Flood Season Division with an Improved Fuzzy C-mean Clustering Method in the Taihu Lake Basin in China

Abstract

The division of the flood season in a basin is a multi-dimension time series clustering problem with unknown clustering numbers. So the clustering method is required to have the capability of dealing with multi-dimensions, time series and clustering validity. A dynamic fuzzy C-means clustering method (simplified as DFCCM) with clustering validity function based time series was developed in the paper. The real coding based accelerating genetic algorithm (simplified as RAGA) is used for solving DFCCM, which is very effective and can overcome the initial value sensitivity from the regular iterative optimal method generally used for the fuzzy C-means clustering. The newly developed method was applied in the Taihu lake basin in China for the division of the basin flood season, and 11 schemes were designed for removing the uncertainty from the initial value of flood season, which could exert influence on the results of the flood seasons division. Application results show that the method is very effective and the division conclusion of flood season in the Taihu Lake Basin is also reasonable.