Flood Feature Identification and Clustering in Wujiang River, South China

Wujiang River, one of the main branches of the Beijiang River in South China, is frequently suffered from flood disasters. Flood clustering becomes one of the critical sub-issues for realizing the different types of flood features. This paper attempts to put forward a flood clustering approach for flood feature identification which is important to the flood risk management and flood forecasting. In this paper we proposed the simulated annealing algorithm to solve the shortcomings of the Fuzzy C-Means which is generally used for flood clustering. The specific design of the clustering model was given. FCM and FCM-SA algorithms are used for clustering the flood features covering a period of 42 years from 1965 to 2006 in Wujiang River. The result showed that both FCM and FCM-SA can be successfully applied to cluster flood features, however, the FCM-SA has more advantages than the FCM. The flood clustering results can help us to identify very quickly the flood characteristics in the previously year. The clustering result showed that the flood situation of Wujiang River is much more serious.