Abductive Reasoning in Environmental Decision Support Systems

Decision trees and rule-based system including variants based on propositional and fuzzy logic have been the method of choice in many applications of environmental decision support systems. Reasons are the ease of use, the capability of representing uncertainty, and the fast computation of results at runtime when using decision trees or other similar means for knowledge representation. Unfortunately there are drawbacks related with these modeling paradigms. For example, the cause-effect relationships between quantities are not captured correctly. The resulting model is well appropriated for a certain purpose but can hardly be re-used. Moreover, maintaining the knowledge base can be an intricate task. In this paper we focus on the problems related with decision trees in the context of environmental decision support systems using an example from the domain. We further present abductive reasoning as an alternative for modeling and show how these technique can be easily implemented using existing techniques.