Figure 1: Spatial process modeling intergrates GIS and process modelling
Figure 2: Visual Basic implementation of SPMS showing partially developed model
Abstract
This paper discusses the background, methods and initial testing of a Spatial Process Modelling System. While the system is intended for general use, a case study of a class of environmental problems is described for which current support is poor. These are the issues involved in regional environmental decision making (REDM).
Computer support for environmental problems
Previous papers have discussed the difficulties with developing computer support for environmental problems, particularly REDM. This has resulted in a set of criteria required for development of useful tools. These criteria are:
Integration
These criteria make explicit the requirement for a linking of spatial and process modelling. There have been many calls for such an integration and taxonomies of the available methods fill the literature, but there remains few systems which have attempted this move. A brief review examines the current state of integration with the intention of finding a path for development.
A distinction can be identified based on a difference in focus; an emphasis on ‘content and structure’ for GIS and ‘content and process’ for modelling. A major conflict is the way in which dynamic processes are represented. An environmental scientist might produce a map of evaporation for a particular day but would find it harder to use a GIS to represent the processes involved in evaporation.
This review concludes that there is a mismatch in interface, in data organisation and in general approach. Most of these problems come down to a difference in what is represented in terms of perceptions; pattern (data) or process (dynamics). This should not be an overwhelming obstacle, much of geography could be described by the interaction of these themes. The task then is to develop a GIS/model hybrid that best mixes the advantages of both pattern and process while overcoming the inherent conflicts. Such a system is referred to as Spatial Process Modelling (Figure 1).
Figure 1: Spatial process modeling intergrates GIS and process modelling
Figure 2: Visual Basic implementation of SPMS showing partially developed model
Focus
The criteria given above also have an emphasis on the importance of encouraging human interaction and thought. The issues of REDM can be characterised by their complexity and tendency to result in conflict between different groups. The underlying goals of this project are to provide a means to lessen this conflict and to allow reasoned analysis, and as such, care is taken over the approaches used in providing a useful platform for modelling.
Various approaches are considered including that used for AEAM workshops, question-template approaches, high level pseudo-english and diagrammatic approaches. These discussions are synthesised in a set of design criteria.
A review of three systems which partially fulfil the design criteria demonstrates both the potential benefits and difficulties inherent in this approach.
A final section describes the Spatial Process Modelling System. First, there is a technical description of the prototype implementation and second, discussion on the initial testing of the system is presented by way of a worked example.
Overview
The SPMS essentially inserts map objects directly into a Stella-like visual modelling toolbox. This significant step has resulted in a very powerful yet flexible tool but has also raised many methodological issues.
In the SPMS users build process diagrams interactively (see Figure 2). At the core of the program, the GIS and modelling functions are combined. Within a graphical environment, process models (emulating objects) are linked together visually. Also available to form part of the model are spatial objects, with inputs and outputs. Objects may be joined to form complex structures allowing feedback mechanisms.
Lines are drawn linking the objects and then the equation for each process model is defined according to the inputs and outputs. This equation may be relatively simple (e.g. add 5 to input 1 when input 2 = 10), include more complex geographical operators or link to a database via a neural network where relationships are not known. The equation is also ‘flagged’ with any assumptions made, for example, ‘this is how tussock grass responds to burning, I know that it is different in very dry years’. This has the duel advantage of clearly laying out assumptions, and in directing areas where further research is needed. The inclusion of technical details in the model but hidden from view also facilitates the integration of research findings from a number of disciplines.
Technical description
A full technical presentation of the methods used in developing the SPMS is given. This includes data flow diagrams and a description of a series of arrays that interpret the diagrammed models and perform the processing. A major component is that which keeps track of the current status of the model and its components.
Worked example
The power and flexibility of the system is demonstrated by example. The significant advances include particularly the inclusion of feedback loops and the ability to modify model structure within a spatial paradigm. Remaining difficulties include the concepts and difficulties involved in hierarchical representation, modular links to other modelling tools, and data management. More work is also required on the representation of error. Testing and model validation will eventually be linked to an automatic report writer that lists known assumptions and inadequacies, along with assessments of sensitivity to variation in model components.
Difficulties encountered when dealing with the environment result in inadequate regional decision making. This situation is not helped by the current generation of computerised aids. This paper outlines the basis for a spatial simulation and modelling system to be used in regional environmental decision making. The SPMS is an inherently flexible yet powerful system and is applied initially to the New Zealand high country. The system relies on the successful integration of spatial processing and display with the characteristics of system dynamic modelling.