水文频率分析中的一种贝叶斯方法

A Bayesian Method in Hydrological Frequency Analysis

在贝叶斯（Ｂａｙｅｓ）框架下将年洪峰分布参数θ＝ＣＳ／ＣＶ选作为随机变量，在对θ的地区分布进行一定概化的基础上给出了由区域信息估计θ先验分布的一种主观贝叶斯方法；同时对Ｐ－Ⅲ型总体下的贝叶斯频率计算问题进行了统计试验研究。结果表明，提高先验分布的估计精度会降低贝叶斯方法估计Ｘｐ的抽样误差，当样本容量较小时效果更明显，这对水文频率分析中使用贝叶斯方法具有一定的参考价值。

arameter θ=CS/CVis taken as a random variable under the Bayesian framework in this paper .Asubjective Bayesian method is proposed to estimateprior distribution by applying regional information .Soms statistic tests have also been performed for the Pearson type Ⅲ(PⅢ)distribution .The outcomes show that the sampling error of Xp can be reduced by improving the estimation of prior distribution, and this is obvious for samples with small sizes .