基于Copula函数的三维非连续序列洪水频率分析  

Trivariate Discontinuous Flood Frequency Analysis Using Copulas

作　　者：崔冬梅[1] 陆宝宏[1,2] 张翰文 赵超[1] 徐琨[1] 毛豆 邓山[1] 吴星鑫 

机构地区：[1]河海大学水文水资源学院,南京210098 [2]河海大学水文水资源与水利工程科学国家重点实验室,南京210098

出　　处：《中国农村水利水电》2015年第4期105-108,112,共5页China Rural Water and Hydropower

基　　金：国家自然科学基金项目(NSFC-50979023);水利部公益科研项目(201201026)

摘　　要：考虑洪峰与洪量之间的相依性和历史洪水资料,建立基于Copula函数的三维非连续序列洪水频率分析模型,探讨洪水三变量的联合概率分布和条件概率分布。以龙滩水库设计洪水为例,用Copula函数构造边缘分布为P?Ⅲ分布的洪水联合分布,分别用适线法和极大似然法估计P?Ⅲ分布和Copula函数的参数,通过K-S法、AIC信息准则、OLS准则和Genest-Rivest方法验证模型的合理性。结果表明模型合理可信,联合设计值均接近并略大于单变量分析方法的成果,且重现期越长越接近;确定条件下的重现期小于超过该条件时的重现期。模型为多变量非连续序列洪水频率分析计算提供了一种新方法。Historical information and dependent relationship among peak and volume are explained .A Copula-based discontinuous flood frequency analysis model is proposed with parameter estimated by three-dimensional maximum likelihood method ,and the three vari‐ables joint probability distribution and the condition probability distribution are analyzed .The design flood of Longtan Reservoir is selected as a case study .The results show that the proposal is reasonable and feasible ,the designed values of copulas are higher than that based on the conventional univariate method ,the return period under fixed condition was less than the one when the conditions were above the value .The model provides a new approach for the multivariate discontinuous flood frequency analysis .

关 键 词：COPULA函数 历史资料 设计洪水 极大似然法 联合分布 

分 类 号：P333[天文地球—水文科学]