传统水文频率分析方法的前提条件是水文极值系列需满足一致性要求，然而由于气候变化和人类活动的影响，使得水文资料的非一致性问题越来越突出，导致现行水文频率分析方法的应用受到挑战。根据重建的淮河流域1470年－1999年共530年夏季面平均降雨量长系列数据，进行分析，通过假设检验表明系列在1534、1724和1923年发生跳跃性变异，因而将系列划分为4个子平稳系列；采用现行频率分析方法估计每个子系列的概率分布函数，再根据混合分布概念由各子系列的概率分布构建一个综合的概率分布函数，作为非一致性降雨系列总体分布的估计。在此基础上，计算了给定设计标准下的设计值，并与基于大样本（530年）计算的经验设计值进行了对比，结果表明混合分布模型对观测系列具有较好的拟合效果。

The traditional hydrological frequency analysis (THFA) method has been widely used to estimate the design flood for a given return period. It relies on the assumption that the series should be stationary. However, climate change and human activities caused the series to lose their stationarity, and thus limit the application of the THFA method. In this paper, we studied the 530-year summer-precipitation series covering the period from 1470 to 1999. At the 0.01 significant level, it was found that the series had three change points: 1534, 1724, and 1923. Therefore, the whole series was divided into four sub-sample series. We estimated the probability distribution function (PDF) of each sub-sample using the THFA method. Then we obtained the mixed probability distribution function (MPDF) by weighting and synthesizing the four PDFs related to the four sub-samples. The MPDF was regarded as the PDF of the whole series and was used to estimate the design precipitation for a given return period. Besides, the 530-year large precipitation series was also used to estimate the empirical value of the design precipitation. Then we compared the observation-sample-based estimations and MPDF-based estimations of the design precipitation. The results showed that the MPDF fit the series well.

国家重点研发计划“水资源高效开发利用”重点专项（2016YFC0402706、2016YFC0402707、2016YFC0402709）