基于广义Pareto分布的洪水频率分析  

Analysis of flood frequency based on generalized Pareto distribution

作　　者：陈子燊[1] 刘曾美[1,2] 路剑飞[1] 

机构地区：[1]中山大学水资源与环境系,广州510275 [2]华南理工大学水利水电工程系,广州510640

出　　处：《水力发电学报》2013年第2期68-73,83,共7页Journal of Hydroelectric Engineering

基　　金：2009年广东省水利创新研究项目(2009-41);国家自然科学重点基金资助项目(50839005)

摘　　要：基于广义Pareto分布理论,深入探讨洪水序列超阈值分布的阈值选择和超定量样本的适用性检验。使用分块抽样方法和经验平均超过函数选择洪水阈值,检验不同洪水超定量的泊松分布、GPD模型的Anderson-Darling检验和超定量样本分布的拟合优度指标。以广东省西江流域高要水文站测量的日流量序列为例,将最优GPD、GEV和P-Ⅲ型模型推算的洪水重现水平做了对比分析,获得以下结论:①超限量洪水GPD模型的确定是一动态择优过程,需要根据多个检验指标加以确定;②洪水极值分布有可能属于具有上限的短尾型分布;③经验平均超过函数图可作为选择洪水GPD阈值的重要参考,但最佳阈值应采用多种指标综合确定;④构建的洪水GPD模型普遍优于GEV和P-Ⅲ型。This paper presents an in-depth study on the selection of flood threshold and the fitness test of excess value sample for a peaks-over-threshold(POT) model of flood series by applying the generalized Pareto distribution(GPD) theory.In a case study of this POT model,the daily runoff series measured at the Gaoyao hydrologic station on Xi River in Guangdong province was analyzed.A method of empirical mean excess function chart was used for the threshold selections,and the Poisson distribution and the fitness indicators of excess values were tested for flood peaks above different high levels,and a contrastive analysis was made to evaluate the flood return periods calculated by the three flood models of optimal GPD,generalized extreme value distribution and Pearson type Ⅲ distribution.The main findings are as follows.The use of GPD flood model needs a dynamic optimization of the flood threshold through calculation of multiple indicators;the extreme value distributions of Gaoyao floods could be of a short-tail type.The optimal threshold of Gaoyao floods should be determined by using multiple indices,while its value calculated by the chart method above can only be taken as a valuable reference.The good-fitness test indicates that the GPD model is generally better than the GEV and Pearson type Ⅲ.

关 键 词：水文学 洪水阈值 广义PARETO分布 经验平均超过函数图 Anderson-Darling检验 参数估计 

分 类 号：TV133[水利工程—水力学及河流动力学]