非参数方法在干旱频率分析中的应用

Nonparametric Approach for Drought Frequency Analysis

运用非参数核密度估计方法研究干旱发生的联合概率、条件概率和重现期等干旱特征,以宁夏盐池站的月降水为例,应用单变量核概率密度函数估计干旱历时D的边缘分布,进行参数方法和非参数方法的拟合效果比较。在此基础上,采用双变量核概率密度函数估计方法构建了历时D与烈度S、历时D与峰值P的两变量联合概率分布,并计算了联合分布的重现期、条件概率与条件重现期。结果表明:与参数方法相比,非参数核密度估计方法能够描述干旱历时D、烈度S和峰值P两两之间的联合分布,是研究干旱频率的另一种新途径。

This paper aims to study drought characteristics such as joint（and conditional） probability distribution and return period of drought variables using nonparametric kernel estimator method.Based on monthly precipitation at the Yanchi station,univariate kernel density function has been used to estimate the marginal distribution of drought duration,and fitting results of parametric and nonparametric methods were also compared.On the basis of this,kernel estimator was employed to construct bivariate joint probability distribution of drought duration and severity as well as drought duration and peak,and in addition,return periods and probabilities of joint（and conditional） distribution were calculated.The study indicates that compared with parametric methods,nonparametric kernel approach is suitable to build the bivariate distribution of drought duration,severity and peak,which is a new alternative method for drought frequency analysis.