关键词: 混合Copula函数, 洪水频率, 函数构成, 粒子群算法, 遗传算法, 重现期

Abstract: Flood frequency calculation is a basis for determining the scale of water-related projects, and it can be improved using a mixed Copula function that better describes the distribution characteristics of multidimensional flood random variables than a single Copula function. This paper constructs two types of mixed Copula functions from single Copula functions, demonstrating their features via the flood data from the Madawask basin of Quebec, Canada. Their weight coefficients are solved using the particle swarm optimization and genetic algorithm, and we optimize them in calculation of flood frequencies. The results show that the largest mean squared errors of the mixed types are 0.000567 and 0.000352, both smaller than those of the single type, and a better effect can be achieved by constructing a mixed function from the rotating Copula functions. When the combination of large variables is calculated, the joint return period estimated by a mixed Copula function is shorter than that of the single function, while the co-occurrence return period is longer.

Key words: mixed Copula function, flood frequency, function structure, particle swarm optimization, genetic algorithm, return period