Search

As it is a critical process of watershed management, classification always faces challenges of inefficiency in handling complexity and uncertainty. This study attempts to fill this gap by developing an integrated adaptive resonance theory mapping system consisting of a two-stage adaptive resonance theory mapping (TSAM) approach and an integrated rule-based fuzzy adaptive resonance theory mapping (IRFAM) approach. To demonstrate their feasibility and efficiency, TSAM and IRFAM were compared with conventional adaptive resonance theory mapping (ARTMap) in two case studies in the Deer River watershed in Manitoba, Canada, which were classifications of watershed subbasins and types of land-cover to support hydrological modeling. Among the three approaches, IRFAM performed best in effectively processing the classification for input patterns with a high level of uncertainty and a wide range of variations, although it required predefined criteria. TSAM performed reasonably well by generating criteria for supervised classification on the basis of the internal relationship of the original data, indicating its advantage in handling an insufficient data situation with a low demand for subjective judgment. Consequently, the two developed approaches can be complementary and improve classification efficiency and robustness in dealing with systematic complexity and uncertainty and supporting watershed hydrological modeling. [ABSTRACT FROM AUTHOR]