Flood Regionalization of Mazandaran Province's Watersheds by Using Fuzzy Clustering Algorithm

One of techniques for estimation of flood quantiles in ungauged watersheds is regional frequency analysis. In the common techniques of regionalization، homogeneous hydrological regions are created based on watershed characteristics such as physiographic characteristics of watershed، geographic location and flood statistics at each station. However، most watersheds only partially resemble other watersheds in a region. Therefore one watershed cannot be completely assigned to one group or another. The fuzzy clustering algorithm allows a watershed to have partial or distributed memberships of all the regions (groups) identified. In this paper، performance of fuzzy clustering algorithm has been evaluated for regionalization of Mazandaran province''s watersheds، and some solutions have been represented to improve homogeneity of the regions that have been clustered by this algorithm and are not statistically homogeneous. In addition، several indices for performance evaluation of fuzzy clustering algorithm have been presented to determine the number of optimum clusters. Then، the Zdist statistic for the region''s stations was calculated by FORTRAN programming language for the 3-parameter distributions and the best distribution was introduced for each region. The results showed that Mazandaran province has been composed of three hydrological homogeneous regions and generalized logistic distribution with different parameters was selected for each of the homogeneous regions.