Nonstationary Flood Frequency Analysis In Changing Climate And Flood Return Period Study

In recent years,because of the aggravation of climate change and human activities,more and more extreme events have happened frequently,such as floods disaster and drought phenomenon activities.The uncertainty existing in the changing climate is more and more significant,and the analysis of hydrologic series which is based on traditional stationary assumption is questioned,and the application of nonstationarity in hydrological field is focused by many researchers nowadays.In this paper Wangkuai reservoir is chosen as the study area,and the design floods were recalculated under nonstationary condition(1)The test of nonstationarity in flood time series of Wangkuai reservoir.By using Mann-Kendall test and Pettitt test,the nonstationarity in flood series was identified at a 5% significant level.There is a decreased trend in the annual maximum flood series,and change point is in 1979.(2)Using GAMLSS model to conduct nonstationary flood frequency analysis for the data of 1956~2007 in Wangkuai reservoir.The results of nonstationary frequency analysis in annual maximum flood series proved that there was a decreased trend in annual maximum flood series from 1955 to 2007,and the change point is in 1979.By using GAMLSS model,Gamma,Gumbel,Weibull and Log-Normal distribution were chosen to fit annual maximum flood series,and at the 95% confidence interval,the fitness of Log-Normal distribution is the best.There are two concepts of flood return period under nonstationarity: Expected Waiting Time(EWT)and Expected Numbers of Events(ENE).with the EWT concepts of return period,the design floods are smaller than those under stationary condition.Because the trend or/and change points will not be existing in flood series for a long time,nonstationary return period can not represent long term flood disaster.(3)The stochastic modeling technique was used to interpolate the missing flood data between the historical flood and the observed data to develop the reliability of flood data.Choosing PDO index as exogenous variable,a nonlinear ARX model was built using annual maximum flood series.Then the nonstationary design floods were calculated considering the historical floods,and the comparison was made with the traditional design floods.