James D. Clark
Robert M. Itami
H. Randy Gimblett
George L. Ball
The University of Arizona
School of Renewable Natural Resources
Tucson, Arizona 85721
This paper will present a conceptual framework for modeling spatial dynamic systems that employs discrete event theory and cellular automata. Discrete event theory is applied to control the model as a set of hierarchical events through a conceptual time space. Cellular automata theory is applied as a method for interpreting the spatial variability of the system. This modeling and simulating framework forms the analytical framework in the ongoing development of the Integrated Dynamic Ecological Analysis System (IDEAS) at the University of Arizona. The IDEAS project focusses on improving the spatial analytic functions of Geographic Information Systems. This paper will discuss the conceptual framework for modeling and simulating spatial dynamics. Research under IDEAS focusses on simulating alternative environmental management scenarios. This approach aids in understanding the temporal implications of environmental decision making. Examples of potential applications of IDEAS and spatial dynamic system modeling will be provided.
The Integrated Dynamic Ecological Analysis System (IDEAS) is a research program currently under way at the University of Arizona. The purpose of IDEAS is to provide research scientists, environmental planners, resource specialists, and managers with a common framework for decision making, analysis and modeling, and data management. It will provide these three functions through integration along three dimensions:
1) Analysis and Modeling Integration - (spatial and non-spatial, static and dynamic, deterministic and stochastic);
2) Analysis Tools Integration (geographic information systems, data base management systems, numerical analysis systems, textual and graphic systems); and
3) data integration (spatial data and non-spatial data).
IDEAS will achieve this integration through advanced multi-tasking workstation environments with powerful user interfaces, high level graphics, high speed communications, and large mass storage devices within a multifacetted object-oriented environment. IDEAS will take advantage of current advances in technology while anticipating future technologies such as parallel processing.
This paper will provide a general description of the functions of IDEAS. It will then describe a modeling framework for use in analysis and modeling integration. Finally, potential applications of the IDEAS and spatial dynamic modeling.
This section will briefly discuss the role of IDEAS in providing its three functions of decision support, analysis and modeling, and data management. This broad scale framework is a part of a long term research and development program. Current research focusses on improving the GIS analytic capabilities by providing spatial dynamic capabilities using existing GIS databases.
IDEAS will function as a decision support system for a variety of users. It is designed as a multi-layered system which represents knowledge in a hierarchical structure for easy access and use. At the outermost layer of IDEAS will be an intelligent graphic user interface which acts as a system manager, to direct users to the most appropriate level of the system. The public may wish to review maps or reports on various natural or cultural features of a natural reserve. Policy makers may be interested in current status of various projects or the status of management plans with related budgets. Planners and managers may be interested in summaries of the conditions of various locations in the study area or may wish to simulate alternative management scenarios as part of the planning process. Resource specialists may need to access data from other specialists or may need to run statistical analyses or evaluate the impact of management proposals on resources in their specialization. Research scientists may need access to different analysis and modeling tools, such as statistical packages or geographic information systems.
All of these users have very different needs yet all benefit from sharing a common data base. It is the purpose of IDEAS to facilitate this integration to reduce redundant data base development, provide better decision making tools, and provide a framework for integrating social, ecological, and economic aspects of a management area.
IDEAS is designed as a modeling and simulation system based on object oriented and multifacetted modeling concepts. Past and on going research has developed and continues to develop methods for modeling environmental processes such as forest dynamics (Grossman and Schaller, 1986; Shugart, 1984), hydrologic processes (Haan et al., 1982; Moore et al., 1988), and recreation behavior (Carls, 1978; Timmermans, 1987; Wagtendonk and Benedict, 1980) with specialized tools such as GIS packages and statistical analysis techniques. It is the task of IDEAS to provide a framework for integration of these individual models in a systems-like simulation environment for modeling the interactions of these components of the system through time. It will use a graphic user interface to allow the user to conceptually model his system of interest in a user friendly, highly flexible environment. In order to simulate a multi-layered integrated system, IDEAS will implement two fundamental modeling methodologies under the multifacetted approach. Discrete event simulation (DEVS) (Zeigler 1976, 1984) is used at the global level for modeling the hierarchical system structure through time. Cellular automata theory is implemented within the DEVS concept, to model the spatial event at the discrete model level. Both of the concepts will be more fully discussed below.
IDEAS will include a long-term framework for data collection of spatial and non-spatial data. Existing digital mapped information in MOSS/MAPS format, Arc-Info, DLG, DMA or GRASS, ERDAS or other formats will transferred to the IDEAS environment. IDEAS is seen as supplementing the analytic capabilities of existing GIS and not as a replacement for them. Custom interfaces to other data formats can be written to link a broad range of data formats to the IDEAS environment.
Recent works by several authors support the use of dynamic modeling in environmental analysis. Macgill (1986) provides an assessment of different modeling styles including a discussion of system dynamics models and their possibilities for scenario testing. Sampson (1985) presents a view of modeling and simulation based on systems methodology. Sheldon (1984) suggests the use of system simulation techniques for determining optimum production strategies for natural resources. Couclelis (1985) suggested the use of cellular automata in modeling geographic systems and Itami (1988) and Gimblett (1989) applied the cellular automata concept within a GIS. Green (1989) has studied the utility of cellular automata and percolation theory in spatial patterns and dynamics in forest ecosystems. All of this work indicates a trend toward the use of modeling system dynamics as a decision support tool in environmental management and decision making. This next section describes a conceptual framework for modeling the spatial dynamics of environmental systems for modeling integration in IDEAS.
An environmental system is viewed as a hierarchically structured set of components. Within the IDEAS framework, these components are viewed as events linked through time. This hierarchy of events can be decomposed, each event being composed of more discrete events. Many of these events are events of spatial change and variability. IDEAS provides the modeler a simple yet powerful method for dealing with the problems linking these spatial events through time into a global system model. This linkage is accomplished by conceptualizing the environmental system in two system spaces, a time system and a physical spatial system. IDEAS deals with these two systems simultaneously by implementing discrete event formalism (Zeigler 1976) as a method for modeling the time system, and cellular automata theory within the discrete event formalism as a method for modeling the physical spatial system (Clark 1989).
Using the DEVS framework(Zeigler, 1976), IDEAS views each component model as an event in time. Discrete categories of data from the environment are viewed as state variables of the event, which together are called the event's state. Each event has a set of inputs, a set of outputs, a set of possible states, and transition functions which define the changes in the state of the event. An internal transition function defines a self induced state change, an external transition function defines a response to an input, an output function defines the appropriate output, and a time advance function determines the time of the next internal transition. These component models can be coupled in a hierarchical fashion, to create a multidimensional system model. This model can be simulated through a space in time, allowing system changes to be monitored at any level. Thus the resources/ecosystem can be modeled at multiple levels, creating a powerful decision making and management tool.
Discrete problems areas can be modeled as a system and simulated in a series of "what if" scenarios to determine the affects of alternative management actions on the system. The time space of the simulation can be varied to look at long term as well as short term consequences of the various alternatives. Thus the modeling environment becomes not only a surrogate laboratory for the scientists, but also a decision support tool to the managers and policy makers, as well as a valuable public education tool.
Having established a method for modeling the interaction of components events of a system, cellular automata theory is implemented to allow modeling of the spatial nature of the individual component event. From work done by Zeigler (1976, 1984) and Couclelis (1985) cellular automata theory can be generalized to fit into the DEVS modeling structure (Clark 1989). This cellular automata framework (CA) places processors on equally spaced grid locations (this is analogous to current raster data structures). In the CA framework, the event model has a set global state which is comprised of the local states of each of the processors. It has a set of inputs (other map overlays), a set of influencers (neighboring cells within the same map matrix or cells in input overlays), a set of transition rules (spatial analytic operators in conventional GIS systems), and a set of outputs (maps generated by application of the transition rules). The transition rules are applied at equal time steps to simulate the change of the event state through time. By implementing this CA framework within the DEVS structure, IDEAS provides a framework for simulating not only the spatial nature of some event, but also the influence of other components of the system on that event and vice versa.
The cellular automata framework for spatial dynamics proposed by Couclelis(1985) (following Ziegler,1976) was demonstrated using Map Analysis Package for the PC (Tomlin,1986) by Itami(1988). In that demonstration, residential site selection behavior including environmental and social rules over 30 iterations was simulated. Changes in the state of the environment were modeled for residential site capability, vegetation cover, and visual quality. Site preferences based on social criteria were generated for low, medium and high income residents. The demonstration showed good potential for Cellular Automata models using conventional raster based GIS. However many limitations were found in neighborhood mathmatical operators, running many iterations on complex sequences of spatial operators, and the long computing times even on small matrices. These problems were re-emphasized in work by Vasconcelos(1988) working on simulating forest fire behaviour using MAP. Many of the calculations had be run externally from the GIS further aggravating the time needed to run the simulation. Many of these problems are being alleviated bye new GIS operators for improved overlay mathematics and cell-neighborhood mathematics (see George Ball's paper in this conference).
Clark(1989) developed the conceptual framework for linking Ziegler's(1976,1984) DEVS framework with the Cellular Automata framework. He demonstrated the DEVS framework using DEVS-Scheme(Ziegler, 1987) for simulating the reactions of managers and recreationist to environmental disasters. While the demonstration showed great improvements in modeling interactions between spatial and non-spatial events, it did not implement the cellular automata modeling framework. The linkage between DEVS and Cellular Automata is the next logical step in developing the IDEAS analytical framework.
The purpose of this paper has been to describe IDEAS and a conceptual framework for modeling and simulating temporal and spatial dynamics, which can be used as a decision support tool in environmental management. First, the framework needs to be programmed and written into existing software for testing. The subsequent testing of the framework should be conducted on a contained and simple environmental system in which the component models have been developed and validated. Most importantly, reliability testing of the model should be conducted over a range of temporal and spatial scales, in order that questions of utility in application can begin to be addressed.
Research behind this work was funded by USDA Forest Service, North Central Forest Experiment Station, Chicago, Illinois, cooperative agreement # 238818 and the USDA Hatch funds.
Carls, G.E. 1978. A Simulation Model of Wild River Use. Leisure Sciences, 1, 3, 209-218.
Clark, J.D. 1989. A Framework for Modeling and Simulating Spatial Dynamics. Master of Landscape Architecture Thesis. School of Renewable Natural Resources, University of Arizona, Tucson, Arizona.
Couclelis, H. 1985. Cellular Worlds: A Framework for Modeling Micro-Macro Dynamics. Environment and Planning A, vol. 17, pages 585-596.
Gimblett, H.R. 1989. Dynamic Modelling in GIS: A Cellular Automaton Approach to Modelling the Growth Behavior of Urban and Rural Development. GIS National Conference '89. Ottawa, Ontario, Canada, February 28 to March 2.
Green, D.G., et al. 1989. A Generic Approach to Landscape Modelling. in proceedings from Eighth Biennial Conference of Simulation Society of Australia, Canberra, Australia.
Grossman, W.D. and Schaller, J. 1986. Geographical Maps on Forest Die-Off, Driven by Dynamic Models. Ecological Modeling, 31, 341-353.
Haan, C.T., Johnson, H.P., and Brakensiek, D.L. 1982. Hydrologic Modeling of Small Watersheds. American Society of Agricultural Engineers Monograph Number 5, St. Joseph, Michigan.
Itami, R.M. 1988. Cellular Automatons as a Framework for Dynamic Simulations in Geographic Information Systems. in GIS/LIS '88 Proceedings. American Society for Photogrammetry and Remote Sensing, Falls Church, Virginia.
Itami, R.M., Clark, J.D., Pereira, J.M.C. 1988. IDEAS - Integrated Digital Ecological Analysis System: A Framework for Dynamic Modeling Using GIS. in Proceedings of Conference on Spatial Data Systems for Management. The University of Arizona, Tucson, Arizona.
Macgill, S.M. 1986. Evaluating A Heritage of Modelling Styles. Environment and Planning A, 18, 1423-1446.
Moore, I.D., O'Loughlin, E.M., and Burch, G.J. 1988. A Contour-Based Topographic Model for Hydrological and Ecological Applications. Earth Surface Processes and Landforms, 13, 305-320.
Sampson, J.R. 1985. Biological Information Processing. J. Wiley and Sons, New York.
Sheldon, R.A. 1984. Optimal Management of Renewable Resources Using Computer Simulation. Journal Environmental Management, vol. 18, pages 89-93.
Shugart, H.H. 1984. A Theory of Forest Dynamics: The Ecological Implications of Forest Succession Models. Springer-Verlag, New York.
Timmermans, H. 1987. Hybrid and Non-Hybrid Evaluation Models for Predicting Outdoor Recreation Behavior: A Test of Predictive Ability. Leisure Sciences, 9, 67-76.
Tomlin, C.D. 1986. The IBM Personal Computer Version of the Map Analysis Package. Laboratory for Computer Graphics and Spatial Analysis, Harvard Graduate School of Design.
Wagtendonk, J.W. van and Benedict, J.M. 1980. Travel Time Variation on Backcountry Trails. Journal of Leisure Research, 12, 2, 99-104.
Zeigler, B.P. 1976. Theory of Modelling and Simulation. Krieger Publishing Company, Malabar, Florida.
Zeigler, B.P. 1984. Multifacetted Modelling and Discrete Event Simulation. Academic Press, London.
Zeigler, B.P. 1987. Hierarchical, modular discrete-event modelling in an object-oriented environment. Simulation, vol. 49, number 5, pages 219-230.