A ranking method for regionalization of watersheds

Abstract

Identification of groups of watersheds with similar flood generation mechanisms and the same hydrological responses is an important step of regional flood frequency analysis which is called regionalization of watersheds. A group of watersheds with similar flood generation mechanisms and the same hydrological responses is called a homogeneous region. Theoretically, the homogeneity of regions is a requirement to fit the regional frequency distribution and estimate flood quantiles. However, it may be observed in some real-world cases that a regionalization with a more appropriate homogeneity condition results in less accurate estimates in comparison with a regionalization with lesszappropriate homogeneity. In this study, a ranking method is proposed to compare various regionalizations and select more suitable ones for regional flood frequency analysis. The proposed method is used to identify a suitable regionalization for regional flood frequency analysis by considering the homogeneity, quantile estimation accuracy, and average region size. The method was applied to Karun-e-bozorg basin in the southwest of Iran as a case study. The results show that the method has an acceptable performance in identifying a regionalization that provides suitable homogeneity, quantile estimation accuracy, and average region size. Also, the study results indicate that the ranking based on the homogeneity ratio is relatively compatible with the ranking based on the quantile estimation accuracy. In addition, according to the results, the most appropriate regionalization of 41 watersheds in the study area includes 3 regions provided by the combination of the clustering algorithms average-linkage and K-means.