中文摘要：

基于淮河中上游主干流两侧13个水文站点1959-2008年逐日径流量的年最大值序列和超门限峰值序列,采用Mann-Kendall趋势检验、21种分布函数和Pearson III分布,分析淮河流域极端径流的强度和频率特征。结果表明：1极端径流强度6个站点呈增加趋势,7个呈减少趋势;极端径流发生频率8个站点呈增加趋势,5个呈减少趋势。2极端径流在径流值上的频率分布,年最大值序列总体较服从Weibull分布,而非普遍采用的广义极值分布;超门限峰值序列总体仍较服从广义帕累托分布。3基于超门限峰值序列和广义帕累托分布估算的50年一遇的极端径流值精度最高,大部分地区误差率低于0.2,精度优于工程标准Pearson III分布。气候变化背景下,极端径流频率特征发生变化,流域上游地区工程标准可能需要调整。受函数形态影响,极值序列最优拟合函数估算精度不如广义极值分布和广义帕累托分布。

英文摘要：

In order to study the intensity and frequency characteristics of extreme runoff over the Huaihe River Basin, this paper used 21 distribution functions to fit the AM（Annual Maximum） and POT（Peak Over Threshold） extreme runoff series at 13 hydrological stations which were located in the coverage area of trunk and tributary of the middle and upper Huaihe River Basin. In addition, M- K test was used to study the variation trend of intensity and frequency of extreme runoff. The results are as follows：（1） Intensity of extreme runoff at 6stations has a positive trend while that at the other 7 stations has a negative trend; temporal frequency of extreme runoff at 8 stations has a positive trend while that at the other 5 stations has a negative trend.（2） In AM series, the frequency distribution of runoff value obeys the Weibull function rather than the widely used GEV function on the whole, while in POT series,the frequency distribution of runoff value still obeys the GP function on the whole.（3）Compared with other 12 stations, the frequency distribution of extreme runoff value of Bantai station is the most intensive with a peak frequency of 0.27, and the frequency distribution of extreme runoff value of Wujiadu station is the most homogeneous with a peak frequency of0.0475.（4） The extreme runoff value is high in the middle basin and Mengwa region in the study area.（5） The error rate of extreme runoff estimated based on GP function is the lowest,whose error rates of most regions is lower than 0.2. In particular, the estimation accuracy of GP function is higher than that of Pearson III function, so that the engineering standard in the upper reaches of the basin may need to be adjusted. Moreover, affected by the function form, the estimation accuracy of best-fit function is lower than that of the GP and GEV functions.