汉江流域上游已建成投运诸多水库及引调水工程,显著改变了洪水的时空分布,造成了洪水资料序列的非一致性。把洪水系列资料延长至2023年,开展非一致性洪水频率分析,以水库系数和调水工程系数作为协变量,建立时变P-III分布模型,定量估算丹江口水库运行期的设计洪水,并与原设计成果对比,结果表明：(1)所选协变量符合水文客观规律,均可反映上游水利工程调蓄对丹江口坝址洪水的影响；(2)丹江口水库运行期1000年一遇设计洪峰流量,3d、7d和15d洪量较原设计成果,分别削减了37%、24%、25%和27%；(3)汉江上游水库调蓄对设计洪水的影响比调水工程更加显著。由于丹江口水库运行期设计洪量显著减少,在防洪标准不变的前提下可酌情提高水库的运行水位,这不仅有利于增发电量,还可提高水资源利用效率和水库蓄满率。

Danjiangkou Reservoir is the key hydraulic project for flood control in Han River basin, and the water source for the middle route of South-to-North Water Diversion Project. The design flood of Danjiangkou Reservoir is an important reference for reservoir operation and comprehensive benefits, which is overestimated due to the short-recorded data series and the uncertainty of historical floods in construction period. Besides, many reservoirs and water diversion projects have been built and put into operation in the upper Han River basin, which have significantly altered the spatiotemporal distribution of floods. The Mann Kendall test indicates a significant decrease for the annual maximum flood series from 1929 to 2023 with the statistical p-values all greater than 0.05, which means that the gauged flood data series is non-stationary. Thus, how to quantitatively estimate design flood of Danjiangkou in operation period has become an urgent scientific and technological challenge. The Generalized Additive Models for Location, Scale and Shape (GAMLSS) model has been widely applied and proved to be effective in non-stationary flood frequency analysis, which is based on the principle of time-varying moment method and assumes that the probability density parameters like the location, scale, and shape would follow certain mathematical relationship with corresponding covariates. In this study, we apply GAMLSS-based time-varying P-III distribution method for non-stationary flood frequency analysis for Danjiangkou Reservoir in operation period. The selection of covariates would significantly determine the effectiveness of time-varying model. The Reservoir Index (RI) is defined with the hydrological characteristics and reservoir storage capacity, which presents superiority in design flood investigation. Meanwhile, Diversion Index (DI) is first proposed to quantify the influence of water diversion projects with mathematical definition inspired by RI. The RI and DI are selected as covariates to construct the time-varying P-III distribution model, and the maximum likelihood estimation method are used to estimate P-III distribution parameters for the single covariate (RI) and the double covariates (RI&DI). Based on the annual maximum flood series with 441-year historical investigation period (1583-2023), the design flood for Danjiangkou Reservoir in operation period is estimated and compared with original design values. The main conclusions are summarized as follows: (1) The RI and DI are selected as covariates which can conform to the hydrological variation laws objectively. The fitting results of time-varying P-III distribution with both covariates can reflect the flood series decreasing trend caused by upstream hydraulic structures. However, there is no significant difference for the time-varying P-III distribution fitted with double covariates (RI&DI) and single covariate (RI). The influence mainly attributes to upstream reservoirs rather than water diversion projects, while the RI only slightly reduce about 1-4% design floods in Danjiangkou Reservoir. (2) The design flood of Danjiangkou in operation period has been significantly reduced. 1000-year design flood peak, 3-day, 7-day, and 15-day flood volumes are estimated as 40700 m3/s, 7.87 billion m3, 13.4 billion m3, and 19.3 billion m3 (compared to the original design values 64900 m3/s, 10.4 billion m3, 17.9 billion m3, and 26.3 billion m3), which have been reduced by 37%, 24%, 25%, and 27%, respectively, (3) The original design flood of Danjiangkou Reservoir was overestimated which resulted discard water in flood season and lower refill rate in dry season. Since the estimated design flood in reservoir operation period has significantly reduced, the reservoir operation water level could be high under the condition of flood prevention standard unchanged, which not only can generate more hydropower, but also can increase the water resources utilization rate as well as reservoir refill rate.

国家自然科学基金长江联合(U2340205)Funds: National Natural Science Found of China(U2340205)