Changing properties of extreme flood in Xiliugou River Basin, China

Based on the annual maximum flood discharge sequence of the Longtouguai hydrological station in the Xiliugou River Basin from 1960 to 2015, the statistical probability characteristics of extreme flood discharge were analyzed by using generalized extreme value (GEV) distribution, generalized Pareto (GP) distribution, Pearson type III (P-III) distribution, general logistic (GLO), and Burr distribution. The parameters were estimated by the L-moments method, and the optimal distributions were selected by some of goodness-of-fit method. The results showed that the annual maximum flood discharge series decreased significantly with an abrupt change point around 2005. From the L-moments and goodness-of-fit test method, Burr distribution can better fit the annual maximum flood discharge before and after the abrupt change. Compared with the sequence before the change, the return period becomes larger and the probability of severe flood is smaller. The designed maximum flood peak discharge for the same return period of the entire sequence is smaller than that of the sequence before the change.