ArcView was customized by using Avenue to establish a link with ArcInfo Grid.
Spatial Decision Support Systems (SDSS) have evolved out the need to provide geographic information systems (GIS) users with the proper tools for which to resolve complex geoanalytical problems. Given these tools, the user is able to make well informed decisions based on the problem at hand. The issue of site planning, including analysis and selection, is a task that is well suited for such a system. A SDSS is currently under development that will provide the casual, desktop GIS user with a user-friendly site evaluation and selection system. The system was developed using AML programming to customize the user interface, and incorporates both the vector and GRID based modeling functions of ArcInfo. ArcView 2.0 serves as the front end for accessing these site selection tools. Topics discussed include needs assessment, system design, graphical user interfaces, incorporation of the spatial model, and cartographic display.
Poor locational decisions in facility siting may result from a variety of factors, the most common of which are uninformed land use planners (O'Hare, et al., 1983). However, recent advances in geographic information systems (GIS) provide decision makers with efficient tools with which to organize and structure the spatial decision making process (Densham, 1992). The research presented here includes the design, prototyping, evaluation, refinement, and implementation of a user-friendly site selection system for use by research scientists and land-use planners at the Savannah River Site. This site selection module will play an integral role in the advancement of spatial decision support systems (SDSS) development at the Savannah River Site.
The Savannah River Site (SRS) lies along the border of the states of South Carolina and Georgia and is one of fourteen Department of Energy (DOE) nuclear industrial sites located within the United States. Construction of the site began in the 1950s with its primary mission being the refinement and production of Tritium and Plutonium-239, radioactive materials for use in nuclear weapons (WSRC, 1990). While the handling of nuclear materials for national defense purposes remains a key initiative of the site, SRS is shifting towards a new focus of waste management and environmental restoration and protection (WSRC, 1993). Most recently, SRS and other DOE facilities have been presented with the task of siting a wide variety of facilities within their boundaries. These include the siting of landfills, new reactors, waste disposal sites, power facilities, and at a much grander scale, facilities such as the Superconducting Super Collider (Siderelis, 1992).
The immediate goals of the environmental monitoring program at SRS include the identification and quantification of the effect of site activities on the environment using state of the art techniques (WSRC, 1993). Currently, SRS researchers and scientists are applying remote sensing and geographic information systems techniques to a variety of site related tasks that are spatial in nature (Jensen, 1986). Among the most current and well established applications of geographic information systems (GIS) are complex forms of spatial analysis such as facility planning, site selection, and land use planning (Tomlin and Johnston, 1988; Tomlinson, 1987). While the use of GIS and automated mapping sciences have been used in previous DOE site planning investigations (WSRC, 1992a; WSRC, 1992b; Siderelis, 1992; Jensen, 1986), hands-on access to the technology remains mostly limited to a small group of trained and technically capable users.
Over the past two years, the Environmental Impact Data Analysis and Retrieval System was developed as a cooperative effort between the University of South Carolina and Westinghouse Savannah River Company. The goals of the project were to integrate into a centralized computer database a diverse array of non-standard digital geographic data layers, airborne video, and bibliographic information for use within a multimedia geographic information viewing system (Bresnahan, et al., 1994a). The system was developed in an effort to provide SRS researchers with the necessary tools to increase efficiency in their everyday geographic decision making. Throughout development of the electronic atlas, various data and media formats were integrated into the site- wide system. An integral part of the system is the provision of a user friendly geographic data browsing system that is standard (i.e., operates and looks the same) across all platforms (Macintosh, UNIX, OS/2, Windows) and can seamlessly access the central data bank using native information transfer protocols (Bresnahan and Cowen, et al., 1994a; Bresnahan and Cowen et al., 1994b). Bibliographic information query and search may be accessed either spatially or aspatially using a document search system (Cowen and Jensen, et al., 1995). In addition, airborne digital video is directly incorporated into the system by means of a geographical tag or "hot link", allowing spatial referencing and point-and-click launching of digital movies.
The first phase of development of the Environmental Impact Data Analysis and Retrieval System was completed in the Fall of 1994. This phase focused on the initial design and implementation of digital data integration and the spatial browsing system. The next phase of the project is concerned with actual data analysis, in particular the design and implementation of a user-friendly site selection system. As initially indicated, key features of the proposed system include ease of use and flexibility in terms of being able to handle a wide variety of siting tasks. By examining both past and present siting problems at SRS, a generic site selection model is currently under development. Of the more recent siting tasks within SRS, two stand out as excellent examples of site evaluation procedures.
The siting of the New Sanitary Landfill and the Replacement Power Facility at the Savannah River Site both employ similar methodology. This methodology includes a decision making process that embodies the objectives of all parties involved, including knowledgeable subject experts. Both site selection processes involved the implementation of several exclusionary criteria that specifically eliminate certain types of geographic features based on regulations or mandates and then are assigned a score based on a ranking and weighting scheme (WSRC, 1992a; WSRC, 1992b).
The primary goal of the proposed system was to design and implement a generic site selection system that is fully compatible with the Environmental Impact Data Analysis and Retrieval System. Current locational decision making processes at SRS do not take full advantage of the spatio-analytic power that GIS has to offer. It is hypothesized that these complex spatial problems may be solved more efficiently by incorporating the analytic modeling capabilities of the application specific spatial decision support system (SDSS). Spatial decision support systems have evolved out the need for high level planners to make efficient and well informed decisions based on complex spatial tasks (Densham, 1992). This decision making process is often ill-structured in nature, allowing knowledge based techniques and heuristics to be especially applicable toward such systems (Cowen and Ehler, 1994; Padgett, 1993; Peterson, 1993; Elmes, 1991).
The concept of spatial decision support systems (SDSS) represents an effort to address complex spatial problem solving and assist spatial decision making. Densham (1992) argues that by providing the user with a flexible problem solving environment, the user is able to increase their awareness and understanding of the problem task, as well as refine his or her knowledge of undesirable solutions. The components of a "true" SDSS as defined by Densham include the integration of a geographic database management system with analytical modeling capabilities, a visualization component or graphical user interface (GUI), and the decision making knowledge of domain experts (Densham, 1992). In addition the key aspect that separates SDSS from GIS are geographic information analysis (GIA) functions. These analysis components include: 1) support of analytical modeling; 2) appropriate spatial data to support the model; 3) flexible graphical (mapping) and tabular output; and 4) incorporation of flexible decision making processes. Lastly, Densham provides a framework for developing spatial decision support systems and provides a theoretical design architecture.
Armstrong and Densham (1990) discuss two groups of decision making approaches that may be incorporated into SDSS: programming techniques and heuristic methods. Programming techniques tend to be computationally intensive but always yield an optimal solution. Heuristic techniques yield sub- optimal solutions, however are able to provide recommendations more efficiently by means of suggesting a range of solution alternatives (Armstrong and Densham, 1990). The authors also discuss the most commonly used databases and their applicability toward SDSS. The rectangular data model, or geographic matrix, was found to be most appropriate when only one level of spatial aggregation is to be considered. Finally, the authors conclude that a newly developed hybrid data model is most suitable to spatial decision support systems, which is based on the entity-category-relationship (ECR) approach and an extended network model (Armstrong and Densham, 1990).
Armstrong, et al. (1993) describe a SDSS that allows the analysis of school redistricting problems by education administrators. The system was designed around a geographic information system, population projection models, and a locational modeling routine. The role of GIS is to manage spatial data and produce reports and maps for use within the model. In effect, the system serves as a redistricting assistant, allowing the user to interact with the SDSS.
Peterson (1993) discusses how the problem of real estate investment can be facilitated through the use of spatial decision support systems. The real estate investment problem is an ill structured task, making it appropriate for SDSS application. Through the identification of the most important spatial tasks, the tracking of minor but important details involved in analysis, and organization of the problem space, SDSS allows the user to make well informed decisions (Peterson, 1993).
A spatial decision support system for coastal wetland permitting has been developed by Ji and Johnston (1994). The study involved customizing ArcInfo GIS to allow resource managers to interact with permit sites and plan future permit activities. The GUI that was designed includes windows, text displays, on-line help, and functional icons. Analytical models containing knowledge based rules were implemented pertaining to rules and regulations of wetlands permitting. The present condition of the system has yet to include mapping and report capabilities, however these functions are currently under consideration.
The Savannah River Site covers approximately 300 square miles of federally owned land and is located along the border of South Carolina and Georgia, including portions of Aiken, Allendale, and Barnwell counties. Sixteen USGS 1:24000 quadrangles contain the boundary of SRS, which serves as the areal extent of analysis for the proposed research. In addition to data currently contained within the Environmental Impact Data Analysis and Retrieval System, other geographic information was integrated into the system for inclusion in the site selection module. Existing data includes USGS Digital Line Graphs (DLG), digitally scanned aerial photography, multi-spectral and panchromatic SPOT imagery, wetlands and soils data, radiological contamination, DLG derived digital elevation models (DEM), and land cover. Additional data included water table information, depth to groundwater, subsurface geology, and the location of threatened and endangered species.
II. Computer Hardware and SoftwareComputer hardware and software decisions were based on the selection of state of the art technology and the status quo at the Savannah River Site and the University of South Carolina. ArcInfo GIS Ver. 7.0 and ArcView Ver. 2.0 from Environmental Systems Research Institute (Esri) of Redlands, California were selected as the geographic information system environments for project development. This decision was based upon the features found in both systems such as their support of a high level programming language (Arc Macro Language and Avenue), ability to be customized, ability to integrate current data sets found within the Environmental Impact Data Analysis and Retrieval System, multi-platform support, and reputation as state of the art technologies and industry standards. A Sparc 5 UNIX workstation from Sun Microsystems was acquired as the primary hardware for use in developing the site selection system. Additional hardware that was used in this research includes a desktop image scanner, digitizing table, 1/4" and DAT tape drives, and various printing and plotting devices including a Hewlett-Packard Model 650 color plotter.
III. The Data ModelWithin ArcInfo, spatial data including both locational data and attribute information is stored as a coverage for vector based data and a grid for cell or raster based data. Currently, data pertaining to the Savannah River Site resides in both formats. The properties of each data model must be taken into consideration before their incorporation into the site selection system. While a full discussion of spatial data models lies outside the scope of this paper, it is important to consider a primary goal of the site selection system -- performance. Users require quick and efficient results so that decisions may be made in a timely manner. As reported by Dangermond (1990), the speed of overlay and buffer operations within a raster based GIS greatly outperform their vector based counterparts. It is for this reason that the raster data model and cell based data processing methods were primarily utilized for this research. The level of spatial resolution of the cell based model was 5 meters, a standard for raster data within the Environmental Impact Data Analysis and Retrieval System.
IV. User NeedsFollowing preliminary discussions between WSRC technical representatives and the project team at the University of South Carolina, a document of user requirements was drafted. Among the many needs that were discussed, some of the most important include the ability to support the user with timely and accurate decisions, ease of use, the ability to provide the casual user with powerful geo-analytical tools, and visualization and reporting capabilities. In addition, the system had to both flexible and general enough to address a variety of site location projects.
V. Location Model DevelopmentMethods of designing spatial models for site planning were first discussed over 30 years ago before the advent of automated geographic information systems. In his seminal paper McHarg (1969) mapped thematic site criteria onto mylar transparencies and, when superimposed, was able to differentiate between acceptable and unacceptable zones. The methods that were developed were applied to a variety of social, economic, and environmental problems. In effect each mylar layer served as an input to the "maximum benefit - minimum cost" model (McHarg, 1969).
Based on previous site location research that was conducted between the University and the Westinghouse Savannah River Company, a generic site selection model was developed. This model is based on the same methodology that was outlined in previous siting efforts at SRS. Exclusionary criteria are incorporated into the model to specifically eliminate certain types of geographic features (WSRC, 1992a; WSRC, 1992b). Examples of exclusionary criteria include the location of wetlands, 100 year flood plain, and sensitive areas such as threatened and endangered species. The resulting potential site locations are then assigned a score based on their suitability to the remaining non-exclusionary, or inclusionary factors (WSRC, 1992a; WSRC, 1992b). A site's individual score or rating is calculated by summing criteria ratings found at each potential site location.
VI. System DesignUser interface design and incorporation of the analytical site selection model was done by customizing ArcView Avenue scripts and ArcInfo AML programming. The user accesses the site selection system from ArcView 2.0, the graphical interface of the Environmental Impact Data Analysis and Retrieval System (Figure 1 [157 Kb]). A link is established between ArcView 2.0 and ArcInfo via Avenue scripting, allowing the user to access the full geo-processing functions of ArcInfo. Within ArcInfo, customization of the main user interface was accomplished by designing a series of pull down menus, "widgets" (interactive graphical objects), and map and text based displays. Next, the user sets the resolution of cell-based analysis (coarse or fine resolution, which determines processing speed and spatial accuracy), enters locational criteria as it pertains to a particular siting task by checking theme widgets on or off (Figure 2 [31 Kb]) and applies ranks and weights for each theme. This location model is then run against the raster based GIS data producing a series of potential site locations. The user may display any of the data layers used in the analysis including the final sites, save the current settings of the model to an ASCII file for future modifications or overnight processing, or return to ArcView to view the selected sites, produce customized output, or to further browse or query the data (Figure 3 [60 Kb]). The analytical processing within the ArcInfo remains totally transparent to the user, requiring no interaction with ArcInfo's text-based command interface. This transparency is a common feature found within Spatial Decision Support Systems (Densham, 1992).
A differentiation between qualitative and quantitative geographic information system evaluations is made by Goodchild and Rizzo (1987). The authors discuss that a qualitative evaluation is based on the system's ability to satisfactorily perform the required functions as set forth by initial user requirements and the evaluation team. A quantitative evaluation addresses the ability of the system to perform the task at hand within the constraints of personnel working time, CPU processing speeds, and data storage limits.
The proposed SDSS will be evaluated on-site by WSRC personnel following a demonstration of the system's features and capabilities. This evaluation will primarily consist of user feedback and criticism and will be mostly qualitative in nature. In addition, a more rigorous quantitative evaluation will be proposed to WSRC technical staff. System performance will be measured by statistical comparison between hypothetical site selection results from the SDSS and those results derived from traditional manual site selection techniques (i.e., from human experts). It is hypothesized that as more criteria are introduced into the model, the accuracy of the decisions made by human site selection experts decreases while the accuracy of the SDSS remains relatively static.
The proposed study will contribute to geographic research in several ways. Spatial knowledge in relation to decision making processes of site planning will be organized, modeled, and formalized, bringing a greater understanding of the site selection process to decision makers. This will yield well informed and efficient locational decisions. Additionally, a methodology for quantitative evaluation of SDSS will be outlined and applied, perhaps sparking further studies into this under-researched topic.
Armstrong, M.P., and P.J. Densham, 1990. Database Organization Strategies for Spatial Decision Support Systems. International Journal of Geographical Information Systems, 4, No. 1, pp. 3-20.
Armstrong, M.P., P. Lolonis, and R. Honey, 1993. A Spatial Decision Support System for School Redistricting. URISA Journal, 5, pp. 40-51.
Bresnahan, P.J., D.J. Cowen, J. Jensen, H.E. Mackey, and G.B. Ehler, 1994a. Integration of Heterogeneous Spatial Data for Hazardous Waste Units at the Savannah River Site, South Carolina. Proceedings, GIS/LIS, Phoenix.
Bresnahan, P.J., D.J. Cowen, G.B. Ehler, E. King, W.L. Shirley, and T. White, 1994b. Using Geographical Data Browsers in a Networked Environment. Proceedings of the 6th International Symposium on Spatial Data Handling, Edinburgh, Vol. 2, pp. 921-932.
Cowen, D.J., J.R. Jensen, C. MacCharles, H. Mackey and W. Holliday, 1995. Incorporating Bibliographic Information into a Spatial Data Query System for the Savannah River Site. Technical Papers, ASPRS/ACSM, Charlotte.
Cowen, D.J., and G.B. Ehler, 1994. Incorporating Multiple Sources of Knowledge into a Spatial Decision Support System. Proceedings of the 6th International Symposium on Spatial Data Handling, Edinburgh, Vol. 2, pp. 921- 932.
Dangermond, J., 1990. A Classification of Software Components Commonly Used in Geographic Information Systems. In Introductory Readings in Geographic Information Systems edited by Peuquet and Marble (New York: Taylor and Francis), pp. 30-51.
Densham, P.J., 1992. Spatial Decision Support Systems. In Geographical Information Systems: Principles and Applications, edited by Maguire, Goodchild, and Rhind (Essex, England: Longmans), 1, pp. 403-412.
Elmes, G.A., and Twery, M.J., 1991. GypsES: A Knowledge Based Decision Support System for the Management of the Gypsy Moth. Technical Report, USDA Forest Service, Morgantown, WV.
Jensen, J.R., 1986. Solid and Hazardous Waste Disposal Site Selection Using Digital Geographic Information Systems Techniques. Science of the Total Environment, 56, pp. 265-276.
Ji, W., and J. Johnston, 1994. A GIS-Based Decision Support System for Wetland Permit Analysis. Proceedings, GIS/LIS, Phoenix, pp. 471-476.
McHarg, I.L., 1969. Design With Nature (Garden City, NY: Doubleday and Company, Inc.).
O'Hare, M., L. Bacaw, and D. Sanderson, 1983. Facility Siting and Public Opposition (New York: Van Nostrand Reinhold Company).
Padgett, D.A., 1993. Technological Methods for Improving Citizen Participation in Locally Unacceptable Land Use (LULU) Decision Making. Computers, Environment, and Urban Systems, 17, No. 6, pp. 513-520.
Peterson, K., 1993. Spatial Decision Support Systems for Real Estate Investment Analysis. International Journal of Geographical Information Systems, 7, No. 4, pp. 379-392.
Siderelis, 1992. Land Resource Information Systems. In Geographical Information Systems: Principles and Applications, edited by Maguire, Goodchild, and Rhind (Essex, England: Longmans), 2, pp. 261-273.
Tomlin, C.D., and K.M. Johnston, 1988. An Experiment in Land-Use Allocation with a Geographic Information System. Technical Papers, ACSM- ASPRS, St. Louis, Vol. 5, pp. 23-34.
Tomlinson, R.F., 1987. Current and Potential Uses of Geographic Information Systems: the North American Experience. International Journal of Geographical Information Systems, 1, pp. 203-218.
Westinghouse Savannah River Company (WSRC), 1993. Summary Pamphlet: Savannah River Site Environmental Report for 1992. WSRC-TR-93- 076.
Westinghouse Savannah River Company (WSRC), 1992a. Replacement Power Facility Site Selection Report. WSRC-RP-92-672.
Westinghouse Savannah River Company (WSRC), 1992b. Preliminary Site Selection Report for the New Sanitary Landfill at the Savannah River Site. WSRC-RP-92-1397.
Westinghouse Savannah River Company (WSRC), 1990. Savannah River Site: Aiken, South Carolina. M9009031.
