考虑历史洪水的超定量(Peak-over-Threshold,POT)洪水频率分析方法能使洪水信息的利用最大化,并有效提高洪水频率分析的合理性。本文以武江流域犁市(二)站为例,以泊松分布为超定量年发生次数分布,用广义Pareto(GP)分布拟合POT样本,线性矩法(L-M)估计不连续POT样本的分布参数,探讨了历史洪水在POT洪水频率分析中的应用。结果表明,武江选取门限值为1079m3/s能兼顾分布稳定性和样本独立性;对连续POT样本和不连续POT样本的洪水频率分析对比得出,对历史洪水的考虑有效改善了POT方法对大洪水的拟合,由不连续POT样本所得的GP频率分布曲线对"2006·07"洪水的重现期估计为501年,而基于连续POT样本的频率分析估计结果为330年,表明超定量洪水频率分析时必须考虑历史洪水;考虑历史洪水条件下,POT方法的拟合优度略优于年最大值(Annual Maximum Series,AMS)方法。
Flood frequency analysis of peak-over-threshold ( POT ) series with historical floods can maxi mize the use of flood information , and effectively improve the accuracy of flood frequency analysis. In this paper , taking Lishi ( 2 ) station in the Wujiang River Basin as an example , use of historical information in POT flood frequency analysis is considered. The POT samples are fitted by the generalized Pareto ( GP ) dis tribution with a Poisson model for arrival , using L-moment ( L-M ) method to estimate distribution parame ters. The result shows that choosing 1 079m 3 /s as the threshold can maximize the stability of the estimation for POT distribution parameters and meet the hypothesis of independence. Comparison between POT method with and without historical floods shows that the application of historical floods does improve the fitting of large flood in POT method. Return period of " 2006·07 " flood estimated by POT method with historical floods is 501 years , while the estimate result of POT method without historical floods is 330 years , which seems that it ’ s necessary to consider historical floods in POT modeling. The goodness of fit of POT method is a little better than flood frequency analysis of annual maximum series ( AMS ), while both using histori cal information.