A knowledge-based approach for classifying lake water chemistry

Abstract

Knowledge-based systems are computer models that facilitate reasoning such that human experience and expertise can be represented and made available to non-specialists. In this paper we describe the application of a knowledge-engineering methodology, using the NetWeaver™ software, to the problem of lakewater acid-base chemistry assessment. We present, and document with examples, the structure, arguments, and criteria values of a knowledge-based decision support system for classifying lakes in five acid-sensitive regions of the United States. We also discuss the significance of this software tool for federal land managers in the management of aquatic resources in national parks, national wildlife refuges, and wilderness areas to protect against water quality degradation associated with atmospheric deposition of sulfur and nitrogen. The Lake Chemistry knowledge bases have undergone repeated testing by members of a lake chemistry domain expert panel. There is agreement among the panel that these regional models provide accurate classifications of lakewater chemistries. The graphical and executable rendering of knowledge bases within NetWeaver™ greatly facilitates the knowledge engineering process, as it permits the inclusion of the domain expert(s) in the knowledge representation process and hence encourages greater participation in the design of the final knowledge-based model. In addition, the inclusion of fuzzy arguments, against which data values can be compared, greatly reduces the potential for combinatorial explosion that often occurs in expert systems that rely on categorical data interpretation, while at the same time providing a robust description of complex systems. It is our expectation that adoption of this approach, and others like it, will stimulate further development of knowledge-based systems for agriculture, natural resource management, and other complex decision support arenas.