How to Design Multicriteria Spatial Decision Support Systems

Spatial decision support systems (SDSS) are generally employed in decision problems, in which spatial dimensions play a significant role. They have a satisfactory record in a wide range of problems and in various circumstances. SDSS Applications include land use planning, nuclear waste disposal facility location, water resource management, habitat site development, health care resource allocation, highway planning, and retail location selection problems. New application areas arise as the technology gains wider acceptance. Today’s spatial decision support systems rely on a geographical information systems (GIS) component. Therefore, this paper explains the GIS technology and provides a framework that integrates it into decision support systems. It also describes the different techniques and approaches used in the design of spatial decision support systems. The theoretical and practical aspect of the design process is discussed and the illustrative examples are presented. Introduction Spatial multicriteria decision problems involve a set of geographically defined alternatives from which choices are made based on a given set of evaluation criteria (Jankowski, 1995; Malczewski, 1996). This paper studies multicriteria spatial decision support systems (MC-SDSS) and describes the underlying technologies and theoretical background that are necessary to develop such systems. The first section defines the MC-SDSS and identifies GIS as a key technology employed in such systems. Multicriteria Spatial Decision Support Systems A multicriteria spatial decision support system (MCSDSS) provides a tool to combine and transform spatial and non-spatial data into a resultant decision. Since MC-SDSS can be viewed as a part of a broader field of decision support systems (DSS), this section begins by explaining the basic characteristics of DSS, applicable to MC-SDSS, before focusing on the specific elements of MC-SDSS. DSS can be considered as a subset of computerbased information systems and it includes a variety of familiar systems such as office automation systems, transaction processing systems, management information systems, and management support systems. DSS are often referred as a type of management support system and have two main elements: human decision makers and computer systems. A DSS can be defined as a computer-based, human-computer decision-making system that supports decision makers to solve problems with varying degrees of structure (e.g. from non-structured or illstructured to very structured) by using data and analytical models. Since the main advantage of using DSS is facilitation of decision processes, DSS focus on effectiveness rather than efficiency in decision. Unlike expert systems, which mimic human decision makers in making repetitive decisions in a narrow domain, DSS do not replace decision makers but rather support them in solving different decision problems, which are often not well-structured. Decision problems that involve geographical data are referred to as geographical or spatial decision problems. Spatial decision support systems are a subgroup of DSS and are generally employed in decision problems, in which spatial dimensions play a significant role. Since almost every modern spatial decision support system relies on a geographical information systems (GIS) component, it is important to develop a sound understanding of GIS concepts in order to design successful spatial DSS. In fact, most experts see GIS as a form of DSS. For example, Cowen (1988) defined GIS "as a decision support system involving the integration of spatially referenced data in a problem solving environment". GIS were originally developed to automate cartography and map production. Early implementations in the 1970s were not able to deliver the benefits of GIS systems effectively, because of the limited computational power available at that time. In the mid 1980s, software integration and more powerful computers were combined to create packaged GIS offering usable digital mapping functionality. Utility companies and mapping agencies were among the first users of GIS systems (Raper, 2000). A GIS system is composed of a geographical data base, an input/output process, a data analysis method, and a user interface. The terms spatial or geographical data are often used interchangeably and describe objects based on two types of characteristics: location and attributes. Location-related data provides spatial position of the object and attribute data includes all other properties of the objects. The data in GIS systems are usually organized by thematic maps or sets of data. Although GIS provide extensive spatial analysis and data visualization power to their users, such systems offer a limited capacity for tackling complex, ill-defined, spatial decision problems. Multicriteria Spatial decision support systems (MC-SDSS) augment the problem-solving capacity of GIS by integrating analytical models into decision process (Densham and Goodchild, 1989). There has been considerable growth in research, development and application of MC-SDSS in the last decade (NCGIA, 1990, 1996). The integration of analytical models and GIS can be achieved by employing different techniques. Nyerges (1992) developed a conceptual framework for the coupling of analytical models with GIS. He designates four levels with increasing intensity of coupling: Level 1: Isolated Application. The GIS module and the analytical models are located in different hardware platforms and the data transfer is performed off-line by ASCII files. The efficiency of this approach is very limited. Level 2: Loose Coupling. The GIS module and the analytical models are run on the same computer or on a number of connected computers. The data transfer is performed online by using ASCII text files. The formatting of the ASCII files is conducted manually by the user. Level 3: Tight coupling. In this scheme, data interfacing is standardized and does not require human intervention. On the other hand, the analytical models and GIS are still separate modules. Level 4: Full Integration. The system performs like a single program and the data interfacing is seamless. A dedicated database management system serves both GIS and the analytical model. The efficiency of the coupling increases as the designed scheme approaches to a full integration. On the other hand, required development time and resources also increases with the level of integration. The next section focuses on the spatial multicriteria decision analysis and provides a general architecture. Developing a Design Methodology for the MC-DSS A number of frameworks for designing MC-SDSS have been proposed including Diamond and Wright (1988), Carver (1991), Eastman et al. (1993), and Jankowski et al. (1997). Despite differences in GIS capabilities and multicriteria decision making (MCDM) techniques, the generic framework contains three major components: a user interface, MCDM models (includes tools for generating value structure, preference modeling, and multiattribute decision rules), and geographical data analysis and management capabilities. Malczewski (1999) provides a framework for spatial multicriteria decision analysis. According to this framework, shown in Exhibit 1, there are three main phases of MC-SDSS: intelligence, design, and choice. This three-phase process is the most widely accepted generalization of the spatial multicriteria decision making and is described in more detail in the following paragraphs. The intelligence phase includes the definition of the problem, constraints, and evaluation criteria. Once the decision problem is identified, spatial multicriteria analysis focuses on the set of evaluation criteria. This step involves specifying a comprehensive set of objectives that reflects all concerns relevant to the decision problem and the measures (or attributes) for achieving those objectives. Attributes that address the system objectives are converted to criterion maps, which are twodimensional representations of evaluation criteria in GIS database. The set of criterion maps represents a particular decision situation related to a particular segment of the real-world geographical system. Evaluation criteria form the basis for the decision matrix and each criterion can have different importance to the decision makers. Consequently, information about the relative importance of the criteria is necessary. One way of deriving criterion maps is to use value/utility approaches that assign to each criterion a weight that indicates the criterion importance relative to the other criteria under consideration. The design phase integrates decision makers’ preferences into the system. Decision alternatives, decision rules, and the decision matrix structure the design phase to accommodate the required functionality. Decision alternatives are the alternative courses of action among which the decision maker must choose. The process of generating decision alternatives should be based on the value-structure and be related to the set of evaluation criteria. The decision matrix compiles feasible decision alternatives while decision rules provide a procedure that allows for ordering alternatives. The values provide a basis for the integration of the criterion map layers and judgments in order to provide an overall assessment of the alternatives. Specifically, the decision rules order the decision space via a one-to-one or one-to-many relationship of outcomes to decision alternatives. Exhibit 1. General architecture of spatial multicriteria decision analysis The choice phase includes sensitivity analysis in the optimization. In the final step of this phase, recommendations are formed based on the findings of the sensitivity analysis. The framework for this step is an iterative process, which allows the readjustment of the decision rules and evaluation criteria based on the outcomes of the recommendations and sensitivity analysis (Malczewski, 1999). The three stages of multicriteria spatial decision making do not necessarily follow a linear path from intelligence, to design and to choice and they can be adapted to different decision situations. The next section studies the scope and range of applications in MC-SDSS. Applications of Multicriteria Spatial Decision Support Systems Examination of MC-SDSS applications provides an understanding of the usefulness of this method and this section presents a number of representative examples. They are selected to cover different fields and address the significant aspects of MC-SDSS. The examples also illustrate typical approaches used in MC-SDSS. MC-SDSS have been developed for a variety of problems, including land use planning (Diamond and Wright, 1988) (Thill Jean-Claude, Xiaobai Yao, 1999) (MacDonald and Faber, 1999), nuclear waste disposal facility location (Carver, 1996), water resource management (Bender and Simonvic, 1995), habitat site development (Jankowski et al., 1997), health care resource allocation (Jankowski and Ewart, 1996), land suitability analysis (Eastman et al., 1995; Fischer et al., 1996), highway planning (Morin, David M., 1997), and retail location selection problems (Arentze, Borgers and Timmermans, 1997). Exhibit 2 summarizes the findings of a literature search on MC-SDSS applications. Most of the MC-SDSS, listed on Table 1 focus on static spatial decision analysis without providing a temporal analysis capability. The necessity to integrate simulation models into spatial decision support system was articulated by Burrough et al. (1988). In many MC-SDSS, temporal decision analysis necessitates the development of simulation models. However, the integration of simulation models poses significant challenges and successful applications are scarce (Colombo, 1992). Ideal integrated systems should feature seamless transitions between simulation modules and other system components and they should also offer user-friendly environments to analyze different scenarios effectively. As illustrated by the applications, given above, MC-SDSS have a satisfactory record in a wide range of problems and in various circumstances. Sensitivity Analysis Evaluation Criteria Recommendation Alternatives Decision Matrix Decision Rules Decision-Maker's Preferences Problem Definition