基于Copula函数的水文频率分析的依赖性模拟.docVIP

  Modeling Dependence in Hydrological Frequency Analysis using Copula# XU Yueping, TONG Yangbin, XU Xiao* 5 10 15 20 25 30 35 40 (Institute of Hydrology and Water Resources Department of Civil Engineering, Zhejiang University, Hangzhou, 310058) Abstract: Although the importance of multivariate analysis has already been recognized, dependences among different variables are often ignored due to the complexity of problem. It is known that doing so may induce great modeling risk that potentially exist, and will jeopardize the successive hydrosystem analysis, design and management. In this paper, the copula approach is proposed to model the dependence among rainfall data of different durations. Several well-known copulas are chosen and compared, among which the Gaussian copula is found to be the best one. In the meanwhile, rainfall data is simulated using the copula approach and it is found out that the simulated rainfall data preserve quite well the statistical characteristics of actual data. Key words: Water resouces protection; Copula; Dependence modeling; Multivariate analysis; Puyang River 0 Introduction Hydrological frequency analysis, one of the most important components for hydrosystem analysis, design and management, often focuses on one single variable, such as flood peak. This is inappropriate since many hydrological processes involve several variables, among which is the rainfall process, usually characterized by rainfall depth (or intensity), duration and interval[1]. The separate modeling of individual variables will over- or under-estimate the risk related to hydrosystem design. One such risk is caused by the crossing problem resulting from modeling hydrological variables separately[2]. Therefore, multivariate distributions are highly recommended to appropriately model complex hydrological processes[3.4]. The uses of copula are currently very popular in many fields and its substantial advantages such as flexibility and easy estimation have been well-kn