The characteristics of Soil Taxonomy are analyzed relative to various techniques for developing expert systems. Special emphasis is placed on computer program features that allow for more consistent application of classification systems and make them more user-friendly and understandable. We studied the functional logic and query processes employed by Soil Taxonomy to identify soil individuals and compared the methods with those used in other natural object classification systems. Numerical and classical identification methods and program features found in recent computer programs were evaluated for use with Soil Taxonomy. The keys in Soil Taxonomy are purely phenetic in nature and single-access in approach. In the absence of rule- and value confidence-weighting factors, the rules must be encoded without sequence modification to preserve the decision logic. Decisions in Soil Taxonomy query a large, often incomplete, and sometimes faulty data set, requiring error-checking of data and the addition of expert rules to the encoded decisions to prevent indecision. Soil Taxonomy rules check within the soil individual for the presence or absence of spatial and nonspatial differentiae, specific property values, or other qualifications. Soil Taxonomy is suitable as the subject of an object-oriented expert system, and planning has begun on development of an automated prototype for the Histosol, Andisol, Spodosol, and Oxisol soil orders. Expert system features coupled with additional models and algorithms can be used to improve the use and user-friendliness of Soil Taxonomy.
All Science Journal Classification (ASJC) codes
- Soil Science