Applying Fuzzy based Inductive Reasoning to Analyze Qualitatively the Dynamic Behavior of An Ecological System

In the last decade a variety of methodologies for representing and evaluating knowledge qualitatively has been developed, particularly within the field of Artificial Intelligence. Qualitative reasoning methodologies represent an alternative to quantitative modeling approaches, if the knowledge about the system of interest is imprecise or incomplete, as it is often the case when dealing with ecological systems. As most of the methodologies have not outgrown toy examples, it re1 Institute of Artificial Intelligence, Department of Computer Science, University of Ulm, D-89069 Ulm, Germany mains challenging to apply those methodologies to real world applications. In Biosphere 2, a closed ecological system, the level of O2 has dropped and the CO2 level has risen continuously during its closure between 1991 and 1993. The mechanisms of carbon cycles have been subject to multiple research efforts, and are therefore formulated as general rules in principle. However, the specific situation within Biosphere 2, a closed ecosystem, might influence the validity of these rules. Thus, the structure of the carbon cycle in Biosphere 2 is not well known, yet abundant data exist on some of the important fluxes and pools. Whereas deductive, quantitative as well as qualitative, methodologies need knowledge about the structure of the system to derive the behavior of the system, the fuzzy-based inductive reasoning methodology FIR derives inductively the behavior model by analyzing time series. The derived behavior model comprises cases and information how to retrieve prototypical cases that can be adapted to the given situation. Thus, FIR combines one-shot inductive and incremental case-based reasoning techniques in analyzing and forecasting dynamic systems.