基于线性矩系数图的日降雨量年最大值系列频率分布选型

Selecting a Probability Distribution for Annual Maximum Series of Daily Rainfall Based on L-moment Ratio Diagram

为提高对暴雨数据的预测精度,探讨城市暴雨频率分布选型,考察线性矩系数图作为城市暴雨分布直观判断工具的可行。以南京市58年日降雨量年最大值系列为例,采用线性矩系数图方法,并配合L-kurtosis的偏差以及K-S检验法,分析5种常的3参数分布函数(即GEV、GLO、LN3、GPA和PE3)的拟合情况,得到计算结果,并绘制线性矩系数图。根据线性矩系数图观判断,皮尔逊III型分布(PE3)拟合最佳,3参数对数正态分布(LN3)次之,该结果与L-kurtosis的偏差和K-S检验法保持很的一致性。结果表明,线性矩系数图作为一种判断工具,应用在国内城市暴雨频率分布选型上有其可行性。

To improve the estimation accuracy of rainfall, the distribution of annual maximum series of daily rainfall based on L-moment ratio diagram was investigated. Selecting 58 years rainfall data of Nanjing as an example, the L-moment ratio diagram was plot due to it is useful for goodness of fit applications. L-kurtosis difference and Kolmogorov-Smirnov tests were used in 5 distribution functions, GEV, GLO, LN3, GPA, and PE3. These fitting curves were plot in the L-moment ratio diagram. The results show that the PE3 distribution is the best fit for the sample data as judged by the L-moment ratio diagram, second is the LN3 distribution. This agrees with the results of L-kurtosis difference and Kolmogorov-Smirnov tests. The L-moment ratio diagram is an attractive tool for distribution fitting.