多变量水文分析计算方法的比较  

Comparative study of multivariate hydrological frequency analysis methods

作　　者：闫宝伟[1] 郭生练[1] 郭靖[1] 张娜[1] 

机构地区：[1]武汉大学水资源与水电工程科学国家重点实验室,湖北武汉430072

出　　处：《武汉大学学报（工学版）》2009年第1期10-15,共6页Engineering Journal of Wuhan University

基　　金：国家自然科学基金项目(编号:50679063);国际科技合作项目(编号:2005DFA20520);水利部科研专项项目(编号:200701015)

摘　　要：简述了多维联合分布在水文分析计算领域的研究进展,结合实例比较分析了多变量水文分析计算3种代表方法,即正态变换法、非参数方法和Copula函数法的适用性和合理性.通过Monte Carlo试验,比较研究这3种方法的统计性能.实例分析及统计试验表明:正态变换方法在其反变换过程中可能会出现部分信息失真,导致计算结果不可靠;非参数法外延的有效性差,不适合作水文事件的尾部分析;Copula函数法的计算过程是可逆的,推求的结果相对可靠,无偏性和有效性等统计性能也最好,可以在多变量水文计算领域推广应用.The multivariate hydrological frequency analysis methods are reviewed briefly. The applicability and reasonability of three representative methods, namely bivariate normalization （BNOR）, bivariate nonparametric （BNP） and Copula function, are comparatively studied. The statistical properties of these methods are further studied by the Monte Carlo method. Application and statistical experiment results indicate that the BNOR method may result in distortions during its inverse transformation which causes the results unreliable. The BNP method seems to be inapplicable to the tail frequency analysis for its limited extrapolation. Calculation process of Copula function method is reversible and so its results are reliable. It performs best with least bias and high efficiency, and is suggested to be used in multivariate hydrological frequency analysis.

关 键 词：多变量 正态变换 非参数方法 COPULA函数 MONTE Carlo 

分 类 号：P333[天文地球—水文科学]