CCRES Systems Simulation Model: User guide. C.S. Smith and R.G. Richards, The University of Queensland

Transcription

1 Capturing Coral Reef and Related Ecosystem Services Project CCRES Systems Simulation Model: User guide C.S. Smith and R.G. Richards, The University of Queensland

2 Acknowledgements We would like to acknowledge the following for their help in developing the Systems Simulation Model: The World Bank Global Environment Facility The University of Queensland Currie Communications DINAS Marine and Fisheries, Selayar El Nido Foundation El Nido Local Government Unit Institut Pertanian Bogor Palawan Council for Sustainable Development Palawan State University University of California, Davis University of the Philippines Marine Science Institute Cover image: The CCRES System Simulation Model can be used to see the impact of different activities on coastal ecosystems, including seagrass (Photo: G Sheehan) Page 2

3 Table of Contents Acknowledgements... 2 INTRODUCTION... 7 ABOUT CCRES... 7 SECTION 1 BACKGROUND TO SYSTEMS THINKING AND SYSTEM DYNAMICS... 8 Systems thinking... 8 Introduction... 8 Systems thinking and policy resistance... 8 Causal loop development... 8 Feedback loops... 9 Structure and behaviour of dynamic systems Introduction Modes of behaviour Inferring structure from behaviour Stocks and flows Introduction Identifying stocks and flows Generic stock and flow structures Creating a stock and flow model example compound interest SECTION 2 INTRODUCTION TO THE SYSTEMS SIMULATION MODEL Introduction Model purpose Model limitations Software requirements to run the model Model architecture SECTION 3 HOW TO USE THE SYSTEM DYNAMICS MODEL Opening Stella Architect Opening the Systems Simulation Model Exploring the model Overview of the modules Model datasets Introduction Linking the model to the data The Systems Simulation Model interface Opening the user interface Navigating the interface Page 3

4 Entering the simulation part of the interface Key indicators Slider menus Case study Adjusting the run specifications of the model Setting up an interface page Adding a slider to the interface Adding a graph to the interface Setting up a new site Introduction Data sources Setting up a data spreadsheet SECTION 4 TESTING THE SYSTEM DYNAMICS MODEL Testing for mass balance Introduction Setting up a mass balance experiment Ghosted variables Comparison with historical data Extreme conditions testing (test / simulation model) Page 4

5 Figures Figure 1. Systems thinking applied a chicken population. The left panel shows a reinforcing loop (what came first, the chicken or the egg?) and the right panel shows a balancing loop (why did the chicken cross the road?)... 9 Figure 2. Common modes of behaviour in dynamic systems Figure 3. Example of how stocks, flows and clouds are combined to make a stock and flow model.. 13 Figure 4. Causal loop diagram of population Figure 5. Stock and flow representation of the births reinforcing loop of the population model Figure 6. Stock and flow representation of the deaths balancing loop of the population model Figure 7. Stock and flow representation of the full population model containing the births reinforcing loop and the deaths balancing loop Figure 8. Generic stock and flow configurations for linear, exponential growth, exponential decay and convergence behaviour Figure 9. Causal loop diagram for compound interest Figure 10. Stock and flow model for compound interest. Also shown is the behaviour over time graph of the stock (money in account) Figure 11. Architecture of the system dynamics model Figure 12. Stella Architect icon as it might appears in the start menu Figure 13. Opening screen for Stella Architect Figure 14. Accessing the file options in Stella Architect to load the System Dynamics Model Figure 15. The model view when the System Dynamics Model is first opened Figure 16. Finding stocks, flows and auxiliary variables using the Find function in Stella Architect.. 24 Figure 17. Input data requirements for the Tourism / Hotels module Figure 18. Example of the properties for an input variable ( tourist length of stay ). The dialogue window indicates that the input variable has a value of 1.34 weeks Figure 19. Click on Model from the toolbar and select Import Data from the drop-down menu. 29 Figure 20. Import data window with no data sets linked Figure 21. Import data window with import link added through clicking on green icon. Then click on Browse button to open file explorer where the data file is located Figure 22. Navigate to the folder where the data file is located and open Figure 23. Data file linked and ready to be imported Figure 24. Click on the interface window button to enter the interface Figure 25. The Model Interface as it appears when opened. The vertically arranged panels on the left are the separate pages that make up the interface Figure 26. Click on the down arrow next to the Interface icon and select Present Fullscreen Figure 27. The Interface operating in fullscreen mode with the menu buttons now active Figure 28. Clicking on any of the top five buttons opens a dialogue window with contextual information. This example shows the dialogue box when About the Model button is pressed Figure 29. The Main Menu of the Interface Figure 30. The Indicator Menu of the Interface Figure 31. The seven key indicator graphs. Note that per capita income (tourism, agriculture, fishing) are at present value Figure 32. Indicator graphs for tourism Figure 33. Marine sliders interface layout Figure 34. Terrestrial sliders interface layout Figure 35. Case study interface screen Figure 36. Case study comparing base case and traditional boat limit (n = 500) Page 5

6 Figure 37. Case study after surveillance effect increased from 0 (no surveillance) to 50% (surveillance has 50% effect in decreasing destructive fishing) Figure 38. Boat damage to reef decreased Figure 39. To navigate from the Interface window to the Model window, click on the model icon Figure 40. The run specifications can be adjusted in the Model window by clicking Model then run specs Figure 41. To add a new interface, use the left scroll area to select where you would like to place the new interface page (e.g. as indicated by the top dashed square) and then click Add Page Figure 42. The new interface page as it appears Figure 43. The icons used to add a chart (left) and a slider (right) Figure 44. A new slider has been added to the interface page. Double clicking on this opens a window where properties can be assigned Figure 45. Variable window showing a list of variables that can be linked to this slider Figure 46. Window showing the elements arrayed under the selected variable Figure 47. Adding a new chart to the interface page Figure 48. The graph window where its properties can be specified. Check the Comparative check box to allow the graph to show multiple runs Figure 49. Simple population model with births, deaths and carrying capacity Figure 50. The population model showing growth in the population (births > deaths) until the population reaches carrying capacity (births = deaths) Figure 51. Configuring the Population model to calculate the mass balance. The stocks appear with dashed lines, indicating that they are ghosted variables Figure 52. Output plot from the mass balance test showing the change in mass. If the model is correctly set up then the output should be zero (as shown) Figure 53. Comparison of a simulation model output with historical data Figure 54. Adjustment of normal birth rate improves model fit to data although there is increasing under-estimation towards the end of the model run Figure 55. Model dynamics when Carrying Capacity is set to zero the population falls rapidly to zero as expected Tables Table 1. Modules of the System dynamics model Page 6

7 INTRODUCTION This User Guide outlines the key features of an interactive process-based model (the Systems Simulation Model - SSM) that simulates the dynamics of a socio-ecological system. This SSM has been developed through the Systems Analysis activity of Component 2 ( Generating robust local economies that capture and sustain marine ecosystem services ) for the Capturing Coral Reef & Related Ecosystem Services (CCRES) project. This aim of this User Guide is to provide the context for the development of the SSM and to highlight to the User its main features. This involves introducing the concepts of systems thinking and system dynamics, which provide the theoretical and methodological bases for the SSM. This User Guide is set out in the following four sections: Section 1 Background information on systems thinking and system dynamics Section 2 Introduction to the Systems Simulation Model Section 3 Description on how to use the Systems Simulation Model Section 4 Common tests used in developing system dynamics models ABOUT CCRES Our coastal ecosystems coral reefs, mangroves and seagrass beds provide fish to eat and sell, support tourism and protect the coastline from storms. Coastal communities rely on these ecosystems for their livelihoods and food security. Unfortunately, these ecosystems are under threat from pollution, overfishing, unsustainable development and climate change. The CCRES project is working to ensure the long-term sustainability of these coastal ecosystems with models, tools and knowledge products to support planning. At the same time this project seeks to unlock new, sustainable income streams for the communities which rely on these ecosystems. The CCRES project has involved local, national and regional communities, businesses and policymakers working with scientists and other experts from a range of fields. Our multidisciplinary approach has involved collaboration between leading centres of discovery, learning and engagement in North America, Australia and the East-Asia Pacific region. Page 7

8 SECTION 1 BACKGROUND TO SYSTEMS THINKING AND SYSTEM DYNAMICS Systems thinking Introduction Systems thinking is a methodology that is used to enhance learning in complex systems, to understand connectedness and interactions, to understand feedback and to understand dynamics (behaviour over time). It is a method to help us learn about dynamic complexity, to understand the sources of policy resistance and to help us design effective policies (policies that have intended consequences and avoid unintended consequences). Learning about complex systems is more than just technical skills and mathematics. It requires: Tools to elicit and represent mental models about the nature of difficult problems Formal models and simulations to test and improve our mental models Methods to improve scientific reasoning and group processes to overcome the defensive routines of individuals and teams (which are very common in politics for example) Systems thinking and policy resistance Using a systems thinking approach is important because policies that we create and decisions we make often have unanticipated side effects and unintended consequences. These unintended consequences lead to policy resistance, which is the tendency for interventions to be delayed, diluted or defeated by the response of the system to the intervention. Complex system behaviour arises from interactions and feedbacks among system components, not from the number of components. The behaviour of all systems arise from the interaction of just two feedback loops positive (or reinforcing) and negative (or balancing). Positive feedback loops reinforce or amplify change while negative feedback loops balance or counteract change. Shifts in feedback loop dominance over time causes complex system behaviour. Causal loop development Causal loop diagrams (CLDs) are used to qualitatively model the causal relationships among a set of variables within a system. They capture our dynamic hypothesis; capture the mental models of individuals and teams; and communicate important feedback loops. A CLD consists of variables connected by arrows. A variable is a condition, situation, action or decision that can influence, or be influenced by, other variables. They can be quantitative (e.g., profit, population, temperature, etc.) or qualitative (e.g., motivation, trust, burnout, etc.). Figure 1 is an example of a basic system describing the number of chickens there are over time. The direction of the causal relationship between interconnecting elements is indicated by the arrow that connects. For example, the left panel of Figure 1 shows two elements (Chickens and Eggs) that represents the relationship between the number of chickens and the number of eggs. There are two causal relationships shown: The number of chicken Eggs directly affects the number of Chickens The number of Chickens directly affects the number of Eggs Page 8

9 The other attribute that is specified for each causal relationship is the direction, known as the polarity, which is either positive () or negative (-). A positive polarity indicates that the interconnected elements move in the same direction whilst a negative polarity indicates that they move in an opposite direction. For example, the right panel shows two elements (Chickens and Road Crossings) that represent the relationship between the number of chickens and how many road crossings they undertake. As the number of Chickens increases, the number of Road Crossings also increases (positive polarity). As the number of Road Crossings increases, the number of Chickens decreases that is the chickens have left the system. Figure 1. Systems thinking applied to a chicken population. The left panel shows a reinforcing loop (what came first, the chicken or the egg?) and the right panel shows a balancing loop (why did the chicken cross the road?) (source: Sterman, J.D. (2000). Business Dynamics. Systems thinking and modeling for a complex world. McGraw-Hill, Boston. Fig 1-5) Feedback loops Figure 1 shows the behaviour over time (BOT) graphs for two feedback loops. The left panel represents a reinforcing loop where the magnitude of the two elements (Eggs and Chickens) increasingly get bigger at a faster rate (as indicated in the chart on the left side). The right panel represents a balancing loop where the increase in Chickens increases Road Crossings, which then results in decreasing the number of Chickens over time. Reinforcing feedback loops Reinforcing feedback loops amplify change within systems over time and are also referred to as vicious and virtuous cycles. Compound interest in a bank account is an example of a virtuous reinforcing loop. Under-investment in training that leads to poor financial performance of a firm, Page 9

10 which in turn leads to further under investment in training is an example of a vicious reinforcing loop. The key rule for all reinforcing loops is that they create either exponential growth or exponential decline over time. Exponential growth = doubling time Exponential decline = halving time The same reinforcing loop can generate either exponential growth or decline depending on how the loop is triggered Balancing feedback loops Balancing feedback loops counteract change within systems over time and are also referred to as counteracting feedback loops. Balancing loops seek to achieve or maintain targets or goals sometimes these targets are explicitly stated as target dangles. A thermostat is an example of a balancing feedback loop. Referring back to Figure 1, the loop on the right panel is a balancing loop. As the chicken population gets higher, the number of Chickens lost to Road Crossings increases, which reduces the number of Chickens. The difference between reinforcing and balancing loops at a glance Reinforcing loops have zero or an even number of negative or - polarities Balancing loops have an odd number of negative or - polarities Structure and behaviour of dynamic systems Introduction The dynamic behaviour of systems arises from their structure. The structure consists of the reinforcing and balancing feedback loops and it is the interaction of these feedback loops and changes in their dominance over time that leads to system behaviour that we observe. Modes of behaviour There is a huge array of systems on Earth, however the dynamics they produce is made up of instances of a small number of fundamental patterns (Figure 2). These are exponential growth, goal seeking and oscillation. Each of these fundamental patterns are generated by a simple feedback loop. Other common modes of behaviour generated by interactions between exponential growth, goal-seeking and oscillation include S-shaped growth, S-shaped growth with overshoot and oscillation, and overshoot and collapse (Figure 2). Page 10

11 Figure 2. Common modes of behaviour in dynamic systems (source: Sterman, J.D. (2000). Business Dynamics. Systems thinking and modeling for a complex world. McGraw-Hill, Boston. Fig 4-1) Exponential growth Exponential growth arises from reinforcing loops. Classic examples are compound interest and the growth of populations. A key property of exponential growth is that the doubling time is constant. For example, the time taken for a population to double remains the same as the population grows exponentially. Reinforcing loops can also generate exponential decline. Goal seeking Goal seeking behaviour is caused by balancing loops. Balancing loops counteract any disturbances that move the state of the system away from the goal. If there is a discrepancy between the desired and actual state of the system, corrective action is initiated to close the discrepancy. Oscillation Oscillations are also caused by goal seeking behaviour, however, delays mean that the state of the system overshoots and undershoots the goal. The time delays cause corrective actions to continue after the state of the system has reached the goal, leading the system to adjust too much. Delays can occur in any part of the loop. Inferring structure from behaviour The link between structure and behaviour is important because it allows inference about the structure of a system from its behaviour over time. If a system exhibits exponential growth, then you know that there must be at least one dominant reinforcing loop in operation. Note that there may also be balancing loops operating as well but the reinforcing loop will be dominant. Similarly, if a system exhibits goal seeking or oscillation behaviour then you know that there must be at least one dominant balancing loop in operation Page 11

12 The behaviour of the system may change over time and switch from say exponential growth to goal seeking. This tells you that there is at least one reinforcing loop and one balancing loop, and their dominance changes over time. Once you have identified the current dominant feedback loops, it is necessary to identify other feedback loops that are currently latent or sub-dominant, because if there is a shift in loop dominance then these latent loops will affect the behaviour of your system. For example, take the chicken population model shown previously (Figure 1). If the Chicken population is growing exponentially we know that the population-births reinforcing loop is dominant (i.e. the left panel in Figure 1). However, we also know that this growth cannot go on forever and as we breach the system s carrying capacity, balancing loops will become more and more dominant to limit population growth this is represented in the Chicken model by the balancing loop in Figure 1. Things like limits to growth can help you to identify balancing loops that may be latent or subdominant now, but will become dominant in the future. Stocks and flows Introduction A key limitation of causal loop diagrams is that they do not discriminate between stocks and flows so they cannot be used to quantitatively model systems. Stock and flow models (SFMs) are what we use to build system dynamic models (SDMs) that help us to quantitatively model system behaviour over time. All variables within a CLD can be classified as either stocks or flows and SFMs can be used to model feedback loops represented within CLDs Stocks - are represented by rectangles and are accumulated quantities, such as population, money in a bank account, etc. Stocks describe the condition of a system and would continue to exist if all flows within a system stopped. Flows - are represented by pipes with valves and are actions that cause stocks to increase (inflow) or decrease (outflow) over time, such as births, deaths, interest earned on a bank account, etc. Flows fill or drain stocks, they do not accumulate quantities themselves. In SFM clouds are sources and sinks for flows and represent stocks that sit outside the boundary of the model and have infinite capacity. Note that the value or accumulation of these cloud stocks are not calculated or modelled. Figure 3 provides a summary of how stocks, flows and clouds can be combined to construct a simple SFM. (source: Sterman, J.D. (2000). Business Dynamics. Systems thinking and modeling for a complex world. McGraw-Hill, Boston. Fig 6.1.2) Page 12

13 Figure 3. Example of how stocks, flows and clouds are combined to make a stock and flow model There are two other important components of SFMs; converters and connectors: Converters - represent auxiliary variables that are contained within feedback loops between stocks and flows. They can store constants, equations and graphical functions. Connectors - used to link variables together to complete feedback loops. For example, they are used to show which variables contribute to a Flow. Page 13

14 Identifying stocks and flows Stocks not only represent amounts of material (such as people, money and things), they can also represent intangible variables such as morale or expectations. Importantly, stocks must obey the conservation of matter law i.e. stuff remains in the stock until it flows out. Just like causal loop diagrams, SFMs also contain feedback loops. That is, flows influence stocks and stocks influence flows via an information feedback link. Take the following simple population CLD (Figure 4), it contains one reinforcing loop and one balancing loop. Birth Rate Death Rate Births R Population B Deaths - Figure 4. Causal loop diagram of population Population is a stock (people), while births and deaths are flows (people/year). In this example the fractional birth rate (called Birth Rate with units 1/year) is a constant. So births add to the population, and the population and the birth rate control the births. This completes the births reinforcing loop (Figure 5). Births R Population Birth Rate Figure 5. Stock and flow representation of the births reinforcing loop of the population model Similarly, the flow of deaths (people/year) reduce the population (people), and the population and death rate (1/year) control the deaths (Figure 6). This completes the deaths balancing loop. Page 14

15 - Population Deaths B Death Rate Figure 6. Stock and flow representation of the deaths balancing loop of the population model The stock and flow models for births and deaths can be joined to represent both the births reinforcing loop and the deaths balancing loop (Figure 7). Birth Rate - Population Births Deaths R B Death Rate Figure 7. Stock and flow representation of the full population model containing the births reinforcing loop and the deaths balancing loop The best way to convert a CLD into a SFM is to: Identify the stocks: every feedback loop must contain at least one stock Identify the flows for each stock Identify the feedback links from stocks to flows, which may include auxiliary variables. These are neither stocks nor flows and are usually influenced by stocks and control flows. Auxiliary variables are often represented by converters (circles) Generic stock and flow structures The previous section included a description of fundamental modes of system behaviour such as exponential growth and goal seeking behaviour. These fundamental modes of behaviour can be produced by generic SFM structures that simulate reinforcing and balancing loops (Figure 8). Page 15

16 Figure 8. Generic stock and flow configurations for linear, exponential Growth, exponential decay and convergence behaviour (source: Fisher, D.M. (2011). Modeling Dynamic Systems. Lessons for a first course. Third Edition. Page 3-41, 3-42) Creating a stock and flow model example compound interest Compound interest means that the money in a bank account will grow exponentially over time if none is spent, because the interest earned adds to the amount of money in the account, which adds to the interest earned. Exponential growth is indicative of reinforcing loops. The CLD for compound interest is shown in Figure 9. The stock and flow model based on this CLD is shown in Figure 10. Page 16

17 Interest Rate (%/year) Money in Account ($) R Interest Earned ($/year) Figure 9. Causal loop diagram for compound interest Interest Earned ($/year) R Money in Account ($) Interest Rate (%/year) Figure 10. Stock and flow model for compound interest. Also shown is the behaviour over time graph of the stock (money in account) Page 17

18 SECTION 2 INTRODUCTION TO THE SYSTEMS SIMULATION MODEL Introduction The Systems Simulation Model (SSM) is a model of a coastal system that simulates interactions between activities on land (such as farming and urban development), activities on water (such as fishing), coastal ecosystems (such as coral reefs and mangroves) and coastal resources (such as fish). To simulate the coastal system, the SSM uses system dynamics, which is a modelling methodology that simulates the movement of material (such as fish) and information (such as fish price) through a system using stocks and flows. The model also simulates feedback loops that control the behaviour of the system over time, for example a reduction in fish supply, which leads to an increase in fish price, which leads to an increase in fish supply again. Model purpose The SSM is designed to simulate the behaviour of a coastal system for up to a 50-year time period at one-week time steps. It can be used to simulate the effect of changes in policies or interventions on the behaviour of the coastal system. For instance, it might be used to assess the impact of land use zoning policies on land use as well as on fish habitat and fish catch. It might also be used to evaluate the impact of fishing management policies, such as a cap on boat numbers, on fish catch and fish price. The SSM simulates the following components of a coastal system: Catchment runoff sediment, nitrogen, phosphorus Crop production rice, cashews, fruit, vegetables, oranges, bananas, coconut, mango livestock production cattle, goats, pigs, carabao, poultry Fish populations herbivores, predators, squid Illegal fishing poison, bomb Legal fishing traditional, pelagic, squid Tourism Coastal habitats reefs (marine protected area, non-marine protected area), mangroves, sea grass Accommodation hotels, domestic Land use crops, livestock, hotels, domestic Human population domestic, tourist Supply, demand and price fish, crops, livestock Income from tourism, fishing, crop production, livestock production Net exports export income and import costs from fishing, cropping, livestock production, tourism, ancillary Jobs from fishing, cropping, livestock production, tourism, ancillary Waste production septic tanks, stormwater, catchment runoff Water use domestic, crops, livestock Water quality algal blooms, suspended sediment, nutrients Model limitations This Systems Simulation Model, like any model, has limitations that should be understood before it is used. These include: Page 18

19 Is a site level model: The model is aggregated to a site, such as an island or a local government area. It does not model the flow of material and information within a site, such as the transfer of goods or money between households. Not an agent-based model: The model does not simulate the behaviour of individual entities such as people, households or boats. Not a spatially explicit model: The model does not distinguish between spatial units or the movement of entities or material between spatial units. It does not simulate the movement of people from one location to another within a site and it does not simulate the difference in the number of people at one location or another within a site, for example. Parametrised for two sites only: The data spreadsheets associated with the model contain parameters for two sites only the main island of Selayar in Indonesia and the LGU of El Nido in the Philippines. Using the model for new sites will require parameters for these sites to be added to the data spreadsheets. Parameterised using average values: Parameters used in the model are based on site averages e.g. the average number of people per household. The model does not specify uncertainty in model parameters although this can be added if known. Contains assumptions: Because data for all parts of a system are difficult to obtain, some of the model parameters are based on assumptions and not on measurement. The source of each parameter used in the model is listed within the data spreadsheets associated with the model. Does not predict absolute values: The model simulates behaviours and trends over time and should not be used to predict absolute values. The model will simulate increases or decreases in fish population in the future, for example, however it cannot provide precise prediction of how many fish there will be in the future. The model can only be run using Stella Architect (available from Software requirements to run the model Stella Architect (available from Computer (min 8GB RAM) Model architecture The model was developed using Stella Architect Software Version 1.5. The overall architect of the model is shown in Figure 11. Page 19

20 Food Production Costs <Demand for Susbsistence Food Production> Crop Production Livestock Production - - Household Income from Food Production - - Ability to Purchase Food Boats Food Prices - Fishing Effort Tourists Fish Catch - - Land Available Hotels Demand for Food from Market Demand for Susbsistence Food Production - Households Domestic Population - Fish Popuation Demand for Timber and Fuel Water Available Rainfall Stormwater Pelagic habitat Septic tanks - Mangrove Coral Reef Seagrass Runoff Water Temperature Contaminant <Crop Production> <Livestock Production> loading - Algae (phyto) Estuary Water Quality - - Figure 11. Architecture of the system dynamics model Page 20

21 SECTION 3 HOW TO USE THE SYSTEM DYNAMICS MODEL Opening Stella Architect The model uses the software platform Stella Architect (available for Mac or PC from When installed navigate to the Stella Architect icon and double click to open i.e. double click on the icon that appears on the desktop, or left-click on the Start button at the bottom left of the screen and Navigate to Stella Architect in the list of programs (Figure 12). The opening screen for Stella Architect appears as shown in Figure 13. Figure 12. Stella Architect icon as it might appears in the start menu Figure 13. Opening screen for Stella Architect Opening the Systems Simulation Model To load the Systems Simulation Model (SSM) through Stella Architect, click on File and then on Open (Figure 14). Then navigate to the location of the file named CCRES Systems Simulation Model. Double click on this file to open the model. Page 21

22 Figure 14. Accessing the file options in Stella Architect to load the System Dynamics Model An alternative approach is to navigate to the model file (CCRES Systems Simulation Model) in the directory, right click and select Open with then select Choose default program and select Stella Architect.exe (or Keep Using Stella Architect.exe if it is already the default program). If the default program is already set to beginning Stella Architect.exe, then you can shortcut this process by double clicking on the file name. Exploring the model The Systems Simulation Model should appear on the screen after a few seconds showing a technical looking arrangement of stocks, flows, auxiliary variables and text labels (Figure 15). The view of the model that can be seen is only a fraction of the total model. To zoom in and out of the map using the following short cut keys: Simultaneously press the Ctrl key and the key to zoom in. Simultaneously press the Ctrl key and the key to zoom out. Use the scroll bars located on the right-hand side and along the bottom to move the viewing window to different parts of the model (also shown in Figure 15). Page 22

23 Side scroll bar Bottom scroll bar Figure 15. The model view when the System Dynamics Model is first opened The SSM can also be explored using the Find function that is built into Stella Architect. Simultaneously pressing Ctrl and the F key opens a search window (Figure 16). This search window can also be opened by selecting Model from the tool bar located at the top of the screen and then selecting Find from the dropdown menu. Enter the name of a variable into the search window and a list of variables containing this word (or phrase if multiple words used) will appear in the pop-out window. Click on the variable name of interest and Stella Architect will navigate you to the location of this variable. For example, Figure 16 shows FISH as the entry to the Find function, which has resulted in a range of variables with this keyword in its title. Note that the search results are not case-sensitive and therefore will include variables with Fish and fish. These results can be filtered to show (or hide) stocks, flows and auxiliary variables. This can be a useful tool for quickly navigating to parts of the model if you know the stocks and flows that are involved. Page 23

24 Figure 16. Finding stocks, flows and auxiliary variables using the Find function in Stella Architect Overview of the modules There are 62 modules in the Systems Simulation Model and these are summarised in Table 1. Table 1. Modules of the Systems Simulation Model Module name Human population Human population calculations Effects on population migration rates Employment Site level savings Description A multi-structured and multi-arrayed sub-model that calculates the population dynamics. This module contains the most stocks of any of the modules and is categorised by: Sex (male, female) Age group (children, young, middle, old, retired) Employment (fishing, non-fishing) Fishery (traditional, squid, pelagic, poison, bomb) Provides summary calculations to be used to update the human population module. Coalesces the different determinants of human migration and selects the most influential determinant to update the human population migration rates. Calculates the number of jobs currently available based on fishing, farming, tourism and population. Calculates the site level savings based on income to the site and costs exiting the site. This is then converted to net present value. Page 24

25 Module name Tourism/Hotels Houses Cost of building and infrastructure Enter / Exit Fishing Move between fisheries Fish population Fishing effort Large boats Large boat motors Small boats Tourist boats Small boat motors Boats summary Small boat control Fish price Fish catch Cost of fishing Cost of fish imports Cost of tourist boats Income from tourism Limits to crop and livestock landuse Crop price Crop production Description Calculates the number of tourists visiting the system, how long they stay for and how many hotels there are to accommodate them. Calculates the number of houses there are to accommodate the domestic population. Calculates the imported cost of building houses and hotels, and associated infrastructure. Coalesces the different effects of people moving into, and out of, fishing to update normal entry/exit rates used in the human population module. Coalesces the different effects of people moving between fisheries to update normal entry/exit rates used in the Human Population Calculations module. Simulates the population of three fish types (herbivores, predators, squid) across two life stages (juvenile, adult) and four habitats (seagrass, mangroves, reef, pelagic). The reef habitat is further separated into MPA and non-mpa. Calculates the weekly boat level fishing effort. This is specified for each Fishery, whether the boat is motorised or not, whether the boat is owned by a poor household or not, and which habitat is targeted. Calculates the number of big fishing boats (pelagic, squid) that are being used. Is integrated with the large boats module to simulate the number of motors being used by large fishing boats. Calculates the number of traditional, poison and bomb fishing boats that are operating within the system. Calculates the number of tourist boats that are operating within the system. Is integrated with the small boats and tourist boats modules to simulate the number of motors being used. Coalesces the data for large fishing boats, small fishing boats and tourist boats. Manages the flow of boats between small fishing boats and tourism boats. Calculates the current price of fish sold locally at traditional market and fish exported to other markets. It includes price elasticity effects of supply and demand on price, and price effects on supply and demand. Calculates the volume (as mass) of fish that is sent to local and export markets. Calculates the imported cost of fishing for fuel use, boat and motor maintenance, boat and motor purchase. Calculates the cost of fish imported to the local market. Calculates the imported cost of tourist boats for fuel use, boat and motor maintenance, boat and motor purchase. Calculates the weekly income from tourism. Determines how land and water availability affects crop and livestock land use. Calculates the current price of the crops sold at the local market. Calculates the volume (as mass) of crops that is produced by the site. Page 25

26 Module name Crop supply Crop demand Crop land use change Cost of crop imports Crop water use Meat price Livestock production Livestock supply Meat demand Cost of meat imports Livestock water use Income from crop exports Income from crop local sales Income from crop sales Present value of crop income Income from meat exports Income from meat local sales Income from meat sales Present value of meat income Grazing landuse change Livestock landuse Interest rates Water Availability Surface and groundwater pollutant mass loadings Pollution loading Description Calculates the amount (by mass) of crops that is supplied to the local and external markets each week. Calculates the weekly demand for crops at the local market. Calculates the change in cropping land use. This includes conversion of vegetated (non-farmed) land to farm land, and the loss of farm land to degraded cropland. Calculates the cost of crops imported to the local market. Manages the availability and use of surface / groundwater to crop irrigation. Calculates the current price of the meat sold at the local market. Calculates the amount of livestock that is produced by the site. Calculates the amount (by mass) of livestock that is supplied to the local and external markets each week. Calculates the weekly demand for meat at the local market. Calculates the cost of meat imported to the local market. Manages the availability and use of surface / groundwater to livestock irrigation. Models the income that is generated from selling crops to external markets. Models the income that is generated from selling crops to local markets. Combines the income from local and external crop sales. Converts the income from selling crops to present day equivalent value. Models the income that is generated from selling meat to external markets. Models the income that is generated from selling meat to local markets. Combines the income from local and external meat sales. Converts the income from selling meat to present day equivalent value. Calculates the change in grazing land use for livestock. This includes the loss of grazing land to degraded grazing land. Coalesces the land use currently used for livestock. Calculates the compounded interest effect on prices. Calculates the current volume of surface water and the current depth of groundwater. Calculates the pollutant concentrations in the water and from this calculates the amount of pollutants that are discharged to the receiving waters. Coalesces the point and diffuse sources of pollutants into a single loading rate. Page 26

27 Module name Septic tank Receiving water contaminants Reef habitat Seagrass habitat Mangrove habitat Stormwater module Phytoplankton Light extinction Water temperature Description Calculates the wastewater treated through septic tanks. From this, the pollutant loading rate from the septic tank system is calculated. Calculates the concentration of pollutants in the receiving water. Manages the dynamics of reef condition and fish carrying capacity. The proportion of these reefs that are located within a marine protected area (MPA) can be specified. This module contains stocks for Reef Condition and Reef Area (both for MPA and non-mpa). Manages the dynamics of seagrass condition, area and associated fish carrying capacity. Manages the dynamics of mangrove habitat, including the potential harvesting of mangroves for fuel and/or timber. Mangrove abundance is modelled as a size-structured model (small and large). Manages the dynamics of stormwater generation and flow generated from urban area runoff and greywater production. Manages the dynamics of phytoplankton biomass production. It can be used as an indicator of potential eutrophication (algal blooms). Calculates the light extinction rate for the receiving water. Calculates the temperature of the receiving water at weekly time steps. Model datasets Introduction The Systems Simulation Model uses many different parameters that control how the model behaves. This model is a process-driven model rather than a data-driven mode, which means that it is not dependent on data for modelling dependencies between components of the model. However, it is a data-heavy model, requiring a substantial number of parameters so that the SSM operates in a manner that reflects the case study area. The following section provides instruction on the types of parameterisation data used in the model and how to identify them. As an example, Figure 17 shows the Tourism / Hotel module, which highlights the three types of parameterisation data required: - Auxiliary input variables - nodes that have a specified value i.e. not calculated from other variables - Auxiliary initial conditions nodes that are used to provide initial conditions for stocks - Graphical functions these are converter variables that relates an input value to an output value. These are typically dimensionless (no units) variables. Note that these are not input variables, rather they are relay variables. To view the data contained in one of these parameterisation nodes (or any node for that matter), simply double click on the variable of interest to open a dialogue window on the right hand side (Figure 18). Page 27

28 Graphical functions Initial conditions for stocks Auxilliary variable parameters Figure 17. Input data requirements for the Tourism / Hotels module Figure 18. Example of the properties for an input variable ( tourist length of stay ). The dialogue window indicates that the input variable has a value of 1.34 weeks. Linking the model to the data The Systems Simulation Model comes complete with data values for all parameters and the model will successfully run with these values. However, the default values in the SSM do not relate specifically to either of the case study sites (El Nido, Selayar) and therefore do not reflect an actual site. The SSM needs to be linked to a spreadsheet file so that it can be run with El Nido or Selayar input data. Figure 19 Figure 23 provides stepwise instruction in linking a model to the data. Note that once the model is linked to the spreadsheet, saved changes in data values in the spreadsheet will automatically be updated in the model. Page 28

29 Formatting is very important in the spreadsheet. This includes using the correct names for variables, including the correct number of values (e.g. if a variable is arrayed for two elements but the dataset shows three elements then there will be an error). The model will provide an error message if there is a parameter name in the spreadsheet that does not appear in the SSM. Note that the spreadsheet data will still be imported (except for the parameter(s) that do not match the parameters in the model). The steps involved: 1. Select the import data option from the menu in Stella Architect (Figure 19) 2. Set up an import link between Stella Architect and the spreadsheet (Figure 20) 3. Navigate to the folder where the data file is located. Click on this and click on the Open button located at the bottom right of this window (Figure 22) 4. The name of the data file should appear in the link at the top of the window. Select the correct Sheet orientation for the file. The preferred orientation for this model is Horizontal orientation, where parameter values are arranged in rows (one parameter per row). If the linked dataset has multiple pages of data (Figure 23) Figure 19. Click on Model from the toolbar and select Import Data from the drop-down menu Page 29

30 Figure 20. Import data window with no data sets linked Figure 21. Import data window with import link added through clicking on green icon. Then click on Browse button to open file explorer where the data file is located Page 30

31 Figure 22. Navigate to the folder where the data file is located and open Figure 23. Data file linked and ready to be imported The Systems Simulation Model interface Opening the user interface The Systems Simulation Model interface is where the operation of the model is controlled. To access the interface window, click on the icon that resembles a monitor with a graph on it (Figure 24). Page 31

32 Figure 24. Click on the interface window button to enter the interface The Interface should open at the title screen (Figure 25). The screen should be dominated by the title (Systems Simulation Model) with six buttons arranged vertically below. To the left you should see the pages of the interface, which shows how the contents of the interface are arranged. These can each be accessed by clicking on the relevant page and the main screen will refresh to show this. Figure 25. The model interface as it appears when opened. The vertically arranged panels on the left are the separate pages that make up the interface The initial view of the interface (Figure 25) is inactive and therefore pressing the buttons on the screen will not result in any model activity. To start the interface so that the model responds to button presses, click on the down arrow next to the Interface icon (see Figure 26) and select Present Fullscreen. The Interface should now appear as shown in Figure 27 and is now active. Page 32

33 Figure 26. Click on the down arrow next to the Interface icon and select Present Fullscreen Figure 27. The Interface operating in fullscreen mode with the menu buttons now active Page 33

34 Navigating the interface The top five buttons shown in the opening page of the interface (Figure 27) provide information about the model. For example, clicking on the About the Model button opens a dialogue window that provides detail on the model itself (Figure 28). To close a dialogue window, click on another information button, which opens a new dialogue window, click on the START! Button, which navigates away from this page, or click on an empty part of the interface. Figure 28. Clicking on any of the top five buttons opens a dialogue window with contextual information. This example shows the dialogue box when the About the Model button is pressed. Entering the simulation part of the interface The simulation part of the Interface is entered by pressing the green button labelled START! This will open a new page titled Main Menu (Figure 29). The purpose of this interface screen is to allow navigation to different parts of the model interface. Page 34

35 Figure 29. The Main Menu of the Interface Key indicators The first two buttons enable navigation to the Key Indicators component of the interface. The User can select Menu, which navigates to an interface where the type of Key Indicator can be selected (Figure 30). There are two types of indicators: Key Indicators these are a small (seven) selection of indicators that provide an aggregated assessment of different parts of the model output. Thematic Indicators these are a larger (12) selection of indicators that provide a more specific assessment of different parts of the model output. Pressing the Menu button on the Main Menu page navigates the User to an interface screen that contains output graphs for seven key indicators (Figure 31): Unprotected reef carrying capacity Protected reef (MPA) carrying capacity Tourism income per capita (at present value) Agriculture income per capita (present value) Traditional fishing income per capita (present value) Domestic population Dependence on Imported food The User can also directly navigate to the page where the seven key indicator graphs are presented (Figure 31) by selecting Graphs from the Main Menu page. Page 35

36 Each thematic indicator button shown in the Key Indicators menu (Figure 30) allows navigation to a specific interface page that contains multiple indicator graphs relevant to a specific theme of the model output e.g. Tourism (Figure 32). Figure 30. The indicator menu of the interface Figure 31. The seven key indicator graphs. Note that per capita income (tourism, agriculture, fishing) are at present value Page 36

37 Figure 32. Indicator graphs for tourism Slider menus There are two buttons on the Main Menu interface page (Figure 29) that allow the User to access the part of the Interface where changes can be made in select model parameters. The motivation for making changes in these parameters is to reflect potential management interventions in the system. The Systems Simulation Model can then be re-run and the trends plotted in the Key Indicator charts can be observed. The two buttons are: Marine sliders interface page that contains a suite of interactive sliders that enable changes to be made to the marine component. Terrestrial sliders interface page that contains a suite of interactive sliders that enable changes to be made to the terrestrial component. Marine sliders The interface layout for the marine sliders is shown in Figure 33. This layout contains 14 sliders grouped into six categories: Boat licence limit allows the number of traditional, pelagic, squid and tourist boats to be specified. Note that these sliders are each matched with a switch that disables/enables the slider setting. E.g. if the traditional boat limit is set to 100 but the associated switch is turned OFF then the boat licence limit is not applied in the SSM. Government grant this slider reflects a potential policy where boats are donated to fishers. This slider allows the number of boats to be specified and is matched with a switch that disables/enables this option. Note that the model applies this grant as an annual grant i.e. will Page 37

38 occur once every year. There is an additional slider that allows specification of how the boat grant is distributed between poor and non-poor households. Destructive fisheries management this is represented by a single slider that allows the effectiveness of surveillance to be specified. The model assumes that 100% surveillance effect results in 100% continued cessation of destructive fishing. Marine protected area this single slider allows the proportion of reef that is protected to be specified. Mooring damage this contains three sliders, one for each of small fishing boats, large fishing boats and tourist boats. These are multiplier sliders, which amplify (slider > 1) or attenuate (slider < 1) the damage rate caused by these boat types. Mangrove restoration there are three sliders that allow the mangrove restoration rate to be manipulated. Figure 33. Marine sliders interface layout Terrestrial sliders The interface layout for the marine sliders is shown in Figure 34. This layout contains 15 sliders grouped into five categories: Tourism there are four sliders that enable the rate that tourists arrive, the current number of tourists, rooms per hotel and land zoned for hotels to be specified. The first two sliders enable explicit adjustments to tourist numbers whilst the second two sliders enable the hotel development footprint to be manipulated. Housing there are two sliders for this category. The first is a zoning parameter where the land available for housing can be specified. The second (desired people per house) allows scenarios to be run based on increasing or decreasing household sizes. Page 38

39 Jobs the slider in this category explicitly addresses the dynamics of alternative livelihoods without specifying what these alternative livelihoods are. Waste management there are six sliders that allow manipulation of septic tank management and grey water production for houses and hotels. Agriculture there are two sliders that allow the zoning area for crops and livestock to be specified. Figure 34. Terrestrial sliders interface layout Case study The bottom button on the Main Menu (Figure 29) is labelled Case Study. This is included to provide an example of how the indicator graphs and sliders can be used to explore the efficacy of multiple management interventions. The theme of this case study is managing coral reef fish and fishing. Currently, many reefs are heavily exploited by fishing and there has been much effort in improving their sustainability. This case study provides an insight into how management interventions can address specific issues but can also result in unexpected issues that also require management. Pressing the Case Study button on the Main Menu page navigates the User to the relevant interface (Figure 35). On the left-hand side of this interface there are two panels, both labelled Control Panel. These outline the sequence of the management interventions used in the case study. The top panel contains only text indicating that it represents the business as usual scenario i.e. no changes are made in the parameters. The bottom panel contains sliders that are used in testing different management options. There are three sliders: Page 39

40 Traditional fishing boats (accompanied by a switch that disables / enables this slider) Surveillance effect Damage rate multiplier small boat These sliders are listed in the order that they are applied in the case study. That is, once the business as usual scenario has been tested, the first management intervention that is tested is to restrict the number of small fishing boats. Six charts populate the middle of the interface these are the indicators used to assess the efficacy of the management interventions: Reef carrying capacity (kg/hectare) this is an indicator of the health of the reef where people fish i.e. excludes marine protected areas Reef fish density (kg/hectare) this is an indicator of the number of fish that are currently living on the reef where people fish Traditional fishing boats this indicates the number of boats currently fishing. These boats tend to apply much of their fishing effort to reef habitats (they may also fish in seagrass and mangrove habitats). Bomb fishing boats the number of illegal fishing boats engaged in bomb fishing Poison fishing boats the number of illegal fishing boats engaged in poison fishing Weekly income per boat the weekly income for traditional fishers presented at the presentday value Figure 35. Case study interface screen Intervention 1 business as usual The first step in the case study example is to run a base case where none of the parameters adjusted. To ensure that this is the case, press the button labelled RESET SLIDERS (this makes sure Page 40

41 that all sliders are back at their default settings) and press the button labelled RESET GRAPHS (this clears all previous trends from the graphs). Press the RUN MODEL button and the model should display the outputs for the six indicators. The base case model output is shown in Figure 36 as a solid blue line. Intervention 2 restricting the number of small fishing boats The first management intervention to test is providing a cap on the number of small fishing boats. Note that this intervention only applies to the legal fishers as it is assumed that the illegal (bomb, poison) fishers will not adhere to this cap given that they are already engaged in illegal activity. Take note of the number of traditional boats in the base case. In our run, the number of boats were about Therefore, set the limit to below this it is recommended that the limit be 50% of the number of boats observed in the base case. We set it to 500 in our example. Now re-run the model. Figure 36 shows the results of this intervention super-imposed on the base case (intervention 2 is represented by dashed red line). The most obvious result is that the number of traditional fishing boats does not exceed 500. This means that there are less traditional fishing boats engaged in fishing, which decreases the pressure on reef habitat and reef fish populations right? Well actually, it is not that straightforward. Note how the number of bomb and poison fishers quickly increase at the start of the second run. The issue is that limiting traditional fishing resulted in increased fishers participating in destructive fishing. The impact on the reef habitat carrying capacity and fish density is that these trajectories are actually worse (steeper decline) than the base case. Note also that weekly income from fishing has changed. At the start of this second run, it looks like there is a benefit in the form of income per boat by reducing the number of traditional fishers. However, this quickly diminishes as the destructive fishing increases. Note that in this version of the model, fish supply to the local market is supplemented by imported fish. Therefore, even if the fish become scarce to local fishers, the price at the market remains reasonably stable as the number of imports increases. Thus an intervention with the intention of reducing fishing pressure has actually exacerbated the problem by increasing destructive fishing damage. Page 41

42 Figure 36. Case study comparing base case and traditional boat limit (n = 500) Intervention 3 increasing surveillance effect The next intervention is to address the destructive fishers. This is done by increasing the surveillance effect from its default value (in our case it is 0, which means that there is no surveillance) to a higher value (in our case it is increased to 50% - this approximates a 50% reduction in destructive fishing due to surveillance). Now re-run the model (Figure 37). Note how the number of bomb and poison fishing boats are now generally lower than they were for the first two runs. There is a period at the start of the model where the number of destructive fishing boats is similar in number as for the base case, but this is due to the model settings showing that at the start of the simulation there is zero bomb and poison fishing boats. Therefore the influence of surveillance is not apparent until the number of bomb and poison fishing boats increases somewhat. Also note how the reef carrying capacity and fish density are improved. The income per boat has also improved because there are more fish being caught per traditional fishing boat, and therefore less dependence on the imported fish. However, there is still unwanted impact of the destructive fishers on reef carrying capacity and fish density, but now the trajectories in these two indicators is starting to change away from decline towards sustained. We could increase the surveillance to 100% and re-run the model. Instead we move onto the final intervention. Page 42

43 Figure 37. Case study after surveillance effect increased from 0 (no surveillance) to 50% (surveillance has 50% effect in decreasing destructive fishing) Intervention 4 Decreasing damage caused by boat mooring The final management intervention is to decrease the effect of mooring damage on the reef habitat. This intervention could represent a technological improvement and/or a change in mooring practices. The slider used in this intervention is a multiplier parameter, which we used to increase (slider setting > 1) or decrease (slider setting < 1) the effect. In this example, we set the multiplier at 0.5, which represents a lowering of the mooring damage rate by 50%. Re-run the model. The case study output is shown in Figure 38. The key observation is that this intervention has not had a noticeable effect on any of the indicators. There are two potential reasons for this: mooring damage is not a major determinant of reef damage because the number of small fishing boats is limited reef damage by destructive fishing is much larger than mooring damage and therefore the former masks the latter. Setting the surveillance effect to 100% would remove the destructive fishing damage, allowing the mooring damage to be assessed. Finally, this example highlighted how multiple interventions can be used to address a suite of problems. However, this is only an insight into this process. For example, we have only looked at a single habitat (reef). What might be the consequences on other habitats of these management interventions? Also, restricting the number of traditional (boat cap) and destructive (surveillance) boats provided an improved reef habitat and fish density but reduced the livelihood options for the local population and presumably increased the dependence on imported fish. This suggests a system that is now more vulnerable to vagrancies of the imported fish supply chain. Page 43

44 Figure 38. Boat damage to reef decreased Page 44

45 Adjusting the run specifications of the model There will be times when the simulation period of the model needs to be changed. The default period is 2000 weeks (approximately 40 years). Changing the simulation period is managed through the model window. If you are in the Interface, then you can return to the model by clicking on the model icon (Figure 39). Then from the top menu select Model, and then Run Specs. Figure 39. To navigate from the Interface window to the Model window, click on the model icon The run specifications can be adjusted in the Model window by clicking Model and selecting run specs (Figure 40). To change the simulation period of the model, change the Start Time or Stop Time. The performance of the model can be adjusted by changing the DT (time step) from its default value of 1/3. This might be considered if the model is having issues with stability that might be fixed by decreasing this value (e.g. to ¼). Figure 40. The run specifications can be adjusted in the Model window by clicking Model then run specs Page 45

46 Setting up an interface page To build a new interface page, return to the Interface window (Figure 41). Use the scroll bar that shows the interface pages to highlight where you would like to add a new page. Click on the existing page that sits directly above the location that you would like to add a new page. Then click on the Add Page at the bottom of the screen. This process is shown in Figure 41 and the result shown in Figure 42. Figure 41. To add a new interface, use the left scroll area to select where you would like to place the new interface page (e.g. as indicated by the top dashed square) and then click Add Page Page 46

47 Figure 42. The new interface page as it appears Adding a slider to the interface To add a slider to this page, click on the slider icon at the top of the screen (Figure 43 icon highlighted on the right). This will generate a slider icon in place of the cursor, which can be dragged using the mouse. Click on the interface page where you want to place the slider and left click (Figure 44). This is an unlinked slider and therefore will not affect the model. To link this slider to the model, double click on the slider to open a dialogue window on the right-hand side (also shown in Figure 44). Figure 43. The icons used to add a chart (left) and a slider (right) To link the slider to a variable, click the button to the right of Variable in the opened dialogue window. This will open a window where properties can be specified to this slider (Figure 44). Clicking Page 47

48 on the button next to the Variable box opens another window where a variable from the model can be selected (Figure 45). Note that not only dangle variables can be selected i.e. stocks and variables that are dynamically calculated through an equation can also be linked to a slider. For stocks, the slider will adjust the initial conditions (i.e. the value that the stock has at the start of the simulation period). For variables that have an underlying equation, connecting to a slider overrides this equation, which can interrupt feedbacks in the model. It is recommended that equation-based variables are not connected to sliders. Figure 44. A new slider has been added to the interface page. Double clicking on this opens a window where properties can be assigned You can scroll through this list of variables to select the variable to link to this slider. However, this is a large list and this process can be slow. An alternative approach is to search for a variable by entering part of the name. Figure 45 shows fish has been entered that results in a filtered list of variables that contains the word fish. If the intention is to link the stock FISH PRICE to this slider then click on FISH PRICE from the list. Note that this stock is arrayed (as are many of the variables in the model) and therefore a second window opens requiring the user to specify which element of the array is to be linked. In this example, FISH PRICE is arrayed by fish type (herbivores, predators and squid) (Figure 46). Page 48

49 Figure 45. Variable window showing a list of variables that can be linked to this slider Figure 46. Window showing the elements arrayed under the selected variable Page 49

50 Adding a graph to the interface To add a graph to an interface page, click on the graph icon at the top of the screen (Figure 43 icon highlighted on the left). This will generate a graph icon in place of the cursor, which can be dragged using the mouse. Click on the interface page where you want to place the graph and left click (Figure 47). This is an unlinked graph and therefore will not show any output plots. To link this graph to the model, double click on the graph to open a dialogue window on the right-hand side (Figure 48). The process of linking a graph is similar to linking a slider. However, multiple variables (series) can be linked by clicking the button and selecting a variable. However, it is often beneficial to only plot a single variable but show it for different runs. This is enabled by clicking the comparative check box (see red dashed circle in Figure 48). Figure 47. Adding a new chart to the interface page Page 50

51 Figure 48. The graph window where its properties can be specified. Check the Comparative check box to allow the graph to show multiple runs Page 51

52 Setting up a new site Introduction The Systems Simulation Model comes set up for two case study sites: El Nido, Palawan, Philippines Selayar, Indonesia To set up the SSM for a new site requires that a new dataset that contains parameter values specific to the new site be established. As stated earlier, the SSM uses many different parameters that control how the model behaves. This model is a process-driven model rather than a data-driven model, which means that it is not dependent on data for modelling dependencies between components of the model. However, it is a data-heavy model, requiring a substantial number of parameters so that the model operates in a manner that reflects the case study area. The following section provides instruction on where data needed for parameterising the model might be found. It also highlights the type of parameters that would ideally need site-specific data versus the type of parameters where data from similar case study sites (e.g. a similar landscape or a neighbouring site) and/or from larger scale areas (e.g. data for provincial, national locations) are likely to be sufficient. Data sources This model requires significant number of parameters to work. The model comes with default parameters so that simulations can be run. The best data is data collected from the case study site that is being modelled. However, it is unlikely that data for all parameters will be able to be obtained from the case-study site. Local government agencies might be able to provide data for populating some or many of the socio-economic parameters. For example, they might have data on food imports and exports, population data and the current number of fishing boats. Data for the model might also be obtained through visiting a local food market, where the sale price of meat, fish and crops can be elicited. Any local ecological and/or socio-economic studies might also provide data that is suitable for use in the parameterisation of this model. Reports associated with these studies might be available at local government buildings, public libraries, NGOs and universities. Site-specific versus generalised data Many of these parameters will be specific to the study area being simulated by the model. For example, parameters representing the initial conditions like habitat size (mangroves, reef, seagrass), human population, land use area, number of buildings, commodity prices (fish, meat, crops) will be site specific. There will also be parameters that are more general and therefore can be sourced from similar case study sites and/or from regional / national studies. For example, parameter values for human birth rates and death rates might be expected to be reasonably similar at regional and national levels. interest rates birth rates death rates Page 52

53 Setting up a data spreadsheet In an earlier section, this manual provided information about linking to the existing spreadsheet that contains the input data for El Nido and Selayar. It is recommended that this spreadsheet be used as a template when creating a spreadsheet for a new site. Simply open the existing spreadsheet, copy the El Nido or Selayar worksheet and change the name to the new site. This sheet provides all the data that is needed for the SSM to be parameterised. Page 53

54 SECTION 4 TESTING THE SYSTEM DYNAMICS MODEL This section provides a brief overview of the model tests that are typically employed in developing system dynamics models. We have endeavoured to employ them in the development of the Systems Simulation Model. Tests such as mass balance testing and extreme condition testing, which are outlined in this section have been applied through the development of the SSM. However, we have been mostly unable to compare our model output against historical data because of the dearth of these trend data sets to carry out this exercise. Testing for mass balance Introduction The objective of mass balance testing is to check for the conservation of mass in the model. If the model is running correctly, the net difference between the inflows and outflows for each stock should be reflected in the change in the stock value. Setting up a mass balance experiment For example, in a simple population model (Figure 49) the Population is controlled by births (people are added to the population) and deaths (people are lost from the population). The other feature of this population model is the Carrying Capacity, which sets the maximum value for the Population. If the number of births is greater than the number of deaths then we expect the Population to increase in relationship to this (Figure 50 the first part of the graph where the line has a steep gradient). However, as the Population gets nearer the Carrying Capacity, the number of births will decrease and the number of deaths will increase (Figure 50 the end of the graph where the line is flat). Figure 49. Simple population model with births, deaths and carrying capacity Page 54

55 Figure 50. The population model showing growth in the population (births > deaths) until the population reaches carrying capacity (births = deaths) The model looks correct at a glance but how do we test the mass balance of the Population stock (or any stock) formally? The solution is to apply the following formula: STOCK initial stock value (stock inputs) (stock outputs) [equation 1] Where initial stock value is the value of the stock at the start of the simulation period (e.g. the initial Population shown in the simulation is 100). The Σ(stock inputs) and Σ(stock outputs) components of this equation are accumulations of the flows entering and exiting the stock respectively. If the model is correctly specified then the answer to this equation should be zero (0), which indicates that mass is not being created or destroyed i.e. there is conservation of mass and therefore it is mass balanced. For the Population stock in the population model the equation becomes: Population initial population births deaths [equation 2] To capture the Σ(births) and Σ(deaths) accumulations, we use additional stocks. In Figure 50, these are inflow stock (captures and stores the number of births over the simulation period) and the outflow stock (captures and stores the number of deaths). To model the mass balance in Stella Architect, we can introduce an additional auxiliary variable that uses equation 2 (Figure 51). Note how the three stocks are connected to the new variable mass balance. The remaining parameter that is needed for equation 2 is the initial population, and this is taken directly from the Population stock using the command INIT(Population). Therefore, we have all the pieces needed to calculate the mass balance. Page 55

56 Figure 51. Configuring the Population model to calculate the mass balance. The stocks appear with dashed lines, indicating that they are ghosted variables To enter the equation, use the mouse to double click on the mass balance variable, which opens the equation editor. The equation used is: Population-INIT(Population)-inflow_stockoutflow_stock Running the model produces the output graph shown in Figure 52. This shows a value of zero throughout the simulation, which indicates that the model is mass balanced. Figure 52. Output plot from the mass balance test showing the change in mass. If the model is correctly set up, then the output should be zero (as shown). Ghosted variables You might have noticed that the stocks appear multiple times, sometimes with a dashed outline (e.g. the stocks in Figure 51). These are ghosted variables and allows a variable to be shown multiple times however, note that there is only a single version of a variable and therefore changing a ghosted variable will change the original variable. Comparison with historical data Comparing a model output or outputs is an important part of highlighting the accuracy of the model as a simulation tool. The simple population model is used once again for this example. Figure 53 shows the model output plotted against historical population data. Even a casual observation indicates that there is not a very good comparison between the model output and the historical data. However, such comparison does allow the model to be calibrated, which is the process of adjusting one or more of the model parameters so that there is a better match between model and data. Often the parameter that is selected is the parameter that is most uncertain i.e. we are unsure Page 56

57 of its true value, so we use calibration to provide an estimate. In this example, we have simply adjusted the parameter normal birth rates to a lower value. The result of this is shown in Figure 54 note, however, that whilst this is an improved fit between model and data, it is clear that there is increasing under-estimation of the model towards the end of the simulation. Figure 53. Comparison of a simulation model output with historical data Figure 54. Adjustment of normal birth rate improves model fit to data although there is increasing under-estimation towards the end of the model run Extreme conditions testing (test / simulation model) The final model test that is covered is extreme condition testing. That is, if the model is subjected to an extreme condition, does it respond in a way that makes sense? For example, again revisiting the simple population model, the population is strongly dependent on a variable called the Carrying Capacity. This is the number of people that an environment can support if the population exceeds Page 57