BFS在洪水预报中的应用研究  

Application Research of BFS to Flood Forecasting

作　　者：张宇[1] 梁忠民[1] 

机构地区：[1]河海大学水文水资源学院,江苏南京210098

出　　处：《水电能源科学》2009年第5期44-47,共4页Water Resources and Power

基　　金：国家自然科学基金资助项目(50779013)

摘　　要：采用贝叶斯预报系统(BFS)水文不确定性处理器(HUP)对水文预报不确定性进行分析,实现概率洪水预报。以新安江流域水文模型为洪水预报模型提供流量初始预报系列,通过亚高斯模型对流量初始预报系列及实测系列分别进行正态分位数变换,由贝叶斯公式得到预报变量的后验概率分布并进行洪水过程的概率预报,采用分布点估值定值预报,并可通过构造置信区间对点估值预报的不确定性进行评估。以南一水库流域为例,将BFS后验概率分布均值与新安江模型预报进行对比,结果表明BFS可提高预报精度。The hydrologic uncertainty processor （HUP） within Bayesian forecasting system （BFS） is employed to investigate the hydrologic forecasting uncertainties,and thus realizing probabilistic forecasting. Xin＇anjiang model is used to yield initial discharge forecasting series. Then both the initial predicted discharge series and the corresponding observations are normalized by Meta-Gaussian model. Finally the predictive posterior probability distribution is acquired through Bayesian formula. Therefore, probabilistic forecasting for a flood event can be accomplished based on the predictive posterior distribution. The point estimation of the distribution,e, g. the mean value can be taken as the quantitative forecasting similar to traditional approaches. Furthermore the uncertainty of the point estimation is evaluated by constructing confidence interval. As an example, BFS is applied to the probabilistic flood forecasting for Nanyi basin,China. The mean of the posterior probability distribution derived from BFS is compared with the predication of XAJ model. It is shown that BFS can improve forecasting accuracy appropriately.

关 键 词：贝叶斯预报系统 概率洪水预报 新安江流域水文模型 亚高斯模型 后验概率分布 

分 类 号：TV122[水利工程—水文学及水资源] P338[天文地球—水文科学]