Decision Support Systems, Environmental Models, Visualisation Systems and GIS

Introduction

Combinations of technologies

Decision Support Systems

Modelling within DSS

Visualisation Systems

Visual Resource Management Systems

Inter-operability

Computer simulation languages

The temporal dimension

Scale

Decision Support Systems

Descriptions of some DSSs

FORPLAN

TERRA-Vision

RELMdss

SYLVATICA

Forest landscape simulation model

Components of Decision Support Systems

Components of specific DSSs

Hierarchical structures

Hierarchical levels in RELMdss

Models

Environmental models without landscape Visualisation

DEVS-Scheme model

GIMS

ECOLECON

Regional land use planning model

Ecological and vegetation models

Vegetation modelling

Tree population modelling

Forest growth gap model

RELMdss optimisation models

Visualisation and Decision Support Systems

Visualisation Systems in DSS

ALBE GIS and AVS

RELMdss and UVIEW

Forest landscape simulation model

Arc/Info and SiteView

SmartForest

Data Transfer

Arc Macro Language (AML)

ASCII files

Geographical Information Systems in Decision Support

Rationale for combining GIS with other systems

GIS and modelling

ECOLECON

TERRA-Vision

SYLVATICA

GIS and Visualisation

RELMdss

GIS and CAD

References

Introduction

Much of the recent change in the planning process is due to heightened concerns to protect habitat, ensure the survival of endangered species, promote biological diversity, provide recreation, and to balance these concerns with economic issues. The focus of management has changed to concentrate on managing ecosystems as a whole rather than focusing on specific species, in order to ensure ecosystem sustainability by protecting habitat and promoting biological diversity (Church et al, 1994a; 1994b).

Host et al (1992) state that:

"It is this integration of space and time in the broader context of the regional landscape that must be the focus of environmental and natural resource management (Host et al, 1992)."

There is a need to combine the technologies of GIS, Visualisation and environmental modelling to produce decision support systems. Examples of DSSs which combine two or more of these technologies are reviewed.

Combinations of technologies

Decision Support Systems

The decision support field has been defined by Sol (1983) as the:

"development of approaches for applying information systems technology to increase the effectiveness of decision makers in situations where the computer can support and enhance human judgment in the performance of tasks that have elements which cannot be specified in advance."

Decision support systems must provide integration of information and feedback loops to support the exploratory nature of the process of scientific discovery. Issues of information integration include the ability to scale information without distorting the technical content, to handle temporal information and to support spatial information throughout the analysis and decision-making process (van Voris et al, 1993).

Modelling within DSS

Modelling is an important component of decision support systems. Accurate predictions of impacts, ecological, economic, social and visual is essential to effective decision support.

Visualisation Systems

User expectations concerning ease of use and clarity of interpretation have increased to the point where it is expected that analysis tools convey the impacts of various management plans using techniques conducive to instant understanding and meaning. Visualisation techniques have proven invaluable in the presentation of analysis results (Church et al, 1994b). These techniques can be crucial in supporting DSS users in gaining new insights into the structure of their problems by generating different views of the decision situation and by exploiting their own visual skills so that they can recognise meaningful alternatives and strategies during the problem-solving process (Angehrn and Luthi, 1990).

Visual Resource Management Systems

Visual Resource Management Systems (VRMS) should be able to predict and assess impacts resulting from potential management alternatives. To facilitate land use management and planning, land managers, planners and designers must be able to ask questions of a VRMS that enable them to assess potential consequences of potential actions and thereby select successful, politically acceptable, cost-effective solutions (Bishop and Hull, 1991).

Inter-operability

To support an effective decision-making process inter-operability must be provided within the decision support system. Inter-operability is the capability to organise and transfer information between scientific models and cross-functional components of the integrated system (van Voris et al, 1993).

In an effective decision support system the decision-maker would need only to identify the source of the information to have the proper modelling parameters input into the system. Inter-operability between models is the ability for individual models to contribute to the resolution of a larger problem. In decision-support components, inter-operability is measured by the ability of the system to orchestrate the acquisition, transformation and presentation of information throughout the decision-making process. The most difficult issue in inter-operability is the ability to use the graphical interface, decision support system and integrated information management component to 'feed' decisions back into the system to perform further iterations and assessments (van Voris et al, 1993).

Computer simulation languages

The evolution of computer languages can be divided into three stages or generations: unstructured; structured; and object-oriented (Liu, 1993). Unstructured languages include FORTRAN and BASIC and are the most primitive of the languages; structured languages, such as PASCAL and C are less primitive and have given rise to the object oriented languages currently used in many decision support systems.

Object-oriented languages have a hierarchical organisation consisting of objects which are sections of code reflecting the characteristics and processes of different entities. Although an object has its own distinct properties, it inherits properties of higher-level objects since an object is one of the nodes in a network hierarchy (Liu, 1993).

The most advanced generation of computer languages is object-oriented programming (OOP) languages, used in the area of artificial intelligence. OOP languages include C++, Smalltalk and LISP. C++ was developed as a translator program that processes C++ source code into C source language. There are three major problems for the OOP feature in C++: encapsulation, inheritance, and polymorphism. Inheritance is the technique of inheriting characteristics from higher-level classes in the class hierarchies; polymorphism in C++ lets programmers use many versions of the same function throughout a class hierarchy, with a specific version to be executed at run time. At present there exists no standard C++ language (Liu, 1993).

The temporal dimension

Natural systems change slowly and impacts on them become evident only with the passage of considerable time. This apparent resilience of the impacted system may mask changes that are, in fact, impossible to halt and irreversible. At later stages in the change process the absolute level of change may be significant, but the evaluator may have habituated to the changing conditions and therefore be less sensitive (Imaging Systems Laboratory, 1995). The RELMdss system (Church et al, 1994b) allows for tracking impacts forward in time. At any given time period, each activity is limited by activities of the previous time periods by relationships that revise threshold limitations and attribute levels in future periods.

Proper treatment of time is as important as proper analysis of structure. A model in which there are distinct time steps for very process is difficult to construct , and even when it is possible to control time steps from the faster to the slower components of the system, there may be many computation steps without significant state transitions of the components involved. In the methodology used by Perestrello de Vasconcelos et al (1993) this problem is avoided by specifying hierarchical models in the DEVS discrete-event formalism.

Scale

The ability to transfer modelling output to other scales without losing the validity of the information is necessary for effective decision making. It is difficult to apply models which simulate the growth of individual trees for areas measured in metres and hectares to landscape and regional scales whose metrics are in kilometres (van Voris et al, 1993).

Landscape in the foreground of a view is viewed at a different scale to that in the distance. In a Visualisation of landscape, the local landscape elements could be enhanced while those in the distance are reduced. The impact and dominance of the same element will be less in the background than in the foreground (Baldwin et al, 1996).

Whilst the shapes and forms of the world surface can be modelled within the GIS environment, it is not so simple to define the specific boundaries of mountains and valleys, plains and plateaus for digital analysis (Baldwin et al, 1996).

Decision Support Systems

Decision support systems can be described as analytical tools which can be used to assist the decision maker in assessing the inter-relationships and potential effects of a policy or decision. Analytical results are presented to the user through the Visualisation and user interface component of the DSS. Ideally, a decision to implement a change would be reflected in a model's raw data, resulting in changes to the analytical data sets; this may invoke different decision support relationships to be constructed which can be reviewed visually and assessed by the decision-maker in a positive feedback loop (van Voris et al, 1993).

Many of today's DSSs focus on problem solving rather than on supporting the modelling process, but the main goal of a DSS should be to provide decision makers with tools for interactively exploring, designing and analysing decision situations (Angehrn and Luthi, 1990). Users should be able to perform the following functions: they can analyse decision situations according to their personal styles and knowledge; they can build and compare various quantitative models; they can adapt these models to changing conditions; and they can evaluate different aspects of their activities using a variety of different means.

For DSSs to be truly integrated, planning teams using them need to have a broad range of expertise across the disciplines of relevance to the planning problem. The decision makers, planners and users must be educated as to how to best utilise the systems for the management of natural resources (Kent et al, 1991).

Many of the DSSs found in the literature are based on forestry planning. Although simple forest treatment rules do not cover the range of situations, computer simulation is commonly used in forest planning to study the consequences of different management alternatives and can be used, in conjunction with Visualisation systems, to simulate the change in forest landscape (Kellomaki and Pukkala, 1989).

Descriptions of DSSs

Several DSSs are mentioned in this review. A brief description of some of them may assist in looking at the uses and capabilities of DSSs.

FORPLAN

FORPLAN is a large-scale linear programming system used to support national forest land management planning for the USDA Forest Service. It consists of a matrix generator and a report writer, both of which interface with a commercial mathematical programming solution package (Kent et al, 1991). The system is used to construct forest models which simultaneously allocate forest land to general management objectives and schedule the treatments and the resulting product flows.

TERRA-Vision

A decision support system for risk assessment of terrestrial environmental resources which combines scientific analysis and the decision making process in a DSS. The goal of the system is to provide a scientifically based method for establishing the potential effects of a policy of decision. The TERRA-Vision prototype was used to identify and evaluate methods of presenting decision-support information graphically. The system combined graphical and mapping capabilities provided by a GIS with a 3D Visualisation system with temporal and real time capabilities (van Voris et al, 1993) along with models representing environmental, atmospheric, economic and political criteria.

RELMdss

RELMdss is a spatial decision support system, developed as a forest planning tool. It has a hierarchically based modelling framework to assist in the development of land range forest management plans. The system attempts to optimise forest operations and the resulting spatial patterns. The main objective of RELMdss is to provide the capability to generate and display different management schemes that maintain spatial and temporal constraints, while achieving scenario based goals (Church et al, 1994a; 1994b).

SYLVATICA

SYLVATICA is an integrated framework for forest landscape simulation; it used a variety of consultation systems to allow the user to visualise the effects of silviculture or other resource management strategies, natural or anthropogenic disturbance, or global climate change over long term horizons (Host et al, 1992). Hypertext systems provided access to an organised and structured scientific knowledge-base and expert systems based on silviculture, wildlife management and forest growth were used. The approach is conceptual, and integrates several resource management technologies in a visual interactive environment.

Forest landscape simulation model

The forest landscape simulation model of Kellomaki and Pukkala (1989) determines the amenity value of forest landscapes subjected to a selected management regime. The model provides quantitative predictions of the temporal development of tree and stand dimensions. The landscape created by computer graphics is composed of tree symbols whose species and size distribution correspond to those in nature. The method is only suitable for management regimes which are sufficiently different.

DSS components

Functionally, a DSS is made up of three components: a language subsystem, a knowledge subsystem, and a problem processing subsystem. The problem-processing component is then responsible for activating the available functions, and generating and conveying the appropriate information to the decision maker. The language component determines whether the functions of a potentially rich system are transparent to the user (Angehrn and Luthi, 1990). Some of the components used in specific DSSs are noted overleaf.

Components of specific DSSs

The conceptual model for the TERRA-Vision proof-of-concept prototype has four main components: analytical modelling, integrated information management, decision support and the Visualisation-based user interactions (van Voris et al, 1993), these components are used together to form a forest management and decision support system. The analytical modelling was based in a resource dynamics forest growth model; the integrated management systems was a GIS. The decision support system examined the effects of temperature on forest composition and total biomass based on predicted trends for global warming. The final component was a multi-level animated graphics display capability.

The components used by RELMdss were similar to those of TERRA-Vision - a treatment based optimisation system, a raster-based GIS and a 3D terrain viewer. The combination of these systems provided a method to simulate, investigate and visualise forest landscapes and watersheds at various geographic scales over time (Church et al, 1994b).

Host et al (1992) describes the "fundamental components required to build a comprehensive geographically based simulation and decision support system":

1. a GIS will be used to create, manipulate and analyse spatial data;

2. a database management system (DBMS) will manage numerical data associated with the forest stands as well as individual tree data;

3. forest simulation models will be used to change forest composition over time according to successional pathways, environmental conditions and management practices;

4. a hypertext system (HS) will be used to structure and provide access to existing knowledge about the system;

5. knowledge-based management systems (KBMS) will be used to provide opinions and advice through simulated consultations with experts in various fields;

6. tutorial systems will be used to watch over the user and provide guidance and direction; and

7. a graphical user interface (GUI) will provide a common link between the user and the model's subsystems.

Hierarchical structures

One of the central ideas of an object-oriented, hierarchical modular modelling techniques is that entities can be identified, whose behaviour can be detailed independently through specification of interactions with their neighbours and environment. In addition to this several levels of aggregation must be studied simultaneously. Hierarchical structuring means that at any given level of resolution, an ecological system is composed of interacting components (lower-level entities) and is itself a component of a larger system (a higher level entity) (Perestrello de Vasconcelos et al, 1993).

Hierarchical levels in RELMdss

RELMdss (Church et al, 1994a; 1994b) used four conceptual levels of analysis (regional, forest, ranger district and operational) and had a design objective to support a hierarchical linkage if the various planning levels. Hierarchical planning in forestry evolved because it mimics the decision making process defined by regulations and case law, where certain planning decisions are delegated to various levels in the hierarchy (Church et al, 1994a). Each level of planning is addressed by the development and application of optimisation models.

The top of the hierarchy involves large scale and regional planning to provide the overall direction in the planning process. Strategic planning defines broad scale regional goals and forest-wide plans to direct the next level in the hierarchy. The forest-wide plans are translated to specific tracts of land for the tactical level of forest planning. The lowest level of the planning hierarchy is based on operational decisions on specific lands (Church et al, 1994a).

Models

Environmental models without landscape Visualisation

Four examples of environmental models which do not use landscape Visualisation are described. They do have some form of Visualisation, through more abstract forms of landscape. All are forms of decision support systems.

DEVS-Scheme model

The understanding of ecosystem structure and function requires incorporating knowledge at several levels of resolution. To study several levels of organisation simultaneously it is necessary to understand how the processes at different levels interact and coherently couple the different times and spatial scales (Perestrello de Vasconcelos et al, 1993). The modelling scheme put forward by Perestrello de Vasconcelos et al (1993) used a discrete event systems (DEVS) framework, which can be used to specify dynamic processes occurring at three different levels of resolution: the key-species, the vegetation patch and the whole landscape (Perestrello de Vasconcelos et al, 1993). The modelling approach can be summarised as follows:

1. Formalisation of species atomic models, and coupling them within a patch;
2. Formalisation of an experimental frame for the simulations;
3. Definition of assumption for interactions between patches. Formalisation of interactions between patches by coupling;
4. Design of system entity structure applicable to simulation if the dynamics of the landscape under study;
5. Pruning for creating a model of a one-patch landscape, the composition tree incorporates only one patch;
6. Pruning for creating the model representing the landscape.

GIMS

GIMS (geographic and information modelling system) was developed by national park management authorities in Australia seeking to implement modelling capabilities for vegetation and fuel dynamics, fire behaviour and its environmental effects in fire-prone national parks and adjacent rural and suburban lands. During its initial development GIMS was seen as an environmental modelling and decision support system rather than as a GIS and the intent was to provide a methodology for modelling the behaviour of fires, for assessing strategies for fire control and for determining the potential impacts of fire (Kessell, 1990).

ECOLECON

ECOLECON, an ecological-economic model, is a spatially-explicit, individual based and object-oriented program which has been developed to simulate animal population dynamics and economic revenues in response to different forest landscape structure and timber management scenarios. ECOLECON can generate artificial forest landscapes or can link with GIS to run simulations on real landscapes. The model predicts population dynamics, spatial distribution, extinction probability of a species under consideration as well as future landscape structure, and economic income from timber harvest based on current tax and timber market situation (Liu, 1993).

Quantitative modelling is one of the most useful approaches for dealing with ecological and economic issues simultaneously. ECOLECON simulates animal population dynamics and economic cash-flows in response to landscape structure and timber harvest in managed forests. (Liu, 1993). Some of the options available to the user of the model are:

1. Options for the number of replicates and the simulation length;
2. Choice of either creating hypothetical landscapes or linking ECOLECON with real landscapes from GISs. To produce an artificial landscape, users can select different landscape shapes, sizes, composition and configurations;
3. Variable demographic parameters;
4. Habitat selection rules;
5. Economic variables;
6. Result output options.

Regional land use planning model

This model determines the land use mixes optimising a pre-defined multi-objective function, including economic and ecological variables, and defines control strategies necessary to achieve the optimal mixes at a given point in time. The model is built on heuristics, network theoretic concepts, dynamic simulation and multi-objective programming (Camara et al, 1986).

The model considers the region represented in a polygon map where each cell stands for a land use. Each cell may be visualised as a node in a network, which represents a region. Arcs between nodes then represent incoming or departing economic, ecological, social or aesthetic flows. Alternative land uses are generated using suitability analysis models; unfeasible alternatives are then eliminated by applying economic, ecological, social and political criteria (Camara et al, 1986).

Ecological and vegetation models

Ecosystem models will be valuable for evaluating long-term impacts of factors such as climate change, increased ozone or other atmospheric pollutant levels, or acidified deposition. By linking these models with the spatial analysis capabilities of GIS and changes in ecosystem boundaries, changes in the landscape may be predicted (Host et al, 1992). An example of such models is that of Host et al (1992) which used forest growth and yield models as a means to evaluate forest growth and the economic viability of different timber harvesting scenarios (Host et al, 1992).

Vegetation modelling

The primary goal in vegetation modelling is to simulate overall landscape texture and pattern rather than specific, detailed vegetation structure. UVIEW, a subsystem of RELMdss, models vegetation patterns to simulate existing or desired landscape conditions. Canopy closure based vegetation modelling provides vegetation patterns over an entire landscape. The method represents differences in stand densities well, but does not represent differences in stand composition and structure. Vegetation modelling based on structure definitions represents both stand density and stand composition (Church et al, 1994b). The Visualisation system Vistapro uses elevation data to assign vegetation types to user-specified elevation ranges, however, in their study, Berger et al (1996) vegetated the landscape based on actual land cover distribution, which is often independent of elevation.

Frequently the need arises to constrain the amount of suitable wildlife habitat on a portion of the forest. This habitat is usually measured with an index that represents the estimated proportion of an acre that is suitable habitat. This index is a function of factors like the type of vegetation present on an acre, the age of this vegetation and the type of management being implemented (Kent et al, 1991).

Tree population modelling

Tree population can be modelled from data gathered in the field. Theoretical tree population for each forest compartment is calculated in two stages. Firstly, the diameter distribution of stand basal area or number of stems is estimated from the field data using the beta function as a theoretical distribution. The diameter distribution is calculated separately for each tree species and canopy layer. Secondly, each distribution is divided into three classes of equal width and the class midpoint is taken to represent the class. This tree is described by species, stem diameter at breast height and age. The initial age is obtained from field data and tree heights are predicted based on stem diameter data (Kellomaki and Pukkala, 1989).

The simulation method can be used to visualise the short term and long term impacts of proposed treatments. In addition to landscape drawings, the simulation produces quantitative predictions on tree removals and changes of stand dimensions. The simulation provides ample information for decision making in economically managed forest areas which are situated in scenically important locations (Kellomaki and Pukkala, 1989). The method is most suitable where forest treatments are significant and cover a reasonably large area, the ideal area for the model is between 5 and 30 hectares. The effects of slight differences in treatments cannot easily be evaluated from the computer drawings.

Forest growth gap model

Gap models simulate forest dynamics by tracking the establishment, annual growth in diameter and eventual mortality of each tree on a small model plot, corresponding to the zone of influence of a canopy dominant tree. This is an explicitly hierarchical approach to simulating a forest stand: trees are the basic building blocks, these interact at the gap scale, and a stand is created from multiple gaps (van Voris et al, 1993). The model is spatial to the extent that competition in the vertical dimension is considered. In the TERRA-Vision model, the gap model was extended to the landscape scale, ignoring explicitly spatial aspects of forest development and focusing on a simple elevation gradient in temperature (van Voris et al, 1993).

The regeneration of seedlings on a plot and their subsequent growth is based on the silvicultural characteristics of each species, including site requirements and sensitivity to environmental factors (water and temperature). The trees in a model plot collectively determine the amount of light available at each height position via their leaf area profiles and heights (van Voris et al, 1993).

At the finest scale, a simple tree icon was developed to represent each tree as a lollipop with its diameter, height, of crown and crown diameter proportionate to these as simulated in the gap model. The tree's foliage density was used to assign a transparency to the crown, indicating its ability to shade other trees on the plot. The Advanced Visualisation System (AVS) was used to generate a 'gap' Visualisation utility which input the ASCII file and generated a three-dimensional display of the individual trees for each time step of the gap model (van Voris et al, 1993).

RELMdss optimisation models

Four optimisation models were developed for RELMdss. The first model is called the Minimum Area model, which involves identifying activities that meet targets or goals as well as a general goal of minimising the acreage that is subject to treatment. The Equivalent Risk model spreads activities across the forest whenever flexibility allows; it determines a solution that optimises one or more objective terms as well as minimises the largest percentage of any threshold constraint reached by the assignment of activity. These models attempt to meet volume targets in each time period (Church et al, 1994b).

The final two optimisation models are multi-objective versions of the previous two. The multi-objective version of the Minimum-Area model can be used to optimise activities in conflict with harvesting, where harvesting can be given a weight that is low enough that, in effect, harvesting will be accomplished only when it helps to meet specified conditions. The multi-objective version of the models can, in fact, be used to produce desired future conditions (Church et al, 1994b).

Visualisation and Decision Support Systems

Visualisation Systems in DSS

Visualisation is important part of decision support in landscape planning and can be crucial in supporting DSS users in gaining new insights into the structure of their problems by generating different views of the decision situation and by exploiting their own visual skills so that they can recognise meaningful alternatives and strategies during the problem-solving process. (van Voris et al, 1993)

Visualisation of input parameters will assist the scientist in checking data for content and correctness. Visualisation of output parameters allows scientists to understand better the resultant data sets as well as their relationships to other data sets. The ability to visualise the information over space and time adds perspective to the scientists' and the decision-makers' understanding (van Voris et al, 1993).

ALBE GIS and AVS

The ALBE GIS is a very flexible, general purpose GIS for the display of a variety of data input from user supplied models. It was developed specifically to support analytical modelling and the development of decision-support applications and has the ability to integrate modelling, data management, Visualisation and user interface capabilities, making it an effective tool for Visualisation (van Voris et al, 1993).

In the TERRA-Vision prototype, each tree species was assigned a rank position along the elevation/temperature gradient; these ranks ranged from 1 to 9, representing warm-site to cold-site species. Biomass was illustrated as stature. Although the actual information presented at each scale (tree, gap, stand, landscape) might be quite different, the kind of information remains the same (stature, composition) and the viewer does not need to readjust his or her visual cues with every change in scale (van Voris et al, 1993).

TERRA-Vision also used AVS (advanced Visualisation system). An AVS 'landscape' Visualisation utility was developed to register the surface to its corresponding digital elevation data. The surface could then be 'draped' over the elevation using height as a z-value to generate a 3D image. Capabilities to provide lighting and shadowing effects were also provided by AVS (van Voris et al, 1993).

RELMdss and UVIEW

RELMdss is a spatial decision support system which has been developed to help relate the various decision making levels and provides a modelling structure to generate and test consistent alternatives at different levels of the hierarchy. UVIEW is a display system designed to be used in conjunction with UTOOLS in order to produce 2 and 3D images of DTMs, attribute data stored in Paradox spatial databases, and vegetation patterns at landscape scales (Church et al, 1994b).

UVIEW provides a flexible system for viewing a DTM with four parameters controlling the appearance of perspective views: head or eye location; focus or target location; camera lens focal length; and vertical exaggeration. UVIEW also provides a variety of methods and resolutions for displaying a DTM: coarse and fine resolution profiles and grids; solid surface representations with hidden surface removal, with and without lighting (Church et al, 1994b).

Forest landscape simulation model

In the model of Kellomaki and Pukkala (1989) a computer landscape is created by placing tree symbols on the surroundings of the grid points; different species and tree sizes are represented corresponding to the theoretical tree populations. The simulation of growth with specific forest treatments is based on the theoretical tree populations created for each compartment on the basis of field data. The growth is simulated by increasing the diameter, height and age of each tree using models and a time step of 5 years. The cuttings are simulated by decreasing the tree density and canopy layers and regeneration is taken into account by adding new trees to the selected compartment through the keyboard (Kellomaki and Pukkala, 1989).

Arc/Info and SiteView

ARC/INFO is an extensive GIS with vector and raster capabilities including import, georeferencing, editing, analysis and output. SiteView is more limited and lacks several of the major components of a GIS, but it is specifically designed for 3D Visualisation and analysis of surface and subsurface site characterisation data. Data imported with the DXF format is for display only, and cannot be attributed beyond its graphics-related information (Kuiper et al, 1996), alternative data transfer techniques have been developed, as noted in the section on data transfer.

SmartForest

SmartForest is a Visualisation system, able to visualise on both regional and local scales, which can be developed interactively using biological models. Time-scale differences can be addressed and the gradual changes over time visualised. Trees can be queried directly using the mouse to display data from the underlying database as well as calculated indices of crowding, tree-to-tree competition and pest hazard. Forest prescriptions can be applied and the results modelled using the incorporated growth models (Imaging Systems Laboratory, 1995).

3D visual modelling is limited by computing time; to keep this time down, the tree symbols are kept fairly simple. A big obstacle is the difficulty and costs of creating databases. The 3D visual modelling approach has many advantages over GIS-based Visualisations. Each tree is an object with a known location in space, changes made by the user are recorded as changes to the database and all new iterations are based on the new changes (Imaging Systems Laboratory, 1995).

Data Transfer

A common problem encountered in GIS modelling is the exchange of data between different software packages to best utilise the unique features of each package (Kuiper et al, 1996). Although the modelling may be very fast, the process of data transfer can be awkward and slow (Bishop and Karadagli, 1996). For example, Berger et al (1996) found that importing GIS data into Vistapro, a Visualisation system, was not straightforward, requiring the development of special products and the use of other software packages to preprocess the data. Several methods of data transfer found in the literature are mentioned below.

Arc Macro Language (AML)

Arc/Info is used in a large range of research. Mayall and Hall (1994) and Bishop and Karadagli (1996) both used AMLs to integrate Arc/Info with other packages. Bishop and Karadagli (1996) ran their model within the GIS as a series of GIS commands in AML. Mayall and Hall (1994) found that GIS and CAD technologies are limited for the representation of regional visual landscapes, but by integrating the two, their relative strengths could be taken advantage of and used with other models to predict and simulate landscape change. They used AMLs to output GDS command files from the graphic and attribute data stored in the GIS.

ASCII files

Like Mayall and Hall (1994), Kuiper et al (1996) tried using DXF (the AutoCAD drawing exchange format) to transfer data between a GIS and a Visualisation system. This method of data exchange was inefficient and limited the combined use of the systems. The alternative approach used was to design an ASCII file format that was uniquely adapted to the Visualisation system, SiteView and its data model. These were coded using Arc/Info software development libraries (ArcSDL) and C code.

The SiteLink translator benefits Arc/Info by adding 3D Visualisation capabilities and SiteView by giving streamlining access to data from a widely used GIS. The transfer file format was designed as a general purpose file for SiteView, but could be implemented for exchange with other systems. SiteLink can also be implemented for general purpose file export for Arc/Info, especially for transfer to object-oriented systems or those having similar data representation designs (Kuiper et al, 1996).

GIS in Decision Support

Rationale for combining GIS and other systems

A fundamental problem in decision support is that highly specialised analytical programs such as GISs or forest simulation models are often used in isolation. For example, most forest growth models use sophisticated algorithms to simulate temporal change, but consider only point estimates of the composition and growth of individual forest stands/ecosystems (Host et al, 1992).

Although a GIS may have a broad set of data input, processing and output capabilities, they often lack 3D Visualisation and certain modelling functions, whereas specialised object-oriented packages designed for Visualisation and modelling can lack many of the other capabilities of a GIS (Kuiper et al, 1996). This inability to produce effective 3D simulations of landscapes for visual landscape representation and assessment is a major problem with GIS in landscape planning support (Mayall and Hall, 1994).

Mayall and Hall (1994) describe a method to combine the technologies of Computer Aided Design (CAD) and Geographic Information Systems (GIS). Their procedure allows an end-user to obtain quite realistic 2D and 3D digital representations of current landscapes and to produce visual simulations of landscape change. This technology allows numerous planning applications, such as facility siting, environmental impact analyses, transportation routing, and resource allocation, to be undertaken more efficiently than is possible using the traditional, manual methods.

GIS and modelling

Geographical information systems have been combined with many models to produce decision support systems with a spatial context. Three such systems are ECOLECON, TERRA-Vision and SYLVATICA.

ECOLECON

ECOLECON used quantitative modelling of ecological and economic variables within forest ecosystems. Combining the model with a GIS allows users to simulate the effects of various management schemes in real forest landscapes and provides computer simulations, which are a useful tool for providing valuable information and insights for policy making and analysis (Liu et al, 1994).

TERRA-Vision

The goal of TERRA-Vision was to investigate the potential of developing a quantitative spatially-distributed approach for analysing environmental processes occurring at the landscape, regional, continental and global scales (van Voris et al, 1993).

The Integrated Information Management (IIM) component provided the automated link between the scientific information and the user-oriented decision support system by orchestrating the identification and integration of, and access to, the diverse data sets created during modelling. The IIM was, to all extents and purposes, a GIS. This DSS component establishes how the diverse scientific data sets are related to the decision-making process (van Voris et al, 1993).

SYLVATICA

The SYLVATICA concept linked ecosystem models with a GIS to combine spatial analysis with models evaluating long term impacts on ecosystems, in order to examine the change in the landscape. The SYLVATICA simulation will be a game-format visual interface around existing mathematical models of forest growth and ecosystem processes, with associated encyclopaedic knowledge bases and rule-based decision support systems (Host et al, 1992).

GIS and Visualisation

GIS, modelling and Visualisation need to operate together in an interactive computational environment. Ideally the modelling will feed the Visualisation, which in turn influences the human operator who can then change the modelling parameters. Not all systems can currently so this. The model of Bishop and Karadagli (1996), for example, has the decision options for the model set prior to the initiation of the Visualisation, but has a future objective to create a direct link between the modelling and the Visualisation, so that adjustment of controls would create new imagery.

A previously distinct activity has been the development of GIS driven Visualisation including realistic simulation (Bishop and Karadagli, 1996). The term `virtual reality' denotes a system which provides the tools for users to interact with a simulated environment, but not necessarily in real time (Berger et al, 1996). Such systems will combine the spatial display capabilities of GIS and GIS based modelling of environmental impact with high performance visual simulation in a multi-channel graphics environment (Bishop and Karadagli, 1996). Landscapes will be rendered as perspective views using actual elevation and land cover data, so they can depict realistic scenery (Berger et al, 1996).

RELMdss

RELMdss combines a GIS (UTOOLS) with a Visualisation system (UVIEW) to assist in displaying the visual impact of management alternatives. The UTOOLS data structure combines detailed GIS data (polygon, line and raster attributes) into a single Paradox spatial database. Spatial data can be used by UTOOLS to calculate additional layers (e.g. slope, aspect, buffers, etc.) from existing database fields. The resulting databases and vector data can then be mapped with UVIEW, which was designed as a display system to be used in conjunction with UTOOLS in order to produce 2D and 3D images of DTMs, attribute data stored in Paradox spatial databases, and vegetation patterns at landscape scales (Church et al, 1994b).

GIS and CAD

CAD technology has allowed the manual construction of landscape models to be taken from the drawing board to the computer terminal, through the creation and arrangement of 3D solid fill models. The customary methods of manually designing objects in CAD do not apply well to the modelling of change in regional landscapes, which contain not only many individual landscape features, but also many types of individual features. Therefore, conventional CAD based object building methods are as inappropriate as GIS for landscape Visualisations.

The solution put forward by Mayall and Hall (1994) is an amalgamation of the relative strengths of GIS and CAD technology in an integrated 3D landscape modelling process. By joining the database handling capabilities of GIS with the 3D modelling capabilities of CAD, changes in the visual landscape can be visualised relatively quickly and effectively.

References

Angehrn, A.A. and Luthi, H-J. (1990) Intelligent Decision Support Systems: A Visual Interactive Approach. Interfaces, 20, 17-28.

Baldwin, J., Fisher, P., Wood, J. and Langford, M. (1996) Modelling Environmental Cognition of the View with GIS. Third International Conference/Workshop on Integrating GIS and Environmental Modeling, Santa Fe, New Mexico, January 1996.

Berger, P., Meysembourg, P., Sales, J. and Johnston, C. (1996) Towards a virtual reality interface for landscape Visualisation. Third International Conference/Workshop on Integrating GIS and Environmental Modeling, Santa Fe, New Mexico, January 1996.

Bishop, I.D. and Hull, R.B. (1991) Integrating technologies for visual resource management. Journal of Environmental Management, 32, 295-312.

Bishop, I.D. and Karadagli, C. (1996) Combining GIS based environmental modeling and Visualisation: another window in the modeling process. Third International Conference/Workshop on Integrating GIS and Environmental Modeling, Santa Fe, New Mexico, January 1996.

Camara, A.S., Mano, A.P., Martinho, M.G., Nunes, J.F., Lopes, T.C. and Cabeleira, A. (1986) An economic-ecological model for regional land-use planning. Ecological Modelling, 31, 293-302.

Church, R.L., Murray, A.T. and Barber, K. (1994a) Designing a Heierarchical Planning Model for USDA Forest Service Planning. Presented at Sixth Symposium on Systems Analysis and Management Decisions in Forestry, Pacific Grove, California.

Church, R.L., Murray, A.T., Figueroa, M.A., Ager, A.A., McGaughey, R.J. and Merzenich, J. (1994b) Artificial Landscape Visualisation of Ecosystem Management Plans. Submitted to AI Applications, June 1994.

Host, G.E., Rauscher, H.M. and Schmoldt, D. (1992) SYLVATICA: an integrated framework for forest landscape simulation. Landscape and Urban Planning, 21, 281-284.

Imaging Systems Laboratory (1995) SmartForest: An Interactive Forest Data Modelling and Visualisation Tool. Department of Landscape Architecture, University of Illinois at Urbana-Champaign.

Kellomaki, S. and Pukkala, T. (1989) Forest Landscape: A Method of Amenity Evaluation based on Computer Simulation. Landscape and Urban Planning, 18, 117-125.

Kent, B., Bare, B.B., Field, R.C. and Bradley, G.A. (1991) Natural Resource Land Management Planning using Large-Scale Linear Programs: the USDA Forest Service experience with FORPLAN. Operations Research, 39, 13 - 27.

Kessell, S.R. (1990) An Australian geographic information and modelling system for natural area management. International Journal of Geographical Information Systems, 4, 333-362.

Kuiper, J., Ayers, A., Johnson, R. and Tolbert-Smith, M. (1996) Efficient data exchange: integrating a vector GIS with an object-oriented, 3-D Visualisation system. Third International Conference/Workshop on Integrating GIS and Environmental Modeling, Santa Fe, New Mexico, January 1996.

Liu, J. (1993) ECOLECON: An ECOLogical-ECONomic model for species conservation in complex forest ecosystems. Ecological Modelling, 70, 63-87.

Liu, J., Cubbage, F.W. and Pulliam, H.R. (1994) Ecological and economic effects of forest landscape structure and rotation length: simulation studies using ECOLECON. Ecological Economics, 10, 249-263.

Mayall, K. and Hall, G.B. (1994) Information Systems and 3-D Modeling in Landscape Visualisation. In Urban and Regional Information Systems Association Annual Conference Proceedings, Vol 1, 796-804.

Perestrello de Vasconcelos, M.J., Zeigler, B.P. and Graham, L.A. (1993) Modeling multi-scale ecological processes under the discrete event systems paradigm. Landscape Ecology, 8, 273-286.

Sol, H.G. (1983) Processes and tools for decision support: Inferences for future developments. In Processes and Tools for Decision Support, (ed. Sol, H.G.), North Holland, Amsterdam, the Netherlands, pp 1-6.

van Voris, P., Millard, W.D., Thomas, J. and Urban, D. (1993) TERRA-Vision - the integration of scientific analysis into the decision making process. International Journal of Geographic Information Systems, 7, 143-164.