This study tested the feasibility of producing an automated expert system for Soil Taxonomy to identify soil order from stored data by building an expert system prototype. Soil Taxonomy rules for the Histosol, Spodosol, Andisol, and Oxisol orders were translated into decision tree format. Seventy independent properties were stored in tabular format for each pedon. Heuristic knowledge (expert rules) was added to the decision trees to query a minimum data set, with 13 field description properties required to contain data for each soil horizon, 20 default values, and three estimated values from lookup tables. The prototype expert system was developed using an object-oriented expert system shell. Twenty-seven subsections were named in the rules to identify the Histosol, Spodosol, Andisol, and Oxisol soil orders. Sixty-seven objects, 70 independent properties, and 135 calculated properties were needed to define these subsections and their properties. The tested prototype quickly and correctly identified the diagnostic horizons, nonspatial differentiae, and the soil order, proving the feasibility of developing an expert system for Soil Taxonomy using existing computer programs and programming methods. We recommend improvements in policy and procedure for recording field description data and development of the expert rules to add dynamic links to outside models and software and incorporate fuzzy logic. The project should be continued to improve the prototype interface and data output features and to complete an expert system to add the remaining soil orders for Soil Taxonomy.
All Science Journal Classification (ASJC) codes
- Soil Science