基于Vine Copula模型的淮河干支流洪水遭遇分析  

Analysis of Floods Encounter of the Mainstream and Tributaries of Huaihe River Based on Vine Copula Function

作　　者：徐鹏程 张志浪 刘春明 方红远[2] 王磊之 XU Peng-cheng;ZHANG Zhi-lang;LIU Chun-ming;FANG Hong-yuan;WANG Lei-zhi(Key Laboratory of Flood and Drought Disaster Defense,the Minstry of Water Resources,Nanjing 210029,Jiangsu Province,China;College of Hydraulic Science and Engineering,Yangzhou University,Yangzhou 225009,Jiangsu Province,China;Jiangyin Yangtze River Dike and Sluice Station Management Center,Wuxi 214437,Jiangsu Province,China)

机构地区：[1]水利部水旱灾害防御重点实验室,江苏南京210029 [2]扬州大学水利科学与工程学院,江苏扬州225009 [3]江阴市江堤闸站管理中心,江苏无锡214437

出　　处：《中国农村水利水电》2024年第12期12-18,26,共8页China Rural Water and Hydropower

基　　金：水利部水旱灾害防御重点实验室开放基金(KYFB202307260031);国家自然科学基金(42301026);江苏省青年基金(BK20220589)。

摘　　要：随着极端气候的加剧,洪水灾害越来越频繁,洪水遭遇事件会加剧洪水对下游地区的危害。采用淮河王家坝站和蚌埠站、史河蒋家集站以及颍河阜阳站共4个站点的从1959到2016年的58年实测日径流量数据,使用年日最大流量法提取出汛期中的洪现时间序列以及洪峰序列。以Von Mises分布构建年最大洪水的洪现时间的边缘分布,以对数正态分布、皮尔逊Ⅲ型分布和广义极值模型分别构建洪峰的边缘分布。通过极大似然法估计各个分布的参数,按照赤池信息量准则(AIC)选取最优模型,并使用KS检验确定边缘分布是否合格。使用Vine Copula分别构建多站点洪现时间的联合分布和洪峰的联合分布,对淮河干支流两两遭遇,三站点遭遇以及当下游发生洪水时,上游发生不同组合洪水的条件概率遭遇等多种情况的遭遇风险进行了定量化的分析。研究结果表明:Von Mises分布尤其是单峰Von Mises分布可以很好的拟合淮河流域的洪现时间;淮河干支流洪水遭遇中站点两两遭遇,以及多站点遭遇的最可能日期均在7月15日左右;干支流两两遭遇中王家坝站点与蒋家集站点遭遇概率最高,阜阳站点与蒋家集站点遭遇概率最小;当下游蚌埠站发生50年一遇洪水时,上游三站点均发生10年重现期以上洪水概率为0.076。研究进一步拓展了洪水遭遇风险的计算方法,对淮河流域防洪减灾具有重要意义。As the severity of extreme weather escalates,the frequency of flood disasters is on the rise and the flood coincidence events may in⁃tensify the adverse effects on downstream regions.This study leverages 58 years of observed daily runoff data from 1959 to 2016 of four stations:Wangjiaba Station and Bengbu Station on the Huai River,Jiangjiaji Station on the Shi River,and Fuyang Station on the Ying River.The study extracts the flood occurrence time series and peak flow series during the flood season using the annual maximum flow method.The marginal distribution of flood occurrence time for the annual maximum floods is constructed using the Von Mises distribution,while the mar⁃ginal distributions of peaks are constructed using the logarithmic normal distribution,Pearson type III distribution,and generalized extreme value model,respectively.The parameters of each distribution are estimated using maximum likelihood estimation,the optimal model is se⁃lected according to the Akaike Information Criterion,and the Kolmogorov-Smirnov test is used to determine whether the marginal distribu⁃tions are qualified.Additionally,the Vine Copula is used to construct the joint distribution of flood occurrence time and peak flow for multi⁃ple stations.Quantitative analysis of encounter risks under various conditions such as encounters between two rivers,encounters between three stations,and the conditional probability of various combinations of flooding events occurring upstream when flooding occurs down⁃stream.The results show that the Von Mises distribution,especially the unimodal Von Mises distribution,can well fit the flood occurrence time in the Huaihe River Basin.For encounters between tributaries of the Huaihe River,the most probable date for pairwise encounters and multiple station encounters is around July 15th.Among pairwise encounters of tributaries,the highest encounter probability is between Wangjiaba Station and Jiangjiaji Station,and the lowest encounter probability is between Fuyang Station and Jiangjiaji S

关 键 词：Vine Copula 多变量频率分析 洪水遭遇 淮河 

分 类 号：TV11[水利工程—水文学及水资源]