Flood Indicators and Their Clustering Features in Wujiang River, South China

It is necessary to figure out the flood indicators and recognize the flood intensity clustering for understanding the flood from management perspective. In particular, flood events are indicated by their peak discharge, peak level, the maximum 24-hours volume, the maximum 72-hours volume, and the total flood volume. In this paper, to avoid the defects of single indicator in describing flood intensity, all these five indicators are calculated and their temporal variations are analyzed with the 53 year's flood records from 1955 to 2007 in Wujiang River. The fuzzy C-means algorithm (FCM) is used for flood intensity clustering to exhibit high internal (within-cluster) homogeneity of the flood and to detect the similar groups among observed floods. Trend analysis for the yearly maximum flood events from 1955 to 2007 shows that the peak flood discharge, total flood volume, and the maximum 72-hours volume have a continuous weak increase tendency that is marginally statistically significant. The flood peak level does not show a statistically significant trend at the 95% significant level. The maximum 24-hours volume exhibits an increasing trend at the 90% significant level. The clustering results show that the number of floods type I and type II varies from the 1950s to the 21st century and exhibits an overall increase trend. The chances of encountering floods type I and type II is greater than encountering flood type III. Floods are found to be getting more serious in Wujiang River after the 1990s.