Copula函数在广西洪涝灾害的降水概率预测中的应用

为了研究洪涝的重要致灾源因子降水的概率,利用广西从1970年-2012年的降水数据,以年降水均值（X）和年降水极值均值（Y）为研究变量,通过Copula函数构建其联合分布.经过OLS,AIC拟合优度评价,采用Copula函数中拟合效果较好的Gumbel-Hougaard Copula函数建立边缘分布为Pearson-Ⅲ型的两变量的联合分布.随后,进行相应的重现期计算,计算结果表明,在联合重现期下的两变量降水设计值比单变量设计值和同现重现期下的设计值都要高,故采用联合重现期下的联合设计值作为防洪标准会更加安全.最后,对比降水数据与灾情数据,研究降水重现期与洪涝灾情的关系.结果表明,降水的重现期越长,洪涝灾害也越严重.

Copulas Application in Precipitation Probability Forecast of Flood Disasters in Guangxi

Based on the precipitation dates of Guangxi from 1970 to 2012, precipitation probability, which is the important source factor to cause flood, was studied. The average values of annual precipitation （X） and annual precipitation extremum （Y） were regarded as researching variables, and their joint distribution by Copula function was established. The result showed that the design values of two variables under the joint return period are greater than the design values of single variable and co-occurrence return period. It means that it is safer to adopt the design values of two variables under the joint return period. The longer return period of precipitation implies that flood damage is more serious.