IRFAM: Integrated Rule-Based Fuzzy Adaptive Resonance Theory Mapping System for Watershed Modeling

Abstract

Watersheds are featured by a variety of hydrological, meteorological, and ecological characteristics. Complexity and uncertainty are usually two major challenges during watershed classification which is one of the key processes in hydrological modeling. This study aims to develop an integrated rule-based fuzzy adaptive resonance theory mapping (IRFAM) system, by incorporating fuzzification and rule-based operation to more efficiently handle the complexity and uncertainty. The developed system has been tested with a case study conducted in the Deer River watershed in Manitoba, Canada. The results are further compared with the ones generated by the conventional adaptive resonance theory mapping (ARTMap) method. All subbasins are classified by IRFAM while ARTMap leaves five subbasins unclassified. Furthermore, another nine subbasins in the juncture between classified groups from the ARTMap classification results are relocated by IRFAM. The IRFAM system can take advantage of fuzzy set theory to generate full criteria combinations to match the input patterns and use the rule-based operation to screen the matched patterns into the target groups. Therefore, the developed system can effectively process the classification for the input patterns with a high degree of uncertainty and wide range in variations. Furthermore, the IRFAM can effectively help resolve the traditional difficulty in criteria generation, which is always affected by uncertainty due to insufficient references and historical records, and complexity occurs due to multifeatures. The improvement of classification efficiency and robustness will be directly beneficial to hydrological modeling and related watershed management which rely on classification results.