START Advanced Institute
August 2003
Systems Thinking/Modeling Sessions
Craig Forster, Eric Scharff and SethMcGinnis
This document is HANDOUT.DOC on the systems session CD.
OVERVIEW OF SESSIONS
UNIT 1 – Introduction to Systems Thinking
1.1Background
- descriptors
- system behaviors
- hypothesis testing
- problem definition
- storytelling
- causal loop (influence) diagrams
1.2Explore pre-constructed, simple, urban CO2 emissions model “CO2 Learn”
- develop story and reference behaviors
- identify problem and hypotheses to be tested
- causal loop (influence) diagram
- introduce STELLA® user interface level
- introduce input-output features
- slidersand graphical input
- buttons
-graphmanipulation
- experience hypothesis testing
- change securityand review model at map level
1.3Evaluate strengths and weaknesses of model
UNIT 2 – Introduction to STELLA® for Building System Dynamics Models
2.1 Introduction
- expand story for urban system with goal to reducing urban CO2 emissions
- expand causal loop diagram (population with migration, sequestration, economic output, pollution,CO2emissions from energy use)
2.2 Population Model
- review Population Modelstory and outline reference behaviors
- build Population Model: a simple, two-stock population model with migration
driven by differentials in economic output
- explorePopulation Model
2.3 Economic Model
- review EconomyModel story and outline reference behaviors
- build Economy Model: a simple one-stock economic output model driven by urban population growth. Growth in economic output is assumed to cause growth in generic pollution flows that are reduced by pollution abatement strategies and technologies.
- explore Economic Model
2.4 Energy-to-CO2 Model
- review Energy-to-CO2 Model story and outline reference behaviors
- build Energy-to-CO2 Model: a simple, no-stock CO2 emissions model driven by urbaneconomic output
- explore Energy-to-CO2 Model
2.5 CO2 Sequestration Model
- review CO2 Sequestration Model story and outline reference behaviors
- build CO2 Sequestration Model: a simple, one-stockCO2 sequestration model driven by energy use
- explore CO2 Sequestration Model
UNIT 3 – STELLA® for Building System Dynamics Models - Continued
3.1 reviewstory for full integrated system
3.2review the integration of the 4 models from UNIT2 as 4 interacting model sectors
3.3 explore final model
3.4 evaluate strengths and weakness of model
WEDNESDAY NIGHT HOMEWORK (Highly Recommended)
1. Explore models from Wednesday afternoon workshop
2. Review next stepsin model evolution to be carried out on Thursday
3. Review causal loop diagram for next step in model evolution
4. Start mapping model structures required for next step in model evolution
ASSIGNMENT FOR REMAINDER OF INSTITUTE (Highly Recommended)
1. Develop a story that approximates a system of interest to you
2. Define problem to be addressed through model exploration
3. Develop causal loop diagram for your system
4. Reassess story and problem
5. Develop a system model starting with simple sectors and expanding as appropriate
6. Tune your system model with data from your system
Instructor/Tutor Availability During START Institute
Craig ForsterAugust 3 to 9 and August 16 to 22: Hotel Room 921
Eric ScharffAugust 4 to 22: NCAR Room 2011 [
Seth McGinnisAugust 4 to 22:NCAR Room 2011 [
Instructor Availability After START Institute
Craig (801) 581-3864
Resources
Introduction to Systems Thinking by High Performance Systems (HPS)
- the 1st few chapters are located in your binder
- the complete PDF file and related STELLA® files are on your STELLA® CD from HPS
START Institute, Systems Thinking/Modeling CD with this document and accompanying STELLA® model files.
The system dynamics project at MIT maintains a Web server devoted to issues in system dynamics. This includes the Roadmap introduction to system dynamics that provides a step-by-step, self-taught course in systems thinking/modeling.
High Performance Systems web page has a number of resources
“Modeling the Environment: Introduction to System Dynamics Modeling of Environmental Systems.” Andrew Ford. Island Press, 1999. Fundamental concepts and citations to important foundation literature.
“Understanding Urban Ecosystems”,Berkowitz, A.R., Nilon, C.H., and K.S. Hollweg, Springer-Verlag, pp. 328-342, 1999.
Acknowledgements & Software License
High Performance Systems generously provided the STELLA® software to the START Institute at reduced pricing. It is important to note that your license is a ‘Student Version’. Although there is no difference between this version and the regular software, you are not entitled to support from High Performance Systems – except as communicated through your instructor. The details of this arrangement are outlined on the HPS web page,
UNIT 1 Background Information
1. Systems Descriptors (from Smith, 2003)
NETWORKS – Every component (part or relationship) is interconnected to every other component in some way. These interconnections can create interdependence and contribute to diversity and complexity. Example – road systems.
BOUNDARIES – The real or abstract separation between a system and its environment or between different levels of scale within systems – systems occur nested within other systems, each linked to the others across scales. Example – city borders
CYCLES – Reoccurring events within or between systems. Cycles allow for the flow of materials and the revitalization and repair of individuals and systems. Examples – seasonal change, economic cycles
FEEDBACK LOOPS – They are chains of events in which an “output” of the event influences the first link at the beginning of the chain, either slowing down or speeding up initiation of the next event in the chain. Examples – urban sprawl, democratic elections
FLOW – The flowing of such things as energy, matter, and information through systems of all sizes of scale. The flow creates an effect. The rates, amounts, and importance of flow-through varies greatly. Examples – water, energy, money, information, power
DEVELOPMENT – Processes at all levels of scale that create growth and generate new forms. Examples – fund-raising, organizational/community infrastructure
DYNAMIC BALANCE – Changes that are continually occurring around an unfixed central point or temporary state of “well-being”. These fluctuations may swing very
widely or in small increments around this point. The focus of this phrase is on dynamic, rather than balance. Examples – population response to stresses (lack of food or water), financial market fluctuations
Smith, G.C., 1999. Systems Thinking and Urban Ecosystem Education, in “Understanding Urban Ecosystems”, (eds. Berkowitz, A.R., Nilon, C.H., and K.S. Hollweg), Springer-Verlag, pp. 328-342.
2. System Behaviors
A principle use of system dynamics models is to compute future ‘behaviors’ of systems of interest. In many cases you might want to compare the consequences of different policy alternatives by comparing the patterns of dynamic change computed over the simulation period. Ultimately, the absolute value of the parameters graphed cannot match reality. Differences observed between the computed patterns, however, can be instructive. Simple behavioral types are shown in the following figures. Suggest ways that systems you know about might lead to the graphed behaviors.
In developing systems models it is important to identify a “Reference Behavior” that is typically your view of the “Business as Usual” case. Once a model is developed you will want to compute the Reference Behavior then change the model settings (representing policy implementations, technology changes, etc.) to map a suite of alternative futures.
3. Hypothesis Testing
Mapping the consequences of alternative futures is one form of hypothesis testing where you attempt to anticipate the outcome of a specific change in model settings. If the results don’t match your hypothesis then one or more of the following may have occurred:
- you have not conceptualized the system correctly when building the model
- although conceptualize correctly, perhaps there is an error in the model code
- you may have discovered an unanticipated consequence
4. Problem Definition
Building a useful model requires a clear statement of the problem to be solved. Building a model of a system without a clear question to be addressed generally leads to unsatisfactoryresults because the model is not structured for optimal exploration of a specific question.
5. Storytelling
Using icon-based system dynamics model-building software enables users unfamiliar with the underlying mathematics and numerical methods to implement the ‘story’ of their system in a form that can be shared with others. The software enables one to illustrate a complex mental model that can be quantified and explored through hypothesis testing. Comparing different mental models and stories almost always leads to improved understanding of the system of interest.
6. Causal Loop Diagrams
A useful step in model building is to construct causal loop (influence) diagrams that map out the key elements of the system story without building model structures. Examples of causal loop diagrams will be discussed in the session.
UNIT 1 - “CO2 Learn” Model Explanation
Problem
An isolated metropolitan area is emitting CO2 at a rate that can be expressed as a per capita emissions rate. The per capita emissions rate has been increasing since 1950. If allowed to continue on the current non-linear trajectory, this growth in per capita emissions rate will lead to a substantial increase in CO2emissions from the metro area. What reductions in per capita emissions rate and/or population growth must be accomplished beginning in 2000 if CO2 emissions are to be maintained at 2000 rates? A simple STELLA® model (CO2_learn.STM on your CD) has been constructed to aid in your strategy explorations.
Facts
City Population
- 1950 = 0.5 million
- grows only through natural growth – migration is assumed negligible
- grows at a fixed growth rate from 1950 to 2000
- future growth rates can be adjusted to fixed values for specified time increments
Per Capita CO2Emissions
- documented from 1950 to 2000 as shown in graphical input
- accounts for all city-caused CO2 emissions internal, or external, to the city
- emissions after 2000 are extrapolated assuming “worst case” conditions
Issues to Consider
1. What are the reference behaviors of per capita CO2 emissions, total CO2 emissions and population?
2. What CO2 emissions trajectory is needed to solve the problem?
3. What impact does slowing population growth have on solving the problem?
3. What causal loop diagram represents the system?
4. What assumptions are made to simplify the system?
5. How might the model be made more realistic?
Model Explorations
A series of model explorations are required to assess the possible consequences of implementing alternative population growth and per capita CO2 emissions strategies. Consider the case where no regulatory limits are placed on CO2 emissions and the goal is to maintain emissions at the 2000 level.
Situation 1: no regulatory limit [USE CO2_learn.STM file]
1. The first strategy to consider is to reduce future per capita CO2 emissions rates for the 50 year period 2000 to 2050 by adjusting the graphical input without changing the natural population growth. Document the different strategies that you tested and report the strategy that achieves the desired goal. Does this seem feasible?
2. The second strategy to consider is to reduce future natural growth rate in order to reduce future populations. Start by assuming that you will maintain per capita CO2 emissions at 2000 levels. Are the necessary reductions in population growth rate feasible? Which strategies are most likely to achieve the desired result (reduced per capita emissions or reduced population growth)?
3. The third strategy is to find a balance between CO2 reductions through a pattern of reduced per capita CO2 emissions and reduced population growth.
UNIT 1 Learning Targets
1. Develop abilities in telling stories about systems.
2. Learn how to create the causal loop diagrams that underlie system stories.
3. Learn how to use STELLA® to explore strategy options.
4. Become familiar with STELLA® input, output and control features.
5. Become familiar with STELLA® programming icons (stocks, flows, converters, information connectors) and model-building strategies
6. Enhanced insight regarding possible strategies to pursue for reducing urban CO2 emissions
UNIT 2–Urban System Stories and Models
Population Model
Consider an isolated metropolitan area with surrounding rural region. As the urban center grows, the urban economic output increases and people migrate from the surrounding region. Thus, there are two population stocks to consider – the urban population and the regional population. We want to account for the way that urban birth and death rates might vary(due to improving or declining health) in addition to accounting for in-migration from the surrounding region. In the surrounding region, however, we consider net natural growth in population and out-migration to the nearby metro area. Migration from the regional to the urban community is assumed to be driven largely by differences in economic output between the regional and urban communities. Although this population model is destined to become one sector in a final urban system model, there are insights to be gained by exploring the dynamics between population and economic output in urban vs regional settings. For example, what level of economic development might be required in the surrounding region to reduce migration to a minimum? Should other factors that drive migration be considered in the model (e.g., social networking between migrants living in the metro area and others who are considering the move)?
Population Model: Step-by-Step
1. isolate relevant part of causal loop diagram
2. introduce model building levels
3. introduce map level icons and manipulation
4.open new STELLA® worksheet and build regional population exponential
growth model (see figure below)
Regional Population (stock)
- initial population = 2 million (expressed in millions)
Regional natural growth rate default = 3.0 % (slider 0% to 4%)
Time Period = 50 years
STELLA® Code
REGION_POPULATION(t) = REGION_POPULATION(t - dt) + (Region_Natural_Growth) * dt
INIT REGION_POPULATION = 2
INFLOWS:
Region_Natural_Growth = REGION_POPULATION*Region_Natural_Growth_Rate_%*0.01
Region_Natural_Growth_Rate_% = 3
5. add user interface slider to control growth rate and graph results
6. save model and remember file name [REGION POP.STM on CD]
7. experiment briefly, but don’t close model
8.using previous STELLA®fileadd an urban population model with
births/deaths (see figure below)
Urban Population (stock)
- initial population = 2 million (expressed in millions)
Urban Birth Rate default = 4.4% (slider 0 to 6%)
Urban Death Rate default = 2.0 % (slider 0 to 4%)
9. modify user interface with new sliders and run/restore buttons
10. save model with NEW filename [URBAN POP.STM on CD]
11. experiment briefly, but don’t close model
12.using previous STELLA®file add migration caused by movement from
regional population to urban
Per Capita Output (see figure below)
- urban default = 15 monetary units per person (slider 0 to 40 MU)
- regional default = 10 monetary units per person (slider 0 to 40)
Migration from region to urban is tied to
Econ Output Ratio = (Urban Econ Output - Regional Econ Output)/Urban Econ Output
Output Ratio / Migration Change (%)0.0 / 0.300
0.1 / 0.450
0.2 / 0.650
0.3 / 0.900
0.4 / 1.275
0.5 / 1.675
0.6 / 2.100
0.7 / 2.500
0.8 / 2.725
0.9 / 2.875
1.0 / 3.025
13. modify user interface with new graphical input and sliders
14. save model with NEW file name [MIGRATION POP.STM on CD]
15. experiment briefly and close model
Economy Model
Population, energy use to CO2 emissions, and CO2 sequestration must be linked to an urban economic output model if we are to explore CO2 emissions reduction strategies for the fully integrated urban system. Urban Economic Output responds directly to changes in the National Economy because the two economies are assumed to be closely linked. Thus, any change in National Economy is transferred directly to a change in Urban Economic Output. Urban Economic Ouput is reduced by the expenditures required to abate pollution/waste flows. As Urban EconomicOutput increases, pollution flows are assumed to increase. Expenditures used to abate pollution flows cause reduced pollution flows and reduced economic output which is, in turn, represented as a computed Net Urban Economic Output.
Economy Model: Step-by-Step
1. isolate relevant part of causal loop diagram
2. open new model file and build economic output growth model
(see figure below)
Time Period = 50 years
Urban Economic Output (stock)
- initial output = 30 million MUs (expressed in millions)
Growth in Urban Economic Output is a ‘biflow’ tied to National Economy
that is, in turn, specified graphically as a function of time.
Time(years) / National Economy
(millions of MUs)
0 / 4250
5 / 4700
10 / 5100
15 / 5150
20 / 4450
25 / 3800
30 / 3200
35 / 2650
40 / 2550
45 / 3050
50 / 4100
Change in National Economy is computed using the ‘builtin’ derivative function DERIVN(National Economy,1)/National Economy where ‘1’ specifies the 1st derivative (d[National Econ]/dt).
3. add graphical input for National Economy to user interface
4. save model and remember filename [URBAN ECONOMY 1.STM on CD]
5. experiment briefly (NOTE: a version with a delayed response to changes in the National Economy is found in URBAN ECONOMY 1 delay.STM)
6. using previous STELLA® Economy file, add pollution abatement component
Pollution Flow depends on the pollution technology funds expended (expressed as a % of Urban Economic Output)
Pollution Technology Expenditures as % Urban Economic Output / Pollution Flow(Pollution Units)
0 / 100
10 / 79
20 / 66
30 / 57
40 / 49
50 / 43
60 / 37
70 / 34
80 / 31
90 / 30
100 / 30
Pollution Technology Expenditures as % Urban Economic Output
- default = 0 (slider 0 to 100 %)
7. modify user interface with new slider
8. save model with NEW filename [URBAN ECONOMY 2.STM on CD]
9. experiment briefly and close model
Energy Use to CO2 Emissions
Gaseous CO2 emissions to be accounted for include those derived from electricity production, internal combustion engines, heating systems, open burning of waste, decomposition of vegetation and combustion in industrial processes. The aggregate need for energy across all sectors of the metro area is driven by the Urban Economic Output. Energy use increases as output increases. Energy conservation strategies help to reduce energy required for each unit of economic output. Energy delivered as electricity is distinguished from energy delivered as the direct consequence of combustion processes. This enables changes in the portion of electricity that is produced by technologies that don’t involve burning fossil fuels (e.g., hydropower, renewable, solar, etc.) to be accounted for. This enables us to track the consequences of choices made in the generation of electrical energy. Can energy conservation or shifts in sources of electrical energy lead to significant cost savings while also reducing CO2 emissions?
Energy Use to CO2 Model: Step-by-Step
1. isolate relevant part of causal loop diagram
2. open new model file and build economic output growth model
(see figure below)
No Stocks
Time Period = 50 years
Energy Conservation %: default = 0 (slider 0 to 100 %)
Ratio of Electrical vs Combustion (derived) Energy %: default = 80%
(slider 0 to 100%)
Percent of Electricity from Combustion %: default = 75%
(slider 0 to 100%)
Energy per output= 1.0 Energy Unit per million MU