Since many water resources problems require more than just the analysis of spatial data, several authors have offered guidelines for the development of spatial decision support systems---the merging of geographic information systems and water resources models in a DSS for problems with a spatial dimension. Armstrong [1992] provided an abstract overview of data design and geometric representations needed for a spatial water resources DSS. Walsh [1993] presented more practical research challenges to be faced in the development of such DSS. He emphasized the need for an open architecture, interdisciplinary collaboration in prototype development, and the development of models which are more spatially aware. Djokic [1993] made a strong case for the use of commercially available software within a spatial decision support system shell. He argued that the one-time effort of developing interfaces between the software components would require much less effort than customizing existing or writing new software. Similarly, Frysinger et al. [1993] stressed the need for an open architecture to facilitate the integration of GIS and water resources and environmental models in a DSS. Kilgore et al. [1994], however, stated that the use of commercial GIS in DSSs would impede integration and even the sharing of data because the GIS and models do not reside in the same environment.
Others have evaluated the use of GIS in DSSs with respect to more specific applications. Fürst et al. [1993] recommended that GIS components be used for supporting model based scenario analysis in DSSs for groundwater management. They identified the advantages of efficient spatial data analysis and display, as well as the disadvantages of poor three-dimensional data representation, poor performance of interactive graphical tasks, and the sometimes tedious procedures required to interface GIS tools with groundwater models.
Leipnik et al. [1993] provided an excellent overview of implementing GIS in water resources planning and management. The process described involves the initial decision to use a GIS, the selection of a system, data-base development and product generation, and installation and training. Kaden [1993] focused on some of the problems of GIS implementation, identifying some common pitfalls in the integration of software components, the acquisition of adequate data, the management of time-varying data, and increasing levels of complexity which may constrain user-friendliness. These problems have led researchers to investigate expert system ( Morse [1992], Hu [1993]) and object-oriented ( Berrill and Moon [1992]) approaches to the implementation of GIS in a DSS. Such approaches considered the GIS to be the primary analytical tool of the DSS, with the integration of other technologies supporting its use as intelligent front ends. This is similar to the approaches of some, but not all, of the research discussed in the next section.