A fuzzy hybrid clustering method for identifying hydrologic homogeneous regions

Abstract

Identification of hydrologic homogeneous regions (HHR) facilitates prioritization of watershed management measures. In this study, a new methodology involving a combination of self-organizing features maps (SOFM) method and fuzzy C-means algorithm (FCM), designated as SOMFCM, is presented to identify HHRs. The case study region is Walnut Gulch Experimental Watershed (WGEW) located in Arizona. The input data consisted of a number of factors that influence runoff generation processes, including ten surface features as well as various rainfall values corresponding to 25, 50, and 100 years return periods. Factor analysis (FA) was applied for the selection of effective surface features along with rainfall value, used in the clustering algorithm. Validation procedure indicated that the best clustering scenario was achieved through merging three layers including TPI (topographic position index), CN (curve number), and P50 (50-year rainfall). The optimum number of clusters turned out to be six while the fuzzification parameter became 1.6. The presented methodology may be proposed as a simple approach for identifying HHRs.