Non-stationary design flood estimation of Three Gorges Reservoir in operation period considering historical information

At-site natural annual maximum flood series and the Pearson Type III (P- III) distribution curve fitting method are used to estimate flood quantiles and determine the reservoir characteristic water levels in the planning and design stage of water conservancy and hydropower projects. However, the stationary assumption of flood data series has been destroyed because of the construction and operation of upstream reservoirs. In this paper, a time-varying moment model for non -stationary flood frequency analysis was built, and the Reservoir Index(RI) of key large-scale reservoirs in the upper reaches of the Yangtze River was calculated. Based on the monotonically increasing property of the distribution function of the continuous flood variable, the time-varying P- HI curve fitting method was proposed to fit the Q-Q plots. The non - stationary design floods during operation period of the Three Gorges Reservoir(TGR) were inferred and compared with those of the maximum likelihood (ML) method. The results show that: (1) The RI effectively reflects the non - stationarity of flood series, serving as covariates in the time-varying P- HI model ; (2) The time-varying P -III curve fitting method yields a smaller AIC value than the ML method, indicating a better fitness for historical flood series ; (3) Compared with the original results, The nonstationary design flood peak, along with the 3-day and 7-day flood volume of TGR are decreased by about 18%, while the 15-day and 30-day design flood volumes are decreased by about 14%. These results demonstrate that the regulation of the cascade reservoirs in the upper Yangtze River directly affects the flood control capacity and characteristic water level of TGR during operation period. © 2024 International Research and Training Center on Erosion and Sedimentation and China Water and Power Press. All rights reserved.