基于Copula函数的广东北江流域极端降水指标概率分布特征

Probability Distribution Behaviors of Extreme Precipitation Indices Based on Copula Function in Guangdong Beijiang River Basin

选用9种概率分布函数对广东北江流域18测站近50 a极端降水指标R1d(年内最大1日降水量)和R5d(年内最大连续5日降水量)进行拟合,采用线性矩法估计参数,根据拟合优度检验,确定出最佳的概率分布函数,并在此基础上对50 a、100 a一遇设计值进行空间分析。通过择优选取GH Copula作为联接函数,对流域R1d-R5d联合概率分布进行研究。结果表明:1)翁源、清远地区出现50 a、100 a一遇的R1d或R5d概率较大;中北部地区50 a、100 a一遇的R1d和R5d同时出现的概率较大。2)特定R1d设计频率条件下,随着R5d设计频率的减小,二者遭遇概率也随之迅速减小;反之,特定R5d设计频率随R1d条件频率的减小,二者遭遇概率随之明显增大。

The Extreme Precipitation Indices R1 d and R5 d at 18 meteorological stations over Guangdong Beijiang River Basin for recent fifty years were analyzed by using 9 probability distribution functions. Estimate of parameters was performed by using the L-moment technique. Test of goodness of fit was done based on K-S, RMSE and PPCC for the optimal linear frequency distribution before evaluating the 50 and 100 years return periods values. Based on the Genest-Rivest method, AIC and the tail dependency of random variable, GH copula were used for R1d-R5 d connecting function. The Kendall rank correlation coefficient method was used to estimate the parameters of GH copula function. Then we comprehensively analyzed the spatial variability of joint probability distribution of the two indices. The results are shown as follows： 1） In terms of the 50 and 100 years return periods precipitation of R1 d or R5 d, the probabilities at Wengyuan and Qingyuan are greater than those in other places; The probabilities of R1 d and R5 d occurring in the same year in north-central areas are greater than those in other areas. 2） The encountering probabilities of R1d-R5 d rapidly decreased along with decreasing design frequency of R5 d under a specific R1 d frequency; whereas the encountering probabilities significantly increased along with decreasing design frequency of R1 d under a specific R5 d frequency.