基于运动扩散模型的含零值降水序列频率计算

摘要

应用运动扩散(KD)模型,研究含零值降水序列频率计算.以陕西省6个测站2月份降水序列为例,在Strupczewski推导矩法、极大似然法参数计算公式的基础上,利用数学变换原理和数值计算原理,推导了KD模型含零值序列分布概率权重矩法参数计算公式.采用AIC准则、OLS准则和残差平方和(RSS)最小准则,进行KD模型、频率比例法和Ⅱ型乘法分布模型拟合效果评价.结果表明:KD模型可以应用于含零值降水序列频率计算,且该模型拟合效果优于频率比例法和Ⅱ型乘法分布法;3种KD模型参数估计方法中,概率权重矩法估计KD模型参数拟合度最优.文中方法以期为我国含零值水文序列频率计算提供一种新的计算方法.%This study developed a frequency analysis method for precipitation samples with zero values using Kinematic Diffusion model.Taking the precipitation series of 6 stations in Shaanxi Province as an example,based on the Strupczewski derivation of the moment method and the maximum likelihood method,the parameter estimation formulas of KD model based on probability weight moment method are deduced by mathematical transformation principle and numerical calculation principle.The fitting effect of KD model,frequency proportional method and type Ⅱ multiplication distribution model is evaluated by using of the AIC criterion and OLS criterion and residual sum of squares criterion.The results show that the KD model can be applied to calculate the frequency of precipitation samples with zeros,and the fitting effect of the model is better than the frequency proportional method and the type Ⅱ multiplication method.Compared the three parameter estimation methods of KD model,the best fitness is the probability weight moment method to estimate the KD model parameters.The method of this paper provides a new calculation method for frequency analysis of precipitation samples with zero values.