft Hydraulics.
YIN, H. 2002. Development of a watershed information
system, M.Sc. Thesis. Ithaca, N.Y., Cornell University,
Civil and Environmental Engineering Department.
Additional References (Further Reading)
ASCE. 1999. National stormwater best management prac-
tices (BMP) data-base: version 1.0. Users Guide and CD.
Washington, D.C. Prepared by Urban Water Resources
Research Council of ASCE, and Wright Water Engineers,
Inc., Urban Drainage and Flood Control District, and URS
Greiner Woodward Clyde, in cooperation with EPA
Office of Water.
BECK, M.B. 1987. Water quality modelling: a review of
the analysis of uncertainty. Water Resources Research,
No. 23, pp. 1393442.
BECK, M.B. and VAN STRATEN, G. (eds.). 1983. Uncertainty
and forecasting of water quality. Berlin, Springer-Verlag.
BISWAS, A.K. (ed.). 1997. Water resources: environmental
planning, management, and development. New York,
McGraw-Hill.
BOWIE, G.L.; MILLS, W.B.; PORCELLA, D.B.; CAMPBELL,
C.L.; PAGENKOPF, J.R.; RUPP, G.L.; JOHNSON, K.M.;
CHAN, P.W.H.; GHERINI, S.A. and CHAMBERLIN, C.E.
1985. Rates, constants, and kinetics: formulations in surface
water quality modelling, 2nd edn. Report EPA/600/3-85/040.
Athens, Ga., US EPA.
CHURCHILL, M.A.; ELMORE, H.L. and BUCKINGHAM,
R.A. 1962. Prediction of stream reaeration rates. Journal of
the Sanitary Engineering Division, ASCE Vol. 88, No. SA4,
No. 1, pp. 146.
DELVIGNE, G.A.L. 1980. Natural reaeration of surface
water. WL | Delft Hydraulics, Report on literature study
R1149. Delft, WL | Delft Hydraulics, (In Dutch.)
DITORO, D.M.; PAQUIN, P.R.; SUBBURAMU, K. and
GRUBER, D.A. 1990. Sediment oxygen demand model:
methane and ammonia oxidation. Journal of Environmental
Engineering, ASCE, Vol. 116, No. 5, pp. 94586.
DOBBINS, W.E. 1964. BOD and oxygen relationships in
streams. Journal of the Sanitary Engineering Division, ASCE,
Vol. 90, No. SA3, pp. 5378.
European Union. 2001a. The EU water framework directive.
http://europa.eu.int/water/water-framework/index_en.html.
Accessed 21 June.
European Union. 2001b. Water protection and management:
framework directive in the field of water policy. http://
www.europa.eu.int/scadplus/leg/en/lvb/l28002b.htm.
Accessed 20 June.
European Union. 2001c. Water protection and management:
urban waste water treatment. http://www.europa.eu.int/
scadplus/leg/en/lvb/l28008.htm. Accessed 20 June.
FAGERBAKKE, K.M.; HELDAL, M. and NORLAND, S.
1996. Content of carbon, nitrogen, oxygen, sulfur and
phosphorus in native aquatic and cultured bacteria.
Aquatic Microbial Ecology, No. 10, pp. 1527.
GROMIEC, M.J.; LOUCKS, D.P. and ORLOB, G.T. 1982.
Stream quality modelling. In: G.T. Orlob (ed.),
Mathematical modelling of water quality. Chichester, UK,
Wiley, pp. 176226.
HORNBERGER, G.M. and SPEAR, R.C. 1981. An approach
to the preliminary analysis of environmental systems.
Journal of Environmental Management, No. 12, pp. 718.
HORNBERGER, G.M. and SPEAR, R.C. 1983. An
approach to the analysis of behaviour and sensitivity in envi-
ronmental systems. In: M.B. Beck and G. van Straten (eds.),
wrm_ch12.qxd 8/31/2005 11:59 AM Page 423
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Uncertainty and forecasting of water quality, Berlin, Springer-
Verlag, pp. 10116.
KARR, J.R. 2000, Health, integrity, and biological assess-
ment: the importance of whole things. In: D. Pimentel,
L. Westra, and R.F. Noss (eds.), Ecological integrity: inte-
grating environment, conservation, and health, Washington,
D.C., Island Press, pp. 20926. and KARR, J.R.; CHU,
E.W. Sustaining living rivers. Hydrobiologia, Nos. 4223,
pp. 114.
KARR, J.R. and DUDLEY, D.R. 1981. Ecological perspec-
tive on water quality goals. Environmental Management,
No. 5, pp. 5568.
LANGBIEN, W.B. and DURUM, W.H. 1967. The aeration
capacity of streams, Circular 542. Washington, D.C.,
USGS.
MAIDMENT, D.R. (ed.). 1993. Handbook of hydrology.
New York, McGraw-Hill.
MASLIEV, I.; SOMLYDY, L. and KONCSOS, L. 1995. On
reconciliation of traditional water quality models and activated
sludge models. Working paper WP 9518. Laxenburg,
Austria, International Institute for Applied Systems Analysis.
MILLS, W.B.; PORCELLA, D.B.; UNGS, M.J.; GHERINI,
S.A.; SUMMERS, K.V.; MOK, L.; RUPP, G.L.; BOWIE, G.L.
and HAITH, D.A. 1985. Water quality assessment: a screen-
ing procedure for toxic and conventional pollutants in surface
and ground water, Parts I and II. EPA/600/6-85/002a,b.
Athens, Ga., Environmental Protection Agency.
MORGAN, M.G. and HENRION, M. 1990. Uncertainty.
New York: Cambridge University Press.
NATIONAL ACADEMY OF PUBLIC ADMINISTRATION.
2000. Transforming environmental protection for the
twenty-first century. Washington, D.C., National Academy
of Public Administration.
NOVOTNY, V. 1999. Integrating diffuse/nonpoint pollu-
tion control and water body restoration into watershed
management. Journal of the American Water Resources
Association, Vol. 35, No. 4, pp. 71727.
NOVOTNY, V. and OLEM, H. 1994. Water quality:
prevention, identification and management of diffuse
pollution. New York, Van Nostrand-Reinhold (distrib-
uted by Wiley). Ohio, Ohio EPA.
OCONNOR, D.J. 1961. Oxygen balance of an estuary.
J. San. Eng. Div. ASCE, Vol. 86, No. SA3, pp. 3555.
424 Water Resources Systems Planning and Management
OCONNOR, D.J. and DOBBINS, W.E. 1958. Mechanism
of reaeration in natural streams. Transactions of the
American Society of Civil Engineers, No. 123, pp. 64166.
OWENS, M.; EDWARDS, R.W. and GIBBS, J.W. 1964.
Some reaeration studies in streams. International Journal of
Air and Water Pollution, No. 8, pp. 46986.
PETERS, R.H. 1991. A critique for ecology. Cambridge,
Cambridge University Press.
REICHERT, P. 1995. Design techniques of a computer
program for the identification of processes and the simu-
lation of water quality in aquatic systems. Environmental
Software and Modeling, Vol. 10, No. 3, pp. 199210.
REICHERT, P. 2001. River Water-quality model no. 1
(RWQM1): case study II. Oxygen and nitrogen conversion
processes in the River Glatt (Switzerland). Water Science
and Technology, Vol. 43, No. 5, pp. 5160.
REICHERT, P.; BORCHARDT, D.; HENZE, M.; RAUCH, W.;
SHANAHAN, P.; SOMLYDY, L. and VANROLLEGHEM, P.
2001. River water-quality model No. 1 (RWQM1): II.
Biochemical process equations. Wat. Sci. Tech., Vol. 43,
No. 5, pp. 1130.
REICHERT, P. and VANROLLEGHEM, P. 2001. Identifi-
ability and uncertainty analysis of the river water-quality
model No. 1 (RWQM1). Water Science and Technology,
Vol. 43, No. 7, pp. 32938.
ROESNER, L.A.; GIGUERE, P.R. and EVENSON, D.E.
1981. Computer program documentation for the stream
quality model QUALII. Report EPA 600/981014.
Athens, Ga., USEPA.
SHANAHAN, P.; BORCHARDT, D.; HENZE, M.;
RAUCH, W.; REICHERT, P.; SOMLYDY, L. and
VANROLLEGHEM, P. 2001. River Water-quality model
no. 1 (RWQM1): I Modelling approach. Water Science and
Technology, Vol. 43, No. 5, pp. 19.
SHANAHAN, P.; HENZE, M.; KONCSOS, L.; RAUCH, W.;
REICHERT, P.; SOMLYDY, L. and VANROLLEGHEM, P.
1998. River water quality modelling: II. Problems of the art.
Water Science and Technology, Vol. 38, No. 11, pp. 24552.
SHEN, H.W. and JULIEN, P.Y. 1993. Erosion and
sediment transport. In: D.R. Maidment (ed.), Handbook of
hydrology, Chapter 12. New York, McGraw-Hill.
SMITH, R.A.; SCHWARZ, G.E. and ALEXANDER, R.B.
1997. Regional interpretation of water quality monitoring
wrm_ch12.qxd 8/31/2005 11:59 AM Page 424
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water Quality Modelling and Prediction 425
data. Water Resources Research, Vol. 33, No. 12,
pp. 278198.
SOMLYDY, L. 1982. Water quality modelling: a com-
parison of transport-oriented and ecology-oriented
approaches. Ecological Modelling, No. 17, pp. 183207.
SOMLYDY, L.; HENZE, M.; KONCSOS, L.; RAUCH, W.;
REICHERT, P.; SHANAHAN, P. and VANROLLEGHEM, P.
1998. River water quality modelling III. Future of the
art. Water Science and Technology, Vol. 38, No. 11,
pp. 25360.
SOMLYDY, L. and VAN STRATEN, G. (eds.). 1986.
Modelling and managing shallow lake eutrophication, with
application to Lake Balaton. Berlin, Springer-Verlag.
SOMLYDY, L. and VARIS, O. 1993. Modelling the quality
of rivers and lakes. In: P. Zannetti (ed.), Environmental
modelling: computer methods and software for simulating
environmental pollution and its adverse effects, Vol. 1,
pp. 21357. Southampton, UK, CMP.
SPANOU, M. and CHEN, D. 2001. Water quality model-
ling of the Upper Mersey River system using an object-
oriented framework. Journal of Hydroinformatics, Vol. 3,
No. 3, pp. 17394.
SPEAR, R. and HORNBERGER, G.M. 1980. Eutrophication
in Peel Inlet II. Identification of critical uncertainties
via generalized sensitivity analysis. Water Research, Vol. 14,
pp. 439.
STUMM, W. and MORGAN, J.J. 1981. Aquatic Chemistry.
New York, Wiley.
THOMAS, H.A. Jr. 1948. Pollution load capacity of streams.
Water and Sewage Works, No. 95, p. 409.
TOXOPEUS, A.G. 1996. An interactive spatial and temporal
modelling system as a tool in ecosystem management.
Enschede, the Netherlands, International Institute for
Aerospace Survey and Earth Sciences, ITC.
ULANOWICZ, R.E. 1997. Ecology, the ascendant perspec-
tive. New York, Columbia University Press.
USEPA. 1993. The watershed protection approach: the
annual report 1992. EPA 840-S-93-001. Washington,
D.C., EPA Office of Water.
USEPA. 1994. Water quality standards handbook, 2nd edn.
EPA 823-B-94-005a. Washington, D.C., EPA Office of
Water.
USEPA. 1995a. Environmental indicators of water quality in
the United States. EPA 841-R-96-002. Washington, D.C.,
Office of Policy, Planning, and Evaluation.
USEPA. 1995b. A conceptual framework to support develop-
ment and use of environmental information in decision-
making. EPA 239-R-95-012. Washington, D.C., Office of
Policy, Planning, and Evaluation.
USEPA. 1995c. Ecological restoration: a tool to manage
stream quality. Report EPA 841-F-95-007. Washington,
D.C., USEPA.
USEPA. 1998. Lake and Reservoir bioassessment and biocri-
teria: technical guidance document. EPA 841-B-98-007.
Washington, DC: EPA Office of Water.
USEPA. 1999. Draft guidance for water quality-based
decisions: the TMDL process, 2nd edn. Washington, D.C.,
EPA Office of Water.
USEPA. 2000. Stressor identification guidance document.
EPA-822-B-00-025. Washington, D.C., EPA Office of
Water and Office of Research and Development.
VAN PAGEE, J.A. 1978. Natural reaeration of surface water
by the wind. Report on literature study R1318-II. Delft, the
Netherlands, WL | Delft Hydraulics. (In Dutch.)
VAN RIJN, L.C. 1984. Bed load transport (part I),
suspended load transport (part II). Journal of Hydraulic
Engineering, Vol. 110, Nos. 101, pp. 143156, 161341.
VAN STRATEN, G. 1983. Maximum likelihood estimation
of parameters and uncertainty in phytoplankton models.
In: M.B. Beck and G. van Straten (eds.), Uncertainty and
forecasting of water quality. Berlin, Springer-Verlag.
YOUNG, P. 1998. Data-based mechanistic modelling of
environmental, ecological, economic and engineering
systems. Environmental Modelling and 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 Modelling and
Software, Vol. 15, pp. 25765.
wrm_ch12.qxd 8/31/2005 11:59 AM Page 425
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
13. Urban Water Systems
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
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
wrm_ch13.qxd 8/31/2005 12:00 PM Page 426
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Urban water infrastructure typically includes water
collection and storage facilities at source sites, water trans-
port via aqueducts (canals, tunnels and/or pipelines) from
source sites to water treatment facilities; water treatment,
storage and distribution systems; wastewater collection
(sewage) systems and treatment; and urban drainage works.
This is illustrated as a simple schematic in Figure 13.1.
Generic simulation models of components of urban
water systems have been developed and are commonly
applied to study specific component design and operation
issues. Increasingly, optimization models are being used
to estimate cost-effective designs and operating policies.
Cost savings can be substantial, especially when applied
to large complex urban systems (Dandy and Engelhardt,
2001; Savic and Walters, 1997).
Most urban water users require high-quality water,
and natural surface and/or groundwater supplies, called
raw water, often cannot meet the quality requirements of
domestic and industrial users. In such situations, water
treatment is required prior to its use. Once it is treated,
urban water can then be stored and distributed within the
427
Urban Water Systems
Today, a simple turn of the tap provides clean water a precious resource.
Engineering advances in managing this resource with water treatment,
supply, and distribution systems changed life profoundly in the twentieth
century, virtually eliminating waterborne diseases in developed nations, and
providing clean and abundant water for communities, farms, and industries.
So states the US National Academy of Engineering on its selection of water supply
systems to be among the five greatest achievements of engineering in the
twentieth century. As populations continually move to urban areas for improved
opportunities and a higher standard of living, and as cities merge to form
megacities, the design and management of water supply systems serving these
urban areas becomes an increasingly important part of regional integrated water
resources planning and management.
13
urban area, usually through a network of storage tanks
and pipes. Pipe flows in urban distribution systems
should be under pressure to prevent contamination from
groundwaters and to meet various user and fire-
protection requirements.
After use, the wastewater is collected in a network of
sewers, or in some cases ditches, leading to a wastewater
treatment plant or discharge site. In many urban areas the
sewage system has a dual function. The sewers collect both
wastewater from households and the runoff from streets
and roofs during storm events. However, the transport
capacity of the sewer network and the treatment facilities
are limited. During intense rainfall, overflows from the
sewage system discharge a mixture of surface runoff and
wastewater to the surface waters. This has a negative
impact on the water quality of urban surface waters.
Wastewater treatment plants remove some of the
impurities in the wastewater before it is discharged into
receiving water bodies or on land surfaces. Water bodies
receiving effluents from point sources such as wastewater
treatment plants may also receive runoff from the
surrounding watershed area during storm events. The
pollutants in both point and non-point discharges will
wrm_ch13.qxd 8/31/2005 12:00 PM Page 427
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
affect the quality of the water in those receiving water
bodies. The fate and transport of these pollutants in these
water bodies can be predicted by using water quality
models similar to those discussed in Chapter 12.
This chapter briefly describes these urban water
system components and reviews some of the general
assumptions incorporated into optimization and simu-
lation models used to plan urban water systems. The
focus of urban water systems modelling is mainly on
the prediction and management of quantity and qual-
ity of flows and pressure heads in water distribution
networks, wastewater flows in gravity sewer networks,
and on the design efficiencies of water and wastewater
treatment plants. Other models can be used for the
real-time operation of various components of urban
systems.
2. Drinking Water
Drinking water issues include demand estimation, water
treatment, and distribution.
428 Water Resources Systems Planning and Management
2.1. Water Demand
A secure water supply is of vital importance for the health
of the population and for the economy. Drinking water
demand depends on:
the number of inhabitants with access to drinking
water
meteorological and climatological conditions
the price of drinking water
the availability of drinking water
an environmental policy that aims at moderate use of
drinking water.
Table 13.1 shows an overview of the total annual water
demand in various countries. The total water demand is
sub-divided into domestic use and agricultural and indus-
trial water use.
Drinking water demand for domestic use shows a daily
and seasonal variation. There is no general formula for
predicting drinking water demand. Drinking water sup-
pliers tend to make predictions on the basis of their own
experience and historical information about water
demand in their region.
treatment
E
0
2
0
8
0
9
n
pollution of water ways
waste of scarce resource
pipes leaking
pollution of groundwater
Figure 13.1. Schematic showing
urban surface water source,
water treatment prior to urban
use, and some sources of non-
point urban drainage and runoff
and its impacts.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 428
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 429
2.2. Water Treatment
Before water is used for human consumption, its harmful
impurities need to be removed. Communities that do
not have adequate water treatment facilities, a common
problem in developing regions, often have high incidences of
disease and mortality due to drinking contaminated water. A
range of syndromes, including acute dehydrating diarrhoea
(cholera), prolonged febrile illness with abdominal symp-
toms (typhoid fever), acute bloody diarrhoea (dysentery) and
chronic diarrhoea (Brainerd diarrhoea). Numerous health
organizations point to the fact that contaminated water leads
to over 3 billion episodes of diarrhoea and an estimated 2
million deaths, mostly among children, each year.
Contaminants in natural water supplies can also include
microorganisms such as Cryptosporidium and Giardia lam-
blia as well as inorganic and organic cancer-causing chemicals
(such as compounds containing arsenic, chromium, copper,
lead and mercury) and radioactive material (such as radium
and uranium). Herbicides and pesticides reduce the suitabil-
ity of river water as a source of drinking water. Recently,
traces of hormonal substances and medicines detected in river
water are generating more and more concern.
To remove impurities and pathogens, a typical municipal
water purification system involves a sequence of processes,
from physical removal of impurities to chemical treatment.
Physical and chemical removal processes include initial and
final filtering, coagulation, flocculation, sedimentation and
disinfection, as illustrated in the schematic of Figure 13.2.
As shown in Figure 13.2, one of the first steps in most
water treatment plants involves passing raw water through
coarse filters to remove sticks, leaves and other large solid
objects. Sand and grit settle out of the water during this
stage. Next a chemical such as alum is added to the raw
water to facilitate coagulation. As the water is stirred, the
alum causes the formation of sticky globs of small particles
made up of bacteria, silt and other impurities. Once these
globs of matter are formed, the water is routed to a series of
settling tanks where the globs, or floc, sink to the bottom.
This settling process is called flocculation.
After flocculation, the water is pumped slowly across
another large settling basin. In this sedimentation or
clarification process, much of the remaining floc and solid
material accumulates at the bottom of the basin. The
clarified water is then passed through layers of sand, coal
and other granular material to remove microorganisms
including viruses, bacteria and protozoa such as
Cryptosporidium and any remaining floc and silt. This
stage of purification mimics the natural filtration of water
as it moves through the ground.
The filtered water is then treated with chemical
disinfectants to kill any organisms that remain after the
filtration process. An effective disinfectant is chlorine, but
its use may cause potentially dangerous substances such
as carcinogenic trihalomethanes.
Alternatives to chlorine include ozone oxidation
(Figure 13.2). Unlike chlorine, ozone does not stay in
the water after it leaves the treatment plant, so it offers
country
E
0
4
0
7
1
2
b
Germany 490
USA 1870
Mexico 800
Egypt 920
Namibia 185
China 439
India 588
demand
m /capita
3
domestic
m /capita
3
agriculture
m /capita
3
industrial
m /capita
3
year
67
213
101
55
52
22
29
2
752
662
792
126
338
18
389
828
38
74
6
78
541
1999
1990
1999
1993
1990
1993
1990
OECD data compendium 2002 World Resources Institute
source: source:
Table 13.1. Annual per-capita water
demand in various countries in the world.
Source: 1) OECD data compendium 2002
and 2) World Resources Institute.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 429
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
no protection from bacteria that might be in the storage
tanks and water pipes of the water distribution system.
Water can also be treated with ultraviolet light to kill
microorganisms, but this has the same limitation as
oxidation: it is ineffective outside of the treatment plant.
Figure 13.3 is an aerial view of a water treatment plant
serving a population of about 50,000.
Sometimes calcium carbonate is removed from
drinking water in order to prevent it from accumulating
in drinking water pipes and washing machines.
In arid coastal areas desalinated brackish or saline
water is an important source of water for high-value uses.
430 Water Resources Systems Planning and Management
The cost of desalination is still high, but decreasing
steadily. The two most common methods of desalination
are distillation and reverse osmosis. Distillation requires
more energy, while osmosis systems need frequent
maintenance of the membranes.
2.3. Water Distribution
Water distribution systems include pumping stations,
distribution storage and distribution piping. The
hydraulic performance of each component depends upon
the performance of the others. Of interest to designers are
chlorination
filtered water
chlorine
c
h
l
o
r
i
n
e
unfiltered
water
chemicals
screening rapid mixing filtration
gravel anthracite
coal
flocculation
oxygen ozone
pump
reclaimed
backwash water
backwash
water backwash water
reclamation pond
ozonation
E
0
2
0
8
0
9
p
Figure 13.2. Typical processes
in water treatment plants.
Figure 13.3. A 6-million
gallon per day water
treatment plant at San Luis
Obispo, located about halfway
between Los Angeles and San
Francisco on the central
coast of California.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 430
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 431
both the flows and their pressures. Leakage of drinking
water from the distribution system is a concern in many
old drinking water systems.
The energy at any point within a network of pipes is
often represented in three parts: the pressure head, p/,
the elevation head, Z, and the velocity head, V
2
/2g.
(A more precise representation includes a kinetic energy
correction factor, but that factor is small and can be
ignored.) For open-channel flows, the elevation head is
the distance from some datum to the top of the water
surface. For pressure-pipe flow, the elevation head is the
distance from some datum to the centre of the pipe. The
parameter p is the pressure, for example Newtons per
cubic metre (N/m
3
), is the specific weight (N/m
2
)
of water, Z is the elevation above some base elevation
(m), V is the velocity (m/s), and g is the gravitational
acceleration (9.81 m/s
2
).
Energy can be added to the system such as by a pump,
or lost by, for example, friction. These changes in energy
are referred to as head gains and losses. Balancing the
energy across any two sites i and j in the system requires
that the total heads, including any head gains H
G
and
losses H
L
(m) are equal.
[p/ Z V
2
/2g H
G
]
site i
[p/ Z V
2
/2g H
L
]
site j
(13.1)
The hydraulic grade is the sum of the pressure head and
elevation head (p/ Z). For open-channel flow, the
hydraulic grade is the water surface slope, since the
pressure head at its surface is 0. For a pressure pipe,
the hydraulic head is the height to which a water
column would rise in a piezometer (a tube rising
from the pipe). When plotted in profile along the length
of the conveyance section, this is often referred to as the
hydraulic grade line, or HGL. The hydraulic grade lines
for open channels and pressure pipes are illustrated in
Figures 13.4 and 13.5.
The energy grade is the sum of the hydraulic grade and
the velocity head. This is the height to which a column of
water would rise in a Pitot tube, but also accounts for
fluid velocity. When plotted in profile, as in Figure 13.5,
this is often referred to as the energy grade line, or EGL.
At a lake or reservoir, where the velocity is essentially
zero, the EGL is equal to the HGL.
Specific energy, E, is the sum of the depth of flow and
the velocity head, V
2
/2g. For open-channel flow, the
depth of flow, y, is the elevation head minus the channel
bottom elevation. For a given discharge, the specific
energy is solely a function of channel depth. There may be
more than one depth with the same specific energy. In
one case the flow is subcritical (relatively higher depths,
lower velocities) and in the other case the flow is critical
(relatively lower depths and higher velocities). Whether
or not the flow is above or below the critical depth (the
depth that minimizes the specific energy) will depend in
part on the channel slope.
Friction is the main cause of head loss. There are many
equations that approximate friction loss associated with
fluid flow through a given section of channel or pipe.
These include Mannings or Stricklers equation, which
is commonly used for open-channel flow, and Chezys
or Kutters equation, HazenWilliams equation, and
V
1
2
/ 2g
Z
1
Z
2
head loss
velocity head
water surface
channel bottom
V
2
2
/ 2g
H
L
HGL
EGL
elevation datum
E
0
2
0
8
0
9
q
Figure 13.4. The energy
components along an open
channel.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 431
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
DarcyWeisbach or ColebrookWhite equations, which
are used for pressure-pipe flow. They all define flow veloc-
ity, V (m/s), as an empirical function of a flow resistance
factor, C, the hydraulic radius (cross-sectional area
divided by wetted perimeter), R (m), and the friction or
energy slope, S H
L
/Length.
V kCR
x
S
y
(13.2)
The terms k, x and y of Equation 13.2 are parameters. The
roughness of the flow channel usually determines the
flow resistance or roughness factor, C. The value of C may
also be a function of the channel shape, depth and fluid
velocity. Values of C for different types of pipes are
listed in hydraulics texts or handbooks (e.g. Chin, 2000;
Mays, 2000, 2005).
2.3.1. Open Channel Networks
For open-channel flow, Mannings or Stricklers equation
is commonly used to predict the average velocity, V (m/s),
and the flow, Q (m
3
/s), associated with a given cross-
sectional area, A (m
2
). The velocity depends on the
hydraulic radius R (m) and the slope S of the channel as
well as a friction factor n.
V (R
2/3
S
1/2
)/n (13.3)
Q AV (13.4)
The values of various friction factors n can be found in
tables in hydraulics texts and handbooks.
432 Water Resources Systems Planning and Management
The energy balance between two ends of a channel
segment is defined in Equation 13.5. For open-channel
flow the pressure heads are 0. Thus, for a channel con-
taining water flowing from site i to site j:
[Z V
2
/2g]
site i
[Z V
2
/2g H
L
]
site j
(13.5)
The head loss H
L
is assumed to be primarily due to friction.
The friction loss is computed on the basis of the
average rate of friction loss along the segment and the
length of the segment. This is the difference in the energy
grade line elevations between sites i and j;
H
L
(EGL
1
EGL
2
)
[Z V
2
/2g]
site i
[Z V
2
/2g]
site j
(13.6)
The friction loss per unit distance along the channel is
the average of the friction slopes at the two ends divided
by the channel length. This defines the energy grade
line, EGL.
2.3.2. Pressure Pipe Networks
The HasenWilliams equation is commonly used to
predict the flows or velocities in pressure pipes. Flows
and velocities are again dependent on the slope, S, the
hydraulic radius, R (m), (which equals half the pipe
radius, r) and the cross-sectional area, A (m
2
).
V 0.849 CR
0.63
S
0.54
(13.7)
Q AV r
2
V (13.8)
V
1
2
/ 2g
Z
1
Z
2
V
2
2
/ 2g
H
L
HGL
EGL
elevation datum
E
0
2
0
8
0
9
r
p
1
/ y
p
2
/ y
Figure 13.5. The energy
components along a
pressure pipe.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 432
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 433
The head loss along a length L (m) of pipe of diameter D
(m) containing a flow of Q (m
3
/s) is defined as
H
L
KQ
1.85
(13.9)
where K is the pipe coefficient defined by Equation 13.10.
K [10.66 L]/[C
1.85
D
4.87
] (13.10)
Another pipe flow equation for head loss is the
DarcyWeisbach equation based on a friction factor f:
H
L
fLV
2
/D 2g (13.11)
The friction factor is dependent on the Reynolds number
and the pipe roughness and diameter.
Given these equations, it is possible to compute the
distribution of flows and heads throughout a network of
open channels or pressure pipes. The two conditions are
the continuity of flows at each node, and the continuity of
head losses in loops for each time period t.
At each node i:
Storage
it
Q
in
it
Q
out
it
Storage
i,t1
(13.12)
In each section between nodes i and j:
H
Lit
H
Ljt
H
Lijt
(13.13)
where the head loss between nodes i and j is H
Lijt
.
To compute the flows and head losses at each node
in Figure 13.6 requires two sets of equations, one for
continuity of flows, and the other continuity of head
losses. In this example, the direction of flow in two
links, from A to C, and from B to C, are assumed
unknown and hence each is represented by two non-
negative flow variables.
Let Q
ij
be the flow from site i to site j and H
i
be the
head at site i. Continuity of flow in this network requires:
0.5 Q
DA
Q
DC
(13.14)
0.1 Q
DA
Q
AC
Q
CA
Q
AB
(13.15)
0.25 Q
AB
Q
CB
Q
BC
(13.16)
0.15 Q
DC
Q
AC
Q
CA
Q
BC
Q
CB
(13.17)
Continuity of heads at each node requires:
H
D
H
C
22*(Q
DC
1.85
) (13.18)
H
D
H
A
11*(Q
DA
1.85
) (13.19)
H
A
H
B
22*(Q
AB
1.85
) (13.20)
H
C
H
A
25*((Q
CA
Q
AC
)
1.85
) (13.21)
H
C
H
B
11*((Q
CB
Q
BC
)
1.85
) (13.22)
Solving these Equations 13.14 to 13.22 simultaneously
for the 5-flow and 4-head variables yields the flows Q
ij
from nodes i to nodes j and heads H
i
at nodes i listed in
Table 13.2. Increasing H
D
will increase the other heads
accordingly.
The solution shown in Table 13.2 assumes no eleva-
tion heads, no storage capacity and no minor losses.
Losses are usually expressed as a linear function of the
velocity head, due to hydraulic structures (such as valves,
Q = 0.5
Q = 0.1
Q = 0.15
Q = 0.25
A B
D C
K = 22
K = 22
K = K = 11 11 K = 25
E020809s
Figure 13.6. An example of a pipe network, showing the
values of K for predicting head losses from Equation 13.10.
Q
Q
Q
Q
Q
Q
Q
H
H
H
H
DA
DC
AC
CA
AB
CB
BC
A
B
C
D
=
=
=
=
=
=
=
=
=
=
=
0.29
0.21
0.07
0.00
0.12
0.13
0.00
0.43
0.00
0.26
1.52
E020903k
Table 13.2. Flows and heads of the network shown
in Figure 13.6.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 433
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
restrictions or meters) at each node. This solution
suggests that the pipe section between nodes A and C
may not be economical, at least for these flow conditions.
Other flow conditions may prove otherwise. But even if
they do not, this pipe section increases the reliability of
the system, and reliability is an important consideration
in water supply distribution networks.
2.3.3. Water Quality
Many of the water quality models discussed in Chapter 12
can be used to predict water quality constituent
concentrations in open channels and in pressure pipes. It
is usually assumed that there is complete mixing, for
example at junctions or in short segments of pipe.
Reactions among constituents can occur as water travels
through the system at predicted velocities. Water resident
times (the ages of waters) in the various parts of the net-
work are important variables for water quality prediction,
as constituent decay, transformation and growth
processes take place over time.
Computer models typically use numerical methods
to find the hydraulic flow and head relationships as
well as the resulting water quality concentrations. Most
numerical models assume combinations of plug flow
(advection) along pipe sections and complete mixing
within segments of each pipe section at the end of each
simulation time step. Some models also use Lagrangian
approaches for tracking particles of constituents within a
network. These methods are discussed in more detail in
Chapter 12.
Computer programs (e.g. EPANET) exist that can
perform simulations of the flows, heads and water quality
behaviour within pressurized networks of pipes, pipe
junctions, pumps, valves and storage tanks or reservoirs.
These programs are designed to predict the movement
and fate of water constituents within distribution systems.
They can be used for many different kinds of application
in distribution systems design, hydraulic model calibra-
tion, chlorine residual analysis and consumer exposure
assessment. They can also be used to compare and
evaluate the performance of alternative management
strategies for improving water quality throughout a
system. These can include:
altering the sources within multiple source systems
altering pumping and tank filling/emptying schedules
434 Water Resources Systems Planning and Management
use of satellite treatment, such as re-chlorination at
storage tanks
targeted pipe cleaning and replacement.
Computer models that simulate the hydraulic and water
quality processes in water distribution networks must
be run long enough for the system to reach equilibrium
conditions, i.e. conditions not influenced by initial
boundary assumptions. Equilibrium conditions within
pipes are reached relatively quickly compared to those in
storage tanks.
3. Wastewater
Wastewater issues include its production, its collection
and its treatment prior to disposal.
3.1. Wastewater Production
Wastewater treatment plant influent is usually a mixture of
wastewater from households and industries, urban runoff
and infiltrating groundwater. The characterization of the
influent, both in dry weather situations and during rainy
weather, is of importance for the design and operation of
the treatment facilities. In general, wastewater treatment
plants can handle pure domestic wastewater better than
diluted influent with low concentrations of pollutants. The
discharge of urban runoff to the wastewater treatment
plant dilutes the wastewater, thus affecting the treatment
efficiency. The amount of infiltrating groundwater can also
be significant in areas with old sewage systems.
3.2. Sewer Networks
Sewer flows and their pollutant concentrations vary
throughout a typical day, a typical week, and over the sea-
sons of a year. Flow conditions can range from free surface
to surcharged flow, from steady to unsteady flow, and from
uniform to gradually or rapidly varying non-uniform flow.
Urban drainage ditches normally have uniform cross
sections along their lengths and uniform gradients.
Because the dimensions of the cross sections are typically
one or two orders of magnitude less than the lengths of
the conduit, unsteady free-surface flow can be modelled
using one-dimensional flow equations.
When modelling the hydraulics of flow it is important
to distinguish between the speed of propagation of the
wrm_ch13.qxd 8/31/2005 12:00 PM Page 434
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 435
kinematic wave disturbance and the speed of the bulk
of the water. In general the wave travels faster than the
water particles. Thus if water is injected with a tracer,
the tracer lags behind the wave. The speed of the wave
disturbance depends on the depth, width and velocity
of the flow.
Flood attenuation (or subsidence) is the decrease in
the peak of the wave as it propagates downstream. Gravity
tends to flatten, or spread out, the wave along the
channel. The magnitude of the attenuation of a flood
wave depends on the peak discharge, the curvature of the
wave profile at the peak, and the width of flow. Flows can
be distorted (changed in shape) by the particular channel
characteristics.
Additional features of concern to hydraulic modellers
are the entrance and exit losses to the conduit. Typically,
at each end of the conduit is an access-hole. These are
storage chambers that provide access to the conduits
upstream and downstream. Access-holes induce some
additional head loss.
Access-holes usually cause a major part of the head
losses in sewage systems. An access-hole loss represents a
combination of the expansion and contraction losses. For
pressure flow, the head loss, H
L
, due to contraction can be
written as a function of the downstream velocity, V
D
, and
the upstream and downstream flow cross-sectional areas
A
U
and A
D
:
H
L
K(V
D
2
/2g) [1 (A
D
/A
U
)]
2
(13.23)
The coefficient K varies between 0.5 for sudden
contraction and about 0.1 for a well-designed gradual
contraction.
An important parameter of a given open-channel
conduit is its capacity: the flow that it can take without
surcharging or flooding. Assuming normal depth flow
where the hydraulic gradient is parallel to the bed of
the conduit, each conduit has an upper limit to the flow
that it can accept.
Pressurized flow is much more complex than free-
surface flow. In marked contrast to the propagation
speed of disturbances under free-surface flow conditions,
the propagation of disturbances under pressurized flow
in a 1 m circular conduit 100 m long can be less than
a second. Some conduits can have the stable situation
of free-surface flow upstream and pressurized flow
downstream.
3.3. Wastewater Treatment
The wastewater generated by residences, businesses
and industries in a community consists largely of water.
It often contains less than 10% dissolved and suspended
solid material. Its cloudiness is caused by suspended
particles whose concentrations in untreated sewage range
from 100 to 350 mg/l. One measure of the strength of
the wastewater is its biochemical oxygen demand, or
BOD
5
. BOD
5
is the amount of dissolved oxygen aquatic
microorganisms will require in five days as they
metabolize the organic material in the wastewater.
Untreated sewage typically has a BOD
5
concentration
ranging from 100 mg/l to 300 mg/l.
Pathogens or disease-causing organisms are also pres-
ent in sewage. Coliform bacteria are used as an indicator
of disease-causing organisms. Sewage also contains
nutrients (such as ammonia and phosphorus), minerals
and metals. Ammonia can range from 12 to 50 mg/l and
phosphorus can range from 6 to 20 mg/l in untreated
sewage.
As illustrated in Figures 13.7 and 13.8, wastewater
treatment is a multi-stage process. The goal is to reduce or
remove organic matter, solids, nutrients, disease-causing
organisms and other pollutants from wastewater before it
is released into a body of water or on to the land, or is
reused. The first stage of treatment is called preliminary
treatment.
Preliminary treatment removes solid materials (sticks,
rags, large particles, sand, gravel, toys, money, or
anything people flush down toilets). Devices such as bar
screens and grit chambers are used to filter the wastewater
as it enters a treatment plant, and it then passes on to
what is called primary treatment.
Clarifiers and septic tanks are generally used to
provide primary treatment, which separates suspended
solids and greases from wastewater. The wastewater is
held in a tank for several hours, allowing the particles to
settle to the bottom and the greases to float to the top. The
solids that are drawn off the bottom and skimmed off
the top receive further treatment as sludge. The clarified
wastewater flows on to the next, secondary stage of
wastewater treatment.
This secondary stage typically involves a biological
treatment process designed to remove dissolved organic
matter from wastewater. Sewage microorganisms cultivated
wrm_ch13.qxd 8/31/2005 12:00 PM Page 435
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
and added to the wastewater absorb organic matter from
sewage as their food supply. Three approaches are com-
monly used to accomplish secondary treatment: fixed-film,
suspended-film and lagoon systems.
436 Water Resources Systems Planning and Management
Fixed-film systems grow microorganisms on sub-
strates such as rocks, sand or plastic, over which the
wastewater is poured. As organic matter and nutrients
are absorbed from the wastewater, the film of micro-
organisms grows and thickens. Trickling filters, rotating
biological contactors and sand filters are examples of
fixed-film systems.
Suspended-film systems stir and suspend microorgan-
isms in wastewater. As the microorganisms absorb
organic matter and nutrients from the wastewater, they
grow in size and number. After the microorganisms have
been suspended in the wastewater for several hours, they
are settled out as sludge. Some of the sludge is pumped
back into the incoming wastewater to provide seed
microorganisms. The remainder is sent on to a sludge
treatment process. Activated sludge, extended aeration,
oxidation ditch and sequential batch reactor systems are
all examples of suspended-film systems.
Lagoons, where used, are shallow basins that hold the
wastewater for several months to allow for the natural
degradation of sewage. These systems take advantage of
natural aeration and microorganisms in the wastewater to
renovate sewage.
raw
sewage
sedimentation
tank
aeration
tank
clarifier chlorination
ammonia
stripping
activated
carbon
absorption
grit
chamber
thickener
digester
drying beds
soil conditioner, fertilizer, land fill
trickling
filter
clarifier precipitation denitrification
filter
land fill released
sludge
removal
secondary treatment
primary treatment stream or tertiary treatment
E
0
2
0
8
0
9
t
Figure 13.7. A typical
wastewater treatment plant
showing the sequence of
processes for removing
impurities.
Figure 13.8. Wastewater treatment plant in Soest, the
Netherlands (Waterschap Vallei en Eem).
wrm_ch13.qxd 8/31/2005 12:00 PM Page 436
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 437
Advanced treatment is necessary in some systems to
remove nutrients from wastewater. Chemicals are some-
times added during the treatment process to help remove
phosphorus or nitrogen. Some examples of nutrient
removal systems are coagulant addition for phosphorus
removal and air stripping for ammonia removal.
Final treatment focuses on removal of disease-causing
organisms from wastewater. Treated wastewater can be
disinfected by adding chlorine or by exposing it to suffi-
cient ultraviolet light. High levels of chlorine may be
harmful to aquatic life in receiving streams, so treatment
systems often add a chlorine-neutralizing chemical to the
treated wastewater before stream discharge.
Sludges are generated throughout the sewage
treatment process. This sludge needs to be treated to
reduce odours, remove some of the water and reduce
volume, decompose some of the organic matter and kill
disease-causing organisms. Following sludge treatment,
liquid and cake sludges free of toxic compounds can be
spread on fields, returning organic matter and nutrients to
the soil.
Artificial wetlands and ponds are sometimes used for
effluent polishing. In the wetlands the natural diurnal
variation in the oxygen concentration is restored.
Furthermore, artificial wetlands can reduce the nutrient
content of the effluent by the uptake of
nitrogen and phosphorus by algae or macrophytes. The
organic matter may be harvested from the ponds and
wetlands.
A typical model for the simulation of the treatment
processes in wastewater treatment plants is the Activated
Sludge Model (Gujer et al., 1999; Henze et al., 1999;
Hvitved-Jacobsen et al., 1998). Activated sludge models
predict the production of bacterial biomass and the
subsequent conversion of organic matter and nutrients
into sludge, CO
2
and N
2
gas.
4. Urban Drainage
Urban drainage involves:
rainfall and surface runoff
surface loading and washoff of pollutants
stormwater sewer and pipe flow
sediment transport
structures and special flow characteristics
separation of solids at structures
outfalls.
These components or processes are briefly discussed in
the following sub-sections.
4.1. Rainfall
Rainfall and the need to collect urban stormwater are
the primary reasons for urban drainage systems. Storms
are a major source of flow into the system. Even sanitary
sewage systems that are claimed to be completely
separate from storm drainage sewers are often influenced
by rainfall through illicit connections or even infiltration.
Rainfall varies over time and space. These differences
are normally small when considering short time periods
and small distances, but they increase as time and
distance increase. The ability to account for spatial differ-
ences in rainfall depends on the size of the catchment area
and on the number of functioning rainfall recording
points in the catchment. The use of radar permits more
precision over space and time, rather as if more rain
gauges were used and they were monitored more
frequently. In practice, spatial effects are not measured at
high resolution, and therefore events where significant
spatial variations occur, such as summer thunderstorms,
are usually not very accurately represented.
There are two categories of rainfall records: recorded
(real) events and synthetic (not-real) events. Synthetic
rainfall comes in two forms: as stochastically generated
rainfall data and as design storms. The latter are derived
from analyses of actual rainfall data and are used to
augment or replace those historical (real) data.
Design events are a synthesized set of rainfall profiles
that have been processed to produce storms with specific
return periods; in other words, how often, on average,
one can expect to observe rainfall events of that magni-
tude or greater. Design events are derived to reduce the
number of runs needed to analyse system performance
under design flow conditions.
4.1.1. Time Series Versus Design Storms
Professionals debate whether design rainfall is better
represented by real rainfall or synthetic design events. The
argument in favour of using synthetic storms is that they
wrm_ch13.qxd 8/31/2005 12:00 PM Page 437
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
are easy to use and require only a few events to assess the
system performance. The argument in favour of a time
series of real rainfall is that these data include a wider
range of conditions, and therefore are likely to contain the
conditions that are critical in each catchment.
The two methods are not contradictory. The use of real
rainfall involves some synthesis in choosing which storms
to use in a time series, and in adjusting them for use on a
catchment other than the one where they were measured.
Time series of rainfalls are generally used to look at
aspects such as overflow spill frequencies and volumes.
On the other hand, synthetic design storms can be
generated for a wide range of conditions including the
same conditions as represented by real rainfall. This is
generally considered appropriate for looking at pipe
network performance.
4.1.2. Spatial-Temporal Distributions
Rainfall varies in space as well as in time, and the two
effects are related. Short storms typically come from small
rain cells that have a short life, or that move rapidly
over the catchment. As these cells are small (in the order
of a kilometre in diameter), there is significant spatial
variation in rainfall intensity. Longer storms tend to come
from large rainfall cells associated with large weather
systems. These have less spatial intensity variation.
Rainfall is generally measured at specific sites using
rain gauges. The recorded rainfall amount and intensity
will not be the same at each site. Thus, in order to use
recorded rainfall data we need some way to account for
this spatial and temporal variation. The average rainfall
over the catchment in any period of time can be more
or less than the measured values at one or more gauges.
The runoff from a portion of a catchment exposed to a
high-intensity rainfall will be more than the runoff from
the same amount of rainfall spread evenly over the entire
catchment.
4.1.3. Synthetic Rainfall
A convenient way of using rainfall data is to analyse long
rainfall records to define the statistical characteristics of
the rainfall, and then to use these statistics to produce
synthetic rainstorms of various return periods and
durations.
438 Water Resources Systems Planning and Management
Three parameters are used to describe the statistics of
rainfall depth.
the rainfall intensity or depth of rain in a certain
period
the length of the period over which that intensity
occurs
the frequency with which it is likely to occur, or the
probability of it occurring in any particular year.
In most of the work on urban drainage and river
modelling, the risks of occurrence are expressed not by
probabilities but by the inverse of probability, the return
period. An event that has a probability of 0.2 of being
equalled or exceeded each year has an expected return
period of 1/0.2 or 5 years. An event having a probability
of 0.5 of being equalled or exceeded has an expected
return period of 1/0.5 2 years.
Rainfall data show an intensitydurationfrequency
relationship. The intensity and duration are inversely
related. As the rainfall duration increases, the intensity
reduces. The frequency and intensity are inversely related,
so that as the event becomes less frequent the intensity
increases.
An important part of this durationintensity relation-
ship is the period of time over which the intensity is
averaged. It is not necessarily the length of time for which
it rained from start to finish. In fact any period of rainfall
can be analysed for a large range of durations, and each
duration could be assigned a different return period. The
largest return period might be quoted as the return
period of the storm, but it is only meaningful when
quoted with its duration.
Intensitydurationfrequency relationships or depth
durationfrequency relationships, as shown in Figure 13.9,
are derived by analysis of a long set of rainfall records.
Intensitydurationfrequency data are commonly avail-
able all over the world and therefore it is important to be
aware of how they are derived and ways they can be used
for simulation modelling.
The depth of rainfall is the intensity times its duration
integrated over the total storm duration.
4.1.4. Design Rainfall
Design rainfall events (hyetographs) for use in simulation
models are derived from intensitydurationfrequency
data.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 438
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 439
The rainfall intensity during an event is not uniform in
time, and its variation both in intensity and in the time
when the peak intensity occurs during the storm can be
characterized by the peakedness of the storm and the
skew of the storm (Figure 13.10).
A design storm is a synthetic storm that has an
appropriate peak intensity and storm profile.
4.2. Runoff
Runoff prediction is often the first step in obtaining
stream and river flow data needed for various analyses.
Often lack of detailed site-specific land cover, soil condi-
tions and precipitation data as well as runoff processes
limit the accuracy of runoff predictions.
4.2.1. Runoff Modelling
The runoff from rainfall involves a number of processes
and events, as illustrated in Figure 13.11, and can be
modelled using various methods. Most of these methods
assume an initial loss, a continuing loss, and a remainder
contributing to the system runoff.
Most models assume that the first part of a rainfall
event goes to initial wetting of surfaces and filling of
depression storage. The depth assumed to be lost is
usually related to the surface type and condition.
Rainwater can be intercepted by vegetation or can be
trapped in depressions on the ground surface. It then either
infiltrates into the ground and/or evaporates. Depression
storage can occur on any surface, paved or otherwise.
Initial-loss depths are defined as the minimum
quantity of rainfall causing overland runoff. The initial-loss
depth of rainfall for catchment surfaces can be estimated
as the intercept on the rainfall axis of plots of rainfall
versus runoff (Figure 13.12). The runoff values shown
in Figure 13.12 were obtained for various catchments in
the United Kingdom (Price, 2002).
As rainfall increases, so does depression storage. The
relationship between depression storage and surface slope
S is assumed to be of the form aS
b
, where S is average
slope of the sub-catchment and a and b are parameters
between 0 and 1. The values of a and b depend in part on
the surface type.
Evaporation, another source of initial loss, is generally
considered to be relatively unimportant. For example, in
the case of a heavy summer storm (25-mm rainfall depth)
falling on hot asphalt (temperature say 60 C falling to
2030 C as a result of sensible heat loss), a maximum
evaporation loss of 1 mm is likely to occur.
Continuing losses are often separated into two parts:
evapotranspiration and infiltration. These processes are
usually assumed to continue throughout and beyond the
storm event as long as water is available on the surface of
the ground. Losses due to plant transpiration and general
evaporation are not particularly an issue for single
events, but can be during the inter-event periods where
catchment drying takes place. This is applicable to
models where time-series data are used and generated.
100
90
80
70
60
50
40
30
20
10
0
1600
i
n
t
e
n
s
i
t
y
(
m
m
/
h
r
)
time (minutes)
1400 1200 1000 800 600 400 200 0
25
10
5
2 years
E
0
2
0
8
0
9
u
Figure 13.9. Rainfall intensitydurationfrequency (return
period) curves.
60
50
40
30
20
10
0
350
i
n
t
e
n
s
i
t
y
(
m
m
/
h
r
)
time (minutes)
200 150 100 50 0
E
0
2
0
8
0
9
w
250 300
skewed design storm
symmetric design storm
Figure 13.10. Storm peak skewness profiles.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 439
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
440 Water Resources Systems Planning and Management
Infiltration is usually assumed to account for the
remaining rainfall that does not enter into the drainage
system. The proportion of this loss can range from 100%
for very permeable surfaces to 0% for completely
impermeable surfaces.
Many models try to account for the wetting of the catch-
ment and the increasing runoff that takes place as wetting
increases. The effect of this is shown in Figure 13.13.
It is often impractical to take full account of the
variability in urban topography and surface conditions.
Fortunately for modellers, impervious (paved) surfaces
are often dominant in an urban catchment, and the loss
of rainfall prior to runoff is usually relatively small in
periods with much rainfall.
Runoff-routing is the process of passing rainfall across
the surface to enter the drainage network. This process
results in attenuation and delay. These parameters are
modelled by means of routing techniques that generally
consider catchment area size, ground slope and rainfall
intensity in determining the flow rate into the network.
i
n
t
e
n
s
i
t
y
time
Q
time
infiltration surface depression
excess rainfall
overland flow
discharge / unit length
gutter flow
discharge to inlets
pipe flow
rain
fall
a
b
c
d
e
a
b
c
d
e
-
-
-
-
-
infiltration
depression storage
overland flow
gutter flow
sewer flow
E
0
2
0
8
0
9
x
rain
flow
Figure 13.11. Schematic
representation of urban
rainfallrunoff processes.
6
4
2
0
6
r
a
i
n
f
a
l
l
(
m
m
)
runoff (mm)
0
E
0
2
0
8
0
9
y
2 4
depression storage
Figure 13.12. Estimation of depression storage based on
data from catchments in the UK (Price, 2002).
wrm_ch13.qxd 8/31/2005 12:00 PM Page 440
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 441
The topography and surface channels, and even upstream
parts of the sewage system, are usually lumped together in
this process and are not explicitly described in a model.
The runoff-routing process is often linked to catchment
surface type, and empirical calibration factors are used
accordingly.
A number of models for rainfallrunoff and runoff-
routing are available and are used in various parts of
the world. Overland runoff on catchment surfaces can be
represented by the kinematic wave equation. However,
direct solution of this equation in combination with
the continuity equation has not been a practical
approach when applied to basins with a large number of
contributing sub-catchments. Simpler reservoir-based
models, which are less computationally and data
demanding, represent the physical processes almost as
accurately as the more complex physically based
approaches (Price, 2002). In practice, models applied to
catchments typically assume an average or combined
behaviour of a number of overland flow surfaces, gutters
and feeder pipes. Therefore, the parameters of a physi-
cally based approach (for example the roughness value)
as applied would not relate directly to parameters
representative of individual surfaces and structures.
Many overland flow-routing models are based on a
linear reservoir-routing concept. A single reservoir model
assumes the outflow, Q(t) (m
3
/s), at the catchment outlet
is proportional to the volume of stormwater, S(t) (m
3
),
present on the ground surface of that catchment, includ-
ing the non-explicitly modelled network that contributes
stormwater to that outlet point of the urban drainage
system. To take into account the effects of depression
storage and other initial losses, the first millimetre(s) of
rainfall may not contribute to the runoff.
The basic equation for runoff Q(t) (m
3
/s) at time t is:
Q(t) S(t) K (13.24)
where K is a linear reservoir coefficient. This coefficient
is sometimes a function of the catchment slope, area, length
of longest sewer and rainfall intensity. For a two linear-
reservoir model, two reservoirs are applied in series for
each surface type with an equivalent storage output rela-
tionship, as defined by Equation 13.24, for each reservoir.
The simplest models rely on fixed runoff coefficients
K. They best apply to impervious areas where antecedent
soil moisture conditions are not a factor.
Typical values for runoff coefficients are given in
Table 13.3 (Price, 2002). Use of these coefficients should
be supported either by field observations or by expert
judgement.
4.2.2. The Horton Infiltration Model
The Horton model describes the increasing runoff from
permeable surfaces while a rainfall event occurs by keep-
ing track of decreasing infiltration as the soil moisture
content increases. The runoff from paved surfaces is
assumed to be constant while the runoff from permeable
surfaces is a function of the conceptual wetting and
infiltration processes.
On the basis of infiltrometer studies on small
catchments, Horton defined the infiltration rate, f, either
on pervious surfaces or on semi-pervious surfaces, as
a function of time, t (hours), the initial infiltration
rate, f
o
(mm/hr), the minimum (limiting or critical) infil-
tration rate, f
c
(mm/hr), and an infiltration rate constant,
k (1/hr).
f f
c
(f
o
f
c
)e
kt
(13.25)
r
u
n
o
f
f
r
a
i
n
f
a
l
l
separately
continously
rainfall events
and 2 analysed 1
Q
f
i
max
1
i
max
2
t
t
E020809z
Figure 13.13. Effect of catchment wetness on runoff Q over
time t (Price, 2002).
wrm_ch13.qxd 8/31/2005 12:00 PM Page 441
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The minimum or limiting infiltration rate, f
c
, is commonly
set to the saturated groundwater hydraulic conductivity
for the applicable soil type.
The integration of Equation 13.25 over time defines
the cumulative infiltration F(t):
F(t) f
c
t (f
o
f
c
)(1 e
kt
)/k (13.26)
The Horton equation variant as defined in Equation 13.26
represents the potential infiltration depth, F, as a function
of time, t, assuming the rainfall rate is not limiting, i.e. it
is higher than the potential infiltration rate. Expressed
as a function of time, it is not suited for use in a con-
tinuous simulation model. The infiltration capacity
should be reduced in proportion to the cumulative
infiltration volume, F, rather than in proportion to time.
To do this, Equation 13.26 may be solved iteratively to
find the time it takes to cause ponding, t
p
, as a function of
F. That ponding time t
p
is used in Equation 13.25 to
establish the appropriate infiltration rate for the next time
interval (Bedient and Huber, 1992). This procedure is
used, for example, in the urban stormwater management
model (SWMM) (Huber and Dickinson, 1988).
A flow chart of the calculations performed in a
simulation program in which the rainfall can vary might
be as shown in Figure 13.14.
Various values for Hortons infiltration model are
available in the published literature. Values of f
o
and f
c
, as
determined by infiltrometer studies, Table 13.4, are
highly variable, even by an order of magnitude on
seemingly similar soil types. Furthermore, the direct
transfer of values as measured on rural catchments to
urban catchments is not advised due to the compaction
442 Water Resources Systems Planning and Management
and vegetation differences associated with the latter
surfaces.
4.2.3. The US Soil Conservation Method (SCS)
Model
The SCS model (USDA, 1972) is widely used, especially
in the United States, France, Germany, Australia and parts
of Africa, for predicting runoff from rural catchments. It
has also been used for the permeable component in a
semi-urban environment. This runoff model allows for
variation in runoff depending on catchment wetness. The
model relies on what are called curve numbers, CN.
The basis of the method is the continuity equation.
The total depth (mm) of rainfall, R, either evaporates or is
otherwise lost, I
a
, infiltrates and is retained in the soil, F,
or runs off the land surface, Q. Thus,
R I
a
F Q (13.27)
The relationship between the depths (mm) of rainfall, R,
runoff, Q, the actual retention, F, and the maximum
potential retention storage, S (not including I
a
), is
assumed to be
F/S Q/(R I
a
) (13.28)
when R I
a
. These equations combine to give the SCS
model
Q (R I
a
)
2
/(R I
a
S) (13.29)
This model can be modified for use in continuous
simulation models.
Numerical representation of the derivative of Equation
13.29 can be written as Equation 13.30 for predicting
paved
paved
paved
paved
permeable
permeable
permeable
high quality paved roads with gullies <100m apart
high quality paved roads with gullies >100m apart
medium quality paved roads
poor quality paved roads
high to medium density housing
low density housing or industrial areas
open areas
1.00
0.90
0.85
0.80
0.55 - 0.45
0.35
0.00 - 0.25
surface
type description
coefficient
K
E
0
2
0
9
0
3
m
Table 13.3. Typical values of the runoff
fraction (coefficient K).
wrm_ch13.qxd 8/31/2005 12:00 PM Page 442
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 443
the runoff, q (mm/t), over a time interval t given the
rainfall r (mm/t), in that time interval.
q r(R I
a
)(R I
a
2S)/(R I
a
S)
2
(13.30)
This equation is used incrementally, enabling the rainfall
and runoff coefficients, r and q, to change during the event.
The two parameters S and I
a
are assumed to be linearly
related by
I
a
kS (13.31)
where 0 k 0.2.
The original SCS approach recommended k 0.2.
However other studies suggest k values between 0.05 and
0.1 may be more appropriate.
no ponding at
start
no ponding
calculate
F and f
ponding occurs
depression
storage / runoff
if f <i
ponding occurs
f = g(F)
if f <i
F known
start
yes no
yes no
end
,
E
0
2
0
8
2
0
a
Figure 13.14. Flow chart of
Horton model infiltration
algorithm used in each time
step of a simulation model.
A
B
C
D
250
200
125
76
25.4
12.7
6.3
2.5
2
2
2
2
SCS
soil group
f
(mm/hr)
f
(mm/hr)
k
( / hr) 1
o c
E
0
2
0
9
0
3
n
Table 13.4. Values for Hortons infiltration model for different
soil groups as defined by the Soil Conservation Service.
Note: see Table 13.6 for SCS soil groups.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 443
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The storage variable, S, is itself related to an index
known as the runoff curve number, CN, representing
the combined influence of soil type, land management
practices, vegetation cover, urban development and
antecedent moisture conditions on hydrological response.
CN values vary between 0 and 100, 0 representing no
runoff and 100 representing 100% runoff.
The storage parameter S is related to the curve number
CN by
S (25400/CN) 254 (13.32)
Curve number values depend on antecedent moisture
condition classes (AMC) and hydrological soil group. The
antecedent moisture conditions are divided into three
classes, as defined in Table 13.5.
The four hydrological soil groups are defined in
Table 13.6.
The CN value can either be defined globally for
the catchment model or can be associated with specific
surface types. CN values for different conditions are
available from various sources. Table 13.7 lists some of
those relevant to urban areas and antecedent moisture
condition class AMC II.
444 Water Resources Systems Planning and Management
Figure 13.15 identifies the CN values for antecedent
moisture condition classes (AMC) I and III based on class
II values.
4.2.4. Other RainfallRunoff Models
The rainfallrunoff element of the popular SWMM model
generally assumes 100% runoff from impermeable
surfaces and uses the Horton or the GreenAmpt model
(low runoff potential)
high infiltration rates even when thoroughly wetted.
chiefly deep, well to excessively drained sands or gravels.
high rate of water transmission.
moderate infiltration rates when thoroughly wetted.
chiefly moderately deep to deep, moderately-well to well drained soils with
moderately-fine to moderately-coarse textures.
moderate rate of water transmission.
slow infiltration rates when thoroughly wetted.
chiefly solids with layer that impedes downward movement of water, or soils
with a moderately-fine to fine texture.sslow rate of water transmission.
(high runoff potential)
very slow infiltration and transmission rates when thoroughly wetted.
chiefly:
clay soils with a high swelling potential
soils with a permanent high water table
soils with a clay pan or clay layer at or near the surface
shallow soils over nearly impervious material
definition soil type
A
B
C
D
E
0
2
0
9
0
3
p
Table 13.6. SCS hydrological soil
groups used in Tables 13.4 and
13.7.
I
II
III
< 12.5
AMC
E
0
2
0
9
0
3
o
total 5-day
antecedent rainfall (mm)
dormant growing
12.5 - 28.0
> 28.0
< 35.5
35.5 - 53.5
> 53.5
season
Table 13.5. Antecedent moisture condition classes (AMC) for
determining curve numbers CN.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 444
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 445
for permeable runoff. The GreenAmpt model is similar
to the Horton model in that it has a conceptual infiltration
rate that varies with time. It is therefore applicable to
pervious or semi-pervious catchments (Huber and
Dickinson, 1988; Roesner et.al., 1988).
4.3. Surface Pollutant Loading and Washoff
The modelling of surface pollutant loading and washoff
into sewage systems is very imprecise. Pollutants that
build up on the surface of an urban area originate
from wind-blown dust, debris that is both natural and
human-made, including vehicle emissions. When rainfall
takes place, some of this material is washed into the
stormwater sewers or gullies as dissolved pollutants and
fine solids. During buildup time, many of the pollutants
degrade.
Deposition of this material is not homogeneous
but rather is a function of climate, geography, land use
open spaces, parks, cemeteries
open spaces, parks, cemeteries
industrial districts
residential:
paved parking lots, roofs, etc.
75 %grass
75 %grass
85 %
72 %
65 %
38 %
30 %
25 %
20 %
land
cover
class
land cover /
landuse/ treatment
stormflow
potential
hydrological soil group
A B C D
39
49
89
81
77
61
57
54
51
98
76
72
74
61
69
92
88
85
75
72
70
68
98
85
82
84
74
79
94
91
90
83
81
80
78
98
89
87
90
80
84
95
93
92
87
86
85
84
98
91
89
92
commercial / business area %area
impervious
lot size 500 m
1000 m
1350 m
2000 m
4000 m
tarred, with storm sewers, curbs
gravel
dirt
dirt-hard surface
streets/ roads:
u
r
b
a
n
&
s
u
b
-
u
r
b
a
n
l
a
n
d
u
s
e
E
0
2
0
9
0
3
q
A B C D
2
2
2
2
2
Table 13.7. Initial CN values for
AMC II with various urban
land use, cover quality and
hydrological soil groups (defined
in Table 13.6).
100
90
80
70
60
50
40
30
20
10
0
r
u
n
o
f
f
c
u
r
v
e
n
u
m
b
e
r
E
0
2
0
8
2
0
b
AMC II
100 60 0 10 20 30 40 50 70 80 90
AMC III AMC I
Figure 13.15. Runoff curve numbers for AMC classes I and III
based on curve numbers for AMC II.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 445
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
and human activity. The mechanism of washoff is
obviously a function of location, rainfall intensity, slope,
flow rate, vehicle disturbance and so on. None of these
factors is explicitly modelled in most washoff and sewer
flow models.
Measurements made of pollutant accumulation
and washoff have been the basis of empirical equations
representing both loading and washoff processes. In
practice the level of information available and the
complexity of the processes being represented make
the models of pollutant loading and washoff a tool whose
outputs must be viewed for what they are merely guesses.
4.3.1. Surface Loading
Pollutant loadings and accumulation on the surface of
an urban catchment occur during dry periods between
rainstorms.
A common hypothesis for pollutant accumulation
during dry periods is that the mass loading rate, m
P
(kg/ha/day), of pollutant P is constant. This assumed
constant loading rate on the surface of the ground can
vary over space and is related to the land use of that
catchment. In reality, these loadings on the land surface
will not be the same, either over space or over time.
Hence, to be more statistically precise, a time series of
loadings may be created from one or more probability
distributions of observed loadings. ( Just how this may be
done is discussed in Chapter 7.) Different probability
distributions may apply when, for example, weekend
loadings differ from workday loadings. However, given all
the other uncertain assumptions in any urban loading and
washoff model, the effort may not be justified.
As masses of pollutants accumulate over a dry period
they may degrade as well. The time rate of degradation of
a pollutant P is commonly assumed to be proportional
to its total accumulated mass M
P
(kg/ha). Assuming a
proportionality constant (decay rate constant) of k
P
(1/day), the rate of change in the accumulated mass
M
P
over time t is
dM
P
/dt m
P
k
P
M
P
(13.33)
As the number of days during a dry period becomes very
large, the limiting accumulation of a mass M
P
of pollutant
P is m
P
/k
P
. If there is no decay, then of course k
P
is 0 and
the limiting accumulation is infinite.
446 Water Resources Systems Planning and Management
Integrating Equation 13.33 over the duration t (days)
of a dry period yields the mass, M
P
(t) (kg/ha), of each
pollutant available for washoff at the beginning of a
rainstorm.
M
P
(t) M
P
(0)e
kpt
[m
P
(1 e
kpt
)/k
P
] (13.34)
where M
P
(0) is the initial mass of pollutant P on the
catchment surface at the beginning of the dry period (that
is, at the end of the previous rainstorm).
Sediments (which become suspended solids in the
runoff) are among the pollutants that accumulate on the
surface of urban catchments. They are important in
themselves, but also because some of the other pollutants
that accumulate become attached to them. Sediments
are typically defined by their medium diameter size value
(d
50
). Normally a minimum of two sediment fractions
are modelled: one coarse, high-density material (grit) and
one fine (organics).
The sediments of each diameter size class are
commonly assumed to have a fixed amount of pollutants
attached to them. The fraction of each attached pollutant,
sometimes referred to as the potency factor of the
pollutant, is expressed as kg of pollutant per kg of
sediment. Potency factors are one method for defining
pollutant inputs into the system.
4.3.2. Surface Washoff
Pollutants in the washoff may be dissolved in the water,
or be attached to the sediments. Many models of the
transport of dissolved and particulate pollutants through
a sewerage system assume each pollutant is conservative
(that is, it does not degrade over time). For practical
purposes this is a reasonable assumption when the time of
flow in the sewers is relative short. Otherwise it may not
be a good assumption, but at least it is a conservative one.
Pollutants can enter the sewage system from a number
of sources. A major source is the washoff of pollutants
from the catchment surface during a rainfall event. Their
removal is caused by the impact of rain and by erosion
from runoff flowing across the surface. Figure 13.16
shows schematically some sources of pollution in the
washoff model.
The rate of pollutant washoff depends on an erosion
coefficient,
P
, and the quantity, M
P
, of available
pollutant, P. As the storm event proceeds and pollutants
wrm_ch13.qxd 8/31/2005 12:00 PM Page 446
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 447
are removed from the catchment, the quantities of
available pollutants decrease, hence the rate of pollutant
washoff decreases even with the same runoff.
When runoff occurs, a fraction of the accumulated
load may be contained in it. This fraction will depend
on the extent of runoff. If a part of the surface loading
of a pollutant is attached to sediments, its runoff will
depend on the amount of sediment runoff, which in
turn is dependent on the amount of surface water
runoff.
The fraction of total surface loading mass contained
in the runoff will depend on the runoff intensity. The
following approximate relation may apply for the fraction,
f
t
R
, of pollutant P in the runoff R
t
in period t:
f
t
P

P
R
t
/(1
P
R
t
) (13.35)
The greater the runoff R
t
the greater will be the fraction f
t
P
of the total remaining pollutant loading in that runoff.
The values of the parameters
P
are indicators of the
effectiveness of the runoff in picking up and transporting
the particular pollutant mass. Their values are dependent
on the type of pollutant P and on the land cover and
topography of particular basin or drainage area. They can
be determined on the basis of measured pollutant mass
surface loadings and on the mass of pollutants contained
in the rainfall and sediment runoff, preferably at the
basin of interest. Since such data are difficult, or at least
expensive, to obtain, they are usually based on experi-
ments in laboratories.
A mass balance of pollutant loadings can define the total
accumulated load, M
Pt1
, at the end of each simulation
time period t or equivalently at the beginning of each time
period t1. Assuming a daily simulation time step,
M
Pt1
(1 f
t
P
) M
Pt
e
kP
m
P
(13.36)
Of interest, of course, is the total pollutant mass in the
runoff. For each pollutant type P in each period t, these
will be f
t
P
(M
Pt
) for the dissolved part. The total mass of
pollutant P in the runoff must also include those attached
fractions (potency factors), if any, of each sediment size
class being modelled as well.
As the sediments are routed through the system, those
from different sources are mixed together. The concentra-
tions of associated pollutants therefore change during
the simulation as different proportions of sediment
from different sources are mixed together. The results are
given as concentrations of sediment, concentrations
of dissolved pollutants and concentrations of pollutants
associated with each sediment fraction.
4.3.3. Stormwater Sewer and Pipe Flow
Flows in pipes and sewers have been analysed exten-
sively and their representation in models is generally
accurately defined. The hydraulic characteristics of
sewage are essentially the same as clean water. Time-
dependent effects are, in part, a function of the change
in storage in access-holes. Difficulties in obtaining
convergence occur at pipes with steep to flat transi-
tions, dry pipes and the like, and additional features
and checks are therefore needed to achieve satisfactory
model results.
E
0
2
0
8
2
0
c
road
roof
foul
surface
r
u
n
o
f
f
Figure 13.16. Some sources of
pollutants in the washoff to
sewage systems from land
surfaces.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 447
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
4.3.4. Sediment Transport
Pollutant transport modelling of both sediment and dis-
solved fractions involves defining the processes of erosion
and deposition, and advection and possibly dispersion.
One-dimensional models by their very nature cannot
predict the sediment gradient in the water column. In
addition, the concept of the sewer being a bio-reactor is
not included in most simulation models. Most models
assume pollutants are conservative while in residence in
the drainage system before being discharged into a water
body. All the processes that take place in transit are
generally either ignored or approximated using a range
of assumptions.
4.3.5. Structures and Special Flow Characteristics
Access-holes, valves, pipes, pumping stations, overflow
weirs and other elements that affect the flows and head
losses in sewers can be explicitly included in deterministic
simulation models. The impact of some of these structures
can only be predicted using two- or three-dimensional
models. However, the ever-increasing power of computers
is making higher-dimensional fluid dynamic analyses
increasingly available to practising engineers. The greatest
limitation may be more related to data and calibration than
to computer models and costs.
4.4. Water Quality Impacts
The quality of water in urban drainage and sewer systems
can impact the flow conditions in those systems as well as
the quality of the water into which these wastewater and
drainage flows are discharged.
4.4.1. Slime
Slime can build up on the perimeters of sewers that contain
domestic sewage. The buildup of slime may have a signifi-
cant effect on roughness. In a combined system the effect
will be less, as the maximum daily flow of domestic sewage
will not usually be a significant part of pipe capacity.
The extent to which the roughness is increased by
sliming depends on the relation between the sewage
discharge and the pipe capacity. Sliming will occur over
the whole of the perimeter below the water level that
corresponds to the maximum daily flow. The slime
448 Water Resources Systems Planning and Management
growth will be heaviest in the region of the maximum
water level. Over the lower part of the perimeter, the
surface will still be slimed, but to a lesser extent than at
the waterline. Above the maximum waterline the sewer
surface will tend to be fairly free of slime.
4.4.2. Sediment
When sediment is present in the sewer, the roughness
increases quite significantly. It is difficult to relate the
roughness to the nature and time-history of the sediment
deposits. Most stormwater sewers contain some sediment
deposits, even if only temporarily. The only data available
suggest that the increase in head loss can range from
30 to 300 mm, depending on the configuration of the
deposit and on the flow conditions. The higher roughness
value is more appropriate when the sewer is part full
and when considerable energy is lost as a result of the
generation of surface disturbances. In practice lower
roughness values are used for flow states representing
extreme events when sewers are operating in surcharge.
4.4.3. Pollution Impact on the Environment
The effects of combined sewer overflows (CSOs) or dis-
charges are particularly difficult to quantify and regulate
because of their intermittent and varied nature. Their
immediate impact can only be measured during a spill
event, and their chronic effects are often difficult to isolate
from other pollution inputs. Yet they are among the major
causes of poor river water quality. Standards and per-
formance criteria specifically for intermittent discharges
are therefore needed to reduce the pollutants in CSOs.
Drainage discharges that affect water quality can be
divided into four groups:
Those that contain oxygen-demanding substances.
These can be either organic, such as fecal matter, or inor-
ganic. (Heated discharges, such as cooling waters, reduce
the saturated concentration of dissolved oxygen.)
Those which contain substances that physically hinder
re-oxygenation at the water surface, such as oils.
Discharges containing toxic compounds, including
ammonia, pesticides and some industrial effluents.
Discharges that are high in suspended solids and thus
inhibit biological activity by excluding light from the
water or by blanketing the bed.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 448
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 449
Problems arise when pollutant loads exceed the
self-purification capacity of the receiving water, harming
aquatic life and restricting the use of the water for
consumption and for many industrial and recreational
purposes. The assimilative capacity for many toxic
substances is very low. Water polluted by drainage
discharges can create nuisances such as unpleasant
odours. It can also be a direct hazard to health, particu-
larly in tropical regions where water-borne diseases such
as cholera and typhoid are prevalent.
The aim of good drainage design, with respect to
pollution, is to balance the effects of continuous and
intermittent discharges against the assimilation capacity
of the water, so as to achieve in a cost-effective way the
desired quality of the receiving water.
Figure 13.17 shows the effect of a discharge that
contains suspended solids and organic matter. The
important indicators showing the effect of the discharge
are the dissolved oxygen in diagram a and the clean
water fauna shown in diagram d. The closeness with
which the clean water fauna follow the dissolved oxygen
reflects the reliance of a diverse fauna population on dis-
solved oxygen. These relationships are used by biologists
to argue for greater emphasis on biological indicators of
pollution, as these respond to intermittent discharges
better than chemical tests which, if not continuous, may
miss the pollution incident. There are a number of
biological indexes in use in most countries in Europe.
In Figure 13.17 the BOD in diagram a rises or is
constant after release despite some of the pollutant
being digested and depleting the dissolved oxygen. This
is because there is a time lag of up to several days while
the bacteria, which digest the pollutant, multiply. The
suspended solids (SS) settle relatively quickly and they
point of discharge
a
b
c
d
lag recovery time
BOD
DO
SS
NO
PO
NH
4
4
3
sewage fungus
bacteria
tubificids chironomids diversity of fauna
E
0
2
0
8
2
0
d
Figure 13.17. Pollution impact
along a waterway downstream
from its discharge.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 449
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
can then be a source of pollutants if the bed is disturbed
by high flows. This can create a subsequent pollution
incident, especially if the suspended solids contain
quantities of toxic heavy metals.
Diagram b of Figure 13.17 shows ammonium ions
(NH
4

) that are discharged as part of the dissolved

pollutants being oxidized to nitrates (NO
3

). The rise in
the ammonium concentration downstream of the dis-
charge is relevant because of the very low tolerance many
aquatic organisms, particularly fish, have to the chemical.
The ammonium concentration rises if the conditions are
anaerobic, and will decline once aerobic conditions return
and the ammonium ions are oxidized to nitrates.
Diagrams c and d show the effect of combined sewer
overflows on flora and fauna. The increased quantities of
phosphate and nitrate nutrients that they consume can lead
to eutrophication. The fauna show perhaps the clearest
pattern of response. The predictability of this response
has led to the development of the many biological indices
of pollution. The rapid succession of organisms illustrates
the pattern of dominance of only a few species in polluted
conditions. For example, tubificid worms can exist in
near-anaerobic conditions and, having few competitors,
they can multiply prolifically. As the oxygen levels
increase these organisms are succeeded by chironomids
(midge larvae) and so on until in clean well-oxygenated
water, there is a wide diversity of species all competing
with each other.
In most circumstances the concentration of dissolved
oxygen (DO) is the best indicator of the health of a water
source. A clean water source with little or no content of
biodegradable pollutants will have an oxygen concentra-
tion of about 910 mg/l when in equilibrium with air in
a temperate environment. This maximum saturation
concentration is temperature dependent. The hotter the
water, the lower the DO saturation concentration.
All higher forms of life in a river require oxygen. In
the absence of toxic impurities there is a close correlation
between DO and biodiversity. For example most game-fish
die when the DO concentration falls below about 4 mg/l.
Perhaps of more pragmatic significance is the fact that
oxygen is needed in the many natural treatment processes
performed by microorganisms that live in natural water bod-
ies. The quantity of oxygen required by these organisms to
break down a given quantity of organic waste is, as previ-
ously discussed, the biochemical oxygen demand (BOD).
450 Water Resources Systems Planning and Management
It is expressed as mg of dissolved oxygen required by
organisms to digest and stabilize the waste in a litre of
water (mg/l). These organisms take time to fully digest
and stabilize the waste. The rate at which they do so
depends on the temperature of the water at the start, and
the quantity of these organisms available in it.
Since the BOD test measures only the biodegradable
material, it may not give an accurate assessment of the
total quantity of oxidizable material in a sample in all
circumstances, e.g. in the presence of substances toxic to
the oxidizing bacteria. In addition, measuring the BOD of
a sample of water typically takes a minimum of five days.
The chemical oxygen demand (COD) test is a quicker
method, and measures the total oxygen demand. It is a
measure of the total amount of oxygen required to stabi-
lize all the waste. While the value of COD is never less
than the value of BOD, it is the faster reacting BOD that
impacts water quality. And this is what most people care
about. However, determining a relationship between BOD
and COD at any site can provide guideline values for BOD
based on COD values.
Tables 13.8 and 13.9 provide some general ranges of
pollutant concentrations in CSOs from urban catch-
ments.
Water quality models for urban drainage are similar to,
or simpler versions of, water quality models for other
water bodies (as discussed in Chapter 12). Often a simple
mixing and dilution model will be sufficient to predict
the concentration of pollutants at any point in a sewage
system. For example, such models may be sufficient for
some toxic substances that are not broken down. Once the
flows in the CSO enter the receiving water body, models
discussed in Chapter 12 can be used to estimate their fate
as they travel with the water in the receiving water body.
A factor that makes predicting the impacts of overflow
discharges particularly difficult is the non-continuous
nature of the discharges and their pollutant concentra-
tions. In the first sanitary or foul flush, the fine sediments
deposited in the pipes during dry periods are swept up
and washed out of the system. Most existing models,
termed constant concentration models, do not account
for this phenomenon. Since many of the most significant
pollution events occur when the river has low flows, and
hence low dilution factors, the quantity of spill in the first
flush may be very important in the overall pollution
impact.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 450
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 451
4.4.4. Bacteriological and Pathogenic Factors
The modelling of pathogenic microorganisms is particu-
larly difficult since there are a very large number of
pathogenic organisms, each usually with a unique testing
procedure, many of which are expensive. Also many
pathogens may present a significant risk to human
health in very small numbers. Incubation periods of over
twenty-four hours are not uncommon and there are
as yet no automatic real-time monitoring techniques in
commercial use.
The detection of pathogens relies heavily on indicator
organisms that are present in feces in far higher numbers
than the pathogenic organisms. Escherichia Coliform
(E. coli) bacteria is the most common fecal indicator, and
is used throughout the world to test water samples for
fecal contamination.
Because of the problems in measuring micro-biological
parameters involved in CSOs, most sophisticated
methods of determining the quality of sewer water restrict
themselves to more easily measured determinants such
as BOD, COD, suspended solids, ammonia, nitrates and
similar constituents.
4.4.5. Oil and Toxic Contaminants
Oils are typically discharged into sewers by people and
industries, or are picked up in the runoff from roads
and road accidents. Since oil floats on water surfaces and
disperses rapidly into a thin layer, a small quantity of oil
discharged into a water body can prevent reoxygenation
at the surface and thus suffocate the organisms living
there. The dispersal rate changes with oil viscosity, and
constituent raw treated overflow
suspended
solids mg/l
BOD mg/l
COD
mg/l
ammonia
mg/l
E
0
2
0
9
0
3
r
300
367
470
39
5
10
15
7
200
300
350
30
Table 13.8. Typical quality of domestic sewage.
constituent highway
runoff
residential
area
commercial
area
industrial
area
suspended
solids mg/l
BOD mg/l
COD
mg/l
ammonia
mg/l
E
0
2
0
9
0
3
s
lead mg/l
28 - 1178
12 - 32
128 - 171
0.02 - 2.1
0.15 - 2.9
112 - 1104
7 - 56
37 - 120
0.3 - 3.3
0.09 - 0.44
230 - 1894
5 - 17
74 - 160
0.03 - 5.1
0.1 - 0.4
34 - 374
8 - 12
40 - 70
0.2 - 1.2
0.6 - 1.2
Table 13.9. Pollutant
concentrations (mg/l) in urban
runoff.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 451
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
the length of time the oil is a problem will partly depend
on the surface area of the receiving water as well.
There are three main sources of toxic contaminants
that may be discharged from CSOs:
Industrial effluents. These could be anything from
heavy metals to herbicides.
Surface washoff contaminants. These may be contami-
nants washed off the surface in heavy rainstorms, and
in agricultural and suburban residential areas will
probably include pesticides and herbicides. In many
cases these contaminants make a larger contribution to
the pollutant load than the domestic sanitary flow.
Substances produced naturally in the sewer. Various poi-
sonous gases are produced in sewers. From the point
of view of water quality, ammonia is almost
certainly the most important, though nitrogen
sulphide can also be significant.
4.4.6. Suspended Solids
Discharges high in suspended solids pose a number of
problems. They almost invariably exert an oxygen
demand. If they remain in suspension they can prevent
light from penetrating the water and thus inhibit photo-
synthesis. If deposited they become a reservoir of oxygen-
demanding particles that can form an anaerobic layer on
the bed, decreasing biodiversity. They can also degrade the
bed for fish (such as salmon) spawning. If these suspended
solids contain toxic substances, such as heavy metals, the
problems can be more severe and complex.
5. Urban Water System Modelling
Optimization and simulation models are becoming
increasingly available and are used to analyse a variety of
design and operation problems involving urban water
systems. Many are incorporated within graphicuser
interfaces that facilitate the use of the models and the
understanding and further analysis of their results.
5.1. Model Selection
A wide range of models is available for the simulation of
hydrodynamics and water quality in urban systems. The
selection of a particular model and the setup of a model
452 Water Resources Systems Planning and Management
schematization depends on the research question at hand,
the behaviour of the system, the available time and
budget, and future use of the model.
The research question and the behaviour of the water
system determine the level of detail of the model schema-
tization. The time scale of the dominating processes and
the spatial distribution of the problem are key elements in
the selection of a model, as is illustrated in Figure 13.18
and Table 13.10.
Figure 13.18 shows the time scales of the driving
forces and their impact in urban water systems. It may be
wise to consider the processes with largely different time
scales separately, rather than joining them together in one
model. For instance, the water quality of urban surface
waters is affected by combined sewer overflows and by
many diffusive sources of pollution. A combined sewer
overflow lasts several hours and the impact of the
discharge on the oxygen concentration in the surface
water lasts for a couple of days. The accumulation of
heavy metals and organic micro-pollutants in the sedi-
ment takes many years and the influx of the diffusive
sources of pollution is more or less constant in time. The
impact of combined sewer overflows on the oxygen con-
centration can be studied with a detailed, deterministic
simulation model for the hydrodynamics and the water
quality processes in the surface water system. A typical
time step in such a model is minutes; a typical length
segment is within the range from 10 to 100 metres. The
accumulation of pollutants in the sediment can be mod-
elled by means of a simple mass balance.
Another example is shown in Table 13.10. In this
example the wastewater collection and treatment system
in an urban area is modelled in three different ways. In
the first approach, only the river is modelled. The
discharge of effluent from a wastewater treatment plant
is taken into account as a boundary condition. This is a
useful approach for studying the impact of the discharge
of effluent on water quality.
In the second approach, a detailed water and mass
balance is made for an urban area. The main routes of
water and pollution are considered. Generic measures,
such as the disconnection of impervious areas from the
sewage system, can be evaluated with this type of model
schematization. In the third, most detailed, approach, a
model schematization is made for the entire sewage
system and, eventually, the wastewater treatment plant.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 452
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 453
This type of model is needed when defining specific
measures concerning storage in the sewage system,
enhanced treatment capacity at the wastewater treatment
plant and the design of stormwater retention ponds.
5.2. Optimization
The use of the storage, transport and treatment capacity of
existing urban infrastructure can be optimized in many
cases. Optimization of urban water systems aims at
finding the technical, environmental and financial best
solution, considering and balancing measures in the
sewage system, the wastewater treatment plant and
the surface water system at the same time. For instance,
the optimum use of the storage capacity in a sewage
system by means of real-time control of the pumps may
eliminate the need for a more expensive increase in treat-
ment capacity at the wastewater treatment plant. Storage
storm
drinking water
(peak)
rainy period
drinking water
demand
evaporation
precipitation
surplus
diffuse sources
of pollution
time scale
d
r
i
v
i
n
g
f
o
r
c
e
minutes hours days weeks seasons years
sewer
overflow
dimensions of WWTP,
drinking water net
discharge
drinking
water volume
water
supply
ground water
volume
sediment
quality
E
0
4
0
7
1
6
p
Figure 13.18. Time scales of
driving forces and impacts in
urban water systems.
E
0
4
0
7
1
6
q
type of schematisation implementation in code examples
measured data of effluent
quantity and quality
boundary conditions end-of-pipe measures
at the treatment plants
water balance and mass
balance
rainfall runoff module,
emissions module
analysis of main routes of
rainfall discharge
detailed
model schematisation
sewer flow module,
water quality module
reduction of emissions via
combined sewer overflows
Table 13.10. Three methods of
making a model schematization
for an urban water system.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 453
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
of runoff from streets and roofs in surface water may be
better for river water quality than transporting the runoff
to the wastewater treatment plant and subsequent
discharging it as effluent. Table 13.11 shows a matrix
with the key variables of storage, transport capacity and
treatment capacity in the sewage system, the wastewater
treatment plant and the surface water.
Methods for finding optimal solutions are becoming
increasingly effective in the design and planning of urban
infrastructure. Yet they are challenged by the complexity
and non-linearity of water distribution networks,
especially urban ones.
Numerous calibration procedures for water distribu-
tion system models have been developed since the 1970s.
Trial and error approaches (Rahal et al., 1980; Walski,
1983) were replaced with explicit type models (Boulos
and Wood, 1990; Ormsbee and Wood, 1986). More
recently, calibration problems have been formulated and
solved as optimization problems. Most of the approaches
used so far are either local or global search methods. Local
search gradient methods have been used by Shamir
(1974), Lansey and Basnet (1991), Datta and Sridharan
(1994), Reddy et al. (1996), Pudar and Liggett (1992),
and Liggett and Chen (1994) to solve various steady-state
and transient model calibration problems (Datta and
Sridharan, 1994; Savic and Walters, 1995; Greco and Del
Guidice, 1999; Vitkovsky et al., 2000).
Evolutionary search algorithms, discussed in Chapter 6,
are now commonly used for the design and calibration of
various highly non-linear hydraulic models of urban sys-
tems. They are particularly suited for search in large and
complex decision spaces, e.g. in water treatment, storage
and distribution networks. They do not need complex
454 Water Resources Systems Planning and Management
mathematical matrix inversion methods and they permit
easy incorporation of additional calibration parameters
and constraints into the optimization process (Savic and
Walters, 1995; Vitkovsky and Simpson, 1997; Tucciarelli
et al., 1999; Vitkovsky et al., 2000).
In addition to calibration, these evolutionary search
methods have been used extensively to find least-cost
designs of water distribution systems (Simpson et al.,
1994; Dandy et al., 1996; Savic and Walters, 1997).
Other applications include the development of optimal
replacement strategies for water mains (Dandy and
Engelhardt, 2001), finding the least expensive locations of
water quality monitoring stations (Al-Zahrani and Moied,
2001), minimizing the cost of operating water distribu-
tion systems (Simpson et al., 1999), and identifying the
least-cost development sequence of new water sources
(Dandy and Connarty, 1995).
These search methods are also finding a role in devel-
oping master or capital improvement plans for water
authorities (Murphy et al., 1996; Savic et al., 2000). In
this role they have shown their ability to identify low-cost
solutions for highly complex water distribution systems
subject to a number of loading conditions and a large
number of constraints. Constraints on the system include
maximum and minimum pressures, maximum velocities
in pipes, tank refill conditions and maximum and
minimum tank levels.
As part of any planning process, water authorities need
to schedule the capital improvements to their system over
a specified planning period. These capital improvements
could include water treatment plant upgrades or new
water sources as well as new, duplicate or replacement
pipes, tanks, pumps and valves. This scheduling process
E
0
4
0
7
1
6
r
surface water WWTP sewer
limited - high none moderate storage
limited - high limited limited transport capacity
limited high none treatment capacity
Table 13.11. Storage, transport and
treatment capacity of sewers,
wastewater treatment plants and surface
waters.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 454
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 455
requires estimates of how water demands are likely to
grow over time in various parts of the system. The output
of a scheduling exercise is a plan that identifies what
facilities should be built, installed or replaced, to what
capacity and when, over the planning horizon. This plan
of how much to do and when to do it should be updated
periodically long before the end of the planning horizon.
The application of optimization to master planning for
complex urban water infrastructure presents a significant
challenge. Using optimization methods to find the
minimum-cost design of a system of several thousand
pipes for a single demand at a single point in time is
difficult enough on its own. The development of least-
cost system designs over a number of time periods that
experience multiple increasing demands can be much
more challenging.
Consider, for example, developing a master plan for
the next twenty years divided into four five-year
construction periods. The obvious way to model this
problem is to include the system design variables for
each of the next four five-year periods, given the
expected demands at those times. The objective func-
tion for this optimization model might be to minimize
the present value of all construction, operation and
maintenance costs. As mentioned previously, this is a
very large problem that is probably unmanageable
with the current state of technology for real water
distribution systems.
Dandy et al. (2002) have developed and applied two
alternative modelling approaches. One approach is to find
the optimal solution for the system for only the final or
target year. The solution to this first optimization
problem identifies those facilities that will need to be
constructed sometime during the twenty-year planning
period. A series of sub-problems are then optimized, one
for each intermediate planning stage, to identify when
each necessary facility should be built. For these
sub-problems, the decisions are either to build or not to
build to a predetermined capacity. If a component is to be
built, its capacity has already been determined in the
target year optimization.
For the second planning stage, all options selected in
the first planning stage are locked in place and a choice is
made from among the remaining options. Therefore, the
search space is smaller for this case. A similar situation
applies for the third planning stage.
An alternative approach is to solve the first optimiza-
tion problem for just the first planning stage. All options
and all sizes are available. The decisions chosen at this
time are then fixed, and all options are considered in the
next planning stage. These options include duplication of
previously selected facilities. This pattern is repeated until
the final target year is reached.
Each method has its advantages and disadvantages.
For the first, build-to-target method, the optimum solu-
tion is found for the target year. This is not necessarily
the case for the build-up method. On the other hand, the
build-up method finds the optimal solution for the first
planning stage, which the build-to-target method does
not necessarily do. As the demands in the first planning
stage are known more precisely than those for the target
year, this may be an advantage.
The build-up method allows small pipes to be placed
at some locations in the first time planning stage, if
warranted, and these can be duplicated at a later time; the
build-to-target method does not. This allows greater
flexibility, but may produce a solution that has a higher
cost in present value terms.
The results obtained by these or any other optimiza-
tion methods will depend on the assumed growth rate
in demand, the durations of the planning intervals, the
economic discount rate if present value of costs is being
minimized, and the physical configuration of the system
under consideration. Therefore, the use of both methods
is recommended. Their outputs, together with engineer-
ing judgement, can be the basis for developing an
adaptive master development plan. Remember, it is only
the current construction periods solution that should be
of interest. Prior to the end of that period, the planning
exercise can be performed again with updated informa-
tion to obtain a better estimate of what the next periods
decisions should be.
5.3. Simulation
Dynamic simulation models are increasingly replacing
steady-state models for analysing water quantity, pres-
sure and water quality in distribution and collection
networks. Dynamic models provide estimates of the
time-variant behaviour of water flows and their
contaminants in distribution networks, even arising
from flow reversals. The use of long time-series analysis
wrm_ch13.qxd 8/31/2005 12:00 PM Page 455
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
provides a continuous representation of the variability
of flow, pressure and quality variables throughout the
system. It also facilitates the understanding of transient
operational conditions that may influence, for example,
the way contaminants are transported within the
network. Dynamic simulation also lends itself well to
statistical analyses of exposure. This methodology is
practical for researchers and practitioners using
readily available hardware and software (Harding and
Walski, 2002).
Models used to simulate a sequence of time periods
must be capable of simulating systems that operate under
highly variable conditions. Urban water systems are
driven by water use and rainfall, which by their natures
are stochastic. Changes in water use, control of responses
and dispatch of sources, and random storms over differ-
ent parts of the catchment, all can affect flow quantities
and direction, and thus the spatial distribution of con-
taminants. Because different water sources often have dif-
ferent quality, changing water sources can cause changes
in the quality of water within the system.
The simulation of water quantities and qualities in
urban catchments serves three general purposes:
Planning/Design. These studies define system configu-
rations, size or locate facilities, or define long-term
operating policies. They adopt a long-term perspective
but, under current practices, use short, hypothetical
scenarios based on representative operating condi-
tions. In principle, the statistical distribution of system
conditions should be an important consideration, but
in practice variability is considered only by analyses
intended to represent worst-case conditions.
Operations. These short-term studies analyse scenarios
that are expected to occur in the immediate future so
as to inform immediate operational decisions. These
are based on current system conditions and expected
operating conditions. These analyses are often driven
by regulations.
Forensics. These studies are used to link the presence
of contaminants to the risk or actual occurrence of
disease. Depending on whether the objective is cast in
terms of acute or chronic exposures, such studies
may adopt short or long-term perspectives. Because
there are often doseresponse relationships and
issues of latency in the etiology of disease, explicit
456 Water Resources Systems Planning and Management
consideration of the spatial distribution, timing,
frequency, duration and level of contamination is
important to these studies (Rodenbeck and Maslia,
1998; Aral, et. al., 1996; Webler and Brown, 1993).
6. Conclusions
Urban water systems must include not only the reservoirs,
groundwater wells and aqueducts that are the sources of
water supplies needed to meet the varied demands in an
urban area, but also the water treatment plants, the water
distribution systems that transport that water, together
with the pressures required, to where the demands are
located. Once used, the now wastewater needs to be
collected and transported to where it can be treated and
discharged back into the environment. Underlying all of
this hydraulic infrastructure and plumbing is the urban
stormwater drainage system.
Well-designed and operated urban water systems
are critically important for maintaining public health as
well as for controlling the quality of the waters into
which urban runoff is discharged. In most urban areas in
developed regions, government regulations require
designers and operators of urban water systems to meet
three sets of standards. Pressures must be adequate
for fire protection, water quality must be adequate to
protect public health, and urban drainage of waste and
stormwaters must meet effluent and receiving water body
quality standards. This requires monitoring as well as the
use of various models for detecting leaks and predicting
the impacts of alternative urban water treatment and
distribution, collection system designs and operating,
maintenance and repair policies.
Modelling the water and wastewater flows, pressure
heads and quality in urban water conveyance, treatment,
distribution and collection systems is a challenging
exercise, not only because of its hydraulic complexity, but
also because of the stochastic inputs to and demands on
the system. This chapter has attempted to provide an
overview of some of the basic considerations used by
modellers who develop computer-based optimization and
simulation models for design and/or operation of parts of
such systems. These same considerations should be in the
minds of those who use such models as well. Much more
detail can be found in many of the references listed below.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 456
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 457
7. References
AL-ZAHRANI, M.A. and MOIED, K. 2001. Locating water
quality monitoring stations in water distribution systems.
Proceedings of the World Water and Environmental
Resources Congress. Orlando, Fla., ASCE, May.
ARAL, M.M.; MASLIA, M.L.; ULIRSCH, G.V. and REYES,
J.J. 1996. Estimating exposure to volatile organic com-
pounds from municipal water supply systems: use of a
better computer model. Archives of Environmental Health,
Vol. 51, No. 4, pp. 3009.
BEDIENT, P.B. and HUBER, W.C. 1992. Hydrology and
floodplain analysis, 2nd ed. Reading, Mass., Addison-Wesley.
BOULOS, P.F. and WOOD, D.J. 1990. Explicit calcula-
tion of pipe-network parameters. Journal of Hydraulic
Engineering, ASCE, Vol. 116, No. 11, pp. 132944.
CHIN, D.A. 2000. Water-resources engineering. Upper
Saddle River, N.J., Prentice Hall.
DANDY, G.C. and CONNARTY, M. 1995. Use of genetic
algorithms for project sequencing. In: M.F. Domenica
(ed.), Integrated water resources planning for the twenty-first
century. Proceedings of the 22nd Annual Conference,
Water Resources Planning and Management Division.
ASCE, Cambridge, Mass., May, 5403.
DANDY, G.C. and ENGELHARDT, M. 2001. The optimal
scheduling of water main replacement using genetic
algorithms. Journal of Water Resources Planning and
Management, ASCE, Vol. 127, No. 4, July/August,
pp. 21423.
DANDY G.; KOLOKAS, L.; FREY, J.; GRANSBURY, J.;
DUNCKER, A. and MURPHY, L. 2002. Optimal staging
of capital works for large water distribution systems. In:
D.F. Kibler (ed.), Proceedings of the ASCE Environmental
and Water Resources Planning Symposium, Roanoke, Va.,
May, 2002. Reston, Va., ASCE Press (CD-rom).
DANDY, G.C.; SIMPSON, A.R. and MURPHY, L.J. 1996.
An improved genetic algorithm for pipe network opti-
mization. Water Resources Research, Vol. 32, No. 2,
pp. 44958.
DATTA, R.S.N. and SRIDHARAN, K. 1994. Parameter
estimation in water-distribution systems by least squares.
Journal of Water Resources Planning and Management,
ASCE, Vol. 120, No. 4, pp. 40522.
GRECO, M. and DEL GUIDICE, G. 1999. New approach
to water distribution network calibration. Journal of
Hydraulic Engineering, ASCE, Vol. 125, No. 8, pp. 84954.
GUJER, W.; HENZE, M.; MINO, T. and VAN LOOS-
DRECHT, M.C.M. 1999. Activated sludge model no. 3.
Wat. Sci. Tech., Vol. 39, No. 1, pp. 18393.
HARDING, B.L. and WALSKI, T.M. 2002. Long time-
series simulation of water quality in distribution systems.
Proceedings of the ASCE Environmental and Water
Resources Planning Symposium. Roanoke, Va., May .
HENZE, M.; GUJER, W.; MINO, T.; MATSUO, T.;
WENTZEL, M.C.; VAN MARAIS, G.M. and VAN LOOS-
DRECHT, M.C.M. 1999. Activated sludge model no. 2D,
ASM2D. Water Science and Technology, Vol. 39, No. 1,
pp. 16582.
HUBER, W.C. and DICKINSON, R.E. 1988. Storm water
management model, version 4: users manual. EPA/600/3-
88/001a (NTIS PB88-236641/AS). Athens, Ga., US EPA,
Also see http://www.epa.gov/ednnrmrl/swmm/ (accesssed
11 November 2004).
HVITVED-JACOBSEN, T.; VOLLERTSEN, J. and
TANAKA, N. 1998. Activated sludge model no. 3. Water
Science and Technology, Vol. 38, No. 10, pp. 25764.
LANSEY, K.E. and BASNET, C. 1991. Parameter estima-
tion for water distribution networks. Journal of Water
Resources Planning and Management, ASCE, Vol. 117,
No. 1, pp. 12644.
LIGGETT, J.A. and CHEN, L.C. 1994. Inverse transient
analysis in pipe networks. Journal of Hydraulic
Engineering, ASCE, Vol. 120, No. 8, pp. 93455.
MAYS, L.W. (ed.). 1999. Hydraulic design handbook.
New York, McGraw-Hill.
MAYS, L.W. (ed.). 2000. Water distribution systems hand-
book. New York, McGraw-Hill.
MAYS, L.W. 2005. Water resources engineering. New York,
Wiley.
MURPHY, L.J.; SIMPSON, A.R.; DANDY, G.C.; FREY, J.P.
and FARRILL, T.W. 1996. Genetic algorithm optimiza-
tion of the Fort Collins-Loveland Water Distribution
wrm_ch13.qxd 8/31/2005 12:00 PM Page 457
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Systems. In: Specialty conference on computers in the water
industry, AWWA, Chicago, Ill., April, pp. 1815.
ORMSBEE, L.E. and WOOD, D.J. 1986. Explicit pipe net-
work calibration. Journal of Water Resources Planning and
Management, Vol. 112, No. 2, 16682.
PRICE, R.K. 2002. Urban drainage modeling: course notes.
Delft, UNESCO International Institute for Infrastructural,
Hydraulic and Environmental Engineering.
PUDAR, R.S. and LIGGETT, J.A. 1992. Leaks in pipe net-
works. Journal of Hydraulic Engineering, ASCE, Vol. 118,
No. 7, pp. 103146.
RAHAL, C.M.; STERLING, M.J.H. and COULBECK, B.
1980. Parameter tuning for simulation models of water
distribution networks. Proceedings of Institution of Civil
Engineers, Part 2, Vol. 69, pp. 75162.
REDDY, P.V.N.; SRIDHARAN, K. and RAO, P.V. 1996.
WLS method for parameter estimation in water distribu-
tion networks. Journal of Water Resources Planning and
Management, ASCE, Vol. 122, No. 3, pp. 15764.
RODENBECK, S.E. and MASLIA, M.L. 1998. Groundwater
modeling and GIS to determine exposure to TCE at Tucson.
Practice Periodical of Hazardous, Toxic, and Radioactive Waste
Management, ASCE, Vol. 2, No. 2, pp. 5361.
ROESNER, L.A.; ALDRICH, J.A. and DICKINSON, R.E.
1988. Storm Water Management Model Users Manual,
Version 4: Addendum I, EXTRAN, EPA/600/3-88/001b (NTIS
PB88236658/AS). Athens, Ga., Environmental Protection
Agency.
SAVIC, D.A. and WALTERS, G.A. 1995. Genetic algorithm
techniques for calibrating network models, Report No. 95/12.
Exeter, UK, University of Exeter, Centre for Systems and
Control Engineering.
SAVIC, D.A. and WALTERS, G.A. 1997. Genetic algo-
rithms for least-cost design of water distribution net-
works. Journal of Water Resources Planning and
Management, ASCE, Vol. 123, No. 2, pp. 6777.
SAVIC, D.A.; WALTERS, G.A.; RANDALL SMITH, M.
and ATKINSON, R.M. 2000. Cost savings on large water
distribution systems: design through GA optimisation.
Proceedings of the Joint Conference on Water Resources
Engineering and Water Resources Planning and
Management. Minneapolis, Minn., ASCE, July.
458 Water Resources Systems Planning and Management
SHAMIR, U. 1974. Optimal design and operation of water
distribution systems. Water Resources Research, Vol. 10,
No. 1, 1974, pp. 2736.
SIMPSON, A.R.; DANDY, G.C. and MURPHY, L.J. 1994.
Genetic algorithms compared to other techniques for
pipe optimization. Journal of Water Resources Planning
and Management, ASCE, Vol. 120, No. 4, July/August,
pp. 42343.
SIMPSON, A.R.; SUTTON, D.C.; KEANE, D.S. and
SHERIFF, S.J. 1999. Optimal control of pumping at a
water filtration plant using genetic algorithms. In: D.A.
Savic and G.A. Walters (eds.), Water industry systems:
modelling and optimization applications, Baldock, UK,
Research Studies Press. Vol. 2, pp. 40715.
TUCCIARELLI, T.; CRIMINISI, A. and TERMINI, D.
1999. Leak analysis in pipeline systems by means of
optimal valve regulation. Journal of Hydraulic Engineering,
ASCE, Vol. 125, No. 3, 1999, pp. 27785.
USDA (US Department of Agriculture). 1972. National
engineering handbook. Washington, D.C., USGPO.
VITKOVSKY, J.P. and SIMPSON, A.R. 1997. Calibration
and leak detection in pipe networks using inverse
transient analysis and genetic algorithms, report no. R 157.
Adelaide, University of Adelaide, Department of Civil
and Environmental Engineering.
VITKOVSKY, J.P.; SIMPSON, A.R. and LAMBERT, M.F.
2000. Leak detection and calibration using transients and
genetic algorithms. Journal of Water Resources Planning
and Management, ASCE, Vol. 126, No. 4, pp. 2625.
WALSKI, T.M. 1983. Technique for calibrating network
models. Journal of Water Resources and Planning
Management, ASCE, Vol. 109, No. 4, pp. 36072.
WEBLER, T. and BROWN, 1993. I.S. Exposure to tetra-
chloroethylene via contaminated drinking water pipes
in Massachusetts: a predictive model. Archives of
Environmental Health, Vol. 48, No. 5, pp. 2937.
Additional References (Further Reading)
AXWORTHY, D.H. and KARNEY, B.W. 1996. Modeling
low velocity/high dispersion flow in water distribution
systems. Journal of Water Resources Planning and
Management, ASCE, Vol. 122, No. 3, pp. 21821.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 458
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Urban Water Systems 459
CHAUDRY, M.H. and ISLAM, M.R. 1995. Water quality
modeling in pipe networks. In: E. Cabrera and A.F. Vela
(eds.), Improving efficiency and reliability in water distribution
systems. Dordrecht, the Netherlands, Kluwer Academic.
CLARK, R.M.; GRAYMAN, W.M.; MALES, R.M. and
COYLE, J.M. 1986. Predicting water quality in distribu-
tion systems. In: Proceedings of the distribution system sym-
posium, Minneapolis, Minn. AWWA, September 710,
pp. 15580.
CLARK, R.M.; MALES, R. and STEVIE, R.G. 1984. Water
supply simulation model, Vols. I, II and III. Cincinnati,
Ohio, USEPA.
GRAYMAN, W.M.; CLARK, R.M. and MALES, R.M.1988.
Modeling distribution system water quality: dynamic
approach. Journal of Water Resources Planning and
Management, ASCE, Vol. 114, No. 3, pp. 295312.
ISLAM, M.R. and CHAUDRY, M.H. 1998. Modeling of
constituent transport in unsteady flows in pipe networks.
Journal of Hydraulic Engineering, ASCE, Vol. 124, No. 11,
pp. 111524.
LINGIREDDY, S. and ORMSBEE, L.E. 1998. Optimal
network calibration model based on genetic algorithms.
Lexington, Ky., University of Kentucky.
MALES, R.N.; CLARK, R.M.; WEHRMAN, P.J. and
GATES, W.E. 1985. Algorithms for mixing problems in
water systems. Journal of Hydraulics Engineering, ASCE,
Vol. 111, No. 2, pp. 20619.
MEADOWS, M.E.; WALSKI, T.M.; BARNARD, T.E. and
DURRANS, S.R. 2004. Computer applications in hydraulic
engineering, 6th Edn. Waterbury, Conn.
MORLEY, M.S.; ATKINSON, R.M.; SAVIC, D.A. and WAL-
TERS, G.A. 2001. GAnet: genetic algorithm platform for
pipe network optimisation. Advances in Engineering
Software, Vol. 32, No. 6, pp. 46775.
ORMSBEE, L.E. 1989. Implicit network calibration.
Journal of Water Resources Planning and Management,
ASCE, Vol. 115, No. 2, pp. 24357.
ROSSMAN, L.A.; BOULOS, P.F. and ALTMAN, T. 1993.
Discrete volume-element method for network water qual-
ity models. Journal of Water Resources Planning and
Management, ASCE, Vol. 119, No. 5, pp. 50517.
ROSSMAN, L.A.; CLARK, R.M. and GRAYMAN, W.M.
1994. Modeling chlorine residuals in drinking-water dis-
tribution systems. Journal of Environmental Engineering,
ASCE, Vol. 120, No. 4, pp. 80320.
SHAMIR, U. and HOWARD, C.D.D. 1968. Water distri-
bution systems analysis. Journal of the Hydraulic Division,
ASCE, Vol. 94, No. 1, pp. 21934.
TANG, K.; KARNEY, B.; PENDLEBURY, M. and ZHANG,
F. 1999. Inverse transient calibration of water distribution
systems using genetic algorithms. In: D.A. Savic and
G.A. Walters (eds.), Proceedings of water industry systems:
modelling and optimisation applications, Baldock, UK,
Research Studies Press. Vol. 1, pp. 31726.
TODINI, E. 1999. Using a Kalman filter approach
for looped water distribution network calibration In:
D.A. Savic and G.A. Walters (eds.), Proceedings of water
industry systems: modelling and optimisation applications.
Baldock, UK, Research Studies Press. Vol. 1, pp. 32736.
USEPA. 1992. CEAM systems development life cycle method-
ology (SDLCM) statement of policy, standards, and guidelines:
Version 1.00. Athens, Ga., USEPA.
VIESSMAN, W. Jr. and WELTY, C. 1985. Water manage-
ment technology and institutions. New York, Harper and Row.
wrm_ch13.qxd 8/31/2005 12:00 PM Page 459
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
14. A Synopsis
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 System 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. Reference 478
wrm_ch14.qxd 8/31/2005 12:01 PM Page 460
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Meeting the Challenge
At the beginning of this twenty-first century, water
developers and managers are being challenged to meet
ever-increasing demands for more reliable, cheaper and
cleaner water supplies for socio-economic development
and, at the same time, ensure a quality environment
and a resilient, sustainable and biodiverse ecosystem.
Problems associated with periodic droughts, floods,
pollution, aging infrastructure, threatened riverine
and coastal environments and degraded ecosystems,
conflicts over multiple purposes or uses, and increas-
ingly issues related to safety and security, are common
throughout much of the world. While there are many
different ways of meeting these challenges, it seems
obvious that water and related land development
projects and management practices within any river
basin should fit into an integrated resources manage-
ment plan for the entire basin, including, in some cases,
its estuarine and coastal zones.
Systems approaches encompassing multiple disci-
plines and multiple spatial and temporal scales are needed
to help define and evaluate alternative plans and policies
for managing the interdependent components of water
resources systems, including their watersheds, as an
integrated system. The condition of river basins in the
next ten, twenty or fifty years will largely depend on how
461
A Synopsis
This book introduced some approaches to modelling in support of water resources
systems planning and management. These modelling tools can help identify and
evaluate alternative plans and management policies in terms of their various
physical, economic, ecological and environmental, and social impacts. Model
outputs information relevant for deciding what to do and when and where are
based on model inputs. Model inputs include data, judgements and assumptions.
Models provide a way to identify just which of these data, judgements and
assumptions are critical and which are not. This sensitivity information can help
focus the political decision-making debate on issues that are most important in
the search of alternatives that best meet desired objectives and goals.
well various water-related economic, environmental and
social interests within these basins have been satisfied in
an integrated, equitable and sustainable way.
2. The Systems Approach to
Planning and Management
Systems approaches, including the use of extensions
of the modelling methods introduced in this book,
can provide an organized framework for resources
management and for estimating the important hydrologi-
cal, geomorphic, ecological, social and economic impacts
and trends over relevant scales of space and time. Within
a systems framework, multiple purposes can be investi-
gated, tradeoffs among competing objectives may be
identified and evaluated, potential adverse impacts can be
assessed, and the various costs and benefits, however
measured, of a project may be estimated and examined.
This can all be done within a context or process that
incorporates the concerns and desires of all those with an
interest or stake in the outcome.
Quantitative models can help inform interested stake-
holders and those individuals or agencies responsible for
recommending or making decisions or policy. The merits
and advantages, as well as the limitations, of various
quantitative methods for analysing various planning or
14
wrm_ch14.qxd 8/31/2005 12:01 PM Page 461
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
management issues are generally recognized throughout
the water resources community. The assumptions and
uncertainties associated with any model-generated impact
predictions should be understood and considered by
those using these model predictions.
2.1. Institutional Decision-Making
Due to institutional, fiscal and political factors, the approach
to decision-making in government water planning and
management agencies is often piecemeal, fragmented and
local-project oriented. The resulting decisions or actions
are not always as effective or efficient as they could be.
Decisions made today at the local level, without consider-
ation of how the entire system works and how it will
respond to such decisions, can lead to problems for
tomorrow. History points to many examples.
The cost of restoring much of the unique, but
degraded, Everglades ecosystem in south Florida in the
United States is now estimated to exceed US$8 billion.
Without considering the root causes of ecological stress
on the Everglades, much of that $8 billion may be poorly
spent. Land development and a rising sea level may
continue to encroach on the natural system, whether
restored or not. This reduces the ability of water managers
to create the regimes of hydrological flows, water levels,
water quality and hydro-periods needed to protect and
enhance the various habitats within the Everglades
(Figure 14.1).
The decision over the past decades to reroute river
sediment at the mouth of the Mississippi River away from
its delta has resulted in a loss of delta land (Figure 14.2).
This delta supports a diverse ecosystem and an infra-
structure used to offload crude oil from tanker ships and
pump it in from offshore oil platforms. An estimated
462 Water Resources Systems Planning and Management
US$14 billion will be needed to protect and restore that
eroding and subsiding delta. In addition, the excessive
nutrient loads in runoff from farmlands throughout much
of the Mississippi Basin have resulted in a hypoxic (dead)
zone of over 7,000 square miles (some 19,000 km
2
) along
the Louisiana coast in the Gulf of Mexico. Eliminating that
dead zone will be a massive political challenge in a basin
where land owners do not appreciate being told how to
manage their lands and their runoff.
Attempts to re-establish fish passage around dams
(Figure 14.3) throughout the Columbia River and its tribu-
taries have already cost an estimated US$3,000 per salmon
based on the current salmon population in the river.
Success in solving these and similar challenges
will depend on the extent to which they are viewed and
managed as part of the entire river basin, estuary and
coastal region, and on the extent of involvement of all
affected stakeholders.
These are among the many water resources
management issues that can be studied through the use
of quantitative modelling methods. Clearly one of the
biggest issues, even for river basins located entirely
within single countries but especially for international
river basins, is the institutional structure and capacity
needed to take advantage of a more comprehensive
systems approach to water resources planning and
management. How can multiple government agencies,
citizen or community groups, non-governmental
organizations and private companies each with inter-
ests, responsibilities and authorities to address a local or
limited range of water management issues and needs
generate broadly supported integrated regional
solutions? Where there exist active and well-staffed river
basin commissions (such as those for the Delaware,
Danube, Mekong and Rhine Rivers, to mention a few),
marine/
estuary
coastal
prairie
mangroves slough
cypress
marl
prairie
hammock pinelands
E040716b
Figure 14.1. Ecosystem habitats in the
Everglades in southern Florida (www.nps.gov).
wrm_ch14.qxd 8/31/2005 12:01 PM Page 462
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A Synopsis 463
there is a greater potential for basin-wide coordination of
multiple projects and management decisions. However,
these commissions are not always able to fund the
studies and analyses they would like to do in concert
with other stakeholders in a participatory planning
process that could result in strategies for guiding
the integrated and sustainable development and
management of their river basins (Chapter 1).
Water resources professionals and the informed public
are increasingly seeking better ways of identifying and
implementing solutions to water resources problems that
can be done in less time and at a lower cost than
traditional engineering projects. Working together in a
participatory planning process, while less efficient in the
speed of decision-making, is perhaps more effective. The
more traditional top-down planning and management
process is now becoming a multi-agency multi-stake-
holder bottom-up planning and management process. Yet
such a bottom-up process needs coordination to ensure
unity of purpose, facilitate collaboration, deal with con-
flicts, and promote the inclusion of the full spectrum of
public and private sector stakeholders. Both coordination
and technical expertise and advice can be provided by
government agencies if they have the ability and authority
to do so. All too often, they do not.
Professionals in various disciplines must provide the
information needed by those involved in the political
planning and management process if they are to make
Mississippi
River
Red River
Sabine
River
Gulf of Mexico
E
0
4
0
7
1
6
c
Figure 14.2. The Mississippi River Delta on the Louisiana coast
in the Gulf of Mexico (NOAA Central Library, www.noaa.gov).
wrm_ch14.qxd 8/31/2005 12:01 PM Page 463
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
informed decisions. This book has introduced some
modelling approaches that are being used to provide
some of this information. As discussed in Chapters 2 and
3, these tools can play an important role in helping
stakeholders become better informed and reach a shared
vision of how their water resources system functions and
perhaps, if possible, even how they want it managed.
2.2. The Water Resources System
Understanding the natural processes as well as economic
and social services or functions that rivers, lakes, reservoirs,
wetlands, estuaries and coasts perform is critical to the
successful and sustainable management of regional water
resources systems. These natural processes involve numer-
ous geomorphological, biological and chemical interactions
that take place among the components of fluvial systems
and their adjacent lands. Those who model and manage
them should be aware of these processes and interactions.
464 Water Resources Systems Planning and Management
The current hydrological, geomorphological, environ-
mental and ecological states of streams, rivers, wetlands,
estuaries and coasts are indicators of past and current
management policies or practices. The condition of such
components of a water resources system is an indicator of
how well they can function. Natural systems are able to
filter contaminants from runoff; store, absorb and gradu-
ally release floodwaters; serve as habitat for fish and
wildlife; recharge groundwaters; provide for commercial
transport of cargo; become sites for hydropower and
provide recreational opportunities beneficial to humans.
Degraded systems do not perform these functions so well,
and thus can result in added costs of providing alternative
ways of meeting these needs.
Today the importance of keeping natural aquatic
systems alive and well, diverse and productive, in
addition to meeting the needs of multiple economic and
social interests, is much better appreciated and
recognized than it was throughout much of our history up
Oregon
Idaho
Washington
Alberta
Montana
Nevada California Utah
British Columbia
P
a
c
i
f
i
c

O
c
e
a
n
R
i
v
e
r
Colum
bia
K
o
o
t
e
n
a
y
Salmon
R
iv
e
r
R
i
v
e
r
E
0
4
0
7
1
6
d
Snake
River
Mica
Revelstoke
Dundan
Keenleyside
Libby
Albeni
Falls
Grand
Coulee
Chief
Joseph
Wells
Rocky Reach
Rock Island
Wanapum
Priest Rapids
Noxon
Kenc
Hungry Horse
Dworshek
Lower Grande
Little Goose
Lower
Monumental
Ice Harbor
McNary
Bonnevile The
Dalles
John
Day Hells Canyon
Oxbow
Brownlee
Boise Projects Jackson Lake
Dams:
Private
Corps of Engineers
Bureau of Reclamation
S
n
a
k
e
Columbia River
Figure 14.3. Obstacles to fish passage on the
Columbia River in Canada and the United States.
wrm_ch14.qxd 8/31/2005 12:01 PM Page 464
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A Synopsis 465
through the first half of the past century. During most
of this time, water resources planners and managers
were involved in conquering nature and taming
its variability. Today, natural system restoration and
sustainability have become major objectives alongside the
usual economic services that water and related land
management can provide. Satisfying all these objectives to
the extent possible requires an understanding of the basic
hydrological, geological, environmental and ecological
interactions that take place in natural aquatic systems.
Appendix A reviews some of the major natural
processes and interactions that occur in watersheds,
river basins, wetlands, estuaries and along coasts. It
provides a background for those who may not be
familiar with this subject and who wish to model such
processes and their interactions for the purposes of
planning and management.
2.3. Planning and Management Modelling:
A Review
Water resources systems are characterized by multiple
interdependent components that together produce multi-
ple economic, environmental, ecological and social
impacts. Planners and managers working toward improv-
ing the performance of, or the services provided by, these
complex systems must identify and evaluate alternative
plans and management or operating policies, comparing
their predicted performance with desired goals or objec-
tives. Constrained optimization and simulation modelling
is the primary way we have of predicting system perform-
ance associated with any plan or operating policy.
Chapters 3 and 4 introduced the development and use of
some of the modelling methods commonly applied for
identifying and analysing the performance of water
resources systems. Chapter 5 showed how these models
can be extended to include problems or issues better
described qualitatively rather than quantitatively.
Models require data. Data sets contain information,
often much more than can be learned from just looking at
various graphical plots of those data. Statistical or data-
based models help us abstract and gain new information
and understanding from these data sets. They also serve
as substitutes for more process-based models in applica-
tions where computational speed is critical or when more
process-oriented models become too complex and/or too
data demanding. Statistical data-based models range from
the commonly used regression models to more recent
ones based on evolutionary or biological processes. In
Chapter 6 three such models were described. They are
among a wide range of data mining tools designed to
identify and abstract information from large sets of data.
Events that cannot be predicted precisely are often
called random. Many if not most of the inputs to, and
processes that occur in, water resources systems are to
some extent random. Hence, so too are the outputs of
predicted impacts, and even peoples reactions to those
outputs or impacts. To ignore this randomness or uncer-
tainty when performing analyses in support of decisions
is to ignore reality. Chapters 7, 8 and 9 reviewed some of
the commonly used tools for dealing with uncertainty in
water resources planning and management. Chapter 9
also addressed the topic of sensitivity analyses: ways of
determining the sensitivity of model output to changes in
the values of model inputs.
Water resources systems provide a variety of eco-
nomic, environmental and ecological services. They also
serve a variety of purposes (such as water supply, flood
protection, hydropower production, navigation, recre-
ation, and waste reduction and transport). Performance
criteria provide measures of just how well a plan or
management policy performs. There are a variety of
criteria one can use to judge and compare system
performance. Some of these performance criteria may be
conflicting. The water that has access to one consumer is
not available to another consumer, and vice versa. In
these cases, tradeoffs exist among conflicting criteria
and these tradeoffs must be considered when searching
for the best compromise decisions. Chapter 10 presented
ways of identifying and working with these tradeoffs in
the political process of selecting the best decision.
Multipurpose river basin development projects typi-
cally involve the identification and use of both structural
and non-structural measures that are designed to, for
example, increase the reliability of municipal, industrial
and agricultural water supplies when and where
demanded, to protect against floods, to improve water
quality, to provide for commercial navigation and recre-
ation, and to produce hydropower, as appropriate for the
particular river basin. Structural measures may include
diversion canals, reservoirs, hydropower plants, levees,
flood proofing, irrigation delivery and drainage systems,
wrm_ch14.qxd 8/31/2005 12:01 PM Page 465
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
navigation locks, recreational facilities, groundwater
wells, and water and wastewater distribution and
collection systems and treatment plants. Non-structural
measures may include land use controls and zoning, flood
warning and evacuation measures, and economic and
legal incentives that affect human behaviour with regard
to water and watershed use. Planning the development
and management of water resources systems involves
identifying just what and when and where various
structural or non-structural measures are needed, the
extent to which they are needed, and their combined
economic, environmental, ecological and social impacts.
Chapters 11 and 12 introduced some modelling
approaches for doing this. Chapter 11 focused on
measures and models for water quantity management.
Chapter 12 focused on measures and models applicable
for water quality management.
Most people in most countries live in urban areas.
These areas have their own special types of water
resources management issues and problems. As popula-
tions continually move to urban areas to seek a higher
standard of living and improved economic opportunities,
and as cities merge to form megacities (with populations
in excess of 10 million), the design and management of
urban water supply systems becomes an increasingly
important part of regional integrated water resources
planning and management. Chapter 13 focused specifi-
cally on measures and models applied to the planning and
management of water and wastewater in urban areas.
Many of these tools are applicable in rural areas as well.
The management of water resource systems to serve
various purposes and to meet various objectives requires
actions with respect to land and water use. Regulating
water quantity and quality often requires engineering
infrastructure. The planning, design and operation of
such infrastructure should be based on the best science
available. Even the best science available is often not
enough to give us the understanding we need to be sure
we are making the right decisions. Furthermore, the goals
and objectives our decisions are intended to meet can,
and usually do, change over time. Given this uncertainty
and changing political and social environment, it makes
sense to monitor the impacts of our decisions and adapt
make changes as appropriate.
All too often there is no follow-up monitoring carried
out to judge the effectiveness of past decisions.
466 Water Resources Systems Planning and Management
Monitoring and adaptive management are increasingly
being advocated when water management decisions are
made on the basis of an uncertain science. But what
should be monitored, and where, and how often and to
what accuracy? These issues are addressed in Appendix B.
Considerable efforts have been (and continue to be)
spent preparing for droughts or floods. Appendices C
and D are specifically devoted to drought and flood
management issues, where models can play an important
role in evaluating the effectiveness of any measures taken
to mitigate damage from these extreme hydrological
events. Models and their associated computer software
such as The Planning Kit of Delft Hydraulics for
floodplain planning or the US Army Corps of Engineers
shared vision modelling approaches to drought
management planning have proven successful at
stakeholder, professional and various governmental
levels in the Netherlands and in the US.
3. Evaluating Modelling Success
There are a number of ways one can judge the extent of
success (or failure) in applying models and performing
analyses in practice. Goeller (1988) suggested three meas-
ures as a basis for judging success:
1. How the analysis was performed and presented (analy-
sis success).
2. How it was used or implemented in the planning and
management processes (application success).
3. How the information derived from models and their
application affected the system design or operation and
the lives of those who use the system (outcome success).
It is often hard to judge the extent to which particular
models, methods and styles of presentation are appropri-
ate for the problem being addressed, the resources and
time available for the study, and the institutional environ-
ment of the client. Review panels and publishing in
peer-review journals are two ways of judging. No model
or method is without its limitations. Two other obvious
indications are the feelings that analysts have about their
own work and, very importantly, the opinions the clients
have about the analysts work. Client satisfaction may not
be an appropriate indicator if, for example, the clients are
unhappy only because they are learning something they
wrm_ch14.qxd 8/31/2005 12:01 PM Page 466
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A Synopsis 467
do not want to accept. Producing results primarily to
reinforce a clients prior position or opinions might result
in client satisfaction, but, most would agree, this is not
an appropriate goal of modelling.
Application or implementation success implies that
the methods and/or results developed in a study were
seriously considered by those involved in the planning
and management process. One should not, it seems to us,
judge success or failure on the basis of whether or not any
of the model results (the computer printouts) were
directly implemented. What one hopes for is that the
information and understanding resulting from model
application helped define the important issues and iden-
tify possible solutions and their impacts. Did the model-
ling help influence the debate among stakeholders and
decision-makers about what decisions to make or actions
to take? The extent to which this occurs is the extent to
which a modelling study will have achieved application or
implementation success.
Outcome success is based on what happens to the
problem situation once a decision largely influenced by
the results of modelling has been made and implemented.
The extent to which the information and understanding
resulting from modelling helped solve the problems or
resolve the issues, if it can be determined, is a measure of
the extent of outcome success.
It is clear that success in terms of the second or third
criteria will depend heavily on the success of the preced-
ing one(s). Modelling applications may be judged
successful in terms of the first two measures but, perhaps
because of unpredicted events, the problems being
addressed may have become worse rather than improved,
or while those particular problems were eliminated, their
elimination may have caused other severe problems. All
of us can think of examples where this has happened.
For example, any river restoration project involving the
removal of engineering infrastructure is a clear indication
of changing objectives or new knowledge. Who knows
whether or not a broader systems study might have
helped earlier planners, managers, and decision-makers
foresee such consequences, but one cannot count on that.
Hindsight is always clearer than foresight. Some of what
takes place in the world is completely unpredictable. We
can be surprised now and then. Given this, it is not clear
whether we should hold modellers or analysts, or even
planners or managers, completely responsible for any lack
of outcome success if unforeseen events that changed
goals, or priorities or understanding did indeed take place.
Problem situations and criteria for judging the extent
of success will change over time, of course. By the time
one can evaluate the results, the system itself may have
changed enough for the outcome to be quite different
than what was predicted in the analysis. Monitoring the
performance of any decision, whether or not based on a
successfully analysed and implemented modelling effort,
is often neglected. But monitoring is very important if
changes in system design, management and operation are
to be made to adapt to changing and unforeseen condi-
tions. (Adaptive management is discussed in more detail
in Appendix B.)
If the models, data, computer programs, documenta-
tion and know-how are successfully maintained, updated,
and transferred to and used by the client institutions,
there is a good chance that this methodology will be able
to provide useful information relevant to the changes that
are needed in system design, management or operation.
Until relatively recently, the successful transfer of models
and their supporting technology has involved a consider-
able commitment of time and money for both the analysts
and the potential users of the tools and techniques. It has
been a slow process. Developments in interactive
computer-based decision support systems that provide a
more easily understood humanmodeldatacomputer
interface have substantially facilitated this technology
transfer process, particularly among model users. These
technology developments have had a major impact on the
state of the practice in using models in support of water
resources planning and management activities.
4. Some Case Studies
What does one do with models used for analysing water
resources planning and management issues? The follow-
ing sections describe a few case studies or projects involv-
ing the application of modelling methods that are
extensions of those described in this book. These projects
were carried out by Delft Hydraulics in cooperation with
other consulting firms and governmental and inter-
national institutions, and are typical of many such studies
sponsored by and often carried out by international
development banks and various UN and national foreign
wrm_ch14.qxd 8/31/2005 12:01 PM Page 467
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
aid agencies. They serve to illustrate only a few of the
wide variety of types of planning and management
problems, issues, opportunities and challenges facing
water resources planners and managers.
4.1. Development of a Water Resources
Management Strategy for Trinidad and Tobago
The goal of this project was to develop a national water
resources management strategy that would lead to
sustainable management of the available water resources
in Trinidad and Tobago. An important element in this
project was the establishment of an effective and finan-
cially autonomous institutional setting that guaranteed the
optimum management of the islands water resources
sector. The project lasted just over one year. The planning
horizons for the study were the years 2015 and 2025.
Trinidad and Tobago are two islands in the West
Indies directly opposite the mouth of the Orinoco River
(Figure 14.5). Although the average precipitation (2,000
mm/year) is quite high, its distribution in time (there is a
wet and a dry season) and space (it is concentrated in the
hills along the northern coast) does not match when and
where it is needed. This mismatch between supply and
demand creates a need for storage and distribution
systems. The last decades have witnessed a deterioration
of both availability and quality. The public water supply
sector has failed to meet the demands for water during the
wet as well as the dry season. Surface water quality poses
a major constraint on the use of the resource. The lack of
468 Water Resources Systems Planning and Management
proper watershed management has resulted in increased
peak flows and reduced discharge capacities of the rivers.
Environmental problems have become increasingly
apparent and will worsen if measures are not taken to
reverse these trends. Without any major investments, the
water resources system will limit the expected future
growth in various demand sectors.
The project was divided into three distinctive phases
(as described in the framework of analysis sections of
Chapter 3 and further in Appendix E), each of which
involved client participation. The Inception Phase defined
the water resources system to be studied and identified the
various problems and issues of concern. The Development
Phase analysed the water-related problems and defined the
scope of the study. This phase passed through a number
of steps, from collection of data and information on all
relevant subjects to a preliminary analysis of the water
resources system. This analysis identified the various
economic growth bottlenecks for present and future
situations. Subsequently a large number of possible
demand-oriented, supply-oriented, and environment-
oriented measures were identified. These measures were
grouped according to five development strategies that
were considered as possible solutions for both present
and future problems in the water resources situation. Each
of these development strategies was examined with the
aid of a river basin simulation model. A number of water
supply, water quality, and environmental criteria were used
to classify the degree of success of each strategy. These
analyses included assessments of the necessary changes in
Figure 14.4. On the shores of
Trinidad and Tobago, a
hardship site for getting any
work done!
wrm_ch14.qxd 8/31/2005 12:01 PM Page 468
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A Synopsis 469
the institutional and legal framework that could lead to an
improvement of the management of the water resources.
The water resources management study required an
ample range of data, varying from discharge and rainfall
data to per-capita water demand and expected trans-
portation losses. The hydrological inputs to the simula-
tion model were based on a statistical extension of the
original historical records by applying time-series genera-
tion tools (as discussed in Chapter 7). The analysis
included the assessment of the water availability (both
surface water and groundwater) and the present and
future (projected) water demand, both for public water
supply (including industry) and irrigated agriculture.
The application of a simulation model allowed for the
assessment of individual measures as well as groups of
measures according to a certain development strategy.
The impact of the various measures on the water flows
and quality were displayed by means of the graphical
tools in the simulation model. Figure 14.6 shows one
such display.
An important aspect in the definition of a water
resources strategy is the expected impact of any recom-
mended measures on environmental and ecological
conditions. Ongoing deterioration of the natural
environment is evident from the major soil loss from hills
due in part to the large-scale deforestation that has
occurred in a number of regions.
On the basis of the information on water availability,
water demand and other water-related aspects in Trinidad
and Tobago, water resources management strategies were
0
Gulf of Paria
Atlantic Ocean
Caribbean Sea
Venezuela
E
0
4
0
7
1
6
e
25 km
TOBAGO
Fullerton
San Juan
Toco
Princes Town
Morugo
San Fernando
Brighton
Port of Spain
ST.
GEORGE
ST. DAVID
ST.
ANDREW
CARONI
NARIVA
MAYARO
VICTORIA
PATRICK
Figure 14.5. Trinidad and Tobago are two
islands located in the West Indies.
wrm_ch14.qxd 8/31/2005 12:01 PM Page 469
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
470 Water Resources Systems Planning and Management
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
fraction calculation over a five-year period
initial
variable inflow
PWS return flow
irrigation return flow
tumpuna inflow
ib: n. oro / matura
ib: groundwater
ib: caroni-a confl.
E
0
4
0
7
1
6
g
Figure 14.6. Composition of the
water at the inflow to the Point
Lisas Industrial Estate.
developed. They focused on different topics and goals.
Each of the suggested strategies consisted of a number of
related measures that in combination met pre-set criteria.
These included supply reliability, environmental protec-
tion, government investment requirements and time
required for implementation. Failure criteria were chosen
such that both the time of failure and the extent of the
failure would be taken into account. The final decision
made by the government of Trinidad is illustrated in
Figure 14.7.
Other aspects taken into account were changes in the
existing legal framework, the impact on other agencies
related to water resources, and the practical implementa-
tion of a new agency within the Ministry of Planning and
Development. This agency should be financially inde-
pendent, with income from licence fees covering the costs
of the organization. A Planning Unit within that agency
will have responsibility for the preparation and periodic
revision of the water resources management plan, and will
take over the planning activities that have been performed
in the project.
4.2. Transboundary Water Quality
Management in the Danube Basin
The Danube River Basin (as described in Chapter 1)
covers some 817,000 square km in parts of eighteen
countries in the heart of central Europe. The basins
population of 85-million is characterised by large socio-
economic differences. The river flows from relatively
rich western European states to relatively poorer former
Soviet Union Republics. The rivers larger tributaries
include the Sava (Slovenia, Croatia, Bosnia-Herzegovina,
Yugoslavia), the Tisa (Ukraine, Slovakia, Hungary, Romania,
Yugoslavia), the Drava (Austria, Slovenia, Croatia, Hungary)
and the Inn (Germany, Austria). The inhabitants of the basin
use the Danubes water intensively (Figure 14.8). The basin
includes many important natural areas, including the
Danube delta, the second largest wetland area in Europe.
On 22 October 1998, the Danube River Protection
Convention (DRPC) became the overall legal instrument
for cooperation and transboundary water management in
the Danube River Basin. The overall objective of the
DRPC is to achieve and maintain the sustainable develop-
ment and use of water resources in the Danube River
Basin. The executing body of the convention is the
International Commission for the Protection of the
Danube River (ICPDR). A modelling project, lasting from
1995 to 2004, has assisted the ICPDR in establishing and
implementing sustainable water quality management on a
basinwide scale.
Events such as accidental spills, flooding and ice jams
may degrade the downstream ecosystem and jeopardize
the supplies of water. In such cases, there is a clear need
for the collection and dissemination of early information
about these events. To safeguard the aquatic environment
and the use of water in the riparian states, the project
developed and implemented an Accident Emergency
Warning and Prevention System (AEWPS) for accidental
pollutant discharges into the river Danube. The principal
aim of the AEWPS is to communicate information about
sudden changes in the water characteristics, such as
wrm_ch14.qxd 8/31/2005 12:01 PM Page 470
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A Synopsis 471
accidental spills or unpredictable changes in the water
level, with special attention to transboundary impacts.
The focal locations of the AEWPS are the Principal
International Alert Centres (PIACs), shown on Figure 14.9.
A Danube Basin Alarm Model (DBAM) was developed
to support decision-making in relation to accidental
spills with a probable transboundary impact. The
January 2000 Baia Mare spill, Figure 14.10, presents a
dramatic example.
The model provides forecasts of the travel time and the
expected peak concentrations in the cloud of pollutants
during its travel down the river. The DBAM was designed
for use in operational conditions, to provide a fast
and first-order assessment of the effects of a spill. It uses
swamp
settlement
new reservoir
new intake
upgrade intake
existing transfer
proposed transfer
main water demand centres
hydrometric area
existing reservoir
0
Gulf of Paria
Atlantic
Ocean
Venezuela
E
0
4
0
7
1
6
h
5 0 5 0 25 km
Toco
Matura
Matura
North
Oropoche
swamp
Hodlis
Salybia
Guanapo
Tacarigua
St.Joseph
Maraval
Port of Spain
Caroni swamp
Caroni Arima
Kelly weir
Arena
Nariva
swamp
North
Oropogche
Cauva
Navet
Navet
San Fernando
Point Lisas
Ortoire
Guayaguayare
Pilot
Moruga
Moruga
South Oropuche
Erin
Erin
Point Fortin
Guapo
Los Blanquizales
Lagoon
Cedros
Inniss
1 1 2
Caribbean Sea
Figure 14.7. Water resources
development strategy for Trinidad.
Figure 14.8. The Blue Danube at Budapest, Hungary.
wrm_ch14.qxd 8/31/2005 12:01 PM Page 471
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
readily available limited-input data. For reasons of
computational speed and accuracy, the model uses
analytical techniques to solve the governing mathematical
advectiondiffusion equation (discussed in Chapter 12).
Experts in ten Danube countries are currently using the
472 Water Resources Systems Planning and Management
model. The water quality component of this model has
been used to support a so-called Transboundary
Diagnostic Analysis for the nutrients nitrogen and
phosphorus, as well as to assess the effectiveness of the
proposed pollution reduction program.
Black
Adriatic
Romania
Austria
Hungary
Bulgaria
Bosnia
Hercegovina
0 250 km
Inn
Mura
Morava
Tisza
Prut
Siret
Olt
Danube
Iskar
v.Morava
Mures
Drava
Slovakia
Vh
Sava
PIAC
AEWS
PIAC
AEWS
PIAC
AEWS
PIAC
AEWS
PIAC
AEWS
PIAC
AEWS
PIAC
PIAC
PIAC
AEWS
AEWS
AEWS
PIAC
PIAC
PIAC
AEWS
AEWS
AEWS
Austria
Czech
Republic
Slovakia Hungary Ukraine Moldova Ukraine Germany
Croatia
Slovenia Romania
Bulgaria
Sea
Sea
E
0
4
0
7
1
6
f
E
0
4
0
7
1
6
n
Figure 14.9. The Principle
International Alert Centres
(PIACs) of the Danube AEWPS.
Figure 14.10. Application of the
DBAM to the Baia Mare spill in
January 2000.
wrm_ch14.qxd 8/31/2005 12:01 PM Page 472
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A Synopsis 473
4.3. South Yunnan Lakes Integrated
Environmental Master Planning Project
This project developed models and provided training aimed
at protecting and rehabilitating four lakes: Fuxian, Xingyun,
Qilu and Yilong. These four lakes are among the nine
plateau lakes in Yunnan Province, China, (Figure 14.11).
Figure 14.12 shows fishing boats on Fuxian Lake.
The South Yunnan lakes are threatened by water pollu-
tion and by decreased storage volumes, resulting in
increased risks of flooding and drought. Water pollution is
caused by wastewater discharges from industrial, urban
Kunming
Fuxian
Xingyun
Oilu
Yilong
Beijing
Yunnan
E
0
4
0
7
1
6
j
Figure 14.11. Lakes of project
area in Yunnan Province, China.
Figure 14.12. Fishing boats
on Fuxian Lake.
wrm_ch14.qxd 8/31/2005 12:01 PM Page 473
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
and agricultural sources. This leads to eutrophication of the
lakes (high nutrient levels and high productivity of water
plants and algae), as shown in Figures 14.13 and 14.14.
The decrease in storage capacity of the lakes is caused
by the accumulation of organic material resulting from
eutrophication, runoff and deposition of eroded soil
material, and the reclamation of land for agricultural
production. Soil erosion in the lake basins is a result of
the collection of fuel wood, and overgrazing by livestock.
The South Yunnan Lakes Integrated Environmental
Master Planning Project was aimed at reversing this
trend, improving the quality of life in the lake basins,
and demonstrating what could be done to restore the
other lakes in the region. A decision support system (DSS)
(Chapter 3) was developed to provide information on the
effectiveness of various measures for accomplishing these
474 Water Resources Systems Planning and Management
project goals. These measures included:
reduction of point and non-point pollution, to protect
and improve the current water quality
flood protection
reduction of water shortages and the improvement of
irrigation management and infrastructure (Figure 14.15)
erosion prevention
reduction of health impacts from pesticides, hazardous
waste and bacteriological pollution
institutional strengthening and awareness raising,
leading to an integrated approach for the management
of the lakes.
An evaluation process using multiple criteria analysis
(Chapter 10) was used to select only the most effective
measures. The first task of the DSS was to analyse the
present water balance, waste loads and water quality.
Second, it was used to estimate future conditions, includ-
ing the lake quality, flood protection, reduction of water
shortages, erosion prevention, biodiversity, public health,
and macro economic development and poverty alleviation
under different water management strategies or scenarios.
The criteria were weighted to reflect their relative
importance to the development of the lake basins. The
determination of the weighted factors, a political rather
than technical issue, was determined for each lake during
sessions held with the representatives of all involved at
the Bureau and County levels (Figure 14.16).
The minimum package of measures needed to achieve
the water quality targets was identified with the aid of the
Figure 14.13. Lotus flowers on Yilong Lake.
Figure 14.14. Children bathing among water hyacinths. Figure 14.15. Water is pumped from Yilong Lake for irrigation.
wrm_ch14.qxd 8/31/2005 12:01 PM Page 474
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A Synopsis 475
DSS. This was followed by the selection of a package of
additional measures to satisfy the other criteria, based on
the multiple criteria analysis. Basic design reports were
prepared for each of the selected measures, including
the assessment of their technical, financial, economic and
social feasibility and their environmental impact. Most
importantly, institutional organizational issues were
addressed and recommendations were made that would,
if implemented, ensure the proper execution of the master
plans and continued training, monitoring, planning and
adaptation to changing goals and new knowledge.
4.4. River Basin Management and Institutional
Support for Poland
A three-year project for the Polish Ministry of Environment,
the agency responsible for water management in Poland,
has involved river basin planning. The objective of this
project was to support Poland with the implementation
of the EU Water Framework Directive in Poland. A pilot
project in the Brda River catchments area served to
improve the skills of the Polish experts and other parties
in the field of integrated water resources management.
The Brda River Basin is located in the northwest of
Poland. The main management issue in the basin is the
improvement of surface water quality to safeguard present
and future drinking water abstractions from the river for
the city of Bydgoszcz, shown in Figure 14.17.
The water quality problems in the basin are mainly
due to agricultural and domestic wastewater discharges.
The process of preparing the entire plan for the Brda
Basin, and implementing a training program, started at
the beginning of 2001 and was completed in June 2003.
The development of an integrated water resources
management plan for the Brda River (Figure 14.18)
served to: first, strengthen the institutional organization
and coordination; second, enhance public interest
through an open planning process; and third, improve
the financial and operational management and the
planning capacity of the ministry.
The Brda Catchment has a population of about
550,000, of which some 300,000 live in the area of
Bydgoszcz, which is in the lower part of the Brda Basin.
The upstream part of the basin is rural, with agricultural,
forestry and some tourism activities, as shown in
Figure 14.17. Forest and agriculture lands occupy about
85% of the basin area; 10% is urban, and 5% is covered
by water bodies. Agriculture, farming and aquaculture
are quite intensive, with many cattle (60,000), pigs
(340,000), poultry (270,000) and sheep (10,000), as well
as fish farms. This intensive agriculture causes water
Figure 14.16. Workshop with local decision-makers about the
weighting factors for the multiple criteria analysis.
forest
rural area
urban area
manure intensively fertilized area
Czyzkowko drinking water intake
protection zone
Bydgoszcz
Osielsko
Dobrcz
Pruszcz
Swiekatowo
Lubiewo
Cekcyn
Sepolno
Krajenskie
Kamien
Krajenski
Rzecznica
Koczala
Studzienice
Wisla
E
0
4
0
7
1
6
k
Figure 14.17. The Brda River Basin in Poland.
wrm_ch14.qxd 8/31/2005 12:01 PM Page 475
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
quality problems that result from runoff of organic waste
and nutrients. Industries are mainly found in and around
Bydgoszcz. They include food production, (basic) metals
and ceramics, and chemical industries (pharmaceutical,
organic-chemical and the like).
Increasing agricultural and industrial activities are
likely to result in increased animal waste loads. An
increase of population will also put additional pressure
on the water quality and the environment. The action
plan is designed to protect the raw water source and
safeguard the public water supply intake on Brda River
for Bydgoszcz City. To estimate and evaluate the impacts
of suggested projects and measures, computer simula-
tions of the Brda Basin were executed. The results of
these modelling studies were used by the Polish
authorities to identify a final list of protective measures
for the basin.
4.5. Stormwater Management in The Hague in
the Netherlands
During periods of heavy rainfall when sewers are not
capable of transporting all the stormwater to wastewater
treatment plants, the excess water is discharged into the
surface water. These sewer overflows degrade the water
476 Water Resources Systems Planning and Management
quality of urban channels and pools and flood the streets
(Figure 14.19).
The objective of this study was to develop a set of
measures with which both the emission standards and
water quality standards for surface water could be met,
at the lowest possible cost. The measures refer to both
the sewage systems and the surface water.
A systems analysis was performed for the sewage and
surface water networks of the Hague and its neighbouring
municipalities. Discharges of sewage systems have to meet
the Dutch emission standards and in addition must not
degrade the quality of the receiving surface water.
Attention was also given to preventing street flooding. The
project scope included the municipalities of the Hague,
Rijswijk, Leidschendam/Voorburg and Wateringen, and
the water control board of Delfland. These municipalities
share a large part of their sewage systems. The water con-
trol board of Delfland is responsible for both the water
quantity and water quality in this region.
In this project, the sewage and surface water systems
were modelled together as a single system. Sewage
systems transport wastewater (and rainwater in the case
of for combined sewage systems) to the wastewater treat-
ment plant (WWTP). An overloaded sewage system
results in overflows into surface water, which harm its
Figure 14.18. Beeches by
the Brda river (Marcin
Kowalski).
wrm_ch14.qxd 8/31/2005 12:01 PM Page 476
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A Synopsis 477
water quality. In the city of The Hague, surface water has
a major role in boat transport and aesthetics (including
tourism).
Within the region studied, the sewage system
(Figure 14.20) serves 36 drainage areas and consists of
about 1,360 km of main sewage pipes, 18 storage areas,
220 sewage overflows, and about 30,000 inspection
sites. The surface water system consists of a large part of
Delflands boezem (storage system) and eleven polders. It
includes about 835 km of canals, 46 pumps, 131 orifices,
594 culverts and 42 km
2
of drained surface area.
The sewage system is divided into eight areas which
together cover some 23,000 manholes and 27,000 pipes. In
the surface water model (Figure 14.21) only those channels
are taken into account which were thought to be important
for judging the effect of sewage overflows. Although many
smaller channels were not modelled, the surface water
model is one of the largest used for water quality evalua-
tions, having over 5,500 water quality segments.
Models for the sewage and surface water systems were
used to evaluate the reference situation. The model of the
sewage system provided insight into situations where
streets are flooded and sewage overflows occur. The effect
of sewage overflows on water quality was simulated with
the surface water model. The analysis of the reference
Figure 14.19. The results of excess
stormwater on the otherwise
picturesque canals and streets of
urban areas in the Netherlands.
E
0
4
0
7
1
6
s
Rijswijk
the Hague
Scheveningen
0
N
o
r
t
h

S
e
a
5 km
Figure 14.20. The main sewer system of the project area.
wrm_ch14.qxd 8/31/2005 12:01 PM Page 477
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
situation highlighted the major problem areas: water on
streets, organic pollutant (BOD) loads, and water qual-
ity of receiving channels. The problems occurring in the
receiving channels are caused by large residence times of
the water and/or large BOD loads.
To analyse this problem, the following measures were
considered:
basic measures, aiming at compliance with emission
standards
integral measures, aiming at reduction of street flood-
ing and compliance with the emission standards and
the water quality requirements
Sewer system measures included:
enlargement of the outflow capacity
enlargement of the storage capacity
construction of facilities at surface overflows, such as
storage basins
displacement of sewage overflows
construction of separate or improved separate sewage
systems
transformation of combined sewer systems into
improved separated sewage systems.
Surface water system measures included:
enlargement of the surface water volume
improvement of the circulation of the surface water for
dilution and displacement of the BOD effluent load.
478 Water Resources Systems Planning and Management
Combined system measures included:
improving the discharge to existing storage basins
addressing causes of sewage overflows with large BOD
loads
subdividing and separating drainage areas to reduce
volumes of overflow at single discharge points.
Other measures involved the enlargement of pipes in the
sewage system, displacement of sewage overflows, the
enlargement of storage capacity in the sewage system, and
the construction of flood control and sedimentation facilities
at sewage overflows. Each of these measures aims at reduc-
ing the risk of flooding and BOD loads to the surface water.
The measures for surface water aim at reducing the
effect of sewage overflows on the water quality. This may
be done by improving the circulation of water near
sewage overflows. Another measure prevents polluted
water from flowing to a part of the surface water system
that has a better water quality.
The models that were used in this project will be used
in future evaluations of alternative measures to further
improve the surface water quality in one of the largest
cities of the Netherlands.
5. Summary
This concluding chapter has attempted to summarize
and, using some case studies, illustrate the modelling
methods introduced in this book for water resources
systems planning and management. It is difficult today to
find any water resources planning or management project
which does not benefit from the use of modelling. Readers
are encouraged to refer to the current journals and other
sources of literature to learn more about this subject, and
then enjoy the professional life of a water resources
systems planner and manager. For us it has been fun,
always challenging, and a profession that can serve and
benefit society as well.
6. Reference
GOELLER, B.F., 1988. A framework for evaluating success
in systems analysis. Report P-7454. Santa Monica, Calif.,
Rand Corporation.
E
0
4
0
7
1
6
t
0
rain water outlet
combined sewage overflow
sewage overflow st. basin
sewage overflow st. pipe
N
o
r
t
h

S
e
a
5 km
Figure 14.21. The main surface water system of the project area.
wrm_ch14.qxd 8/31/2005 12:01 PM Page 478
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions
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
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
wrm_appA.qxd 8/31/2005 12:08 PM Page 480
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
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
wrm_appA.qxd 8/31/2005 12:08 PM Page 481
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
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
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 545
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
wrm_appA.qxd 8/31/2005 12:08 PM Page 482
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
The hydrological, geomorphological, environmental and
ecological state of streams and rivers, and of their down-
steam estuaries and coasts, are indicators of past and
current management policies or practices. The condition
of a stream, river, estuary or coast is an indicator of how
well it, as a system, can function. Natural systems can
filter contaminants from runoff; store, absorb and gradu-
ally release floodwaters; serve as habitat for fish and
wildlife; recharge groundwater; provide for commercial
transport of cargo; become sites for hydropower; and
provide recreational opportunities beneficial to humans.
Degraded systems do not perform these functions as well
as non-degraded systems.
Today the importance of keeping natural aquatic
systems alive and well, diverse and productive, in addi-
tion to meeting the needs of multiple economic and social
interests, is much better appreciated and recognized than
it was when water resources planners and managers were
involved in conquering nature and taming its variabili-
ties. Natural system restoration and sustainability have
become major management objectives, along with the
maintenance of the usual economic services that water
483
Natural System Processes
and Interactions
Understanding the natural processes as well as the economic and social services
or functions that rivers, lakes, reservoirs, wetlands, estuaries and coasts fulfill is
critical to the successful and sustainable management of these hydrological
systems. These natural processes involve numerous physical and biological
interactions that take place among the components of fluvial systems and their
adjacent lands. This appendix briefly views some of the important natural
processes and interactions that occur in watersheds, river basins, estuaries and
coasts. Those who manage them should be aware of these processes and
interactions as they build and use their models to analyse, plan and evaluate
alternative management policies and practices.
Appendix A:
and related land management can provide. Satisfying all
these objectives as far as possible requires an understand-
ing of the basic hydrological, geological, environmental
and ecological interactions that take place in natural
aquatic systems.
This appendix is divided into five main sections. The
next section will focus on rivers. It is followed by sections
describing lakes and reservoirs, wetlands, estuaries, and
finally coasts. Multiple books have been written on each
of these water resources system components. What is
presented in this appendix is thus only an outline of the
main features of these water bodies and how they can be
managed. Its purpose is to introduce some vocabulary
and serve as a primer for those not familiar with this back-
ground information.
2. Rivers
Rivers are driven by hydrological, fluvial, geomorphologic,
water quality and ecological processes that occur over a
range of temporal and spatial scales. The plant and animal
communities in rivers and their floodplains are dependent
upon change: changing flows, moving sediments and
wrm_appA.qxd 8/31/2005 12:08 PM Page 483
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
shifting channels. They depend on inputs of organic
matter from vegetation in the riparian zone. They depend
on the exchange of nutrients, minerals, organic matter
and organisms between the river and its floodplain. This
is provided by variable flows and sediment transport. All
of these factors influence the structural character of the
stream or river and its aquatic and terrestrial ecosystems
their species distribution, diversity and abundance.
Assessments of the effect of human activities on river
systems require indicators relating cause to effect. A
causeeffect chain is illustrated in Figure A.1. Insight into
connections between processes and structures and their
temporal and spatial scales leads to a more integrated
interdisciplinary approach to river system monitoring and
management.
2.1. River Corridor
A river corridor can be viewed as a hierarchical series
of river segments, from upstream headwater streams
to large downstream rivers, as illustrated in Figure A.2.
River corridors include the river channels and the river
margins (the waterland interfaces), and both are influenced
by surface watergroundwater interactions. These environ-
ments are characterized by hydrological, geomorphological,
environmental and ecological interactions. River margin
interactions influence surrounding terrestrial landscapes.
Features that influence the structure and functioning
of river systems occur at various spatial scales. A stream
or river, for example, has an inputoutput relationship
484 Water Resources Systems Planning and Management
with the next higher scale, the stream or river corridor.
This corridor scale, in turn, interacts with the landscape
scale, and so on up the hierarchy. Similarly, because
each larger-scale system contains the smaller scale ones,
the structure and functions of the smaller systems affect
the structure and functions of the larger.
Investigating relationships between structure and scale
is a key first step for planning and designing stream or
river system management plans. Landscape ecologists use
four basic terms to define spatial structure at a particular
scale.
These spatial landscape, component types in river
basins, illustrated in Figure A.3 are:
Matrix. The dominant and interconnected land cover
in the basin.
Patch. A different type of land cover found on smaller
areas within the matrix.
Corridor. A land cover type that links other patches in
the matrix. Typically, a corridor is elongated in shape,
such as a stream or river.
Mosaic. A collection of isolated patches.
The watershed scale that includes the stream corridor is
a common scale of management, since many functions of
the stream corridor are closely tied to drainage patterns.
While the watershed scale is often the focus of river
restoration and water resources management, especially
for non-point pollutant discharge management, the other
spatial scales should also be considered when developing
a stream or river system management policy or plan. The
E
0
2
0
8
0
1
h
functional characteristics
flux of matter
retention
structural characteristics
species diversity
gradients and
zonations in species
hydrology
geomorphology
water quality
abiotic system
biotic system
river system
land use emissions
river regulation
dams, levees
waste water discharges
human system
Figure A.1. Cause and effect
chain of factors influencing a
river system.
wrm_appA.qxd 8/31/2005 12:08 PM Page 484
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 485
exclusive use of watersheds for the large-scale manage-
ment of stream corridors, however, ignores the materials,
energy and organisms that move across and through
landscapes independent of water drainage. A more
complete large-scale perspective of the stream and
river system management is achieved when watershed
hydrology is combined with landscape ecology and when
actions in problem sheds rather than only in drainage
basins are being considered.
2.1.1. Stream Channel Structure Equilibrium
Nearly all channels are formed, maintained and altered by
flows and sediment loads. Channel equilibrium involves
the relation among four basic factors: sediment discharge,
Q
s
; sediment particle size, D; streamflow, Q
w
; and stream
slope, S. Lane (1955), using median particle size, D
50
,
expressed this relationship qualitatively as:
Q
s
D
50
Q
w
S (A.1)
reach scale
stream corridor scal
e
stream scale
landscape scal
e
region scale
watershed
local region
river watershed
mixed landscape
suburban
agricultural
forest cover
stream corridor
reservoir
watershed
reach
E
0
2
0
8
0
1
j
Figure A.2. River basin components
viewed at multiple spatial scales,
from the large regional scale to the
local stream segment scale.
wrm_appA.qxd 8/31/2005 12:08 PM Page 485
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
This relationship states that a measure of sediment load
(sediment discharge Q
s
times median particle size D
50
) is
proportional to a measure of the streamflow power
(streamflow Q
w
times slope S). Channel equilibrium
occurs when the streamflow power is constant over the
length of the stream. If this occurs, no net changes in
the channel shape will occur. If a change occurs in either
the left or right-hand side of Equation A.1, the balance and
hence equilibrium will be temporarily lost. If one variable
changes, one or more of the other variables must change
appropriately if equilibrium is to be maintained. Reaching
equilibrium typically involves erosion and/or deposition.
Assuming increasing flows from runoff in the down-
stream direction, the channel slope has to be decreasing
in the downstream direction. If the slope is too steep,
sediment is deposited to reduce that steepness. This is
why stream channels that experience increasing down-
stream flows have decreasing slopes in the downstream
direction.
If streams in equilibrium have constant streamflow
power, Q
w
S, over distance, from Equation A.1, the
sediment load, Q
s
D
50
, must also be constant. Hence,
if sediment deposition is occurring in the downstream
direction to decrease stream slopes, the median particle
size, D
50
, will be decreasing and the sediment discharge,
Q
s
, along with streamflow, Q
w
, will be increasing. This
is typically observed in channels with increasing
downstream streamflows.
A stream seeking a new equilibrium tends to erode
more sediment and larger particle sizes. Alluvial streams
that are free to adjust to changes in these four variables
486 Water Resources Systems Planning and Management
generally do so and re-establish new equilibrium condi-
tions. Non-alluvial streams such as those flowing over
bedrock or in artificial, concrete channels are unable to
maintain this equilibrium relationship because of their
inability to pick up additional sediment.
The stream balance expressed in Equation A.1 can be
used to make qualitative predictions about the impacts
of changes in runoff or sediment loads from a watershed.
Quantitative predictions, however, require the use of
more complex simulation or physical models.
2.1.2. Lateral Structure of Stream or River Corridors
Stream and river valleys are created over time by the stream
or river depositing sediment as it moves back and forth
across the valley floor. These processes of lateral migration
and sediment deposition, usually occurring during flood
flows, continually modify the floodplain. Through time, as
the channel migrates, it will maintain the same average size
and shape as long as the channel stays in equilibrium.
One can distinguish two types of floodplains. The
hydrological floodplain is the land adjacent to the base-
flow channel residing below bank-full elevation. It is
inundated about two years out of three. Not every stream
corridor has a hydrological floodplain. The topographic
floodplain is the land adjacent to the channel, including
the hydrological floodplain, that is flooded by a flood
peak of a given frequency (for example, the 100-year
flood the flood that is equalled or exceeded once every
100 years on average defines the 100-year floodplain).
Higher flood-peak flow return periods define wider
topographic floodplains. These two types of floodplains
are shown in Figure A.4.
bankfull
elevation
topographic floodplain
hydrological floodplain
E
0
2
0
8
0
1
m
Figure A.4. Two types of floodplains, the hydrological and
topographic.
Figure A.3. River basin landscapes made up of matrix, patch,
corridor and mosaic components at various scales.
E
0
2
0
8
0
1
k
matrix
patch
patch
matrix
matrix
mosiac
patch
patch
patch
wrm_appA.qxd 8/31/2005 12:08 PM Page 486
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 487
Floodplains provide temporary storage space for
floodwaters and sediment. This lengthens the lag time
of a flood. This lag time is the time between the middle of
the rainfall event and the runoff peak. If a streams
capacity for moving water and sediment is diminished, or
if the sediment loads become too great for the stream to
transport, the valley floor will begin to fill.
Topographic features on the floodplain, as illustrated
in Figure A.5, include:
Meander scroll. A sediment formation marking former
channel locations.
Chute. A new channel formed across the base of a
meander. As it grows in size, it carries more of the flow.
Oxbow. A severed meander after a chute is formed.
Clay plug. A soil deposit at the intersection of an
oxbow and the new main channel.
Oxbow lake. A water body created after clay plugs
separate the oxbow from the main channel.
Natural levees. Formations built up along the bank
of some streams that flood. As sediment-laden water
spills over the bank, the sudden loss of depth and
velocity causes coarser-sized sediment to drop out of
suspension and collect along the edge of the stream.
Splays. Delta-shaped deposits of coarser sediments
that occur when a natural levee is breached. Natural
levees and splays can prevent floodwaters from return-
ing to the channel when floodwaters recede.
Backswamps. A term used to describe floodplain
wetlands formed by natural levees.
These different features provide a variety of habitats for
plants and animals.
2.1.3. Longitudinal Structure of Stream or River
Corridors
The processes that determine the characteristic lateral
structure of a stream corridor also influence its longitudi-
nal structure. For streams and rivers whose flows increase
with distance downstream, channel width and depth
also increase downstream due to increasing drainage area
and discharge. Even among different types of streams, a
common sequence of structural changes, as shown in
Figure A.6, is observable from headwaters to mouth.
The longitudinal profile of many streams can be divided
into three zones. The changes in the three zones are
characterized in Figure A.6. Zone 1, the headwater zone,
has the steepest slopes. In this zone sediments erode from
slopes of the watershed and move downstream. The rivers
in hilly regions are characterized by the swiftness of the
flow in restricted and/or steep channels, the occurrence of
landslides and the formation of rapids along their courses.
The control of rivers in the upper reaches is known as
torrent control. Zone 2, the transfer zone, receives some of
these sediments and hence is usually characterized by
wider floodplains and more meandering channel patterns.
The flatter slopes in zone 3 receive most of the coarser sed-
iments. River training methods are often adopted for
managing alluvial rivers in this most downstream zone.
A
C
B
D
F
E
E
G
H
F splay -
G
H
backswamp
natural levee
-
-
A oxbow lake -
B chute -
E meander
scrolls
-
C clay plug -
D oxbow -
E
0
2
0
8
0
1
n
Figure A.5. Topographic
features of a meandering
stream on a floodplain.
wrm_appA.qxd 8/31/2005 12:08 PM Page 487
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Though Figure A.6 displays headwaters as mountain
streams, these general patterns and changes apply to
watersheds with relatively small topographic relief from
the headwaters to the mouth. Erosion and deposition
occur in all zones, but the zone concept focuses on the
most dominant process.
2.2. Drainage Patterns
One distinctive aspect of a watershed or river basin when
observed from above (birds-eye view) is its drainage
pattern. Drainage patterns are primarily controlled by
topography and geologic structure. Figure A.7 shows a
method of classifying, or ordering, the hierarchy of natural
channels within a watershed or basin. This is a modified
method based on the one proposed by Horton (1945).
488 Water Resources Systems Planning and Management
The uppermost channels in a drainage network (head-
water channels with no upstream tributaries) are desig-
nated as first-order streams down to their first confluence.
A second-order stream is formed below the confluence of
two first-order channels. Third-order streams are created
when two second-order channels join, and so on. The
intersection of a channel with another channel of lower
order does not raise the order of the stream below the
intersection, e.g. a fourth-order stream intersecting with,
a second-order stream is still a fourth-order stream below
the intersection.
Within a given drainage basin, stream order correlates
well with other basin parameters, such as drainage area or
channel length. Consequently, knowing what order a stream
is can provide clues to other characteristics, such as its
longitudinal zone and its relative channel size and depth.
(~downstream distance
drainage area
2
)
i
n
c
r
e
a
s
e
headwater transfer deposition
b
e
d
m
a
t
e
r
i
a
l
g
r
a
i
n
s
i
z
e
s
l
o
p
e
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
t
r
e
a
m
d
is
c
h
a
r
g
e
channel width
channel depth
mean flow velocity
r
e
l
a
t
i
v
e
v
o
l
u
m
e
o
f
s
t
o
r
e
d
a
l
l
u
v
iu
m
E
0
2
0
4
2
2
e
zone 1
zone 2
zone 3
Figure A.6. Typical changes in the
stream channel characteristics
along its length.
wrm_appA.qxd 8/31/2005 12:08 PM Page 488
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 489
2.2.1. Sinuosity
Sinuosity (Figure A.8) is a term indicating the amount of
curvature in the channel. The sinuosity of a reach is its
channel centreline length divided by the length of the
valley centreline. If the channel-length/valley-length ratio
is more than about 1.4, the stream can be considered
meandering in form. Sinuosity is generally related to
streamflow power (slope times flow). Low to moderate
levels of sinuosity are typically found in zones 1 and 2 of
the longitudinal profile (Figure A.6). Sinuous streams
often occur in the broad, flat valleys of zone 3.
2.2.2. Pools and Riffles
Most streams share a similar attribute of alternating, regu-
larly spaced, deep and shallow areas called pools and riffles
(Figure A.8). Pools and riffles are associated with the
deepest path along the channel (thalweg). This deepest
path meanders within the channel. Pools typically form in
the thalweg near the outside bank of bends. Riffle areas
usually form between two bends at the point where the
thalweg crosses over from one side of the channel to
the other. The pool-to-pool or riffle-to-riffle spacing, where
they exist, is normally about five to seven times the chan-
nel width at bank-full discharge (Leopold et al., 1964).
2.3. Vegetation in the Stream and River
Corridors
Vegetation typically varies along and across stream and
river corridors. In zone 1, flood-dependent or tolerant
plant communities tend to be limited but provide veget-
ative organic matter along with the sediment to zones 2
and 3 downstream. Woody debris from headwaters
forests can be among the important features supporting
food chains and instream ecological habitats in rivers
(Maser and Sedell, 1994).
Zone 2 is typically a wider and more complex floodplain
and larger channel than zone 1. The lower gradient, larger
stream size and less steep terrain in zone 2 often allow more
agricultural or residential development. This development
may restrict the diversity of the natural plant communities
in the middle and lower reaches, especially when land uses
involve clearing and narrowing the corridor. Such actions
alter stream processes involving flooding, erosion/deposi-
tion, and import or export of organic matter and sediment.
This affects stream corridor geomorphology, habitat
diversity, and water quantity and quality regimes.
The lower gradient, increased sediment deposition,
broader floodplains and greater water volume in zone 3
typically lead to different plant communities than those
found in the upstream zones. Large floodplain wetlands
can develop on the flatter terrain.
The changing sequence of plant communities along
streams from source to mouth is an important source of
1
2
3
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
3
3
4
E
0
2
0
8
0
1
o
4
2
2
4
Figure A.7. Stream ordering in a drainage network showing
first-order streams down to fourth-order streams.
sinuous
straight
pool
thalweg line
riffle or cross over
E
0
2
0
8
0
1
p
Figure A.8. Sequence of pools and riffles in straight and
sinuous streams.
wrm_appA.qxd 8/31/2005 12:08 PM Page 489
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
ecosystem resiliency. A continuous corridor of native
plant communities is desirable. Restoring vegetative
connectivity in even a portion of a stream will usually
improve conditions and enhance the ecosystems beneficial
functions.
2.4. The River Continuum Concept
The river continuum concept identifies biological connec-
tions between the watershed, floodplain and perennial
stream systems, and offers an explanation of how biolog-
ical communities develop and change along perennial
stream or river corridors. The concept is only a hypo-
thesis, yet it has served as a useful conceptual model for
describing some of the important features of perennial
streams and rivers.
The river continuum concept assumes that, because of
forest shading in many first to third-order headwater
streams, the growth of algae, periphyton and other
aquatic plants is limited. Since energy cannot be created
through photosynthesis (autotrophic production),
aquatic organisms in these lower order streams depend
on materials coming from outside the channel such as
leaves and twigs. Consequently, these headwater streams
are considered heterotrophic (that is, dependent on
the energy produced in the surrounding watershed). The
relatively constant temperature regimes of these streams
tend to limit biological species diversity.
Proceeding downstream to fourth, fifth and sixth-
order streams, the channel widens, which increases the
amount of incident sunlight and average temperatures.
Primary production increases as a response. This shifts
many stream organisms to internal autotrophic produc-
tion and a dependence on materials coming from inside
the channel (Minshall et al., 1985). Species richness of the
invertebrate community increases due to the increase in
the variety of habitats and food resources. Invertebrate
functional groups, such as the grazers and collectors,
increase as they adapt to both out-of-channel and in-
channel sources of food.
Mid-ordered streams also experience increasing
temperature fluctuations. This tends to further increase
biotic diversity. Larger streams and rivers of seventh to
twelfth order tend to increase in physical stability, but
undergo significant changes in structure and biological
function. Larger streams develop increased reliance on
490 Water Resources Systems Planning and Management
primary productivity by phytoplankton, but continue to
receive heavy inputs of dissolved and fine organic parti-
cles from upstream.
Large streams frequently carry increased loads of
clays and fine silts. These materials increase turbidity,
decrease light penetration, and thus increase the signifi-
cance of heterotrophic processes. The frequency and
magnitude of temperature changes decrease as stream-
flows increase, and this in turn increases the overall
physical stability of the stream as well as species compe-
tition and predation.
2.5. Ecological Impacts of Flow
Streamflow regimes have a major influence on the physi-
cal and biological processes that determine the structure
and dynamics of stream ecosystems (Covich, 1993). High
flows are important in terms of sediment transport. They
also serve to reconnect floodplain wetlands to the
channel. Floodplain wetlands provide habitat for aquatic
plants as well as fish and waterfowl. Low flows promote
fauna dispersment, thus spreading populations of species
to a variety of locations. The life cycles of many riverine
species require an array of different habitat types, whose
temporal availability is determined by the variable flow
regime. Adaptation to this environment allows riverine
species to persist during periods of droughts and floods
(Poff et al., 1997).
2.6. Geomorphology
The major large-scale physical characteristics of most
streams and rivers have resulted from hydro-geological
processes that have occurred over periods ranging from
several decades to hundreds or even thousands of years. At
smaller spatial scales, interactions between the hydrological
cycle and the land that can alter the geometry of these water
bodies take place in much shorter times, possibly ranging
from a few minutes, hours or days to several years. Land use
changes also influence the shape and flow directions of
streams, rivers, estuaries and coastlines. Understanding how
this happens requires a knowledge of how land cover and
land-cover changes influence the partitioning of precipita-
tion into runoff, infiltration, soil water storage, groundwater
flow and storage, and discharge, and what impact all this
has on the erosion, transport and deposition of sediment.
wrm_appA.qxd 8/31/2005 12:08 PM Page 490
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 491
2.6.1. Channel Classification
Streams and rivers are classified on the basis of their
dimensions, configurations and shapes. Classifying chan-
nel morphology helps in the development of reproducible
descriptions and assessments of channel flows, movement
and sediment transport potential. Rosgen (1996) classi-
fied stream or river morphology based on bank-full
width, channel sinuosity, slope and channel material size.
While classifying a stream or river channel is helpful in
understanding the behaviour of the channel and how
the behaviour might change due to land use changes
or modifications to the channel, it is only an aid in the
management of a stream or river or in the development of
an engineering design for its channel. Classification itself
does not directly provide the detailed information needed
for an engineering design solution.
The ratio of the flood-prone area width to the
bank-full width is called the entrenchment ratio. The
width-to-depth ratio is the ratio of the bank-full width to
the mean bank-full depth. The mean hydraulic depth is the
cross-sectional area (darker blue area marked as depth in
Figure A.9) divided by the wetted perimeter.
2.6.2. Channel Sediment Transport and Deposition
The energy contained in flowing water is typically
expended by eroding and transporting sediment. The
source of sediment comes in the runoff from the drainage
area or from the channel bed and banks (Trimble and
Crosson, 2000). The sediment transported in water can
be dissolved, suspended and pushed along the bed by
saltation and traction. The latter two processes form the
bedload. Figure A.10 illustrates these types of sediment
loads in a river channel.
Sediments range from clays to gravel and, in extreme
events, even boulders. They include clay (0.004 mm),
silt (0.0040.062 mm), sand (0.0621.000 mm), gravel
(granules and pebbles, 164 mm), cobbles (64250 mm),
and boulders (250 mm). Each of these size classes can
be subdivided from fine to coarse. In addition, fine
sediments can be cohesive (tending to attach to other
sediments) or non-cohesive.
Energy is required to erode and transport sediment.
The heavier the sediment particles, the more energy
required to erode and transport them. The energy avail-
able to do this is a function of the rate of flow. Figure A.11
E
0
2
0
9
1
6
a
channel bankfull width
depth
bank
bank
flood-prone area width
wetted perimeter
Figure A.9. Channel cross-section
schematic.
flow direction
dissolved
load
suspended
load
saltation
traction
river bed
bed load
E
0
3
0
5
2
6
a
Figure A.10. Types of sediment loads in a river channel.
wrm_appA.qxd 8/31/2005 12:08 PM Page 491
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
shows the approximate flow rates, in centimetres per
second, needed to transport sediment particles of various
sizes. Note that the ability to erode and transport cohesive
fine sediments is much less than non-cohesive sediments.
If the energy of flowing water exceeds that needed to
carry the existing sediment load, additional erosion can
occur to add to the sediment load. Flowing water will
seek to achieve its equilibrium sediment load. Conversely
if the energy is insufficient to carry the existing load in the
water, some of it will be deposited.
Considerable efforts have been made over the past
decades to estimate, prevent and control soil erosion and
sediment runoff from watersheds. To the extent that these
prevention and control efforts have worked, increased
stream channel and bank erosion have occurred.
Controlling watershed erosion does not reduce the sedi-
ment carrying capacity of flows. If sediment is available in
or on the banks of the stream or river channel, it will
replace what otherwise might have come from watershed
runoff. Especially in urbanizing watersheds, where water-
shed contributions to sediment loads have been substan-
tially reduced, most of the total sediment yield can be due
to channel and bank erosion.
Sediment is transported as bedload and/or as suspended
load. The latter consists of fine materials such as clay, silt
and fine sand that usually make the flow look muddy. Flows
from snow or ice melt tend to carry fine particles of rock that
492 Water Resources Systems Planning and Management
can make the flow look emerald green. Lake Louise in
Alberta, Canada is a classic example of this. If fine materials
are available, the suspended load can be as high as 95% of
the total sediment load carried by the stream or river flow.
This fine material settles in areas of reduced velocity.
Sediment deposition processes can eventually change the
channel dimensions and even its location.
Bedload transport of coarser sediments involves a
combination of sliding, rolling and saltation (bouncing).
These transport processes begin when the flow velocity
reaches a critical value for the particular size of the bed
material. This critical velocity corresponds to a critical
shear stress. The shear stress associated with a flow is an
important parameter for sediment transport.
Suspended sediment particle settling (or fall) velocities
influence the rate of deposition. These velocities are
affected by particle shapes, the density or specific weight of
the sediment particles relative to that of the water, and the
chemical attraction (cohesion) of the particles, especially in
clays. Equations have been proposed for estimating the ter-
minal fall velocity of a single particle in quiescent, distilled
water. While equations and graphs exist in handbooks for
these and other individual particle size fall velocities under
quiescent distilled conditions, measurements of fall veloci-
ties under natural conditions are much more reliable.
Erosion depends on tractive stress or shear stress that
creates lift and drag forces at the soil surface boundaries in
v
e
l
o
c
i
t
y

(
c
m
/
s
)
grain size (mm)
1000
800
500
200
100
50
20
10
5
2
1
0.001 0.005 0.02 0.05 0.1 0.2 0.5 1 2 10 20 50 100 5
clay silt
fine
sand
medium
to coarse
sand granules pebbles
c
o
h
e
s
i
v
e
c
l
a
y
a
n
d
s
i
l
t
non-cohesive fine clay and silt
erosion of particles from bed
t
r
a
n
s
i
t
i
o
n
z
o
n
e
sedimentation of particles onto bed
cobbles and boulders
E
0
3
0
5
2
6
b
Figure A.11. Relation between
flow rate and sediment particle
size erosion and transport.
wrm_appA.qxd 8/31/2005 12:08 PM Page 492
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 493
fields and along the bed and banks. Shear stress varies as
a function of flow depth and slope. The larger the soil
particle, the greater the amount of shear stress needed to
dislodge and transport it downstream. The energy differ-
ential that sets sediment particles into motion is created by
faster water flowing in the main body of the channel next
to slower water flowing at the boundaries. The momentum
of the faster water is transmitted to the slower boundary
water. The resulting shear stress moves bed particles in a
rolling motion in the direction of the current.
Sediment transport rates can be computed using
various models, but the uncertainty associated with any
sediment load prediction can be considerable. Predicting
sediment transport is one of the most enduring challenges
facing hydrologists and geomorphologists.
Stream channels and their floodplains are constantly
adjusting to the water and sediment in them. Channel
response to changes in water and sediment yield may
occur at differing times and locations, requiring differing
levels of energy. Daily changes in streamflow and sediment
load result in frequent adjustment of bedforms and rough-
ness in many streams with movable beds. Streams also
adjust periodically to extreme high and low-flow events.
Both flood and drought flows often remove vegetation as
well as creating and increasing vegetative potential along
stream and river corridors. Long-term adjustments in
channel structure and vegetation may come from changes
in runoff or sediment yield from natural causes, such as
climate change or wildfire, or from human activities such
as cultivation, overgrazing or rural-to-urban conversions.
Changes in vegetation can also affect streamflow and
sediment deposition. Bio-geomorphology is the term used
for the study of these plant, soil and water interactions.
2.6.3. Channel Geometry
Like all physical systems, stream and river flows and their
sediment loads will attempt to reach an equilibrium. The
equilibrium between flows and sediment erosion, uptake
and deposition, is sometimes referred to as regime theory.
Leopold and Maddock (1953) derived regime relations in
stream and river channels in alluvial (sedimentary) basins.
These relations predicted the stream width, w; mean
hydraulic depth, h; velocity, U; sediment concentration,
Q
s
; slope, S; and Mannings n friction coefficient as a
power function of the discharge, Q, in the stream or river.
Each of these relations i is of the form a
i
Q
bi
. Leopold and
Maddock found the values of these different coefficients a
i
and b
i
for a variety of rivers in the United States (also in
Richardson et al., 1990).
The values of the coefficients a
i
and b
i
may differ along
the length of the stream or river (Leopold, 1994). In the
downstream direction the relative change in the width
will increase more rapidly than the relative change in
velocity. Hence, the coefficients a
i
and b
i
associated with
width, depth and velocity will change. However, since the
product of width, depth and velocity equals the flow Q,
and each equation a
i
Q
bi
for width, depth and velocity are
functions of the flow Q, both the product of the three
coefficients a
i
and the sum of the three exponent coeffi-
cients b
i
must equal 1.
Other studies have found approximate relationships
between bank-full channel dimensions of alluvial streams
and rivers and their bank-full discharge. The channel
dimensions, pattern and profile are primarily related to
the effective or bank-full discharge. The magnitude of the
bank-full discharge in the main channel typically corre-
sponds to the 1.5- to 2-year expected return period flow
event based on annual peak flows. The bank-full stage is
lower than the top of the bank. It is commonly identified as
a bench, a change in bank material and vegetation, or the
top of a point bar (Rosgen, 1996; Ward and Elliot, 1995).
The loglog plot of Figure A.12 shows a relationship
between flow rates and depths in relation to bank-full
flow rates and depths for thirteen rivers in the eastern
United States (Leopold et al., 1992).
The dimensionless rating curve presented in Figure A.12
shows that the annual discharge is less than the bank-full
flow Q / bankfull Q
B
d
e
p
t
h
D

/

b
a
n
k
f
u
l
l

d
e
p
t
h

D
B 8.00
4.00
2.00
1.00
0.50
0.25
0.125
0.1 10 1.0
mean annual Q
bankfull depth
bankfull flow
1.5 5 10 25 50 100
return period (years)
E
0
2
0
2
2
6
m
Figure A.12. Dimensionless regression curve based on
thirteen gauging stations in the eastern United States
(Leopold, Wolman and Miller, 1992).
wrm_appA.qxd 8/31/2005 12:08 PM Page 493
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
discharge and occurs at a stage of about a third of the
bank-full depth. A 50- to 100-year return period discharge
occurs at a stage that is about double the maximum bank-
full depth. In developing a stream classification system,
Rosgen (1996) defined a flood prone width as one that
occurs at twice the maximum bank-full depth. Therefore, it
might be expected that this depth corresponds to about the
100-year return period discharge.
2.6.4. Channel Cross Sections and Flow Velocities
Typically, flow velocity across a channel will vary from one
bank to another. This lateral variation will change along a
stream or river channel. For example, the velocity near the
outer bank of a meander will be higher than near the inner
bank because water near the outer bank has to travel
further. Also, the velocity decreases towards the bottom or
sides of the channel due to the surface roughness of the
channel. Water in contact with the bottom and sides will be
stationary. These relationships are shown in Figure A.13.
The velocity distribution in a channel can change from
one cross section to the next. Typically, at any cross sec-
tion in a channel, there will be a portion of the flow above
the deepest point (the thalweg) that is moving the fastest.
Along the thalweg the velocity increases in riffles,
decreases in pools, moves from side to side, up and down
with depth, and rotates as it moves around bends. The
494 Water Resources Systems Planning and Management
rotation of the flow on the outer bend might also create
undercut concave banks.
The velocity decreases near the bottom or sides of
the channel due to the roughness of its surface. Water in
contact with the bottom and sides will be stationary. The
changing velocities from riffles to pools will cause turbu-
lence. Also, as water moves through pools, it will push
towards the banks and away from its direction of flow. As
much as a third of the flow might circulate back
upstream.
The bank-full discharge is most effective in forming a
channel, benches (active floodplains), banks and bars. It
is this discharge rate that, as shown in Figure A.14,
transports the largest total sediment load over time. The
sediment concentration in the water is a function of the
flow discharge rate and of course the available sediment.
Low discharge rates are ineffective in transporting
sediment, while high ones have very high sediment
transport rates. However, extremely high discharge events
occur less frequently, so the total sediment load they carry
over a period of many years is not the largest.
A measure of the total sediment load carried by a
particular discharge equals the event frequency multi-
plied by the transport rate for that discharge. The maxi-
mum value of that product for all discharges is called the
effective discharge. This effective discharge is the bank-full
discharge. It has the longest-term impact on an alluvial
streams or rivers equilibrium morphology.
relative channel flow velocities
straight channel
3.0
2.5
2.0
1.5
1.0
0.5
0.0
average channel bend
E
0
2
0
2
2
6
d
Figure A.13. Typical channel relative velocities at bank-full
flows. The velocities at the outside of channel bends (the
right-hand side of the lower channel) are greater than on the
inside of the bends.
E
0
2
0
2
2
6
e
sediment transport rate
frequency of occurrence
effective sediment discharge
maximum effective discharge
stream or river flow
bankfull discharge
Figure A.14. The bank-full discharge typically has the largest
total sediment load.
wrm_appA.qxd 8/31/2005 12:08 PM Page 494
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 495
Discharges larger than the mean annual discharge
typically transport more than 95% of the sediment in a
river system.
2.6.5. Channel Bed Forms
Bed forms result from the interaction of the fluid forces
and sediment particles. Bed forms in alluvial streams and
rivers also depend on channel shape, size of bed material,
bed vegetation and the viscosity of the fluid. In flumes
having a limited range in depth and discharge, changing
the slope is the principal means of changing the bed forms.
In streams and rivers, however, where the slope is
relatively constant, a change in bed form can occur with a
change in discharge and/or a change in viscosity. Viscosity
is affected by a change in sediment size distribution and/or
temperature. An increase in viscosity resulting from a
decrease in temperature or an increase in fine sediments
such as clays can change a dune bed to washed-out dunes,
plane beds or antidunes. For example, the Missouri River
in the United States along the border between Iowa and
Nebraska has temperatures between 21 and 27 C in the
summer and between 15 and 17 C in the autumn. In
summer, its bed form is characterized by dunes. In autumn,
its bed form becomes washed-out dunes (USACE, 1969).
2.6.6. Channel Planforms
Channel planforms are what one sees when looking at
them from above. Their geographic shapes can be
broadly grouped into straight, meandering and braided
planforms (Lagasse et al., 1991; Leopold and Wolman,
1957; Schumm, 1972, 1977; Richardson et al., 1990).
These three types of channels are shown in Figure
A.15. More detailed classifications have been used
(Brice and Blodgett, 1978; Culbertson et al., 1967), but
these three basic types are the ones most commonly
considered.
A meandering stream is characterized by sinuous
S-shaped flow patterns. The sinuosity of a channel of fixed
meandering straight braided
point bar
crossing
a a
b
c
b
c
d d
e e
f
f
a a
b b
c c
d d
e
f
e
f
E021029n
Figure A.15. Types of channel
planforms (after Richardson
et al., 1990).
wrm_appA.qxd 8/31/2005 12:08 PM Page 495
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
length is the ratio between its length and the straight-line
distance between channel end points, or valley length.
Channels can be classified according to their sinuosity. If
the sinuosity is 1.0 the channel is called straight, a condi-
tion that rarely occurs in natural alluvial channels except
over short distances. If the sinuosity is greater than 1.0 the
channel is called sinuous. If the sinuosity is greater than
about 1.4 the river is said to meander, and if the sinuosity
exceeds 2.1 the degree of meandering is tortuous.
Figure A.16 defines some geometric parameters of
meandering channels.
A braided stream or channel consists of a number of
subordinate channels. At normal and low flow rates,
these subordinate channels are separated by bars, sand-
bars or islands. The shape and location of the bars and
islands can change with time, sometimes with each runoff
event. In an anabraided stream or river, the subordinate
channels are more permanent and more widely and dis-
tinctly separated than those of a braided stream.
A stream or river channel can have different planforms at
different locations along its length. Channel planforms can
affect channel dimensions, flows, bed material, floodplain
size and plant cover. Channel planforms in turn are affected
by geology, topography, the drainage area of the contribut-
ing watershed, flow velocity, discharge, sediment transport,
sediment particle distribution, channel geometry, vegetation
cover and any geomorphologic controls on the system.
2.6.7. Anthropogenic Factors
Stream and river morphology is affected by human
activities such as changing land use, bank protection,
navigation, and construction and operation of hydraulic
496 Water Resources Systems Planning and Management
infrastructure. Deforestation and cultivation of watershed
areas, water management initiatives and floodplain devel-
opment affect the runoff and the sediment yield to the
river; they also influence the hydraulic roughness and
the trapping of sediment during floods.
Urbanization increases the fraction of impervious land
in a drainage area. This in turn reduces infiltration and
causes more runoff and higher peak discharges.
Sediment loads typically increase during construction
and decrease following construction. All of these factors
affect the equilibrium in the river and can cause it to
widen and/or deepen. Channel modifications such as
straightening a reach of the channel will also increase
the sediment-carrying capacity of the discharge and
might cause bank and bed scour.
Urbanization increases the frequency as well as the
amounts of water associated with different runoff events
(Figure A.17). Increasing imperviousness increases the
number of expected runoff events each year. The soil
moisture that exists at the beginning of a storm also influ-
ences the frequency of runoff events. In areas of Ohio
in the United States, Ward (personal communication)
found at the start of a storm event in a rural area with
wet soil conditions there might be only one 1-cm runoff
event annually and a 2-cm runoff event every few years.
However, a low-density urban area or an urban area with
wavelength
b
e
l
t

w
i
d
t
h
a
m
p
l
i
t
u
d
e
radius
of curvature
bend length =
0.5 wavelength
E
0
2
0
2
2
6
g
Figure A.16. Geometry of a meandering stream (Rosgen,
1996).
e
x
p
e
c
t
e
d

e
v
e
n
t
s

p
e
r

y
e
a
r
runoff
E
0
2
0
2
2
6
a
urban
urban
rural
(high, wet)
(low, dry)
(wet)
Figure A.17. Expected annual occurrences of runoff
events showing the effect of high and low urban and rural
development densities under wet and dry soil moisture
conditions at the beginning of a storm (Ward, personal
communication).
wrm_appA.qxd 8/31/2005 12:08 PM Page 496
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 497
dry soil conditions at the start of a storm event might on
average experience six or seven 1-cm runoff events, two
or three 2-cm events, one 2.5-cm event per year, and even
an event with 3.5-cm every few years. High-density urban
areas, or urban areas with wet soil conditions at the start
of a storm event will on average experience more than
twenty 1-cm runoff events, six 2-cm events, two 2.5-cm
events, and one 3.5-cm event per year. These were for
this study area in Ohio; for other areas elsewhere the
numbers will differ, but the general relationships shown
in Figure A.17 will apply.
The effect of urbanization on discharges is also
illustrated in Figures A.18 and A.19. Rural areas will have
little if any impervious cover, while urban areas will have
considerably more. The plot shows the relative discharge
(percentage of the rural discharge) increase that occurs as
the percentage of impervious cover increases. Plots are
presented for flow return periods of 2, 10 and 100 years.
Note that the biggest impact is on the smaller, more
frequent storm events. The two-year return period
discharges, Q
2
, in highly urbanized areas can be over two-
and-a-half times larger than those on rural areas. In addi-
tion, as imperviousness increases, so does channel width.
2.7. Water Quality
Establishing an appropriate flow regime in a stream or
river corridor may do little to ensure a healthy ecosystem
if the physical and chemical characteristics of the water
are damaging to that ecosystem. For example, streams or
rivers with high concentrations of toxic materials, high
temperatures, low dissolved oxygen concentrations or
other harmful physical/chemical characteristics cannot
support healthy stream corridor ecosystems.
Figure A.20 illustrates some of the key water quality
processes affecting the oxygen content and hence the biology
of surface waters. (The modelling of these and other chemi-
cal and biological processes is discussed in Chapter 12.)
Dissolved oxygen (DO) is a basic requirement for a
healthy aquatic ecosystem. Most fish and aquatic insects
breathe the oxygen dissolved in the water body. Some
fish and aquatic organisms, such as carp and sludge
worms, are adapted to low oxygen conditions, but most
sport-fish species, such as trout and salmon, suffer when
DO concentrations fall below a concentration of 34 mg/l.
Larvae and juvenile fish are more sensitive and require
even higher concentrations of DO.
Many fish and other aquatic organisms can recover
from short periods of low oxygen concentrations in
the water. However, prolonged episodes of depressed
dissolved oxygen concentrations of 2 mg/l or less can
result in dead (anaerobic) water bodies. Prolonged
exposure to low DO conditions can suffocate adult fish
or reduce their reproductive survival rate by suffocating
sensitive eggs and larvae, or starve fish by killing aquatic
insect larvae and other sources of food. Low DO
concentrations also favour anaerobic bacteria that
produce the noxious gases often associated with polluted
water bodies.
d
i
s
c
h
a
r
g
e

i
n
c
r
e
a
s
e

(
%
)
relative imperviousness
E
0
2
0
2
2
6
b
Q
Q
Q
2
0
00
1
1
Figure A.18. Relative discharge increases, for flows having 2-
10- and 100- year return periods, as a function of degree of
impervious surface area.
channel width increase (%)
relative imperviousness
E
0
2
0
2
2
6
c
100
80
60
40
20
0
Figure A.19. Relative increase in channel width as a function
of impervious surface area.
wrm_appA.qxd 8/31/2005 12:08 PM Page 497
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water absorbs oxygen directly from the atmosphere,
and from plants as a result of photosynthesis. The amount
of oxygen that can be dissolved in water is influenced by
its temperature and salinity. Water loses oxygen primarily
by respiration of aquatic plants and animals, and by the
mineralization of organic matter by microorganisms.
Discharges of oxygen-demanding wastes or excessive
plant growth (eutrophication) induced by nutrient load-
ing followed by death and decomposition of vegetative
material can also deplete oxygen.
In addition to oxygen and water, aquatic plants require a
variety of other elements to support their bodily structures
and metabolism. Just as with terrestrial plants, nitrogen
and phosphorus are important among these elements.
Additional nutrients, such as potassium, iron, selenium and
silica, are also needed by many species but are generally not
limiting factors to plant growth. When any of these elements
498 Water Resources Systems Planning and Management
are limited, plant growth may be limited. This is an impor-
tant consideration in ecosystem management.
Nutrients cycle from one form to another depending
on nutrient inputs, as well as temperature and available
oxygen. The nitrogen cycle is illustrated in Figure A.21 as
an example. Table A.1 lists some common sources of
nitrogen and phosphorus nutrient inputs and their typical
concentration ranges.
Management activities can interact in a variety of
complex ways with water quality. This in turn can affect
ecosystem species, as shown in Figure A.22.
2.8. Aquatic Vegetation and Fauna
Stream biota are often classified in seven groups: bacteria,
algae, macrophytes (higher plants), protists (amoebas,
flagellates, ciliates), microinvertebrates (such as rotifers,
E
0
2
0
8
0
1
q
algae
photosynthesis respiration
reaeration
carbonaceous
deoxygenation
oxygen
demand
NH
+
4
NO
NO
-
-
2
3
dissolved
oxygen
nitrification
Figure A.20. Some of the basic
chemical and biological
processes affecting dissolved
oxygen in waters.
wrm_appA.qxd 8/31/2005 12:08 PM Page 498
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 499
litter input
dissolved organic
nitrogen NH
NO
3
3
biota
cyanobacteria
and microbial
populations
benthic algae
particulate
organic
matter
and
associated
microbes
NH
3
N
2
NH
3
O
NO
3
nitrogen
fixation
nitrogen
fixation
NO
2
NH
3
excretion
decomposition
NH
3
accumulation
decomposition
/excretion
NH
NO
3
3
dissolved
organic
nitrogen
assimilation
N
2
assimilation NO
3
n
i
t
r
i
f
i
c
a
t
i
o
n
d
e
n
i
t
r
i
f
i
c
a
t
i
o
n
o
x
y
g
e
n

c
o
n
c
e
n
t
r
a
t
i
o
n
riparian
vegetation
N
2
import from
upstream
atmospheric N
2
sediment
surface
inter-
stitial
water
groundwater dissolved organic nitrogen NO
3
E
0
2
0
8
0
1
r
downstream export to
Figure A.21. Dynamics and
transformations of nitrogen in a
stream ecosystem.
urban runoff
livestock operations
atmosphere
90 % forest
50 % forest
90 % agriculture
untreated waste water
treated waste water
(wet deposition)
source total
nitrogen (mg/l)
total
phosphorus (mg/l)
3-10
6-800
0.9
0.06-0.19
0.18-0.34
0.77-5.04
35
30
0.2-1.7
4-5
0.015
0.006-0.012
0.013-0.015
0.085-0.104
10
10
E
0
2
0
9
0
3
e Table A.1. Sources and concentrations of
pollutants from common point and non-point
sources (USDA, 1998). These data show little
or no impact on nutrient removals from basic
wastewater treatment facilities.
wrm_appA.qxd 8/31/2005 12:08 PM Page 499
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
copepods, ostracods, and nematodes), macroinvertebrates
(such as mayflies, stoneflies, caddisflies, crayfish, worms,
clams, and snails) and vertebrates (fish, amphibians,
reptiles, and mammals). The river continuum concept
provides a framework for describing how these organisms
change from lower to higher-order streams.
Much of the spatial and temporal variability of type,
growth, survival and reproduction of aquatic organisms
reflects variations in water quality, temperature, stream-
flow and flow velocity, substrate content, the availability
of food and nutrients, and predatorprey relationships.
These factors are often interdependent.
500 Water Resources Systems Planning and Management
2.9. Ecological Connectivity and Width
Healthy ecosystems also depend on conductivity
and width. Connectivity is a measure of how spatially
continuous a corridor or a matrix is (Forman and Godron,
1986). A stream corridor with connections among its
natural communities promotes transport of materials
and energy and movement of flora and fauna.
Connectivity is illustrated in Figure A.23.
Width is the distance across the stream and its zone of
adjacent vegetation cover. Factors affecting width are
edges, community composition, environmental gradients
fine particulate
organic matter
dissolved
organic matter
i
n
v
e
r
t
e
b
rate shre
d
d
e
r
s i
n
v
e
r
t
e
brate scr
a
p
e
r
s
i
n
v
e
r
t
e
b
rate pre
d
a
t
o
r
s v
e
r
t
e
b
rate pred
a
t
o
r
s
i
n
v
e
r
t
e
b
rate co
lle
c
t
o
r
s
m
ic
ro
organism
s
e
p
ilithic alga
e
la
rger plan
ts
m
o
sses, red alg
a
e
c
o
a
r
s
e
p
a
r
t
ic
ulate org
a
n
ic
m
a
t
t
e
r
e
.
g
.
h
y
p
homycete
fu
n
g
i
flocculation
micro organisms
E
0
2
0
8
0
1
s
Figure A.22. Interactions
among aquatic organisms and
their sources of energy.
(Dashed lines reflect weaker
interactions.)
wrm_appA.qxd 8/31/2005 12:08 PM Page 500
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 501
and disturbance effects of adjacent ecosystems, including
those due to human activity. Width and connectivity
interact throughout the length of a stream corridor.
Corridor width varies along a stream and may have gaps.
Gaps across the corridor can interrupt and reduce
connectivity. Ensuring connectivity and adequate width
can provide some of the most useful ways to mitigate
disturbances.
2.10. Dynamic Equilibrium
In constantly changing ecosystems like stream and river
corridors, the stability of a system is its ability to persist
within a range of conditions. Within this range the system
is resilient. This phenomenon is referred to as dynamic
equilibrium.
The maintenance of dynamic equilibrium requires a
series of self-correcting mechanisms in the stream
corridor ecosystem. These mechanisms control the
responses to external stresses or disturbances within
certain ranges. The threshold levels associated with these
ranges are often difficult to identify and quantify. If they
are exceeded, the system can become unstable. Corridors
may then undergo a change toward a new steady-state
condition, usually after some time for readjustment has
occurred.
Many stream systems can accommodate some distur-
bances and still return to functional conditions once the
sources of the disturbances are removed. Ecosystems tend
to heal themselves when external stresses are removed.
The time it takes to do this, of course, depends on the
level of stress.
2.11. Restoring Degraded Aquatic Systems
Some principles found useful in the management of
restoration activities in degraded stream or river systems
are listed below. These principles apply from early
planning to post-implementation monitoring. They focus
on scientific and technical measures and their likely
impacts. Restoration activities can include upland
best-management practices for agriculture, stream channel
restoration, removal of exotic species and increasing native
plant cover, establishing windbreaks and shelterbelts,
improving upland corridors and riparian habitat, and
wetland enhancement for water quality improvements. As
in all management activities, the presence or absence of
public support for a restoration project can make the
difference between success and failure.
Preserve and Protect Aquatic Resources
Existing ecosystems that are relatively intact provide the
natural materials needed for the recovery of impaired
systems. It is not usually necessary to import species.
Restore Ecological Integrity
Restoration should re-establish the ecological integrity of
degraded aquatic ecosystems. Ecological integrity refers
to the natural structure, composition and processes of
biotic communities and the physical environment. Key
processes, such as nutrient cycles, succession, water
levels and flow patterns, and the dynamics of sediment
erosion and deposition, function within the natural range
of variability. Restoration strives toward ecological
integrity by taking actions that favour the desired natural
processes and communities that can be sustained
through time.
Restore Natural Structure and Function
Many aquatic resources in need of restoration have
problems caused by past changes in channel form or
other physical characteristics that led to problems such as
habitat degradation, changes in flow regimes and
siltation. Stream channelization, ditching in wetlands,
disconnection from adjacent ecosystems and shoreline
modifications are examples of such adverse changes.
E
0
2
0
8
0
1
t
Figure A.23. Landscapes with high (left picture) and low (right
picture) degrees of connectivity.
wrm_appA.qxd 8/31/2005 12:08 PM Page 501
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Structure and function are closely linked in river
corridors, lakes, wetlands, estuaries and other aquatic
resources. Re-establishing the appropriate natural struc-
ture can bring back beneficial functions. For example,
restoring the bottom elevation in a wetland can help
re-establish the hydrological regime, natural disturbance
cycles and nutrient fluxes. Monitoring the extent to which
desired functions have been re-established can be a good
way to determine the effectiveness of a restoration project.
Work Within the Watershed and Broader Landscape
Context
Restoration requires a design based on the entire watershed
or river basin, not just the part of it that is degraded. A local-
ized restoration project may not be able to change what goes
on in the whole watershed, but it can be designed to accom-
modate watershed effects better. New and future urban
development may, for example, increase runoff volumes,
streambed down-cutting, bank erosion and pollutant load-
ing. Restoration may help mitigate these adverse effects. For
example, in choosing a site for a wetland, stream or river
restoration project, planners should consider how the
proposed project may be used to further related efforts in
the watershed, such as increasing riparian habitat continu-
ity, reducing flooding and/or enhancing downstream water
quality. Beyond the watershed, the broader landscape
context also influences restoration through factors such as
interactions with terrestrial habitats in adjacent watersheds,
or the deposition of airborne pollutants from other regions.
Understand the Natural Potential of the Watershed
A watershed has the capacity to become only what its
physical and biological setting its climate, geology,
hydrology and biological characteristics will support.
Establishing restoration goals for a water body requires
knowledge of the natural range of conditions that existed
on the site prior to degradation and of what future condi-
tions might be. This information can then be used in
determining appropriate goals for the restoration project.
Address Ongoing Causes of Degradation
Restoration efforts are likely to fail if the causes of
degradation persist. Therefore, it is essential to identify the
causes of degradation and eliminate or mitigate them
502 Water Resources Systems Planning and Management
wherever possible. Degradation can be caused by one
event, such as the filling of a wetland, or it can be caused
by the cumulative effect of numerous events, such as
gradual increases in the amount of impervious surfaces in
the watershed that alter the streamflow regime. In identi-
fying the sources of degradation, it is important to look at
upstream and upslope activities as well as at direct impacts
on the immediate site where damage is evident. In some
situations, it may also be necessary to consider down-
stream modifications such as dams and channelization.
Develop Clear, Achievable and Measurable Goals
Goals direct implementation and provide the standards
for measuring success. The chosen goals should be
achievable given the natural potential of the area and the
available resources and the extent of community support
for restoration. Goals provide focus and increase project
efficiency. They can also change (adapt) over time.
Focus on Feasibility
Particularly in the planning stage, it is critical to focus on
whether any proposed restoration activity is feasible, taking
into account scientific, financial, social and other consider-
ations. Community support for a project is needed to ensure
its long-term viability. Ecological feasibility is also critical.
For example, a wetland, stream or river restoration project
is not likely to succeed if the hydrological regime that
existed prior to degradation cannot be re-established.
Use a Reference Site
Reference sites are areas that are comparable in structure
and function to the proposed restoration site before it
was degraded. They may be used to identify targets for
restoration projects, and as yardsticks for measuring
the progress of the project. While it is possible to use
historic information on sites that have been altered or
destroyed, it may be most useful to identify an existing,
relatively healthy, similar site as a benchmark.
Anticipate Future Changes
Although it is impossible to plan for the future precisely,
foreseeable ecological and societal changes can and
should be factored into restoration design. For example,
wrm_appA.qxd 8/31/2005 12:08 PM Page 502
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 503
in repairing a stream channel, it is important to take into
account potential changes in runoff resulting from
projected increases in upstream impervious surface area
due to development. In addition to potential impacts
from changes in watershed land use, natural changes such
as plant community succession can also influence restora-
tion. Long-term, post-project monitoring should take into
account successional processes in a stream corridor when
evaluating the outcome of the restoration project.
Involve the Skills and Insights of a Multidisciplinary
Team
Restoration can be a complex undertaking that integrates
a wide range of disciplines, including ecology, aquatic
biology, hydrology and hydraulics, geomorphology, engi-
neering, planning, communications and social science.
The planning and implementation of a restoration project
should involve people with experience in the disciplines
needed for that particular scheme. Complex restoration
projects require effective leadership to bring the various
disciplines, viewpoints and styles together as an effective
team.
Design for Self-Sustainability
Perhaps the best way to ensure the long-term viability of
a restored area is to minimize the need for continuous
operation, maintenance and repair costs, vegetation
management or frequent repair of damage done by
high-water events. High-maintenance approaches make
long-term success dependent upon human and financial
resources that may not always be available. In addition to
limiting the need for maintenance, designing for self-
sustainability also involves favouring ecological integrity.
An ecosystem in good condition is more likely to have the
ability to adapt to changes.
Use Passive Restoration, When Appropriate
Before actively altering a restoration site, determine
whether simply reducing or eliminating the sources of
degradation and allowing time for recovery will be
enough to allow the site to regenerate naturally. There are
often reasons for restoring a water body as quickly as
possible, but there are other situations when immediate
results are not critical. For some rivers and streams,
restoring the original hydrological regime may be enough
to let time re-establish the native plant community, with
its associated habitat value.
Restore Native Species and Avoid Non-Native
Species
Many invasive species out-compete native species because
they are expert colonizers of disturbed areas and lack
natural controls. The temporary disturbance present
during restoration projects invites colonization by
invasive species that, once established, can undermine
restoration efforts and lead to further spread of these
invasive species. Special attention should be given to
avoiding the unintentional introduction of non-native
species at the restoration site when the site is most
vulnerable to invasion. In some cases, removal of non-
native species may be among the primary goals of the
restoration project.
Use Natural Fixes and Bioengineering Techniques,
Where Possible
Bioengineering is a method of construction combining live
and dead plants or inorganic materials, to produce living,
functioning systems to prevent erosion, control sediment
and other pollutants, and provide habitat. Bioengineering
techniques can often be successful for erosion control and
bank stabilization, flood mitigation and even water
treatment. Specific projects can range from the creation of
wetland systems for the treatment of stormwater to the
restoration of vegetation on riverbanks to enhance natural
decontamination of runoff before it enters the river.
Monitor and Adapt Where Changes
are Necessary
Every combination of watershed characteristics, sources of
stress, and restoration techniques is unique and, therefore,
restoration efforts may not proceed exactly as planned.
Adapting a project to at least some change or new infor-
mation should be considered normal. Monitoring before
and during the work is crucial for finding out whether
goals are being achieved. If they are not, adjustments
should be undertaken. Post-project monitoring will help
determine whether additional actions or adjustments are
needed, and can provide useful information for future
wrm_appA.qxd 8/31/2005 12:08 PM Page 503
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
restoration efforts. This process of monitoring and
adjustment is known as adaptive implementation or
adaptive management (Appendix B). Monitoring plans
should be feasible in terms of costs and technology,
and should provide information relevant to meeting the
project goals.
3. Lakes and Reservoirs
Lakes and reservoirs are components of many river
systems. They are typically dramatic and visually pleasing
features of a watershed or basin. They range from
pond-sized water bodies to lakes stretching for hundreds
of kilometres. Referred to by some as pearls on a river,
lakes and reservoirs can have a significant effect on the
quantity and quality of the freshwater that eventually
reaches the oceans.
Seen from the shoreline, a large natural lake looks
much the same as a large artificial reservoir, and both
often contain the word lake in their name. Furthermore,
the same principles of biology, chemistry and physics
apply to both. Indeed, it may be difficult to discern any
obvious differences between a lake and reservoir, but
differences as well as similarities exist.
Lakes are water bodies formed by nature whereas
reservoirs are artificial ones constructed by humans,
either by damming a flowing river or by diverting water
from a river to an artificial basin (impoundment). Some
reservoirs are made by increasing the capacity of natural
lakes. Many characteristics of lakes and reservoirs are a
function of the way in which they were formed and how
humans use their waters.
Lakes and reservoirs are important sources of fresh-
water for agriculture, industry and municipalities. At the
same time, they provide habitats for a variety of species of
plants and animals. They are the sources of fish, areas for
migratory birds to feed, reproduce or rest, and places we
all go for enjoyment and recreation. Considerable money
as well as technical and scientific expertise is often
required to keep them clean and healthy.
As human populations grow, greater demands are
placed on the services lakes and reservoirs provide. The
water levels of many lakes and reservoirs have become
consistently lower as a result of higher consumption
by upstream agriculture, households and industries.
Increasing numbers of people enjoy the recreational
504 Water Resources Systems Planning and Management
activities these bodies can support, but this alters their
shores, changes the surrounding land use and cover, and
increases the amount of soil or sediments as well as
nutrients reaching their waters. Finally, pollution from
adjacent lands and various point sources may increase
eutrophication processes and produce other non-desirable
effects such as increased concentrations of toxic algae,
reduced dissolved oxygen and generation of foul odours.
3.1. Natural Lakes
Lakes are naturally-formed, usually bowl-shaped, depres-
sions in the land surface that have filled with water over
time. These depressions were typically produced as a result
of glaciers, volcanic activity or tectonic movements. The
age of most permanent lakes is usually of a geological time
frame. Some ancient lakes may be millions of years old.
The most significant past mechanism for the formation
of lakes in temperate areas was the natural process of
glacial scour, in which the slow movement of massive
volumes of glacial ice during and after the Ice Age
produced depressions in the land surface that subse-
quently filled with water. The North American Great
Lakes (Superior, Michigan, Huron, Erie and Ontario),
lakes in the Lake District of the United Kingdom, and
the numerous lakes in Scandinavia and Argentina are
prominent examples of this type of lake formation. Some
smaller kettle lakes, as found on Cape Cod in
Massachusetts, for example, were formed by the deposi-
tion and subsequent melting of glacial ice blocks.
Another major lake formation process was tectonic
movement, in which slow movements of the earths crust
gradually produced depressions that were subsequently
filled with water. Lake basins also formed as a result of
volcanic activity, which also produced depressions in the
land surface. Most of the earths very deep lakes resulted
from either volcanic or tectonic activity. Lake Baikal in
Russia, the worlds deepest lake, which contains approxi-
mately 20% of the worlds liquid freshwater, and the
African Rift Valley lakes are prominent examples of this
tectonic type of lake formation.
Other natural processes that produced lake basins
include seepage of water down through layers of soluble
rock, erosion of the land surface by wind action, and
plant growth or animal activity (such as beaver dams) that
resulted in blocking the outlet channels from shallow
depressions in the land surface.
wrm_appA.qxd 8/31/2005 12:08 PM Page 504
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 505
There are literally millions of small lakes around
the world, concentrated largely in the temperate and
sub-arctic regions. These regions are also characterized
by a relative abundance of freshwater. Many more
intermittent lakes occur in semi-arid and arid regions.
3.2. Constructed Reservoirs
In contrast to natural processes of lake formation, reservoirs
are water bodies that are usually formed by constructing a
dam across a flowing river. A dam may sometimes also
be constructed on the outlet channel of a natural lake as a
means of providing better control of the lakes water level
(examples include Lake Victoria in Africa, and Lake Tahoe
in the United States). However, these latter water bodies
typically retain their natural lake characteristics.
Reservoirs are found primarily in areas with relatively
few natural lakes, or where the lakes do not satisfy human
water needs. Reservoirs are much younger than lakes, with
life spans expressed in terms of historical rather than geo-
logical time. Although lakes are used for many of the same
purposes as reservoirs, a distinct feature of the latter is that
they are usually built to address specific water needs. These
include municipal and drinking water supplies, agricultural
irrigation, industrial and cooling water supplies, power
generation, flood control, sport or commercial fisheries,
recreation, aesthetics and/or navigation. Small reservoirs
are sometimes built for fire protection as well.
The reasons for constructing reservoirs are ancient in
origin, and have focused on the need of humans to ensure a
more reliable water supply and to protect themselves during
periods of floods. Accordingly, reservoirs are usually found
in areas of water scarcity, or where a controlled water facil-
ity was necessary. Small reservoirs were first constructed
some 4,000 years ago in China, Egypt and Mesopotamia,
primarily to supply drinking water and for irrigation
purposes. Simple small dams were constructed by blocking
a stream with soil and brush, in much the same manner as
beavers dam a stream. Larger reservoirs were constructed by
damming a natural depression, or by forming a depression
along the river and digging a channel to divert water to it
from the river. Early irrigation practices were linked largely
to land adjacent to streams. They required the construction
of larger dams that allowed humans to impound larger
volumes of water. Later reservoirs were also used as sources
of power, first to drive waterwheels and subsequently to
produce hydroelectric power.
3.3. Physical Characteristics
Like lakes, reservoirs range in size from pond-like to very
large water bodies. The variations in type and shape,
however, are much greater than for lakes. The term
reservoir includes different types of constructed water
bodies and/or water storage facilities. These are (1) valley
reservoirs, created by constructing a barrier (dam) per-
pendicular to a flowing river; and (2) off-river storage
reservoirs, created by constructing an enclosure parallel
to a river, and subsequently supplying it with water either
by gravity or by pumping from the river. The latter reser-
voirs are sometimes also called embankment or bounded
or pumped-storage reservoirs. They have controlled
inflows and outflows to and from one or more rivers. For
example, much of the water in the river above Niagara
Falls between Canada and the United States is diverted to
a pumped storage reservoir during the night and released
through hydroelectric generators during the day when the
energy demand, and hence price, is higher.
In addition to single reservoirs, reservoir systems also
exist. These may be (1) cascade reservoirs, consisting of a
series of reservoirs constructed along a single river; or (2)
inter-basin transfer schemes, designed to move water
through a series of reservoirs, tunnels and/or canals from
one drainage basin to another. These types are illustrated
in Figure A.24.
Much of our current limnological knowledge (includ-
ing that used to manage lakes and reservoirs) has come
from studies of lakes over many decades. Although we
now have a reasonable understanding of physical and
biological processes that take place in lakes, we are
less advanced in our understanding of these processes
in reservoirs. Many early reservoir studies focused on
sediment loading from drainage basins. The rationale for
this was that the rate at which a reservoir filled with
sediment is a major determinant of useful operational life.
Comparatively little attention was given to the environ-
mental and socioeconomic issues associated with
reservoir construction. That situation has changed.
3.3.1. Shape and Morphometry
The shape of lakes and reservoirs is determined largely
by how they were formed. This also affects some of their
fundamental characteristics. Because lakes are naturally-
formed, bowl-shaped depressions typically located in the
wrm_appA.qxd 8/31/2005 12:08 PM Page 505
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
central part of a drainage basin, they usually have a more
rounded shape than reservoirs. As in a bowl, the deepest
part of a lake is usually at its centre. The shallowest part of
the water basin is usually located near the outflow channel.
In contrast, a reservoir often has its deepest part near
the dam. Moreover, because a river often has a number of
streams or tributaries draining into it, when it is dammed
the impounded water tends to back up into the tribu-
taries. As a result, some reservoirs have a characteristic
dendritic shape, with the arms radiating outward
from the main body of the reservoir as illustrated in
Figure A.25. In contrast, a reservoir formed by damming
a river with high banks will tend to be long and narrow.
Depending on how they were constructed, off-river
storage reservoirs can have many shapes.
The dendritic or branching form of many reservoirs
provides a much longer shoreline than associated with
lakes of similar volume. Reservoirs also usually have larger
drainage basins. Because of their larger basins and multi-
ple tributary inputs, the flow of water into reservoirs is
more directly tied to precipitation events in the drainage
basin than it is in lakes. Also, the fact that the deepest parts
of most reservoirs are just upstream of the dam facilitates
the possibilities for draining the reservoir.
Damming a river inundates land previously above
water, and sometimes forces the relocation of inhabitants
and wildlife living around the river. The presence of a
dam downstream also allows a greater degree of control of
water levels and volumes for reservoirs than for lakes.
Constructing water discharge structures at different levels
506 Water Resources Systems Planning and Management
in the dam allows withdrawal or discharge of water from
selected depths in a reservoir. Selective withdrawal
(as shown in Figure A.26) has major implications for the
water-mixing characteristics and increases flushing
possibilities for reservoirs.
3.3.2. Water Quality
The characteristics of river water typically undergo
changes as the water enters the lake or reservoir. Primarily
because of reduced velocities, sediment and other
E
0
2
0
8
0
1
x
Figure A.25. Characteristic dendritic shape of a large
reservoir created by a dam to the right of this figure.
A B C
impoundment
dam
cascade
reservoir
urban
and rural use
E
0
2
0
8
0
1
w
Figure A.24. Types of reservoir
arrangements, ranging from
(left to right) a single reservoir
to a cascade of reservoirs along
a river, to pumped storage
reservoirs adjacent to rivers.
wrm_appA.qxd 8/31/2005 12:08 PM Page 506
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 507
materials carried in the water settle out in the lake or
reservoir. The structure of the biological communities also
changes from organisms suited to living in flowing waters
to those that thrive in standing or pooled waters. Pooled
waters provide greater opportunities for the growth of
algae (phytoplankton) that can lead to eutrophication.
Reservoirs typically receive larger inputs of water, as
well as soil and other materials carried in rivers, than do
lakes. As a result, they usually receive larger pollutant
loads. However, because of greater water inflows, flushing
rates are typically more rapid than in lakes (Figure A.27).
Thus, although reservoirs may receive greater pollutant
loads, they have the potential to flush the pollutants more
rapidly. Reservoirs may therefore exhibit fewer or less
severe negative water quality or biological impacts than
lakes for the same pollutant load.
Although there are many variables of limnological
significance, water quality is typically characterized on the
basis of such variables as water clarity or transparency,
concentration of nutrients and algae, oxygen concentra-
tion, concentration of dissolved minerals and acidity.
Waste chemical compounds from industry, some
with toxic or deleterious effects on humans and/or
water-dependent products, can be part of the pollution
load discharged into lakes and reservoirs. These loads
can kill aquatic organisms and damage irrigated crops.
The quantity of bacteria, viruses and other organisms
in waste-receiving waters are a primary cause of water-
borne disease. Although such organisms present a risk to
human health worldwide, such risks are particularly
severe in developing countries.
There are some major differences between deep and
shallow lakes and reservoirs. Deep water bodies, particularly
in non-tropical regions, usually have better water quality in
their lower layers. Shallow water bodies do not exhibit this
depth differentiation in quality. Their more shallow, shore-
line areas have relatively poorer water quality, because those
areas are where pollutant inputs are discharged and where
there is a greater potential for disturbance of bottom muds.
Thus, the water quality of a natural lake usually improves
from the shoreline to the deeper, central part. In contrast,
the deepest end of a reservoir is immediately upstream of
the dam, so that water quality usually improves along the
length of a reservoir, from the shallow inflow end to the
deeper, lake-like end near the dam.
Reservoirs, particularly the deeper ones, are also
distinguished from lakes by the presence of a longitudinal
gradient in physical, chemical and biological water
quality characteristics from the upstream river-end to
the downstream dam-end. Thus, reservoirs have been
characterized as having three major zones: an upstream
riverine zone, a downstream lake-like zone at the dam-
end, and a transitional zone separating these two zones
(Figure A.28). The relative size and volume of the three
zones can vary greatly in a given reservoir.
3.3.3. Downstream Characteristics
The construction of a dam can change the downstream
river channel below it. Because reservoirs act as sediment
and nutrient traps, the water at the dam-end of a reservoir
is typically of higher quality than the water entering it.
E
0
2
0
8
0
1
y
thermocline
sedim
ent
well-mixed (oxic) layer
anoxic layer
A
B
C
Figure A.26. Multiple outlets from a dam permit some control
of water quality as well as quantity downstream.
rivers lakes
reservoirs
small impoundments
off-river basins
increasing hydraulic residence time
large impoundments
E
0
2
0
8
0
1
z
Figure A.27. A comparison of typical hydraulic residence
times of rivers, reservoirs and lakes.
wrm_appA.qxd 8/31/2005 12:08 PM Page 507
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
This higher-quality water subsequently flows into the
downstream river channel below the dam. The smaller
the quantity of sediments and other materials transported
in the discharged water, the greater the quantity of sedi-
ment that can be picked up and transported as it moves
downstream. This scouring effect can have a negative
impact on the flora, fauna and biological community
structure in the downstream river channel. The removal
and deposition of sediments from floodplains can also
affect the biological character of those floodplains.
Many reservoirs, especially those used for drinking
water supplies, have water release or discharge structures
located at different vertical levels in their dams
(Figure A.26). This allows for the selective withdrawal or
discharge of water from different layers within the reser-
voir. Depending on the quality of the water discharged,
selective withdrawal can affect water quality within
the reservoir itself, as well as the chemical composition
and temperature of the downstream river. The ability
to regulate or schedule water, temperature and silt
discharges can provide some control on the hydrological
regimes, affecting both flora and fauna.
508 Water Resources Systems Planning and Management
Constructing a reservoir to protect downstream areas
from floods often has significant social and economic
implications, including the potential for stimulating
urban and agricultural development adjacent to, and
below, the reservoir. This can have both positive and
negative impacts, depending on the nature and size of
development. Appendix D discusses this aspect of
flood management in more detail.
Table A.2 summarizes some of the major characteris-
tics of lakes and reservoirs.
3.4. Management of Lakes and Reservoirs
Because drainage basins are simultaneously the sources of
water, the places where the water is usually used, and the
points where human activities affect both water quantity
and quality, they are usually considered the logical
management unit for lakes and reservoirs. However, activi-
ties that generate pollutants (such as urbanization, industri-
alization and agricultural production) may occur outside the
affected drainage basin. They still need to be considered
when developing management plans for lakes or reservoirs.
riverine zone transitional zone lacustrine zone
narrow, channelized basin
relatively high flow rates
light-limited primary
productivity
high suspended solids
nutrient supply by
advection
low light availability at
depth
relatively high nutrient
levels
cell losses primarily
by sedimentation
organic matter sypply
primarily allochthonous
more eutrophic
broader, deeper basin
reduced flow rates
primary productivity
relatively high
reduced suspended solids
advective nutrient supply
reduced
light availability at
depth
cell losses by grazing
and sedimentation
intermediate
intermediate
broad, deep, lake basin -like
low flow rates
nutrient-limited primary
productivity
relatively clear
nutrient supply by
internal recycling
high light availability at
depth
relatively low nutrient
levels
cell losses primarily
by grazing
organic matter sypply
primarily autochthonous
more oligotrophic
E
0
2
0
8
0
9
a
Figure A.28. Characteristics of
three longitudinal zones in a
reservoir.
wrm_appA.qxd 8/31/2005 12:08 PM Page 508
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 509
Point sources of pollutants are pipeline discharges to
receiving waters. These are relatively easy to identify and
isolate. In contrast, non-point pollution results from
storm runoff or snowmelt that transports polluting
materials diffusely and over urban or agricultural lands to
streams, rivers, lakes and reservoirs. Non-point source
pollution is closely tied to precipitation and runoff events.
It is less predictable and more variable than pollutants
from point sources and groundwater flows. Because of
their diffuse nature, pollutants from non-point sources
are more difficult to identify and control. Chapter 13
contains a more detailed discussion of this problem,
especially from urban areas.
Of particular importance in addressing lake and
reservoir management problems is the need to consider the
affected ecosystems as well as the direct economic benefits
lakes and reservoirs can provide. Managers must balance
the water needs of economic development with the need to
protect and preserve the environment and its ecosystem.
Many environmental and ecosystem processes in
reservoirs and downstream rivers are complex, long term
and non-traditional from the perspective of current under-
standing of lake and river limnology. Some non-traditional
processes exist because reservoirs are often intermediate
aquatic systems representing transitions between flowing
rivers and lakes (Figure A.28). Our limnological under-
standing of these processes is relatively young.
The presence of a reservoir in a drainage basin where
no such water body previously existed can obviously
affect the watercourse, its flora and fauna, and the human
inhabitants in the drainage basin. These potential impacts
should be identified and analysed prior to reservoir
construction. The results of procedures to identify and
properly evaluate potential environmental, social and
economic consequences of reservoir construction are
included in environmental impact assessment reports. In
many countries such assessments are now required by law
for all new dam construction.
especially abundant in glaciated areas; orogenic
areas are characterized by deep, ancient lakes;
riverine and coastal plains are characterized by
shallow lakes and lagoons
located worldwide in most landscapes, including
tropical forests, tundra and arid plains; often
abundant in areas with a scarcity of natural lakes
generally circular water basin elongated and dendritic water basin
drainage: surface area ratio usually <10:1 drainage: surface area ratio usually >10:1
stable shoreline (except for shallow, lakes in
semi-arid zones)
shoreline can change because of ability to
artificially regulate water level
water level fluctuation generally small
(except for shallow lakes in semi-arid zones)
water level fluctuation can be great
long water flushing time in deeper lakes water flushing time often short for their depth
rate of sediment deposition in water basin is
usually slow under natural conditions
rate of sediment deposition often rapid
variable nutrient loading usually large nutrient loading their depth
slow ecosystem succession ecosystem succession often rapid
stable flora and fauna (often includes endemic
species under undisturbed conditions)
water outlet is at surface water outlet is variable, but often at some depth
in water column
water inflow typically from multiple,
small tributaries
water inflow typically from one or more large
rivers
variable flora and fauna
lake reservoir
E
0
2
0
9
0
3
h
Table A.2. Comparison of major
characteristics of lakes and
reservoirs.
wrm_appA.qxd 8/31/2005 12:08 PM Page 509
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
3.5. Future Reservoir Development
Nearly all major river systems in the world have reservoirs
in their drainage basins. A number of river systems (such as
the Angara, Columbia, Dnieper, Missouri, Mosel, Parana
and Volga) also have cascades of reservoirs within their
basins. Reservoirs exist on all continents (except Antarctica)
and in all countries, although their distribution within
specific countries and regions is irregular. Construction
of new reservoirs has all but ceased in North America and
Europe. In contrast, reservoir construction is continuing
in developing countries, with nearly all new reservoirs
scheduled to enter operation in the twenty-first Century
located in the Middle East, Asia, Africa and Latin America.
Reservoirs represent important components of the
social and economic infrastructure of both developed and
developing countries. In some cases, these have generated
public and international concern, as in the cases of the
Sardar Sarovar Dam in India and the Three Gorges Dam
in China. Proponents of large dams argue that they bolster
local economies, improve electrical energy supplies,
provide needed flood control, and help humans manage
the worlds water resources more effectively. Opponents
of large dams say that they cause significant damage to the
environment and the local culture, and produce little
overall economic gain. These conflicting points of view
require attention early in the planning stage to ensure that
they are properly considered by all relevant parties and
interests prior to initiation of construction activities.
The Sanmenxia Dam on the Yellow River, China,
provides an example of problems that were not
sufficiently considered prior to dam construction.
Finished in 1960, the projects goals were to prevent
floods, provide water for irrigation and produce hydro-
electric power. However, significant silt loads in the
Yellow River were not adequately considered in the
planning stage. The reservoir water basin was largely
filled with silt only four years after construction, and the
reservoir was subsequently taken out of operation.
Another example is the construction of the Aswan High
Dam that impounds the Nile River in southern Egypt. The
dam has now been in operation for about half a century,
and both positive and negative impacts have resulted from
its construction. The positive economic impacts include an
improvement of summer crop rotations and guaranteed
availability of irrigation water for agricultural production,
510 Water Resources Systems Planning and Management
expanded rice cultivation, conversion of about a million
acres from seasonal to perennial irrigation, an expansion
of about 486 thousand hectares of new agricultural and
industrial land due to increased water availability, protec-
tion from high floods and droughts, generation of
significant quantities of hydroelectric power, improved
navigation possibilities and increased tourism.
The negative environmental and social impacts
include declining water levels at Nile River barrages
downstream of the dam, rising water levels upstream of
the Delta Barrage, increased riverbank erosion and river
meandering, production of river channel scour holes
downstream of existing river barrages, decreased water
quality due to increased industrial and agricultural dis-
charges, increased reservoir siltation, increased reservoir
eutrophication, increased water evaporation, increased
coastal erosion at the mouth of the Nile River, decreased
human health due to increased incidence of schistosomi-
asis and spread of water-related vectors, and inundation
of historical monuments.
The overall conclusion of most observers is that the
Aswan High Dam has had an overall positive effect,
although it contributed to some significant environmental
problems as well. Continued studies will undoubtedly
provide additional information and guidance to those
considering the construction of large dam projects in
future years, particularly in developing countries.
Table A.3 summarizes some of the possible effects,
both beneficial and damaging, of large reservoirs.
4. Wetlands
Wetlands are areas of frequent and prolonged presence of
water at or near the soil surface. Swamps, marshes and
bogs are well-recognized types of wetlands. Others less
known include vernal pools (pools that form in the spring
rains but are dry at other times of the year), playas (areas
at the bottom of undrained desert basins that are
sometimes covered with water) and prairie potholes.
Some well-known wetlands, such as the Everglades in
South Florida and the Mississippi bottomland hardwood
swamps in the United States, can be completely dry at
times. In contrast, many upland areas that are not
considered wetlands can be very wet during and shortly
after wet weather.
wrm_appA.qxd 8/31/2005 12:08 PM Page 510
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 511
4.1. Characteristics of Wetlands
When the upper part of the soil is saturated with
water at growing season temperatures, organisms
consume the oxygen in the soil and cause conditions
unsuitable for most plants. Such conditions also cause the
development of so-called hydric soils. Plants that can
grow in these conditions, such as marsh grasses, are
called hydrophytes. The presence of hydric soils and
hydrophytes can indicate the existence of wetlands.
Wetlands are classified as either marine, estuarine,
lacustrine, riverine or palustrine. Marine and estuarine
wetlands are associated with the ocean and include
coastal wetlands, such as tidal marshes. Lacustrine
wetlands are associated with lakes, while riverine wet-
lands are found along rivers and streams. Palustrine wet-
lands may be isolated or connected wet areas and include
marshes, swamps and bogs (Cowardin et al., 1979). A
palustrine system can exist directly adjacent to or within
lacustrine, riverine or estuarine systems.
Wetlands store precipitation and surface water and
then slowly release the water into associated surface water
resources, groundwater and the atmosphere. Wetland
types differ in this capacity based on a number of
physical and biological characteristics, including land-
scape position, soil saturation, the fibre content/degree of
positive benefits negative effects
E
0
2
0
9
0
3
j
displacement of local populations following
inundation of reservoir water basin
excessive human immigration into reservoir
region, with associated social, economic and
health problems
deterioration of conditions for original population
increased health problems from increasing
spread of waterborne disease and vectors
loss of edible native river fish species
loss of agricultural and timber lands
loss of wetlands and land/water ecotones
loss of natural floodplains and wildlife habitats
loss of biodiversity, and displaced wildlife
populations
n
housing
eed for compensation for loss of agricultural
lands, fishery grounds and
degradation of local water quality
decreased river flow rates below reservoir,
and increased flow variability
decreased downstream temperatures,
transport of silt and nutrients
decreased concentrations of dissolved oxygen
and increased concentrations of hydrogen
sulfide and carbon dioxide in reservoir bottom
water layer and dam discharges
barrier to upstream fish migration
loss of valuable historic or cultural resources
(e.g., burial grounds, relic sites, temples)
decreased aesthetic values
increased seismic activity
production of energy (hydropower)
increased low-energy water quality
improvement
retention of water resources in the
drainage basin
creation of drinking water and water
supply resources
creation of representative biological
diversity reserves
increased welfare for local population
enhanced recreational possibilities
increased protection of downstream river
from flooding events
increased fishery possibilities
storage of water for use during low-flow
periods
enhancement of navigation possibilities
increased potential for sustained
agricultural irrigation
Table A.3. Possible positive
and negative effects of large
reservoir construction.
wrm_appA.qxd 8/31/2005 12:08 PM Page 511
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
decomposition of the organic soils, vegetation density and
type of vegetation.
4.1.1. Landscape Position
Landscape position affects the amount and source of water
in a wetland. For example, wetlands that are near the top
of a watershed, such as a mountain bog, will not receive as
much runoff as marshes in low areas. Wetlands can
result from precipitation, groundwater or surface flow.
Precipitation-dominated wetlands can be in flat or slightly
elevated areas in the landscape, where they receive little or
no surface runoff. Generally such wetlands have a clay and
peat layer that retains the precipitation and also prevents
discharge from groundwater. Wetlands also form in areas
of active groundwater discharge, particularly at the base
of hills and in valleys. These groundwater-dominated
wetlands may also receive overland flow but they have a
steady supply of water from and to groundwater. Many
wetlands in low points on the landscape are dominated
by overland flow. Such riverine, fringe (marsh) and tidal
wetlands actively influence the landscape since they
come in contact with, store or release large quantities of
water and act upon sediments and nutrients. Surface water
provides the major source of water for these wetlands.
512 Water Resources Systems Planning and Management
4.1.2. Soil Saturation and Fibre Content
Soil saturation and fibre content affect the capacity of a
wetland to retain water. As in a sponge, as the pore spaces
in wetland soil and peat become saturated by water, they
are able to hold less additional water and release the water
more easily. Clay soils retain more water than loam or
sand. They hold the water particles more tightly through
capillary action, since pore spaces are small and the water
particles are attracted to the negatively-charged clay. Pore
spaces between sand particles are larger and water drains
more freely, since less of the water in the pores is close
enough to be attracted to the soil particle.
Water in wetlands flows over or close to the surface in
the fibric layer and root zone. Wetlands with sapric
peat (mostly decomposed, unrecognizable fibres) and clay
substrate will store water but will have no groundwater
inflow or outflow.
4.1.3. Vegetation Density and Type
Plants growing in wetlands transpire water. Their stems
reduce the velocity of water flowing through the wetland.
As vegetation density increases, flow velocity decreases.
Sturdy plants, such as shrubs and trees, cause more
friction than do grasses.
E
0
2
0
8
0
9
b
Figure A.29. Wetlands support
a productive food web, from
microscopic algae and
submerged vascular plants to,
in some areas, great blue
herons and otters.
wrm_appA.qxd 8/31/2005 12:08 PM Page 512
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 513
Plant transpiration reduces the amount of water
in wetland soil and increases the capacity to absorb
additional water. As a result, water levels and outflows
from the wetland are less during the growing season than
when plants are dormant. Obviously larger plants
and plants with greater surface area will take up and
transpire more water than will smaller or less dense plant
communities.
4.1.4. Interaction with Groundwater
Some wetlands in low-lying areas are fed by groundwater
discharges. Most other wetlands are a source of ground-
water recharge. The extent of groundwater recharge
by a wetland is dependent upon soil, vegetation, site,
perimetervolume ratio and water table gradient.
Groundwater recharge occurs through mineral soils
found primarily around the edges of wetlands; the soil
under most wetlands is relatively impermeable. A high
perimetervolume ratio, as found in small wetlands,
means that the surface area through which water can
infiltrate into the groundwater is relatively high.
Groundwater recharge is typical in small wetlands such as
prairie potholes.
4.1.5. OxidationReduction
The fluctuating water levels (also known as hydrological
flux) that are characteristic of wetlands control the
oxidationreduction (redox) processes that occur. These
redox processes, governed by the frequency, duration and
timing of wetland flooding (hydroperiod), play a key role
in nutrient cycling, availability and export; pH; vegetation
composition; sediment and organic matter accumulation;
decomposition and export; and metal availability and
export.
When wetland soil is dry, microbial and chemical
processes occur, using oxygen as the electron acceptor.
When wetland soil is saturated with water, microbial
respiration and biological and chemical reactions con-
sume available oxygen. The soil changes from an aerobic
to an anaerobic, or reduced, condition. As conditions
become increasingly reduced, other electron acceptors
than oxygen must be used for reactions. These acceptors
are, in order of microbial preference, nitrate, ferric iron,
manganese, sulphate and organic compounds.
Wetland plants are adapted to changing redox condi-
tions, and often contain spongy tissue with large pores in
their stems and roots. This allows air to move quickly
between the leaf surface and the roots, from which
oxygen is released to oxidize the root zone (rhizosphere)
and allow processes requiring oxygen, such as organic
compound breakdown, decomposition and denitrifica-
tion, to occur.
4.1.6. Hydrological Flux and Life Support
Changes in the frequency, duration and timing of
hydroperiods may affect spawning, migration, species
composition and food chain support of the wetland and
associated downstream systems. Normal hydrological flux
allows exchange of nutrients and detritus and passage of
aquatic life between systems.
Values of, or services provided by, wetlands as a result
of the functions of hydrological flux and storage include
enhanced water quality, water supply reliability, flood
control, erosion control, wildlife support, recreation, and
cultural and commercial benefits.
4.2. Biogeochemical Cycling and Storage
Wetlands may serve as sinks for nutrients, organic
compounds, metals and components of organic matter.
They may also act as filters of sediments and organic
matter. A wetland may be a permanent sink for these
substances if the compounds become buried in the sub-
strate or are released into the atmosphere. Alternatively, it
may retain them only during the growing season or under
flooded conditions. Wetland processes play a role in the
cycles of carbon, nitrogen and sulphur by transforming
them and releasing them into the atmosphere (see, for
example, the nitrogen cycle in Figure A.21).
Decomposition rates vary across wetland types, partic-
ularly as a function of climate, vegetation types, available
carbon and nitrogen, and pH. A pH above 5.0 is necessary
for bacterial growth and survival. Liming, to increase pH,
accelerates decomposition, causing the release of carbon
dioxide from wetlands and land subsidence.
The nutrients and compounds released from decom-
posing organic matter may be exported from the wetland
in soluble or particulate form, incorporated into the
soil, or eventually transformed and released to the
wrm_appA.qxd 8/31/2005 12:08 PM Page 513
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
514 Water Resources Systems Planning and Management
Dissolved phosphorus is processed by wetland soil
microorganisms, plants and geochemical mechanisms.
Microbial removal of phosphorus from wetland soil or
water is rapid and highly efficient; following cell death,
however, the phosphorus is released into the water again.
Similarly, for plants, litter decomposition causes a release
of phosphorus. Burial of litter in peat can provide long-
term removal and storage of phosphorus. Harvesting of
plant biomass is needed to maximize biotic phosphorus
removal from the wetland system.
4.2.3. Carbon (C)
Wetlands store carbon within peat and soil, and so play an
important role within the carbon cycle, particularly given
observations of increasing levels of carbon dioxide in
the atmosphere and concerns about global warming. When
wetlands are drained the oxidizing conditions increase
organic matter decomposition, thus increasing the release
of carbon dioxide. When wetlands are preserved or
restored, the wetlands act as a sink for carbon since organic
matter decomposition is stabilized or slowed.
4.2.4. Sulphur (S)
Wetlands are capable of reducing sulphate to sulphide.
Sulphide is released to the atmosphere as hydrogen,
methyl and dimethyl sulphides or is bound in insoluble
complexes with phosphate and metal ions in wetland
sediments. Dimethyl sulphide released from wetlands
may act as a seed for cloud formation. Sulphate may exist
in soils or enter wetlands through tidal flow or atmos-
pheric deposition.
4.2.5. Suspended Solids
Wetlands filter suspended solids from water that comes
into contact with wetland vegetation. Stems and leaves
cause friction that affects the flow of the water, thus
allowing settling of suspended solids and removal of
related pollutants from the water column. Wetlands may
retain sediment in the peat or permanently as substrate.
Sediment deposition varies across individual wetlands
and wetland types, as deposition depends upon the
rate and type of water flow (channelized or sheet flow),
particulate size and vegetated area of the wetland.
atmosphere. Decomposed matter (detritus) forms the
base of the aquatic and terrestrial food web.
Decomposition requires oxygen and thus reduces
the dissolved oxygen content of the water. High rates of
decomposition such as after algae blooms can reduce
water quality and impair aquatic life support.
The values of wetland functions related to biogeo-
chemical cycling and storage include water quality and
erosion control.
4.2.1. Nitrogen (N)
The biological and chemical process of nitrification and
denitrification in the nitrogen cycle transforms the major-
ity of organic nitrogen entering wetlands, causing
between 70% and 90% to be removed.
In aerobic substrates, organic nitrogen may mineral-
ize to ammonium, which plants and microbes can use,
adsorb to negatively charged particles (e.g. clay), or
diffuse to the surface. As ammonia (NH
3
) diffuses to
the surface, the bacteria Nitrosomonas can oxidize it
to nitrite (NO
2
). The bacteria Nitrobacter in turn oxi-
dizes nitrite to nitrate (NO
3
). This process is called
nitrification. Conversely, plants or microorganisms can
assimilate nitrate, or anaerobic bacteria may reduce it
(denitrification) to gaseous nitrogen (N
2
) when nitrate
diffuses into anoxic (oxygen depleted) water. The
gaseous nitrogen volatilizes and the nitrogen is
eliminated as a water pollutant. Thus, the alternating
reduced and oxidized conditions of wetlands complete
the needs of the nitrogen cycle and increase
denitrification rates.
4.2.2. Phosphorus (P)
Phosphorus can enter wetlands attached to sediment or
in dissolved form. Its removal from water in wetlands
occurs through uptake of phosphorus by plants and soil
microbes; adsorption by aluminium and iron oxides
and hydroxides; precipitation of aluminium, iron, and
calcium phosphates; and burial of phosphorus adsorbed
to sediments or organic matter. Wetland soils can,
however, reach a state of phosphorus saturation, after
which phosphorus may be released from the system.
Phosphorus is released into surface water as organic
matter decomposes.
wrm_appA.qxd 8/31/2005 12:08 PM Page 514
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 515
4.2.6. Metals
All soils contain at least a low concentration of metals, but
in some locations human activities have resulted in metal
levels high enough to cause health or ecological risks.
Metals may exist in wetland soils or enter wetlands
through surface or groundwater flow.
Wetlands can remove metals from surface and ground-
water as a result of the presence of clays, humic materials
(peats), aluminium, iron and/or calcium. Metals entering
wetlands bind to the negatively ionized surface of clay
particles, precipitate as inorganic compounds (including
metal oxides, hydroxides and carbonates controlled by
system pH), interact with humic materials, and adsorb or
occlude to precipitated hydrous oxides. Iron hydroxides
are particularly important in retaining metals in salt
marshes. Wetlands remove more metals from slow-
flowing water since there is more time for chemical
processes to occur before the water moves out of the area.
Burial in the wetland substrate will keep bound metals
immobilized. Neutral pH favours metal immobilization in
wetlands. With the exception of very low pH peat bogs,
as oxidized wetland soils are flooded and reduced, pH
converges toward neutrality (6.5 to 7.5), whether the
wetland soils were originally acidic or alkaline.
4.3. Wetland Ecology
Wetlands are productive ecosystems. Immense varieties
of species of microbes, plants, insects, amphibians,
reptiles, birds, fish and other wildlife depend on them in
some way. Wetlands with seasonal hydrological pulsing
are the most productive.
Wetland plants play an integral role in the ecology of
the watershed. They provide breeding and nursery sites,
resting areas for migratory species, and refuge from
predators. Decomposed plant matter (detritus) released
into the water is the source of food for many invertebrates
and fish both in the wetland and in associated aquatic
systems. Physical and chemical characteristics such as
climate, topography, geology, hydrology, and inputs of
nutrients and sediments determine the rate of plant growth
and reproduction (primary productivity) of wetlands.
The greater the amount of vegetation, the more the
wetland vegetation will intercept runoff and be capable of
reducing runoff velocity and removing pollutants from
the water. Wetland plants also reduce erosion as their
roots hold the soil particles of streambanks, shorelines or
coastlines.
The inundated or saturated conditions occurring in
wetlands limit plant species composition to those that can
tolerate such conditions. Beaver, muskrat and alligators
create or manipulate their own wetland habitat, which
other organisms, such as fish, amphibians, waterfowl,
insects and mammals, can then inhabit.
Wetland shape and size affect the habitat of the wildlife
community. The shape affects the perimeterarea ratio,
which is important for the success of interior and edge
species. Shape is also important for movement of animals
within the habitat and between habitats. Wetland size is
particularly important for larger and wider-ranging
animals that use wetlands for food and refuge, such as
black bear or moose, since in many regions wetlands may be
the only undeveloped and undisturbed areas remaining.
4.4. Wetland Functions
Only recently have scientists begun to understand the
importance of the functions that wetlands perform. Far
from being useless disease-ridden places, wetlands
provide benefits that no other ecosystem can, including
natural water quality improvement, flood protection,
shoreline erosion control, opportunities for recreation
and aesthetic appreciation, community structure and
wildlife support, and natural products at reduced costs
compared with other alternatives. Protecting wetlands in
turn can increase human safety and welfare as well as
enhance the productivity of aquatic ecosystems.
4.4.1. Water Quality and Hydrology
Wetlands have important filtering capabilities for intercept-
ing surface water runoff from higher dry land before it
reaches open water. As the runoff water passes through, the
wetlands retain excess nutrients and some pollutants, and
reduce sediment that would otherwise clog waterways and
affect fish and amphibian egg development. In performing
this filtering function, wetlands offset the costs of waste-
water treatment. In addition to improving water quality
through filtering, some wetlands maintain streamflow
during dry periods, and many replenish groundwater
supplies.
wrm_appA.qxd 8/31/2005 12:08 PM Page 515
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
4.4.2. Flood Protection
Wetlands function as natural sponges. They trap and
slowly release surface water, rain, snowmelt, groundwater
and floodwaters. Trees, root mats and other wetland
vegetation slow the speed of floodwaters and distribute
them over the floodplain. This combined water storage
and braking action lowers flood heights and reduces
erosion. Wetlands within and downstream of urban
areas are particularly valuable, counteracting the greatly
increased rate and volume of surface water runoff from
pavement and buildings.
The holding capacity of wetlands helps control floods
and prevents waterlogging of crops. Preserving and
restoring wetlands, together with other water retention
measures, can often provide the level of flood control that
would otherwise require expensive dredging operations
and levees. The bottomland hardwood-riparian wetlands
along the Mississippi River once stored significantly more
floodwater than they can today because most have been
filled or drained.
4.4.3. Shoreline Erosion
The ability of wetlands to control erosion is so valuable
that communities in some coastal areas are restoring
wetlands to buffer the storm surges from hurricanes and
tropical storms. Wetlands at the margins of lakes, rivers,
bays and the ocean protect shorelines and stream banks
against erosion. Wetland plants hold the soil in place with
their roots, absorb the energy of waves, and break up the
flow of stream or river currents.
4.4.4. Fish and Wildlife Habitat
Many animals and plants depend on wetlands for
survival. Estuarine and marine fish and shellfish, various
birds and certain mammals depend on coastal wetlands
for survival. Most commercial and game fish breed and
raise their young in coastal marshes and estuaries.
Menhaden, flounder, sea trout, spot, croaker and striped
bass are among the more familiar fish that depend on
coastal wetlands. Shrimp, oysters, clams, and blue
and Dungeness crabs likewise need these areas for food,
shelter and breeding grounds.
For many animals and plants, like wood ducks,
muskrat, cattails and swamp rose, inland wetlands are the
516 Water Resources Systems Planning and Management
only places they can live. Beaver may actually create their
own wetlands. For others, such as striped bass, peregrine
falcon, otter, black bear, raccoon and deer, wetlands
provide important food, water or shelter. Many breeding
bird populations including ducks, geese, woodpeckers,
hawks, wading birds and many songbirds feed, nest and
raise their young in wetlands. Migratory waterfowl use
coastal and inland wetlands as resting, feeding, breeding
or nesting grounds for at least part of the year. Indeed,
an international agreement to protect wetlands of inter-
national importance was developed because some species
of migratory birds are completely dependent on certain
wetlands and would become extinct if those wetlands
were destroyed.
4.4.5. Natural Products
We use a wealth of natural products from wetlands,
including fish and shellfish, blueberries, cranberries,
timber and wild rice, as well as medicines that are derived
from wetland soils and plants. Many fishing and shell-
fishing industries harvest wetland-dependent species. In
the southeast part of the United States, for example,
nearly all the commercial catch and over half of the recre-
ational harvest consists of fish and shellfish that depend
on the estuary-coastal wetland system. Wetlands are
habitats for fur-pelted animals like muskrat, beaver and
mink, as well as reptiles such as alligators.
4.4.6. Recreation and Aesthetics
Wetlands have recreational, historical, scientific and
cultural value. More than half of all US adults hunt, fish,
go birdwatching or photograph wildlife. Painters and
writers continue to capture the beauty of wetlands on
canvas and paper, or through cameras, and video and
sound recorders. Others appreciate these wonderlands
through hiking, boating and other recreational activities.
Almost everyone likes being on or near the water; part of
the enjoyment is the fascinating variety of lifeforms.
5. Estuaries
Estuaries are places where freshwater mixes with salt-
water under the influence of tides, creating a unique
environment supporting a rich and diverse ecosystem.
wrm_appA.qxd 8/31/2005 12:08 PM Page 516
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 517
Estuaries are capable of gathering and holding an abun-
dance of life-giving nutrients from the land and from the
ocean. They are also important sites for human economic
and recreational activities. An estuary can provide a
laboratory for lessons in biology, geology, chemistry,
physics and social issues.
Estuaries are partially enclosed bodies of water formed
where freshwaters from rivers and streams flow into and
mix with the saline seawater. They and the lands
surrounding them are places of transition from land to
sea, and from fresh to saltwater. Although influenced by
the tides, they are generally protected from the full force
of ocean waves, winds and storms by the reefs, barrier
islands or fingers of land, mud or sand that define their
seaward boundaries.
Estuarine geometry can vary substantially, and the
physical characteristics of estuaries affect the distribution
of their constituents. The shape and geomorphology of
the estuary influence the flows, circulation and mixing
of the waters. The flow pattern and the density structure
in an estuary govern the advective transport and diffu-
sion/dispersion characteristics that determine the fate and
distribution of waterborne constituents.
The word estuary comes from the Latin aestus, mean-
ing tide. Traditionally, the upstream limit of an estuary is
defined in terms of the limit of penetration of saltwater. Salt
concentrations move upstream under the influence of the
ocean tide. A commonly used definition is that of Pritchard
(1952), who defined an estuary as a semi-enclosed coastal
body of water which has a free connection with the open
sea and within which sea water is measurably diluted with
fresh water derived from land drainage.
A broader definition of an estuary would take into
account the diversity and spatial variability of its fauna
and flora. Hutchings and Collett (1977) define estuaries
as the tidal portions of river mouths, bays and coastal
lagoons, irrespective of whether they are dominated by
hypersaline, marine or freshwater conditions. Included
in this definition are inter-tidal wetlands, where water
levels can vary in response to the tidal levels of the
adjacent waterway, together with perched freshwater
swamps, as well as coastal lagoons that are intermittently
connected to the ocean.
Perched freshwater swamps, or swamps located above
the regional water table, can occur in the lower catchment
areas of an estuary. Although tides or ocean salinity does
not affect them, they can form an important component
of the estuarine habitat. For this reason, such swamps
need to be included in estuarine management plans and
policies.
5.1. Types of Estuaries
Every estuary is unique, yet all share certain features.
Some are similar enough to be grouped or classified
according to their shared characteristics. Others are
typical of specific regions. All are greatly affected by
geology, climate and many other factors, including the
ever-growing human population.
Depending on their geological characteristics, estuaries
can be classified as:
Coastal plain estuary (or drowned river valley). These are
commonly found along coastlines that have relatively
wide coastal plains. They were created by the gradual
rise in sea level following the last glacial period, some
10,000 years ago. Usually, a stretch of the freshwater
river is subject to tidal oscillations. Typically, the
estuary is funnel-shaped, widening gradually in the
downstream direction, often with inter-tidal mud flats.
Sometimes a large delta system has formed with many
natural channels. Chesapeake Bay is the largest estuary
of this type of in the United States.
Fjord-type estuary (formed by a glacier). These estuaries
generally have U-shaped cross sections with steep sides
and are often quite deep (300 or 400 m). They are
common in the Northern Hemisphere above 45 degrees
latitude. In North America they are most spectacular
along the coast of Alaska and British Columbia.
Bar-built estuary or lagoon. These are formed when off-
shore barrier sand islands and sand pits become higher
than sea level and extend between headlands in a
chain, broken by one or more inlets. Lagoons are
usually situated parallel to the coastline. Many have
narrow outlets to the sea and minimal freshwater
inflow, often creating higher salinity levels than in
coastal-plain and other estuaries. In North America,
lagoons occur mainly along the Gulf Coast. Smaller
lagoons occur along the West Coast.
Estuaries produced by tectonic processes. Estuaries
created by processes such as landslides, faulting and
volcanic eruptions are found along coasts where such
wrm_appA.qxd 8/31/2005 12:08 PM Page 517
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
activity is or has been common, such as the Pacific
Coast of North America. These estuaries vary greatly
and often share characteristics with other types. The
largest in the United States is San Francisco Bay.
Any existing estuary can be a composite estuary evolving
from an overlap of two or more of these basic types.
Estuaries can also be classified on the basis of stratifi-
cation and circulation:
A salt-wedge estuary is highly stratified. Saltwater
moves into it in the shape of a wedge, with freshwater
flowing over it. The velocity of the freshwater outflow
is usually greater than that of the saltwater inflow. This
keeps the saltwater from extending very far up the
estuary. The Mississippi River Estuary is an example of
this type, where the difference in water level between
the high and low tides (the tidal range) on the seaward
side of the estuary is small and incapable of mixing the
stratified layers.
The vertically homogenous or well-mixed estuary is
characterized by low inflow of freshwater and large
tidal ranges. In such estuaries, saltwater and freshwater
tend to mix vertically, and sometimes laterally as well.
Intermediate estuaries are partly mixed. They exhibit
circulation patterns that are somewhere between those
of the salt-wedge and vertically homogenous estuaries.
This type of estuary, typical along the US East Coast,
has a moderate inflow of freshwater and moderate-to-
large tidal range.
Estuaries area also classified based on their circulation
the interaction of tidal currents and river flow (Bowden,
1967):
A salt wedge estuary has only a small amount of
friction between the layers. The steep density gradient
at the interface reduces the turbulence and mixing to
a low level. The effect of the Coriolis force causes the
lateral sloping of the interface downward to the right
in the Northern Hemisphere when looking towards
the sea.
A vertically homogeneous estuary can exhibit lateral
variation due to Coriolis force in wide estuaries, or be
laterally homogeneous where the ratio of width to
depth is relatively small. Coriolis force causes a net
seaward flow of lower-salinity water on the right
hand of the estuary, and a compensating flow of
518 Water Resources Systems Planning and Management
higher-salinity water on the left-hand side. In this
sequence, the tidal currents are stronger in relation to
the river flow.
In a two-layer flow with entrainment, the entrainment
is due to breaking internal waves. Many fjords have
this type of circulation.
In a two-layer flow with vertical mixing, the volume of
water involved in this type of circulation may be many
times the river discharge. Examples include the Tees
and Thames estuaries in the UK.
Estuaries are constantly affected by tidal and wave action,
prevailing and changing winds, local and distant weather
systems, and variations of rainfall runoff and river
discharge.
5.2. Boundaries of an Estuary
With some exceptions, such as those found in the
Mediterranean Sea, tidal motions are significant features
in estuaries. The upstream boundary or head of an estu-
ary is the limit of tidal influence. The tidal limit can be a
considerable distance upstream from the salinity limit.
The actual limit of tidal influence varies with time,
depending upon freshwater flows and the natural vari-
ability of tides. It can be difficult to determine the tidal
limit by mere observation. Towards the limit of tidal
influence, flood flows may persist for as little as an hour
or less, and may be so small as to go unnoticed.
As well as short-term cyclical changes in response to
the changing ocean tides, there may be long-term
variations in the upstream boundary of an estuary due
to both natural processes and artificial disturbance. The
downstream boundary is not always obvious, but it
corresponds to the seaward limit of the entrance bar in
most situations.
The lateral boundaries of an estuary are often defined
in ecological rather than hydraulic terms. The shallow
and inter-tidal margins of an estuary support the diversity
and productivity of marine life, such as sea-grass mead-
ows, mangrove forests and saltmarshes, all generating
large amounts of organic detritus that form the founda-
tion of the estuarine food chain.
The lateral extent of an estuary includes all wetlands
salt, brackish and fresh that interact with tidal and flood
flows. They also include those marshes that are inundated
only during extreme tides or flood events.
wrm_appA.qxd 8/31/2005 12:08 PM Page 518
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 519
5.3. Upstream Catchment Areas
An estuary acts as a funnel to convey freshwater runoff
from the land into coastal waters. Catchment activities
can affect the volume and quality of this runoff.
Floodwaters and stormwater runoff often contain
significant quantities of suspended solids, natural and
artificial nutrients, pesticides and other constituents,
some of which can be detrimental to estuarine
ecosystems.
Upstream catchment activities are the single most
important factor in determining the nutrient balance and
water quality of estuaries. The impact of external activities
occurring beyond the strictly defined estuarine limits
points to the need for a total catchment management
approach in managing estuaries. What we do upstream
affects what happens downstream, including, for most
rivers, the estuary.
5.4. Water Movement
Water moves along an estuary under the influence of
two primary forcing mechanisms: freshwater inflows
from rivers and the regular tidal movement of seawater
into and out of the estuary. In addition, differences in
tides, winds and salinity densities within the estuary
generate secondary currents that, while of low velocity,
can be significant with respect to mixing and sediment
transport.
Freshwater inflows fluctuate in a more or less random
fashion in response to surface runoff from tributary catch-
ments. The construction of major dams in upstream
catchment areas reduces both the volume of freshwater
runoff and the freshwater flushing of estuaries. In times
of drought, freshwater inflows may be absent, or may be
mainly the discharge of sewage effluent and other waste-
waters.
Freshwater inflows result in a net seaward excursion of
water particles over each tidal cycle, and thus promote
estuary flushing.
The movement of seawater in and out of an estuary is
predominantly influenced by the tides. Freshwater effects
are generally small (but often significant with respect to
water quality), except during floods.
Coastal water levels fluctuate in a regular and
predictable fashion in response to gravitational effects,
primarily of the moon and sun, on the oceans of the earth.
The tidal range varies from one tidal cycle to another in
response to the changing relative positions of these
celestial bodies. In response to the monthly orbit of the
moon around the earth, the tidal range undergoes a
regular fourteen-day cycle, increasing to a maximum over
a week (spring tides) and then decreasing to a minimum
over the following week (neap tides). Solstice tides, or
king tides occur in June and December of each year, when
the sun is directly over the Tropics of Cancer and
Capricorn respectively.
Tides along many coastlines are semi-diurnal in
nature; that is, high water and low water occur about
twice daily (the actual period of a tidal cycle is about 12.5
hours). They are sinusoidal in shape and have a
pronounced diurnal inequality (successive high tides
differ markedly).
The tidal rise and fall of ocean water levels propagates
along an estuary as a wave. The speed of travel or celerity
of this wave the speed at which high water and low
water travel upstream from the estuary entrance varies
with water depth: the deeper the water, the faster the
wave celerity. The propagation of the tide along an
estuary is affected by the geometry of its bed. Tidal
propagation in tidal rivers is very sensitive to water
depths. In certain estuaries, the ocean tidal range is
increased in upstream reaches of the estuary because of its
geometry. The celerity of the flood tide is higher than that
of the ebb tide.
5.4.1. Ebb and Flood Tides
The vertical rise and fall of tides (Figure A.30) produce
horizontal flows in the form of tidal currents. The incom-
ing or rising tide is traditionally referred to as the flood
tide because it floods the channel. The outgoing tide is
referred to as the ebb tide. The strength of the ebb and
flood tide velocities varies diurnally and over springneap
cycles in exactly the same way as tidal water levels do.
Spring tides produce the fastest tidal currents. Depending
on the tidal range, entrance characteristics and the depth
of the estuary, tidal currents can have velocities of up to a
metre a second.
If the tidal range is appreciable compared to the
mean depth of the estuary, the speed of propagation of
the tide at high water will be significantly faster than at
wrm_appA.qxd 8/31/2005 12:08 PM Page 519
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
low water. This causes the shape of the tidal wave
to become progressively more distorted as it moves
landward. This tidal distortion results in a saw-tooth
tide curve, in which the rise of the tide is noticeably
faster than its fall. As a consequence, peak flood tide
velocities are generally greater than peak ebb tide
velocities. This is of significance with respect to sediment
transport.
The period of quiet water when the tide reverses from
flood to ebb or vice versa is referred to as slack water.
High-water slack is the name given to the tide change
from flood to ebb; low-water slack refers to the tide
change from ebb to flood. The duration of slack water
varies from one estuary to another and can last from
twenty minutes to almost an hour.
5.4.2. Tidal Excursion
The total distance travelled by a water particle from low-
water slack to high-water slack and vice versa is referred
to as the tidal excursion. This is the maximum distance
travelled by a water particle during the rising or falling
limb of the tide.
Tidal excursion is not to be confused with the distance
travelled by the tide wave itself (e.g. high water) that
propagates from the ocean to the end of the estuary each
tide cycle.
520 Water Resources Systems Planning and Management
Freshwater inflows impose a net seaward movement
on water particles over a tide cycle. In these circum-
stances, the ebb tide excursion is greater than the flood
tide excursion. This is illustrated in Figure A.31, which
also shows the net effect of oscillatory tidal flows and
seaward-draining freshwater flows in flushing a parcel
of water out to sea.
5.4.3. Tidal Prism
The total volume of water moving past a fixed cross section
of the estuary during each flood tide or ebb tide (i.e. slack
water to slack water) is referred to as the tidal prism. The
larger the tidal range within the estuary and the greater the
dimensions of the estuary, the larger is the tidal prism. On
average, the ebb and flood tidal prisms are equal.
The size of the tidal prism is not an indication of
the amount of flushing. The movement of water contained
in the tidal prism is largely oscillatory, as depicted in
Figure A.31. The net seaward flushing action over a tidal
cycle results from two processes: freshwater advection and
longitudinal dispersion. Freshwater advection is the net
seaward displacement caused by Freshwater. Freshwater
longitudinal dispersion results mainly from the tides.
5.4.4. Tidal Pumping
The higher celerity of the flood tide compared to the ebb
tide results in a tendency for greater upstream movement
of water on the flood tide compared to downstream move-
ment on the ebb tide. This leads to a dynamic
trapping of water in the upper reaches of the estuary.
This effect is sometimes referred to as tidal pumping:
tidal distortion results in the tide pumping water upstream.
In shallow estuaries, tidal pumping can affect the
distribution of dissolved pollutants along an estuary.
5.4.5. Gravitational Circulation
The presence of salt in an estuary produces a longitudinal
density gradient, with water densities around the mouth
of the estuary being greater (because of higher salt con-
centrations) than densities around the head of the estuary.
This results in the enhancement of flood-tide velocities
near the bed and ebb-tide velocities near the surface.
When averaged over a tidal cycle, this behaviour leads to
mean
sea level
low
tide lag
fall time rise time
high tide lag
ocean tide
tide at
head of estuary
E
0
2
0
2
1
3
f
w
a
t
e
r

l
e
v
e
l
time
Figure A.30. Tidal characteristics at the entrance and head of
a tidal river.
wrm_appA.qxd 8/31/2005 12:08 PM Page 520
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 521
residual currents, in which saline water flows upstream
along the bottom of the estuary and less salty, even fresh
water, flows seawards near the surface (see Figure A.32).
This pattern of residual flows is referred to as gravitational
circulation (it is driven by the gravitational forces result-
ing from density differences). Gravitation circulation can
give rise to discharges considerably greater than fresh-
water inflows. Gravitational circulation is an important
mechanism of upstream sediment transport and the
longitudinal dispersion of salt in an estuary.
5.4.6. Wind-Driven Currents
Wind shear at the water surface produces surface
currents, counterbalancing currents at the bottom of the
water column, and large-scale lateral circulation. Wind
speeds up to about 6 to 7 m/sec are the most effective in
moving surface waters. Wind-driven surface currents
have speeds of about 2% of the wind speed. Wind-driven
currents are a major mechanism in the transport and
distribution of floating pollutants such as oil.
In sluggish tidal systems or bays, where surface tidal
flows are weak, wind-driven currents can be among the
main agents leading to effective water movement and
mixing within the main water body. When a constant wind
blows over a basin of variable depth, a laterally varying
surface current is induced, flowing with the wind in shallow
areas and as a return flow against the wind in deeper areas.
5.5. Mixing Processes
Parcels of water mix as they move along the estuary under
the influence of freshwater flows, tidal flows and second-
ary currents. Mixing not only involves an exchange of
water mass, but also of any substance dissolved in it, such
as salts and various pollutants. Mixing processes affect the
d
i
s
t
a
n
c
e

a
l
o
n
g

e
s
t
u
a
r
y
time
parcel of water
parcel
flushed
out to
sea
freshwater advection
h
i
g
h

w
a
t
e
r

s
l
a
c
k
l
o
w
w
a
t
e
r
s
l
a
c
k
ebb
tide
flood
tide one tidal cycle
ocean tides
freshwater inflows
estuary
E
0
2
0
2
1
3
g
Figure A.31. Movement of a
parcel of water down an
estuary.
Q
f
Q
g
0.1 1.0 5 10 20
tidal limit salinity limit salinity (ppt)
Q
f
Q
g
- freshwater inflow
- gravitational circulation
- isolines (ppt)
E
0
2
0
2
1
3
h
Q
f
Q
g
+
Figure A.32. Gravitational
circulation in a partially mixed
estuary.
wrm_appA.qxd 8/31/2005 12:08 PM Page 521
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
distribution of salinity and water quality constituents
throughout the estuary.
5.5.1. Advection and Dispersion
Advective and dispersive transport are the major processes
by which dissolved matter is distributed throughout an
estuary. As water flows along an estuary it transports
dissolved matter with it. This is advective transport.
During this process, water mixes with neighbouring
parcels of water. This mixing, or dispersive transport,
leads to a net transport of dissolved materials from regions
of high concentration to regions of lower concentration.
The mixing of parcels of water is principally caused by
lateral and vertical variations in velocity along an estuary.
These variations create velocity shear. Velocity variations
result in faster parcels of water together with their
dissolved loads moving ahead of their slower neigh-
bours, as depicted in Figure A.33. Turbulence and eddies
generated by the velocity shear between neighbouring
parcels result in the interchange of water and dissolved
matter between parcels.
5.5.2. Mixing
Mixing occurs in all three directions vertically laterally
and longitudinally. The principal mechanism of vertical
mixing is the velocity shear caused by flow moving across
the bed of the estuary. The surface and mid-depth flow
velocities are faster than bed-flow velocities. This results
in turbulence, principally at the bed, which promotes the
mixing of overlying waters. Vertical mixing is responsible
for well-mixed salinity regimes.
The major mechanisms promoting the mixing of
waters laterally across an estuary are the velocity shear
associated with lateral velocity gradients, wind shear,
lateral tidal flows, large-scale eddies generated by
obstructions, and bends and meanders in the alignment
of major channels.
Figure A.33 shows the lateral exchange of parcels of
water associated with the lateral variation of longitudinal
velocities across the estuary. Lateral variations in velocity
are caused by the presence of the banks of the estuary and
changes in water depth. Deeper waters flow faster than
shallow waters. This lateral exchange promotes mixing
across the estuary.
522 Water Resources Systems Planning and Management
Wind shear and lateral tidal flows (such as the filling
of inter-tidal areas on the flood tide) both result in the
advection of water across an estuary and associated
mixing. The presence of bends and meanders also results
in water flowing from one side of a channel to the other,
and associated mixing across the channel.
Obstructions such as rock bars and shoals also
promote the flow of water across an estuary, both by
the advection associated with redirected flows and by the
dispersion associated with lateral shear.
Large-scale boundary effects, tidal trapping, tidal loops
and interconnections between tidal systems, and gravita-
tional circulation are among a number of factors that
promote longitudinal mixing. Large-scale boundary
effects relate to the mixing caused by the presence of bays
and channels along an estuary. Waters are temporarily
stored in these areas on the rising tide and released when
it falls. This separation of the trapped waters from other
waters moving along the estuary facilitates mixing by
velocity shear and advective transport.
The presence of shoals generates strong velocity shear
between fast-moving main channel flows and the
longitudinal
velocity variation
across estuary
position
of parcels
at time zero
position
of parcels
at time T
position
of parcels
at time 2x T
E
0
2
0
2
1
8
a
Figure A.33. Mixing caused by the lateral variation of velocity
across an estuary.
wrm_appA.qxd 8/31/2005 12:08 PM Page 522
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 523
slow-moving waters that move in and out of shoal areas
on the rising and falling tide. This behaviour leads to tidal
trapping, whereby a discrete body of water is separated
and trapped over shoal areas on the flood tide, to be
released on the ebb. This behaviour facilitates mixing
by velocity shear and advective transport, supplementing
that resulting from large-scale boundary effects.
Finally, the presence of tidal loops in a large delta or
the interconnection of two tidal systems can create a
complicated pattern of residual flows that further enhances
advective and dispersive transport along the estuary. To
summarize, mixing will be greatest in those estuaries where
lateral and vertical velocity gradients are greatest, that is, in
estuaries where deep channels thread their way through
shallow flats, where there are extensive shoals and peri-
pheral bays and channels, and where tidal flows are fast.
5.6. Salinity Movement
Seawater consists of a dilute solution of a mixture of salts.
The term salinity refers to the total salt concentration.
Seawater has a worldwide average salt concentration of
about 35 kg/m
3
or 35 parts per thousand (ppt). Saline
coastal waters are carried into an estuary by the tides;
freshwater inflows tend to wash the saltwater back out to
sea. Tidal flows are very effective in moving salinity
upstream. This is seen in the relatively quick migration of
salinity upstream in an estuary after a major flood has
washed it downstream. The balance between the land-
ward transport of salt by tidal processes and its seaward
return by freshwater discharges determines the limit of
salinity intrusion along an estuary.
Salt and other dissolved substances are transported
along an estuary by the longitudinal advection associated
with large-scale water movements and secondary cur-
rents, and by the longitudinal dispersion associated with
velocity shear. Freshwater inflow is the major factor
affecting the limit of saline intrusion along an estuary. Salt
will not penetrate far along channels with a net seaward
residual velocity in excess of 0.1 m/sec.
5.6.1. Mixing of Salt- and Freshwaters
The density of seawater is greater than that of freshwater
and varies depending on both salinity and temperature.
At a temperature of 20 C, seawater has a density of about
1,025 kg/m
3
, whereas freshwater at that temperature has
a density of 1,000 kg/m
3
. Although the difference in den-
sity is relatively small, it still affects estuarine circulation.
Because of its lower density, freshwater tends to float
on top of the seawater (stratification). Turbulence gener-
ated by the movement of the water over the bed of an
estuary causes vertical mixing, which tends to break
down any salinefreshwater stratification. The faster the
water moves, the greater its turbulence and resultant
mixing. High tidal velocities produce strong vertical
mixing, resulting in little variation in salinity from the top
to the bottom of the water column. Low tidal velocities
are insufficient to cause complete vertical mixing
and stratified conditions can develop (i.e. bottom
salinity concentrations are greater than surface salinity
concentrations).
5.6.2. Salinity Regimes
In well-mixed estuaries, the salinity distribution is almost
uniform with depth. In partially mixed estuaries, the
salinity varies continuously through the depth of the
water column, with no evidence of a marked interface
between the upper and lower layers. The salinity can
vary over the depth by as little as 1 ppt or as much as
10 ppt.
Stratified conditions are characterized by an abrupt
increase in salinity over the water depth. In the absence of
significant tidal velocities, vertical mixing in an estuary is
weak. This allows the saltwater to move into the estuary
as an arrested salt wedge that can penetrate a long
distance into a deep estuary.
The freshwater overlying the wedge tends to entrain
saltwater as it moves seaward, thereby becoming more
brackish as it approaches the estuary mouth. These
conditions are shown in Figure A.34.
5.6.3. Variations due to Freshwater Flow
If there is considerable variation in freshwater inflows
over time, the estuaries can display all of the foregoing
salinity regimes (shown in Figure A.34) at different
times. During periods of extended dry weather, when
freshwater flows are small, estuaries tend to be well
mixed. When freshwater inflows to estuaries are low, the
salinity of water in the estuary may exceed the salinity of
wrm_appA.qxd 8/31/2005 12:08 PM Page 523
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
the open ocean (hyper-saline conditions). During periods
of wet weather, partially mixed conditions prevail. Under
full flood flows, stratified conditions occur in entrance
reaches and most of the salt may be washed out to sea.
5.7. Sediment Movement
Sediment is a major component of most estuaries. Its
sources, movements and impacts are numerous.
5.7.1. Sources of Sediment
Different types of sediment are delivered to an estuary by
a variety of sources, as illustrated in Figure A.35.
Riverbank erosion and general catchment runoff produce
quantities of sand, silt and clay. Catchment runoff also
delivers organic matter to the river/ estuary. Littoral
processes in coastal waters can supply large quantities of
524 Water Resources Systems Planning and Management
sand to an estuary. Wind action on dunes and inter-tidal
sandbanks carry fine sand into an estuary.
5.7.2. Factors Affecting Sediment Movement
The currents caused by freshwater inflows and tidal
behaviour are the main mechanisms moving sediment in
estuaries. The faster these currents, the greater will be
the shear stress and turbulence generated at the bed, and
the greater will be the movement of sediment by bedload
and suspended load transport.
Sediment movement depends on flow, turbulence and
the physical characteristics of the sediment particles.
Sediment transported in water will settle at times or
places of low wave or tidal activity. The rate of settling
depends upon the grain size of sands and upon the
mineralogy and chemistry of muds.
Q
Q
Q
f
f
f
Q
f
Q
g
Q
f
Q
g
+
+
Q
Q
g
g
6
6
6
12
12
24
24
32
32
salinity (ppt)
salinity (ppt)
salinity (ppt)
Q
f
Q
g
- freshwater inflow
- upstream gravitational flow at bed
- isohalines (ppt)
E
0
2
0
2
1
8
b
dry season (low freshwater inflow), well mixed salinity regime
wet season, partially mixed salinity regime
high freshwater inflow, stratified conditions
Figure A.34. Possible salinity
regimes in tidal rivers.
wrm_appA.qxd 8/31/2005 12:08 PM Page 524
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 525
Sediment on the bed will be eroded and transported
when the shear stress exerted on the bed by waves, tides
and freshwater action, acting either alone or together, exceeds
a critical minimum value. The critical shear stress also varies
according to sediment size, mineralogy and chemistry.
If sediment is deposited in locations where the critical
shear stress is not exceeded, or is exceeded only infre-
quently, then the sediment will slowly consolidate,
increasing in both density and strength. As bed density
increases, the stress threshold for erosion will increase,
and the sediment deposit will become more stable and
less likely to be eroded by natural forces.
The difference between peak flood flow and peak
ebb flow velocities (tidal distortion) and gravitational
circulation facilitate a net upstream movement of
suspended solids, in other words, a residual suspended
load flux. Because of tidal distortion, the duration of high-
water slack is different from low-water slack, thereby gen-
erating differences in the slack water settling opportunity
for suspended particles. This, coupled with differences in
current behaviour around the two slacks, can impart a net
displacement to suspended sediment each tide cycle.
Bedload transport is the principal means by which sand
is moved along the estuaries. Suspended load transport of
sands occurs usually in the immediate vicinity of estuary
entrances, where high velocities (greater than 1 m/s) and
wave action promote the suspension of sand grains.
Bedload transport is quite sensitive to small changes in
velocity, such as those brought about by tidal distortion
and gravitational circulation. These two mechanisms
generate residual bedload flux in the upstream direction.
Together, these two processes result in a net upstream
transport of marine sand that forms shoals and deltas in
lower estuarine reaches.
Freshwater flows during large floods can transport
sand downstream at very high rates, as evidenced by the
substantial scouring of shallow sand shoals that tends to
occur during floods. The sand transported downstream
by freshwater flows is deposited on the seaward face of
the entrance bar (the ebb terminal lobe of Figure A.36).
After the flood dissipates, this sand is reworked by the
action of waves and nearshore currents and is returned to
the shoreline over a period of many months.
The action of wind on exposed sand dunes can transport
considerable quantities of sand into an estuary. Transport of
sand by wind is one of the dominant factors in the sediment
budget of exposed entrances and may be the principal
reason for the tendency of such entrances to close.
5.7.3. Wind Effects
Some transport of sand grains from inter-tidal sand shoals
exposed by the ebb tide also occurs through wind action.
Wind-generated waves can have a significant effect on sedi-
ment production and movement along estuarine foreshores.
Within the relatively narrow confines of river estuar-
ies, wind-generated waves are small and of short wave
length (typically, the significant wave height is 0.3 m or
less). Nevertheless, these waves can cause extensive river-
bank erosion when the dominant wind direction coin-
cides with a long, straight and wide stretch of the estuary.
Wind waves are often the dominant transport mecha-
nism along the foreshores of tidal lakes (tidal and fresh-
water velocities in the body of the lake are generally low).
Inter-tidal sand and mudflats around the peripheries of
tidal lakes are formed as a result of wind waves and wind
induced currents.
5.7.4. Ocean Waves and Entrance Effects
The movement of sand into and out of the entrance of an
estuary is a complex phenomenon because of the interac-
tion of tidal, wave and freshwater transport processes.
Figure A.36 depicts the principal sediment pathways
around an estuary entrance.
S
LB
S
D
S
O
S
P
S
W
S
LB
S
A
S
P
S
P
S
A
estuary
c
o
n
t
r
o
l
v
o
l
u
m
e
E
0
2
0
2
1
8
c
S
LB
-
-
-
-
-
-
-
S
D
S
O
S
P
S
W
S
A
S
R
littoral drift and/or bank erosion
return of dredged spoil
entrance marine shoals and tidal delta sands
disposal of domestic and industrial effluents and solid wastes
wind erosion of coastal dunes and drying inter tidal shoals
decomposition and excretions of marine and river plants and animals
erosion of upper catchment and transport by rivers and streams
Figure A.35. Sources of sediment in an estuary (McDowell
and OConnor, 1977).
wrm_appA.qxd 8/31/2005 12:08 PM Page 525
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Waves breaking on the entrance bar bring into sus-
pension considerable volumes of sand that can be carried
into the estuary on the flood tide. Much of this sand is
usually flushed back out of the estuary on the ebb tide,
but a small amount may be deposited in the estuary,
where it will be transported upstream by the net flood-
tide transport caused by the tidal distortion.
Most of the sand transported out of the entrance
by floods will be deposited on or near the entrance
bar, from where it will be carried to the updrift coast-
line by a process known as littoral bypassing (see
Figure A.36). In addition to their stirring up sediments
from the entrance bar and promoting suspended
sediment transport, the significance of ocean waves to
estuarine sediment transport depends upon the entrance
conditions.
Where there is a well-developed offshore bar, ocean
waves will break on it and little wave energy will
penetrate the lower reaches of the estuary. In these
526 Water Resources Systems Planning and Management
circumstances, ocean waves will have little effect on
sediment transport within an estuary.
In contrast, ocean waves can readily penetrate the
lower reaches of drowned river valley estuaries, where
they combine with tidal flows to produce a relatively
complex pattern of sediment movement. In some tidal
lakes and rivers with relatively wide mouths, tidal sedi-
ment transport is enhanced by the penetration of waves
through the entrance.
5.7.5. Movement of Muds
Muds are cohesive materials, and hence the processes of
settling, consolidation and resistance to erosion are
dependent upon inter-particle bonding and the hydro-
dynamic environment.
Mud particles and aggregations (flocs) of mud particles
have small settling velocities. Turbulent mixing keeps
mud particles in suspension throughout the tide, except
main
ebb channel
E
0
2
0
2
1
8
d
ebb-tide
dominated
sedimentary
processes
enhanced littoral transport in
marginal flood tide channel
south to north littoral
drift on open beach
swash
platform
net on-shore transport
areas swash platform
2
3
4
5
6
8
7
4
3
2
9
ebb currents concentrate into
low tide channels around shoals
dominant flood tide currents
across top of inner bay shoal
build-up of inner bay shoals due
to flood tide transport, and
overall shoal sizes constant
when regime reached
average ebb and flood tide
transport equal when estuary
is in regime
localised reversal of littoral
drift due to marginal flood tide
channel and wave shadow area
marginal flood channel
littoral bypassing
outer portion of main channel
has strong net ebb tide transport
2 3
4
5
6
7
main ebb channel
8
terminal lobe
0
9
3 4
Figure A.36. Sediment
pathways at an estuary
entrance.
wrm_appA.qxd 8/31/2005 12:08 PM Page 526
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 527
at slack water when a concentrated layer of mud forms
near the estuary bed. Tidal distortion and gravitational
circulation result in a net upstream movement of mud
particles that are deposited around the limit of the gravi-
tational circulation current.
Muds will remain suspended in a body of moving
water when the vertical mixing due to turbulence is equal
to or greater than the rate of settling of individual parti-
cles. As velocities decrease, so does the vertical mixing
and deposition will increase.
The settling velocity of muddy sediment is dependent
upon its concentration. At high concentrations, more
collisions occur between particles and the resulting flocs
are larger and settle faster than the smaller particles. At low
concentrations, there is little or no interaction between
particles. In these cases, the settling velocity tends to be constant.
At very high concentrations, the particles actively interfere
with each other and hinder settling.
Like any sediments, muds begin to erode when the
shear stress imposed by water movement is equal to the
shear strength of the surface layer of mud. The resistance
of new mud deposits to subsequent erosion increases
with time. As consolidation occurs the cohesion among
individual particles and flocs increases, and hence their
resistance to erosion increases. Armouring also increases
with time and renders the deposit more erosion resistant.
In some cases, the strength obtained through
consolidation and armouring will be insufficient to resist
the erosive forces of the next half-tide cycle. In these
circumstances, all the material will be resuspended. In
other cases, some of the deposit will be eroded and
the remainder will be largely undisturbed. The residual
deposit may then remain largely undisturbed until the
next spring tide cycle, when it may or may not have gained
sufficient strength to resist the stronger tidal currents.
The pattern of mud deposition within an estuary
varies seasonally and spatially in response to the variable
nature of the salinity limit. Mud particles brought into
suspension are transported upstream by the gravitational
circulation. This produces a turbidity maximum near the
upstream limit of the circulation, leading to increased
deposition of muds at slack water. In other words, the
gravitational circulation leads to a trapping of muds in the
estuary. The location of the zone of maximum deposition
will vary with freshwater flow and the resulting pattern of
salt intrusion.
The mud trapping efficiency of an estuary drops off
quickly as freshwater flows increase. During major floods,
the high turbulence and flushing of the flood flows is
usually sufficient to keep the very high suspended load in
suspension and flush it entirely out to sea. Tidal lakes,
however, act as settling basins, and mud accumulates on
the lake bed after each flood.
Drowned river valley estuaries and artificially deep-
ened port areas in estuaries trap most of the mud brought
into the estuary by small floods. This can be a major con-
cern for the maintenance of shipping channels. However,
during major floods most of the mud is discharged to the
ocean. This is just as well, otherwise port areas might be
buried in mud by a single major flood event.
5.7.6. Estuarine Turbidity Maximum
Most estuaries contain a specific region having an extra
high concentration of suspended particles. The turbulent
area in the river where the tidal forces interact with
the fast-flowing fresh river waters creates a cloud of
suspended particulate matter (SPM): a mix of inorganic
sediment, organic detritus and living organisms such as
algae and zooplankton. This region is called the estuarine
turbidity maximum or ETM. The ETM is very important
for supporting biogeochemical, microbial and ecological
processes that sustain a dominant pathway in the
estuarys food web.
The turbidity maximum generally occurs near the
upstream limits of the salinity intrusion due to particle
trapping caused by circulation phenomena induced by
tides and density-driven (gravitational) currents. Tidal
shear also traps particulate matter. The turbidity maxi-
mum varies in location, spatial extent and particulate
matter concentration, depending on the tidal and river
flow conditions.
5.7.7. Biological Effects
Estuarine organisms influence both the settling and
resuspension of sediments. Sea-grass beds reduce current
velocities and dampen wave action near the bed, thereby
increasing the settling of fine material. Filter feeders such
as oysters ingest considerable amounts of fine suspended
organic and mineral particulate matter that would other-
wise remain in suspension. They eject waste in the form
wrm_appA.qxd 8/31/2005 12:08 PM Page 527
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
of pellets, which settle to the bottom. The activities of
some other animals have the opposite effect. The feeding
behaviour of many fish, birds and invertebrates disturbs
bed sediments and resuspends fine particles, thereby
increasing water turbidity.
The physical properties and stability of the surface
layer of bottom sediments are affected by the burrowing,
particle sorting and tube building activities (i.e. bio-
turbation activities) of benthic fauna such as crustaceans,
bivalves and polychaetes. Benthic invertebrates, bacteria
and diatoms produce mucus that binds sediment particles
together (Madsen et al., 1993; Meadows et al., 1990;
Rhoads, 1974).
These activities alter the texture, surface roughness,
density, water content and shear strength of bed
sediments, particularly muds (Rhoads and Boyer, 1982).
Many deposit feeders ingest sediment particles at some
depth and eject waste particles at the sediment surface.
This waste material may consist of fine particles that are
easily resuspended by currents or compacted fecal pellets
that behave in a similar way to sand grains.
The remains of dead organisms can also affect
sediment properties. In some areas bivalves are very
abundant, and their accumulated shells form a dense
layer that armours the sediment against erosive forces
(Rhoads and Boyer, 1982).
5.8. Surface Pollutant Movement
The previous discussion on advective and dispersive
transport and flushing processes that can take place in
estuaries applies to dissolved substances, such as salinity
and dissolved pollutants, as well as sediments. The behav-
iour of floating pollutants, such as oil, however, is quite
different.
Oil has a very low surface tension, hence its usefulness
as a lubricant. This, coupled with its hydrophobic (water-
repelling) nature, means that it tends to spread over the
waters surface as an unbroken thin film. Oil, when float-
ing on a water surface, undergoes three types of transport:
spreading
surface-current advection
wind-driven advection.
As oil spreads out and is transported across the surface of
an estuary, the more volatile fractions evaporate and the
528 Water Resources Systems Planning and Management
more water-soluble fractions dissolve or emulsify with
the water mass. Breaking waves facilitate emulsification.
In addition to spreading out over the water surface, the oil
film is carried along (advected) by surface water currents
associated with freshwater and tidal discharges. Variation
in surface velocities will result in additional spreading of
an oil slick by lateral and longitudinal dispersion.
Finally, wind effects are of major significance to the
spreading and transport of floating oil. Table A.4 shows
the spread of about 500 litres of oil under a wind speed
of 69 m/s. After three hours the spread covered 35 ha,
whereas under calm conditions it was estimated that the
slick would have covered only 2 ha.
The transport of some pollutants can occur in
the upper surface layers of the water column (the top
millimetre or so) or via surface slicks (compressed
micro-layers). Organochlorines and heavy metals have
been recorded in micro-layers at concentrations of up to
10,000 times greater than that which would normally
occur in the water column (Hardy et al., 1990; Szekielda
et al., 1972). Such concentrations may have adverse effects
on planktonic organisms that gather at the sea surface.
Assessing the extent, concentrations and possible impacts
of micro-layers is both difficult and expensive, but their
role in estuarine pollution is receiving increasing attention.
5.9. Estuarine Food Webs and Habitats
The tidal, sheltered waters of estuaries support unique
communities of plants and animals that are adapted for
life at the margin of the sea. Estuarine environments are
time
(hrs)
extent
of slick (ha)
1 2.5
3 35.0
5 78.0
10 265.0
E
0
2
0
2
1
8
e
Table A.4. Spread of 500 litres of oil under windy conditions
(wind speed 69 m/s) (Cormack and Nichols, 1977).
wrm_appA.qxd 8/31/2005 12:08 PM Page 528
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 529
among the most productive anywhere, creating more
organic matter each year than comparably sized areas of
forest, grassland or agricultural land. The productivity
and variety of estuarine habitats foster an abundance and
diversity of wildlife. Shore birds, fish, crabs and lobsters,
marine mammals, clams and other shellfish, marine
worms, sea birds and reptiles are among the animals that
make their homes in and around many estuaries. These
animals are linked to one another and to an assortment of
specialized plants and microscopic organisms through
complex food webs and other interactions. For example,
shore birds use their long bills to obtain fish, worms,
crabs or clams. Within the mud, silt, sand or rock
sediments live microscopic bacteria, a lower level of the
food web, consuming decaying plants.
Many different habitat types are found in and around
estuaries, including shallow open waters, freshwater and
salt marshes, sandy beaches, mud and sand flats, rocky
shores, oyster reefs, mangrove forests, river deltas, tidal
pools, sea-grass and kelp beds, and wooded swamps.
The resident and visiting organisms in estuaries must
be able to tolerate frequent and rapid changes in salinity,
currents, temperatures and water levels. At high tide,
seawater changes estuaries, submerging the plants and
flooding creeks, marshes, mudflats or mangroves. The
incoming waters seemingly bring back to life organisms
that have sought shelter from their temporary exposure to
the non-aquatic environment. As the tides ebb, organisms
return to their protective sediments and adjust to chang-
ing temperatures.
Food webs include primary producers and primary,
secondary and tertiary feeders, ranging from single-celled
algae to the highly efficient predators at the top of
the chain. Ecologists refer to these increments as trophic
levels. The term trophic means, simply, pertaining to
food. The ways in which food is consumed that is, the
pattern of consumption and how it changes with time is
called trophic dynamics.
Inasmuch as an estuary is an environment character-
ized by rapid and frequent change, which leads to biolog-
ical diversity, food webs and trophic dynamics in
an estuary are complex. Unlike the open ocean, where
phytoplankton species are usually the sole primary
producers, estuarine systems usually contain several types
of primary producers. In addition to phytoplankton, these
include sea-grasses and salt-marsh plants. Zooplankton
graze on phytoplankton. These, in turn, become food
for plankton-eating fishes, such as herring, smelt, and
the larvae and young of larger fishes. These, then, become
food for carnivores and omnivores nearer the top of the
food web.
Some animals graze on the larger estuarine plants, but
such plants are probably more important food sources
after they die and begin to decompose. Here, bacteria
and fungi promote the breakdown of the dead plant
material. This organic detritus is an essential source of
nutrition for detritus-eating animals and supports a
detrital food web.
Benthic, or bottom-dwelling, and bottom-oriented
organisms are other important components in estuarine
food webs. Clams, for example, reside in the bottom
sediments and feed on plankton and other organic matter
which they filter from the water. Oysters and mussels are
other filter feeders.
Bottom deposit feeders, such as the various kinds of
worms found in the estuary, move over and through
bottom sediments where they find food deposited in or
on the sediments. Shrimps, crabs and other invertebrates
are well adapted to bottom feeding, as are many of
the estuarine fishes, such as sculpin, flounder, sole and
sturgeon. In fact, most fish species that reside in estuaries
or move into them on feeding forays are bottom oriented
in their feeding patterns.
Near the top of the estuarine food web are various
carnivores and omnivores. Some show marked food
preferences, while others are opportunistic feeders. These
may include diving birds, wading birds, waterfowl, gulls,
terns, pelicans, ospreys, trout, perch, striped bass, sharks
and salmon. At the top of the food web can be seals, sea
lions, whales, larger sharks and humans.
5.9.1. Habitat Zones
Figure A.37 illustrates various ecosystem habitats
associated with an estuary. Shown on the figure are nine
different habitats.
Fish habitats are mainly underwater. Although not
readily visible, they play an important role in an estuary
and in the economy of the region. There are many
different types of habitats on which fish depend for their
survival, ranging from deep ocean water to the shallower
areas near the coast to rivers and lakes far inland. Like one
wrm_appA.qxd 8/31/2005 12:08 PM Page 529
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
giant mosaic, habitat pieces are linked together and
support one another. Habitat connectivity is critical.
What affects one habitat will in turn eventually affect
other, connected habitats.
Coastal waters can support commercial and recre-
ational fishing industries. They can also support a variety
of aquaculture activities such as shellfish culture or fish
farms. In many communities, fishing is an important
contributor to the regular diet of the local population.
Freshwater habitats are not much different from their
marine counterparts in supporting a wide and diverse fish
population. Lakes provide deep and stable habitats of
rocky, sandy and muddy bottoms. Aquatic plants provide
food, shelter and spawning areas, as they do in the oceans
and estuaries. Unpolluted rivers and streams can offer a
habitat of clean, well-oxygenated water. Insects living and
falling in the water are a major food source for fish. Fish
use the rivers to travel between spawning, rearing and
feeding habitats.
Pelagic fish such as mackerel, tuna and herring that
live their entire lives just below the surface of the ocean
often follow the ocean currents, which are like rivers in
the ocean. The fish habitat in these areas can change
rapidly due to the actions of the wind, rain, sun and air
temperature. The fish follow these rivers in the ocean to
530 Water Resources Systems Planning and Management
areas that are suitable for them to reproduce and feed.
They lay their eggs in the open sea or shallow coastal
waters. When the eggs hatch, the young drift with the
ocean currents. Typically, out of millions that hatch from
eggs, only a few survive to become adults.
Many nutrients brought down from streams and rivers
pass through estuaries and salt marshes before reaching
the oceans. These are among the most fertile areas found
anywhere for plants and animals. Coasts that contain
extensive estuaries and salt marshes support the fisheries
on which many depend. The nutrients fertilize large
expanses of the plants along shores. These are the main
feeding, growing and sheltering areas for young fish and
shellfish. After hatching in freshwater, young smelt drift
downstream to the estuaries where they hide and feed
among aquatic plants.
Cod, halibut and other fish that live on or near the
bottom of the ocean are called benthic fish or ground-
fish. Others that are harvested from these deep-water
areas can include snow crabs and scallops. In the waters
nearer to the shore, the fish habitat is a mixture of rock
and mud. This near shore mixture can be the home to
lobsters, rock crabs, shellfish and smaller fish. The
physical and chemical conditions in these areas typically
do not change very quickly. If any sudden changes do
fish habitat coastal waters freshwater habitat
pelagic zone waterways estuaries
& salt marshes
benthic zone inter tidal zone coastal shallows
fish habitat coastal waters freshwater habitat
pelagic zone waterways estuaries
& salt marshes
benthic zone inter tidal zone coastal shallows
E
0
2
0
2
1
9
a
Figure A.37. Typical habitats of
an estuary and its surroundings.
wrm_appA.qxd 8/31/2005 12:08 PM Page 530
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 531
happen, the effects on the fish habitat and the fish
themselves can be severe.
Shallow water habitats are usually a mixture of mud,
rocks and gravel. Animals like oysters, clams and mussels
that filter the water for food can find abundant food in
this habitat. They feed on microscopic animals and plants
called plankton. They spend their lives in one place, so
their habitat is extremely important for shelter and
feeding. Urban or commercial developments along coasts,
poorly designed wastewater treatment systems, sewage
outfalls and infilling of marshes and bays can all degrade
the ecosystem habitat in these areas. The filter-feeding ani-
mals in these areas are very sensitive to changes in their
habitats such as a lack of oxygen, increases in tempera-
tures or being buried by mud.
In the shallower waters and rocky shores, the
increased temperature, sunlight and wave action results
in more types of habitats that support more kinds of
plants and animals. Plants become very important in this
area, transferring the suns energy through plant-eating
animals. This inshore area supports a wide variety of
animals that use the plants for food, refuge, hunting
grounds and reproduction. For example, periwinkles,
urchins and starfish scrape the algae off the rocks and
plants, and also lay their eggs on them. These eggs are
sources of food for crabs, lobsters and other fish.
5.10. Estuarine Services
Besides serving as important habitats for wildlife, estuar-
ies and the wetlands that often fringe them perform other
valuable services. Water draining from the uplands carries
sediments, nutrients and other pollutants. As the water
flows through fresh and salt marshes, much of the sedi-
ment and pollution is filtered out, creating cleaner and
clearer water that benefits both people and marine life.
Wetland plants and soils also form a natural buffer zone
between the land and ocean, absorbing floodwaters and
dissipating storm surges. This helps protect upland
organisms as well as valuable real estate from storm and
flood damage. Salt marsh grasses and other estuarine
plants also help prevent erosion and stabilize the shore-
line.
Among the cultural benefits that estuaries offer are
recreation, a source of scientific knowledge, education
and aesthetic aspects. Boating, fishing, swimming, surfing
and bird watching are just a few of the recreational activ-
ities people enjoy there. Estuaries are often the cultural
centres of coastal communities, serving as the focal points
for local commerce, recreation and culture. They also
provide aesthetic enjoyment for the people who live,
work or holiday in and around them.
Finally, estuaries provide tangible and direct economic
benefits. Tourism, fisheries and other commercial activities
thrive on the wealth of natural resources estuaries supply.
The protected coastal waters of estuaries also support
important public infrastructure, serving as harbours and
ports vital for shipping, transportation, and industry.
To maintain and enhance these and other services and
benefits derived from estuaries, they must be managed.
This often includes protection and restoration.
5.11. Estuary Protection
The economy of many coastal areas is based primarily on
the natural beauty and bounty of estuaries. When those
natural resources are imperilled, so too are the economies
and well-being of people who live and work there. In
many countries that have coastlines, most of their popu-
lations are concentrated along them. Populations in
coastal regions are growing faster than in inland regions
throughout most of the world.
Population pressures and the impact of human
encroachment on estuaries, as shown in Figure A.38, is
evident. In North America, for example, some 70% of the
original sea-grass meadows and salt marshes have been
lost in Puget Sound, Galveston Bay and Narragansett Bay.
Over 90% of original wetlands have been lost in San
Francisco Bay, Chesapeake Bay, Hudson-Raritan Estuary,
and Tampa Bay (USEPA, 2001).
This increasing concentration of people living along
coastlines tends to change the natural balance of estuarine
ecosystems and threatens their integrity. Channels are
dredged, marshes and tidal flats are filled, waters are
polluted, and shorelines are reconstructed to accommo-
date human housing, transport and recreational needs.
Stresses caused by overuse of resources and unchecked
land use practices in some areas have resulted in unsafe
drinking water, closing of beaches and shellfish beds,
harmful algal blooms, unproductive fisheries, loss of
habitat, death of fish and wildlife, and a host of other
human health and natural resource problems.
wrm_appA.qxd 8/31/2005 12:08 PM Page 531
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Polluted runoff from rural, suburban and urban areas
upstream of estuaries can also cause damage. Stormwater
can pick up contaminants from roads, vehicles, lawns and
construction sites, and discharge them into streams, lakes
and wetlands. Everything that happens on land in the
watershed that drains into the estuary can affect its habitat.
When stream channel habitat is degraded, say from
agriculture, construction or forestry activities, fish die
because their nesting and feeding areas are destroyed. In
urban harbours, in particular, polluted runoff creates hot
spots of toxic contamination where relatively few species
can live. For example, polluted runoff from agricultural
chemicals is responsible for the dead zone off the coasts
of Texas and Louisiana, which extends some 11,000 km
2
into the Gulf of Mexico. Airborne pollution that falls on
estuaries also contributes to their contamination.
Dams blocking river flows can restrict the upstream
and downstream passage of migrating fish, isolating them
from spawning and feeding areas. The construction of
dams accounts for significant and ongoing loss of habitat
in the watersheds of many of the worlds rivers and estu-
aries. Without healthy streams, estuaries cannot receive
their normal allocation of nutrients. The result is usually
reduced productivity. If fewer fish return to the estuary,
many living organisms in the food web that depend on
healthy fish populations will suffer.
532 Water Resources Systems Planning and Management
Excessive discharges of pollutants in wastewater
effluent from city and industrial sewage treatment plants
can degrade estuary ecosystems that require clean water
to survive and thrive. This is evident in many if not most
of Western Europes major estuaries; examples include
the Tyne, Tees, Humber, and Thames Estuaries in the UK;
the Scheldt Estuary in Belgium and the Netherlands; the
Ems Estuary in Germany and the Netherlands; the Weser
and Elbe Estuaries in Germany; and the Seine Estuary
in France. Wastewater treatment can help restore some of
the damage done by excessive pollution. When the
treatment plants discharging into Tampa Bay (Florida)
were upgraded to advanced technology, the sea-grasses
85% of which had been destroyed began to grow back,
and along with them the fish and other creatures that
depend on those grasses. Estuaries can be restored, but it
takes time as well as money.
Shifts in climate and the altering of stream channels
by humans cause an annual loss of tens of thousands of
hectares of estuary habitat. This is of concern along the
Louisiana coast of the United States. Changing the course
of the Mississippi River as it enters the Gulf of Mexico along
the coast of Louisiana has substantially reduced the yearly
input of sediment to its delta. The result has been an open-
sea encroachment of both the estuaries and dry land.
Louisiana estuaries lose about 12,000 hectares of land each
Figure A.38. Not all estuaries
remain natural. Some show the
impacts of development and the
desire for humans to live near and
on them. (South Florida Water
Management District, West Palm
Beach, Florida.)
wrm_appA.qxd 8/31/2005 12:08 PM Page 532
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 533
year to erosion and subsidence (actual sinking of the land
into the water). Delta erosion and loss has also been a prob-
lem in the River Nile Estuary since the building of the High
Aswan Dam reduced the sediment loads normally carried
by the Nile River to the Mediterranean Sea.
Estuarine habitat and freshwater loss is a major chal-
lenge to estuarine management throughout the developed
and developing world. Population growth and economic
development will almost certainly lead to additional loss
of habitat and freshwater diversion in many estuaries. The
same social driving factors, combined with a continuing
migration of human populations to coastal urban centres,
require major expansions of engineering infrastructure.
Nutrient emissions from human activities can greatly
exceed even those naturally carried by large rivers. An
increasing number of coastal areas are already manifesting
serious effects of nutrient over-enrichment, including
bottom-water hypoxia or anoxia, undesirable algal
blooms, and the loss of sea-grasses and coral reefs.
5.12. Estuarine Restoration
The most cost-effective route to saving estuaries is to
prevent habitat alteration and destruction in the first
place. But because of the substantial loss of vital estuarine
habitat, and the habitat that continues to be destroyed,
restoration is often necessary.
Restoration means returning an area of estuary habitat
to a successful, self-sustaining ecosystem with both clean
water and healthy habitats that support fish and wildlife
and human uses of the estuary, such as swimming,
boating and recreational and commercial fishing and port
activities. Ecological restoration does not usually focus on
a single species but strives to replicate the original natural
system. The goal is to help rebuild a healthy, functioning
system that works as it did before it was polluted or
destroyed. Restoration also means an actual increase in
the quantity of high-quality estuary habitats, as measured
both by their surface area and by their ability to support
a resilient healthy ecosystem.
Restoration activities in estuaries range from the
simple to the complex. They may include, singly or in
combination:
baseline assessments
setting performance standards and defining long-term
monitoring and conservation plans
restoring the physical and hydrological conditions
through engineered activities, often involving heavy
equipment and the returning of tidal waters
reducing inflows of nutrients, BOD and other
pollutants
chemical cleanup of toxic substances
revegetation of an area through native plantings or
natural regrowth.
Restoration of an estuary is most effectively done by
communities that live in the watersheds that impact that
estuary. These communities should take a watershed
approach to estuary restoration. Estuaries are often at the
heart of local economies and traditions. Although they
may need additional financial and technical assistance
from federal, state and local governments, the people who
live near estuaries are the ones who will determine just
how successful any restoration effort will be.
5.13. Estuarine Management
To achieve estuarine protection and restoration and all
the benefits that can be derived from them, estuaries
must be managed. Management must begin with some
attention to the impacts of daily activities of people who
live in watersheds that drain into the estuaries. These
activities can affect the ecological habitat of waterways
and estuaries. Everything that enters into a stream or
river or any other type of waterway that drains into an
ocean will eventually find its way to the estuaries and
oceans, and thus can eventually affect the estuarine
environment. Pollution or even mud runoff in the water
can affect aquatic habitats downstream and interfere
with plant and animal reproduction. For example, con-
taminants washing downstream can cause the closure of
shellfish beds for human consumption or kill large
numbers of fish.
Although each estuary is unique, many face similar
environmental problems and challenges that include
urban and commercial development, over-enrichment of
nutrients, pathogen contamination, toxic chemicals, alter-
ation of freshwater inflow, loss of habitat, declines in fish
and other wildlife, and introduction of invasive species.
The following discussion provides a brief overview of
the most common problems facing estuary management
and restoration. These problems are primarily caused by
human activities.
wrm_appA.qxd 8/31/2005 12:08 PM Page 533
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
5.13.1. Engineering Infrastructure
Many engineering developments such as harbours, train-
ing walls, navigation channels, reclamation and dredging
typically take place in the mouths of estuaries. By virtue
of their ability to alter depths significantly in the most
tidally sensitive reach, such developments can affect tidal
behaviour along the entire estuary. Often complex hydro-
dynamic-ecological models are needed to estimate the
impacts of such infrastructure.
5.13.2. Nutrient Overloading
Nutrients such as nitrogen and phosphorus are necessary
for the growth of plants and animals, and to support a
healthy aquatic ecosystem. In excess, however, these
nutrients can contribute to fish disease, red or brown tide,
algae blooms and low levels of dissolved oxygen. The
condition where dissolved oxygen is less than 2 mg/l is
referred to as hypoxia. Many species are likely to die
below that level. The concentration of dissolved oxygen in
healthy waters is at least 5 or 6 mg/l.
Nutrients can come from both point sources and non-
point sources such as sewage treatment plant discharges,
stormwater runoff from lawns and agricultural lands,
faulty or leaking septic systems, sediment in runoff,
animal wastes, atmospheric deposition originating from
power plants or vehicles, and groundwater discharges.
Excessive nutrients stimulate the growth of algae. As the
algae die they decay, and this decreases the dissolved
oxygen in water. Algae also reduce sunlight penetration
in the water. Fish and shellfish deprived of oxygen, and
underwater sea-grasses deprived of light, will die.
Animals that depend on sea-grasses for food or shelter
leave the area or die. In addition, the excessive algal
growth may result in brown and red tides that have been
linked to fish kills, manatee deaths and negative impacts
on scallops. Decaying algae may also cause foul smells
and decreased aesthetic value.
5.13.3. Pathogens
Pathogens such as viruses, bacteria and parasites, as well
as certain types of algae, can be toxic and may cause
diseases. When found in marine waters they can pose a
health threat to swimmers, surfers, divers and seafood
534 Water Resources Systems Planning and Management
consumers. Fish and filter-feeding organisms such as
shellfish concentrate pathogens in their tissues, and
people eating them may become ill. Pathogen contamina-
tion can result in the closure of shellfishing areas and
bathing beaches.
Sources of pathogens include urban and agricultural
runoff, boat and marina waste, faulty or leaky septic
systems, sewage treatment plant discharges, combined
sewer overflows, recreational vehicles or campers, illegal
sewer connections, and waste from pets or wildlife.
5.13.4. Toxic Chemicals
Toxic substances such as metals, polycyclic aromatic
hydrocarbons (PAHs), polychlorinated biphenyls (PCBs),
heavy metals, and pesticides are a concern in the estuar-
ine environment. These substances can enter waterways
through stormdrains; discharges from industry and
sewage treatment plants; runoff from lawns, streets and
farmlands; and deposition from the atmosphere. Many
toxic contaminants are also found in sediments and are
resuspended into water by dredging and boating
activities. Bottom-dwelling organisms are exposed to these
chemicals and if consumed may pose a risk to human
health. As a result there may be fishery and shellfish bed
closures, and bans on the consumption of certain foods.
5.13.5. Habitat Loss and Degradation
The continued health and biodiversity of marine and
estuarine systems depends on the maintenance of high-
quality habitats. The same areas that often attract human
development also provide essential food, cover, migratory
corridors and breeding/nursery areas for a broad array of
coastal and marine organisms. In addition, these habitats
perform important functions such as enhancement of
water quality and flood protection, and water storage.
Estuarine ecosystems can be degraded through loss
of habitat such as the conversion of a sea-grass bed to an
island of dredged material or through a change or degra-
dation in structure, function or composition. Threats to
habitat include commercial and residential development,
highway construction, marinas, dyking, dredging, damming
and filling. Wetland loss and degradation caused by
dredging and filling reduces the amount of habitat available
to support healthy populations of wildlife and marine
wrm_appA.qxd 8/31/2005 12:08 PM Page 534
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 535
organisms. All of these activities may result in an increase in
the runoff of sediments, nutrients and chemicals.
5.13.6. Introduced Species
Intentional or accidental introduction of invasive species
can often result in unexpected ecological, economic and
social damages. Through predation and competition,
introduced species have contributed to the eradication of
some native populations and substantially reduced others,
fundamentally altering the food web. Overpopulation of
some introduced herbivores has resulted in overgrazing
of estuarine vegetation and the resultant degradation and
loss of marsh. Other impacts can include:
alteration of water tables
modification of nutrient cycles or soil fertility
increased erosion
interference with navigation, agricultural irrigation, sport
and commercial fishing, recreational boating, beach use
possible introduction of pathogens.
Sources of introduced species include ship ballast, mari-
culture and aquarium trade.
5.13.7. Alteration of Natural Flow Regimes
Alteration of the natural flow regimes of tributaries can
have significant effects upon water quality and the distri-
bution of living organisms within estuaries. Freshwater is
an increasingly limited resource in many areas. Human
management of this resource has altered the timing and
volume of inflow to some estuaries. Too much or too little
freshwater can adversely affect fish spawning, shellfish
survival, bird nesting, seed propagation, and other sea-
sonal activities of fish and wildlife. In addition to changing
salinity levels, inflow provides nutrients and sediments
that are important for overall productivity of the estuary.
5.13.8. Declines in Fish and Wildlife Populations
The distribution and abundance of estuarine fish and
wildlife depend on factors such as light, turbidity, nutri-
ent availability, temperature, salinity and food availability.
Natural and human-induced events that disturb or
change environmental conditions affect the distribution
and abundance of estuarine species. Declines in fish and
wildlife populations have resulted from fragmentation
and loss of habitats and ecosystems, pollution and
decreased water quality, over-exploitation of resources,
and the introduction of exotic or nuisance species.
Habitat loss and degradation can lead to decreases in the
stocks of sport and commercial fish and shellfish, changes
in the populations of fur-bearing and waterfowl species,
and decreasing habitat for migratory birds and other
species. Pollutants such as herbicides, pesticides and other
wastes pose a threat to living resources by contaminating
the food chain and eliminating food sources. Runoff from
farms and cities, and toxic releases, can alter aquatic habi-
tat, harm animal health, reduce reproductive potential, and
render many fish unsuitable for human consumption.
Other threats to wildlife include oil spills, bioaccumu-
lation of toxins, outbreaks of contagious and infectious
diseases, and algal blooms such as red and brown tides.
Over-exploitation occurs when fisherman, trappers,
hunters or collectors take so many individuals of a species
that its ability to maintain stable population levels is
impaired. Introduced species compete with native species
for food and habitat. Other causes of decline in fish and
wildlife populations include agricultural and logging
activities, trawling, boat disturbances, entanglement from
marine debris, and changes in freshwater inflow.
6. Coasts
Many factors threaten the resilience of naturally dynamic
and generally fragile coastal systems. Most originate from
land development and urbanization, pollution, overuse
and accelerated sea level rise. The challenge in coastal
zone management is to maintain the natural physical
variability and ecosystem diversity present in, and all
the resulting benefits derived from, coastal systems, while
simultaneously satisfying multiple stakeholder interests
and competing resource uses and values that often favour
the stabilization of such zones.
6.1. Coastal Zone Features and Processes
The coast is where land and ocean interact. It is a
zone extending inland to the limit of tidal or sea spray
influence. The relative levels of the sea and land
determine coast locations. In relation to the land, sea
wrm_appA.qxd 8/31/2005 12:08 PM Page 535
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
levels fall and rise as glaciers grow and melt over periods
of thousands of years. When sea levels rise, as is being
observed today, the coasts move inland unless prevented
from doing so by barriers. Most efforts to protect the
coasts are oriented toward preventing this inland
movement and destruction of property property that is
often very desirable and hence very valuable. But coastal
beaches are naturally dynamic, adjusting themselves,
through erosion and deposition, to the variable and often
random forces they are subjected to. Coasts are shaped
by land and marine processes, which in turn are increas-
ingly being influenced by human activities and structures
that tend to constrain these natural dynamic processes.
Preventing erosion, for example, will eventually destroy
the beaches, and beaches provide much more than just
recreational benefits, as will be discussed below.
Land processes have primarily shaped the irregular,
glacially scoured, rocky coasts typically found in higher
latitudes. These are characterized by long narrow embay-
ments carved into bedrock. Other coasts have a combined
marine and land origin, or a primarily marine one. The
Outer Banks of North Carolina on the US Atlantic coastline
are barrier islands formed by the rising sea. The Pamlico
and Albemarle Sounds behind them, on the other hand,
are simply a series of flooded river valleys. The Mississippi
Delta comprises sediment that is transported by the
Mississippi River into the Gulf of Mexico, where it is
sculpted by waves into long fingers of land protruding into
the sea. The same is true of the Nile River Delta in Egypt.
Coasts formed by marine processes include much
of the US Pacific Coast, with its cliffs formed by wave
erosion. Outer Cape Cod, jutting into the Atlantic Ocean
in Massachusetts, is another coast formed by the wave
erosion of a deposit of glacial sands and gravel. A
predominantly marine-formed coast in North America
extends from Long Island, New York, to Mexicos
Yucatan, with only a few interruptions. These barrier
islands just offshore of the eastern coast of the United
States occupy more shoreline distance than any other
open-ocean shoreline type. The islands are formed by the
waves, the wind and the offshore currents.
6.1.1. Water Waves
Every coastal project such as beach nourishment or
harbour design requires information on the wave condi-
tions in the region of interest. Increasing demands for
536 Water Resources Systems Planning and Management
accurate design wave conditions and for input data for
the investigation of sediment transport and surf-zone
circulation have resulted in improved wave-prediction
models during the last two decades. Liu and Losada
(2002), Dean and Dalrymple (1984), Dingemans (1997),
Mei and Liu (1993), Horikawa (1978), Ippen (1966),
Kinsman (1965), and Sarpkaya and Isaacson (1981)
discuss various theoretical wave models and cite original
papers.
In spite of model improvements, the prediction of
waves and their effects is still relatively primitive in com-
parison with the complexity of the real system.
Numerical models are still based on simplified governing
equations, boundary conditions and numerical schemes,
imposing different restrictions to practical applications.
The computational effort required to solve a truly three-
dimensional wave propagation problem, involving numer-
ous physical processes resulting in hundreds or more
wavelengths with different temporal and spatial scales, is
still too large to be feasible in engineering design practice at
this time.
Local currents, accelerations and pressure fluctuations
often accompany water waves. Their simplest form is
sinusoidal (Figure A.39).
Small amplitude wave theory is based on the
sinusoidal wave shown in Figure. A.39. A right-hand
coordinate system is commonly used, with its origin at
still-water level. Still-water level, SWL, is the water surface
that would exist in the absence of any wave action. The
x-axis is horizontal and parallel to the direction of wave
propagation and no variation is assumed in the y direc-
tion, perpendicular to the x-axis.
The sinusoidal water surface level, , at any time t at
location x may be described by:
(A.2)
where a is the amplitude of the wave, x is distance in the
direction of wave propagation, t is time, L is the wave-
length, (pi) radians is 180 and T is the wave period.
The maximum vertical distance between crest and
trough of the wave is the wave height, H (2a). The dis-
tance over which the wave pattern repeats itself is the
wavelength, L. The waves propagate with a velocity, C,
and the time required for a wave to pass a particular
location is the wave period T (T L/C). The inverse of
the wave period is the wave frequency f (f 1/ T). The

a
x
L
t
T
cos
2 2

wrm_appA.qxd 8/31/2005 12:08 PM Page 536

WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 537
ratio of wave height to wavelength (H/ L) is called wave
steepness.
The high water levels are the wave crests; the low
levels are the wave troughs. The crest of waves is
measured in the y direction. Swell is normally charac-
terized by long crested waves, each extending in the y
direction over a hundred metres or so. Sea normally has
short-crested waves with local peaks.
Finally, waves are said to be in deep water when
the depth of the water per unit wavelength, L, is greater
than 0.5 and in shallow water when it is less than
0.05. Between these conditions, the water depth is called
transitional.
Water waves range from capillary waves that have
very short wave periods (order 0.1 second) to tides,
tsunamis (earthquake generated waves) and seiches
(basin oscillations), where the wave periods are
expressed in minutes or hours. Waves also vary in
height, from a few millimetres for capillary waves to tens
of metres for the long waves. Gravity or wind-generated
waves are in the middle of that range. They have periods
from 1 to 30 seconds and wave heights that are seldom
greater than 10 m and mostly of the order of 1 m. They
are generated by wind against the gravitational force
attempting to restore the still-water level. These gravity
or wind-generated waves account for most of the total
available wave energy.
The actual shape of a water surface subjected to wind
does not look like Figure A. 39. It is a complex combina-
tion of many waves of many different wave heights and
periods. These waves propagate more or less in the wind
direction. Local peaks in the water levels occur when and
where the two waves combine, and lower water levels
occur when and where the waves separate, resulting in
the irregular wave pattern at any particular time and loca-
tion. Even when the first puffs of wind touch a flat water
surface, the resulting distortions present non-linearities
that make rigorous analysis extremely difficult, if not
impossible. When the first ripples generated by these
puffs are subsequently strengthened by the wind and
interact with each other, the result is known as a confused
sea. As the wind speed and duration increase, the com-
plexity of the waves will increase. Coastal zone modellers
need to understand this confusion.
Wind waves generated by a storm over the sea or
ocean usually consist of a wide range of wave frequencies.
Wave components with higher wave frequencies propa-
gate at slower speeds than do those with lower ones. As
they propagate towards the coast, long waves lead the
wave group and are followed by short waves. In deep
water, wind-generated waves are not affected by the
bathymetry. Upon entering shallow waters, however, they
are either refracted by bathymetry or current, or diffracted
around abrupt bathymetric features such as submerged
ridges or canyons. A part of wave energy is reflected back
to the deep sea. Continuing their shoreward propagation,
waves lose some of their energy through dissipation near
the bottom. Nevertheless, each wave profile becomes
steeper with increasing wave amplitude and decreasing
wavelength.
In very shallow water the front face of a wave moves at
a slower speed than the wave crest, causing the overturn-
ing motion (the breaking) of the crest. Such an over-
turning motion usually creates a jet of water, which falls
near the base of the wave and generates a large splash.
Turbulence associated with breaking waves is responsible
not only for the energy dissipation but also for the
sediment movement in the surf zone.
On large bodies of water, waves can travel well beyond
the area in which they are generated. For example, waves
generated by a storm off the coast of Newfoundland in
Canada may travel easterly toward Europe, eventually
arriving with much less energy on the coast of Portugal.
Wave energy frequencies are reduced. The resulting
waves arriving in Portugal, for example, will be more
orderly than the initial sea generated off Newfoundland,
with longer wave periods (1020 seconds), smaller wave
heights and more pronounced wave grouping. Waves
generated some distance away are called swell.
wave
crest
wave
trough
L
H
a
C
SWL x
y
z
E
0
2
0
8
0
9
c
Figure A.39. Basic wave components and characteristics
include wave height, H, length, L, amplitude, a, relative
to still-water level, SWL, and velocity C.
wrm_appA.qxd 8/31/2005 12:08 PM Page 537
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
6.1.2. Tides and Water Levels
Water levels influence both flooding and wave exposure.
Most damage to structures along coastlines occurs when
the water levels are high. Rising water levels expose shore
structures to larger waves because the water depth deter-
mines where waves break and thus where they release
most of their energy. The increased forces and over-
topping of water may damage coastline structures and
areas behind it. Conversely, when the water level drops,
the same structures may not be exposed to waves at all.
Similarly, high water levels can cause sandy shores to
retreat, even if the shores are backed by substantial dunes.
Higher water levels allow larger waves to come closer in
to the shore. These waves erode dunes and upper beach
and deposit the sand offshore. If the water level rise is
temporary, most of this loss will be regained at the next
low water. Permanent water level rise, however, will
result in permanent loss of sand. Shorelines consisting of
bluffs or cliffs of erodible material, such as glacial till or
soft rock, are continually eroded by wave action. High
water levels, however, will allow larger waves to attack the
bluffs directly, increasing the rate of shoreline recession.
6.1.3. Coastal Sediment Transport
Although there are many important aspects of coastal
zone management, the most important driver and funda-
mental criterion in a coastal zone management plan is
often the movement of sand and other sediment.
The transport of sediment, moved by waves and
wind, may be divided into cross-shore and long-shore
components. Sediment movement can result in erosion or
accretion: removal or addition of volumes of sand.
Erosion normally results in shoreline recession (move-
ment of the shoreline inland). Deposition of sediments on
beaches (accretion) causes the shoreline to move out to
sea. Most protection schemes do not function well with
too much cross-shore sediment movement. In particular,
if the main cause of shoreline recession is systematic
movement of sand offshore, protection becomes difficult.
6.1.4. Barrier Islands
Barrier islands, as illustrated in Figure A.40, are ribbons
of sand formed by rising sea levels on coastal plains
with low, flat surfaces extending to the edge of the sea,
538 Water Resources Systems Planning and Management
a large supply of sand, and waves large enough to move
the sand about. Barrier island migration towards the
mainland occurs as the barrier island rolls over itself
like a wheel. The ocean side of the island retreats as
storms push sand across the island to form sand over-
wash fans. Overwash fans often extend into the lagoon
behind the island and may cause the island to widen in
a landward direction. Simultaneous shoreline retreat of
the side of the island open to the ocean results in island
migration.
6.1.5. Tidal Deltas and Inlets
Inlets usually form during storms when stormwaters that
have been forced into estuaries behind a barrier island
rush across the island on their return to the sea, cutting
a channel through the island. Figure A.40 illustrates a
natural inlet.
Tidal deltas, as illustrated in Figure A.40, are bodies of
sand formed at inlets by tidal currents. Flood tidal deltas
are formed by sand forced into the lagoon between bar-
rier islands and the mainland by incoming or flood tides.
The ebb tidal delta is the body of sand pushed seaward by
the outgoing or ebbing tide. Salt marshes or mangroves,
where mangroves grow and eventually add to the width
of the barrier islands, colonize flood tidal deltas.
Overwash fans also add to the width of barrier islands
once the inlet migrates or opens to the sea.
When an inlet is jettied to stabilize a navigation
channel, the ebb tidal delta is broken up. For a few years
or decades after jetty construction, the sand from the ebb
tidal delta ends up being deposited on adjacent shore-
lines. Eventually a new and completely submerged tidal
delta forms far offshore at the end of the jetties, taking
sand permanently away from the beach.
6.1.6. Beaches
Beaches are strips of unconsolidated material found at the
seaward margin of the coast. The location of the beach
shoreline, the wetdry boundary of the beach, moves up
and down with the tide. (Maps usually show the mid-tide
line or mean sea level.) A beach on a barrier island or spit
is called a barrier beach, and beaches between rocky
headlands are called pocket beaches.
Beach materials are subjected to wind, waves and
currents, which shape the beaches accordingly. They
wrm_appA.qxd 8/31/2005 12:08 PM Page 538
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 539
As shown in Figure A.41, beaches are steeper during
calm weather, and flatter during storm events. All beaches
with a reasonably good sand supply and no seawall will
respond to strong storm waves by either flattening or
forming offshore bars. Beaches that flatten in response to
large waves cause breaking waves to dissipate their energy
over a broader surface. Waves that reach cliffs before
having their energy dissipated must dissipate their energy
over a relatively short distance. The result of course is a
spectacular rise of water into the air. Offshore bars are
small ridges of sand parallel to the shore that can also help
dissipate the energy of breaking waves. Their presence is
evident when incoming waves break one or more times
before reaching the shore.
Beaches usually recover after each storm or storm
season. The sand that moved offshore under the
influence of the larger storm waves moves onshore
under the influence of the smaller fair-weather waves.
Offshore bars flatten or gradually move onshore, and the
beach between the low- and high-tide lines tends to
become steeper over time. The onshore distance
depends on the size of the storm or severity of the storm
season.
Beach grain size also affects the slope of the beach.
High-permeability gravel for example absorbs much of
the water that falls on it; hence, the return flow energy
is much less than the incoming flow energy. In this
case the forces pushing beach material up onto the beach
are stronger than those moving material seaward in the
backwash. The beach becomes steeper in response.
On beaches of fine sand, the permeability of the material
is much less than for gravel, so relatively little water from
breaking waves is absorbed. The backwash of the wave
has essentially the same volume of water as the upwash.
This backwash volume tends to move sand in a seaward
direction thereby flattening the beach.
6.1.7. Dunes
Beaches can be major sources of material for the land as
well. Sand blown inland from the beach piles up as
sand dunes, sometimes forming large dunes that move
further inland. Some of these large dunes can be big
enough to cover up trees and roads as the dunes move
inland, as occurs on the US Pacific Coast near Florence,
Oregon.
land
sea
flood tidal delta
barrier island barrier island
eb tidal delta
main
channel
E
0
2
0
8
0
9
d
Figure A.40. Barrier island main channel inlets cut through
a barrier island and tidal deltas.
constantly adjust to wave and tidal energy, the quality and
quantity of the sediment supply, and level of the sea. One
of the challenges confronting coastal zone managers is to
predict the future behaviour or movements of beaches.
Beaches can be classified by the size of the material on
them (e.g. mud, sand, gravel, boulders), or by the type of
material (e.g. volcanic, coral, quartz). Beaches can also be
referred to as high energy, moderate energy or low
energy beaches according to some average regional wave
height.
Storms dramatically increase wave energy, which is
proportional to wave height, as well as a rise in sea level.
Beaches adjust their shapes in response to this higher
energy environment. Sand will be moved, offshore bars
may form, or the beach may flatten out. The storm-related
rise in sea level helps waves reach dunes or bluffs, wash-
ing dune sand back to the beach. This addition of sand
and sediment on the beach further buffers the waves and
wind. The amount of sand movement and the ultimate
shape taken by the beach will depend in part on this
sediment supply and the grain size of the beach material.
During a storm the beach flattens as sand moves from
the upper to the lower beach. This dissipates wave energy
over a broadened surface relative to the beach before the
storm. After the storm the sand will gradually return to
nearly its pre-storm profile.
wrm_appA.qxd 8/31/2005 12:08 PM Page 539
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
6.1.8. Longshore Currents
Sand moves because of wind, waves and currents. The
largest volume of sand is often carried by longshore
currents. Such currents are created as waves strike the
shoreline at an oblique angle, forcing some of the surf-
zone water to move laterally along the shore, as illustrated
in Figure A.42. Breaking waves suspend sand, and the
currents move it. Sandy beaches are actually rivers of
sand.
Beach material comes from the land as well as the
seabed. Rivers supply sediment that eventually reaches
the coast and ends up on beaches. Eroding cliffs along the
coast are also sources of beach material. For most of
the US East Coast, the continental shelf is an important
source of sand, which is brought to the shore by the
circular motion of the waves. Calcareous marine organ-
isms are often contained in this continental shelf sand. In
tropical waters the skeletal material of calcareous marine
organisms can make up the entire beach, providing the
sparkling white or pink colour characteristic of, for
example, some Caribbean beaches.
6.2. Coasts under Stress
Many coastal areas are under pressure from a number of
causes, mostly stemming from increasing human popula-
tions. Coastal regions tend to have substantially higher
population densities than elsewhere. Today about half the
540 Water Resources Systems Planning and Management
normal tide
pre storm profile
after storm profile E
0
2
0
8
0
9
e
Figure A.41. Possible response of a beach to a major storm.
Cross sections show steeper pre-storm profile, a possible
shallower after-storm profile (denoted by the dark red line),
and the recovering beach profile, showing some loss of its
original steepness, sometime after the time of the storm.
longshore
current direction
direction of
incoming waves
wave crest
E
0
2
0
8
0
9
f
Figure A.42. The energy from breaking waves at the shoreline
can cause longshore currents. These currents are responsible
for transporting beach sand laterally along the beach.
US population lives near (within easy driving distance of )
the coast. In countries like Australia and Canada a much
higher percentage of the population lives along a coast.
Most of the worlds major cities are located along coasts.
Historically the economy of coastal zones depended on
fishing, boating or boat building. Todays populations are
engaged in a much greater range of occupations that give
them a higher standard of living and quality of life, even
in increasingly expensive and crowded coastal regions.
Continued migrations of people to coastal regions for
jobs, recreation, more moderate weather and a higher
quality of life have increased the stress on coastal zone
resources. Resorts along the Mediterranean coast, South
Florida and the Caribbean islands, for example, have
witnessed large increases in the tourism and recreation
industries. Coastal areas along the western border of
North America have seen substantial increases in
residential development for people desiring the lifestyle
associated with the coast.
Coastal zones are relatively scarce commodities. They
are just narrow strips of land along a coast. The high
demand for this limited real estate invariably results in
relatively high land prices and in the costs of using various
facilities located along the shore. In addition, many coastal
lands are fragile. They erode easily. This puts a high
priority on protecting and maintaining what is there.
6.3. Management Issues
Individuals like to live, work and play along coasts.
This is evidenced by the continued higher proportion
of population growth and economic development along
wrm_appA.qxd 8/31/2005 12:08 PM Page 540
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 541
the coastal regions of just about every country in the
world that has them. Most of the worlds mega-cities
(populations in excess of 10 million) are coastal. As
suggested by Figure A.43, coastal areas provide attractive
environments for residential as well as economic develop-
ment in spite of their occasional natural hazards and
changing landforms. Changing land cover and use that
accompany population growth in coastal zones affect the
geomorphological and ecological processes that take place
along coastal shorelines.
Coastal zone management is the management of
conflicts among multiple conflicting uses of natural and
human resources in highly populated and economically,
ecologically and often geopolitically important coastal
zones. Coastal zone management focuses on the eco-
nomic development and use of the environmental and
ecological resources of coastal regions in ways that
protect, and in some cases restore, them.
Coastal zone management historically focused on
providing protection of shoreline properties against
floods, and the design and construction of harbours and
marinas (that is, the infrastructure associated with com-
mercial and recreational boating). Today, coastal zone
management involves much more than protection of land
from being eroded by the sea and transportation on the
sea. Issues such as water quality, dispersion of pollutants
and the proper management of the complete coastal
ecosystem have become important. Structural design is
now only a small aspect of coastal management.
Current coastal zone management issues include
reducing the risk of erosion from coastal storms,
protecting bays, sloughs (channels) and sounds where
rivers meet the sea, preserving and restoring beaches and
ecological habitats, enhancing economic development
and recreation opportunities along urban waterfronts,
and controlling polluted runoff the major source of
pollution in coastal waters today.
The development of technically, economically and
environmentally appropriate management plans requires
an understanding of the complexity of coastal zone
processes that probably no single person has mastered. It
requires an understanding of the effects of the natural
variability of wind and waves and temperatures as well as
the effects of structural development, over-fishing, exces-
sive pollution, habitat destruction and human population
shifts. Inputs are needed from many individuals who have
expertise in one or more aspects of coastal zone manage-
ment and who are willing to work with others who have
different expertise and experiences.
This necessary expertise and experience may have to
include river basin management as well as coastal zone
management. Some coasts are affected by what happens
in upstream river basins. If those basins drain into coastal
estuaries, it may be important to consider the effects of
basin management on the coastal zones. For example,
much of the sediment along some of North Americas
Pacific Ocean coasts is brought to those coasts by rivers
that discharge their waters and sediment into the Pacific
Ocean. Damming these rivers typically diminishes the
sediment load, that is, the sand being transported to
the coast. Much of the sediment that would otherwise
end up on beaches is trapped behind the dams, even
though these dams can be many hundreds of kilometres
upstream of the beaches.
Figure A.43. Encroaching
residential development along a
coastal beach and estuary.
wrm_appA.qxd 8/31/2005 12:08 PM Page 541
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Consider, for example, the impact the High Aswan
Dam has had on the Nile Delta. The policy of keeping the
downstream flows from reaching flood stage has resulted
in a shrinking of the delta at the mouth of the river as it
discharges into the Mediterranean Sea. Similarly, reduced
sediment loads in the Mississippi River, along with sea
level rise, have contributed to the reduction of the delta of
the Mississippi River on the Gulf Coast of the State of
Louisiana.
6.3.1. Beaches or Buildings
For most of the worlds beaches, if there were no people
there would be no coastal zone management problems. As
humans place obstacles such as houses, apartment
buildings, highways and seawalls along a beach, the
dynamic processes that maintain natural beaches are
constrained. Shoreline erosion, and hence its retreat
inland, is blocked. The beach jams up against these struc-
tural objects. This causes it to narrow, which in turn leads
to a reduction in the supply of sand to adjacent beaches.
Since there is a worldwide tendency for all but very
rocky coasts to erode, protecting and maintaining them is a
high priority, particularly given the high economic value of
coastal land. The economics of coastal zone protection and
maintenance are complex. But is it seldom economically
feasible to invest in shoreline protection unless there is very
dense development or extensive tourism. Areas such as
Miami Beach in southern Florida in the United States, the
Gold Coast in Australia, Scheveningen in the Netherlands,
Copacabana in Brazil and the Chicago and Toronto
waterfronts on the Great Lakes are prime candidates for
coastal protection. Agricultural areas and areas of single-
family residential properties, on the other hand, are not. Yet
protection decisions are not just based on economics.
Decisions about what to protect are often political as well.
Consider, for example, the coastal management
strategy of the Netherlands. Much of the Netherlands,
being below sea level, depends on coastal protection. One
single storm event in 1953 breached the dykes along the
coast and drowned nearly 2,000 people. The loss of live-
stock and property was extensive. The result was a deci-
sion to provide protection from sea storms as severe as
those expected once in every 10,000 years. This is a storm
intensity having a probability of 0.00001 of occurring in
any single year.
542 Water Resources Systems Planning and Management
Alternative strategies for flood protection in the
Netherlands include withdrawal from areas where further
erosion may occur, selective erosion control, full erosion
control, and expansion in the seaward direction by artifi-
cial beach nourishment where the coastal defences are
considered to be weak. Full erosion control and perhaps
expansion in threatened areas would appear to be almost
a necessity for this densely populated and flood-
vulnerable country. Full control involves placing 610
million m
3
of sand along the coast annually.
6.3.2. Groundwater
Freshwater in maritime coastal regions usually comes
from two sources: rivers and lakes, and groundwater
aquifers. When precipitation is sufficient to keep the
groundwater levels higher than the surrounding sea level,
the higher density saltwater does not mix with the fresh-
water. A floating freshwater lens will lie on top of the
denser saltwater.
If withdrawals of freshwater from groundwater aquifers
along coastal regions lower the freshwater level to below
the seawater level, a flow of saltwater from the sea into
the groundwater aquifer can occur. Since the flow rates of
groundwater are very small, there is little mixing of the salt
and fresh groundwater. The freshwater reservoirs of small
island communities are therefore very susceptible to
saltwater intrusion if population pressures along the coasts
cause over-pumping of the aquifer for freshwater supplies.
Similar damage can be caused by cutting away the dunes
and thus lowering the water table, by increasing the
amount of impervious surfaces that increases runoff and
reduces aquifer recharge, and by dredging of rivers that
can increase saltwater intrusion upstream.
6.3.3. Sea Level Rise
A potentially significant change in sea level may come
from changes in the global climate. Global warming since
the last glaciation has resulted in a sea level rise of 100 to
150 metres through melting of the polar ice caps and
thermal expansion of the water in the ocean. This process
continues at a rate estimated to be about 1 to 1.5 mm/yr.
Any additional warming would increase this rate.
Global climate change models estimate the water level
rise resulting from increases in greenhouse gases and
wrm_appA.qxd 8/31/2005 12:08 PM Page 542
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 543
hence increases in temperature, glacier ice melt and
ocean thermal expansion. Such numerical models have
produced a range of scenarios, and these predictions of
water level increases are by no means precise. There
are many uncertainties in the estimates of production
of greenhouse gases. Probably the most uncertain is the
extent to which gas emissions will be controlled, a
matter that is unpredictable and largely political. The
methods used to translate these uncertain atmospheric
pollution figures first into global warming and then into
water level rise also involve many assumptions that are
uncertain. Nevertheless, the impacts of sea level rise are
fairly clear:
More severe storms (tornadoes and hurricanes) will
occur more often.
While storm surge will decrease a little because of the
larger water depths, it will also increase because of
the more severe storm activity.
Offshore, the waves will be higher, because of more
severe storms.
Tides appear not to be significantly affected.
Tidal prisms will increase, because the surface area of
the bays and estuaries will increase.
Waves breaking on shores and structures will be
higher because larger depths all the way into shore will
reduce bottom friction losses, as well as permitting
large breaking waves to come closer into shore.
The result:
Structures will be subjected to higher stress from the
higher waves; factors of safety will decrease.
Structure runup and overtopping will increase, adding
to the risk of flooding and damage by overtopping. As
an example, the Delta Project in the Netherlands raised
all dykes in response to the 1953 storm surge, yet a
1-m sea level rise would reduce the present margin of
safety by about 90% (Wind, 1987).
Sandy shorelines will retreat.
Barrier islands will roll back more rapidly and marshes
behind the barrier islands will disappear.
Deltas will not build out at the same rate; they may
even retreat.
Bluffs and cliffs will retreat more rapidly.
Sediment transport rates will increase, possibly
filling currently stable inlets and harbour entrance
channels.
Saltwater intrusion into groundwater tables will
increase.
Many wetland areas will be inundated and disappear.
How can everyone prepare? Flood protection, shore pro-
tection and navigation structures can be strengthened and
raised to cope with the rise in water level. If all else fails,
a properly executed retreat can be planned in which
buildings are moved back from the shore or abandoned.
The magnitude of the problem can be huge.
Population densities along the ocean shore are already
high and rapidly increasing. Sustainable development of
the coastal areas in the face of rising sea levels will be an
important issue and challenge in this century. Moreover,
although flood defences such as dykes can be raised with
presently available technology, the risk to people and
properties behind those dykes increases.
The main casualties will be the already limited wetland
areas. Their development can keep up with slowly rising
sea levels and move inland, but they may have problems
adjusting to a more rapid rise in water levels. Also, since
most of the properties behind the wetlands are usually
dedicated to human activities and uses, it is unlikely that
wetlands will be allowed to intrude into this valuable
real estate. Damage will also occur to agricultural areas
because of the additional saltwater intrusion.
Sustainable shoreline development through maintain-
ing the existing shorelines by retro-fitting, retreat by
moving infrastructure inland, and meeting the concerns
for wetlands and agriculture, will require a complete
restructuring of most current political and policy-making
processes. They typically are not designed to deal with
slowly developing impacts over very large areas.
6.3.4. Subsidence
Although subsidence can occur naturally, it is often
caused by human activities, for example, pumping
groundwater, petroleum and natural gas out of the
ground. Subsidence of coastal landmass results in
increased flooding. It exacerbates the effects of rising sea
levels, since the relative sea level rise with respect to the
land will be even greater. One of the best known exam-
ples of subsidence coupled with a rising sea level can be
seen in the city of Venice, where increased pumping of
both water and natural gas has caused an accelerated rate
of subsidence. The result is more frequent flooding of an
wrm_appA.qxd 8/31/2005 12:08 PM Page 543
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
increasingly large portion of the city. As a result, the city
and its mediaeval monuments are subjected more and
more regularly to aqua alta or high water.
6.3.5. Wastewater
Coastal waters have traditionally been used for waste-
water disposal. Increased discharges of sewage and chem-
ical effluents, particularly since the Second World War,
have polluted many coastal waters. Although large oceans
and lakes are often thought to have almost unlimited
dilution capacity for wastewater, they do not. In many
cases those dilution limits have already been reached, and
many are showing serious overload and eutrophication
problems. Even though some offending chemical dump
sites have been cleaned up, the toxic chemical content of
organisms in the aquatic ecosystem, including the fish for
example, can remain at high levels for considerable times
after cleanups are completed.
Cleanup of any coastal area is a complex undertaking,
but even more so if it requires altering the pollutant
discharge actions of many nations. There is still the
prevailing philosophy of out of sight, out of mind, when
it comes to the discharge of pollutants into maritime
coastal waters.
Many maritime communities still discharge raw
sewage, often into the near-shore zone. Even communi-
ties that have adequate sewage treatment prior to dis-
charge often discharge sewage into coastal waters during
heavy rainstorms due to combined sewer overflows
(CSOs). With heavy rainfall, the combined sewage and
stormdrain flows can exceed the hydraulic and treatment
capacities of the sewage treatment plant, and the excess
water (overflow) containing some raw sewage is dis-
charged directly without treatment.
Beach pollution is less likely if the sewage outfall is
piped and diffused into the coastal waters many kilo-
metres offshore, as occurs at Sydney, Australia. Many
outfalls are mere open channels that discharge storm and
wastewaters relatively close to shore. The discharged
pollutants, trapped by currents and wave action, can flow
right along the coastline, leading to unacceptably high
pollutant concentrations near the shore.
People have discharged raw sewage into the sea
because it has been a relatively cheap way to dispose of it.
If beach closures and the accompanying economic losses
544 Water Resources Systems Planning and Management
happen often enough, the added costs of wastewater
treatment and disposal methods might be a less costly
alternative. Tertiary wastewater treatment costs about ten
times as much as dumping raw sewage. Incineration, if
acceptable, costs even more.
Two recent examples of actions taken to clean up
coastal zones, at considerable cost, have been those for
reducing the discharge of pollutants into the estuary adja-
cent to the city of Boston on the northeastern Atlantic
Coast of the United States (Figure A.44), and into the
Tagus Estuary adjacent to the city of Lisbon in Portugal.
6.3.6. Other Pollutants
Runoff from farming can also be harmful to coastal
waters, particularly bays and lagoons. Runoff that
contains fertilizers used to promote the growth of agricul-
tural products also promotes the growth of algae and
aquatic weeds. Pesticides contain high levels of toxic
substances, such as heavy metals. Even without the chem-
icals, runoff from farmlands can be undesirable. The
manure from the high-density populations of cattle and
pigs causes high levels of nitrates in both the groundwater
and the surface waters. The runoff of fine sediment mate-
rials from soil erosion resulting from the conversion of
forests to agricultural land has also caused problems for
many maritime organisms. The death of coral reefs in
many tropical countries can be attributed, at least in part,
to the sediment that has entered the water column since
the land was cleared for agriculture, as early as the
eighteenth century in some cases.
Oil spills resulting from transportation and explo-
ration of oil close to shore, such as the spill from the
Exxon Valdez in 1989 (Figure A.45), have been the
cause of well-known and costly disasters. The Prestige oil
spill off the northwest coast of Spain in November 2002
involved twice as much oil as the Exxon Valdez spill.
Areas along major shipping routes are particularly
vulnerable, as witnessed by the spills of the Torrey
Canyon in 1967, the Amoco Cadiz in 1978 and the Erika
in 1999, all along the northwestern coast of France.
Although such major incidents cause some government
action, much larger volumes of oil are routinely released
with virtually no restrictions each year into the oceans
from ships and petroleum production platforms and
refineries.
wrm_appA.qxd 8/31/2005 12:08 PM Page 544
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 545
Hazardous solid waste is another major source of
pollution of coastal waters. Although ocean dumping
legislation severely limits the dumping of solid waste in
many countries, the oceans continue to be the recipient of
all sorts of solid hazardous wastes, including hospital
waste, contaminated dredge spoil and nuclear material.
6.3.7. Mining of Beach Materials
Mining of sand and other beach materials is one of the
major causes of beach erosion in some beaches. Sand
mining for construction and other uses occurs on many
scales, but the accumulated impact can be substantial
along some coasts.
6.4. Management Measures
Coastal zone management is the management of the land
and its uses along the coast. Thus it is the management of
conflicts. It requires legislation and enforcement of meas-
ures to protect and enhance the sustainable use of lands
along shorelines. Coastal zone managers need scientific,
technical and political skills, skills based on an under-
standing of geology and morphology, biology, ecology,
law, planning and engineering among other disciplines.
Although coastal zone management is interdisciplinary, it
is usually the engineers who are asked to make and imple-
ment crucial technical decisions. To do this well, they
need to be properly informed, and work with specialists
from many other disciplines.
Management options include structural as well as
non-structural measures such as zoning, regulation
Figure A.44. Deer Island Sewage
Treatment Plant in Boston Harbour
treats and discharges an average
daily flow of about 380 million
gallons (1.43 million m
3
) of
wastewater. (With permission of
Piping Resources Partnership of
New England.)
E
0
3
0
4
1
7
a
site of the grounding
of the Exxon Valdez
Prince
William Sound
approximate range of
spilled oil after 5 days
Figure A.45. Site of Exxon Valdez oil spill in the US State of
Alaska, 1989.
wrm_appA.qxd 8/31/2005 12:08 PM Page 545
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
enforcement, public awareness activities, negotiation and
consultation. These tools need to be part of an appropri-
ate adaptive decision-making process.
6.4.1. Conforming Use
Coastal zones have traditionally supported many activi-
ties, all of which compete for limited space and may or
may not conflict with each other. Some of the more
important uses of coastal lands include tourism, residen-
tial, recreational, industrial and commercial activities,
agriculture, transportation, waste disposal, aquaculture,
fishing, nature reserves, and military and security
functions. (Not all of these may be present along any
given coast.)
Given the high demand for limited space along a coast,
one approach to coastal zone resource management is to
limit development to only those conforming uses that
need to be along the coast. Examples could include
swimming beaches, fishing ports and marinas.
Amusement parks, casinos, theatres and car parking areas
that are often found along a coast can be located more
inland from the beach. Development (building) permits
for beach property could be limited to those projects that
need to be there.
Implementing a restrictive zoning policy is not simple.
For instance, developments that attract people to an area,
such as harbours or marinas, are normally required by
law to supply sufficient infrastructure to support their
operation. Does a car park suddenly become a conform-
ing use, if it is needed in support of a conforming use?
Similarly, money generators such as casinos are usually
permitted to locate along the coast regardless of conform-
ing use. The installation of adequate public transportation
from inland casinos, hotels and parking lots to coastal
shorelines could alleviate some of this political as well as
economic pressure for non-conforming uses right along
shorelines.
In many coastal regions, abandoned older facilities are
being taken over by interests catering to tourism.
Facilities such as factories, ports and railway lines are
being converted to more conforming uses. Warehouses
and loading terminals are being transformed into
apartment buildings, abandoned railway lines are serving
as biking and hiking trails, and small commercial
ports are becoming marinas and condominiums. These
546 Water Resources Systems Planning and Management
transformations are taking advantage of unique oppor-
tunities, but they also present unique challenges.
In environmentally sensitive regions, tourism can be a
positive influence. Proper coastal management is, in part,
aimed at enhancing the intrinsic value of the area for
tourism. The definition of intrinsic value of the coast is
also changing. Highways, parking lots, hotels and fast-
food restaurants are typical features of many tourist areas,
but an increasing number of tourists demand other
aspects, especially along coastlines. They seek bicycle
paths, dunes, wetlands, clean water, birds and fish.
Increasing numbers of tourists are more interested in
nature and prefer physical activities such as hiking, bik-
ing, birding, boating and fishing to simply sitting on the
beach soaking up the sun and a few six-packs. The coastal
environment they want is much more natural than the
traditional amusement park type of environment
dominated by concrete and French fries. (Obviously,
many tourists like both the concrete and the sun and
sand.)
Whatever tourists seek, tourism development is clearly
motivated by economics. Enhanced recreational (as well
as environmental and ecosystem) planning must be done
within an economic framework that sustains the contin-
ued coastal zone development, maintenance, protection
and restoration activities.
6.4.2. Structures
Many types of structures are built along coastlines to meet
the needs of people who visit or live there. The focus here
is only on those that are built to protect property from
damage by waves and currents along the coasts.
There are three available responses to local shoreline
erosion problems:
hard stabilization, such as seawalls (Figure A.46),
offshore breakwaters (Figure A.47) and groynes
(Figure A.48)
soft stabilization, such as beach sand replenishment
relocation of threatened structures.
Each response has its advantages and disadvantages. Hard
stabilization alternatives involve armouring the shoreline
with structures designed to hold it in place. This may be
the best way to save buildings, but not beaches. Removing
buildings or relocating them further inland is the best way
wrm_appA.qxd 8/31/2005 12:08 PM Page 546
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 547
to save beaches, but obviously not the buildings in their
current locations. Hard stabilization combined with
beach replenishment may provide necessary protection
for both buildings and beaches, but it will be necessary to
continually add sand to the beach over time.
6.4.3. Artificial Beach Nourishment
Artificial nourishment is an alternative to structural meas-
ures for protection. It has the least impact on adjacent
properties and the environment. Instead of harming the
surroundings, a beach fill can benefit adjacent eroding
properties. Apart from its cost, the diffusion and advec-
tion of a beach fill only presents a problem when water
depth needs to be maintained at the adjacent properties,
such as in navigation channels, or when the sand added
to the system threatens valuable habitat.
Although this approach has been used for many years,
the technology is still very much intuitive. Artificial beach
nourishment is subject to the same erosion that caused
the need for replenishment in the first place. The design
of such programmes must therefore be concerned not
only with how much sand to place, but also with how
often it needs to be replenished. Artificial nourishment
in most areas then becomes a beach maintenance
programme based on annual costbenefit estimates. If the
site erodes more rapidly as a result of offshore conditions,
such as a locally steeper shoreline or a convergence of
wave energy, the artificially placed fill will also be
subjected to the same conditions and will not perform
well.
Since a major objective of most artificial nourishment
schemes is to provide protection as well as additional
recreational beach space, most schemes consider beach
fills in combination with shore face nourishment. Beach
fill normally requires rehandling of the sand so that it can
be placed by pipeline dredge and perhaps be reshaped by
earthmoving equipment. The onshore sand is usually
placed with a steep seaward slope. The wave action on
such a fill will shape the most seaward part of the fill mass
into a beach profile. During this adjustment period and
at any later time, when other beach material further land-
ward is redistributed fine grain sizes will be winnowed
out of the mass of sand and lost to deep water, until the
grain size distribution of the remaining sand mass is
similar to the native distribution. Once the fill has been
re-adjusted by the waves to form a beach profile, a steep
scarp may have formed at the top of the beach.
The combination of artificial nourishment with struc-
tures such as groynes or offshore breakwaters will help
contain the fill material. Structures also provide an oppor-
tunity to use beach fills in areas that would never be stable
with artificial nourishment alone. Examples are Hilton
Head on the US Atlantic Coast (Bodge et al., 1993) and
Norderney on Europes North Sea Coast (Kunz, 1993).
Biologically, a beach is relatively unproductive. Benthic
communities that are covered by a beach fill seem to
re-establish relatively quickly after nourishment. The
surrounding ecosystem, however, can be affected and thus
scarped dune is evidence of eroding
shoreline
narrowing of beach
steepening
of offshore slope
no beach
offshore
slope has steepened
as slope steepens,
wave size increases,
a higher wall is needed
E
0
2
0
8
0
9
k
Figure A.46. Seawalls designed to protect property. Larger
and stronger seawalls are needed over time as the beach
disappears in front of them.
wrm_appA.qxd 8/31/2005 12:08 PM Page 547
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
548 Water Resources Systems Planning and Management
should be carefully considered and monitored (CUR,
1997; NRC, 1995; Simm, 1996).
7. Conclusion
This concludes an overview of the many components,
functions and processes that take place in watersheds and
river basins, wetlands, estuaries and along coasts, and the
ways in which these processes can be affected by land and
water management policies and practices. This appendix
provides an introductory background for those who
have not had separate courses in these subjects and
who are now being introduced, through this book, to the
Figure A.47. Eastern Scheldt Storm
Surge Barrier along the North Sea
coast of the Netherlands, a seawall
eight kilometres long. Its sixty-two
openings are closed off with sliding
barriers at high tide.
updrift
accretion
wave direction
local
accretion
downdrift
erosion
original
shoreline
groyne
E
0
2
0
8
0
9
g
local
erosion
Figure A.48. Groynes are often used to reduce shoreline
erosion. They change the longshore sediment transport rates
and result in accretion updrift of the groynes within the groyne
field and downdrift erosion.
wrm_appA.qxd 8/31/2005 12:08 PM Page 548
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 549
art of building and solving water resources planning and
management models. The purpose of these models is
to predict how, and to what extent, these system
components, functions and processes are affected by
alternative ways of managing and using these land and
water resources.
8. References
BODGE, K.R.; OLSON, E.J. and CREED, C.G. 1993.
Performance of beach nourishment at Hilton Head
Island, S.C.: beach nourishment engineering and
management considerations. In: Proceedings of the Eighth
Symposium on Coastal and Ocean Management (Coastal
Zone 93), New Orleans, La., 1923 July 1993. New York,
ASCE, pp. 66876.
BOWDEN, K.F. 1967. Circulation and diffusion. In:
G.H. Lauff (ed.), Estuaries. Washington, D.C., American
Association for the Advancement of Science (Publ. No. 83).
pp. 1536.
BRICE, J.C. and BLODGETT, J.C. 1978. Countermeasures
for hydraulic problems at bridges. Volume 1: Analysis and
assessment. Volume 2: Case histories for sites 1283. Federal
Highway Administration report (FHWA-RD-78-162 and
FHWA-RD-78-163).
CORMACK, D. and NICHOLS. 1977. The concentrations
of oil in seawater resulting from natural and chemically
induced dispersion of oil slicks. In: J.A. Nichols (ed.),
Proceedings of the Oil Spill Conference (Prevention, Behaviour
Control, Clean-up), New Orleans, March 1977. Washington,
D.C., American Petroleum Institution (Publication 4284).
pp. 3815.
COVICH, A.P. 1993. Water and ecosystems. In: P.H.
Gleick (ed.), Water in crisis: a guide to the worlds freshwa-
ter resources. Oxford, UK, Oxford University Press.
COWARDIN, L.W.; CARTER, V.; GOLET, F.C. and
LAROE, E.T. 1979. Classification of wetlands and deep
water habitats of the United States. Washington, D.C., US
Fish and Wildlife Service.
CULBERTSON, D.M.; YOUNG, L.E. and BRICE, J.C.
1967. Scour and fill in alluvial channels. Washington, D.C.,
USGS (Open-File Report).
CUR. 1997. Beach nourishment and shore parallel structures.
Gouda, Netherlands, CUR, (Rep 97-2).
DEAN, R.G. and DALRYMPLE, R.A. 1984. Water wave
mechanics for engineers and scientists. Englewood Cliffs,
N.J., Prentice Hall.
DINGEMANS, M.W. 1997. Water wave propagation over
uneven bottoms. Advanced series on ocean engineering.
Singapore, World Scientific Publishing. Vol. 13.
FORMAN, R.T.T. and GODRON, M. 1986. Landscape
ecology. New York, Wiley.
HARDY, J.T.; CRECELIUS, E.A.; ANTRIM, L.D.; KIESSER,
S.L. and BROADHURST, V.L. 1990. Aquatic surface
microlayer contamination in Chesapeake Bay. Marine
Chemistry, No. 28, pp. 33351.
HORIKAWA, K. 1978. Coastal engineering. Tokyo,
University of Tokyo Press.
HORTON, R.E. 1945. Erosional development of streams
and their drainage basins: hydrophysical approach to
quantitative morphology. Geological Society of America
Bulletin, No. 56, pp. 275370.
HUTCHINGS, P.A. and COLLETT, L.C. 1977. Guidelines
for the protection and management of estuaries and estu-
arine wetlands. Toowong, Queensland, Australian Marine
Sciences Association.
IPPEN, A. 1966. Estuary and coastline hydrodynamics. New
York, McGraw-Hill.
KINSMAN, B. 1965. Wind waves. Englewood Cliffs, N.J.,
Prentice Hall.
KUNZ, H. 1993. Sand losses from artificially nourished
beach stabilized by groynes, beach nourishment engi-
neering and management considerations. In: Proceedings
of the Eighth Symposium on Coastal and Ocean Management
(Coastal Zone 93), New Orleans, La., 1923 July 1993. New
York, ASCE, pp. 66876.
LAGASSE, P.F.; SCHALL, J.D.; JOHNSON, F.;
RICHARDSON, E.V.; RICHARDSON, J.R. and
CHANGE, F. 1991. Stream stability at highway structures.
McLean, Va., HEC-20 (publ. No. FHWA-IP-90-014).
LANE, E.W. 1955. The importance of fluvial morphology
in hydraulic engineering. Proceedings of the American Society
of Civil Engineers, No. 81, Paper 745, pp. 117.
wrm_appA.qxd 8/31/2005 12:08 PM Page 549
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
LEOPOLD, L.B. 1994. A view of the river. Cambridge,
Mass., Harvard University Press.
LEOPOLD, L.G. and MADDOCK, T. Jr. 1953 The hydraulic
geometry of stream channels and some physiographic implica-
tions. Washington, D.C., USGS (Professional Paper 252).
LEOPOLD, L.B. and WOLMAN, M.G. 1957. River chan-
nel patterns: braided, meandering, and straight. In:
Physiographic and hydraulic studies of rivers. Washington,
D.C., US Government Printing Office (Professional paper
US Geological Survey, Vol. 282B), pp. 3985.
LEOPOLD, L.B.; WOLMAN, M.G. and MILLER, J.P.
1964. Fluvial processes in geomorphology. San Francisco,
Calif., W.H. Freeman.
LEOPOLD, L.B.; WOLMAN, M.G. and MILLER, J.P.
1992. Fluvial processes in geomorphology. Mineola, N.Y.,
Dover Publications.
LIU, P.L.F. and LOSADA, I.J. 2002. Wave propagation
modeling in coastal engineering. Journal of Hydraulic
Research, Vol. 40, No. 3, pp. 22940.
MADSEN, K.N.; NILSSON, P. and SUNDBACK, K. 1993.
The influence of benthic microalgae on the stability of a
subtidal sediment. Journal of Experimental Marine Biology
and Ecology, No. 170, pp. 15977.
MASER, C. and SEDELL, J.R. 1994. From the forest to the
Sea: the ecology of wood in streams, rivers, estuaries and
oceans, St. Lucie Press, Delray Beach, Florida, p 200.
MCDOWELL, D.M. and OCONNOR, B.A. 1977. Hydraulic
behaviour of estuaries. London, MacMillan.
MEADOWS, P.S.; TAIT, J. and HUSSAIN, S.A. 1990.
Effects of estuarine infauna on sediment stability and par-
ticle sedimentation. Hydrobiologia, No. 190, pp. 2636.
MEI, C.C. and LIU, P.L.F. 1993. Surface waves and
coastal dynamics. Annual Review of Fluid Mechanics,
No. 25, pp. 21540.
MINSHALL, G.W.; CUMMINS, K.W.; PETERSEN, R.C.;
CUSHING, C.E.; BRUNS, D.A.; SEDELL, J.R. and VAN-
NOTE, R.L. 1985. Developments in stream ecosystem
theory. Canadian Journal of Fisheries and Aquatic Sciences,
No. 42, pp. 104555.
NRC (National Research Council). 1995. Beach nourishment
and protection. Washington, D.C., National Academy Press.
550 Water Resources Systems Planning and Management
POFF, N.L.; ALLAN, J.D.; BAIN, M.B.; KARR, J.R.;
PRESTEGAARD, K.L.; RICHTER, B.D.; SPARKS, R.E. and
STROMBERG, J.C. 1997. The natural flow regime: a par-
adigm for river conservation and restoration. BioScience,
No. 47, 1997, pp. 76984.
PRITCHARD, D.W. 1952. Estuarine hydrography.
Advances in Geophysics, No. 1, pp. 24380.
RICHARDSON, E.V.; SIMONS, D.B. and JULIEN, P.Y.
1990. Highways in the river environment. Washington,
D.C., FHWA, (revision of 1975 publication by E.V.
Richardson, D.B. Simons, S. Karki, K. Mahmood and
M.A. Stevens; Pub. No. FHWA-HI-90-016).
RHOADS, D.C. 1974. Organismsediment relations on
the muddy sea floor. Oceanography Marine Biology Annual
Review, No. 12, pp. 263300.
RHOADS, D.C. and BOYER, L.F. 1982. The effects of
marine benthos on physical properties of sediments. In:
P.L. McCall and M.J.S. Tevesz, (eds.), Animalsediment
relations: the biogenic alteration of sediments. New York,
Plenum, pp. 352.
ROSGEN, D. 1996. Applied river morphology. Pagosa
Springs, Colo., Wildland Hydrology.
SARPKAYA, T. and ISAACSON, M. 1081. Mechanics of
wave forces on offshore structures. New York, Van
Nostrand-Reinhold.
SCHUMM, S.A. (ed.). 1972. River morphology: benchmark
papers in geoplogy. Stroudsburg, Pa., Dowden, Hutchinson
and Ross.
SCHUMM, S.A. 1977. The fluvial system. New York, Wiley.
SIMM, J. 1996 Beach management manual. London,
Construction Industry Research and Information
Association (CIRIA) (Rep. 154).
SPARKS, R. 1995. Need for ecosystem management of
large rivers and their flooodplains. BioScience, Vol. 45,
No. 3, p. 170.
SZEKIELDA, K.H.; KUPFERMAN, S.I.; KLEMAS, T. and
POLIS, D.F. 1972. Element enrichment in organic films
and foams associated with aquatic frontal systems. Journal
of Geophysical Research, Vol. 77, pp. 527884.
TRIMBLE, S.W. and CROSSON, P. 2000. U.S. soil erosion
rates: myth and reality. Science, Vol. 289, pp. 24850.
wrm_appA.qxd 8/31/2005 12:08 PM Page 550
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 551
USDA (United States Department of Agriculture). 1998.
Stream corridor restoration: principles, processes and practices.
Washington, D.C. Interagency Stream Restoration Working
Group; (See also Federal Interagency Stream Restoration
Working Group, (FISRWG), Stream corridor restoration:
principles, processes, and practices, GPO Item No. 0120-A;
SuDocs No. A 57.6/2:EN3/PT.653. ISBN-0-934213-59-3
2001.)
USEPA (United States Environmental United States
Environmental Protection Agency) 2001. www.usepa.gov
(accessed 12 November 2004).
USACE (US Army Corps of Engineers). 1969. Missouri
river channel regime studies. Omaha, Nebr., USACE (MRD
Sedimentation Series No. 13.B).
WARD, A.D. and ELLIOT, W.J. 1995. Environmental
hydrology. Boca Raton, Fla., CRC Lewis 1995.
WIND, H.G. 1987. Impact of sea level rise on society.
Rotterdam, Balkema.
Additional References (Further Reading)
ABBOTT, M.B. and Price, W.A. 1994. Coastal, estuarial
and harbour engineers reference book. London and New
York, E.&F.N. Spon.
ABERNETHY, B. and RUTHERFORD, I.D. 1998. Where
along a rivers length will vegetation most effectively sta-
bilize stream banks? Geomorphology, Vol. 23, pp. 5575.
ABT, S.R.; CLARY, W.P. and THORNTON, C.I. 1994.
Sediment deposition and entrapment in vegetated
streambeds. Journal of Irrigation and Drainage Engineering,
Vol. 120, No. 6, pp. 1098111.
ACKERS, P. and CHARLTON. 1970. Meandering geome-
try arising from varying flows. Journal of Hydrology, Vol.
11, No. 3, pp. 23052.
ACKERS, P. and WHITE, W.R. 1973. Sediment transport:
new approach and analysis. Journal of the Hydraulics
Division, ASCE 99, No. HY11 (American Society of Civil
Engineers Proc. Paper 10167), pp. 204160.
ADAMS, W.J.; KIMERLE, R.A. and BARNETT, J.W. Jr.
1992. Sediment quality and aquatic life assessment.
Environmental Science and Technology, Vol. 26, No. 10,
pp. 186475.
ALDRIDGE, B.N. and GARRETT, J.M. 1973. Roughness
coefficients for stream channels in Arizona. Washington,
D.C., US Government Printing Office (US Geological
Survey, open-file report).
ALLAN, J.D. 1995. Stream ecology: structure and function of
running waters. New York, Chapman and Hall.
BAYLEY, P.B. 1995. Understanding large river-floodplain
ecosystems. BioScience, Vol. 45, No. 3, p. 154.
BESCHTA, R.L.; PLATTS, W.S.; KAUFFMAN, J.B. and
HILL, M.T. 1995. Artificial stream restoration: money well
spent or an expensive failure? In: Proceedings on environ-
mental restoration:, UNCOWR 1994 annual meeting.
Southern Illinois University, Carbondale, Ill. pp. 76104.
BIEDENHARN, D.S.; ELLIOTT, C.M. and WATSON, C.C.
1997. The WES stream investigation and streambank
stabilization handbook. Vicksburg, Miss., prepared for US
Environmental Protection Agency by US Army Corps of
Engineers, Waterways Experiment Station.
BIRD, E.C.F. 1984. Coasts, third edn. Oxford, UK, Basil
Blackwell.
BIRD, E.C.F. 1985. Coastline changes: a global review. New
York, Wiley.
BIRD, E.C.F. 1993. Submerging coasts: the effects of a rising
sea level on coastal environments. New York, Wiley.
BISHOP, P.L. 1983. Marine pollution and its control. New
York, McGraw-Hill.
BLENCH, T. 1957. Regime behaviour of canals and rivers.
London, Butterworths Scientific.
BOOTH, D. and JACKSON, C. 1997. Urbanization of
aquatic systems: degradation thresholds, stormwater detec-
tion and the limits of mitigation. Journal of the American
Water Resources Association, Vol. 33, No. 5, pp. 107789.
BREUSERS, H.N.C. and RAUDKIVI, A.J. 1991. Scouring.
Rotterdam, Balkema.
BRICE, J.C. 1984. Planform properties of meandering rivers.
In: C.M. Elliott (ed.), River meandering, pp. 115. Proceedings
of the 1983 Conference on Rivers. New Orleans, ASCE.
BROWN, A.G. 1997. Biogeomorphology and diversity in
multiple-channel river systems. Global Ecology and
Biogeography Letters, Vol. 6, Floodplain forests special issue,
pp. 17985.
wrm_appA.qxd 8/31/2005 12:08 PM Page 551
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
BROWN, A.G.; HARPER, D. and PETERKEN, G.F. 1997.
European floodplain forests: structure, functioning and
management. Global Ecology and Biogeography Letters,
Vol. 6, Floodplain forests special issue, pp. 16978.
BROWN, C.B. 1950. Sediment transportation. In: H. Rouse
(ed.), Engineering hydraulics, pp. 769858. New York,
Wiley.
BROOKES, A. 1990. Restoration and enhancement of
engineered river channels: some European experiences.
Regulated Rivers: Research and Management, Vol. 5, No. 1,
pp. 4556.
BROOKES, A. and SHIELDS, F.D. Jr. (eds.). 1995. River
channel restoration: guiding principles for sustainable proj-
ects. Chichester, UK, Wiley.
CHESAPEAKE BAY PROGRAM OFFICE. 2001.
www.chesapeakebay.net / wqcmodelling.htm (accessed
12 January 2004).
COLBY, B.R. 1964. Discharge of sands and mean-velocity in
sand-bed streams. Washington, D.C., USGS (Professional
Paper 462-A).
CUMMINS, K.W. 1974. The structure and function of
stream ecosystems. BioScience, Vol. 24, pp. 63141.
DALRYMPLE, R.A.; KIRBY, R.T. and HWANG, P.A. 1984.
Wave diffraction due to area of energy dissipation. Journal
of Waterway, Port, Coastal, and Ocean Engineering, ASCE,
Vol. 110 , No. 1, pp. 6779.
DEAN, R.G. 1983. Principles of beach nourishment. In:
P. Komar (ed.), Handbook of coastal processes and erosion.
Boca Raton, Fla., CRC Press, pp. 21732.
DELFT HYDRAULICS; DHI. 1996. Qualitative assessment
of possible morphological impacts of FAP implementation.
Delft, the Netherlands, Water Resources Planning
Organization, Government of Bangladesh (River Survey
Project Special Report No. 5).
DELFTHYDRAULICS; DHI and EGIS. 1996. Spatial repre-
sentation and analysis of hydraulic and morphological data.
Delft, the Netherlands, Water Resources Planning
Organization, Government of Bangladesh (River Survey
Project Special Report No. 17).
DE VRIEND, H.J. 2003. On the prediction of aggregated-
scale coastal evolution. Journal of Coastal Research,
Vol. 19, No. 4, pp. 7579.
552 Water Resources Systems Planning and Management
DOWNS, P.W. 1995. River channel classification for
channel management purposes. In: A. Gurnell and
G. Petts (eds.), Changing river channels. New York, Wiley,
pp. 34765.
DRUERY, B.M.; DYSON, A.R. and GREENTREE, G.S.
1983. Fundamentals of tidal propagation in estuaries.
Proceedings of the Sixth Aust. Conference on Coastal and
Ocean Engineering. Barton, Engineers Australia.
DUNNE, T. and LEOPOLD, L.B. 1978. Water in environ-
mental planning. San Francisco, Calif., W.H. Freeman.
DYER, K.R. 1989. Estuarine flow interaction with topog-
raphy, lateral and longitudinal effects. In: B.J. Neilson,
A. Kuo and J. Brubaker (eds.), Estuarine circulation.
Clifton, N.J. Humana. pp. 3959.
DYER, K.R. 1990. The rich diversity of estuaries.
Estuaries, No. 13, pp. 5045.
GIS and DELFT HYDRAULICS. 1997. Morphological
dynamics of the Brahmaputra-Jamuna River. Delft,
Environmental and GIS support project for water sector
planning (EGIS).
EINSTEIN, H.A. 1950. The bed-load function for sediment
transportation in open-channel flow. Washington, D.C.,
USDA-SCS (Technical Bulletin 1026).
ENGELUND, 1974. F. Flow and bed topography in chan-
nel bends. Journal of the Hydraulics Division, ASCE, Vol.
100, No. HY11, pp. 163148.
EXNER, F.M. 1925. ber die Wechselwirkung zwischen
Wasser und Geschiebe in Flssen, pt.IIa, Bd.134. Wien,
Sitzungsberichte Akademie der Wissenschaften.
FERGUSON, B.K. 1991. Urban stream reclamation. Journal
of Soil and Water Conservation, Vol. 46, No. 5, pp. 3248.
FORMAN, R.T.T. 1995. Land mosaics: the ecology of land-
scapes and regions. Cambridge, UK, Cambridge University
Press.
FRISSELL, C.A.; LISS, W.J.; WARREN, C.E. and HURLEY,
M.D. 1986. A hierarchical framework for stream habitat
classification: viewing streams in a watershed context.
Environmental Management, Vol. 10, No. 2, pp. 199214.
GERMANOSKI, G. and SCHUMM, S.A. 1993. Changes in
braided river morphology resulting from aggradation and
degradation. Journal of Geology, Vol. 101, pp. 45166.
wrm_appA.qxd 8/31/2005 12:08 PM Page 552
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 553
GOODWIN, C.N.; HAWKINS, C.P. and KERSHNER, J.L.
1997. Riparian restoration in the western United States:
overview and perspective. Restoration Ecology, Vol. 5,
No. 4S, pp. 414.
GORDON, N.D.; MCMAHON, T.A. and FINLAYSON,
B.L. 1992. Stream hydrology: an introduction for ecologists.
Chichester, UK, Wiley.
GRAF, W.H. 1971. Hydraulics of sediment transport. New
York, McGraw-Hill, New York.
GRAF, W.L. (ed.). 1987. Geomorphic systems of North
America: Centennial special volume 2. Boulder, Colo.,
Geological Society of America.
GREGORY, K.J. 1992. Vegetation and river channel
process interactions. In: P.J. Boon, P. Calow and G.E. Petts
(eds.), River conservation and management. Chichester, UK,
Wiley, pp. 25569.
GREGORY, K.J. and GURNELL, A.M. 1988. Vegetation
and river channel form and process. In: H.A. Viles (ed.),
Biogeomorphology. Oxford, Basil Blackwell, pp. 1142.
GREGORY, S.V.; SWANSON, F.J.; MCKEE, W.A. and
CUMMINS, K.W. 1991. An ecosystem perspective of
riparian zones: focus on links between land and water.
BioScience, Vol. 41, No. 8, pp. 54051.
GURNELL, A.M. 1997. The hydrological and geomor-
phological significance of forested floodplains. Global
Ecology and Biogeography Letters, Vol. 6, Floodplain forests
special issue, pp. 21929.
GURNELL, A.M. and GREGORY, K.J. 1995. Interactions
between semi-natural vegetation and hydrogeomorpho-
logical processes. Geomorphology, Vol. 13, pp. 4969.
HOFFMANS, G.J.C.M. and VERHEIJ, H.J. 1997. Scour
manual. Rotterdam, Balkema.
HUPP, C.R. and OSTERKAMP, W.R. 1996. Riparian veg-
etation and fluvial geomorphic processes. Geomorphology,
Vol. 14, pp. 27795.
INGRAM, J.J. 1988. Total sediment load measurement using
point source suspended sediment data. Vicksburg, Miss.,
USACE (Hydraulics Laboratory Technical Report HL-88-
28, Waterways Experiment Station).
INTERAGENCY ECOSYSTEM MANAGEMENT TASK
FORCE. 1995. The ecosystem approach: healthy ecosystems
and sustainable economies; Vol. I, Overview. Washington,
D.C., Council on Environmental Quality.
JAGERS, H.R.A. 1996. Behaviour oriented modelling of
braided rivers. Delft, the Netherlands, Delft Hydraulics
and University of Twente (research report), January.
JAGERS, H.R.A. 1997. On prediction models for plan-
form changes of braided rivers. In: Abstracts 6th Int. Conf.
Fluvial Sedimentology, University of Cape Town, Cape Town,
226 September, p. 93.
JAGERS, H.R.A. 2000. Draft of Ph.D. thesis. Netherlands,
University of Twente.
JANSEN, P.P., VAN BENDEGOM, L.; VAN DEN BERG, J.;
DE VRIES, M. and ZANEN, A. 1979. Principles of river
engineering: the non-tidal alluvial river. London, Pitman.
KALKWIJK, J.P.T. and DE VRIEND, H.J. 1980.
Computation of the flow in shallow river bends. Journal of
Hydraulic Resesearch, IAHR, Vol. 18, No. 4, pp. 32742.
KAMPHUIS, J.W. 2000. Introduction to coastal engineering
and management. Singapore, World Scientific Publishing.
KARAMBAS, T.V. and KOUTITAS, C. 1992. A breaking
wave propagation model based on the Boussinesq equa-
tions. Coastal Engineering, No. 18, pp. 119.
KAWAMURA, T. 1998. Numerical simulation of 3D
turbulent free-surface flows. Hrsholm, Denmark,
International Research Center for Computational
Hydrodynamics (ICCH).
KELLERHALS, R.; CHURCH, M. and BRAY, D.I. 1976.
Classification and analysis of river processes. Journal of the
Hydraulics Division, ASCE, Vol. 102, No. HY7, pp. 81329.
KENNEDY, A.B.; CHEN, Q.; KIRBY, J.T. and DALRYM-
PLE, R.A. 2000. Boussinesq modelling of wave transfor-
mation, breaking and runup, I: 1D. Journal of Waterway,
Port, Coastal, and Ocean Engineering, ASCE, Vol. 126,
No. 1, pp. 3948.
KENTULA, M.E.; BROOKS, R.E.; GWIN, S.E.; HOLLAND,
C.C.; SHERMAN, A.D. and SINFEOS, J.C. 1992. An approach
to improving decision-making in wetland restoration and
creation. Washington, D.C., Island Press.
KERCHNER, J.L. 1997. Setting riparian/ aquatic restora-
tion objectives within a watershed context. Restoration
Ecology, Vol. 5, No. 45, pp. 45971.
wrm_appA.qxd 8/31/2005 12:08 PM Page 553
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
KERN, K. 1992. Restoration of lowland rivers: the
German experience. In: P.A. Carling and G.E. Petts (eds.),
Lowland floodplain rivers: geomorphological perspectives.
Chichester, UK, Wiley, pp. 27997.
KIRKBY, M. 1990. The landscape viewed through
models. Zeitschrift fur Geomorphologie, Suppl.-Bd., No. 79,
pp. 6381.
KIRKBY, M.J. (ed.). 1994. Process models and theoretical
geomorphology. New York, Wiley.
KLAASSEN, G.J. and MASSELINK, G. 1992. Planform
changes of a braided river with fine sand as bed and
bank material. In: Proceedings of the Fifth International
Symposium on River Sedimentation, pp. 45971, Karlsruhe.
KLAASSEN, G.J.; MOSSELMAN, E. and BRHL, H.
1993. On the prediction of planform changes in braided
sand-bed rivers. In: S.S.Y. Wang (ed.), Advances in
Hydro-Science and Engineering. Oxford, Mass, University
of Mississippi, pp. 13446.
KLIJN, F. 1997. A hierarchical approach to ecosystems and
its implications for ecological land classification: with
examples of ecoregions, ecodistricts and ecoseries of the
Netherlands. Ph.D. thesis, Leiden University.
KLIMAS, C.V. 1987. River regulation effects on floodplain
hydrology and ecology. In: D.D. Hook (ed.), The ecology
and management of wetlands, Chapter 4. London, Croom
Helm.
KNIGHTON, A.D. and NANSON, G.C. 1993. Anastomosis
and the continuum of channel pattern. Earth Surface
Processes and Landforms, BGRG, Vol. 18, No. 7, pp. 61325.
KNOPF, F.L.; JOHNSON, R.R.; RICH, T.; SAMSON, F.B.
and SZARO, R.C. 1988. Conservation of riparian systems in
the United States. Wilson Bulletin, No. 100, pp. 27284.
LARGE, A.R.G. and PETTS, G.E. 1994. Rehabilitation of
river margins. In: P. Calow and G. Petts (eds.), The rivers
handbook, pp. 40118. Oxford, UK, Blackwell Scientific.
LEWIS, R.E. 1997. Dispersion in estuaries and coastal
waters. Chichester, UK, Wiley.
LEWIS, R.R. III. 1989. Wetland restoration/creation/enhance-
ment terminology: suggestions for standardization. Washington,
D.C., US Environmental Protection Agency (Wetland
Creation and Restoration: The Status of the Science, Vol. II.
EPA 600/3/89/038B).
554 Water Resources Systems Planning and Management
LI, R.M. and SHEN, H.W. 1973. Effect of tall vegetations
on flow and sediment. Journal of the Hydraulics Division,
Proceedings of the American Society of Civil Engineers, Vol. 99,
No. HY5, pp. 793814.
LIN, P. and LIU, P.L.F. 1998. A numerical study of break-
ing waves in the surf zone. Journal of Fluid Mechanics,
No. 359, pp. 23964.
LORENZ, E. 1963. Deterministic nonperiodic flow.
Journal of Atmospheric Sciences, Vol. 20, pp. 13041.
MALANSON, G.P. 1993. Riparian landscapes. Cambridge,
UK, Cambridge University Press.
MANLEY, P.A.; HAZELHURST, S.; MAGARY, F. and
HAWK, K.S. 1995. Sustaining ecosystems: a conceptual
framework. San Francisco, Calif., USDA Forest Service,
Pacific Southwest Region.
MARSTON, R.A.; GIREL, J.; PAUTOU, G.; PIGAY, H.;
BRAVARD, J.P. and ARNESON, C. 1995. Channel meta-
morphosis, floodplain disturbance, and vegetation devel-
opment in a river in France. Geomorphology, Vol. 13,
pp. 12131.
MARTIN, J.L. and MCCUTCHEON, S.C. 1999.
Hydrodynamics and transport for water quality modelling.
Boca Raton, Fla., Lewis.
MOSSELMAN, E. 1989. Theoretical investigation on
discharge-induced river-bank erosion. Communications on
Hydraulic and Geotechnical Engineering. Delft, the
Netherlands, Delft University of Technology. No. 89.3.
MOSSELMAN, E. 1992. Mathematical modelling of
morphological processes in rivers with erodible cohesive
banks. Communications on Hydr. and Geotech. Engrg. Delft
University of Technology. No. 92.3.
MOSSELMAN, E. 1995. A review of mathematical mod-
els of river planform changes. Earth Surface Processes and
Landforms, BGRG, Vol. 20, No. 7, pp. 66170.
MOSSELMAN, E.; HUISINK, M.; KOOMEN, E. and
SEYMONSBERGEN, A.C. 1995. Morphological changes
in a large braided sand-bed river. In: E.J. Hickin (ed.),
River geomorphology. Chichester, UK, Wiley, pp. 23547.
MOSSELMAN, E. and WENSINK, G.J. 1993. River bathym-
etry observation with radar remote sensing. Proceedings of
the International Workshop on Morphological Behaviour
of the Major Rivers in Bangladesh, Dhaka,.November.
wrm_appA.qxd 8/31/2005 12:08 PM Page 554
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 555
Dhaka, Government of Bangladesh and Commission of
the European Communities, pp. 24350.
MURRAY, A.B. and PAOLA, C. 1994. A cellular model of
braided rivers. Nature, No. 371, pp. 547.
MURRAY, A.B. and PAOLA, C. 1997. Properties of a cel-
lular braided-stream model. Earth Surface Processes and
Landforms, BGRG, Vol. 22, pp. 100125.
NAIMAN, R.J.; LONZARICH, D.G. BEECHE, T.J. and
RALPH, S.C. 1992. General principles of classification and
the assessment of conservation potential in rivers. In: P.J.
Boon, P. Calow and G.E. Petts (eds.). River conservation
and management. Chichester, UK, Wiley, pp. 93123.
NAKAGAWA, H.; TSUJIMOTO, T. and SHIMIZU, Y.
1992. Sediment transport in vegetated bed channel. In:
Larsen and N. Eisenhauer (eds.), Proceedings of the Fifth
International Symposium on River Sedimentation, Karlsruhe,
610 April.
NANSON, G.C. and CROKE, J.C. 1992. A genetic classifi-
cation of floodplains. Geomorphology, Vol. 4. pp. 45986.
NANSON, G.C. and KNIGHTON, A.D. 1996. Anabranching
rivers: their cause, character and classification. Earth Surface
Processes and Landforms, Vol. 21, pp. 21739.
NATIONAL ACADEMY OF SCIENCES. 1995. Wetlands:
characteristics and boundaries. Washington, D.C., National
Academy Press.
NEILL, C.R. and GALAY, V.J. 1967. Systematic evaluation
of river regime. Journal of Waterways and Harbors Division,
ASCE, Vol. 93, No. WW1, pp. 2553.
NELSON, J.M. and SMITH, J.D. 1989. Evolution and sta-
bility of erodible channel beds. In: S. Ikeda and G. Parker
(eds.), River meandering. Washington, D.C., AGU (Water
Resources Monograph 12). pp. 32177.
NEWSON, M.D. and NEWSON, C.L. 2000
Geomorphology, ecology and river channel habitat:
mesoscale approaches to basin-wide challenges. Progress
in Physical Geography, Vol. 24, No. 2, pp. 195217.
NIENHUIS, P.H. 1992. Eutrophication, water manage-
ment and the functioning of Dutch estuaries and coastal
lagoons. Estuaries, Vol. 15, pp. 53848.
NRC (National Research Council). 1983. An evaluation
of flood-level prediction using alluvial river models.
Washington, D.C., National Academy Press.
NRC (National Research Council). 1992. Restoration of
aquatic ecosystems: science, technology, and public policy.
Washington, D.C., National Academy Press.
ODGAARD, A.J. 1981. Transverse bed slope in alluvial
channel bends. Journal of the Hydraulics Division, ASCE,
Vol. 107, No. HY12, pp. 167794.
ODUM, E.P. 1989. Ecology and our endangered life-support
systems. Sunderland, Mass., Sinauer.
OLESEN, K.W. 1987. Bed topography in shallow river
bends. Communications on Hydraulic and Geotechnical
Engineering. Delft, the Netherlands, Delft University of
Technology. No. 871.
ONEILL, R.V.; JOHNSON, A.R. and KING, A.W. 1989.
A hierarchical framework for the analysis of scale.
Landscape Ecology, Vol. 3, Nos. 34, pp. 193205.
PARK, J.K. and JAMES, A. 1988. Time-varying turbulent
mixing in a stratified estuary and the application to a
Lagrangian 2-D model. Estuarine, Coastal and Shelf Science,
Vol. 27, pp. 50320.
PARK, J.K. and JAMES, A. 1989. A unified method of
estimating longitudinal dispersion in estuaries. Water
Science and Technology, Vol. 21, pp. 98193.
PETERS, J.J. 1988. tudes rcentes de la navigabilit. In:
Symp. Laccs maritime au Zare, Bruxelles, 5 Dcembre
1986. Acadmie Royale des Sciences dOutre-Mer,
pp. 89110.
PETERSEN, M.S. 1986. River engineering. Englewood
Cliffs, N.J., Prentice-Hall.
PHILLIPS, J.D. 1995. Biogeomorphology and landscape
evolution: the problem of scale. Geomorphology, Vol. 13,
pp. 33747.
PIGAY, H. 1997. Interactions between floodplain forests
and overbank flows: data from three piedmont rivers
of southeastern France. Global Ecology and Biogeography
Letters, Vol. 6, Floodplain forests special issue, pp. 18796.
PILKEY, O.H. and DIXON, K.L. 1996. The Corps and the
shore. Washington, D.C., Island Press.
POINCAR, J.H. 1914. Science et mthode. Paris,
Flamarion, 1908. (First Book, Chapter 4, Section II.)
Authorized German Edn., Wissenschaft und Methode.
Teubner, F.L. Lindeman.
wrm_appA.qxd 8/31/2005 12:08 PM Page 555
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
PONCE, V.M. 1989. Engineering hydrology: principles and
practices. Englewood Cliffs, N.J., Prentice-Hall.
PRITCHARD, D.W. 1955. Estuarine circulation patterns.
Proceedings of the American Society of Civil Engineers,
No. 81, pp. 111.
PRITCHARD, D.W. 1967. What is an estuary? Physical
viewpoint. In: G.H. Lauff (ed.), Estuaries. Washington, D.C.,
American Association for the Advancement of Science
(Publ. No. 83), pp. 35.
PRITCHARD, D.W. 1989. Estuarine classification: a help
or a hindrance. In: B.J. Neilson, A. Kuo and J. Brubaker
(eds.), Estuarine Circulation. Clifton, N.J., Humana.
pp. 138.
PROSSER, I.P.; DIETRICH, W.E. and STEVENSON, J.
1995. Flow resistance and sediment transport by
concentrated overland flow in a grassland valley.
Geomorphology, Vol. 13, pp. 7186.
RADEMAKERS, J.M.G. and WOLFERT, H.P. 1994. Het
Rivier-Ecotopen-Stelsel; een indeling van ecologisch relevante
ruimtelijke eenheden ten behoeve van ontwerp- en belei-
dsstudies in het buitendijkse rivierengebied. Publikaties en
rapporten van het project Ecologisch Herstel Rijn en
Maas, No. 61. Lelystad, RIZA.
RICHARDS, K.S. 1982. Rivers: form and process in alluvial
channels. London, Methuen.
RICHARDSON, E.V. 1997. Sedimentation. In: A.K. Biswas
(ed.), Water resources: environmental planning, management
and development. New York, McGraw-Hill, pp. 7398.
RICKLEFS, R.E. 1990. Ecology. New York, W.H. Freeman.
RILEY, A.L. 1998. Restoring stream in cities: a guide for
planners, policy-makers, and citizens. Washington, D.C.,
Island Press.
ROBERSON, J.A. and CROWE, C.T. 1996. Engineering
fluid mechanics, 6th edn., New York, Wiley.
RUST, B.R. 1978. A classification of alluvial channel sys-
tems. In: A.D. Miall (ed.), Fluvial Sedimentology. Calgary,
Canadian Society of Petroleum Geologists (Memoir 5),
pp. 18798.
SARKER, M.H. 1996. The morphological processes in the
Jamuna River. M.Sc. thesis H.H.290. Delft, IHE.
556 Water Resources Systems Planning and Management
SCHUMM, S.A.; HARVEY, M.D. and WATSON, C.C.
1984. Incised channels: morphology, dynamics and control.
Littleton, Colo., Water Resources Publications.
SEEHORN, M.E. 1992. Stream habitat improvement
handbook. Atlanta, Ga., US Department of Agriculture,
Forest Service (Technical Publication R8-TP 16).
SHIMIZU, Y. and ITAKURA, T. 1985. Practical computa-
tion of two-dimensional flow and bed deformation in alluvial
streams. Sapporo, Japan, Hokkaido Development Bureau
(Civil Engineering Research Institute Report).
SHIMIZU, Y. and ITAKURA, T. 1989. Calculation of
bed variation in alluvial channels. Journal of Hydraulic
Engineering, ASCE, Vol. 115, No. 3, pp. 36784.
SILVESTER, R. and HSU, J.R.C. 1997. Coastal stabiliza-
tion, advanced series on ocean engineering, Vol. 14,
Singapore, World Scientific Publications.
SIMONS, D.B. and SENTRK, F. 1992. Sediment transport
technology. Littleton, Colo., Water Resources Publications.
STANLEY, S.J. and SMITH, D.W. 1992. Lagoons and ponds.
Water Environment Research, Vol. 64, No. 4, June, pp. 36771.
STARKEL, L. 1990. Fluvial environment as an expression
of geoecological changes. Zeitschrift fur Geomorphologie,
Suppl.-Bd. 79. Gebruder Borntraeger, pp. 13352.
STATZNER, B. and HIGLER, B. 1985. Questions and
comments on the river continuum concept. Canadian
Journal of Fisheries Aquatic Sciences, Vol. 42, pp. 103844.
STRUIKSMA, N.; OLESEN, K.W.; FLOKSTRA, C. and DE
VRIEND, H.J. 1985. Bed deformation in curved alluvial
channels. Journal of Hydraulic Research, IAHR, Vol. 23,
No. 1, pp. 5779.
THOMANN, R.V. and MUELLER, J.A. 1987. Principles of
surface water quality modelling and control. New York,
Harper and Row.
THORNE, C.R. 1990. Effects of vegetation on riverbank
erosion and stability. In: J.B. Thornes (ed.), Vegetation and
erosion: processes and environments Chichester, UK, Wiley,
pp. 12544.
THORNE, C.R.; HEY, R.D. and NEWSON, M.D. (eds.).
1997. Applied fluvial geomorphology for river engineering
and management. Chichester, UK, Wiley.
wrm_appA.qxd 8/31/2005 12:08 PM Page 556
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix A: Natural System Processes and Interactions 557
THORNE, C.R.; RUSSELL, A.P.G. and ALAM, M.K. 1993.
Planform pattern and channel evolution of the
Brahmaputra River, Bangladesh. In: J.L. Best and C.S.
Bristow (eds.), Braided rivers. London, Geological Society
(Spec. Publ. No. 75), pp. 25776.
THORNES, J.B. 1990 The interaction of erosional and veg-
etational dynamics in land degradation: spatial outcomes.
In: J.B. Thornes (ed.). Vegetation and erosion: processes and
environments. Chichester, UK, Wiley, pp. 4153.
TICKNER, D.; ARMITAGE, P.D.; BICKERTON, M.A. and
HALL, K.A. 2000. Assessing stream quality using infor-
mation on mesohabitat distribution and character. Aquatic
Conservation: Marine and Freshwater Ecosystems, No. 10,
pp. 17996.
TOOTH, S. and NANSON, G.C. 1999. Anabranching
rivers on the Northern Plains of arid central Australia.
Geomorphology, Vol. 29, pp. 21133.
TORREY, V.H. III. 1995. Retrogressive failures in sand
deposits of the Mississippi River. In: C.R. Thorne, S.R. Abt,
F.B.J. Barends, S.T. Maynord and K.W. Pilarczyk (eds.). River,
coastal and shoreline protection; erosion control using riprap and
armourstone. Chichester, UK, Wiley, 1995, pp. 36177.
TOWNSEND, C.R. 1996. Concepts in river ecology: pattern
and process in the catchment hierarchy. title Archiv fur
Hydrobiologie, Supplement 113, Large Rivers, No. 10, pp. 321.
TRIMBLE, S.W. and MENDEL, A.C. 1995. The cow as
a geomorphic agent; a critical review. Geomorphology,
Vol. 13, pp. 23353.
TSUJIMOTO, T. 1999. Fluvial processes in streams with
vegetation. Journal of Hydraulic Research, Vol. 37, No. 6,
pp. 789803.
TURNER, M.G.; DALE, V.H. and GARDNER, R.H. 1989.
Predicting across scales: theory development and testing.
Landscape Ecology, Vol. 3, Nos. 34, pp. 24552.
TURNER, R.K. and BATEMAN, I.J. (eds.). 2001. Water
resources and coastal management. Cheltenham, UK,
Edward Elgar.
USEPA (United States Environmental Protection Agency).
1995. Ecological restoration: a tool to manage stream quality.
Washington, D.C., USEPA.
USEPA (United States Environmental Protection Agency).
2002. Wetlands, oceans and watersheds. www.epa.gov/owow
(accessed 13 November 2004).
VAN BENDEGOM, L. 1947. Some considerations on river
morphology and river improvement. De Ingenieur, Vol. 59,
No. 4, (in Dutch; English transl.: National Research
Council of Canada, Tech. Translation 1054 1963).
VANNOTE, R.L.; MINSHALL, G.W.; CUMMINS, K.W.;
SEDELL, J.R. and CUSHING, C.E. 1980. The river con-
tinuum concept. Canadian Journal of Fisheries and Aquatic
Sciences, Vol. 37, No. 1, pp. 1307.
VANONI, V.A. (ed.). 1975. Sedimentation engineering.
New York, American Society of Civil Engineering Press.
WARD, J.V.; TOCKNER, K. and SCHIEMER, F. 1999.
Biodiversity of floodplain river ecosystems: ecotones and
connectivity. Regulated Rivers: Research and Management,
Vol. 15, pp. 12539.
WILLIAMS, G.P. and WOLMAN, M.G. 1984.
Downstream effects of dams on alluvial rivers. Washington,
D.C., US Geological Survey (Professional Paper 1286).
WOLMAN, M.G. 1954. A method of sampling coarse
river bed material. Transactions of the American Geophysics
Union, Vol. 35, No. 6, pp. 9516.
WOLMAN, M.G. and LEOPOLD, L.B. 1957. River flood
plains: some observations on their formation. Washington,
D.C., USGS (Professional Paper 282C).
YANG, C.T. 1996. Sediment transport theory and practice.
New York, McGraw-Hill.
ZEDLER, J. 1996. Ecological issues in wetland mitigation:
an introduction to the forum. Ecological Applications,
Vol. 6, No. 1, pp. 337.
YEH, G.T.; CHENG, H.P.; CHENG, J.R. and LIN, H.C.J.
1998. A numerical model simulating flow, contaminant, and
sediment transport in watershed systems. Vicksburg, Miss.,
USACE (Waterways Experiment Station Technical Report
CHL-98-15, WASH12D).
ZWOLINSKI, Z. 1992a. Sedimentology and geomorphol-
ogy of overbank flows on meandering river floodplains.
Geomorphology, Vol. 4, No. 6, pp. 36779.
ZWOLINSKI, Z. 1992b. Sedimentology and geomor-
phology of overbank flows on meandering river flood-
plains. In: G.R. Brakenridge and J. Hagedorn (eds.),
Floodplain Evolution. Geomorphology, Special Issue, Vol. 4,
No. 6, pp. 36779.
wrm_appA.qxd 8/31/2005 12:08 PM Page 557
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management
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
wrm_appB.qxd 8/31/2005 12:08 PM Page 558
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Monitoring is the process of observing what is happening.
Managers of water resources systems need to know
what is taking place in their systems, both over time
and over space. This usually requires sampling one or
more elements or features of the system according to pre-
arranged schedules, using comparable methods for data
measuring or sensing, recording, collection and analysis.
Monitoring provides information on the state of the system.
Adaptive management is the action taken in response to
that information with the aim of improving how the system
performs.
Adaptive management can be active or passive. Active
adaptive management, defined by Holling (1978),
involves managing the system in ways that maximize the
understanding obtained from the monitored data. Active
adaptive management involves experiments aimed at
discovering the limits of system vulnerability and
resilience. Such management can be risky. It is possible
for management experiments to degrade the system, for
example by killing some of the species in an ecosystem
that is managed to protect them. For this reason managers
559
Monitoring and
Adaptive Management
Monitoring the impacts or outcomes of any water management policy provides a
way of assessing just how well the policy meets expectations. Developing a
monitoring plan requires the identification of performance indicators and how
frequently and accurately they will be measured over time and space. The major
challenge in monitoring is to make the information obtained fit the information
needed and then to act on it, as and when appropriate. This is called adaptive
management: what most of us do throughout our lives. Over time managers should
be able to improve their management and their monitoring policies on the basis of
what they learn about the system they are managing. Just how much and how well
they learn will largely depend on the effort given to developing and implementing
an effective monitoring and adaptive management strategy.
Appendix B:
are often reluctant to assume those risks just to learn
more. Hence, if adaptive management is implemented, it
is often passive adaptive management, sometimes called
adaptive assessment or adaptive implementation. Passive
adaptive management strategies strive to manage the
system in the best way possible, correcting mistakes and
implementing changes on the basis of what is learned as
it is learned, but without making possibly damaging
experiments.
The questions to be addressed in establishing a
monitoring programme for adaptive management are
what to measure, where and how often to measure or
record each parameter, with what accuracy, and why. The
challenge is to know when the information derived from
a monitoring programme is sufficient to act upon, as well
as what action to take. Often the results, especially the
ecological impacts, of a change in a management policy
will be observable only long after the change has been
implemented.
These questions are addressed in the various stages
of a monitoring programmes life cycle, as shown in
Figure B.1. The process of monitoring and adaptive
management is more than just measuring system
wrm_appB.qxd 8/31/2005 12:08 PM Page 559
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
560 Water Resources Systems Planning and Management
water management:
adaptive assessment
information
objectives, and needs
information use
monitoring plan
start
reporting results
network design data analyses
data sampling,
collection, storage
E
0
2
0
2
1
9
e
Figure B.1. Chain of activities in
monitoring and assessment
(Adriaanse and Lindgaard-
Jorgensen, 1995).
attributes, collecting, storing, analysing and publishing
the measured data, and then acting on the results. It is a
sequence of related activities that starts with the identifi-
cation of information needs, and ends with the use of
the information. All too often time and money is invested
to obtain data before sufficient thought has been given
to how those data will or could be used and their value
compared to their cost.
Modelling can help identify just what data are needed
for the decisions being considered, and how accurate
those data need be. (Some modelling examples in
Chapters 4 and 7 illustrate this.) A dilemma, of course,
is that if data obtained from a current monitoring
programme are intended to be of value to future
managers, it is difficult to know today just what data and
what precision those future managers may want.
The design of a monitoring system starts with
defining the information needed for decision-making.
The information needed determines the attributes to
be measured the types of data to be collected and
the kinds of analyses to be applied to them. The
monitoring plan specifies these data, their required
accuracy and their frequency of measurement.
Frequency of measurement and the density of monitor-
ing sites are in part dependent on the variability of an
attributes or parameters value over time and/or space,
and just how important it is to capture this temporal or
spatial variability.
Once the network design has been defined, data
collection, storage and analysis procedures need to be
specified, along with plans for reporting and disseminat-
ing the results. This information should be included in
the monitoring plan.
The last element in Figure B.1, information use, is
input to the managers of the system. Actions taken to
manage the system more effectively on the basis of this
new information may lead to changes in the information
needs. As information needs change, this chain of activi-
ties will repeat itself. Each component of the monitoring
cycle is subject to change and enhancement over time,
reflecting changes in knowledge or goals, improvements
in methods and instrumentation, and budgets.
wrm_appB.qxd 8/31/2005 12:08 PM Page 560
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 561
2. System Status
Information needs depend on the issues and problems
facing water managers. To identify management priorities
several activities are needed. As suggested in Figure B.2,
these include identifying the functions and the uses
served by the system being managed. This in turn
involves carrying out inventories and assessments of
available and accessible information, making field surveys
if information is lacking, identifying criteria and targets,
and evaluating the use, costs and benefits of additional
data.
Specifications of information needs should be
based on the analysis of water management issues
and opportunities. These in turn are determined from
inventories, surveys, stakeholder concerns about what
needs attention, and failures to meet standards, targets, or
management criteria. Issues and targets can include
existing or future problems or threats, e.g. flooding, toxic
contamination, water supply shortage. They can include
the full range of qualitative and quantitative aspects in
multipurpose system management (see Table B.1).
Additional surveys and monitoring are needed if
sufficient data are not available to identify the causes of
known problems. Surveys and monitoring generate new
data that can relate to a broad range of subjects, such as
the evaluation of site conditions (e.g. post-flood surveys),
the variability of monitoring parameters in space and
time, or the screening of the occurrence of pollutants or
toxic effects in water and sediments.
information
needs
inventory of
available information
surveys to
supplement available info
identification of
uses, functions, issues
criteria for
uses, functions, issues
analyses of issues
and information
E
0
2
0
2
1
9
f
Figure B.2. Water management activities to
identify information needs.
Table B.1. System functions and uses.
f
l
o
o
d
i
n
g
s
c
a
r
c
i
t
y
e
r
o
s
i
o
n
/
s
e
d
i
m
e
n
t
a
t
i
o
n
b
i
o
d
i
v
e
r
s
i
t
y
c
o
n
t
i
n
u
i
t
y
s
a
l
i
n
i
t
y
a
c
i
d
i
f
i
c
a
t
i
o
n
p
o
l
l
u
t
i
o
n
e
u
t
r
o
p
h
i
c
a
t
i
o
n
issues
functions & uses
human
health
ecosystem
functioning
fisheries
recreation
drinking
water
irrigation
industrial
use
hydro
power
transport medium
& navigation
E
0
2
0
2
1
9
g
wrm_appB.qxd 8/31/2005 12:09 PM Page 561
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Water quality surveys can give additional insight
into the functioning of the aquatic ecosystem and the
incidence of pollution and toxic effects in the water.
Investigating the structure of the macro-invertebrate
community, the upstream and downstream differences
in the river reach, and the changes that occur over time
can provide an assessment of the biological quality of
the aquatic zone of a river. Chemical screening of
surface water, sediment and effluents at hot spots and
key locations can allow an assessment of the chemical
quality of the aquatic zone of a river. Additionally,
specific target compounds, toxic effects in surface
water, sediments and effluents that might be expected
can be measured and analysed. By means of ecotoxico-
logical tests, the concentrations of a broad range of
chemicals and variation in sensitivity among species
can be assessed.
2.1. System Status Indicators
System status or performance indicators should ideally
be measurable parameters. Indicator parameters should
sufficiently characterize or represent functions and
uses of water bodies and/or be of value for testing
the effectiveness of management decisions. For exam-
ple, total phosphorus can be a good indicator of
the eutrophication status of the river. The presence of
salmon is an indicator for the ecological state of rivers
such as the Columbia and Rhine Rivers. Such indicators
should include those that the public care about
and thus suitable for communicating with policy
makers or the public. Important criteria for choosing
an indicator are:
Communication. The indicator should be appealing to
those who will use it.
Simplification. The indicator should provide insight
into the situation, without having to go into great
detail. The oxygen concentration in a river, for exam-
ple, is a good indication of its quality for aquatic
organisms; there is no need to identify the concentra-
tions of all oxygen-demanding substances.
Data availability. Sufficient data for the indicator
should be available, otherwise its information content
may be unreliable.
562 Water Resources Systems Planning and Management
3. Information Needs
As many monitoring programmes are data rich, but
information poor, attention should be directed towards
the end product of monitoring: information. The ultimate
goal of monitoring is to provide information needed to
answer specific management questions. Thus, a critical
step in developing a successful, tailor-made and cost-
effective monitoring programme is the clear definition and
specification of information needs. These must be known
before design criteria can be derived for the development
or planning of a monitoring and assessment system.
Information needs come from management objectives.
Management objectives might be stated in terms like 20%
reduction of pollution in the next five years, or No more
interruptions in the intake of drinking water within two
years. There should be an element of relativity (percent-
age reduction of ) or quantity (no more or less
than ) in the specification. Next to that, an element of
time ( within two years) is imperative. Consider for
example the Salmon 2000 slogan that summarized the
efforts to restore the ecology of the River Rhine this past
half decade. Getting the salmon back into the Rhine in
the year 2000 is a goal that can be measured. There is
an element of quantity (a more or less stable population
of salmon) and an element of time (the year 2000). These
elements can be converted into a monitoring strategy. The
effectiveness of a monitoring network can only be tested
when the information needs related to one or more
management objectives are defined.
Five different approaches to defining information
needs can be distinguished:
The effect approach. There may be an adverse effect of
some kind that should be reduced within a certain
period. The element of relativity can be used here.
The source approach. There are sources that cause
adverse impacts. Their effects have to be reduced, for
example by reducing their loads on the environment.
This is closely related to the effect approach.
The achievement approach. There is a goal to be
achieved within a given time period. This approach
gives an impression of the effects of the intended
actions after a stated time. Salmon back in the River
Rhine In the year 2000 is an example of this approach.
wrm_appB.qxd 8/31/2005 12:09 PM Page 562
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 563
The background approach. There may be no change in
a given parameter, or the river has to be back in
its original state by a given time. This is usually com-
parable to the ecological function of a water resource
system.
The function approach. The water system has to fulfil a
specific function, such as being fit for salmon and/or
swimming.
The different approaches listed above are often inter-
related. Nevertheless, each of them can help facilitate the
identification and specification of information needs.
These needs should be identified and described in
sufficient detail to ensure the design of a monitoring
and assessment system that will meet them. Examples of
such specified information needs include:
The identification of appropriate parameters and/or
indicators.
The definition of criteria for assessment, such as
considerations for the setting of standards or criteria
for the choice of alarm or trigger conditions for early
warning in the event of floods or accidental pollution.
Requirements for reporting and presentation of the
information, e.g. visualization, degree of aggregation,
indices.
Relevant error margins specified for each monitored
indicator. What detail is relevant for decision making?
Response times identifying when specified information
is needed. In early-warning procedures, information
is needed within hours, whereas for trend detection
information is needed weeks or even months after
sampling.
Reliability requirements. To what extent is false infor-
mation allowed? It is often impossible or prohibitively
expensive to have 100% reliability. Depending on
the consequences of error, more or less reliable infor-
mation should be required. Together with the relevant
margins of error, these reliability requirements may
be a determining factor when selecting locations,
frequencies and methodologies in the design of moni-
toring programmes.
Information needs should be comparable between places
and situations, and should be linked to specific issues,
which are in turn linked to specific management needs.
Interested stakeholders should be involved in the process
of specifying information needs alongside institutions
responsible for the management and use of water
resources. Both information users and information
producers should be identified and should interact
closely.
3.1. Information Objectives and Priorities
Information objectives indicate the intended use or
purpose of the information and the management concern.
The information may be needed for compliance with
established standards or targets, for planning, for early
warning of hazards, or for scientific understanding of
natural processes or impacts.
As information needs are derived from issues, the pri-
oritization of issues leads to a prioritization of information
needs. Information is mostly needed on high-priority
issues. If the same information need arises from various
issues, this information need should be given increasing
priority. By collecting this information once, a variety of
issues may be addressed.
Information objectives evolve as water management
develops, targets are met or policies change. Consequently,
monitoring strategies often need to be adapted to
changing information needs over time. Information
needs require a regular rethinking (revision) of the
information strategy in order to update the concept.
When revising monitoring strategies for time-series
measurements, one should not neglect the need for
continuity (in parameters being measured, in locations
where data have been collected, in the analytical methods
used and so on). This continuity is needed to detect
significant and reliable trends in system performance
characteristics.
4. Monitoring Plans
A monitoring plan provides the basis or rationale for the
design of monitoring networks. Monitoring plans should
specify what has to be measured (also in terms of
accuracy, type 1 and type 2 errors, etc.) and why. The
network design specifies how and where it should be
measured. The monitoring plan should also include the
data analysis and reporting procedures that in turn can
influence network design requirements.
wrm_appB.qxd 8/31/2005 12:09 PM Page 563
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Elements of a monitoring plan are:
the information needs that will be covered by the
monitoring programme and, equally important, the
information needs that will not be covered by
the monitoring strategy
the type of monitoring (physical, chemical, biological,
hydrological, early warning, effluent), the indicator
variables to be measured and the preconditions for
selecting locations (minimum/maximum distance
from border, intake point, etc.) and sampling frequen-
cies (in terms of reliability)
the calculation methods, and the graphical, statistical
and other tools (such as indices) to be used
the preconditions, suppositions, assumptions, and
descriptions of the area, relevant industries, major
demands and so on
the organizational responsibilities for the monitoring
programme
a plan for the design and implementation of the
monitoring network
an analysis of the risks and the possible problems that
can lead to the failure of the monitoring programme
Monitoring plans are the bridges between information
needs and monitoring networks.
The selection of the parameters to be monitored is
usually based on their indicative character, their occur-
rence and the hazards they present. For reasons of
efficiency, the number of monitored parameters should
be restricted to those whose uses are explicitly identified.
The benefits derived from measuring any additional
parameter should be compared to the cost incurred.
Since the benefits will probably not be expressed in
monetary terms, this usually has to be a qualitative
comparison, using judgement.
Integrated water management involves the considera-
tion of all aspects of a water resources system. This
includes its watersheds, its aquifers, rivers, lakes, reser-
voirs and wetlands, its estuaries and coastal waters, its
natural ecosystems, its regulatory measures for environ-
mental media, its management and monitoring strategies,
and its relations to social and economic factors. An inte-
grated approach eschews a focus on only localized sepa-
rate components of the system in isolation. Monitoring
plans should reflect these interdependencies and facilitate
an integrated approach to water management.
564 Water Resources Systems Planning and Management
An integrated management approach includes
humans as a central element in the system. This implies
recognition of social, economic, technical and political
factors that influence the ways in which human beings
use and affect the system. These factors should be
assessed because of their ultimate effect on the systems
integrity. For example, in trying to restore the hydrology
and ecosystem of the Everglades region of South Florida
to what it was like a half century ago, one cannot ignore
the addition in recent decades of some 20 million people
who now make their home there for at least part of the
year. The needs for reliable water supplies and flood con-
trol were much less fifty years ago, as were the pollutants
discharged from municipalities, agriculture and industry.
These extra people, together with their pollutants, are not
going away. To manage integrated systems, managers
need to know their condition and how their condition
reacts to people and their activities. This in turn requires
monitoring not only of water quantities and qualities and
ecological indicators, but also of their major drivers:
humans and their activities.
5. Adaptive Monitoring
One approach to monitoring when the precise level of
detail or precision is not known is an adaptive stepwise or
phased plan, proceeding from coarse to fine assessments.
At the conclusion of each step, an evaluation can be made
of whether or not the information obtained is sufficient.
Such stepwise testing strategies can result in a reduction in
unnecessary data collection. In general, a phased approach
to monitoring, going from broad to fine and from simple
to advanced, may also be cost effective. Additionally, for
developing countries or countries in transition, stepwise
monitoring strategies going from labour-intensive to
technology-intensive methods might be appropriate. In
many cases, the lack of consistent and reliable data and the
lack of a baseline against which progress can be measured
are additional arguments for a phased approach.
5.1. Risk Assessments For Monitoring
Risk assessment can help considerably in prioritizing
monitoring activities. For example, consider flood protec-
tion and water quality management.
wrm_appB.qxd 8/31/2005 12:09 PM Page 564
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 565
The central question in flood prevention is what protec-
tion is available at what price, and what remaining risk
has to be accepted by society. Risk assessment (or,
more comprehensively, flood risk management that
includes risk assessment, mitigation planning and the
implementation of measures) will show which hydro-
logical, meteorological and other data should be
monitored or observed.
The quality of water in a small, sparsely populated
catchment is unlikely to pose a risk to human health.
Conversely, if there are refuse dumps or industrial
plants in a catchment, there may be a high risk to
human health and/or aquatic ecosystems. Thus, by
using risk assessment one can decide which of all the
monitoring activities have higher or lower priority.
Identify priorities by asking what may go wrong when
insufficient information is available (because of a lack
of monitoring). What loss is likely when less than
optimal decisions are made because of insufficient
information or money? The same questions can be
asked in the design or optimization of monitoring net-
works. What are the consequences for decision-making
if there are no or only limited results from monitoring?
Risk assessment can also be used to prioritize specific pol-
lutants, on the basis of their physico-chemical properties
and toxicity. Risk assessment, regarding both biological
agents and chemical substances, can also help in setting
priorities for establishing health-related monitoring
and/or early-warning systems in general, and in selecting
appropriate parameters for monitoring in particular.
Although still to be developed, good systems will include
hazard identification, doseeffect relationships, exposure
assessment and risk characterization (both qualitative and
quantitative).
5.2. Use of Models
Models (numerical, analytical or statistical) can assist in
developing a monitoring and assessment plan. They can
help in screening (the preliminary evaluation of) alterna-
tive policies, in optimizing monitoring network design, in
assessing the effectiveness of implemented measures, and
in determining the physical and health impacts on
humans and ecosystems. Computer models linked with
geo-referenced databases can be used to analyse the
impact of proposed measures, for example by simulating
the flow and water level variations in a river and on
floodplains during floods. Models can play an important
role in early-warning systems (flood forecasting, travel
time computations in emergency warning systems in
the event of accidental pollution). They can be used in
addition to monitoring to help understand what, where,
and how often and how accurately to monitor.
Successful mathematical modelling for planning
monitoring programmes is possible only if the modelling
activities are integrated with data collection, data process-
ing and other techniques and approaches for identifying
system characteristics. One cannot calibrate and verify
models without data, yet even uncalibrated models can
often help identify just what data are needed and how
accurate they need to be for the purposes of management
and decision making.
6. Network Design
The design and operation of monitoring networks
includes the selection of attributes or parameters to be
measured, the locations where they are measured, and
their sampling frequencies. The network design should
meet the requirements specified in the monitoring plan.
The type and nature of the system being monitored
should be understood (most frequently through
preliminary surveys), particularly its spatial and temporal
variability. Monitoring of the quality and biology of
the aquatic environment should be coupled with the
appropriate hydrological quantity monitoring. Finally,
arrangements should be made to ensure the quality of
data. This requires the periodic checking and mainte-
nance of the monitoring network as well as the computer
data management and storage system.
The design of a monitoring network is influenced
by its purpose or purposes. Table B.2 summarizes some
of the design considerations that will vary for different
purposes.
It will often be necessary to carry out a preliminary
sampling and analysis programme to obtain a better
understanding of the parameters to be monitored. The
objective in the survey, a better understanding of the
processes, is not directly related to the information need,
but will give information for the monitoring network
design. For instance, a survey to find out the distance it
wrm_appB.qxd 8/31/2005 12:09 PM Page 565
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
takes for the water at the confluence of two rivers to mix
completely will be useful when choosing water quality
sampling locations.
6.1. Site Selection
The desired spatial coverage of a monitoring network
depends on the spatial variability of the data being meas-
ured, and how important it is to capture or measure that
variability. For monitoring meteorological parameters on
watersheds, the shorter the distances between adjacent
sampling sites, the greater probability of measuring any
spatial variation that may exist. However, more monitor-
ing equipment will be needed and hence costs will
increase. There is a tradeoff between the accuracy of the
estimates of spatially varying parameter values and cost.
These relationships are shown in Figure B.3.
Referring to Figure B.3, if rainfall is being recorded
at the monitoring sites 1 through 9 the estimate of the
average rainfall over the entire watershed would be much
more accurate than if only site 2 or 8 existed, or even if
both existed and were used to make that estimate.
Similarly for streamflow and quality gauges a, b, and c.
Whatever the number of stream gauges to be used, they
566 Water Resources Systems Planning and Management
should be placed where one knows significant changes
are likely to occur, such as just upstream and downstream
of the confluence of tributaries.
6.2. Sampling/Measurement Frequencies
Water quantity and quality, sediment characteristics and
biota vary over time as well as space. This variation over
time affects decisions about the frequency of sampling. The
objectives of monitoring strongly influence the time scale of
interest, e.g. long-term variations for trend detection, short-
term changes for flood forecasting and early warning. The
required frequencies and methods of sampling (continuous
sampling, grab sampling, composite sampling, etc.) will be
dictated by both the temporal variability and the monitor-
ing objectives. The costaccuracy tradeoff relation shown
in Figure B.3 applies to temporal sampling frequency as
well as to spatial coverage.
6.3. Quality Control
Quality control should be performed to ensure the
achievement of an acceptable standard of accuracy and
precision.
variables long term interest/
policy goals
included
in the standard
depending
on possible
accident spills
locations significant
and independent
locations
representative for
the water system
hot spots for
accidental spills
and interests
reliability/
accuracy
no discontinuities
in time-series
representative for
the test period
reliability
more important
than accuracy
sampling
frequency
relatively low,
depending on dynamics
of the aquatic system
relatively low,
depending on dynamics
of the aquatic system
high frequency,
depending on dispersion
of pulse inputs
response
time
not important less important crucial
objectives
design
aspects
trend
detection
testing for
compliance
early warning
E
0
2
0
2
2
0
a Table B.2. Network design aspects for
various monitoring objectives.
wrm_appB.qxd 8/31/2005 12:09 PM Page 566
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 567
6.4. Water Quantity Monitoring
The main hydrological and hydro-meteorological param-
eters such as precipitation, snow cover, water level,
river flow, suspended and bed load sediment discharges,
evaporation and transpiration, soil moisture, and data on
ice conditions are measurable. The use of such data
has increased during the past decades, due largely to
developments in the use of models and forecasting
systems that require these data.
Spatial representativeness is crucial in the selection of
monitoring sites for hydro-meteorological parameters.
The most common locations for gauging stations are the
lower reaches of rivers, immediately upstream of the river
mouth or where the rivers cross borders, near the conflu-
ence with tributaries and at major cities along the river. In
general, a sufficient number of gauging stations should be
located along the main river to permit interpolation of
water level and discharge between the stations. Water
balances require observation stations at small streams
and tributaries as well. Gauges on lakes and reservoirs
are normally located near their outlets, but sufficiently
upstream to avoid the influence of drawdown.
Hydraulic conditions are an important factor in select-
ing sites on streams, particularly where water levels are
used to compute discharge using stage-discharge rating
curves. Unambiguous relationships are found at stations
located on streams with natural regimes, not affected by
variable backwater at the gauge caused by downstream
tributaries or reservoir operations or by tidal effects.
The frequency of measurements, data transmission
and forecasting depends on the variability of the hydro-
logical characteristics and the response time require-
ments. Systematic water-level recordings, supplemented
by more frequent readings during floods, are appropriate
floodplain
hill side
wetlands
groundwater
precipitation
tributary
headwaters
stream channel
watershed boundary
monitoring site
E
0
2
0
2
2
1
a
a
b
c
2
3
5
4
6 7
8 9
Figure B.3. Density of monitoring sites influences estimates of average conditions over a watershed or along a stream. The
density of monitoring sites also affects the cost of monitoring.
distance between adjacent sites
estimation
error
cost
E
0
2
0
2
2
1
b
wrm_appB.qxd 8/31/2005 12:09 PM Page 567
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
for most streams. The installation of water-level recorders
will usually be required for streams whose levels are
subject to abrupt fluctuations. For flood forecasting or
flood management, telemetric systems may be used to
transmit data whenever the water level changes by a pre-
determined amount. Continuous river flow records may
be necessary in the design of water-supply systems, and in
estimating the sediment or chemical loads of streams,
including pollutants.
Factors to be considered in determining the number
and distribution of discharge measurements within the
year include:
the stability of the stage-discharge relationship
seasonal discharge characteristics and variability
accessibility of the gauge in various seasons.
At new stations, many discharge measurements at differ-
ent flow levels are needed to define stage-discharge
relationships. At existing stations the frequency of
measurements is dictated in part by the number needed
to keep the stage-discharge relationship up to date.
Adequate determination of discharge during floods and
under ice conditions can be difficult, but is of prime
importance where applicable. In situations where the
channel shape can readily change during high-flow
conditions, keeping an up-to-date stage-discharge
relationship is a challenging (or perhaps indeed unattain-
able) goal.
6.5. Water Quality Monitoring
When water quality is monitored, water quantity (flow)
should also be monitored at the same sites and times.
For specific human uses, standards may dictate the
water quality parameters that need monitoring.
Management issues may also dictate the parameters of
interest. For ecological functioning, parameters are speci-
fied by the selected method of assessment (indices, habitat
factors) and regional reference communities. The selection
of hazardous pollutants as monitoring parameters will
usually depend on the specific problem substances
produced and/or discharged into the water and on their
probability of occurrence. In practice, this should be based
on results of site-specific preliminary surveys.
Nationally and internationally recognized lists of prob-
lem substances can often be used as the starting point for
568 Water Resources Systems Planning and Management
the selection of monitoring parameters. They are among
or indicative of the pollutants that are of general concern.
The availability of reliable and affordable analytical and
measurement methods may also influence the selection of
monitoring parameters.
In general, the selection of sampling sites is based
on how representative they are. The distance between
sampling locations can be critically evaluated from their
degree of correlation by statistical analysis of time series of
parameters. However, this is possible only if these time
series are available. In transboundary rivers, sampling
should preferably be performed at or near to border cross-
ings. Sampling in the river and in the main tributaries
upstream of the confluence can show the contribution
(e.g. pollution load) of different tributaries. The selection
of sampling sites downstream of a confluence should avoid
the uncertainties related to incomplete mixing. Mixing
zones can be several kilometres long, depending on the
widthdepth ratio and the turbulence of the main river.
Considerations of the local representativeness of the
sampling point at the river site may be based on prelimi-
nary surveys, taking into account the hydrology and
morphology of the river. In general, locations in the main
flow of the river will be chosen for water and suspended
solid sampling. Bottom sediment can best be sampled
in regions where the suspended material settles. As a
consequence, most sediment samples are taken near
riverbanks and in the downstream sedimentation area.
The number of sampling points for sediment monitor-
ing strongly depends on the objectives. For trend
detection, a low number of sampling points or mixing
samples into composite samples can sometimes yield
enough information. If spatial information is to be
estimated, the number of sites will increase and com-
posite samples will normally not be used.
The selection of the sampling frequency for surface
water quality parameters should be based on the variabil-
ity in parameter values, and on the statistical significance
and accuracy required for specific objectives. Examples
of specific objectives include trend detection, load calcu-
lation and compliance testing.
Sampling frequencies for suspended solids are very
similar to surface water sample frequencies. For load cal-
culation, a higher sampling frequency is recommended
during the start of flooding periods, when the main load
of suspended solids is transported. The precision of the
wrm_appB.qxd 8/31/2005 12:09 PM Page 568
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 569
estimates obtained by sampling at regular intervals
depends mainly upon the distribution of the total load
over the year. The reliability of load estimates obtained
with current monitoring equipment can be improved by
increasing the sampling frequency.
The need to obtain information that is integrated or
differentiated over time and space should determine
the selection of methods for measurement and sampling
of water and sediment quality and biota. There are
various possible methods, including grab sampling,
depth integrated sampling, time-proportional composite
sampling and space composite sampling. Monitoring of
the biological status implies measurement and sampling
of biotic groups. Each group requires specific sampling
and measurement methods; fine mesh nets collect phyto-
plankton and zooplankton, for example, and waterfowl
can be measured through field observations. Some
indicative parameters like dissolved oxygen, pH, water
temperature and redox potential are best measured
in situ, using sensor-based instruments. Such instruments
require frequent calibration.
6.6. Ecological Monitoring
Monitoring of ecological parameters can be carried out
on the level of species, communities or ecosystems. For
many purposes, monitoring the habitat of communities or
ecosystems is appropriate and much less demanding
than doing so on a species scale. A habitat is the place
where an organism or a community of organisms lives,
and comprises all living and non-living factors or
conditions of the surrounding environment. Habitat
descriptions consider the physical environment together
with the representative floral and faunal assemblages
present. They are often suitable for environmental impact
assessments.
Habitat and community descriptions can be based on
two types of scales. The DAFOR scale (dominant, abun-
dant, frequent, occasional, rare), adapted from the Joint
Nature Conservancy Council (UK), can be used to assess
habitat types. It is designed to allow consistent recording
of habitat features within and between sites. This habitat
detail can be used with species lists produced from site
sampling. Only one feature may be recorded as dominant,
while any number of features may be recorded as
abundant, frequent, occasional or rare.
Species data recorded on site can be assessed using
the SACFOR scale (superabundant, abundant, common,
frequent, occasional, rare). Fauna and flora species can be
identified and recorded where possible. When sampling
is not applied, the most abundant species observed
and any noteworthy or rare species at each site can be
identified in situ, with a further list of species expected or
typical of such habitats.
6.7. Early-Warning Stations
Measurement systems in an early-warning station are
either substance- or effect-oriented. Chemical analysis
screening methods can detect increases in concentrations
of specific substances. Biological early-warning systems
can detect deterioration in water quality through the
biological effects on fish, daphnia, algae, bacteria and
other species.
The pollutants that may occur in hazardous concen-
trations should be monitored for early warning.
Automatic in situ sensors can measure simple indicative
parameters such as dissolved oxygen, pH, or oil. If the
detection of specific micro-pollutants (e.g. pesticides) is
needed, advanced, but more expensive, analytical systems
based on gas chromatography with mass spectrometry
(GC-MS), high-performance liquid chromatography
(HPLC) can be used. Toxicological effects in organisms on
various trophic levels can be measured with automated
biological early-warning systems.
Early-warning equipment puts high demands on
operation characteristics such as speed of analysis, identi-
fication capacity and reliability of operation. Characteristics
such as the precision and reproducibility of the analysis
are less critical.
Early warnings should provide enough time for
emergency measures to be taken. The relation between
response time (the interval between the moment of
sampling and the alarm) and the travel time of the
pollution plume in the river, especially in high flows,
from the warning station to the site where the water is
used (e.g. water intake for drinking water) influences the
location of an early-warning station.
The sampling sites should obviously be chosen in such
a way that no pollutants are missed in the sampled water.
The measurement frequency should be determined by the
expected size of pollutant plumes (elapsed time for the
wrm_appB.qxd 8/31/2005 12:09 PM Page 569
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
plume to pass the station). Dispersion of the plume
occurs between the discharge location and the sampling
location due to the discharge characteristics of the river.
Furthermore, the frequencies should provide sufficient
time to take action in the event of an emergency.
Additional (intensified) sampling is recommended after
the first indication of accidental pollution.
6.8. Effluent Monitoring
The selection of effluent monitoring parameters and their
priorities can be based on risk assessments. Existing
national or international priority lists of chemical
substances can also be helpful. In addition to specific
pollutants, an emphasis should be placed on aggregate
parameters and total effluent toxicity testing.
Sampling frequencies and sampling methods for
effluent discharges should be based on the amount and
variability of the effluent. Surveys of restricted duration
using continuous or high-frequency sampling can be
performed to gain the required insight into the discharge
characteristics of batch and continuous effluent genera-
tion processes. The statistical significance and accuracy
required for specific objectives (e.g. for compliance testing
or load calculation) can be a basis for selecting the
sampling frequencies and sampling methods.
7. Data Sampling, Collection and
Storage
7.1. Overview
Sampling is the first stage in the actual collection of
information. Methods of sampling include spot, periodic,
continuous and large-volume sampling. Which method is
most appropriate will depend on the variable of interest
and on the characteristics of the watershed or water body.
Sampling equipment should be designed to minimize the
contact time between the sample and the sampler and
the likelihood of sample contamination.
There are a number of sampling decisions that must be
made. The first is the choice of the precise sampling
site. This can be affected by the conditions at the site,
the distribution of what is being measured, and the ease
of access to the sampling site with the needed equipment.
570 Water Resources Systems Planning and Management
Second, the frequency and time of sampling must be
determined. Different time-based effects can influence this
decision. Natural cycles may occur, as well as production
and discharge cycles of industries or other facilities just
upstream. Third, there is the choice of the sampling
method. Some methods, even if recommended in the
network design plan, may not be usable in particular situ-
ations, for example where the water is too shallow. Fourth,
decisions must be made regarding the transporting,
stabilizing and storing the samples. Fifth, quality-control
procedures must be implemented. All sampling methods
should be periodically tested using field-based quality-
control and audit procedures specifically designed to
reveal the effectiveness of the entire sampling programme,
including those aspects relating to the transportation,
stabilization and storage of samples prior to analysis.
Finally, safety requirements have to be met.
To make monitored data rapidly and conveniently
available to users, they are almost always stored in
computerized data files. They include the measurement
data and associated meta data. The latter identify what
the measured data are, when and where they were
measured, what methods of measurement or laboratory
analyses were used and by whom, and so on.
Often the weakest link within the monitoring
programme chain is the proper storage of data. If these
data are not accessible and complete with respect to the
conditions and qualifiers pertaining to their collection
and analysis, or are not properly validated, then the data
are not likely to satisfy any information need.
Computer hardware and software used to store and
manage data must be tested, maintained and upgraded
regularly. The software has to insure against data loss.
It must identify the correct secondary data. Furthermore,
it should perform internal checks on the measured data,
such as correlation analysis and application of limit
pairs. Examples of software control functions are 0 pH
14.0, orthophosphate-P (total-P), dissolved heavy
metals (total heavy metals), and calculation of the 10%
(lower) and 90% (upper) limit pairs. The software should
give users a warning when data fall outside these ranges.
All such calculations should be tested for accuracy before
using a computer database management system.
Clear procedures should be agreed upon for the
interpretation and validation of the measurement data.
These will include how to deal with:
wrm_appB.qxd 8/31/2005 12:09 PM Page 570
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 571
data limitations such as missing values
sampling frequencies that change over the period of
record
multiple observations within one sampling period
uncertainty in the measurement procedures
censoring the measurement signals
small sample sizes
outliers (values that do not conform with the general
pattern of a data set)
measurement data rounding
data at or below the limit of detection.
7.2. Remote Sensing
A variety of remote sensing techniques using aerial
photography or satellite images are available for
monitoring some parameters. They may be used for the
identification of different vegetation types, biotopes and
landscape elements. Laser-altimetry provides a useful
technique for monitoring forest and grassland structure
and sediment bank development.
7.2.1. Optical Remote Sensing for Water Quality
A number of satellite-based optical sensors can be used
for monitoring of water quality. The spatial resolution of
these sensors renders this method of monitoring suitable
for large inland water bodies or coastal waters. For remote
monitoring of smaller water bodies where higher spatial
resolution is needed, aircraft-based sensors can be used.
When natural sunlight, including the visible spectrum,
hits the surface of a water body, some of it may be directly
reflected at the surface, some may enter the water where
it is absorbed and scattered by particles, and some may be
transmitted through the water. The amount of absorption
and scattering are the main factors influencing the reflec-
tion of light from the water, and thus the colour of the
water. The substances present in the water often have
unique characteristics with respect to the absorption and
scattering of light. Thus the colour of water varies with
the concentration of different substances in it.
Deep ocean water has a distinct blue colour because in
clear water there is a lot of scattering of blue light (wave-
length of blue) compared to the other wavelengths in the
visible spectrum. Light absorption is relatively low. If a
water body contains algae, then the light scattering by the
algae cells dominates the light scattering by water. Also,
there will be more light absorption at the wavelengths of
blue and red light. As a result, there is a relatively high
reflection of green light and the water looks green. Due to
different pigments present in different algae types, algae-
dominated water bodies can also take on a brown or
reddish colour. Dissolved organic substances (such as
humic acids) tend to give water a yellow colour, and
suspended sediments rich in organic matter (e.g. dead
cells) tend to give water a brownish colour.
The amount of reflection of the different colours can
be used to identify which substances are present in water
and, along with advanced analysis techniques, to deter-
mine the concentrations of different substances. The
substances that can be quantified on the basis of optical
measurements of reflected light include algae (chloro-
phyll), dissolved organic carbon (humic acids) and
suspended sediment.
Optical remote sensing can be applied for assessment
of water quality and classification of inland and coastal
waters. Sediment plumes can easily be observed and can
often give a good indication of the spreading of river
water (typically with high sediment concentrations) in
coastal seas. The extent of eutrophication can also be
monitored. With several consecutive images, the changes
in water quality over time can be followed. One dis-
advantage of remote sensing is that the frequency of
available images is inflexible, and on cloudy days no
images can be collected.
By using specialized software and computer models,
the measured spectral reflection data can be reworked
into water quality concentrations, as indicated in
Figure B.4. An important aspect in this process is to
correct for atmospheric influences on the measured
reflection. The measured reflection values are often very
low, so processes such as atmospheric scattering of light
can have a considerable influence on the measurements.
It is therefore important to remove the atmospheric influ-
ence by means of atmospheric correction procedures.
Sensors can measure the surface temperature as well as
reflected light. The method based on the reflection of
natural sunlight is called passive remote sensing, as
opposed to active remote sensing that measures the
reflection of a beam sent out by the sensor itself (e.g.
radar). It is possible to convert the measured spectral
reflection data from an airplane or satellite sensor into
wrm_appB.qxd 8/31/2005 12:09 PM Page 571
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
an image or map of reflection values. For water quality
information, the reflected light signal can be converted to
concentrations of water quality parameters, such total
suspended matter or chlorophyll. The steps involved in
this are shown in Figure B.4.
7.2.2. Applications in the North Sea
The water quality of the North Sea and inland waters is
based on concentrations of certain water quality param-
eters such as algae and suspended matter. These concen-
trations have traditionally been identified by collecting
and analysing water samples from fixed monitoring
platforms or from ships. Analyses of the samples are made
in the field or at a laboratory. This method has a number
of disadvantages. The spatial coverage of the sea is limited
due to high costs of collecting and analysing the many
samples required from each sampling site. Furthermore,
the conditions at a sampling site may be changing during
572 Water Resources Systems Planning and Management
the time that it takes to collect the required number of
samples.
Remote sensing is now being used as a relatively
inexpensive monitoring method to supplement traditional
water quality monitoring, by providing spatial coverage
of the water body at different particular times.
8. Data Analyses
Data analysis converts raw data into usable information.
Routine data analysis is commonly directed toward
obtaining information on average conditions, trends or
changing conditions, or to test for compliance with a
standard. To compare and trace information obtained
from the raw data, protocols for data analysis have to be
developed. Using these protocols, many data analysis
procedures can be automated.
Data analysis protocols should include:
A statement of the information to be produced. This is
directly related to the specified information need.
Procedures for preparing a raw data record for
graphical and statistical analysis, including how data
limitations such as missing values or outliers are to be
addressed before data analysis proceeds.
Means to visually summarize the behaviour of the
monitored data. Graphical presentations of the data
often serve to give a better understanding of data value
variability over time and space and help to interpret
statistical results.
Recommended statistical methods that yield the
desired information. The selection of methods should
match the statistical characteristics of the data being
analysed as well as the information need.
Data analysis protocols should be established before any
data are collected and analysed. Otherwise arguments can
develop over the analysis methods. Statisticians can and
do disagree over what statistical procedures are most
appropriate. Whatever methods are used, one should
understand the important assumptions that underlie
them, whether these assumptions are reasonable in the
particular application, and the consequences of violations
of the assumptions. If a data analysis protocol is agreed
upon, any subsequent discussion can focus on the result-
ing information.
atmospheric
model
remote sensing image
raw sensor data
filter for
atmospheric influence
above water
reflection data
filter for
air-water interface
optical model
and database
air-water
interface model
E
0
2
0
2
2
1
h
below
water reflection
filter for absorption
and scattering
maps of
water quality
Figure B.4. Steps for obtaining water quality data from optical
remote sensing.
wrm_appB.qxd 8/31/2005 12:09 PM Page 572
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 573
Water quality and biological samples often require
laboratory analyses. This is not the place to go into what
is required for the analysis of different types of param-
eters, except to note that whatever analyses are performed
they should be scientifically acceptable and validated.
Laboratory equipment should be properly maintained
and calibrated with the use of reference materials. The
laboratory should undergo effective internal as well as
independent quality-control audits and participate in
inter-laboratory check sample schemes. Laboratory
personnel should be properly trained. These and other
basic elements of quality assurance should be followed
and enforced to obtain reliable, verifiable results.
A major quality-control issue in data analysis is
traceability. It must be possible to trace back to the raw
data used in the analysis as well as to the exact analysis
method. Reproduction of any previously performed
analysis should lead to the same result.
Geographical information systems (GIS) are useful
tools for the interpretation of spatial data such as those
found on maps and satellite pictures. Integrating spatial
data with time-series data, each possibly originating from
different agencies/sources, into one system is not easy.
Standardized interfaces should be used to interconnect
databases and provide for integration with a GIS.
Relational databases can be used together with GIS and
data processing models. Data processing based on
accepted, compatible standards will make assessment and
reporting comparable, even when the software used is not
the same.
9. Reporting Results
Selecting the method or methods of presenting the data is
not a trivial issue. What methods are best depends to a
large measure on the target audience. Possible presenta-
tion techniques, from a detailed presentation to an
aggregated overview, include:
Tables that list measurement data. No data are lost
but the reader has to glean the needed or desired
information from the data.
Statistically-processed measurement data are transfor-
mations of the original data into values that make
changes in time and/or space more visible.
Graphs providing a medium in which, for instance,
trends can be recognized at a glance. Showing stan-
dards or other references in the graph puts the system
status in perspective. The amoeba-type presentation
discussed in Section 9.2 is an example of this. Graphs
may be line graphs, histograms, pie charts or various
other forms.
Geographically presented information shown on a
map. Different data from multiple locations can be
displayed as multiple layers of geographically referenced
information. This often provides a better understand-
ing of the spatial distribution of the parameters
involved.
Aggregated information for rapid interpretation of
large amounts of data, for example using indices.
Quality indices are often used for biological quality
assessments.
9.1. Trend Plots
Trends are often displayed as time-series plots, examples
of which are shown in Figure B.5. If a significant
long-term trend exists, it may be apparent upon visual
examination of a plot of the raw data. Many parameters
exhibit strong seasonality and therefore a typical range of
values can also be shown. In Figure B.5, the typical range
was defined to encompass approximately 70% of the
values for each month. In this case, 15% of the values
should exceed the high value and 15% should be lower
than low value. The shaded areas in Figure B.5 represent
the typical range across the middle of each plot.
Comparing the data points to that shaded area makes any
trend more apparent should it exist.
If seasonal variability is too great to use the time-series
plots to identify long-term trends, a twelve-month
moving mean can be calculated for each site. The moving
mean smoothes out seasonality in many cases and makes
long-term trends easier to see. If the moving mean plot
suggests that a trend might exist, the raw data can then be
analysed further for trends.
9.2. Comparison Plots
One way of comparing data is through the use of amoeba
plots. An amoeba plot is a schematic representation of
a given condition compared to the natural average or
wrm_appB.qxd 8/31/2005 12:09 PM Page 573
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
baseline condition. For the water body under study, a set
of parameters considered to be representative of the water
bodys condition is chosen. The reference system is
represented by plotting the value of the parameters
under natural conditions on a circle. The present values
of the selected parameters are plotted relative to the
circle. This provides an amoeba-like figure, representing
deviations from the reference or normal state, as illus-
trated in Figure B.6.
The figure shows stream health as indicated by eight
measures of stream bugs (benthic macroinvertebrates).
Stream bugs are excellent indicators of stream health.
They are relatively easy and inexpensive to collect. They
play a crucial role in the stream nutrient cycle, and their
populations affect the whole ecosystem. The presence or
absence of pollution-tolerant and intolerant bug types
can indicate the condition of the stream. Population
fluctuations might indicate that a change (positive or
negative) has occurred in the stream. One can detect
population fluctuations in a short period of time.
The circle in Figure B.6 represents the normal
(healthy) value of each parameter. Deviations from that
circle are expressed as percentages of these normal values.
Log scales are sometimes convenient.
Amoeba plots can be used to compare data at a given
site or compare the data at one site with those of other
sites in a specified region.
574 Water Resources Systems Planning and Management
Figure B.7 illustrates another way to compare the
baseline averages measured in one stream to the median
levels for all tributaries measured in the region. The shaded
area represents the range in which the middle 50% of all
caddisfly
taxa richness
intolerant
taxa richness
clinger
taxa richness
percent
predator
percent
dominance
(3 taxa)
stonefly
taxa richness
mayfly
taxa richness
total
taxa richness
E
0
2
0
2
2
1
c
Figure B.6. Amoeba diagram showing status of a system with
respect to the target or normal state of, in this example, eight
specified parameters (bug species).
Jan 80
Jan 80
82
82
84
84
86
86
88
88
90
90
92
92
94
94
96
96
98
98
2.0
1.5
1.0
0.5
0.0
20
10
0
81
81
83
83
87
87
85
85
89
89
91
91
93
93
95
95
97
97
99
99
m
g
/
l
o
C
E
0
2
0
2
2
1
j
north creek nitrate + nitrite
swamp creek temperature
Figure B.5. Examples of trend
plots. The shaded areas are the
ranges of values in a specified
percent of the data. A moving
mean plot is shown in the lower
graph for temperature data.
wrm_appB.qxd 8/31/2005 12:09 PM Page 574
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 575
site averages fell, i.e. 25% were higher and 25% were
lower than the shaded area. Along with these charts can
be tables, not shown here, that list the average, minimum
and maximum values for each parameter for each stream.
Other ways of displaying data are shown in Table B.3.
In this case the data are being summarized with respect to
the percentage of values that met specified standards.
Scorecards can also be used to compare data. Consider
a biotic integrity index composed of ten different indica-
tors of stream biology. Each indicator characterizes some
aspect of the community that responds to degradation.
The actual value of each indicator is calculated, and from
that value, a score of 1, 3, or 5 is assigned to the indica-
tor. A score of 5 indicates little or no degradation, a score
of 3 moderate degradation, and a score of 1 severe
degradation. The ten metric scores are then added
to produce the overall score that ranges from 10 to 50.
(If any indicator values are missing no score is given.)
The resulting scorecard is shown in Table B.4.
Data can be interpreted by risk assessments as well.
This refers to the comparison of measured, modelled or
predicted values with target values. Target (desired)
values of various parameters will be based on specific
functions of waters, such as use for drinking water or
recreation. The measured or predicted data are referred to
as Predicted Values (PV), and the target or desired values
are referred to as Target Values (TV). For the former,
terms like Predicted Environmental Concentrations (PEC)
of pollutants are sometimes used. For the latter, terms like
Predicted No Effect Concentrations (PNEC) or Maximum
12
11
10
9
8
7
6
75
50
25
0
17
15
13
11
9
7
150
100
50
0
10
8
6
4
2
0
50
40
30
20
10
0
8.5
8.0
7.5
7.0
6.5
6.0
2400
2000
1600
1200
800
400
0
D
O

(
m
g
/
l
)
t
u
r
b

n
t
u
p
H
O
R
T
H
-
p

(
u
g
/
l
)
T
D
T
-
p

(
u
g
/
l
)
N
H
3

(
u
g
/
l
)
N
O
3

(
u
g
/
l
)
t
e
m
p
e
r
a
t
u
r
e

(
o
C
)

E
0
2
0
2
2
1
d
Figure B.7. Comparison plots showing how
selected parameter values at one site compare
with the values at other sites in the area. The
grey bands identify the middle 50% ranges. The
small red squares are the specific site values.
wrm_appB.qxd 8/31/2005 12:09 PM Page 575
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Permissible Concentration (MPC) or function-related
directives are also used. The risk quotient is the ratio of
predicted (PV) over target (TV). This ratio will indicate
the relative priority of that parameter.
The outcome of a sediment quality assessment could
be expressed as a PEC/PNEC ratio. If this ratio is 1,
little priority is given to the potential risk derived. If this
ratio is 1, a certain risk is indicated. Classifying the
responses might help to visualise the estimated risks in
time trends or spatial gradients. The more function-
related the quality criteria used, the more specific the
conclusions that can be drawn on which function might
be impeded due to the pollution present.
9.3. Map Plots
Two examples of map plots are shown in Figures B.8
and B.9.
10. Information Use: Adaptive
Management
Management decisions will always be made on the basis
of uncertain information. These uncertainties in our
ability to predict the impacts of our management
decisions motivate the use of adaptive approaches to
management. Adaptive management is the process by
which management policies change in response to new
576 Water Resources Systems Planning and Management
knowledge gained from research and new information
obtained from monitored data about the system being
managed. Adaptive management requires a monitoring
programme to detect changes in the system, the ability to
evaluate trends in system performance and, finally, the
authority and willingness to modify management
decisions in response to those trends in an effort to
improve system performance.
E
0
2
0
2
2
1
f
Figure B.8. Map display showing lake shore areas, in red, that
are at risk of bank failure.
E
0
2
0
2
2
1
m
D.O.
Temperature
Turbidity
pH
Enterococcus
Fecal Coliforms
231
245
234
231
127
235
parameter
collected
16
17
15
1
64
172
7.1
6.9
6.4
0.4
50.4
73.2
#
of samples
#
of samples
not meeting
criteria
%
of samples
not meeting
criteria
Table B.3. Ways of displaying information pertaining to
whether or not sample data met the standards for the
selected parameters.
E020221k
1
2
3
4
5
6
7
8
9
10
11
Nicki Creek
Bear Creek
Eelco Creek
Simon Creek
Bug Creek
Jos Creek
Sus Creek
Jen Creek
Beaver Creek
Wolf Creek
Erik Creek
44
36
25
35
32
18
36
34
26
<30
<30
22
<30
<30
26
28
<30
26
26
28
dry
30
26
24
18
20
32
30
26
20
dry
22
28
24
28
dry
22
34
22
16
14
dry
28
24
26
24
dry
22
28
name
site
c
o
d
e
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
Table B.4. Scorecard for indices of biotic integrity.
wrm_appB.qxd 8/31/2005 12:09 PM Page 576
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 577
Models of the hydrological and ecological responses to
management decisions, together with monitored observa-
tions, are essential components of any adaptive approach
to management. An effective research strategy can lead to
improved monitoring designs, improved interpretation of
monitored parameter values, and improved predictive
power of models and other assessment tools used in
management. An integrated approach to monitoring,
modelling, research and management can lead to an
improved understanding of how the overall system
functions and how best management practices can be
implemented. Each component continually needs refin-
ing, as our understanding of the system being managed
increases.
Adaptive management and decision-making is a
challenging blend of scientific research, monitoring and
practical management that provides opportunities to act,
observe and learn, and then react. Both monitoring and
management actions need to be adaptive, continually
responding to an improved understanding that comes
from the analysis of monitored data in comparison to
model predictions and scientific research. It is a cycle, as
illustrated in Figure B.10. Adaptive management requires
explicit consideration of system structure and function,
well-defined management goals and actions, and assess-
ment of the anticipated system response to management
decisions.
The design of an adaptive management plan is best
accomplished in cooperation with policy-level personnel
who have the authority to commit the resources and
technical personnel needed to identify scientific issues
and evaluate monitoring data. Management that is to be
adaptive has to rely on monitored data, and given the
response times of many hydrological and ecosystem
performance criteria used to judge success or failure,
there must be a long-term commitment of resources to
monitoring and all that it entails. But monitoring costs
money and hence should be efficient. In the interest
of cost savings, some water management agencies
SI = 0.82 SI = 0.62 SI = 0.53 SI = 0.50
fish periphyton alligator Tree islands
habitat suitability comparisons
E
0
2
0
2
2
1
g
Figure B.9. Map displays
showing distribution of habitat
suitability index (SI) values for
four ecosystem indicators
applicable to southern Florida in
the United States. Separate
colors represent ranges of SI
values. The average index value
is shown for each indicator.
(Based on Tarboton et al., 2004).
E
0
2
0
2
2
1
e
r
e
search
monitor
system response
implement
management strategy
data interpretation,
model analysis
and improvement
model prediction
and management
plan improvement
Figure B.10. The continually updating process of an adaptive
approach to management.
wrm_appB.qxd 8/31/2005 12:09 PM Page 577
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
intensively monitor river basin processes and parameters
only periodically, say every five years, and carry out more
routine, low-intensity monitoring in the intervening
period.
11. Summary
The degree of success of any monitoring and adaptive
management programme can be expressed using the two
terms effectiveness and efficiency. Effectiveness is the
extent to which the information obtained from monitor-
ing meets the information needs of management, and the
extent to which management decisions reflect this
increased information and understanding. Efficiency is
concerned with obtaining the information at the least
possible cost in funding and personnel. The two aspects
address the two fundamental questions:
Are the data being collecting the right data?
Can the required information be obtained at a lower
cost?
The answers to these questions are strongly related to the
available budget for monitoring, and to the issue of
whether more or other information is needed. If more
information is required (e.g. biological monitoring in
addition to chemical; or ambient water and sediment
monitoring in addition to effluents) and no more money
is available, then certain aspects of the existing
programme must be reduced. For example, parameters
reflecting the combined impacts from a number of sepa-
rate parameters can be considered instead of single water
quality variables. The number of stations or variables,
or the frequency of sampling can be altered. If the need
for other information means reductions in current data
monitoring (e.g. measuring surface water quality data
instead of effluent data, or using biological classification
instead of chemical classification), then conflicts may
arise with existing regulations and data users.
The total benefits and costs related to the monitoring
information system are not only dependent on the num-
ber of sampling locations or the equipment and personnel
required for laboratory analyses. Benefits also include any
increased system performance due to the additional
monitoring information. To quantify the benefits in finan-
cial terms, it is necessary to calculate the consequences
578 Water Resources Systems Planning and Management
of a shortfall in the monitoring design, for example
the financial implications of incorrect decisions, or the
cost related to the absence of specific information on
a particular water system.
The design of a monitoring programme has to be
based on the information requirements, which in turn are
related to the needs of water managers. Monitoring
programmes should be iterative in character. The design
of the future monitoring system should be based on infor-
mation collected by the existing monitoring programme.
Monitoring programmes must be constantly designed,
specified, detailed, described, or documented and
updated to be sure that the system continually produces
the information desired.
12. References
ADRIAANSE, M. and LINDGAARD-JRGENSEN, P.
1995. Information system for water pollution control. Draft
paper.
HOLLING, C.S. (ed.). 1978. Adaptive Environmental
Assessment and Management. New York, Wiley.
TARBOTON, K.; LOUCKS, D.; DAVIS, S. and OBEYSEK-
ERA, J. 2004. Habitat suitability indices for evaluating water
management alternatives. West Palm Beach, Fla., South
Florida Water Management District, (Draft Technical
Report).
Additional References (Further Reading)
ADRIAANSE, M.; NIEDERLNDER, H.A.G. and
STORTELDER, P.B.M. 1995. Chemical monitoring: moni-
toring water quality in the future, Vol. 1. Lelystad, the
Netherlands, Institute for Inland Water Management and
Waste Water Treatment (RIZA).
BUISHAND, T.A. and HOOGHART, J.C. (eds.). 1986.
Design aspects of hydrological networks. The Hague,
Proceedings and Information/TNO Committee on
Hydrological Research (no. 35).
CHAPMAN, D. (ed.). 1992. Water quality assessments: a
guide to the use of biota, sediments and water in environmen-
tal monitoring. London, Chapman and Hall (published on
behalf of UNESCO, WHO and UNEP).
wrm_appB.qxd 8/31/2005 12:09 PM Page 578
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix B: Monitoring and Adaptive Management 579
COFINO, W.P. 1993. Quality assurance in environmen-
tal analysis. In: D. Barcel (ed.), Environmental analysis:
techniques, applications and quality assurance. Amsterdam,
Elsevier Science, pp. 35981.
COFINO, W.P. 1994. Quality management of monitoring
programmes. In: M. Adriaanse, J. van de Kraats,
P.G. Stoks and R.C. Ward (eds.), Monitoring Tailormade.
Proceedings of the international workshop, Beekbergen,
the Netherlands, 2023 September 1994. Beekbergen,
RIZA, pp. 178187.
EUROWATERNET. 1998. Technical guidelines for imple-
mentation: The European Environment Agencys monitoring
and information. Copenhagen, Network for Inland Water
Resources (Technical Report No. 7).
GILBERT, R.O. 1987. Statistical methods for environmental
pollution monitoring. New York, Van Nostrand Reinhold.
GROOT, S. and VILLARS, M.T. 1995. Organizational
aspects of water quality monitoring: monitoring water quality in
the future, Vol. 5. Delft, the Netherlands, Delft Hydraulics.
HARMANCIOGLU, N.B.; FISTIKOGLU, O.; OZKUL, S.D.;
SINGH, V.P. and ALPASLAN, M.N. 1998. Water quality
monitoring network design. Dordrecht, the Netherlands,
Kluwer Academic.
HOFSTRA, M.A. 1994. Information is vital for the national
decision maker. In: M. Adriaanse, J. van de Kraats, P.G.
Stoks and R.C. Ward (eds.), Monitoring Tailormade.
Proceedings of the international workshop, Beekbergen,
the Netherlands, 2023 September 1994. RIZA, pp. 4354.
KING COUNTY, WASHINGTON, USA. 2004. http://dnr.
metrokc.gov/wlr/waterres/streams/streamsites.htm
(accessed15 November).
SANDERS, T.G. and LOFTIS, J.C. 1994. Factors to con-
sider in optimization of a monitoring network. In: M.
Adriaanse, J. van de Kraats, P.G. Stoks and R.C. Ward
(eds.), Monitoring Tailormade. Proceedings of the interna-
tional workshop, Beekbergen, the Netherlands, 2023
September 1994. RIZA, pp. 14652.
UN/ECE. 2000. Guidelines on monitoring and assessment
of transboundary rivers. Lelystad, the Netherlands, RIZA,
Institute for Inland Water Management and Waste
Water Treatment, UN/ECE Task Force on Monitoring
and Assessment project-secretariat.
UN/ECE. 1993. Guidelines on the ecosystems approach
in water management. In: Protection of water resources and
aquatic ecosystems. New York and Geneva, United Nations
(Water Series No. 1, ECE/ENVWA/31, United Nations
Economic Commission for Europe).
WALTERS, C. 1986. Adaptive management of renewable
resources. New York, McMillan.
WARD, R.C.; LOFTIS, J.C. and MCBRIDE, G.B. 1990.
Design of water quality monitoring systems. New York, Van
Nostrand Reinhold.
WINSEMIUS, P. 1986. Guest in own house: considerations
about environmental management. Alphen aan de Rijn,
Samson H.D. Tjeenk Willink. (In Dutch.)
wrm_appB.qxd 8/31/2005 12:09 PM Page 579
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management
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
wrm_appC.qxd 8/31/2005 12:09 PM Page 580
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Droughts are normal, recurrent, yet relatively infrequent
features of climate. Their timing and severity are unpre-
dictable. They can occur virtually anywhere that precipita-
tion occurs. Arid and semi-arid areas where precipitation is
minimal or nonexistent are indeed dry, but they are not
considered to be in a perpetual state of drought. Droughts
occur when water supplies are substantially below what is
usually experienced for that place and time. Just what is
considered substantially below is rather arbitrary, depend-
ing on the location and what features of a drought cause the
most stress or loss. For an often-cited example, a drought on
the coast of Libya would result when annual rainfall is less
than 180 mm. In Bali a drought might be considered to
occur after a period of only a week without rain! This is why
it is difficult to define droughts in a consistent way that
applies to all situations. However defined, droughts are
larger and longer-lasting water supply deficits than those
associated with the usual short-term variations of climate.
Droughts can be related to the timing of rainfall and/or
the amount or effectiveness of the rains. Other climatic
factors such as high temperature, high wind and low rel-
ative humidity can significantly aggravate their severity.
Figure C.1 shows the drought susceptible regions of the
world and the relative levels of stress that could occur
581
Drought Management
Throughout history droughts have had far-reaching effects on humans and their
civilizations. Droughts have caused crop failures and the death of natural
vegetation, livestock, wildlife and people. The World Health Organization estimates
that droughts and their effects cause half the deaths worldwide due to all natural
disasters. Economic losses from prolonged droughts often exceed those from other
more dramatic natural hazards. Humans have choices: they can increase the
adverse impacts of droughts or they can take measures to reduce them. Droughts
are not going to be preventable. Humans need to learn how better to live with and
manage them, as with all other types of natural events.
Appendix C:
should substantially less water be available than what is
considered normal.
Some areas shown as having a relative low or negligible
stress in Figure C.1 can experience droughts. Figure C.2
shows the areas of the world that experienced drought
conditions in 1982 and 1983.
2. Drought Impacts
Droughts are characterized by dry, cracked soils on river
beds and lakes, dust, and thirsty plants and animals. The
Australian Bureau of Meteorology provides some pictures
of these impacts, shown in Figure C.3.
Damage from droughts can exceed that resulting from
any other natural hazard. In the United States their annual
expected impacts are estimated to exceed $6 billion.
Drought primarily affects agriculture, transportation,
recreation and tourism, forestry and energy sectors. Social
and environmental impacts are also significant, although
it is difficult to assign a monetary value to them.
When a drought begins, the agricultural sector is usu-
ally the first to be affected because of its dependence on soil
moisture, which can be rapidly depleted during extended
dry periods. If precipitation deficiencies continue, then
users dependent on other sources of water will begin to feel
wrm_appC.qxd 8/31/2005 12:09 PM Page 581
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
the effects of the shortage. Those users who rely on surface
water (reservoirs and lakes) and groundwater, for example,
are often the last to be affected. A short-term drought that
persists for three to six months may have little impact on
these water users, depending on the characteristics of the
hydrological system and water-use requirements.
When precipitation returns to normal, the sequence is
repeated for the recovery of surface and subsurface water
supplies. Soil water reserves are replenished first, fol-
lowed by streamflow, reservoirs and lakes, and ground-
water. Drought impacts may diminish rapidly for
agriculture because of its reliance on soil moisture, but
can linger for months or even years for other users who
582 Water Resources Systems Planning and Management
are dependent on stored surface or subsurface supplies.
Groundwater users, often the last to be affected by
drought during its onset, may be the last to experience a
return to normal water levels. The length of the recovery
period is a function of the intensity of the drought, its
duration and the quantity of precipitation received as the
episode terminates.
The socio-economic impacts of a drought can follow
all the other impacts just described. These impacts occur
when the demand for an economic good exceeds supply.
In Uruguay, for example, the 198889 drought resulted
in significantly reduced hydropower production because
power plants were dependent on streamflow rather than
stress
high none
E
0
3
0
5
2
6
r
Figure C.1. Map showing the
relative stress caused by water
supply deficits should a
prolonged drought occur in any
of the drought-prone regions of
the world.
E
0
2
0
9
1
7
c
droughts:
1982
1983
1982 + 1983
Jan - Dec
Jan - Aug
Figure C.2. Map showing areas
of drought conditions in 198283
(National Drought Mitigation
Center, 2000).
wrm_appC.qxd 8/31/2005 12:09 PM Page 582
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 583
storage for power generation. Reducing hydroelectric pro-
duction required the government to use more expensive
(imported) petroleum and implement stringent energy
conservation measures to meet the nations power needs.
Estimates of the economic impacts of the 198283
droughts, perhaps the most widespread drought event in
recorded history (see Figure C.3), are given in Table C.1
(NOAA, 1994).
In North America one of the worst drought periods
was from 198789. Economic losses from that drought in
the United States exceeded $39 billion (OTA, 1993). This
damage can be compared to that caused by the most
costly flood, earthquake and tropical storm events.
The worst storm event in US history was Hurricane
Andrew. On 24 August 1992, this costliest natural disas-
ter, as it is called, hit South Florida and Louisiana. The
storm killed sixty-five people and left some 200,000
others homeless. Approximately 600,000 homes and
businesses were destroyed or severely damaged by the
winds, waves and rain. Large parts of South Floridas
E
0
3
0
5
2
6
w
Figure C.3. Signs of drought:
dry dirt, thirsty animals and dust
(courtesy of the Australian
Bureau of Meteorology.
http://www.k12.atmos.
washington.edu/~gcg/RTN/
Figures/adrought.html).
Mexico & Central America
Southern Peru & Western Bolivia
Australia
Indonesia
Philippines
Southern India, Sri Lanka
Southern Africa
Iberian Peninsula, Northern Africa
United States
location
n/a
n/a
71 dead, 8,000 homeless
340 dead
n/a
n/a
disease, starvation
n/a
agricultural crop loss
major social impacts
$
$
$
$
$
$
$
$
$
costs
million
million
billion
million
million
million
billion
million
billion
600
240
500
450
150
200
2.5
1
10-12
E
0
2
1
0
1
4
a
Table C.1. Effects of the 198283
worldwide droughts (from the New
York Times, 2 August 1983; data
updated 7 February 1996).
wrm_appC.qxd 8/31/2005 12:09 PM Page 583
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
communications and transportation infrastructure were
significantly damaged. There was loss of power and
utilities, water, sewage treatment and other essentials, in
some cases up to six months after the storm ended.
Andrew also damaged offshore oil facilities in the Gulf of
Mexico. It toppled thirteen platforms and twenty-one
satellites, bent five other platforms, and twenty-three
other satellites, damaged 104 other structures and
resulted in seven pollution incidents, two fires, and five
drilling wells blown off location. The damage caused by
Andrew in both South Florida and Louisiana totalled
some $26 billion dollars.
The costliest earthquake in US history was the Loma
Prieta earthquake. At 5 p.m. on 17 October 1989 the San
Andreas Fault system in northern California had its first
major quake since 1906. Four minutes later, as over
62,000 fans filled Candlestick Park baseball stadium
for the third game of the World Series and as the
San Francisco Bay Area evening commute moved into
its heaviest flow, a Richter magnitude 7.1 earthquake
struck. The Loma Prieta earthquake was responsible for
62 deaths, 3,757 injuries, and damage to over 18,000
homes and 2,600 businesses. About 3,000 people were
left homeless. This 20-second earthquake, centred about
sixty miles south of San Francisco, was felt as far away as
San Diego in southern California and Reno in western
Nevada. Damage and interruptions to business cost about
$10 billion, with direct property damage estimated at
$6.8 billion.
The most devastating flood in US history occurred
in the summer of 1993. All large midwestern streams
flooded, including the Mississippi, Missouri, Kansas,
Illinois, Des Moines and Wisconsin rivers. The floods dis-
placed over 70,000 people. Nearly 50,000 homes were
damaged or destroyed and fifty-two people died. Over
31,000 km
2
(12,000 square miles) of productive farmland
were rendered useless. Damage was estimated between
$15 and 20 billion.
Hurricane Andrew, the Loma Prieta earthquake, and
Mississippi flood events were sudden and dramatic.
Droughts, on the other hand, are usually neither sudden
nor dramatic; they are not given names. They can never-
theless be much more costly. As mentioned earlier, the
cost of the 198889 drought exceeded $39 billion.
Drought planning and implementing mitigation measures
can help reduce those costs.
584 Water Resources Systems Planning and Management
3. Defining Droughts
Operational definitions of droughts attempt to identify
their beginning, end and degree of severity. Knowing
exactly when a drought is beginning, so that water con-
servation measures can be implemented, is difficult since
at the time it begins, whenever that is, the droughts
severity and duration are unknown. Usually the begin-
ning date is defined when some arbitrary drought index
threshold is exceeded. Such a threshold might be, for
example, observing less than 75% of the average precipi-
tation in a two-year period. Alternatively, it could be
based on the total volume of water in reservoirs used for
water supply. The threshold or index that defines the
beginning of a drought is usually established arbitrarily,
depending on the hydrological characteristics of the basin
and the specific demands for water in the region.
Meteorologists usually consider droughts in terms of
their relative dryness and dry-period duration. What is
considered abnormally dry will differ in different
regions, since the atmospheric conditions that result in
deficiencies of precipitation vary from region to region.
Some measures of dryness identify the number of days
with precipitation less than some specified threshold.
This measure is appropriate for regions having a humid
tropical or sub-tropical climate, or a humid mid-latitude
climate characterized by year-round precipitation
regimes. Examples are locations such as Manaus, Brazil;
New Orleans, Louisiana (USA); and London, England.
Regions such as the central United States, Europe,
northeast Brazil, West Africa and northern Australia are
characterized by seasonal rainfall patterns. Extended
periods without rainfall are common in, for example,
Omaha, Nebraska (USA); Fortaleza, Cear (Brazil); and
Darwin, Northwest Territory (Australia). In these cases a
definition based on the number of days with precipita-
tion less than some specified threshold is unrealistic.
Actual precipitation departures from average amounts
on monthly, seasonal or annual time scales may be more
useful.
Farmers are mostly concerned about the susceptibility
of crops to soil moisture deficits during different stages
of crop development. Plant water demand depends on
prevailing weather conditions, the biological characteris-
tics of the specific plant, its stage of plant growth, and the
physical and biological properties of the soil. Deficient
wrm_appC.qxd 8/31/2005 12:09 PM Page 584
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 585
topsoil moisture at planting may hinder germination,
leading to low plant population densities and a reduced
final yield. However, if topsoil moisture is sufficient for
early growth requirements, deficiencies in subsoil mois-
ture at this early stage may not affect final yield, provided
subsoil moisture is replenished as the growing season
progresses or rainfall meets plant water needs.
A drought threshold for farmers might involve daily
precipitation values together with evapotranspiration
rates that are used to determine the rate of soil moisture
depletion and its effect on plant behaviour (growth and
yield) at various stages of crop development. This meas-
ure of drought severity could be tracked over time, con-
tinually re-evaluating the potential impact of drought
conditions on final yield.
Those who manage water for multiple uses at regional
scales are primarily concerned about future water sup-
plies. These hydrologists are interested in the effects of
precipitation (including snowfall) shortfalls on surface or
subsurface water supplies, as evidenced by streamflow,
reservoir and lake levels, and groundwater table eleva-
tions. The frequency and severity of such a hydrological
drought is often defined on a watershed or river basin
scale. Although all droughts originate with a deficiency of
precipitation, hydrologists focus on how this deficiency
affects the hydrological system.
Hydrological droughts usually occur some time after
precipitation deficiencies show up in soil moisture. It
takes longer for streamflows or groundwater and surface
reservoir levels to react. It takes even longer for these
effects to affect various economic sectors. For example, a
precipitation deficiency may result in a rapid depletion of
soil moisture that is almost immediately discernible to
farmers, yet the impact of this deficiency on reservoir lev-
els may not affect hydroelectric power production or
recreational uses for many months. Also, water in lakes,
reservoirs, and rivers is often used for multiple and com-
peting purposes (irrigation, recreation, navigation,
hydropower, wildlife habitat and so on), further compli-
cating the sequence and quantification of impacts.
Competition for water in these storage systems escalates
during droughts, and hence conflicts among water users
may increase as well.
Developing a history of drought frequency, severity,
duration and impacts for a region provides a greater
understanding of the regions drought characteristics and
the probability of drought recurrence at various levels of
severity. Even though these characteristics could be chang-
ing along with a changing climate, information of this type
is beneficial in the development of drought response and
mitigation strategies and preparedness plans. These plans
need to be kept current as conditions change.
Figure C.4 summarizes the sequence and conse-
quences of drought events, indicating which of these are
of primary interest to meteorologists, agriculturalists and
hydrologists.
4. Causes of Droughts
Understanding what causes droughts helps us predict
them. This section briefly outlines the relation of droughts
to normal variations in regional weather, and how
weather is related to global climate patterns.
The immediate cause of droughts is the downward
movement of air (subsidence). This causes compres-
sional warming or high pressure that inhibits cloud for-
mation and results in lower relative humidity and less
precipitation. Most climatic regions experience high-
pressure dominance, often depending on the season.
Prolonged droughts occur when large-scale high-pres-
sure anomalies in atmospheric circulation patterns per-
sist for months or seasons (or longer). The extreme
drought that affected the United States and Canada in
198889 resulted from the persistence of a large-scale
atmospheric circulation anomaly. Regions under the
influence of semi-permanent high pressure during all or
a major portion of the year are usually deserts, such as
the Sahara and Kalahari deserts of Africa and the Gobi
Desert of Asia.
The ability to predict droughts is limited and differs
by region, season and climatic regime. It is difficult to
predict a drought a month or more in advance for most
locations. Predicting droughts depends on the ability to
forecast precipitation and temperature, which are inher-
ently variable. Anomalies of precipitation and tempera-
ture may last from several months to several decades.
How long they last depends on airsea interactions, soil
moisture and land surface processes, topography, inter-
nal dynamics, and the accumulated influence of dynam-
ically unstable synoptic weather systems at the global
scale.
wrm_appC.qxd 8/31/2005 12:09 PM Page 585
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
4.1. Global Patterns
Global patterns of climatic variability tend to recur
periodically with enough frequency and with similar char-
acteristics over a sufficient length of time to offer opportu-
nities for improved long-range climate predictions. One
such pattern is called El Nio/Southern Oscillation
(ENSO). Off the western coast of South America every two
to seven years, ocean currents and winds shift. This brings
warm water westward to displace the nutrient-rich cold
water that normally wells up from deep in the ocean. The
invasion of warm water disrupts both the marine food
chain and the economies of coastal communities that
586 Water Resources Systems Planning and Management
depend on fishing and related industries. Because
the phenomenon peaks around the Christmas season, the
fishermen who first observed it named it El Nio
(the Christ Child or the little boy).
In recent decades, scientists have recognized that
El Nio is linked with other shifts in global weather pat-
terns. The upper half of Figure C.5 illustrates the El Nio
conditions compared to normal conditions.
The intensity and duration of El Nio events are var-
ied and hard to predict. They typically last anywhere from
fourteen to twenty-two months, but they can be much
shorter or longer. El Nio often begins early in the year
and peaks between the following November and January,
natural climate variability
precipitation deficiency
(amount, intensity, timing)
high temperature, high winds, low
relative humidity, greater sunshine,
less cloud cover
reduced infiltration, runoff, deep
percolation, groundwater recharge
increased evaporation and
transpiration
soil water deficiency
plant water stress, reduced
biomass and yield
reduced streamflow, inflow to
reservoirs, lakes, and ponds;
reduced wetlands, wildlife habitat
economic impacts social impacts environmental impacts
t
i
m
e

(
d
u
r
a
t
i
o
n
)
m
e
t
e
o
r
o
l
o
g
i
c
a
l
d
r
o
u
g
h
t
a
g
r
i
c
u
l
t
u
r
a
l
h
y
d
r
o
l
o
g
i
c
a
l
E
0
2
0
9
1
7
d
Figure C.4. A summary of the
sequence and consequences of
drought events, and the way
meteorologists, agriculturalists
and hydrologists view them.
wrm_appC.qxd 8/31/2005 12:09 PM Page 586
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 587
but no two events in the past have behaved in the
same way.
Often accompanying El Nio is the Southern
Oscillation, a seesaw of atmospheric pressure between
the eastern equatorial Pacific and Indo-Australian areas
(Glantz et al., 1991). The term Southern Oscillation
refers to the atmospheric component of the relationship,
shown in Figure C.5, and El Nio represents the oceanic
property in which sea surface temperatures are the main
factor.
During an El Nio/Southern Oscillation event, the
Southern Oscillation is reversed. Generally, when
pressure is high over the Pacific Ocean, it tends to be low
in the eastern Indian Ocean, and vice versa (Maunder,
1992). ENSO events are those in which both a Southern
Oscillation extreme and an El Nio occur together.
La Nia is the counterpart of El Nio and represents
the other extreme of the ENSO cycle. In this event, the sea
surface temperatures in the equatorial Pacific drop well
below normal levels and advect to the west, while the
trade winds are unusually intense rather than weak.
La Nia cold event years often (but not always) follow
El Nio warm event years. The lower half of Figure C.5
compares La Nia and normal conditions.
normal conditions
120
120
o
o
E
E
80
80
o
o
W
W
equator
convective
circulation
thermocline
El Nio conditions
normal conditions
equator
convective
circulation
thermocline
La Nia conditions
120
120
o
o
E
E
80
80
o
o
W
W
thermocline
thermocline
E
0
2
0
9
1
7
f
Figure C.5. Normal, El Nio,
and La Nia conditions in the
mid-Pacific Ocean.
wrm_appC.qxd 8/31/2005 12:09 PM Page 587
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
4.2. Teleconnections
Shifts in sea surface temperatures off the west coast of
South America are just one part of the coupled interac-
tions of atmosphere, oceans and land masses. Statistically,
El Nio/Southern Oscillation can account at most for
about half of the inter-annual rainfall variance in eastern
and southern Africa (Ogallo, 1994). Yet many of the more
extreme anomalies, such as severe droughts, flooding and
hurricanes, have strong teleconnections to ENSO events.
Teleconnections are defined as atmospheric interactions
between widely separated regions (Glantz, 1994).
Understanding these teleconnections can help in forecast-
ing droughts, floods and tropical storms (hurricanes).
During an ENSO event, drought can occur virtually
anywhere in the world, though the strongest connections
between ENSO and intense drought are in Australia,
India, Indonesia, the Philippines, Brazil, parts of eastern
and southern Africa, the western Pacific Basin islands
(including Hawaii), central America and various parts of
the United States. Drought occurs in each of the above
regions at different times (seasons) during an ENSO event
and in varying degrees of magnitude.
In the Pacific Basin, Indonesia, Fiji, Micronesia and
Hawaii are usually prone to droughts during ENSO
events. Virtually all of Australia is subjected to abnormally
dry conditions in these periods, but the eastern half has
been especially prone to extreme drought. Bush fires and
crop failures usually follow. India too has been subjected
to drought through a suppression of the summer mon-
soon season that seems to coincide with ENSO events in
many cases. Eastern and southern Africa also show a
strong correlation between ENSO events and a lack of
rainfall that brings on drought in the Horn region and
areas south of there. Central America and the Caribbean
islands are also abnormally dry during El Nio events.
ENSO events seem to have a stronger influence on
regions in the lower latitudes, especially in the equatorial
Pacific and bordering tropical areas. The relationships in
the mid-latitudes are less pronounced and less consistent
in the way El Nio influences wet or dry weather patterns.
The intensity of the anomalies in these regions is also less
consistent than those of the lower latitudes.
Incorporating some of these teleconnections within
numerical computer models of global climate can
improve their predictive capabilities. Predicting when an
588 Water Resources Systems Planning and Management
ENSO event might occur is much easier than predicting
its duration and intensity, however. Predictive models are
becoming more sophisticated and more effective in many
respects due in part to the expanded data sets that are
available for the equatorial Pacific region. Some models
have predicted ENSO events a year or more in advance.
ENSO forecasts can help countries anticipate and mit-
igate droughts and floods, and can be useful in agricul-
tural planning. Countries that are in latitudes with strong
El Nio connections to weather patterns, such as Brazil,
Australia, India, Peru and various African nations, use
predictions of near-normal conditions, weak El Nio con-
ditions, strong El Nio conditions or a La Nia condition
to help agricultural producers select the crops most likely
to be successful in the coming growing season. In coun-
tries or regions with a famine early warning system
(FEWS) in place, ENSO forecasts can play a key role in
mitigating the effects of floods or droughts that can lead
to famine. Famine, like drought, is a slow-onset disaster,
so forewarning may enable countries to substantially
reduce, if not eliminate, its worst impacts.
ENSO advisories are used to a lesser extent in planning
in North America and other extratropical countries
because the links between ENSO and weather patterns are
less clear in these areas. As prediction models improve,
the role of ENSO advisories in planning in mid-latitude
countries will undoubtedly increase.
4.3. Climate Change
Climate change is associated with the greenhouse effect.
The greenhouse effect is a naturally occurring phenome-
non necessary to sustain life on earth. In a greenhouse,
solar radiation passes through a mostly transparent piece
of glass or plastic and warms the inside air, surface and
plants. As the temperature increases inside the green-
house, the interior radiates energy back to the outside and
eventually a balance is reached.
The earth is in a greenhouse: its atmosphere is the
piece of glass or plastic. Short-wave radiation from the
sun passes through the earths atmosphere. Some of this
radiation is reflected back into space by the atmosphere,
some is absorbed by the atmosphere, and some makes it
to the earths surface. There, it is either reflected or
absorbed. The earth, meanwhile, emits long-wave radia-
tion outward. Gases within the atmosphere absorb some
wrm_appC.qxd 8/31/2005 12:09 PM Page 588
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 589
of this long-wave radiation and re-radiate it back to the
surface. It is because of this greenhouse-like function of
the atmosphere that the average global temperature of the
earth is what it is, about 15 C (59 F). Without the
atmosphere and these gases, the average global tempera-
ture would be a chilly 18 C (0 F). Life that exists
on earth today would not be possible without these
atmospheric greenhouse gases. They include carbon
dioxide (CO
2
), water vapour (H
2
O), methane (CH
4
),
nitrous oxide (N
2
O), chlorofluorocarbons (CFCs), and
ozone (O
3
).
Increasing concentrations of greenhouse gases will
increase the temperature of the earth. Currently
both are increasing. Carbon dioxide is responsible for
approximately half of the greenhouse gas increase.
Concentrations of CO
2
have been monitored at the
Mauna Loa Observatory in Hawaii since 1958.
Measurements of CO
2
concentrations from before 1958
are based on air bubbles within ice cores taken from
around the world. In the middle of the eighteenth cen-
tury, levels of CO
2
were around 270 parts per million
(ppm). By the end of the twentieth century these levels
were about 370 ppm. They are currently increasing at
about 0.4% per year (IPCC, 2001). About 7090% of the
CO
2
added to the atmosphere is due to the burning of
fossil fuels, and much of the rest is from deforestation
(OTA, 1993). Methane, N
2
O, and CFCs are increasing
at similar rates. At these rates the concentration of CO
2
will double pre-industrial concentrations by about 2100
(IPCC, 2001).
Scientists incorporate this information within large-
scale models of the atmosphere called general circula-
tion models (also called global climate models, or
GCMs). These models are designed to simulate global
atmospheric conditions and make projections of the
future climate. Although there are differences among
the GCM projections, the models are in general agree-
ment that, as a result of increasing greenhouse gas con-
centrations, the average global temperature will
increase by about 1.45.8 C (2.5210.44 F) by 2100
(IPCC, 2001). In the past century, the global average
surface temperature has increased 0.60 C (1.08 F).
This increase by itself is within the normal variability
and, although it may be a result of climate change, it
cannot be used as definitive proof that recent human
activities have contributed noticeably to global warming.
Between 1860 and 2000, however, the nine warmest
years for the global temperature occurred after 1980
(IPCC, 2001).
With the projected global temperature increase, the
global hydrological cycle could also intensify. GCMs
indicate that global precipitation could increase 715%.
Meanwhile, global evapotranspiration could increase
510% (OTA, 1993). The combined impacts of increased
temperature, precipitation, and evapotranspiration would
affect snowmelt, runoff, and soil moisture conditions. The
global climate models (GCM) generally show that precip-
itation will increase at high latitudes and decrease at low
and mid-latitudes. If this happens, evapotranspiration in
mid-continental regions will be greater than precipitation.
This could lead to more severe, longer-lasting droughts in
these areas. In addition, the increased temperatures alone
will expand ocean waters and this will lead to an esti-
mated sea level rise of about half a centimetre per year
(OTA, 1993).
One of the weaknesses of GCM climate change pre-
dictions is that they cannot adequately take into account
factors that might influence regional and river basin cli-
mates, such as the local effects of mountains, coastlines,
large lakes, vegetation boundaries and heterogeneous
soils, or the ways that human activities affect those pre-
dictions. Thus, it is difficult to transfer GCM predictions
to regional and basin scales. GCMs also cannot predict
changes in the frequency of extreme events. Some
believe the occurrence of extreme events such as
droughts is increasing. Although the losses due to these
events have increased worldwide, this is often the result
of increased vulnerability rather than an increased num-
ber of events.
No one knows who will lose or who will benefit from
a CO
2
-induced climate change. Clearly, coastal regions
will be likely to experience a slow rise in average sea
levels. But some interior regions might benefit from gains
in agricultural production resulting from the indirect
effects of a warmer climate and adequate precipitation,
especially in higher latitudes across Canada and Russia.
The increased CO
2
might also directly increase plant
growth and productivity as well. This is known as the
CO
2
fertilization effect. Laboratory experiments have
shown that increased CO
2
concentrations potentially
promote plant growth and ecosystem productivity by
increasing the rate of photosynthesis, improving nutrient
wrm_appC.qxd 8/31/2005 12:09 PM Page 589
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
uptake and use, increasing water-use efficiency, and
decreasing respiration, along with several other factors
(OTA, 1993). Increased ecosystem productivity could
draw CO
2
from the atmosphere, thereby diminishing con-
cerns about global warming (OTA, 1993). Whether any
benefits would result from the CO
2
fertilization effect
within the complex interactions of natural ecosystems is
still unknown, however.
4.4. Land Use
Although climate is a primary contributor to droughts,
other factors such as changes in land use (e.g. deforesta-
tion), land degradation and the construction of dams all
affect the hydrological characteristics of a basin. Changes in
land use upstream may reduce infiltration and thereby
increase runoff rates, resulting in more variable streamflow,
less groundwater retention and a higher incidence of
hydrological drought downstream. Bangladesh, for exam-
ple, has shown an increased frequency of water shortages
in recent years because of land use changes that have
occurred within its territory and in neighbouring countries.
Land use change is one of the ways human actions can
alter the frequency of water shortage even when no statis-
tical change in the frequency or intensity of precipitation
has been observed.
5. Drought Indices
Drought indices combine various indicators of drought
into a single index to help in identifying the onset as well
as the severity of a drought. Different indices have been
proposed and are used for different purposes. This sec-
tion identifies a few of the more commonly used indices.
No index is perfect for all situations. Hence, most water
supply planners find it useful to track more than one
index before making decisions that will affect peoples
welfare.
5.1. Percent of Normal Indices
The percent of normal indices are easily calculated and well
suited to the needs of weathercasters and the general pub-
lic. They can be associated with any hydrological variable,
such as precipitation, soil moisture, groundwater level or
590 Water Resources Systems Planning and Management
reservoir storage volume. The percent of normal index
based on precipitation is one of the simplest measure-
ments of rainfall or snowfall for a location. Such analyses
are effective when used for a single region or a single sea-
son. The index value is 100 times the actual value divided
by the average value (typically considered to be a thirty-
year mean). Normal for a specific location is considered to
be 100%.
This index can be calculated for a variety of time
scales. Usually these range from a single month to a group
of months representing a particular season, to an annual
or water year.
One potential disadvantage of using the percent of
normal value is that the mean, or average, is often not the
same as the median (the value exceeded 50% of the time
in a long-term climate record). This is because most ran-
dom hydrological values on monthly or seasonal scales
are not normally distributed. Some people assume the use
of the percent of normal comparison is based on median
rather than mean values, and this can cause confusion.
The long-term precipitation record for the month of
January in Melbourne, Australia, can serve to illustrate the
potential for misunderstanding. The median January
precipitation is 36.0 mm (1.4 in.), meaning that in half
the January months less than 36.0 mm has been recorded,
and in the other half of the January months more than
36.0 mm has been recorded. However, a monthly January
total of 36.0 mm would be only 75% of the historical
mean, which is often considered to be quite dry (Willeke
et al., 1994).
5.2. Standardized Precipitation Index
The realization that a precipitation deficit has different
impacts on groundwater, reservoir storage, soil moisture,
snow pack and streamflow led McKee et al. (1993) to
develop the Standardized Precipitation Index (SPI). This
index quantifies the precipitation deficit for multiple time
scales that reflect the impact of drought on the availabil-
ity of different sources of water. Soil moisture conditions
respond to precipitation anomalies on a relatively short
time scale. Groundwater, streamflow and reservoir stor-
age reflect the longer-term precipitation anomalies. For
these reasons, McKee et al. (1993) originally calculated
the SPI for three, six, twelve, twenty-four, and forty-eight-
month time scales.
wrm_appC.qxd 8/31/2005 12:09 PM Page 590
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 591
The SPI calculation for any location is based on the long-
term precipitation record fitted to a probability distribution.
This distribution is then transformed into a normal distri-
bution so that the mean SPI for the location and desired
period is zero (Edwards and McKee, 1997). Positive SPI
values indicate greater than median or mean precipitation,
and negative values indicate less than median or mean
precipitation. Because the SPI is normalized, wetter and
drier climates can be represented in the same way, so wet
periods can also be monitored using the SPI.
McKee et al. (1993) used the classification system
shown in Table C.2 to define drought intensities result-
ing from the SPI. They also defined the criteria for a
drought event for any of the time scales. A drought event
occurs any time the SPI is continuously negative and
reaches an intensity of 1.0 or less. The event ends
when the SPI becomes positive. Each drought event,
therefore, has a duration defined by its beginning and
end, and an intensity for each month of the drought
event. The absolute sum of the negative SPI for all the
months within a drought event can be termed the
droughts magnitude.
Using an analysis of stations across Colorado in the
western United States, McKee determined that, as defined
by the SPI, western Colorado is in mild drought 24% of the
time, in moderate drought 9.2% of the time, in severe
drought 4.4% of the time, and in extreme drought 2.3% of
the time (McKee et al., 1993). Because the SPI is
standardized, these percentages are expected from a
normal distribution of the SPI. The 2.3% of SPI values
within the extreme drought category is a percentage that is
typically expected for an extreme event (Wilhite, 1995). In
contrast, the Palmer index, to be discussed next, reaches
its extreme category more than 10% of the time across
portions of the central Great Plains. This standardization
allows the SPI to determine the rarity of a current drought,
as well as the probability of the precipitation necessary to
end the current drought (McKee et al., 1993).
5.3. Palmer Drought Severity Index
The Palmer drought severity index (PDSI) is a soil mois-
ture index designed for relatively homogeneous regions.
Many US government and state agencies rely on it to trig-
ger drought relief programmes. The index measures the
departure of the soil moisture supply from demand
(Palmer, 1965). The objective of the PDSI is to provide
standardized measurements of moisture conditions so
that comparisons using the index can be made between
locations and between months. Complete descriptions of
the equations can be found in the original study by
Palmer (1965) and in the analysis by Alley (1984).
The PDSI responds to weather conditions that have
been abnormally dry (or abnormally wet). It is calculated
with the use of precipitation and temperature data, as well
as the local available water content (AWC) of the soil. It
does not consider streamflow, lake and reservoir levels, or
other longer-term hydrological variable values that indeed
may still show a drought (Karl and Knight, 1985). Human
impacts on the water balance, such as irrigation, are also
not considered.
Palmer developed the PDSI to include the duration of
a dry or wet spell period. His motivation was as follows:
an abnormally wet month in the middle of a long-term
drought should not have a major impact on the index,
and a series of months with near-normal precipitation
following a serious drought does not mean that the
drought is over. Therefore, Palmer developed criteria for
determining when a drought or a wet spell begins and
ends. Palmer (1965) described this effort and gave exam-
ples. It is also described in detail by Alley (1984).
In near-real time, Palmers index is no longer a meteo-
rological index but becomes a hydrological one, referred
to as the Palmer hydrological drought index (PHDI). It is
based on moisture inflow (precipitation), outflow and
storage, and does not take into account the long-term
trend (Karl and Knight, 1985).
E
0
2
1
0
1
4
b
2.0+
1.5 to 1.99
1.0 to 1.49
SPI values
extremely wet
very wet
moderately wet
near normal
moderately dry
severely dry
extremely dry
0.99 -0.99
-1.00
-1.50
-2.00
to
to
to
and less
-1.49
-1.99
Table C.2. Standardized precipitation index.
wrm_appC.qxd 8/31/2005 12:09 PM Page 591
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A modified method to compute the PDSI was pro-
posed by Heddinghaus and Sabol, 1991. This modified
PDSI differs from the PDSI during transition periods
between dry and wet spells. Because of the similarities
between these Palmer indices, the terms Palmer index and
Palmer drought index have been used to describe general
characteristics of the indices.
The Palmer index varies between roughly 6.0 and
6.0. Palmer arbitrarily selected the classification scale of
moisture conditions, shown in Table C.3, on the basis of
his original study areas in the central United States
(Palmer, 1965). Ideally, the Palmer index is designed so
that a 4.0 in one region has the same meaning in terms
of the moisture departure from a climatological normal as
a 4.0 in another region (Alley, 1984). The Palmer index
has typically been calculated on a monthly basis, and a
long-term archive (from 1895 to present) of the monthly
PDSI values for every climate division in the United States
exists at the National Climatic Data Center. In addition,
weekly Palmer index values (modified PDSI values) are
calculated and mapped (Figure C.6) for the climate divi-
sions during every growing season.
The Palmer index has been widely used for a variety of
applications across the United States. It is most effective
at measuring impacts sensitive to soil moisture condi-
tions, such as agriculture (Willeke et al., 1994). It has also
been useful as a drought-monitoring tool and has been
used to trigger actions associated with drought contin-
gency plans (Willeke et al., 1994). Alley (1984) identified
592 Water Resources Systems Planning and Management
three positive characteristics of the Palmer index that con-
tribute to its popularity: it provides decision-makers with
a measurement of the abnormality of recent weather for a
region; it provides an opportunity to place current condi-
tions in historical perspective; and it provides spatial and
temporal representations of historical droughts. Several
states in the USA, including New York, Colorado, Idaho
and Utah, use the Palmer index as part of their drought
monitoring systems.
The Palmer index has some limitations, and these are
described in detail by Alley (1984) and Karl and Knight
(1985). Also, while the index has been applied within the
United States it has little acceptance elsewhere (Kogan,
1995; Smith et al., 1993). It does not do well in regions
where there are extremes in the variability of rainfall or
runoff, such as Australia and South Africa. Another weak-
ness in the Palmer index is that the extreme and severe
classifications of drought occur with a greater frequency
in some regions than in others (Willeke et al., 1994). This
limits the accuracy of comparisons of the intensity of
droughts between two regions and makes planning
response actions based on a given intensity more difficult.
5.4. Crop Moisture Index
Whereas the PDSI monitors long-term meteorological wet
and dry spells, the crop moisture index (CMI) was
designed to evaluate short-term moisture conditions
across major crop-producing regions. It is not intended to
assess long-term droughts. It reflects moisture supply in
the short term across major crop-producing regions.
The CMI uses a meteorological approach to monitor
week-to-week crop conditions. It is based on the mean
temperature and total precipitation for each week within
a climate division, as well as the CMI value from the
previous week. The index responds rapidly to changing
conditions, and is weighted by location and time so that
maps, which commonly display the weekly CMI across a
region, can be used to compare moisture conditions at
different locations.
5.5. Surface Water Supply Index
The surface water supply index (SWSI, pronounced
swa zee) is designed to complement the Palmer where
mountain snow pack is a key element of water supply. It
E
0
2
1
0
1
4
c
Palmer classifications
-0.5
-1.9
-2.0
-3.0
-4.0
4.0
3.0
2.0
1.0
0.5
0.49
or more
to
to
to
to
to
3.99
2.99
1.99
0.99
-0.49
-0.99
-1.99
-2.99
-3.99
to
to
to
to
or less
extremely wet
very wet
moderately wet
slightly wet
incipient wet spell
near normal
incipient dry spell
mild drought
moderate drought
severe drought
extreme drought
Table C.3. The Palmer drought severity index.
wrm_appC.qxd 8/31/2005 12:09 PM Page 592
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 593
is calculated for river basins on the basis of snow pack,
streamflow, precipitation and reservoir storage (Shafer and
Dezman, 1982; Dezman et al., 1982). The index is unique
to each basin, which limits interbasin comparisons.
Shafer and Dezman (1982) designed the SWSI to be an
indicator of the conditions of surface water in which
mountain snow pack is a major component. The objective
of the SWSI was to incorporate both hydrological and
climatological features into a single index value resem-
bling the Palmer index for each major river basin in the
state of Colorado. These values could be standardized to
allow comparisons between basins. Four inputs are
required within the SWSI: snow pack, streamflow, pre-
cipitation and reservoir storage. Because it is dependent
on the season, the SWSI is computed with only snow
pack, precipitation and reservoir storage in the winter.
During the summer months, streamflow replaces snow
pack as a component within the SWSI equation.
To determine the SWSI for a particular basin, monthly
data are collected and summed for all the precipitation
stations, reservoirs and snow pack/streamflow measuring
stations over the basin. Each summed component is
normalized using a frequency analysis gathered from a
long-term data set. The probability of non-exceedance
(the probability that subsequent sums of that component
will exceed the current sum) is determined for each com-
ponent on the basis of the frequency analysis. This allows
comparisons of the probabilities to be made between the
components. Each component has a weight assigned to it
depending on its typical contribution to the surface water
within that basin. These weighted components are
summed to determine a SWSI value representing the
entire basin. Like the Palmer index, the SWSI is centred
on zero and has a range between 4.2 and 4.2.
5.6. Reclamation Drought Index
Like the SWSI, the reclamation drought index (RDI) is
calculated at the river basin level, incorporating tempera-
ture as well as precipitation, snow pack, streamflow and
reservoir levels as input. By including a temperature com-
ponent, it also accounts for evaporation. It is used by the
US Bureau of Reclamation as a trigger to release drought
emergency relief funds. The RDI classifications are listed
in Table C.4.
The RDI was developed as a tool for defining drought
severity and duration, and for predicting the onset and
end of periods of drought. The impetus to devise it came
extreme drought extremely moist near normal
0 -1 +1 -2 -3 -4 -5 +2 +3 +4 +5
E
0
2
0
9
1
7
e
Figure C.6. Palmer index
values for the United States
during a week in August,
2000.
wrm_appC.qxd 8/31/2005 12:09 PM Page 593
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
from the US Reclamation States Drought Assistance Act of
1988, which allows states to seek assistance from the US
Bureau of Reclamation to mitigate the effects of drought.
The RDI differs from the SWSI in that it builds a tem-
perature-based demand component and duration into the
index. The RDIs main strength is its ability to account for
both climate and water supply factors.
5.7. Deciles
This index groups monthly precipitation occurrences into
deciles so that, by definition, much lower than normal
weather cannot occur more often than 20% of the time.
These deciles are shown in Table C.5.
Arranging monthly precipitation data into deciles is
another drought-monitoring technique. Developed by
Gibbs and Maher (1967), it removes some of the weak-
nesses in the percent of normal approach. The technique
divides the distribution of occurrences over a long-term
precipitation record into tenths of the distribution. Each
of these categories is a decile. The first decile is the rain-
fall amount not exceeded by the lowest 10% of the pre-
cipitation occurrences. The second decile is the
precipitation amount not exceeded by the lowest 20% of
occurrences. These deciles continue until the rainfall
amount identified by the tenth decile is the largest
precipitation amount within the long-term record. By def-
inition, the fifth decile is the median; it is the precipitation
amount not exceeded by 50% of the occurrences over the
period of record. The deciles are grouped into five classi-
fications, as shown in Table C.5.
The decile method was selected as the meteorological
measurement of drought within the Australian Drought
Watch System because it is relatively simple to calculate and
594 Water Resources Systems Planning and Management
requires fewer data and assumptions than the Palmer
Drought Severity Index (Smith et al., 1993). In this system,
farmers and ranchers can only request government assis-
tance if the drought is shown to be an event that occurs only
once in twenty or twenty-five years (deciles 1 and 2 over a
100-year record) and has lasted longer than twelve months
(White and OMeagher, 1995). This uniformity in drought
classifications, unlike a system based on the percent of nor-
mal precipitation, has assisted Australian authorities in
determining appropriate drought mitigation responses.
5.8. Method of Truncation
The method of truncation uses historic records of stream-
flow, precipitation, groundwater drawdown, lake eleva-
tion and temperature (Chang and Kleopa, 1991). The
historic data are sorted in ascending order, and
trigger levels are determined from the truncation level
specified (Figure C.7).
For example, if stage 1 drought is defined as the 70%
level, this corresponds to streamflows that are less than
70% of all flows. When using this method, drought events
of higher severity are nested inside drought events of
lower severity: that is to say, a 90% drought implies the
occurrence of a 70% and 80% drought (if those are the
trigger levels selected). The highest level of severity deter-
mines the decisions to make, as a water management
system must deal with the most severe shortage that occurs.
5.9. Water Availability Index
The water availability index (WAI) relates current water
availability to historical availability during periods of
drought by measuring the deviation-from-normal rainfall
E
0
2
1
0
1
4
d
RDI classifications
or more
to
to
to
to
4.0
1.5
1.5
extremely wet
moderately wet
normal to mild wetness
normal to mild drought
moderate drought
extreme drought
4.0
1.5
1
0
-1.5
-4.0
-4.0
or less
Table C.4. RDI classifications.
lowest 20 %
lowest next 20 %
middle 20 %
highest next 20 %
highest 20 %
E
0
3
0
5
2
6
o
2:
4:
6:
8:
10:
1
3
5
7
9
-
-
-
-
-
deciles
deciles
deciles
deciles
deciles
much below normal
below normal
near normal
above normal
much above normal
decile classifications
Table C.5. Decile classifications.
wrm_appC.qxd 8/31/2005 12:09 PM Page 594
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 595
over the prior four months (Davis and Holler, 1987). The
WAI is multiplied by a constant to make the index fall
between 0 and 10, with zero representing normal condi-
tions and ten a severe drought. Hoke (1989) included a
factor for minimum and maximum desirable reservoir
pool elevation, and Raney (1991) included volume and
drainage area factors to account for multiple reservoirs
within one watershed.
5.10. Days of Supply Remaining
The days of supply remaining (DSR) index includes
current reservoir storage, forecast future inflows from
precipitation and snowmelt, and predicted demands
for municipal supply and instream flows (Fisher and
Palmer, 1997). It provides an objective and easily
understood measure of the supply of water. As new
sources of supply or new demands occur, the index
remains appropriate.
DSR is calculated by predicting future inflows and
demands and determining when the inflows and existing
supply will no longer be adequate to meet demands.
A simplified example shown in Table C.6 illustrates this
calculation.
The DSR values are calculated at the beginning of each
time step, for example, a week. At that time, forecasts are
made of the subsequent weeks inflows and demands.
Inflows are added and demands subtracted for each suc-
cessive week until the remaining supply is less than the
demand.
In the example shown in Table C.6, there are 100
units of storage currently available. Forecasts show that
the projected demand cannot be met after six weeks from
now. There is adequate supply for the current week plus
the following five weeks. Thus, only forty-two days of
supply remains. A water management agency presented
with the data in this example could implement restric-
tions to try to reduce demand by 5%. If successful that
would maintain adequate levels of storage until the
expected reservoir refill begins in seven weeks.
Alternatively, a new supply source could be sought, or
other demand and supply management options could be
analysed. Determining just what actions one should take
(given the uncertainty of supply and demand forecasts)
can be based, in part, on such analyses, including simula-
tions of alternative actions given supply and demand
situations experienced in the past.
Palmer et al. (2002) describe doing this for the City of
Atlanta in southeastern United States. Determining in
advance what actions are appropriate under what specific
conditions or values of DSR and getting them accepted
makes drought management substantially easier when
under the stress of an actual drought.
The particular values of DSR or any other index value
that trigger management decisions are called drought
triggers.
drought events
truncation level
time
i
n
d
i
c
a
t
o
r

s
u
c
h

a
s

p
r
e
c
i
p
i
t
a
t
i
o
n
,
s
t
r
e
a
m
f
l
o
w
,

r
e
s
e
r
v
o
i
r

l
e
v
e
l
s
,

e
t
c
.
E
0
3
0
5
2
6
p
Figure C.7. Method of truncation to define drought conditions
(Fisher and Palmer, 1997).
E
0
3
0
5
2
6
q
week
begin
+ 1
+ 2
+ 3
+ 4
+ 5
+ 6
+ 7
100
100
120
100
80
40
20
-10
40
60
20
20
0
20
10
60
40
40
40
40
40
40
40
40
100
120
100
80
40
20
-10
10
no
7
14
21
28
35
42
b
e
g
i
n
n
i
n
g

s
t
o
r
a
g
e
v
o
l
u
m
e
d
e
m
a
n
d

m
e
t
?
e
n
d
i
n
g

s
t
o
r
a
g
e
v
o
l
u
m
e
d
e
m
a
n
d
v
o
l
u
m
e

p
e
r

w
e
e
k
i
n
f
l
o
w
v
o
l
u
m
e

p
e
r

w
e
e
k
D
S
R
d
a
y
s
Table C.6. Calculation of DSR index value based on weekly
time steps and forecast inflow and demands (from Fisher and
Palmer, 1997).
wrm_appC.qxd 8/31/2005 12:09 PM Page 595
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
6. Drought Triggers
Drought management is tricky. Without knowing just
how severe a drought may become, it is difficult to know
what actions are best to take, if any, when a drought is
likely to have begun. Water managers must take appro-
priate action to reduce future losses and not increase them
by imposing unnecessary restrictions when they are not
needed (Wright et al., 1986). Taking actions that impose
hardships on the public and the water utility when it later
turns out they were unnecessary is not likely to engender
compliance with water saving actions the next time they
are ordered. One way to make drought management eas-
ier is to establish a sequence of increasingly strict conser-
vation measures based on a sequence of drought triggers,
and seek the publics approval.
A drought trigger is the specific value of a drought
indicator that activates a management response. For
example, a drought trigger could be a reservoir decreasing
below 50% of its active storage capacity. In a drought
contingency plan, trigger levels can be varied to alter the
sensitivity of the response and the effectiveness of the
plan. Defining drought triggers can be difficult. Trigger
levels change over time: that is, an appropriate trigger
level for a particular system may change dramatically if
that systems infrastructure is expanded or if water
demands change dramatically. Urban water triggers are
often quite different from agricultural drought triggers, as
the urban infrastructure can often mitigate the impacts of
short-term droughts.
Short-term responses may include the initiation of
outdoor water use bans, an increase in the price of water
or the use of printed media to inform the public of water
supply problems. Drought management plans for many
urban areas are often developed with four to five levels
of responses, all of which encourage different levels of
demand reduction or supply augmentation. The appro-
priateness or effectiveness of particular drought meas-
ures is dependent upon the type of user. An outdoor
water ban, for instance, may be effective for a residential
community but not for a heavy industrial complex.
Specific values of any of the drought indices just
reviewed in the previous section could be used for a
drought trigger. Since each has its limitations, triggers
based on multiple indices appropriate for the region will
probably be the most reliable and robust (Alley, 1985;
596 Water Resources Systems Planning and Management
Dracup et al., 1980; Johnson and Kohne, 1993; Mather,
1985; Titlow, 1987).
Numerous utilities and government organizations in
many countries have explored the use of triggers. Those
using triggers to support decision-making in the United
States include the Massachusetts Water Resources Authority
(which serves the Boston area), Northern New Jersey, the
Interstate Commission on the Potomac River Basin (which
facilitates supply operations in Washington, D.C., northern
Virginia and southern Maryland), the Bureau of Water
Works for the City of Portland, and the Seattle Water
Department and Denver Water (Fisher and Palmer, 1997;
Hrezo et al., 1986; Nault et al., 1990; USACE, 1991).
Deciding on what the trigger indices should be, what
the trigger values of those indices should be, and what
decisions to make when those trigger index values have
been reached is not a trivial exercise. Performing simula-
tions of alternative actions given supply and demand situ-
ations experienced in the past can help identify the
possible range of likely impacts associated with any partic-
ular set of trigger values and associated decisions (Palmer
et al., 2002). Performing these analyses in a real-time vir-
tual drought simulation exercise involving all relevant
agencies and decision-makers may be even more effective.
7. Virtual Drought Exercises
Drought management requires more than just a plan. Its
implementation needs public support and financial
resources. The plan needs to be kept up to date. People
who are going to be implementing it need to be ready
to do that when a drought comes. Virtual drought exer-
cises are one way to accomplish and maintain a state of
readiness.
A virtual drought is a fire drill, a test during which
water managers, public health officials and other stake-
holders put a drought management plan into practice
using a hypothetical, computer-generated drought event.
A good virtual drought exercise will raise important issues
and provide experience that can be applied in real
droughts. It will also keep those from different agencies
who must work closely together during times of droughts
in closer contact with each other.
For over twenty years, the Interstate Commission on
the Potomac River Basin (ICPRB) has been conducting
wrm_appC.qxd 8/31/2005 12:09 PM Page 596
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 597
annual virtual drought exercises for the Washington
metropolitan area. During a drought, the Commission
coordinates water supply operations of three major
water utilities in Washington, D.C., and the adjacent
suburbs in the states of Maryland and Virginia. In the
drought exercises, the Commission and the utilities
practise the operations of the system as they would
during an actual drought. Box C.1 gives a history of how
this began.
Virtual drought exercises typically have three stages:
briefing, gaming and debriefing. During the briefing,
the objectives of the exercise are stated, the rules of the
exercise are described, and the initial conditions of
the simulated drought scenario are introduced. During the
gaming, participants discuss management strategies and
negotiate decisions to respond to the simulated drought
conditions as they progress. Finally, during debriefing, the
outcomes of the exercise are evaluated and drought man-
agement plans are modified if and as appropriate.
The briefing portion of the exercise provides a founda-
tion for the day, establishing a common understanding
of the objectives and how the gaming will take place.
Briefing materials are distributed in advance of the
exercise to give participants time to clarify questions or
misunderstandings. Typically the briefing will:
define the objectives of the exercise
introduce participants
review the roles of participants
clarify the format of the exercise
specify the rules for decision-making
explain the background information
demonstrate the virtual drought process.
During the gaming portion of the exercise, the participants
manage the water resources system during a drought
scenario. This requires discussion among the participants
and joint decision-making. Guidelines are provided to
define options and modes of decision-making, but the nego-
tiation and acceptance of decisions is one of the by-products
of this drought rehearsal. Typical responses include water
use restrictions, modifying instream flow requirements,
revising water pricing, emergency source acquisitions,
Box C.1. A Recent History of the Potomac River
In 1963, the US Army Corps of Engineers conducted a comprehensive study of the Potomac River basin to identify
solutions to the anticipated demand deficits during low river flows. They proposed the construction of sixteen
large multipurpose reservoirs in the Potomac River basin. This proposal was not accepted by the public, nor by
Congress who would pay the bill. Flow shortages in which unrestricted water use would cause the river to run dry,
coupled with increasing population in the 1960s and 1970s, underlined the need to continue to look for solutions.
One of the sixteen proposed reservoir projects was built: Jennings Randolph Lake (originally called Bloomington
Lake).
Other structural solutions were examined. Interbasin transfers were analysed. A pilot treatment plant and an
emergency pumping station were constructed.
Concurrently, an academic study was being conducted that combined distribution areas of the three major
Washington metropolitan area utilities into a single regional demand centre. Through simulation modelling, the
study investigated the benefits derived from the coordinated operation of all the resources then available to all
three utilities. The study showed that coordinated management of the water resources from a total integrated
systems perspective led to gains in the reliability of the water resource. The results of the latter analysis and its
lower cost non-structural features led to the adoption of its results with the signing of the Water Supply
Coordination Agreement in 1982.
The management objectives embodied in the agreement are to keep the reservoir storage volumes balanced
while meeting environmental requirements and municipal demands for water. The coordinated operations of each
utility are expected to meet the expected demands without imposing restrictions through the year 2015 even
under a repeat of the drought of record. This is possible because of synergistic gains in total yield realized under
the cooperative management strategies. Each utility gives up a small measure of autonomy in order to gain the
substantial benefits of reduced capital costs through coordinated cooperative operations of its individually and
jointly owned resources. A committee of the water utilities and the Interstate Commission on the Potomac River
Basin oversee the coordinated operations.
wrm_appC.qxd 8/31/2005 12:09 PM Page 597
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
modifying surface and groundwater system operations,
public involvement and regional coordination measures.
During the gaming, the drought event unfolds over
time. The participants intervene with actions when neces-
sary to manage the drought. Information to support
and/or influence decision-making is provided throughout
the exercise. Such information can include the existing
drought plan, forecasts of future conditions, system status
reports, results of technical analyses, media news briefs
and legal actions.
Debriefing is the reflective and evaluative portion of
the exercise. The participants actions and perceptions are
discussed and translated into lessons learned. Specific
insights on how the participants managed the drought are
brought out and recommendations are made. A facilitated
group discussion is the best form of debriefing, providing
the participants an opportunity for self-observation.
There are clear links between the general drought
planning and preparedness process and virtual drought
exercises. Virtual drought exercises test a groups abilities
to resolve conflicts and make decisions; they allow partic-
ipants to experience a drought event and practise its man-
agement without the dangers of a real event, and they
provide an effective means of evaluating a drought plans
likely effectiveness.
The more realistic the virtual drought exercise is, the
more beneficial it will be. It is important that those indi-
viduals who are responsible for drought management par-
ticipate, that the participants represent a wide range of
perspectives, that the objectives of the exercise are clear to
all, and that the exercise be made as realistic as possible.
This makes it possible to capture a true sense of how the
resource is likely to be managed during the stress of an
actual drought.
Virtual drought exercises should be viewed as natural
complements to traditional drought planning activities. If
properly designed and implemented, these exercises can
keep plans up to date, stakeholders informed and water
managers skills current.
8. Conclusion
Droughts are unusually severe water shortages. We can-
not predict when they will occur or how long they will
last, but they will happen and when they do the economic
as well as social costs can be substantial. Being prepared
598 Water Resources Systems Planning and Management
for droughts can help to mitigate some of those costs and
damages. It costs money to plan, but planning to be pre-
pared for droughts in advance of when they occur saves
much more that it costs.
Societys vulnerability to droughts is affected by popu-
lation density and growth (especially in urban regions),
changes in water use trends, government policy, social
behaviour, economic conditions, and environmental and
ecological objectives. Changes in all of these factors tend
to increase the demand for water, and hence societys vul-
nerability to droughts.
Although droughts are natural hazards, society can
reduce its vulnerability and therefore lessen the risks
associated with drought events. The impacts of droughts,
like those of other natural hazards, can be reduced through
planning and preparedness. Drought management is
decision-making under uncertainty. It is risk management.
Planning ahead to identify effective ways of mitigating
drought losses gives decision-makers the chance to reduce
both suffering and expense. Reacting without a plan to
emergencies in a crisis mode during an actual drought
generally decreases self-reliance and increases dependence
on government services and donors.
Planning for droughts is essential, but it may not come
easily. There are many constraints on planning. For
example, it is hard for politicians and the public to be
concerned about a drought when they are coping with a
flood or any other more immediate crisis. Unless there
is a drought emergency, it is often hard to get support for
drought planning. There are always more urgent needs
for money and peoples attention. Where coordination
among multiple agencies can yield real benefits, it is not
easy to get it to happen until it becomes obviously
essential, for example during a severe drought. Multi-
agency cooperation and coordination must be planned
for, and perhaps practised in virtual drought management
exercises, in advance of the drought. Getting multiple
agencies to work together only in a crisis mode is
never efficient. Crisis-oriented drought response efforts
have been largely ineffective, poorly coordinated,
untimely, and inefficient in terms of the resources
allocated.
Drought planning will vary from one city or region to
another, simply because resources, institutions and
populations differ. Although drought contingency plans
may vary in detail, they all should specify a sequence
of increasingly stringent steps to either augment supplies
wrm_appC.qxd 8/31/2005 12:09 PM Page 598
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 599
or reduce demand as the drought becomes more
severe that is, as the water shortage increases. This
should happen in such a way as to minimize the adverse
impacts of water shortages on public health, consumer
activities, recreation, economic activity and the environ-
ment in the most cost-effective manner possible.
Drought plans provide a consistent framework to pre-
pare for and respond to drought events. The plans should
include drought indicators, triggers, and responses. They
should also include provisions for forecasting drought
conditions, monitoring and enforcement (Werick and
Whipple, 1994). Drought plans should consider a wide
range of issues and be compatible with the political and
social environments that can affect just what measures
can be implemented.
Developing a drought plan and keeping it current is a
continuing process that should include an informed pub-
lic. Drought plans should also include measures to edu-
cate the public and keep them aware of the potential risks
of droughts, and the measures that will be implemented
to mitigate those risks. A comprehensive public informa-
tion programme should be implemented to achieve pub-
lic acceptance of and compliance with the plan. At the
same time, enforcement measures are often necessary to
encourage the public to abide by the water-use restric-
tions. Enforcement measures traditionally include penal-
ties for noncompliance. They can also include economic
incentives such as rebates on low flow showerheads and
faucets, and cheaper water rate charges for lower con-
sumption rates.
9. References
ALLEY, W.M. 1984. The Palmer drought severity index:
limitations and assumptions. Journal of Climate and
Applied Meteorology, No. 23, pp. 11009.
ALLEY, W.M. 1985. The Palmer drought severity index
as a measure of hydrological drought. Water Resources
Bulletin, Vol. 21, No. 1, pp. 10514.
CHANG, T.J. and KLEOPA, X.A. 1991. A proposed
method for drought monitoring. Water Resources Bulletin,
Vol. 27, No. 2, pp. 27581.
DAVIS, P.C. and HOLLER, A.G. 1987. Southeastern
drought of 1986: lessons learned. In: A.D. Feldman (ed.),
Proceedings of the Engineering Hydrology Symposium.
Williamsburg, Va., 37 August 1987. New York, ASCE,
pp. 61621.
DEZMAN, L.E.; SHAFER, B.A.; SIMPSON, H.D. and
DANIELSON, J.A. 1982. Development of a surface water
supply index: a drought severity indicator for Colorado,
In: A.I. Johnson and R.A. Clark (eds.). Proceedings of the
International Symposium on Hydrometeorology, 1317 June
1982, Denver, Colorado. Bethesda, Md., American Water
Resources Association, pp. 33741.
DRACUP, J.A.; LEE, K.S. and PAULSON, E.G. 1980.
On the definition of droughts. Water Resources Research,
Vol. 16, No. 2, pp. 297302.
EDWARDS, D.C. and MCKEE, T.B. 1997. Characteristics
of twentieth century drought in the United States at multiple
time scales. Fort Collins, Colo., Colorado State University,
Department of Atmospheric Sciences (Climatology Report
Number 972).
FISHER, S.M. and PALMER, R.N. 1995. Managing water
supplies during drought: the search for triggers. In:
M.F. Domenica (ed.), Proceedings of the 22nd Annual
National Conference, 711 May 1995, Cambridge, Mass.
Cambridge, Mass., Water Resources Planning and
Management Division of ASCE, pp. 10014.
GIBBS, W.J. and MAHER, J.V. 1967. Rainfall deciles as
drought indicators. Bureau of Meteorology Bulletin, No. 48.
Melbourne, Commonwealth of Australia.
GLANTZ, M. (ed.). 1994. Usable science: food security,
early warning, and El Nio. Proceedings of the Workshop
on ENSO/FEWS, Budapest, Hungary, October 1993.
Nairobi and Boulder, Colo., UNEP and NCAR.
GLANTZ, M.; KATZ, R. and NICHOLLS, N. (eds.). 1991.
Teleconnections linking worldwide climate anomalies.
Cambridge, Cambridge University Press.
HEDDINGHAUS, T.R. and SABOL, P. 1991. A review of
the Palmer drought severity index and where do we go
from here? In: Proceedings of the Seventh Conference on
Applied Climatology, 1013 September 1991, Salt Lake
City, Utah. Boston, Mass., American Meteorological
Society, pp. 2426.
HOKE, J.P. Jr. 1989. Savannah River Basin drought con-
tingency planning. In: Proceedings of the ASCE National
Water Conference, pp. 18895. Newark, Del., 1720 July
1989. New York, ASCE.
wrm_appC.qxd 8/31/2005 12:09 PM Page 599
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
HREZO, M.S.; BRIDGEMAN, P.G. and WALKER, W.R.
1986. Managing droughts through triggering mecha-
nisms, Journal AWWA, Vol. 78, No. 6, pp. 4651.
IPCC (Intergovernmental Panel on Climate Change).
2001. Climate change 2001: the scientific basis. In: J.T.
Houghton, Y. Ding, D.J. Griggs, M. Noguer, P.J. van der
Linden, X. Dai, K. Maskell and C.A. Johnson (eds.).
Contribution of Working Group I to the Third Assessment
Report of the Intergovernmental Panel on Climate Change.
Cambridge, Cambridge University Press.
JOHNSON, W.K. and KOHNE, R.W. 1993. Susceptibility
of reservoirs to drought using Palmer index. Journal of
Water Resources Planning and Management, Vol. 119,
No. 3, pp. 36787.
KARL, T.R. and KNIGHT, R.W. 1985. Atlas of monthly
Palmer hydrological drought indices (19311983) for the
contiguous United States. Asheville, N.C., National
Climatic Data Center (Historical Climatology Series 37).
KOGAN, F.N. 1995. Droughts of the late 1980s in the
United States as derived from NOAA polar-orbiting satel-
lite data. Bulletin of the American Meteorological Society,
Vol. 76, No. 5, pp. 65568.
MATHER, J.R. 1985. Drought indices for water managers.
Publications in Climatology, Vol. 38, No. 1, pp. 177.
MAUNDER, W.J. 1992. Dictionary of global climate change.
New York, Chapman and Hall.
MCKEE, T.B. and DOESKEN, N.J.; KLEIST, J. 1993. The
relationship of drought frequency and duration to time
scales. In: T.B. McKee, N.J. Doesken and J. Kleist (eds.).
Preprints, 8th Conference on Applied Climatology,
pp. 17984. 1722 January 1993, Anaheim, Calif.
Boston, Mass., American Meterological Society.
NAULT, R.J.; GAEWSKI, P. and WALL, D.J. 1990.
Drought indication and response: optimizing the resources for
water management. In: R.M. Khanbilvardi and T.C. Gooch
(eds.). Proceedings of the 17th Annual National Conference,
1721 April 1990, Fort Worth, Texas. New York, ASCE,
pp. 72025.
NOAA. 1994. El Nio and climate prediction: reports to
the nation on our changing planet. A publication of
the University Corporation for Atmospheric Research
pursuant to National Oceanic and Atmospheric
600 Water Resources Systems Planning and Management
Administration Award No. NA27GP023201. Boulder,
Colorado, UCAR.
OGALLO, L.A. 1994. Validity of the ENSO-related
impacts in Eastern and Southern Africa. In M. Glantz
(ed.). Usable science: food security, early warning, and El
Nio, Proceedings of the Workshop on ENSO/FEWS,
Budapest, Hungary, October 1993. Nairobi and Boulder,
Colo., UNEP and NCAR, pp. 17984.
OTA (Office of Technology Assessment). 1993. Preparing
for an uncertain climate, Vol. I. OTA-O-567. Washington,
D.C., US Government Printing Office.
PALMER, R.N.; KUTZING, S.L. and STEINEMANN, A.C.
(eds.). 2002. Developing drought triggers and drought
responses: an application in Georgia. Proceedings of the
EWRI Symposium in Water Resources Planning and
Management, 1922 May, Roanoke, Va.
PALMER, W.C. 1965. Meteorological drought. Washington,
D.C., US Department of Commerce Weather Bureau
(Research Paper No. 45).
RANEY, D.C. 1991. A water availability index for reser-
voir management. In: P.A. Krenkel (ed.), Proceedings of the
Environmental Engineering Conference, 1012 June 1991,
Reno, Nev. New York, ASCE, pp. 68691.
SHAFER, B.A. and DEZMAN, L.E. 1982. Development of
a surface water supply index (SWSI) to assess the severity
of drought conditions in snowpack runoff areas. In:
Proceedings of the 1982 Western Snow Conference, 1982,
Reno, Nev. Fort Collins, Colo., Colorado State University,
pp. 16475.
SMITH, D.I. HUTCHINSON, M.F. and MCARTHUR, R.J.
1993. Australian climatic and agricultural drought:
payments and policy. Drought Network News, Vol. 5,
No. 3, pp. 112.
TITLOW, J.K. III. 1987. A precipitation-based drought
index for the Delaware River Basin, Publications in
Climatology, Vol. 40, No. 2, p. 68.
USACE (US Army Corps of Engineers). 1991. The national
study of water management during drought: a research assess-
ment. Alexandria, Va., Institute for Water Resources (IWR
Report 91-NDS-3).
WERICK, W.J. and WHIPPLE, W. Jr. 1994. National
study of water management during drought: managing water
for drought. Alexandria, Va., USACE, Water Resources
wrm_appC.qxd 8/31/2005 12:09 PM Page 600
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix C: Drought Management 601
Support Center, Institute for Water Resources (IWR
Report 94-NDS-8).
WHITE, D.H. and OMEAGHER, B. 1995. Coping with
exceptional droughts in Australia. Drought Network News,
Vol. 7, No. 2, pp. 1317.
WILHITE, D.A. 1995. Developing a precipitation-based
index to assess climatic conditions across Nebraska.
Lincoln, Nebr., final report submitted to the Natural
Resources Commission.
WILLEKE, G.; HOSKING, J.R.M.; WALLIS, J.R. and
GUTTMAN, N.B. 1994. The national drought atlas.
Alexandria, Va., USACE (Institute for Water Resources
Report 94-NDS-4).
WRIGHT, J.R.; HOUCK, M.H.; DIAMOND, J.T. and
RANDALL, D. 1986. Drought contingency planning. Civil
Engineering Systems. Vol. 3, No. 4, pp. 21021.
Additional References (Further Reading)
BLAIKIE, P.; CANNON, T.; DAVIS, I. and WISNER, B.
1994. At risk: natural hazards, peoples vulnerability, and
disasters. London, Routledge.
DAVIS, C.P.; WATSON, R.M. III and HOLLER, A.G.
1989. Southeastern drought of 1988: a federal perspec-
tive. In: T.A. Austin (ed.). Proceedings of the National Water
Conference, 1720 July 1989, Newark, Del. New York,
ASCE, pp. 1729.
FISHER, S. and PALMER, R.N. 1997. Managing water
supplies during drought: triggers for operational
responses. Water Resources Update, Vol. 3, No. 108,
pp. 1431.
HAYES, M. 1998. Drought indices. Lincoln, Nebr.,
National Drought Mitigation Center. www.drought.unl.
edu/whatis/indices.htm (accessed 15 November 2004).
KNUTSON, C.; HAYES, M. and PHILLIPS, T. 1998.
How to reduce drought risk: a guide prepared by the
Preparedness and Mitigation Working Group of the Western
Drought Coordination Council. Lincoln, Nebr., National
Drought Mitigation Center. www.drought.unl.edu/plan/
handbook/risk.pdf (accessed 15 November 2004).
PALMER, W.C. 1968. Keeping track of crop moisture
conditions, nationwide: The new crop moisture index.
Weatherwise, No. 21, pp. 15661.
PHILANDER, S.G. 1990. El Nio, La Nia, and the
Southern Oscillation. San Diego, Calif., Academic Press.
WILHITE, D.A. and VANYARKHO, O. 2000. Drought:
pervasive impacts of a creeping phenomenon. In D.A.
Wilhite (ed.). Drought: a global assessment, Vol. I, Ch. 18.
London, Routledge.
wrm_appC.qxd 8/31/2005 12:09 PM Page 601
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management
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
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
wrm_appD.qxd 8/31/2005 12:10 PM Page 602
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Introduction
Just as droughts are defined as unusual scarcities or
deficits of water, floods are defined as unusual surpluses
or excesses of water. The result is higher than usual water
levels that extend out onto floodplains. They occur in
river basins, usually as the result of a lot of rain. They
occur along coastal areas, usually as the result of severe
storms and occasionally as the result of earthquakes.
Floods are natural events. Flood damage usually results
because humans and their structures occupy the space
needed to store the excessive amounts of water. When a
flood comes, as it will, these objects get inundated. To
the extent that structures occupy some of the space
floods need, flood levels increase, thereby flooding
larger areas. The occurrence of flood events is not often
enough to keep reminding people that floodplains can
and do get wet. Flooded buildings and sandbags, such as
shown in Figure D.1, are symbolic of what happens on
occasion when floodplains are used, and often benefi-
cially used, for other activities incompatible with flood
prevention.
In mid-August 2002 Central Europe received an
abnormal amount of rain. Floodwaters inundated
603
Flood Management
Floods are always in conflict with floodplain development. No amount of structural
protection will remove the risk of being damaged by flood flows and the mud,
debris and pollution that accompany them. Hence the challenge is to be prepared
to manage floods and mitigate the resulting damage when floods occur. In short,
this means allowing space for floods when they occur and keeping that space
available when they are not occurring. The challenge is to find the mix of open
space and development on floodplains that maximizes the net expected monetary,
environmental and social benefits derived from them. This chapter discusses some
floodplain management approaches being considered in Europe and in North America,
both motivated by the extreme flood events that occurred in the mid-1990s and early
2000s. Time will tell if any of these approaches are successfully implemented.
Appendix D:
floodplains along three major rivers, claiming around 100
lives and causing over 10 billion euros of damage to prop-
erty in Germany, Russia, Austria and the Czech Republic.
The upper left image in Figure D.1 illustrates a small but
particularly damaging part of that flood in the historic
portion of Dresden, Germany. The map in Figure D.2
shows the affected rivers.
The top image in Figure D.1 shows the River Elbe
(Figure D.2) flooding the historic portion of Dresden in
mid-August 2002. This flood of record exceeded the
highest of floods in 157 years of records. Central Europe
was not the only place getting flooded in the summer of
2002. By August of that year, flooding across Asia had
claimed an estimated 1,800 lives. The waters of Dongting
Lake and the Xiangjiang River (Figure D.3), which flows
through the provincial capital of Changsha, were near
all-time highs. If the sandbag dykes (middle image of
Figure D.1) had failed around the 2,800 square-kilometre
(1,070 square-mile) Dongting Lake that acts as a giant
overflow for the flood-prone Yangtze River, over 10
million people and 667,000 hectares (1.6 million acres)
of fertile crop land would have been flooded.
In the summer of 2002, millions of people in
Bangladesh, India, Nepal, Thailand and Vietnam were
wrm_appD.qxd 8/31/2005 12:10 PM Page 603
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
displaced from their homes by flooding. At least 900
people died in eastern India, Nepal and Bangladesh just in
July and August of 2002 after heavy monsoon rains
triggered widespread flooding, landslides and disease.
In Venezuela at least 45,000 people lost their homes
in flooding in the southwestern state of Apure as rivers
604 Water Resources Systems Planning and Management
overflowed and created lakes of stagnant, polluted water.
In Algiers, the capital of Algeria, the flood toll reached
597 lives when muddy waters swept down a main road.
In Indias state of Assam, thousands of homes were
destroyed by floodwaters. In Cambodia, water levels on
the Mekong River rose to above emergency levels in two
central towns following heavy rains upstream. Floods
from rain falling across southern China inundated
Vietnams northern provinces, including streets in the
capital Hanoi (lower image in Figure D.1).
In spite of considerable study and advice over the past
half century on how to plan and manage floods, and
in spite of increasing amounts of money being spent on
flood protection, annual flood damage is increasing
almost everywhere. In some years, like 2002, the damage
can be quite substantial.
Of the major floods that occurred in the last decade
of the last century, one (in 1993) occurred on the
Mississippi River and some of its tributaries in the central
United States, and two (in 1993 and 1995) occurred
on the Rhine River. Both rivers are major multipurpose
waterways serving many different interests, including
E
0
3
0
6
0
3
h
Brandenburg
Dresden
Leipzig
Prague
Bratislava
Vienna
Germany
Austria
Slovakia
Czech Rep
Vitava
Elbe
Danube
E021001a
Figure D.2. The Rivers Danube and Elbe in Central Europe
experienced their most extreme flood of record in August 2002.
Pearl
Yangtze
China
Laos
Vietnam
Thailand
Burma
India
Tibet
Thaiwan
South
China Sea
East
China
Sea
Shanghai
Hunan
Chenzhou
Yellow Sea
Dongting Lake
Zhishui
Xiangjiang
Yueyang
Changsha
E
0
2
1
0
0
1
b
Figure D.3. Map showing Dongting Lake, which receives
excess flow from the Yangtze River in China.
Figure D.1. Scenes of flooding, the efforts made to prevent
flood damage, and coping at times of high flows and water
levels.
wrm_appD.qxd 8/31/2005 12:10 PM Page 604
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 605
barge transport, which is critical to the economy of the
regions they flow through.
How can floods be managed? How can humans live
with floods that, some believe, are becoming more
frequent and more severe? This appendix discusses
approaches to floodplain management in the Netherlands
and in some parts of the United States. In both cases the
goal is to avoid the continuing spiral of increasing costs,
both for flood protection and then for flood damage
recovery when the protection measures fail. (Additional
information on flood modelling and risk assessment is in
Chapters 7 and 11.)
2. Managing Floods in
the Netherlands
Much of the land surface in the Netherlands is below sea
level; thus, it is not a surprise that those who live in the
Netherlands place a high priority on safety against flooding.
2.1. Flood Frequency and Protection
Levees (dykes) protect the economically important
low-lying part of the Netherlands: roughly the western
half of the country. The design levels of these levees
are linked to the frequency of occurrence of a certain
flood stage. The particular frequencies of occurrence, or
risk levels, are determined by the Dutch Parliament.
Translating those risks to levee heights is the job of
engineers.
Levees along the coasts of densely populated and highly
industrialized parts of the country are to be designed to pro-
tect against all storms whose magnitudes would be
exceeded only once in 10,000 years on average. In other
words, the levees could fail but only for storms that exceed
that 10,000-year storm. The probability of such a failure
happening in any year is 1/10,000 or 0.0001. Of course,
there is no guarantee it could not happen two successive
years in a row. On the basis of methods discussed in
Chapter 7, the probability of having at least one storm that
exceeds the design capacity of the levee at least once in a
fifty-year period is 0.005. While considerably greater than
the probability (0.0001) of it happening in any particular
year, this is still a low probability. The Dutch are not risk
prone when it pertains to floods.
For the less densely populated coastal areas, the design
risk level is increased to storms expected once in 4,000
years, that is, those having a probability of 0.00025 of being
exceeded in any given year. Along the Rhine and Meuse
Rivers, the flood frequency is once per 1,250 years, or a
probability of 0.0008 of being exceeded in any given year.
These so-called design floods also constrain all landscape
planning projects in the floodplain. Proposed river works
for nature restoration, sand mining or any other purpose,
need formal approval as laid down in the River Act.
The condition of flood control works, levees and
fairways is monitored regularly. Every five years a
formal report on flood safety is made. This involves
re-determining the design floods using statistical analysis of
river flows in the period 1900 to date. Furthermore, data
regarding river cross sections and vegetation types and
densities are updated. With the aid of that information, the
design flood levels are assessed, taking into account effects
of wind set-up and a freeboard margin of half a metre
(twenty inches) for overtopping of the levee crests. If pre-
established flood risk level tolerances are being exceeded,
action must be taken to reduce these excessive risks.
While the principal objective is to protect places
where people live, work and spend their leisure time, a
secondary objective is to preserve the quality of the spatial
environment, including natural as well as cultural and
historical sites. The socio-economic interests of many sec-
tors of society are considered when designing alternatives
for flood risk reduction.
2.2. The Rhine River Basin
The Rhine, one of the largest rivers in Europe, extends
1,320 km through Switzerland, France, Germany and
the Netherlands. It enters the Netherlands at Lobith and
travels another 170 km until it reaches the North Sea. The
Rhine catchment area, shown in Figure D.4, covers some
185,000 km
2
, draining parts of nine countries.
The River Rhine now has a combined rainfall
snowmelt driven flow regime. The winter season shows the
largest discharges originating from precipitation in
the German and French parts of the basin. Summer
discharges originate mainly from melting snow in the Swiss
Alps when evaporation surpasses precipitation in the
lowland region. Modern climate scenarios show a temper-
ature rise combined with a rainfall increase during winter
wrm_appD.qxd 8/31/2005 12:10 PM Page 605
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
in the basin. According to these scenarios the River Rhine
could change from a combined snowmeltrain-fed river
into an almost completely rain-fed river. If that happens,
both floods and dry spells could become more frequent.
To maintain the same flood safety standards, additional
flood protection measures will be necessary. Water supplies
in the future will be limited when they are most needed, and
dry spells are likely to occur more often. Also, water quality
may decrease due to both higher concentrations of pollu-
tants and longer residence times of the water behind weirs.
606 Water Resources Systems Planning and Management
This means that preservation of water for use in dry periods
could become necessary and that floodplains may be more
frequently inundated.
The Netherlands is at the end of the Rhine and contains
the Rhine Delta, which comprises more than half of the
Netherlands (Figure D.4). A river often divides into mul-
tiple branches in its delta and the Rhine is no exception.
In this case, the so-called Rhine Branches are the Waal, the
Neder-Rijn/Lek and the Ijssel, as shown in Figure D.5.
Along the Rhine branches, flood levels are entirely
aare
alpenrhein
hochrhein
moezel
neckar
main oberrhein
lahn
sieg
mittelrhein
wupper
ruhr
lippe
niederrhein
nahe
thur
bodensee
neder-
ijsselmeer
ijssel
waal
lek rijn
The
Netherlands
Germany
Luxembourg
France
Austria
Switzerland
Belgium
Lobith
Kln
Bingen
Rheinfelden
E
0
2
1
0
0
1
c
Figure D.4. River Rhine drainage basin.
The river enters the Netherlands at Lobith.
wrm_appD.qxd 8/31/2005 12:10 PM Page 606
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 607
determined by the Rhine discharge. The area where the
water levels are no longer determined solely by river
discharge, but are also under the influence of the tidal
effect of the sea, is referred to as the lower river region. In
total, the low flow channel and the floodplains in the
Netherlands cover an area of approximately 500 km
2
.
To make and keep inhabitable that portion of the
Netherlands that is below sea level, dykes have been
built. As early as the mid-fourteenth century, an almost
completely connected system of dykes arose that created
the landscape of the Netherlands as it is today. There
are two dykes on each side of a floodplain: the summer
and the winter (flood season) dykes. These are shown in
Figure D.6.
Design flow water levels in the river are mostly
determined by the room that is present in the riverbed
and by the flow resistance that the water encounters from
E
0
2
1
0
0
1
e
Lobith
Gorinchem
Schoonhoven
Zutphen
Amerongen
Pannerden
Nijmegen
Hagestein
Zwolle
Deventer
Tiel
Arnhem
Driel
St. Andries
Boven-Rijn
Waal
Neder-Rijn
Lek
ljssel
Pannerdensch
Kanaal
lake
low flow channel
flood plain
area that can be inundated
Ijssel
Figure D.5. The Rhine branches and their
floodplains in the Netherlands. The purple
areas are subject to flooding and hence
need dyke protection.
a
b c
d
e
E
0
2
1
0
0
1
m
a - winter dike
b - flood plain
c - river channel
d - groyne or wing dike
e - summer dike
Figure D.6. Two-tier dyke structures and groynes for flood
protection along the rivers in the Netherlands.
wrm_appD.qxd 8/31/2005 12:10 PM Page 607
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
embankments and vegetation and other obstacles in the
floodplain. The relation between design discharge and
design water levels is therefore not entirely fixed, but
depends upon the width and depth of the riverbed, the
height of the floodplains, the flow resistance, and the
wind and wave effects. Finally, an extra height is added to
the dyke to equip it with an inspection path that keeps it
passable, and to compensate for any subsidence of the
embankment. In the lower river region, the design water
levels are calculated with the aid of computer simulation
models based on a large number of combinations of Rhine
and Maas discharges, the sea level and the possible failure
of the coastal flood barriers in the Rotterdam area.
The design height of a dyke is the water level whose
chance of occurrence is linked to the level of protection
that has been chosen for the protected regions. As already
mentioned, practically all of the dykes along the Rhine
Branches have a protection level of 1/1,250 per year. This
means that the chance that the water level in a particular
year will be higher than the design water level is less than
1/1,250. In the western part of the Netherlands, the
protection levels are higher, namely 1/2,000, and rise to
1/10,000 per year for Central Holland, where large cities
such as Rotterdam and The Hague are located. Here, the
population density and economic interests are larger, but
so also is the difficulty of predicting a storm at sea. Floods
from seawater can result in substantial casualties and
economic damage due not only to water, but also to salt.
Two movable gates protect the entrance to the
Rotterdam harbour, as shown in Figure D.7. Each gate is
about as long as the Eiffel Tower is tall and contains over
twice the towers weight of steel.
In 1995 after a long period of relative freedom from
floods the River Rhine region witnessed a flood. It was
608 Water Resources Systems Planning and Management
not only the highest since 1926, but also one that was
long in duration. Approximately 250,000 people were
evacuated for a little under a week due to the questionable
stability of saturated dykes that had been exposed to
long-term flooding. This flood together with the flood
of 1993 and the comparable events along the River Maas
(Meuse), the second largest river in the Netherlands
motivated government agencies to give some priority to
flood safety and flood risk reduction in the river regions.
Thanks in part to its location in the Rhine Delta, the
Netherlands has been able to flourish economically.
Rotterdam is one of the largest harbours in the world
with a large and rich hinterland. Agriculture in the
Netherlands has been able to profit from the rich soils
that the Rhine has deposited. The economically strong
position of the Netherlands is also due in part to the
dykes. At the same time, the dykes have come to repre-
sent a sort of Achilles heel for the country.
A large part of the Netherlands lies below the
high-water level of the major rivers. At present, the sea
level is gradually rising as a result of climate change. This
same climate change, together with watershed land use
changes within and upstream of the Netherlands, may
cause the peak discharges on the Rhine and Maas to
increase. In the meantime, the area protected by the dykes
has been sinking, primarily due to subsidence and oxida-
tion of peat, because the soil is so well drained. This has
increased the difference between water levels in the area
protected by the dykes and the area outside of the dykes.
The population in the area protected by the dykes, the
land use intensity and the capital investment have all
rapidly increased. As a result, the potential adverse eco-
nomic and emotional consequences of a flood, or even an
evacuation due to flood risk, have increased substantially.
Figure D.7. A drawing and photograph
of the waterway storm surge barrier in
the closed position at Rotterdam in the
Netherlands.
wrm_appD.qxd 8/31/2005 12:10 PM Page 608
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 609
Yet there is a limit to just how much more heightening
and reinforcing of dykes is reasonable. The public has
made it clear they do not want dyke heights increased.
Many feel too boxed in as it is, in this crowded country.
Today a more natural solution is sought.
2.3. Problems and Solutions
As a result of the floods of 1993 and 1995, the design
once-in-1,250-year discharge that must be contained or
controlled within the floodplain is now higher than it was
prior to these events. As shown in Figure D.8, the design
discharge of the Rhine is now established at 16,000 m
3
/s,
an increase of 1,000 m
3
/s. This design discharge height
determines the design height of the dykes.
Without further measures, this means higher dykes.
Due to climate change, the design discharge will most
probably further increase to 18,000 m
3
/s in the second
half of this century. At the same time, the sea level is
expected to rise, causing backwater effects in the estuaries
and rivers.
Given this increase in the design water level, one could
argue that the dykes should be made higher and stronger.
If this is not done, the probability of floods topping the
dykes will increase. However, the higher the dykes,
the more severe the impacts will be once a dyke fails. The
dykes often protect low-lying polders, whose potential
flooding depths already exceed several metres. If nothing
is done to change the available room that exists for the
Rhine River flood flow, this will lead to 2030-cm higher
design water levels in the short run, and up to 90-cm
higher design water levels in the long run, factoring in
climate change impacts. This will make higher dykes
necessary in order to maintain the current level of flood
protection. The higher the river water level is, the larger
the consequences are, should a major flood occur that
exceeds the dyke capacity. In reality, however, higher
flood protection dykes are not going to be a problem,
since the public has made it clear they are against that
option. So, how to provide the extra protection needed to
meet the design risk criteria established by the Dutch
Parliament?
The challenge is to create and execute measures
that despite the higher design discharge that must be
contained avoid the need for additional dyke heighten-
ing. The only way to do that is to create more space
for water to flow at existing dyke levels. This means
increasing the widths or depths of the rivers or their
floodplains, and providing flood overflow areas outside
the dyked floodplain for use when the space within the
dykes is full. Specific floodplain planning and designs
needed to cope with the design flow of 16,000 m
3
/s are to
be implemented by 2015.
2.4. Managing Risk
There is always a possibility that the water level along a
river will exceed the design dyke capacity. The desired
level of safety is a matter of societal choice. If society
wanted complete and total safety, the only way to achieve
it would be to move out of the floodplain. For natural
phenomena such as wind, rain and river discharge, there
are no absolute known upper limits. There are, thus, no
absolute protection guarantees; a finite risk of flooding
will always exist. What society can do however is protect
to the extent people deem appropriate, while preparing
for flood events that exceed the design capacity of any
implemented protection works. Controlled flooding in
emergency overflow areas is one way to manage flood
events that exceed the design water levels.
The notion of flood risk considers both the chance of
a flood and its adverse consequences. Minimizing
the chance of inundation is not the only possible risk
reduction option. Diminishing the adverse consequences
24000
22000
20000
18000
16000
14000
12000
10000
8000
6000
4000
2000
10000 1000 200 100 50 20 5 2 1
1926
1995
1993
expected return period (years)
1250
Lobith flow (m /s)
3
workline
15000
16000
E
0
2
1
0
0
1
f
Figure D.8. Re-evaluation of design flood flows for the River
Rhine in the Netherlands.
wrm_appD.qxd 8/31/2005 12:10 PM Page 609
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
of the event also reduces risk. This idea in particular
underlies the design of emergency overflow areas: it is
better to have a controlled flood with minor damage than
an uncontrolled flood with major damage.
It is technically possible to continue raising dyke
heights, now and on into the future, just as has been done
in the past. But Dutch society has made it clear that this
is undesirable. Dyke reinforcements bring with them
increasingly negative consequences for landscape, nature,
and cultural and historical values. And as the difference in
height between the water level in the river and the area
protected by (behind) the dykes continues to increase, the
flood risk also increases. The larger population and
capital investment in the area protected by the dykes
already make uncontrolled floods more drastic events
than they were some twenty to fifty years ago. Moreover,
the perception of safety increases as the dykes are made
higher and more heavily reinforced, which leads to an
increase in infrastructure investment, which leads to
greater land value and hence potential loss should a
flood occur, which justifies further flood protection
investments, and so on.
It is for this reason that a policy is needed to break this
cycle. Such a policy needs to:
anticipate floods instead of merely reacting to them
make more room for water, rather than relying only on
technical measures such as dyke heightening
prevent the transfer of water problems to downstream
areas by means of a detainstoredischarge sequence.
This requires river-widening and deepening measures to
prevent new design flood flow water levels from exceed-
ing current dyke heights.
Past human interventions have resulted in the erosion of
the low flow channel in upstream sections due to mainte-
nance work on the navigation channel; the sedimentation
in the low flow channel in the downstream sections after
estuaries were closed off; and the silting-up of floodplains
due to the construction of dykes and summer embank-
ments. These processes have increased the elevation of the
floodplains in relation to the low flow channel. At the same
time, the area protected by the dykes has been subsiding
due to drying out and oxidation of the peat soils. All of
this has resulted in river levels that are higher above the
surface of the land protected by (behind) the dykes than
they are above the land adjacent to the river in front of the
dykes.
610 Water Resources Systems Planning and Management
There are several ways to maintain design water levels
as design discharges increase in the Rhine:
keep water in the catchment area upstream of the
Netherlands
store (excess) water along the Rhine Branches in the
Netherlands
discharge (excess) water via the Rhine Branches.
The first alternative attempts to ensure that the upstream
precipitation does not lead to higher discharges down-
stream at a later time. This requires measures in the catch-
ment area upstream of the Netherlands, namely Germany,
such as changes in land use or detention. Comparable
measures in the Netherlands may also help to reduce the
additional flow to the Rhine Branches from tributary
streams and canals.
Alternatives for storing and discharging floodwaters can
be implemented in the Netherlands. The storage of river
water in detention areas along the Rhine Branches leads to
lower peak flows, lowering the water levels downstream of
the detention areas. Conversely, measures that increase the
discharge capacity of the riverbed reduce the water levels
upstream of the measure. Examples include the removal of
obstacles in the winter bed such as high lying areas, ferry
ramps or bridge abutments, excavation of the floodplains,
lowering of groynes or wing dykes, dredging of the low
flow channel, and setting back the dykes. Some of these
options are illustrated in Figures D.9, D.10 and D.11.
An important condition for such projects is that they
must allow at least an increase in the design flow from
15,000 to 16,000 m
3
/s. A further condition is that they
cannot change the proportions of river flow that enter
each of the Rhine Branches in the Netherlands.
2.4.1. Storage
Storage reduces discharge, and is helpful for practically
the entire area downstream of the storage basin.
Emergency overflow and temporary storage areas are
being considered for controlled flooding in a way that
limits damage. A segment of the discharge peak could be
attenuated by temporarily storing it in a dyked area. After
the flood peak has passed through, the temporarily stored
water can be released.
Since the desired effect of a detention area occurs down-
stream from the area itself, a location as far upstream as
wrm_appD.qxd 8/31/2005 12:10 PM Page 610
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 611
high water
8
1
2
3
4
lowering groynes
deepening low flow channel
removing hydraulic obstacles
lowering flood plains
E
0
2
1
0
0
1
d
-
-
-
-
5
6
7
8
setting back dikes locally
setting back dikes globally
using detention reservoirs
reducing lateral inflow
-
-
-
-
7
1
1
2
low water
5, 6
3
4
E
0
2
1
0
0
1
n
Figure D.10. Alternatives to dyke heightening are,
from top to bottom, setting back dykes, deepening
low flow channel, development of green rivers,
removal of hydraulic bottle necks, lowering
groynes, and lowering floodplains.
Figure D.9. Alternatives to dyke heightening
for increasing flood flow capacity of rivers.
wrm_appD.qxd 8/31/2005 12:10 PM Page 611
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
possible is preferred. In the Netherlands, detention areas
close to the German border are thus the most appealing.
The total detention (storage) capacity necessary to
maintain a given flood height depends on the shape
and duration (or length) of a flood wave. For example, a
storage capacity of more than 150 million m
3
would be
required to reduce the peak flow by 1,000 m
3
/s of a flood
wave lasting several days and having an average shape.
This in turn would require a detention area of some
3,000 hectares (30 km
2
) flooded to a depth of five metres.
In areas with less depth, the surface area must be propor-
tionately larger.
Detention areas require enclosure dykes and intakes,
both of which can be extensive. Implementing detention
areas usually also involves landowner evacuation, and
compensation for any damage occurring during occa-
sional flooding. All of these factors add to the cost of
detention measures.
Detention areas seldom need to be used, and in this
case the probability of them being needed in any year is
approximately 1/500, as seen from Figure D.8. This low
probability can be a problem. The less often an area
overflows, the more the societal pressure will be to
develop it and safeguard it against flooding. After many
years without floods, people may begin to think that these
detention areas will never be needed for flood detention.
The result: flood flow restricting activities will begin to
encroach on that land. The additional flood protection
provided by the detention area would then be decreased
if not entirely lost. Perhaps this suggests the need to
periodically let some water flow into these areas during
high river flows, even when it might not have been
necessary, and/or alternatively to allow it to become an
open recreation area or nature reserve.
For detention areas to be effective during flood events,
a reasonable amount of precision in their operation is
required. They must not fill up too soon, because they
612 Water Resources Systems Planning and Management
run the risk of being full before the actual peak discharge
has arrived. If they cannot store some of the peak flow,
when it arrives, they will have no effect in lowering the
maximum flood stage. A similar danger exists for a very
lengthy, flattened flood wave, or when a second peak
occurs soon after the first, and the detention area has not
emptied.
Detention basins must also not fill up too late, because
the discharge peak will already be over. All these
considerations imply that accurate predictions for the
timing and shape of the discharge must be available if
detention basins are to be used effectively.
In the past, detention basins were usually filled via a
fixed sill, a lowered section of the dyke that allows overflow
to take place in a controlled manner, ideally uniformly
over the entire length of the sill. It is still very difficult to
design a sill that will ensure a uniform overflow over a
substantial length. This problem may be overcome by using
a regulated intake under human control. However, people
have had bad experiences with regulated intakes. The
residents of an area are never in favour of inundation, and
if some person is making it occur a conflict with the
responsible organization or agency, if not the operator, is
inevitable.
2.4.2. Discharge-Increasing Measures
Measures taken to increase the discharge capacity can also
reduce water levels while maintaining the same design
discharge. In contrast to storage, however, increasing the
discharge capacity is only advantageous for a limited river
section.
With discharge-increasing measures, it is not only the
height reduction but also the distance a measure covers
that is important. The distance depends on factors such as
the steepness of the water level slope, the location of
the dykes on the floodplain, other obstacles in the area,
E030603b
Figure D.11. Bridge abutments and
supporting structure, before (left) and
after (right) floodway modification.
wrm_appD.qxd 8/31/2005 12:10 PM Page 612
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 613
the hydraulic roughness of the vegetation in floodplains
and so on.
There are many measures available for reducing the stage
of a particular design discharge. Three major ones include:
increasing the flow capacity in the low flow channel
increasing the flow capacity in the floodplains
providing flood storage capacity in the areas protected
by (behind) the dykes (by, for example, setting back
dykes or creating detention basins).
As mentioned earlier, sedimentation occurs in downstream
sections and this makes regular dredging necessary.
Dredging to deepen the low flow channel in the down-
stream sections can lead to a water level reduction.
However, it can also accelerate erosion upstream. Thus, to
maintain the desired design water level, continual dredging
may be required.
Alternatives to dredging include groynes or wing
dykes (Figures D.6, D.9, and D.10). Groynes were
constructed in the past to ensure that the river retained a
sufficient depth without continual dredging. They guide
the river flow to the middle of the channel and ensure that
the depth of the river is maintained for a pre-determined
width. This is particularly important for navigation. They
also tend to prevent sand banks.
Removal of the groynes would result in a decrease in
flow velocities and the river would become shallower
in places. Sandbanks might even form in the middle of
the river. With few exceptions shortening or removing
groynes is an option only if the shipping function of the
river were to be discontinued, and of course in the River
Rhine it is not.
It appears from simulations that lowering of groyne
heights could contribute to a reduction varying from 5 to
15 cm in the water level on the Waal and the Ijssel. On
the Neder-Rijn, the maximum reduction is 10 cm. This
may not seem like much, but on the other hand, the costs
of groyne lowering are relatively low; thus, this measure
is relatively cost effective.
Excavation of the floodplains and the removal of
hydraulic bottlenecks can also be considered. Floodplain
excavation is a measure by which the gradual height
increase caused by sedimentation on floodplains may be
counteracted. Floodplain excavation may be combined
with clay mining, and dyke reinforcement, and/or with
developing nature reserves. Nature development is often
a reasonable use of an excavated floodplain since excava-
tion makes land less profitable for agriculture, particularly
if the summer embankments are also removed in the
process. Moreover, it appears that after excavation, such
development produces a result that is valued by many.
Creating nature reserves without floodplain excavation,
however, pushes water levels upward since rough
vegetation, scrub, and wooded areas can slow down the
discharge. Thus, it requires additional excavation to
compensate for the backwater effect of the vegetation.
While floodplain excavation is an effective way to
reduce flood heights, it is also the most expensive measure.
If it were to be implemented along all three Rhine Branches,
the costs involved would total 3 to 4 billion. Soil excava-
tion is expensive in and of itself, but it is the necessary stor-
age and containment of contaminated soil that substantially
increases the costs. Roughly 15 to 20% of the soil on
the Rhine floodplains is actually contaminated and another
40 to 50% is unusable as building material.
Storing the contaminated soil safely and locally in exist-
ing deep ponds or in sand excavation pits after usable
material has been removed, so-called earth-swapping, can
achieve substantial cost savings, from 1 to 1.5 billion.
The removal of hydraulic obstacles in the floodplain
is another way to increase its discharge capacity without
increasing its water level. Examples of hydraulic
bottlenecks include ferry ramps, bridge abutments
(Figure D.11), high-lying areas (Figure D.12), summer
E
0
3
0
6
0
3
c
Figure D.12. Removal of high-lying
areas on a floodplain.
wrm_appD.qxd 8/31/2005 12:10 PM Page 613
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
embankments that are high and/or perpendicular to the
flow direction, narrowing of winterbeds (Figure D.13)
and other obstacles. Hydraulic bottlenecks may be identi-
fied by studying the water level slope of the river.
Typically, there is a direct relation between structures and
a change in the water level slope.
Removing bottlenecks can decrease design flow water
levels. This decrease can be calculated, as can its cost. The
costs of replacing bridge abutments by bridge sections
(Figure D.11) and the removal of ferry ramps vary from
less than 2.5million to more than 75million for a large
bridge. The costs of removing embankments and small-
scale setting back of dykes (Figure D.13) are usually on
the order of 5 million per project, but they can run up
to over 15million if many houses must be expropriated.
The costs for removing high-lying areas (Figure D.12) can
amount to 30 million, especially in those cases where
land is contaminated.
Substantial water level reductions may be achieved by
widening and deepening measures at an urban bottle-
neck. Such measures are typically very expensive, as a
sizeable area is often needed in times of flooding. Despite
their high costs, these measures can be cost effective at
the urban bottlenecks due to the relatively large reduction
in water levels that may be achieved.
On average, the removal of about sixty bottlenecks can
reduce the water level by 20 cm on the Waal and 10 cm
on the Neder-Rijn/Lek and the Ijssel. However, actual
water level reductions may vary considerably over the
length of the river branches.
There are some forty locations, not including urban
bottlenecks, where large-scale setting back of dykes can
lead to a substantial decrease in the water levels. Setting
back of dykes is particularly effective in situations where
the winter bed is very narrow and causes backwater
effects quite a distance upstream. In such a case, this
decrease in the water level also continues to work
614 Water Resources Systems Planning and Management
relatively far upstream. Some setting back of dykes can, in
fact, decrease the water level up to half a metre. There are
other sites where such measures produce only several
centimetres worth of reduction.
To set back a single dyke section of a few kilometres in
length costs anything from 5 million to 100 million.
While rather expensive, particularly if considerable urban
or industrial development is present on or just behind the
dyke, these measures are cost effective compared to many
other hydraulic bottleneck reductions. Along the Waal
and the Neder-Rijn/Lek, the setting back of all the dykes
together could result in a maximum reduction of 60 cm.
2.4.3. Green Rivers
If it is not possible to set back a dyke, or if the effect is too
small, then a so-called green river alternative may be an
effective way to reduce flood levels. Green rivers are
floodplains between two dykes where water would flow
only during floods. Green rivers may be used for agricul-
tural purposes or may be designed for nature reserves
and/or recreational areas; they are, in short, green. This
does not preclude the possibility of digging a channel or
lake into such an area for the sake of recreation, for exam-
ple. How such a green river may be designed depends
upon the location. Figure D.14 shows a section of a river
floodplain converted to a green river.
Green rivers can lead to significant reductions in water
levels at and upstream of those locations. This means that
other flood height reducing measures along these river
sections may not be necessary.
Green rivers offer options for agriculture, water-based
recreation and nature development. The land is seldom
flooded and, if it is, this usually occurs off-season (out-
side the agricultural season) just as it currently does in the
floodplains. One could say that the land is temporarily
loaned to the river every year, but is otherwise available
E
0
3
0
6
0
3
a
Figure D.13. Opening up a narrowing
by setting dykes further back, and at
the same time creating a nature
reserve.
wrm_appD.qxd 8/31/2005 12:10 PM Page 614
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 615
for other compatible activities. There is thus a definite
practical value derived from green rivers.
Additionally, these measures offer what one could
refer to as a future value for river management. Efforts
must be made to prevent such space from being used for
incompatible residential construction, business parks,
greenhouse complexes and similar land developments
that would incur substantial damage, and indeed increase
the flood height, should a flood occur. While this limits
the land use possibilities at this moment, it offers the
option to implement other river widening and deepening
measures in the future, such as floodplain excavation.
Similar far-reaching measures may be taken to improve
the quality of the surroundings, such as clean-ups of
industrial areas and development of recreational areas.
2.4.4. Use of Existing Water Courses
Where existing canals, brooks or creek remnants run
parallel to the river, these may be connected to the
river channel to take on a portion of the discharge.
Opportunities to do this have been examined in the lower
Rhine Branches region in particular. New, still-to-be
excavated channels are also being studied with respect to
their effectiveness and costs. In practically all cases,
however, water is guided in an entirely different direction
and this places increasing burdens on other rivers or
sections situated further downstream.
2.4.5. The Overall Picture
It would appear that the large-scale setting back of dykes,
construction of detention basins/green rivers and lowering of
groynes would result in the greatest water level reduction
per euro invested. The removal of hydraulic bottlenecks
obtains an average score, as does dredging the low flow
channel. Floodplain excavation is the most expensive, and
in this respect, the least desirable type of measure.
Some measures, such as lowering of groynes and
floodplain excavation, are only really possible in upstream
sections. Others, such as dredging of the low flow channel,
are more feasible downstream. Large-scale setting back of
dykes and green rivers relieve certain bottlenecks only,
albeit with substantial carry-over upstream.
Finally, cost effectiveness is only one criterion.
Sometimes floodplain excavation can involve multiple
objectives; nature development and even sand and clay
mining may also profit from it. The extent to which simi-
lar multiple objectives may be served by various flood
capacity enhancing alternatives needs to be explored.
Combinations of flood-height-reducing measures will
undoubtedly be undertaken along the Rhine River
Branches. Models will be needed to assess their overall
effectiveness. The overall reduction in flood heights will
not be simply the sum of their individual reductions in
isolation. It is not possible to simply add up the water
level reducing effects of the different measures. The dis-
charge of a river is, after all, determined by the function-
ing of the whole: one single bottleneck can negate the
effect of a package of measures. On the other hand, some
measures are synergistic; their overall effectiveness can be
greater than the sum of all their individual water level
reductions. For this reason, a systems view is necessary to
effectively lower water levels across the entire length
of the Rhine Branches. Not only the combination of
measures but also the possible changes in the shape of the
flood wave must be taken into account.
There are an innumerable number of alternatives that
could be put together to safely contain the design flow of
16,000 m
3
/s. It is a matter of preference which measures
will be applied first or most often.
To handle a discharge of 18,000 m
3
/s safely, it appears
that large-scale measures in the dyke-protected area
would be necessary, such as setting back dykes and creat-
ing detention areas and green rivers.
2.5. Dealing With Uncertainties
There are many uncertainties in predicting flood levels,
and sometimes there is little one can do about them. One
uncertainty involves the design discharge itself. Others
Figure D.14. Existing situation (left) of the Rhine floodplain
near the Dutch town of Haaften, and artists impression (right)
of a green river north of the town.
wrm_appD.qxd 8/31/2005 12:10 PM Page 615
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
arise over the shape of the flood wave, the distribution of
River Rhine flow over the three Rhine Branches, the bed
level of the river and the roughness of the vegetation in
the floodplains along the three branches. All these uncer-
tain factors affect the design water levels; hence, the
design water level is itself uncertain.
Moreover, we are at the mercy of changes that may
occur in future and these are by definition uncertain. We
know the climate will change, but we do not know how
quickly or to what extent. All the climate models
currently used predict warming and more frequent
extremes, but the variations between predictions remain
rather substantial. All of these issues present quite a
dilemma for management: on the one hand, safety is so
vital that the river manager has to anticipate higher
discharges that must be accommodated, yet the speed
with which situations may change is very unpredictable.
To wait and see how new floods influence design flow risk
statistics seems unacceptable. Further research can do
nothing to change that.
Essentially, decisions must be made in spite of these
uncertainties.
The design discharge for the Rhine is based upon
an extrapolation of measurement data from the past.
Previously, the period was 1901 to 1991, but this has
now been extended to 1995. The extension of only four
years, two of which had high discharges (namely 1993
and 1995), has caused the slope of the graph to change
somewhat, as shown in Figure D.8, above. The new
design discharge has become 16,000 m
3
/s, an increase of
1,000 m
3
/s. It appears that several floods resulting from
relatively wet years can substantially influence the design
discharge.
The most important reason for the large change in the
design discharge is that it applies for events that occur
once every 1,250 years, well beyond what has been
observed during the 100 years of measurements. This
means that the graph must be extrapolated beyond the
measured data to estimate the design flow associated
with that 1/1,250 risk level. This can result in strange
effects. For example, a plot of the annual peak flows of
the Oder River catchment area in Poland from 1901 to
1985 produces a fairly straight line without any large
deviations. In 1997 however, a discharge of 3,300 m
3
/s
was measured, the largest flow on record. That one event
led to the new extrapolation line causing the 1/1,250
616 Water Resources Systems Planning and Management
discharge to rise from 2,500 to 2,600 m
3
/sec. This is still
far below the 3,300 m
3
/s discharge that actually occurred.
While extreme discharges, perhaps such as that on the
Oder, may be rare, they can nonetheless occur in any
year.
What is clear is that any extrapolation line, such as the
one shown in Figure D.8, is uncertain. It could be higher,
or it could be lower. Figure D.15 shows the 90% confi-
dence interval band of uncertainty associated with the
extrapolation in Figure D.8. This shows that there is a
90% chance that the design discharge on the Rhine (that
is, expected on average once in 1,250 years) will be
between 13,000 and 18,500 m
3
/s. Not only the new
design flow of 16,000 m
3
/s but also the old design flow of
15,000 m
3
/s, as well as the possible future design flow of
18,000 m
3
/s, are in this range.
Figure D.15 is based on historical flow data that may
not be indicative of future flows. There are various climate
change scenarios, each with high and low estimates.
Assuming a high estimate of a 4

C rise in temperature in
the year 2100, the design discharge on the Rhine could
increase by 20%. This added to the current 16,000 m
3
/s
would produce a design discharge of more than 19,000
m
3
/s, assuming that Germany is successful in keeping this
discharge within the dykes.
In this respect, there are thus two types of uncertain-
ties: first, what precisely would change in terms of
Figure D.15. Range of uncertainty associated with estimated
design flows of various return periods. The true design flow
has a 90% chance of being within the blue band.
24000
22000
20000
18000
16000
14000
12000
10000
8000
6000
4000
2000
10000 1000 200 100 50 20 5 2 1
expected return period (years)
1250
Lobith flow (m /s)
3
E
0
2
1
0
0
1
g
wrm_appD.qxd 8/31/2005 12:10 PM Page 616
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 617
climate, and second, how the other Rhine States upstream
of the Netherlands would react to these changes. No one
today can answer that.
How can flood management proceed, given this uncer-
tain future? The answer is: by building into any adopted
management strategy both flexibility (robustness) and
resilience.
Flexibility is the ability to adapt with minimal cost
to a wide range of possible futures. Building in this
flexibility may cost more, but may be still be desirable
insurance against risks society does not want to take.
Regulating the discharge distribution over the Rhine
Branches could increase flexibility. So too could building
temporary and emergency detention areas.
The term resilience, on the other hand, specifically
involves the speed of recovery after a flood and its accom-
panying damage has occurred. This is achieved much
more easily if the consequences of an above design level
flood are not permanent, but remain limited and may be
easily rectified. This requires that no uncontrolled flood
should occur one that would be accompanied by possi-
bly extremely severe damage or even social disruption
but only controlled flooding that will cause the least
amount of damage. In this manner, resilience could be
built into a flood safety system through disaster facilities,
for example in the form of emergency spill areas or by
dividing large dyke rings up into smaller sections com-
partmentalizing to limit flood damage.
It is a major challenge for river managers and also for
the water and spatial-planning policy-makers to develop
a strategy that will minimize future pain by taking into
account the fact that there will always be uncertainties
about the expected discharges in rivers.
2.6. Summary
Visions for future developments on the River Rhine in
and upstream of the Netherlands currently concentrate
on flood mitigation and ecological restoration. Relatively
little effort is devoted to dealing with low flows that have
implications for future water quality and navigation
requirements.
Until recently, the strategy for flood prevention was to
raise dykes (embankments) along the floodplains.
Currently this strategy has met social resistance and is
thought to be too inflexible to cope with an uncertain
future. Alternative solutions focus on reducing water
levels during floods, using retention basins along the
River Rhine in Germany and lowering floodplains in the
Netherlands to enlarge the cross section of the river. At
the same time, these floodplains are to be designed in
such a manner that more natural morphological and
ecological processes in the floodplains can take place.
In the delta of the River Rhine, future adaptation
visions focus on a further widening of the floodplains and
the planning of green rivers. These green rivers will only
be inundated during floods. In the upstream sections in
Germany, the focus is on landscape planning so that
water will flow less quickly to the river.
3. Flood Management on the
Mississippi
The Mississippi River, shown in Figure D.16, is the major
river of North America and the United States. It is over
3,700 km (2,300 miles) in length, flowing from the north-
western part of the state of Minnesota south to
the Gulf of Mexico, just below the city of New Orleans.
It drains 3.2 million km
2
(1.2 million square miles), or
about 40% of the United States. The Mississippi is the
sixth largest river in the world in terms of discharge. Its
annual average flow rate is 14,000 m
3
/s and its freshwater
discharge onto the continental shelf is 580 km
3
/yr. It is a
significant transportation artery, and when combined with
its major tributaries (the Missouri and Ohio rivers), it is
the third largest river transportation system in the world.
Van Meier
St. Louis
Mississippi
Missouri
E
0
2
1
0
0
1
h
Mississippi river drainage basin
region affected by flooding
Figure D.16. Mississippi River Basin in the United States and
the area affected by the 1993 flood.
wrm_appD.qxd 8/31/2005 12:10 PM Page 617
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Typically the discharge at the mouth of the Mississippi
River into the Gulf of Mexico varies over the seasons, with
the highest flow occurring between March and May and
the lowest between August and October. In August and
September of 1993, however, it rained heavily. Monthly
mean discharges in April and May 1993 were approxi-
mately 50% higher than their long-term monthly mean
values. August and September 1993 discharges exceeded
the highest of those recorded in the previous sixty-three
years. The result was a flood.
The States of Minnesota, Iowa, Illinois, and Missouri
were hardest hit (the darker blue area of Figure D.16).
At St. Louis, the river crested at about 15 m (49.6 ft),
some 6 m (over 19 ft) above flood stage. This was about
two metres (more than six feet) above the old record
set in 1973. The Mississippi remained over flood stage at
St. Louis for over two months. Farther north, record
flooding occurred on the Des Moines River, a tributary
of the Mississippi (Figure D.17). At one point flooding
disabled a major water plant, and Des Moines, Iowa, a
city of nearly 200,000 people, was without safe drinking
water.
The discharge of 35,700 m
3
/s (1,070,000 ft
3
/s) was
the greatest ever measured during more then 130
years of data, exceeding the previous high by 26%.
Flood elevations exceeded the design flood stage of 9 m
(30 ft) on the St. Louis gauge for eighty consecutive days
during the main portion of the flood, and for 148 days
during the calendar year. The previous record was sev-
enty-seven days above flood stage, both consecutively
and annually.
The duration of flooding at high stages was unprece-
dented. The flood was 3 m (10 ft) or more over the design
flood stage for thirty-six days, exceeded the fifty-year
618 Water Resources Systems Planning and Management
flood stage for twenty-three days and exceeded the
100-year flood stage for eight days. Before 1993, there
had only been twelve days total in the entire period of
record, dating back to 1861, when floods exceeded the
design flood stage by this extent or more. The total water
volume passing St. Louis from June 26 to September 13
could cover the entire State of Missouri to a depth of
0.75 m (2.5 ft). The peak discharge at St. Louis has been
estimated to have a 150200 year average recurrence
interval. At some upstream stations the flood exceeded
the 500-year average recurrence interval.
Transportation and industry along the Mississippi
River was disrupted for months. Damage to surface and
river transportation in the region was the worst ever
incurred in the United States.
Over 1,000 of the 1,300 levees designed to hold
back floodwaters failed, though major cities along the
rivers, like St Louis, were protected from flooding by
massive floodwalls. Floods displaced over 70,000 people,
and damaged or destroyed nearly 50,000 homes and
over 31,000 km
2
(12,000 square miles) of productive
farmland. Fifty-two people died. The estimated damage
was between $12 and 20 billion (1993 US dollars),
depending on what is included in those estimates.
This flood provided a trigger, once again, for an
evaluation of the existing flood protection schemes and
procedures for managing the rivers and floodplains in
the Mississippi Basin. At the same time, the future use of
the rivers and floodplains, especially in the context of
economic development, coupled with the improvement
of ecological habitats, became important political issues.
Historically considerable attention has been given to
solving problems associated with particular interests.
For the Mississippi Basin, these include navigation,
Figure D.17. Confluence of Des Moines and
Mississippi Rivers during the 1993 flood.
wrm_appD.qxd 8/31/2005 12:10 PM Page 618
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 619
hydropower generation, nature conservation and provid-
ing a certain flood safety for agricultural, industrial and
urban areas. A large number of structural and non-
structural measures have been implemented over the
years, many of which have contributed to the economy of
the region, the safety of the inhabitants and the protection
of natural values. What was less strongly developed, how-
ever, was a clear, concrete and widely accepted perspective
(a policy or strategy) on the long-term future water
resources development in the Mississippi Basin as a whole.
Studies carried out by the Federal Interagency
Floodplain Management Review Committee (known as
the Galloway Committee), further supported by US Army
Corps of Engineers (USACE) studies, all prompted by the
flood, have contributed substantially to debates over
future policies for river management. Various alternatives
for floodplain management were evaluated, but detailed
concrete proposals were not asked for and were not
included within the scope of this floodplain management
assessment.
Identifying alternative future river management strate-
gies that could be implemented raises two questions:
Should the river management strategy be based on a
continuation of the historic practice of river engineer-
ing works, using reservoirs and levees to provide
a certain flood safety level, in combination with
dredging works for navigation? Or alternatively,
should a future strategy focus more on the use of the
floodplain and the river in such a way that environ-
mental values are enhanced, while at the same time the
potential flood damage is reduced?
How and to what extent can the river and floodplain
resources be further used by society for the develop-
ment of the regional and national economy without
compromising the river ecosystem?
Associated with the need for a strategy for river basin
management, additional questions include:
How can federal and state agencies be organized in
such a way that river basin management can take place
effectively?
How can management and development planning
be structured so that stakeholders (industry, environ-
mental groups, farmers, urban population and others)
are actively involved in the process?
3.1. General History
A sketch of some aspects of the history of water resources
development in the United States may provide some con-
text for what has happened in the Mississippi River.
For more than a century, the Federal Government has
been involved in developing water projects for a variety
of purposes including flood control, navigation, power
generation, and irrigation and settlement of the West. Most
major water projects, such as large dams and diversions,
were constructed by government agencies the Bureau
of Reclamation or the U.S. Army Corps of Engineers.
Traditionally, the Corps has built and maintained projects
designed primarily for flood control, navigation and
power generation, whereas Bureau projects were designed
primarily to enhance storage capacity and provide reliable
supplies of water for irrigation and some municipal and
industrial uses.
The Corps operates hundreds of flood control, naviga-
tion, and multipurpose works throughout the country.
While its navigational activities date back to foundation of
the Federal Government, the Corps did not become
involved in flood control until the mid-to-late 1800s, and
then was primarily concerned only with studies and
investigations. Although Congress authorized the Corps
to construct levees throughout the country in 1917, the
modern era of federal involvement in flood control did
not get fully underway until the enactment of the Flood
Control Act of 1936, when Congress declared that it was
in the national interest to assist states with flood control
measures and that this was a proper federal activity. Prior
to this time, flood control had been largely viewed as a
local responsibility. In addition to its flood control
responsibilities, the Corps continued to construct naviga-
tional improvements, and later expanded its activities to
include construction of multipurpose water projects.
For decades, the Federal Government took a largely
structural approach to flood control, floodplain manage-
ment and water supply development. While the Corps
was active in building dams and levees throughout the
country, channelling meandering rivers and streams to
move floodwaters quickly and efficiently out of harms
way, the Bureau was building some of the largest water
supply projects in the world. Thus, the Corps and
the Bureau are substantially involved in the management
of some of the countrys largest river systems, including
wrm_appD.qxd 8/31/2005 12:10 PM Page 619
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
the Colorado, the Columbia, the Missouri and the
Mississippi.
Project construction for all types of water works
expanded greatly during the 1930s and 1940s and con-
tinued rapidly until the late 1960s and early 1970s. By the
late 1960s, a combination of changing national priorities
and local needs, increasing construction costs and the
decreasing number of prime locations for water works all
contributed to a decline in new construction of major
water works nationwide. Water supply for traditional
off-stream uses such as public supply, domestic,
commercial, industrial and agricultural uses was
increasingly in direct competition with a growing interest
in allocating water to maintain or enhance in-stream uses
such as recreation, scenic enjoyment, and fisheries and
wildlife habitat.
During the 1970s, construction of new projects
slowed to a handful of major works, culminating in the
completion of the Tellico Dam project in Tennessee and
the Tennessee Tombigbee Waterway through Alabama
and Mississippi. These projects pitted conservation and
environmental groups, as well as some fiscal conserva-
tives, against the traditional water resources development
community. New on the scene was the National
Environmental Policy Act of 1970 (NEPA), which for the
first time required an assessment of the environmental
effects of federal projects and provided for more public
scrutiny of such projects.
In 1978, President Carter announced that future
federal water policy would focus on improving water
resources management, constructing only projects that
were economically viable, cooperating with state and local
entities, and sustaining environmental quality. The subse-
quent Reagan administration continued to oppose large
projects, contending they were fiscally unsound. Federal
water research and planning activities were also reduced
during the early years of the Reagan Administration (early
1980s), which felt that states should have a greater role in
carrying out such activities. Consistent with this outlook,
President Reagan abolished the Water Resources Council,
an umbrella agency established in 1968 to coordinate
federal water policy and to assess the status of the nations
water resource and development needs.
Congress subsequently scaled back several remaining
authorized projects, changed repayment and cost-share
structures, and passed laws that altered project operations
620 Water Resources Systems Planning and Management
and water delivery programmes. For example, in 1982
Congress passed the Reclamation Reform Act, which
altered the Bureaus water pricing policies for some users.
The Act revised acreage limitation requirements and
charges for water received to irrigate leased lands.
Congress soon passed another landmark law, the Water
Resources Development Act of 1986, which requires local
entities to share the construction costs of water resources
projects built by the Corps. This act did little to enhance
comprehensive systems-wide planning but did allow
projects to be built, especially if members of Congress
wanted them.
Over the last decade, both the Corps and the Bureau
have undertaken projects or programmes aimed at
mitigating or preventing environmental degradation due
in part to the construction and operation of large water
projects. The agencies have pursued these actions
through administrative efforts and congressional man-
dates, as well as in response to court actions. Currently,
the Federal Government is involved in several restoration
initiatives, including the Florida Everglades, the
California Bay-Delta, the Mississippi Delta, Lake Ontario
and St. Lawrence River, and the Columbia and Snake
River Basins in the Pacific Northwest. Degradation
processes in these river, lake and coastal systems have
been occurring over many years.
These restoration initiatives are not without contro-
versy. Each involves many stakeholders at the local and
regional level (water users, landowners, farmers, com-
mercial and sports fishermen, urban water suppliers and
users, navigational interests, hydropower customers and
providers, tourists and environmentalists), all with their
own objectives and desires. At the same time, demand for
traditional or new water resources projects continues
particularly for ways to augment local water supplies,
maintain or improve navigation, and control or prevent
floods and shoreline erosion.
Now let us turn to the Mississippi
Natures design of the upper Mississippi was imperfect for
human use. So Congress, with the help of the US Army
Corps of Engineers, remade it. For over a century, the
Corps has responded to Congressional requests to pro-
vide for transportation and flood control and most
recently for ecology as well in the upper Mississippi.
wrm_appD.qxd 8/31/2005 12:10 PM Page 620
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 621
Natures upper Mississippi was often less than four
feet deep. Todays canalized version handles the standard
upper-Mississippi barge: 2.75 m (9 ft) deep, 10.7 m (35 ft)
wide and some 60 m (195 ft) long. They are tied three
across (32 m or 105 ft) and five deep (297 m or 975 ft),
and pushed (not towed) by a towboat, as shown in
Figure D.18. Most of the locks are 33.5 m (110 ft) wide and
183 m (600 ft) long. So a standard set of barges is divided
into two lengths, each taken separately through the lock,
and then re-tied. A single lock-passage takes about twenty
minutes. Waiting, because of traffic, can sometimes take
twenty hours. This is one of the reasons why the Corps of
Engineers has been asked to build more lock capacity.
Managing floods on the upper Mississippi, to the
extent possible, is done by using reservoirs and levees.
They will no doubt continue to be used.
The upper Mississippi has twenty flood-control
reservoirs, controlled by three Corps of Engineer
Districts: St Paul (16), Rock Island (3), and St Louis (1).
There are also more than a hundred smaller dams and
reservoirs on upper-Mississippi tributaries for local use
and not for flood control on the main river. There are no
flood control reservoirs on the river; the locks and dams
on the river above St. Louis control water levels primarily
for transportation (Figure D.19).
The Corps-controlled reservoirs serves several pur-
poses. At each, a permanent minimum level is kept for
recreation and wildlife. In dry times, water is released to
maintain minimum levels in the tributary and the river.
On a tributary that can flood any time, water is released
after a flood, to be ready for the next one. On a tributary
with well-defined flood seasons, floodwater is held
for release as needed by the tributary and the river.
Hydropower is produced at some reservoirs, and provides
approximately 9% of the combined energy used in the
Mid-continent Area Power Pool, which includes Iowa,
Minnesota, Nebraska, North Dakota, South Dakota, and
portions of Illinois, Montana and Wisconsin.
Levees are used for local flood control. They reduce
flooding for one area but typically cause problems
elsewhere. A levee on one side causes higher floods on the
other. Levees on both sides narrow the channel, causing
higher floods upstream. Levees protecting portions of the
fertile floodplain keep floodwaters from occupying and
soaking into that natural floodplain land. This increases
the flood flow downstream.
Figure D.18. Barged cargo passing through a lock on the
upper Mississippi River.
Figure D.19. The upper Mississippi River locks and navigation
dams, known locally as pools.
Minneapolis
Cape Girardeau
Iowa
Missouri
St. Louis
Chicago
Minnesota
Illinois
Wisconsin
North
Dakota
South
Dakota
Nebraska
Kansas
Oklahoma
Arkansas
I
l
l
i
n
o
i
s
Mississippi
R
i
v
e
r
Mississippi
25
26
24
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
5a
E
0
3
0
6
0
3
o
wrm_appD.qxd 8/31/2005 12:10 PM Page 621
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The overall plan of reservoirs and levees includes the
possibility that, in extreme situations, less valuable areas
will be flooded to reduce the risk of losses in more
valuable ones. According to the basic plan for the upper
Mississippi, submitted to Congress by the Corps in 1940:
The reservoir system proposed is not of sufficient capacity
to completely control floods in the upper Mississippi
River. The best method of operation of reservoirs
often requires that reservoirs be emptied as rapidly as
practicable after a flood has passed in order to be in
readiness for a possible second large flood in the same
season. Large discharges might be superimposed upon
the crest of a major flood at some point downstream.
Hydrological regimes in watersheds have been changed
significantly by the construction of dams, levees and
channels. Drainage systems and other land use changes
have influenced the runoff pattern. Watershed alterations
promoted human welfare, and policy-makers saw
opportunities to use water development as an engine for
economic prosperity. The achievements were impressive:
improved safety against floods, expanded navigation on
the rivers, availability of floodplains for agricultural and
commercial uses, production of hydroelectric power and
other benefits.
However, as in many rivers that are highly controlled,
the floodplains of the upper Mississippi, Illinois and
Missouri Rivers have been reduced in size by the construc-
tion of levees. The remaining floodplains have slowly been
rising due to silt deposition. River runoff has increased due
to the loss of upland cover in the basin. The construction
of wing dykes has made the low flow channel narrower and
deeper. Plants and animal species have slowly disappeared.
Wetland areas have decreased gradually as land was
converted to agricultural use. However, even today the
upper Mississippi, Illinois and Missouri Rivers still contain
extensive and important ecological and landscape values.
Figure D.20 illustrates land cover changes over a century at
two sites along the upper Mississippi.
Comparison of the land-cover/land use data between
1891 and 1989 in the dammed portion of the upper
Mississippi River showed that open water and marsh
habitats generally increased, mostly at the expense of
forest and agriculture (upper images in Figure D.20).
Where river dykes have been built, wetland and woody
areas tended to be converted to agricultural areas over the
same period (lower images in Figure D.20).
622 Water Resources Systems Planning and Management
In all, nearly 2,000 km (1,240 miles) of levees now
isolate more than 400,000 ha (988,000 acres) from the
river during all but the highest discharge rates. Wing dams
and levees, along with other changes to the watershed,
have also had a major effect on habitats by changing
the relationship between discharge and water-surface
elevations. Wing dams have narrowed and deepened the
main channel so that water elevations at low discharges
are now lower than they were historically. Levees restrict
flows and result in higher water elevations during
high discharges. Water surface elevations at relatively low
discharges (1,700 m
3
/s or 60,000 ft
3
/s) have dropped
about 2.4 m (8.0 ft) over the recorded 133-year period at
St. Louis, Missouri, 0.5 m (1.5 ft) over the 52-year record
at Chester, Illinois, and 1.5 m (5.0 ft) over the 60-year
record at Thebes, Illinois. Water-surface elevations at
relatively high discharges (22,000 m
3
/s or 780,000 ft
3
/s),
however, have risen about 2.7 m (9 ft) over the record
period at St. Louis, 1.5 m (5.0 ft) at Chester, and 1.1 m
(3.6 ft) at Thebes.
In the upper Mississippi Basin, protection levels
against flooding vary, largely depending on the use of the
floodplain. Major urban areas are generally protected by
levees that are strong and high enough to withhold a
Figure D.20. Land-cover/land use comparisons at two sites
on the upper Mississippi. The left-hand images depict land
uses in 1891 and the right-hand images show the land uses at
the same sites nearly 100 years later (Wlosinski et al., 2002.).
open water
marsh
green.........
woody..........
sand / mud
agriculture
urban development
E
0
3
0
6
0
3
m
wrm_appD.qxd 8/31/2005 12:10 PM Page 622
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 623
flood like the 1993 flood. Many small towns and agricul-
tural areas, however, are protected by 50-year levees, and
many of these were overtopped or failed in 1993.
The 1993 flood reminded everyone of how damaging
a river can be. The damage caused by this 1993 flood was
substantial but was less than it might have been without
the upstream flood control works (reservoirs and levees),
upland soil conservation measures, terraces and ponds.
Yet those flood damage reduction measures were not
sufficient. The issue addressed by Gilbert White (1945)
over a half century ago, the issue today, and no doubt the
issue on into the future is how to effectively reduce the
threat of future flood damage while supporting continued
economic development in the affected areas.
A variety of options are available to reduce flood risks.
Many are similar to those being considered in the
Netherlands. They include raising or strengthening levees,
relocating or setting back levees, dredging the main and/or
side channels, and lowering floodplains. These may have
major or minor effects, be expensive or relatively inexpen-
sive, be in line or out of line with the current regulations
or policies, and be more or less socially acceptable. For
sustainable protection against flooding, measures in the
upstream and downstream part of the river need to be
considered together as a single system.
3.2. Other Considerations
The Mississippi is an important waterway for inland
navigation, and the regions economy depends on the
inexpensive transport of its bulk commodities. Much of
the transport of goods to and from the Midwest states is
by barge through an extensive lock and dam system.
Periodic dredging is necessary at several locations on the
river to maintain the navigation system.
Safe, speedy and cheap navigation demand a channel
of adequate depth and width, and locks large enough to
accommodate the fifteen-barge tows (Figure D.18). For
such tows, the length of the present locks on the upper
Mississippi River and Illinois is limiting, and often causes
considerable delays, as mentioned above. Furthermore,
the continued growth of commercial and recreational
traffic is approaching maximum lock capacities. The
width of the navigation channel is insufficient for
two-way navigation at many locations. This further delays
navigation and adds to transportation costs. However,
reliability and low cost still make inland navigation a
competitive means of transport. Other positive attributes
of inland navigation are the relatively low energy
consumption and the large transport capacity with few
nuisance and safety problems. Air pollution by navigation
per ton of cargo is far less than that produced per ton by
trucks, or even by rail transport.
Agriculture and industry compete for space on the
floodplain. Agriculture is the leading commercial user of
the floodplains along the rivers in the upper Mississippi
Basin. About 70% of the total 2.4 million ha (6 million
acres) of floodplain is used for agricultural production.
Due to the fertile soils, the yield in well-drained flood-
plains is usually substantially higher than in the upland
areas.
A small but highly valuable portion of the floodplain,
mainly in the direct vicinity of towns, is occupied by
industries. Floodplains well protected by levees offer
industry the advantage of immediate access to inland
water transport to bring in raw materials and transport
finished products, and immediate access to cooling and
process water.
Nature also competes for space on the floodplain.
Large wetland areas used to exist in the basin. These are
unique links between land and water. Some are almost
continuously under water, whereas others may be flooded
for only a short period. This results in a variety of
wetlands and their specific habitat types and functions.
Through these special conditions, wetlands are among
the most biologically productive natural ecosystems in the
world.
In addition to being vital to the survival of various
animals and plants, including threatened and endangered
species, wetlands play a role in the reduction of peak
water levels during smaller flood events (occuring every
few years) because of their capability to store floodwater
and release it slowly. For large events like the 1993 flood,
the effect of wetlands on high water levels is much less
because wetlands are usually saturated by rainfall or
already flooded in an earlier stage.
From the mid-1800s the wetland areas were gradually
decreased mainly to reduce the incidence of malaria and
other waterborne diseases. Wetlands were converted to
agricultural land and to a lesser extent to residential and
industrial uses. Currently, wetlands account for about
10% of the floodplain in the upper Mississippi Basin.
wrm_appD.qxd 8/31/2005 12:10 PM Page 623
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Wetlands, storage lakes and the river itself provide
considerable opportunities for recreation. Popular
activities are hunting, fishing, camping, boating, sightsee-
ing and bird watching. River-related recreation is
important for the economy of local communities. The
locks and dams provide many recreation lakes and the
opportunity for boating, skiing, fishing and hunting.
Floodplains along the rivers also contain numerous
archaeological and historic sites. These sites include
historic architectural and engineering features and struc-
tures, and resources of traditional cultural or heritage
significance to Native Americans and other social or
cultural groups. Examples are forts, quarries, potteries
and burial sites. Construction activities in and along the
river, streambank erosion and extreme floods all can
adversely affect these values.
Some of these multiple uses of the floodplain are
compatible with flood management and obviously others
are not. Yet all must be considered when developing a
flood management plan.
The river itself is a source of water for municipal and
industrial use: drinking water, processing and cooling
water, and irrigation. The number of people who obtain
drinking water from the river and its tributaries is quite
large, yet the amount of water extracted for these
purposes is relatively small when compared with the river
discharge. As long as the flow withdrawn for water uses
(after proper treatment) is returned to the rivers, this
water use can be considered of minor importance in the
overall water balance of the rivers. The export of water to
other basins (diverting Missouri water all the way to
Denver, for example) is another matter.
3.3. Interactions Among User Groups
Just as there are multiple purposes for the upper
Mississippi River water and floodplain, some of which are
complementary and others of which conflict, the same is
true of the different user groups or stakeholders.
Table D.1 provides a rough indication of the ranges
of some of the potentially more conflict-causing interac-
tions. The tables columns are the activities that affect
the activities shown in the rows. Scores of 1 and 2
indicate complementarities and positive interactions, 0
is neutral and 1 and 2 are competing or negative
interactions. Participatory planning can help to push
624 Water Resources Systems Planning and Management
these relative numbers towards the positive side of the
ranges, representing fewer conflicts and more positive
interactions.
There are other important uses of water that are not
mentioned in Table D.1 because the interactions are
obvious and straightforward. Supplying municipal and
domestic public water systems is one. Hydropower is
another.
Some of the categories of conflict indicated in Table D.1
create serious problems in the upper Mississippi Basin.
Some of the potential positive interactions are only weakly
developed. The sharpest conflicts appear to be those
between environmentalists and agriculturalists, and with
those who would develop industrial sites through flood
control and increased river shipping, but are constrained
by legislation limiting land development in the floodplain.
A significant number of farmers consider environmen-
tal concerns by-and-large as expressions of urban
outsiders, who have somehow gained the right to criticize
the farmers activities on their own land. In some cases,
environmental regulations are viewed as threats to farm
livelihood. Furthermore, when the reasons for regulation
involve the preservation of animal or plant species that
seem either insignificant or even nuisances to the agricul-
tural community, the regulations are perceived to be both
insulting and injurious. Whatever the pros and cons of
the regulations, when the conflict of interests has reached
the stage that some of the parties feel threatened, willing
compliance with regulations drops and conflict resolution
becomes far more difficult.
Table D.1 indicates only direct interactions. Indirect
interactions are not listed, although they may also be very
important. For instance, the effect of industrial develop-
ment in or near the floodplain on river ecology is scored
at 0 to 2 because industrial effects may vary between
environmentally neutral, moderately polluting to so badly
polluting that they destroy habitats. Industry is rarely
directly complementary or positive to environmental
concerns in floodplains. However, if the commitment is
made to use some of the revenues generated by industrial
development, through taxes or other means, to enhance
environmental conditions (e.g. using some funds to
establish conservation areas) then the indirect effects
of the industrial development on the environment may
be positive (score 1). Such indirect effects can be
important in negotiations among user groups.
wrm_appD.qxd 8/31/2005 12:10 PM Page 624
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 625
There are serious conflicts over the degree and extent
of flood control that should be provided and the ways this
should be done. There are various possibilities between
two extremes: those who feel that flood protection levels
should be raised significantly, primarily by increasing the
heights of levees all along the river; and those who believe
that development in the floodplain should be actively
discouraged, and existing uses that restrict flood flows
should be reduced. Arguments for the first option are
largely that economic development in the region, particu-
larly farming and industry, will be greatly served by
protection from floods. The arguments for the second
approach are primarily that floodplain protection and
compensation for flood damage are very expensive, and
that there is no compelling reason for the nation to
subsidize floodplain development when alternative sites
for agriculture and industry exist outside the floodplain.
Floodplain protection measures are also frequently not
environmentally friendly. Decreased floodplain develop-
ment allows a return to more natural conditions.
Both arguments are valid. In some cases, flood control
measures are complementary to environmental interests,
particularly when the measures involve setting back levees,
lowering floodplains or making parallel, unregulated chan-
nels or levees that protect or encourage the re-establishment
of certain kinds of habitat. In general, however, environ-
mental interests favour less engineering and lower rather
than higher levees. River dredging is highly controversial.
This puts environmental groups in strong conflict with
those advocating floodplain development.
f
l
o
o
d

c
o
n
t
r
o
l
s
t
r
u
c
t
u
r
e
s
a
g
r
i
c
u
l
t
u
r
e
i
n

t
h
e

f
l
o
o
d
p
l
a
i
n
i
n
d
u
s
t
r
y

i
n

o
r
n
e
a
r

f
l
o
o
d
p
l
a
i
n
s
h
i
p
p
i
n
g

c
h
a
n
n
e
l
d
r
e
d
g
i
n
g
r
i
v
e
r

b
a
s
e
d
r
e
c
r
e
a
t
i
o
n
h
i
s
t
o
r
i
c
a
l

a
n
d
c
u
l
t
u
r
a
l

p
r
e
s
e
r
v
a
t
i
o
n
e
n
v
i
r
o
n
m
e
n
t
a
l
p
r
o
t
e
c
t
i
o
n
on
flood control
on floodplain
agriculture
on industry in
or near floodplain
on river
navigation
on river-based
recreation
on historical &
cultural property
on condition
of river ecology
1
2 -
+ to 1
1 -
+ to
1 -
0 to 1
1 -
+ to 0 1
1 -
+ to
2
1
+
+
to 1
2 -
+ to 1
1 -
+ to 0
1 -
to 2
1
+
+
to 0
1 -
to
2 + 0
2 -
to 0
1 -
to 0 2
1
+
+
to 1
0
+ to
1
1
-
+
to 0
1 -
to 0 0
2 -
to 1
0
+ to 0
1
1 -
+ to 2
1
+
+
to 0
1 -
to 0
2 -
to 0
1 -
to
2
2 -
+ to 2
0
+ to 1
0
+ to 0
1 -
to 0
2 -
to 0
1 -
to
0
2 -
to 2
0
+ to 1
1 -
+ to 0
1 -
to 0
2 -
to 1
1 -
+ to
scoring the effects of
2
1
0
1
2
-
-
+
+
=
=
=
=
=
necessary, or highly complementary and positive effects
generally positive
neutral: no particular advantage nor conflict (small interaction)
implementation of one means a restriction or damage to the other
mutually exclusive thus completely in competition, or highly damaging
E
0
3
0
6
0
3
n
1
1
+ to
-
Table D.1. Interactions among potentially
conflicting river activities.
wrm_appD.qxd 8/31/2005 12:10 PM Page 625
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Negotiating conflicts between different interest groups
requires extensive data collection and expert analyses of
the impacts of different measures to control flooding. The
results of the analyses need to be presented in ways that
are accessible and understandable to stakeholders and
decision-makers, and there should be a process that
encourages debate, negotiation and compromises among
stakeholders.
3.4. Creating a Flood Management Strategy
The current (2004) approach to river management on the
upper and middle Mississippi River is not as integrated or
as well balanced as it could be. This situation is not
unique to this river and some of the reasons for it could
be eliminated if policies that restrict comprehensive
integrated approaches were changed. The 1993 flood in
the upper Mississippi River Basin focused national
attention on this problem, at least for a while. The
concern it aroused presented an opportunity for reassess-
ment and the creation of an integrated water resources
management strategy. One of the conclusions of the
Galloway Committee (1994) was: The United States has a
rare opportunity to make a change in floodplain management.
It should not be missed. Many of the recommendations
made by this Committee have yet to be seriously consid-
ered. It seems, once again, that the concerns generated by
the flood and identified by the committee are fading from
national attention. What began as a useful debate follow-
ing the 1993 flood seems to be dissipating without many
substantial decisions that will make it easier to manage or
mitigate the damage when the next flood occurs.
An argument could be made that the resources offered
by the river could be used considerably more intensively
than they are now without compromising the rivers
ecological integrity. The basins potential for increased
economic and ecological benefits is not fully developed. A
more active management of ecological values, including
using options for ecological restoration, can serve
environmental protection and at the same time allow
sustainable economic development. Active management
of both economic and ecological values requires some
compromises, but can also provide some important gains.
The economic advantage the river offers to the
communities along it and to the region could and should
be increased along with its ecological benefits. Navigation
626 Water Resources Systems Planning and Management
could be increased, even from an environmental point of
view, given the relatively low air pollution and high safety
levels of river freight compared with trucking and even
rail. River transport of containerized cargo in addition to
bulk agricultural products, agro-chemicals, coal and steel
could be stimulated. There is potential to develop
additional river ports and inter-modal transport facilities.
Such development would enhance the economies of the
communities along the river.
Similarly, the use of the river and the riverbanks for
recreation (boating, fishing, hunting) could be increased.
This would benefit local economies and could lead to the
provision of guarantees for environmental conservation.
Increased river-related recreation helps support environ-
mental integrity since a high-quality environment is a
necessary condition for this activity.
An integrated water management strategy would
address flood management issues and uncertainties such
as how levees influence upstream flood stages, how
raising or setting back levees affect water levels elsewhere,
how floodplain vegetation (particularly trees) influences
flood stages, and so on. In addition, the flood safety levels
for different land uses on the floodplain would be
selected. The Galloway Report (1994) emphasized flexi-
bility and variations in flood protection levels. Questions
remain as to what levels of flood protection are appropri-
ate for urban areas/industrial sites and agricultural lands,
and how agreements on such levels are to be reached.
3.5. The Role of the Government and NGOs
State and federal agencies are major players in water
resources management in this as in most major river
basins in the United States. Without their leadership,
not much planning or management takes place or is
implemented. Yet there are limits as to what any particu-
lar agency can do. Thus, it is absolutely necessary for
them to work together with each other and with private
non-governmental organizations (NGOs and other
stakeholder groups) if any approximation of an integrated
water management plan or strategy is to be developed. If
institutional change can facilitate that process, such
change should be considered. Independent river basin
commissions representing all stakeholders and taking
responsibility for integrated river management can be
effective, assuming their recommendations are sought
wrm_appD.qxd 8/31/2005 12:10 PM Page 626
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 627
and would be seriously considered in the higher-level
political decision-making process.
Present regulations regarding benefitcost analysis
need to be revised to allow more types of values to be
considered alongside monetary values. The nations inter-
est in the economic development of the upper Mississippi,
Illinois and Missouri River Basins is not reflected in
the restricted way future benefits are calculated in the
benefitcost considerations required for federal invest-
ment plans in the river and floodplains. The omission of
any estimates of the growth of future economic benefits
can undervalue economic development. Since environ-
mental benefits are not easily translated into monetary
terms, they tend not to be fully considered. The economic
benefits derived from natural well-functioning ecosystems
that help to justify environmental protection are not
made explicit either. The manner in which future benefits
are included in the benefitcost calculation should be
consistent for different types of projects.
A balanced river-basin development plan for the
upper Mississippi, Illinois and Missouri River Basins,
to be drafted interactively with wide participation by
all parties concerned, would improve understanding of
the complex interrelationships between environmental
protection, resource use, and river and floodplain devel-
opment. Such a development plan would not be static,
but would serve as a foundation to be revised on a regular
basis in line with new developments and insights. In such
an adaptive way, the plan could enhance the effective as
well as efficient use of the river resources.
4. Flood Risk Reduction
There are two types of structural alternatives for flood risk
reduction. One is flood storage capacity in reservoirs
designed to reduce downstream peak flood flows. The
other is channel enhancement and/or flood-proofing
works that will contain peak flood flows and reduce
damage. This section introduces methods of modelling
both these alternatives for inclusion in either benefitcost
or cost-effectiveness analyses. The latter analyses apply to
situations in which a significant portion of the flood
control benefits cannot be expressed in monetary terms
and the aim is to provide a specified level of flood
protection at minimum cost.
The discussion will first focus on the estimation of
flood storage capacity in a single reservoir upstream of a
potential flood damage site. This analysis will then be
expanded to include downstream channel capacity
improvements. Each of the modelling methods discussed
will be appropriate for inclusion in multipurpose river
basin planning (optimization) models having longer
time-step durations than those required to predict flood
peak flows.
4.1. Reservoir Flood Storage Capacity
In addition to the active storage capacities in a reservoir,
some capacity may be allocated for the temporary storage
of flood flows during certain periods in the year. Flood
flows usually occur over time intervals lasting from a
few hours up to a few days or weeks. Computational
limitations make it impractical to include such short time
durations in many of our multipurpose planning models.
If we modelled these short durations, flood routing equa-
tions would have to be included in the model; a simple
mass balance would not be sufficient. Nevertheless, there
are ways of including unknown flood storage capacity
variables within longer-period optimization models.
These capacities can be determined on the basis of
economic efficiency within benefitcost analyses, or they
can be based on constraints specifying maximum flood
risk protection levels.
Consider a potential flood damage site along a river. A
flood control reservoir can be built upstream of that site.
The question is how much flood storage capacity, if any,
should it contain. For various assumed capacities, simula-
tion models can be used to predict the impact on the
downstream flood peaks. These hydraulic simulation
models must include flood routing procedures from the
reservoir to the downstream potential damage site and the
flood control operating policy at the reservoir. For various
downstream flood peaks, economic flood damages can
be estimated. To calculate the expected annual damage
associated with any upstream reservoir capacity, the
probability of various damage levels being exceeded in
any year needs to be calculated.
The likelihood of a flood peak of a given magnitude or
greater is often defined by its expected return period.
How many years would one expect to wait, on average, to
observe another flood equal to or greater than some
wrm_appD.qxd 8/31/2005 12:10 PM Page 627
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
specified magnitude? This is the reciprocal of the proba-
bility of observing such a flood in any given year. A T-year
flood has a probability of being equalled or exceeded in
any year of 1/T. This is the probability that could be
calculated by adding up the number of years an annual
flood of a given or greater magnitude is observed, say in
1,000 or 10,000 years, divided by 1,000 or 10,000
respectively. A one-hundred year flood or greater has
a probability of 1/100 or 0.01 of occurring in any
given year. Assuming annual floods are independent, if
a 100-year flood occurs this year, the probability that
a flood of that magnitude or greater occurring next year
remains 1/100 or 0.01.
If PQ is the random annual peak flood flow and PQ
T
is
a particular peak flood flow having a return period of T
years, then by definition the probability of an actual flood
of PQ equalling or exceeding PQ
T
is 1/T;
Pr[PQ PQ
T
] 1/T (D.1)
The higher the return period (i.e. the more severe the
flood), the lower the probability that a flood of that
magnitude or greater will occur. Equation D.1 is plotted
in Figure D.21.
The exceedance probability distribution shown in
Figure D.21 is simply 1 minus the cumulative distribution
function F
PQ
() of annual peak flood flows. The area
under the function is the mean annual peak flood flow,
E[PQ].
The expected annual flood damage at a potential flood
damage site can be estimated from an exceedance
628 Water Resources Systems Planning and Management
probability distribution of peak flood flows at that site
together with a flow or stage damage function. The peak
flow exceedance distribution at any potential damage site
will be a function of the upstream reservoir flood storage
capacity K
f
and the reservoir operating policy.
The probability that flood damage of FD
T
associated
with a flood of return period T will be exceeded is
precisely the same as the probability that the peak flow
PQ
T
that causes the damage will be exceeded. Letting FD
be a random flood damage variable, its probability of
exceedance is
Pr[FD FD
T
] 1/T (D.2)
The area under this exceedance probability distribution is
the expected annual flood damage, E[FD];
E[FD]
0

Pr[FD FD
T
]dFD
T
(D.3)
This computational process is illustrated graphically in
Figure D.22. The analysis requires three input functions
that are shown in quadrants (a), (b), and (c). The dashed-
line rectangles define point values on the three input
functions in quadrants (a), (b), and (c) and the corre-
sponding probabilities of exceeding a given level of
damages in the lower right quadrant (d). The distribution
in quadrant (d) is defined by the intersections of
these dashed-line rectangles. This distribution defines the
probability of equalling or exceeding a specified level of
damage in any given year. The (shaded) area under the
derived function is the annual expected damage, E[FD].
The relationships between flood stage and damage,
and flood stage and peak flow, defined in quadrants
(a) and (b) of Figure D.22, must be known. These do not
depend on the flood storage capacity in an upstream
reservoir. The information in quadrant (c) is similar to
that shown in Figure D.21 defining the exceedance
probabilities of each peak flow. Unlike the other three
functions, this distribution depends on the upstream
flood storage capacity and flood flow release policy. This
peak flow probability of exceedance distribution is
determined by simulating the annual floods entering the
upstream reservoir in the years of record.
The difference between the expected annual flood
damage without any upstream flood storage capacity
and the expected annual flood damage associated with a
flood storage capacity of K
f
is the expected annual flood
damage reduction. This is illustrated in Figure D.23.
Figure D.21. Probability of annual peak flood flows being
exceeded.
E
0
2
0
1
2
1
u
P
r

[
>
]
P
Q
P
Q
T
peak flood flow
0
0
1
1/T
PQ
T
wrm_appD.qxd 8/31/2005 12:10 PM Page 628
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 629
Knowing the expected annual damage reduction associ-
ated with various flood storage capacities, K
f
, permits the
definition of a flood damage reduction function, B
f
(K
f
).
If the reservoir is a single-purpose flood control
reservoir, the eventual tradeoff is between the expected
flood reduction benefits, B
f
(K
f
), and the annual costs,
C(K
f
), of that upstream reservoir capacity. The particular
reservoir flood storage capacity that maximizes the
net benefits, B
f
(K
f
) C(K
f
), may be appropriate from a
national economic efficiency perspective but may not be
Figure D.22. Calculation of the expected
annual flood damage shown as the
shaded area in quadrant (d) derived
from the expected stage-damage
function (a), the expected stage-flow
relation (b), and the expected probability
of exceeding an annual peak flow (c).
exceedance probability
flood stage
E
0
2
0
1
2
1
v
flood flow damage $
high high
high
high
low
Figure D.23. Calculation of expected
annual flood damage reduction benefits,
shown as the darkened portion of
quadrant (d), associated with a specified
reservoir flood storage capacity.
exceedance probability
flood stage
K = 50
K = 0
expected
annual
flood
damage
reduction
benefits
E
0
2
0
1
2
1
w
flood flow damage $
f
f
wrm_appD.qxd 8/31/2005 12:10 PM Page 629
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
best from a local perspective. Those occupying the poten-
tial damage site may prefer a specified level of protection
from the reservoir storage capacity, rather than the
protection that maximizes expected annual net benefits,
B
f
(K
f
) C(K
f
).
If the upstream reservoir is to serve multiple purposes,
say for water supply, hydropower and recreation as well
as for flood control, the expected flood reduction benefit
function just derived could be a component in the overall
objective function for that reservoir.
Total reservoir capacity K will equal the sum of dead
storage capacity K
d
, active storage capacity K
a
and flood
storage capacity K
f
, assuming they are the same in each
period t. In some cases they may vary over the year. If
the required active storage capacity can occupy the flood
storage zone when flood protection is not needed, the
total reservoir capacity K will be the dead storage, K
d
, plus
the maximum of either 1) the actual storage volume and
flood storage capacity in the flood season or 2) the actual
storage volume in non-flood season.
K K
d
S
t
K
f
for all periods t in flood season
plus the following period (that represents the end
of the flood season) (D.4)
K K
d
S
t
for all remaining periods t (D.5)
In the above equations, the dead storage capacity, K
d
, is a
known variable. It is included in the capacity Equations
D.4 and D.5, assuming that the active storage capacity is
greater than zero. Clearly, if the active storage capacity
were zero, there would be no need for dead storage.
4.2. Channel Capacity
The unregulated natural peak flow of a particular design
flood at a potential flood damage site can be reduced by
upstream reservoir flood storage capacity, or it can be
contained within the channel at the potential damage site
by levees and other channel-capacity improvements. In
this section, the possibility of levees or dykes and other
channel capacity or flood-proofing improvements at a
downstream potential damage site will be considered. The
approach used will provide a means of estimating combi-
nations of flood control storage capacity in upstream
reservoirs and downstream channel capacity improve-
ments that together will provide a pre-specified level of
flood protection at the downstream potential damage site.
630 Water Resources Systems Planning and Management
Let QN
T
be the unregulated natural peak flow in the
flood season having a return period of T years. Assume
that this peak flood flow is the design flood for which
protection is desired. To protect from this design peak
flow, a portion QS of the peak flow may be reduced by
upstream flood storage capacity. The remainder of the
peak flow QR must be contained within the channel.
Hence, if the potential damage site s is to be protected
from a peak flow of QN
T
, the peak flow reductions due to
upstream storage, QS, and channel improvements, QR,
must at least equal to that peak flow;
QN
T
QS QR (D.6)
The extent to which a specified upstream reservoir flood
storage capacity reduces the design peak flow at the
downstream potential damage site can be obtained by
routing the design flood through the reservoir and the
channel between the reservoir and the downstream
site. Doing this for a number of reservoir flood storage
capacities permits the definition of a peak flow reduction
function, f
T
(K
f
):
QS f
T
(K
f
) (D.7)
This function is dependent on the relative locations of the
reservoir and the downstream potential damage site, on
the characteristics and length of the channel between the
reservoir and downstream site, on the reservoir flood
control operating policy, and on the magnitude of the
peak flood flow.
An objective function for evaluating these two
structural flood control measures should include the cost
of reservoir flood storage capacity, Cost
K
(K
f
), and the cost
of channel capacity improvements, Cost
R
(QR), required to
contain a flood flow of QR. For a single-purpose, single-
damage site, single-reservoir flood control problem, the
minimum total cost required to protect the potential
damage site from a design flood peak of QN
T
, may be
obtained by solving the model:
minimize Cost
K
(K
f
) Cost
R
(QR) (D.8)
subject to
QN
T
f
T
(K
f
) QR (D.9)
Equations D.8 and D.9 assume that a decision will be
made to provide protection from a design flood QN
T
of
return period T; it is only a question of how to provide the
required protection.
wrm_appD.qxd 8/31/2005 12:10 PM Page 630
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 631
Solving Equations D.8 and D.9 for peak flows QN
T
of
various return periods T will identify the riskcost trade-
off. This tradeoff function might look like what is shown
in Figure D.24.
4.3. Estimating Risk of Levee Failures
Levees are built to reduce the likelihood of flooding on
the floodplain. Flood flows prevented from flowing over a
floodplain due to a levee will have relatively little effect on
users of the floodplain, unless of course the levee fails to
contain the flow. If any of the flow in the stream or river
channel passes over, through or under the levee and
onto the floodplain, the levee is said to fail. This can
result from flood events that exceed (overtop) the design
capacity of the levee, or from various types of geo-
structural weaknesses. The probability of levee failure
along a river reach is in part a function of the levee height,
the probability distribution of flood flows in the stream or
river channel and the probability of geo-structural failure.
The latter depends in part on how well the levee and its
foundation were constructed. Some levees are purposely
designed to fail at certain sites at certain flood stages to
reduce the likelihood of more substantial failures and
flood damage further downstream.
The probability of levee failure given the flood stage
(height) in the stream or river channel is often modelled
using two flood stages. The US Army Corps of Engineers
calls the lower stage the probable non-failure point, PNP,
and the higher stage the probable failure point, PFP
(USACE, 1991). At the PNP, the probability of failure is
assumed to be 15%. Similarly, the probability of failure at
the PFP is assumed to be 85%. A straight-line distribution
between these two points is also assumed, as shown in
Figure D.25. Of course, these points and distributions are
at best only guesses, as not many, if any, data will exist to
base them on at any given site.
To estimate the risk of a flood in the floodplain
protected by a levee due to overtopping or geo-structural
levee failure, the relationships between flood flows and
flood stages in the channel and on the floodplain must be
defined.
Assuming no geo-structural levee failure, the flood
stage in the floodplain protected by a levee is a function
of the flow in the stream or river channel, the cross-
sectional area of the channel between the levees on either
side, the channel slope and roughness, and the levee
height. If floodwaters enter the floodplain, the resulting
water level or stage in the floodplain will depend on the
topological characteristics of the floodplain. Figure D.26
illustrates the relationship between the flood stage in the
channel and the flood stage in the floodplain, assuming
no geo-structural failure of the levee. Obviously once the
flood begins overtopping the levee, the flood stage in the
floodplain begins to increase. Once the flood flow is of
sufficient magnitude that its stage without the levee is
the same as that with the levee, the existence of a levee has
only a negligible impact on the flood stage.
Figure D.24. Tradeoff between minimum cost of flood
protection and flood risk, as expressed by the expected
return period.
E
0
2
0
1
2
1
x
c
o
s
t

o
f

f
l
o
o
d
p
r
o
t
e
c
t
i
o
n
flood return period T
0
0
Figure D.25. Assumed cumulative probability distribution of
levee failure expressed as function of flood stage in river
channel.
E
0
2
0
1
2
1
y
p
r
o
b
a
b
i
l
i
t
y

o
f

l
e
v
e
e

f
a
i
l
u
r
e
flood stage in channel (height)
1.00
0.85
0.60
0.40
0.00
1.20
PNP PFP
0.15
wrm_appD.qxd 8/31/2005 12:10 PM Page 631
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Figure D.27 illustrates the relationship between flood
flow and flood stage in a floodplain with and without flood
levees, again assuming no geo-structural levee failure.
Combining Figures D.26 and D.27 defines the
relationship between reach flow and channel stage. This is
illustrated in the upper-left quadrant of Figure D.28.
Combining the relationship between flood flow and
flood stage in the channel (upper-left quadrant of Figure
D.28) with the probability distribution of levee failure
(Figure D.25) and the probability distribution of annual
peak flows being equalled or exceeded (Figure D.21),
provides an estimate of the expected probability of levee
failure. Figure D.29 illustrates this process of calculating
632 Water Resources Systems Planning and Management
the expected probability of levee failure in any year from
overtopping and/or geo-structural failure. This expected
probability is shown as the blue shaded area in the lower-
right quadrant of the figure.
The channel flood stage function, S(q), of peak flow q
shown in the upper-left quadrant of Figure D.29 is
obtained from the upper-left quadrant of Figure D.28.
The probability of levee failure, PLF(S), a function of
E
0
2
0
1
3
0
a
f
l
o
o
d

s
t
a
g
e

i
n

f
l
o
o
d
p
l
a
i
n
flood stage in stream or river channel
3
3
4
4
2
1
2
1 stage at levee capacity
no levee
stage at levee base
stage at which levee 'fails'
Figure D.26. Influence of a levee
on the flood stage in floodplain
compared to stream or river
flood stage. The channel flood
stage where the curve is vertical
is the stage at which the levee
fails due to overtopping or from
geo-structural causes.
E
0
2
0
1
3
0
b
f
l
o
o
d

s
t
a
g
e

i
n

f
l
o
o
d
p
l
a
i
n
flood stage in stream or river channel flow at
base of levee
flow begins to
overtop levee
flow
corresponding to
maximum levee
height stage
without levee
with levee
flood flow
in stream or river reach
flood stage
in floodplain
f
l
o
o
d

s
t
a
g
e
i
n

c
h
a
n
n
e
l
f
l
o
o
d

f
l
o
w
i
n

s
t
r
e
a
m

o
r

r
i
v
e
r

r
e
a
c
h
without levee
with levee
E
0
2
0
1
3
0
c
Figure D.27. Relationship between flood flow and flood stage
in a floodplain with and without flood levees, again assuming
no geo-structural levee failure.
Figure D.28. Deriving the relation (shown in the upper left
quadrant) between flood flow and flood stage in the channel.
wrm_appD.qxd 8/31/2005 12:10 PM Page 632
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 633
flood stage, S(q), shown in the upper-right quadrant is
the same as in Figure D.25. The annual peak flow
exceedance probability distribution, F
Q
(q), (or its inverse
Q(p)) in the lower-left quadrant is the same as Figure
D.30 or that in the lower-left quadrant (c) of Figure D.22.
The exceedance probability function in the lower-right
quadrant of Figure D.29 is derived from each of the other
three functions, as indicated by the arrows, in the same
manner as described in Figure D.22.
In mathematical terms, the annual expected probabil-
ity of levee failure, E[PLF], found in the lower-right
quadrant of Figure D.29, equals
E[PLF]
0

PLF[S(q)]f(q)dq
0
1
PLF[S(Q(p))]dp

0
1
PLF(p)dp (D.10)
where PLF(p) is the probability of levee failure associated
with a flood stage of S(q) having an exceedance probabil-
ity of p.
Note that if the failure of the levee was only due to
channel flood stages exceeding the levee height (i.e. if
the probability of geo-structural failure were zero), the
expected probability of levee failure would be simply
the probability of exceeding a channel flow whose stage
equals the levee height, as defined in the lower-left quad-
rant of Figure D.29. This is shown in Figure D.30.
Referring to Figure D.30, if the levee height is increased,
the horizontal part of the curve in the upper-right quadrant
would rise, as would the horizontal part of the curve in the
upper-left quadrant as it shifts to the left. Hence, given the
same probability distribution as defined in the lower-left
quadrant, the expected probability of exceeding an
increased levee capacity would decrease, as it should.
4.4. Annual Expected Damage From
Levee Failure
A similar analysis can provide an estimate of the expected
annual floodplain damage for a stream or river reach.
Consider, for example, a parcel of land on a floodplain
at some location i. If an economic efficiency objective
were to guide the development and use of this parcel, then
the owner would want to maximize the net annual eco-
nomic benefits derived from its use, B
i
, less the annual
(non-flood damage) costs, C
i
, and the expected annual
probability
of levee failure
f
l
o
o
d

s
t
a
g
e
i
n

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

e
x
c
e
e
d
a
n
c
e
E
0
2
0
1
3
0
d
flood flow in reach
area under curve is
expected probability
of levee failure
probability of flow
exceeding levee capacity
0.0
1.0
1.0
Figure D.29. Derivation of the probability of exceeding a given
probability of levee failure, shown in lower right quadrant. The
shaded area enclosed by this probability distribution is the
annual expected probability of levee failure.
Figure D.30. Calculation of annual probability of equalling or
exceeding any specified probability of levee overtopping,
shown in the lower right quadrant. The shaded area in this
quadrant is the expected probability of levee overtopping,
assuming there is no geo-structural failure.
probability
of levee failure
f
l
o
o
d

s
t
a
g
e
i
n

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

e
x
c
e
e
d
a
n
c
e
flood flow in reach
probability of flow
exceeding levee capacity
0.0
1.0
E
0
2
0
1
3
0
e
area under curve is
expected probability
of levee failure
1.0
wrm_appD.qxd 8/31/2005 12:10 PM Page 633
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
flood damage, EAD
i
. The issue of concern here is the esti-
mation of the expected annual flood damage.
The degree of damage at location i resulting from a
flood will depend in part on the depth of flooding at that
location and a host of other factors (flood duration, veloc-
ity of and debris in flood flow, time of year and so on).
Assume that the flood damage at location i is a function of
primarily the flood stage, S, at that location. Denote this
potential damage function as D
i
(S). Such a function is
illustrated in Figure D.31.
Integrating the product of the annual exceedance
probability of flood stage, F
s
(S), and the potential dam-
ages, D
i
(S), over all stages S will yield the annual expected
damages, E[D
i
], for land parcel i;
E[D
i
] D
i
(S)F
s
(S)ds (D.11)
The sum of the expected damage estimates over all the
parcels of land i on the floodplain is the total expected
damage that one can expect each year, on average, on the
floodplain.
(D.12)
Alternatively, the annual expected flood damage could be
based on a calculated probability of exceeding a specified
flood damage, as shown in Figure D.22. For this method,
the potential flood damages, D
i
(S), are determined for
various stages S and then summed over all land parcels i
for each of those stage values S to obtain the total poten-
tial damage function, D(S), for the entire floodplain,
defined as a function of flood stage S:
EAD E D
i
[ ]
i

634 Water Resources Systems Planning and Management

(D.13)
This is the function shown in quadrant (a) in Figure D.22.
Levee failure probabilities, PLF(p), based on the
exceedance probability p of peak flows, or stages, as
defined in Figure D.29 can be included in calculations
of expected annual damage. Expressing the damage func-
tion, D(S), as a function, D(p), of the stage exceedance
probability p and multiplying this flood damage function
D(p) times the probability of levee failure, PLF(p) defines
the joint exceedance probability of expected annual dam-
ages. Integrating over all values of p yields the expected
annual flood damage, EAD:
EAD
0
1
D(p)PLF(p)dp (D.14)
Note that the floodplain damage and probability of levee
failure functions in Equation D.14 both increase with
increasing flows or stages, but as peak flows or stages
increase, their exceedance probabilities decrease. Hence,
with increasing p the damage and levee failure probability
functions decrease. The effect of levees on the expected
annual flood damage, EAD, is shown in Figure D.32. The
without levee function in the lower right quadrant of
Figure D.28, is D(p). The with levee function is the
product of D(p) and PLF(p). If the probability of levee
failure, PLF(p) function were as shown in Figure D.28,
(that is, if it were 1.0 for values of p below some overtop-
ping stage associated with an exceedance probability p*,
and 0 for values of p greater than p*), then the function
would appear as shown with levee, overtop only in
Figure D.32.
4.4.1. Risk-Based Analyses
Risk-based analyses attempt to identify the uncertainty
associated with each of the inputs used to identify the
appropriate capacity of various flood risk reduction
measures. There are numerous sources of uncertainty
associated with each of the functions shown in quadrants
(a), (b) and (c) in Figure D.21. This uncertainty translates
into uncertainty associated with estimates of flood risk
probabilities and expected annual flood damage reduc-
tions obtained from reservoir flood storage capacities and
channel improvements.
D( )
i
S D S
i
( )

potential flood damage

to land parcel i
D
i
(S)
floodplain flood stage S
E
0
2
0
1
3
0
f
base elevation of land parcel i
maximum
potential
damage
Figure D.31. An example function defining the damages that
will occur given any flood stage S to a parcel of land l.
wrm_appD.qxd 8/31/2005 12:10 PM Page 634
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 635
Going to the substantial effort and cost of quantifying
these uncertainties, which themselves will be surely be
uncertain, does provide additional information. The design
of any flood protection plan can be adjusted to reflect the
attitudes of stakeholders toward the uncertainty associated
with specified flood peak return periods or equivalently
their probabilities of occurring in any given year.
For example, assume a probability distribution captur-
ing the uncertainty about the expected probability of
exceedance of the peak flows at the potential damage site
(as shown in Figure D.22) is defined from a risk-based
analysis. Figure D.33 shows that exceedance function
together with its 90% confidence bands near the higher
flood peak return periods. To be, say, 90% sure that
protection is provided for the T-year return period flow,
PQ
T
, one may have to design for an equivalent expected
T year return period flow, PQ
T
; that is, the flow
having a (1/T) expected probability of being exceeded.
Conversely, protection from the expected T year peak
flood flow will provide 90% assurance of protection from
flows that will occur less than once in T years on average.
If society wanted to eliminate flood damage it could do
so, but at a high cost. It would require either costly flood
control structures or the elimination of economic
activities on floodplains. Both reduce expected economic
returns from the floodplain, and hence, are not likely to
be undertaken. There will always be a risk of flood dam-
age. Analyses such as those just presented help identify
the risks. They can be reduced and managed but not
eliminated. Finding the best levels of flood protection and
flood risk, together with risk insurance or subsidies
(illustrated in Figure D.34) is the challenge for public and
private agencies alike. Flood risk management is as much
concerned with good things not happening on
floodplains as with bad things happening on them.
5. Decision Support and Prediction
To assist decision-makers in evaluating alternative
solutions for problems encountered in integrated river
management, and in communicating these solutions and
their consequences to the public, interactive computer-
based decision support systems (DSS) can be of consider-
able help. A DSS could consist of up-to-date river and
floodplain information (using geographic information
systems and database managers) and dynamic river
floodplain
damage d
f
l
o
o
d

s
t
a
g
e
i
n

f
l
o
o
d
p
l
a
i
n
s
p
r
o
b
a
b
i
l
i
t
y
o
f

e
x
c
e
e
d
a
n
c
e
p
E
0
2
0
1
3
0
g
peak flood flow
in channel
0.0
with levee
without levee
with levee, overtop only
1.0
Figure D.32. Calculation of expected annual flood damage
taking into account probability of levee failure.
E
0
2
0
1
3
0
h
P
r
[
?
]
P
Q
P
Q
T
peak flood flow
0
0
1
1/T
PQ
T
PQ
T+
1/T -

90% confidence
estimates
Figure D.33. Portion of peak flow probability of exceedance
function showing contours containing 90% of the uncertainty
associated with this distribution. To be 90% certain of
protection from a peak flow of PQ
T
, protection is needed from
the higher peak flow, PQ
T
expected once every T years,
i.e. with an annual probability of 1/(T ) or (1/T) of being
equalled or exceeded.
wrm_appD.qxd 8/31/2005 12:10 PM Page 635
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
models of the complete river network. It could be used to
simulate flooding scenarios in which water managers and
stakeholders could be involved in deciding what to do
and noting the impacts or consequences. Periodic simula-
tion exercises such as this could serve as a reminder to the
public, especially during dry years, that floods can and
will occur, and some may have to be allowed to overflow
levees and cause damage.
5.1. Floodplain Modelling
Efforts to identify improved flood management policies
are almost always based on predictions, admittedly uncer-
tain, of the impacts of specific structural or non-structural
measures. These impact predictions come from models
and their databases. Modelling and the communication of
model results to interested property owners, government
planners and insurance companies are becoming increas-
ingly important components of flood management.
In providing the desired information, one must deal
with uncertainties. Flood studies imply extrapolation of
model parameters and boundary data to situations that
have never been recorded. This raises the question of
the reliability of current procedures of model calibration.
The quality of model results depends to a large extent
on available data and their relevance for future planning.
Two-dimensional (2D) models require accurate digital
terrain models (DTM). The cost of collecting data of
636 Water Resources Systems Planning and Management
sufficient accuracy for flood studies has decreased
significantly through the use of airborne laser altimetry.
Moreover, research has started to provide floodplain
roughness data through complementary analysis of laser
altimetry data. Laser altimetry techniques are now also
applied to the recording of flood levels, which can further
support the calibration of flood simulation models. The
new data collection techniques are also complemented by
hydraulic laboratory investigations, such as the derivation
of laws and parameters describing flow resistance
resulting from various classes of vegetation.
New data collection techniques also show a distinct
change in the type of data underlying model development
and model use. In general, there is a tendency to move from
the collection of point data to spatially-distributed data.
Whereas design flood hydrographs used to form the basis
for flood control studies, the rainfall radar data, together
with river catchment models, allow the use of design storm
patterns. The development of 2D models based upon digital
terrain models and roughness maps is becoming easier and
more cost-effective than the development of 1D models
based upon point information of cross sections measured at
selected intervals along the river axis.
This does not mean that the traditional 1D modelling
of flood events will be fully replaced by 2D model
applications. Although the development of an accurate
2D model supported by the new data collection and
processing techniques is becoming easier and cheaper for
many studies, 2D models require more computer time, and
this is not going to change within a decade. Flood manage-
ment planning and risk assessments typically require the
simulation of many floodplain management alternatives
over large numbers of probabilistic or stochastic events.
Both 1D and 2D models can be used together as
appropriate. They can be integrated in many applications
to provide local refinement of the description, in particu-
lar the flooding of normally dry land. A 2D model can be
used as a support tool for the development of accurate
and extrapolatable 1D models through the transfer of
discharge conveyance and storage functions.
The shift from a policy of protecting as much
agricultural, industrial and urban land as possible by
constructing dykes or levees close to the main river, to
one of creating, if possible, green rivers and larger
floodplains or areas with lower protection against
flooding results in a need to understand and predict the
e
x
p
e
c
t
e
d

r
e
t
u
r
n
flood risk
maximum acceptable risk
level without / with
insurance or subsidy
E
0
2
0
1
3
0
j
Figure D.34. Relationship between expected economic
return from floodplain use and risk of flooding. The lowest
flood risk 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_appD.qxd 8/31/2005 12:10 PM Page 636
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 637
flood flows and heights resulting from such measures.
This requires data on the local topography and flow
resistance.
The hydraulic roughness of floodplains can be
assessed with the aid of maps showing vegetation or land
use classes. Such habitat or land use compositions can be
related to flow roughness descriptions. Assessments of
roughness values associated with each vegetation class are
usually based upon a combination of hydraulic laboratory
experiments and expert judgement. The spatial distribu-
tion of vegetation or land use classes subsequently leads
to the assignment of roughness values at the grid cells of
2D models. Flow resistance for each vegetation class can
be constant or expressed as a function of stage (flood
height).
The 2D models incorporating these floodplain
roughness parameters can be used to assist in the cali-
bration of 1D models. Both types of model can then be
used for a variety of water management studies and pol-
icy analyses.
The link between vegetation or land use classes and
roughness provides a description of the energy losses as a
function of the flow-blocking area in a specific region.
This provides a more accurate description of the losses as
a function of the water level. Flow through reed fields,
for example, meets a high resistance as long as the
vegetation is not overtopped or bent-over and lying on
the bottom. Other, more dense vegetation may even
create laminar flow before overtopping takes place. After
drowning, its flow resistance reduces significantly. A
relation between vegetation or land use and flow
resistance also allows for the introduction of seasonal
variations in flow resistance.
One software product developed at WL | Delft
Hydraulics is the so-called Delft-1D2D system. This is
currently in use for the simulation of large numbers
of scenarios of flooding resulting from potential
dyke breaches. The approach followed in this software
provides many possibilities for the integrated simulation
of river floods, tidal flow and urban storms.
5.2. Integrated 1D2D Modelling
There are various ways in which one can couple hydro-
logical and hydraulic system simulation models. These
include:
Sequential coupling of data flows if one process is
completely independent of another process, for
example, from a rainfallrunoff module to a channel
flow module.
On-line explicit coupling at a selected frequency in
time, such as the link between a surface water flow
module and a water quality module. Such coupling
works well if the mutual dependence of processes
described by the individual modules is not important:
for example, if the time scale of one process is quite
different from the time scale of the other process. If
process time scales are similar, constraints in time
stepping may have to be satisfied.
On-line implicit coupling of modules. This approach
requires not only the exchange of data between various
modules but also the exchange of complete sets of
equations, which are to be solved simultaneously.
The WL | Delft Hydraulics general 0D, 1D and partly 2D
modelling system called SOBEK is designed to integrate
processes, such as rainfallrunoff; river, channel and pipe
flow; water quality and morphology. The hydrodynamic
part handles sub- and supercritical flows, including tran-
sitions such as hydraulic jumps and tidal bores.
This approach to integration offers a range of model-
ling options. In an urban drainage model, one can com-
bine pipe flow, river flow, tidal flow, street flow, and flow
over parks and parking places, for example (Bishop and
Catalano, 2001). Another example is dam break analysis
that allows for any suitable combination of 1D and 2D
flow, including zones with supercritical flow. Finally, the
creation of retention storage in flood control is modelled
more realistically than on the basis of a combined 0D and
1D schematization by itself.
The 1D2D integration can be illustrated by examining
alternative flood scenarios for the Geldersche Vallei in
the centre of the Netherlands (Figure D.35). The area
modelled measures approximately 35 20 km, schema-
tized with a 2D grid of 100 100 m and coupled to a
1D schematization of the principal waterways and road
culverts in the area. The model study had the objective of
assessing flood damage and establishing evacuation
plans for a number of flood scenarios. The most severe
situation is caused by a possible breach in the Grebbedijk,
protecting the area from floods conveyed along the Lower
Rhine.
wrm_appD.qxd 8/31/2005 12:10 PM Page 637
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
The coupled model also allows for the extension of
the 2D model along the Lower Rhine. This river has been
schematized in 1D over a length of 19 km, with the
breach coupled a few kilometres downstream of the
upstream boundary of the Lower Rhine branch. The flow
through the breach significantly influences water levels
upstream, and consequently the discharges generated
through it. It is here that another advantage of implicit
coupling shows up. Inclusion of the Lower Rhine into the
2D model would have given only approximately a 3 to 10
grid-point resolution over its width, which is not suffi-
cient to model its flow satisfactorily. Another and often
principal advantage of 1D and 2D coupling is that the grid
cells of the 2D model part can generally be significantly
larger than when pure 2D modelling is applied. Apart
from the flexibility in schematization, this usually also
leads to considerable savings in simulation times.
6. Conclusions
In the wake of the unusually severe floods on the River
Rhine and on the upper Mississippi River in the early and
638 Water Resources Systems Planning and Management
mid-1990s and the equally severe floods in many parts of
the world since then, including those in Central Europe
and in Asia that occurred in 2002 and 2004, an increas-
ing consensus is emerging. Humans must provide space
for floods where and when they occur, and floods must be
managed better. This is not a new idea. The only sustain-
able way to reduce the continuing cycle of increasing
flood damage and increasing flood protection costs is to
restore the natural floodplains that contain the rivers
floodwaters. Floods have become a reason to restore
rivers rather than dam or dyke them. Many who have
traditionally advocated structural solutions to flood
protection are increasingly embracing the idea that rivers
need to be given space to flood. Flood protection has
joined clean drinking water and recreation as a reason for
restoring river ecosystems.
Governments can motivate more responsible flood-
plain management. National governments can assume
leadership, define clear roles and provide fiscal support
for flood management at lower levels. As happened in the
upper Mississippi River Basin, governments can
provide relocation aid and buyout funding for those
living in floodplains
stop disaster relief payments and subsidized flood
insurance for those who continue to live or develop in
flood prone areas
make flood insurance mandatory where offered, and
have the premiums reflect flood risks
enforce building code requirements on floodplains.
In the United States, the Army Corps of Engineers Flood
Control Programme has had a major influence on flood-
plain development. Many now might claim it to be a
negative influence, but the Corps did what Congress, and
the public, asked them to do. Hundreds of dams and
thousands of miles of Corps levees and floodwalls were
built for flood protection. The result, of course, was to
encourage further development in flood-prone areas.
Existing Corps projects continue to influence the man-
agement of most major rivers and their floodplains,
including those of the Mississippi, Missouri, Ohio and
Columbia River Basins. Although local government is
ultimately responsible for decisions regarding land use in
the United States, flood control projects constructed by
the Corps provide incentives for floodplain development.
E
0
2
1
0
0
1
k
Figure D.35. Coupled 1D2D SOBEK model of the Geldersche
Valley in the Netherlands.
wrm_appD.qxd 8/31/2005 12:10 PM Page 638
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 639
The Corps analyses of benefits and costs have traditionally
strongly favoured structural flood control projects.
Engineers like to build things. Many within the Corps
now recognize that their regulations and incentives need
to be changed to enable them to take a more balanced
view and allow non-structural flood control projects to
compete with structural ones.
In the Netherlands, a similar story can be told.
There is an expression the Dutch like to say: God
made the Earth, but the Dutch made the Netherlands.
Without structures, much of the country would be
permanently under water. But here, as in the United
States and in most of the rest of the world where floods
occur, flood management needs to become a part of
integrated water and land management. Multiple uses,
multiple sectors of the economy, and the interests of
multiple stakeholders need to be considered when
developing comprehensive land use and water manage-
ment plans and policies.
Options available to local governments for floodplain
management include land use zoning. Land zoning
should be based on comprehensive land use plans that
define a vision of how a community should be developed
(and where development should not occur). Through
these plans, uses of the land can be tailored to match the
lands economic benefits as well as its hazards. For exam-
ple, flood hazard areas can be reserved for parks, golf
courses, backyards, wildlife refuges, natural areas or
similar uses that are compatible with the natural flooding
process.
Open space preservation should not be limited to
floodplains, because some sites in the watershed (but out-
side the floodplain) may be crucial to controlling runoff
that adds to the flood problem. Areas that need to be
preserved in a natural state should be listed in land use
and capital improvement plans.
Zoning and open space preservation are ways to
keep damage-prone development out of hazardous or
sensitive areas. Floodplain development regulations
can include construction standards governing what
can and cannot be built in the floodplain. They can
serve to help protect buildings, roads and other
projects from flood damage, and prevent development
from aggravating the flood problem. The three most
common types of floodplain regulations are subdivision
ordinances, building codes, and stand-alone floodplain
ordinances.
Several measures can help reduce runoff of stormwater
and snowmelt throughout the watershed. Retention and
detention regulations, usually part of a subdivision ordi-
nance, require developers to build retention or detention
basins to minimize the increases in runoff caused by
new impervious surfaces and new drainage systems. Best
management practices (BMPs) reduce polluted runoff
entering waterways. Pollutants in runoff may include
lawn fertilizers, pesticides, farm chemicals, and oils from
street surfaces and industrial areas.
Wetlands filter runoff and adjacent surface waters to
protect the quality of lakes, bays and rivers, and protect
many of our sources of drinking water. They can store large
amounts of floodwater, slowing and reducing downstream
flows. They can protect shorelines from erosion. They also
serve as a source of many commercially and recreationally
valuable species of fish, shellfish and wildlife.
Moving a flood-prone building to higher ground is the
surest and safest way to reduce its risk from flooding.
Acquisition of flood-prone property is undertaken by a
government agency, so the cost is not borne by the
property owner. After any structures are removed, the
land is usually converted to public use, such as a park, or
allowed to revert to natural conditions. There are a variety
of funding programmes that can support a local acquisi-
tion project. For example, more than 8,000 homes were
acquired or relocated by the US Government after the
1993 Mississippi Flood.
Based on lessons learned from floods and flood
protection efforts throughout the world, the following
principles are suggested:
Restore river systems and functions that improve flood
management, while at the same time restoring the
natural waterway and its ecosystems:
Restore to a meaningful extent the historic capacity
of rivers and their floodplains to better accommo-
date floodwaters by setting back levees to widen the
floodway of the river channel.
Increase wetland and riverside forest habitat within
the widened river zone.
Increase the use of planned floodplain flooding to
reduce downstream flood peaks.
wrm_appD.qxd 8/31/2005 12:10 PM Page 639
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Strengthen existing and properly sited levees at
high risk that protect high value floodplain uses
that cannot be relocated from the floodplain.
Re-assess the operations of reservoirs and water-
works to ensure efficient, reliable and prudent use
of flood control space. In some cases, dams and
waterworks need to be structurally modified to
improve their ability to release water to avoid
downstream flooding.
Improve use of weather forecasting and monitor-
ing upstream conditions so as to have a better
early warning system of when a flood could be
coming.
Manage the uses of floodplains to minimize taxpayer
expense and maximize environmental health:
Eliminate incentives or subsidies for development
in the most dangerous parts of the floodplain. No
more people should be put in harms way.
Reform floodplain mapping programmes so
that they accurately portray the risks and conse-
quences of anticipated flooding. Ensure that people
understand the risk of flooding where they are
located.
Ensure that new structures that have to be built in
floodplains are designed to resist damage from
foreseeable future floods.
Educate people in the risks of living, working or
farming in areas prone to floods, and make sure
they are willing to bear the appropriate financial
responsibility for such use.
Endeavour to relocate the most threatened people
and communities who volunteer to move to safer
locations.
Ensure that state and local governments responsi-
ble for floodplain land use decisions bear an
increased financial responsibility for flood recovery
efforts.
Manage the entire watershed to provide the most
protection from floods in an environmentally sensitive
way:
Discourage development in remaining wetlands
and floodplains. Wetlands and functioning flood-
plains act as giant sponges to absorb and slow the
progress of floodwaters.
640 Water Resources Systems Planning and Management
Use acquisition and easement programmes to
restore historical wetlands and floodplain acreage
and to promote the functional restoration of associ-
ated river systems.
Discourage clearcutting and road building in areas
prone to landslides.
Where possible, replace non-native hillside annual
vegetation with native perennials to improve rain-
water absorption and reduce hillside erosion.
Newly developed technologies are facilitating the
development of models for studying the propagation
of floods through floodplains, coastal zones and urban
areas. They include the use of airborne laser altimetry to
provide cheaper and more accurate digital terrain
models. In addition, the technology is being used for
an efficient assignment of flow resistance parameters
to flood models. It also offers further support to model
calibration by monitoring water levels during the passage
of flood waves.
The descriptive capabilities of flood models have
been extended significantly through the implicit
coupling of 1D and 2D schematizations. Integrated
1D2D surface and subsurface modelling can be applied
to floodplain and urban flooding investigations. These
applications can lead to a better understanding of the
flood phenomena, more accurate predictions and better
planning.
7. References
BISHOP, W.A. and CATALANO, C.L. 2001. Benefits of
two-dimensional modelling of urban flood projects. In:
Sixth International Conference on Hydraulics in Civil
Engineering, 2830 November 2001, Hobart, Australia.
Barton, Australia, Institution of Engineers.
INTERAGENCY FLOODPLAIN MANAGEMENT REVIEW
COMMITTEE (Galloway Committee). 1994. Sharing the
challenge: floodplain management into the twenty-first century.
Washington, D.C., US Government Printing Office.
USACE. 1991. Proceedings of a Hydrology and Hydraulics
Workshop on Riverine Levee Freeboard, 2729 August 1991,
Monticello, Minnesota. Davis, Calif., USACE Hydrological
Engineering Center.
wrm_appD.qxd 8/31/2005 12:10 PM Page 640
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix D: Flood Management 641
WHITE, G. 1945. Human adjustments to floods. Chicago,
Ill., University of Chicago Press.
WLOSINSKI, J.; OLSEN, D.A.; LOWENBERG, C.;
OWENS, T.W.; ROGALA, J. and LAUSTRUP, M. 2002.
Habitat changes in the Upper Mississippi River floodplain.
National Biological Service, http://biology.usgs.gov/st/
noframe/m2136.htm (accessed 16 November 2004).
Additional References (Further Reading)
ALVAREZ, F. and VERWEY, A. 2000. Simulation of wind
set-up on lakes in a one-dimensional model schematisa-
tion. Fourth International Conference on Hydroinformatics.
2327 July 2000, Cedar Rapids, Iowa. Iowa City, Iowa
Institute for Hydraulic Research (CD-rom).
ASSELMAN, N.E.M. 2001. Laser altimetry and hydraulic
roughness of vegetation. Arnhem, the Netherlands,
Report Q2701 prepared for RIZA of the Directorate-
General for Public Works and Water Management.
ASSELMAN, N.E.M.; MIDDELKOOP, H.; RITZEN, M.R.;
STRAATSMA, M.W.; VAN VELZEN, E.H. and JESSE, P.
2002. Assessment of the hydraulic roughness of river
floodplains using laser altimetry. In: F. Dyer, M.C. Thoms
and J.M. Olley (eds). The Structure, Function and
Management Implications of Fluvial Sedimentary Systems.
IAHS International Symposium on the Structure,
Function and Management Implications of Fluvial
Sedimentary Systems, 26 September 2002, Alice
Springs, Australia. International Association of
Hydrological Sciences (Publication no. 276).
CARLSON, E. 1996. Turn environmental confrontation
into cooperation. LandOwner Newsletter, 9 Dec.
COLLINS, N. and CATALANO, C.L. 2001. Specialised
2D modelling in floodplains with steep gradients. Sixth
International Conference on Hydraulics in Civil Engineering,
2830 November 2001, Hobart, Australia. Barton,
Australia, Institution of Engineers.
DUEL, H.; PEDROLI, B. and LAANE, W.E.M. 1996. The
habitat evaluation procedure in the policy analysis of
inland waters in the Netherlands: towards ecological
rehabilitation. In: M. Leclerc, H. Capra, S. Valentin,
A. Boudreault and Y. Ct (eds). Ecohydraulics 2000,
Proceedings of the Second IAHR Symposium on Habitat
Hydraulics, Qubec (INRS-Eau, Sainte-Foy, Qc), June
1996. Qubec, INRS-Eau. pp. A61930.
DUFF, I.S.; ERISMAN, A.M. and REID, J.K. 1986. and
Direct methods for sparse matrices. Oxford, UK, Claredon.
FAPRI (Food and Agricultural Policy Research Institute).
1994. The economic impact of a loss of navigation on the
Missouri River on Missouri agriculture. Ames, Iowa, Iowa
State University.
FRANK, E.; OSTAN, A.; COCCATO, M. and STELLING,
G.S. 2001. Use of an integrated one dimensionaltwo
dimensional hydraulic modelling approach for flood haz-
ard and risk mapping. In: R.A. Falconer and W.R. Blain
(eds). River basin management, Southampton, UK, WIT.
pp. 99108.
GOLUB, G.H. and VAN LOAN, C.F. 1983. Matrix compu-
tations. Oxford, UK, North Oxford Academic.
GOMES PEREIRA, L.M. and WICHERSON, R.J. 1999.
Suitability of laser data for deriving geographical informa-
tion: a case study in the context of management of
fluvial zones. ISPRS Journal of Photogrammetry and Remote
Sensing, No. 54, pp. 10514.
HARDIN, G. 1968. The tragedy of the commons. Science,
No. 162, pp. 124348.
SAST (Scientific Assessment and Strategy Team). 1994. A
blueprint for change, part v: science for floodplain management
into the twenty-first century. Washington, DC, SAST.
SILVA, W.; KLIJN, F. and DIJKMAN, J.P.M. 2001. Room
for the Rhine Branches in the Netherlands. Arnhem and
Delft, the Netherlands, Directorate-General for Public
Works and Water Management and WL | Delft
Hydraulics.
STELLING, G.S.; KERNKAMP, H.W.J. and LAGUZZI,
M.M. 1998. Delft flooding system: a powerful tool
for inundation assessment based upon a positive
flow simulation. In: V. Babovic and L.C. Larsen (eds.),
Hydroinformatics 98, Rotterdam, Balkema. pp. 44956.
USACE. 1992. Final Upper Mississippi River navigation
study: reconnaissance report, revised version. Rock Island,
Ill., USACE.
wrm_appD.qxd 8/31/2005 12:10 PM Page 641
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
USACE. 2001. Missouri River master water control manual
review and update: draft environmental impact statement,
executive summary. Omaha, Neb.
VERWEY, A. 1994. Linkage of physical and numerical
aspects of models applied in environmental studies.
Keynote lecture. In: Proceedings of the Conference on
Hydraulics in Civil Engineering. International Conference on
Hydraulics in Civil Engineering, 1517 February 1994,
Brisbane, Australia. Australian Institution of Engineers
(National Conference Publication 94/1).
VERWEY, A. 2001. Latest developments in floodplain
modelling: 1D/2D integration. In: Proceedings of the Sixth
642 Water Resources Systems Planning and Management
Conference on Hydraulics in Civil Engineering, 2830
November 2001, Hobart, Australia. Hobart, Institution of
Engineers.
WELTZ, M.A.; RITCHIE, J.C. and FOX, H.D. 1994.
Comparison of laser and field measurements of vegetation
height and canopy cover. Water Resources Research,
No. 30, pp. 131119.
WHITE, G. 1986. Geography, resources, and environ-
ment. In: W. Kates and I. Burton (eds). Volume I: Selected
writings of Gilbert White, Chicago, Ill.: University of
Chicago Press. pp. 44359.
wrm_appD.qxd 8/31/2005 12:10 PM Page 642
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis: Putting it All Together
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
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
wrm_appE.qxd 8/31/2005 12:11 PM Page 644
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
1. Basic Concepts and Definitions
The approach taken for water resources planning involves
a number of terms and concepts related to planning, man-
agement and the role of systems analysis. These terms and
concepts are commonly used in water resources planning,
but with varying connotations. To avoid confusion, we
begin this appendix by indicating what we mean when
using these terms and concepts.
1.1. The Water Resources System
A water resources system (WRS) can be considered to
consist of:
The natural resources system (NRS):
the natural sub-system of streams, rivers, lakes and
their embankments and bottoms, and the ground-
water aquifer
the infrastructure sub-system, such as canals, reser-
voirs, dams, weirs, sluices, wells, pumping plants
and wastewater treatment plants (including the
operation rules for elements of this sub-system)
645
Project Planning and
Analysis: Putting it All Together
The main purpose of water resources management is to ensure the best use of
available water resources. In addition to supporting life itself, water is used in the
production of economic goods and services that are needed to meet national and
regional development goals. Planning projects are often needed to determine how
best to develop and manage these resources. This appendix describes the general
approach used by WL | Delft Hydraulics to assess water resources systems and
develop management strategies for them. Each water resources system is
different and has different problems, and the specific application of any planning
approach should address the particular issues involved. What is important in all
cases is the comprehensive and systematic process of analysis, together with
constant communication among planners, decision-makers and the interested
and affected public.
the water itself, including its physical, chemical and
biological components in and above the soil, often
referred to as the ABC components: abiotic or
physical, biological, and chemical.
The socio-economic system (SES):
water-using and water-related human activities.
The administrative and institutional system (AIS):
the system of administration, legislation and regu-
lation, including the authorities responsible for
managing and implementing laws and regulations.
This definition of a WRS covers the aspects that are essen-
tial for natural resources management: supply, demand
and means to manage the resources. The NRS incorpo-
rates the supply side of the system (resource base), and
the SES the demand side. The management of both the
supply and the demand sides is provided by the AIS.
The external natural boundaries of an NRS usually con-
sist of the water divides of the catchment area, boundaries
of the groundwater aquifer(s) belonging to it, and the
point where the river or canal discharges into the sea.
Beyond the latter point, one usually speaks of coastal
waters. There is a (brackish) transition zone between the
Appendix E:
wrm_appE.qxd 8/31/2005 12:11 PM Page 645
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
two systems. When discharges are large, the brackish
water zone moves seaward. The extent to which the brack-
ish water zone is explicitly included in the NRS depends
upon the interests involved, the problems encountered
and the management objectives to be attained.
The geographical boundaries of the SES and the AIS
vary depending on what part of the socio-economic
system is considered essential for managing and assessing
the impacts of the NRS. For example, in order to analyse
the relations between an NRS and a region with an open
economy, one would have to examine the relationship of
this region with the economy outside it so as to estimate
future developments in the region.
1.2. Functions of the Water Resources System
A general framework of functions is presented in
Table E.1. The classification distinguishes between the
functions and uses that are tangible and those that are
not. Tangible functions, such as hydropower generation
or municipal water supply, may be assigned a monetary
value; intangible functions are activities such as nature
conservation. In between are environmental functions,
some of which may be given a direct value and others val-
ued indirectly, by using a shadow price or other valuation
method (see Chapter 10). The self-purification process of
a river, for example, may be assigned a shadow price by
comparing this work done by nature with the costs of
constructing and operating a wastewater collection and
646 Water Resources Systems Planning and Management
treatment system. The functions in Table E.1 are
explained below.
1.2.1. Subsistence Functions
Local communities depend to a large extent on ambient
water resources for household uses, and for irrigating
home gardens and village irrigation plots. They may also
use the streams, paddy fields, ponds and lakes for fishing.
These uses are often neglected in national economic
accounts, as they are not marketed or otherwise assigned
a monetary value. However, if the WRS becomes unable
to provide these products, this may well be considered an
economic loss, as the people who are dependent on these
products now have to buy them. An example is the cost
of providing purified drinking water where water quality
has deteriorated and become undrinkable.
1.2.2. Commercial Functions
Commercial uses of water resources are reflected in
national economic accounts because they are marketed or
otherwise given a monetary value, e.g. the price to be
paid for domestic water supplies. Catching fish for sale by
individuals and commercial enterprises is an example.
These uses have a commercial value and most are also
consumptive in nature.
The concept of non-consumptive use should be
regarded with certain reservations. Non-consumptive
water use may alter the performance of the WRS in way
Function Description Examples
Subsistence functions Local communities make use of
water and water-based products
which are not marketed
Commercial functions Public or private enterprises
make use of water or water-
based products that are
marketed or otherwise given a
monetary value
Environmental functions Regulation functions
Non-consumptive use
Ecological values Values of the WRS as an
ecosystem
local drinking water supply
traditional fishing
subsistence irrigation
urban drinking water supply
industrial water supply
irrigation
hydro-power generation
commercial fishing
transportation
purification capacity
prevention of salt intrusion
recreation and tourism
integrity
gene pool, biodiversity
nature conservation value
Table E.1. Functions of
the water resources
system.
wrm_appE.qxd 8/31/2005 12:11 PM Page 646
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 647
that constrains or increases the cost to other users.
Hydroelectric generation is an example of a partly non-
consumptive use that affects the system in various ways.
First, reservoirs built for hydropower increase evaporation
losses, and hence reduce the amount of water available for
downstream users. Second, operation of the reservoir for
the production of peak power may alter the flow regime
downstream, and this can adversely affect downstream
habitats and users. Finally, water quality problems related
to reservoirs may seriously affect the ecosystems both
upstream and downstream of the reservoir.
Another example of partly non-consumptive use is
inland water transportation. Oil and chemical pollution
caused by water transport activities can affect other users
and the ecosystem that depend upon the water resources.
Moreover, inland water transportation may involve a real
consumptive demand for water. If water depths are to be
maintained at a certain level for navigational purposes,
releases from reservoirs may be required which provide
no value to other water users. An example is the Lower
Nile system, where water is released from Lake Nasser to
enable navigation (and energy generation) during the so-
called winter closure. This water could otherwise remain
stored for (consumptive) use by agriculture during the
growing season.
1.2.3. Environmental Functions
The drainage basin of a river fulfils a series of functions
that require no human intervention, and thus have no
need of regulatory systems. These environmental functions
include the self-purification capacity of the water system,
and recreational and tourism uses. It is sometimes difficult
to assign a value to environmental functions. They may be
assessed by using a shadow price, calculated as the costs
of providing similar functions in other ways, e.g. the cost
of additional wastewater treatment. Recreational and
tourism values may be determined by assessing the
economic benefits accruing from the use of tourist facilities
like hotels, and/or the revenues from fishing licenses.
1.2.4. Ecological Values
Water is a substance that is essential for life. Rivers,
streams and lakes not only offer an environment for
aquatic species, but are often bordered by wetlands such
as reed beds, floodplains and marshes. These landwater
ecotones (transition areas between adjacent ecological
communities) are known to harbour a rich assemblage of
species, and are also important for the diversity of adja-
cent ecological communities. These ecological entities
have a value of their own, irrespective of the actual or
potential human use, an intrinsic ecological value. There
are many concepts and expressions that describe this
ecological value: heritage value, aesthetic value, nature
value, option value, existence value among others.
Environmental functions relate to the benefits of the
natural environment for humans, and ecological values
include the intrinsic value of nature. In view of the increas-
ing emphasis on sustainability, this distinction is useful
in water resources planning and management. It points
clearly to the need for the continuous care for the natural
conditions of our planet and the maintenance of an accept-
able and sustainable level of environmental quality.
1.3. Policies, Strategies, Measures and
Scenarios
In planning the terms policy goal, strategy, measure and
scenario are frequently used. In popular use they are often
treated as interchangeable, and this can be confusing. In
this appendix, and indeed in this book, the following
meanings are used:
A policy goal has to do with identifying the needs,
prioritizing issues and setting targets for sectors or
regions. A policy goal does not by itself lay down
specific actions; it merely sets the targets and
constraints for the actions (levels, timing and budget).
Policies also may specify in general terms how a
certain target should be achieved: for example, by
applying user-oriented demand management meas-
ures rather than relying on large-scale water supply
infrastructure development. In specifying how targets
should be achieved, other policies are taken into
account (to do with foreign trade perhaps, or fiscal or
income tax requirements, or environmental controls).
Government policies are usually considered as given
Box E.1. Definitions
Policy goal: where do we want to go
Strategy: how do we want to get there
Measure: what are we going to do
Scenario: external development, affecting our strategy
wrm_appE.qxd 8/31/2005 12:11 PM Page 647
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
for any WRS study, but the results of the study may
lead to the adoption of new government policies.
A strategy is defined as a logical combination of indi-
vidual measures or decisions that provides a solution
to the WRS problem, for example, the construction of
a reservoir plus the widening of the canal down-
stream and the increase of the intakes of the irrigation
system. Together, these measures will reduce the risk
of damage to the agriculture sector in a drought-
stricken area. An alternative strategy might be to
implement a cropping pattern that uses less water.
A measure is an individual management action or deci-
sion. A distinction can be made between:
Technical (or structural) measures: modifications of
elements of the water resources infrastructure such
as canals, pumping stations, reservoirs, and fish
stairs or ladders. Technical measures often include
managerial measures as described below.
Managerial measures to improve the (daily) operation
of the system, such as better ways of using the
infrastructure (reservoirs, gates, weirs, sluices, etc.).
Ecological (non-structural) measures to improve the
functioning of the ecosystem, for example by intro-
ducing fry in spawning areas, or large herbivores.
Economic incentives to induce water consumers to
use the water resources in a socially desired manner
by changing the price of the resource use (through
charges, taxes or subsidies).
Regulation measures to restrict uncontrolled use of
the water resources (through land-use zoning,
permits, pollution control and other forms of
restrictive legislation).
Institutional arrangements specifying which govern-
mental agencies are responsible for which functions of
the WRS, and specifying the necessary interactions
between public and private sectors involved.
A scenario is defined as a development exogenous to
the water system under consideration: in other words,
developments that cannot be controlled by the
decision-makers involved in the system. Examples of
scenario variables are climate, climate change, demo-
graphical trends and changes, and economic growth.
What should be treated as a scenario and what as
a (potential) measure may depend on the system
boundaries that have been set. In real integrated
water resources management studies (see Chapter 1,
Section 5.1.3), restrictions on demographic and
648 Water Resources Systems Planning and Management
economic developments could be treated as potential
actions. In that case they are not scenario variables
anymore but should be treated as measures.
1.4. Systems Analysis
In relation to the systems approach to planning, the
literature often emphasizes the mathematical techniques
at the core of this approach. The use of mathematical
tools, however, does not of itself constitute a systems
approach. The approach, designed especially for complex
systems of many interdependent components, involves:
building predictive models to explain system behaviour
devising courses of action (strategies) that combine obser-
vations with the use of models and informed judgments
comparing the alternative courses of action available to
decision-makers
communicating the results to the decision-makers in
meaningful ways
recommending and making decisions based on the
information provided
monitoring and evaluating the results of the strategies
implemented.
Systems analysis and policy analysis are often considered
to be one and the same. If a distinction should be made,
one might define systems analysis as an activity that does
not apply only to policy problems. It can be applied to
any system one wants to analyse for whatever reason.
System diagrams or conceptual models are important
tools in systems analysis. A system diagram represents
causeeffect relations between elements or sub-systems of
the overall system. An example of the use of system dia-
grams in analysing water resources problems is presented
in Figure E.1.
As the figure shows, water-using activities may cause two
problems. First, the quantity demanded is greater than the
supply; second, they affect the natural system (e.g. generate
pollution or alter the water level) with undesired effects. The
perception of these problems can be a trigger for analysis
and planning activities, which in turn can result in manage-
ment actions. The figure shows that the problem can be
addressed in two ways: either by implementing demand-
oriented measures (addressing the water use, i.e. SES), or by
developing the water resources (i.e. NRS). Demand-oriented
measures aim to reduce water use and effluent discharge per
unit of output. Supply-oriented measures on the other hand
wrm_appE.qxd 8/31/2005 12:11 PM Page 648
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 649
are aimed at increasing the water supply so that shortages
are reduced or at increasing the assimilative capacity of the
receiving water bodies. Which measure or combination of
measures is most effective depends on the criteria used by
the implementing authority.
The relations between the elements of the systems can
be expressed quantitatively. The system diagram is an
outline of the computational framework to be used to
analyse the performance of the system.
2. Analytical Description of WRS
Water resources management aims to match the usage of
the WRS (the socio-economic system or SES) with the
natural resources system (the NRS). This matching is
based upon a common notion that scarce resources should
be used for society in an optimal way. Just what is
optimum is determined by those who make decisions. The
interaction between the use and availability of resources is
controlled by the administrative and institutional system
(the AIS) through which this management is implemented.
These three entities are depicted in Figure E.2.
The AIS manages the other two entities, using
information about the present or expected future state
of each system. The management actions of this system
are depicted by the single arrows leading towards the
two interacting systems: by supply-oriented measures
(infrastructure related) for the NRS and demand-oriented
measures for the SES. The single arrow is not meant to
imply there is no feedback from the NRS and the SES to
the AIS; there must be feedback in the form of information
on the state/performance of the two systems, otherwise
effective management would be impossible. The arrows in
Figure E.2 represent only the actions, not the information
flows. The interaction between the user system (SES) and
the resources system (NRS) is a physical one, depicted by
the double arrow between the two systems. Resources
provided by the NRS are used to support the SES, and the
SES has physical effects on the NRS.
Each of the three systems is embedded within its own
environment. The NRS is bounded by climate and
(geo)physical conditions; the SES is formed by the
demographic, social and economic conditions of the
E
0
4
0
7
2
0
d
demand
supply
(I) input WRM
NRS
resource
input
supply
equal
demand ?
state
of natural system
perception
of WRM problem
planning and analysis
effects
SES
Figure E.1. Identification of a water
resources management (WRM) problem.
l
e
g
is
la
tion / organ
iz
a
t
io
n
NRS SES
impacts
resources use
AIS
legislation,
regulation
infra structure
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
Figure E.2. Context for water resources planning.
wrm_appE.qxd 8/31/2005 12:11 PM Page 649
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
surrounding economies; and the AIS is formed and
bounded by the constitutional, legal and political system. It
is important to analyse these boundary conditions.
Boundary conditions are usually considered fixed, but may
not be as rigid as sometimes thought. For example, climatic
conditions are mostly taken as fixed, and the analysis is
based on historical information about precipitation,
temperature and so on. These conditions may be considered
to be subject to change, due to global warming. Whether to
consider this possibility or not should be decided at the start
of any analysis of a water management problem.
Economic growth is another example of a boundary
condition that is often treated as fixed. If the resources
system cannot sustain these foreseen developments (or
only at very high costs), it may be appropriate to recon-
sider this fixed condition. By showing the consequences
of unrestricted growth, the planning agencies can
question the desirability of such development at a higher
(usually national) planning level. This is represented in
Figure E.2 by the border frame socio-economic develop-
ment plans. In fact, the arrow pointing inwards to the
socio-economic system is reversed in such a case: the
analysis provides information to a higher planning level
that determines the (original) boundary conditions.
2.1. System Characteristics of the Natural
Resources System
The natural resources system (NRS) is defined by its
boundaries, its processes and its control measures.
2.1.1. System Boundaries
The study area will often coincide with an administrative
unit (region, district, province, etc.), because the admin-
istrative system usually requires an analysis of the func-
tioning of the water resources within its administrative
boundaries. The system boundaries of the NRS, however,
depend on its physical characteristics. The NRS must
include the administrative area, but may extend over a
larger area, depending upon the physical boundaries.
In many studies it may be useful to subdivide the NRS
into smaller units with suitable boundaries. Examples are
subdivisions into a groundwater and a surface water sys-
tem, subdivision of a surface water system into catchments
and sub-catchments, and subdivision of a groundwater
system into different aquifers or aquifer components. The
650 Water Resources Systems Planning and Management
definition of (sub)systems and their boundaries should be
done in such a way that the transport of water across the
boundaries can be determined properly.
2.1.2. Physical, Chemical and Biological
Characteristics
The physical processes in a NRS are transport and storage
within (sub)systems and transport between them. For the
surface water system, a distinction is made between the
infrastructure of rivers, canals, reservoirs and regulating
structures (the open channel network) and the catch-
ments draining to the open channel network.
The chemical characteristics define the composition of
groundwater and surface water and the processes that
may influence this composition, such as transport,
degradation and adsorption. The level of detail to which
chemical characteristics are considered has to be in line
with the requirements and threats of water-using and
water-related activities, the problems and possible meas-
ures to be considered, and the institutional setting.
A useful concept for describing the biological character-
istics is the ecosystem. An ecosystem is a combination of
abiotic and biotic elements that influence each other. It
includes all the organisms in a given area and their inter-
action with the physical environment, so that a flow of
energy leads to clearly defined trophic structure, biotic
diversity and material cycles within the system. The prime
driving force for the flow of energy and matter within the
system is ultimately the energy from the sun. Through an
often complex series of steps, the solar energy is used for
production and reproduction, consumption and decom-
position, succession and recolonization. Ecosystems are
not static or mechanistic. Rather, they are dynamic
networks of interrelated parts, which to a certain extent
are capable of auto-repair when some parts are lost. This
dynamic character of nature has proven to be essential in
evolutionary terms, as it provides enormous resilience in
coping with physical disturbances such as volcanic
eruptions and global climate change. With respect to the
sustainable development of water resources, the dynamic
character of ecosystems guarantees that human activities
can achieve equilibrium with the environment, albeit at
the expense of the quality of the ecosystem in terms of
species diversity and the like.
In ecosystems, a distinction can be made between the
aquatic, terrestrial and landwater sub-systems:
wrm_appE.qxd 8/31/2005 12:11 PM Page 650
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 651
The aquatic sub-system may be defined as all perma-
nent water bodies: lakes, rivers, streams, reservoirs,
ponds and so on. A characteristic ecological parameter
is, for example, the flow velocity.
The terrestrial sub-system may be defined as the area that
is predominantly dry and receives water only from rainfall
and groundwater. Soil characteristics, altitude, slope and
climate are major parameters for its living conditions.
The landwater sub-system (wetland) is the interface
between the aquatic and terrestrial subsystems. It is
characterized by a highly dynamic environment,
caused by the frequent (often seasonal, but also day-
to-day and annual) changes in the water level.
Examples are the floodplains along rivers, reed belts
fringing lakes and streams, intertidal areas (man-
groves) and salt marshes.
In addition to this broad classification into three
sub-systems, it is often appropriate to distinguish differ-
ent landscape or vegetation units, such as habitats. As a
first approximation, different habitat types can be identi-
fied on the basis of homogenous vegetation structure.
Once the study area has been divided into a series of
sub-systems, habitats or other ecosystem classes, it is
important to assess the degree of spatial relationships
between these subsystems. Ecosystems are rarely
completely isolated from one another, but are linked by
physical transport processes (e.g. water and soil
transport), nutrient pathways and migration patterns, as
discussed in Appendix A.
2.1.3. Control Variables: Possible Measures
Some of the system parameters described above can be
used to control the system. By adding or changing system
parameters, water resources managers can change the
state of the NRS in ways they desire. An example is the
rule curve describing the operation of a reservoir (when to
release how much water for what purpose). Another
example is the dimension of feeder canals. Increasing the
dimension of these canals permits greater allocations of
water to farmers. Control or decision variables can
include those relating to the condition, design and opera-
tion of the WRS. An example of non-physical control that
changes the state of the biotic system is the release of
predator fish in reservoirs to reach a desired balance of
species in the ecosystem.
2.2. System Characteristics of the
Socio-Economic System
Like the NRS, the socio-economic system (SES) has its
boundaries, processes, and control measures.
2.2.1. System Boundaries
The economic system generally does not have a physical
boundary like that of the natural system. Activities in a
river basin, for example, are connected to the surround-
ing economies through the exchange of goods and
services. Also, from a social point of view, the boundaries
of the natural and economic systems rarely coincide.
The factors that determine the socio-economic
activities in a study area now and in the future should be
analysed in the context of the problems being considered.
They could relate to larger systems, such as the national
or even the international economy, e.g. when prices on
the international market are important for local agricul-
tural activities. The boundary conditions for the socio-
economic system are those factors that are beyond the
control of the decision-makers. Examples of such bound-
ary conditions are the state of the world economy, the
value of the US dollar or the price of oil.
2.2.2. System Elements and System Parameters
The socio-economic part of the WRS can be defined by iden-
tifying the main water-using and water-related activities, the
expected developments in the study area, and the parameters
that determine these developments. Examples of activities or
economic sectors that may be relevant and of the type of
information that has to be obtained to be able to describe the
socio-economic system are:
Agriculture and fisheries: present practice, location
and area of irrigated agriculture, desired and potential
developments, water use efficiency and so on.
Power production: existing and planned reservoirs
and power stations, operation and capacity, future
demand for electric energy.
Public water supply: location of centres of population and
industrial activities, expected growth, alternative resources.
Recreation: nature and location, expected and desired
development, water quality conditions.
Navigation: conditions of the water depth in relevant
parts of the open channel system.
wrm_appE.qxd 8/31/2005 12:11 PM Page 651
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Nature conservation: location of valuable and vulnerable
areas and their dependence on water quality and quantity.
Some examples of important system parameters of
the SES are: population dynamics (life expectancy, birth rates,
death rates, etc.), labour force and wage rates, price levels in
relation to national and international markets, subsidies, effi-
ciency of production and water use, and income distribution.
When identifying and analysing activities in the study
area, it is important to consider possible discrepancies
between the opinions of individual actors and their rep-
resentatives. For example, individual farmers may have
different interests than suggested by the official agricul-
tural organizations.
2.2.3. Control Variables: Possible Measures
The functioning of the SES can be influenced by legislative
and regulatory measures, and the price of water may be
a particularly important factor in deciding how much
is demanded. This price can be influenced by the water
resources managers and used as a control variable. When
the cost of water use represents only a small portion of the
total cost of an activity, however, an increase in its price
may have little if any impact on water use. In some cases
water use is a necessity of life no matter how high the costs,
and in such cases, the price of water (or taxation for waste
water discharges) is not a proper control variable.
2.3. System Characteristics of the
Administrative and Institutional System
The administrative and institutional system (AIS), like the
NRS and SES, has its elements that define its boundaries
(its authority or limits) and its processes including its
ways of reorganizing for improved performance.
2.3.1. System Elements
To characterize the administrative and institutional setting,
the responsible institutes at the national, regional and local
levels have to be identified. In most countries, the following
elements in the institutional framework can be distinguished:
the central government, divided into sectors such as
public works, irrigation, agriculture, forestry, environ-
ment, housing, industry, mining and transport
a coordinating body, for example a national water
board, to coordinate actions by various sectors of the
national government
652 Water Resources Systems Planning and Management
regional bodies, based upon the normal subdivisions
of government, for example provinces, districts, cities
and villages.
regional bodies, based on a division according to the
physical characteristics of the area, such as river basin
authorities
water-user organizations, representing the interests of
directly involved stakeholders, for example in
irrigation systems.
The following information needs to be made available:
which ministries and coordinating bodies have
authority and responsibilities related to water
resources management
which agencies are involved in the preparation of
water resources development plans
which national and regional water resources develop-
ment plans are available
which authorities are responsible for implementing
these plans (regulation, construction and operation of
infrastructural works)
what is the existing legislation (laws and regulations)
concerning water rights, allocation of water resources,
water quality control and the financial aspects of water
resources management.
Other sources of information about the administrative and
legal setting include the policy documents and plans of other
water-related sectors such as environment, agriculture, econ-
omy, transportation, physical planning and energy.
2.3.2. Control Variables: Possible Measures
From a systems point of view, the decision or control vari-
ables in the AIS are less clear than in the case of the NRS
and SES. But in the case of the AIS too, measures can be
taken to improve the functioning of the system, for exam-
ple by establishing coordinating bodies when these are
not present, shifting responsibilities towards lower levels
of government, privatization and other measures.
3. Analytical Framework: Phases
of Analysis
A water resources planning study generally comprises
several phases. Although we do not suggest the use of a
rigid framework, some distinct phases and activities can
be recognized and used to structure the analysis as a
logical sequence of steps. The description of these phases,
wrm_appE.qxd 8/31/2005 12:11 PM Page 652
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 653
the activities in them and the interactions between
activities is referred to as the analytical (or conceptual)
framework. A coherent set of models is used for the quan-
titative analysis of the water resources system, measures
and strategies. This set of models and related databases
will be referred to as the computational framework.
The purpose of a water resources planning analysis
is to inform and support decisions. A decision process
is not a simple linear sequence of steps, but involves feed-
backs to earlier steps. Part of the process is thus iterative.
A distinction is made between comprehension loops and
feedback loops. A comprehension loop improves the
understanding of a complex problem by cycling within or
between steps. Feedback loops imply a return to previous
phases of the process. They are needed when:
solutions fail to meet criteria
new insights change the perception of the problem
and its solutions
essential system components and links have been
overlooked
situations change (political, international, develop-
ments in society).
The general analytical framework for water resources
management studies is depicted in Figure E.3. The three
main phases of the framework are: inception, development
and selection. Communication and interaction with the
decision-makers are essential throughout the process.
Regular reporting (inception and interim reports etc.)
helps in effective communication, but a continuous
dialogue is necessary at all stages of the analysis.
The first phase of the process is the inception phase. Here,
the subject of the analysis (what is analysed under what con-
ditions), the objectives (the desired results of the analysis)
and constraints (its limitations) are specified. On the basis of
this initial analysis, during which intensive communication
with the decision-makers is essential, the approach for the
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 E.3. Analytical
framework for water resources
planning studies.
wrm_appE.qxd 8/31/2005 12:11 PM Page 653
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
remainder of the analysis is specified. The results of the
inception phase are presented in the inception report, which
includes the work plan for the other phases of the analysis.
In the development phase, the tools for the analysis and
(elements of) solutions to the water resources problems are
developed. Major activities are data collection and model-
ling. A preliminary analysis will ensure that the developed
tools are suited to the generation and assessment of meas-
ures to solve water resources problems. Individual measures
will be developed and screened in this phase. A gradual
improvement of the understanding of various characteristics
of the WRS is obtained, going from limited data sets and
simple tools to more detailed data and the full set of mod-
els. Interaction with the decision-makers will be greatly
enhanced if they are involved in the preliminary analysis as
part of the analysis team. The formal interaction is struc-
tured through the presentation of results in interim reports.
The purpose of the selection phase is to generate a
limited number of promising strategies that, after detailed
assessment of their effects in terms of the evaluation
criteria, will be presented to the decision-makers who
with inputs from the public will decide on their
preferred line of action. Important activities in this phase
are strategy design, impact assessment, evaluation of strate-
gies, sensitivity and scenario analysis, and presentation.
The results of this phase are included in the final report,
together with a summary of the results of the development
phase. Continuous interaction with the decision-makers is
essential in the phase. They or their representatives should
be part of the team that carries out the analysis. During this
phase, public consultations will also be needed to inform
and involve stakeholders and get their reactions and inputs.
Reference is made above to the decision-makers.
Although it is clear that the analysis will eventually
support decisions, it is not always clear who (or which
unit) will make the final decisions. If an outside consultant
performs the analysis, careful selection of the coordinating
agency is essential for the successful implementation of the
project. Interaction with the decision-making arena
usually takes place through steering committees (which
can act as an interdepartmental forum), technical advisory
committees and public participation processes.
The inception phase is a very important component of
this analytical framework. It determines to a large extent
what will be done in the analysis. This inception phase
should therefore get ample attention. Many commercial con-
tract terms of reference (ToRs) for planning studies include
654 Water Resources Systems Planning and Management
an obligation for the consultant to submit an inception
report a few weeks after the starting date of the assignment.
An inception report is simply an update of the proposal, cov-
ering the time between the submission of the proposal and
the starting date, and addresses only the changes that need to
be taken into account. Depending on the complexity of the
system and required interaction with decision-makers and
stakeholders, an estimate of the percentage of the total proj-
ect resources (money and time) that should be devoted to the
inception phase can amount to 30 or even 40%.
4. Inception Phase
Water resources planning studies are often triggered
(left oval in Figure E.3) by specific management problems
such as the need to increase power production, the occur-
rence of droughts or floods, or the threat of water quality
deterioration. The need for water resources planning in
relation to other sector planning efforts may also be a
trigger. Which parts of the WRS are studied and under
what conditions follows primarily from the objectives of
the study (and from the available budget, information and
time). The initiators of the study generally have more or
less concrete ideas about the objectives and subject of the
analysis. They react to triggers as mentioned above and,
being aware of problems or issues, have taken the initia-
tive for a planning analysis.
The clients ideas about the problems and issues to
be addressed will usually be described in a Project
Formulation Document (PFD) or ToR (terms of reference).
The very first activity of the project is to review and discuss
these documents. If the subject (what) and objectives (what
for) of the analysis are adequately described in the ToR, the
first step of the study is to specify the approach (how). In
many situations however, the first task of the analysts is to
assist the decision-makers in further specifying the objec-
tives and subject of the analysis. For this activity, intensive
communication is required with authorities involved in
water resources planning and the stakeholders. They can
provide information on the requirements of various inter-
est groups related to water and on expected problems. It is
not uncommon to have the stated objectives of a study dif-
fer from the actual (often unstated) objectives of the client.
Furthermore, objectives change over time. As emphasized
in Chapter 2, constant and effective communication
between analysts and their clients is absolutely essential.
wrm_appE.qxd 8/31/2005 12:11 PM Page 654
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 655
This will contribute to the development of scenarios
for the analyses. Scenarios are defined as developments
exogenous to the water system under consideration (see
Section 1.3).
4.1.2. Problem Analysis
Based upon the previous steps a problem analysis can be
carried out. This should be done based on the facts and
what the stakeholders and decision-makers perceive as a
problem. A simple example of a problem analysis is
presented below in Box E.2.
A problem analysis should be expressed as far as
possible in terms of the socio-economic and environmental
or ecosystem impacts that have meaning to the decision-
makers. Not all stakeholders may be able to relate to pre-
dicted changes in flows, water levels or pollutant
concentrations. Some may want to know how much money
is involved (e.g. crop production, repair costs), the rate of
shore line erosion, the relative change in fish population, or
the number of people affected by flooding. Expressing out-
comes in terms of socio-economic impacts makes it easier
to relate the problems to the (socio-economic) develop-
ment objectives that decision-makers have formulated for
the particular region or system under consideration.
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
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
and computation
framework
characteristics of WRS
activities & developments
policies & institutions
problems & measures
objectives and criteria
data availability
initial analysis
Figure E.4. Components of initial analysis.
The inception phase starts with an initial analysis, to
clarify the subject and the problems and objectives
involved. Building on the understanding that developed
from this initial analysis, a more or less formal description
can be given of the approach that the planning study will
follow, including a description of the analysis steps that
will be taken, the computational framework that will be
used and the assumptions and conditions under which
the planning study will be undertaken.
4.1. Initial Analysis
The purpose of the initial analysis is to better understand
the system (NRS, SES and AIS), its problems, manage-
ment objectives and possible measures that can be taken
to improve it (see Figure E.4). The analyst has to structure
the activities and indicate the consequences of consider-
ing or neglecting various aspects. Brainstorming sessions
with various specialists and stakeholders should help to
identify and balance the different aspects of the study.
The aim of these activities is to ensure that the analyst has
the same perception of what needs to be done as the
decision-maker and stakeholders.
4.1.1. Inventory of Characteristics, Developments
and Policies
The initial analysis starts with an inventory of the
characteristics of the WRS. This is not an easy task because
it requires the reduction of a complex reality into a
comprehensible description of system components.
Choices have to be made about what (the detail that)
should be included and what can be ignored. Such choices
require engineering and economic skills in combination
with a good understanding of the problems and possible
measures.
The next step will be an inventory of the activities and
ongoing developments that will determine how the system
will function in the future and what kind of additional
activities can be expected. This can include autonomous
developments (such as population and urbanization
growth) as well as policy decisions that have been or may
be taken that will influence the WRS and/or its perform-
ance. An inventory of policies and institutions is needed
to identify who is involved in the management and
development of the system (and hence who should
be involved in the analyses) and what their plans are.
wrm_appE.qxd 8/31/2005 12:11 PM Page 655
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
A good problem analysis will also indicate the meas-
ures that can be taken to eliminate, reduce or alleviate the
identified problems. The problem analysis in Box E.2
includes such measures. The identification of measures
does not only help to clarify the problems and possible
solutions; such early identification is also needed for the
design of the computational framework and the data
collection activities (see Section 5.1). These activities
should be designed in such a way that the measures can
be evaluated in the analysis phases of the study.
4.1.3. Objectives and Criteria
An essential activity in the inception phase is the
translation of general objectives, as described in policy
documents, into operational objectives that can be
656 Water Resources Systems Planning and Management
assessed in a quantitative way. Some examples of
objectives and criteria are listed in the Box E.3. The objec-
tives and criteria used in a water resources management
study in West Java, Indonesia are presented as an
illustration at the end of this appendix.
To incorporate sustainability as an objective in the study,
attention should be given to all aspects of sustainable
development. The decision-makers and analysts must be
aware of how sustainable development aspects can be incor-
porated in the water resources planning study. Criteria
should be laid down to assess the sustainability of develop-
ments and measures that may contribute to sustainable
development should be identified.
4.1.4. Data Availability
The last activity in the initial analysis is to investigate
the availability of data and other information required for
the study, in particular for the description of the NRS
(quantity and quality) and the SES. The availability of
data determines the level of detail and accuracy that can
Box E.2. Example of Problem Analysis
A reservoir is in operation for power production and
water supply to an agricultural area in a coastal zone.
Farmers in the area are complaining that the water
supply is inadequate. As the general objective for
water resources planning is to develop the availability
of water resources to obtain maximal economic
benefits, an analysis is required.
A first analysis of the available information shows
that the water demand for irrigation in August cannot
be satisfied because the farmers have shifted to
crops with a higher water consumption and the irri-
gated area has been extended. The constraint (bot-
tleneck) is the capacity of the feeder canal to the
area. If the capacity of the feeder canal is increased,
the reservoir may become empty in dry years, unless
the operation of the reservoir is modified. The use of
groundwater could be considered as an alternative to
surface water irrigation. Possible measures are:
increasing the capacity of the feeder canal
improving the operation of the reservoir
increasing the withdrawal of groundwater from
the coastal aquifer
reducing the agricultural water demand in
August.
A practical criterion to compare the effectiveness of
the measures is their net benefit. Other factors, like
the effect of groundwater withdrawal on saltwater
intrusion or the loss of operational flexibility of the
reservoir, may also have to be considered in the
assessment of alternative solutions.
Box E.3. Examples of Objectives and Criteria
General objectives:
- to develop the potential of the water resources
in such a way that the expected value of the
(net) benefits to the national and regional
economy will be maximized, while minimizing
negative impacts on public health, welfare and
the environment
- to provide good drinking water to the population.
Operational objectives:
- to meet the given demands for public water
supply (for population and industry) in a cer-
tain year
- to realize a certain level of agricultural pro-
duction in a given time
- to meet certain predetermined water quality
standards.
Criteria:
- cost-benefit ratio
- amount (or value) of agricultural production
- amount of energy production
- area of aquifer lost to saltwater intrusion
- changes in groundwater levels
- environmental standards (BOD, DO, etc.).
wrm_appE.qxd 8/31/2005 12:11 PM Page 656
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 657
be achieved in the analysis. If few data are available, a
more qualitative analysis may have to be performed.
The required level of detail will primarily be
determined by what is needed to analyse the problems,
objectives and measures involved. Data availability might
be a constraint, but if abundant data are available they
need not be all used (by applying very detailed computer
models, for example). The relevant problems and
measures might very well be addressed by a less detailed
approach or even a qualitative analysis.
On completion of the initial analysis, the analyst
should have a clear idea about what will be studied, for
what purpose and under what conditions. The informa-
tion related to those questions will be documented in the
second part of the inception phase, the phase that defines
the scope of the study.
4.2. Specification of the Approach
Once it is clear what will be analysed and why, it is the
task of the analysts to specify how this will be done.
Figure E.5 highlights the elements involved. The speci-
fication starts with the steps that will be followed in
the analysis, enabling the stakeholders to follow the
process. Next the system will be described, laying out
what will be analysed and what will not. To assess the
effects of measures or combinations of measures, a com-
putational framework is required. The analysis condi-
tions specify the assumptions and conditions under
which the analysis will be performed. All required activ-
ities will be combined in a work plan. Some important
elements in the specification of the approach are dis-
cussed below.
4.2.1. Analysis Steps
The specification of the analysis steps has both an internal
(to the project) and an external function. The internal
function is that it will tell the various (often interdiscipli-
nary) team members what the position of their contribu-
tion is in the overall analysis approach. This is necessary
to schedule activities of all the members and to ensure (or
at least enhance the probability) that they will provide the
right kind of information to each other at the right times.
The external function relates to the interaction with the
decision-makers and stakeholders; it specifies when and
how they will be involved in the analysis.
The specification of the analysis steps can be based on
Figure E.3. This generic figure should be made more
specific by addressing the particular aspects of the system
and problems under consideration.
4.2.2. Delineation of System
The next step is to determine the components that will be
taken into account in the analysis. The boundaries of the
(sub)systems need to be determined on the basis of their
natural boundaries, physical in case of the NRS, adminis-
trative in case of AIS, etc.
Required Level of Detail
An often difficult aspect to address in the inception
phase is the level of detail that should be chosen for the
components and subsystems of the system under study.
To comply with the overall purpose of the study the
preparation of a plan that addresses the problem and
provides the best use of goods and services provided by
the WRS one does not have to know the use of every
square metre of land in the study area or the water use of
-
Figure E.5. Components of the inception approach.
wrm_appE.qxd 8/31/2005 12:11 PM Page 657
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
each leather tannery in the area. One has to use values
averaged over areas and groups of users with just enough
detail to provide the information required to support
decision-making. The information is used to discriminate
between alternative courses of actions and has to be
relevant for the evaluation process. It should thus
be expressed in terms of the selected evaluation criteria.
One of the main tasks of a project leader is to manage
the experts from various disciplines in their enthusiasm
to find the answers to all possible questions. Not staying
focused on the appropriate level of detail is one of the
most common causes for project failure. If the needed
level of detail is underestimated at the start of the proj-
ect, the study has to go into more detail to be able to
solve the problem. Sometimes the right level of detail is
chosen, but team members spend too much time
addressing more detailed questions and fail to come up
with answers within the available time. Changing the
level of detail is one of the main reasons for feedback
loops in the analysis process.
The level of detail has to be specified for the following
aspects:
Spatial disaggregation of the study area: type and size
of units. The spatial disaggregation usually depends
on the types of water demand, the system of rivers and
canals, and the characteristics of groundwater aquifers.
Sectoral disaggregation of water users. The ISIC
classification (International Standard Industrial
Classification) distinguishes economic activities in
various sectors. At the two-digit level for example,
agriculture is treated as one sector. At more detailed
levels the agricultural sector is broken down in smaller
subsectors such as aquaculture, dairy farming and
cotton growing.
Time steps for the analysis of processes. It is important to
select the most appropriate time scale. One generally
has to consider various time steps. In a system with
reservoirs with over-year storage, one should run a
sequence of years to study the performance of the
reservoir. For the analysis of the agricultural sector one
often has to account for seasonal variability.
Quality of the environment. Not all species can be used
as indicator species. A choice has to be made to show
the impacts of measures on the quality of the environ-
ment and the ecological integrity.
658 Water Resources Systems Planning and Management
4.2.3. Computational Framework
The results of the preliminary analysis will define the
computational framework that is needed for the analy-
sis. This framework comprises mathematical models,
databases, GIS and the like. These must describe the
system (NRS, SES) and evaluate possible measures and
strategies under different scenarios. (The different kinds
of models and approaches that can be used are
described in the main text of this book.) In general, a
combination of simulation and optimization models will
be useful.
For the development of the computational framework,
the study area may be subdivided into smaller units
considered to be homogeneous with respect to their
characteristic parameters. Each unit is represented in the
mathematical model(s) by a computation element. The
number of elements required for the analysis depends on
the issues being addressed, the complexity of the study
area, the kind of measures to be studied and the
availability of data. It generally is wise to start with a
preliminary schematization with the minimum number of
elements. If more spatial detail is required in a later stage,
elements can be subdivided.
Use/Adaptation/Development of Computer Programs
Computer programs may be used to analyse various
aspects of the WRS. A decision should be made about
whether to use existing programs or to adapt or develop
new ones. As development of computer programs is
a very time consuming activity, existing ones should
be used whenever they meet the requirements of the
study.
Hydrometeorological Boundary Conditions
Boundary conditions for models of the NRS comprise
precipitation, evapotranspiration and flow across the
model boundaries. Parts of the NRS that are not influ-
enced by water management (natural catchments) are
usually not included in the computational framework.
The discharges of these (sub)catchments are derived
from historical streamflow data. If the availability of
streamflow data is insufficient, rainfallrunoff models
may be used to generate streamflow records from
precipitation data.
wrm_appE.qxd 8/31/2005 12:11 PM Page 658
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 659
Three approaches may be used to account for the
stochastic character of the hydrometeorological conditions
for which water resources strategies have to be evaluated:
a. Simulation of a strategy for a complete set of
hydrometeorological data, e.g. forty consecutive years,
including dry and wet periods.
b. Simulation for a selected sample of the available data,
each one representative for a certain situation, such as
an average year, a 10% dry year or a 2% dry year.
c Simulation of just one representative year, such as an
average year.
Which approach is preferred will depend on the specific
problem situation, variability of the hydrometeorological
conditions, availability of data and the specific analysis
that is required. It is clear that in cases where substantial
over-year storage in reservoirs (surface water and ground-
water) is involved, approaches b) and c) are not viable.
Approach c) is only applicable in situations with very lit-
tle variability over the years. Approach b) is often a good
compromise. Figure E.6 illustrates the approach. The
simulations are performed for three representative years:
an average year (D
50
), a moderate dry year (D
10
) and an
extreme dry year (D
5
). The expected damage in this exam-
ple is approximated with the following formula:
Expected damage 0.125
*
D
5
(0.350 0.125)
*
D
20
(0.650 0.350)
*
D
50
In the development phase, when many alternative
measures (solutions) have to be screened, simulations will
be made for selected conditions only. For screening
(see Section 5.2.3), the system is generally studied under
stress conditions, for example a 10% dry year. The
minimum time interval taken to represent a certain
situation depends on the memory of the system. In a
system comprising reservoirs with considerable over-year
storage, representative periods should be more than a
single year. In the selection phase, simulations will be
made for promising strategies only, so that the complete
set of historical data can be applied to them.
Requirements for Data Collection
The analytical approach described above is usually
formulated in an iterative way, taking into account data
requirements, data availability and available resources
(budget, manpower, models and analytical techniques).
When the approach has to be modified in view of
constraints (usually budgetary), the data requirements are
reassessed. Once the analytical approach is specified, the
exact requirements for data collection can be formulated:
which data need to be collected, when, and by whom.
Opportunities to collect additional data from field
observations will generally be limited by the available
budget and time.
4.2.4. Analysis Conditions
The assumptions and conditions under which the
analysis is undertaken have to be specified in close coop-
eration with the decision-makers. These assumptions and
conditions include:
(Sub)systems boundaries. Definition of the WRS and a
possible subdivision in sub-systems are based on
physical (NRS) and/or administrative boundaries (AIS,
see also Section 4.2.2).
Base year. The analysis will be made for the present
condition and one or more points in time. The present
situation may be defined by the situation in the most
recent year for which a reasonably complete set of data
is available (the base year).
Time horizons. These will be selected on the basis of
existing planning horizons usually ten and twenty-
five years or be related to the characteristic time
scales of subsystems.
frequency of occurrence (e.g. flood, drought)
e
s
t
i
m
a
t
e
d

d
a
m
a
g
e
5 20 35 50 65 80 95 12.5
D
5
D
20
D
50
E
0
4
0
7
2
9
a
0
Figure E.6. Expected damage calculation based on three
representative years.
wrm_appE.qxd 8/31/2005 12:11 PM Page 659
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Scenario assumptions concerning factors external to
the WRS, such as the growth of population, food con-
sumption and of water-related economic sectors and
prices (energy or crop prices, for example). In many
cases an analysis of socio-economic data will be
required to obtain reasonable ranges for scenario vari-
ables (see Section 5.1.2).
System assumptions. These concern factors internal to
the WRS, such as the response of crop production to
improved cultivation practices, or the effectiveness of
price incentives on per capita water consumption.
These system assumptions can be subject of additional
(sensitivity) analysis in the selection phase (see
Section 6.3).
Time and budget constraints. The study has to be
executed within constraints of available time and
budget.
Data availability. The study must be based on the data
that can be made available within the time and budget
available.
The choice of the time horizon is often given insufficient
attention. Official planning horizons (e.g. ten and
twenty-five years) are often used as time horizons for all
elements of the analysis. As well as the planning horizon
of the traditional government planning framework,
however, one should also consider the time scales of the
system and the processes within it. System components
will have characteristic time scales. For example:
User functions (economic activities) have life cycles
that are determined by the amortization period of the
investments. The time horizon of the planning process
is based on these conditions.
Social institutions have time horizons that depend on
the pace of legal/institutional and political decision-
making.
Physicalchemical systems have time scales that
depend on the response or restoration times of the
systems. Restoration of polluted rivers, for example,
may be achieved within a few months, while the
restoration of a polluted groundwater aquifer may take
decades.
Ecosystems may have a time scale of a few weeks
algae blooms) or tens of years (degradation of man-
grove forests), depending on the type of process or
intervention.
660 Water Resources Systems Planning and Management
To study the sustainability and ecological integrity of the
resource system, time horizons should be tuned to the
response times of the system rather than to a planning
horizon only. Although more attention is now paid to
sustainability, no operational procedure has been
developed to give long-term effects their proper place in
the evaluation process. Decision-makers tend to take
short-range decisions, involving possible risks in the long
term, because their political jobs are often limited to (or
renewable in) short terms and hence they prefer
short-term political gains.
4.2.5. Work Plan
The results of the inception phase are documented in an
inception report (Figure E.3), which will be used as a
reference during the execution of the study. An essential
part of the report is the work plan, in which time, budget
and human resource allocations to various activities are
specified. This work plan typically includes bar charts for
activities and staffing, time schedules for deliverables,
milestones, reporting procedures and similar features.
It should include a communication plan that describes
the interaction between the decision-makers and
stakeholders and the analysis team.
4.3. Inception Report
The inception report has to contain all the results of the
inception phase. It should make clear what will be stud-
ied, why and how. In many cases it will also specify what
will not be studied and why. The content of the inception
report follows the subjects mentioned above under the
initial analysis and approach:
Description of the water resources system
NRS: components
SES: activities and developments
AIS: institutions involved, legal aspects including
a delineation of the components that will be
studied.
Problem analysis and problem statement.
Objectives and criteria.
Analytical approach (steps, models and databases).
Analysis conditions (time horizon, scenarios, assump-
tions and so on).
Work plan.
wrm_appE.qxd 8/31/2005 12:11 PM Page 660
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 661
4.4. Communication with Decision-Makers and
Stakeholders
The inception report is a specific and concrete result of
the inception phase. It is an important product because it
contains all that has been learned in this first phase and
that has been agreed upon between the analyst and
the client (the decision-makers and the stakeholders).
A possibly even more important result of the inception
phase, however, is the interaction that should take
place during this phase between the analyst and the
client. The clients views about problems, objectives and
other aspects should develop alongside the view of the
analyst. The analyst must understand the clients con-
cerns, problems and objectives and not just his or her
own. Clients should feel they own the results of the
inception phase and view the inception report as their
own product, not merely a report of the analysts or
consultants.
To achieve such ownership, frequent interaction must
take place among the analysts, the decision-makers and
stakeholders, to a much greater extent than is indicated
in Figure E.3. This can be done in specific workshops,
such as those devoted to the problem statement or to the
specification of objectives and criteria, public consulta-
tion meetings.
5. Development Phase
The development phase includes simultaneous model
development and data collection followed by a preliminary
analysis to identify possible solutions for the problems
being addressed.
5.1. Model Development and Data Collection
In the development phase, a coherent set of computa-
tional tools is developed for the analysis of the WRS and
SES, as specified in the inception phase. Possible solu-
tions for the water resources problems are also identified
in this phase. The result of the data collection and
modelling activity is a computational framework that
represents the characteristics of the WRS at an adequate
level of detail. The framework is designed to quantify
the effects of individual measures or combinations of
measures, expressed in values for the evaluation criteria
that correspond to the objectives of water resources
management. For the schematization of the study area,
experience from previous applications may be helpful.
However, each study is different and the schematization
should always be tailored to its specific requirements.
If computer programs have to be developed from
scratch, or if existing computer programs have to be
adapted in a significant way, a considerable effort may
be required and the activity may consume a large part
of the available budget and time. Careful selection of the
phenomena to be represented by the models, tuned to the
needs of the analysis, can result in considerable savings.
In the course of the modelling activity, more information
on the study area and the type of measures to be consid-
ered may become available. Sometimes this may lead to
a simplification of the modelling activity. The models
should therefore be flexible and adaptable to new
information. Activities related to data collection andmod-
elling will comprise all three components of the WRS (see
Figure E.7). The emphasis is generally on the NRS, and to
analyse this, a distinction can be made between physical,
chemical and biotic components.
-
Figure E.7. Components of data collection and analysis.
wrm_appE.qxd 8/31/2005 12:11 PM Page 661
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
5.1.1. Analysis of the Natural Resources System (NRS)
Physical Components
The NRS comprises the natural and human-made infra-
structure, including the hydrometeorological boundary
conditions. Models can be used to simulate the processes of
water distribution through the infrastructure, taking into
account the storage of water and withdrawal from
the network to satisfy the demands of the water-using
activities. (Chapter 11 provides an overview of possible
models involved.)
The main water allocation related functions of the
surface water system are storage of water in reservoirs
and diversion of water from the rivers and main canals to
branch canals. Typical time steps in planning studies are
periods of a week, ten days, fifteen days or a month. In
most river basin simulation models (e.g. RIBASIM,
WEAP, MIKEBASIN), the surface water system is
represented by a network of nodes and branches. At the
nodes, natural runoff can be supplied to the system or
water can be abstracted to satisfy specified water demands.
Storage is represented by reservoir nodes, where reservoirs
are operated according to user-specified rule curves.
Facilities are available to account for power generation,
diversion of water to branch canals, return flow from irri-
gated agriculture, drainage from districts, minimum flow
requirements and so on. The allocation of water to the
demand nodes is in upstreamdownstream sequence,
accounting for user-specified priorities. Most river basin
models contain algorithms to include simple flow routing
in the analysis. If necessary, more detailed hydrodynamic
models such as MIKE-11 or SOBEK can be used.
Hydrometeorological data are analysed to provide
time series of inflows to the surface water system. Typical
elements of the analysis are: historical streamflow data,
rainfall data and rainfallrunoff relations. The reliability
and availability of historical streamflow data are often
insufficient for the purpose of the study. Streamflow data
may then have to be supplemented with synthetic data
derived from observed precipitation and rainfallrunoff
relations. A typical period for which data would be
required is twenty-five years. Chapters 7, 11, and 13
introduce tools for an analysis of hydrometeorological
data and rainfallrunoff simulations.
A main parameter characterizing the groundwater
system is the groundwater level or head. In coastal zones,
662 Water Resources Systems Planning and Management
where seawater intrusion may occur, the salinity of the
groundwater has to be considered as well. The time scale
for effects of withdrawals on the groundwater head is in
the order of ten to fifty years; saltwater intrusion is even
slower and effects may have a time scale of a century or
more. For such computations, steady flow conditions are
often assumed and a comparison is made of the effects of
different abstraction alternatives. Important aspects in the
analysis of the groundwater system are the identification
of the system of aquifers and aquitards and the determi-
nation of the recharge to the aquifers. Groundwater
models can be used to test the consistency of collected
data. Use of models to quantify the effects of future
groundwater abstraction is only suggested in cases where
there is sufficient information to calibrate the models for
reproduction of the existing situation. Many general
groundwater programs (e.g. MODFLOW) are available to
compute the groundwater heads in a system of aquifers.
Chemical Components
The analysis of chemical components in the water system
is used to study the influence they have on the user func-
tions or the biological system. The components and
processes that are to be considered in the analysis will
have been selected in the inception phase. Their concen-
trations are computed in discrete segments of the surface
or groundwater system, and this will require data on
the influx of substances across segment boundaries and
on the chemical processes in the segments. (Chapter 12
gives more detailed descriptions of available modelling
approaches.)
The results of the water quantity modelling may be
the inputs for water quality models, and are used to deter-
mine the water balance for each segment. To assess the
mass balance of the chemical components, data on con-
centrations in the inflows to the segments are required. In
addition to the inflow terms in the water balance, some
extra terms have to be specified. For example, the dis-
charge of sewage water may be negligible for the overall
water balance but have a significant impact on the water
quality of a river segment.
Biotic Components
The analysis of the biological system aims to determine
the response of the ecosystems to water resources
wrm_appE.qxd 8/31/2005 12:11 PM Page 662
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 663
management. Because there is too little exact information
on the biotic components in most cases, and because
ecosystems largely depend on habitat parameters, the
analysis may have to be limited to the essential relation-
ships between ecosystems and physical and chemical
parameters. The selection of key factors for ecosystem
performance depends on the type of ecosystem, the kind
of analysis for which they will be used and the type(s) of
activities or human interventions involved.
For example, consider the impacts of dam construc-
tion on downstream ecosystems. Key parameters include:
hydrodynamic regime (discharge characteristics and
flooding pattern, such as duration, depth and frequency)
water quality (e.g. sediment load, temperature,
oxygen, salinity).
It is likely that these parameters will change due to the
dams construction, which may thus affect the downstream
ecosystems. A combination of information on the present
ecosystems and physical conditions and the direction
and order of magnitude of change in the parameters
mentioned above often enables one to make an expert
judgment on the significance and direction of change of the
ecosystems.
However, if one wants to make a more accurate and
detailed prediction of ecosystem changes, for example by
using mathematical modelling, much more information on
the conditions determining fluvial ecosystems is needed.
For example, a vegetation development model of the
downstream areas liable to flooding needs in addition to
the key parameters mentioned above information about:
local topography (relief, dikes and culverts, channels,
oxbows etc.)
local geomorphology or soil types
vegetation patterns.
Using known relationships between vegetation and
physical factors such as the inundation frequency and
soil/geomorphology, a preliminary model can be made
that predicts changes in the vegetation pattern caused by
changes in river flooding behaviour. Such an approach
can be greatly enhanced by the use of geographically
oriented data, in the form of digitized maps of the topo-
graphy of the area, the soils, inundation and vegetation.
For aquatic habitat types, the approach may be quite
different. Aquatic flora and fauna in rivers and lakes are
affected by such factors as flow velocity, shear stress, water
depth, bottom type and water quality. The use of one-
dimensional hydraulic and water quality models is there-
fore often indispensable, but needs a concerted effort
between ecologic and hydraulic modelling experts. It is
especially important to identify data requirements as early
as possible. In most cases, special physical or chemical
modelling calculations are needed to supply the required
data for habitat prediction. For example, hydraulic model-
ling often concentrates on high and low water levels,
whereas floodplain ecosystems also depend on intermedi-
ate levels and duration curves. Time steps may differ and
additional detail may be required in the model.
In most cases, predictions of ecosystem changes have
a relatively large margin of uncertainty. Causes of this
uncertainty include the often limited availability of data
and the gaps in our fundamental knowledge of how
ecosystems really work and of their complexity and
dynamic behaviour. Creating detailed descriptions of
species composition remains very difficult, especially
when new species may appear. These factors also make it
very difficult to assess the significance of the changes.
The evaluation of ecosystems generally requires informa-
tion on species diversity and the occurrence of rare or
endangered species, and this is often lacking.
The significance of the effects of water management on
ecosystems is increasingly acknowledged. Specific tools
for predicting these effects are being developed, mostly on
a regional basis. For example, models can be used for the
prediction of vegetation development in relation to
inundation characteristics, soil type and nature manage-
ment practice in floodplains, or in relation to ground-
water management. For biological water quality, models
are widely available, including models for algal composi-
tion and algal bloom (Chapter 12).
Scenario Development
Depending on the scenarios defined in the inception phase,
it might be necessary to develop scenarios for the natural
resource system. In nearly all cases, these willcertainly be
needed to take into account possible climate change that
may affect the physical components (rainfall, flow, etc.), and
very probably also the chemical and biotic components.
Scenarios for the natural system are also needed when
the system boundaries, as defined for the analysis, do not
wrm_appE.qxd 8/31/2005 12:11 PM Page 663
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
cover the full river basin and, for example, developments
upstream might influence the inflow into the system. This
is typically the case in national studies where the system
boundary has been set at the national border.
5.1.2. Analysis of the Socio-Economic System (SES)
Developments in the SES determine the way demands on
the NRS will develop. Conversely, the development of
economic activities within the study area may depend on
the availability of water. For example, good supplies of
relatively cheap surface water may stimulate the develop-
ment of irrigated agriculture, or attract industrial activities
that require large quantities of water for their production
processes. Another example is the development of water-
based recreation around a reservoir. These socio-economic
system developments in turn increase the water demands.
It is the task of the economist or planner to estimate
the future levels of the activities that require water
resources for their development, as well as the resulting
water demands (and discharges). The estimate is made for
the time horizon of the study, within the geographical
boundaries of the area. The main activities are:
assessment of the present economic situation
activity analysis
scenario specification.
Assessment of the Present Economic Situation
The starting point for an analysis of the SES is an
assessment of the present economic situation with respect
to the water-related activities and the factors that
determine these activities. Past trends should be analysed
to provide information on factors that have been decisive
in bringing about the present situation and that may give
clues about future developments. The analysis should
focus on present and future major water use categories,
which are identified at an early stage in the analysis,
rather than trying to estimate water demands and
discharges for all possible water use categories.
Ones attention, then, should be on the most impor-
tant factors that determine relevant water-related activi-
ties rather than on an analysis of the total economy. The
difficulty in forecasting economic developments is, how-
ever, that one does not know a priori which factors will be
decisive for these developments. As population dynamics
664 Water Resources Systems Planning and Management
are explanatory factors for many economic developments,
demographic analysis (population growth, where and at
what rate) is generally one of the first activities in this
phase. Other important parameters for the development
of water-related activities are:
sectoral distribution (primary, secondary and tertiary
sectors)
industrialization
labour force and wage development
the price of oil and other commodities
the balance of trade (imports and exports)
the stability of local currency.
In the analysis of the present economic situation, attention
should be given to the possible effects of changes in the
WRS on the regional or national economy. To determine
how effects related to water users may influence major
economic indicators such as gross national product, balance
of trade or equity considerations, a macro-economic analy-
sis of the relation between the water-related sectors and the
rest of the economy and its surroundings (e.g. foreign trade)
is often required.
An example of a macro-economic model that could be
used for such an analysis is an inputoutput model. Such
a model shows the relationships between various sectors
of the economy. Using an inputoutput model, one can
analyse how effects generated in one sector influence
other sectors of the economy, not only in terms of value
added, but also in terms of labour requirements. Such
economic models have been extended to include environ-
mental and resource-use aspects (national resource
accounting models). Inclusion of these aspects, however,
requires a rather large effort (in terms both of data and of
manpower) and should therefore only be considered
when major effects on the national economy are expected.
Activity Analysis
The activity analysis focuses on the relation between the
economic activities and their water use. It focuses on the
relations that determine the type and amount of water used
by various activities. An activity analysis must answer the
following questions with respect to each identified activity:
What are the amounts of water (quantity and quality)
demanded during which periods of the year and at
which locations?
wrm_appE.qxd 8/31/2005 12:11 PM Page 664
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 665
What are the amounts of water discharged and the
pollution loads during which periods of the year and
at which locations?
What are the benefits to the user if these amounts are
made available?
What is the damage to the user if these amounts cannot
be made available?
Which costs can be recovered by having the user pay
for the water?
Will cost recovery influence the water use pattern of
the water user (both at the intake and the discharge
side of his activity)?
Estimates of future water demands and wastewater dis-
charges are generally based on the determination of unit
water use and wastewater discharge per unit of activity.
Looking at combined trends in unit water demands and
economic developments can provide insight in possible
future water demands. The final result of the activity
analysis should be water demand functions, relating water
demand to exogenous parameters.
As well as the level of activities and the resulting
water demands, the geographical location of water-using
activities (the pattern of activities) must also be known or
estimated. If the pattern of activities is not expected to
change, the analysis can be focused on the present situation
in the study area. If new activities are expected to develop
within the study area, it may be necessary to analyse the
water-use characteristics of similar activities in other
regions.
The resulting water demands should not be consid-
ered as given. Water-use coefficients can be changed
through measures such as water pricing that aim at
reaching a socially preferred use pattern. Technological
developments may result in less water use and pollution
load per unit of product. If supplies and demands are
matched before the effects of such incentives are analysed
then one may end up with an overcapacity, because the
given demands may be lower if water users are con-
fronted with the costs of the use of the water resources.
This type of feedback needs to be considered in the study.
Estimating benefits and costs related to water use
provides information to evaluate water resources manage-
ment strategies. One of the criteria in this evaluation is the
net economic benefit of the measures for various water
users. One needs to know what benefits water users
derive from the use of the water resources, and what costs
they have to pay in order to obtain these benefits. The
activity analysis should produce therefore benefit and cost
functions related to various water uses. These functions
can be used to evaluate the proposed strategies.
Scenario Specification
Because of the multitude of factors that determine water
demand, perfect forecasting of economic development
and resulting water demands is a utopian dream. Future
water demands are often dependent on future scenarios.
A water demand scenario is a logical combination of basic
parameters of the economy, their effects on water-related
activities, and the resulting water demands. An under-
standing of the functioning of the socio-economic system
developed through the assessment of past and present
trends should be used to formulate a limited number of
consistent scenarios.
For analytical purposes, it is useful to define an
autonomous development scenario (i.e. the development of
the socio-economic system) and its water demands, as it
would develop independently from changes in the WRS. An
autonomous development scenario does not consider possi-
ble constraints in the supply of water, and is thus very use-
ful for analysing bottlenecks in the water resources system.
In a regional water resources management analysis,
national economic growth and technological develop-
ment are usually taken as exogenous parameters for the
autonomous development of the region. Regional
measures can however influence water use and may divert
economic activities away from the area of study, or even
influence the growth of the national economy, which was
originally seen as an exogenous parameter.
Since many uncertain factors influence the develop-
ment of water demands, the future cannot be projected by
using a single autonomous development scenario. Rather
than constructing many alternative scenarios through per-
mutation of the uncertain parameter values, one may have
to consider only two or three different development paths.
Scenario building is not a straightforward activity; it is
partly an art that depends largely on the views of the
analysts and local experts about which trends are likely to
prevail in the future. The developed scenarios have to be
internally consistent with respect to the magnitude
and direction of the developments. Exogenous variables
(population growth, economic growth) and water-related
wrm_appE.qxd 8/31/2005 12:11 PM Page 665
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
variables (for example per capita water consumption) are
combined into logical, coherent water-demand scenarios.
An increase in per capita income could cause an increase
in per capita water consumption. A scenario which
presumes high economic growth should therefore also
presume high per capita water-use coefficients.
Scenarios are often made top-down: that is, first one
projects the basic parameters of an economy (population,
gross national product, imports and exports, etc.) for
the whole country or large parts of it. The economic
developments projected at the national level are then
broken down for regions, usually using past trends in
regional economic growth related to the national average
growth trend. These regional projections should be in
line with population projections, because economic
development also depends on population growth. One
should use information available at the local level to
check the assumptions made in disaggregating national
projections.
5.1.3. Analysis of the Administrative and
Institutional System (AIS)
An analysis of the AIS is required for three reasons:
The information required for the study has to be
obtained from the agencies involved.
The legal framework may impose certain constraints
on the development of water-using activities.
666 Water Resources Systems Planning and Management
The implementation of measures may be constrained
by institutional bottlenecks.
In the inception phase, a preliminary analysis is made
of the institutional setting of water resources manag-
ment. The responsible authorities and institutes in
the water resources sector and water-related sectors are
identified at the national, regional and local level. In the
analysis phase, the responsibilities and instruments of
the authorities are reviewed. Attention must be given to
the interaction between various authorities involved in
water resources management and to the effectiveness of
the AIS.
The responsibilities and authorities of various agencies
are generally described in national legislation. An analysis
of the legal framework should identify the various agencies
responsible for (aspects of ) water resources managment,
their authority, their instruments and their coordination
with other agencies. Arrangements made in the past con-
cerning the use of water (water rights) should be carefully
studied, since these may significantly constrain the options
for water resources development.
Water resources management studies are often limited
to the preparation of policies for a certain agency. In this
situation, the analysis of the AIS will mainly serve to iden-
tify measures that can be implemented effectively by that
agency. The responsible agency should be aware of the
possible solutions and the role they may have in solving
the managementproblem. Sometimes, the analysis of the
AIS may result in recommendations for institutional and
legal changes.
5.1.4. Integration into a Computational Framework
The various models and components developed for the
NRS and SES describe parts of the total system. Some
models may produce output that is needed as input for
another model. Some may run interactively. This means
that all models will have to be linked in a computational
framework. Depending on the models and the problem
situation, such linkage can be very strict and formalized
in a kind of decision support system. In other cases, a
clear description of information flow from one model to
the other may be sufficient. In optimization models the
linkage of the various components is taken care of by the
structure of the model.
Box E.4. Example
The development of water demand in agriculture is
autonomous only to a limited extent, as it depends
largely on the availability of land and water
resources. The demand for agricultural products,
however, will develop in an autonomous way. If the
availability of water resources in a region is limited,
the autonomous development of agriculture will be
limited as well, and one would predict a small
increase in agricultural water demand. If the demand
for agricultural products increases considerably and
self-sufficiency in food production is an objective,
then the desired agricultural development to meet
this objective may be considerable. The water
demand corresponding to this desired agricultural
development will show the need for further
development of the water resources in the region.
wrm_appE.qxd 8/31/2005 12:11 PM Page 666
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 667
Figure E.8 gives an example of a computational frame-
work in which various simulation models are combined
to analyse a river basin under drought conditions. The
reservoirs in the system require special attention for the
quality of the water they contain, sedimentation and
the optimization of hydropower generation. The core of
this computational framework is formed by the core
models block in the upper right corner of Figure E.8. In
this block the demand for water is determined, followed
by a balancing of supply and demand and the determina-
tion of the implications of the water allocation for the
use functions. The models and modules in this core will
typically be combined into one unit with automated
linkages between the models and modules. The other
models are linked through file transfer. This applies to
the required input on macro-economic and hydrometeo-
rological conditions (generated by scenarios) as well as
the side analysis with respect to the modelling of the
sedimentation and water quality in the reservoirs. The
last parts of the computational framework are the mod-
ules that determine the financial and economic aspects
(investments, operation and maintenance, benefitcost,
etc.) and support the overall multi-criteria analysis. At
various places in this computational framework, measures
can be included by changing input parameters. Scenarios
can be analysed by changing the macro-economic and
hydrometeorological conditions.
Figure E.8 is just an example. Other problem situa-
tions may require different computational frameworks.
Computer modellers and analysts have a tendency to
make these computational frameworks rather compli-
cated and fancy. Sometimes this complexity is necessary,
but in most cases a simple computational framework will
be sufficient and preferable. The basic approach is to start
as simple as possible and only add details when they
prove necessary to carry out a proper analysis.
E
0
4
0
2
2
3
e
demand
determination
water balance
and distribution
core models
determination
impacts
agriculture
aquaculture
DMI-demand
district model
river basin
model
macro-
economics
input / output
model
population and
urbanisation
models
water quality evaluation
investments
(B/C)
hydro-power water quality
models
stratification
model
multi-criteria
evaluation
energy
optimization
agriculture
aquaculture
DMI-demand
hydrological
database
rainfall runoff
surface water
models
groundwater
models
hydrology /
hydraulics
reservoir
sedimentation
morphology
erosion model
erosion
measures
measures measures scenarios
Figure E.8. Typical computational
framework of simulation models.
wrm_appE.qxd 8/31/2005 12:11 PM Page 667
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
5.2. Preliminary Analysis
The preliminary analysis includes the application of the
knowledge obtained from data collection and modelling
(Figure E.9) to identify possible solutions for the
water resources management problem. At the start of
the development phase, the analysis is focused on
problems in the present situation (base case analysis) and
the future situation (reference or bottleneck analysis).
Knowing the problems that are to be expected under
certain conditions, measures to solve them can be
defined.
Water resources development should be guided by
the principle of sustainable development. It should lead
to a sustainable economy, in which natural renewable
resources are used without being depleted (in either their
carrying capacity or environmental utilization space).
Sustainable economic development can also be supported
by the development and adaptation of knowledge and
improvement of organizational and technical efficiency.
Many alternative measures can be identified in general,
each making a contribution to the (partial) solution of a
particular problem. To avoid spending time on less effec-
tive measures, the effects of individual measures should
668 Water Resources Systems Planning and Management
be compared and promising measures selected for further
analysis. Preliminary versions of the computational
framework can be used to get a first idea of the effects of
different measures. During the development phase, when
it becomes clear which types of measure are promising
and which elements in the computational framework
are critical for the assessment of their effects, the
computational framework can be modified.
5.2.1. Base Case Analysis
In the inception phase, a provisional problem statement is
formulated on the basis of the results of the initial analy-
sis. In the development phase, computational tools are
used to further elaborate the problem statement. This is
done by specifying two cases: the base case and the
reference case.
In the base case analysis, the performance of the WRS
is studied for the infrastructure and water demands in
the base year, which is the most recent year for which a
complete set of data can be collected. The base case
describes the performance of the WRS in the present
situation. A comparison of this base case with targets as
specified in the management objectives produces a quan-
tified problem statement.
From a modelling point of view the base case analysis
also functions as a kind of calibration of the models. The
results can be compared with measured information and,
when necessary, parameters can be adjusted to better
reflect the reality. Sufficient attention should be given to
proper calibration and validation procedures.
5.2.2. Bottleneck (Reference Case) Analysis
An essential activity is an analysis of the WRSs perform-
ance if present policies are continued (no change alterna-
tive). This is called the bottleneck or reference case
-
Box E.5. Definitions
Base year: present situation.
Time horizon: future situation.
Base case: performance of WRS in present
situation.
Reference case: performance of WRS in future
situation if no additional measures are taken.
Figure E.9. Overview of activities in the preliminary analysis.
wrm_appE.qxd 8/31/2005 12:11 PM Page 668
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 669
analysis. The latter term is used here because, in a later
stage of the analysis, the effects of measures and strategies
will be determined by comparing the results of that
strategy with the reference case situation. The perform-
ance of the measure or strategy is assessed by using
evaluation criteria related to the management objectives.
For the future (the planning horizons), the reference
case comprises the present infrastructure, to which
measures are added that have already been executed
or decided, together with projected demands.The effects
of autonomous developments in the SES are accounted
for in these projected demands.
5.2.3. Identification and Screening of Measures
Once the base case and reference case have been defined,
and problems and bottlenecks identified, measures are
developed to alleviate the problems. In the identification
of measures, a distinction is made between various
categories of measure, as listed in Section 1.3.
An inventory of all possible kinds of action that can be
taken will in general result in hundreds of measures. In
most cases it will not be possible to analyse all of them in
detail. A screening process is needed to select the most
promising ones. This can be done in several ways. As
mentioned in various chapters of this book, separate
optimization models can be used to limit the solution
space. It can also be done by using the computational
framework developed for the project but limiting
the analysis to a few criteria, such as benefit and costs.
The measures with the highest benefitcost ratio will be
labelled as the most promising ones. A third kind of
screening analysis is to apply a qualitative judgment, for
example in terms of criteria effectiveness, efficiency,
legitimacy and sustainability. Box E.6 explains these
criteria.
The aim of the screening process is to get some feeling
of how likely the measures are to alleviate the present and
expected problems. No final judgment is given and no
measures will be discarded in the process.
The screening of measures is a cyclic process.
Assessing the measures will contribute to a better
understanding of their effectiveness and new ones may
be identified (comprehension loop). Combinations of
measures may be considered for specific parts of the
WRS, for instance for solving the water quality problems
in a sub-basin. The result of the screening process is a set
of promising measures that can be used for strategy
design. The whole process of base case and reference case
analysis and screening is depicted in Figure E.10.
5.2.4. Finalization of the Computational Framework
During the screening of measures, adjustments in the
computational framework may be necessary. Some com-
ponents may need to be defined in more detail than was
realized in the inception phase, while for others a simpler
approach may be justified. Additional evaluation criteria
may also be introduced to better discriminate the effects
of different measures.
Box E.6. Criteria
Effectiveness. Meaning that the measures to be
taken are those which solve the most serious prob-
lems and have the highest impact on the objectives.
Measures to prevent the problem will be preferred to
those that solve it. Similarly, measures that solve the
problem are preferred to those that only control it.
Efficiency. The measures to be taken should not meet
the explicit objectives at the expense of other implicit
objectives. The cost-benefit analysis (at the national
level) is one indicator of efficiency. An example is to
issue a law that forces industrial firms to incur the
full cost of end-of-pipe treatment. In Egypt, this
would improve the Nile-system water quality, and
thus improve health and avoid environmental dam-
age, but on the other hand it might impose high costs
to the firms, possibly resulting in loss of employ-
ment. An efficient decision may be to opt only for cost
sharing rather than full cost recovery.
Legitimacy. Measures to be included in the strategy
should not rely on uncertain legal/institutional
changes. Measures should also be as fair as possible,
thus reducing public opposition so that they will be
favoured by as many stakeholders as possible. For
instance, reducing drainage water pollution through
subsidizing non-chemical pesticides will be more
legitimate than applying penalties on excessive use
of chemical pesticides.
Sustainability. The measures to be taken are those
that improve (or at least maintain) the present envi-
ronmental and socio-economic conditions for future
generations.
wrm_appE.qxd 8/31/2005 12:11 PM Page 669
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
670 Water Resources Systems Planning and Management
6. Selection Phase
In the selection phase, promising measures are combined
into strategies. The effects of various strategies are
assessed and a limited set of promising ones is defined.
For these promising strategies, the effects are assessed in
more detail. The sensitivity of these effects to (uncertain)
assumptions in the analysis is then assessed, and finally
the results of the selected strategies are presented to the
decision-makers. Figure E.11 highlights these steps, and
the selection process is depicted in Figure E.12.
6.1. Strategy Design and Impact Assessment
Strategy design is the development of coherent combina-
tions of promising measures to satisfy the management
objectives as far as possible. As there are generally many
criteria related to these objectives, and probably many
expressed in different units, the strategy design is not a
simple process. Relations among combinations of
measures and their scores on the evaluation criteria are
usually non-linear. As there are no general methods to
compensate a lower score on one criterion with a higher
score on another, the optimum itself is not uniquely
defined. Sustainability should be considered in the
assessment of measures and strategies.
Figure E.10. Elements of the
preliminary analysis.
The design of strategies is an iterative process. One
could start by developing strategies on the basis of a
single objective like, for example, food production or
maximum net economic benefit. These strategies define
Figure E.11. Components of the selection phase.
-
wrm_appE.qxd 8/31/2005 12:11 PM Page 670
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 671
the boundaries of the solution space. Comparison of the
impacts of these strategies can lead to the construction of
compromise strategies by changing elements in the
strategy. A resulting loss with respect to one criterion is
then compared with gains to another.
The developed computational framework may be used
for the assessment of selected strategies. Input data sets
will be prepared for the computations, defining the mod-
ification of the water resources situation (strategy) and the
assumed water demand (scenario). Such a data set will be
referred to as a case. Each case is associated with a unique
combination of a WRS, its demand situation (scenario)
and its hydrometeorological boundary conditions.
6.2. Evaluation of Alternative Strategies
Strategies are compared on the basis of their scores on
the evaluation criteria that were chosen to characterize the
management objectives (see Section 4.1.3). To facilitate
the comparison, the number of evaluation criteria should
be limited. Criteria have to be comprehensive (suffi-
ciently indicative of the degree to which the objective
is met) and measurable, i.e. it should be possible to
assign a value on a relevant measurement scale. Where
possible, criteria should be aggregated; for example,
some financial criteria might be processed into a single
value when distribution issues are not important. It is
usually impossible to express all criteria in a single
measurement scale such as a monetary value. Criteria
related to the quality of the environment can often be
expressed only in non-monetary terms. This should
however be done in such a way that a ranking is possi-
ble on the basis of the criteria.
Generally, there will not be a single strategy that is
superior to all other ones with respect to all criteria
used in the assessment. That means that an evaluation
method is required for the ranking of alternative strategies
(see Chapter 10). The evaluation can be an intuitive one
that gives implicit weights to various criteria and, depend-
ing on these weights, selects a preferred strategy.
In multi-criteria evaluation methods, formal decision
rules are used to obtain a ranking. The weighted summa-
tion is one simple method. The best alternative is the one
that gives the best weighted score.
6.3. Scenario and Sensitivity Analysis
Before drawing conclusions from the impact assessment,
it is necessary to analyse the effects on the results of
changes in the assumptions about the scenario variables
and model parameters. The analysis of the influence of
the exogenous variables is called scenario analysis; analy-
sis of the influence of model parameter values (the system
assumptions mentioned in Section 4.2.4) is called sensi-
tivity analysis (Chapter 9).
Scenario analysis is done by repeating the impact
assessment for the selected strategies under different sce-
nario assumptions. If the selection of a different scenario
would significantly change the attractiveness of the
selected strategies, then additional study is required to
reduce the uncertainties in the scenario assumptions. The
sensitivity of the results for a change in model parameters
and assumptions is done in a similar way.
6.4. Presentation of Results
Presentation of the selected promising strategies to
decision-makers may be by means of briefings, presenta-
tions, summary reports or other media. The level of detail
and the way results are presented have to be selected care-
fully. The presentation should give a good overview of the
-
Figure E.12. Activities in the selection phase.
wrm_appE.qxd 8/31/2005 12:11 PM Page 671
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
672 Water Resources Systems Planning and Management
results, without an overdose of information or detail.
Visual aids such as score cards and interactive computer
presentations of study results are extremely helpful for a
discussion of the results of the analysis.
The results of selected strategies can be presented in
matrix form on scorecards. The columns of the scorecard
represent the alternative cases used in the analysis. The
rows represent the impact of different alternatives with
respect to a given criterion. An example is depicted in
Figure E.13. Scorecards can contain numbers only, or the
relative value of the criteria can be expressed by a colour
or shading to obtain a visual picture of the relative order
of the alternatives for various criteria. This can also help
to detect clusters of criteria for which alternatives have a
consistently better score. The presentation of the results
in scorecards allows a decision-maker to give each impact
the weight considered most appropriate.
Two examples of objectives and criteria are shown in the
following two boxes. Box E.7 lists the objectives and criteria
adopted in a water resources management study in West
Java. Box E.8 is a scorecard used in Egypt for the
development of the National Water Resources Plan. This
example does not show alternative strategies; only the
chosen strategy Facing the Challenge has been given.
Including the values of the base case (present situation)
and reference case (future situation if no new measures are
taken) provides a good overview of the new strategys
impacts.
7. Conclusions
This appendix has described the approach used by
WL | Delft Hydraulics to assess water resources systems
and develop plans and management strategies for
them. Like many other such organizations, WL | Delft
Hydraulics has been actively involved in numerous
regional water resources planning and management
projects throughout the world, mostly in developing
countries. The approach described in this appendix
illustrates how these projects are conducted, and the
major factors that are considered while conducting them.
The effects and impacts of some projects are only local
and require consideration of only a few sectors of the
economy. Other, more comprehensive projects have had
national or international impacts. Clearly each water
resources system is unique with respect to its manage-
ment issues, problems and institutional environment.
Figure E.13. Example of a scorecard.
wrm_appE.qxd 8/31/2005 12:11 PM Page 672
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 673
Box E.7. Example 1: Objectives and criteria adopted in West Java WRM study
OBJECTIVES EVALUATION CRITERIA
SOCIO-ECONOMIC OBJECTIVES AND CRITERIA
1. Improve employment Increase of employment by WRM strategies:
number of permanent jobs (#)
number of temp. jobs (mn/yr)
2. Increase income of people
improve income position of farmers farmer net income (Rp/yr)
improve equity in income distribution. difference in benefits of WRM
strategies per capita between:
kabupatens (%)
urban/rural areas (%)
income groups (%)
3. Increase the non-oil export production export value (Rp/yr)
(shrimps, tea and rubber).
4. Support economic development in an total annual. benefits (Rp/yr)
economically efficient way. total annualized costs (Rp/yr)
B/C ratio (-)
- IRR (%)
NPV (Rp/yr)
total capital required (Rp)
foreign currency required (%)
total construction costs (Rp)
total O&M costs (Rp)
sectoral value added (Rp/yr)
GRP (Rp/yr)
USER-RELATED (SECTORAL) OBJECTIVES AND CRITERIA
1. Increase agricultural production padi (ton/yr)
(3% per year) palawija (ton/yr)
export value of crops (or import
substitution) (Rp/yr)
unit costs water supply (Rp/m
3
)
% failure meeting demand (%)
2. Increase power production installed capacity (MW)
power production (GWh/yr)
failure meeting firm power (%)
price of power prod. (Rp/Kwh)
energy production value (Rp)
3. Increase fish production fish produced (ton/yr)
fish pond area (ha)
export value (Rp/yr)
4. Support industrial development amount of supply (m
3
/s)
water supply for industry (full supply) cost of water supply (Rp/yr)
provision of opportunity for discharge unit costs water supply (Rp/m
3
)
of waste water level of failure to meet demand (%)
cost to maintain water quality
standards (Rp/yr)
5. Enhance water-related recreation
(Contd.)
wrm_appE.qxd 8/31/2005 12:11 PM Page 673
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
674 Water Resources Systems Planning and Management
Box E.7 Example 1: Continued
OBJECTIVES EVALUATION CRITERIA
ENVIRONMENTAL AND PUBLIC HEALTH RELATED OBJECTIVES AND CRITERIA
1. Improve public health supply (l/day/ capital)
improve drinking water supply % of people connected (Rp/m
3
)
urban: BNA, IKK and major city programs: price of drinking water (%)
60 l/cap/day, serving 70% % failure meeting demand (%)
rural: 55%
improve flushing (1 litre/s/ha in urban area) volume of flushing water (m
3
/s)
unit costs (Rp/m
3
)
% failure meeting demand (%)
2. Improve/conserve natural resources and area severely eroding (ha)
environment erosion (mm/year)
erosion and sedimentation control sediment yield (tons/yr)
(erosion 1 mm/yr) reforested area (ha)
conservation of nature replanted area (ha)
water quality terraced area (ha)
ratio of external wood supply to total
wood demand (%)
concentration water quality parameters (ppm)
3. Provide flood protection return period (years)
(return period: depending on value of flood alleviation benefits (reduced damage) (Rp/year)
endangered area) flood control cost (Rp/year)
number of people in endangered areas (#)
flooded area (ha)
PLANNING AND IMPLEMENTATION RELATED OBJECTIVES AND CRITERIA
1. Take care of maximum agreement with existing deviations from/conflicts with existing policies
policies in other fields of planning (e.g. economic
regional planning).
2. Maximize flexibility of proposed strategy. degree to which strategy can be adjusted to changes
in demands, standards, technological innovations
3. Maximize reliability of proposed strategy. degree of certainty with which proposed strategy
will meet the realization of objectives
4. Provide sufficient acceptance of proposed degree of acceptance by parties involved
strategy by public, interest groups and executing
authorities.
5. Take care of maximum agreement of proposed deviations from/conflicts with existing competence
strategy with existing competence and and responsibilities
responsibilities of agencies concerned.
Note: GRP = Gross regional product
IRR = Internal rate of return
B/C = Benefit-cost ratio
NPV = Net present value
O&M = Operation and maintenance

kabupaten = Indonesian administrative unit

padi = rice crop

palwija = non-rice crop

Rp = Rupiah
wrm_appE.qxd 8/31/2005 12:11 PM Page 674
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Appendix E: Project Planning and Analysis 675
Box E.8. Example 2: Score-card for Egyptian National Water Resources Plan study
2017 Strategy
Unit 1997 Reference Facing the
Base Case challenge
General (middle scenario)
Population Million 59.3 83.1 83.1
Urbanization Ratio 0.44 0.48 0.48
GDP at economic growth of 6% Billion LE 246 789 789
Economic development objectives
Agriculture: irrigation area M feddan 7.985 11.026 10.876
Gross production value Billion LE 34.46 35.76 38.50
Crop intensity Ratio 2.1 1.5 1.7
Net value production per feddan LE/feddan 2,812 2,075 2,153
Net value production per unit of water LE/m
3
0.64 0.66 0.60
Export/import value Ratio 0.09 0.12 0.20
Industry: Costs of polluted intake water LE/m
3
0.651.10 0.651.10 2.00
Wastewater treatment costs LE/m
3
0.220.50 0.220.50 1.00
Fishery: production (index 100 in 1997) index 100 86 95
Tourism: navigation bottlenecks Index 100 114 0
Social objectives
Create living space in desert areas % of total pop 1.5% 23% 22%
Employment and income
Employment in agriculture M pers.year 5.01 6.24 7.30
Employment in industry M pers.year 2.18 4.99 4.99
Average income of farmers LE/yr 5,362 4,629 4,309
Drinking water supply
Coverage Percentage 97.3% 100% 100%
Sanitation
Coverage Percentage 28% 60% 60%
Equity
Equity water distribution in agriculture , 0, 0
Self sufficiency in food: cereals Percentage 73% 53% 46%
Meeting water needs
Water resources development
Available Nile water BCM 55.8 55.5 55.5
Abstraction deep groundwater BCM 0.71 3.96 3.96
Water-use efficiency of Nile system
Outflow to sinks from Nile system BCM 16.3 17.6 12.5
Overall water-use efficiency of Nile system Percentage 70% 67% 77%
Water in agriculture
Supply-demand ratio (1997 assumed 1.0) Ratio 1.00 0.80 0.92
Water-availability per feddan in Nile system m
3
/feddan/yr 4,495 3,285 3,866
Public water supply
UFW losses Percentage 34% 34% 25%
Supply/demand ratio ratio 0.67 0.76 1.00
Health and environment
Pollution and health
E-coli standard violation (1997 100) Index 100 121 110
Water quality of shallow groundwater , 0, 0
Ecology and sustainability
Sustainability: use of deep groundwater abstr/pot 0.15 1.00 1.00
Condition of Bardawil (Ramsar site) , 0,
Condition of Coastal Lakes , 0, 0 0
Note: UFW Unaccounted for water (the water that is lost in the system)

feddan = 0,42 ha

LE = Egyptian pound
wrm_appE.qxd 8/31/2005 12:11 PM Page 675
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
676 Water Resources Systems Planning and Management
Project planning and analysis approaches must adapt to
these situations. Hence, each project will differ, and will
no doubt need to deviate from the general approach
described in this appendix. Other approaches are pos-
sible and may be equally effective. What remains impor-
tant in all cases is the establishment of a comprehensive,
systematic process of analysis together with constant
communication among planners, decision-makers and
the interested and affected public. The end result should
be an improved, sustainable and equitable water
resources development plan and management policy,
appropriate for the region and its people.
wrm_appE.qxd 8/31/2005 12:11 PM Page 676
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Index
A
Adaptive management 24, 257, 322, 421, 466,
504, 559, 576
Advective transport 385, 522
Aeration 436
Agriculture 11, 16, 51, 113, 325, 405, 458, 475,
501, 564, 581, 675, 608
Algae
Eutrophication 3, 15, 393, 45059, 47478
507520
Modelling 30, 59, 412, 497514
Allocations 20, 60, 83, 85, 138, 156, 236, 303,
356, 364, 389
Alternatives 23, 45, 52, 59, 214, 284, 286,
29596, 313, 35960, 613
Anaerobic 393, 452, 497
Annual value 82
Aquatic systems 386, 464, 501, 656
Aquifer 6, 50, 51, 199, 333, 542, 656
Artificial neural networks 41, 147, 148
Autocorrelations 198, 201
Autoregressive-moving average (ARMA)
models 213
Average values 170, 283
B
Benefits 3, 46, 60, 81, 85, 293, 342, 513, 597,
627, 629
Benefitcost
Analyses 298, 360, 627
Ratio 674, 669
Bias 150, 176
Biochemical oxygen demand (BOD) 390, 449,
451, 478
Boundary conditions 74, 260, 536
Box-Jenkins models 213
Budget 302, 368, 525, 578
C
Calibration 69, 118, 120, 148, 257, 260, 441
Capacity expansion 61, 98, 368
Capital 16, 82, 454
Capital recovery factor 83
Case Studies 48, 467478
Censored data 19395
Central limit theorem 276
Chance constraints 235
Climate change 4, 257, 542, 58890, 608
Coasts
Beaches 538
Management 540
zones 535
Coefficient of skewness 174, 176, 178, 190
Coefficient of variation 174, 178, 189
Coliform
Bacteria 378, 391, 435
Count 379
Collective goods 300
Compensation criterion 300
Competitive conditions 299
Concave functions 84, 89, 130
Concentration gradient 385
Confidence intervals 184, 219, 260
Constraint method 31120
Constraints 39, 60, 64, 87, 266, 296
Constituents
Conservative 390
First-Order 390
Continuous variables 70
Convex functions 86, 130
Correlation 232, 275, 276, 450, 568
Cost-effectiveness 298, 360, 627
Cost 10, 53, 62, 82, 98, 123, 130, 281, 299301
Covariance 206, 271
Critical value 169, 401
Crop 581, 583, 592
Cross-correlation 196
D
Data
Collection 560, 564, 570
Management 565
Mining 163
Dead storage 342, 343, 364, 366
Decision-making 22, 25, 29, 47, 92, 265,
29495, 46264
Decision
Support systems 26, 4749, 78
Uncertainty 24, 72, 258, 260
Variables 60, 81
Deficit value 315
Degrees of freedom 76
Demand
Agricultural 357
Function 358
Industrial 35657
Municipal 356
Depression storage 439440
Depth 102, 154, 173, 261, 268, 306,
402, 491
Design 255, 272, 388, 43739, 456, 493, 509,
560, 565
Deterministic equivalent 236
Deterministic modelling 231
Differential equations 89, 416
Notes: text in boxes is indicated by bold page numbers and in figures and tables by Italic page numbers.
wrm_index.qxd 8/31/2005 12:07 PM Page 677
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Differential persistence 209
Dimensionality 97
Disaggregation 209, 211
Discount factors 82
Dispersion
Bulk 399
Dispersive transport 385, 522
Dissolved oxygen 41, 39093, 435
Distribution systems 427
Diversions 83, 234, 239, 355, 364
Dominance 31011, 585
Droughts
El Nio and La Nia 587
Indices 59095
Management 581, 596
Triggers 596
Virtual exercises 596
Dynamic expansion 368
Dynamic Programming
Backward-moving 92
Deterministic 24041
Dimensionality 97
Forward-moving 9596
Network 90
Stochastic 133
E
Ecology
Ecological criteria 306308
Economics
Criteria 298
Long-run benefits 303
Loss functions 305
Short-run benefits 303
Economies of scale 98
Ecosystems 3, 21, 26, 306, 377, 500
Effective 26, 71, 331, 381, 559
Environmental
Criteria 305306
Impacts 303
Epilimnion 407
Equity 28, 142, 309, 388, 675
Equivalent 8283, 284, 343
Error 175, 176, 257, 260, 276, 277
Estimation 17982, 299300, 339, 348
Estimators 17376
Eutrophication 252, 287, 393, 45059,
47478
Evaporation 326, 332, 34652, 355
Excess value 127
Expansion 16, 368
Expectations 4, 173, 559
Expected value 169, 173, 248, 321
F
Failure 45, 217, 218, 348, 63135, 674
Feasible 63, 240, 24647
Finite difference 416
Firm yield 347
First-order 53, 198, 238, 27071
Fixed costs 98, 130
Flood
Damage 342, 361
Flood storage capacity 359, 343
Flood dikes and levees 607, 632
Management 51
Protection 43, 360, 362
Reservoir 169, 170, 216
Return period 197, 362, 366
Risk 170, 359
Storage 326
Flow augmentation 295, 296
Flux 386, 402, 513
Fractional factorial design 272
Fuzzy
Models 142
Optimization 135
Sets 138, 140
G
Generalized likelihood estimates 181, 192
Genetic algorithms 147, 156, 158, 341
Genetic programming 148, 15963
Global optimum 89, 158
Goal attainment 31415
Goal programming 315
Goodness-of-fit 176, 186, 277
Groundwater
Flows 326, 33435, 509
Interactions 336, 513
Management 540, 542
Modelling 336
Growth 9, 16, 22, 195, 390, 414, 468
H
Hazen plotting position 183
Heat 389, 390
Heavy metals 39899, 452
Hill-climbing methods 84
Hydroelectric power 18, 35759
Hydrological
Models 32829
Parameters 329
Processes 329
Groundwater 333
Surface water 329, 331
Hypolimnion 407
Hypothesis 282, 446
I
Impervious 380, 440, 496
Incremental flow 84, 419
Index-flood method 195, 196
Indifference analysis 284, 313
Inequality constraints 89
Infiltration 328, 329, 437, 441, 496
Inflation 70, 198
Interest 8283
Internal rate of return 674
Interquartile range 173
Investments 27, 28, 59, 83, 660
Irreversibility 284
Irrigation 3, 4, 70, 211, 298, 357, 474, 648
J
Joules 359
K
Kilowatts 359
Knowledge uncertainty 258, 260, 420
Kolmogorov-Smirnov 183
L
Lagrange multipliers 71, 8386
Lake
Outflow 406
Quality 406
Recreation 362
Land 59, 204, 300, 377, 590
Latin hypercube sampling 275, 278
Level 20, 48, 103, 169, 380, 501, 542, 650
Lexicography 313
Light 32, 400, 413, 430, 571
Linear programming
Piecewise linearization 126, 128
Linearizations 127
Loading 10, 41, 380, 445, 446
Local optima 89, 162
Log-likelihood 180
Long-run 23, 303, 305, 368
Loss 6, 21, 272, 341, 377, 534, 584
M
Management
Adaptive 24, 31
Approaches 24
Models 28, 30
Marginal 82, 172, 208, 299
Market prices 299, 300301
Markov
Chains 198200
Processes 198200
Mass diagram 343, 344
Mass transport 385
Maximum likelihood estimates 186, 188
Mean 103, 169, 174, 265, 271, 320, 351, 393
Median 169, 173, 282, 590
Method of moments 180, 181
Michaelis-Menten kinetics 393, 395
Mineralization 408, 498
Model
Adequacy 18284
Calibration 118120, 121
Data 465
Development 45, 46, 71
Dynamic 153
Errors 259, 260
Fuzzy 138
Journal 72
Linear 70, 113
Output 255, 258, 283
Optimization 6465, 83
Selection criteria 38081
Shared-vision 31
Simulation 64
Steady-state 38688
Structure 255, 260
Success 46667
Verification 41, 72, 155
Moments
L-moments 17678
678 Index
wrm_index.qxd 8/31/2005 12:07 PM Page 678
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Index 679
Monitoring
Network 564, 565
Plan 563
Sampling 564, 566
Monte Carlo
Sampling 273, 274, 275
Simulation 23335, 274
Multi-criteria analyses
Constraint method 31113
Dominance 310
Goal-attainment 314
Goal-programming 315
Indifference analysis 313
Interactive methods 315
Lexicography 313
Satisficing 313
Weighting method 311
Multiple criteria or objectives 321
Multiple purposes 20, 231, 355, 461, 624
Multiple yields 347, 352
Multiplier 88
Multivariate models 209
N
National economic development 309
Navigation 21, 163, 164, 356, 613, 619, 623
Network 90, 148, 432, 434, 565
Nitrogen
Cycle 409
Models 381, 409
Non-inferior 310, 312
Non-point
Models 296
Quality 296
Runoff 356, 427
Normal distributions 275
Nutrients
Cycling 408
Models 41213
O
Objective 43, 59, 63, 138, 156, 179, 256, 293,
364, 605, 656
OMR Costs 83
Operating policy 59, 60, 107, 214, 216, 248
Operating rule 352
Operational hydrology 225
Operations Research 56, 133
Opportunity cost 27, 302
Optimal 24, 54, 64, 81, 161, 248, 264, 310
Optimization 59, 60, 81, 135, 239, 309, 368, 453
Organics
Micropollutants 397
Over-year 203, 348
Oxygen 41, 297, 378, 390, 435, 450, 514, 562
P
Parameter
Calibration 120, 150, 260, 263, 333
Uncertainty 260
Value 260
performance measures 59, 103, 135, 281, 352
Peak flow 360, 362, 486, 628, 630
Performance criteria 20, 62, 293, 295
Phosphorus
Cycle 410, 411
Model 410
Photosynthesis 392, 415, 452, 490
Plan
Formulation 6163
Implementation 32, 75, 164, 467
Selection 61, 6364
Planning
Approaches 24
Economic 2728
Financial 27
Integrated 26, 31
Plant capacity 358
Plant factor 358
Plotting positions 183, 186
Policy 5, 19, 26, 86, 107, 164, 242, 311, 322,
379, 421, 428, 598, 647
Population 3, 6, 175, 204, 319, 405, 517, 531,
574, 598, 652
Precipitation 70, 205, 257, 326, 399, 590
Present value 63, 82, 298, 369, 674
Price 86, 220, 300, 302, 428, 664
Primary treatment 435
Principal of optimality 97
Probability
Conditional
Distribution 172
Beta 182
Cumulative 171
Density 171
Exponential 184
Functions 179
Gamma 18789
Gumbel and GEV 190
Lognormal 18687
Log-Pearson Type 3 18990
Normal 18687
Exceedance 183
Joint 171
Marginal 172, 208
Transition 199
Unconditional 200
Production 3, 7, 18, 86, 298, 309, 357, 386,
434, 673
Project
Management 72
Scheduling 368
Public 5, 12, 43, 83, 266, 295, 456, 463, 546,
596, 609, 645
Q
Quality
Impacts 448
Quantity
Flows 271, 325
Storage 27, 327
Velocities 325
Quantiles 173, 176, 278
R
Radioactive substances 396, 400
Radionuclides 400
Rainfall
Design rainfall 43839
Distribution in space and time 438
Models 44445
Synthetic rainfall 438
Random variables
Independent 172, 180
Rank 283, 313, 348
Rate 81, 82, 83, 153, 170, 306, 331, 380, 432,
491, 585, 617
Reaction 7, 386, 393, 415
Re-aeration 384, 392, 415
Recreation 325, 362, 516
Recurrence 348, 585, 618
Recursive 90, 94, 97, 207
Redundant constraints 89
Regional 8, 44, 59, 176, 255, 298, 427, 462,
517, 568, 585, 619
Regionalization 176, 195
Regret 31
Reliability 45, 294, 321
Reservoir
Active storage volume 103
Capacity 115
Dead storage volume 353
Flood storage volume 353
Operation 102
Optimization 68
Over-year storage 203
Release rule 102
Simulation 68, 352
Yield 114
Within-year storage 353
Resilience 111, 321, 535, 559
Respiration 392, 393, 415, 498
Response 4, 77, 153, 268, 302, 368, 380, 444,
490, 559, 585, 620
Return period 197, 362, 438, 493, 674
Revenue 28, 87
Risk 359, 451, 609, 627, 631, 634
Rivers 48384
River basins
Management 328
Simulation 215
Root zone 332, 513
Routing 341, 342, 440, 627, 662
Rule curves 102, 248, 353
Runoff 439441, 444
S
Safe yield 347
Sampling
Monte Carlo 27375
Uncertainty 27576
Satisficing 313
Scale
Processes 75, 76
Spatial 22, 327
Temporal 23, 327
Scarcity 20, 298, 505
Scheduling 368370, 455
Screening 30, 251, 359, 562, 569, 669
Search 7, 63, 97, 147, 192, 319, 454, 461
Second-order conditions 180
wrm_index.qxd 8/31/2005 12:07 PM Page 679
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Sediment
Bed load 403
Burial 402
Resuspension and Sedimentation 401
Settling 396, 397, 398
Suspended load 492
Seepage 103, 257, 336, 504
Sensitivity
Analyses 261
Coefficients 267
First-order analysis 270
Sequent peak analyses 343
Settling
Detritus 409
Inorganic 409
Sediment 398
Sewer systems 448, 478
Silica Cycle 411
Simulation
Accuracy 416
Methods 416
Uncertainty 420
Social criteria 24, 308
Stakeholder 4, 45, 295, 364, 463, 535,
561, 626
State variable 90, 97
Static analyses 327
Stationarity 197
Statistical 41, 69, 147, 183, 231, 259, 320,
329, 380, 438, 465, 564
Steady-state 107, 200, 353, 386
Step method 73, 315
Stochastic
Model 216
Optimization 239
Processes 197
Simulation 214
Variables 214
Storage zones
Active 343
Dead 362
Flood 360
Over-year 353
Stratified 407
Within-year 353
Storage-yield
Functions 344, 346
Storms
Stormwater 447
Stratification 407
Streamflow
Estimation 33940
Gauge 328, 340
Generation 203
Routing 341
Synthetic values 203
Streeter and Phelps 390
Students t-distribution 222
Substitution 173, 210, 298
Sustainability 20, 23, 322, 483, 503
Synthetic Streamflow 203, 205
Systems analysis 12, 30, 476
Systems Approach 461
T
Target 104, 278, 575
Temperature 76, 38990
Thermocline 407
Thomas-fiering model 213
Time periods 327
Time series 197, 259, 437
Time value 82
Toxic chemicals 396, 534
Tradeoffs 59, 280, 293, 29697, 355, 388
Transition probability 199, 238
Transpiration 326, 332, 439, 513
Transport 14, 118, 162, 293, 367, 385, 448,
464, 491, 626
Trial and error 63, 103, 157, 214, 421, 454
U
Uncertainty 254, 258
Analyses 261
Model 262
Parameter 260, 262
Uniform 83, 157, 214, 232, 277, 333, 434,
523, 612
Urban
Drainage 437
Model 441, 442
Pollutant loading 445
Runoff 439
V
Validation 72, 151, 570
Variability 258, 259
Variables
Deterministic 169
Random 170, 174, 214
Variance 174, 175, 232, 255, 320, 588
Verification 16, 41, 69, 154, 384
Vulnerability 111, 220, 319, 321, 559, 598
W
Wastewater
Collection 427, 434
Discharge 434
Treatment 43536
Water
Groundwater 124, 333, 336, 513, 542
Polluted 21
Quality 434, 497, 506, 122
Quantity 567
Surface water 331, 592
Watershed 23, 76, 103, 148, 193, 261, 297,
328, 379, 427, 466, 585, 608
Weibull plotting position 183, 185
Weighting method 311
Wetlands 510
Wilcoxon test 283
Willingness to pay 82, 299
Wilson-Hilferty transformation 189
Withdrawals 3, 355
Within-year
Periods 347, 354
Storage 347
Y
Yield 113, 114, 344, 347
Z
Zooplankton 410, 569
680 Index
wrm_index.qxd 8/31/2005 12:07 PM Page 680
WATER RESOURCES SYSTEMS PLANNING AND MANAGEMENT ISBN 92-3-103998-9 UNESCO 2005
Droughts, oods and pollution are frequently viewed as constraints to economic and
social development. How too little, too much or over-polluted water is managed can
determine the extent to which this critical resource contributes to human welfare.
A variety of management tools and approaches exist, however, that can help water
and watershed managers identify development plans and management policies
that best meet societys economic, environmental and social goals.
Over the past several decades both authors have been applying these tools to water
resource system planning, development and management projects worldwide.
This book is based on experiences teaching at universities and in using these tools
in practice. It uses case studies to introduce both the methods and demonstrates
their practical application. It will serve as a valuable reference for both consultants
involved in planning and development projects and as a text, with exercises, for
university students.
About the authors:
Daniel P. Loucks, Professor of Civil and Environmental Engineering at Cornell University
in the United States, has been actively involved in both the development and
application of water resources systems models for over three decades. As a consultant
to private, governmental and international organizations he has participated in
regional water resources planning, development and management projects in Africa,
Asia, Australia, Eastern and Western Europe, and North and South America.
Eelco van Beek, water resources specialist and manager of the freshwater systems
group within WL / Delft Hydraulics, The Netherlands, has been involved in river
basin planning and management projects all over the world, together with model
software development and use, for over three decades. As a professor he headed the
integrated water resources management program in the Faculty of Civil Engineering
and Geosciences at the Technical University in Delft.
www.unesco.org/publishing
WL | delft hydraulics