Nonstationary Flood Frequency Analysis And Its Effect On Flood Control For Longmen Reservoir

Due to climate change and frequent human activities,The underlying surface of Longmen Reservoir Control Basin has undergone significant changes,resulting in the fact that the measured flood sequence does not meet the consistency assumption.The design results based on the traditional P-III distribution hydrological frequency analysis without considering the variation have certain risks in engineering design.This paper takes the Longmen Reservoir,the main stream of the Caohe River in the Daqing River Basin,as the research object,and studies the bivariate nonstationary flood frequency analysis and its effect on flood control for Longmen Reservoir.The main research aspects and results are as below:(1)The annual flood peak series(AMFP),annual maximum flood volume(1-day,3-day and 6-day AMF V)of Longmen Reservoir were used as the basic data.The results show that the hydrological alteration diagnosis system was used to obtain AMFP series of Longmen Reservoir without variation;AMFV series varied to different degrees varied to different degrees showed moderate variation,and the variation point was in 1979.(2)According to the variation diagnosis results,building CPD model and MD model analyzed and calculated the nonstationary flood frequency respectively.K-S test was used for fitting distribution test.OLS and AIC were used to choose optimal fitting model.The results reveal that:after fitting,MD model as the optimal distribution was employed to analyze the non-station flood time series.The results show that the non-consistent bivariate design flood results are greater than the design flood results for the univariate non-consistent MD distribution.Compared to the design flood of the traditional Pearson type Ⅲ(P-Ⅲ)distribution without considering the variance,the design flood of the univariate mixed distribution under all return periods.showed an overall trend of decrease.(3)Building the joint distribution of AMFP and 1-day AMFV series based on their respective marginal distributions by the Copula method.Based on the joint distribution,the joint return periods and the conditional probability were calculated.In addition,the designed flood hydrograph was derived from the amplification method.The results show that:Compared with the non-station univariate distribution,the design flood values of AMFP and AMFV of the bivariate distribution are larger,and the increased magnitudes of flood peak is 2.29%and 4.75%respectively,and the design values of 1-day AMFV series increased by 2.51%and 4.29%.(4)Based on the flood routing regulations,the design flood hydrograph derived from the mixed distribution and the joint diatribution were the inflow hydrographs of Longmen Reservoir,they were used to imolement flood routing and compare the flood routing results under various conditions.The results show that:(1)the results of flood regulation under the non-stationary mixed distribution are smaller than the stationary conditions by the conventional method;the maximum reservior water level with 2000 year return period deduced by the mixed distribution is decreased by 0.14m,and the maximum reservoir water level with 100 year return period decreased by 0.19m;the maximum discharge with the 2000 return period is decreased by 2.46%,the maximum discharge with the 100 return period is decreased by 1.80%.(2)the results of flood regulation under the non-stationary mixed distribution are smaller than those of the stationary conditions by the conventional method;the maximum reservior water level with 2000 year return period deduced by the joint distribution is decreased by 0.18m,and the maximum reservoir water level with 100 year return period decreased by 0.37m;the maximum discharge with the 2000 return period is decreased by 3.36%,the maximum discharge with the 100 return period is decreased by 14.25%.