应用高阶概率权重矩法估计广义极值分布参数

Estimation of GEV Distribution Parameters by Higher Probability Weighted Moments

在洪水频率分析计算中,当点绘年最大洪峰流量序列的经验频率时,经验点据常常出现两段或多段的分散区,重现期较大的洪水估算一般是根据大中洪水值的趋势进行外推。本文根据高阶概率权重矩法原理和广义极值分布估算参数模型,进行陕北地区神木水文测站的年最大洪峰流量频率分布参数计算。结果表明:高阶概率权重矩法能赋予大洪水值更多的权重。蒙特卡洛模拟试验表明:并非阶数越高越好,适当提高阶数可以减小误差,但阶数过高反而会增大误差。

When an annual maximum flow series is displayed in a probability plot for the analysis of flood frequency, the data often exhibit two or more distinct segments. For estimating floods of large return periods, it is probably extrapolated by the trend of large and medium-sized flood values. The principle of higher probability weighted moments (HPWMS) was employed to estimate parameters of generalized extreme value (GEV) distribution. The results show that the higher the order are, the better fitting of the GEV distribution to annual maximum flows in larger segments. The Monte Carlo simulations also show that moderate order of HPWMS may reduce the estimation errors rather than higher order, and conversely, over-higher orders may increase the estimation errors.