System Dynamics Modelling for Dryland Salinity Strategic Management: A Methodological Investigation

By

Naeem Ullah Khan B.Sc. (Hons) Agriculture (Faisalabad, Pakistan) M.Sc. Natural Resources Planning and Management (AIT-Thailand) M.Sc. Sustainable Agriculture (Aberdeen, UK)

A thesis submitted for the Degree of

Doctor of Philosophy

School of Engineering and Information Technology

University College University of New South Wales Canberra, Australia

2012

ABSTRACT

This thesis examines the nature of dryland salinity, its causes and the difficulties that arise in its management. Why such problematic situations are confounding to manage is examined, noting particularly that humankind’s actions over two centuries have played a dominant role in creating dryland salinity as we know it in Australia today.

This research aims to define an improved methodological approach for designing management strategies that are likely to produce enduring results. It also aims to identify the likely effectiveness of alternate strategies and the timeframes over which results might be achieved.

To do this, a multi-methodology approach is taken, one which combines an understanding of feedback dynamics and one which sets out to improve the rigour of investigations into complex systems. The key methodologies are System Dynamics Modelling and Systems Engineering.

How these methodologies were applied in an attempt to design highly effective strategies is described. Lessons for applying the methodologies individually and in combination are described, as are the candidate strategies for alleviating dryland salinity both in Australia and in other parts of the world where its effects are similarly devastating.

ii PUBLICATIONS

Khan, N.U., McLucas, A.C. and Linard, K. (2004). Development of a Reference Mode for Characterization of the Salinity Problem in the Murray Darling Basin. In Proceedings of 22nd International System Dynamics Society Conference, Oxford, UK, System Dynamics Society.

Khan, N.U. and McLucas A.C. (2006). Development of a Strategic System Dynamics Model of Dryland Salinity. In: Proceedings of 24th the International System Dynamics Society Conference, Nijmegen, The Netherlands, System Dynamics Society.

Khan, N.U. and McLucas A.C. (2008). A Case Study in Application of Vee Model of Systems Engineering to System Dynamics Modelling of Dryland Salinity in Australia. In: Proceedings of the 26th International System Dynamics Society Conference, Athens, Greece, System Dynamics Society.

Khan, N.U. and McLucas A.C. (2008). Application of Systems Engineering Vee Model to Enhance System Dynamics Modelling Methodology. In: Proceedings of the Systems Engineering and Test and Evaluation (SETE) Conference, Canberra, Australia, Systems Engineering Society of Australia.

iii ACKNOWLEDGEMENTS

I am grateful to my supervisor Dr Alan McLucas for his encouragement, tireless review of my written work and guidance on the direction of this research, particularly at the times when there was a changing focus consequent upon the outcome of preceding research. He provided an environment for original thinking and critical review of different systems methodologies for strategy development. He also provided me with opportunities for teaching system dynamics courses to graduate students. This was critical in self-analysis about the methodology through presentation, interactions with graduate students and feedback, and it also provided some financial assistance in a time of need.

I am thankful to Keith Linard, the former Director of the UNSW Center for Business Dynamics and Knowledge Management and Robert Niven for their valuable time and thought provoking discussions in the initial conceptualisation of the project.

For prompt response to my requests for information and data, I am thankful to Dr Mark Littleboy, NSW Department of Land and Water Conservation; Colin Mues of the Australian Bureau of Agricultural and Resource Economics; Brad Powel Murrumbidgee Irrigation; Andrew Close, Matt Kendall and Judy Swan from the Murray Darling Basin Commission; Garry Richard and Ashley Fuller from the Australian Greenhouse Office; Numerous data providers from the Australian Bureau of Statistics and Andrew Hollis from the National Climate Centre, Bureau of Meteorology.

A major part of the study was accomplished as a part-time student being in full-time employment at the Policy and Planning Branch of the Department of Immigration and Citizenship. I am thankful to Daniel Caldwell, Kristina Crnagoj, Robert Johnston, Christopher Lanspeary, Jan Tankiang and Andrew Bleeze for intermittent study leave approvals. It is a pleasure to work in an organisation that values an understanding of individual needs, encourages personal development and work life balance.

iv I am also thankful to Professor Michael Frater, Head of the School of Information Technology and Electrical Engineering and Professor Steve Yeoman, the then Head of the School of Civil Engineering for providing a workplace for this study.

I thank the anonymous reviewers of my papers and audience at the International System Dynamics Society Conference at Oxford, Nimjegen and Athens, and Systems Engineering Test and Evaluation Conference in Canberra for their critical and useful comments and feedback. I am also thankful to Ms Katie Poidomani for proofreading of the final manuscript.

I am also grateful to my elder brother Naseem Ullah Khan for his encouragement towards higher ideals in life and particularly during completion of my doctoral studies.

For her patience, understanding and love during this particularly busy period, I am thankful to Afroz. The cheerful smiles of Aleena and Talal kept reminding me to successfully finish my studies.

v Contents

Chapter 1 Introduction 1-1 1.1 Motivation 1-2 1.2 Research Gap, Challenges and Benefits 1-8 1.3 Purpose of Inquiry 1-9 1.4 Scope of Study 1-10 1.5 Study Area 1-11 1.6 Publications 1-12 1.7 Thesis Structure 1-13

Chapter 2 Dryland Salinity - A Complex Problem 2-1 2.1 Defining Dryland Salinity 2-2 2.1.1 Classification of Salt Affected Lands 2-3 2.2 Extent of the Dryland Salinity Problem 2-4 2.2.1 Global Extent 2-4 2.2.2 Dryland Salinity in Australia 2-6 2.3 Significance of the Dryland Salinity Problem 2-6 2.4 Varying Explanations about Causes of Dryland Salinity 2-8 2.4.1 The FAO Model 2-9 2.4.2 PMSEIC Model 2-10 2.4.3 Acworth and Jankowsky Model 2-14 2.5 Other Factors Affecting Dryland Salinity 2-15 2.5.1 Climate 2-15 2.5.2 Land cover/Land use 2-16 2.5.3 Salt Stores 2-18 2.5.4 Hydrogeology Characteristics 2-19 2.5.5 Local Flow System 2-21 2.5.6 Regional Flow System 2-22 2.6 Control Options 2-24 2.6.1 Engineering Options 2-25 2.6.2 Plant-Based Options 2-29

vi 2.6.3 Living with Salts 2-35 2.6.4 Other dryland salinity policies and unused points of entry 2-35 2.7 Summary and Conclusion 2-38

Chapter 3 Models, Tools and Decision Support Systems for Dryland Salinity Management 3-1

3.1 Defining A Model, Tool or Decision Support System 3-2 3.1.1 Model 3-2 3.1.2 Decision Support Systems and Tools 3-2 3.2 Existing Models, Tools and Decision Support Systems 3-3 3.2.1 Models of Physical Processes 3-7 3.2.1.1 Soil Water and Groundwater Simulation (SWAGSIM) 3-7 3.2.1.2 Grassland Water District Decision Support System 3-10 (Quinn and Hanna, 2003) 3.2.2 Spatial Models 3-12 3.2.3 Bio-economic Models 3-12 3.2.3.1 Salinity and Land Use Simulation Analysis (SALSA) 3-13 3.2.4 Dynamic Models 3-14 3.2.5 Integrated Modelling Frameworks and the Practice of Using Multiple Models 3-17 3.2.6 Information Frameworks 3-18 3.3 Issues in Modelling For Management of Dryland Salinity 3-18 3.3.1 Model Boundary and Varying Levels of Focus 3-19 3.3.2 Emergence 3-20 3.3.3 Dynamic Complexity 3-21 3.3.4 Time Delays 3-24 3.3.5 Complexity and Modelling 3-24 3.4 Summary and Conclusions 3-26

Chapter 4 System Dynamics 4-1 4.1 What is System Dynamics? 4-2 4.1.1 Definitions 4-2 4.1.2 Purpose of Inquiry in a System Dynamics Study 4-3

vii 4.2 Philosophy Underlying System Dynamics 4-5 4.2.1 Principles of Systems 4-6 4.2.2 Rationality in System Dynamics 4-7 4.3 System Dynamics Methodology 4-9 4.4 System Dynamics Practice 4-15 4.4.1 A Brief History of System Dynamics Applications 4-15 4.4.2 Quantitative and Qualitative Debate 4-18 4.5 Summary and Conclusion 4-19

Chapter 5 A Multi-Methodology Framework 5-1 5.1 Need for Multi-Methodology 5-3 5.2 Multi-Methodology Philosophy 5-3 5.3 Current Use of System Dynamics with other Systems Methodologies 5-12 5.4 A Multi-Methodology Framework to Enhance System Dynamics Modelling Process 5-13

5.4.1 Cognitive Mapping/Concept Mapping 5-14 5.4.2 System Dynamics 5-17 5.4.3 Systems Engineering 5-17 5.4.4 Strengths of using Systems Engineering in conjunction with System Dynamics 5-22

5.5 Summary and Conclusions 5-25

Chapter 6 Reference Mode Development 6-1 6.1 Why Reference Modes are needed? 6-3 6.2 Method for Reference Modes Development 6-3 6.3 Data for Developing Reference Modes 6-6 6.4 Domain Boundary 6-7 6.4.1 Snapshots of Current Situation 6-7 6.4.2 Concerns about Salinity Data 6-8 6.4.3 Climate 6-9 6.4.4 Land Use 6-9

viii 6.4.5 River Water Diversions and Salt Carrying Capacity of Rivers 6-10 6.4.6 Delays 6-11 6.5 Key Variables 6-12 6.5.1 Time Series Data 6-12 6.5.2 Farmers’ Perceptions about the Past and Future of Dryland Salinity 6-19 6.5.3 Delineation of the Domain Boundary 6-24 6.6 Preliminary System Boundary 6-24 6.6.1 Step 5-Examination of Variables 6-25 6.6.2 Steps 6 and 7 6-26 6.6.3 Step-8 6-26 6.7 Preliminary Model Boundary 6-27 6.7.1 Step-9 Examine Selected Group of Patterns 6-27 6.7.2 Step-10 Aggregate at the Desired Level 6-27 6.7.3 Step-11 Graph Inferred Behaviour of the Aggregated and Abstract Variables 6-29

6.7.4 Step-12 Assemble Graphed Historical Pattern into Fabric of Model Variables 6-29

6.8 Model Boundary 6-30 6.8.1 Step-13 Examine Selected Group Patterns 6-30 6.8.2 Step-14 Infer Stocks Missing in the Data 6-30 6.8.3 Step-15 Graph Behaviour of the Additional Stocks Missing in Data 6-32

6.8.4 Step-16 Assemble Historical Patterns in the Decomposed Variables into Fabric of Model Variables 6-32

6.9 Reference Mode 6-33 6.9.1 Step 17 Examine Past Behaviour of the Variables 6-33 6.9.2 Projections into Future 6-34 6.9.3 Step 20 Ensure Logical Consistency 6-37 6.10 Issues in Overall Process of Developing Reference Modes 6-38 6.10.1 Sources of Data 6-38 6.10.2 Stakeholders Involvement 6-39 6.10.3 Aggregation of Data 6-39

ix 6.10.4 All Variables in Reference Mode or a Few? 6-40 6.10.5 Main Framework of the Method 6-40 6.10.6 All Steps in Series or Transition to Alternate Steps Possible 6-41 6.11 Summary and Conclusions 6-41

Chapter 7 Tools for Qualitative Analysis in System Dynamics 7-1 7.1 Role of Diagrams in System Dynamics 7-2 7.2 Developing Concept Maps for Dryland Salinity 7-3 7.2.1 Cognitive map of a Murray Darling Basin Farmer 7-5 7.3 A Composite Concept Map 7-9 7.3.1 Central/Key Concepts 7-9 7.3.2 Densely Linked Concepts 7-9 7.3.3 Potent Concepts 7-10 7.4 Feedback/Causal Loops 7-12 7.4.1 Reinforcing Loop or Positive Feedback Loop 7-12 7.4.2 Balancing Loop or Negative Feedback Loop 7-13 7.4.3 Developing Feedback-Loop Diagrams for Dryland Salinity 7-14 7.4.4 Feedback loop cleared Land Availability 7-15 7.4.5 Threats to Land Availability 7-16 7.4.6 Feedback Loop - Farming Land Availability 7-18 7.4.7 Feedback Loop – Productivity 7-20 7.4.8 Feedback Loop – Income and Cost of Production 7-21 7.4.9 Composite Causal Loop Diagram 7-23 7.4.10 Physical Processes leading to Dryland Salinity 7-25 7.5 Mitigation Strategies 7-27 7.6 Influence Diagram 7-30 7.7 Summary and Conclusion 7-36

Chapter 8 Simulation Modelling in System Dynamics 8-1 8.1 Why Simulation Is Necessary in System Dynamics 8-2 8.2 Need for a Modular Approach for Developing System Dynamics Models 8-4

x 8.3 Overview of the Dryland Salinity System Dynamics Model 8-8 8.3.1 Model Purpose and Intended Users 8-8 8.3.2 Basic Requirements for the Model 8-9 8.3.3 Model Structure 8-10 8.3.4 Module Integration 8-26 8.4 A Detailed Framework 8-36 8.5 Use of the Simulation Modelling in Dryland Salinity Strategic Management 8-40

8.5.1 Strategy Development 8-40 8.5.2 Training of Decision Makers 8-44 8.5.3 Learning at the Grass-Root Level about a Real World Problem 8-44 8.6 Summary and Conclusion 8-46

Chapter 9 Verification and Validation of System Dynamics Models9-1 9.1 Model Validation in System Dynamics 9-3 9.1.1 Model Simplification as A Validation Strategy 9-6 9.1.2 Scientific Modelling 9-7 9.1.3 Model Calibration as a Testing Strategy 9-9 9.1.4 Model Validation as an Integrated Social Process 9-11 9.2 Verification and Validation in Systems Engineering 9-16 9.3 Validation and Verification Processes in Systems Engineering vs Validation of System Dynamics Models 9-23

9.4 Verification and Validation of the Dryland Salinity Model 9-28 9.4.1 Boundary Adequacy 9-31 9.4.2 Structure Assessment 9-33 9.4.3 Sensitivity Analysis 9-35 9.4.4 Parameter Assessment 9-38 9.4.5 Flow Calculation Sequence 9-40 9.4.6 Mass Balance Test 9-41 9.4.7 Dimensional Consistency 9-42 9.4.8 Behaviour Reproduction Tests 9-43 9.5 A New Process for validation 9-45

xi 9.6 Summary and Conclusion 9-48

Chapter 10 Summary and Conclusions 10-1 10.1 Summary of Main Findings and Their Limitations 10-1 10.2 Specific Contribution and Limitations of This Research 10-13 10.3 A Paradox or Just A Matter For Further Inquiry 10-15

Bibliography Bib-1

xii List of Figures

Figure 1.1 Location map of Murray Darling Basin. 1-15 Figure 2.1 Salt affected pastures near Kingston, South Australia. 2-2 Figure 2.2 A conceptual model of a recharge and seepage area. 2-10 Figure 2.3 Linkages between land clearing and rise in watertable. 2-13 Figure 2.4 Leakage through different land-covers. 2-17 Figure 2.5 Relationship between soil salinity and crop yield. 2-18 Figure 2.6 Conceptual framework of ground water processes. 2-20 Figure 2.7 A conceptual framework of a local scale groundwater flow system. 2-22

Figure 2.8 A conceptual framework of regional groundwater flow system. 2-24

Figure 3.1 Conceptual framework for SWAGSIM. 3-9 Figure 3.2 Treatment of land units in SALSA. 3-14 Figure 3.3 Feedback structure of the dynamic model of salinisation. 3-16 Figure 3.4 Event oriented view of the world. 3-22 Figure 3.5 Feedback view of the world. 3-23 Figure 4.1 Bounded rationality in Systems Dynamics. 4-9 Figure 4.2 Reiterative Process of Systems Dynamics. 4-11 Figure 4.3 Overview of Systems Dynamics Method. 4-12 Figure 4.4 Articles published in the Systems Dynamics Review with respect to author affiliation. 4-17

Figure 5.1 Habermas’ three worlds. 5-5 Figure 5.2 Conceptual relationship between a pluralist methodology and a single methodology. 5-9

Figure 5.3 Vee Diagram of the Systems Engineering. 5-21 Figure 6.1 Broad Framework of Saeed’s Method. 6-4 Figure 6.2 Method for Development of Reference Modes (cycles 1-3). 6-5 Figure 6.3 Method for Development of Reference Modes (cycles 4-5). 6-6 Figure 6.4 Storage capacity in the Murray Darling Basin. 6-13

xiii Figure 6.5 River salinity in the NSW. 6-14 Figure 6.6 Total diversions in the Murray Darling Basin. 6-14 Figure 6.7 Historical trends in the agricultural industry development converted to dry sheep equivalents with major influences. 6-15

Figure 6.8 Australian farm incomes. 6-16 Figure 6.9 Examples of change in landuse intensity index. 6-17 Figure 6.10 Time-graph of the land clearing from 1988 to 1998. 6-18 Figure 6.11 Increase in dryland salinity. 6-20 Figure 6.12 Increase in dryland salinity over time- trends . 6-20 Figure 6.13 Decrease in dryland salinity on individual properties. 6-21 Figure 6.14 Decrease in dryland salinity over time. 6-21 Figure 6.15 Increase in land clearing in the Murray Darling Basin. 6-22 Figure 6.16 Decrease in land clearing in the Murray Darling Basin. 6-22 Figure 6.17 Impacts of dryland salinity on individual properties 6-23 Figure 6.18 Relationship between dryland salinity and depth to groundwater. 6-23 Figure 6.19 Pattern of behaviour of different variables overtime. 6-26 Figure 6.20 Aggregated total land. 6-29 Figure 6.21 Behaviour of missing stocks. 6-32 Figure 6.22 Farmers’ perspective on future trends 6-35 Figure 6.23 A reference mode for salinity problem. 6-36 Figure 7.1 Concept map of a Murray Darling Basin Farmer-Causes 7-6 Figure 7.2 Concept map of a Murray Darling Basin Farmer-Impacts 7-7 Figure 7.3 Concept map of a Murray Darling Basin Farmer-Remedies 7-8 Figure 7.4 A concept map of dryland salinity in the Murray Darling Basin. 7-11

Figure 7.5 Positive or Reinforcing loop. 7-11 Figure 7.6 Behaviour of Reinforcing loop. 7-11 Figure 7.7 A balancing of Negative feedback loop. 7-12 Figure 7.8 Behaviour of Balancing loop. 7-12 Figure 7.9 Feedback loop – Cleared Land Availability. 7-15 Figure 7.10 Feedback Loop – Threats to Land Availability. 7-17

xiv Figure 7.11 Feedback Loop – Farming Land Availability. 7-19 Figure 7.12 Feedback loop – Productivity. 7-21 Figure 7.13 Feedback Loop – Income and Cost of Production. 7-22 Figure 7.14 Composite feedback loop diagram of the dryland salinity problem in Australia. 7-24

Figure 7.15 Feedback Loop – depth to groundwater. 7-26

Figure 7.16 Salts in soil mass. 7-26

Figure 7.17 Composite Feedback Loop including physical aspects

and mitigation strategies. 7-29

Figure 7.18 Identification of influences- example 7-32

Figure 7.19 Influence diagram of the dryland salinity problem in the Murray Darling Basin, Australia. 7-33

Figure 8.1 Structure of a generic module. 8-11 Figure 8.2 Sub-model land under natural vegetation. 8-16 Figure 8.3 Sub-model cleared land neither salt affected nor at the risk of becoming salt affected. 8-17

Figure 8.4 Sub-model land either salt affected or at the risk of becoming salt affected. 8-18

Figure 8.5 Rate of land clearing. 8-19 Figure 8.6 Table Function - land clearing fraction. 8-20 Figure 8.7 Random time delay in land clearing. 8-21 Figure 8.8 Rate of land becoming salt affected. 8-21 Figure 8.9 Table function/graph of the fraction land becoming salt affected. 8-22

Figure 8.10 Time delay in cleared land becoming salt affected. 8-23 Figure 8.11 Rate of land reclamation. 8-23 Figure 8.12 Rate of land returning to natural vegetation cover. 8-25 Figure 8.13 Strategic framework – stock and flow diagram. 8-27 Figure 8.14 Graph window. 8-30 Figure 8.15 Inputs control window. 8-31

xv Figure 8.16 Sub-model providing initial values of the land under natural vegetation. 8-34

Figure 8.17 Sub-model that provides initial values to the cleared land neither salt affected nor at risk of becoming salt affected. 8-35

Figure 8.18 Sub-model providing initial values to the stock land either salt affected or at the risk of becoming salt affected. 8-36

Figure 8.19 Physical processes in dryland salinity. 8-38

Figure 8.20 A detailed framework. 8-39

Figure 8.21 Business as usual scenario, i.e., without any further intervention - simulation settings. 8-42

Figure 8.22 Scenario: 90% effective control treatment is applied at the 50% of the area. 8-42

Figure 8.23 45% effective treatment applied over 50% of area. 8-43 Figure 8.24 45% effective treatment is applied over 50% of the land at risk of becoming salt affected and 20% of the land returning to natural vegetation. 8-43 Figure 9.1 Two kinds of validation processes used in System Dynamics. 9-13

Figure 9.2 Relationship between verification and validation in Systems Engineering. 9-18

Figure 9.3 Verification Planning within Vee model of Systems Engineering. 9-21

Figure 9.4 Baseline verification – the second leg of Vee. 9-22 Figure 9.5 Simplified Systems Engineering verification and validation processes. 9-25

Figure 9.6 Systems Engineering process as applied to System Dynamics modelling. 9-28

Figure 9.7 Model boundary. 9-32

xvi Figure 9.8 Response of the stock ‘land under natural vegetation to varying time delays. 9-37

Figure 9.9 Rate of land clearing – conforming to physical laws. 9-37 Figure 9.10 Inventories conforming to physical laws. 9-38 Figure 9.11 Flow of materials. 9-41 Figure 9.12 Delta check model. 9-42 Figure 9.13 Features of the proposed validation process. 9-47

xvii List of Tables

Table 2.1 Dryland salinity classes. 2-4 Table 2.2 Global extent of dryland salinity. 2-5 Table 2.3 Areas with high dryland salinity potential in Australia. 2-6 Table 2.4 Average annual water balance for a native vegetated and an agricultural catchment. 2-16 Table 2.5 Merits and limitations of various engineering options. 2-27 Table 2.6 Tactics and management options for reducing recharge. 2-29 Table 2.7 Advantages and limitations of landcover/agronomic options for recharge control. 2-31

Table 2.8 Focus of the salinity related policies. 2-37

Table 3.1 Models, tools and decision support systems for dryland salinity management. 3-5

Table 5.1 The objective versus the subjective poles on the nature of social science axis and the meaning of the vocabulary. 5-7

Table 5.2 Triads of methods. 5-10 Table 5.3 Pairs of methods. 5-11 Table 7.1 Definition of variables used in the influence diagram. 7-28 Table 8.1 Available calendars and time units within PowersimTM Studio. 8-14

Table 8.2 Units of measurement used in the model. 8-15 Table 9.1 Tests for assessment of system dynamics models. 9-15 Table 9.2 System Dynamics modelling activities organised according to Vee model of Systems Engineering. 9-30

xviii List of Annexes

Annex A Glossary Annex B Major Dryland Salinity Policy Documents Annex C Analysis of Concept Maps Annex D Conventions for Influence Diagrams Annex E Model Equation

Supplements

Supplement I Farmers’ Perspective on Causes, Impacts and Potential Remedial Measures for Dryland Salinity Supplement II A Tutorial

xix CHAPTER 1 INTRODUCTION

This thesis presents a multi-methodology approach for developing strategies to address a seriously confounding problem namely dryland salinity which is a major problem in many parts of the world. Based on the FAO/UNESCO Soil Map of the World, the total area of saline soils is 397 million hectares and that of sodic soils is 434 million hectares (FAO, 2000). These soils are not necessarily arable but cover salt-affected lands throughout the world. A saline soil is the one which has an accumulation of free salts at the soil surface and/or within the soil profile to a level where it can affect plant growth and/or land use. Salinity is a measure of the total concentration of soluble salts in a soil. This condition is generally attributed to changes in land use or natural changes in drainage or climate, which affects the movement of water through the landscape (Isbell, 1996). In contrast, sodic soils are the ones which have high concentrations of exchangeable sodium salts. Both salinity and sodicity refer to excessive concentrations of salt (CRC Soil and Land Management, 1999) which renders them marginally or completely unsuitable for agriculture. Detailed discussion of the nature of salt affected soils is contained in Chapter 2.

In Australia, 32 million hectares are now salt affected. 29.2 million hectares of these salt affected areas have primary salinity, while 2.5 million hectares have secondary salinity and 5.7 million hectares have high potential to develop dryland salinity (NLWRA, 2001). The term ‘primary salinity’ refers to naturally occurring salinity and includes playa lakes, saline lakes and the erosion of saline soils. The term ‘secondary salinity’ refers to the salinity induced by human activity such as land clearing, farming practice, land use change, etc.

The Australian National Land and Water Resources Audit (NLWRA, 2001) estimated that the area with a high potential to develop dryland salinity will reach 17 million hectares by 2050. Over the past 30 years, the Australian Government has invested extensively in attempts to understand, mitigate and control the problem of progressively worsening dryland salinity.

1-1 This chapter introduces the problem to be addressed in different parts of the thesis. The first part of the chapter describes the motivation behind research, identifies the research gap and describes both the challenges faced and benefits deriving from this research. The later part of this chapter provides a road map of the thesis. The roadmap provides an overview of chapters and explains the sequence in which various parts of this thesis are presented.

1.1 MOTIVATION

The need for this research arose from a personal observation that complex problems in agriculture and environment are frequently addressed in ways that do not produce effective or enduring results. These problems are often interwoven and have severe long-lasting impacts on the social, economic and cultural fabric of a society. Such problems are complex and as such can never be completely solved. Governments apply enormous efforts and commit valuable resources and commitment to addressing such problems, often with marginal prospects for producing the desirable improvements. Problems that respond favourably in the short term often produce highly unfavourable responses in the long term and efforts to correct or improve such problems give rise to new problems. In some situations, these problems appear to be more problematic than problems faced earlier. The challenge here is to identify better ways of addressing such problems and becoming proficient in implementing remedial strategies.

Complex problems are sometimes referred to as “wicked”, “ill defined” or “ill- structured” problems. They have certain characteristics that make them difficult to address and practically impossible to solve and such problems provide little or no information on how best to develop strategies for addressing them (Cameron, 2000). People have different perceptions for such problems, referring to such messy problems, Vennix (1996:13) suggests:

One of the most pervasive characteristics of many problems is that people hold entirely different views on (a) whether there is a problem, and if they agree there is, (b) what the problem is. In that sense, messy problems are quite intangible, and as a result various authors [Ackoff (1981), Checkland

1-2 (1981), Checkland and Scholes (1990), Eden, Jones and Sims (1983) and Bryant (1989)] have suggested that there is no objective problem, only situations defined as a problem by people.

Rittel and Webber (1973) describe such problems as wicked problems, not because they considered these problems as ethically deplorable, instead to differentiate them from another class of problems termed ‘tame’ that are relatively easy to solve. To differentiate wicked problems from the tame ones, Rittel and Webber (1973:160) explain:

The problems that scientists and engineers have usually focused upon are mostly “tame” or “benign” ones. As an example, consider a problem of mathematics, such as solving an equation; or the task of an organic chemist in analyzing the structure of some unknown compound; or that of the chess player attempting to accomplish checkmate in five moves. For each the mission is clear. It is clear, in turn, whether or not the problems have been solved. Wicked problems, in contrast, have neither of these clarifying traits; and they include nearly all public policy issues whether the question concerns the location of a freeway, the adjustment of a tax rate, the modification of school curricula, or the confrontation of crime.

Rittel and Webber (1973) present the following characteristics of wicked problems:

 Wicked problems have no definitive formulation. Every formulation of a wicked problem corresponds to a statement of solution, and vice versa. Understanding the problem goes hand in hand with solving it.  There is no stopping rule for wicked problems. There is always room for improvement. Potential consequences are played out indefinitely. The planner terminates work on wicked problems because of external consideration rather than the logic of the problem.  Solutions to wicked problems are not true or false but good or bad. There is no single criterion, system or rule to judge if the solution is true or false. Solutions can only be relative to one another.  There is no immediate and no ultimate test of a solution to a wicked problem.

1-3  Every solution to a wicked problem is a one shot operation; there is no opportunity to learn by trial and error, every attempt counts significantly.  There is no exhaustive, enumerable list of permissible operations to be used for solving a wicked problem nor is there a well described set of permissible operations that may be incorporated into the plan.  Every wicked problem is essentially unique. Once a solution is attempted, one can never undo what he/she has already done. There is no room for trial and error.  Every problem can be considered as a symptom of another problem. It has no single definitive root cause. Since curing symptoms does not cure problems, one is never sure that the problem is being attacked at the proper level.  The existence of a discrepancy representing a wicked problem can be explained in multiple ways. The choice of explanation identifies the resolution. There are many possible explanations for the same discrepancy. Depending on which explanation one chooses, the solution takes a different form.  The planner or the wicked problem solver has ‘no right to be wrong’. He has ethical responsibility for what he is doing. The purpose of a wicked problem solver is not to find the truth, but to find some characteristics of the world people live in. Since there is no way of knowing when a wicked problem is solved, very few people are praised for grappling with them.

Sterman (2000) uses the term complex dynamic problems to differentiate a particular class of difficult-to-solve problems, and considers that natural and human systems have levels of dynamic complexity. Sterman (2000) suggests that the following characteristics of systems give rise to dynamics complexity:

 Systems change over time. What appears to be unchanging, if observed over a longer period seems to vary.  The actors in the system are tightly coupled with each other, i.e., they strongly interact with each other.  Systems are governed by feedback, that is, actions today alter the state of the system that in turn caused others to act. This gives rise to new situations and the world we face for new decisions.

1-4  Since the start of System Dynamics, it has focused on feedback thinking in which feedback loops produce non-linear behaviour (Forrester, 1961).  Actions taken within a system are irreversible and depend upon the history, that is, of the actions taken previously.  Dynamics of the systems depends upon their internal feedback structure.  Systems are adaptive in the sense that the capabilities and decision rules of agents change over time either through learning or other factors. The systems adapt to any change in circumstances.  In complex systems cause and effect can be separated both in space and in time. This confounds common intuition where one looks for patterns of causes and events being juxtaposed. Consequently, the behaviour of such systems is counter-intuitive.  The complexity of feedback systems outmatches our ability to understand them. Consequently, systems show resistance to obvious policy and strategy intervention.  Systems behaviour can be quite different when observed over the short-term compared to the long-term. Effective policies, generally, are a trade-off between the benefits gained over a short-term and those over a long-term. Alternatively stated short-term gains are often achieved at the expense of long-term benefits even to the point of precluding the latter being achieved at all.

Review of literature described in Chapter 2 indicates that dryland salinity in Australia has many of the characteristics of complex dynamic problems and wicked or ill-structured problems, alike. Dryland salinity is a complex dynamic problem that has a long history of development and detrimental human intervention stretched over a period of around 200 years. There is no single root cause of this problem and stakeholders perceptions differ and frequently conflict. There is no common or broad understanding of the cause and effect mechanisms that underpin the complex problem that dryland salinity is. To demonstrate this, causal models proffered by different researchers and stakeholders are explained in Chapter 2.

1-5 Dryland salinity is not just a physical problem but also a socio-ecological problem. Humans are important actors and people directly perform land management operations, and develop policies that change the land. The human desire to produce food and wealth has resulted in inexorable changes to the land and the environment. There is a surprising paucity of information about how much land is actually salt affected. The Nationl Land and Water Resources Audit (NLWRA, 2001) determined land to be at risk of becoming salt affected when the water table rises to be within two meters of the land’s surface. Previous land and water management policies have exacerbated the problem with salt affected areas continuing to grow. Since the 1970s several policy approaches to managing dryland salinity have been applied but dryland salinity is increasing. This is typical of a complex dynamic problem.

To establish the context for the research described in this thesis it was necessary to review the literature covering various aspects of dryland salinity and the current methodologies employed to address these complex dynamic problems was needed. The literature review spanned ecological sustainability, dryland salinity, existing models of dryland salinity, land management, and systems approaches to managing complex dynamic problems. Dryland salinity is an enormous problem affecting much of the world’s land mass and the literature which describes the problem is consequently very large. The literature review was necessarily limited by the defined scope of research; however, the literature review showed that:

 Dryland salinity poses a major management problem in many dryland areas. It is a major threat to the land and water resources in several countries and has introduced agricultural management problems in Australia, North America, South Africa, and Asia. Dryland salinity has also been reported in South Africa, Iran, Afghanistan, Turkey, Thailand, India and Pakistan (Pannell, 2006; White, 1997; Ghassemi, 1995). In Australia, dryland salinity poses a serious problem in vast tracts of land across South Australia, Victoria, New South Wales and Western Australia. In Canada it occurs extensively in the prairie provinces of Manitoba, Saskatchewan and Alberta. The United States of America is affected in the states of Montana, and North and South Dakota (Ghassemi, 1995; FAO, 2000). Further detail of the extent of the dryland salinity problem is given in Chapter 2.

1-6  A review of literature on dryland salinity modelling presented in Chapter 3 indicates that physical models have been developed to support decisions and model the physical processes through which salinity manifests itself. Over time, these models have been integrated with economic models by various researchers at the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Australian Bureau of Agricultural and Resources Economics (ABARE) and various academic institutions. The literature review of previously existing models, tools or decision support systems for dryland salinity did not provide evidence of any use of System Dynamics modelling for understanding the complex dynamic dryland salinity problem in Australia, or elsewhere in the world. This is critically important as it will be shown that dryland salinity, its remedial measures, its causes and its potential remediation involve complex dynamic interactions which are not adequately explained by the various physical process models.  Proponents of System Dynamics suggest that it is a rigorous methodology highly effective in enabling the analysis of complex dynamic problems. The detail of the System Dynamics method along with its strengths and weaknesses is described in detail in Chapter 4. The philosophy and theories behind the System Dynamics methodology presume that complexity can only be understood by taking a top-down, holistic or systems thinking view, and consequently effective interventions can only be designed when cause-and- effect mechanisms that exist in any problem situation have been completely understood. Because feedback causality can often produce counter-intuitive dynamic behaviours, System Dynamics uses particular tools for mapping and analysis of feedback loops and provides a framework/language for developing computer models of dynamic problems. The feedback loops explicitly focus the analyst’s attention on the policy or strategy leverage points where intervening can make a big difference to the outcome. Such problem structuring frameworks provide necessary tools for understanding problems. However, emphasis on closed systems or closing the feedback loop might result in some of the important variable being left out of the model (Warren, 2005). Caution is needed to ensure that the most important variables and the most influenced feedback mechanisms are modelled to determine the extent

1-7 of their influence. Computer simulations derived from System Dynamics modelling provides a laboratory for learning through experimentation. The computer model development process in System Dynamics focuses on model improvement through iterations of inquiry, hypothesis formulation and testing to a point where the model can be considered to be necessary and sufficient representation of the dynamic problem being investigated. The literature review revealed that real challenges arise when attempting to comprehensively validate System Dynamics models, which is particularly so as the complexity of the problem under investigation increases. Whilst difficulties in validating System Dynamics models of complex problems means that such models cannot be taken as unequivocally depicting “truth”, they have an important utility in enabling learning about the operation of feedback mechanisms and delays, and associated counter-intuitive behaviours. In support of System Dynamics modelling, some lead authors (Schwaninger, 2008) have suggested it can be strengthened by its integration or synergistic combining with other methodologies.

1.2 RESEARCH GAP, CHALLENGES AND BENEFITS

The literature review described in Chapters 2, 3 and 6, and discussed above, suggests that the dryland salinity problem is characteristically a wicked, complex dynamic problem. Chapter 3 explains that contemporary models have been developed to address the physical as well as bio-economic problems associated with dryland salinity. The methodologies specific to complex dynamic problems have not been previously applied to dryland salinity and the design of strategies for its remediation. Chapter 4, in which System Dynamics method is described, suggests that whilst System Dynamics is a methodology for structuring complex dynamic problems and may be valuably applied to the development of strategies for managing dryland salinity, model validation challenges are significant. This suggests a research gap might be closed by the synergistic use of systems approaches suitable for developing

1-8 strategies for dryland salinity in Australia. Identification of this research gap suggests the following tasks, to:

 Review current and past research on dryland salinity modelling in Australia, and identify how effectively such research has coped with problem complexity.  Review and analyse the current use of systems methodologies and develop a framework for synergistic use or the enhancement of methodologies. The scope of work described in this thesis has been limited to detailed problem structuring using System Dynamics modelling as the primary methodology. This choice has been influenced by the utility of System Dynamics modelling in addressing complex systemic problems.  Apply the integrated framework to dryland salinity problems and identify the consequent strengths and weaknesses of the proffered framework.  Compare and contrast results of the study with others and suggest a pathway for future research.

Expected benefits of successfully addressing this research gap are:

 Improved understanding of the potential of System Dynamics for its synergistic use with other methodologies.  Identifying the particular opportunities for integrating with Systems Engineering and cognitive mapping.  Identifying the benefits of integrating System Dynamics into a Multi- methodology framework.  Improved understanding of the complexity which might derive from feedback thinking about dryland salinity, and applications of System Dynamics and allied approaches to it.

1.3 PURPOSE OF INQUIRY

The purpose of this research is to investigate how we might improve understanding about the synergistic use of System Dynamics modelling with other systems methodologies in addressing the relationships between feedback structure and

1-9 dynamics behaviour within the context of complex socio-ecological systems which rise to dryland salinity. Improved understanding may lead to the development of strategies for improving the problematic behaviour. To accomplish this objective, the following guiding research questions were devised and answered during this research:

 What causal relationships contribute to the growth of the dryland salinity problem? How do these causal drivers operate?  What are the current models of dryland salinity? How do they help in informing an understanding of dryland salinity? What are their merits and demerits?  What are the merits and demerits of System Dynamics in address problems as complex as dryland salinity? How can we enhance System Dynamics capabilities in the context of this particular class of problems?  Which conceptual frameworks assist in combining systems methodologies? What are the merits and demerits of multi-methodology practice? How can it help in System Dynamics modelling of dryland salinity?  What is the contribution of qualitative frameworks and simulation modelling in understanding complex dynamic problems, particularly dryland salinity?  How can we improve confidence in System Dynamics modelling through improvements to model verification and validation?

1.4 SCOPE OF STUDY

This study sets out to identify the opportunities for a multi-methodology framework which might improve understanding about the feedback structure, dynamic behaviour and the linkages between the two purposes of; developing highly effective long term strategies, and for managing dryland salinity. The main methodologies identified are concept mapping, System Dynamics and Systems Engineering.

This study does not attempt to predict the future state of dryland salinity. Rather, the frameworks and tools proffered are intended to enhance learning about dryland salinity and how effective long term strategies might be developed.

1-10 This study uses the term ‘System Dynamics’ to refer to the System Dynamics method that was introduced by Jay Wright Forrester (1918- ) and is developed further through the efforts of System Dynamics researchers and practitioners.

This study considers dryland salinity as a complex problem and examines dominant feedback loops and only includes those and associated variables identified as being most influential in dryland salinity. The rationale for identifying such feedback mechanisms and strongly influential variables are derived from System Dynamics theory and practice.

This study considers Australia’s Murray Darling Basin as a complex system and does not, in any way attempt to model the whole Murray Darling System.

1.5 STUDY AREA

The Murray Darling Basin was selected for study on the basis of it being a unique ecosystem having a long history of development, strategically important land and water resources, as well as being subjected to massive land clearings and water diversions over many decades with consequent rising dryland salinity.

The Murray Darling Basin is considered one of Australia’s most important assets. Despite its unequivocal values as a national asset it is in a state of serious decline. From a political perspective, the parlous state of the management of the Murray Darling Basin is considered to be a national disgrace.

The Murray Darling Basin is situated in the eastern region of Australia. Figure 1.1 shows the location and extent of the Murray Darling Basin in Australia.

1-11

Figure 1.1 Location of the Murray Darling Basin in Australia

1.6 PUBLICATIONS

A three step approach to publications has been adopted to expose this research to peer-review and rigour in the research. Several draft chapters formed the basis for peer-reviewed papers presented at high profile conferences. After each presentation, the feedback received from reviewers guided subsequent work and development of future journal papers.

Parts of this research were presented at the International System Dynamics Society Conferences (Oxford, 2004; Nijmegen, 2006; and Athens, 2008) organised by the System Dynamics Society, and Systems Engineering Test and Evaluation Conference (SETE, 2008) organised by the Systems Engineering Society of

1-12 Australia. The feedback received was incorporated into the relevant chapters of this thesis.

1.7 THESIS STRUCTURE

The outline of this research is presented in four parts, as described below:

1.7.1 Part one

This part answers the following basic questions:

 What is dryland salinity?  How did it develop over time and what contributions previous policies had in making it a complex problem?  What are the factors most affecting dryland salinity? Why is dryland salinity a massively complex problem rather than a ‘tame’ problem? What research has been done so far to accommodate the inherent complexity, thereby leading to in-depth understanding of this problem and potential remedial strategies?  What models, tools or decision support systems currently exist? How do they help us understand the dryland salinity problem? What are their strengths and weaknesses? What opportunities exist to improve these?

This part consists of Chapters 2 and 3:

 Chapter 2 examines the dryland salinity problem in Australia, its historical development, basic concepts, differing perceptions, the confounding consequences of the existence of differing perceptions, current policies and their outcome, current research, and factors contributing to on-going dryland salinity.  This chapter establishes the features of the dryland salinity problem and uses it as an example of a complex and ill-defined real world problem that defies remedial action. The chapter establishes why some of the approaches have failed in addressing some of the key characteristics of dryland salinity.

1-13  Chapter 3 is a review of literature on the models, tools and decision support systems for managing dryland salinity. This chapter also highlights the characteristics of complex problems that need to be addressed through these models for strategy development purposes.

1.7.2 Part two

This part presents methodologies used for this research. First the System Dynamics method is described and is followed by a discussion on multi-methodology framework. This part specifically answers the following questions:

 Why is System Dynamics needed for this study?  What is the method of System Dynamics?  What are its merits and demerits for complex dynamic problems?  How will this method be applied?  What are its limitations?  How will this method be applied for this study?  What synergies of other systems approaches will be used to enrich the analysis?

The review describes background literature on System Dynamics and Multi- methodology.

This part consists of Chapters 4 and 5:

 Chapter 4 presents a literature review on the System Dynamics methodology, its applications and other systems methodologies that support the analysis. Critical review of the System Dynamics method is presented highlighting its philosophy, methodology and practice.  Chapter 5 presents details of concepts underlying multi-methodology and proffers a multi-methodology framework for improving problem conceptualisation and rigour of System Dynamics modelling.

1-14 1.7.3 Part three

This part presents qualitative and quantitative aspects of System Dynamics modelling, and describes results of the application of System Dynamics to dryland salinity. The results are organised in two levels, i.e., problem conceptualisation, and quantitative model and scenario analysis. This part answers the following questions:

 What are the conceptual foundations underlying qualitative frameworks?  How can such frameworks help in understanding dryland salinity for development of improved strategies?  What role simulation modelling plays in development of strategies and policies? How can such concepts be applied to dryland salinity problem?

This part is organised into four chapters:

 Chapter 6 explains the concepts underlying reference models development, highlights difficulties in development of reference modes and demonstrates the use of learning cycles approach for reference modes development.  Chapter 7 presents output of the application of problem conceptualisation approaches. Approaches applied include concept mapping, influence diagrams and learning cycles for reference modes. This chapter explains the concepts underlying these qualitative approaches and demonstrates their application to the dryland salinity problem.  Chapter 8 explains the concept underling simulation modelling in System Dynamics and demonstrates these concepts through building a simulation model for dryland salinity. The model developed has been described with stocks, rate, and auxiliaries.  Chapter 9 provides an account of the model validation concepts in System Dynamics and Systems Engineering and demonstrates their synergistic use for simulation model development. A framework for validation of System Dynamics models has been put forward.

1-15

1.7.4 Part four

This part presents synthesis of the entire research effort, compares results with other studies leading to logical conclusions and recommendations for further research. The key questions this part answers are:

 What are the key findings of this research?  What is the specific contribution of this research?  What recommendations do this research puts forward for further research?

This part consists of Chapter 10 that presents conclusions drawn from this research and recommendations for future research.

1-16

CHAPTER 2 DRYLAND SALINITY - A COMPLEX PROBLEM

We are limited by the basic theorem of ecology, 'We can never do merely one thing'. Garett Hardin (1915-2003)

Dryland salinity has emerged as a consequence of interactions among physical, biological and human activity systems over a long period. Once land has become saline, agricultural activity is adversely affected and the degradation may be irreversible or even permanent. Depending upon the extent of salinity, the choice of crops in a region, the choice of construction materials and age of infrastructure may be reduced. The problem is geographically distributed in the most parts of the world particularly arid and semi-arid regions where the pressure for food production has led to intensive farming using irrigation and demands for higher levels of production from marginally productive arable lands.

The primary purpose of this chapter is to present a literature review on the diverse aspects of dryland salinity and to describe the dryland salinity problem from a systems perspective. The description starts with the basic concepts about dryland salinity, followed by a description of the problem in terms of its geographical extent, significance and impacts. A review of the causal models of dryland salinity and control options has been also included in the description.

2-1

2.1 DEFINING DRYLAND SALINITY

Salts occur in nature as a result of the natural process of weathering and soil formation. The accumulation of salts in the top soil to the level at which it starts to adversely affect crop production systems is referred as salinity. When salinity occurs in non-irrigated or dryland areas, it is called dryland salinity (Department of Sustainability and Environment, 2008).

Dryland salinity is sometimes defined based on perceived causes, for example, the Australian National Land and Water Resources Audit (NLWRA, 2001:5) in its assessment of dryland salinity based all its investigations on the basis of the groundwater depth and trend, and the risk of shallow water tables is derived from these two attributes (Short and McConnell, 2000). This Audit assumed dryland salinity was caused by shallow water tables. However, not all shallow water tables will be saline.

The obvious symptoms of the areas affected by dryland salinity may include dead trees, salty rivers, bare white patches (as shown in Figure 2.1), crumbling roads, and reduced agricultural production (NSW Department of Land and Water Conservation, 2000).

Figure 2.1 Salt-affected pastures near Kingston, South Australia (adopted from the NLWRA, 2008)

2-2

In the laboratory sense, soil salinity is a ratio of salts to water in a certain environment. Soil salinity is generally measured in terms of salinity of the soil water that is extracted from a soil sample using different procedures in a soils laboratory.

A criterion of salinity is the electrical conductivity (EC) of the soil saturation extract. Saline soils are defined as those with an EC of greater than 1.5 dS/m for a 1:5 soil water extract, and greater than 4 dS/m for a saturation extract. It can be interpreted in terms of the salinity tolerance of plants. Soils are considered saline if their EC exceeds 4 dS/m (Abrol et al., 1988).

2.1.1 Classification of salt affected lands

Dryland salinity is generally classified into two broad forms—primary dryland salinity and secondary dryland salinity. Primary dryland salinity occurs naturally as a part of natural weathering processes, e.g., deposition, removal and translocation, and transformation. Secondary dryland salinity is the salinisation of land due to its use by humans (NLWRA, 2001). Regardless of the types of dryland salinity, the accumulation of salts on or near the land surface or in the water to a level that can adversely affect the human use of that land or water resource is termed secondary dryland salinity.

From the perspective of process of salt accumulation, dryland salinity can be dry scalds that are caused by the removal of topsoil by wind or water or salt pans, salt seepages and salt scald that are related with the rise in the water table. Their main cause is the rise in the water table.

The Soil Survey staff of the United States Department of Agriculture (Soil Survey Staff, 1997) specified three classes of salt-affected soils as defined in terms of their electrical conductivity and percentage of exchangeable sodium:

 A saline soil has a saturation extract conductivity of 4 mmhos/cm or greater and has a low percentage of exchangeable sodium.  A sodic soil has 15% per cent of exchangeable sodium or greater but has a low salt content.

2-3

 A saline-sodic soil has both the salt concentration to qualify as saline and sufficient exchangeable sodium to qualify as sodic.

Ayer and Westcot (1989) and the American Society of Civil Engineers (1990) have classified salt affected soils into four classes: low dryland salinity land, medium dryland salinity land, high dryland salinity land, and extremely saline land (characteristics of these classed are presented in Table 2.1). Saline soils become more problematic if they turn into alkaline soils as alkalinity affects physical and chemical properties of soil as well as deteriorates soil structure.

Table 2.1 Soil salinity classes

Dryland salinity level Electrical conductivity Cultivation

(dS/m)

Non-saline land <2 All crops

Low salinity land 2-4 Restricted growth of sensitive plants

Medium salinity land 4-8 Restricted growth of many plants

High salinity land 8-16 Only tolerant plants grow

Extremely >16 Only salt bush or very tolerant saline/abandoned land plants grow (Ayers and Westcot, 1989; ASCE, 1990)

2.2 EXTENT OF THE DRYLAND SALINITY PROBLEM

2.2.1 Global extent

Dryland salinity is a major threat to the land and water resources of several countries and poses a major management problem in many dryland areas where crops are grown under rain-fed conditions. It has introduced agricultural management problems in Australia, North America, South Africa, and Asia. Dryland salinity has also been reported in South Africa, Iran, Afghanistan, Turkey, Thailand, India and Pakistan (Pannell and Ewing, 2006; White, 1997; Ghassemi et al., 1995). In Australia, it poses a serious problem in South Australia, Victoria, New South Wales

2-4

and Western Australia. In Canada it occurs extensively in the prairie provinces of Manitoba, Saskatchewan and Alberta, and in the United States in the states of Montana, and North and South Dakota (Ghaseemi, 1995).

On a global scale, the total area of saline soils is 397 million hectares and that of sodic soils is 434 million hectares (based on the FAO/UNESCO Soil Map of the World). These soils are not necessarily arable but cover all salt-affected lands at a global level. However, there are marked differences between areas reported by different authors and agencies. For example, Ghassemi et al. (1995) reported around 77 million hectares of salt affected land around the world. The figures he presented are reported in Table 2.2. Oldeman et al. (1991) reported that 45 million hectares of the 230 million hectares (in 1991) of irrigated land (that constitutes 19.5 percent) are salt-affected. 32 million of the salt-affected soils (2.1 percent) of 1,500 million hectares of dryland agriculture are salt affected to varying degrees by human-induced processes.

Table 2.2 Global extent of dryland salinity (Ghassemi et al., 1995)

Region Light Moderate Strong Extreme Total (Mha) (Mha) (Mha) (Mha) (Mha) Africa 4.7 7.7 2.4 - 14.8 Asia 26.8 8.5 17 0.4 52.7 South America 1.8 0.3 - - 2.1 North and Central 0.3 1.5 0.5 - 3.8 America Europe 1.0 2.3 0.5 - 3.8 Australasia 0.5 - 0.4 0.9 Total 34.6 20.8 20.4 0.8 76.6 Note: Mha refers to million hectares.

2-5

2.2.2 Dryland salinity in Australia

According to the National Land and Water Resources Audit (NLWRA, 2001), there are 5.7 million hectares that have high potential to develop dryland salinity that is expected to reach 17 million hectares in 2050. Main states affected by dryland salinity include New South Wales, Victoria, Queensland, South Australia, Western Australia and Tasmania. Western Australia ranks first with around 4.3 million hectares (as at 1998/2000) having high potential to develop dryland salinity. Victoria, South Australia and New South Wales stand second, third and fourth with 0.67, 0.39 and 0.18 million hectares respectively. Table 2.3 shows the potential of the area in different states to develop dryland salinity.

Table 2.3 Areas (hectares) with a high potential to develop dryland salinity in Australia (NLWRA, 2000).

State/Territory* 1998/2000 2050 New South Wales 181 000 1 300 000 Victoria 670 000 3 110 000 Queensland not assessed 3 100 000 South Australia 390 000 600 000 Western Australia 4 363 000 8 800 000 Tasmania 54 000 90 000 Total 5 658 000 17 000 000

*The Northern Territory and the Australian Capital Territory were not included as the dryland salinity in these territories was considered to be minor by the National Land and Water Resources Audit.

2.3 SIGNIFICANCE OF THE DRYLAND SALINITY PROBLEM

The impacts of increased dryland salinity levels can be diverse and can impact both farmlands and the lands set aside for non-agricultural purposes. It can cause the

2-6

physical environment to deteriorate, increase the costs of production, decrease biodiversity and can increase multiple flow-on social impacts for example employment.

The “on-farm” impacts include reduction in crop yields, a need for intensive farm management to achieve limited crop production. Crops can reduce their yield as crops become increasingly sensitive to salinity in the soil (Rhoades et al., 1992).

Adverse impacts of increased dryland salinity levels on the physical infrastructure can be manifold. Dryland salinity can affect physical infrastructure in two ways: firstly the saline water supplies can have detrimental impacts on household and industrial usage of water and the systems that support/facilitate like storage and supply devices (Barnet, 2000; Wilson, 1999). Secondly, the high water tables associated with dryland salinity-prone areas can have detrimental impacts on the communication as well essential supplies networks.

Water tables within two metres of the land surface have detrimental impacts on the communication networks such as roads and bridges, casings/pipes of the telephone lines etc. It can also cause corrosion of the railway lines and underground fuel and water storage tanks. In turn this increases the cost of maintenance and operation of these services (ABARE, 1996; Barnet, 2000; Bar, 1999).

Dryland salinity can also have impacts on the foundations and structures of buildings and other concrete structure. These impacts can take the form of a rise in dampness that would increase the need for more salt-resistant building materials. High water tables can also affect the sewerage networks as they penetrate into sewerage pipe works and increase the volume of effluent that has to be treated. Salinity in streams and water catchments can also increase the treatment cost of making potable water available (ABARE, 1996).

Increased salinity of land and water can adversely impact on the physical environment. It can negatively impact on the biodiversity by causing a decline in native vegetation, loss of nesting sites for birds and some animals. The quality of habitat can be reduced in turn through plants and animals living in and dependent on

2-7

those habitats. It can reduce the value of tourism and aesthetic resources and can introduce marked changes in the stream and wetland ecology (Barnet, 2000).

Increasing dryland salinity can have a number of indirect socio-cultural impacts, which may include declining income of the business supplying agricultural goods and services leading to joblessness, business closures and population decline as a result of out-migrations from saline areas. Dryland salinity may also have adverse impacts on human health. Jardine et al. (2007) identify the following impacts of dryland salinity:

 wind-borne dust and respiratory health,  altered ecology of the mosquito borne Ross River virus, and  mental health consequences of salinity induced environmental degradation.

Biggs and Moutram (2008) suggested that there is a logical link between the increasing prevalence of dryland salinity, increased habitat for saltwater mosquito species, and increased likelihood of transmission of Ross River Virus. This observation is also supported by Lindsay et al. (2007).

2.4 VARYING EXPLANATIONS ABOUT CAUSES OF DRYLAND SALINITY

There are many possible explanations for causal mechanisms for dryland salinity with the literature review indicating contradictory opinions about the causes. These causes are often linked to physical, biological, socio-economic and managerial aspects of dryland salinity. The hydro-geological realities and human activities leading to secondary salinity are complex and may involve spatial affects and time- lags.

Three main models describe the physical processes giving rise to dryland salinity. These models are, hereby, designated as the Food and Agriculture Organization (FAO) model, Prime Ministers Science, Engineering and Innovation Council (PMSEIC) model and Acworth and Jankowsky model.

2-8

2.4.1 The FAO model

The Food and Agriculture Organization (FAO) model (Abrol et al., 1988) considers dryland saline seepage as the main cause of salt accumulation. The main points of the model are:

 salt is accumulated in the seepage spots at low points or side slopes;  infiltration water carries these salts to some impermeable layer and moves laterally down-slope; and  at lower elevations, water seeps laterally out at the soil surface and evaporates leaving the salts behind.

A conceptual framework of the model is shown in Figure 2.2. According to this model development of the saline seeps involves two areas in the field: the recharge and discharge areas. In recharge areas, water in access of the retention capacity of the root zone soil percolates beyond the root zone. In well-drained soils, sub-soils and underlying strata, the excess water percolates down and reaches ground water. If there is an impermeable layer or the layers with lower hydraulic conductivity, water moves laterally down slope. In both cases, the water travelling to the discharge areas dissolves salts from the soil. In the discharge areas, the ground water or perched water rises to the soil surface creating a seep.

2-9

Rainfall

Land surface Salts Watertable Saline seepage area

Impermeable layer

Figure 2.2 FAO Conceptual model of a recharge and a seepage area (redrawn from Abrol et al., 1988)

2.4.2 PMSEIC model

The model of dryland salinity presented in the Australian Prime Minister’s Science, Engineering and Innovation Council’s report (PMSEIC, 1999) is widely accepted and being used by current policy makers in Australia. National Land and Water Resources Audit (NLWRA, 2001), and the Murray Darling Basin Commission (MDBMC, 2000) also used this model to describe dryland salinity. However, some authors (Acworth and Jankowski, 2001a; Bann and Field, 2006) have expressed their reservations about this model which will be explained in later sections.

In this model, dryland salinity occurrences are considered associated with rise of water tables. A brief account of it as presented by PMSEIC (1999) is given below:

Changes in land use since European settlement have significantly altered the hydrology of the Australian landscape. In particular, there has been large scale clearing of native vegetation which has been replaced with shallow-rooted annual crops and pastures. This has substantially increased the amount of water entering groundwater systems.

2-10

As saline groundwater evaporates, salt is left, causing land salinisation. The salt can then increase surface water salinity when it is moved by rain into waterways and river systems. Water leaking beyond the root zone may also move laterally through soils and discharge into streams, rather than enter the groundwater.

This model is in line with the previous explanations of dryland salinity. For example, White (1997) describes dryland salinity causes as:

Extensive removal of deep rooted native vegetation has occurred in the south-western sector of Western Australia, in South Australia, in Victoria and southern New South Wales, and on parts of the western slopes of the Great Dividing Range northwards into Queensland. Its replacement with crops of pastures has upset the water balance over large areas. Rising groundwater has mobilized salts causing salinization of soils. Seepage of groundwater into rivers has increased their dryland salinity, polluting the major source of domestic and irrigation water supplies.

An example of the narration of causes of dryland salinity by the CRC for Plant-based Management of Dryland Salinity (Pannell et al., 2004) is described below:

Salt, mainly sodium chloride, occurs naturally at high levels in the sub- soils of most Australian agricultural land. Some of the salts in the landscape have been released from weathering rocks (particularly marine sediments) (NLWRA, 2001) but most have been carried inland from the oceans on prevailing winds and deposited in small amounts (20- 200 kg/ha/year) with rainfall and dust (Hingston and Gailitis, 1976). Over tens of thousands of years, it has accumulated in sub-soils and in Western Australia for example, it is commonly measured at levels between 100 and 15,000 tonnes per hectares (McFarlane and George, 1992). Prior to European settlement, groundwater tables in Australia were in long-term equilibrium. In agricultural regions, settlers cleared most of the native vegetation and replaced it with annual crop and pasture species, which allow a larger proportion of rainfall to remain

2-11

unused by plants and to enter the groundwater (George et al., 1997; Walker et al., 1999). As a result, groundwater tables have risen; bringing dissolved accumulated salt to the surface (Anonymous, 1996). Patterns and rates of groundwater change vary widely but most bores show a rising trend, except where they have already reached the surface or during periods of low rainfall. Common rates of rise are 10 to 30 cm/year. (Ferdowsian, 2001).

The Australian Dryland Salinity Assessment (NLWRA, 2001:46; Pannell et al., 2004) is also based on this PMSEIC model. The main approach described in this report is presented below:

 Australia is an ancient continent. Over millennia, its land surfaces and rocks have eroded, mobilised and accumulated sediments and salts.  Some of the salts are released from its weathering rocks (particularly marine sediments) but most are carried from surrounding oceans in rain.  Salt stores have been developed because there is a little capacity to drain the continent of salt and water.  Salt is distributed widely across the semi-arid and arid landscapes of Australia.  Australia’s natural dryland salinity has been exacerbated by changes in land use since European settlement. Native vegetation has been replaced with crops and pastures with shallow roots.  Water leaking beneath the root zone and entering internal drainage and groundwater systems has increased so that it now exceeds the capacity of the system to discharge additional water to rivers and streams.  Since more water is entering the system than is leaving it, the water table rises bringing dissolved salts with it.

This process is depicted in Figure 2.3.

2-12

Runoff

Water leaking to the water table from beneath Drainage from landscape deep-rooted plants (0.5-5.0 mm) (0.5-5.0 mm/yr)

Land clearing and Time delay European farming practices

Runoff

Water leaking to the water table from beneath Water table Drainage from landscape shallow-rooted plants rising (0.5-5.0 mm/yr) (15-150 mm/yr)

Figure 2.3 Linkages between land clearing and rise in water table (prepared from NLWRA, 2001; White, 1997; White, 1994; White, 1994).

The above abstracts and diagrams indicate that the PMSEIC model attributes land clearing and change in land cover as a major cause of dryland salinity. It assumes presence of the salt stores (built from oceanic salt sources) in soils that are mobilised with elevating water tables in the discharge areas and become near and on the land surface. This causal relationship is simplified in Figure 2.3. The concepts of this model have been used in developing the causal loop diagram that is presented in Chapter 6.

Bann and Field (2006) heavily criticised this model on the grounds that it is simplistic and does not account for many natural and anthropogenic processes that also influence salinity outbreaks. Such influences include climate, geology, soils, salt types, landscape topology.

2-13

2.4.3 Acworth and Jankowsky model

Acworth and Jankowski (2001) do not support the PMSEIC hypothesis presented above. To investigate the sources of salt causing dryland salinity, they conducted a detailed study involving drilling, geophysics, hydro-geochemistry and ground water monitoring for 10 years in a small catchment in New South Wales. The catchment they selected was within two kilometres of the top of the regional groundwater and surface water divide and remains substantially tree covered.

Acworth and Jankowski (2001) present an alternate view of salinisation of land based on their evidence of land salinisation in an upland catchment in the southern tablelands of New South Wales. They emphasise the possible sources of salts in any model of dryland salinity. Their concluding remark about their perspective is presented below:

The simple conceptual model of dryland salinity proposed by PMSEIC (1999) is not supported by the data from Jinchille. This is a significant observation because the location of the catchment at Jinchille, close to regional surface water and groundwater divide is a good test of PMSEIC model. It is apparent that even at that small scale and close to the catchment divide, there is considerable heterogeneity in the distribution of salt that a more specific mechanism of salt emplacement is required to explain the observed data.

…have proposed that the source of salt within the colluvium is an Aeolian dust carried from arid interior. It is further proposed that the dust will have varying amounts of salt entrained within it, depending upon the sources of silt. Kiefert (1997) noted that the modern dust can contain as much as 50% salt. ….The salt contained within the clayey silt overlying fractured bedrock occurs in thin layers (often <1 m thick) of reworked material. The deep groundwater discharge destabilises these materials, causing dispersion, the development of scalding, and release of salts to surface run off. If there was no groundwater discharge, the

2-14

scalding would not occur, provided the natural vegetation were not removed or interfered with.

The interpretation of geophysical data is much easier by the recognition that the high bulk electrical conductivity is not associated with the rising salty water table but with the clayey silts that contain salts and has low hydraulic conductivity.

Acworth and Jankowski (2001) suggest a change in the management options and further suggest identifing the unstable clayey units in the landscape, stabilising them and preventing their saturation by groundwater or contact with rainwater. They suggested the following measures to reduce the groundwater pressure in the discharge areas.

Shallow inclined bores through the clayey silt and allow discharge to a surface drainage channel. It is noted that at sites where dryland salinity has led to the formation of deep gullies, the water table in the remaining clayey silt is lowered and the material again became stable. The challenge is therefore to achieve this without significant erosion.

2.5 OTHER FACTORS AFFECTING DRYLAND SALINITY

2.5.1 Climate

Natural climate is one of the most important factors influencing dryland salinity. Climate affects dryland salinity through two ways: first, by volume of precipitation and second, by evaporation. In areas where rainfall exceeds evapo-transpiration, the groundwater recharge increases (Coram et al., 2001). In discharge areas, evapo- transpiration provides a suction force for movement of capillary water to the land surface. Seasonal variability of rainfall determines seasonal variations in dryland salinity. In Australia, dryland salinity has been realised in the summer and winter dominant rainfall areas (Coram et al., 2001). However it is also realised in the summer dominant rainfall areas where rainfall does not normally exceed evapo- transpiration. Gordon et al. (2003) estimated change in water vapour flow and

2-15

suggest that over the last 200 years, there was 10% decrease in water vapour flows in the Australian continent.

2.5.2 Land cover/land use

Land cover influences groundwater recharge through interception, infiltration and use of the precipitation (Abrol et al., 1988). Rates of recharge vary with the type of land use. Table 2.4 indicates the difference between recharge from a land surface covered with natural vegetation and agricultural crops.

Table 2.4 Average annual water balance for a native vegetated and an agricultural catchment north-east of Newdegate, Western Australia (Mcfarlane, 1991).

Evapotrans- Inter- Rainfall Runoff Recharge piration ception (mm) (mm) (mm) (mm) (mm) Native 370 0 359 11 0 Vegetation Agriculture 370 18 319 7 26 Difference 0 18 40 4 26

Land cover/use affects the recharge to groundwater through the depth of plant roots and the life span of plants.

Deeper roots allow plants to extract water from deeper soil horizons, reducing the opportunity for water to leak below them. Even with deep rooted crops, leakage can occur during early growing stages when the plants have established only shallow roots. Perennial plants grow all the year and keep their foliage. They can also send deep roots to use extra water from deeper horizons in soil. Overall perennials have more opportunity to use more water than annuals. Figure 2.4 shows a scatter-graph depicting difference of annual leakage among the annuals, perennials and trees (Walker, 1999).

2-16

180 Annuals Perennials 160 Trees

140

120

100

80

Annual leakage (mm) (mm) Annual leakage 60

40

20

0 0200 400 600 800 1000 1200 Annual rainfall (mm)

Figure 2.4 Leakage through different land covers (Walker et al., 1999).

Growth of a certain plant cover is affected by the level of dryland salinity in the soil. Figure 2.5 indicates the marked reduction in the potential yield of grain crops that can occur at different levels of soil salinity (Rhoades et al., 1992). For the moderately sensitive crops like rice and maize, potential yield may start to reduce at the Electrical Conductivity of soil saturation extract as low as 1.5 dS/m. For moderately tolerant crops like wheat, yield reduction may start at 3dS/m. For tolerant crops like barley, yield reduction may start at 7dS/m. Rhoades et al. (1992) suggest the soils with ECe beyond 9 dS/m as unsuitable for grain crops.

2-17

100

80

60

% of Crop% of Yield 40 Unsuitable for Crops 20 Sensitive Moderately Moderately Tolerant Sensitive Tolerant

0 5 10 15 20 25 30 35 Electrical Conductivity of Soil Saturation Extract (Ece) dS/m

Figure 2.5 Relationship between soil salinity and crop yield (Rhoades et al., 1992).

2.5.3 Salt stores

Salts are present in rocks and thereby in the soil through the weathering processes of the soil parent materials (Buol et al., 1980). Salts are usually present in the regolith and are mobilised by the movement of groundwater. The regolith is defined as ‘the soil, sediments, and weathered bedrock, that lies between fresh air and fresh bedrock’ and represents the ‘major salt store in the landscape’. These soil stores provide the main source of salts that appear on the land surface (Buol et al., 1980).

Presence of salt stores in the sub-soil and underlying strata increases vulnerability of a piece of land to become salinised through subsequent groundwater rise. There may be various sources of salts, for example, salts blown from land in a wind erosion

2-18

process or these might have been deposited through rain or wind carrying salts from the ocean. These salts are spatially distributed and can be identified through soil survey and laboratory investigations. For example, in Australia salts are distributed widely across the arid and semi-arid landscapes and are stored in patchy, complex patterns reflecting earlier geological events (HRSCSI, 2004). Salt stores exist:

…in a huge arc from northern Australia, south by the Great Dividing Range, then broadening and sweeping south-west across the Murray- Darling Basin to take in the Riverina and Mallee regions of NSW, Victoria and South Australia. In Western Australia, massive amounts of salt are stored in an arc that sweeps south and east across the semi-arid and arid landscapes of south-western Australia (NLWRA, 2001:44).

2.5.4 Hydrogeology characteristics

The capacity of a catchment to accommodate and transmit infiltrating water largely depends upon hydro-geological characteristics of the rocks. The main hydro- geological characteristics important in dryland salinity are related to the types of groundwater aquifers and how much water can be transmitted through the aquifer material.

Water beneath the land surface occurs in two principal zones, the unsaturated zone and the saturated zone. In the unsaturated zone, the spaces between particle grains and the cracks in rocks contain both air and water. Although a considerable amount of water can be present in the unsaturated zone, this water cannot be pumped by wells because capillary forces hold it too tightly (Alley et al., 1991).

In contrast to the unsaturated zone, the voids in the saturated zone are completely filled with water. The approximate upper surface of the saturated zone is referred to as the water table. Water in the saturated zone below the water table is referred to as ground water (Alley et al., 1991).

2-19

Between the unsaturated zone and the water table is a transition zone, the capillary fringe. In this zone, the voids are saturated or almost saturated with water that is held in place by capillary forces (Alley et al., 1991). This conceptual model is shown in Figure 2.6.

Evapo-transpiration Precipitation

Soil zone

Unsaturated zone

Recharge to water table Water table Capillary fringe

Saturated zone Below the water table (Groundwater)

Figure 2.6 Conceptual framework of the groundwater processes (Redrawn from Alley et al., 1991)

An aquifer is the material below ground surface that can store or transmit groundwater. Aquifers generally occur in sand gravels, limestone, sandstone or highly fractured rocks. Confined aquifer is the one that is bound both above and below by the hard rocks that inhibit groundwater recharging and discharging from the aquifer. An unconfined aquifer is the one that contains the water table and is normally exposed to the surface. Occasionally, there may be a layer overlying it

2-20

protecting it from the surface. Confined aquifers are often unconfined at their recharge sites.

The quantity of water that can be transmitted through an aquifer (often referred to as through flow capacity) is related to it transmissivity, thickness and hydraulic conductivity (Alley et al., 1991). This through flow is often described as the groundwater flow systems that are classified as local, intermediate and regional flow systems. This groundwater through flow can determine the rise in water table and subsequent salinisation according to the PMSEIC model.

2.5.5 Local flow system

Figure 2.7 shows a simple conceptual diagram of a local flow system. In this local scale groundwater flow system, inflow of water from areal recharge occurs at the water table. Outflow of water occurs as (1) discharge to the atmosphere as ground- water evapo-transpiration (transpiration by vegetation rooted at or near the water table or direct evaporation from the water table when it is at or close to the land surface), and (2) discharge of groundwater directly through the streambed. Short, shallow flow paths originate at the water table near the stream. As distance from the stream increases, flow paths to the stream are longer and deeper. For long-term average conditions, inflow to this natural groundwater system must equal outflow (Alley et al., 1991).

2-21

Transpiration by vegetation

Unsaturated zone

Water table Water table

Stream

Unconfined aquifer

Low hydraulic conductivity confining unit

Confined aquifer

Bedrock (very low hydraulic conductivity)

Figure 2.7 A conceptual framework of a local scale groundwater flow system (Source: Alley et al., 1991)

2.5.6 Regional flow system

The definition of a groundwater flow system is to some extent subjective and depends in part on the scale of a study. The extent of groundwater flow systems can vary from a few square miles or less to tens of thousands of square miles (Alley et al., 1991). The length of groundwater flow paths ranges from a few feet to tens, and sometimes hundreds, of miles. A deep groundwater flow system with long flow paths between areas of recharge and discharge is termed a regional flow system.

Figure 2.8 shows a conceptual framework of the regional groundwater flow system. Significant features of this depiction of part of a regional groundwater flow system include:

2-22

 Local groundwater flow subsystems in the upper watertable aquifer that discharge to the nearest surface-water bodies (lakes or streams) and are separated by groundwater divides beneath topographically high areas.  A subregional groundwater flow subsystem in the water table aquifer in which flow paths originating at the water table do not discharge into the nearest surface-water body but into a more distant one.  A deep, regional groundwater flow subsystem that lies beneath the water table subsystems and is hydraulically connected to them. The hydro-geologic framework of the flow system exhibits a complicated spatial arrangement of high hydraulic-conductivity aquifer units and low hydraulic-conductivity confining units. The horizontal scale of the figure could range from tens to hundreds of miles (Alley et al., 1991).

2-23

Low hydraulic High hydraulic conductivity conductivity aquifer Surface water-body confining units Unsaturated zone Water table

Bedrock

Figure 2.8 Conceptual diagram of the regional groundwater flow systems that subsystems at different scales and a complex hydro-geologic framework.

Note: 1, 2 and 3 indicate local, subregional and regional groundwater flow systems respectively. (Source: Alley et al., 1991)

2.6 CONTROL OPTIONS

A large number of dryland salinity control and remediation options are suggested in the literature. These options range from engineering to agronomic to socio-economic solutions. All these options focus on either reducing the groundwater recharge in the recharge areas of the catchment or managing the discharge areas. Engineering and agronomic measures focus on direct reduction of recharges or treatment of saline water and salt removal. The socio-economic measures focus on cost-effectiveness and facilitate the adoption of agronomic and engineering approaches conducive to dryland salinity remediation. There are no statistics available to ascertain the extent these options can be helpful in Australia. The then Chairman of the Australian Land

2-24

and Water Resources Research and Development Corporation, Graeme Robertson (Robertson, 1995) suggested:

It is unlikely there will ever be a total solution to the salinity problem in Australia unless we return the landscape to its original state. Even then it will take centuries to restore the original hydrologic balance. The acceptance of this reality places the management of salinity in a new paradigm. One that acknowledges the impact land use has had on landscape processes. Further management will not be about simple fixes, but about making choices, choices between retaining farms or retaining forests; between saline land or foregoing income from wheat; between saline wetlands or forgoing income from wool.

2.6.1 Engineering options

Engineering options provide a means of quick response to the physical aspect of the dryland salinity problem by lowering water tables. These include different types of surface, subsurface and groundwater drains, groundwater pumping and on-farm options. NLWRA (2001) has classified engineering options for recharge and discharge management into two major groups:

 fairly simple, largely on the paddock surface water management measures

(e.g., banks, drains); and

 more expensive, often larger area measures like drains, subsurface drains,

pumps, interception and diversion systems.

Earlier research on open drains in Western Australia found that they reduce ground water levels only within a few meters of the drain on high clay soils and rarely more than 40 meters on favourable soils (Ferdowsian et al., 1997). More recent research in Western Australia found positive impacts over considerably greater distances (Pannell, 2001; Pannell and Ewing, 2006).

2-25

Major engineering interventions are suitable for only a narrow range of dryland situations. Aspects such as: scale, assets at risk, economics, social acceptance, offsite impacts, climate change, uncertain results, private gain, costs of investigation and implementation are significant impediments (Gill, 2006). Gill (2006) also considers engineering options to be suitable for the following scenarios:

 where a single landholder is keen, cashed up and can store the salt up slope; or  where a significant public asset is being degraded and the high costs of an

attempted technical solution can be justified.

NLWRA (2001) suggested that deep surface drainage and pumping can be cost effective only in situations where:

 the land is very valuable (towns, nature reserves, and infrastructure);

 the soils and aquifers are permeable and in hydraulic connection (water flow

is connected); and

 there is a safe option for effluent proposal.

A comparison of the advantages and limitations of different engineering options is shown in Table 2.5

2-26

Table 2.5 Merits and limitations of various engineering options

Engineering Merits Limitations Options Surface drains Low cost measures that are cheaper to construct and maintain than Value for controlling waterlogging may be limited in some areas. it is to take land out of production. Effluent may cause problems offsite (salts, nutrients). Cause minimal disruption to land use. Cannot be built in sodic soils. Sub-surface Effective for removal of groundwater from waterlogged areas. Much more costly than surface drains. drains Can lower water table effectively to enable productive use of Storage and disposal of collected water may be costly. affected land. Clogging risks and relatively high maintenance costs.

Groundwater Interception trenches are a relatively cheap water storage and Deep drains are costly to build. drains interception strategy. Storage and disposal of collected water can be costly unless it has a Can lower water tables. beneficial use.

Enable unusable land to be reclaimed for production.

2-27

Groundwater If pumping works, salinisation problems can be minimised and Capital cost is high. pumps production of conventional agricultural systems can be boosted at the same time. Application areas are limited to high yielding aquifers.

Capital assets in the town can be protected by removing the Only treats the systems. problems associated with dryland salinity and waterlogging. Limited areas of influence in most systems.

If using water for irrigation it must be carefully managed.

Water storages that leak can recharge the system.

Industry infrastructure is required for irrigated crops.

Disposal of water to the environment.

(NLWRA, 2001:102-103)

2-28

2.6.2 Plant-based options

Land cover change approach is based on the ability of different cropping systems to extract and use different quantities of water from soil. These options include conservation of native vegetation, establishment and improvement of perennial pastures, encouragement of low recharge cropping systems and farm forestry. Details of these options are given in Table 2.6.

Table 2.6 Tactics and management options for reducing recharge (Powell, 2004)

Main Tactics Management Options Remnant vegetation Retain remnant vegetation. Perennial pastures Sustainable grazing pastures:  Basics of the water cycle and water balance.  Critical management factors. Low recharge cropping Phase farming with lucerne:  Intercropping with lucerne.

 Opportunity cropping.

 Other research into low recharge farming

systems.

 National workshop on managing water in

farming systems.

Farm forestry Water use compared with grassland:  Farm forestry design for dryland salinity

management.

Among the annual crops, there may be multiple ways to help reduce recharge. NLWRA (2001) suggests these options to include improvement in crop rotations, facilitation of the root growth, elimination of the fallow periods, and opportunity

2-29

cropping. A comparison of the advantages and disadvantages of land cover change and agronomic practices is given in Table 2.7.

2-30

Table 2.7 Advantages and limitations of land cover/agronomic options for recharge control.

Land cover/agronomic option Advantages Limitations Maintenance of the remnant native vegetation Helps in recharge control: Only modest benefits to the catchment scale recharge control:  Preserved areas help to set recharge control targets.  Only incomplete recharge control.  Areas of remnant native vegetation that  Too scattered to regenerate effectively. remain are the most important systems in  Requires ongoing management. terms of biodiversity outside tropical  Longer delay periods in regeneration and rainforests. effective control.  High heritage and conservation values are maintained. Annual crops and pastures High rate of adoption: It is difficult to significantly reduce discharge using annuals:  Improving agronomy to reduce recharge  Cost effective.  Removing impediments to root growth  Does not require substantial land use  Skilful management and widespread  Elimination of fallow periods change. commitment on every farm is required.  Opportunity cropping  Likely to increase productivity.  Annuals do not use water that falls in  Phase cropping–rotations with lucerne intense or prolonged rainfall events, or that

2-31

pastures falls when plants are not growing.  Alley cropping  Reducing recharge under annuals is a problem in well drained (e.g., deep sands), low fertility and water logged soils and on heavily grazed pastures.  When annual rainfall exceeds 600mm the scope for recharge control by modified agronomic practices becomes limited.  Continuous cropping practices may be unsustainable where there is soil structure decline, acidification and herbicide resistance.

Traditional perennial species Deep roots of trees, shrubs and some perennial May only deliver marginal benefits for recharge: pasture species give greater water use potential than annual, shallow rooted plants:  When changing from annuals to perennials, better management is required (e.g.,  Land use is complementary. restricted grazing).  Adoption rates are higher than for more substantial land use change.

2-32

 Increased water use.

Changing to new perennial species: May prove effective for controlling recharge and re- Uncertainty about market prospects of the crops establishing water balance in low rainfall areas:  Deep rooted perennial woody crops such as  Time-lag in establishing crops can place Jojoba, oil mallee species and broom brush  May provide economically viable additional financial strains. alternative to traditional crops.  Deep-rooted perennials which thrive in a  Mixtures of trees/woody plant crops and new environment always have greater annual plants may be more productive than potential to invade remnant vegetation. the monoculture of either. Trees Returns to both landowners and community: Initial planting costs are high:

 Trees can extract large quantities of water  Cash flows: long delays, for example 25 from the soil by transpiration, and can years. directly intercept and evaporate rainfall.  Difficulties with integrating trees into They actively use water for a greater part of farming systems. the year than most crops and pastures.  Lack of species and systems suitable for  Aesthetic improvements to the appearance lower rainfall areas. of land.  Need for infrastructure to support  Biological weed and pest control. agroforestry enterprises e.g., Sawmills and

2-33

 Enhancement of regional biodiversity. markets.  Risk of fire.  Long time lag before trees start to have an impact on recharge.

(NLWRA, 2001:98-101)

2-34

2.6.3 Living with salts

Living with salt is a strategy that arises from the acceptance of dryland salinity as a problem and recognition of our serious limitations in controlling/containing it. It focuses on the productive use of the salt lands. The main approaches consider establishing salt land grazing and cropping systems, i.e., introduction and use of plants that are salt tolerant but have high economic value. Ten case studies (Anon, 2004) in various Australian states suggested that a well-managed salt land can significantly generate higher returns than improved pastures.

Although living with salts presents an option to get production from saline lands, the impacts of dryland salinity extend beyond agricultural production. It has long term impacts with serious consequences for infrastructure, human health, and ecological sustainability. Any option should be evaluated in the breadth of impacts on the physical, social and human health.

2.6.4 Other dryland salinity policies and unused points of entry

A list of dryland salinity programs is in Annex A. An analysis of the policies/programs indicates that the focus of dryland salinity policy development has been based on the following elements:

 Improving water allocation through pricing and operational rules, i.e., water policy;  Awareness raising, education and community group facilitation policies, i.e., land care, water wise on the farm and agricultural extension policies;  Salinity targets;  Behaviour change and governance policies;  Vegetation restoration policies; and  Saline industries.

Table 2.8 presents an historical perspective on the policy focus of major programs. The programs have used the most points of entry. These include from engineering to plant-based options, governance to behaviour change policies, community based to

2-35

catchment scale options. A question that needs to be addressed is at what points in the dryland salinity system do these options impact? Is the impact towards improvement in the state of the system/problem? And what frameworks exist that help us to understand these impact points?

2-36

Table 2.8 Focus of the salinity related policies in the Murray Darling Basin

Policy Focus Supporting Program/Document Year

Improvement in water use NSW Water Management Bill 2000 through: COAG Water reforms  Allocation QLD Water Act 2000  Water sharing QLD Water Resources Act 1989  Water pricing

Salinity reduction through GHD 1970 dilution by releasing water from reservoirs

Pumping of groundwater GHD 1970 Basin Salinity Management 2001 Strategy Pumping of saline water and its GHD 1970 diversion into evaporation Basin Salinity Management 2001 basins Strategy

Community based groups National Action Plan for Salinity 2000 and Water Quality Salinity targets

Improved governance framework

Behaviour change

Awareness raising and National Landcare Program 1989 education

Catchment planning processes for groups of farmers

Across the entire catchment. Integrated Catchment Management 1980s

2-37

2.7 SUMMARY AND CONCLUSION

This chapter presents a report of the literature review about dryland salinity, extent of the problem around the world, its predominant causes, available mitigation technologies and their affiliated problems; and literature based analysis of the complexity of the problem.

Dryland salinity is a complex problem. It arises as a consequence of a number of biological, physical and socio-economic causes that may be further linked through feedback mechanisms.

Dryland salinity has far reaching impacts on the physico-economic cum socio- cultural system. These impacts are both on the farm as well as off farm. On-farm impacts include loss in agricultural production, damage to farm infrastructure, farm management problems, reduced water quality, environmental degradation, secondary land degradation and reduction in land values. The off farm impacts include impacts on urban/rural infrastructure, environment and employment patterns.

In terms of physical reasons, there appear to be two schools of thought with differing hypotheses. One school of thought gives a conceptual model that European farming practices coupled with land clearing introduced hydrological imbalances within the catchments. This hydrological imbalance increased net groundwater recharge rates which results in elevation of water levels. This elevating water level dissolves salts in the soil profiles and brings them into the root-zone near the surface. The second school of thought does not support this model on the basis of a few experimental results. This group is of the opinion that there is no hydro-geochemical evidence to suggest evaporative or transpirative concentration of salts in the groundwater. The short flow path from top of catchment cannot provide a significant source of salts from bedrock weathering. The alternative model of salt accumulation this school of thought proposes is based on salt importation during dust storms. This school of thoughts stresses that in any model of salinisation, source of salts must be included along-with the processes of accumulation. The former school of thought has suggested the reduction in net recharge rate through land cover and land use practices while the later school has suggested prevention of the saturation of clayey silt units by groundwater or rainwater and to stabilise the clayey silt materials.

2-38

The biological reasons include the change in land cover that has affected the groundwater recharge rates through changes in evapo-transpiration or through the change in root characteristics of the vegetation. Change in vegetation might have also caused the destabilisation of the clayey silt units that Acworth has mentioned.

Socio-economic reasons may include hidden private costs and benefits of the current farming systems to farmers, market incentives for a certain farm pattern/crop mix, legislative framework over the past century and adoption of dryland salinity control practices.

The potential policy measures that have either been adopted or suggested in the literature for adoption include pumping of groundwater to lower water tables and pumping of saline water and its diversion into evaporation basins, perennial plantations in the recharge areas, raising awareness through community action and community based groups, establishing dryland salinity targets, improved governance framework, behaviour change, awareness raising through education, and catchment planning processes for groups of farmers across the entire catchment.

Any preferred option should be evaluated in the context of a broader problem boundary that includes impacts on physical, biological and social systems and human health.

The complexities of the dryland salinity problem discussed in this chapter are further used to enable a discussion on the current approaches to address dryland salinity in the next chapter.

2-39 CHAPTER 3 MODELS, TOOLS AND DECISION SUPPORT SYSTEMS FOR MANAGING DRYLAND SALINITY

Models are used for a wide range of purposes in most disciplines to augment human capabilities in understanding and subsequently responding to complex and messy problems. In Australia, various models are used for different purposes in relation to the dryland salinity. Although, there is not a large number of models specifically designed to address dryland salinity, URS (2002) reported more than 100 models that can be used for salinity management for the National Action Plan for Salinity and Water Quality. These models include tools, simulation models, and decision support systems and they focus on different processes and aspects of salinity.

New developments in modelling frameworks and the complexities associated with dryland salinity provide an opportunity to review the current use of models, tools and decision support systems. Within the scope of this Chapter, it is not possible to review each and every model; rather a review of the approaches underlying these models can provide insights into the way current models are designed and their capabilities to address dryland salinity.

The primary purpose of this Chapter is to provide a report on the review of literature on the tools, models and decision support systems currently being used for dryland salinity or the ones that have the potential for such usage. A critical review of the current modelling approaches and the models is presented along with a discussion on their merits and demerits for addressing dryland salinity. Issues to be addressed by the future models supporting dryland salinity management are discussed. While the main focus of this Chapter is on dryland salinity models, some generic salinity models are also presented.

3-1 3.1 DEFINING A MODEL, TOOL OR DECISION SUPPORT SYSTEM

3.1.1 Model

The term model is used in a large number of disciplines and with a variety of definitions. After a review of model definitions, Muller (2008) suggests that each definition of model is insufficient as it covers only a small range of the reach of use. According to the Oxford Dictionary (2003) the term model refers to:

A simplified mathematical description of a system or process, used to assist calculations and predictions.

Forrester (1968) defined a model as a substitute for a real system. The term ‘model’ also refers to abstractions, concepts and theories. In the remainder of this Chapter the term model will be used in the meaning similar to the definitions provided in the systems thinking and System Dynamics literature.

3.1.2 Decision support systems and tools

System thinking, System Dynamics and decision support literature contains numerous definitions of the term decision support system. It has been used to describe a variety of approaches to the provision of information for decision making for many types of systems, including health, environmental and business systems (Letcher, 2002a). In the early 1960s, the focus of these definitions was predominantly management information systems (Power, 2003). Over time, the focus of decision support systems has changed towards complex model-based decision support, however, in the foreseeable future this concept is likely to take new turns. A few of the decision support system definitions are presented below to demonstrate the rich context in which the concept of decision support system is portrayed.

A decision support system is a computer system that assists decision makers in choosing between alternative beliefs or actions by applying knowledge about the decision domain to arrive at recommendations for various options. It incorporates an explicit decision procedure based on a set of theoretical principles that justify the rationality of this procedure (Fox and Das, 2000)

3-2 A decision support system is computer software that facilitates and accepts input of a large number of facts and methods to convert them into meaningful comparisons, graphs, and trends that can facilitate and enhance decision makers’ decision making abilities (Bhatt and Zaveri, 2002:297)

Turban et al. (2001) presented a working definition of the decision support system:

A DSS is an interactive, flexible and adaptable CBIS [Computer Based Information System] specially developed for supporting the solution of a non-structural management problem for improved decision making. It uses data, provides easy user interface, and can incorporate the decision maker’s own insights. In addition, a DSS may use models, is built by an interactive e process (often by end users), supports all phases of decision making, and may include a knowledge component.

After a review of roles that a DSS can perform and context of decision-making, Marakas (2002) defines a decision support system as:

…a system under control of one or more decision makers that assists in the activity of decision making by providing an organized set of tools intended to impose structure on the portions of the decision making situation and to improve the ultimate effectiveness of the decision outcome

3.2 EXISTING MODELS, TOOLS AND DECISION SUPPORT SYSTEMS FOR DRYLAND SALINITY

Although there are a limited number of models specifically designed for the management of dryland salinity, the models addressing bio-physical relations that can be used for dryland salinity are numerous (Littleboy et al., 1998).

The modelling frameworks specifically developed for dryland salinity address the assessment of the spatial extent of dryland salinity (Malins and Metternicht, 2006), effects of land use change on water and salt delivery (Vaze et al., 2004; Tuteja et al., 2003), viability of recharge reduction (Dawes et al., 2002), evaluation of land management options (Daamen et al., 2002) and mechanisms for salt delivery to streams (Summerell et al., 2006).

3-3 The decision support for salinity management was indirectly provided by catchment models. Only recently the name of decision support systems has been used in association for salinity (Letcher, 2002b). A list of the models addressing either dryland salinity or an aspect of it is given in Table 3.1

3-4 Table 3.1 Models, tools and decision support systems for dryland salinity management

Model/DSS Relation to Decision Level Geographical References Brief Description Salinity Focus/Applicat Direct/Indirect ions

NDSP Decision Support Tool Evaluation of NDSP (2008) Provides information sheets. salinity options SMAC Dryland salinity Catchment Australia Letchure It links farm level decision making and catchment specific/Liverpool (2001) scale hydrological processes and uses mathematical plains programming to investigate optimal strategies for resources uses. MODFLOW Indirect Catchment to Littleboy et Evaluates the effects of management options on regional al. (2001) aquifer behaviour including effects of recharge regimes due to land use changes. FLOWTUBE Catchment Littleboy et It is a simple groundwater model for examining the al. (2001) effects of a range of recharge and discharge options on catchment groundwater. SWAGSIM (Soil water and Indirect/through Operational/farm Australia/ Prathapar et It is capable of modelling spatial and temporal ground water simulation) water tables Murrumbidgee al. (1996) fluctuations of shallow water tables, identifying Irrigated areas recharge and discharge zones within irrigation areas, and assisting the evaluation of drainage options at field and regional scale.

3-5 Model/DSS Relation to Decision Level Geographical References Brief Description Salinity Focus/Applicat Direct/Indirect ions

TCM-Manager (Total Prediction of Direct Australia Martens and It provides an accessible, realistic and user friendly catchment management) water sediment DiBiase graphical platform from which preliminary catchment and nutrient (1996) impact assessment, planning and management can be flux/water performed. quality Watershed DSS (Water, soil, Indirect USA, only one Osmond et al. It can help in identification of water quality problems and hydro environmental water resource, (1995) and select best management practices. decision support system) water use or impairment CEDARE-EIADSS Direct/Salinity Strategic/program/ Not applicable Abu-Zeid et Provides a tool for comparison between irrigation (Environmental impact as an impact of project level al. (2000) project alternatives based on relevant aspects of assessment decision support irrigation surrounding environment. Salinity has been treated as system for irrigation projects) development an environmental impact of irrigation projects. SALMOD (Salinity and Direct Specific irrigation South Africa/ WRCSA It calculates the profit maximising crop choice and leaching model for optimal Economic costs area/farm Vaal and Rivet (2002) distribution over the farm, matching the crop choice irrigation development) Rivers with soil type, drainage status and irrigation system. SALSA (Salinity and land Economic Basin scale Murray Darling Heaney and Compares costs of alternative land use options use simulation analysis) Basin, Australia Bell (2001) For detail see Section 3.2.3

3-6 3.2.1 Models of physical processes

Most of the current models included in the national dryland salinity database-PRISM fall into this category. These models generally represent hydrological processes and in some cases represent integrated hydrological models, including routing of flows, salt, sediment and or nutrients, groundwater models or the models of the biological processes.

These models are generally used for establishing salt water balances and for assessment of the physical or hydrological implications of the land and water resources policies. They are also used to measure the environmental implications of scenarios where land use patterns and rates of policy adoption are assumed.

Within the dryland salinity context, these models do not consider economic or social adjustment processes and generally make assumptions about the level of policy adoption, often assuming 100% of adoption of different management practices. In the catchment context, the behaviour response of farmers or individuals to changing policies is not included in such models (Letcher, 2001).

Such models are, due to their narrow problem boundaries limited in their ability to provide an understanding at the systems level of how these particular physical processes and subjects of the models are linked to other processes or the systems behaviour.

3.2.1.1 Soil Water and Groundwater Simulation (SWAGSIM)

SWAGSIM (Prathapar et al., 1996) was developed by CSIRO scientists with the aim to facilitate the evaluation of shallow water table management options with minimal data and computing requirements. SWAGSIM addresses the problem of rising water tables in the irrigation areas. In older irrigation areas like Murrumbidgee irrigation area (Australia), over 80% of the landscape has water tables within two meters of the ground surface. SWAGSIM can help in estimation of various factors that are important in managing shallow water tables in irrigated areas.

3-7 It has a basic soil water and groundwater simulation model at its core and was developed by Prathapar, Meyer et al. (1996). SWAGSIM framework takes into account physical relationships between soil and water (with certain assumptions) that include run-off, macro-pore recharge, net water flux, water flow in unsaturated zones, and interactions between confined and unconfined aquifers. The conceptual framework of SWAGSIM (Prathapar et al., 1996) is shown in Figure 3.1. A brief review of the conceptual framework is presented below. Further details are contained in Prathapar, Meyer et al. (1996).

3-8

iration iration

p oration -flow -flow p p p u y Rain Rain Eva Mole Drains Pum Trans Irrigation Tile Drains Tile Drains Crack Recharge Crack Recharge Run off illar Rivers/Drains Rivers/Drains Basins Evaporation SOIL SURFACE p Soil Water Recharge Recharge Soil Water Ca Root-zone SUB-SOIL Water table WATER TABLE

Unconfined Aquifer Groundwater Flow Semi Permeable Aquitard

Confined Aquifer Downward leakage leakage Downward Upward leakage

Figure 3.1 Conceptual framework of SWAGSIM (Redrawn from Prathapar et al., 1996)

3-9 SWAGSIM can analyse where recharge and discharge are occurring in an irrigation sub-region by tracking groundwater movement at sub-regional levels of irrigation bays. SWAGSIM can also determine how much pumping is needed to control soil salinisation and to estimate rising groundwater levels and discharge into streams. These modules allow SWAGSIM to model regional water table fluctuations to locate recharge or discharge zones, and to calculate the rates of these processes. It can also determine how much pumping is needed to control salinisation and estimate groundwater discharge into streams.

SWAGSIM utilises two different grids: the first grid represents homogenous land use units (e.g., yields, paddocks, irrigation bays) within the irrigation areas. The land use units are used to determine recharge or discharge throughout the soil surface on a particular day. SWAGSIM assumes that a unit has a single land use, with rain and irrigation supplied. The second grid is used to estimate other vertical flux components (i.e., recharge, groundwater pumping, tile and mole drainage, river recharge, leakage) to solve the finite difference approximation of the groundwater flow equation. SWAGSIM incorporates variable grid sizing within the finite difference mesh.

3.2.1.2 Grassland Water District Decision Support System (Quinn and Hanna, 2003)

Grassland Water District Decision Support System (GWDSS) was developed to aid the US Environmental Protection Agency’s efforts in regulating salinity discharges from non-point sources in the San Joaquin Valley of California, USA. It was developed to improve management of seasonal wetlands. Seasonal wetlands of the grassland water district are over 20,000 hectares of privately owned wetlands that provide a habitat for wildfowl on the pacific flyway. These wetlands are managed to meet habitat requirements by flooding and releasing waters in autumn and spring respectively. Spring releases are discharged into tributaries of the Lower San Joaquin River (SJR). These releases, in combination with agricultural drainage contain varying amounts of stressors like dissolved solids, boron and selenium that frequently exceed the water quality objectives of San Joaquin River.

3-10 GWDSS was developed to provide a set of analytical tools for assistance in calculation of the GWD water requirements, estimation of wetland salinity load in seasonal wetlands and in the selection of the best management practices. The decision support system was designed under four main constraints: 1) simple in design, 2) intuitive, 3) similar to the data management tools typically used by the Grassland Water District, and 4) able to work in synergy with the existing SJR water quality forecasting models and software.

The decision support system consists of a core water quality model designed in a Microsoft Excel spreadsheet. This water quality model works with another model, the Delta Simulation Model II (DSM-2). It develops salinity balances both at the regional as well as local scale and incorporates weekly water use requirements of the major wetland habitats types and adjacent areas. It provides wetland outflows and salinities to the DSM-2 for use in the river forecasts, then it uses DSM-2 output (SJR assimilative capacities as a input).

The decision support system uses four types of input data: static, annually constant, annually varying, and real time. The static data includes soil properties, land classifications, acreages, drainage basin allocations, precipitation and evapo- transpiration qualities. Annually constant data includes crop coefficients, best management practices and water table depth. Annually varying data includes precipitation, water year classification, air, water and soil temperatures, and water table depth. Annually varying data includes precipitation, water year classification, air, water, and soil temperatures, irrigation schedule, and wetland flood-up schedule. Real time data includes supply of water quantity and quality, evapo-transpiration, precipitation and assimilative capacity of the San Joaquin River. Much of the static and annually constant data are assumptions.

The decision support system works on two geographic scales (i.e., regional and local). The regional scale concentrates on deliveries to and exports from GWD while the local scale focuses on individual wetland units where more intensive monitoring is conducted. By running scenarios of different weekly wetland fill and release schedules and annual changes in vegetation type and water bird usage, the managers

3-11 are able to plan operations to minimise water quality impacts on SJR while maximising wildlife benefits.

3.2.2 Spatial models

Spatial models of dryland salinity aim at mapping the extent of dryland salinity. They use remote sensing and satellite images or aerial photographs in a geographic information system (GIS) framework to generate such maps. These models are useful in understanding the spread of dryland salinity at a certain point in time or change in area under dryland salinity. However, these models do not address the underlying causes of dryland salinity or provide a framework for policy analysis. An example is the fuzzy landscape analysis GIS (FLAG) method for dryland salinity assessment (Roberts et al., 1997). FLAG uses a digital elevation model to drive data for the shape and curvature of landscapes. These data are then combined with the fuzzy set theory and water-cycle assumptions to predict the likely extent and location of dryland salinity (Roberts et al., 1997).

Malins and Metternicht (2006) presented an application of the fuzzy modelling approach to predict the distribution of the secondary dryland salinity. They extended the fuzzy landscape analysis Geographic Information System (Roberts et al., 1997) by incorporating geological and vegetation data. The model assumes that salinity occurrences in the landscape is a function of topography, closeness of the water table to the soil surface and its relative correspondence to variations in the topographic surface, and the uphill area contributing to potential discharge at a point (Malins and Metternicht, 2006). Fuzzy logic uses graded or qualified statements rather than the ones that are strictly true or false (Zadeh, 1965). Fuzzy models can deal with imprecise, uncertain or ambiguous data sets and knowledge or relationships among data (Malins and Metternicht, 2006).

3.2.3 Bio-economic models

Bio-economic models came out of economic modelling circles and have tended to be used for theoretical considerations (Litcher, 2001). They are usually based on optimisation frameworks in which a biophysical system is represented by a series of

3-12 coefficients using various mathematical programming algorithms. It uses theory underlying hydrological process as well as that of economic analysis. Often in these models physical processes are linearised MIDAS. Examples of such models are Model of Integrated Dryland Agricultural Systems (MIDAS) and Salinity and Landuse Simulation Analysis (SALSA). MIDAS is a whole-farm linear programming model with an emphasis on biology and economics. The MIDAS model has been applied to a diverse range of issues including supply elasticities of farm commodities, evaluation of new technologies, and trade-offs between economic and environmental objectives (Abadi and Pannell, 1998).

3.2.3.1 Salinity and land use simulation analysis (SALSA)

SALSA (Heaney and Bell, 2001) was developed at the Australian Bureau of Agricultural and Resource Economics (ABARE) with the collaboration of Commonwealth Scientific and Industrial Research Organisation (CSIRO). The basic purpose of the model was to evaluate salinity management options such as land use change, improvements in irrigation practices and engineering in terms of economic costs.

The model has two components: agro-economic and hydrology. The land management problem in the agro-economic component of the model is maximising the economic return from the use of agricultural land by choosing between alternatives and steady state land use activities in each year. There are seven possible land use activities: irrigated crops, irrigated pastures, irrigated horticulture, dryland crops, dryland pasture, alternative cropping/pasture activity with reduced recharge and plantation forestry.

Each region is assumed to allocate its available land between these activities each year to maximise net returns from the use of land in production subject to constraints on the overall availability of irrigation water (from rivers, groundwater) and suitable land.

This basin scale model is composed of a series of 14 catchments and 11 irrigation areas in the Victorian Mallee and South Australian river-land linked through main river channels in the Murray Darling Basin. Surface and groundwater discharges

3-13 from a catchment combine to contribute to a salinity level. In turn, each catchment is divided, on the basis of the type of the groundwater system present, into sub- catchments or land management units. The land management units are linked to each other through surface flows in the form of runoff and groundwater discharge into streams and rivers. The model assumes that the externalities will be generated between but not within, land management units, hence the land management unit is represented by a single land manager.

Figure 3.2 Treatment of land management units in SALSA (Heaney and Bell, 2001).

After using SALSA in the grapevine industry, Alexander and Heaney (2003) considered SALSA underestimates the impacts of the rising stream salinity on the yield and producer’s returns from grapevine. Moreover SALSA is used by ABARE, and because of the complexity of the system, users are not allowed to use SALSA themselves.

3.2.4 Dynamic Models

In contrast with the models described above which attempt to explain the mechanisms by which salinity arises, dynamic models provide time-based causal analysis and options for simulations in which many years of elapsed time can be simulated in minutes. This offers the opportunity to examine suitability of alternate

3-14 causal explanations. Dynamic models are typically represented with difference equations or differential equations and use a different approach to model development. These models are based on investigations of cause-and-effect and how these produce changes over time. Some models are based on System Dynamics.

An example of such models is the Dynamic Model of Salinization (Saysel and Barlas, 2001) developed at the Bogazici University in Turkey. It is based on System Dynamics (Sterman, 2000, Forrester, 1968), a methodology specifically designed for long-term, chronic, dynamic management problems and will be discussed in detail in Chapter 4.

Saysel and Barlas’s (2001) model is constructed by variables categorised as stocks, flows and convertors. Stocks are state variables and they represent accumulations in the system. Flow variables represent the rates of change that cause variations in stock variables (accumulations or mathematical integrations) to occur. Rate controlling variables represent the activities, which fill in or drain the stock. Convertors are variables used for miscellaneous calculations.

The model aims at a generic long-term process of salt accumulation in the irrigated areas and proposed a System Dynamics model of salinity consisting of a simulation model and feedback loops. Feedback loops underlying model are presented in Figure 3.3.

3-15 Critical discharge level Critical watertable level

+ Discharge Level - Discrepancy water table level + discrepancy +

1 (-) Watertable 2 (-) Level + - + - + Groundwater discharge Groundwater intruding ro o tz o ne

Percolation - drainage

Salinity Salinity + salinity groundwater groundwater irrigatio n + + discharge + water Groundwater discharge Salinity drainage water 2 (+) +

Drainage Salinity infiltratio n 2 (+) + + Groundwater intrusion Salinity intruding 3 (+) Salinity ro o tz o ne groundwater + + + Salinity evapo rating groundwater

+

Figure 3.3 Feedback structure of the dynamic model of salinisation (Saysel and Barlas 2001).

3-16 In this model, three positive feedback loops represent the salinisation processes that are activated by drainage discharge, groundwater discharge and groundwater intrusions into the soil root zone. As the root zone salinity rises, it increases the salinity of infiltration water, salinity of drainage water and consequently the salinity of irrigation water increases through drainage discharge into freshwater supplies. As the salinity of irrigation water increases, so does the salinity of infiltration water and this in turn further increases the root zone salinity. The water table level is controlled by two negative feedback loops representing groundwater discharge and intrusion.

3.2.5 Integrated modelling frameworks and the practice of using multiple models

Integrated frameworks use two or more different disciplinary models and link them for a certain purpose. Such models in the salinity context include an economic component that account for the direct costs of management decisions (Letchure, 2001). Models derived from economic or social sciences tend to contain fairly simplistic mathematical representations of physical and hydrologic processes. There are a limited number of modelling efforts where the treatment of different disciplines has been complex, so that the disciplinary models used for each of the component processes are largely acceptable to the disciplinary proponents of each of these areas.

An example of such integrated frameworks is the spatial optimisation model for analysing catchment management (SMAC) which was developed to analyse land management options on dryland salinity in eastern Australia. SMAC links farm level decision making and catchment scale hydrological processes and uses mathematical programming to investigate optimal strategies for resources uses.

There also exists practice of using multiple models in which the models are not interrogated to each other but are used for separate purposes at different stages of research. Daamen, Hoxley et al. (2002) used two models, Soilflux (a model of vertical soil water movement) and MODFLOW (a groundwater model) and airborne geophysical data to evaluate land management options for dryland salinity in Kialla East, Shepparton, Victoria.

3-17 3.2.6 Information frameworks

Information frameworks do not present any specific model but aim to help decision maker by providing information on multiple aspects in the form of fact sheets, maps, charts etc. An example of such frameworks is the National Dryland Salinity Decision Support tool.

National Dryland Salinity Program (NDSP) of Australia introduced a decision support tool to create awareness and help farmers in selection of engineering options for their land. It consists of a number of case studies (fact sheets), a groundwater map, site evaluation tool and a questionnaire.

Case studies cover groundwater pumping, shallow surface drains and seepage interceptor drains, deep surface drains and subsurface drains. The NDSP DST provides nine fact sheets that describe different types of drains and provides guidance and puts technical information at the disposal of the decision maker. However, it does not address the feedback interactions in a systemic way.

3.3 ISSUES IN MODELLING FOR MANAGEMENT OF DRYLAND SALINITY

Most of the current models address physical processes and/or economic consideration, or integrate models to each other for increasing modelling capability. Some of these models address the static complexity in the optimisation framework. These models are developed at different scales, for example, property scale, catchment or sub-catchment scale, river basin scale or regional scale.

Most of the models used simplifying assumptions about the processes (Littleboy and Vertessy et al., 2001), however, while developing simplifying assumptions, these models ignore many of the core requirements of the systems enquiry. It may be due to the different views of the world or the ontology these methods are based on. A review of these models from systems perspective indicates that these models have narrow boundaries, don’t attempt to capture emergent properties, dynamic complexity, time delay or the complexity associated with those processes. While an absence of these issues from modelling frameworks reduces the research problem to

3-18 a measureable phenomena in which data can be generated and observations made, the usefulness of such models in managing dryland salinity is seriously compromised as these processes do not work in isolation rather as a part of other processes as well as other parts of the system. A brief overview of such issues is presented in the following paragraphs.

3.3.1 Model boundary and varying levels of focus

A model boundary contains the variables used in building that model. The insights drawn from the model, future policies based on the model, and the outcome of implementation of those policies all depend upon the boundary assumptions of the model. Forrester (1994) suggested that variables important for the problem must be included in the model. Checkland (1981:312) defined boundary as:

In the formal systems model, the area within which the decision taking process of the system has power to make things happen, or prevent them from happening. More generally, a boundary is a distinction made by an observer which marks the difference between an entity he takes to be a system and its environment.

Churchman (1970; cited in Midgley, 2000) considered boundaries as personal or social constructs, as different problem boundaries are perceived by different stakeholders of a problem, for example, managers, farmers, policy analysts, environmentalists or animal rights workers see different boundaries around the same problem. Midgley (2000:150) suggested that consideration of the boundaries as real world entities or personal or social constructs will depend upon the boundaries of analysis and the theories used. As the boundary of a model changes, stakeholders, actors and their roles will change as will the level of interconnectedness, and the consequent results drawn from the model.

Most of the models discussed in the preceding sections draw a very narrow boundary around the dryland salinity problem and include the variables that can be easily measured. This approach can help in understating a process in isolation but has limited utility in understanding functioning when these processes are interconnected with other parts of the system. In order to be useful, the modelling for dryland

3-19 salinity should consider the boundary issue very seriously. Systems philosophy can certainly help in defining problems boundaries.

3.3.2 Emergence

Within systems way of thinking, there exists a concept of a hierarchy of levels of organisation, i.e., each level is more complex than the level below (Checkland 1981, Garner 1991). Each level is characterised by emergent properties that do not exist below. Dubrovsky (2004:114) suggests:

The doctrine of emergence makes a much stronger statement that unity emerging out of the relationship of parts is something ontologically ‘new’, ‘additional to’ and different from parts and their relationship. It can neither be predicted from the examination of the constituent parts and relationship nor reduced to them. (Dubrovsky, 2004:114).

Further elaborating the concept, Senge (1990) suggests that dividing an elephant into two does not create two elephants. In this sense, emergent properties are the characteristics of a certain level of an organisation and these properties cannot be observed at other levels of the organisation. (Dubrovsky 2004:114) further suggests that:

Because the emerging unity is always new and unpredictable from its parts and relationship, the only way to determine it is to have it already emerged and known, and, therefore in actual investigation it is never new or unpredictable. (Dubrovsky 2004:114)

It implies that secondary dryland salinity in Australia is a change in the behaviour of the food production system resulting in reduction in land and water quality leading to a decrease in agricultural production and damage to rural and regional infrastructure. The growth of dryland salinity is an emergent property of the level of food production systems where physical, biological, social, economic and political processes interact to produce food. The models, discussed above, prepared to predict individual processes are not equipped to identify those emergent properties that are attributable to the level where food production system changes its behaviour.

3-20 3.3.3 Dynamic Complexity

In most management situations, the real leverage lies in understanding dynamic complexity, not detail complexity (Senge 1990). Gaining useful insights into the complex dynamic nature of dryland salinity requires the use of methods that are suitable for addressing dynamic complexity. Senge (1990:71-72) suggested that dynamic complexity exists in situations where:

 cause and effect are subtle, and where effects over time of intervention are not obvious;  when the same action has dramatically different effects in the short run and the long;  when an action has one set of consequences locally and a very different set of consequences in another part of system; and  when obvious interventions produce non-obvious consequences.

Sterman (2000, 2001, 2006) suggests that the dynamic complexity arises because of the following characteristics of the systems:

 Continually changing;  Tightly coupled: it implies that the actors in the system interact with one another and with the natural world in close relationship, i.e., change in one part of the system brings change in the other parts.  Governed by feedback: systems exhibit intricate networks of feedback processes (Sterman 2006). A feedback system is influenced by its own past behaviour (Forrester 1968).  Non-linear: historically limitations of mathematical analysis forced exclusion of most non-linearities from the models of social systems (Forrester 1987). It implies that the effect of a certain cause is rarely proportional to its cause, and what happens locally in a system, does not happen in other states of the system. Non-linearity often arises out of the basic physics of the system.  History dependant: systems exhibit history dependence. Many actions are irreversible and path dependant. Adopting a certain course of action precludes other actions and may determine the destination.

3-21  Self-organising: it implies that systems organise themselves over time. The dynamics of the systems generally arises from their internal structures. Small random changes are amplified and augmented by the feedback structure that generate certain patterns in space and time.  Adaptive: within a complex system, actors interact with the system and the systems responds. As a result of such interaction, the capabilities and behaviours of the actors change over time or as a result of some other factors such as short term objectives, maximising goals or beliefs. This change in actors’ behaviours enables them to adapt to the changes in the system.  Characterised by trade-offs: an intervention into the system brings both short- term and long-term impacts. The actors within a system make trade-offs between one set of impacts and others. These trade-offs are due to long- and short-term impacts. Sterman (2001, 2002, 2006) suggests that generally high leverage policies generate worse results before they improve the system behaviour.  Counter-intuitive: the behaviour of the complex systems is counter-intuitive in a way that it does not conform to conventional wisdom (Forrester 1994). The reasons for this counter-intuitive behaviour are attributed to cause and effect being distant in time and space while there is a tendency among researchers/managers to look for causes near the events that are subject of explanation.  Policy resistant: it is the state of a system when generally considered good policies do not bring good outcomes. The main cause of this policy resistance is that the complexity of a system overwhelms human ability to understand it.

Traditional problem solving approaches that are part of early education and training have developed an event view of the world (Sterman 2001). Within this world view, problem are defined as a gap between a perceived situation and a desired goal (Sterman, 2001; Kepner and Tregoe, 1981) that leads to a decision after evaluation of options. Decisions are implemented giving an outcome or result. This view is presented in Figure 3.4.

3-22 Feedback view of the world differs from traditional world view as it presents the endogenous view of the problem, i.e., a decision alters environment leading to new decisions. Decisions also trigger side effects, delayed reactions, changes in goals and intervention by others and such feedbacks may lead to unanticipated results and ineffective policies (Sterman 2001). The results of interventions define future situations. This feedback view is presented in Figure 3.5.

Generally, people recognise few feedbacks (Sterman 2000:28-29) and a failure to focus on feedback in policy design has critical consequences.

Goals

Problems Decisions Results

Situations

Figure 3.4 Event oriented view of the world (Sterman 2001)

Decisions

Goals Side Effects

Environment

Goals of Other Agents

Actions of Others

Figure 3.5 Feedback view of the World (Sterman 2000, 2001, 2006)

3-23 3.3.4 Time delays

Time delays are a source of dynamics. In complex systems, the results of an intervention may appear immediate, in a short time and over a long time or in a different part of the system. Time delays between an action and a system’s response are among the sources of dynamics. Delays in feedback loops may introduce instability into a system and may also increase its tendency to oscillate (Sterman 2000, 2001). Time delays are part of natural systems, for example, delay between sowing of seeds of a crop and its harvesting, or time delays between intervention into an ecosystem, for example, deforestation and its response on the water quality and its flow into streams. While being part of the natural processes, these time delays are often ignored in the conventional modelling of dryland salinity.

3.3.5 Complexity and modelling

Kline (1995) developed an index of complexity to describe the various levels of complexity on the basis of independent variables needed to describe the state of the system, independent parameters needed to distinguish the system from other systems in the same class and the control feedback loops both within a system and connecting the system to its surroundings. Kline’s complexity index provides a way to differentiate between different problems on the basis of their relative levels of complexity.

On the basis of complexity, Kline (1995) classified systems into the following classes:

 Paradigmatic systems of physics, chemistry and simple engineering analyses are simple systems used to study a particular case. Within such systems there are no feedback loops and the number of parameters is always fixed. Complexity index of such systems is more than 105;  Systems of human designed hardware combine the simple physics, chemistry and simple engineering analyses to create relatively more complex behaviours. Such systems have more numbers of parameters and therefore their complexity index systems is more than 106;

3-24  A single human being is much more complex. A human being has a greater number of parameters, and feedback loops. Complexity index of such systems is more than 109;  Human social systems contain networks of humans and are relatively more complex as these contain more independent variables and the complexity index of such a system is more than 1011;  Socio-technical systems are complete systems of coupled social and technical parts which humans erect and operate primarily to control the environment and tasks that cannot be done without such systems. Complexity index of such systems usually exceed 1013.

On the basis of complexity levels, Kline (1995) suggests that when the number of independent variables is more than five, investigators hold the values of parameters fixed and seek simpler models to analyse some components of the system or some aspects of the complete system even when the number of feedback loops is zero. It implies that a socio-technical system is very complex for full analysis via analytical or computer processes as the adequate system representations cannot be developed for the entire system. For such complex systems, Kline (1995:62) presents the following guideline for inquiring into complex systems:

In very complex systems, such as socio-technical systems, we have no theory for the entire system, and must therefore create, operate and improve such systems via feedback that is, repeated cycles of human observations plus trials of envisaged improvements in the real system. In such very complex systems, data from a wide variety of cases, therefore, become the primary basis for understanding and judgements, and should take precedence over results from theory based on cuts through the hyperspace (called the primacy of data).

The food production system exhibiting secondary dryland salinity is a highly complex system that falls into Kline’s (1995) category of socio-technical system in which humans use production technologies and makes decisions on the use of resources. It involves humans, complex hardware of many kinds and feedback loops in which the past actions of producers, policy makers, perceived natural resources capacities and limits and ecosystems services and functions shape the current state of

3-25 affairs. The feedback loops in the dryland salinity problem are presented and discussed in detail in Chapter 6.

Estimation of the complexity index of such problems is very difficult (Kline 1995), however, the concept of a complexity index can aid in appreciation of the level of complexity implicit in the dryland salinity problems and in identification of analytical tools for its strategic management. On Kline’s (1995) complexity index scale, the dryland salinity problem is likely to be in the class of problems that have a complexity index of more than 1013. That means the conventional problem solving methods and dryland salinity modelling—though useful in understanding salt accumulation processes and costs associated with remediation measure—would be inadequate in gaining using insights for strategic management of dryland salinity. Such models are also limited in ability to provide understanding at the systems level how these particular physical processes that are subject of a certain model are linked to other processes or the systems behaviour, or the emergent properties of the system.

3.4 SUMMARY AND CONCLUSION

This Chapter presented examples of the current models of dryland salinity in Australia along with examples of two models from other countries. Various physical, spatial, bio-economic, dynamic, and integrated models were examined.

While these models provide understanding at the process level, their ability to provide insights for managing the dryland salinity problem is seriously compromised. These models do not address the issues that are important for understanding the dryland salinity problem in Australia. These issues include complexity, problem boundary, emergent properties, dynamic complexity and time delays. These issues need to be considered, if useful insights are to be gained to assist the management of dryland salinity.

Moreover these models are based on the methodology derived from theory and come along with assumptions underlying the theory. In such instances, philosophical analysis can inform about those assumptions as will as provide guidance in identifying problems boundaries and decision space (boundary judgement) available within a model as suggested by Midgley (2000) and Churchman (1994).

3-26 As argued in Chapter 2, dryland salinity in Australia has some characteristics of complex dynamic problems. Successful intervention into complex dynamic systems requires more than technical tools and mathematical models (Sterman 2001). In the next two Chapters, methodological approaches that are suitable for addressing complex dynamic problems are discussed. The next Chapter reviews the System Dynamics method.

3-27 CHAPTER 4 SYSTEM DYNAMICS

We cannot solve our problems with the same thinking we used when we created them. (Albert Einstein)

The issues in addressing complex problems have challenged human minds and attempts were started to address such problems through different ways of thinking. With the introduction of computers in the middle of the 1900s such efforts took a leap. System Dynamics concepts were introduced in early 1950 at MIT’s servomechanisms laboratory, as a part of Jay Wright Forrester’s (1918- ) research on dynamic problems. Afterwards such concepts were compiled overtime and System Dynamics emerged as a discipline for addressing complex dynamic problems.

Some of the characteristics of the dryland salinity problem in Australia, as discussed in the preceding chapters, involve change in multiple aspects of the problem like ecology, human use of land and water, land clearing etc. over time. Such changes, the complexity of the dryland salinity problem, and the issues addressed in the complex problems encourage us to look for a method that can help in problem conceptualisation that treat dynamics complexity and help in the identification of strategies and policies that endure. System Dynamics is one of those methods that have been suggested in the literature reviewed on addressing complex dynamic problems. Further knowledge about the capabilities of System Dynamics may help in understanding its potential for problems like dryland salinity.

In this chapter, results of a literature review on multiple aspects of the System Dynamics method are presented. First the basic concepts about System Dynamics are described, then the discussion leads to the philosophy of and the purpose of inquiry in System Dynamics. An overview of System Dynamics methodology is presented (the details about different components are described in the concerned chapters that lead to a brief history of System Dynamics practice).

4-1 4.1 WHAT IS SYSTEM DYNAMICS?

4.1.1 Definitions

System Dynamics is a computer modelling technique that has its origins in control theory, cybernetics, organisational theory, behavioural psychology, economics, and digital computer simulation. It is used to build models of systems that are experiencing problems and/or exhibiting behaviours that are not well understood. The completed models are used as "laboratories" for testing policy changes aimed at improving system behaviour (Richardson, 1991; 2002).

Jay Wright Forrester introduced System Dynamics in the 1960s through his industrial dynamics and world dynamics models. Since then, System Dynamics has been used for understanding complex problems in all manner of domains (Coyle, 2000). Using this method fully specified quantitative models are built in which problems are simulated and important insights are generated into policies to improve system behaviour.

Since the 1960s, System Dynamics underwent different phases of development, from applications of control theory to a complex problem analysis framework. Overtime, the definition of System Dynamics changed with recent definitions describing it in terms of its framework for studying complex problems. The System Dynamics Society (SDS, 2005) has defined System Dynamics as below:

System dynamics is a methodology for studying and managing complex feedback systems, such as one finds in business and other social systems. In fact it has been used to address practically every sort of feedback system. While the word system has been applied to all sorts of situations, feedback is the differentiating descriptor here. Feedback refers to the situation of X affecting Y and Y in turn affecting X perhaps through a chain of causes and effects. One cannot study the link between X and Y and, independently, the link between Y and X and predict how the system will behave. Only the study of the whole system as a feedback system will lead to correct results.

4-2 Jay Wright Forrester (2001) defined it as:

System Dynamics deals with how things change through time, which includes most of what most people find important. It uses computer simulation to take the knowledge we already have about details in the world around us and to show why our social and physical systems behave the way they do. System dynamics demonstrates how most of our own decision-making policies are the cause of the problems that we usually blame on others, and how to identify policies we can follow to improve our situation.

System Dynamics deals with development of understanding how complex systems change over time and how internal feedback loops within the structure of the system influence the entire system behaviour. It has a graphical presentation along with a background mathematical model. The graphical model and the explicit presentations of the circular cause and effects, helps clarify concepts, and to communicate with a wider community across a variety of disciplines.

Forrester Winch (2000) suggested that an analysis in the absence of multi-loop, multi-state, non-linear characteristics of the feedback system is not System Dynamics but the expedient use of a convenient software tool. The practice of developing computer simulation models is widely emphasised, the soft systems thinking involving feedback thinking, capturing mental models, derived causal loop diagrams or the stock and flow diagrams is also considered as System Dynamics because of its close proximity to original concepts.

4.1.2 Purpose of inquiry in a System Dynamics study

The purpose of a System Dynamics inquiry is to improve understanding of complex systems. The System Dynamics modelling process fosters learning at each step from development of reference modes to the model development and validation. System Dynamics models are not specifically designed to generate predictions. The issue of the modelling purpose has been debated extensively in System Dynamics literature. Richardson (1981) writes:

4-3 The purpose of building System Dynamics models is to enable improved understanding of the relationships between feedback structure and dynamics behaviour of a system, i.e. problem, so that policies for improving the problematic behaviour may be developed.

A clear understanding of the purpose of a problem guides through the model building process and helps in deciding what variables are to be included in the problem boundary. Learning in complex systems can be difficult as it is hampered by a number of factors. Sterman (2000) described the following barriers to learning that the System Dynamics approach may help to reduce:

 Dynamic complexity.  Limited information.  Confounding variables and ambiguity.  Bounded rationality and misperceptions of feedback.  Flawed cognitive maps.  Erroneous inferences about dynamics.  Unscientific reasoning/judgemental errors and biases.  Defensive routines and interpersonal impediments to learning.  Implementation failure.

The process of problem analysis and model building helps to reduce complexity through alleviating these barriers.

To aid learning, the system dynamic models provide broader understanding about the particular problems in focus. System Dynamics models also work as learning laboratories or micro-worlds where cognitive capacities of managers can be enhanced.

4-4 4.2 PHILOSOPHY UNDERLYING SYSTEM DYNAMICS

System Dynamics needs a broader and deeper debate about its underlying philosophy, the contrast with alternative philosophies, the nature of knowledge, the role of subjective and observational information and the criteria for judging validity. (Forrester, 1980:15 cited in Lane 1990)

System Dynamics origin in engineering sciences and its evolution into a methodology aiding complex managerial decision-making has strongly influenced the process of System Dynamics inquiry, terminology and sets of beliefs. The general tradition in System Dynamics is to design projects that aim at solving specific problems—that means to model a problem not the system. Within elementary system texts, Forrester (1961:44, 60) advocates focusing on goals andthe specific questions to be answered. It makes the system dynamic enquiry process purposeful and interesting. Later System Dynamics literature restricts focus on undesirable system behaviour (Forrester, 1994; Grobler, 2008). Forrester usually calls ‘the first phase of a modelling project’ as ‘describe the system’ (Grobler, 2008) and in later System Dynamics literature Richardson and Pugh (1981), Roberts et al. (1983), Coyle (1996), and Sterman (2000) explicitly concentrate on ‘problem articulation’ or ‘problem definition’ as a first step in modelling. This may imply that in a ‘describe the system’ phase of System Dynamics enquiry, the problem specific system parameters are explored or in Sharp’s (1977) words “embedding a problem within a wider framework”. Sharp (1977) suggested that embedding a problem within a wider framework helps to avoid improving the performance of the one part of the system at the expense of the others. It also enables a problem that is fairly intractable to be located within a framework of policies that managers actually apply.

The fundamental idea of System Dynamics is that socio-economic and business systems can be regarded as continuous feedback control systems that have self- regulating properties by virtue of the technological and accounting relationships between system variables and the policies that are used to manage the system.

4-5 4.2.1 Principles of systems

The basic concepts of System Dynamics provide the basis for development validation and use of the System Dynamics models. These concepts have been described throughout the System Dynamics literature ranging from Forrester’s principles of systems (Forrester, 1968) to the most recent literature in System Dynamics. These fundamental concepts have been used within different contexts. In the road maps (MIT, 1994) these concepts have been compiled into 26 systems principles. These principles provide necessary guidance during model development.

In the following paragraphs, these principles are adopted from the road maps (MIT, 1994). Additional elaborations are added where required. These principles are used as a guiding principle in Chapters 6, 7 and 8 while describing the model:

1. The feedback loop is the basic structural element of systems. 2. Levels and rates are fundamental to loop substructure. 3. Levels and rates are not distinguished by units of measure. 4. Levels are accumulations (integrations). 5. Levels are changed only by the rates. 6. Levels exist in conservative subsystems. 7. Rates depend only on levels and constants. 8. Decisions are always within feedback loops. 9. Every equation must have dimensional equality. 10. First-order loops exhibit exponential behaviour. 11. Levels completely describe the system condition. 12. Variables have the same units within conservative subsystems. 13. Solution interval DT is in all level equations and no others. 14. Simple, second-order negative loops exhibit sinusoidal oscillation. 15. Goal, observation, discrepancy, and action create a system substructure.

4-6 16. Level variables and rate variables must alternate. 17. Higher-order, positive-feedback loops usually show exponential behaviour. 18. Conversion coefficients are identifiable within real systems. 19. Time constant of a first-order loop relates a level to a rate.

20. Rates are not instantaneously measurable. 21. Every system has a closed boundary. 22. Information links connect levels to rates. 23. Decisions (rates) are based only on available information. 24. Auxiliary variables lie only in the information links. 25. Mathematical simulation models belong to the broad class of abstract models. 26. Model validity is a relative matter.

4.2.2 Rationality in System Dynamics

System Dynamics models are presented as abstract representations of the actual physical and information flows in a system, their feedback implying that decisions are not entirely considered as freewill but strongly conditioned by the environment. Bounded rationality assumptions of System Dynamics gain strength from Simon’s (1957) following principle of rationality agreed to and repeated by Sterman (2000:598):

4-7 The capacity of the human mind for formulating and solving complex problems is very small compared to the size of the problem whose solution is required for objectively rational behaviour in the real world or even for a reasonable approximation to such objective reality.

In System Dynamics, models attempt to capture the real world decision making. In real-world decision making, people try to achieve some, occasionally conflicting, goals (which can change), use heuristics, and acquire only a certain amount of information about the issue they have to decide on (Grobler, 2004). The reasons for bounded rationality are clearly articulated by Sterman (2000:598). He suggests:

…the bounded rationality results from limitations on our knowledge, cognitive capabilities, and time. Our perceptions are selective, our knowledge of the real world is incomplete, our mental models are grossly simplified and imperfect, and our powers of deduction and inference are weak and infallible. Emotional, subconscious, and other non-rational factors affect our behaviour. Deliberations take time and we must make decisions before we are ready.

Grobler (2004) suggests that limitations and restrictions to rationality can affect the modelling process and learning from simulations in System Dynamics. Grobler (2004) further suggests two types of rationality issues in System Dynamics modelling practice, content and process bounded rationality. Grobler’s content view of bounded rationality implies that the artefacts of bounded rationality occurring in the real world must be represented in the formal model. The understanding improved through simulation experiments is also bounded by the process of modelling and simulation. Figure 4.1 depicts the content and process view of bounded rationality.

4-8

Figure 4.1 Bounded rationality in System Dynamics (Grobler, 2004)

The field of System Dynamics is evolving and improving its ability to address complexity prevalent in the social and economic systems. Since 1970, there has been growing development within the System Dynamics approach itself. There is a shift in focus from hard systems engineering to soft systems analysis and inclusion of causal loop diagrams. Coyle in his keynote address to the International System Dynamics Society Conference (1999) in Wellington raised some research questions about the quantification of qualitative models. He also suggested finding some formal measure of the extent to which uncertainties in formulating equations or obtaining data affect the reliability of the model.

4.3 SYSTEM DYNAMICS METHODOLOGY

System Dynamics uses an iterative process for model development a process which has been considerably changed since its inception. These major changes occurred in the ways models are built, interpreted and used. These changes have evolved into the

4-9 current form of the problem analysis and model development processes in System Dynamics. As a result model development processes are variedly described by different System Dynamics pioneers. In its current form, the main features of the System Dynamics method are described by the System Dynamics (2005) society as below:

 Identifies a problem,  Develops a dynamic hypothesis explaining the cause of the problem,  Builds a computer simulation model of the system at the root of the problem,  Tests the model to be certain that it reproduces the behaviour seen in the real world,  Devises and tests in the model alternative policies that alleviate the problem, and  Implements this solution.

The method can only be better understood with reference to its context of development. Various System Dynamics authors have outlined the model development process. The initial form of the method can be traced back to the pioneering work of the Jay Wright Forrester and the like principle of systems, industrial dynamics and urban dynamics. Within the context of industrial dynamics, Forrester has described the method as below:

 Identify a problem,  Isolate the factors that appear to interact to create the observed systems,  Trace the cause and effect information feedback loops that link decision to actions to resulting information changes and new decisions,  Formulate acceptable decisions policies that describe how decisions result from available information streams,  Construct a mathematical model of the decision policies, information sources, and interactions of the system components  Generate the behaviour through time of the system as described by the model  Compare results from all available pertinent knowledge about the actual system

4-10  Revise that model until it is acceptable as representation of actual system  Redesign, within the model, the organisational relationships and policies which can be altered in the actual system to find the changes which improve system behaviour.

Altering the real system in the directions that model experimentation has shown will lead to improved performance.

During the 1960s and 1970s, there were no significant changes in the methodology as Forrester described in early texts. Its focus remained on development of computer models. However, it was in 1980 when more attention was focused on the problem analysis part that causal loop diagrams came into existence. In a recent article, Forrester presents a graphical form of the method. Figure 4.2 depicts the reiterative

Step-1 Step-2 Step-3 Step-4 Step-5 Step-6 Convert Design Implement Describe description to Simulate alternative Educate changes in the system level and rate the model policies and and debate policies and equations structures strucutre

Figure 4.2 Reiterative process of System Dynamics Modelling (Source: Forrester, 1994) process of System Dynamics modelling procedure starting from empirical evidence (system description) and problem conceptualisation to model building (simulation) and validation.

Sterman’s (2000) description of the System Dynamics method is closely aligned to the outline of the method at the System Dynamics society website. He emphasises the iterative nature of modelling that starts from problem articulation and proceeds through formulation of dynamic hypothesis and subsequent model and leads to the

4-11 model testing and policy analysis and evaluation. A simplified loop of Sterman’s method is shown in Figure 4.3.

1. Problem articulation

5. Policy Formulation and 2. Dynamic Evaluation Hypothesis

4. Testing 3. Formulation

Figure 4.3 Overview of System Dynamics Method (Sterman, 2000)

Features that distinguish Forrester’s method from Sterman’s include causal loop diagrams and dynamics hypothesis. In Forrester’s initial as well as recent writings, there is no indication of the causal loop diagrams or the dynamic hypothesis that is emphasised by Sterman and other later authors.

Rapid growth of models over the last three decades has increased expectations from models. Keys (1988, 1990) termed System Dynamics as an amalgam of objective and subjective approaches and considered it as breaking through the paradigmatic incommensurability. Lane (1999) responded to Keys’s criticism by describing the nature of System Dynamics as a method and emphasising the need to specify the paradigm in which SD is being used with each context. The concept of paradigmatic incommensurability relates to Kuhn’s (1970) thesis that competing paradigms are incommensurable in that they cannot be compared by recourse to reason. Literature

4-12 (Every, 1998; Puttnam, 1988; Myerson, 1994; Feyerabend, 1975) indicates that there is difference of opinion about the radical incommensurability of paradigms. Puttnam (1988) concludes that there must be some correlation (or translation) of concepts and objects which different groups share that is definable to the extent that interpretation succeeds. The problems associated with paradigmatic incommensurability themselves originate from a particular school of thought (Kuhn, 1970; Feyerabend, 1975). One such response to paradigmatic incommensurability is dialogic rationalism. The proponent (Meyerson, 1994) of dialogic rationalism argues that despite heterogeneous epistemological paradigms, people ‘in the world’ still inter- relate on an everyday level. In System Dynamics models, the question of incommensurability can be handled in two ways; first by considering System Dynamics as a method and specifying the paradigm in which it is being used as Lane (1999) suggested; second by considering a room for paradigmatic dialogue on lines similar to dialogic rationalism.

Most economists’ (Page, 1973; Nordhaus, 1972) criticisms of System Dynamics stem from the Meadow’s Limits to Growth Model, as the concepts of System Dynamics and Limits to Growth Model were presented in the same decade. The process of System Dynamics was viewed from the same focal point as the Limits to Growth Model. Perman et al. (1996) summarises the criticism to the Limits to Growth as:

 Unreasonable growth projections.  Poorly specified feedback loops as it failed to take account of the behavioural adjustments operating through the price mechanisms.

While addressing this criticism, Perman et al. (1996) described economists’ criticisms as based on the assumption of the presence of well-functioning markets. In the absence of well-functioning markets, Perman et.al. (1996) concluded, price- induced substitution effects may not take place.

Inherent strengths and weaknesses of computer models have crucial implications for their applications in foresight and policy analysis (Sterman, 1991). Literature indicates the following sources of modelling errors:

4-13  Input data and difficulties in obtaining it (Humphrey, 1997; Sprauge and Watson, 1989).  Modelling approach itself (Baoguo, 1997).  Difficulty for the users to create their own models (Sprauge and Watson, 1989).  High predictive expectations from the models (Forrester, 1994; Sterman, 1991; Nuthmann, 1994).  Interpretation of results and more dependency on computers (Coyle, 1996).  Difficulties in keeping models up-to-date (Sprauge and Watson, 1989).

Some of these issues, like errors due to input data, can be addressed by preparing a quality assurance plan at the project conceptualisation stage and subsequent adherence to the plan during model development process. To remain understandable models should be simple (Saeed, 1994). Forrester (1994) and Sterman (1991) suggested the use of modelling for understanding reality rather than prediction. A possible answer to the questions comes from the work of Coyle (2000), Kleignen (1995), Forrester (1994), Sterman (1991), and Saeed (1986). The reservations (Saeed 1986, Forrester 1994, Sterman 1991) stem from the proliferation of ready-made methods and theories that limit mental involvement in decision making. Forrester (1994) mentioned that the purpose of modelling should be the understanding of reality which would in turn; provide a foresight that would help to identify different policy options and selection of an appropriate course of action. The use of models as a learning laboratory would help shape intuition and judgement and would develop a shared understanding of the problem that, in turn, would improve quality of decisions. The second response is from Coyle (2000) and Kleignen (1995) they suggest optimising System Dynamics models. Coyle (2000) further suggests the use of statistical methods for constructing reference modes.

4-14 4.4 SYSTEM DYNAMICS PRACTICE

4.4.1 A brief history of System Dynamics applications

System Dynamics emerged at the Massachusetts Institute of Technology (MIT), USA through the work of Jay Wright Forrester (1918- ) who worked on the feedback control systems (Forrester, 1989). Early texts include principles of systems, industrial dynamics (Forrester, 1961), Urban Dynamics (Forrester, 1969), World Dynamics (Forrester, 1971), Limits to Growth (Forrester, Meadows et al., 1972) and Principles of Systems (Forrester, 1968) were among the early texts on the subject that explain the method in its early development as well as the results of its applications. Industrial Dynamics was a textbook intended for students of management for understanding how policies, decisions, structures and time delays are interrelated in influencing the growth and stability of industrial organisations. Urban Dynamics described the dynamic model advancing understanding about the growth, stagnation and revival of cities. World Dynamics and Limitations to growth were written in response to the Club of Rome project “Predicament of Mankind”, and highlighted the resource limits to growth in global populations. The early days of System Dynamics were in development of discipline and its applications to the problems that were important at that time.

Overtime, System Dynamics has been widely used in policy analysis and resource management in North America, Europe, the Middle East and some parts of Asia (Saeed, 1979, 1982; 1987; Moxnes, 2000; Lane, 2000; CEC, 1998). One of the main uses of dynamic modelling had been to identify how feedback, non-linearity and delay interact to produce troubling dynamics that persistently resist solutions (Sterman, 1985; Morecroft, 1983; Forrester, 1969). Different policy options are tried and insights generated. A few examples of the use of System Dynamics for analysis of complex problems are given in the following paragraphs.

In the 1970s, Kumm (1975) employed SD for a study of development of economy and environment in Germany. The purpose of the project was to identify the environmental policy measures that help sustain quality of life. In 1978, the US

4-15 Department of Energy used System Dynamics models to evaluate each option during the congressional debate over natural gas deregulation. As a result, President Carter’s original proposal was modified dozens of times before the final compromise was passed (Department of Energy, 1979; Sterman 1991).

In the 1980s, there was a minor effort to use System Dynamics in macro- or socio- economic sciences. Vester (1990) used System Dynamics for development and testing of long term strategies for the Ford Company. System Dynamics was extensively used with much success for the study of ecological and biological systems. In USA, Costanza and his group used System Dynamics for the description of ecosystems (Bockstael et al., 1995).

The Commission of the European Communities (CEC) used System Dynamics to analyse the long-term non-marginal effects of the assessment of common transport strategies (CEC, 1998). In Australia, System Dynamics has gathered a body of knowledge and practitioners with applications in a variety of disciplines. For example Wolfendon (1999) used it for the complex ecological issues of catchment management. Linard (1991, 1996) used it for: a) analysis of policies in health and aged care; b) project management, and c) defence. Moreover grey (unpublished or not peer reviewed) literature indicates a considerable quantity of information on the use of System Dynamics for policy analysis in the public sector.

A search of the published literature on System Dynamics indicates that it has been applied to address problems in the following broad categories:

 Growth and stagnation of cities.  Global development.  Software development.  Emerging economies.  Health care policy.  Organisational learning.  Implementation of TQM initiatives.  Environmental issues.  Human Resources Management.

4-16  Problem analysis and risk management.

System dynamics has been introduced in school education where systems viewpoint and System Dynamics simulation provides a ‘learner directed environment’ to foster learning during concept building stages of individuals.

Despite these applications and the ability of System Dynamics to address the dynamic nature of problems, there exists a perception among the System Dynamics community that the method of System Dynamics has not been used on a wider scale by managers, strategists and policy makers in the public sector. The practice of System Dynamics had been relatively confined to universities and research institutes. Winch (2000:12) analysed data about the articles published in the System Dynamics Review, the official journal of the System Dynamics Society, between 1995 and 2000 and found that 75% of the published articles included at least one author from a university or a research institute. Results of Winch’s (2000) study are presented in Figure 4.4 and show authors’ affiliation with respect to the representing institution. Only seven per cent (7%) of the articles were authors from government agencies. System Dynamics literature is also published in other journals but the System Dynamics Review is considered a major specialist source of information for System Dynamics Practitioners.

80 70 60 50 40 30 20 10 0 Company Univeristy/ Govt Agency NGO/NfP Research inst

Figure 4.4 Articles published in the System Dynamics Review with respect to author affiliation. (Winch, 2000)

4-17 4.4.2 Quantitative and qualitative debate

The tradition in the main stream System Dynamics had been that problems can only be analysed and provide understanding through quantified models (Forrester, 1968). This tradition has been seriously criticised by certain analysts including Coyle (2000) who suggested the choice of the model—either qualitative or quantitative depends— upon the nature of a problem. He suggested that quantification of certain problems can pose potential risks and that these risks arise from the method’s emphasis on quantification and the use of different functions like multipliers that introduce risks of double accounting. Coyle (2000) gave examples of a number of case studies in which qualitative models can produce equal or superior insights as from those of the quantitative model. These case studies include: Maya civilization of Central America, treatment of psycho-geriatric patients, year 2000 problems of a major utility company, and Angola problems. In such problems, Coyle favoured the use of influence diagram and suggested the following benefits of influence diagrams:

 They put a very complex problem (that might need many pages of narrative explanation) onto one piece of paper.  They may be used as a helpful reminder during discussions.  Feedback loops from the diagram may help in explaining the behaviour.  Study of the diagram may provide a wider context to the modelling task.  A correctly drawn influence diagram is the basis for a quantified model and is easily transformed into equations.

The complexity of the problems that System Dynamics initiated to address has lead to a multitude of discussions among its practitioners. On one hand they started to question the value that a quantitative model can add to analysis, and on the other hand, methods started to crop up to overcome the limitations in the method. One example of such a method is the one proposed by Vennix (Vennix, Andersen et al., 1992; Vennix and Gubbels, 1992; Vennix, Scheper et al., 1993; Vennix, 1994; Vennix, 1995; Vennix, 1996; Vennix, Akkermans et al., 1996; Vennix, Richardson et al., 1997; Vennix, Thijssen et al., 1997; Vennix, 1999; Vennix and Rouwette, 2000)

4-18 who suggest group model building to enrich the problem analysis and modelling building efforts.

Quantitative model building in System Dynamics faces a number of challenges. As one move from a qualitative phase into a quantitative phase, the focus of approach becomes narrower and one faces very difficult trade-offs. In such situations, there remains a danger of ignoring some qualitative information/feedbacks in the process of improving quantitative precision/robustness. Such examples have been successfully demonstrated by Coyle (2000). Forrester (1994) suggested a variable that is important for the problem under analysis must be included in the model regardless of its ability to be quantified or not.

Issue of the take-up of System Dynamics is highly important for the System Dynamics community and there is a general feeling that it is not being taken up by managers. Warren (2004) criticised causal loop, and stock and flow diagrams of System Dynamics and suggested that there are inherent flaws.

4.5 SUMMARY AND CONCLUSION

This chapter described the method of System Dynamics along with its philosophical assumptions, methodology and applications. System Dynamics provides a framework for improving an endogenous point of view within a conceptualised boundary of a system. Key strengths of System Dynamics are its strong conceptualisation of dynamics involved and experimental approach to learning through simulation modelling. However, System Dynamics needs to attain the confidence of the modellers, model users, strategy developers and policy analysts to be able to effectively contribute towards complex problems. We need to ensure rigour in our thinking, analysis and strategy development. This may lead to characterisation of an improvement in the methodology.

The next chapter describes a multi-methodology framework in which System Dynamics can be used along with other systems disciplines for enhancing development and validation of System Dynamics models.

4-19 CHAPTER 5 A MULTI-METHODOLOGY FRAMEWORK

Systems ideas provide surest foundation for management practice. We have to acknowledge our limitations and always critically reflect on the nature and potential outcomes of the methods and techniques we adopt in seeking to improve social system Jackson 1995:39,41).

Allison and Hobbs (2006:81) reviewed natural resources management policies within the context of the Western Australian wheat belt and concluded that existing natural resources policies have failed to develop sustainable land management practices that can mitigate natural resources degradation. They partly attributed this failure to the uncertainty in asking the right questions and usage of methodological tools suitable for addressing such problems. The use of a single method might restrict the analysis within the theoretical framework, assumptions and analytical requirements of that method. Finding mitigation strategies for complex natural resource management problems like the WA wheat belt or dryland salinity might have analytical requirements that go beyond the scope of a single method.

Increasingly within the management science field, a variety of methods having utility in their own right are being combined in multi-methodology or multi-method research. The term multi-methodology is used to refer to the use of more than one method in addressing management problems frequently in an integrated way. The term methodology refers to a set of theoretical ideas that justifies the use of a particular method while the term method refers to a set of techniques used in a sequence to achieve a particular purpose (Midgley, 2000:105).

Multi-methodology or multi-method has its theoretical foundations in pluralistic thinking. Pluralistic thinking within systems movement emerged as a cure to the fragmented nature of management science (Torlak, 2001). Jackson (1987) provides early work on the conceptual foundations for pluralist thinking. Pluralist thinking appreciates different strands, encourages their epistemological development and suggests the ways in which these strands can be useful in addressing specific problems (Torlak, 2001). Methodological pluralism acknowledges the limits of individual methods, and hence the appropriateness, of specific methods to specific

5-1 questions (Norgaard, 1989). Midgely (2000) stresses that methodology must be based on sound theories which are inextricably linked to philosophy. Further lessons from applying the methodology or methodologies used in practice inform the revision of theories and development of the methodologies.

Mingers (1997:2) explains multi-methodology as:

[It is] combining together more than one methodology (in whole or part within a particular intervention. Thus it is not the name of a single methodology or even of a specific way of combining methodologies together. Rather it refers to the whole area of utilizing a plurality of methodologies or techniques or techniques within the practice of taking action in problematic situations.

Mixing of methodologies to address a particular class of problems is conceptually simple. However, the task could become complex, for example, when mixing ‘parts’ of methodologies from different paradigms (Watkins, 2003) and may lead to theoretical issues concerning sets of beliefs within paradigms. The users of multi- methodology must be aware of conflicts that might arise from incompatible theories and philosophies or at least that incongruity might exist. Researchers and practitioners need to remain cognisant of the theoretical, methodological and practical issues that arise when mixing and matching methods (Midgley, 2000).

The purpose of this chapter is to describe the multi-methodology framework developed and employed for this study and provide background knowledge and information to support the explanation of how the research was conducted, which follows in later chapters. Firstly, a review of literature on mixing and matching of methodologies, and their underpinning philosophies and issues is presented. This is followed by a discussion on candidate systems methodologies for inclusion in the framework. This discussion then enables the conceptual framework to be described. How it was applied is explained in subsequent chapters.

5-2 5.1 NEED FOR MULTI-METHODOLOGY

Mingers (2001) suggests the following three distinct benefits of multi-methodology:

 Real world problems have multiple dimensions. To address those problems one needs an approach that can address such multiple-dimensions as physical, social, political, and personal. The individual methods address different parts of problem situations. Multi-methodology is necessary to address problem situations as a whole with full richness of the methods available.

 Intervention process proceeds through different phases and these phases pose a variety of tasks. Within this process, some methodologies might be useful at a certain stage of intervention while others are applicable at other stages. Multi-methodology approach can provide appropriate methods for each stage.

 Combining different methods even if they perform similar functions provides an avenue for cross-verification through synthesis and integrated analysis of data from multiple sources for decision making. Such “triangulation” can increase both researcher and user confidence in the methods and, the consequent results.

In addition to the benefits that Mingers (2001) mentioned, there may be additional benefits such as method enhancement. Methods keep on evolving as a result of on- going research and applications, and are, therefore, at different levels of development and user confidence. Synergistic applications of methods can enhance their usefulness and in some situations, strengths of one method can reinforce strengths of other methods.

5.2 MULTI-METHODOLOGY PHILOSOPHY

Philosophical underpinnings of multi-methodology research trace back to Habermas’ three worlds that influenced the motivation for multi-methodology researchers like John Mingers and Gerald Midgely. Habermas (1984) takes a philosophical view that there are three worlds; firstly, there is a physical world that existed without humans and might exist without human actions; the second world, called the social world is

5-3 the one that humans have created as a result of their interactions with each other and nature—this is the world that humans share and participate in—it consists of language, meaning associated with language, social relations, practices, norms, rules and resources that both enables and constrains our actions; the third world is called a personal world. It is the world of an individual’s thoughts, emotions, values, beliefs and experiences. Habermas’s three world’s framework is presented in Figure 5.1.

5-4 Our social world The material world Inter-subjectivity Objectivity observation participation Reproduces Moulds

Actions Language

Enables and Constrains constrains

Appreciates Emotions Express

My personal world Subjectivity experience

Figure 5.1 Habermas’s three worlds (adapted after Mingers, 2001)

5-5 Mingers (2001) found Habermas’s three world framework very useful in analysing problems with multiple dimensions and suggests that it is always wise to utilise a variety of approaches. However, the use of a variety of approaches brings several potential problems that need to be addressed and the user of multi-methodology needs to remain cognisant of these theoretical and related issues. Such issues include, philosophical, cultural and psychological (Mingers, 2000).

Burrell and Morgan (1985) argue that all social theories are based upon a philosophy of science as well as upon a theory of society. Burrell and Morgan also contend that there are philosophical problems of using methods across paradigms of knowledge (cited in Pruyt, 2006). Burrell and Morgan (1979:85) based on human nature and the ontological, epistemological, and methodological assumptions (Table 5.1), suggest that the social sciences create the objective and subjective views that are diametrically opposed. The nature of society opposes those theories and methodologies which have a radical view of society compared to those that have a regulative view of society (Pruyt, 2006). Burrell and Morgan (1979, 1985) argue that crossing these subjective and objective views, four sets of fundamentally different assumptions are obtained, which constitute the four irrevocably incommensurable paradigms:

 Radical Humanism: social world as a psychological prison of economic alienation.

 Radical Structuralism: social world as a prison of structural economic forces.

 Interpretive Sociology: social world is what agents interpret it to be and can be observed.

 Functionalist Sociology: external social world exists, and laws and structures can be uncovered.

5-6 Table 5.1 The objective versus the subjective poles on the nature of social science axis and the meaning of the vocabulary. (Burrell and Morgan, 1979, 1985; Lane, 2001 and Pruyt, 2006) Subjective View Objective View

Ontology: what is the Nominalist: real world exists Realist: external world exists ‘nature’ of phenomena? as a product of appreciation outside of appreciation Epistemology: what Anti-positivist (humanistic) Positivist: causal laws, ‘knowledge’ can we obtain? knowledge is subjective deducted by objective And how? meaning observer Human nature: what is the Voluntarist: free will allows Determinist: humans react nature of human actions? humans to shape their mechanically to their environment environment Methodology: how can we Ideographic: access unique Nomothetic: measurement of obtain knowledge? individual insights and general concepts interpretations

Pruyt (2006) suggests that, paradigms and paradigmatic frameworks are nothing but artefacts of the human mind and they evolve as philosophical and scientific theories and thoughts and help to structure thinking or tools to classify ideas. Paradigms influence thinking and our mental models. Meadows and Robinson (1985:20) suggest that paradigm frameworks can also negatively influence modelling process:

Different modelling paradigms cause their practitioners to define different problems, follow different procedures, and use different criteria to evaluate the results. Paradigms deeply bias the way modellers see the world and thus influence the contents and shapes of models. This means that different frameworks could be (and are) proposed for specific purposes and in specific contexts, and that these frameworks could be extended and adapted when justified.

Morgan (1983, 1986) and Hassard (1991) cited in Mingers and Brocklesby (1997) suggest the need for a conscious pluralism in research practice due to the ontological and epistemological uncertainties associated with any single paradigm. Hassard (1991) also suggests that moving between paradigms in a single piece of research or paradigm mediation is difficult but not impossible. Individuals can be trained into new ways of thinking.

5-7 Within the context of multi-methodology, the issues of paradigmatic incommensurability has been debated strongly by Mingers (1997), Mingers and Brocklesby (1997), and Midgley (2000) and it is useful to be cognisant and appreciative of such philosophical issues about paradigms of knowledge while practicing multi-methodology. Further debate about paradigmatic incommensurability is beyond the relevance to the purpose and scope of this chapter.

Mingers (2001) describes the following types of multi-methodology research:

 Methodology combinations: Such combinations consist of more than one method in single intervention.  Methodology enhancement: In this type of research only one main methodology is used that is enhanced by using methods and parts of the methods.  Single paradigm multi-methodology: All the methods employed pertain to the same paradigm.  Multi-paradigm multi-methodology: In this type of usage, different methods from different paradigms are used.

The user of multi-methodology engages with a number of methods as opposed to the user of single methodology. Midgley (2000) calls the use of a single methodology as isolationist methodology as the user of isolationist methodology engages with only one methodology as opposed to a pluralist methodology (Midgley, 2000). Figure 5.2 shows the relationship of a pluralist methodology to the isolationist (a single methodology) and multiple methods as described by Midgley (2000). The circles show the type of methodologies, i.e., either pluralist or isolationist. The term isolationist is used here in line with Midgley (2000) and means the approaches that favour a particular method as opposed to the practice of using multiple methods.

5-8

Isolationist Pluralist Isolationist Methodology Methodology Methodology

Method 3 Method 5 Method 1

Method 2 Method 4

Figure 5.2 Conceptual relationships between a pluralist methodology and a single methodology (Midgley, 2000).

Within multi-methodology practice integrating or concurrently using a number of different systems methodologies offer opportunities for synthesis to be explored whilst improving the value of analysis. Munro and Mingers (2000) suggest a list of methods that are used in triads or pairs, such triads and pairs are shown in Table 5.2 and Table 5.3 respectively. For details about individual method, please refer to Rosenhead and Mingers (2001).

5-9 Table 5.2 Triads of methods (adapted from Mingers, 2001:305)

Method 1 Method 2 Method 3

Strategic choice Soft systems methodology Interactive planning Mathematical modelling Simulation Statistics Mathematical Modelling Simulation Heuristic Statistics Influence diagrams Cognitive mapping Statistics Strengths, weaknesses, Soft systems methodology opportunities and threats (SWOT) Statistics Soft systems methodology Cognitive mapping Statistics Project networks Forecasting Statistics Forecasting Inventory Soft systems methodology Viable system model Strategic choice Soft systems methodology Viable system model Total systems intervention Soft systems methodology Critical Systems Viable system model Soft systems methodology Interactive planning Critical systems heuristic Soft systems methodology Scenarios Critical systems heuristics Cognitive mapping Delphi Scenarios Hyper-games Delphi Scenarios Cognitive mapping Delphi Systems dynamics Cognitive mapping Decision analysis Strategic choice Cognitive mapping Influence diagrams Systems dynamics

5-10 Table 5.3 Pairs of methods (adapted from Mingers, 2001:305)

Method 1 Method 2

Simulation Statistics Forecasting Statistics SWOT Soft systems methodology Simulation Soft systems methodology Influence diagram Soft systems methodology Strategic choice Soft systems methodology Critical systems heuristics Soft systems methodology Soft systems methodology Interactive planning Soft systems methodology Cognitive mapping Statistics Soft systems methodology Viable system model Soft systems methodology Mathematical modelling Statistics Mathematical modelling Simulation Structured analysis and design Soft systems methodology Mathematical modelling Heuristics Decision analysis Strategic choice Decision analysis Cognitive mapping Statistics Cognitive mapping Influence diagrams Viable system model Influence diagrams Soft systems methodology Strategic choice Cognitive mapping Interactive planning Critical systems heuristics Strategic choice Interactive planning

5-11 5.3 CURRENT USE OF SYSTEM DYNAMICS WITH OTHER SYSTEMS METHODOLOGIES

Synergies of System Dynamics with other systems methodologies are being investigated increasingly by researchers faced with the need to address complex dynamic problem situations. A recent example is the work of Schwaninger (2004) on the synergistic use of System Dynamics and viable systems thinking. Constructing a Viable System Model involves defining management functions of an organisation and their inter-relationship in a highly structured way as the necessary conditions for the viability of any human or social system (Beer, 1981). Beer’s (1981) approach has been used for diagnosis and supporting design for social systems (Schwaninger, 2006).

Schwaninger and Rios (2008:169) report that System Dynamics and Viable System Model cannot be fused in an algorithmic way but that both methodologies can be used in a synergistic way as a powerful couple for following the reasons:

 both System Dynamics and Viable System Model are rooted in the systems approach;

 both methodologies are highly generic and therefore applicable to a great variety of situations;

 the objectives of both methodologies are complementary and in harmony;

 both methodologies are individually incomplete and mutually exclusive but collectively rather comprehensive; and

 both methodologies are connectable in functional and virtuous ways.

Another recent example of the current research into synergistic use of systems methodologies is synergies between System Dynamics and soft systems methodology. Rodriguez-Ulloa and Paucar-Caceres (2005) reports the combination of the soft systems methodology with System Dynamics into a framework they call the Soft System Dynamics Methodology. Rodriguez-Ulloa and Paucar-Caceres

5-12 (2005:331) reported the following benefits of combining System Dynamics with the soft systems methodology:

 It explicitly introduces the observer’s weltanschauung and the observer’s role. Weltanschauung is one of six elements in Checkland’s (1981) description of a root definition and refers to an outlook or framework or image that makes a particular root definition meaningful. It is an image or model of the world that makes a particular human activity system meaningful (Checkland, 1981:319).

 The soft System Dynamics methodology proposes and allows the implementation of desirable and feasible changes in the real world.

 The synergistic use allows, through the computer simulation over time, to measure and assess the kind and intensity of impacts, due to the behaviour of the variables studied in the problem situation as well as in the solving situation.

 The synergistic use of both methodologies allows the analysis of different possible interpretations of the problematic and solving behaviour of a situation in the real world.

5.4 A MULTI-METHODOLOGY FRAMEWORK TO ENHANCE SYSTEM DYNAMICS MODELLING PROCESS

A multi-methodology framework was developed to enhance the System Dynamics modelling practices through strengthening the problem conceptualisation and simulation model validation. The framework brings together cognitive mapping, System Dynamics and Systems Engineering for gaining strategic insights in addressing complex dynamic problems. A detailed account of all the three methodologies is provided along with a discussion on merits of including the three methodologies into the pro-offered multi-methodology framework.

5-13 5.4.1 Cognitive mapping/concept mapping

Cognitive mapping is a technique which supports the capture and graphical depiction of the mental models or thoughts of a particular individual about a problem (Eden, 1988). Theoretically, cognitive mapping is based on Kelly’s (1955) Theory of Personal Constructs. Such mapping is consistent with Habermas’s (1984) three worlds’ paradigm in that it specifically accommodates the personal subjective experience of individuals. Kelly (1955) suggests:

Man as a scientist continually checks the sense he makes of his world by using his current understanding (construct system) to anticipate and reach out for the future.

Cognitive mapping is a process for identifying those personal constructs. Eden and Ackermann (1998) suggest the following benefits of cognitive mapping:

 additional richness could be ascertained,

 the map is immediately useful to both the mapper and the interviewee, and

 the process is not constrained by any formal structure and can follow a natural conversation.

Within the context of the cognitive mapping method, McLucas (2003:210) defines a concept as a general notion and concepts range from hard, physical ideas or notions, to fuzzy ideas that involve or describe entities without measurable units (see also Eden and Ackermann, 1998:195-209). Concepts are interrelated in a causal structure using three main types of links; these are causal, connotative and conflict. Methods for developing concept maps are discussed in Chapter 6, section 6.2.

Concept mapping is a technique to elicit and visually represent information about complex concepts, and organise concepts within a certain domain for clarity and effective communication of those concepts.

The term cognitive mapping refers to cognition that belongs to an individual or one person and not to a group or organisation, while the term concept map is used to refer

5-14 to a map representing the views of several individuals in a single map (McLucas, 2003:9, 210).

Cognitive and concept mapping is used to map out the structure, to facilitate analysis and to interpret the perspectives that stakeholders have about a particular problem. Cognitive mapping can also be used in recording concepts during interviews (McLucas, 2003:9).

Cognitive and concept mapping has several advantages, according to Trochim, Jackson, (Trochim, 1989; Jackson and Trochim, 2002), McLucas (2003), Eden and Ackermann (1998), Gains and Shaw (1995), Seaman (1990), and Plotnic (1997), these advantages include:

 the process of developing cognitive maps keeps individuals (in case of cognitive maps) and groups (in case of concept maps) focused on objectives, a specific problem, and can be used for generating ideas, for example, brainstorming;

 both cognitive and concept maps are immediately available to the mapper, interviewee, or the group involved;

 results are quickly provided in an interpretable conceptual framework that can be used to design complex structures, for example, long texts and complex decision-making situations;

 the process of mapping is not constrained by formal structure but can follow a natural conversation (Brown, 1992 cited in Eden and Ackermann, 1998:295);

 expresses this framework in the language of the participants that can provide help in communicating complex ideas;

 results in a pictorial product that simultaneously shows all major ideas and their relationships. This pictorial product can aid in learning by explicitly integrating new and old ideas; and

5-15  can enhance the problem-solving phases of generating alternative solutions and options and often improves group or organisational cohesiveness.

The qualitative analysis in System Dynamics using causal feedback loops and stock has been criticised by Warren (2004) considering these diagrams having theoretical and pedagogical and managerial flaws. Warren (2004) specifically highlights that there is undue emphasis on seeking and closing feedback loops in System Dynamics. Warren (2005) concludes that assigning undue priority feedback loops may understate the importance of exogenous factors that would have been identified, had attention instead focused on identifying resource flows and their drivers:

Factors flowing in from, or out to, the environment are not only exceedingly common in practice, but are often highly influential. For example, the development of motor vehicle sales in Indonesia in the mid-1990s appeared to be exhibiting a tipping point (Gladwell, 2000), with rapidly accelerating growth. On closer examination, this tipping point was not significantly due to reinforcing feedback (e.g., word-of-mouth amongst new and prospective owners) even though a feedback simulation fitted the data exceedingly well. Rather, it was being strongly driven by increasingly large numbers of people crossing the income threshold at which they could afford new vehicles.

The strength of this mechanism was demonstrated when the economy turned down by just a small percentage, and new vehicle sales fell by over half.

Nothing in the business architecture of the suppliers in this market could conceivably have had any significant impact on this powerful exogenous factor. Nor could any reinforcing feedback amongst car buyers, whether deliberately encouraged by dealers or not, have had any significant impact on the collapse of sales.

The use of cognitive mapping/concept mapping can bring rigour to the qualitative analysis used in System Dynamics as there is no requirement of the feedback links to be closed in a concept map and as many concepts as available can be mapped. Moreover, it brings a method for bringing in the mental models of the participant if

5-16 used in individual consultations. The use of cognitive mapping to enrich qualitative analysis is demonstrated in Chapter 7.

5.4.2 System Dynamics

Details of the System Dynamics methodologies have already been presented in Chapter 4.

5.4.3 Systems Engineering

Sufficient rigour is needed to provide confidence in application of methods (Midgley, 2000) and such rigour might be provided to the System Dynamics simulation model development by the structured processes of Systems Engineering. Systems Engineering emphasises a top-down approach, a life cycle orientation, systems requirements, and an interdisciplinary team approach (Blanchard, 2004). The top-down approach helps to view the system as a whole and helps to gain an understanding how different components fit together. Life cycle orientation emphasises consideration of all phases of system development and ensures the efforts are not only focused on one or two phases compromising the other phases of the system development cycle.

There exist differences between Systems Engineering and System Dynamics in terms of modelling philosophy, methodology and practice. To develop a context to enable discussion of the likely benefits of using Systems Engineering in System Dynamics modelling, an overview of Systems Engineering is provided in the following paragraphs. The overview briefly discusses the Systems Engineering definition, its purpose, its modelling beliefs, its approach to model validation and the use of Systems Engineering models.

5.4.3.1 What is Systems Engineering?

Smith (1965) suggests Systems Engineering as an important tool for the managerial profession. Systems Engineering is a management function that controls the total system development effort for the purpose of achieving an optimum balance of all system elements. It is a process that transforms an operational need into a description

5-17 of system parameters and integrates those parameters to optimise the system effectiveness (DSMC, 1990). In a system development project, multiple experts and groups work and Systems Engineering facilitates their integration into a collective effort forming a structured development process that proceeds from concept to production to operation.

In Systems Engineering, a system has very specific attributes which withstand rigorous testing. The International Council on System Engineering (INCOSE, 2008) defined a system as:

…a construct or collection of different elements that together produce results not obtainable by the elements alone. The elements, or parts, can include people, hardware, software, facilities, policies, and documents; that is, all things required to produce systems-level results. The results include system level qualities, properties, characteristics, functions, behaviour and performance. The value added by the system as a whole, beyond that contributed independently by the parts, is primarily created by the relationship among the parts; that is, how they are interconnected (Rechtin, 2000).

Definitions of Systems Engineering are varied and represent an array of situations where it can be applied. The International Council on Systems Engineering (INCOSE) has provided a comprehensive definition of Systems Engineering that suits most situations. INCOSE (2008) defines Systems Engineering as:

…an interdisciplinary approach and means to enable the realisation of successful systems. It focuses on defining customer needs and required functionality early in the development cycle, documenting requirements, then proceeding with design synthesis and system validation while considering the complete problem [in] operation, performance, disposal, test, training and support, cost and schedule.

5-18 5.4.3.2 Goals of Systems Engineering

In Systems Engineering the main aim is to design something on the basis that we can define a boundary with interfaces and everything within that boundary exists to deliver some defined functionality. Blanchard (2004:258) provides the following general goals of Systems Engineering:

 Ensure that the requirements for system design and development, test and

evaluation, production, operation and support are developed in a timely

manner through a top-down, iterative requirements analysis.

 Ensure that system design alternatives are properly evaluated against

meaningful, quantifiable criteria that relate to all of the desired

characteristics; for example performance factors, effectiveness factors,

reliability and maintenance ability factors, supportability characteristics and

lifestyle costs.

 Ensure that all applicable design disciplines and related speciality areas are

appropriately integrated into the total design effort in a timely and effective

manner.

 Ensure that the overall system development effort progresses in a logical

manner with established configuration baseline, formal design reviews, the

proper documentation supporting design decisions, and necessary provisions

for corrective action as required.

 Ensure that the various elements (or components of the system are

compatible with each other and are combined to provide an entity that will

perform its required functions in an effective and efficient manner.

5-19 5.4.3.3 Models of Systems Engineering process

There are various models that describe Systems Engineering processes. The most prominent are the waterfall model, spiral model and the generic ‘V’ model (Forsberg and Mooz, 1994; Blanchard, 2004; Forsberg, Mooz et al., 2005). All these models list System Engineering processes with varying levels of detail and feedback processes. Both the waterfall model and the spiral model were introduced in the 1980s (Blanchard, 2004). The waterfall model shows the software life cycle in a progressive or ladder like structure, while the spiral model emphasises the continuous monitoring of objectives, strategies, design alternative, and validation methods (Forsberg and Mooz, 1994; Blanchard, 2004; Forsberg, Mooz et al., 2005).

As compared to waterfall and spiral models, the V model is relatively new. It was introduced in the early 1990s and reflects both the top-down and bottom-up approaches to system development (Blanchard 2004). The V model is usually represented in the shape of the English letter “V” and is also known as a “Vee diagram”. The first leg of the vee diagram shows the activities involved in the development of a system. These activities include concept development, needs analysis or requirements engineering, design and implementation. The second leg of the vee diagram shows the activities that are involved in the system integration, validation and operation. Examples of such activities include testing, subsystem verification, acceptance and operation and maintenance (Forsberg, Mooz et al., 2005). A generic Vee model is shown in Figure 5.3

5-20

Figure 5.3 Vee diagram of the Systems Engineering (adapted from Neudorff, Randall et al., 2003)

The Systems Engineering process can help in accomplishing critical tasks in a project’s success. According to Neudorff, Randal et al. (2003), these tasks include the following:

 identification and evaluation of alternatives;

 management of uncertainty and risk;

 incorporation of quality into the system; and

 successful handling of the program management issues.

Within the multi-methodology framework for this study, the V model was used. The ways in which it was applied is described in detail in Chapter 9.

5-21 5.4.4 Strengths of using Systems Engineering in conjunction with System Dynamics

Strengths of using Systems Engineering process for System Dynamics simulation model development accrue from:

 the modelling beliefs about the ‘system’ that Systems Engineering brings to the framework;

 Systems Engineering’s approach towards model validation; and

 the use of the models developed employing Systems Engineering processes.

There are philosophical differences between System Dynamics methodology and Systems Engineering in respect of validation and verification. The importance of verification and validation and independent verification and validation, where an agent is commissioned to ensure the processes one has followed are rigorous, has been long recognised in Systems Engineering. The same rigour does not appear to be needed in System Dynamics modelling or so it would appear from the System Dynamics literature on validation. This is also a consequence of having a ‘system’ as a convenient construct for the particular purposes of enabling System Dynamics analysis. System Dynamics represents a system view of the problematic situations that are system like but are not systems as perceived by systems engineers.

Part of the reason for this is that System Dynamics practitioners and researchers build models to help explain to themselves (i.e., learn) how various causal (feedback and delay mechanisms and system responses operate). Having established the insights they move on, often to build further models. The problems that System Dynamics practitioners and researchers often consider are highly complex (Forrester, 1975) and the best that might be done is to develop a much better understanding of the phenomena (usually involving a mix of hard and soft variables). In contrast, in Systems Engineering researchers and practitioners face the challenge of designing and building something (a physical system) that is relatively low in complexity but actually works. If a Systems Engineering practitioner does not apply rigour then this will be obvious as:

5-22  the physical system is ultimately built to satisfy the functional requirements established at the outset; and

 the system is expected to sustain function over a length of time.

Such functionality should also be required from System Dynamics models if these are to be used for management flight simulators or micro-worlds for training of executives in complex dynamics decision making. Such use of System Dynamics models have been advanced by (Morecroft, 1988; Senge and Lannon, 1990; Sterman, 1994, 1996, 1998, 2000).

In many Systems Engineering projects, there can be a long journey between establishing the requirements and final acceptance of a fully tested system, therefore, processes are critical. In System Dynamics modelling interventions, iterations are short-lived, the requirements to build a simulation model or conceptualise are less likely to be clear and there may be disagreements about what the problem is other than achieving some insights (Coyle, 1999), there is no definitive test of acceptability of any interim or final, modelling product.

System Dynamics models are transitioned objects to enable our thinking (Coyle 1999) and having served their purpose in aiding our learning and understanding; they are frequently discarded or become redundant. In Systems Engineering the models built are often inextricably linked to testing or processes which enable the design of strategies for testing the system to be built.

In Systems Engineering, there is greater emphasis on the validity of any model that is built. Indeed models are frequently converted into code that drives the performance of high fidelity simulations. The fidelity can and is exhaustively tested. Key decisions to proceed or not to proceed to the next stage of the design project rests on passing certain prescribed tests. This is part of the rigour offered by Systems Engineering however, the same rarely applies to System Dynamics modelling interventions, and this is a very important matter for models that are to be used subsequently for gaining policy insights or to be used for training executives in complex dynamic decision-making.

5-23 From the Systems Engineering perspective, System Dynamics offers something that Systems Engineering practitioners often find difficult and that is to conceptualise problems that have certain “systems” characteristics, but where the component parts cannot be isolated and replicated. In System Dynamics modelling, we recognise the opportunities for conceptualising even when we may never be able to fully or exhaustively test (Coyle, 2000).

In System Dynamics, models are frequently built as dynamic hypothesis or more correctly representations of dynamic hypotheses (Oliva, 1996; Sterman, 2000). We may need to formulate many dynamic hypotheses in a search for plausible explanations of dynamic behaviours we might have observed in a complex real world. The extent to which we might test these hypotheses is where Systems Engineering offers an opportunity to improve the rigour of System Dynamics modelling. For example, the learning cycles approach (Saeed, 1998; Saeed, 2001) for development of reference modes attempts to provide additional rigour to the System Dynamics modelling process. In highly complex problem situations there are no exhaustive or definitive tests to the reference modes that are either necessary or sufficient. Whereas in Systems Engineering we often design exhaustive tests or at least do a much more complete job of it.

By adopting Systems Engineering philosophy and methodology we would be seriously thinking about improving the rigour of System Dynamics simulation models. Systems Engineering offers a structured way for development of simulation models. Development of the simulation model of feedback structure is one of the core components of the System Dynamics method (Forrester, 1968; Forrester, 1994; Sterman, 1994; Sterman, 2000; Richardson, 2001; Sterman, 2001) advanced by the mainstream authors in System Dynamics. As discussed above, the development of System Dynamics models using a Systems Engineering approach can bring new rigour to the System Dynamics simulation models and it can also help in developing robust and ‘responsive to purpose’ models. The value that Systems Engineering processes can add to the model verification and validation is further explained in Chapter 10 where it is illustrated by applying Systems Engineering process to the System Dynamics modelling of dryland salinity.

5-24 5.5 SUMMARY AND CONCLUSIONS

In this chapter, a multi-methodology framework was introduced. The multi- methodology framework used three systems methodologies namely, cognitive mapping, System Dynamics and Systems Engineering for conceptualisation, development and validation of System Dynamics models for gaining strategy level insights in addressing complex dynamic problems.

The use of cognitive mapping along with other diagramming conventions like causal loop diagram can further strengthen the problem conceptualisation, through providing a framework for exploration of ideas and for preparing individual’s cognitive maps. The use of multiple diagramming/mapping approaches at the conceptualisation stage can also provide additional rigour to the qualitative analysis through creating opportunities for triangulation of the insights.

While having a strong conceptual basis for problems that have certain “systems” characteristics, but where the component parts cannot be isolated and replicated, System Dynamics simulation model development processes are not as rigorous as are the ones used in Systems Engineering. This rigour generally leads to exhaustive model testing and sound model verification and validation practices. This differing rigour in processes is linked to differing modelling philosophies, the purposes of the modelling activity, intended use of the models and the practices underlying both methodologies. Synergistic use of both methodologies can strengthen problem conceptualisation as well as model verification and validation.

The multi-methodology framework developed in this chapter was applied to System Dynamics modelling of dryland salinity for strategic management. The next chapter describes the study area and various components of this approach as applied to the dryland salinity problem are described in Chapters 7, 8 and 9.

5-25 CHAPTER 6 REFERENCE MODE DEVELOPMENT

System Dynamics modellers seek to characterise the problem dynamically, that is, as a pattern of behaviour, unfolding over time, which shows how the problem arose and how it might evolve in future. Reference modes help you and your clients break out of the short-term event oriented worldview so many people have. (John Sterman, 2000:90)

A reference mode is essentially a qualitative and intuitive concept because it represents a pattern rather than a precise description of a series of events. A reference mode also subsumes history, extended experience, and a future inferred from projecting the inter-related past trends. (Khalid Saeed, 2003:412)

In System Dynamics, the term ‘reference mode’ is used to denote a pattern of graphs that present the idealised or actual behaviour of different variables over time. The other terms used include behaviour over time (BOT) graph, reference behaviour or reference conditions. These terms fundamentally refer to the same thing, that is, behaviour over time and specifying patterns that characterise those changes.

The concept of reference modes is not new, being used in a variety of disciplines and for a variety of purposes. The disciplines in which it has been used include basic sciences, social sciences and management sciences. In chemistry laboratories equipment (atomic absorption spectrophotometers, colorimeters etc.) is calibrated against standard solution concentrations. These standard solutions, with known concentrations, work as reference solutions for calibrating the equipment. In physics laboratories, instruments are calibrated against the standards provided by the Standards Institute. The measurements held at the Standards Institute act as the reference for calibration of equipment (weights and weighing balances, measuring devices etc.). Within management sciences, widely used approaches for problem analysis (for example, Kepner and Tregoe (1981)) define a problem as a deviation from the “SHOULD BE” conditions. Here the “SHOULD BE” conditions provide the reference conditions for comparison of the present conditions and for

6-1 characterisation of the problems. In control theory, it is assumed that there exists a mathematical model describing the dynamic behaviour of the underlying process, which can be changed from existing to the desired behaviour (Ozbay, 1999). In this case, the existing behaviour can act as reference behaviour to identify the desired behaviour.

Although within different disciplines, the concepts of reference mode have been used, its use was confined to the past or the known behaviour of a variable that depends upon observations made in the past. In System Dynamics modelling, the models are calibrated against the recent observed behaviour of the system. The behaviour is represented by a pattern arising from the combination and interaction of variables in sets of feedback structures.

The System Dynamics Modelling literature has made a marked contribution to the development of the reference modes concept. Whilst the processes of building reference modes start from time series data, the reference mode is more than a graph of the time series. Saeed (2002) termed it as an abstract concept that represents a fabric of trends and shows how different variables change with respect to each other over time. In particular, a time series graph simply represents correlation. The developer of the reference modes is presenting dynamic hypothesis (Helsinki School of Economics, 1981; Oliva, 1996; Maliapen, 2000; Sterman, 2000; Raimondi, 2001) of the causality based on empirical data and local knowledge.

In this chapter, a method suggested by Saeed (2002) was applied to understand the dryland salinity problem in the Murray Darling Basin. Sources and availability of data for key salinity parameters are evaluated and insights gained from the application of Saeed’s method are discussed. Shortcomings of the method encountered have been addressed and ways to improve this method are suggested. Reference modes developed in this chapter are used in the later chapters for developing a stock and flow model.

6-2 6.1 WHY REFERENCE MODES ARE NEEDED?

Effective policy interventions deliver sustained and desirable changes to system of purposeful human activity (Saeed, 1998). In problem solving approaches such as that proffered by Kepner and Tregoe (1981), problems are considered to be deviations from the pre-existing conditions and to improve our understanding of the reasons for deviation, problems need to be characterised (Saeed, 2002). This characterisation includes articulation of the problem according to the available modes of understanding. Problem articulation refers to initial characterisation of the problem in terms of time horizon, stakeholders perceptions of the problem, observable symptoms, the perceived causes of the problem, and factors affecting it (Saeed, 2001; 2002; 2003). Problem characterisation is usually done through a combination of discussions with the client team, archival research, data collection, interviews and direct observation or participation (Saeed, 1998; Sterman, 2000). Two valuable processes in problem articulation are establishing reference modes and explicitly setting the time horizon. Saeed (1998) suggests that half of the understanding about the problem is achieved through the learning processes involved in the development of the reference modes. Facilitating such understanding is fundamental to the System Dynamics discipline.

The reference modes are used for two purposes: first learning about the problem and its definition; and second, for building confidence in the model through testing the hypothesised causality. Development of reference modes is a learning process that leads the effort of the modeller through to identification of model variables.

6.2 METHOD FOR REFERENCE MODES DEVELOPMENT

A methodology consisting of learning cycles approach (Saeed, 2003) was adopted. Over a decade, Saeed (2001; 2002; 2003) developed a technique for ‘problem slicing’ and development of reference modes in a 20 step method (Figures 6.2 and 6.3). His method is based on Kolb’s (1984) model of experiential learning that includes a learning cycle: feeling, watching, thinking and doing. Figure 6.1 shows the broad framework of Saeed’s (2002) method.

6-3 Domain Boundary Preliminary System Boundary Preliminary Model Boundary Model Boundary

Reference Mode

Figure 6.1 Broad framework of Saeed’s method

This method describes a process involving five learning cycles. Each learning cycle consists of four steps and the method starts with the examination of available time series data. During the process, each learning cycle yields an intermediate product. The intermediate products of this learning process are domain boundary, preliminary system boundary, preliminary model boundary and model boundary. At the end of learning cycle five, a reference mode is produced. This reference mode consists of a graph showing a pattern of the past behaviour as well as likely future behaviour of the variables.

Each step of this method is described in Figures 6.2 and 6.3 along with the learning cycle to which it pertains.

6-4

Learning Cycle 1

1 examine problem description sanpshots of current situation

4 2 OUTPUT identification of key variables Select a subgroup from the past trends Domain Boundary that best represents problem history

3 Collect and plot time series data

Learning Cycle 2

5 examine multiple set of complex historical time series

8 OUTPUT 6 Select a subgroup of patterns Preliminary Decompose each set of representing the behavior of interest system boundary

complex patterns into simpler parts and discard remaining subgroups (PSB)

7 graph various componentsf o decomposed patterns

Learning Cycle 3

9 examine selected group of patterns in PSB

OUTPUT 12 Preliminary 10 Assemble graphed historical pattern aggregate at the desired level model boundary into fabric of model variables (PMB)

11 graph inferred behaviour of aggregated and abstract variables

Figure 6.2 Method for development of reference modes (Saeed, 2001; 2002), Cycles 1-3

6-5 Learning Cycle 4

13 Examine Selected group patterns in PMB

16 OUTPUT 14 Assemble historical patterns infer behavior of stocks model boundary in the decomposed variables (MB) missing in the data into fabric of model variables

15 graph behavior of the additional variable conceived

Learning Cycle 5

17 Examine past behavior of variables in the extended model boundary

20 18 review past and inferred OUTPUT make intelligent projections future trendsof the model and Reference Mode of the future behaviour of model policy variables as a fabric and variables in the extended boundary ensure logical consistency

19 Graph inferred future trends for the variables in the extended model boundary

Figure 6.3 Method for development of reference modes (Saeed 2002, 2001), Cycles 4 and 5

6.3 DATA FOR DEVELOPING REFERENCE MODES

Data for identifying reference modes for the proposed dryland salinity model was acquired from a variety of sources. Datasets came in different formats having been prepared for disparate customers for a variety of reasons, using different methods and at different times. The climate data was acquired from the Bureau of Meteorology, Australia. Temperature and rainfall data exists in the form of a time series dataset, whilst some of the data acquired was in the form of printed graphs. Data was generated from these graphs by using the software GrapherTM and the water

6-6 resources data was acquired from different published material. Most of these were in hardcopy and an ExcelTM dataset was generated through GrapherTM. Agricultural land use and socio-economic data was acquired through Agstat, a database in MicrosoftTM with access held by the Australian Bureau of Statistics. The gaps in data were addressed through direct contact with local agencies and land clearing data was acquired through the Australian Greenhouse Office. The quality of the data lies with the source of the data and integrity of the methods of collection. The original datasets have been aggregated at a variety of levels according to their original purpose and they also differ in their precision. Further, alternate sources of data have not been discussed.

To complement data gaps and seek farmers’ perspectives, a field survey of Murray Darling Basin farmers was conducted on the causes, impacts and potential remedial options. Results are presented in the supplement attached.

6.4 DOMAIN BOUNDARY

This learning cycle consists of four steps and results in delineation of the domain boundary of the salinity problem. Each step is described below:

6.4.1 Snapshots of current situation

In the following paragraphs, different snapshots of salinity in the Murray Darling Basin are described as cited in the literature. The term ‘snapshot’ has been used consistent with its dictionary meaning:

 A short description of a small amount of information that gives you an idea of what something is like. (online Oxford English dictionary).

 An isolated observation. (Houghton, 2000).

 An impression or view of something brief or transitory. (Merriam-Webster, 2003).

6-7 The snapshots described below include the past and current status of salinity in the Basin and the literature perceived causes of salinity. These snapshots have been described as an isolated observation and may or may not have relationships with the previous or the next snapshot.

6.4.2 Concerns about salinity data

There are several concerns about the data including a) the trends the data shows are based on the risk associated with the rise of the water table (NLWRA, 2001) rather than actual salinity, b) the new advancement in data collection methods has identified the salt affected lands that actually existed before but were not identified due to the limitations of the methods used for data collection at that time. Other than implementation, the paucity of data can also restrict identification of the appropriate remediation measures. The initiative of salinity and land use mapping is quite recent and there are expected considerable delays in generation of consolidated data-sets. The following excerpts from literature depict the status and the quality of existing salinity data.

 The paucity of data available to accurately characterise the state of degradation of Australia's irrigated and dryland areas is a major restriction on the implementation of appropriate and priority remediation programs (Evans et al., 1996).

 Even where the data is available (e.g., Victoria, southwest Australia), the forecasted groundwater levels to 2020 and 2050 are based on straight-line projection of recent trends in groundwater level. Due to inadequacies in current methods, accurate groundwater surfaces cannot be developed with the existing distributed data (NLWRA, 2000).

6-8 6.4.3 Climate

Climate affects salinity in multiple ways. First, fluctuations in rainfall can affect the recharge and discharge of groundwater. An increase in evapo-transpiration from soil surface can enhance the transport of salts towards the surface and ultimate formation of a salt crust. Climate change can impact water resources in the MDB in two ways:

 Significant reductions in stream flow in the MDB,

o 0 to 20% reduction by 2030 (Jones et al., cited in Howden, Hartle et al. 2003); and

o 10 to 35% reduction by 2050 (Arnell, 1999 cited in Howden, Hartle et al,. 2003).

 Increase in water demand due to increased evaporation rates, lower rainfall. Frequency of general and high security water allocations for environmental flows not being met will increase (Howden, Hartle et al., 2003)

Based on their analysis, Howden et al., (2003) stressed finding ways to integrate the effects of climate change into policies and practices.

6.4.4 Land use

The landscape in the Murray Darling Basin has been changing and will continue to change, in part due to major human induced impacts such as settlement and land clearing for agricultural, urban and industrial uses. Agriculture is one of the major sectors for land use change (Crabb, 1997):

 In Australia, 20% of the total land area is under forest (major ecosystems are shrub land, savannah and grasslands). The cropland and crop/natural vegetation mosaic covers 6% of the area. A major expansion in agricultural development during the 1950s to 1980 had been due to extensive clearing and an increase in the cultivated area. (World Resources Institute, 2003)

6-9  Land clearing started in Australia many years ago and it is still continuing. The term land clearing refers to removal of the natural cover (e.g. forest) from the land for alternative uses. Within various studies conducted by the Australian Government, different terms have been used for land clearing. In “Land Clearing a Social History” (AGO, 2000), the term land clearing is used with the above meaning but in “Carbon Accounting System” (AGO, 2000) the terms land conversion and re-clearing have been used. Land conversion refers to the first time clearing of the forest while re-clearing refers to the clearing of re-growth and conversion to an alternate land use. The current motivators for land clearing include land availability, clearing controls, environmental and social influences, financial and institutional incentives, agricultural research and development, and market forces (AGO 2000).  Graetz et al., (1995) assessed that 1,029,640 sq km have been thinned and cleared within intensive land use zones and most of this is in the Murray Darling Basin. One of the causes of land clearing was conditional purchases, for example from the 1860s to 1960s leases and conditional purchases were issued on the proviso that a certain percentage of tree cover was to be removed each year (BRS, 2000).  The native vegetation regimes evolved to make the best use of available rainfall while avoiding the salts. All vegetation pumps water from the soils and transpire a component to the atmosphere (through the evapo-transpiration process). Any change in vegetation density or type (e.g. a change in vegetation’s water pumping capabilities) will alter the volume of water reaching the saturated zone below. Clearing of native vegetation disturbed the current balance (Evans et al., 1996).

6.4.5 River water diversions and salt carrying capacity of rivers

The diversions from the Murray Darling River have increased. The amount of water presently taken from rivers is not ecologically sustainable and a new balance between the environmental requirements and the consumptive use will have to be struck (Toyne, 1995 cited in Crabb, 1997:53). These river extractions have increased multi-fold over the last five decades. For example, in 1960 diversions from the

6-10 Barwon Darling and the New South Wales and the Queensland tributaries were 50,000 ML while in 1990-91 they were 1.4 million ML. Moreover, the increase in diversion has been primarily due of the cotton industry and the use by growers of large on-farm water storage (Crabb, 1997). The following excerpts from literature depict the picture:

The continuing saga of the extraction of massive amounts of irrigation water from inland rivers to satisfy the escalating demands of the irrigation industry is Australia’s most serious and ultimately most disastrous water related issue. (White, 2000)

The impacts of land clearing and management of the Murray Darling Basin Waters by construction of dams and canals have lessened the variations in flow and salinity. However, the exploitation of the waters has reduced the capacity of the rivers to carry salt to the sea without prejudice to water users in the downstream reaches and has delivered a far higher salt load to the river systems. This has occurred through saline water drainage directly to the rivers and through increased groundwater flows. (Evans et al., 1996)

6.4.6 Delays

Delays in the groundwater system’s response to disturbances exacerbate our understanding of the dynamics mechanisms and exactly how they contribute to dryland salinity as it is evident from the following excerpt:

The groundwater system responded very slowly to these massive disturbances, so the full consequences of the human impacts have begun to be felt only after decades, or even a century. In fact, we know that the incipient degradation processes will often continue for centuries to come. We have now released a time-bomb with a slow fuse. (Evans et al., 1996)

6-11 6.5 KEY VARIABLES

In this learning cycle, the following variables have been identified from the problem description described above:

 Agricultural land (hectares).  Land clearing (hectares).  Climate.  Rainfall (mm/year).  Evapo-transpiration (mm/year).  Temperature.  Stream-flow (mean annual stream flow).  Stream salinity.  Salt affected rivers.  Value of agricultural production.

6.5.1 Time series data

In this step, the data was acquired from different secondary sources that include Bureau of Meteorology, the Murray Darling Basin Commission, National Land and Water Resources Audit, Greenhouse office, Australian Bureau of Statistics and published literature. The data about important variables is presented below in the form of graphs in Figures 6.4 to 6.10.

6-12

35000

30000 Total

25000

20000

15000 MDBC

10000 NSW Storage Capacity (Giga Litres) VIC 5000 QLD

1920 1930 1940 1950 1960 1970 1980 1990

Figure 6.4 Storage capacity in the Murray Darling Basin (Crabb, 1997).

6-13 2500

2000

1500

1000

Averge River Salinity (EC) 500

0 1980 2000 2020 2040 2060 2080 2100 2120 Years

Murrumbidgee River Darling River (Menindee) Bogan Narromine Namoi Gwyder

Figure 6.5 River salinity in NSW (MDBC, 1999)

Natural Flow at Murray

Total Diversions (GL/year) Mouth Barrages (GL/year) Mean 13,754 Median 11,883

Figure 6.6 Total diversions in the Murray Darling Basin (excluding Queensland) Source: (Crabb, 1997).

6-14 600 Grain Cattle Sheep 500

400

300

200

Total dry sheep equivalents (millions)

100 Extensive Land Clearing

0 1860 1880 1900 1920 1940 1960 1980 2000 Year

Figure 6.7 Historical trends in the agricultural industry development converted to dry sheep equivalents (DSE) (adopted from NLWRA, 2001).

6-15 Figure 6.8 Australian farm incomes (AGO, 2000)

6-16 Figure 6.9 Examples of change in land use intensity index (NLWRA, 2001)

6-17 30

25

20

NSW 15 South Australia Victoria

10

5

0 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

Figure 6.10 Time-graph of land clearing (x 1000' hectares) from 1988 to 1998 (AGO, 2002)

6-18 6.5.2 Farmer’s perception about the past and future of dryland salinity

A field survey was carried out to seek farmers’ perspectives on the extent of dryland salinity using a structured questionnaire. Details about the field survey and its results are presented in the Supplement. The survey explored farmers perceptions about dryland salinity at individual properties as well at the Murray Darling Basin level, past and future trends in dryland salinity, remedial measures, and factors affecting and impacts of dryland salinity.

The graphs representing participating farmers’ responses on dryland salinity trends are reproduced below. For detail, please refer to the Supplement.

6-19 60

50

40

30

Respondents (%) 20

10

0 Last 10 yrs 1980 to 2000 1950 to 1980 None

Figure 6.11 Increase in dryland salinity (Question 5)

70

60

50

40

30 Respondents (%) 20

10

0 The increase in area The increase in area The increase in area The increase in area affected by dryland affected by dryland affected by dryland affected by dryland salinity between 2000 and salinity between 2000 and salinity between 1980 and salinity between 1980 and 2010 was more than the 2010 was less than the 2000 was more than the 2000 was less than the increase between 1980 increase between 1980 increase between 1950 increase between 1950 and 2000 and 2000 and 1980 and 1980

Figure 6.12 Increase in dryland salinity over time – trends (Question 6).

6-20 60

50

40

30

Respondents (%) 20

10

0 Last 10 yrs 1980 to 2000 1950 to 1980 No

Figure 6.13 Decrease in dryland salinity on individual properties (Question 7)

60

50

40

30

Respondents (%) Respondents 20

10

0 The decrease in area The decrease in area The decrease in area The decrease in area affected by dryland affected by dryland affected by dryland affected by dryland salinity between 2000 and salinity between 2000 and salinity between 1980 and salinity between 1980 and 2010 was more than 2010 was less than the 2000 was more than 2000 was less than the thedecrease between decrease between 1980 thedecrease between decrease between 1950 1980 and 2000 and 2000 1950 and 1980 and 1980 Figure 6.14 Decrease in dryland salinity over time – trends (Question 8)

6-21

40

35

30

25

20

15 Respondents (%)

10

5

0 Last 10 yrs 1980 to 2000 1950 to 1980 No

Figure 6.15 Increase in land clearing in the Murray Darling Basin (Question 17)

60

50

40

30

Respondents (%) 20

10

0 Last 10 yrs 1980 to 2000 1950 to 1980 No

Figure 6.16 Decrease in land clearing in the Murray Darling Basin (Question 18)

6-22 35

30

25

20

15 Respondents (%)Respondents

10

5

0 Obvious salt Reduction in Impairment of Impairment of Other None crust/patches crop yields infrastructure water-supplies

Figure 6.17 Impacts of dryland salinity on individual properties (Question 9)

45

40

35

30

25

20

Respondents (%) 15

10

5

0 Area affected by dryland Severity of impacts of Area affected by dryland Don’t know salinity increases as dryland salinity increases salinity or severity of its depth to groundwater as depth to groundwater impacts is not affected by decreases decreases the depth to groundwater

Figure 6.18 Relationship between dryland salinity and depth to groundwater

6-23

6.5.3 Delineation of the domain boundary

All the variables identified in step two and plotted in step three are considered important and have been retained within the problem domain. At this stage the following variables are included in the problem domain:

 Agricultural land (hectares).  Land clearing (hectares).  Climate: o Rainfall (mm/year). o Evapo-transpiration (mm/year). o Temperature.  Stream-flow (mean annual stream flow).  Stream salinity.  Salt affected rivers.  Value of agricultural production.

6.6 PRELIMINARY SYSTEM BOUNDARY

In this learning cycle, changing patterns of the variables identified in the preceding learning cycle are analysed. Complexities of the patterns have been further simplified to aid analysis. This learning cycle consists of four steps. First, the variables identified in learning cycle one have been re-examined from the perspective of their relationship with salinity. The nature of most data suggests that further decomposition of the patterns into simpler patterns through Fourier series analysis is difficult. However the general trends amongst the data have been simplified visually and plotted against time for comparison with each other. The output of this learning cycle is the preliminary system boundary.

6-24 6.6.1 STEP 5 - Examination of Variables

The variables identified above have direct logical links with salinity. In the following paragraphs, these relationships are discussed.

Climate is the main driver that determines the water fluxes between atmosphere and the land. The parameters considered here are i) temperature, ii) rainfall/precipitation, and iii) evapo-transpiration. Temperature affects the weather and seasonal patterns affecting both evapo-transpiration and precipitation. Water provides the medium for the movement of salt through the biosphere and under the forces of evaporation, salts in the groundwater move upwards through soil capillaries towards the soil surface and it is at the soil/ground surface that water evaporates and leaves the salt crust. Climate is a very important model variable and must be included in the model.

Land clearing replaces the vegetation with clear land and affects the net evapo- transpiration and ultimate water balance in the region. Land clearing has been a very significant event in the Murray Darling Basin and must be included in the model. Generally land clearing has been done for the purpose of agricultural development.

Agricultural land hereby refers to land under crops or kept for agricultural purposes like grazing/animal husbandry providing a base for the growth of plants/crops. The root system, nutrient use pattern and root-shoot ratio determines the water-uptake by crops, moreover, crops vary in their tolerance to salts. This is a variable that cannot be left out from a viable model of salinity.

Stream salinity is an indicator of catchment health, that is, how much salts are leaking or being exported from the land to the water. Moreover, water taken from streams for irrigation purposes depends upon the salinity of streams. It can be included in the model as an indicator of catchment health.

Value of agricultural production/farm incomes indicates the economic viability of agricultural land. Salinity affects the land productivity and hence crop yields and it provides a good indicator of land problems.

6-25 6.6.2 STEPS 6 and 7

Steps six and seven have been combined, as the processes of simplification and the processes of graphing are not different in this case. The graphed patterns are shown in Figure 6.11.

Rainfall Cleared Land

Irrigated Land

Agricultural land

River Diversions

Farm Income

Now Time

Figure 6.19 Pattern of behaviour of different variables over time

6.6.3 STEP-8

In this learning cycle, data about different variables was simplified and graphed. Time series data about the evapo-transpiration over the Basin was not available and will be addressed in the following learning cycles. The temperature is indirectly reflected in the evapo-transpiration; therefore, it has not been included in the system

6-26 boundary. However upon analysis, the following variables have been found to constitute the system boundary:

 Rainfall,  Evapo-transpiration,  Agricultural land,  Cleared land,  Irrigated land,  River diversions, and  Farm income.

6.7 PRELIMINARY MODEL BOUNDARY

6.7.1 STEP-9 Examine selected group of patterns

Data presented in learning cycle 1 and 2 has been collected from secondary sources and represents state of those parameters at different geographical levels. In Figure 6.5, average salinity forecasts are also on the individual river Basin scale but these rivers represent a major part of the Murray Darling Basin. Time graphs are shown in Figure 6.10, with land clearing at the state level. Storage capacities represented in Figure 6.4 are at the state level. River diversions shown in Figure 6.6 are summed up at the Basin level while Queensland has been excluded.

The problem of data collection at the Basin level is in reported studies for example the Australian Bureau of Statistics do not take Murray Darling Basin as one category, therefore, there are problems in the aggregation of data.

6.7.2 STEP-10 Aggregate at the desired level

The concept of aggregation has been used in two perspectives. First it is used for combining the data from different geographical regions. Second, it is used for combining different variables to create new variables at a different level of system. In the following paragraphs, first the problems in geographical aggregation are discussed and thereafter follow the problems in variables aggregation.

6-27 At the Basin level, parts of all the four states (not the whole state) are involved. To work at the Basin level, data from all states will be required but the state figures cannot be used as none the whole state is part of the Basin.

Climate information represents two geographical focuses, i) at the national scale and ii) at the regional scale. The climate change phenomenon is occurring over the whole continent and it is no different for the Murray Darling Basin, therefore the trends shown over eastern Australia have been adopted. Agricultural production figures presented in learning cycle one represent national figures. Over the last 100 years the major development of agriculture had been in the Murray Darling Basin, major water reservoirs have been built and more land has been brought under cultivation as well as the area under irrigation has increased. Therefore the agricultural development curve may be relatively steeper in the Murray Darling Basin than the overall Australia figures. Land use intensity shown in Figure 6.9 is at the catchment scale. Multiple land use intensity curves have been shown in different areas of the MDB, however, these don’t represent the whole MDB figures but they do indicate an increasing trend in land use intensity. The only results available are from the studies for specific projects; therefore, these are the figures that will have to be taken into account while looking at land use intensity. Multiple time series shown in Figure 6.10 for land clearing provide data at the state level. Figure 6.4 shows aggregated data of water storages.

The three variables, irrigated land, agricultural land, and cleared land are types of land uses. Land remains the same but uses change, therefore the three variables are aggregated to constitute a variable, total land. All transformation of land uses remain within the total land.

6-28 6.7.3 STEP-11 Graph inferred behaviour of the aggregated and abstract variables

Total land Cleared Land

land under forest/bush

Agricultural land

Now Time

Figure 6.20 Aggregated total land

6.7.4 STEP-12 Assemble graphed historical pattern into fabric of model variables

So far the following variables have been identified to be in model:

 Total land: o Land under forest/bush. o Cleared land. o Agricultural land.  Rainfall.  Evapo-transpiration.  River diversions.

6-29 6.8 MODEL BOUNDARY

In this learning cycle, the variables selected in learning cycle five have been examined and on the basis of this examination, variables missing in the examined data have been identified. The graphs presented in the following sections represent our hypothesis about the over time behaviour of the missing variables. Our hypothesis is based on the ancillary data. The output of this learning cycle is the boundary of salinity model.

6.8.1 STEP-13 Examine selected group patterns

The patterns expressed in the above learning cycles indicate that land clearing is increasing while the total land is constant. River diversions have increased and farm incomes have a general decreasing trend throughout Australia. However these variables present a partial picture. The problem in focus is salinity and the data does not lead to develop a mode of behaviour for the salinity problem. In the following sections, such variables that are missing in data are identified and described.

6.8.2 STEP-14 Infer stocks missing in the data

The following stocks are missing in the data examined above. The term salinity refers to the amount of salts in water, in other words it is a ratio of salts to water in a solution. Therefore the variables describing the physical stocks need to be included in the list of main variables:

Stock of salts: salt stock is the main important variable as the slats are naturally a part of the earth crust. The movement of water moves salts through different phases of the soil, water and plants, e.g. stock of salts in soil, stock of salts in groundwater, stock of salts in rainwater.

Stock of water: in the water cycle, water moves through the atmosphere passing through land into groundwater, rivers or sea. The snapshots described in Chapter 2 indicate that over time the water table has come up and has brought salts with it, which indicates an increasing trend in the stock of the groundwater.

6-30 Depth to water table, the generic model that explains salinity states that a rise in the water table provides the medium for salts solubility and brings salts to the soil surface. Any model of salinity should examine the depth of the water table or the elevation of the water table. When the water table becomes within two meters from the soil surface, the yields of the crops start to decrease depending upon the salt tolerance of the crop.

Other than that, the rates of profit, rates of gross and net rent will need to be included in the specific modules concerning profitability of the farming enterprise. There exists a correlation between the worsening terms of trade and declining value of production.

6-31 6.8.3 STEP-15 Graph behaviour of the additional stocks missing in data

Total Salts

Salts on soil surface

Salts Dissolved

Salts undissolved

Underground water

Now Time

Figure 6.21 Behaviour of missing stocks

6.8.4 STEP-16 Assemble historical patterns in the decomposed variables into fabric of model variables

At this level, the following variables have been suggested to be included in the model boundary:

 Climate: o Rainfall. o Evapo-transpiration.  Total land:

6-32 o Land under forest/bush. o Cleared land. o Agricultural land.  Salts: o Land surface salts. o Undissolved salts. o River salts.

6.9 REFERENCE MODE

In this learning cycle, the past behaviour of the output of learning cycle four has been examined. Projections have been made into the future based on logical relationships and graphs of this behaviour are presented. The output of this analysis becomes the reference mode. It contains the variables identified through data, variables missing in data (abstract) and the future trend of the variables.

6.9.1 STEP 17 Examine past behaviour of the variables

The past behaviour of the variables has already been discussed. Here only the past behaviour of the missing variables is discussed. The salt stock in land has not changed, the only its location has changed, and its location has been influenced by the stock of water. An increase in recharge has increased the groundwater that has resulted in elevation of the water table.

6.9.2 Projections into the future

For few variables like the future of dryland salinity in the Murray Darling Basin and land clearing, the farmers’ responses provide guidelines for trends used in developing this reference mode. For other variables, the current trends have been extended into the future. A composite reference mode is presented in Figure 6.23.

In the field survey a couple of questions were asked about the future of dryland salinity in the Murray Darling Basin with the majority of respondents suggesting that there will be a decrease in dryland salinity between 2010 and 2020 and it will further

6-33 decrease between 2020 and 2050. None of the respondents indicated any increase in dryland salinity between 2020 and 2050. These responses are presented in Figure 6.22. The detail about the questions and responses is presented in the Supplement.

6-34 45

40

35

30

25

20 Respondents (%) Respondents

15

10

5

0 Dryland salinity problem Dryland salinity problem Dryland salinity problem Dryland salinity problem The increase in dryland The decrease in dryland in the Murray Darling in the Murray Darling in the Murray Darling in the Murray Darling salinity between 2010 salinity between 2010 Basin will increase Basin will decrease Basin will increase Basin will decrease and 2020 will be more and 2020 will be less between 2010 and 2020 between 2010 and 2020 between 2020 and 2050 between 2020 and 2050 than the increase than the decrease between 2020 and 2050 between 2020 and 2050

Figure 6.22 Farmers’ perspective on future trends of dryland salinity

6-35

Total Salts Rainfall

Total Land Cleared Land

Land under Forest/Bush Irrigated Land

Dry Cultivated Land

River Diversions

Salts on soil surface

Farm Income

Rural debt

Now Time Figure 6.23 A reference mode for salinity problem

6-36 6.9.3 STEP 20 Ensure logical consistency

The behaviour shown in Figure 6.14 draws its logic from the salinity processes described in the literature (Chapter 2). These processes provide a basis for examining logical consistency and may include:

Due to a rise in the water table, the salts in the subsoil become dissolved in the water and move towards soil surface where the water evaporates and leaves the salt crust. Over time, due to an increase in groundwater recharge due to land clearing, the water table has increased its elevation and it is still rising at 0.5 meters/year in Wagga Wagga. With the overland flow of the rainwater, these salts wash out to the rivers. Diversions from river for human consumption decrease the quantity of available water downstream for dilution of these slats. The graph in Figure 6.6 shows an increase in river diversion.

River diversions are likely to maintain their current status as shown in Figure 6.14, due to two main reasons i) the awareness about the salinity problem has increased which in turn has increased public pressure for environmental flows, ii) the major irrigation schemes have already been built and area has been brought under irrigation. There are no planned further irrigation schemes.

The future projections shown in Figure 6.14 indicate an increase and then smoothening in land clearing. The increase is based on the current status of land clearing which is still continuing in Queensland and on the margins of the wheat-belt in New South Wales.

The total salts content, available land and the climatic variables have been considered constant. Although the climate change is happening, it is a very long-term process of change. According to the Australian Bureau of Meteorology, there is a slight upward trend in the rainfall but some authors have accrued it to the difference in measurement and analysis techniques. For the purpose of this model, the rainfall trend has been considered as the current.

6-37 6.10 ISSUES IN OVERALL PROCESS OF DEVELOPING REFERENCE MODES

In preceding sections of this chapter, a learning cycles-based reference modes development process (Saeed, 2001; 2003) was applied step-by-step. The application of this method for characterisation of the salinity problem provides several insights about the potential and limitations in development of reference modes for the dryland salinity problem. The method provides a valuable means for focusing analysts’ attention to the main issues. The question of starting “from where” in the fuzzy complex system is a prime one and Saeed’s method provides a starting point. However, difficulties were encountered in the use of this method for the dryland salinity problem in Australia, the main issues around these difficulties are:

6.10.1 Sources of data

Saeed’s method relies on published statistical data as the primary source of information to start with and proceeds to final development of the reference modes. Uses of other sources of data are not envisaged in early phases. In situations where, the published statistical data is not available, sources other than such numerical data will typically need to be considered. These sources may include the knowledge within the mental models of the systems expert/players of day-to-day systems functioning or written descriptions (Forrester 1994; Ford and Sterman 1998; Sterman 2000).

The development of the reference modes was strongly constrained by the lack of consistent numerical data. This was supported by descriptive information from various reports published by the Murray Darling Basin (included in the reference list). This was accomplished by employing the approach that Jay Wright Forrester (1918- ) has consistently emphasised that key dynamical properties find their origin in system structure and the policies that guide decisions (e.g., Forrester, 1995). In general, such information is not amenable to statistical collection and must come from the alternative sources of information, for example the mental models of the key role players.

6-38 6.10.2 Stakeholders involvement

The involvement of stakeholders in the problem articulation stage is important not only because the stakeholders are:

 a part of the problem space as they influence the current system’s behaviour through their decisions;  an integral part of learning cycles; and  knowledgeable people essential to activities directed at identifying missing variables, missing data or explaining variations in quality of data..

The process of stakeholder involvement in the problem articulation stage also is important in developing the foundation for the subsequent validation process. Ford and Sterman (1997) stressed the inclusion of process knowledge of experts into System Dynamics models as without their involvement the procedure of problem articulation will remain incomplete. Without their involvement in characterising the modes of the problem behaviour the identification of reference modes from patchy deficient statistical data becomes a painful process of looking into a magic bowl. The shortcomings identified in this instance might be reduced through extensive and close involvement of stakeholders’ right from the earliest stages when attempting to identify the preliminary model boundary.

6.10.3 Aggregation of data

The aggregation of data faced some limitations that arise first because of the nature of the salinity problem, and second because of the nature of available statistical data. First the available statistical data is not enough to characterise the modes of salinity behaviour based just on the time series data. The data that is available is not classified according to the catchment it is collected and classified according to administrative boundaries (statistical units). Second, the concept of catchment has different connotations. The surface water catchment does not necessarily coincide with the groundwater catchments. The social catchment, that is the areas influencing, the adoption of resource use practices does not coincide with these either. In those situations, the aggregation of secondary data may not represent the reference

6-39 behaviour of the salinity problem. Available statistical data is from localised studies in different periods in time a consolidated time series data is not available. Even the Australian Bureau of Statistics’ Agstats provides data only from 1984 to 1997 with some parameters or data missing. The problem of dryland salinity scale has spread over 200 years or so and over time, the methods of investigation change as do the beliefs about data, its collection methods, storage and retrieval. In such situations alternatives to the aggregate data may also be highlighted.

6.10.4 All variables in reference mode or a few?

Within the method used, it appears that all the variables in the model must be represented in the reference mode as the process of model development and development of reference modes is not different. Literature presents a different picture. Ford (1999:185) used only one variable (deer population) as a reference mode in his Kebab deer herd. He used only a descriptive reference mode (Ford, 1999:226) while his models contained more than one variable. In the company profit problem, Albin (1997) used only two variables (profit and inventory) to designate reference modes of the problem while his model contained more than two variables. Albin (1997) also used four variables to designate reference mode while his model contained more than four variables in the Heroine crime system. This raises the question approximately what proportion of model variables must be represented in the reference modes to be useful for model validation?

6.10.5 Main framework of the method

The learning cycle approach provides a phased approach for advancement in understanding complex situations. At each phase some data is discarded. The domain is reduced to preliminary system, and the preliminary system, in turn is reduced to preliminary model and reference mode (Figure 6.1). During the whole process, both the model and the reference modes are the product of the same process and same sources of data. Only the addition of likely future pattern differentiates a reference mode from the model. Each learning cycle leads to a next level of hierarchy in learning. The diagrams representing the methodology (Figures 6.2 and 6.3) do not explicitly show the linkages between the output of each learning cycle and its place

6-40 in the entire learning process as the learning can occur other than in participative processes. Some processes can occur at once and stages can be jumped or missed out completely depending upon the purpose of modelling, agenda one follows, tools one uses or available information. The process of developing reference mode can be further improved by including the agenda one follows and the methods or tools one employs in support of this method.

6.10.6 All steps in series or transition to alternate steps possible

The learning cycles approach suggests that all steps must be followed to reach a reference mode. In some situations where the available data or the information provided may require the use of alternate steps and arrangements may be specified for the smooth transition from one learning cycle to the next when some steps cannot be performed due to the nature of available information or available tools of analysis.

6.11 SUMMARY AND CONCLUSIONS

This chapter revisited the concept of reference modes and applied a learning cycles approach for developing reference modes for the dryland salinity problem and critically evaluated the approach on the basis of experience gained through this application.

The study indicates that the learning cycles approach is a valuable starting point when identifying variables that might be incorporated into the model. The method helps in identifying data shortcomings and it helps keep the modeller well focused on the problem being addressed. However, difficulties were encountered in application of this method. In part, the difficulties stem from expectation that a sound statistical database will be available for key variables. This was not the case in this problem nor, it is suggested, in the generality of dynamic based problems. Variations in the level of aggregation of available data proved problematic, as did a lack of involvement of the many opposite stakeholders in defining the problem space. Whilst most appropriate when consistently aggregated data is available for a specific problem space, Saeed’s method was found lacking here. The shortcomings in this instance might be reduced through extensive and close involvement of stakeholders’

6-41 right from the earliest stages when attempting to identify the preliminary model boundary. It is based on the beliefs that stakeholders are i) a part of the problem space as they influence the current system’s behaviour through their decisions, ii) an integral part of learning cycles, and iii) knowledgeable people essential to activities directed at identifying missing variables, missing data or explaining variations in quality of data. By addressing these issues, a template can be prepared based on the learning cycle’s method that may help the modellers in problem articulation and development of reference modes.

The variables identified during reference modes development and described in this chapter are used to develop a System Dynamics simulation model that is the subject of the next chapter. The behaviour of variables was used to build confidence in the simulation model. The way it was used is described in Chapter 9 on model verification and validation.

6-42 CHAPTER 7 TOOLS FOR QUALITATIVE ANALYSIS IN SYSTEM DYNAMICS

Anyone who doubts the dominant scope of remembered information should imagine what would happen to an industrial society if it were deprived of all knowledge in people's heads and if actions were guided only by written policies and numerical information. (Forrester 1992:56)

A picture is better than 1000 words. (A famous quote)

Diagrams are generally used to convey abstract ideas and concepts, to present models, help to visualise complex processes, and to recall information or to substitute lengthy text description. Some systems approaches use diagrams as the most formal expression for that particular approach (Lane, 2008). Diagrams in System Dynamics are used for the qualitative analysis. In System Dynamics, system’s description and causal loop diagrams form the qualitative part of a System Dynamics study and it these words and arrows diagrams that present dynamics hypothesis. Similar diagrams are used in systems thinking and are called influence diagrams.

The qualitative analysis in System Dynamics aims at providing illumination, improving understanding of the problem situation and creating opportunities for extrapolation to similar situations in line with the characteristics of qualitative research suggested by Heopfl (1997). Due to limitations in the use of diagrams as a stand-alone tool for System Dynamics analysis (Homer and Olive, 2001), Lane (2008) suggests the use of diagramming ideas within related disciplines in addition to diagramming convention used in System Dynamics, for example, cognitive science that uses cognitive mapping. Using different diagrams for the same purpose creates an opportunity for triangulation of insights gained from such diagrams, and hence can improve confidence in those diagrams. Triangulation can be a strategy or test for improving the validity or evaluation of findings (Golafshani, 2003) and may include multiple methods for data collection and data analysis. The methods chosen in triangulation to test the validity and reliability of a study depends on the criterion for that research (Golafshani, 2003).

7-1 The main purpose of this chapter is to describe the results of the qualitative analysis employed in conceptualisation of the dryland salinity problem in the Murray Darling Basin. The qualitative analysis used three main approaches: concept mapping, causal loop diagrams and influence diagrams. Firstly, a concept map of the dryland salinity problem developed from the literature is described. Secondly, causal loop diagrams developed for dryland salinity in the Murray Darling Basin are described along with relevant conventions and during this analysis, feedback loops concerning the dryland salinity problem were explained. The third angle of analysis was problem conceptualisation through influence diagrams that is described in the last part of this chapter.

7.1 ROLE OF DIAGRAMS IN SYSTEM DYNAMICS

In mainstream System Dynamics diagrams are used for multiple purposes, i.e., as a communication tool, to elicit ideas, information, organise information, and to succinctly present and elaborate ideas and to support ideas presented in text and words. Though System Dynamics favours development of formal simulation models based on differential equations, causal loop diagrams are also used in problem structuring phases (Wolstenholme, 1982; Wolstenholme and Coyle, 1983; Wolstenholme, 1985) or in the explanation of model structure (Forrester, 1994). Forrester (1961:81) and Richardson (1986) consider that a diagram is an intermediate transition between verbal description and equations. Lane (2008) considers communication as the main role of diagrams in System Dynamics. Randers (1980) considers that a possible use of diagrams is in model conceptualisation. Within this context, diagrams can be used as tools for thinking and elicitation of ideas. Lane (2008:5-6) considers the following roles for diagrams in the System Dynamics method:

 As a communication tool, diagrams convey the polarities of the constitutive links in the model. It makes the assumptions behind variable relationships explicit.  Diagrams can be used as a support to simulation model development;  Diagrams convey the structural source of complex behaviour; and  Diagrams show various types of variables.

7-2 Coyle (2000:240-241) suggest that simulation is not necessary for certain problems that can only be better understood through qualitative diagrams. In such cases, quantifying some relationships for simulation can be either difficult or adds only little value to the insights gained. He suggests the following roles of influence diagrams (Coyle, 2001, 2002):

 Diagrams can put on one piece of paper a view of a problem that might otherwise require many pages of text to explain.  Diagrams can serve as an effective agenda for problem articulation.  Diagrams may help during discussions; effectively a form of agenda that, unlike the normal serial agenda, shows the relationships between the items being discussed.  Diagrams show feedback patterns and observing such patterns can be helpful in gaining insights into how the problem emerged. Even though dynamics cannot be predicted from the study of these diagrams alone (Homer and Oliva 2001), the use of diagrams in a System Dynamics study may help in identifying the scope of the study or the wider contexts of a modelling task, if any.  A correctly drawn influence diagram or causal loop diagram can become the basis for a quantified model and is easily transformed into equations whether by a text editor or an iconic package. This is similar to the role of diagrams which Forrester (1961) suggests as an intermediate step in transition between problem description and equations.

7.2 DEVELOPING CONCEPT MAPS FOR DRYLAND SALINITY

The methods to draw concept maps range from informal or semi-structured to formal or highly structured methods. The former category includes maps as simple lists accompanied by some illustration in the form of images and/or hierarchical or tree structure. Later category includes highly structured methods that sometimes use quantitative approaches including statistical methods.

7-3 Concept maps were prepared using concepts embedded in literature discussed in Chapter 2 with farmers’ response in a field survey and a farmer’s cognitive maps. Decision ExplorerTM was used to prepare maps of those concepts. Concept maps are developed for different purposes and in this chapter, the purpose of developing concept maps is to identify variables that are important in developing a System Dynamics model and the relationships among these variables.

For the purpose of developing concept maps in this chapter, concepts are ideas or notions that may have links to other ideas. In developing and presenting concept maps, a combination of conventions described by McLucas (2001) and Eden and Ackerman (1998) and the capability of Decision ExplorerTM were used. Three types of the links among concepts are used:

 Causal links: these are represented by solid arrows. Direction of the arrow is in line with the direction of causality.  Connotative links: these are represented by dashed lines. The direction of causality is not depicted and it can be both ways.  Conflicting link: these are represented by an arrow with ‘C’. Presence of such links indicates that there exists a conflict between these concepts.

Key benefits of concept maps are:

 They help in structuring messy problems (McLucas, 2003; Eden and Ackerman, 1998) that includes clarifying of aims and objectives of the problem solving effort. Options can also be evaluated for their effectiveness.  Dilemmas, feedback loops and conflicts can be quickly distinguished, explored and worked upon.  It increases the user's understanding of the issue through questioning the chains of argument and figuring out where isolated chunks of data fit in. (Ackerman, 1991).  All factors relevant to a certain concept can be graphically depicted on a single paper (McLucas, 2001). It helps to see through massive details that can otherwise conceal core concepts. It improves accessibility to concepts.

7-4  Concept maps can be used as a thought crystallising tool and/or as a front-end for developing System Dynamics models.  Concept maps can also be used for strategy development and risk assessment (McLucas 2001, 2003; Eden and Ackerman, 1992).

7.2.1 Cognitive map of a Murray Darling Basin farmer

Subsequent to the field survey described in Chapter 6, cognitive maps of a Murray Darling Basin Farmer were prepared. The purpose of cognitive mapping was to explore in detail the way a farmer thinks about the causes, impacts and potential remedial options for dryland salinity.

The cognitive maps were prepared during an interview. The farmer had previously participated in the field survey in relation to this study and provided responses to the questionnaire (Supplement – Annex S1). After the interview, the cognitive maps were immediately shared with the farmer for validation. These maps are presented in Figures 7.1, 7.2 and 7.3.

7-5 3 Clearing of deep rooted perennials , replacement with 10 Federal shallow rooted government pasture legislation on 8 Federal and state clearing policies – the 1972 taxation system had incentives for 4 Government policies people who drain wetlands 17 Land clearing 12 On a state level SA 13 Periodic wet brought in clearing period bans in 1990 so contractors went to 15 Better Victoria technologies, more efficient use of water 1 Dryland salinity 14 Farmers feel 5 Poor understanding threatened ... going of soil moisture to loose viable relationships of water allocations 6 Best practice different soils today might not be the best practice in the past 11 Rapidity/fast 16 Impacts of tracking of installing irrigation irrigation system

7 Knowledge about farming...why it is not good, Govt. 9 Delivery losses departments tend to from open channel follow single theme 2 Poor irrigation system practices- flood irrigation inappropriate

Figure 7.1 Concept map of a Murray Darling Basin farmer – dryland salinity causes

7-6 8 Loss of biodiveristy

6 Restriction in the 7 precipitates- range of secondary impacts on agricultural infrastructure commodities that we can produce 5 Loss of income 1 Dryland saliniy affected areas, as compared to T non-saline areas 2 Loss of Productivity

10 feel about place what ever they do 9 Social impacts does not has 3 Water quality impacts natural supplies

Figure 7.2 Concept map of a Murray Darling Basin farmer – Impacts

7-7

2 Greater cooperation between all levels of 3 Understanding government 8 Better management Soil and land scape techniques that lead resources to better infiltration

1 Dryland salinity 4 Soil water balance

9 Reducing flood irrigation 7 Schemes which assist young farmers to take up career in agriculture 5 Better match between vegetation 6 Schemes for people types selected and who take on better soil resources technologies

Figure 7.3 Concept map of a Murray Darling Basin farmer –Potential mitigation options

7-8 7.3 A COMPOSITE CONCEPT MAP

Forty (40) concepts are presented in the concept map as shown in Figure 7.4 and these concepts have 52 links among them with 18 loops. The main concepts presented are around land clearing, dryland salinity, farm incomes, agricultural production and dryland salinity mitigation strategies.

The concept map was subjected to domain, loop detect, central, potent and co-tail analyses. These analyses were designed to identify density of links and to elaborate the concepts that are central or key in the map. A description of these analyses and the resulting lists are in Annex C. Conclusions drawn from those tests are presented in the below paragraphs.

7.3.1 Central/key concepts

The following are the first five concepts among 40 concepts that are considered central to dryland salinity. All maps and models designed to address dryland salinity must include these concepts:

 Land reclaimed from dryland salinity.  Area on which dryland salinity mitigation strategies are applied.  Loss of productive land due to dryland salinity.  Dryland salinity.  Increased costs of agricultural production.

7.3.2 Densely linked concepts

Density of links was identified through ‘Domain Analysis’ in the Decision Explorer™. Density of links helps to identify the level of dependency among different concepts.

‘Dryland salinity mitigation strategy’ and ‘dryland salinity’ are the most densely linked concepts in the map whilst the dryland salinity mitigation strategy has the highest density of links, while dryland salinity had the second highest density of

7-9 links. Other densely linked concepts are ‘shortage of land for agricultural production’ and increase in requirements of agricultural production.

7.3.3 Potent concepts

The concepts that appear in a higher number of Hierarchical Set Clusters (Hiesets) are considered potent concepts. Hiesets based on a specified set of concepts are obtained as a result of Hiest analysis in the Decision Explorer™.

The highest number of ‘Hiesets’ is 4 in this concept map (Figure 7.1). The following concepts appear in 4 ‘Hiesets’ and are considered the most potent concepts:

 Dryland salinity.  Rise in water table.  Climate change.  Different causal theories of dryland salinity.  Salt deposit carried away with rain, groundwater.

The second most potent concepts are:

 Degraded land quality.  Decrease in agricultural production.  Increase in requirements for agricultural production.  Increase in agricultural exports.  Increase in domestic consumption.

These concepts appear in 3 ‘Hiesets’.

7-10 7 Increased 22 Increased cost of 18 Decrease in value construction costs 21 decrease in Agricultural of land either salt 38 Farm assett's of urban and biodiversity production affected ot at the Value transport risk of becoming infrastructure salt affected 23 Increase in value of productive land neither salt 17 Climate change affected nor at the risk of becoming 5 Degraded land salt affected 78 Farm Debt 16 Decrease in quality rainfall due to 6 Decrease in drought 1 Dryland Salinity agricultural production 36 Farm income 30 salt deposit and 12 Increase in agricultural exports C 3 rise in watertable carried away with 29 diff erent causal rain, groundw ater 37 Farmers theories of dryland w illingness to apply 10 Increase in 2 Increase in salinity dryland salinity groundw ater recharge requirement for control treatments agricultural 28 Cost of dryland 15 Required increase 19 Loss of salinity control in land to production productive land due treatments to dryland salinity accommodate increase in requirements f or 4 Increase in land agricultural 11 Current 13 Increase in clearing 31 Salt interception production agricultural domestic consumption schemes production

39 Restrictions on 26 Effectiveness of land clearing dryland salinity control treatments 32 Agronomic 25 Land reclaimed Solutions 20 Required increase from dryland 9 Availability of in productive land salinity land for clearing 27 Dryalnd salinity control treatments 33 Plantations 14 Current land under agricultural production 24 Area on w hich dryland salinity 35 Stabilization of 8 Shortage of land 34 Targetted control treatments salty patches/ for agricultural reforestations are applied saltsotres production Figure 7.4 A concept map of dryland salinity in the Murray Darling Basin

7-11 7.4 FEEDBACK/CAUSAL LOOPS

Causal loop diagrams are graphical representations of the cause and effect relationships. These contain variables, direction of the causal relationships and emphasis of the relationship. A system in causal loop diagrams is represented by positive and negative feedback loops.

Causal loop diagrams are widely used in System Dynamics analysis as an aid to simulation model building and for some problems these have been used as a stand- alone analysis tool. Richardson (1986) has criticised causal loop diagrams considering them inadequate on the grounds of their inability to represent the stock and flow structure that is fundamental to System Dynamics analysis. The use of causal loop diagram Richardson (1986) suggests is to explain the model for public consumption. After a review of initial problems, Richardson (1997) supports the old conventions of writing causal loop diagrams (with positive and negative signs) in contrast to popular convention with polarity demonstrated by ‘O’ (opposite direction) and ‘S’(same direction) symbols. Richardson (1997: 251) concludes:

I submit that our current enthusiasm for S’s and O’s is significantly misguided and needs to be curbed. Positive and negative polarities stand up much better to the deep thinking we are striving to facilitate in and about complex systems.

7.4.1 Reinforcing loop or positive feedback loop

A reinforcing loop is the structure which feeds itself to produce growth or decline. In other words an important variable accelerates up or down with exponential growth or collapse and growth or collapse continues at an ever increasing rate (Senge, Ross et al., 1994; Bellinger, 2004). The structure and behaviour of this loop is shown in Figures 7.5 and 7.6.

7-12 +

A B

+

Figure 7.5 Positive or reinforcing loop

t t

Figure 7.6 Behaviour of reinforcing Loop (Senge et al., 1994).

7.4.2 Balancing loop or negative feedback loop

A balancing loop or negative loop attempts to move some current state of the system to a desired or reference state of the system through some action. The structure may begin with the current state greater or less than the desired state, in which the current state adjusts itself according to desired state. The structure and behaviour of this loop is shown in Figures 7.7 and 7.8.

7-13 +

A B

-

Figure 7.7 A balancing or negative feedback loop

t t

Figure 7.8 Behaviour of balancing loop (Senge et al., 1994)

7.4.3 Developing feedback-loop diagrams for dryland salinity

Causal loop also called feedback loop diagrams were prepared based on the relationships inferred from literature, theoretical principles, secondary data, reference modes and concept maps in the preceding sections. The main causal loop diagram at

7-14 the strategic level is shown in Figure 7.14. It consists of fourteen variables that are depicted in five feedback loops. Each feedback loop containing definitions of variables, relationship between variables and reasoning behind that relationship is described below.

7.4.4 Feedback loop – cleared land availability

This loop (Figure 7.9) consists of two endogenous variables and two exogenous variables to this loop:

Endogenous variables

 land available for clearing; and  cleared land.

Exogenous variables

 restrictions on land clearing; and  demand for good quality land for agriculture.

The relationship between ‘land available for clearing’ and ‘cleared land’ is depicted as positive which means if there is more land available for clearing it will increase the rate of land clearing. Historically in the Australian landscape, land clearing rates were highest in the first half of the last century. During that period, there were fewer restrictions on land clearing and demand for good quality land was high due to increase in population and export potential of agricultural commodities. Until the 1970s, land leases in NSW were made on the proviso that the occupier will clear a certain portion of the land (NSW 2000). On the basis of this historical information, the relationship between is depicted as positive.

In the last three decades, restrictions on land clearing have progressively increased. The relationship between ‘restrictions on land clearing’ and ‘cleared land’ has been depicted as negative, i.e., an increase in restriction will decrease land clearing. It is the general logic underlying imposition of restrictions and for the purpose of this

7-15 feedback loop, it is assumed that this general logic is valid and landowners comply with retractions on land clearing.

The relationship between demand for good quality land and land clearing is depicted as positive on the proviso that the way to fulfil demand for good quality land for agriculture is through land clearing. As the demand for agricultural land increases, it causes increased land clearing.

The relationship between ‘cleared land’ and ‘land available for clearing’ is depicted as negative on the assumption that there is a certain amount of land in the landscape that is available for clearing. As a portion of it is cleared, it decreases the inventory of the land available for clearing.

Restrictions on land clearing

- L1 + Land available Cleared land for clearing (-) - +

Demand for good quality land for agriculture

Figure 7.9 Feedback loop – cleared land availability

7.4.5 Feedback loops – threats to land availability

This is a positive feedback loop shown in Figure 7.10. It has five variables that are endogenous to the loop. These variables are:

 Cleared land: already defined in Section 7.4.4.

7-16  Groundwater recharge.  Land either salt affected or at the risk of becoming salt affected.  Good quality land discrepancy.  Demand for good quality land for agriculture: already defined in Section 7.4.4.

The relationship between land clearing and groundwater recharge is depicted as positive on the assumption that land clearing reduces evapo-transpiration and increases the ground recharge. This increased recharge, along with the groundwater movement and based on groundwater response times, adds to the ground water in different areas and moves the watertable close to the land surface.

The relationship between groundwater recharge and land either salt affected or at the risk of becoming salt affected is positive on the grounds that a rise in the water table may dissolve salts in different soil layers. It may cause land to become either salt affected or at the risk of becoming salt affected (when the water table reaches within two meters of the land surface). The relationship between natural vegetation and the hydrological cycle is explained by Evans, Newman et al. (1996):

The native vegetation regimes evolved to make the best use of available rainfall while avoiding the salts. All vegetation pumps water from the soils and transpire a component to the atmosphere (through the evapo-transpiration process). Any change in vegetation density or type (e.g., a change in vegetation’s water pumping capabilities) will alter the volume of water reaching the saturated zone below. Clearing of native vegetation disturbed the current balance.

The relationship between ‘land either salt affected or at the risk of becoming salt affected’ and good quality land discrepancy is depicted as positive on the assumption that as the land becomes at risk of becoming salt affected, it reduces good quality land and therefore increases the discrepancy of good quality land.

The relationship between good quality land discrepancy and demand for land clearing is depicted as positive. As the discrepancy increases, through information

7-17 flows it increases demand for good quality agricultural land considering the production expectations are unchanged.

The relationship between demand for good quality agriculture and land clearing is considered positive as if the land is cleared a demand for agriculture exists.

+ Land either salt affected GW recharge or at the risk of becoming salt affected +

(+) Cleared land +

+ Go o d quality land discrepancy

Demand for good + quality land fo r agriculture

Figure 7.10 Feedback loop – threats to land availability

7.4.6 Feedback loop – farming land availability

This loop has four endogenous and three exogenous variables:

Endogenous variables

 Cleared land: already defined in Section 7.4.4.  Available good quality land for agriculture: the land that has water table below two meters from land surface.  Good quality land discrepancy: already defined in Section 7.4.5  Demand for good quality land for agriculture: already defined in Section 7.4.4.

7-18 Exogenous variables

 Mitigation strategies for dryland salinity: these may consist of engineering options, agronomic options, and other options or in a combination. These options are described in Chapter 2.  Land either salt affected or at the risk of becoming salt affected: already defined in Section 7.4.5.

The relationship between ‘available good quality land for agriculture’ and ‘good quality land discrepancy’ is considered negative. If more good quality land is available for agriculture, it will reduce the discrepancy of good quality land as land discrepancy is the difference between available good quality land for agriculture and the land either salt affected or at the risk of becoming salt affected.

All other relationships in the loop have been described in earlier sections. The feedback loop is shown in Figure 7.12.

Treatments for dryland salinity

+

Available go o d quality land for agriculture +

(-) - Goo d quality land Cleared land discrepancy + +

Demand for good + quality land fo r Land either salt affected agriculture or at the risk of becoming salt affected

Figure 7.11 Feedback loop – farming land availability

7-19 7.4.7 Feedback loop - productivity

This is positive feedback loop consisting of six endogenous variables:

 Revenue: the sale value of agricultural production.  Farmer income: this is the net margins from agricultural production that is revenue minus costs.  Money available for investment on mitigation: this is the portion of a farmer’s income that is available and he/she is willing to invest in mitigation strategies for dryland salinity.  Mitigation strategies for dryland salinity: as defined earlier.  Available good quality land for agriculture: as defined earlier.  Agricultural production: this is the total agricultural production.

The relationship between ‘revenue’ and ‘farm income’ is positive on the assumption that revenue increases farmer income considering other determinants of farmer income remain unchanged. The relationship between farmer’s income and money available for investment on mitigation strategies is positive on the assumption that the portion of farmers income that goes into dryland salinity mitigation will increase as income increases. Money available for mitigation strategies increases the application of mitigation strategies for dryland salinity that in turn increases available good quality land for agriculture and consequently increases agricultural production. The feedback loop is shown in Figure 7.9.

7-20 + Farmer incomes Revenue

+

+ (+) Money available for Agricultural investment on land production treatment +

+

Available go o d quality land fo r agriculture Treatments for dryland + salinity

Figure 7.12 Feedback loop - productivity

7.4.8 Feedback loop – income and cost of production

This is a negative feedback loop consisting of four endogenous and three exogenous loop variables:

Endogenous variables

 Farmer income: already defined in Section 7.4.7.  Money available for investment on mitigation strategies: already defined in Section 7.4.7.  Mitigation strategies for dryland salinity: already defined in Section 7.4.6.  Cost of production.

Exogenous variables

 Revenue: already defined in Section 7.4.7.  Available mitigation strategies for dryland salinity: this is an inventory of the effective mitigation strategy for dryland salinity available for application on salt affected land.

7-21  Land either salt affected or at the risk of becoming salt affected: already defined in Section 7.4.5.

The relationship between ‘application of mitigation strategies for dryland salinity’ and cost of production is positive, this means that the cost of production on salt affected land increases with the application of mitigation strategies. Increased cost of production decreases farmer income considering other determinants of farmer income remain unchanged. Therefore, the relationship between cost of production and farmer income is negative.

The relationship between availability of mitigation strategies and application of mitigation strategies is positive as an increased inventory of mitigation strategies improves the choice for application of those strategies. The feedback loop is shown in Figure 7.13.

Revenue +

Farmer incomes -

+

Money available for Cost of production (-) investment on land + treatment

+ Application of mitigation strategies for dryland salinity

+ +

Land either salt Available control affected or at the risk treatments for of becoming salt dryland salinity affected

Figure 7.13 Feedback loop – income and cost of production.

7-22 7.4.9 Composite causal loop diagram

In a dryland salinity system, land is considered under different covers/uses. The primary land cover is natural vegetation and based on the demand for land clearing, the land under natural vegetation is cleared if it is available for clearing. Higher rates of land clearing reduce the total land under natural vegetation, Figure 7.11 shows this relationship. Land is cleared according to Land Clearing Discrepancy (LCD). As the LCD increases, more land is cleared subject to its availability. LCD is a difference between the demand for cleared land and already cleared land. The demand is considered exogenous to the model as it may vary depending upon multiple factors like the conditions of the external global market, commodity prices, political objectives of the government, e.g. settlement in a certain areas. Land clearing operations depending upon their time-lags convert land cover/use from a naturally vegetated land to cleared land available for multiple uses.

After a certain time-lag, the impacts of land clearing start to emerge, and the decision makers implement strategies or control measures to combat dryland salinity. These strategies after a delay, depending upon their effectiveness, help to lower water tables either by improving land cover or through their water utilisation (depending upon the root/shoot characteristics of plants). In this way the entire system locks itself into land degradation from land under natural vegetation into land that is either salt affected or at the risk of becoming salt affected. Over time as the land availability for clearing becomes less and less the rates of land clearing decreases.

7-23 Revenue + +

Farmer incomes - + Productivity (+) Income and Cost Agricultural production Cost of production of Production Money available for + (-) investment on land treatment +

Treatments for dryland + Restrictions on Available go o d quality + salinity land clearing + land for agriculture +

Farming Land (-) Land either salt affected Availability or at the risk of becoming salt affected + + GW recharge - - Land available Cleared Land + Good quality+ land Availability Cleared land for clearing (-) Threats to Land discrepancy - Availabilbity + (+)

Demand for good quality land for + agriculture Figure 7.14 Composite feedback loop diagram of the dryland salinity problem in Australia.

7-24 7.4.10 Physical processes leading to salinity

Physical processes that lead to dryland salinity include the ones affecting the water table and the salts either present in the water or in the soil mass. The feedback relation involved in the management of the ground water table and the movement of salts are presented in Figures 7.15 and 7.16.

The physical model identifies the relationships around maintaining the depth to water table in the surface soil. This feedback loop has three sub-loops presenting depth to water table as influenced by a depth to water table discrepancy and the ground water intruding root zone. The main variables are defined below:

 Depth to water table: is the distance between land surface and the water table.  Evapo-transpiration: this is the amount of water that is taken up by plants and transpired into air.  Depth to water table discrepancy.  Groundwater intruding root zone – annual quantity of groundwater intrusion.  Precipitation: annual rainfall.  Deep percolation: this is the part of infiltrated water that adds into the groundwater.  Groundwater discharge.

The feedback loop diagram shows that the water table (depth to water table) is maintained by the rising groundwater, receding groundwater, precipitation, deep percolation and evapo-transpiration. The direction of influences is presented in Figure 7.15. As has been discussed in Chapters 2 and 6, when the water table becomes within 2 meters of land surface, the land is considered at the risk of becoming salt affected (NLWRA, 2001).

7-25 Critical depth to groundwater Critical discharge level

- + Groundwater Depth to water table discharge discrepancy + discrepancy +

-

Evapo-transpiration Depth to water table + - - + + + - Groundwater Groundwater intruding discharge root zone

Percolation Precipitation

Figure 7.15 Feedback loops - depth to watertable. Note: Parts of this feedback loop have been adopted from Saysel and Barlas (2001).

The second part of the physical processes involves salt movement. A feedback loop of the salts movement between soil and water is presented in Figure 7.16.

Salts in wind + + Salts on land surface

- Salts in precipitation + + Salts in infiltration water Salts in the evaporating water +

+ + Salts in the soil mass - +

+

Salts in capillary water + Salts in the percolating water +

-

+ Salts in the ground water

Figure 7.16 Salts in the soil mass

7-26 7.5 MITIGATION STRATEGIES

Various options for dryland salinity have been discussed in Chapter 2. This Section addresses the linkages of mitigation strategies identified through the literature review, cognitive/concept mapping, and causal analysis and various aspects of dryland salinity. Such linkages are presented in Figure 7.17. The following groups of mitigation strategies were used in the analysis:

 Plant based option: this set of options includes improvement in on farm agronomic practices farming systems that effectively utilise groundwater. This strategy aims to improve farm incomes and improves use of water through causal relations depicted in Figure 7.14. The relationship between agronomic practices, farm incomes and agricultural production has been depicted as positive as better agronomic practices are supposed to improve agricultural productivity. Salt bush and trees of various species also included in the plant based options. Cognitive maps of the Murray Darling Basin Farmer (Figure 7.3) identified that levers that help to manage soil water balance include: o identification of better management techniques that lead to better infiltration; and o better match between vegetation types and soil resources.  Stabilisation of salt patches: this strategy originates from Acworth and Jankowski’s (2001) hypothesis about causes of dryland salinity. Acworth and Jankowski (2001) suggest a change in the management options and further suggest that to identify the unstable clayey units in the landscape, stabilise them and prevent their saturation by groundwater or contact with rainwater. This stabilisation may include targeted reforestation, or inclined bores as discussed in Chapter 2.  Reforestation/targeted reforestation: this option includes reforestation in areas with specific hydro-physical characteristics. It aims at increasing the volume of water transpired and therefore lowering the watertable.  Engineering options: such as deep drainage, deep open drains and pumping (also explained in Chapter 2) have relatively high costs of entry, and difficulty with trialling on a small scale. There are also issues with the disposal of

7-27 pumped/drainage water. Uncertainty around the achievable hydrological impact of engineering works prior to installation places a large disincentive on investment. Engineering options focus on physically removing saline water from problem areas.  Adaptation to saline environment: this option includes development of saline industry and farming systems that thrive in saline environments. This strategy aims at maintaining farmers’ incomes from land as well as ameliorating the saline environment through reducing the dryland salinity risk by lowering water tables.  Capacity building: includes improving the capacity of farmers to implement mitigation by raising awareness about basic solutions, education and training, and demonstration centres. Demonstration centres can be used to show a combination of salinity management practices.  Reducing the cost of dryland salinity remediation measures.  Addressing labour shortages: a scheme/program that encourages young farmers to stay on land or adopt agriculture as a profession.

7-28 Revenue + + + Development of + salt industry EducationAgronomic and Awareness Farmer incomes + - trainingngoptions+ compaigns + + Productivity (+) Income and Cost + Agricultural production Cost of productionof Production Money available for Demonstration + (-) investment on land centres treatment +

Mitigation + Restrictions on Available good quality + strategies land clearing + land for agriculture + - Farmers' w illingness to apply mitigation strategies + (-) Farming Land Land either salt affected Availability or at the risk of + becoming salt affected - + Reducin + + GW recharge Land available - Go o d quality land- + Cleared Land Cleared land + + for clearing discrepancy Availability Threats to Land - (-) g cost of Availabilbity + dryland salinity (+) remediation measures Demand for good Critical discharge level Critical depth to quality land fo r + groundwater agriculture

- + + Groundwater discharge Depth to water table + + discrepancy + + discrepancy Targetted - Stablisation of reforestation salt PatchesSalts in wind Salts in precipitatio n + + Salts on land surface Depth to water table Evapo-transpiration - - + + + + + - + + Groundwater+ intruding Groundwater discharge - Salts in infiltratio n water rootzone Salts in the evapo rating - Agronomic + + water + Precipitatio n options

+ Percolation + Salts in capillary w ater Salts in the so il mass + Salt interception + schemes - - + + + Salts in the perco lating Figure 7.17 Compositewater feedback loop including physical aspects and mitigation strategies

Salts in the gro und water 7-29 7.6 INFLUENCE DIAGRAMS

Wolstenholme (1999:422) suggests that causal loop diagrams are referred to as influence diagrams in the United Kingdom community of systems thinkers. These are also words and arrows diagrams, however, different conventions (from causal loop diagrams) are used to draw and interpret them. Wolstenholme (1990:13) suggests that an influence diagram is an alternative name for the causal loop diagram. Due to separate conventions governing the drawing of influence diagrams and causal loop diagrams, it is easier to locate stock and flows within an influence diagram as compared to stock and flow diagrams. This is the key difference between a causal loop diagram and an influence diagram. Richardson (1986, 1997) suggests that one of the problems with causal loop diagrams is that the causal loop diagrams do not show which variables within the diagram are stocks and which are flows. This issue may be addressed by using influence diagrams as an intermediate step between a causal loop diagram and the stock and flow diagram. In the influence diagrams, the influences of flows on stocks are specified using arrows, delays are marked with a symbol ‘D’ and the polarity (either + or -) is specified while the sources and sinks are omitted (Wolstenholme, 1990:13-15).

An influence diagram of the dryland salinity problem was developed using the relationships described in the literature, i e. Chapters 2, 3 and 5. Figure 7.19 shows this influence diagram. The relationships between twenty variables are represented in the influence diagram. Definitions of the variables used are given in Table 7.1. Physical flows are represented by the unbroken direct arrows while the influences are represented by the broken lines. The direction of the arrows indicates flow direct. The conventions for influence diagram are described in Annex D.

For developing an influence diagram, major influences on various variables, their direction of flow and the type of flow were mapped. An example is shown in Figure 7.18.

Key concepts represented in the influence diagram are land clearing, salinisation, land discrepancy and mitigation strategies. There exists a physical flow between

7-30 different land covers—the land available for clearing—the land under forest or bush, with either primary or secondary vegetation converted into cleared land. This process conveys an influence to the inventory of the land available for clearing. With the increase in land clearing, the inventory of land available for clearing decreases. The cleared land over time becomes salt affected due to the groundwater recharge in the upland areas which reduces the quantity of the cleared productive land. The rate of land becoming salt affected increases the land either salt affected or at the risk of becoming salt affected through a positive physical flow.

A key advantage of the influence diagram over a causal loop diagram is that it helps to differentiate between the stocks and flows in the system.

7-31 Land clearing

Effectiveness of the treatment Depth to watertable

+ + Land either salt affected or at the risk - Land abandoned due of becoming salt Application of to dryland salinity affected (Low land) control treatment +

-

Crop Yields/ production

Cost of production Money available for + investment for land + Farmers incomes treatment

Crop prices

Figure 7.18 Example of the identification of influences on a variable during the process of developing an influence diagram

7-32

Rate of land Rate of land Rate of land abandonment due to becoming salt D clearing D dryland salinity affected

Land either salt Land abandoned Land available Cleared land affected or at the risk dryland sali for clearing (upland) Uptake by of becoming salt n plants affected (low land) Evapo- Effectiveness of treatments Precipitation transpiration Percolation Rate of land Groundwater reclamation Depth to recharge watertable

Movement of salts in soil mass

Rate of land returning to natural vegetation Demand for cleared land

Land discrepancy

land required to maintain current level of agricultural production

Figure 7.19 Major influences affecting the dryland salinity problem in the Murray Darling Basin, Australia.

7-33 Table 7.1 Definition of variables used in the influence diagram Variables Definitions Rate of land clearing. It is the rate, i.e., hectares per year at which the land under primary/secondary vegetation is cleared to convert it into agricultural land. Rate of land becoming salt affected. It is the rate at which cleared land becomes salt affected. The processes are explained in Chapter 2. Rate of land abandonment due to dryland It is the rate at which the salt affected salinity. land is so badly scarred that it can no longer be profitably used for agriculture and is rendered abandoned. Cleared land. This is the inventory, stock of the cleared land at a certain point in time. Land either salt affected or at the risk of It is the stock of land under which the becoming salt affected. groundwater surface is within 2 meters from the ground surface. It is defined consistent with the Land and Water Resources Audit (NLWRA, 2000). GW recharge. It is portion of the precipitation that percolates down and is added to the groundwater though surface or subsurface drainage. Land available for clearing. It is the stock of land that is under natural vegetation and has not been subject to land clearing. Demand for cleared land. It is the land required to meet production objectives. Rate of land returning to natural It is the land abandoned or otherwise left vegetation. unattended and returns to natural vegetation cover over time.

7-34 Rate of land reclamation. It is the rate, i e. hectares per year that is reclaimed from dryland salinity. Effectiveness of mitigation strategies. This is the quality of a dryland salinity mitigation strategy. Land discrepancy. It is the difference between the land required to meet expected production and the land available for agriculture. Depth to water table. It is the distance between the land surface and watertable.

Percolation. It is the portion of infiltration that adds to the groundwater.

Salts in soil mass. The quantity of salts in the top-soil and the sub-soil.

7-35 7.7 SUMMARY AND CONCLUSION

In this chapter, qualitative approaches used in the study are described. This included development of concept maps, causal loop diagrams and influence diagrams. All three analyses were based on the secondary data and literature review described in earlier chapters. The concept map presented main concepts and factors involved in the lands becoming salt affected. The causal loop diagrams presented a dynamic hypothesis while the influence diagrams presented the cause and effect relationship amongst those factors.

The three diagramming conventions described capture views of the problems at different levels of aggregation in an attempt to represent the complexity of the dryland salinity problem and its management. The main groups of variables of interest that a model of dryland salinity should include land clearing, dryland salinity, land quality, revenue and mitigation strategies/remedial measures to address problems. All three diagramming conventions represented core variables in the diagrams; however, diagrams differ in detail. Some variables, if not included in the model can provide context of the problem for descriptions. All three diagramming conventions are useful as they depicted the core concept.

The variables identified through qualitative analysis will be considered for inclusion into the simulation model described in Chapters 8.

7-36 CHAPTER 8 SIMULATION MODELLING IN SYSTEM DYNAMICS

Our intuitive judgement is unreliable about how these systems will change with time, even when we have good knowledge of the individual parts of the system. (Forrester 1961:14)

When experimentation in the real system is infeasible, simulation becomes the main, and perhaps the only way, learners can discover for themselves how complex systems work. (Sterman 2000:45)

Within early texts in System Dynamics, Forrester (1968:3-5) provided the following brief description of simulation:

…equations described how the system changes and these changes were accumulated step-by-step to unfold the behaviour pattern of the system. But the equations did not tell how to go directly to some distant future without first computing through all of the intermediate stages. This process of step-by-step solution is called simulation.

Within the System Dynamics modelling methodology, problem conceptualisation progressively leads to development of a simulation model utilising the insights gained during the problem conceptualisation stage.

A simulation model in System Dynamics generally consists of a computer model with:

 a graphic model using the System Dynamics language (stocks and flows, auxiliary variables, constants etc.), directions of material and information flows and time delays within processes;

 a set of mathematical equations describing how different components are linked together to represent the conceived problem situation;

 a database consisting of input and output data, if needed; and

8-1  a user interface that consists of input controls like slider bars, buttons, graphs and the output controls like graphs, bar charts, gauges and other data analysis tools and presentation tools.

Within the mainstream System Dynamics, simulation modelling is considered an essential component of the methodology (Forrester 1963; Forrester 1968; Richardson and Pugh 1981; Forrester 1985; Richardson 1991; Sterman 1994; Lane 2000; Sterman 2000; Homer and Oliva 2001; Richardson 2001) to an extent that System Dynamicist David Lane (Coyle 2000) considered System Dynamics without quantification as an oxymoron and called it System Dynamics lite [sic]. On the other hand, Wolstenholme (1990:48) suggests that as the modelling moves from hard to softer areas of the system spectrum, quantification becomes more difficult. Homer (1992) emphasises the capabilities of human mind in capturing realistic details and dynamics that may be overlooked by mathematical models and suggested that this fact is appreciated by government decision makers who often reach their own conclusions rather than rely on mathematical models. Homer’s (1992) position is supported by Coyle (1999, 2000) who presented a few case studies to support his conclusion that there exist difficulties in quantification of soft variables in System Dynamics models and qualitative models in such cases can provide useful insights. McLucas (2001) makes a similar observation.

Rationale and benefits of System Dynamics qualitative modelling were discussed in Chapter 7. This chapter explains System Dynamics views on simulation and critically analyses the value that simulation adds in addressing complex dynamic problems. This enables the discussion on the development of a simple conceptual simulation model for strategic analysis of the dryland salinity problem in Australia using a modular approach. In the later parts of this chapter, the insights gained in developing this model are discussed and the process of model development is critically analysed.

8.1 WHY SIMULATION IS NECESSARY IN SYSTEM DYNAMICS

Early works in System Dynamics consider simulation as the only practical way to test mental models (Forrester 1961; Forrester 1968; Richardson and Pugh 1981) and

8-2 suggest that without simulation even the best qualitative maps or diagrams can be biased or driven by some ideology (Sterman 1994, 2000). Formal mathematical models are unambiguous in their assumptions and produce results reliably consistent with these assumptions, and this is a clear advantage not shared by qualitative mental models (Homer 1992).

Forrester (1968), Richardson (1981, 1991) and Sterman (1994, 2000) emphasise the important role of simulation modelling in System Dynamics. Schwaninger (2008:439) observes:

Nevertheless, it must be noted that the use of computer-based simulation modelling of continuous flows of decisions and influencing actions— derived from the field’s servo-mechanistic engineering roots—has been seen as a distinctive attribute of System Dynamics.

System Dynamics literature (Forrester 1961, 1963, 1968; Sterman 1994, 2000) presents the following reasons for using simulation modelling:

 Most problem structuring methods yield qualitative models that represent the mental models of people involved in developing these models. These models omit causal relationships, parameters, function forms, exogenous variables and initial conditions. Such mental models are numerically deficient and omit feedback, time delays, accumulations and non-linearities. Simulation is a way to test these qualitative models (Sterman 1994, 2000).

 There are severe limits in the cognitive ability of the human brain to process multivariate problems without computer simulation (Wolstenholme, 1990). Due to cognitive limits and in the absence of convenient and trusted tools to help us deal with dynamics behaviour, we cannot mentally manage all the facets of a complex system at one time. Therefore, we tend to break the system into its components and draw conclusions independently from the other components. Such fragmentation fails to show how these components or subsystems interact to produce a specific behaviour (Forrester 1968:3).

 Using simulation models to test the qualitative models often leads to radical changes in the way we perceive reality and carve problems out of messes

8-3 (Ackoff 1979; Sterman 1994, 2000). Simulation modelling exposes the assumptions within systems and brings these assumptions as a subject to debate, and therefore, speeds up learning and improves mental models and qualitative maps. A detailed analysis of relationship within models forces attention on the issue of difficulty in quantification (Wolstenholme 1990:48).

 Processes of accumulation are important in understanding dynamics. System Dynamics employs simple numerical methods based on differential equations to represent the process of accumulation. Stocks are the numerical integration [accumulation] of rates over a period of time (Wolstenholme 1990:48-49).

 In most cases, simulation is needed to assess the importance of reinforcing loops, to deduce resulting behaviour, and to provide a basis for interventions in a complex system (Lane 2008).

 Development of simulation models and subsequent experimentation speeds up learning feedback. In the absence of simulation models, even the best maps, models or diagrams can only be tested and improved through learning feedback through the real world. Such feedback is slow and rendered ineffective by dynamic complexity, time delays, inadequate and ambiguous feedback, poor reasoning skills, defensive reactions (Argyris, 1982) and costs of experimentation (Sterman 1994, 2000).

8.2 NEED FOR A MODULAR APPROACH FOR DEVELOPING SYSTEM DYNAMICS MODELS

The development process for a simulation model starts from simple to more complex systems as Richardson (1996:142) suggests:

…understanding connections between model structure and behaviour comes from a sequential modelling process that moves from simpler formulations to a more complex structure.

McLucas (2005) favours a modular approach for development of System Dynamics models in which the term ‘module’ refers to a small specifically focused component of model. McLucas (2005:30) uses the terms module, sector or model to refer to conceptual objects with varying levels of complexity and suggests:

8-4 The modules, sectors and models are intellectual devices or transient conceptual objects, which we use to describe, communicate ideas about, and represent parts of the real world around us. They are necessary simplifications of the real world produced for a specific purpose, such as providing the basis of analysis and informing our understanding of the past behaviour and what future behaviour might look like.

McLucas’ (2005) representation of the model, sector or model is based on realist ontology evidenced by the criterion ‘necessary and sufficient representation of real world’ and represents the engineers’ systemic view of the world. Although in the early work of Jay Wright Forrester, there is emphasis on the real world representation, the subsequent work on the philosophy of System Dynamics leads more towards a relativist or constructivist view used in socials sciences (Lane 2000; Pruyt 2006). The reference modes are based on data or the description or the knowledge contained within the mental models or social constructs to which System Dynamics models are compared to gain confidence in the model. However, within the pluralist world view that allows the use of multi-methodology, methods varying paradigmatic assumptions can be used (for a pluralist world view, please refer to Chapter 5, also see Midgley, 2000).

Engineers strive to achieve an end which is defined and documented at the start. Checkland (1981:32-33) describes the following generic process of an engineer’s work:

Engineering thinking is teleological; it asks what is the purpose served by the object/system? The engineer works back from the purpose, or objective, and creates an object or system which will achieve that objective.

A specification is produced which gives a careful description of what is required, whether a physical object (for example, a particular kind of valve for an oil rig) or a complete system (for example a petrochemical complex). The professional skill of an engineer is then used to meet the specifications in the most efficient, economic and elegant way.

8-5 Within the systems engineers’ worldview of models Stevens, Brook et al. (1998: 255) suggest that the characteristics of an ideal model include fidelity, simplicity, validity bands, resolution, clarity, neutrality and traceability. Both Stevens, Brook et al. (1998) and McLucas (2005) emphasise that the model should be as simple as possible and suggest the use of Occam’s Razor as a guiding principle to prevent unnecessarily complex models. Occam’s Razor also referred to as Ockham’s Razor is a principle attributed to the fourteenth-century English logician William of Ockham. Merriam-Webster Online Dictionary (2009) defines Occam’s Razor as:

…a scientific and philosophic rule that entities should not be multiplied unnecessarily which is interpreted as requiring that the simplest of competing theories be preferred to the more complex or that explanations of unknown phenomena be sought first in terms of known quantities.

Williams (2002:43 cited in McLucas, 2005) interprets as if few entities or reasons are sufficient to explain a phenomenon, the explanation is preferable to the one using many entities. The application of this principle in System Dynamics models favours the use of simple models that are appropriate to the task (McLucas, 2005).

As discussed above, one of the purposes of building System Dynamics models is to enhance understanding. McLucas (2005:38) suggests that:

...to achieve a valid understanding and derive valid learning outcomes from modelling and simulation, the models must satisfy important criteria. The models must be both a necessary and sufficient representation of the real world to be useful and to inform valid conclusions: to do this they must be error-free.

One of the Executive Editors of the System Dynamics Review—the official journal of the System Dynamics society—Winch (2000:12) expresses the state of the practice of System Dynamics as:

Regrettably, some of the articles submitted to the SDR (System Dynamics Review) and seen in conferences contain fundamental errors in equation

8-6 structures or dimensions, for example that must have meant the analyst had no real understanding of the nature of feedback structures and of the relationships between stocks and flows. Yet these people still confidently present important conclusions, and one has to challenge that their false basis means that they are not System Dynamics, nor frankly even acceptable science.

This may be due to the fact that System Dynamics intends to model complex problems within the system where problems are difficult to be separated from the system. System Dynamicists strive to identify feedback loops that are difficult to grasp because these feedback loops can be distant in time and space. A System Dynamicist’s major time and effort is spent in causal analysis as conception of these feedback processes or generation of a dynamic hypothesis is an iterative process. Once these feedback loops are identified, System Dynamicist’s efforts shift towards building a computer simulation model. As the problems that System Dynamics intends to address are complex and intentions of System Dynamicists are to conceptualise the whole system structure and resulting dynamics due to various sources, the models built with good intentions may have technical errors that result in models that are difficult to validate or contain errors in their formulation. The System Dynamics generic structures or famous models used within System Dynamics are a result of the knowledge accumulated by experts over 50 years of history of the method through model development and experimentation. The experiential knowledge that is available to System Dynamics pioneers and experts might not be available to the enthusiastic novice modeller or the people from other disciplines who want to extend their knowledge domain by learning or adopting System Dynamics.

Once the feedback structure is identified through qualitative analysis, a breakdown of the system structure into manageable components (which McLucas (2005) refers to as modules) can help in quality assurance, error checking, debugging and model verification. These modules in this instance provide simple structures whose functional architecture and equations can be readily verified, thereby, facilitating learning about model structure and dynamic behaviour. Once developed, such

8-7 modules can be integrated with other modules to achieve the level of integration suitable to modelling purposes.

8.3 OVERVIEW OF THE STRATEGIC FRAMEWORK FOR SYSTEM DYNAMICS MODELLING FOR DRYLAND SALINITY

For this study, a simple model was developed using the software PowersimTM Studio 2005. The selection was based on the key characteristics of the software, its abilities to model causal models, versatile graphic tools for easy communication and availability and knowledge accumulated about PowersimTM.

In the mid-1980s the research aimed at improving the quality of high school education using System Dynamics models resulted in Mosaic, an object oriented system aimed primarily at the development of simulation based games for education. PowersimTM Constructor was later developed as a Windows based environment for the development of System Dynamics models. It facilitates packaging as interactive games or learning environments (Eberlein 2009).

8.3.1 Model purpose and intended users

The purpose of this modelling effort was to develop a model that helps to understand the impacts of delays on different land categories undergoing dryland salinity. However, the model does not intend to predict the quantity of actual salt affected lands or the quantity of salt at a certain geographical location. The model incorporates some of the variables identified in the various causal loop diagrams. The model should be concise and simple enough to be used for communication purposes and it should provide a user interface to allow users to change the inputs. The model should exhibit past behaviour close to the one identified in the reference modes (Chapter 6). As the reference modes were prepared using descriptive data, the model output is not expected to provide statistical correlation but the visual patterns of dynamics behaviour similar to the one presented in Chapter 6.

This method creates insights into the problems addressed and therefore informs the development and virtual testing of policy options, although the method does not

8-8 provide optimal policy choices (Richardson and Pugh 1981). The model provides an opportunity for understanding the causal mechanisms underlying the perceived system, and it should not be used as a framework for statistically based inferences.

The model is developed keeping in the view the farmers and other persons interested in understanding the dryland salinity problem as the main users of the model.

8.3.2 Basic requirements for the model

Chapters 6 and 7 presented the qualitative analysis of the dryland salinity problem that presents a large number of variables that are important in understanding this problem. The simulation model takes a few key variables related to the land cover and to guide the model development, criteria was developed that is described below. The model must conform to the following expectation:

 The model should enable the development of a set of strategic views of the problems and should be able to be used in conjunction with the causal loop and influence diagrams, concepts maps and the reference modes.

 The model should help in understanding how the impacts of time delays are involved in land progression from different stages.

 The model should aid learning about the impacts of land use on dryland salinity.

 The model should provide policy levers for experimentation, that is, virtual testing of the efficacy of alternate strategies.

 As it stands, the model is a research tool that has emerged through a synergistic use of System Dynamics and Systems Engineering approaches in model development. The model has evolved from a basic generic module; however, additional detail has been added to aid analysis and subsequent learning. Further modules can be added should they be required in subsequent studies.

 The model should use the important variables that can directly influence dryland salinity, time frames and model boundary requirements.

8-9

8.3.3 Model structure

8.3.3.1 Basic building block: a generic module

The model structure was developed using a generic module specified by McLucas (2003) and defined by its boundary and functionality. According to McLucas (2003) the functionality of a module means the operations it performs on the inputs, e.g., accumulating, draining, etc. and the outputs from a module are either lost across a boundary or made available to another module.

This generic module consists of one stock and two flows. One flow is into the stock and accumulates stock and the other is out of the stock and drains it. Flows and their determinants are within a module boundary. Across the boundary, there are physical and information flows as well as datasets that provide it connectivity to other modules and its environment. The generic module is shown in Figure 8.1.

8-10 Physical Outflow Physical Inflow Export to Dataset Definition to include: Definition to include: Definition to include: flow type discrete or flow type discrete or write format;write continous;flow continous;flow direction;conversion direcion;max flow direcion;max flow factors;units of rate; dt;simulation rate; dt;simulation measurement. time step;simulation time step;simulation time horizon;units; time horizon;units; dimentions dimentions Constant_2

Rate_1Level Rate_2

Constant_1

Import from Dataset Definition to include: read format;read Information Inflow direction;conversion actors; units of Information Outflow Definition to include: measurement. Definition to include: sampling rate;dt; sampling rate;dt; simulation time ste p; simulation timestep; simulation time simulation time horizon;calendar. horizon;calendar.

Figure 8.1 Structure of a generic module. Redrawn from McLucas (2005:179)

System Dynamics requires a specific language to be used when dealing with the basic building blocks of the model. These building blocks are then used in different ways to model the relationships amongst them. These building blocks include:

 Level: also called stocks, accumulators and tubs, stocks are the conceptual representations that describe the state of the system at any particular time. The stocks accumulate as a result of actions within system at any time. It is a variable with a memory and it accumulates values as a bathtub collects the water running from the faucet (Powersim Software AS 2003; McLucas 2005).

Level makes a model dynamic. It accumulates flows going into it and subtracts flows going out of it. Levels give a snapshot view of reality their values inform us how the system is doing at any given point in time. If time suddenly stops, levels would remain and be observable and measurable.

8-11 Levels do not change instantly. Stocks characterise states of the system and generate the information upon which decisions and actions are based. Levels create delays by accumulating the difference between the inflow process and its outflow. By decoupling rates of flow, stocks are the sources of dis- equilibrium dynamics within systems (Sterman 2000). A level may seem to be changing instantaneously, but there is nonetheless a delay, no matter how small. Stocks are usually represented by a rectangle (Figure 8.1).

 Rate: also called flows, ‘rates’ are the variables that determine how fast the stocks change (Ford 1999). Thus, flows represent a system's activity and may depend on the values of the levels. Levels are increased or decreased only by their associated flows and flows may depend on other levels.

 Auxiliaries: are used to combine or reformulate information. These are the additional variables that help to define a system (Sterman 2000) and clearly communicate the model. These can be a part of a rate equation (or provide aid in the formulation of flow rate equations) to achieve a certain level of detail.

An auxiliary has no standard form. It is an algebraic computation of any combination of levels, flow rates, or other auxiliaries (Powersim Software AS 2003). Auxiliaries are used to model information and not the physical flow of goods, so they change with no delay, instantaneously. They can be inputs to flows, but never directly to levels, because flows are the only variables that change their associated levels. Levels, however, can be inputs to auxiliaries (Powersim Software AS 2003). In most System Dynamics software, a circle is used to represent auxiliary variables (Figure 8.1). The use of auxiliaries is critical to effective modelling (Sterman 2000) because:

 It can improve model communication. Ideally each equation should convey one main idea for clarity purposes (Sterman 2000). An auxiliary provides space in a model to show each equation or part of equation separately. These equations often contain the parametric values needed for numerical consistency of units used in the model.

 Auxiliaries provide flexibility for input controls: business rules can be changed without editing/disturbing many other equations.

8-12  Auxiliaries can provide unit consistency and demonstrate how the unit conversion takes place to achieve dimensional consistency.

 Auxiliaries can always be eliminated from a system and model be reduced to a set of stocks and flows (Sterman 2000).

 Constants: are those state variables that change so slowly that they are considered constant over the time horizon of interest in the model (Sterman 2000). A diamond represents these constants. A constant is defined by an initial value and maintains this value throughout the simulation, unless the user changes the value manually (Powersim Software AS 2003). For instance, in a one-year simulation, a company may have an essentially fixed workforce that can be represented as a constant auxiliary.

8.3.3.2 Project Settings

Project settings are used to specify time units for further use throughout calculations, output reports and the calendar settings for the entire model. There are three available calendars in the Powersim Studio:

 Bank calender: consists of 360 days with 51 weeks and three days. It has 90 days in a quarter.

 Fiscal: consists of 364 days with exactly 52 weeks.

 Gregorian: consists of 365 days with each 4th year a leap year with 366 days. A year has 52 weeks and 1 or 2 days.

These calendars differ in the number of days and the permissible time units. Selection of a calendar depends upon reporting requirements and the time unit to be used in the simulation. Table 8.1 presents the available calendars and time units in the Powersim Studio.

For this simulation the fiscal calendar was used as it provides years as a time unit. While this calendar would provide a day less in a leap year, it provides the flexibility of generating quarterly or annual reports. Depending upon the purpose and in line with the modelling philosophy used in this thesis, this one day difference in leap

8-13 years is not likely to affect the model behaviour as well as the understanding being generated about different variables important in the dryland salinity system.

Table 8.1 Available calendars and time units within PowersimTM Studio.

Unit Calender

Gregorian Bank Fiscal s (second) @_second @_second @_secon min (minute) 0@s+60s 0@s+60s 0@s+60s hr (hour) 0@s+60min 0@s+60min 0@s+60min da (day) 0@s+24hr 0@s+24hr 0@s+24hr wk (week) 0@s+7da 0@s+7da 0@s+7da mo (month) Undefined 0@s+30da Undefined qtr (quarter) Undefined 0@s+90da 0@s+91da yr (year) Undefined 0@s+360da 0@s+364da

8.3.3.3 Units of Measurement

Dimensional consistency is emphasised in System Dynamics literature. The software being used is designed to specifically force the modeller to ensure that the units used are compatible between variables and, thereby, consistent throughout the model. Table 8.2 presents the units of the main variables used in the model.

PowersimTM Studio’s system of units consists of local units as well as global units. The global units are available to all components of the model. The local units are defined only for specific components and are only available to those components.

8-14 Table 8.2 Units of measurement used in the model.

Variable Unit of Measure

Land under natural vegetation/forest/bush. Hectare (ha)

Rate of land clearing. Hectare/year (ha/yr)

Cleared land neither salt affected nor at the risk of Hectares (ha) becoming salt affected. Time delays. Year (yr)

Cleared land either salt affected or at the risk of Hectare (ha) becoming salt affected. Rate of land becoming salt affected. Hectare/year (ha/yr)

Rate of land at risk of becoming salt affected Hectare/year (ha/yr) returning to natural vegetation. Rate of land reclamation. Hectare (ha/yr)

8.3.3.4 Stocks and flows

The model presents three land stocks, land under natural vegetation, cleared land neither salt affected nor at the risk of becoming salt affected, and land either salt affected or at the risk of becoming salt affected. These stocks are linked by four flows, ‘rate of land clearing’, ‘rate of land becoming salt affected’, ‘rate of land reclamation’ and the ‘rate of land at risk of becoming salt affected that is returning to a natural vegetation cover’. A brief description of these stocks and flows follows.

The qualitative analysis (Chapters 6 and 7) identified a large number of variables. The modular approach adopted for this model encourages starting simple and then adding details as necessary.

Land stock 1: land under natural vegetation.

This stock represents land either bush or forest that has not been cleared for agricultural purposes under land clearing operations. This stock is represented in the model diagram by a rectangle. Land clearing rate drains it while rate of land

8-15 becoming salt affected returning to natural vegetation adds to this stock. This simple formulation is represented in Figure 8.2. A negative feedback loop manages the level stock. As the rate of land clearing increases, it decreases the stock of land under natural vegetation.

LAND UNDER NATURAL VEGETATION

+ +

Rate of land at risk - Rate of land of becoming salt clearing affected returning to natural vegetation

Initial land under natural vegetation

Figure 8.2 Sub-model land under natural vegetation

Land stock 2: cleared land neither salt affected nor at risk of becoming salt affected.

Cleared land represents land that was previously under natural vegetation either bush or forest and is cleared for the purposes of bringing it under agricultural production. The ‘rate of land clearing’ compounds this stock while the rates ‘rate of land becoming salt affected’ and ‘rate of reclamation’ drains it. There is a negative feedback loop that manages this stock. The rate reduces the stock (shown in dashed lines) and low stock causes a reduced ‘rate of land becoming salt affected’ as rate equation are formulated as a fraction of the stock. Rate formulations are discussed in the following sections. The inflows and out flows of this stock are presented in Figure 8.3.

8-16 Rate of land reclamation +

CLEARED LAND NEITHER SALT AFFECTED NOR AT RISK OF BECOMING SALT AFFECTED - + +

Rate of land - Rate land becoming Clearing Salt Affected

Initial cleared land not at risk of becoming salt affected

Figure 8.3 Sub-model cleared land neither salt affected nor at the risk of becoming salt affected. (Note: bold Arrows show direction of flow).

 Land stock 3: cleared land either salt affected or at the risk of becoming salt affected.

As it is clear from the stock name, this stock represents land that is already salt affected or is at the risk of becoming salt affected, i.e., water table is at or within two meters below the ground surface. This stock has one inflow that compound the stock and two outflow that drain this stock. A simple representation of the stock is presented in Figure 8.4. The inflow is ‘rate of land becoming salt affected’.

There are two outflows, ‘rate of land at risk of becoming salt affected returning to natural vegetation’ and the rate of land reclamation. There are two negative feedback loops that manage this stock (shown with broken lines in Figure 8.4).

8-17 Rate of land reclamation +

CLEARED LAND EITHER SALT AFFECTED OR AT RISK OF BECOMING SALT AFFECTED -

+ +

Rate land becoming - Rate of land at risk Salt Affected of becoming salt affected returning to natural vegetation

Initial land at risk of becoming salt affected

Figure 8.4 Sub-model land either salt affected or at the risk of becoming salt affected.

 Rate of land clearing

A sub-model providing the land clearing rate is shown in Figure 8.5. For this sub- model, rate of and clearing is defined as a fraction of the land under natural vegetation. The formulation is consistent with the formulation of rate equations described by Sterman (2000).

Rate of land clearing = land under natural vegetation * Fraction of land under natural vegetation/time delay in land clearing.

8-18 Rate of Land clearing

Fraction of land LAND UNDER under natural NATURAL VEGETATION vegetation being cleared

Random time delay in land clearing Time delay in land Random 1 clearing

Figure 8.5 Rate of land clearing

Note: the ‘picture corners’ symbols placed around the rectangle representing ‘LAND UNDER NATURAL VEGETATION’ stock indicates that this is an image of the original stock which appears on another diagram in the same model.

The fraction of land under natural vegetation being cleared is modelled on the basis of the historical data of land clearing developed from different references (Chapters 2, 6 and 7). The fraction is provided as a table function shown in Figure 8.6.

The fraction of land under natural vegetation that is being cleared is considered to be varying overtime. The input data is given through the following graph (Figure 8.6). The maximum rate is considered between 30-35% during the middle of the last century. Under current environmental pressures and data provided by the Australia Greenhouse Office (AGO 2000), it was considered that during the latter part of the last century, land clearing rates were started to decline.

Time delay in land clearing is a user defined variable and includes the time that is consumed in planning, land acquisition, getting permissions for land clearing, arrangements for the machinery, acquisition and movement of machinery and felling and export of logs from the area. As there may be varying times for different areas, land clearing operations, communities to check sensitivities, a random variable is used.

Time delay in land clearing provides for the time spent in planning for land clearing, getting approvals/permissions, accessibility to the area and finally clearing the land

8-19 of its natural vegetation either bush or forest. Time delay is a user controlled parameter whose default value is 10 years. As the actual time delay will vary over the simulation period, a random number has been used that fluctuates between 6 and 10 years. The following formulation for the Random 1 was used:

Random 1 = 1-(0.5*Random())

This formulation returns random numbers between 0 and 1, with a new seed used during each simulation run. The same formulation is used for random numbers in the model to ensure consistency. Time delays are shown in Figure 8.7.

0.35 ation

0.3

0.25

0.2

0.15 et veg natural under of land Fraction being Cleared

1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 8.6 Table function-land clearing fraction.

8-20 yr 10

9

8

7

6 Random Time Delay in Land Clearing

1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 8.7 Random time delay in land clearing.

 Rate of land becoming salt-affected

Rate of land becoming salt affected is depicted in Figure 8.8 and defined as:

Rate of land becoming salt affected = cleared land neither salt affected nor at the risk of becoming salt affected*Fraction of land not at the risk of becoming salt affected

Rate land becoming salt affected CLEARED LAND NEITHER SALT AFFECTED NOR AT RISK OF BECOMING SALT AFFECTED

Time delay in land Fraction of land not Random 2 becoming salt at risk of becoming affected salt affected that becomes at risk

Figure 8.8 Rate of land becoming salt affected

Fraction of cleared land that becomes salt affected is provided by history developed in the process of developing reference modes. The fraction near negligible at the start of the simulation, reaches a peak of 0.3 around 2000 and reduces to 0.1 near the end of the simulation period. As accurate estimates of the rates could not be obtained, a

8-21 random number given below, fluctuates the fraction around the points in the table function.

Random 2 = 1-(0.5*Random())

Figure 8.9 shows the behaviour of the fraction over the simulation period.

0.3

0.25

0.2

0.15

0.1

0.05 Fraction of land not at riskof becoming salt affected that risk at becomes

1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 8.9 Table function/graph of the fraction land becoming salt affected.

Time delay in a land becoming salt affected or at the risk of becoming salt affected is not actually known. It would vary according to land policies, specific geo-physical and social set-up and market forces. Time delay in land becoming salt affected is a user controlled parameter. The default value is a rough estimate of 30 to 40 years. A random variable (Random()) fluctuates this time delay between 0 and 40 years over the simulation period. The behaviour of the time delay is shown in Figure 8.10.

8-22 yr

40

30

20

Random time delay 10

0 1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 8.10 Time delay in cleared land becoming salt affected.

 Rate of land reclamation

In this model, the rate of reclamation has been defined as function of the fraction of salt affected Or at the risk of becoming salt affected on which a control treatment is applied, time delay and effectiveness of the control treatments. The model that provides the rate of land becoming salt affected is shown in Figure 8.11.

Rate of land reclamation

CLEARED LAND Effectiveness of EITHER SALT control treatments AFFECTED OR AT RISK OF BECOMING SALT AFFECTED

Fraction of land at risk of becoming Time delay between salt affected on land at risk of which control becoming salt treatment is applied affected and land not at risk of becoming salt affected under a certain control treatment

Figure 8.11 Rate of land reclamation

8-23 For simplicity, a single category of salt affected land or land at risk of becoming salt affected is used. However it is acknowledged, the process of land becoming salt affected is gradual. The actual statistics/number of salt affected land is not available. The National Land and Water Resources Audit (NLWRA 2001) used a parameter ‘salt affected land or land at risk of becoming salt affected’ that means a land that has water table within two meters below the ground surface. As stated before, this model uses the parameter that is consistent with the one used by the National Land and Water Resources Audit (NLWRA 2001).

The fraction of land either salt affected or at risk of becoming salt affected is a user controlled parameter. The default value is 0.1 which means a land control treatment is applied at 10% of the salt affected or at risk of becoming salt affected land. Effectiveness of a control treatment is a treatment specific parameter and depends upon the scientific studies of the relationship of a certain control treatment and the capability of that treatment to address dryland salinity. In the model, it is also a user- controlled parameter. The default value is 0.5 (50% effective). Various agronomic and engineering control treatments for dryland salinity are listed in Chapter 2.

Time delay in a piece of land going out of risk of becoming salt affected is also user controlled. The default value is kept at 30 years.

Rate of land reclamation=cleared land either salt affected or at the risk of becoming salt affected*friction of land at risk of becoming salt affected on which control treatment is applied*effectiveness of control treatments/time delay between land at risk of becoming salt affected and land not at risk of becoming salt affected under a certain control treatment.

 Rate of salt affected land returning to natural vegetation cover.

The model accommodates another pathway for the salt affected land, i e. land left out of agricultural operations. No further land reclamation control treatment is applied. Overtime, the unattended land starts to return to the bush/natural vegetation cover. The formulation that provides this rate is graphically shown as in Figure 8.12.

8-24 Rate of land at risk of becoming salt affected returning to natural vegetation

Time delay between CLEARED LAND land at risk of EITHER SALT AFFECTED OR AT RISK becoming salt affected and OF BECOMING SALT returning to natural AFFECTED Random fraction of vegetation land at risk of becoming salt affected that is returning to natural vegetation

Fraction of land at Random 4 risk of becoming salt affected that is returning to natural vegetation

Figure 8.12 Rate of land returning to natural vegetation cover

The rate of land returning to natural vegetation is defined as:

Rate of land returning to natural vegetation cover = cleared land either salt affected or at the risk of becoming salt affected*fraction of land at risk of becoming/time delay between land either salt affected or at the risk of becoming salt affected and returning to natural vegetation.

Both the fraction and the time delay are user controlled parameters. The default value is 0.05, i.e., 5% the actual fraction may vary over the period of simulation. A random number generator (Random()) fluctuates this fraction.

Time delay may vary based on a number of factors, e.g., location of a piece of land, type of vegetation and other geo-physical conditions. It is a user-controlled parameter whose default value, i.e., the maximum time a piece of land takes in returning to its natural vegetation cover is kept at 50 years.

8-25 8.3.3.5 Delays

All time delays in the model are considered as material delays. Although the symptoms of land becoming salt affected, e.g., reduction in crop yields, salt rust, surface appearance and change in vegetation cover may appear after a certain time, the processes involved in a piece of land becoming salt affected start in an early phase. The simple formulation consistent with Sterman (2000:416) models these delays:

Outflow = material/average delay time

8.3.4 Module integration

Integration means combining two or more modules into a functioning higher level

System Dynamics model. As at this stage various modules start to work as the components of the main model. McLucas (2005) suggests that the risks at the time of combining modules can be minimised by:

 ensuring that material is neither created nor destroyed at the interfaces between modules. This is usually done through a test called delta check that is discussed in detail in Chapter 9;

 ensuring consistency of the specified time increment (dt), simulation time step, simulation time horizon and units;

 maintaining the order in which calculations are to be conducted; and

 ensuring appropriate and consistent levels of aggregation.

Individual modules were prepared using consistent dt, time step, time horizon and units. Therefore modules were consistent and did not pose any specific problem at the time of integration. A stock and flow diagram of the strategic level model is shown in Figure 8.13.

8-26 Random Fraction of Fraction of land at land at risk o f risk of becoming salt becoming salt affected that is affected that is returning to natural returning to natural vegetation vegetation

Time delay between land at risk o f Time delay between land at risk of becoming salt Rate of land at risk of becoming salt affected and becoming salt affected and land not returning to natural affected returning to Fraction of land at at risk of becoming vegetation natural vegetation Mitigation strategy is salt affected under a applied mitigation strategy Effectiveness of Mitigation Strategy

CLEARED LAND EITHER SALT AFFECTED OR AT RISK OF BECOMING SALT CLEARED LAND NEITHER SALT AFFECTED NOR AT RISK OF Rate of land LAND UNDER NATURAL VEGETATION AFFECTED BECOMING SALT AFFECTED reclamation

Rate of Land Clearing Rate Land becoming Salt Affected Initial land under Random 2 natural vegetation Time delay in land Fraction of land not Random Time Delay Fraction of land Initial Cleared land becoming salt at risk of becoming in Land Clearing under natural not at risk of affected salt affected that vegation being becoming salt becomes at risk Initial land at risk o f Fraction of area at Total Area Cleared affected becoming salt affected risk of becoming salt Random 1 Total Area affected at start of Fraction of area Time Delay in Land Total Area simulation Fraction of area Clearing cleared at the start under natural of simulation vegetation at the start of simulation

Figure 8.13 Strategic framework - stock and flow diagram of the dryland salinity model

8-27

8.3.4.1 Simulation settings

Simulation settings specify time-frames, time step, integration method, number of runs, history settings and report settings for the model.

Timeframe for the simulation is 200 years, i.e., from 1900 to 2100. This timeframe includes the major periods of dryland salinisation, and accommodates time delays to allow remediation strategies to take effect. However, this timeframe is not the rigid one and has the flexibility to be changed to suit experimentation.

Time step is a value that specifies the time increment for each step of the simulation (Powersim 2003). General rule of thumb is to adjust time step according to delays. Time step for this simulation is kept at one year keeping in view the time delays in land clearing and growth plants, if used in a plant based management of dryland salinity.

There are several methods, or algorithms for performing the integration during a simulation. The integration method is used for adding the sum of flow rates to the current value of their connected levels at each time step during a simulation (Powersim AS 2003). Sterman (2000:911) suggests the following guidelines for numerical integration:

 Select a time step for your model that is a power of 2, such as 2,1, 0.5, 0.25 etc.

 Make sure your time step is evenly divisible into the interval between data points or other periodic exogenous events.

 Select a time step one fourth to one tenth as large as the smallest time constants in your model.

 Test for integration error by cutting the time step into half and running the model again. If there are no significant differences (judged relative to your purpose), then the original value is fine. If the behaviour changes continue to cut the time step in half until the differences in behaviour no longer matter.

8-28

 Note that Euler integration is almost always fine in social and human systems where there are large errors in parameters, initial conditions, historical data, and especially model structure. Test the robustness of your results to Euler by running the method with high-order methods such as fourth order Runge- Kutta, if there are no significant differences, Euler is fine.

Both time step and integration methods are very easy to change in PowersimTM Studio. This utility of PowersimTM Studio helps in experimentation with the model as well as in determining the appropriate values during model development.

8.3.4.2 Data input and output

The model provides users with flexibility to control inputs. Slider bars are provided in the input control window and the inputs that can be changed are initial values for stocks, land fractions, time delays and effectiveness of control treatments. With these flexibilities, the model can be adjusted for use for any river basin up to 1.5 million hectares in size. Land fractions can be from 0 to 1.0. Time delay input sliders provide flexibility to include time delays up to 100 years. Effectiveness of control treatment is on a scale of 0 to 1.0. Figure 8.15 shows the input control window for the model.

A graph control provides a window where output from various stocks can be compared with each other. An output window is shown in Figure 8.14 where a green line shows the area under natural vegetation, the brown line shows cleared land while the red line shows the area that is becoming salt affected.

8-29

ha

80,000,000

Cleared land

60,000,000

Area either salt affected or at the risk of becoming salt 40,000,000 affected

20,000,000

Area under natural vegetation

1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 8.14 Graph window

8-30

Initial Values: these are the values of the stocks that are used at the start of simulation period Time Delays: this is time consumed in a process or change of state

Total Area T ime Delay in Land Clearing

0 200,000 400,000 600,000 800,000 1,000,000 1,200,000 1,400,000 0 1020304050 sq km yr Non-commercial use onl y! Non-commercial use onl y! Fraction of area under natural vegetation at the start of simulation Time Delay between non saline land becoming salt affected

Initial 0 102030405060708090100 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 yr Non-commercial use onl y! Non-commercial use onl y! Fraction of area cleared at the start of simulation T ime delay in land becomingInitial salt affected

0 102030405060708090100 0 0.10.20.30.40.50.60.70.80.91 yr Non-commercial use onl y! Non-commercial use onl y! Fraction of area at risk of becoming salt affected at start of simulation Time delay between land at risk of becoming salt and becoming land not at risk under a certain control treatment

0 0.10.20.30.40.50.60.70.80.91 0 102030405060708090100 yr Non-commercial use onl y! Non-commercial use onl y! T ime delay between land at risk of becoming salt affected and returning to natural vegetation

0 102030405060708090100 yr Land fraction: this is fraction of a certain category of land Non-commercial use onl y!

Fraction of land at risk of becoming salt affected that is returning to natural vegetation Effectiveness of Control Treatment Initial 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Non-commercial use onl y! Fraction of land at risk of becoming salt affected on which control treatment is Effectiveness of ControlInitial treatments applied

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Non-commercial use onl y! Non-commercial use only!

Figure 8.15 Inputs control window

8-31

The model exhibits the behaviour shown in Figure 8.14 when given the default values (Figure 8.15). Figure 8.14 indicates that land under natural vegetation decreases and starts to stabilise in the last decades of the last century. One reason for this decrease is the large-scale land clearing for agricultural uses as documented in Chapters 2 and 7. The causes for the stabilisation after 1980 are two-fold. First, the extensive land clearing in the first half of the century reduced the land available for clearing and reduced the rate of increase in land clearing. Secondly, the impacts of land clearing started to be realised and the Commonwealth and State Governments undertook steps to discourage land clearing. The behaviour is close to the reference modes depicted in Chapter 7.

With the decrease in natural vegetation, cleared land that is neither salt affected nor at the risk of becoming salt affected (Figure 8.14, blue curve) increases. After a certain time delay, the cleared land started to become salt affected (Figure 8.14, red curve) near the 1950s and steadily increases until 2050, after which, it starts to stabilise until 2100. The increase in cleared land from 1900 to 1950 was due to increased land becoming cleared as a result of an increase in demand for land clearing. Around the 1950s, as the cleared land became salt affected or at the risk of becoming salt affected, the inventory of cleared land decreased. Between 2000 and 2100, the cleared land is likely to be influenced by two factors: a decrease in the rates of land clearing and an increase in the rates of salinisation. This causes a decrease in the cleared land stock. This relationship is depicted by the reference modes showing the cleared land and land under forest in Figure 8.14.

8-32

8.3.4.3 Initial values

Initial values are the values of stocks that are used at the start of a simulation. The sub-models that provide initial values are shown in Figures 8.16, 8.17 and 8.18. The initial values of the stocks are user controlled and provide a room for experimentation. Initial values of stocks have two components: total area of the Basin and the fraction of that area pertaining to that particular stock.

The initial value at start of the simulation for the land under natural vegetation is the percentage of the area of the Murray Darling Basin that was uncleared at the start of the last century. The simulation considers that 90% of the area was under natural vegetation at the start of the century. This functionality is provided by the auxiliary ‘initial land under natural vegetation’ (Figure 8.16).

Initial land under natural vegetation = total area*fraction area under natural vegetation. Where total area is the geographical area of the Basin, i.e., 1,061,469 square kilometres, and the fraction of area under natural vegetation is any number from 0 to 1 the fraction used as the default value is 0.9. Both total area and the fractions provide the user with flexibility to change the number.

8-33

LAND UNDER NATURAL VEGETATION

Initial land under natural vegetation 955,322.10 sq km

1,061,469.00 sq km

Total area 0.90

Fraction of area under natural vegetation at the start of simulation

Figure 8.16 Sub-model providing initial values for the land under natural vegetation.

The initial value for the land that is cleared but is neither salt affected nor at the risk of becoming salt affected, is provided by the auxiliary ‘initial cleared land not at risk of becoming salt affected. This auxiliary is defined as:

Initial cleared land not at the risk of becoming salt affected = Total area*fraction cleared land where the fraction cleared is any number between 0 and 1. The default value is 0.09 that corresponds to the 9% cleared area in 1900, i.e., at start of the simulation (Figure 8.16)

8-34

CLEARED LAND NEITHER SALT AFFECTED NOR AT RISK OF BECOMING SALT AFFECTED

5,307,345.00 ha

5,307,345.00 ha

Initial cleared land not at risk of 1,061,469.00 sq km becoming salt affected 0.05 Total area Fraction of area cleared at the start of simulation

Figure 8.17 Sub-model that provides initial values to the cleared land neither salt affected nor at risk of becoming salt affected.

Initial value for the land either salt affected or at the risk of becoming salt affected is provided by an auxiliary ‘initial land at risk of becoming salt affected’ that is in turn a product of the geographical area of the Basin and fraction of the land at risk of becoming salt affected. The fraction can be any number from 0 to 1. The default value for the sub-model is 0.01 which means that at start of the simulation 1% of the land are either salt affected or at the risk of becoming salt affected.

8-35

CLEARED LAND EITHER SALT AFFECTED OR AT RISK OF BECOMING SALT AFFECTED

Initial land at risk of 1,061,469.00 ha becoming salt affected

1,061,469.00 sq km 0.01

Total area Fraction of area at risk of becoming salt affected at start of simulation

Figure 8.18 Sub-model providing initial values to the land either salt affected or at the risk of becoming salt affected.

Initial values can be changed through input control window.

8.4 A DETAILED FRAMEWORK

The strategic framework, described in Section 8.3, is designed for basin scale information where there may be an infinite number of catchments that have physical and biological processes. It is user oriented and instead of inscribed/build-in input values (e.g. fractions, parameters, constants, etc.) it provides users with opportunity to try various combinations of policies. At the high level of abstraction and intended uses of simulation modelling as described in Section 8.5, the model structure is kept simple and easy to understand and experiment with.

At the detailed level, there are numerous biophysical processes influencing dryland salinity that may be considered while developing a model at the catchment scale. The inclusion of such processes depends upon of the purpose of the model and level of focus on the problem. Some of the processes are described in the causal loop

8-36

diagrams presented in Chapter 7. A large number of models of the biophysical processes do exist, as discussed in Chapter 3, an example is a model presented by Saysel and Barlas (2001) that was prepared for irrigated areas. This model describes salinity and water table levels. Details of the model has already been discussed in Chapter 3. Another example is provided by Khan et al. (2009) for a farm based hydrological system.

Despite that Saysel and Barlas’s (2001) model was developed for the irrigated areas in the south-eastern Anatolian region, some of the general model relationships can be useful for gaining insights into the dryland salinity problem in Australia. Figure 8.19 present two modules in which such relations are used. The first module describes changes in the depth to water table as a result of an increase due to deep percolation and decrease dues to groundwater discharge and intrusion. Other factors like soil porosity and drainage also influence these rates. The second module presents salinity of the surface soil or the plant-root zone. This salinity level changes as a result of root zone salinity increase and decrease rates. The rate of salinity increase is influenced by the salinity of evapo-transpiring water and the quantity of evapo- transpiring water. The salinity decrease rate depends on the quantity of infiltration water and the salinity of infiltration water. Both salinity increase and decrease rates are influenced by soil porosity and root zone depth. The definitions of the variables are in given in the Annex-A Glossary.

These modules can provide input to the strategic model described above. Figure 8.20 presents a framework of the stock and flow diagram in which these modules can provide input to the strategic model. Within this framework both land surface and the water table models can be used to provide conditions to model for determining the rate of land becoming salt affected or at the risk of becoming salt affected.

As the thesis focus is on a strategic level, further development and full validation of quantified physical models is out of the scope of this thesis and is, therefore, recommended for future research.

8-37

Groundwater discharge fraction

Critical discharge porosity below level rootzone Watertable decrease discharge Discharge level discrepancy

Infiltration percentage Depth to watertable Depth to Watertable Depth to increase Watertable decrease intrusion

Percolation INI Depth to Watertable Watertable discrepancy

Infiltration Critical watertable level Drainage

Groundwater Precipitation Drainage efficiency intruding rootzone Intruding groundwater fraction Effective precipitation

Effective precipitation percentage

Rootzone depth

INI Root zone salinity

Rate of rootzone Rate of rootzone salinity increase surface soil or rootzone salinity salinity decrease Fraction salinity Salinity decrease increase due to due to salts taken salts deposited by up by plants wind evapotranspiring water Rootzone porosity

Salinity of infiltration water Intruding groundwater ratio salinity of Salinity of Adsorption fraction evapotranspiring groundwater water Initial salinity of groundwater Precipitation ratio Salinity of Precipitation

Figure 8.19 Physical process involved in dryland salinity – depth to water table and surface/root zone salinity (After Saysel and Barlas 2001)

8-38

Random Fraction of Fraction of land at land at risk o f risk of becoming salt becoming salt affected that is affected that is returning to natural returning to natural vegetation vegetation

Time delay between land at risk o f Time delay between land at risk of becoming salt Rate of land at risk of becoming salt affected and becoming salt affected and land not returning to natural affected returning to Fraction of land at at risk of becoming vegetation natural vegetation Mitigation strategy is salt affected under a applied mitigation strategy Effectiveness of Mitigation Strategy

CLEARED LAND EITHER SALT AFFECTED OR AT RISK OF BECOMING SALT CLEARED LAND NEITHER SALT AFFECTED NOR AT RISK OF Rate of land LAND UNDER NATURAL VEGETATION AFFECTED BECOMING SALT AFFECTED reclamation

Rate of Land Clearing Rate Land becoming Salt Affected Initial land under Random 2 natural vegetation Time delay in land Fraction of land not Random Time Delay Fraction of land becoming salt Initial Cleared land RZS or WTL at risk of becoming in Land Clearing under natural not at risk of affected salt affected that vegation being becoming salt becomes at risk Initial land at risk o f Fraction of area at Total Area Cleared affected becoming salt affected risk of becoming salt Random 1 Water table level Rootz one salinity Total Area affected at start of Fraction of area switch switch Time Delay in Land Total Area simulation Fraction of area Clearing cleared at the start under natural of simulation vegetation at the start of simulation Watertable level Rootz one salinity

Figure 8.20 A detailed framework - stock and flow diagram including strategic level model and physical components

8-39

8.5 USE OF THE SIMULATION MODELLING IN DRYLAND SALINITY STRATEGIC MANAGEMENT 8.5.1 Strategy development

Scenario planning is a useful tool for strategic thinking (Schoemaker, 1995; Schoemaker et al., 1992; Schoemaker 1991). Wolstenhome and Stevenson (1994:22) suggest systems thinking and modelling provide a unique framework for developing a balanced view between strategy, organisation structure and business processes. Schoemaker 1995: 25) further elaborates the utility of scenario planning:

Among many tools a manager can use for strategic planning, scenario planning stands out for its ability to capture a whole range of possibilities in rich detail. By identifying basic trends and uncertainties, a manager can construct a series of scenarios that will help to compensate for the usual errors in decision making – overconfidence and tunnel vision.

A large number of research articles are available (for example, Schoemaker 1991; Schoemaker and Van der Heijden 1992; Schoemaker 1993; Schoemaker 1995; Schnaars and Ziamou 2001) to provide guidance on development and write-up of scenarios; length of a scenario and the span and the number of scenarios that are useful in strategy development. Development or (write-up) of scenarios poses special challenges that are highlighted and addressed by Schnaars and Ziamou (2001). To elaborate on such challenges Schnaars and Ziamou (2001) suggest:

Scenario writing is an idiosyncratic practice. Just as fiction writers are free to write whatever type of novel they care to, so too are scenario writers free to propose virtually any procedure they wish and call it “scenario analysis”. There simply is no single standard for writing good scenarios. Furthermore, the softer almost intuitive nature of writing scenario writing makes it difficult to state explicitly just what steps should be followed to create acceptable scenarios. Instead many expositions are loaded with flowery and vague assertions that are difficult to practice.

8-40

The challenges increase in terms of development of scenarios for dynamic problems. Simulation models help in such situations as they provide room for iterative development of scenarios and experimentation with the model. The System Dynamics dryland salinity model described in the preceding sections has an easy to use data input tool shown in Figure 8.15. This is a conceptual model that can be further developed and subsequently used in iterative development and testing of various scenarios of remedial measures. It provides input controls that a user can change, conduct experiments with the model and observe the impacts of those changes. The whole process of model development and its use to answer what-if scenario questions focuses on fostering leaning.

Figures 8.21 to 8.24 provide an example of such use of the model. The following four scenarios were developed:

 Business as usual, i.e., without any further information.

 90% of the mitigation strategies/control treatment is applied over 50% of the area.

 45% of the control treatment is applied over 50% of the area.

 45% effective control treatment is applied over 50% of the land at risk of becoming salt affected and 20% of the land returning to natural vegetation.

These scenarios were tested using the simulation model. These scenarios help in responding to ‘what if’ questions using different combinations of the control treatments, area on which control treatments are applied and effectiveness of the control treatments.

Such use of a model can help decision makers in iterative development of ‘what-if’ questions and then answering such questions and progressing their thinking in identifying strategies. To demonstrate the model building process using Systems Engineering processes, the model was kept simple; however, further sectors/subsystems can be added based on the specific purpose and the direction of policy development. The simple structure of the model provides opportunities for the decision makers to change/amend the model to suit the purpose of their analysis.

8-41

However, Repenning (2003) and Gary et al. (2008) emphasise the benefits of a simple System Dynamics model and theories in communicating key insights.

ha

80,000,000 Cleared land

60,000,000

40,000,000 Land either salt affected or at the risk of becoming salt 20,000,000 affected

Land under natural vegetation

1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 8.21 Business as usual (BAU) scenario, i.e., without any further intervention

ha Cleared land

80,000,000

60,000,000

40,000,000 Land either salt affected or at the risk of becoming salt affected 20,000,000

Land under natural vegetation

1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 8.22 Scenario: 90% effective control treatment is applied on 50% of the area

8-42

ha

80,000,000 Cleared land

60,000,000

40,000,000 Area either salt affected or at the risk of becoming salt

20,000,000 affected

Land under natural vegetation

1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 8.23 45% effective treatment applied over 50% of area.

ha

Cleared land 80,000,000

60,000,000

40,000,000 Area either salt affected or at the risk of becoming salt 20,000,000 affected Land under natural vegetation

1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 8.24 45% effective treatment is applied over 50% of the land at risk of becoming salt affected and 20% of the land returning to natural vegetation.

8-43

8.5.2 Training of decision makers.

The model can be further developed into modules for use in training of decision makers in understanding the impacts of non-linearities, time delays and feedback loops. Management Flight Simulators, also known as Microworlds or Virtual Worlds are an example of such tools. These are the formal models or simulations in which decision makers can refresh decision making skills, conduct experiments and play (Sterman 2000:34). Such tools can take the form of role plays or interactive games. In dynamically complex problems, Sterman (2000:34-35) suggests the use of computer simulations.

The dryland salinity model has a user friendly data input window as shown in Figure 8.15. It can be further improved to include the functionality required for interactive experimentation. Such a management flight simulator can be a low cost laboratory for learning through experimentation. In this chapter, the process of developing a module using a structured process was demonstrated. This process can be used to develop other required modules.

8.5.3 Learning at the grass-root level about a real world problem

System Dynamics like other academic disciplines feel pressures that can move it away from the real world issues. Forrester (2007:360) observes:

System dynamics started 50 years ago with academic programs that focused on the outside world with emphasis on major issues outside of academia. However, the pressures inherent in academic institutions are driving our field back into academic journals and away from the public that we should be serving. This scenario leading to reduced relevance has been travelled ahead of us by the field of operations research. Operations research grew out of real problems of the military in World War II. Based on such real-world successes, classes and research programs were launched in universities and a journal was started. Criticism soon arose about how real-world relevance was being lost to the pressure to write for academic journals. As a result, a new society,

8-44

the Institute of Management Sciences, was formed as the applications branch of the field. However, the same thing happened the supposedly practical branch reverted to writing for academia and now the two societies are effectively merged. System Dynamics is threatened by these academic pressures to retreat from major real-world issues.

To keep System Dynamics focused on the real world issues, to keep it in the public domain, and provide opportunities for learner centred learning, System Dynamics studies programs have been suggested from Kindergarten through Grade 12 education since 1991 (for example, see Forrester 1992).

For increasing public understanding of the systems, Forrester (1991:27) suggests that we should seek general insights and make connections to where the same themes have already appeared.

The dryland salinity model presented in the preceding pages demonstrates dryland salinity dynamics with a simple model with the stocks and uses the terminology consistent with the literature and the public understanding of such terminology. The model development process also demonstrates a structured way of developing such a model. If used for learning at early stages of problem conceptualisation in the public domain or for Kindergarten through Grade 12 education, this model can help in building understanding of the changes in the dryland salinity problem as linked to the variety of control treatments and varying levels of their effectiveness; decisions about the selection of control treatments, land use/land cover and long-time delays in land use/land cover change. Such use of the model can foster learning.

Whether it is for the purpose of enriching executives’ knowledge for dryland salinity policy making or farmers or Kingergarten to Grade 12 education, such learning cannot only be gained through playing with the already built models but as Sterman (2000: 43) suggests:

In practice effective learning from models occurs best—perhaps only— when the decision makers participate actively in the development of the model.

8-45

8.6 SUMMARY AND CONCLUSION

This chapter discussed the rationale for using simulations in System Dynamics, explained the need for a modular approach for developing System Dynamics models, and described modules of a simple System Dynamics simulation model for dryland salinity. The structure of the model was represented using stock and flow diagrams.

The process of model development was started from the problem description (Chapter 2) and proceeded through qualitative analysis (Chapter 6), reference modes development using learning cycles approach (Chapter 7) and development of a System Dynamics simulation model. The preceding chapters demonstrated progressive problem analysis. The qualitative models were rich in identifying variables, interactions among variables, the flow of influences and feedback loops. As the process moved from qualitative analysis to simulation model development, the numbers of variables decreased. For, example, the concept map (Chapter 6) has more variables than the model described in this chapter. This position is also supported by Wolstenholme (1990:48) who suggest that as the modelling moves from hard to softer areas of the system spectrum, quantification becomes more difficult.

Simulation plays an important role in addressing the dynamically complex problems. It speeds up learning and provides an avenue for multiple hypotheses building and testing through experimentation and can help in visualising the outcome of mental models. Both qualitative maps and simulation models used synergistically can improve each other in the iterative cycles.

A simple model of dryland salinity was developed using a modular approach. The model consists of three stocks progressing from the land under natural vegetation that becomes cleared for agricultural purposes and subsequently becomes salinised and abandoned from agriculture due to high salinity levels. The relationships of the model were used from the literature review and qualitative analysis described in Chapters 2, 3, 6 and 7.

8-46

The model developed a parameter ‘effectiveness of control treatments’ that is specific to the type of mitigation strategies/control treatment applied. It helps in conveying the concept that all mitigation strategies/control measures are not equally effective and should be treated in the model corresponding to their effectiveness in addressing dryland salinity. The value-add of the simulation model development and experimentation includes the opportunity for developing alternative hypothesis (within the model boundaries) using different combinations of effectiveness of control treatment, area on which such treatments are applied and the time delays between application of a treatment and the results of that application. Such experimentation could not be performed with qualitative models.

The model can be further developed, calibrated and used in identifying strategies that respond to the natural time delays involved in the dryland salinisation processes. The model can also be used to evaluate the effectiveness of mitigation strategies or control measures, for the study of effects of delays in groundwater response, natural land recovery, and reclamation efforts on the respective stocks. The model can also provide a tool for communicating these impacts between researchers, policy groups and farming communities.

The use of a model is linked to the confidence that developers and users of the model have in the value of that model. This leads discussion to verification and validation of the model that is the subject of next chapter in which issues in validation of System Dynamics models are discussed and the approach applied for validation of the dryland salinity model is presented.

8-47 CHAPTER 9 VERIFICATION AND VALIDATION OF SYSTEM DYNAMICS MODELS

No model has ever been or ever will be thoroughly validated…‘Useful’, illuminating, or ‘inspiring confidence’ are more apt descriptors applying to models than ‘valid’. (Greenberger et al., 1976)

The validity and usefulness of dynamic models should be judged, not against an imaginary perfection, but in comparison with the mental and descriptive models which we would otherwise use. (Forrester 1968:3)

As a general rule, we cannot create models that will accurately predict outcomes of complex systems. This awareness has profound impacts for Systems Engineering efforts, specially the many efforts that involve humans and organisations as essential components of the ‘system’. (Sage 2001)

The terminology used and the processes developed for building confidence in a certain model are shaped by:

 the meaning that modellers and model users attribute to the terms verification and validation;  the information and cognitive limits that bound the rationality of the modeller for gaining perceptions about the real world and selecting model parameters; and  beliefs about real world (ontology) and beliefs about accumulated knowledge (epistemology).

Kleindorfer and O’Neil et al. (1998:1096) consider objectivism and relativism fuel the debate on validation of simulation models. Objectivism was advocated by Naylor and Finger (1967) while relativism was advocated by Forrester (1968), Sterman (2000) and Barlas and Carpenter (1990). Kleindorfer and O’Neil et al. (1998:1098) suggest that both relativism and objectivism deny openness in which one can conduct meaningful dialogue. Kleindorfer and O’Neil et al. (1998:1098) suggest a validation environment in which models can be compared to each other, model builders are free

9-1 to increase credibility of their models and the stakeholders are meaningfully engaged.

The modelling process involves a number of decisions such as selection of model components, parameters and their values during model conceptualisation and quantification processes and the selection of model evaluation and validation criteria. Such decisions are made within bounded rationality. Bounded rationality implies that the rationality of individuals (modeller and model users in this case) is limited by the available information, cognitive limitations, and the time to make decisions. Simon (1957:198) articulates the concept of bounded rationality as:

The capacity of human mind for formulating and solving complex problems is very small compared with the size of the problem whose solution is required for objectively rational behaviour in the real world or even for a reasonable approximation to such objective rationality.

Daniel Kahneman (1934- ) in his Nobel Prize lecture (Kahneman 2003) suggested that some concepts come to mind easily whereas others might take some efforts while some ideas come to mind at a particular time. Kahneman (2003:699) further suggested:

Absent a system that reliably generates canonical representations; intuitive decisions are shaped by the factors that determine the accessibility of different features of the situation. Highly accessible features influence decisions, whereas features of low accessibility are largely ignored. Unfortunately, there is no reason to believe that the most accessible features are also the most relevant to a good decision.

Model validation had been a concern for System Dynamicists since the inception of System Dynamics, though there appears to be substantial efforts needed (Barlas and Carpenter 1990; Barlas 1996). There is an ongoing debate on model validity (Barlas 1989; Kleijnen 1995; Coyle and Exelby 2000) and such debate is evident from the work of Forrester (1968, 2007), Richardson and Pugh (1981), Sterman (2000), Saeed (1992) and McLucas (2005). In order to understand the foundations underlying this

9-2 debate, it is imperative to understand the System Dynamics world view of model validation.

This Chapter presents a review of validation approaches generally used for System Dynamics models and critically discusses the efficacy and fidelity of such approaches. The discussion builds on the concepts used in Systems Engineering for verification and validation of designed systems to inform the modelling process in System Dynamics. A case study is then presented demonstrating the synergistic use of System Dynamics and Systems Engineering approaches for verification and validation of the dryland salinity model described in Chapter 8. The lessons learned from the case study are discussed.

9.1 MODEL VALIDATION IN SYSTEM DYNAMICS

In fact, verification and validation of models is impossible. The word ‘verify’ derives from the Latin versus – truth; Webster’s defines ‘verify’ as ‘to establish the truth, accuracy, or reality of’. ‘Valid’ is defined as ‘having a conclusion correctly derived from premises…Valid implies being supported by objective truth. By these definitions, no model can ever be verified or validated. Why? Because all models are wrong. All models, mental or formal are limited, simplified representations of the real world. They differ from reality in ways large and small, infinite in number. (Sterman 2000:846)

Sterman’s (2000) above statement describes the foundation stone of the System Dynamics approach to model validation. This approach is deeply ingrained in bounded rationality, relativist paradigm and utility of models. Such views are evident from the seminal work of Jay Wright Forrester in “Principles of Systems” (Forrester 1968), G.P. Richardson in “Introduction to System Dynamics Modelling with Dynamo” (Richardson and Pugh 1981), John Sterman’s “Business Dynamics” (Sterman 2000) and Barlas and Carpenter (1990). Sterman (2000) further elaborates on the causes of impossibility of the verification and validation of models as:

The impossibility of validation and verification is not limited to computer models. Any theory that refers to the world relies on imperfectly measured data, abstractions, aggregations, and simplifications, whether the theory is

9-3 embodied in a large scale computer model, consists of simple equations, or is entirely literary. The differences between analytic theories and computer simulations are differences of degree only. (Sterman 2000:847)

Sterman (2000) also considers the efforts to prove the model right has detrimental effects on learning that is obvious from the following excerpt:

Unfortunately, testing is often done to prove that the model is right, an approach that makes learning difficult and ultimately erodes the utility of the model and the credibility of the modeller. (Sterman 2000:845-846)

System Dynamicists do not claim their models to be a perfect representation of reality rather a System Dynamics model is considered an abstraction from reality (Forester 1968:3-1). Therefore, model validity is considered as a relative matter that is strongly linked to the purpose of the model and the ontological position of the analyst. Forrester (1968) suggests a relativist approach for model justification and emphasises validation mechanisms that focus on clarity of structure, exposure of underlying assumptions, ease of communication and time varying consequences of the statements.

Within System Dynamics there is a strong appreciation of bounded rationality. Forrester (1968) also suggests utility of a model as an important criterion in judging the quality of a model. Such beliefs within System Dynamics are evident from the following excerpts from “Principles of Systems”:

…the validity and usefulness of dynamic models should be judged, not against an imaginary perfection, but in comparison with the mental and descriptive models which we would otherwise use. (Forrester 1968:3-3)

There is nothing in either the physical or social sciences about which we have perfect information. We can never prove that any model is an exact representation of ‘reality’. Conversely among those things of which we are aware of, there is nothing of which we know absolutely nothing. So we always deal with information which is of intermediate quality–it is better than nothing and short of perfection. (Forrester 1968:3-4)

9-4

The representation [model] need not be defended as perfect, but only that it clarifies thought, captures and records what we do know, and allows us to see the consequences of our assumption, whether those assumptions be perceived as ultimately right or wrong. A model is successful if it opens the road to improving the accuracy with which we can represent reality. (Forrester 1968 3-5)

The purpose of validation of a System Dynamics model is to improve confidence in the model as it fulfils the purpose for which it was built. System Dynamics models are usually intended to design policies to solve problems or improve performance and/or analyse contingencies for which experience does not provide further guidance (Sterman 2000, Barlas and Carpenter 1990).

Sterman (2000) suggests that a fundamental yardstick in the model validation process is the perspective that System Dynamics does not attempt to model a system but it attempts to model a problem. Therefore the model only presents variables that are important for understanding the dynamics that gives rise to a particular problem.

Yaman Barlas, in his seminal work on model validation, analysed System Dynamics model validation processes from the perspective of philosophy of science (Barlas and Carpenter 1987; Barlas 1989; Barlas 1989; Barlas and Carpenter 1990; Barlas 1994; Barlas 1996), and warned the critics who fault System Dynamics for being unscientific because its validation procedures are not sufficiently objective, formal or quantitative. After an analysis of philosophy of science, he suggested that the traditional (logical empiricist) view of scientific objectivity and formalism is not the only and unchallenged philosophy of science. Rather, there exists an alternate, widely held philosophy of science that is in agreement with how System Dynamicists view model validity (Barlas and Carpenter 1987), i.e., relativist philosophy of science.

System Dynamics literature suggest a few emerging approaches to model validation. Below is an overview of such approaches.

9-5 9.1.1 Model simplification as a validation strategy

Saysel and Barlas (2004, 2006) suggest model simplification as a final step of model development. They also suggest that among System Dynamists, there is a tendency to build large models. This tendency has been developed in response to criticism on small models. The reasons for building complex models are evident from the following excerpt from the presidential address of Yaman Barlas to the System Dynamics Society Conference 1998:

The ultimate set of insights offered by a System Dynamicist about a dynamic feedback problem is in general much larger than the results directly "produced" by the model. The analyst bases her conclusions on a much richer database of dynamic intuition, acquired as a result of years of experience with many other models. Through the exercise of building and analysing many models, the analyst acquires a rich set of dynamic analysis skills. This is not to undermine the importance of models; it is just that the "last" model cannot by itself do the entire job. That is probably why to most critics a given model typically looks "unrealistic" (i.e. lacking many intangible factors that they believe are important in addressing the problem). This type of criticism may tempt the modeler to make the model larger, more and more detailed. But this path often makes the situation worse rather than resolving it as new additions to the model will still not satisfy the critic and the model now becomes large, complicated and unrealistic.

Eberlein (1989) presents a formal theory of model simplification as a means of increasing model understanding, which identifies important feedback loops in linearized models with respect to a selected dynamic behaviour. Weak feedbacks that generate this behaviour and the stock variables embedded in these loops are eliminated and the original model is collapsed to a sub-structure that can create the intended dynamic behaviour. Saysel and Barlas (1994) consider limited applicability of such simplification to the System Dynamics models as Eberlain’s (1989) method is restricted to linearized models.

9-6 Saysal and Barlas (2004, 2006) also suggest that simplification typically involves aggregation of parameters, formulations and stock-flow processes, which requires informal reasoning beyond formal methods.

Beyond increasing the quality and understanding of the existing models, the commitment of the System Dynamics field to the idea of creating integrative theories of seemingly separate, case-specific management problems motivates simplification practice. Through simplification, a case-specific, large and parameterised model of a dynamic problem can be reduced to a generic representation of the same problem, suitable for transferring knowledge in the same domain and useful for disseminating the essential structures responsible for the problematic behaviour and mismanagement.

9.1.2 Scientific modelling

Homer (1996:3) advocates scientific modelling for bringing rigour to System Dynamics modelling and describes scientific modelling as:

In scientific (or “refutationist”) modelling the evaluation is done carefully and in depth, with a critical eye and an insistence on empirical evidence (Bell and Bell 1980). A concerted effort is made to gather historical data on as many relevant variables as possible for use as reference behaviors. In regard to the gathering of evidence in general, the modeller delves deeply into all relevant storehouses of data and experience, be they numerical, written, or mental (Forrester 1980). The initial model is based on accepted concepts and relationships, rather than speculations that run far afield from what is known. All discrepancies between model and evidence are investigated and their causes isolated to determine whether the model can not only reproduce history, but also do so for the right reasons.

For justification of the System Dynamics method on scientific grounds Oliva (2001) favoured the popper’s tradition of model falsification. He considered model as a laboratory and the design of experiments as a falsification exercise to test dynamic hypothesis. About testing of a dynamic hypothesis, Oliva (2001) states:

9-7 For any given dynamic hypothesis (DH), there will be many rivalry hypotheses - other structural explanations that might be capable of generating the observed behaviour. Clearly, no single test will be able to rule out all the (other) possible alternative explanations.

Oreskes et al. (1994) suggest that it seems impossible to verify a hypothesis, yet science has refined a systematic approach to increasing the confidence in a stated hypothesis, i.e., subject your assumption to tough tests rather than soft ones (Bunge, 1967, Oliva 2001). Oliva (2001) considers that the essence of the scientific method is captured in its ethos for experimentation, i.e., strive to reject the hypothesis. To gain a justified confidence in the causal argument stated in a dynamic hypothesis, the testing procedure has to be based on experimentation that can yield falsificatory implications (Oliva 2001). About experiments designed for falsification of a theory, Sterman (2000:848) writes:

Tests of any theory take place at a particular time, with particular equipment and instruments. No matter how carefully an experiment is done, an infinite number of possible sources of uncontrolled variations always exist; therefore, there are always an infinite number of auxiliary hypotheses that can be invoked to save any theory from disconfirmation. This realisation, known as Kuhn-Duhem thesis means, all theories can be adjusted to accord with any data whatsoever without discarding the core proposition.

Despite the above concerns of Sterman, some of the authors, for example Oliva (2001) have applied an approach similar to scientific methods. According to Oliva (2001), System Dynamics procedures strive for laboratory experimentation and a System Dynamics model is considered as a laboratory where a dynamic hypothesis is tested with respect to its (model’s) objective. The System Dynamics modelling process includes articulating dynamic hypothesis, building a model and testing the dynamic hypothesis (Oliva 2001).

There had been attempts to justify System Dynamics models and their validation processes from the perspective of scientific inquiry. But Sterman (2000) suggested that System Dynamics practitioners do not have to be apologetic for not meeting a utopian criterion of scientific inquiry.

9-8 While advocating scientific modelling, Homer (1996) suggests predictability as one of the benefits of scientific modelling. Homer (1996) explains:

The result is not necessarily a large model, but a model which takes into account a wide range of known details and which is therefore capable of making predictions with levels of confidence and insight greater than those of an exploratory model. (Homer 1996 3)

One aim of System Dynamics modelling put forward by the mainstream system dynamicists, is that System Dynamics models are built for the purpose of understanding and not for prediction, it is also obvious from Kahneman (2003) discussed at the start of this Chapter. The reasons for reliable predictions are not necessary in the model building process, but also in the nature of complexity itself and the constraints that bounded rationality puts on the model developer while selecting model variables, parameters and their estimation.

9.1.3 Model calibration as a testing strategy

Oliva (2003) suggests model calibration as a stringent test of a hypothesis linking structure to behaviour. Lyneis and Pugh (1996) and Reichelt and Lyneis et al. (1996) also favour the used of model calibration approaches. Oliva (2003) defines model calibration as:

…the process of estimating the model parameters to obtain a match between observed and simulated behavior. Calibration explicitly attempts to link structure to behavior, which is why it is a more stringent test than solely matching structure or behavior. Confidence that a particular structure, with reasonable parameter values, is a valid representation increases if the structure is capable of generating the observed behavior. If the structure fails to match the observed behavior, then it can certainly be rejected; the calibration exercise constitutes a test for the DH [dynamic hypothesis]. (Oliva 2003:554)

9-9 Oliva (2003) also suggests the following reasons for calibration as a testing strategy for a dynamic hypothesis:

A DH explicitly posits a causal link between structure and behaviour. Although it is impossible to verify a hypothesis, science has refined a systematic approach for increasing the confidence in a stated hypothesis and ruling out alternative explanations, namely, experiments designed to falsify the hypothesis. SD models are well suited for this experimental approach since they are logically sound and need to be relevant to the problem situation, i.e. they are empirically testable.

About the processes for model calibration, Oliva (2003) further suggests:

The calibration of the model...is typically done ‘‘by hand’.’ In this iterative process, the modeler examines differences between simulated output and data, identifies possible reasons for those differences, adjusts model parameters in an effort to correct the discrepancy, and re-simulates the model, looping back to the first step. The entire parameter estimation process therefore relies on the expertise and experience of the modeler. (Oliva 2003:555)

As evident from the above citation, the parameter assessment process is dependant on the expertise and experience of the modeller, therefore, Lyneis and Pugh (1996:317) suggest that ‘by hand’ parameter estimation attracts the following criticism:

 In complex feedback systems parameter estimation and calibration is very difficult and relies on the experience and intuition of the modeller. Hence it becomes arts rather than a science.  The process and results are not replicable as different modellers may come up with differing parameter values.  The modeller cannot be sure that their final calibration is the best.  Hand calibration makes sensitivity analysis and confidence bands cumbersome and less robust.

9-10

9.1.4 Model validation as an integrated social process

Richardson (2005) favours the traditional model validation processes used in System Dynamics (Forrester 1968, 2007, Richardson and Pugh 1981, Saeed 1992) and further suggests that the validation is integrated within the social process of model development. Richardson (2005:9) further adds that validation is present at every step of the System Dynamics modelling process, for example:

 Conceptualisation of the answers to the questions, like do we have right people, the right dynamics problem definition, the right level of aggregation?  Mapping efforts focus on developing promising dynamic hypotheses.  Model formulation focuses on clarity, logic and extremes.  At simulation, the focus is on the right behaviour for the right reasons.  Deciding implement able conclusions.  Implementation requires conviction.

Figure 9.1 shows the model validation processes as embedded in the model development processes. The left part of the diagram shows the structure validation processes while the right side of the diagram shows the behaviour validation processes.

The structure validation processes start from the empirical evidence that lead to system conceptualisation that subsequently leads to model formulation. Diagramming as well as description tools are used to represent model structure. On the other hand, empirical evidence leads to perceptions of system structure enriched by mental models, experience and literature. The model structure is then compared and reconciled with the perceptions of the system structure. This leads to improved system conceptualisation.

The left part of Figure 9.1 shows the behaviour validation processes. The empirical evidence leads to empirical and inferred time series obtained through literature or experience. The model formulation leads to the deduction of model behaviour with the help of computing aids. The comparison and reconciliation between the empirical

9-11 and inferred time series and the deduced model behaviour lead to the systems conceptualisation. Through these processes confidence is gained in both the model structure and model behaviour.

9-12

Mental Models, Literature, Experience, Experience Empirical Literature Evidence Empirical and Perceptions of Inferred Time System Structure Series System Conceptualization Comparison and Comparison and Structure Reconcilation Validating Reconciliation. Processes Model Formulation Representation of BehaviorDeduction Of Validating Model Structure Model Behavior Processes Diagramming and Computing Aids Description Tools

Figure 9.1 Two kinds of validation processes used in System Dynamics (adapted after Richardson 2006)

9-13 Richardson (2005:13) also suggests the following action for pursuing validation during the mapping:

 Think causally, not correlationally.  Think stocks and flows, even if you don’t draw them.  Use units to make the causal logic plausible, even if you don’t write them down.  Be able to tell a story for every link and loop.  Move progressively from less precise to more precise–from an informal map to a formal map.

Richardson (2005:20) suggests the following points to keep in mind while pursing validity in writing equations:

 Recognisable parameters.  Robust equation forms.  Phase relations.  Richardson’s rule: every complicated, ugly, excessively mathematical equation and every equation flaw saps confidence in the model.

Summarising Richardson’s (2005) model validation, it can be concluded that System Dynamics pursues validity at every step of model development using a rigorous tradition guiding model creation, formulation, exploration and implications. System Dynamics claims to have a powerful, intimidating battery of tests of model structure and behaviour and model based conclusions that make it through all this deserve the confidence of everyone in the process. Sterman (2000) suggests a list of tests for this purpose as shown in Table 9.1.

9-14 Table 9.1 Tests for assessment of System Dynamics models

Test Qualifying Questions 1. Boundary Are the important concepts for addressing the problem endogenous to adequacy the model? Does behaviour of the model change significantly when boundary assumptions are relaxed? Do the policy recommendations change when the model boundary is extended? 2. Structure Is the model structure consistent with relevant descriptive knowledge assessment of the system? Is level of aggregation appropriate? Does the model conform to basic physical laws? Do the decision rules capture the behaviour of actors in the system? 3. Dimensional Is each equation dimensionally consistent without the use of consistency parameters having no real world meaning? 4. Parameter Are the parameter values consistent with relevant descriptive and assessment numerical knowledge of the system? Do all parameters have real world counterparts? 5. Extreme Does each equation make sense even when its inputs take on extreme conditions values? Does the model respond plausibly when subjected to extreme policies, shocks and parameters? 6. Integration Are the results sensitive to the choice of time step or numerical error integration method? 7. Behaviour Does the model create behaviour of interest in the system reproduction (qualitatively and quantitatively)? Does it endogenously generate the symptoms of difficulty motivating the study? Does the model generate the various modes of behaviour observed in the real system? Do the frequencies and phase relationships among the variables match data? 8. Behaviour Does anomalous behaviour result when assumptions of the model are anomaly changed or deleted? 9. Family Can the model generate the behaviour observed in the other instances member of the same system? 10. Surprise Does the model generate previously unobserved or unrecognised behaviour behaviour? Does the model successfully anticipate the response of the system to novel conditions? 11. Sensitivity When assumptions about parameters, boundary, and aggregation are analysis varied over the plausible range of uncertainty do the:  numerical values change significantly?  modes of behaviour generated by the model change significantly?  policy implications change significantly? 12. System Did the modelling process help change the system for the better? improvement

Source: Sterman (2000:858-861)

9-15 9.2 VERIFICATION AND VALIDATION IN SYSTEMS ENGINEERING

To understand the Systems Engineering’s view of model testing and evaluation, it is imperative to understand the meaning of simulation in Systems Engineering. Early writers in Systems Engineering, for example, Smith Jr. (1962:34) suggest:

The simulation (or mathematical) model possesses at best a relatively small number of properties which correspond (always in an approximate way) to reality.

This implies that a model is always less than reality and the properties that represent reality always approximately correspond to reality. To elaborate the process of simulation further, Smith Jr. (1962:34) further adds:

In order to use simulation, the systems engineer must: 1. Determine those properties of reality which he is interested. 2. Determine those properties of reality which could conceivably have a significant influence upon those in which he is interested. 3. Specify the relationships between the above properties. 4. Use a modelling tool (mathematical theory, general purpose system simulator language, etc.) to build a model. 5. Establish a univalent correspondence between reality and the identified entities and relationships of the model. 6. Manipulate the model. 7. Interpret the results of the simulation.

The Systems Engineering process ensures the production of a system that is verified against the documentation produced during the Systems Engineering process and validated against the original needs (Faulconbridge and Ryan, 2005:119). Sage (2005:6) explains this process further as:

Systems Engineering is a process that is comprised of a number of activities that will assist in definition of requirements for a system, transform this set

9-16 of requirements into a system through development efforts and provide for deployment of the system in an operational environment.

In Systems Engineering confidence in a model or designed system is progressively gained through verification and validation. The purpose of verification and validation is to ascertain whether the designed system meets its original specifications (Stevens, Brook et al., 1998:159).

Verification involves checking products against their specifications and ensuring the consistency of designed systems. During verification, questions such as ‘are we building the model right?’ are answered (McLucas 2005:151, Forsberg, Mooz et al., 2005:366). Verification ensures the model is built correctly, i.e., it does not have internal inconsistency or flaws in its equation formulations and is built conforming to the baseline/requirements either identified at the start of modelling or evolved during the process.

Within the Systems Engineering process, verification is performed for multiple purposes such as design verification, reliability verification and/or quality verification and is accomplished through testing (direct measurement of performance), demonstration (witnessing actual operation), analysis (performance assessment using logical, mathematical or graphical techniques) and/or inspection (compliance to easily observable requirements (Forsberg, Mooz et al., 2005:366- 375)).

Validation processes concern the user satisfaction and seek to answer the question ‘have we built the right model? (McLucas 2005:151). Forsberg, Mooz, et al. (2005:376) further elaborate on validation as:

Validation is proof that the users are satisfied, regardless of whether the specifications have been satisfied or not. Occasionally, a product meets all specifications but is rejected by the user and does not validate.

Verification and validation processes are embedded within the Systems Engineering processes. Figure 9.2 shows the relationship of verification and validation at the multiple phases of a system development. The rectangles within Figure 9.2 indicate

9-17 different stages in a system development while the arrows indicate the sequence in which verification and validation processes are applied. Verification proceeds through various stages of system development while validation is used to confirm that the operational capability corresponds to the user requirements.

User Operational requirements Validation capability

Verification

System Verification Integrated requirements system

Verification

Architectural Verification Integrated design subsystems

Verification

Verification Components Integrated development components

Figure 9.2 Relationship between verification and validation in Systems Engineering (adopted from Stevens et al., 1998:160).

The Systems Engineering process focuses on verification and validation processes for engineering of designed systems. This process coincides with a systems lifecycle and lists key steps in the system development from concept to development, integration, and testing and deployment of the designed system. Baselines are developed during early design phases which are further verified during systems development and operation. Forsberg and Mooz et al. (2005:427) define baseline as:

The gate controlled step-by-step elaboration of business, budget, functional, performance, and physical characteristics, mutually agreed by buyer and

9-18 seller, and under change control. Baseline can be modified between formal decision gates by mutual consent through the change control process. Typical baselines are contractual baseline, budget baseline, schedule baseline, user requirements baseline, concept baseline, system specification baseline, system specification baseline, design to baseline, build to baseline, as built baseline, as tested baseline, and as fielded baseline.

Traceability is another concept governing verification of designed systems. Forsberg and Mooz et al. (2005:161) define traceability as:

The identification and management of the proper “parentage” (parent-child relationship) from the highest-level system requirements to the lowest-level configuration item requirements and to verification requirements and methods is referred to as requirements traceability - a requirements management responsibility.

Stevens, Brook et al. (1998:269-270) suggest that:

Traceability uses resources and tends to slow down our ability to make change. We should strive to minimise the necessary traceability, because of this restrictive effect. Traceability should therefore always be a compromise reflecting cost and benefits of linkage. Not everything needs to be traced; only do it where the traceability information is useful.

These Systems Engineering processes coincide with a systems lifecycle and list key steps in the system development from concept to development, integration, testing and deployment of the designed systems. This process is usually represented in a diagram that is in the shape of the English letter “V” and is also called a “Vee Model”. A traditional Vee is based on project cycle and represents a progressive system development process. It is important to note that the Vee model does not suggest a single pass from identifying a requirement through to demonstrating that those requirements have been met, rather a progressive development of the system through verification through backward and forward traceability at each stage.

Falconbridge and Ryan (2005:19-20) suggest that the basic Systems Engineering process is the analysis-synthesis-evaluation loop that is iteratively applied throughout

9-19 the systems life cycle. Homer (1996:1-19) emphasises the importance of iteration (that involves a certain amount of trial and error to reach a desirable solution) in reaching modelling rigour.

Systems Engineering lead authors, for example, Blanchard (2004), Forsberg and Mooz et al. (2005), Stevens and Brook et al. (1998) use multiple variants of the Vee model. Despite variations, Vee models present similar activities though with different terminology and the use of various levels of decomposition and integration. The first leg of the Vee model shows the activities involved in planning for the development of a system, e.g., concept, requirements or expectations form the system and design. A system is first decomposed to identify requirements. Components are designed and then integrated to progressively improve performance and compatibility of all components of the system.

Forsberg and Mooz et al. (2005) present detailed versions of Vee model for decomposition, integration and verification. Model verification and validation is embedded in the model develop process rather than a retrofit at the end of the modelling process. Figure 9.3 shows decomposition and definition aspects of the Vee model.

Decomposition involves identifying the constituent elements and relationships of the system or subsystem (Seven and Book et al., 1998:360). Forsberg and Mooz et al. (2005:110) define decomposition and definition as:

Decomposition: the hierarchical, functional, and physical partitioning of any system into hardware assemblies, software components, and operator activities that can be scheduled, budgeted, and assigned to a responsible manager. Definition: the design to, build to, and code to artefacts that define the physical and functional contents of every entity.

A baseline is developed at the initial stages of system development, based on the user discussions, plans, specifications and products. The baseline is continuously consulted during the model development and if there is a need to change the baseline,

9-20 appropriate approvals are sought. The baseline verification is designed to ensure that the model is built right.

“Off-Core” User Discussions and Approvals - In-Process Validation” “Are the proposed solutions acceptable”

“Time Now” (Vertical Line) Approved With upward and downward Baseline iterations as required

Baseline Verification Baseline Being Planned Considered “How to prove Verification that the solution Core of the has been built “Vee” Plans, right?” Specifications and Products under Configuration Baselines to Baselines to be Management be Considered Verified (Core of the (Core of the “Off-Core” Opportunity “Vee”) “Vee”) and Risk Management Investigations and Actions” “How are the opportunities and risks of the proposed Time and Baseline Maturity solutions being managed?”

Figure 9.3 Verification planning within Vee model of Systems Engineering re-drawn from Forsberg, Mooz et al. (2005:111).

Figure 9.4 shows the critical process involved in integration and verification processes. Integration involves the successive combining of the components into sub- systems and sub-systems into system. Forseberg and Mooz et al. (2005:164) define integration as:

The successive combining and testing of system hardware assemblies, software components, and operator tasks to progressively prove the performance and compatibility of all entities of the system.

9-21 “Off-Core” User Approval of Waiver or Deviation “Can the user live with the performance”

Defined Baseline System Validation Performance “Is the right solution being built?” Baseline Defined Verification Entity Baselines “Is the solution Performance being Verified being built right?” Core of the Core of the “Vee” “Vee” Plans, Specifications, “Time Now” and Products are (Vertical Line) under Project With upward and Baselines downward iterations Configuration Verified Management as required

“Off-Core” Problem Time and Baseline Maturity Investigation and Resolution “Is the problem cause understood?”

Figure 9.4 Baseline verification – the second leg of Vee. Redrawn from Forsberg and Mooz et al. (2000:115).

The baselines defined during the decomposition and definition phases, are verified with the defined entity performance and defined system performance. The key to verification and validation in Systems Engineering is that the plan for verification and validation is prepared at the initial stages.

Management of the Systems Engineering process is critical for directing Systems Engineering activities, monitoring and reporting those activities to other appropriate areas and reviewing and auditing the effort at critical stages in the entire process (Faulconbridge and Ryan, 2005:20). Blanchard (2004:40) suggests that the successful implementation of the Systems Engineering principle and concepts is dependant not only on the technology issues but on management issues. Central to this management function are the systems of review and audit, test and evaluation,

9-22 risk management, the use of specifications and standards, integration and program planning that are designed to ensure the quality of the designed system.

An extensive body of knowledge is available to practitioners in the form of Systems Engineering Body of Knowledge (SEBoK) to provide guidance on how complexity is managed in the engineering of designed systems. The purpose and intended benefits of the Systems Engineering Body of Knowledge (INCOSE, 2003) are:

The SEBoK is a comprehensive resource for understanding the extent of the practice of Systems Engineering. Accomplished systems engineers will use it to find just in time performance support and to access in-depth information in particular discipline areas. Those new to the practice of Systems Engineering will use it to expand their knowledge and increase their effectiveness in implementing Systems Engineering. Project managers will use it to find answers to questions about interacting with systems engineers as well as engineering their organization (system) of engineers and managers.

Concepts in Systems Engineering, for example, traceability, planning for verification and validation at the early stage of system design, development of baselines, use of process models such as Vee model, Systems Engineering management functions emerged to support the development of relatively simple physical systems. The availability of a Body of Knowledge further supports the system development process.

9.3 VALIDATION AND VERIFICATION PROCESSES IN SYSTEMS ENGINEERING VS VALIDATION OF SYSTEM DYNAMICS MODELS.

There are significant philosophical differences between terms describing validation and verification used in Systems Engineering as compared to System Dynamics modelling. For example, in the engineering of designed systems, what we refer to as “Systems Engineering”, the underlying assumption is that which is being designed and built can be exhaustively tested to assure that it is necessary and sufficient in the way it provides the intended desired functionality. Hard systems (and their

9-23 components/assemblies/sub-systems and etc.) can be exhaustively tested to ensure required functionality is delivered.

There are fundamental differences between Systems Engineering and System Dynamics modelling, particularly in the use of the term "validation", which is closely linked to what a "system" is considered to be in the both subjects. For example, in System Dynamics "system" refers to a convenient construct that we use to help us explain a particular problematic situation. Systems Engineering assumes that a system is no more than the sum of its parts; that if one can complexly specify all the sub-systems, the total system will automatically be specified (Hall, 1962). For example, for a Systems Engineer a bridge is a system which contains component parts which exist for a particular purpose, that is, to carry traffic from one side of an obstacle to the other side. Each component part can be modelled in detail and exhaustively tested as can be the interfaces between each of those parts.

Figure 9.5 shows a conceptual framework of the Systems Engineering process. User requirements originate from the real world (Systems Engineering world view) and elicited through stakeholders engagement or other studies during the concept design stage. System specifications are prepared from the user requirements and then provided to the contractor who performs verification through test and evaluation. This leads to operational testing and evaluation. At this stage the system is validated under certain constraints by comparing it with the user requirements. In a designed physical system if we were to remove any critical part the system it is likely to fail that particle functionality.

9-24 Contractor Real World

Verification Validation User Require Test & Operational Testing System Specification me nts Evaluation and Evaluation

Constraints

Figure 9.5 Simplified Systems Engineering verification and validation processes (Source: Personal communication with Dr Mike Ryan, UNSW in April 2009)

Problematic systemic situations that we use System Dynamics to analyse are generally much more complex than physical systems. In System Dynamics we consider (or conceptualise) problematic situations as comprising components which are interconnected in ways (systemic structures) that produce characteristic dynamic behaviours.

For the most part, System Dynamics models are conceptual even System Dynamics models which we present as computational (quantitative) System Dynamics models in applications such as Powersim, Vensim etc. Only in rare circumstances, we can claim that our System Dynamics models are valid (high fidelity) representations of real world problem situations.

In modelling, confidence that other stakeholders have in the model is also important, particularly, the models used in consultancy (Coyle and Exelby, 2000). Referring back to the questions regarding validation, the issue is more than building confidence. Discussion in the System Dynamics literature (e.g., Barlas and Carpenter, 1990; Sterman, 2000; Richardson and Pugh, 1981) on model validation is set out to demonstrate that having confidence in our models is the only way to decide about the stage when the model is ready to use. System Dynamics models are not,

9-25 generally, validated in the way validation is conducted in Systems Engineering. This is largely due to the usage of System Dynamics for complex dynamic problems.

Complexity that System Dynamics models present is dynamic complexity and only presents an endogenous explanation of the problem situation. While the complexity of the real world (or our mental construct in case of constructivist paradigm) contains both the dynamic and detail complexity. As Systems Engineering addresses complexity of relatively simpler engineering systems that have much less complexity on Kline’s (1995) complexity index than a socio-ecological system, stringent verification and validation is possible. This has implications for the ways validation is treated in the both system’s disciplines.

Sterman (2000: 81) comments on the process of validation:

Validation is a continuous process of testing and building confidence in the model...Models are not validated after they are completed nor by any one test such as their ability to fit historical data. Clients (and modellers) build confidence in the utility of a model gradually, by constantly confronting the model with data and expert opinion - their own and others'. Through this process both the model and expert opinions will change and deepen. Seek out opportunities to challenge the model's ability to replicate a diverse range of historical experiences.

Forrester (1961:115-129) goes further to explain that the model must:

…generate behaviour that does not differ significantly from the real [world] system, and explain real world behaviour through the structure and equations which reflect the real causal relationships in the real-world.

Wittenberg (1992: 22-23) explains:

...one's validation strategy depends largely on model purpose ... [for] real- world systems, whose purpose is to solve a particular problem or understand a particular mode of behaviour ... model and reference mode development are largely independent activities. Thus, while it is true that without a mental model and a purpose one would not know what behaviour is significant, it is

9-26 also true that one can observe and record that behaviour without any knowledge of its underlying causal mechanisms. Choosing which variables appear in the reference mode is surely model-dependent; the shape of the reference mode is not.

Here again, the purpose of the model has to be made explicit. This will be related to the approach we might take to demonstrate the extent to which the model might be considered "valid". Very often our System Dynamics models can be seen as necessary and sufficient representations of the real world for our purpose of learning about the real world, but they do not have the purpose of being predictive. A pioneer Systems Engineer Andrew Sage (2001) also recognised the limits of the models in predicting outcomes of complex systems.

Verification and traceability concepts used in Systems Engineering and discussed in the earlier sections are absent from the System Dynamics model development process. Figure 9.6 shows a conceptual framework for application of verification and traceability concepts to the System Dynamics modelling process. Figure 9.6 suggests that in a System Dynamics study, once the problem of concern is identified, the qualitative models are developed and such qualitative models should lead to specifications for the simulation model on the basis of which the simulation model should be developed. The qualitative model should be traceable to the problem of interest. The specifications of the simulation model should be traceable to the qualitative models and simulation model should be traceable to the specifications. A key concern is that the System Dynamics models are identified through cycles of iteration (Homer, 1994), for example, qualitative models help understand the boundaries of the problem of interest.

9-27 Traceability

Traceability/ Traceability/ Traceability/ Real World Verification Verification Verification

Problem of Qualitative models Specifications of Simulation Model interest simulation model

Validation Time, detail and dynamic parameters

Figure 9.6 Systems Engineering process as applied to System Dynamics modelling (Source: Personal communication with Dr Mike Ryan, UNSW in April 2009)

System Dynamics has value in enabling the development of "insights" and learning about the underlying causes of dynamic behaviour. Understanding these underlying causes is critical to developing strategies that are likely to work in an environment of a complex world. The Systems Engineering has value in improving the internal consistency of System Dynamics models through formal verification and validation processes.

9.4 VERIFICATION AND VALIDATION OF THE DRYLAND SALINITY MODEL.

Whether as modeler or client, we must recognize that no one test is adequate. (Sterman, 2002:521)

A combination of model verification and validation approaches used in Systems Engineering and System Dynamics were applied to build confidence in this model. The verification and validation approach used here aims at uncovering the errors in the model and increasing confidence in the model. For this Chapter, a general approach to Vee model presented by Forsberg and Mooz et al. (2005) is adopted.

9-28 Key activities of the System Dynamics modelling process were grouped and presented according to this Vee model. A System Dynamics model was considered as a system to be developed. The knowledge gained through qualitative modelling process, e.g., learning cycles, reference modes and concept mapping constitutes the baseline requirements for model development and informs the decomposition and definition for identification of modules.

In development of a System Dynamics model, verification may mean that the governing business rules have been correctly identified and coded and the structure in which those rules operate results in correct replication of the reference modes of behaviour identified in earlier stages (McLucas, 2005).

Table 9.2 shows the Systems Dynamics modelling activities used for this study against Vee model. A number of tests from the System Dynamics and Systems Engineering traditions were applied during the model building process. It is important to note that none of the tests is considered as a single criterion for model verification as the tests were used to highlight flaws/inconsistencies in the model. A list of the tests is also given in Table 9.1.

The purpose of each test and its application procedure is described along with answers to questions raised during testing procedures.

9-29 Table 9.2 System Dynamics Modelling activities organised according to Vee model of Systems Engineering.

Vee Model Components System Dynamics Modelling activities Architecture - System level decomposition- Qualitative System Dynamics: decomposition and concept of operation/ requirements - Learning cycles; definition elicitation. - Reference modes; -Concept maps; -Systems arch-types analysis; and - Causal loop diagrams. Sub-system level decomposition. - Identification of stocks and flows. Development of specifications. Identification of the individual modules needed to represent the problem. Planning for integration, - Stock and flow diagrams. verification and validation of sub- systems. LCI- Lowest configuration items Identification of auxiliaries, constants. development. Architecture - integration, System realisation. Integration of individual modules into a whole model including interfaces, verification and validation policy levers, input controls, output objects, data transfer facilities etc. Sub-system realisation. - Built-up of individual modules with stocks, flows, auxiliaries, constants using a System Dynamics modelling software, e.g., Powersim Studio. - Development of interfaces and model input controls, e.g, slider bars, switches, gauges etc. - Development of the output objects, e.g., graphs, tables, gauges, - Development of individual modules. LCI solution system realisation. Development of individual rate models with auxiliaries and constants. Compliance to baseline verification - Examination of each equation to verify that it represents real world and validation. counterparts and follows the model logic. - Model validation tests.

9-30 9.4.1 Boundary adequacy

The boundary adequacy test is generally conducted to check the adequacy of the model boundary to address the modelling purpose. While performing boundary adequacy tests, the boundaries in terms of levels of focus, time, geography, and variables are examined with respect to the purpose of model building. McLucas (2005) suggests that defining boundaries and subsequent testing must consider potential impacts of overly or unnecessarily extending model building activities. Helpful tools include model boundary charts, systems diagrams, bull's eye diagrams, cognitive maps, problem analysis sheets.

The aim of the simulation model described in Chapter 8 was to develop a research tool for comparison of different dryland salinity scenarios. A boundary adequacy test was performed to assess the endogenous nature of the key concepts and impacts of relaxing model boundary on model produced behaviour. The important concepts as elicited through different systems thinking methods including concept mapping, causal loop diagramming and influence diagramming were including in the model (Chapter 6).

The concept maps and causal loop diagrams were richer in number of identified variables. The model boundary diagram is presented in Figure 9.7. The inner circle shows the variables used with the quantitative model. The outer circle shows the variables within the qualitative models. At the current level of model development and the purpose of this model, the boundary was considered adequate. The process of simulation model development started from a simple one stock and one flow model and complexity was added progressively. It is important to note that the model boundary changed as we moved from a qualitative model towards a quantitative model.

9-31 Drought Land available for clearing Climate change Biodiversity Restrictions on land clearing Agricultural exports Rate of land clearing Land under natural vegetation Agricultural Land returning to natural Money available for investment production vegetation on land treatment Land either salt affected or at Effectiveness of the risk of becoming salt land treatments affected Abandoned land due to dryland salinity Demand for good quality Rate of land becoming Domestic consumption land for agriculture Land neither salt affected nor at the salt affected risk of becoming salt affected Rate of land reclamation Construction cost of urban Cost of control and transport infrastructure treatments Groundwater recharge Revenue Treatments for Land values dryland salinity Cost of production Salt interception Land quality schemes Farm income

Figure 9.7 Model boundary – the inner circle shows variables within simulation model while both the inner and outer circles indicate the variables within qualitative models.

9-32

9.4.2 Structure assessment

Structure of a System Dynamics model consists of cause and effect relationships that are endogenous and represents the building blocks and interval connections of a system. It is the way in which system elements are organised or interrelated (System Dynamics in Education Project 2002). This hypothesised structure needs to be consistent with the part of the reality suitable for the problem.

McLucas (2005:153) suggests structural assessment tests to determine the consistency of the model with the real world behaviour represented through modules, sectors and models. He further adds:

…design and conduct of testing is informed by our knowledge of functional components and how do they contribute to the dynamic behaviour.

Sterman (2000:863) suggests examination of the model in regards to aggregation level with respect to model purpose, physical laws, and conformity of decision rules with the behaviour of the actors in the system.

The qualitative models presented in Chapters 6 and 7 are concept maps, causal loop diagrams, influence diagrams and reference modes. Concept maps contain 40 different concepts. The composite causal loop diagram consists of five causal loops. The influence diagram carries twelve variables and the reference modes developed using learning cycles approach present the behaviour over time graph for 12 variables. The simulation model consists of the three main stocks and four rates.

During model development, descriptive knowledge was collected through a review of the published literature and was described in Chapter 2. The collected knowledge was subjected to rigorous qualitative analysis (Chapter 7). The structure of the simulation model was identified through collation and analysis of the relevant descriptive knowledge of the dryland salinity problem in Australia.

During structure assessment, the following questions consistent with Sterman (2000:859) were answered about the model structure.

9-33 Is level of aggregation appropriate?

The level of aggregation was kept at the Murray Darling Basin level. Keeping in view the purpose of the model, the aggregation level at the level of the Murray Darling Basin was considered appropriate based on the level at which strategic decisions are made. However, the model has the flexibility to be adjusted for conditions in other catchments. The model offers a simple structure and was therefore considered appropriate for the purpose and provides flexibility for users to easily change the size of the Basin etc. By changing variables, the model can be used for different size river basins provided that the issues and the causal factors are similar.

Is the time-step appropriate?

An appropriate time step is needed for compilation/integration of dynamic models Sterman (2000:872-873) suggests that the model results can be sensitive to the time step and can introduce serious errors if time-step is not appropriate. Sterman (2000:872) suggests that:

Always test for dT error by cutting the time step in half and running the model again. If the results change in ways that matter, the time step was too large. Continue the results are no longer sensitive to the choice of the time step.

Selection of a suitable time step should be part of good modelling practice and such experimentation was done.

Does the model conform to basic physical laws?

The model conforms to the basic physical laws applicable to stocks. It was ensured that the stocks do not go below zero where it was not logically possible. For example, a land stock cannot go negative. If it could it would negate the law of conservation of matter and would indicate serious formulation flaws in the model equations.

The following function was used to ensure non-negativity of stocks and logical flow direction:

9-34  Max(Input1, Input2), where input 2 equals zero

The model was subjected to varying flow levels and their impacts were observed on stocks. Two examples of such test are described below:

In the first test, a random number was used for the land fraction under natural vegetation. Figure 9.8 shows the over time behaviour of the stock of land under natural vegetation. Despite the large variation in the clearing fraction, the land under natural vegetation does not show spontaneous or erroneous behaviour or negative stocks. Rather it shows a progressive decrease in the land under natural vegetation that slows as it reaches the bottom line.

In the second example, a random input was used for time delay in land clearing. Time delay varied between zero and thirty years. Figure 9.9 shows the over time behaviour of the ‘rate of land clearing’. Despite such large variations, the rate of land clearing never went below zero.

Behaviour of the three land stocks, i.e., land under natural vegetation, ‘cleared land not at the risk of dryland salinity’ and cleared land either saline or at the risk of dryland salinity, is shown in Figure 9.10. None of the stocks went below zero at any point in time

9.4.3 Sensitivity analysis

It is an important tool in System Dynamics model building process and is based on a belief within the System Dynamics community that system structure and not parameter values have the most influence on the system behaviour. It helps to understand that how conclusions from a model based study change when the basic assumptions underlying that model are relaxed or changed (Sterman, 2000:883).

Sterman (2000:883) has suggested the three types of sensitivities that need to be analysed in order to build confidence in a model:

 A model is said to be numerically sensitive when a change in assumptions changes the numerical values of the results.

9-35  A model is said to be behaviourally sensitive when a change in assumptions patterns of the behaviours (shapes of the curves) produced from that model.  A model is said to have policy sensitivity when a change in assumptions reverses the impact or desirability of the proposed policy.

Selection of a sensitivity concern to be analysed depends upon the purpose or intended use of that model. It is generally used to test the parameters that are both highly uncertain and likely to be influential. Sterman (2000:884) considers comprehensive sensitivity analysis impossible even when restricted to parameter sensitivity as it would require testing all combinations of assumptions over their plausible range of uncertainty. The number of combinations can be very large even in the models of the modest size.

A Sensitivity Analysis was performed during structure assessment while ensuring that the model does not violate basic physical laws as described in Section 9.3.2. A wide range of parametric values were given using random numbers. The stocks behaviour did not show sensitivity to changes in parametric values. Two examples of such tests are already described in the above section and the stock behaviour shown in Figures 9.6 and 9.8.

9-36 ha

80,000,000

60,000,000

40,000,000

20,000,000 LAND UNDER NATURAL VEGETATION NATURAL UNDER LAND

0 1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 9.8 Response of the stock ‘land under natural vegetation to varying time delays’

ha/yr

3,000,000 g

2,000,000 learin C d

ate of Lan 1,000,000 R

0 1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 9.9 Rate of land clearing – conforming to physical laws, i.e., rate of land clearing never goes below zero.

9-37 ha

80,000,000

Cleared land

60,000,000

Land at risk of dryland salinity 40,000,000

20,000,000

Land under natural vegetation 0 1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure 9.10 Inventories conforming to physical laws. None of the stocks go below zero at any point during simulation. Note: Green line represents the land under natural vegetation, brown line represents the cleared land neither saline nor at the risk of becoming saline, and the red line represents the land that is either saline or at the risk of becoming saline.

9.4.4 Parameter assessment

Sterman (2000) presents two aspects of parameter assessment: a) each variable and constant should have a real life meaning or a real world counterpart, b) the parameter (whether constant or variable) should be assessed through a valid way of parameter assessment.

McLucas (2005:154) refers to it as a parameter verification test and emphasises consistency between parameter values and the available knowledge. It is evident from the following excerpt:

The values of each parameter included in a module, sector or model must be tested for consistency with what is known about that those parameters. In the

9-38 absence of hard numerical data, parametric verification must be based at least on descriptive knowledge combined with reliable estimates of specific parameter values.

The suggested methods for parameter assessment include statistical data and elicited knowledge. Despite stress on the hard data for parameter assessment, it is specially emphasised in the System Dynamics literature that any important parameters must not be eliminated from a model on the grounds of the unavailability of data (Forrester, 1994; Sterman, 2000). Therefore the System Dynamics method presents a trade-off between the perceived importance of a parameter and the robustness of the assessment method for that parameter. Shreckengost (1997) accepts that many required parameter values in a model may not exist and must be developed.

The parameter assessment tests imply that the values of the parameters (whatever they are derived from) are subject to a rigorous and demanding environment as they contribute to the confidence in the model.

During the parameter assessment, individual parameters, their values, the integration method and the time step were evaluated.

The parameter values were selected based on the review of literature (Chapter 2) and information discovery process used for developing reference modes (Chapter 7). Therefore, parameter values are derived from relevant descriptive and numerical knowledge of the system e.g., time delays in land clearing, fractions of land clearing, and land stocks. The model provides an input window for the changes in the parameter values for experimentation with the model.

The Euler integration method was used which is also a default integration method for Powersim Studio 7. The Euler's integration first calculates the initial values of the levels and flows then it uses the flows to update the levels, the new values of the levels to recalculate the flows, and so on. Euler's method assumes that a flow is constant over the interval of the time step.

In the equations below, Lt represents the value of the level L at time t, and F(Lt, t) represents the value of the flow F into (or out of) level L at time t.

9-39 Powersim (Powersim Corporation, 2007) executes the following two steps to calculate the integral over an interval from T to T+dt:

1. Calculate the derived flow when t=T

Flow=F(LevelT, T)

2. Calculate the value of Level at t=T+dt based on LevelT and Flow

LevelT + dt = LevelT + dt*Flow

Time step was changed and sensitivity of the model to time step was tested until the model was no more sensitive to the changes in the time step.

9.4.5 Flow calculation sequence

A flow sequence test was conducted to ensure the integrity of stocks and that the outflows do not precede inflows.

The sequence of flow used within each module is shown in Figure 9.13. It ensures that the outflows from the stock do not start prior to inflows.

9-40 Rate_4 Rate_3

2 3

1 4

Rate_1 Rate_2 Level

Figure 9.11 Flow of material: the bold circles indicate the sequence of flows in the model.

9.4.6 Mass balance test

Mass balance tests belong to same category of verification tests that ensures the appropriate sequence of calculations. Moreover, they ensure that data is neither created nor destroyed during calculations.

A mass balance test was performed to ensure that the algebraic operators do not result in inadvertent creation or destruction of flows. The method described by McLucas (2005) was used. A new variable called Delta was created for each module. Delta represents the sum of all flows into a stock and the structure of the Delta check model is shown in Figure 9.14. The generic structure suggested by McLucas (2005) was used as a guide while developing this model. The delta check indicated zero mass balance error.

9-41 0.00 ha

95,532,210.00 ha

0.00 ha Count cleared land Rate of Land LAND UNDER Clearing NATURAL VEGETATION

955,322.10 sq km 0.00 ha Delta check - land under natural vegetation module

Initial land under Count land returned to natural natural vegetation Rate of land at risk vegetation of becoming salt 0.00 ha 5,307,345.00 ha affected returning to natural vegetation 0.00 ha CLEARED LAND NEITHER SALT Delta check - cleared land module AFFECTED NOR AT RISK OF BECOMING SALT AFFECTED Count reclaimed land Rate of land reclamation 5,307,345.00 ha

0.00 ha 0.00 ha Initial Cleared land not at risk of becoming salt Count land at risk of becoming affected Rate Land becoming salt affected Delta check - land either salt Salt Affected affected or at risk of becoming salt affected module

1,061,469.00 ha 1,061,469.00 ha

CLEARED LAND Initial land at risk of EITHER SALT becoming salt AFFECTED OR AT RISK affected OF BECOMING SALT AFFECTED Figure 9.12 Delta check model

9.4.7 Dimensional consistency

All algebraic expressions, arithmetic integrations and parametric values must be consistent both in terms of units involved and in case of arrays, have consistent and compatible dimensions (McLucas, 2005). Every equation must be dimensionally consistent without the inclusion of arbitrary scaling factors that have no real world meaning. Such fudge factors can be identified by direct examination of equations. Parameters with meaningless names, strange combination of units, or non- dimensionless parameters with values of unity are suspected (Sterman, 2000).

Dimensional consistency of the model was checked to ensure the integrity of units. Land stocks are in hectares while all rates of change are hectares/acres.

9-42 Dimensional consistency has been assured by:

 using software that does not allow to simulate if there is dimensional inconsistency noting that Powersim Studio has been used to confirm dimensional consistency;  each equation was individually analysed to ensure dimensional consistency; and  a simple model was built without using complex technicalities including multi-dimensional arrays.

9.4.8 Behaviour reproduction tests

The proper use of the behaviour reproduction tests is to uncover the flaws in the structure or parameters of the model and assess relevance of such flaws to the modelling purpose. There exists a difference of opinion about the significance of the behaviour reproduction tests. For example, McLucas (2005:153) suggests:

The module, sector or model must generate behaviour modes, phasing and frequencies and general characteristics of behaviour which sufficiently and faithfully represents behaviour of the real world systems.

Sterman (2000) suggests:

Beware of a modeller who asserts that the model’s ability to fit the data indicates that the model is valid or confirms the model.

Despite of this difference of opinion, the overarching purpose of behaviour reproduction tests is to find out asymmetries in the model. Generally it is done by plotting the simulated behaviour and actual behaviour and then examining the changes differences in the actual and simulated behaviour against the causal factors introduced. This poses a serious challenge in case of problems for which statistical data is not available for plotting actual behaviour. To address such problems reference modes (Chapter 6) are developed within System Dynamics method using problem description, qualitative and quantitative data (Khan, McLucas et al., 2004).

9-43 Saeed (2002) defined a reference mode as a fabric of patterns of behaviours of different variables. To expose this pattern of behaviour, several variables are plotted against each other to identify reference modes. The emerging picture contains a fabric of patterns of behaviour that is prepared from actual data/experts judgement etc. that is compared against fabric generated by a pattern of behaviour that emerges from the simulated data. Comparison and contrast of the two will unveil the relationship amongst the actual and simulated behaviours. The interpretation of this comparison will be used as behaviour reproduction test.

This test is generally used for models whose purpose is to reproduce very accurately the real world system by comparing simulation results and real historical data. This test does not apply here due to two main limitations:

The statistical data of the real system about main model parameters, for example, land clearing and dryland salinity is not available. The National Land and Water Resources Audit (NLWRA, 2001) used a variable ‘land either salt affected or at the risk of becoming salt affected’. The land becomes at the risk of becoming salt affected when he groundwater reaches within 2 meters of the land surface. The same variable has been used in this model.

In the absence of the statistical data, the best guess was the behaviour pattern identified through qualitative information, and developed further through reference modes. The behaviour pattern indicated that dryland salinity has increased since the start of the century (this behaviour pattern is described in Chapter 6 in detail).

The purpose of the model is not to replicate the real data or provide predictions about the level of dryland salinity in the Murray Darling Basin. The key utility of the model is insights into the ways model variables interact to generate a behaviour pattern giving rise to dryland salinity that is similar to the qualitative information.

The model is based on a simple model of the phases through which a piece of land passes through, changes its cover and faces impacts of those changes and helps in identifying the impacts of the various land reclamation strategies.

9-44 9.5 A NEW PROCESS FOR VALIDATION - SYNERGISTIC USE OF V- MODEL AND SYSTEM DYNAMICS MODEL VALIDATION TECHNIQUES

System Dynamics models are developed for learning about complex problems and to be effective the learning needs to be error free. A new process involving the use of Systems Engineering V-model for validation of System Dynamics models was described in the preceding sections. Key features of this process are:

 a modular approach for building basic units of the model;  the use of V- model for streamlining model building process; and  verification testing.

Features of the proffered approach are presented in Figure 9.13. There are certain model building aspects, like requirements identification, sub-system planning and model verification, in which the application of Vee model can improve the quality of a System Dynamics model. Qualitative models like reference modes, causal loop diagrams, concept maps or influence diagrams provide requirements for the model. These requirements can be transitioned into specifications that will provide a baseline for model validation.

Moreover, the application of Vee model can enhance the model development process and can help in development of robust and ‘responsive to purpose’ models. This is also supported by the results of a study by McLucas and Ryan (2005) that highlight the strengths of Systems Engineering in design, building and testing quantitative System Dynamics models including the detailed transition from conceptual representation to quantified model. They examined the System Dynamics modelling process and suggested the following benefits by using the Systems Engineering process:

 deliberate and careful management of the complexity introduced at each stage of the model building process;  discipline and rigour associated with requirements engineering;  aid in managing and coping with complexity through a top-down approach; and

9-45  rigour in validation and verification.

The use of the Vee model can also provide a mechanism for model verification as it emphasises planning and evaluation in almost all stages of model development. The case study presented above, highlights the synergies between Systems Engineering and System Dynamics and provides an avenue for further exploration of such synergies between the two methods to improve model quality.

Moreover, Systems Engineering has a published body of knowledge available to the system developer. Such a ‘Body of Knowledge’ does not exist in System Dynamics for guidance of novice system dynamist.

9-46 Modular approach V-model Verification and testing

Physical Outflow “Off-Core” User Discussions and Approvals - In-Process Validation” Physical Inflow Export to Dataset Definition to include: “Are the proposed solutions acceptable” flow type discrete or Definition to include: Definition to include: Verification – building the model right continous;flow flow type discrete or write format;write direcion;max flow continous;flow direction;conversion “Time Now” (Vertical Line) rate; dt;simulation Validation – building the right model direcion;max flow factors;units of time step;simulation rate; dt;simulation measurement. Approved With upward and downward time horizon;units; time step;simulation Traceability dimentions Baseline iterations as required time horizon;units; dimentions Baseline Constant_2 Verification Mass balance test Baseline Being Planned Considered “How to prove Verification Boundary adequacy that the solution Core of the has been built Family member test “Vee” Plans, right?” Rate_1Level Rate_2 Specifications Sensitivity analysis and Products Flow calculation sequence/test Constant_1 under Configuration Baselines to Baselines to be Behaviour tests Management be Considered Verified Definition to include: (Core of the Import from Dataset (Core of the read format;read “Off-Core” Opportunity “Vee”) “Vee”) direction;conversion actors; units of Information Inflow and Risk Management measurement. Definition to include: Informationsampling rate;dt; Outflow Definition to include: Investigations and simulation timestep; sampling rate;dt; Actions” simulation time simulation timestep; horizon;calendar. simulation time “How are the opportunities horizon;calendar. and risks of the proposed Time and Baseline Maturity solutions being managed?”

Figure 9.13 Features of the proposed validation process for System Dynamics models

9-47 9.6 SUMMARY AND CONCLUSION

This Chapter presents a combination of Systems Engineering and System Dynamics approaches for model validation that were applied for validation of the model presented in Chapter 8. This Chapter also critically describes the opportunities where the Systems Engineering Vee model can help enhance verification of a System Dynamics computer simulation model development.

System Dynamics proffers various tests for increasing user’s confidence in the model. These tests are both applied during the process as well as at the end of the simulation model development. The computer simulation model building faces high costs and difficulty in quantifying some of the difficult to quantify variables. A major part of System Dynamics community strongly believes in the value that computer simulation modelling adds to qualitative models. This provides opportunities for other systems disciplines to contribute to System Dynamics modelling to build robust simulation models.

The application of the Vee model for developing a strategic dryland salinity model made modelling simpler through a step-by-step process. Once the specifications for the model are written, the development of the model equations becomes a relatively easier task. The whole process ensures the important model building steps including model verification are not ignored and further ensures model quality. The process can as well be useful to the novice System Dynamicist during their experimentation and learning as by following the specifications, they have the opportunity to learn about each verification and validation step while ensuring their early happy and successful experience with System Dynamics.

Further exploration of such synergies between the two systems approach can help improve model building processes and hence the model quality on which the policy exploration depends.

This was the last Chapter describing the application of System Dynamics modelling. The next Chapter provides a synthesis of the argument leading to a logical conclusion and recommendations for future research on this issue.

9-48 CHAPTER 10 SUMMARY AND CONCLUSIONS

This dissertation has investigated the nature of dryland salinity, its causes and the difficulties that arise in its management. Why such problematic situations are confounding to manage is examined, noting particularly that humankind’s actions over two centuries have played a dominant role in creating dryland salinity as we know it in Australia today.

The purpose of the current study was to define an improved methodological approach for designing management strategies that are likely to produce enduring results. It also aims to identify the likely effectiveness of alternate strategies and the time frames over which results might be achieved. A pluralist multi-methodology approach was undertaken, that combines an understanding of feedback dynamics and one which sets out to improve the rigour of investigations into complex systems. The key methodologies are System Dynamics Modelling and Systems Engineering.

This Chapter summarises main findings and limitations of this research; and broadly examines opportunities for future research on methodological issues in the application of System Dynamics to complex environmental problems and concludes this presentation.

10.1 SUMMARY OF MAIN FINDINGS AND THEIR LIMITATIONS 10.1.1 Dryland salinity is a complex dynamic problem

It is a global problem present in different ecological regions in almost all continents. Dryland salinity in the Murray Darling Basin has developed over a long time, at least over 200 years. It is not only a physical problem, rather a large number of biological, physical and socio-economic system causes can be attributed to it. It has far reaching impacts on physical, economic, social and cultural systems.

The potential policy measures that have either been adopted or suggested in the literature and described in detail in Chapter 2 and Annex B are based on models of either physical or biological processes or in some instances physical models with a biological or economic layer. The causes that need to go into the design of effective management strategies are not clear. Regarding physical causes, there exist two

10-1 schools of thoughts with opposing hypothesis about the physical causes of dryland salinity (for details, please see Chapter 2, pages 2-8 to 2-15).

The complexity of the problem further increases when such causes are linked through feedback mechanisms. Causal loop diagrams of the dryland salinity problem in the Murray Darling Basin are presented in Chapter 7 (pages 7-14 to 7-29). Any change in one part of the system may cause side effects or impacts on another part of system (that was not in focus while designing policies) often after long delays. For example, from the 1860s to 1960s leases and conditional purchases were issued on the proviso that a certain percentage of tree cover was to be removed each year (BRS, 2000). This land clearing, afterwards was attributed to causes of dryland salinity, greenhouse impacts and climate change. At the time of issuance of land permits it might not be in the cognition of policy makes that imposing such conditions will cause serious ecological problems after a couple of decades.

10.1.2 While existing models of dryland salinity in the Murray Darling Basin provide understanding at the process level, their ability to provide insights for strategic management of dryland salinity is seriously compromised.

Examples of the existing models for dryland salinity management were described in Chapter 3. These models do not address the complexities involved in addressing dryland salinity, boundary issues while defining dynamic problems, emergent properties, dynamic complexity and time delays. Such issues are characteristics of complex dynamic problems (sometimes referred to as wicked problems) that dryland salinity also shares. These issues need to be considered, if useful insights are to be gained to assist the management of dryland salinity. Successful intervention into such problems requires more than technical tools and mathematical models; it needs an understanding of the problem at a system level before designing strategies. Such an understanding can be gained through processes and tools that foster learning.

10-2 10.1.3 Opportunities for understanding and learning in the complex dynamic systems are limited. Effective management needs to make benefit use of whatever opportunities are available.

Learning in complex dynamic systems faces several impediments. These include dynamic complexity, limited information, ambiguity, confounding variables, flawed cognitive maps of causal relations, judgemental errors, bounded rationality and many others. Chapter 4 provides a detailed account of such impediments. Experimentation with the real system may involve negative consequences for humans or the physical and ecological systems and might not be feasible in some instances or even be impossible to realise results due to long time delays. Such reduced opportunities for controlled experimentation prevent one from learning from the consequences of one’s actions and distort the outcomes of the feedback one receives. Every link in the feedback loops by which we might learn can be weakened by a variety of structures (Sterman, 1994, 2000).

Such impediments to learning increase the challenges to design effective strategies. Efforts to address complex dynamic problems need a good understanding of both the system and the problem to help devise strategies that endure. There is a need to benefit from whatever opportunities are available. Overcoming such impediments needs synthesis of many methods and disciplines from mathematics and computer science to psychology and organisational theory (Sterman, 1994).

10.1.4 Specific insights into the dryland salinity problem in the Murray Darling Basin provided by the multi-methodology framework

The specific insights provided by multi-methodology are presented in detail in Table 10.1.

10-3 Table 10.1 Specific insights about dryland salinity provided by the pro-offered multi-methodology framework

Methodology Specific Insights about Dryland Salinity Field survey Responses to the field survey questionnaire provide farmers perceptions about the dryland salinity problem in the Murray Darling Basin. The way farmers’ view this problem is central towards identification of implementable mitigation strategies. Below are the specific insights gained through the analysis of the field survey responses:  The extent of dryland salinity on individual properties is less than 20%.  The increase in dryland salinity affected areas was at a maximum in 1980 to 2000  During the last 10 years (2000-2010), the extent of dryland salinity has decreased.  Major impacts of dryland salinity include obvious salt- crust/patches, reductions in crop yields, loss of biodiversity.  Impairment of infrastructure/water supplies.  Area under dryland salinity and the severity of its impacts increase as the depth to groundwater decreases.  Salt-tolerant crops, plantations, and changes in crop husbandry are among the effective mitigation/control options.  Land clearing is one of the secondary reasons for dryland salinity.  Dryland salinity will decrease in the next 40 years. The decrease between 2020 and 2050 will be more than the one between 2010 and 2020. Reference modes Data needed to build reference modes for the dryland salinity development using problem is either not available, and in cases where available learning cycles lacks consistency to build time-graphs over a long period that

10-4 approach is necessary for simulation to run.

Development of reference modes of dryland salinity will need to rely on descriptive information, mental models/interviews of stakeholders and logical relationships.

The reference modes of dryland salinity for System Dynamics intervention will need to include over time trend/behaviour of land clearing, land under forest/bush, salt stores, water table, and land stocks (i.e., land under natural vegetation, cleared or river diversions and farm incomes). Cognitive and Dryland salinity is a complex problem, has multiple causes, concept mapping severe impacts and solution to the problem partly lies in government policy addressing:  conflicts among government policies, conflicts among interstate policies;  labour shortage;  need for incentives for young farmers to stay on land,  farmers feeling that whatever they do does not bring results, i.e., effectiveness and efficacy of implementation practices.  identification of better match between soil types and vegetation.

Concept maps presents more than 40 concepts related to dryland salinity linked to each other in more than one ways. It presents a web in which dryland salinity problem exists. Any remediation strategy to be effective will need to interact within this concepts web to generate results.

Such a veracity of relationships makes dryland salinity a difficult problem to model or gain only model-based insights

10-5 leading to identification of high leverage strategies for sustainable resource management.

This phase identifies:  central concepts, densely linked concepts and potent concepts and provides guidance on the variables to be included in a model. Detail is provided in Chapter 7 and Annex C.  remediation strategies as salt interception schemes, agronomic solutions, plantations, targeted reforestation and stabilisation of salt patches/salt stores. Causal  The feedback loop diagram (Chapter 7) presents 43 analysis/causal- variables involved in dryland salinity system that are loop diagrams either involved in reinforcing loops or balancing loops.  The diagram suggests that the intractable nature of the dryland salinity problem is due to different variable interactions with each other and some variables interacting with multiple variables giving rise to multiple loops that involve time delays. It helps to explain why the causes and symptoms of the dryland salinity problem are distant in space and time.  Presents the endogenous views of the problem and shows the points of impact of various mitigation strategies with varying leverage on the problem: o Capacity building strategies. o Awareness campaigns. o Education and training. o Adoption of agronomic options. o Demonstration centres. o Reducing costs of dryland salinity remediation. o Stabilisation of salt patches. o Targeted reforestation.

10-6  A communication strategy that focuses on information that is directly useful for the farmers affected by dryland salinity.

Influence An influence diagram helps to identify stock and flows in the diagramming dryland salinity system and the direction of influences.

Simulation Finding effective mitigation strategies is a key towards modelling improving state of the system yielding dryland salinity. Information about effectiveness of various dryland mitigation strategies or remedial measure is lacking. This is the critical information needed for developing simulation model.

Because of the complexity of dryland salinity, no conclusive mitigation strategies for dryland salinity can be identified only using simulation modelling.

Validation of System Dynamics simulation model faces serious challenges for model verification and validation. Prior to its useful application for strategic management of dryland salinity, the issues in model validation need to be addressed.

Systems Systems Engineering provides an objective oriented process for Engineering developing System Dynamics simulation models of dryland salinity and subsequent verification and validation. A Systems Engineering V model has been presented along with conventional System Dynamics model validation processes to improve modelling rigor.

10-7 10.1.5 Contribution of multi-methodology to strategic management of dryland salinity

Multi-methodology makes the following contributions to the strategic management of the dryland salinity:

 Multi-methodology helps to brings out the complexity of dryland salinity that is essential for identifying strategies that can bring sustainable results. Multiple dimensions and complexity of dryland salinity has been discussed in Sections 10.1 to 10.3. Strategic management of dryland salinity needs an approach that can help explore complexity, multiple problem dimensions, views of the stakeholders; robust methodologies with appreciation of strengths and weaknesses of such methodologies; a framework to complement weaknesses of one methodology with strengths of other methodologies;  Chapter 7 presented concept maps showing a web of interactions between concepts. Multi-methodology provides a framework for using methods that highlight the points of interaction of remediation strategies with the other aspects of dryland salinity. Appreciation of such interactions is central in understanding why some strategies cannot be implemented or why they do not bring results.  The individual methods (e.g., field survey, cognitive mapping, System Dynamics, Systems Engineering) address various individual aspects of dryland salinity. Multi-methodology helps to address dryland salinity as a whole with the full richness of the capabilities of field survey, cognitive/concept mapping, System Dynamics, Systems Engineering.  The models proposed in the preceding chapters are complex ones. Building such models, validation and gaining useful insights face serious challenges. The learning process employed in this modelling provides the basis for further research or limited scale intervention. The intervention processes for dryland salinity proceed through different phases and these phases pose a variety of tasks. Within this process, some methodologies are useful at a certain stage of intervention while others are applicable at other stages and the multi-methodology approach provides such utility. Methods keep on

10-8 evolving as a result of on-going research and applications are, therefore, at different levels of development and user confidence.  Combining different methods even if they perform similar functions provides an avenue for cross verification through synthesis and integrated analysis of data from multiple sources for decision making. Such “triangulation” can increase both researcher and user confidence in the methods and, the consequent results.

10.1.6 The use of System Dynamics methodology in a multi-methodology framework can enhance the capabilities of System Dynamics in terms of problem conceptualisation and verification and validation of System Dynamics models

System Dynamics framework consists of tools for conceptualisation of feedback relations in the form of causal-loop diagrams, and a medium for learning through experimentation with the specifically developed simulation model. Chapter 4 provides a detailed account of the System Dynamics method.

Despite being a potential method for researching dynamic complexity, problems like dryland salinity still remain out of its sphere of influence and there are relatively less applications of System Dynamics to strategy development problems. This arises from the criticism such as the focus of System Dynamics on closing feedback loops carries a risk of ignoring exogenous factors that would have been identified should the focus be on identifying resource flows and their drivers. Such exogenous factors, in some instances can have critical influences. Moreover, the validation approaches in System Dynamics have also been criticised for being unscientific. Though the System Dynamics community responded to such criticism and suggested a relativist paradigm, there remain suggestions for System Dynamics to learn from other disciplines.

The use of cognitive mapping along with other diagramming conventions like causal loop diagram can strengthen the problem conceptualisation, by means of providing a framework for exploration of ideas and for preparing individual’s cognitive maps. The use of multiple diagramming/mapping approaches at the conceptualisation stage

10-9 can also provide additional rigour to the qualitative analysis through creating opportunities for triangulation of the insights.

While having a strong conceptual basis for problems that have certain “systems” characteristics, but where the component parts cannot be isolated and replicated, System Dynamics simulation model development processes are not as rigorous as are the ones used in Systems Engineering. This rigour generally leads to exhaustive model testing and sound model verification and validation practices. This differing rigour in processes is linked to differing modelling philosophies, the purposes of the modelling activity, intended use of the models and the practices underlying both methodologies. Synergistic use of both methodologies can strengthen problem conceptualisation as well as model verification and validation.

10.1.7 Qualitative frameworks such as concept maps, causal loop diagrams and influence diagrams help in learning about complex problems.

In Chapter 7 qualitative approaches used in the study are described, this included development of concept maps, causal loop diagrams and influence diagrams. All three analyses were based on the secondary data and literature review described in earlier chapters. The concept maps present the main concepts and factors involved in the lands becoming salt affected. The causal loop diagrams presented dynamic hypothesis while the influence diagrams presented the cause and effect relationship amongst those factors.

All three diagramming conventions used different levels of detail about the dryland salinity problem but covered all the important variables. The main groups of variables of interest that a model of dryland salinity should include are land clearing, dryland salinity, land quality, revenue, and control treatments/remedial measures to address problems. All three diagramming conventions represented core variables in the diagrams; however, diagrams differ in detail. Some variables, if not included in the model can provide context of the problem for descriptions. All three diagramming conventions are useful as they depicted the core concept, however, they do not depict the behaviour of variables over time that is often required for building and validating System Dynamics models.

10-10 10.1.8 Development of reference modes is a difficult task. The historical data for problems developed over a long time, for example 200 years in the case of dryland salinity, might not be available or reliable, even if available, as the data collection purposes and methods evolve over time with advances in knowledge. Learning cycles approach, a method suggested by Saeed (2002), enhances learning about the complex problems but needs a better involvement of stakeholders.

Learning cycles approach is a valuable starting point when identifying variables that might be incorporated into the model. The method helps in identifying data shortcomings and helps keep the modeller well focused on the problem being addressed. However, difficulties were encountered in application of this method. In part, the difficulties stem from expectation that a sound statistical database will be available for key variables. This was not the case in this problem nor, it is suggested, in the generality of dynamic based problems. Variations in the level of aggregation of available data proved problematic, as did a lack of involvement of the many opposite stakeholders in defining the problem space. Whilst most appropriate when consistently aggregated data is available for a specific problem space, learning cycles method was found lacking here.

Reference modes presented in Chapter 6 are based on the published literature and farmers’ responses during a field survey. The shortcomings in this instance might be reduced through extensive and close involvement of stakeholder’s right from the earliest stages when attempting to identify the preliminary model boundary. It is based on the beliefs that stakeholders are i) a part of the problem space as they influence the current system’s behaviour through their decisions, ii) an integral part of learning cycles, and iii) knowledgeable people essential to activities directed at identifying missing variables, missing data or explaining variations in quality of data. By addressing these issues, a template can be prepared based on the learning cycles method that may help the modellers in problem articulation and development of reference modes. This is an issue for the future research to take-up.

10-11 10.1.9 Simulation modelling provides tools for learning through controlled experimentation with the computer model. It does not guarantee a definite change in flaws in mental models. Effective learning occurs when decision makers actively participate in the development of the model.

Simulation plays an important role in addressing the dynamically complex problems. It speeds up learning and provides an avenue for multiple hypotheses building and testing through experimentation and it can help in visualising the outcome of mental models. Both qualitative maps and simulation models used synergistically can improve each other in the iterative cycles.

The model developed a parameter ‘effectiveness of control treatments’ that is specific to the type of control treatment applied. It helps in conveying the concept that all control measures are not equally effective and should be treated in the model corresponding to their effectiveness in addressing dryland salinity. The value-add of the simulation model development and experimentation includes the opportunity for developing alternative hypothesis (within the model boundaries) using different combinations of effectiveness of control treatment, area on which such treatments are applied and the time delays between application of a treatment and the results of that application. Such experimentation could not be performed with qualitative models.

One of the limitations of the model is that it was kept simple for the purpose of investigations into the model building process. The model can be further developed, calibrated and used in understanding the ecological system giving rise to dryland salinity, time delays, dynamics, linkages of dryland salinity with other parts of the food production system and their impact on dryland salinity, and for identifying strategies that respond to natural time delays involved in the dryland salinisation processes.

10.1.10 System Dynamics models can be taken to the next level of confidence through the use of verification and validation concepts in Systems Engineering while taking into account the philosophical difference between physical systems and the systems of interest within System Dynamics.

System Dynamics proffers various tests for increasing user’s confidence in the model. These tests are both applied during the process as well as at the end of the simulation model development. Despite this literature raises concerns about the

10-12 errors in System Dynamics models (even published in the referred journals), particularly those built by novice modellers in System Dynamics. This may arise from the problems within the validation processes of the System Dynamics models or there may be a need to learn from verification and validation processes used in other disciplines.

System Dynamics models are primarily built for learning. One of the uses of System Dynamics models, advocated by the system dynamic literature, is as management flight simulators or virtual worlds for the training of managers. In such cases, the effective learning outcomes need to be error free.

The application of Systems Engineering process for developing a strategic dryland salinity model made modelling simpler through a step-by-step process. The whole process ensures the important model building steps including model verification are not ignored and further ensures model quality. The process can as well be useful to the novice System Dynamicist during their experimentation and learning as by following the specifications, they have the opportunity to learn about each verification and validation step while ensuring their early happy and successful experience with System Dynamics. For this purpose, there exists an easily accessible ‘Systems Engineering Body of Knowledge’ which is not available in the case of System Dynamics.

Further explorations of such synergies between the two systems approaches can help improve model building processes and hence the model quality on which the policy exploration depends.

10.2 SPECIFIC CONTRIBUTION AND LIMITATIONS OF THIS RESEARCH

The primary utility of a theory lies in its generality and transferability. Ohm’s law in electricity would have little usefulness if it applied only to one specific electrical circuit and another to be discovered for the next circuit. (Forrester, 1983 6)

We frequently face problems that resist policy interventions and challenge our abilities to address them. Such problems are in ecology, business, economy,

10-13 engineering, agriculture or at the boundary of all those disciplines. Such problems, in some situations might not be solved. The only advance we can make is to address these problems through refined strategies for a better future. In such situations, methodologies and their philosophical underpinning become important as these help us:

 in identifying better ways of thinking than that which resulted in the current problem;  in ensuring rigour in thinking, analysis and strategy development; and  by provide practitioners a suite of improved methods

Unlike previous work, this research activity proffers a multi-methodology approach to enable better understanding of the dynamic complexity faced in managing the dryland salinity problem in the Murray Darling Basin. This is considered to be an important contribution to dryland salinity research and System Dynamics body of knowledge because this analysis uses a feedback systems approach to enhance our understanding of dryland salinity. To accomplish this, the following further contributions were made to the extant body of knowledge:

 Development of a multi-methodology framework by utilising graphical visualisation of systems thinking using concept mapping, System Dynamics modelling and Systems Engineering together to improve human capability in addressing complex problems leading to simulation model development.  Development of the reference modes of the dryland salinity problem guided by the learning cycles approach presented by Saeed (2002);  Development of qualitative models, i.e., concept map, influence diagram and causal loops envisioned in the dryland salinity problem;  Development of a strategic System Dynamics model of the dryland salinity problem (a detailed framework has been proposed and recommended for future research);  Review of System Dynamics method and enhancement to that method, particularly in the rigour of model development; and  Synergistic use of System Dynamics and Systems Engineering approaches to improve System Dynamics modelling rigour.

10-14 The outcome of this research was limited by the scope of study as described in Chapter 1 (pages 10 to 11), resources and availability of data as described in Chapter 6 (pages 7 to 8). Limitations of individual findings are discussed in the preceding section along with findings.

10.3 A PARADOX OR JUST A MATTER FOR FURTHER INQUIRY

This research has highlighted the issues underlying complex dynamic problems and the capabilities of System Dynamics in a multi-methodology framework to address such issues. System Dynamics pioneers and practitioners present it for complex dynamic problems. System Dynamics has existed for around 50 years—in 2008, the System Dynamics Society celebrated its 50th anniversary of the field of System Dynamics. Systems Thinking and Systems Engineering have also existed for a similar time-span. Why then has a suitable approach not been applied to complex dynamic problems like dryland salinity? or why are researchers and managers seemingly happy to base their knowledge generating systems on various forms of models which are a gross simplification of hydrological processes?

Churchman (1994:106) also makes a similar observation on the use of models by managers and the implementation of recommendations derived from such models and suggests:

The reason why management often resists a recommendation, no matter how strongly it is supported by models or facts, can be found in the recent history of addiction, which uses the term denial to describe how the addict builds a view of reality to defend his addictive behavior, even when the view is patently false to the experts or his friends. I'm suggesting that many managers act similarly: they adopt a view of reality that enables them to deny the facts and analyses of their professional and friendly aids.

Schoemaker (2009:88) treats this issue in a slightly different way and relates it the selection of information by a manager for a certain purpose. In such situations, weak signals (a term used by Schoemaker to refer to the level of robustness or a manager’s perceived level of robustness of information) get ignored despite their significance:

10-15 There is a major difference between taking in signals and realizing what they mean. Managers as well as organizations tend to see the world in a certain way and confuse their mental maps with the territory. Weak signals that don’t fit are often ignored, distorted or dismissed, leaving the company exposed.

In any given week—especially lately—the popular press is full of examples of managers missing weak signals. The major problem is that managers are insufficiently aware of the cognitive and emotional biases that can cloud their judgment when interpreting weak signals. When ambiguity is high, we can easily torture the weak data until it confesses to whatever we want to believe. Countering these insidious tendencies requires leadership as well as the mastery of various tools to combat the pernicious filters that obscure and distort important weak signals.

Though Schoemaker used an example from business, the issues of the manager’s confusion between mental maps and the territories is not alone in the business world. It is also relevant in progressing understanding of complexity within strategy development for the dryland salinity problem. This issue is confounded by the years of dryland salinity modelling as it links theory to practices (the issue has been discussed in Chapter 3). Such models produce valuable knowledge about physical processes and/or the applications of economic theory. Such research provides the basis for development of mental maps (a term used by Schoemaker above) or a view of reality (a term that Churchman uses above) that researchers and managers have. Problem boundaries in such models do not cover dynamics, time delays and linkages to other processes that are important in finding out strategies to improve such situations.

Midgley (2000:104-105) suggests the use of philosophy considering that philosophical analysis can reveal assumptions hidden in the methodologies. Churchman (1994) also favours the use of philosophy for complex problems. Midgley (2000:21) also suggests philosophy, methodology and practice as mutually supportive areas of study where a problem in one may signal problems in the other two. The use of philosophical analysis in line with Midgley (2000:104-105), as

10-16 discussed in Chapter 3 may help in identifying problem boundaries and assumptions embedded in methodologies derived from theories.

Other than denial (a term used by Churchman to refer to a manager’s resistance to model based recommendations), the gap between human ability to cope with complexity (limited human cognitive ability), the complexity in our conceptualisation of dryland salinity and the complexity of the real world itself play important roles in the current state of confidence in models and recommendations derived from those models. Limited cognitive ability allows conceptualisation of certain aspects of complexity at a time. While recognising limits on human cognitive abilities, the conceptualisation of reality (including the robust and week signals) can be facilitated with certain tools as suggested by Schomaker (2009). Such tools are available in various systems disciplines such as System Dynamics, Systems Thinking and Systems Engineering. Such tools have been discussed in earlier Chapters. Chapter 6 demonstrates such a synergistic use of concept mapping, causal loop diagrams and influence diagrams and also demonstrates development of reference modes while Chapter 8 demonstrates the development of simulation models. A pluralist multi-methodology framework puts all these tools in a way to aid sequential and progressive learning about the dryland salinity problem. Here synergistic use of such tools in a framework is emphasised to facilitate problem conceptualisation recognising that different frameworks have varying strengths.

Moreover, conceptualisation of the dryland salinity problem as a dynamic problem may facilitate the consideration of dynamics, time delays, other exogenous and endogenous processes and soft variables. In summary, it broadens the problem boundary. Involvement of the decision makers into model development processes is a must, if they are not to confuse a mental map and territory. This position has also been advanced by Sterman (1994, 2001).

The issue of the uptake of System Dynamics is also important but it is not alone, other systems methodologies are facing similar issues as Churchman (1913-2004) has suggested in the above excerpt. It raises some questions around the methodology, its applications and use. Below is a brief overview of the questions that need to be addressed through further research:

10-17  Is the System Dynamics methodology and the way it is presented accessible to the cognition of the practitioners on ecological issues like dryland salinity?

When faced with massively complex problems, one needs both better ways of thinking than that which resulted in the current problem and one also needs to ensure rigour in thinking, analysis and strategy development. Kahneman (2003) suggests framing and substitution effects (details are provided in Chapter 9, pages 9-2 to 9-3) can possibly impact on one’s accessibility to and interpretation of information.

The most suitable methods may be left out from contributing towards improvement, if they are not closer to the accessibility modes of practitioners. Effectively addressing problems like dryland salinity is contingent upon development of a methodology that is richer in problem conceptualisation and is closer to the accessibility modes of the practitioners working on such problems.

A key philosophical question here is how such components of accessibility influence the methodological development, propagation and take-up of System Dynamics or is there any systemic problem with the presentation of this method. This dissertation has demonstrated the use of modelling concepts from a different discipline, i.e., Systems Engineering, to inform model verification and validation processes in System Dynamics. In this process, it encountered differences between the two approaches at a philosophical level about the way the relationship between a model and reality is perceived. These differences are described in Chapter 9.

Such differences might also exist between philosophers and practitioners of System Dynamics and those of ecological or natural resources management. And, this question is open to further research. Answers to such a question can throw more light on methodological improvements in System Dynamics for its application to complex problem involving ecology, resources and people, and also can identify the gap between the System Dynamics way of thinking and that of natural resources management.

At the practice level, there exist questions like:

10-18  What confidence do natural resources strategists and policy analysts have in System Dynamics?  Are there threats to strategies and policies derived from System Dynamics? What are those threats?

This dissertation has examined the System Dynamics methodology and has not looked into the perceptions of natural resources strategist and policies analysts about System Dynamics. But the perceptions of natural resources decision makers are essential in understanding the accessibility of System Dynamics concepts to their frameworks and their modes of learning. Such research might take the form of cognitive mapping, a methodology discussed in Chapter 7.

Research on further improvement of methodology for reference modes and as an extension of this research to other ecological problems like greenhouse gas emissions is recommended. Such an agenda of research should be viewed as a part of continuous modelling processes as a companion to the judgement and human decision making in line with the Forrester (1985:133-134):

In fact, for any particular real-life implementation we can expect that there will be a series of models simultaneously existing and simultaneously in evolution. Different models will address themselves to different issues. The various issues will evolve and become clearer. New issues will arise which require new models, or combinations of models which previously had existed separately. Rather than stressing the single- model concept, it appears that we should stress the process of modeling as a continuing companion to, and tool for, the improvement of judgment and human decision making. (Forrester,1985:133-134)

10-19 Bibliography

(2003). Oxford English Dictionary, electronic resource, http://oxforddictionaries.com

Abadi, A. and D. J. Pannell (1998). Bioeconomic modelling with end-users in mind. Bioeconomics Workshop, University of New England, Armidale, NSW.

ABARE (1996). Costs of Salinity to government agencies and public utilities in the Murray Darling Basin, Research Report No. 96.2, Australian Bureau of Agricultural and Resource Economics, GPO Box 1563 Canberra ACT 2601, Australia.

ABARE, (2001). Alternative Policy Approaches to Natural Resources Management, Australian Bureau of Agricultural and Resource Economics GPO Box 1563, Canberra 2601, Australia.

Abrol, I. P., Yadav, J. S. P. and Massoud, F. I. (1988). Salt affected soils and their management.,FAO Soils Bulletin 39, Food and Agriculture Organization (FAO) of the United Nations, Rome.

Abu-Zeid, K. M., M. N. Bayoumi, et al. 2000, Assessment of Environmental Impacts for Irrigation Projects, A Decision Support System. Cape Town - South Africa, The World Commission on Dams.

Ackerman, F., T. M. Williams, et al. (1994). Using Cognitive Mapping to Develop a Large Forensic System Dynamics Model. 1994 International System Dynamics Conference, Sterling, Scotland, System Dynamics Society.

Ackerman, J. and A. duPlessis (2000). Systems Thinking And Scenario Planning: A South African Case Study Of The Strategic Transformation Of A Manufacturing Company. 18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Ackermann, F., D. Andersen, et al. (2000). The TPI Framework: An Integrated Approach To Strategy Development. 18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Bibliography-1 Ackoff, R. (1979). ‘The Future of Operational Research is Past’ Journal of the Operational Research Society 30: 93-104.

Ackoff, R. (1981). Creating the Corporate Future, John Wiley & Sons, Chichester.

Acworth, R. I. and Jankowski, J. (2001). ‘Salt Sources for Dryland Salinity-evidence from an upland catchment and southern tablelands of New South Wales’, Australian Journal of Soil Research, 39, 39-59.

Adelman, L. (1992). Evaluating Decision Support and Expert Systems. New York, John Wiley and Sons.

AFFA (2001a). Overview of Water Reform Policy, Agriculture, Fisheries and Forestry Australia.

AFFA (2001b). Our Vital Resources: National Action Plan for Salinity and Water Quality. Agriculture, Fisheries and Forestry Australia, (website: www.affa.gov.au).

AGO (2000). Land Clearing: A social History, The Australian Greenhouse Office.

AGO (2002). Greenhouse Gas Emissions from Land use Change in Australia: Results of the National Carbon Accounting System. Canberra, Australia, Australian Greenhouse Office.

Albin, S. (1997). “Building A System Dynamics Model, Part 1: Conceptualization”, In Road Maps 8, MIT, USA.

Alexander, F. and A. Heaney (2003). Potential Impact of Saline Irrigation Water on the Grape Industry in the Murray Darling Basin, , The Australian Bureau of Agricultural and Resource Economics, Canberra, Australia.

Alley, W. M., Reilly, T. E. and Franke, O. L. (1991). Sustainability of ground-water resources (U.S. Geological Survey circular: 1186)., US Geological Survey, Denver, USA.

American Society of Civil Engineers (1990). Agricultural Salinity Assessment and Management, New York, USA.

Bibliography-2 Andriole, S. J. (1989). Handbook of Decision Support Systems, Tab Publishers, Blue Ridge Summit, Penn.

Anon (2004). Insights: Case studies on how farmers are successfully managing saltland for profit and sustainability, Land and Water Australia.

Anonymous (1991). “The New System Dynamics Lab: A Cooperative Computing Venture”. The University (of Vermont). Computing Newsletter 15(2): 1-3.

Anonymous (1996). Salinity: A Situation Statement for Western Australia, A Report to the Minister for Primary Industry, Minister for the Environment, Government of Western Australia Perth.

Ayers, R. S. and Westcot, W. S. (1989). Water Quality for Agriculture Food and Agriculture Organization of the United Nations, Rome, pp. 174.

Bann, G. R. and Field, J. B. (2006). “Dryland Salinity in South East Australia: Which Scenario makes more sense?” In proc. of Australian Earth Sciences Convention (AESC)., Melbourne.

Baoguo, W 1997 Prediction of Sand Production from Oil Wells, M.Sc. Thesis, Michigan Technological University, Houston, MI, USA.

Barlas, Y. (1989 b). “Tests of Model Behavior that Can Detect Structural Flaws: Demonstration with Simulation Experiments” Computer-Based Management of Complex Systems: International System Dynamics Conference, Stuttgart, International System Dynamics Society.

Barlas, Y. (1989a). "Multiple Tests for Validation of System Dynamics Type of Simulation Models." European J. of Operations Research 42(1): 59-87.

Barlas, Y. (1994). “Model Validation in System Dynamics”, In Proc. International System Dynamics Conference, Sterling, Scotland, System Dynamics Society.

Barlas, Y. (1996). "Formal Aspects of Model Validity and Validation in System Dynamics." System Dynamics Review 12(3): 183-210.

Bibliography-3 Barlas, Y. (1998). Presidential address to the 16th International System Dynamics Society Conference, Quebec, Canada, System Dynamics Society.

Barlas, Y. and K. Kanar (2000). “Structure-oriented behavior tests in model validation. In Proc.18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Barlas, Y. and S. Carpenter (1987). “A fundamental philosophy of science questions and validation of system dynamics models”. International Conference of the System Dynamics Society, Shanghai, International System Dynamics Society.

Barlas, Y. and S. Carpenter (1990). "Philosophical roots of model validation: two paradigms." System Dynamics Review 6(2): 148-166.

Barnett, S.R. (2000). Extent and Impacts of Dryland Salinity in South Australia, PIRSA Report Book 2000/00045, Department of Water Resources.

Barr, N. (1999). “Social aspects of natural resources management” In Proceedings of the National Agricultural and Resource Outlook Conference, Canberra 17-18 March , Vol.1, Commodity Markets and Resource Management, Australian Bureau of Agricultural and Resource Economics, Canberra, pp133-140.

Beare, S & A. Heaney (2001). Irrigation, Water Quality and Water Rights in the Murray Darling Basin, Australia, ABARE Conference Paper 2001.15 (www.abareconomics.com).

Beer S. (1981). Brain of a firm, Wiley, Chichester.

Bell, J. A. and Bell, J. F. (1980). “System Dynamics and Scientific Method.” In Elements of the System Dynamics Method(Ed, Randers, J.). Productivity Press, Cambridge MA, pp. 3-21.

Bell, R & Beare, S. (2000). “Salinity targets in the Murray Darling Basin”, In Australian Commodities, Vol.7, no.2., Australian Bureau of Resource and Agricultural Economics, canberra, Australia.

Bibliography-4 Bell, R & Heaney A. (2000). A Basin Scale Model for Assessing Salinity Management Options: Model documentation, ABARE Technical Working Paper 2000.1 (www.abareconomics.com).

Bellinger, G. (2004). Archetypes: Interaction Structures of the Universe.

Bhatt, D. and J. Zaveri (2002). "The enabling role of decision support systems in organizational learning." Decision Support Systems 32(3): 297-309.

Biggs, A. J. W. and Mottram, P. (2008). "Links between dryland salinity, mosquito vectors, and Ross River Virus disease in Southern Inland Queensland - an example and potential implications." Australian Journal of Soil Research, 46, 62-66.

Blanchard, B. S. (2004). Systems engineering management, John Wiley, Hoboken, NJ.

Bockstael N., Costanza, R., Strand, I., Boynton, W., Bell, K., Waingner, L. (1995). “Ecological Economic Modelling and Valuation of Ecosystems”, Ecological Economics (14): 143-159.

BOM (2003 a). Annual Mean Rainfall, Bureau of Meteorology, Commonwealth of Australia. 2003.

BOM (2003 b). Regional Temperature Trends, Bureau of Meteorology, Commonwealth of Australia. 2003.

Borenstein, D. (1998). Towards a Practical Method to Validate Decision Support Systems, Decision Support Systems 23 (1998).227-239.

Bourguet, R. E. and R. Soto (2002). Qualitative Knowledge Acquisition Using Fuzzy Logic and System Dynamics. Proceedings of the 20th International Conference of the System Dynamics Society, Palermo, Italy, The System Dynamics Society.

Brady, N.C.and Weil, R. R. 1996 The Nature and Properties of Soils, 11th ed. Prentice Hall.

Bibliography-5 Brown, G. S. (1992). Improving education in public schools: innovative teachers to the rescue, System Dynamics Review, 8, 83-90.

BRS (1998). GIS Database of Murray Basin Hydrogeology. Bureau of Resource Science, Commonwealth of Australia.

BRS (1999). Salinity Problem, Bureau of Rural Sciences, Canberra.

BRS (2000). Landcover Change in Australia. Canberra, Bureau of Resource Science.

Bunge, M. (1967). Scientific Research. Springer-Verlag. New York, USA.

Buol, S. W., Hole, F. D. and R.J., M. (1980). Soil genesis and classification, Ames: Iowa State University Press,.

Burrell, G. and Morgan, G. (1985). Sociological Paradigms and Organisational Analysis: Elements of the Sociology of Corporate Life, Gower, Aldershot, UK.

Burrell, G., & Morgan, G. (1979). Sociological Paradigms and Organizational Analysis: Elements of the Sociology of Corporate Life. London

Cameron, M. A. (2000). A Problem Model for Decision Support Systems. Canberra, Australian National University: 250.

Cavana, R. Y. and E. D. Mares (2004). "Integrating critical thinking and systems thinking: from premises to causal loops." System Dynamics Review 20(3): 223-235.

CEC (1998). Assessment of transport Strategies: Design and Specifications of a System Dynamics Model. D2, Project No: ST-97-SC.1049 Commission for European Communities.

CEC 1998 Assessment of transport Strategies: Design and Specifications of a System Dynamics Model. D2, Project No: ST-97-SC.1049 Commission for European Communities.

Checkland, P. (1981). Systems Thinking, Systems Practice, John Wiley and Sons.

Bibliography-6 Checkland, P. and Scholes, J. (1990). Soft Systems Methodology in action, John Wiley and Sons, Chichester.

Churchman, C. W. (1994). "Management Science: Science of Managing and Managing of Science." Interfaces 24(4): 99-110.

Clark, T. D., Jr. and H. Kurono (1995). “A Conversion Table of DYNAMO into STELLA II”. In Proc. System Dynamics Society Conference, Tokyo, Japan.

CLE (1992). "Brattleboro Union High School-A Start Down the Road." The Creative Learning Exchange 1(2): 1, 3-4.

Clemson, B., Y. Tang, et al. (1995). "Efficient Methods for Sensitivity Analysis." System Dynamics Review 11(1): 31-49.

Close, A. F. (1989). River Murray Water Resources Planning, In Proc. National Workshop on Planning and Management of Water Resources Systems, Department of Primary Industries and Energy.

Close, A.F. (1986). “Computer Modelling of the River Murray, Proc. River Basin Management”, Hydrology and Water Resources Symposium, Brisbane, 25-27 November 1986.

Colebatch, H.K. (1998). Policy, University of Minnesota Press, Minneapolis.

Coram, J. E., Dyson, P. R., Houlder, P. A. and Evans, W. R. (2001). Australian Groundwater Flow Systems contributing to Dryland Salinity, National Land and Water Resources Audit.

Corben, D. A. (1994). “Integrating Archetypes and Generic Models into a Framework for Model Conceptualisation”. 1994 International System Dynamics Conference, Sterling, Scotland, System Dynamics Society.

Coyle, G. 1999 Qualitative modelling in system Dynamics or What or the wise limits of Quantification. Keynote address to the International system Dynamics Society Conference, Wellington, New Zealand.

Bibliography-7 Coyle, R G. 2001 System Dynamics Review Vol. 17, No. 4, (Winter 2001): 357– 363

Coyle, R. G. (2000). "Qualitative and Quantitative Modelling in System Dynamics: Some Research Questions." System Dynamics Review 16(3): 225-144.

Coyle, R. G. (1996). System Dynamics Modelling: A Practical Approach. London, Chapman and Hall.

Coyle, R. G. (2001). "Rejoinder to Homer and Oliva." System Dynamics Review 17(4): 357-363.

Coyle, R. G. (2002). A Possible Method for Assessing the Relative Values of Alternative System Dynamics Models. Proceedings of the 20th International Conference of the System Dynamics Society, Palermo, Italy, The System Dynamics Society.

Coyle, R. G. and Exelby, D. (2000). “The validation of commercial system dynamics models”, System Dynamics Review, 16, 27-41.

Crabb, P. (1997). Murray Darling Basin Resources, Murray Darling Basin Commission, Canberra, Australia.

CRC for Soil and Land Management (1999). The Costs of Sodicity, Acidity and Alkalinity to Local and State Governments and Agencies in the Murray Darling Basin, PBM2, Glen Osmand 5064, Australia.

CSIRO (2000) Management of Dryland Salinity: Future Strategic Directions in the Context of National Policy for Natural Resources Management., Commonwealth Scientific and Industrial Research Organization, Carlingford, Victoria, Australia.

Daamen, C. C., G. P. Hoxley, et al. (2002). "Evaluation of land-management options for dryland salinity: the next step after airborne geophysical interpretation." Exploration Geophysics 33: 127-135.

Bibliography-8 Dawes, W. R., M. Gilfedder, et al. (2002). "Assessing the viability of recharge reduction for dryland salinity control: Wanilla, Eyre Peninsula." Australian Journal of Soil Research 40(8): 1407-1424.

Dawes, W.R., Walker, G.R. and Stauffacher, M. (2001). Practical Modelling in Data- Limited Catchments, Mathematical and Computer Modelling 33(2001).625-533.

Department of Agricultural and Rural Affairs (1989). Salinity: the underground flood: Can it be controlled in Shepparton area.

Department of Energy (1979). National Energy Plan II. DOE/TIC-10203 Washington DC.

Department of Sustainability and Environment (2008). Dryland Salinity.

DLWC (2000). NSW Dryland Salinity, Department of Land and Water Conservation, NSW, Australia.

DSMC (1990). Systems engineering management guide. Washington D C, Defence Systems Management College, US Government Printing Office.

Dubrovsky, V. (2004). "Toward System Principles: General System Theory and the Alternative Approach." Systems Research and Behavioural Science, 21, 109-122.

Eberlein, R. L. (1989). “Simplification and understanding of models”, System Dynamics Review, 5, 51-68.

Eberlein, R. L. (2009). System Dynamics Software info., electronic resource, http://www.vensim.com/sdmail/sdsoft.html

Eden, C. (1988). “Cognitive mapping a review”, European J. of Operational Research, 36, 1-13.

Eden, C. (1992). “On the nature of cognitive maps”, Management Studies, 29, 261- 265.

Bibliography-9 Eden, C. (1994). “Cognitive Mapping and Problem Structuring for System Dynamics Model Building”, System Dynamics Review, 10, 257-276.

Eden, C. (2004). Analysing cognitive maps to help structure the problems, European J. of Operational Research, 159, 673-686.

Eden, C. and Ackermann, F. (1998). Making Strategy: the journey of strategic management, Sage Publications.

Eden, C. and Ackermann, F. (2004). “Cognitive mapping expert views for policy analysis in the public sector”, European J. of Operational Research, 152, 615-630.

Ellis, R. E. (2001). “The Use of System Dynamics in a Large-Scale System Engineering Effort: Case Study of the Deepwater Project”, In Proc.The 19th International Conference of the System Dynamics Society, Atlanta, Georgia, System Dynamics Society.

Encyclopædia Britannica (2005). Encyclopædia Britannica Online.

ENS (2001). Audit Reveals Australia’s Most Serious Green Issue, Environmental News Service (www.ens.com/ens/apr2001/2001L-04-10-11.htm).

Evans, W. R., B. Newman, et al. (1996). “Groundwater and Salinisation in the Murray-Darling Basin”, Chapter 4 in Managing Australia's Inland Waters. Roles for Science and Technology. Canberra, Prime Minister's Science and Engineering Council, Department of Industry, Science and Tourism: 72-88.

Every, P. (1998). Incommensurability, Relativism, Methodology. A web publication, Coventry School of Art and Design, UK.

FAO (2000). Extent and causes of salt-affected soils in participating countries., Food and Agriculture organization of United Nations.

Faulconbridge, R. I. and Magpie Applied Technology. (2001). Systems engineering: body of knowledge, Magpie Applied Technology, Kingston, ACT, Australia.

Bibliography-10 Faulconbridge, R. I. and Ryan, M. J. (2003). Managing complex technical projects: a systems engineering approach, Artech House, Boston, Mass.

Faulconbridge, R. I. and Ryan, M. J. (2005). Engineering a System: Managing complex technical projects, Argos Press.

Ferdowsian, R., Pannell, D.J., McCaron, C., Ryder, A. and Crossing, L. (2001). “Explaining groundwater hydrographs: Separating atypical rainfall events from time trends”, Australian Journal of Soil Research, 39, 861-875.

Ferdowsian, R., Ryder, A. and Kelly, J. (1997). Agriculture Western Australia, Perth, WA.

Feyerabend, P. (1978). Science in a Free Society. NBL. London.

Fisher, D. M. (1999). Lessons for a First Course in System Dynamics Modeling Using the STELLA Software. Tigard, OR, Summer Creek Press.

Ford, A. (1999). Modelling the Environment: An Introduction to system dynamics models of the environmental systems, Island Press, Washington D.C.

Ford, D. N. and J. D. Sterman (1998). "Expert Knowledge Elicitation to Improve Formal and Mental Models." System Dynamics Review 14(4): 309-340.

Ford, D. N. and Sterman, J. D. (1997). Expert Knowledge Elicitation to Improve Mental and Formal Models, In 15th International System Dynamics Conference: "Systems Approach to Learning and Education into the 21st Century", Vol. 1 (Eds, Barlas, Y., Diker, V. G. and Polat, S.). Bogazici University Printing Office, Istanbul, Turkey, pp. 3-6.

Forrester J.W. 1992 Policies, Decisions and Information Sources for Modeling, European Journal of operations Research 59 (1): 42-60.

Forrester, J W 2007 System dynamics—the next fifty years, System Dynamics Review Vol. 23, No. 2/3, (Summer/Fall 2007): 359–370

Bibliography-11 Forrester, J. W. (1994 a). "System Dynamics, Systems Thinking, and Soft OR." System Dynamics Review 10(2/3): 245-256.

Forrester, J. W. (1961). Industrial Dynamics. Cambridge, MA, USA, MIT Press:.

Forrester, J. W. (1963). “Industrial Dynamics.” In The Encyclopaedia of Management. (Eds.)C. Heyel. New York, Reinhold Publishing Company: 313-319.

Forrester, J. W. (1968). Principles of Systems, Productivity Press, Cambridge MA.

Forrester, J. W. (1969). Urban Dynamics. Cambridge MA, Productivity Press.

Forrester, J. W. (1971). World Dynamics. Massachusetts 02142, USA, Wright Allen Press.

Forrester, J. W. (1975). Collected Papers of Jay W. Forrester, Productivity Press, Cambridge MA, pp. 211-244.

Forrester, J. W. (1980). “Information Sources for Modeling the National Economy.” Journal of the American Statistical Association: 75, 555-574.

Forrester, J. W. (1983). “Computer Models in Education for the 21st Century.” In Nebraska Alumnus Magazine, Vol. 79, pp. 30-31.

Forrester, J. W. (1985). "'The' Model Versus a Modeling 'Process'." System Dynamics Review 1(1): 133-134.

Forrester, J. W. (1989). The Beginning of System Dynamics, System Dynamics Group, Sloan School, Massachusetts Institute of Technology.

Forrester, J. W. (1994 b) Learning Through System Dynamics as Preparation for 21st Century, Keynote Address for the Systems Thinking and Dynamic Modeling Conference for K-12 Education, 27-29, 1994 at Concord Academy, MA,USA.

Forrester, J. W. (1995). Designing Corporations for Success in the 21st Century, Pegasus Communications, Waltham, MA.

Bibliography-12 Forrester, J. W. (2001). System Dynamics Definition. System Dynamics in the Elevator, Stwardship Modeling.

Forrester, J. W. and Senge, P. M. (1980). “Tests for Building Confidence in System Dynamics Models.” In System Dynamics, Vol. 14 (Eds, Legasto, A. A., Jr. and al., e.). North-Holland, New York, pp. 209-228.

Forsberg, K., H. Mooz, et al. (2005). Visualizing project management: models and frameworks for mastering complex systems. Hoboken, Wiley.

Forsberg, K., H. Mooz, et al. (2000). Visualizing project management: a model for business and technical success. New York ; Chichester, Wiley.

Forsberg, K., Mooz, H. and Cotterman, H. (1996). Visualizing project management, Wiley, New York ; Chichester.

Forsberg, K., Mooz, H. and Cotterman, H. (2005). Visualizing project management: models and frameworks for mastering complex systems, Wiley, Hoboken.

Forsberg, Kevin and H. Mooz (1994). The Relationship of System Engineering to the Project Cycle. The 12th Internet World Congress on Project Management. Oslo, Norway, Centre for Systems Management.

Fox, J. and S. Das (2000). Safe and Sound. Artificial Intelligence in Hazardous Application, AAAI/The MIT Press.

Fu, M. (1994). “Optimization vs Simulation: A review”, Annals of Operations Research No. 53.

Gains, B. and M. Shaw (1995). "Concept maps as hypermedia components", International Journal of Human-Computer Studies 43(3): 323-361.

Gartner, J. A. (1991). A set of lectures for 30 hour course on a systems approach to agricultural development, Asian Institute of Technology, Bangkok, Thailand.

Bibliography-13 Gary, M.S, Kunc, M., Morecraft, J. D. and S.F. Rockart (2009). System Dynamics and Strategy, System Dynamics Review, 24 (4) 407-429.

George, R. J., McFarlane, D. J. a. and Nulsen, R. A. (1997). “Salinity threatens the viability of agriculture and ecosystems in Western Australia”, Hydrogeology, 5, 6- 21.

Ghassemi, F., Close, A.F. and Kellett, J.R. (1995a)“Numerical Methods for Management of Land and Water Resources Salinization”. Proc. International Congress on Modelling and Simulations, 3: 10-16.

Ghassemi, F., Jakeman, A. J. and Nix, H. A. (1995 b). Salinization of Land and Water Resources, CRES, Australian National University, Canberra, ACT, Australia.

GHD (1970) Murray Valley Salinity Investigations, Gutteridge, Haskins and Davey, River Murray Commission, Australia.

Gill, B. (2006). Engineering Options for Dryland Salinity - Are they a silver bullet?, In Australian Earth Sciences Convention (AESC), Melbourne.

Gilmore, J.K., Croke, B.F. and Watson W.,D. (2000) An integrated Modelling Approach for Assessing Water Allocation Rules, Proc. 3rd International Hydrology and Water Resources Symposium of the Insitute of Engineers, Australia, Vol.2 pp947-952.

Glass-Husain, W. (2000). Introduction to System Dynamics modeling. 18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Golafshani, N. (2003). Understanding reliability and validity in qualitative research. , The Qualitative Report, 8, 597-606.

Gordon, l., Dunlop, M. and Foran, B. (2003). Philosophical Transactions: Biological Sciences, 358, 1973-1984.

Bibliography-14 Graetz, R.D. Wilson, M.A. and S.K. Campbell (1995) Landcover Disturbance over the Australian Continent: A contemporary Assessment. Biodiversity series paper no. 7, Department of the Environment, Sport and Territories, Commonwealth of Australia.

Greenburger, M., Crenson, M. A. and Crissey, B. L. (1976). Models in the Policy Process: Public Decision Making in the Computer Era, Russell Sage Foundation, New York.

Grobler, A. (2008). "System Dynamics Modelling as an Inductive and Deductive Endeavour. Comment on the Paper by Schwaninger and Gro¨sser." Systems Research and Behavioral Science 25: 467-470.

Grobler, A., P. M. Milling, et al. (2004). "Perspectives on rationality in system dynamics—a workshop report and open research questions." System Dynamics Review 20: 75-87.

GRSPPAC (1989) Shepparton Irrigation Region Land and Water Management Plan, Goulburn Region Salinity Pilot Program Advisory Council, Departments of Government of Victoria and Dyer Leslie Pty Ltd., Australia.

Guba, E.G. (1990) The Paradigm Dialogue, Sage Publications, Newbury Park, London, New Delhi.

Habermas, J. (1984). The theory of communicative action, Beacon Press, Boston.

Hall, A. D. (1962). A methodology for systems engineering, Van Nostrand, Princeton, N.J.

Hammond, P.H. (1958). Feedback Theory and its Applications, The English Universities Press Ltd, 102 Newgate Street, London, UK.

Hardin, G. (1968). The Tragedy of the Commons, Science, 162(1968):1243-1248.

Harich, J. (2001). A New Approach to Solving Sustainability Problem, Electronic Resource, Thwink.org.

Bibliography-15 Harries, K. (1999). Mapping Crime: Principle and Practice.

Heaney A, & S. Beare (2001). Property Rights and Externalities in Water Trade, ABARE Conference Paper 2001.16 (www.abareconomics.com).

Heaney, A. and R. Bell (2001). A Basin Scale Model for Assessing Salinity Management Options: Model Documetnation. Canberra, Australian Bureau of Agricultural and Resource Economics.

Heaney, A., S. Beare, et al. (2001). "Evaluating improvements in Irrigation efficiency as a salinity mitigation option in th eSouth Australian Riverland." Australian Jounal of Agricultural and Resource Economics 45(3): 477-493.

Helsinki School of Economics (1981). Dynamic or Dynamic Hypothesis? The System Dynamics Research Conference, Institute of Man and Science, Rensselaerville, NY.

Hingston, F. J. and Gailitis, V. (1976). “The geographic variation of salt precipitation over Western Australia”, Australian Journal of Soil Research,, 14, 319-335.

Hoepfl, M. C. (1997). “Choosing Qualitative Research: A Primer for Technology Education Researchers”, Journal of Technology Education, 9, 47-63.

Homer, J. B. (1992). A System Dynamics Model of National Cocaine Prevalence, System Dynamics Review, 9, 49-78.

Homer, J. B. (1996). “Why We Iterate: Scientific Modeling in Theory and Practice”, System Dynamic Review, 12, 1-19.

Homer, J. B. and R. Oliva (2001). "Maps and Models in System Dynamics: A Response to Coyle." System Dynamics Review 17(4): 347-355.

Hooper, B. P. (1996). “The concept of Integrated Resource Management and its applications to the Dawson Valley, Queensland”, A paper presented to the Future Developments in the Dawson Valley Conference, 17-19 April 1996, Wandowan, Dueensland.

Bibliography-16 Houghton, M. C. (2000). The American Heritage® Dictionary of the English Language, Houghton Mifflin Company.

Hovmand, P. (2001). Community Assessment Tool: An Evaluation of Websims for Violence Prevention. The 19th International Conference of the System Dynamics Society, Atlanta, Georgia, System Dynamics Society.

Howden, M., K. Hartle, et al. (2003). Climate Change, Climate Variability and Water. Water Matters Conference, Parliament House, Canberra.

HRSCSI (2004). Science Overcoming Salinity: Coordinating and extending the science to address the nation’s salinity problem, Commonwealth of Australia, House of Representatives Standing Committee on Science and Innovation.

Humphrey, C. 1997 Development and Implementation of QA/QC Protocols for Sample Processing Components of the MHRI Agency Bioassessment Program, Land and Water Resources Research and Development Corporation, Canberra ACT, Australia.

Humphreys, P., Ed. (1996). Implementing Systems for Supporting Management Decisions: Concepts, Methods and Experiences. London, Chapman & Hall.

INCOSE (2007). Systems Engineering Handbook Version 3.1, International Council On Systems Engineering.

INCOSE (2007). Systems Engineering Handbook Version 3.1, International Council On Systems Engineering.

Isbell, R. (1996). Australian Soil Classification, CSIRO, Canberra.

Jackson, K. and W. Trochim (2002). "Concept mapping as an alternative approach for the analysis of open-ended survey responses." Organizational Research Methods 5(4): 307-336.

Jackson, M. C. (1995). "Beyond the Fads: Systems Thinking for Managers'." Systems Research Vol. 12(No. 1): 25-42.

Bibliography-17 Jackson, M. C. (1987). In New Directions in Management Science(Eds, Jackson, M. C. and Keys, P.). Aldershot, Gower, pp. 131-162.

Jackson, M. C. (1995). “Beyond the Fads: Systems Thinking for Managers”, Systems Research, Vol. 12, 25-42.

Jardine, A., P. Speldewinde, et al. (2007). "Dryland Salinity and Ecosystem Distress Syndrome: Human Health Implications." EcoHealth 4: 10 -17.

Jardine, A., P. Weinstein, et al. (2006). "The impact of dryland salinity and waterlogging on mosquito-borne disease in Western Australia." Epidemiology 17(6): S132-S133.

Johnson, S. T. and R. L. Eberlein (2002). “Alternative modeling approaches: A case study in the oil and gas industry”, Proceedings of the 20th International Conference of the System Dynamics Society, Palermo, Italy, The System Dynamics Society.

Kahneman, D, 2003 Maps of Bounded Rationality: Psychology for Behavioral Economics, The American Economic Review, Vol. 93, No. 5 (Dec., 2003), pp. 1449- 1475.

Keiller, D.C. and Close, A.F. (1985) Modeling Salt Transport in a Run River System, proc. 21st IAHR Congress, Melbourne Australia 19-23 August 1985.

Kelly, G. A. (1955). The Psychology of Personal Constructs, Norton, New York.

Kepner, C. H. and B. B. Tregoe (1981). The new rational manager. Princeton, N.J. [P.O. Box 704, Research Rd., Princeton 08540], Princeton Research Press.

Keys, P. (1988). “System dynamics: A methodological perspective”, Transactions of the Institute of Measurement and Control 10(4)., 218-224.

Keys, P. (1990). "System Dynamics as a Systems-Based Problem-Solving Methodology." Systems Practice 3: 479-493.

Bibliography-18 Keyworth, S.W. (1996). “Sustainable irrigation in the Murray-Darling Basin”, In. Focus (Australian Academy of Technological Sciences and Engineering), No. 90, 5- 9.

Khamesan, A. (2005). Software of concept mapping.

Khan, H. (2001). “Workshop: Powersim Studio: The Next Generation in System Dynamics Simulation.” In Proc. The 19th International Conference of the System Dynamics Society, Atlanta, Georgia, System Dynamics Society.

Khan, N. and A. Mclucas (2008). A Case Study in Application of Vee Model of Systems Engineering to System Dynamics Modelling of Dryland Salinity in Australia. In Proc.International System Dynamics Society Conference, Athens, Greece.

Khan, N.U. 1995, A Systems Analysis for Improving Economic Efficiency of Irrigation in the Indus Basin Irrigation System. M.Sc. Thesis, University of Aberdeen, UK.

Khan, S., Stubbs, C and D. McLaughlin (undated). A Regional Hydrologic Economic Framework for Investigating Sustainable Irrigated Landscape Future, CSIRO.

Kim, D. H. (1992). "System Archetypes at a Glance." The Systems Thinker 3(4): 5.

Kim, D. H. (1992). System Archetypes as a Diagnostic Tool: A Field-based Study of TQM Implementation. Proceedings of the 1992 International System Dynamics Conference of the System Dynamics Society, Utrecht, the Netherlands, The System Dynamics Society.

Kim, D. H. (1995). "Systems Archetypes as Dynamic Theories." The Systems Thinker 6(5).

Kim, D. H. (1995). "Using Archetypes as Different 'Lenses'." The Systems Thinker 6(3).

Bibliography-19 Kim, D. H. and G. Burchill (1992). Systems Archetypes as a Diagnostic Tool: A Field-based Study of TQM Implementations, MIT System Dynamics Group.

Kleijnen, J.P.C. (1995). “Sensitivity analysis and optimization of system dynamics models: regression analysis and statistical design of experiments”, System Dynamics Review 11(4): 275-288

Kleijnen, J.P.C., and E. Gaury (2000) Ethical Issues in Modelling: Some Reflections, Department of Information Systems, Center for economic research, Tilburg University, Netherlands.

Kleijnen, J.P.C., and E. Gaury (2001). Optimization versus Robustness in Simulation: A Practical Methodology, with a Production-management Case Study, Department of Information Systems, Center for economic research, Tilburg University, Netherlands.

Kleindorfer, G.B., O'Neill, L., Ram Ganeshan, R., 1998 Validation in Simulation: Various Positions in the Philosophy of Science, Management Science, Vol. 44, No. 8 (Aug., 1998), pp. 1087-1099.

Kline, S. J. (1995). Conceptual Foundations for Multidisciplinary Thinking, Stanford University Press, Stanford, California, USA.

Kolb, D. A. (1984). Experiential Learning, Englewood Cliffs, NJ: Prentice Hall.

Kuhn, T. S. (1984). “Paradigm, Tacit Knowledge and Incommensurability”, in Modernism, Criticism, Realism.

Kuhn, T.S. 1984 Paradigm, Tacit Knowledge and Incommensurability, in Modernism, Criticism, Realism.

Kuhn, T.S., (1970). The Structure of Scientific Revolutions, University of Chicago Press, Chicago, USA.

Kumm, J. (1975). Economic Growth - Environmental Protection – Quality of life, Deutsche Verlags-Anstalt, Stuttgart.

Bibliography-20 Lane, D. (2001). “Rerum cognoscere causas: Part I - How do the ideas of system dynamics relate to traditional social theories and the voluntarism/determinism debate?”, System Dynamics Review, 17, 97-118.

Lane, D. C. (1990). "Modelling as Learning." Strategic Planning in Shell Series 9.

Lane, D. C. (2000). "Should System Dynamics Be Described As A 'Hard' Or 'Deterministic' Systems Approach?" Systems Research and Behavioral Science 17(1): 3-22.

Lane, D. C. (2000). “Should System Dynamics Be Described As A 'Hard' Or 'Deterministic' Systems Approach?” Systems Research and Behavioral Science 17(1): 3-22.

Lane, D. C. (2000). How Accurate Should Model Output Be? Three Questions For Model Builders. Proc. 18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Lane, D. C. (2008). “The Emergence and Use of Diagramming in System Dynamics: A Critical Account”, Systems Research and Behavioral Science, 25, 3-23.

Lane, D.C. (1999). “Social Theory and System Dynamics Practice”, European Journal of Operational Research 113(1999).501-527.

Lane, D.C. and Oliva, R. (1998). “The Greater Whole: Towards a Synthesis of System Dynamics and Soft Systems Methodology”, European Journal of Operations Research 107 (1998). 214-235

Letcher, R. (2002). Issues in integrated assessment and modelling for catchment management, Ph D thesis, Australian National University, Canberra: 318.

Letcher, R. (2002). Issues in integrated assessment and modelling for catchment management, Ph D thesis, Australian National University, Canberra: 318.

Bibliography-21 Linard, K. T, J. Yoon, et al. (2000). “A Dynamic Balanced Scorecard Template For Public Sector Agencies”, Proc. 18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Linard, K. T. (1996). System dynamics and systems thinking in defence & government Australian Defence Force Academy, Canberra, A.C.T, Australia

Linard, K.T. 1996 System dynamics and systems thinking in defence & government Australian Defence Force Academy, Canberra, A.C.T, Australia

Linard, K.T. and White A.S. (1991). Project Management in Public Administration. Australian Defence Force Academy, Canberra, ACT, Australia.

Linard, K.T. and White A.S. 1991 Project Management in Public Administration. Australian Defence Force Academy, Canberra, ACT, Australia.

Lindsay, M.D.A., A. Jardine, C. A. Johansen, A. E. Wright, S. A. Harrington, and P. Weinstein. 2007. Mosquito (Diptera: Culicidae) fauna in inland areas of south-west Western Australia. Aust. J. Entomology. 46: 60-64.

Little, J. D. (1970). "Models and Managers: The concept of Decision Calculas." Management Science 16(8): B466-485.

Littleboy, M., R. Vertessy, et al. (2001). "An overview of modelling techniques and decision support systems and their application for managing dryland salinity."

Ludowyk, F. and Moore B. (1999). The Australian Oxford Paperback Dictionary, Oxford university Press.

LWRRDC (1998). National Dryland Salinity Program: Management Plan 1998- 2000, Land and Water Resources Research Council, Canberra, Australia

Lyneis, J. M. and Pugh, A. L. P. (1996). Automated vs 'Hand' Callibration of System Dynamics Models: An Experiment with a Simple Project Model, In 1996 International System Dynamics Conference, Vol. 1 (Eds, Richardson, G. P. and Sterman, J. D.). System Dynamics Society, Cambridge, Massachusetts, pp. 321-324.

Bibliography-22 Maier, M. and Rechtin, E. (2000). The art of systems architecting, CRC Press, Boca Raton, FL.

Maliapen, M. (2000). The Application Of Business Intelligence Tools In System Dynamics Model. In Proc. 18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Malins, D. and G. Metternicht (2006). "Assessing the spatial extent of dryland salinity through fuzzy modeling." Ecological Modeling 193: 387-411.

Marakas, G. M. (2002). Decision support systems: in the 21st century. Upper Saddle River, N.J., Prentice Hall.

Martens, D. M. and J. F. DiBiase (1996). "TCM-Manager: A PC based total catchment management decision support system." Environmental Software 11(1-3): 1-7.

Mcfarlane, D. (1991). “A review of secondary salinity in agricultural areas of Western Australia”, In Land and Water Research News, 11, 7-16.

McFarlane, D. J. and George, R. J. (1992). “Factors affecting dryland salinity in two wheatbelt catchments in Western Australia”, Australian Journal of Soil Research, 30: 85-100.

McLucas, A. C. (2001). An investigation into the integration of qualitative and quantitative techniques for addressing systemic complexity in the context of organisational strategic decision-making. PhD Thesis, School of Civil Engineering, University College. Canberra, University of New South Wales, Australia.

McLucas, A. C. (2003). Decision making: risk management, systems thinking and situation awareness. Canberra, Argos Press.

McLucas, A. C. (2005). System Dynamics Applications: A Modular Approach to Modelling Complex World Behaviour, Argos Press, Canberra, Australia.

Bibliography-23 Mclucas, A. C. and C. J. Ryan (2005). Meeting Critical Real World Challenges in Modelling complexity: What system dynamics might learn from systems engineering. In Proc. 23rd International System Dynamics Society Conference, Boston, USA, System Dynamics Society.

MDBC (1997) Salt Trends: Historic Trend in Salt Concentration and Saltload of Stream Flow in the Murray-Darling Drainage \Division, Dryland Technical Report no.1. Murray Darling Basin Commission.

MDBC (1999). The Salinity Audit of the Murray Darling Basin: A100 years perspective, Canberra, Murray Darling Basin Commission.

MDBC (1999a) Water and Land Salinity Murray Darling Basin Commission (www.mdbc.gov.au/education/encyclopaedia/water_and_land_salinity.htm ).

MDBC (1999b) Salinity Impacts Study, Final Report. Murray Darling Basin Commission (www.mdbc.gov.au ).

MDBC (2001a). Basin Salinity Management Strategy 2001-2015, Murray Darling Basin Commission. (www.mdbc.gov.au).

MDBC (2001b) Murray Darling Basin Initiative, Murray darling Basin Commission Canberra. (www.mdbc.gov.au).

MDBC (2006). Basin Statistics, Murray Darling Basin Commission, Canberra.

MDBMC 1987 Murray Darling Basin: Environmental Resources Study. Murray Darling Basin Ministerial Council Canberra, Australia.

Meadows, D. H., D. L. Meadows, J. Randers, & W.W. Behrens III (1972). The Limits to Growth,: A Report for the Club of Rome’s Project on the Predicament of Mankind. New York: Universal Books, USA.

Meadows, D. and J. Robinson (1985). "The Electronic Oracle." Computer Models and Social Decisions.

Bibliography-24 Meadows, D. H. (1989). “System Dynamics. Meets the Press”, System Dynamics Review 5(1).68-80.

Meadows, D. H., D. L. Meadows, et al. (1972). The Limits to Growth: A Report for the Club of Rome's Project on the Predicament of Mankind. New York, Universe Books.

Merriam-Webster Online Dictionary (2009). Occam's razor.

Merrium-Webster (2003). Collegiate online dictionary, Merrium- Webster.

Midgley, G. (2000). Systemic intervention: philosophy, methodology, and practice. New York, Kluwer Academic/Plenum.

Midgley, G. (2000). Systemic intervention: philosophy, methodology, and practice, Kluwer Academic/Plenum, New York.

Mingers, J. (1997). Chapter 1: Multi-paradigm multimethodology. In Multimethodology: The theory and practice of combining management science methodologies. J. Mingers and A. Gill (eds.), John Wiley and Sons.

Mingers, J. (2001). “Multimethodology - Mixing and Matching Methods”, In Rational Analysis for a problematic world revisited. J. Rosenhead and J. Mingers. (eds.) West Sussex England, John Wiley & Sons Ltd.

Mingers, J. (2006). "A critique of statistical modelling in management science from a critical realist perspective: its role within multimethodology." Journal of Operational Research Society 24(2): 202-219.

Mingers, J. and Brocklesby, J. (1997). “Multimethodology: Towards a framework for mixing methodologies”, Omega, Internation Journal of Management Science, 25, 489-509.

Mingers, J. and Gill, A. (eds.) (1997). “Multimethodology: the theory and practice of combining management science methodologies, Wiley, Chichester ; New York.

Bibliography-25 MIT (1994). Road Maps: A Guide to Learning System Dynamics, System Dynamics in Education Project. Massachusetts Institute of Technology, USA.

MIT (1994a). Glossary, in Road Maps: A Guide to Learning System Dynamics, System Dynamics in Education Project. Massachusetts Institute of Technology, USA.

Morecroft, J. D. W. (1988). "System Dynamics and Microworlds for Policymakers." European Journal of Operational Research (Netherlands). 35(3): 301-320.

Morecroft, John D. W. (1983). System Dynamics: Portraying Bounded Rationality. Omega.

Moxnes, E. (2000). “Not only Tragedy of Commons: Misperceptions of Feedback and Policies for Sustainable Development”, System Dynamics Review 16(4): 325- 348.

Muller, G. (2005). The informal nature of Systems Engineering. Fm 2005: Formal Methods, Proceedings. 3582: 552-556.

Muller, R 2008 The Concept of Model: Definitions and Types, published on Muller Science website.

Munro, I. and Mingers, J. (2000). The use of multimethodology in practice - results of a survey of practitioners, Warwick Business School, Coventry.

Myerson, G. (1994) Rhetoric, Reason and Society. Sage, London.

Naylor T.H. and J.M. Finger. 1967. Verification of Computer Simulation Models. Management Science, 14: B92-B101.

NDSP (2001). Decision Support Tool. 2008.

Neudorff, L. J., J. E. Randall, et al. (2003). Freeway Management and Operations Handbook, Washington DC, Office of transportation management, Federal Highway Administration.

Bibliography-26 New South Wales Government (2000) NSW Salinity Strategy.

NLWRA (2000). Australian Water Resources Assessment 2000, National Land and Water Resources Audit 2000,

NLWRA (2001). “Land Use Change, Productivity and Diversification”, Canberra, National Land and water Resources Audit: 155.

NLWRA (2001). Australian Dryland Salinity Assessment 2000, National Land and Water Resources Audit, Canberra, ACT, Australia.

NLWRA (2008) Land Salinity –Status of Information for Reporting Against Indicators Under the National Natural Resource Management Monitoring and Evaluation Framework , National Land and Water Resources Audit, Canberra, ACT, Australia.

Nordhaus, W. D. (1972). “World Dynamics: Measurements without Data: Cowles Foundation Discussion Paper”, Reprinted in Economic Journal, December 1973 p1156-1183

Norgaard, R.B., 1989. The case for methodological pluralism. Ecol. Econ., 1: 37-57.

Nørgard, J. S. and N. I. Meyer (1979). National Energy Systems Analysis for Denmark up to Year 2030. Intl. Conf. on Energy System Analysis, Dublin.

NRE (1999). Reversing the Trend: Northeast Salinity Strategy, Department of Natural Resources and Environment, Victoria, Australia

NRE (2000). Victoria Salinity Management Framework, Department of Natural Resources and Environment, Victoria, Australia

NSW Department of Land and Water Conservation (2000). Dryland salinity

Nuthmann C (1994) Using Human Judgement in System Dynamic Models of Social Systems, System Dynamics Review, Vol. 10, No. 1, pp1-19.

Bibliography-27 Nuthmann C. (1994). “Using Human Judgement in System Dynamic Models of Social Systems”, System Dynamics Review, 10(1):1-19.

Oldeman, L. R. (1991). Global extent of soil degradation, International Soil Reference and Information Centre. Hinkeloord, Agricultural University, Wageningen, The Netherlands.

O'Brien, R. (2001). An Overview of the Methodological Approach of Action Research]. In Roberto Richardson (Ed.), Theory and Practice of Action Research. João Pessoa, Brazil: Universidade Federal da Paraíba. (English version) Available: http://www.web.ca/~robrien/papers/arfinal.html (Accessed 15/8/2011)

Oliva, R. (1996). A Dynamic Theory of Service Delivery Implications for Service Quality, Sloan School of Management, Massachusetts Institute of Technology.

Oliva, R. (1996). Empirical Validation of a Dynamic Hypothesis. In proc. 1996 International System Dynamics Conference, Cambridge, Massachusetts, System Dynamics Society.

Oliva, R. (2001). Model Calibration as a Testing Strategy for Dynamic Hypothesis, working paper No 01-047, Version 2.0, Harvard Business School, Boston, MA 02163, USA.

Oliva, R., J. D. Sterman, et al. (2003). "Limits to Growth in the New Economy : Exploring the 'Get Big Fast' Strategy in E-Commerce." System Dynamics Review 19(2): 83-117.

Oliva, R., Sterman, J. D. and Giese, M. (2003). Limits to Growth in the New Economy: Exploring the 'Get Big Fast' Strategy in E-Commerce, System Dynamics Review, 19: 83-117.

Oreskes, N., Shreder-Frechette K., Belitz K. (1994) “Verification, Validation, and confirmation of numerical models in the earth sciences”, Science 263:641-646.

Bibliography-28 Osmond, D. L., D. E. Line, et al. (1995). WATERSHEDSS: Water, Soil and Hydro- Environmental Decision Support System,. USA, Water Quality Group, North Carolina State University.

Otto, C.J. Undershultz, J.R., Hennig, A.L. and Roy, V.J. (2001). “Hydrodynamics Analysis of Flow Systems and Fault Seal Integrity in the North West Shelf of Australia”, APPEA Journal, pp347-363

Ozbay, H. (1999). Introduction to feedback control theory. New York, USA, CRC Press LLC.

Page, T. (1973). The Non-Renewable Resources Subsystem. In HSD Cole, C. Freeman, M. Jahoda and K.L.R. Pavitt (eds.). Thinking About Future: A critique of the Limits to Growth. Sussex University Press, London.

Pannell, D. J. (2001). Dryland Salinity: Inevitable, inequitable, Intractable, presidential address, 45th Annual Conference of the Agricultural and Resource Economics Society, adelaide, 23-25 January 2001

Pannell, D. J. (2001). Australian Journal of Agricultural and Resource Economics, 45, 517-546.

Pannell, D. J. and M. A. Ewing (2006). "Managing secondary dryland salinity: Options and challenges." Agricultural Water Management 80(1-3): 41-56.

Pannell, D. J., Ewing, M. A. and Ridley, A. M. (2004). “Dryland salinity in Australia: Overview and prospects.” In Dryland Salinity: Economic Issues at Farm, Catchment and Policy Levels (Eds., Graham, T., Pannell, D. J. and White, B. J.). Cooperative Research Centre for Plant-Based Management of Dryland Salinity, Perth, Australia.

Pannell, D.J., McFarlane, D.J. and Ferdowsian, R. (2001). “Rethinking the externality issue for dryland salinity in Western Australia”, Australian Journal of Agricultural and Resource Economics 45(3): 459-475

Bibliography-29 Peck, A.J., Thomoas J.F. and Williamson D.R. (1983). Effects of Man on Salinity, Water 2000 Consultants Report Vol.8, Department of National Development and Energy, Canberra.

Pecker, D.W. and Glass-Husain, W. (1997). “Designing Interactive Multi-user Learning Laboratories.” In Proc. International System dynamics Society Conference, Istanbul, Turkey.

Perman, R., MA, Y. and McGilvray 1996 Natural resources and Environmental Economics, Longman, London and New York.

Perman, Roger, Yue Ma, James McGilvray and Michael Common (1996), Natural. Resource & Environmental Economics, Longman, London

Phillips, C. L. and R.D. Harbour (1996) Feedback Control Systems, 3rd edition, Prentice Hall International Inc.

Picardi, A. C. and K. Saeed (1979). "The Dynamics of Water Policy in Southwestern Saudi Arabia." Simulation 33(4): 109-119.

Picardi, A. C., and Saeed, K. 1979. Dynamics of Water Policy in South-western Saudi Arabia. Simulation. 33(4).

Plotnic, E. (1997). "Concept mapping a graphical system for understanding the relationship between concepts." ERIC Digest, ED407938.

PMSEIC (1999) Moving Forward in Natural Resources Management. Prime Minister’s Science, Engineering and Innovation Council, Canberra, Australia.

Powell, J. (2004). Dryland salinity: On Farm Decisions and Catchment Outcomes. Aguide for leading producers and advisors, Land and Water Australia, Canberra, ACT.

Power, D. J. (2003). A Brief history of Decision Support Systems., DSS Resources.Com, Woorld Wide Web, http://DSSResources.Com/History/DSS History.html.

Bibliography-30 Powersim Corp. (1996). Powersim 2.5 Reference Manual. Herdnon, VA20170, USA, Powersim Corporation Inc.

Powersim Corporation (2007). Powersim Studio 7 Manual.

Powersim Software AS (2003). Powesim Studio 2003: User Guide.

Prathapar, S. A., W. S. Meyer, et al. (1996). "A Soil water groundwater simulation model: SWAGSIM." Environmental Software 11(1-3): 151-158.

Pruyt, E. (2006). What is System Dynamics? A Paradigmatic Inquiry. In Proc. International System Dynamics Society Conference.

Puttnam, H. (1988) Representation and Reality., MIT Massachuesetts, USA.

Quinn, N. W. T. and W. M. Hanna (2003). "A decision support system for adaptive real-time management of seasonal wetlands in California." Environmental Modelling & Software 18(6): 503-511.

Quinn, N. W. T. and W. M. Hanna (2003). "A decision support system for adaptive real-time management of seasonal wetlands in California." Environmental Modelling & Software 18(6): 503-511.

Radzicki, M. J. (1994). Using POWERSIM to Teach Sustainable Economic Development, Department of Social Science and Policy Studies, Worcester Polytechnic Institute.

Radzicki, M. J. (2002). What is System Dynamics?, WPI, USA.

Raimondi, F. V. (2001). Organized Crime and Economic Growth: A

Randers, J. (1980). “Guidelines for Model Conceptualization.” In Elements of system dynamics method(Ed, Randers, J.). MIT Press, Cambridge, MA.

Randers, J. (2000). "From Limits to Growth to Sustainable Development or SD (Sustainable Development). in a SD (System Dynamics). Perspective." System Dynamics Review 16(3): 213-224.

Bibliography-31 Rechtin, E. (1991). Systems Architecting: Creating and Building Complex Systems, Prentice-Hall, Paramus.

Reichelt, K. S., Lyneis, J. M. and Bespolka, C. G. (1996). Calibration Statistics: Selecting a Statistic and Setting a Standard, In Proc.1996 International System Dynamics Conference, Vol. 2 (Eds, Richardson, G. P. and Sterman, J. D.). System Dynamics Society, Cambridge, Massachusetts, pp. 425-428.

Repenning, N. (2003). Selling System Dynamics to (other) Social Scientists,System Dynamics Review, 19, 4: 303-327.

Rhoades, J. D., Kandiah, A. and Mashali, A. M. (1992). The use of saline waters for crop production - FAO irrigation and drainage paper 48 Food And Agriculture Organization of The United Nations, Rome

Richardson G. & A.L. Pugh (1981). System Dynamics Modeling with Dynamo, Productivity Press, Portland, OR, USA

Richardson, G (1991) Feedback Thoughts in Social Sciences and Systems Theory. Philadephia: University of Pennsylvania Press

Richardson, G. P. (2001). “Mapping versus Modeling: The Answer to the Debate” In Proc. The 19th International Conference of the System Dynamics Society, Atlanta, Georgia, System Dynamics Society.

Richardson, G. P. (1986). "Problems with causal-loop diagrams (1976)." System Dynamics Review 2(2): 158-170.

Richardson, G. P. (1991). System Dynamics: Simulation for Policy Analysis from a Feedback Perspective, In Qualitative Simulation Modeling and Analysis. (eds) P. A. Fishwick and P. A. Luker. New York, Springer Verlag.

Richardson, G. P. (1996). "Problems for the future of system dynamics " System Dynamics Review 12(2): 141 - 157.

Bibliography-32 Richardson, G. P. (1997). "Problems in Causal Loop Diagrams Revisited." System Dynamics Review 13(3): 247-252.

Richardson, G. P. (2001). Mapping versus Modeling: The Answer to the Debate. In Proc. 19th International Conference of the System Dynamics Society, Atlanta, Georgia, System Dynamics Society.

Richardson, G. P. (2005). Model Validation as an Integrated Social Process., In Invited presentation at the Workshop on System Dynamics Modeling of Physical and Non-Physical Systems for National Security, sponsored by Booz, Allen, Hamilton, Inc., held in Washington, DC, 21-22 April 2005.

Richardson, G. P. (2006). Dynamic Complexity, In 50th Annual Meeting of International Society for the Systems Science.

Richardson, G. P. and A. L. Pugh, III (1981). Introduction to System Dynamics Modeling with DYNAMO. Cambridge MA, Productivity Press.

Richardson, G. P., D. F. Andersen, et al. (2002). Misattribution in Welfare Dynamics : The Puzzling Dynamics of Recidivism. Proceedings of the 20th International Conference of the System Dynamics Society, Palermo, Italy, The System Dynamics Society.

Rittel, H. W. J. and Webber, M. M. (1973). “Dilemmas in a general theory of planning”, Policy Sciences, 4, 155-169.

Roberts, D. W., Dowling, T. I. and Walker, J. (1997). FLAG: A fuzzy landscape Analysis GIS method for dryland salinity assessment, Technical Report No. 8/97, CSIRO, Canberra, Australia.

Roberts, N. H., D. F. Andersen, et al. (1983). Introduction to Computer Simulation: The System Dynamics Modeling Approach. Reading, MA, Addison-Wesley.

Robertson, G. (1995). In Making catchment management happen. Liverpool Plains Land Management Committee, Gunnedah.

Bibliography-33 Rodriguez-Ulloa, R. and A. Paucar-Caceres (2005). "Soft System Dynamics Methodology (SSDM): Combining Soft Systems Methodology (SSM). and System Dynamics (SD)." Systemic Practice and Action Research Vol. 18(No. 3).

Rodriguez-Ulloa, R. and Paucar-Caceres, A. (2005). Systemic Practice and Action Research, Vol. 18.

Rosenhead, J. and J. Mingers (eds.)(2001). Rational analysis for a problematic world revisited: problem structuring methods for complexity, uncertainty and conflict. Chichester ; New York, Wiley.

Saeed, K. (1998). Constructing Reference Mode. In Proc.16th International Conference of the System Dynamics Society Quebec '98, Quebec City, Canada, System Dynamics Society.

Saeed, K. (2001). Defining developmental problems for policy intervention or Building a reference moded in 20 steps over 5 learning cycles, Worcester Polytechnic Institute, Worcester, USA: 1-33.

Saeed, K. ( 2002). System Dynamics for Discerning Developmental Problems. In "Systems Dynamics: Systemic Feedback Modeling for Policy Analysis". In Encyclopaedia of Life Support Systems (EOLSS). Oxford, UK: EOLSS Publishers.

Saeed, K. (1980). Rural Development and Income Distribution: The Case of Pakistan, M. I. T.

Saeed, K. (1982). Rural Poverty and Public Policy: A System Dynamics Analysis of a Social Change Effort in Pakistan. Technological Forecasting and Social Change. 21(4).

Saeed, K. (1986). Minds over Method, System Dynamics Review No.2 150-156.

Saeed, K. (1987). "The World Hunger Problem: Do We Really Understand it?" System Dynamics Rev. 3(l): 36-44.

Bibliography-34 Saeed, K. (1994). Development Planning and Policy Design, Avebury, aldershot, USA.

Saeed, K. (2002). Articulating developmental problems for policy intervention: A system dynamics modelling approach. Worcester, USA, Worcester Polytechnic Institute, Worcester, MA 01609, USA: 1-31.

Saeed, K. (2003). Articulating developmental problems for policy intervention: A system dynamics modeling approach, Simulation & Gaming, 34, 409-436.

Sage, A. P. (2005). Systems of Systems: Architecture Based Systems Design and Integration, Department of Systems Engineering and Operations Research, George Mason University, Fairfax, VA 22030-4444.

Sage, A. P. and Olson, S. R. (2001). “Whither Simulation Based Acquisition ?Part II: Modeling and Simulation in Systems Engineering”, Simulation, 76, 283-286.

Saysel, A K and Barlas, Y. 2006 Model simplification and validation with indirect structure validity tests, System Dynamics Review, 22, 241–262.

Saysel, A. K. (1999). Dynamic Simulation Model for Long Term Comprehensive Environmental Analysis of GAP (Regional Development)., Bogazici University, Istanbul, Turkey.

Saysel, A. K. and Barals, Y. (2001). “A dynamic model of salinization on irrigated lands”, Ecological Modelling, 139, 177-199.

Saysel, A. K. and Barlas, Y. (2006). Model simplification and validation with indirect structure validity tests, System Dyamics Review, 22 (3) 241-262.

Schactman, D. E. (1994) Directory of concepts environmental informatics, USA.

Schnaars, S. and P. Ziamou (2001). "The essentials of scenario writing." Business Horizons July-August 2001.

Schoemaker, P. J. H. (1991). "When and how to use scenario planning: A heuristic approach with illustration." Journal of Forecasting 10(6): 549-564.

Bibliography-35 Schoemaker, P. J. H. (1993). "Multiple Scenario development: Its conceptual and behavioural foundation." Strategic Management Journal: 193-213.

Schoemaker, P. J. H. (1995). "Scenario planning: A tool for strategic thinking", Sloan Management Review 36(2): 25-40.

Schoemaker, P. J. H. and C. A. J. M. Van der Heijden (1992). "Case Study: Integrating scenarios into strategic planning at Royal Dutch/Shell." Planning Review 20(3): 41-46.

Schwaninger, M (2006) System Dynamics and the Evolution of the Systems Movemen, Systems Research and Behavioral Science, 23, 583-594.

Schwaninger, M 2006, System dynamics and the evolution of the systems movement, Volume 23, Issue 5, pages 583–594, September/October 2006

Schwaninger, M. and J. P. Ríos (2008). "System dynamics and cybernetics: a synergetic pair." System Dynamics Review 24(145-174).

Schwaninger, M. (2000). "Managing Complexity : The Path Towards Intelligent Organizations." Systemic Practice and Action Research 13(2): 207-241.

Schwaninger, M. (2002). “Short- Versus Long-Term Dynamics: The Example of Ecological Management”, In Proceedings of the 20th International Conference of the System Dynamics Society, Palermo, Italy.

Schwaninger, M. (2004). “Methodologies in conflict: Achieving synergies between system dynamics and organizational cybernetics”, Systems Research and Behavioral Science, 21, 411-431.

Schwaninger, M. and L. Jones (2000). “Learning to cope with complexity: management teaching supported by system dynamics models.” In Proc. 18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Bibliography-36 SDS (2005). What is System Dynamics. Albany, NY 12222, USA, System Dynamics Society.

Seaman, T. (1990). "On the high road to achievement: Cooperative concept mapping." Eric Digest ED335140.

Senge, P. M. and C. Lannon (1990). "Managerial Microworlds." Technology Review(July).

Senge, P. M. (1990). The Fifth Discipline: the Art and Practice of the Learning Organization. New York, Doubleday/Currency.

Senge, P. M., R. B. Ross, et al. (1994). The Fifth Discipline Fieldbook. London, UK, Nicholas Brealey.

Sharp, J. A. (1977). "System Dynamics Applications to Industrial and Other Systems." Operations Research Quarterly 28(3): 489-504.

Short, R. and McConnell, C. (2000). Extent and Impacts of Dryland Salinity, National Land and Water Resources Audit.

Shreckengost, R. (1997) Dynamic Simulation Models: How Valid Are They? (D- 4463) in Road Maps, System Dynamics Society.

SIL (1999) Glossary of Literacy Terms, Version 4.0 Summer Institute of linguistics International, (www.sil.org).

Simon, H. A. (1957). Models of Men, Wiley.

Sixtensson, A., S. Montán, et al. (2000). “Business Simulation For Daily Decision Support: A Case Study.” In Proc. 18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Smith Jr, E. C. (1962). “Simulation in Systems Engineering”, IBM Systems Journal, 33-50.

Bibliography-37 Smith, D. B. (1965). “Systems Engineering: Implications for Management”, Financial Analysts Journal, 21, 119-127.

Soil Survey Staff (1997). Keys to Soil Taxonomy USDA, Soil Conservation Service, Washington D.C, USA.

Sprague, R.H. Jr. and Watson, H.J. (1989) Decision Support Systems: Putting Theory into Practice, 2nd ed. Prentice Hall, Englewood Cliffs, New Jersey 07632.

Standing Committee on Environment and Heritage (2000). Coordinating Catchment Management: Report of Inquiry into Catchment Management, House of Representatives, The Parliament of the Commonwealth of Australia.

Steed, M. (1992). "Stella, a simulation construction kit: cognitive process and educational implications." Journal of Computers in Mathematics and Science Teaching 11(1): 39-52.

Sterman, J. D. (1994). "Learning in and about Complex Systems." System Dynamics Review 10(2/3): 291-330.

Sterman, J. D. (1994). "The Meaning of Models." Science 264(15 April): 329-330.

Sterman, J. D. (1996). Learning in and about complex systems. In Modelling for Management. (Ed).G. P. Richardson. Aldershoh, UK, Dartmouth Publishing Company. 1: 89-128.

Sterman, J. D. (1998). "Modeling Managerial Behavior: Misperceptions of Feedback in a Dynamic Decision Making Experiment." Japanese Journal of System Dynamics 2.

Sterman, J. D. (2000). Business dynamics: systems thinking and modeling for a complex world. Boston, Irwin/McGraw-Hill.

Sterman, J. D. (2000). Business Dynamics: Systems Thinking and Modeling for a Complex World, McGraw-Hill Higher Education.

Bibliography-38 Sterman, J. D. (2001). "System Dynamics Modeling: Tools for Learning in a Complex World." California Management Review 43(41): 8-27.

Sterman, J. D. (1985). “A Skeptic's Guide to Computer Models.” In Proc. 1985 International Conference of the Systems Dynamics Society, Keystone, Colorado, International System Dynamics Society.

Sterman, J. D. (1991). A Skeptic's Guide to Computer Models. Managing a Nation. G. O. Barney and e. al. Boulder CO, Westview Press.

Sterman, J. D. (1994). "Learning in and about Complex Systems." System Dynamics Review 10(2/3): 291-330.

Sterman, J. D. (1996). “Learning in and About Complex Systems.”In Modelling for Management, Vol. 1 (Ed, Richardson, G. P.). Dartmouth Publishing Company, Aldershoh, UK, pp. 89-128.

Sterman, J. D. (2000). "Learning in and about complex systems." Reflections 1 (3): 24-51.

Sterman, J. D. (2002). "All Models are Wrong: Reflections on Becoming a Systems Scientist." System Dynamics Review 18(4): 501-531.

Sterman, J. D. (2006). “Learning from evidence in complex world.”, American Journal of Public Health, 96, 505-513.

Sterman, J.D. 1991 A Sceptic’s Guide to Computer Models, In Barney G.O. et.al. (eds.). Managing a Nation: The Micro-computer Software Catalogue. Boulder, CO: Westview Press, 209-229.

Stevens, R., P. Brook, et al. (1998). Systems engineering: coping with complexity. London, Prentice Hall.

Summerell, G. K., N. K. Tuteja, et al. (2006). "Contrasting mechanisms of salt delivery to the stream from three different landforms in South Eastern Australia." Journal of Hydrology 330: 681-697.

Bibliography-39 Torlak, G. N. (2001). “Reflections on multimethodology: Maximising flexibility, responsiveness and sustainability in multimethodology interventions through a theoretically and practically improved version of Total Systems Intervention (TSI).” Systemic Practice and Action Research, 14 (3).

Toyne,P.(1995). Water Use and Environmental Flows in the Murray Darling Basin. In Proc. Water use and envioronmental Flows Workshop, 22-23 August 1995, Murray Darling Basin Commission.

Trochim, W. (1989). "An introduction to concept mapping for planning and evaluation." A Special Issue of Evaluation and Program Planning 12,: 1-16.

Turban, E. and J. E. Aronson (1998). Decision support systems and intelligent systems. Upper Saddle River, N.J, Prentice-Hall.

Turban, E., J. E. Aronson, et al. (2001). Decision support systems and intelligent systems. Upper Saddle River, NJ, Prentice Hall.

Tuteja, N. K., G. Beale, et al. (2003). "Predicting the effects of landuse change on water and salt balance - a case study of a catchment affected by dryland salinity in NSW, Australia." Journal of Hydrology 283(1-4): 67-90.

URS Australia (2002). National Action Plan for Salinity and Water Quality – Evaluation of models and tools. Milestone 3 Report for the Natural Resource Management Business Unit, Agriculture, Fisheries and Forestry Australia. A10009, URS Australia.

US Bureau of Reclamation (1985). Colorado River Simulation System: System Overview. Denver, Colorado, USA.

USDA (1960). Diagnosis and Improvement of Saline and Alkali Soils, Handobook 60, United States Department of Agriculture, USA.

Vanclay, F. (1997) The Social Basis of Environmental Management in Agriculture: A background for understanding Landcare, in Lookie, S. and Vanclay F. (eds.).

Bibliography-40 Critical Landcare, Key Papers, Series-5, Center for Rural Social Research, Charles Sturt University, Wagga Wagga, pp9-27.

Vaze, J., P. Barnett, et al. (2004). "Modelling the effects of land-use change on water and salt delivery from a catchment affected by dryland salinity in south-east Australia." Hydrological Processes 18(9): 1613-1637.

Vennix, J. A. M. (1994). Building Consensus in Strategic Decision-Making: Insights from the Process of Group Model-Building. In Proc. 1994 International System Dynamics Conference, Sterling, Scotland, System Dynamics Society.

Vennix, J. A. M. (1995). "Building Consensus in Strategic Decision Making: System Dynamics as a Group Support System." Group Decision and Negotiation 4(4): 335- 355.

Vennix, J. A. M. (1996). Group Model Building: Facilitating Team Learning Using System Dynamics. Chichester, Wiley.

Vennix, J. A. M. (1999). "Group Model-Building: Tackling Messy Problems." System Dynamics Review 15(4): 379-401.

Vennix, J. A. M. and E. A. J. A. Rouwette (2000). Group Model Building: What Does The Client Think About It Now? In Proc.18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Vennix, J. A. M. and J. W. Gubbels (1992). "Knowledge elicitation in conceptual model building: A case study in modeling a regional Dutch health care system." European Journal of Operations Research 59(1): 85-101.

Vennix, J. A. M., C. M. Thijssen, et al. (1997). Group Model Building: A Decision Room Approach. In Proc. 15th International System Dynamics Conference: "Systems Approach to Learning and Education into the 21st Century", Istanbul, Turkey, Bogazici University Printing Office.

Bibliography-41 Vennix, J. A. M., D. F. Andersen, et al. (1992). "Model-building for group decision support and alternatives in knowledge elicitation." European Journal of Operations Research 59(1): 28-41.

Vennix, J. A. M., G. P. Richardson, et al., Eds. (1997). “Group Model Building.” System Dynamics Review.

Vennix, J. A. M., H. A. Akkermans, et al. (1996). "Group Model-Building to Facilitate Organizational Change: An Exploratory Study." System Dynamics Review 12(1): 39-58.

Vennix, J. A. M., W. Scheper, et al. (1993). “Group Model-Building: What Does the Client Think of It?” In Proc. International System Dynamics Conference, Cancun, Mexico, System Dynamics Society.

Vertessy R.(2001) Integrated Catchment Management Science, CSIRO Land and Water, Technical Report No. 21/01, Canberra

Vester, F. (1990) Ausfarht Zukunft (enlish: Exit to Future), 2nd edition Wilhem Heyne Verlag, Munchen,

Walker, G., Gilfedder, M. and Williams, J. (1999). Effectiveness of Current Farming Systems in the Control of Dryland Salinity, CSIRO Land and Water, Canberra.

Wallace, S., M. Saleh, et al. (2000). Optimization Under Uncertainty In System Dynamics. In Proc. 18th International Conference of the System Dynamics Society, Bergen, Norway, System Dynamics Society.

Warren, K. (2004). "Why has feedback thinking struggled to influence strategy and policy formulation?" System Research and Behavioral Science 21(4): 331-347.

Warren, K. (2005). "Improving strategic management with the fundamental principles of system dynamics." System Dyamics Review 21(4): 329-350.

Bibliography-42 Watkins, J. A. (2003). Multimethodology: An Alternative Management Paradigm to Process Quality Improvement. Ph D thesis. Faculty of Economics and Management Sciences. Johannesburg, Rand Afrikaans University.

White, M. E. (1994). After the Greening: the Browning of Australia, Kangaroo Press, Kenthurst, NSW.

White, M. E. (1997). Listen-our land is crying: Australia's environment: problems and solutions, Kangaroo Press, Kenthurst [N.S.W.].

White, M. E. (2000). Running Down: Water in a Changing Land, Kangaroo Press, 20 Barco Street, East Roseville, NSW 2069.

Williams, D. W. (2002). Integrating System Dynamics Modeling and Case Study Research Method : A Theoretical Framework for Process Improvement. Proceedings of the 20th International Conference of the System Dynamics Society, Palermo, Italy, The System Dynamics Society.

Wilson, S. M. (1999). Dryland Salinity: What are the impacts and how do we value them? Murray Darling Basin Commission, Canberra.

Winch, G. W. (2000). System dynamics: from theory to practice. First International Conference on Systems Thinking for Management.

Wingham, P.A. and Young W.J. (2001). “Modelling River and Floodplain interactions for Ecological Response.” Mathematical and Computer Modelling 22(2001).635-647.

Wittenberg, J. (1992). On the very idea of a system dynamics model of Kuhnian science, System Dynamics Review, 8, 21-33.

Wolfenden, J.A. J, (1999). A Trans-disciplinary approach to integrated Resource Management: A pragmatic application of Ecological Economics, Ph.D Thesis, University of New England, Australia.

Bibliography-43 Wolstenholme, E. F. (1982). “System Dynamics in Perspective”, Journal of the Operational Research Society (UK)., 33, 547-556.

Wolstenholme, E. F. (1985). A Methodology for Qualitative System Dynamics, In Proc. of the 1985 International Conference of the Systems Dynamics Society, Vol. 2 International System Dynamics Society, Keystone, Colorado, pp. 1049-1058.

Wolstenholme, E. F. (2003). "Towards the Definition and Use of a Core Set of Archetypal Structures in System Dynamics." System Dynamics Review 19(1): 7-26.

Wolstenholme, E. F. and Coyle, R. G. (1983).” The Development of System Dynamics as a Methodology for System Description and Qualitative Analysis.” Journal of the Operational Research Society (UK)., 34, 569-581.

Wolstenholme, E.F. (1990). System Enquiry: A System Dynamics Approach. John Wiley & Sons, Chichester (UK):

Wolstenholme, E.F. and Al-Ausi (1987). “System Dynamics and Heuristic Optimization in Defense Analysis.” System Dynamics Review 3,2, 102-105

World Resources Institute (2003). EarthTrends: The Environmental Information Portal, Available at http://earthtrends.wri.org. Washington DC: World Resources Institute.

WRCSA (2002). The Economic Impact of Changing Water Quality On Irrigated Agriculture In The Lower Vaal And Riet Rivers, Water Research Council of South Africa.

Young, W. J., A. C. Scott, et al. (2003). Murray Flow Assessment Tool: A Technical Description (version 1.4). Canberra, CSIRO Land and Water,.

Zadeh, L. A. (1965). “Fuzzy sets.” Information control, 8, 338-353.

Bibliography-44 ANNEX A

GLOSSARY

Accumulation: The collection of some quantity over time. Examples of accumulation include water in a bathtub, savings in a bank account, inventory (MIT 1994a:1). In System Dynamics modelling Software, accumulations are represented as stock or level and a rectangle is used in model diagrams.

Action Research: “learning by doing” - a group of people identify a problem, do something to resolve it, see how successful their efforts were, and if not satisfied, try again (O’brien 2001).

Aggregation: The incorporation of numerous distinct system components into one variable. Aggregation is done for simplicity when combination generates the same behaviour of interest as representing the components separately (MIT 1994a:1).

Baseline: The software component versions and hardware modification states which form the basis for continued system development.

Boundary (of a system): Border enclosing only the parts of system structure needed to generate the behaviour of interest. In other words, the system boundary excludes all components not relevant to the problem behaviour of the system (MIT 1994a:1). Boundaries delineate a system from its environment and thereby maintain the system’s identity as it changes.

Causal loop diagram: Diagram representing a closed loop of cause-effect linkages (causal links) which is intended to capture how the variables interrelate (MIT 1994a:2).

Computer model: A set of descriptions that tell the computer how each part of the system acts. A good computer model captures the dynamic essence of the system it represents. It explicitly contains the assumptions being made about the system (MIT 1994a:2).

A1 Counterintuitive behaviour: A surprising result of policies devised to remedy a problem. Often the presumed “solutions” result in counter-productivity. Thus as troubles increase, efforts are intensified which actually worsen the problem (MIT 1994a:2-3).

Decision function: Also known as a rate equation. It is a policy statement that determines how the levels are related to the decisions (rates) (MIT 1994a:3).

Decomposition: The activity of identifying the constituent elements and relationships of a system or subsystems (Stevens et al., 1998: 360).

Delay: A phenomenon where the effect of one variable on another does not occur immediately. Delays result from decisions that often require a long period of time to be effective (MIT 1994a:3).

Dynamic model: A model that deals with relationships that vary with time (MIT 1994a:3).

Electrical Conductivity: It is the ability of a liquid to carry an electrical current. It is usually expressed in microsiemens per centimetres or decisiemens per meter (dS/m).

Equilibrium: A situation in a dynamic system where the inflows and corresponding outflows balance, and the levels cease to change (MIT 1994a:3).

Feedback system: A closed system influenced by its past behavior. Feedback systems have feedback loop structure that consist of closed paths of cause and effect. They are self-regulating and can be either a positive feedback system or a negative feedback system (MIT 1994a:3). A feedback loop is system structure that causes output from one node to eventually influence input to that node. For example, the work output of a population can increase the goods and services available to that population, which can increase the average life expectancy, which can increase the population, which can increase the work output still more, and the loop starts all over again (Harich 2011).

A2 Generic structure: A structure that can be applied across different settings due to fundamentally same underlying structures and relationships (MIT 1994a:4).

Goal-seeking behavior: Behavior that results from a balancing loop which drives a system toward a specified goal. The farther the system from the goal, the quicker it changes towards that goal, the equilibrium homeostasis point, and as it approaches the goal, the growth/decay slows down (MIT 1994a:4).

Leverage point: The term is generally used in Systems thinking and system dynamics and refers to a policy/decision rule which can yield large changes in a system (MIT 1994a:5). Leverage is the ratio of change in output to change in input. A leverage point is a place in a system where force can be applied. A low leverage point is a place in a system where a small amount of force causes a small change to system behaviour. A high leverage point is a place in a system where a small amount of change force (the effort required to prepare and make a change) causes a large amount of predictable, favourable response (Harich 2011).

Limits to growth: A resource constraint, an external or internal response to growth. A growth caused by reinforcing feedback processes begins to slow and eventually come to a halt at the limit, and may even reverse itself and collapse (MIT 1994a:5).

Mental model: A model representing the relationships and assumptions about a system held in a person’s mind. Mental models are often correct in system structure, but frequently draw wrong conclusions about system behaviour (MIT 1994a:5).

Negative feedback: Feedback that works to cancel deviations from a goal. It exhibits goal-seeking behaviour. The control decision attempts to adjust a system level to a value given by a goal introduced from outside the loop (MIT 1994a:5).

Parameters: Numerical values that describe relationships in a system and are considered constant, at least during the computation span of one model run (MIT 1994a:5). Parameters are shown by a circle in System Dynamics modelling software.

A3 Policy analysis: Analysis employed to evaluate the causes of undesirable behaviour in a system. It allows the model-builder to compare how a system would react to different policies through simulation (MIT 1994a:6).

Positive feedback: Feedback that contains reinforcing loops which produce exponential change. Change in one direction results in more and more change in the same direction. Positive feedback produces growth (MIT 1994a:6)

POWERSIM: MS Windows-based modelling software package for system dynamics models. (MIT 1994a:6)

Reference mode:- A system dynamics tool that shows how certain variables change over time. Several variables can appear on the same graph for comparison and Time is shown on the horizontal axis. Organized historical information and inferred future (Saeed 1994).

Simulation: Conducting dynamic experiments on a model instead of on the real system (MIT 1994a:6).

Structure: The building blocks and interval connections of a system. It is the way in which system elements are organized or interrelated (MIT 1994a:7).

System: A collection of parts that interact to function as a whole. A system is almost always defined with respect to a specific purpose. Systems often contain circular patterns of cause and effect called feedback loops (MIT 1994a:7)

System Dynamics: A field for understanding how things change through time. System dynamics deals with how the internal feedback-loops within the structure of a system create behaviour. Computer simulation models are used to achieve better understanding of system behaviour over time. With a better comprehension of systems, one can redesign structure or policies to improve the behaviour. The field of system dynamics was created by Jay Forrester beginning in 1956 (MIT 1994a:7).

Traceability: The ability within the system development to relate a feature of design or implementation to a parent requirement ( Stevens et al., 1998: 362)

A4 Verification: A process that ensures that a model meets its specification.

Validity: Judgment of a model’s suitability for a particular purpose. A model is valid when it accomplishes what is expected of it (MIT 1994a:8). Systems Engineering literature (for example, Stevens et al. 1998) defines it as actions that confirm that the behaviour of a system meets user needs.

Vicious cycle: Reinforcing, amplifying process that yields undesirable results (MIT 1994a:8).

Weltanschauung: Unquestioned image or model of the world that makes a particular human activity system, with its particular transformation process, a meaningful one to consider (Checkland 1993: 312-319)

A5 ANNEX B

MAJOR DRYLAND SALINITY RELATED POLICY DOCUMENTS IN THE

MURRAY DARLING BASIN

The dryland salinity policy development processes started in 1988 and progressed through the last decade. It culminated in the Basin Dryland salinity Strategy 2001- 2015 (MDBC 2001a). During 1990s, three other policies were developed that also contributed to salinity management. These policies are Balancing Water Use Policy (1995), MDB Environment Policy (1994) and Water Quality Policy (1990). Dryland salinity in the MDB has initiated a response both at national as well as state levels over the last two decades. The following plans/ programs constitute the basis of the dryland salinity policy in the MDB:

• COAG Water Reforms (AFFA 2001a) • Our Vital Resources, National Action Plan for Dryland salinity and Water Quality (AFFA 2001b) • Basin Dryland salinity Management Strategy 2001-2015 (MDBMC 2001) • National Dryland Dryland salinity Program (LWRRDC 1998) • Management of Dryland salinity: Future Strategic Directions (CSIRO 2000) • NSW Dryland salinity Strategy (DLWC 2000) • Victoria’s Dryland salinity Management Framework (NRE 2000)

The policy approaches employed in these documents are compared in following Table.

B1

Table 3.11 Dryland Salinity Management Initiative in the Murray Darling Basin

Direct Initiatives Plans that can affect dryland salinity indirectly National Dryland Dryland salinity Murray Darling NSW State Dryland Victoria State Dryland COAG Water Reforms National Action Plan Program Phase II Basin Dryland salinity Strategy salinity Framework for salinity and Water salinity Quality Management Strategy • A wide range of communication • Dryland salinity • Targets for catchments. • Outcomes and • Full cost recovery • Targets and activities and research programs to targets. • Market based solutions targets. based water pricing standards for natural better understand the complex • Managing trade- and strategic • Partnership for and transparency of resources interrelationship between managed offs. investment. integrated catchment cross subsidies. management. ecosystems, rural landscapes and • Dryland salinity • Dryland salinity related management. • Comprehensive • Integrated hydrogeological systems; management business opportunities • Understanding systems of water catchment/ regional • Options for improved management of plans. • Regulation. catchment processes allocation. management plans landscapes threatened by salinisation • Redesigning. • Government advice • Building skills and • Integrated • Capacity building to maintain their potential for farming system. • Information. capacity for change. catchment for communities. productive use and biodiversity • Salt interception • Scientific knowledge. • Efficient water-use management • An improved conservation; works. • Planning System. and regional growth. approach. governance • Principles and practices, that enable • Basin-wide • Cost sharing and • Greater framework. the beneficial use or rehabilitation of accountability. accountability. responsibility at • Clearly articulated salinised landscape resources; local levels for role for the • Options for creation of economic, water resources commonwealth, social, institutional or legal management. state, local incentives or mechanisms that • Public education government and the encourage prevention of dryland about water-use and community. salinity and management of its public consultation • Public impacts; and in implementation communication • Causes, costs and consequences of of water reforms. program dryland salinity as it relates to industry, biodiversity, regional communities and governments.

B2

National Action Plan for Salinity and Water Quality

The National Action Plan for Dryland salinity and Water Quality was published in 2000 (AFFA 2001) with the main goal of motivating and enabling regional communities to use coordinated and targeted action to:

Prevent, stabilise and reverse trends in salinity, particularly dryland salinity, affecting the sustainability of production, the conservation of biological diversity and viability of infrastructure; and Improve water quality and secure reliable allocations for human uses, industry and the environment.

The action plan proposes six main actions, i.e., targets and standards for natural resources management; integrated catchment/ regional management plans; capacity building for communities; an improved governance framework; clearly articulated role for the commonwealth, state, local government and the community; and public communication program.(AFFA 2001)

The National Action Plan for Salinity and Water Quality was ceased on 30 June 2008. It was replaced by the ‘Caring for our country’ a new program for natural resources management.

Murray Darling Basin Salinity Management Strategy

Basin Dryland salinity Management Strategy is a 15 year plan that follows the lines similar to those of integrated catchment management. It focuses on nine program areas (MDBMC 2001):

• Constructing salt interception works. • Identifying values and assets at risks. • Implementing dryland salinity and catchment management plans. • Setting dryland salinity targets. • Redesigning farming systems. • Managing trade offs with the available within-valley options. • Ensuring Basin-wide accountability: monitoring, evaluating and reporting.

B3

• Developing capacity to implement strategy. • Targeting reforestation and vegetation management.

The National Dryland salinity Program

The National Dryland Dryland salinity Program (NDSP) was started in July 1993 as a means of improving the coordination of Australia's research, development and extension (R,D&E) effort directed towards better management of dryland dryland salinity across rural Australia. The first phase of the Program was completed in June 1998. The focus of the first phase was on improving the understanding of the causes of dryland dryland salinity and on establishing a collaborative national focus on the R&D effort. The second phase, started in 1998 has the following objectives (LWRRDC 1998):

• To develop options for operating environments which encourage the prevention of dryland dryland salinity and the appropriate management of its impacts. • To develop understanding and demonstrate principles and practices to address the cause, costs and consequences of dryland dryland salinity. • To develop an understanding, and demonstrate principles and practices, which enable the beneficial use or rehabilitation of landscape resources impacted by dryland salinity. • To develop an understanding of landscape processes and ecosystem functions in areas affected by, or at risk from, high watertables and dryland salinity.

The second phase is built around the following seven themes: • Audit and monitoring. • Policy and operating environment. • Grains and other industries. • New, emerging and other industries that use saline resources productively. • Environmental protection and rehabilitation. • Infrastructure management.

B4

• State, regional and community initiatives.

Examples of the NDSP projects are:

• Extent and impacts of dryland dryland salinity. • Watertable change in western slopes cropping area of the NSW. • Regional case studies to assess water balance and management options. • Dryland salinity management optimisation framework case study.

B5

Annex C Concept Map Analysis

All Definitions and descriptions of the analytical functions of the Decision Explorer, described in this Annex, are adopted form the help menu of the Decision Explorer Version 3.1.2 Academic (Banxia 1991-2000) with little modifications where needed to suit the purpose of this Annex.

C1 List of all concepts 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 7 Increased construction costs of urban and transport infrastructure 8 Shortage of land for agricultural production 9 Availability of land for clearing 10 Increase in requirement for agricultural production 11 Current agricultural production 12 Increase in agricultural exports 13 Increase in domestic consumption 14 Current land under agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 16 Decrease in rainfall due to drought 17 Climate change 18 Decrease in value of land either salt affected or at the risk of becoming salt affected 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 21 decrease in biodiversity 22 Increased cost of Agricultural production

C1

23 Increase in value of productive land neither salt affected nor at the risk of becoming salt affected 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 26 Effectiveness of dryland salinity control treatments 27 Dryland salinity control treatments 28 Cost of dryland salinity control treatments 29 different causal theories of dryland salinity 30 salt deposit and carried away with rain, groundwater 31 Salt interception schemes 32 Agronomic Solutions 33 Plantations 34 Targeted reforestations 35 Stabilization of salty patches/ salt-stores 36 Farm income 37 Farmers willingness to apply dryland salinity control treatments 38 Farm assett's Value 39 Restrictions on land clearing 78 Farm Debt 40 concepts

C2

C 2 List of links 1 > 7 21 19 .17 5 2 > 3 3 > 1 4 > 2 5 > 22 18 6 6 > 36 10 8 > 24 4 9 > 4 10 > 15 11 12 > 10 13 > 10 14 > 8 11 15 > 23 8 16 > 2 17 > .1 18 > 38 19 > 20 20 > 8 22 > 78 36 23 > 38 24 > 25 25 > 19 26 > 25 27 > 35 34 33 32 31 28 25 28 > 36 22 29 > 30 3 30 > 1 36 > 37 37 > 24 38 > .78 37

C3

39 > 4 78 > .38 C 37

C4

C3 Domain Analysis

Domain analysis provides some idea of which concepts are key issues and may warrant further examination. In domain analysis the domain of each concept in the model is analysed, and a list the number of inward, outward, connotative and total links around that concept is produced.

The Domain analysis examines each concept and calculates how many concepts are immediately related to it (i.e. directly linking in or out of the concept). Through this it is possible to be able to identify which concepts are the best elaborated or have a high density of links around them. Furthermore it is possible to focus on only those concepts which are the most densely linked.

Results of Domain Analysis All concepts in descending order of value

7 links around 27 Dryland salinity control treatments

6 links around 1 Dryland Salinity

5 links around 8 Shortage of land for agricultural production 10 Increase in requirement for agricultural production

4 links around 4 Increase in land clearing 5 Degraded land quality 78 Farm debt 22 Increased cost of Agricultural production 25 Land reclaimed from dryland salinity

C5

36 Farm income 37 Farmers willingness to apply dryland salinity control treatments 38 Farm assett's Value

3 links around 2 Increase in groundwater recharge 3 rise in watertable 6 Decrease in agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 19 Loss of productive land due to dryland salinity 23 Increase in value of productive land neither salt affected nor at the risk of becoming salt affected 24 Area on which dryland salinity control treatments are applied 28 Cost of dryland salinity control treatments

2 links around 11 Current agricultural production 14 Current land under agricultural production 18 Decrease in value of land either salt affected or at the risk of becoming salt affected 20 Required increase in productive land 29 different causal theories of dryland salinity 30 salt deposit and carried away with rain, groundwater

1 link around 7 Increased construction costs of urban and transport infrastructure 9 Availability of land for clearing 12 Increase in agricultural exports 13 Increase in domestic consumption 16 Decrease in rainfall due to drought 21 decrease in biodiversity

C6

26 Effectiveness of dryland salinity control treatments 31 Salt interception schemes 32 Agronomic Solutions 33 Plantations 34 Targeted reforestations 35 Stabilization of salty patches/ salt-stores 39 Restrictions on land clearing

0 links around 17 Climate change

C7

C4 HIESET - Hierarchical Set Clustering and Potent Analyses

HIESET analysis looks around all of the 'root' concepts in the Set specified, and traces all of the explanations of each concept until either a tail or another concept in the Set is reached. An individual Hieset Set is created for each concept in the parameter Set. Therefore, if there are three concepts in the parameter Set, Hieset1, Hieset2, and Hieset3 will be produced, containing the concepts that are lower in the model's hierarchy than the particular root concept for each Set (subordinate concepts).

The aim of HIESET analysis is to produce hierarchical sets or groups based on a specified set of concepts (concepts that ‘seed’ the hierarchical sets). Each hierarchical set will contain one of the specified set and all the concepts that explain it - i.e. are the means by which the concept in question can be achieved. However, if the analysis, in tracing down the chain of argument, reaches another of the specified seed set it will stop and not proceed further down that particular chain of argument.

Potent analysis is used in conjunction with the Hieset analysis as it takes its information from the sets created by Hieset. Thus it cannot be run if there are no Hiesets in existence.

By examining the contents of each of the sets created by the Hieset analysis, Potent identifies which concepts appear in the most number of sets and thus determines their ‘potency’ within the model. A list of these ‘potent’ concepts is then displayed in descending order of potency. It is also possible to identify either all or a specified number of potent concepts from one particular set.

Results of Potent Analysis

Top 41 concepts in descending order of value

4 Hiesets with 1 Dryland Salinity

C8

3 rise in watertable 17 Climate change 29 different causal theories of dryland salinity 30 salt deposit and carried away with rain, groundwater

3 Hiesets with 5 Degraded land quality 6 Decrease in agricultural production 10 Increase in requirement for agricultural production 12 Increase in agricultural exports 13 Increase in domestic consumption

2 Hiesets with 15 Required increase in land to accommodate increase in requirements for agricultural production 27 Dryalnd salinity control treatments

1 Hieset with 2 Increase in groundwater recharge 4 Increase in land clearing 78 Farm Debt 8 Shortage of land for agricultural production 9 Availability of land for clearing 11 Current agricultural production 14 Current land under agricultural production 16 Decrease in rainfall due to drought 18 Decrease in value of land either salt affected or at the risk of becoming salt affected 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 22 Increased cost of Agricultural production

C9

23 Increase in value of productive land neither salt affected nor at the risk of becoming salt affected 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 26 Effectiveness of dryland salinity control treatments 28 Cost of dryland salinity control treatments 36 Farm income 37 Farmers willingness to apply dryland salinity control treatments 38 Farm asset’s Value 39 Restrictions on land clearing

0 Hiesets with 7 Increased construction costs of urban and transport infrastructure 21 decrease in biodiversity 31 Salt interception schemes 32 Agronomic Solutions 33 Plantations 34 Targeted reforestations 35 Stabilization of salty patches/ salt-stores

C10

C5 Central Analysis

The Central analysis takes an input set as its second parameter, and with each concept in turn, analyses the concepts away from each band of the central concept. Each concept is weighted according to how many concepts are traversed in its Band level. All the concepts found at level one are divided by one, all concepts at level two are divided by two, and so on, up to the specified Band level (or the default of Band level 3). Each band score is added together to give a total overall score for each concept in the input set. The central score is given first, with the total number of concepts traversed given second.

The Central analysis looks at concepts to the specified band level that are linked to each preceding concept, irrespective of direction. Therefore connotative are treated as merely a link and not bi-directional. Any concepts encountered previously on the route are ignored.

For example, the link chain: 77 -> 23 <- 32 - 12; will report concept 77 at level one, 23 level 2, 32 level 3 and 12 level 4. These are all irrespective of link direction and only take into account the distance away from the central specific concept, in terms of band level.

Results of Central Analysis Cent Scores Calculated: 25 Land reclaimed from dryland salinity 13 from 27 concepts.

24 Area on which dryland salinity control treatments are applied 13 from 29 concepts.

19 Loss of productive land due to dryland salinity 13 from 28 concepts.

C11

1 Dryland Salinity 13 from 25 concepts.

22 Increased cost of Agricultural production 12 from 25 concepts.

8 Shortage of land for agricultural production 12 from 25 concepts.

5 Degraded land quality 12 from 26 concepts.

36 Farm income 11 from 24 concepts.

27 Dryalnd salinity control treatments 11 from 19 concepts.

6 Decrease in agricultural production 11 from 25 concepts.

37 Farmers willingness to apply dryland salinity control treatments 10 from 20 concepts.

28 Cost of dryland salinity control treatments 10 from 20 concepts.

20 Required increase in productive land 10 from 23 concepts.

C12

15 Required increase in land to accommodate increase in requirements for agricultural production 10 from 22 concepts.

10 Increase in requirement for agricultural production 10 from 19 concepts.

4 Increase in land clearing 9 from 18 concepts.

3 rise in watertable 9 from 19 concepts.

78 Farm Debt 8 from 15 concepts.

38 Farm assett's Value 8 from 15 concepts.

18 Decrease in value of land either salt affected or at the risk of becoming salt affected 8 from 19 concepts.

2 Increase in groundwater recharge 8 from 18 concepts.

30 salt deposit and carried away with rain, groundwater 7 from 14 concepts.

23 Increase in value of productive land neither salt affected nor at the risk of becoming salt affected 7 from 18 concepts.

C13

14 Current land under agricultural production 7 from 17 concepts.

21 decrease in biodiversity 6 from 14 concepts.

17 Climate change 6 from 14 concepts.

11 Current agricultural production 6 from 13 concepts.

7 Increased construction costs of urban and transport infrastructure 6 from 14 concepts.

35 Stabilization of salty patches/ saltsotres 5 from 12 concepts.

34 Targetted reforestations 5 from 12 concepts.

33 Plantations 5 from 12 concepts.

32 Agronomic Solutions 5 from 12 concepts.

31 Salt interception schemes 5 from 12 concepts.

C14

29 different causal theories of dryland salinity 5 from 11 concepts.

26 Effectiveness of dryland salinity control treatments 5 from 14 concepts.

39 Restrictions on land clearing 4 from 10 concepts.

13 Increase in domestic consumption 4 from 10 concepts.

12 Increase in agricultural exports 4 from 10 concepts.

9 Availability of land for clearing 4 from 10 concepts.

16 Decrease in rainfall due to drought 3 from 8 concepts.

C15

C6 Loop Detect Analysis

This command is used to find Loops of concepts in a model. Each Loop of concepts found is placed in an individual Set, so that, if three Loops are found, then LoopSet1, LoopSet2, and LoopSet3 would be created.

Loops are caused when a circle of links is formed, often in a complex chain of argumentation in large models. The LOOP command, in identifying these Loops, allows the user to decide where/whether to break them.

A Loop might be formed by the following link sequence, for example:

15+16, 16+17, 17+22, 22+23, 23+15 The right hand side of the following diagram would be the resulting contents of the LoopSet when using the LOOP command:

18 17

17 16 22

16 15 22

23 15

14 23

Note that this command does NOT class concepts linked connotatively as Loops.

Results of loop detect analysis Loop1 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge

C16

3 rise in watertable 4 Increase in land clearing 8 Shortage of land for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land

Loop2 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 8 Shortage of land for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 22 Increased cost of Agricultural production 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 37 Farmers willingness to apply dryland salinity control treatments 78 Farm Debt

Loop3 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 8 Shortage of land for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 22 Increased cost of Agricultural production

C17

24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 36 Farm income 37 Farmers willingness to apply dryland salinity control treatments

Loop4 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 8 Shortage of land for agricultural production 18 Decrease in value of land either salt affected ot at the risk of becoming salt affected 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 37 Farmers willingness to apply dryland salinity control treatments 38 Farm assett's Value

Loop5 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 24 Area on which dryland salinity control treatments are applied

C18

25 Land reclaimed from dryland salinity 36 Farm income 37 Farmers willingness to apply dryland salinity control treatments Loop6 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 10 Increase in requirement for agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 23 Increase in value of productive land neither salt affected nor at the risk of becoming salt affected 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 37 Farmers willingness to apply dryland salinity control treatments 38 Farm assett's Value

Loop7 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production

C19

10 Increase in requirement for agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production

Loop8 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 8 Shortage of land for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity

Loop9 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 10 Increase in requirement for agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity

C20

Loop10 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 10 Increase in requirement for agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land

Loop11 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 22 Increased cost of Agricultural production 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 36 Farm income 37 Farmers willingness to apply dryland salinity control treatments 78 Farm Debt

C21

Loop12 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 22 Increased cost of Agricultural production 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 36 Farm income 37 Farmers willingness to apply dryland salinity control treatments

Loop13 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 18 Decrease in value of land either salt affected ot at the risk of becoming salt affected 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 36 Farm income 37 Farmers willingness to apply dryland salinity control treatments

C22

38 Farm asset's Value

Loop14 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 10 Increase in requirement for agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 22 Increased cost of Agricultural production 23 Increase in value of productive land neither salt affected nor at the risk of becoming salt affected 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 37 Farmers willingness to apply dryland salinity control treatments 38 Farm asset’s Value 78 Farm Debt

Loop15 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production

C23

8 Shortage of land for agricultural production 10 Increase in requirement for agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 22 Increased cost of Agricultural production 23 Increase in value of productive land neither salt affected nor at the risk of becoming salt affected 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 36 Farm income 37 Farmers willingness to apply dryland salinity control treatments 38 Farm asset’s Value

Loop16 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 10 Increase in requirement for agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 18 Decrease in value of land either salt affected or at the risk of becoming salt affected 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 23 Increase in value of productive land neither salt affected nor at the risk of becoming salt affected

C24

24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 37 Farmers willingness to apply dryland salinity control treatments 38 Farm asset’s Value

Loop17 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production 8 Shortage of land for agricultural production 10 Increase in requirement for agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 22 Increased cost of Agricultural production 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 37 Farmers willingness to apply dryland salinity control treatments 78 Farm Debt

Loop18 set contains: 1 Dryland Salinity 2 Increase in groundwater recharge 3 rise in watertable 4 Increase in land clearing 5 Degraded land quality 6 Decrease in agricultural production

C25

8 Shortage of land for agricultural production 10 Increase in requirement for agricultural production 15 Required increase in land to accommodate increase in requirements for agricultural production 19 Loss of productive land due to dryland salinity 20 Required increase in productive land 22 Increased cost of Agricultural production 24 Area on which dryland salinity control treatments are applied 25 Land reclaimed from dryland salinity 36 Farm income 37 Farmers willingness to apply dryland salinity control treatments

C26

C7 COTAIL Analysis

This command traces the path from each tail concept until a branch point is reached (i.e. a concept has more than one consequence). All branch points reached are then sorted according to their style and displayed.

Cotail (standing for Composite tails) is an analysis which searches through the model to find those ‘potential’ options which have more than one outcome, i.e. more than one consequence leading from them. Starting from the bottom of the model - the tails - the analysis traces up each chain of argumentation until it meets one of these branch points. It flags this as a composite tail and starts with the next chain of argumentation. Note that a concept that is a tail which is itself a branch point can thus be a cotail. Once the analysis has identified all those concepts that fit the requirement, they are then listed on the screen according to their style. Any concepts which are already in a set called ‘option’, if any, are listed first followed by the other styles. The analysis is accessed by typing cotail at the command line.

The results of this analysis are useful for identifying potential options. It is sometimes worth comparing these results with those concepts identified through the ‘Potent’ analysis in order to discover which appear in both, thus confirming their potency.

Note: Concepts of style ‘Option’, if any, are always displayed first. Concepts in the display can be selected and their style changed using the style selector.

Cotail results

Branch points of style standard 8 Shortage of land for agricultural production 14 Current land under agricultural production 29 different causal theories of dryland salinity

C27

Branch points of style Issue 1 Dryland Salinity 10 Increase in requirement for agricultural production 27 Dryland salinity control treatments

Reference: Decision Explorer Help. ©1994-99 Matthew Jones, Banxia Software Ltd.

C28 Conventions for Influence Diagrams

(Coyle 1996 )

Types of link in an Influence Diagram

Physical Flow

Information transmission Link have + or - signs Control action Behaviour of nature

Parameter Sign may be omitted, but a + or – may also be used

Delays in Physical Flow

D denotes delay, subscript identifies a particular delay, sign is always D1 positive

Smoothing of Information Annex D

VARIABLE AVERAGE VALUE OF VARIABLE

TIME OVER WHICH VARIABLE IS AVERAGED

External Forces These are forces outside the system over which the systems’s ‘managers’ FORCE have no control. They may be physical or influences from the behaviour of or nature. The link has a + or – sign. FORCE

Constrained Flows

Physical constrained flows are used to indicate that flow is only possible whilst material remains available to flow in direction of the arrow.

A dashed line constrained flow is used to remotely control when physical flows can occur, that is, where a physical flow is constrained by activity which originates elsewhere in the system Annex E Model Equations mainmodel Component 1 { aux Adsoption fraction variability { autotype Real def RANDOM(0.1,0.9,0.5) } const Adsorption fraction { autotype Real init 0.5 } aux Auxiliary_1 { autotype Real unit ha def 6 } aux Auxiliary_2 { autotype Real unit 'sq km' def Auxiliary_1 } const Critical discharge level { autotype Real unit mm init -4500 } aux Critical watertable level { autotype Real unit mm def 500 } level Depth to Watertable { autotype Real autounit mm init 'INI Depth to Watertable' inflow { autodef 'Depth to watertable increase' } outflow { autodef 'Watertable decrease discharge' } outflow { autodef 'Depth to Watertable decrease intrusion' } } aux Depth to Watertable decrease intrusion { autotype Real unit mm/yr def IF('Depth to Watertable'>0<>,'Groundwater intruding rootzone'/'porosity below rootzone',0<< mm/yr>>) } aux Depth to watertable increase { autotype Real unit mm/yr def IF(Percolation/'porosity below rootzone'>0<>,Percolation/'porosity below rootzone',0<>) } aux Discharge level discrepancy { autotype Real unit mm def 'Critical discharge level'-'Depth to Watertable' } aux Drainage { autotype Real unit mm/yr def Infiltration*'Drainage efficiency' } aux Drainage efficiency { autotype Real def RANDOM(0.1,1.0,0.05) } aux Effective precipitation { autotype Real autounit mm/yr def Precipitation*'Effective precipitation percentage'

1 } aux Effective precipitation percentage { autotype Real def RANDOM(0,1,0.35) } aux evapotranspiring water { autotype Real unit mm/yr def IF('Effective precipitation'+'Groundwater intruding rootzone'>0<>,'Effective precipitation'+ 'Groundwater intruding rootzone',0<>) } aux Fraction salinity increase due to salts deposited by wind { autotype Real def 0.2 } aux Groundwater discharge fraction { autotype Real def RANDOM(0.10,0.20,0.4) } aux Groundwater intruding rootzone { autotype Real unit mm/yr def 'Watertable discrepancy'*'Intruding groundwater fraction'*'porosity below rootzone' } aux Infiltration { autotype Real autounit mm/yr def 'Infiltration percentage'*Precipitation } aux Infiltration percentage { autotype Real def 0.5 } const INI Depth to Watertable { autotype Real unit mm init 5000 } const INI Root zone salinity { autotype Real unit mg/l init 1000 } aux Initial salinity of groundwater { autotype Real unit mg/l def 600 } aux Intermediate Groundwater Systems { autotype Real def GRAPHCURVE((TIME-STARTTIME)/TIMESTEP,0,20,{0,4,11.6,28.4,51.6,73.5,86,92,95.5,97.4,100// Min:0;Max:100//}) } aux Intruding groundwater fraction { autotype Real autounit yr^-1 def 0.5<<1/yr>> } aux Intruding groundwater ratio { autotype Real def 'Groundwater intruding rootzone'/'evapotranspiring water' } aux Percolation { autotype Real unit mm/yr def Infiltration-Drainage }

2 aux porosity below rootzone { autotype Real def 0.4 } aux Precipitation { autotype Real unit mm/yr def RANDOM(50<>,1000<>,0.5) } aux Precipitation ratio { autotype Real def IF('Effective precipitation'/'evapotranspiring water'>0,'Effective precipitation'/'evapotranspiring water',0) } aux Rate of rootzone salinity decrease { autotype Real unit mg/l/yr def IF('surface soil or rootzone salinity'>0<>,((Infiltration/'Rootzone depth'/'Rootzone porosity')* 'Salinity of infiltration water')+(Infiltration/'Rootzone depth'/'Rootzone porosity')*'Salinity of infiltration water'*'Salinity decrease due to salts taken up by plants',0<>) } aux Rate of rootzone salinity increase { autotype Real unit mg/l/yr def IF(('evapotranspiring water'/'Rootzone depth'/'Rootzone porosity')*'salinity of evapotranspiring water'+( ('evapotranspiring water'/'Rootzone depth'/'Rootzone porosity')*'salinity of evapotranspiring water'* 'Fraction salinity increase due to salts deposited by wind')>0<>,('evapotranspiring water'/ 'Rootzone depth'/'Rootzone porosity')*'salinity of evapotranspiring water'+(('evapotranspiring water'/ 'Rootzone depth'/'Rootzone porosity')*'salinity of evapotranspiring water'*'Fraction salinity increase due to salts deposited by wind'),0<>) } aux Rootzone depth { autotype Real unit mm def 500 } aux Rootzone porosity { autotype Real def 0.4 } aux Rootzone porosity variability { autotype Real def RANDOM(0.1,0.9,0.5) } aux Salinity decrease due to salts taken up by plants { autotype Real def 0.1 } aux salinity of evapotranspiring water { autotype Real unit mg/l def 'Salinity of groundwater'*'Intruding groundwater ratio'+'Salinity of Precipitation'*'Precipitation ratio' } aux Salinity of groundwater { autotype Real unit mg/l def 'Initial salinity of groundwater'+'Salinity of infiltration water' } aux Salinity of infiltration water { autotype Real unit mg/l def 'surface soil or rootzone salinity'*(1-'Adsorption fraction') } const Salinity of Precipitation { autotype Real unit mg/l init RANDOM(0<>,500<>,0.05) }

3 aux Starttime for a variable { autotype Real def RAMP(45<<1/yr>>,STARTTIME+70<>) } level surface soil or rootzone salinity { autotype Real unit mg/l init 'INI Root zone salinity' inflow { autodef 'Rate of rootzone salinity increase' } outflow { autodef 'Rate of rootzone salinity decrease' } } const Time Delay between non saline land becoming salt affected { autotype Real unit yr init 30 } aux Watertable decrease discharge { autotype Real unit mm/yr def IF('Depth to Watertable'>0<>,-('Discharge level discrepancy'*'Groundwater discharge fraction')/ 1<>,0<>) } aux Watertable discrepancy { autotype Real unit mm def 'Depth to Watertable'-'Critical watertable level' } } unit gm { def ATOMIC doc Gram note System mass Unit } unit ha { def ATOMIC doc Hecatres note System Area Unit } unit Kg { def 1000gm doc Kilogram -mass } unit km { def 1000*m doc Kilometer - Length } unit l { def (0.1*m)^3 doc Liter - Volume } unit m { def __METER doc Meter - length } unit mg { def ATOMIC doc Milligram - mass } unit mm { def ATOMIC doc Millimeter - Length } unit PPM { def ATOMIC doc Parts per millions note Standard unit used to express quantity of salts in soil or soil extract }

4 unit sq km { def 100*ha doc Square Kilometers note System Area Unit } unit Sq m { def m*m doc Square meters note System Area Unit }

5

Supplement I

Primary Producers’ Perspective on the Extent of Dryland Salinity, Current Trends and Potential Remedial Measures

1 Table of Contents

1 INTRODUCTION ...... 4

2 RESEARCH OBJECTIVES ...... 4

3 SURVEY METHODOLOGY ...... 4

3.1 THE QUESTIONNAIRE ...... 4 3.2 STUDY AREA ...... 5 3.3 RECRUITMENT OF PARTICIPANTS ...... 5

4 RESULTS ...... 6

4.4 COGNITIVE MAPS OF A FARMER ...... 17

List of Annexes

Annex S1 Dryland Salinity Questionnaire Annex S2- Participant Information Statement

2 List of Figures

Figure S1 Respondents property sizes Figure S2 Length of time on property- owned or knew the property well Figure S3 Respondents with properties affected by dryland salinity Figure S4 Extent of dryland salinity on properties Figure S5 Increase in dryland salinity Figure S6 Increase In dryland salinity overtime – trends Figure S7 Decrease in the dryland salinity on individual properties Figure S8 Decrease in dryland salinity over time – trends Figure S9 Impacts of dryland salinity on individual properties Figure S10 Relationship between dryland salinity and depth to groundwater Figure S11 Extent of dryland salinity in the Murray Darling Basin Figure S12 Increase in dryland salinity in the Murray Darling Basin Figure S13 Decrease in dryland salinity in the Murray Darling Basin Figure S14 Impacts of dryland salinity in the Murray Darling Basin Figure S15 Options for managing dryland salinity in the Murray Darling Basin Figure S16 Future trend of dryland salinity in the Murray Darling Basin Figure S17 Increase in land clearing in the Murray Darling Basin Figure S18 Decrease in land clearing in the Murray Darling Basin Figure S19 Relationship between land clearing and dryland salinity Figure S20 Cognitive map of a Murray Darling Basin farmer –dryland salinity causes Figure S21 Cognitive map of a Murray Darling Basin farmer –dryland salinity impacts Figure S22 Cognitive map of a Murray Darling Basin farmer –potential solutions to dryland salinity

3

1 INTRODUCTION

This supplement presents the results of a field survey and farmers’ consultation on multiple aspects of dryland salinity problem with the main purpose to support the development of reference modes and subsequent system dynamics analysis presented in the main thesis.

2 RESEARCH OBJECTIVES

The main goal of this supplementary research was to seek farmers (primary producers)’s perspective about the extent, trends and potential remedial measures for dryland salinity.

This information was needed to inform, validate and or revise the reference modes of dryland salinity problem presented in the Chapter 6. Moreover some information was needed to validate relationships amongst model variables presented in Chapters 7 and 8.

3 SURVEY METHODOLOGY 3.1 THE QUESTIONNAIRE

A questionnaire was developed seeking farmer’s perspectives on the following aspects of dryland salinity problem.

 Property level information: o Extend of dryland salinity on individual property. o Trends in dryland salinity (increase or decrease) – development of the problem over time. o Impacts of dryland salinity on individual properties. o Relationship of dryland salinity to depth to groundwater.

 Basin Scale information: o Extent of the dryland salinity problems.

4 o Trends (increase/decrease) in dryland salinity. o Impacts of drylad salinity. o Remedial measures. o Likely future trends (increase/decrease) for dryland salinity. o Likely future trends for land clearing. o Impact of land clearing on dryland salinity.

The questionnaire contained 19 multiple choice questions taking 15 minutes of farmer’s time to fill in the questionnaire. The questionnaire was tested in Canberra prior to sending it to field locations. A copy of the questionnaire is at Annex S1.

3.2 STUDY AREA

The questionnaires along-with a participants information statement and pre-addressed and postage paid envelop were distributed among primary producers in the upper, middle and lower parts of the Murrumbidgee catchments.

The Murrumbidgee River is a major river in the state of New South Wales, and the Australian Capital Territory (ACT). The word Murrumbidgee means "big water" in the local Aboriginal language. Murrumbidgee catchments cover 84000 square kilometers, and provides 25% of NSW’ fruit and vegetable production.

3.3 RECRUITMENT OF PARTICIPANTS

Participation in the survey was voluntary. Primary producers who wish to participate in the survey self-completed the written questionnaire and returned in the pre-addressed, postage paid envelop.

Survey questionnaires were distributed among primary producers through district agronomists or farmers’ organizations in the dryland salinity affected areas of upper,

5 middle and lower Murrumbidgee. This did not require participants to be identified by name or address and ensure strict observance of the privacy principles and ethical conduct of research.

4 RESULTS

Seven randomly selected participants filled in the questionnaire. The results of data analysis are presented in the following graphs (Figures S1 to S19) referring to a particular question number in the dryland salinity questionnaire (Annex S1). Most graphs are self explanatory. Additional explanations are provided where considered necessary.

Results of a detailed interview with a participating Murray Darling Basin farmer are presented in the form of cognitive maps in Figure 20 to 22.

50

45

40

35

30

25

20

Respondents (%) 15

10

5

0 Less than 100 100 hectares to 500 hectares to More than 1000 hectares 499 hectares 1000 hectares hectares

Figure 1 Property sizes of respondents (Question 1)

6 50

45

40

35

30

25

20 Respondents (%) 15

10

5

0 less than 10 yrs 10 to 20 yrs 20-30 yrs 30 to 50 yrs More than 50 yrs

Figure S2 Length of time on property- owned or knew the property well (Question 2).

No 38%

Yes 62%

Figure S3 Respondents with properties affected by dryland salinity (Question 3).

7 100

90

80

70

60

50

40 Respondents (%) 30

20

10

0 Less than 20% 20% to 50% 50% to 80% More than 80%

Figure S4 Extent of dryland salinity on properties (Question 4)

60

50

40

30

Respondents (%) 20

10

0 Last 10 yrs 1980 to 2000 1950 to 1980 None

Figure S5 Increase in dryland salinity (Question 5)

8 70

60

50

40

30 Respondents (%) 20

10

0 The increase in area The increase in area The increase in area The increase in area affected by dryland affected by dryland affected by dryland affected by dryland salinity between 2000 and salinity between 2000 and salinity between 1980 and salinity between 1980 and 2010 was more than the 2010 was less than the 2000 was more than the 2000 was less than the increase between 1980 increase between 1980 increase between 1950 increase between 1950 and 2000 and 2000 and 1980 and 1980

Figure S6 Increase In dryland salinity overtime – trends (Question 6).

60

50

40

30

Respondents (%) 20

10

0 Last 10 yrs 1980 to 2000 1950 to 1980 No

Figure S7 Decrease in the dryland salinity on individual properties (Question 7)

9 60

50

40

30

Respondents (%) 20

10

0 The decrease in area The decrease in area The decrease in area The decrease in area affected by dryland affected by dryland affected by dryland affected by dryland salinity between 2000 and salinity between 2000 and salinity between 1980 and salinity between 1980 and 2010 was more than 2010 was less than the 2000 was more than 2000 was less than the thedecrease between decrease between 1980 thedecrease between decrease between 1950 1980 and 2000 and 2000 1950 and 1980 and 1980

Figure S8 Decrease in dryland salinity over time – trends (Question 8)

35

30

25

20

15 Respondents (%)

10

5

0 Obvious salt Reduction in Impairment of Impairment of Other None crust/patches crop yields infrastructure water-supplies

Figure S9 Impacts of dryland salinity on individual properties (Question 9)

10 45

40

35

30

25

20

Respondents (%) 15

10

5

0 Area affected by dryland Severity of impacts of Area affected by dryland Don’t know salinity increases as dryland salinity increases salinity or severity of its depth to groundwater as depth to groundwater impacts is not affected by decreases decreases the depth to groundwater

Figure S10 Relationship between dryland salinity and depth to groundwater (Question 10)

70

60

50

40

30 Respondents (%)

20

10

0 Less than 20% 20% to 50% 50% to 80% More than 80% Don't know

Figure S11 Extent of dryland salinity in the Murray Darling Basin (Question 11)

11 50

45

40

35

30

25

20 Respondents (%) 15

10

5

0 Last 10 yrs 1980 to 2000 1950 to 1980 No

Figure S12 Increase in dryland salinity in the Murray Darling Basin (Question 12)

90

80

70

60

50

40

Respondents (%) 30

20

10

0 Last 10 yrs 1980 to 2000 1950 to 1980 No

Figure S13 Decrease in dryland salinity in the Murray Darling Basin (Question13)

12 30

25

20

15 Respondents (%) Respondents

10

5

0 Obvious salt Reduction in crop Impairment of Impairment of water Other None crust/patches yields infrastructure supplies

Figure S14 Impacts of dryland salinity in the Murray Darling Basin (Question 14)

13 30

25

20

15

Respondents (%) 10

5

0 Plantation Salt tolerant Salt Changes in None of these Other crops interception crop schemes husbandry

Figure S15 Options for managing dryland salinity in the Murray Darling Basin (Question 15)

14 45

40

35

30

25

20 Respondents (%) Respondents

15

10

5

0 Dryland salinity problem Dryland salinity problem Dryland salinity problem Dryland salinity problem The increase in dryland The decrease in dryland in the Murray Darling in the Murray Darling in the Murray Darling in the Murray Darling salinity between 2010 salinity between 2010 Basin will increase Basin will decrease Basin will increase Basin will decrease and 2020 will be more and 2020 will be less between 2010 and 2020 between 2010 and 2020 between 2020 and 2050 between 2020 and 2050 than the increase than the decrease between 2020 and 2050 between 2020 and 2050

Figure S16 Future trend of dryland salinity in the Murray Darling Basin (Question 16)

15 40

35

30

25

20

15 Respondents (%)

10

5

0 Last 10 yrs 1980 to 2000 1950 to 1980 No

Figure S17 Increase in land clearing in the Murray Darling Basin (Question 17)

60

50

40

30

Respondents (%) 20

10

0 Last 10 yrs 1980 to 2000 1950 to 1980 No

Figure S18 Decrease in land clearing in the Murray Darling Basin (Question 18)

16

60

50

40

30 Respondents (%) 20

10

0 Land clearing is one of the main Land clearing is one of the secondary Land clearing has not affected the causes of dryland salinity in the causes of dryland salinity in the extent of dryland salinity in the Murray Darling Basin Murray Darling Basin Murray Darling Basin

Figure S19 Relationship between land clearing and dryland salinity (Question 19)

4.4 COGNITIVE MAPS OF A FARMER

As a quality assurance on a representative farmer who volunteered to participate in the cognitive mapping was interviewed. The outcomes of interview were mapped and consulted with the farmer. Such cognitive maps are presented in the Figures S20 to S22.

17 3 clearing of deep rooted perennials , replacement with 10 Federal shallow rooted government pasture , legislation on 8 Federal and state clearing policies ... 1972 taxation system had incentives for 4 Government policies people who drain wetlands 17 Land clearing 12 On state level SA 13 periodic wet brought in clearing period bans in 1990 contractors went to 15 better Victoria contractors technologies , more went to Victoria efficient use of water 1 Dryland salinity 14 Farmers to feel 5 Poor understanding threatened ... going of soil moisture to loose viable relationships of water allocations 6 best practice different soils today might not be the best practice in past 11 Rapidity/fast 16 Impacts of tracking of putting irrigation irrigation system

7 Knowledge about farming ... why it is not good, Government 9 delivery losses departments tend to from open channel follow single theme 2 Poor irrigation system practices- flood irrigation inappropriate

Figure S20 Cognitive map of a Murray Darling Basin farmer – dryland salinity causes

18

8 Loss of biodiveristy

6 Restriction in the 7 precipitates- range of secondary impacts on agricultural infrastructure commodities that we can produce 5 Loss of income 1 Dryalnd saliniy affected areas, as compared to T non-saline areas 2 Loss of productivity

10 feel about place what ever they do 9 Social impacts does not has 3 Water quality impacts, natural supplies

FigureS20 Cognitive map of a Murray Darling Basin farmer –dryland salinity impacts

19

2 Greater cooperation between all levels of 3 Understanding government 8 Better management soil and land scape techniques that lead resources to better infiltration

1 Dryland salinity 4 Soil water balance

9 Reducing flood irrigation 7 Some schemes which will assist young farmers to take up career in 5 Better match agriculture between vegetation 6 Schemes for people types selected and who take on better soil resources technologies

Figure S22 Cognitive map of a Murray Darling Basin farmer –potential solutions to dryland salinity

20 Annex S1 Dryland Salinity Questionnaire

Postcode/s of the agricultural property ______

Dear Participant,

This survey seeks to help us understand how dryland salinity is affecting farmers. Your input is important.

This questionnaire consists of two parts. Part-A seeks your opinion on dryland salinity problem on your property. Part-B seeks your opinion on the dryland salinity problem at a broader scale, i.e., Murray Darling Basin. Please respond to the both parts.

Completing this questionnaire will take approximately 15 minutes. If you wish to participate in this survey, please respond to the questions below and return the questionnaire to the following address in the attached pre-addressed and postage paid envelop by 31 July 2010:

Dryland Salinity Project School of Engineering and Information Technology, University of New South Wales, Canberra Campus, Australian Defence Force Academy, Northcott Drive, Canberra, ACT 2600

Part A – Individual Property

1. What is the size of your agricultural property?

Less than 100 hectares. 100 hectares to 499 hectares.

500 hectares to 1000 hectares. More than 1000 hectares.

2. For how long have you been on this agricultural property or know this property well?

less than 10 years. 10 to 20 yrs. 20-30 yrs. 30 to 50 yrs. More than 50 yrs.

3. Is your property affected by dryland salinity?

Yes

No – Please proceed to Part B at page 2.

4. Approximately how much of your property area is currently affected by dryland salinity.

Less than 20%. More than 20% but less than 50%.

More than 50% but less than 80%. More than 80%.

5. Has the area affected by dryland salinity increased on your property?

In the last 10 years. From 1980 to 2000. From 1950 to 1980. None, Proceed to Q6.

6. If increased, please select the statement/s that describes the situation on your property.

The increase in area affected by dryland salinity between 2000 and 2010 was more than the increase between 1980 and 2000.

The increase in area affected by dryland salinity between 2000 and 2010 was less than the increase between 1980 and 2000.

The increase in area affected by dryland salinity between 1980 and 2000 was more than the increase between 1950 and 1980.

The increase in area affected by dryland salinity between 1980 and 2000 was less than the increase between 1950 and 1980.

1 7. Has the area affected by dryland salinity has decreased on your property?

In the last 10 years. From 1980 to 2000, From 1950 to 1980. No. Proceed to Q8.

8. If decreased, please select the statement/s that describes the situation on your property.

The decrease in area affected by dryland salinity between 2000 and 2010 was more than the decrease between 1980 and 2000.

The decrease in area affected by dryland salinity between 2000 and 2010 was less than the decrease between 1980 and 2000.

The decrease in area affected by dryland salinity between 1980 and 2000 was more than the decrease between 1950 and 1980.

The decrease in area affected by dryland salinity between 1980 and 2000 was less than the decrease between 1950 and 1980.

9. Which of the followings are impacts of dryland salinity on your property?

Obvious salt crust/patches.

Reduction in crop yields.

Impairment of property infrastructure, e.g., buildings, roads etc.

Impairment of water supplies.

Other --- Please specify______

None.

10. Relationship between dryland salinity and depth to ground water. Please select one of the followings:

On my property area affected by dryland salinity increases as depth to groundwater decreases, i.e., watertable is elevating.

On my property severity of impacts of dryland salinity increases as depth to groundwater decreases, i.e., watertable is elevating.

Area affected by dryland salinity or severity of its impacts on my property is not affected by the depth to groundwater, i.e. elevating watertable.

Don’t know.

Part B – Dryland Salinity Problem in the Murray Darling Basin.

11. Approximately how much of the area of the Murray Darling Basin is currently affected by dryland salinity.

Less than 20%. More than 20% but less than 50%.

More than 50% but less than 80%. More than 80%.

12. Has the area affected by dryland salinity increased in the Murray Darling Basin?

In the last 10 years. From 1980 to 2000. From 1950 to 1980. No.

13. Has the area affected by dryland salinity decreased in the Murray Darling Basin?

In the last 10 years. From 1980 to 2000. From 1950 to 1980. No.

2 14. Which of the following impacts of dryland salinity have occurred in the Murray Darling Basin?

Obvious salt crust/patches

Reduction in crop yields

Impairment of property infrastructure, e.g., buildings, roads etc

Impairment of water supplies.

Other --- Please specify______

None

15. Which of the following actions you consider suitable for tackling dryland salinity problem?

Plantation

Salt tolerant crops

Salt interception schemes

Changes in crop husbandry

None of these

Other, please specify______

16. Future Trend of dryland salinity problem in the Murray Darling Basin. Please select as many as you consider likely future scenarios of dryland salinity

Dryland salinity problem in the Murray Darling Basin will increase between 2010 and 2020.

Dryland salinity problem in the Murray Darling Basin will decrease between 2010 and 2020.

Dryland salinity problem in the Murray Darling Basin will increase between 2020 and 2050.

Dryland salinity problem in the Murray Darling Basin will decrease between 2020 and 2050.

The increase in dryland salinity between 2010 and 2020 will be more than the increase between 2020 and 2050.

The decrease in dryland salinity between 2010 and 2020 will be less than the decrease between 2020 and 2050.

17. Has land clearing increased in the Murray Darling Basin?

In the last 10 years. From 1980 to 2000. From 1950 to 1980. No.

18. Has land clearing decreased in the Murray Darling Basin?

In the last 10 years. From 1980 to 2000. From 1950 to 1980. No.

19. Please select one of the following statements that you consider appropriate:

Land clearing is one of the main causes of dryland salinity in the Murray Darling Basin.

Land clearing is one of the secondary causes of dryland salinity in the Murray Darling Basin.

Land clearing has not affected the extent of dryland salinity in the Murray Darling Basin.

Thank you for participating in this survey. 3

Supplement II

Tutorial:

1. Linkages between dryland salinity policies, actions and the rate of change in land affected by dryland salinity in the model presented in Chapter 8.

2. A step-by-step guide on how the models can be used for system dynamics analysis of dryland salinity.

i Contents

1. Introduction ...... 1

2. Dryland salinity mitigation options ...... 1

3. Rate of change in land affected by dryland salinity ...... 3

4. The linkage between mitigation options and the model parameters ...... 6

5. A step-by-step guide on using system dynamics modelling for dryland salinity analysis...... 10

5.1 Step I - Clarify objectives of analysis...... 12

5.2 Step II - Collect information and populate the initial conditions for simulation...... 13

5.3 Step III - Estimate and populate time delays...... 14

5.4 Step IV - Identify mitigation options...... 15

5.5 Step V - Run simulation and observe graphs in the observation/output window...... 17

5.6 Step VI - Debate the output and improve the model ...... 18

ii List of Figures Figure SII 1 Farmers’ perceptions about suitable options for mitigating dryland salinity…….2 Figure SII 2 Rate of land clearing…………………………………………………………...... 3 Figure SII 3 Table function - land clearing fraction………………………………………...... 4 Figure SII 4 Rate of land reclamation………………………………………………………….5 Figure SII 5 Linkages between land clearing controls and the rate of land clearing…………..6 Figure SII 6 Linkages between potential policy options and the rate of land reclamation...... 8 Figure SII 7 Overall structure of model use methodology……………………………………11 Figure SII 8 Objectives worksheet……………………………………………………………12 Figure SII 9 Initial values data input window...... 13 Figure SII 10 Time delays data input window...... 15 Figure SII 11 Effectiveness of control treatments window...... 17 Figure SII 12 Land fraction data input window...... 17 Figure SII 13 Output/observation window...... 18

List of Tables Table SII 1 Template for options matrix……………………………………………………..16

iii 1. Introduction

While the focus of the thesis is methodological improvements and the focus of the model presented in Chapter 8 (pages 8-10 to 8-35) is on gaining strategic level insights, linkages between policy options, farm level actions and the rates of change used in the model can be demonstrated. The main goal of this tutorial is to explain how the policy options and farm level actions can be linked to the rates of change in dryland salinity presented in the model (pages 8- 10 to 8-35) and to provide a step-by-step guide on how the model can be used for system dynamics analysis of dryland salinity.

Supplement I described the results of a farmer survey and cognitive mapping that suggested actions that farmers consider beneficial at both farmer and government policy levels. This tutorial uses those options to identify intermediate steps between dryland salinity policies, actions taken at farm level and the rates of change presented in the model in Chapter 8.

The main policy options previously discussed in Chapters 2, 6 and 7 are listed followed by a recapture of rates of change in dryland salinity used in the model presented in Chapter 8. Then linkages between the policies, actions and the rates of change are demonstrated using diagrams depicting intermediate steps and a step-by-step guide is provided to help lead readers through the use of the system dynamics analysis of dryland salinity.

2. Dryland salinity mitigation options

A detailed discussion on dryland salinity policies is presented in Chapter 2 (pages 2-24 to 2-37) and Chapter 7 (pages 7-8 to 7-29). A list of dryland salinity programs is available in Annex B (pages B1 to B5). An analysis of the policies/programs indicates that the focus of dryland salinity policy development has been on the following: • Land clearing controls; • Improving water allocation through pricing and operational rules, i.e. water policy; • Awareness raising, education, and community group facilitation policies, i.e. Landcare, Waterwise on the farm, and agricultural extension policies that focus on behavioural change towards land and water-use; • Engineering options;

1 • Agronomic options and vegetation restoration policies;

A farmers’ survey presented in Supplement I captured farmers’ views about suitable options for tackling dryland salinity. Results are illustrated in Figure SII 1 below.

30

25

20

15

Respondents (%) 10

5

0 Plantation Salt tolerant Salt Changes in None of these Other crops interception crop schemes husbandry

Figure SII 1: Farmers’ perceptions of suitable options for mitigating dryland salinity.

A cognitive map of a Murray Darling Basin farmer (Figures S20-S22, Supplement I pages: 18- 20) identified the following additional options: • Schemes which assist young farmers to take up careers in agriculture. • Schemes for people who take on better technologies. • Greater cooperation between all levels of government. • Better understanding of soil and landscape resources as well as soil water balance. • Better match between vegetation types and soil resources.

The dryland salinity policies and actions described above constitute a varied agenda. Some are of a qualitative nature and support other options, for example, greater cooperation between all levels of government may support other programs through better identification, management and measurement of success of program.

2

3. Rate of change in land affected by dryland salinity

The model presented in Chapter 8 conceptualises rates of change in dryland salinity in terms of rate of land clearing and the rate of land reclamation. Detailed discussion on these parameters is provided in Chapter 8 (pages 8-18 to 8-20 and 8-23 to 8-24). Figure 8.5 (page 8-19) is reproduced below for quick reference.

Rate of land clearing

Fraction of land LAND UNDER under natural NATURAL VEGETATION vegetation being cleared

Random time delay in land clearing Fluctuations in Time delay in land time delay clearing

Figure SII 2: Rate of land clearing (reproduced from Figure 8.5).

A sub-model providing the land clearing rate is shown in Figure SII 2. For this sub-model rate of land clearing is defined as a fraction of the land under natural vegetation.

Rate of land clearing = land under natural vegetation * fraction of land under natural vegetation/time delay in land clearing.

The fraction of land under natural vegetation being cleared is modelled on the basis of the historical data of land clearing developed from different references (Chapters 2, 6 and 7). The fraction is provided as a table function shown in Figure SII 3.

Time delay in land clearing is a user defined variable and includes the time that is consumed in planning, land acquisition, obtaining permission for land clearing, arrangements for machinery, acquisition and movement of machinery, and felling and export of logs from the area. As there may be varying times for different areas due to land clearing operations and checking communities’ sensitivities, a random variable is used to represent fluctuations in time delay.

3 Fluctuations in time delay = 1-(0.5*Random())

This formulation returns random numbers between 0 and 1, with a new seed used during each simulation run.

0.35

0.3

0.25

0.2

0.15 Fraction of land under natural cleared vegetation being 1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure SII 3: Table function-land clearing fraction.

4

Rate of land reclamation

CLEARED LAND Effectiveness of EITHER SALT control treatments AFFECTED OR AT RISK OF BECOMING SALT AFFECTED

Fraction of land at risk of becoming Time delay between salt affected on land at risk of which control becoming salt treatment is applied affected and land not at risk of becoming salt affected under a certain control treatment

Figure SII 4: Rate of land reclamation.

In this model, the rate of reclamation has been defined as a function of the fraction of land salt affected or at the risk of becoming salt affected on which a control treatment is applied, time delay and effectiveness of the control treatments. The fraction of land either salt affected or at risk of becoming salt affected is a user controlled parameter. Effectiveness of a control treatment is a specific parameter and may depend upon the scientific studies of the relationship of a certain control treatment and the capability of that treatment to address dryland salinity.

Rate of land reclamation = cleared land either salt affected or at the risk of becoming salt affected * fraction of land at risk of becoming salt affected on which control treatment is applied * effectiveness of control treatments/time delay between land at risk of becoming salt affected and land not at risk of becoming salt affected under a certain control treatment.

5 4. The linkage between mitigation options and the model parameters

Figures SII 4 and 5 depict the linkage between potential mitigation options, suggest intermediate steps and demonstrate how certain policy options can influence the rate of change in dryland salinity. Preventive measures such as land clearing controls through restrictions and permits can influence the fraction of land being cleared and/or can delay the onset of dryland salinity.

One set of mitigation options contains land clearing controls through bans, permits and raising public awareness about the impacts of land clearing on dryland salinity. Such policies when effective can reduce the land being cleared and therefore provide input to the variable fraction of land being cleared in the model. Such policies can also increase time delays in land clearing due to the time consumed in clearing applications, appraisals, approvals, permits and commissioning of land clearing operations. These linkages are shown in Figure SII 5 with bold brown arrows.

Land clearing control: bans, permits etc.

Rate of land clearing

Fraction of land under natural vegetation being cleared LAND UNDER NATURAL VEGETATION

Random time delay in land clearing

Time delay in land clearing Fluctuations in time delay

Figure SII 5: Linkages between land clearing controls and the rate of land clearing.

6 The set of policies that may enhance remediation or mitigation includes policies aimed at addressing the availability of farm labour, improving water-use on farm, awareness raising, community group facilitation programs (for example, Waterwise on the farm) and agricultural extension programs that focus on behavioural change towards land and water use. Details of various control options and policies are presented in Chapter 2 (pages 2-24 to 2-37), Chapter 7 (pages 7-8 to 7-29) and Annex B (pages B1 to B5). For analysis of these policies using the model presented in Chapter 8 (pages 8-10 to 8-35), the information about the effectiveness of these policies, the time delay in implementation of the policies and the impacts generated and geographical coverage will be needed. This information may feed into the model parameters, effectiveness of control strategy, fraction of the land at risk of becoming salt affected on which control strategy is applied and time delay between land at risk of becoming salt affected and shown in the Figure SII 5 by the bold brown arrows.

7

Engineering Schemes for options,salt Effecti people who interception veness take on better technologies Delay schemes Covera ge Effecti veness Agronomic options and vegetation restoration Rate of land reclamation Delay policies Covera ge

effectiveness of control Better match strategy between CLEARED LAND EITHER SALT vegetation AFFECTED OR AT THE RISK OF types and soil BECOMING SALT AFFECTED resources Better understanding of Time delay between soil and landscape land at risk of becoming Fraction of land at risk of resources as well salt affected and land becoming salt affected on not at risk of becoming which control strategy is as soil water salt affected under a applied balance certain control strategy

Covera Delay ge Effecti veness Awareness raising, education, community Improving water Schemes which group facilitation policies, allocation assist young i.e., land care, water wise through pricing farmers to take on the farm and and operational up career in agricultural extension rules, i.e., water agriculture policies that focus on policy behavioral change towards land and water-use Figure SII 6: Linkages between potential policy options and the rate of land reclamation presented in Chapter 8 (pages 8-23 to 8-24).

8 The second set of policies consists of schemes for people who take on better technologies, programs to achieve better match between vegetation types and soil resources, and improving understanding of soil and landscape resources. Such policies aim at encouraging farmers to take on better technologies and/or improvement in attitudes towards effective use of land and water resources. Linkages between these policies and the model parameters are shown in Figure SII 5 by bold brown arrows.

The third set of programs and policies includes engineering options such as salt interception schemes and/or the programs aimed at improving land drainage. Details about such options are discussed in Chapter 2 (pages 2-25 to 2-28) and linkages to the model parameters are shown in Figure SII 5.

If the model is to be calibrated to present day values, it will put forward a research agenda that needs to measure the relative effectiveness and time delays for each option in situ. For example, how effective are agronomic measures as compared to engineering options, or how much time lag is there between implementation of a certain program, for example, a certain crop mix, water- use efficiency and realisation of expected results. It will include field evaluation of effectiveness of awareness raising campaigns. Development of such a research agenda is out of scope of this research and is recommended for future research. A step-by-step process demonstrating how the model can be used for system dynamics analysis of dryland salinity is described in the following section.

9 5. A step-by-step guide on using system dynamics modelling for dryland salinity analysis

Understanding the purpose of system dynamics modelling is central to successfully using such models. The purpose of a system dynamics inquiry is to improve understanding about dynamics of problems in complex systems. It deals with the development of understanding how complex systems change over time and focusses on goals and the specific questions to be answered. The system dynamics modelling process fosters learning at each step and the models are not specifically designed to generate predictions. To aid learning, the system dynamic models provide broader understanding about a particular problem in focus. System dynamics models can also work as learning laboratories or micro-worlds for training where cognitive capacities of managers can be enhanced.

The dryland salinity model presented in Chapter 8 (8-10 to 8-35) was developed using a multi- methodology approach to understand the dynamics of dryland salinity mitigation options. The model incorporates some of the variables identified in the various causal loop diagrams and provides opportunities for what-if analysis of various dryland salinity mitigation measures. The model does not intend to predict the quantity of actual salt affected lands or the quantity of salts at a certain geographical location and it should not be used as a framework for statistically based inferences. The model provides a communication tool for generating and advancing discussions about system dynamics of dryland salinity, mitigation options and the improvement in modelling methodologies.

Figure SII 7 shows the overall structure highlighting major steps on how to use this model. In the later parts of this tutorial each step is explained in detail providing references to the sources of further information/guidance where needed.

10

Group discussions Stakeholder Step-VI presentations Step-1 Clarify Educate and objectives for debate analysis

Step-II Step-V Establish Initial Simulate and conditions for compare options simulation

Step-IV Step-III Identify Estimate time mitigation option delays

Research Case studies Group discussions

Figure SII 7: Overall structure of the model use methodology.

11 5.1 Step I - Clarify objectives of analysis

The first step in understanding and using the model is to identify the objectives of the analysis, for example, to learn more about a certain dryland salinity mitigation option or a combination of options or impacts of time delays on dryland salinity over time. If there is more than one objective, then a list of the objectives and a brief description of each objective can be prepared.

The purpose of a model may usually include clarification of knowledge and understanding of the dryland salinity problem to discover policies that will improve system behaviour as a communication and unifying medium. Figure SII 8 provides a guiding template.

Figure SII 8: Objectives worksheet

Main goal of the analysis ------Objectives ------Uses of the analysis ------

12 5.2 Step II - Collect information and populate the initial conditions for simulation

Establish initial conditions at the start of simulation and input data in the initial values window through the slider bars as shown in Figure SII 9. • Total area: This is the size of the area under study. For example, to investigate an area of 1000 square kilometres, it is 1000. It may be an area of a river basin and/or its catchments. • Fraction of area under natural vegetation at the start of simulation: This is the benchmark area, for example, if an area of 1000 square kilometres under investigation has 40% of its area (400 square kilometres) under natural vegetation, the fraction will be 0.4.

Initial Values: these are the values of the stocks that are used at the start of simulation period

Total Area

0 200,000 400,000 600,000 800,000 1,000,000 1,200,000 1,400,000 sq km Non-commercial use only! Fraction of area under natural vegetation at the start of simulation

Initial 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Non-commercial use only! Fraction of area cleared at the start of simulation

0 0.10.20.30.40.50.60.70.80.91

Non-commercial use only! Fraction of area at risk of becoming salt affected at start of simulation

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Non-commercial use only!

Figure SII 9: Initial values data input window.

• Fraction of area cleared at the start of simulation: This is the area that has already been cleared of natural vegetation. Using the above example, if an area of 1000 square

13 kilometres under investigation has 60% of the area that had already been cleared, then the fraction becomes 0.6. • Fraction of area at risk of becoming salt affected: This is the area of land that is already salt affected or at the risk of becoming salt affected at the start of simulation. For example if an area under investigation (1000 square kilometres in the above example) has 20% (200 square kilometres) of its area that is salt affected, this fraction will be 0.2.

5.3 Step III - Estimate and populate time delays

The details about types of time delays and how to work them out are provided in Chapter 11 under Business Dynamics (Sterman 2000). Specific details about time delays used are explained along with individual modules in Chapter 8 (pages 10 to35). Using the slider bar provided in the time delays data input window shown in Figure SII 10 enter data about time delays using the input control window:

• Time delay in land clearing: It is a user defined variable and includes the time that is consumed in planning, land acquisition, getting permissions for land clearing, arrangements for the machinery, acquisition and movement of machinery, and felling and export of logs from the area. As there may be varying time delays for different areas, land clearing operations and communities, a random variable is used in the model that provides varying fluctuations for testing sensitivities. • Time delay in non-saline land becoming either salt affected or at the risk of becoming salt affected: It would vary according to land policies, specific geo-physical and social set-up and market forces. Time delay in land becoming salt affected is a user controlled parameter. The default value is a rough estimate of 30 to 40 years. Actual time delay in land becoming salt affected may be assessed through experimental research. To answer what if questions multiple time delay scenarios can be developed and tested. • Time delay in land reclamation: Time delay between land at risk of becoming salt affected and becoming land not at risk under a certain control treatment. Time delay in a piece of land going out of risk of becoming salt affected is also user controlled. The default value is kept at 30 years. • Time delay between land at risk of becoming salt affected and land returning to natural vegetation cover may vary based on a number of factors including: location of a piece of

14 land, type of vegetation and other geophysical conditions. It is a user controlled parameter whose default value, i.e. the maximum time a piece of land takes in returning to its natural vegetation cover is kept at 50 years.

Time Delays: this is time consumed in a process or change of state

Time Delay in Land Clearing

0 1020304050 yr Non-commercial use only! Time Delay between non saline land becoming salt affected

0 102030405060708090100 yr Non-commercial use only! T ime delay in land becomingInitial salt affected

0 102030405060708090100 yr

Non-commercial use only! Time delay between land at risk of becoming salt and becoming land not at risk under a certain control treatment

0 102030405060708090100 yr Non-commercial use only! Time delay between land at risk of becoming salt affected and returning to natural vegetation

0 102030405060708090100 yr Non-commercial use only!

Figure SII 10: Time delay - data input window.

5.4 Step IV - Identify mitigation options

Dryland salinity mitigations options also referred to as control treatments may be single option or a combination of different options. This could be better identified through research, results of field studies and/or through brainstorming and/or knowledge elicitation sessions with focus groups. The procedures for knowledge elicitation are explained in Ford and Sterman (1998). The collaborative model building is explained in Vennix and Richardson, et al. (1997).

15 Prepare a list of the potential mitigation options. For each mitigation option, determine/estimate effectiveness and the area on which this option can be implemented. Vary values for the effectiveness and area. The options may arise from intuitive insights from experience and/or proposals advanced by people working on dryland salinity. This will provide a wide range of scenarios. Enter these estimates in the options matrix shown in Table SII 1.

Table SII 1: Template for generating options matrix Option Option name- description Effectiveness Area for implementation identifier (0 to 1) Example: Change in land-cover 0.4 (40%) 100 sq km Option A

16 Using the slider bars as shown in Figures SII 11 and SII 12, input effectiveness and area values for options to be tested.

Effectiveness of Control Treatment

Effectiveness ofInitial Control treatments

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Non-commercial use only!

Figure SII 11: Data input window for effectiveness of control treatments/mitigation strategies.

Land fraction: this is fraction of a certain category of land

Fraction of land at risk of becoming salt affected that is returning to natural vegetation

Initial 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Non-commercial use only! Fraction of land at risk of becoming salt affected on which control treatment is applied

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Non-commercial use only!

Figure SII 12: Land fraction data input window.

5.5 Step V - Run simulation and observe graphs in the observation/output window

The observation/output window provides a medium for observing the changes in the stocks as a result of the changes in model parameters. It presents information in a graph format. Multiple variables can be brought to and observed in this window. An example is illustrated in Figure SII 13 that shows the behaviour of stocks. Any variables can be brought to this window and changes observed. Initially, the behaviour of these stocks is presented in the observation window. However, it can be increased or decreased depending upon the analytical needs. Run the model

17 with different scenarios, varying sets of initial value and time delays and observe the input in the output window and conduct a comparative analysis of various options listed in the options matrix. Such observations identify the options that show better promise. Such observations may also identify new options for testing.

ha

80,000,000 Cleared land

60,000,000

40,000,000 Land either salt affected or at the risk of becoming salt 20,000,000 affected

Land under natural vegetation

1 Jan 1900 1 Jan 1950 1 Jan 2000 1 Jan 2050 1 Jan 2100 Non-commercial use only!

Figure SII 13: Example of the output/observation window.

5.6 Step VI - Debate the output and improve the model

Debate the strengths and weaknesses of both the options and the model and improve knowledge and the model and inform decisions about mitigation options. This presents the greatest challenge and provides avenues for critical review. The debate will highlight aspects for model improvement and may also identify further options to be tested.

18