Consisting of a set of rules and user-supplied data which interact through an inference engine, an expert or knowledge-based system is able to derive or deduce new facts or data from existing facts and conditions. In the past four years, expert-system shells have become more widely available, allowing users to define the data base and rule base without using artificial intelligence programming languages. Many water resources DSS designers have thus seen expert systems as a powerful complement to numerical and spatial analysis tools.
Fedra [1993a] reviewed the use of expert systems in water resources and identified three types of applications: purely knowledge driven systems, expert systems components in an intelligent front end, and fully embedded expert systems. Of these, intelligent front ends have been the most common. In general, they assist the user in selecting the appropriate numerical model or technique, specifying input parameter values, and interpreting model output. Lam and Swayne [1993] presented such an approach to the integration of virtually any computer technology useful to water resources planners. The role of the expert system was to provide an intelligent interface between the model and data, as well as descriptive dialogue between the user and machine. Palmer and Spence [1992] used the programming language PROLOG and natural language to represent knowledge about water resources management. Their purpose was to help users who were unfamiliar with formal database management or computer programming to access hydrologic and other data. Other examples of expert systems as intelligent front ends were given by Simonovic [1991] for open channel flow measurement, Simonovic [1992] for reservoir management, Bender and Simonovic [1994] for long-range water supply forecast modeling, and Crowe and Mutch [1994] for assessing groundwater contamination potential. Besides aiding decision makers, each of these systems could be useful in training inexperienced analysts.
Related to intelligent front ends, which help the user select the appropriate model, are knowledge-based model support systems, which help the user build the appropriate model. Rozenblit and Jankowski [1991] proposed such an approach to simulation modeling of natural systems and emphasized the following advantages: modular model specification facilities; high degree of model reusability; and support for model selection and coupling. Jankowski [1992] discussed the role of these model support systems within a DSS, illustrating their ability to create new models quickly and easily, to integrate model building blocks, and to interrelate models through a database.
Fully embedded or hybrid expert systems are typically problem-oriented rather than methodology-oriented. Whereas intelligent front ends enhance the use of models in a DSS, fully embedded expert systems enhance model results. An example of such a system was given by McKinney et al. [1993]. They developed an expert-GIS for long-term regional water-resources planning in which the expert system aided the user in analyzing the social, legal, and political aspects of the problem. Hidden from the user, the rules invoked by the expert system eliminated planning options which did not meet certain qualitative constraints supplied by the user. Other embedded expert systems were developed for irrigation systems planning ( Nir [1991]) and for crop planning during droughts ( Raman et al. [1992]). Each of these agricultural expert systems was used to enhance simulation and optimization results.
Finally, purely knowledge driven systems, based solely on a qualitative, causal understanding of how things work, have not been so common in water resources. Arnold and Rouve [1991] illustrated the concept and application potential of this type of system for water resources protection. Examples of other potential uses are operational control of a wastewater treatment plant and hazardous waste site assessment ( Fedra [1993a]).