基于Copula的汾河上游水文干旱频率分析

Copula-based Analysis of Hydrologic Drought Frequency in the Upper Reaches of Fenhe River

运用游程理论对汾河上游4个水文站的月径流资料进行干旱识别;选择4种分布函数拟合干旱特征变量,通过Kolmogorov-Smirnov检验法优选单变量边缘分布;利用Copula函数建立干旱特征变量的二维联合分布,计算重现期。结果表明,汾河上游干旱特征变量的最佳边缘分布是对数正态分布;汾河上游大多数干旱重现期小于5年,且干旱历时与烈度的单变量重现期与二维重现期较为接近;干旱历时与烈度峰值、烈度与烈度峰值的单变量重现期与二维重现期差距较大。

Monthly runoff data taking from four hydrometric stations in the upper reaches of Fenhe River are applied to identify the droughts by using Run Theory. The drought characteristic variables are fitted by four distribution functions, and the Kolmogorov-Smirnov method is employed to evaluate the goodness-of-fit of univariate marginal distributions. The Copula functions are employed to construct two-dimensional joint distribution models of drought characteristic variables, and then their return periods are determined. The results show that： （a） the lognormal distribution is the candidate marginal distribution function with highest goodness-of-fit in fitting drought characteristic variables in the upper reaches of Fenhe River; （b） the return periods of most of the droughts are less than 5 years, and the univariate return periods and the two-dimensional return periods of drought duration and severity are close to each other; and （c） there are large differences between the univariate return periods and the two-dimensional return periods of drought duration and peak intensity, as well as the drought severity and peak intensity.