基于AM-MCMC算法和Nash模型的概率洪水预报  

Probabilistic Flood Forecast Based on AM-MCMC and Nash Model

作　　者：邢贞相[1,2] 芮孝芳[3] 冯杰[3] 

机构地区：[1]东北农业大学水利与建筑学院,哈尔滨150030 [2]东北农业大学农业工程博士后科研流动站,哈尔滨150030 [3]河海大学水文水资源学院,南京210098

出　　处：《武汉理工大学学报（交通科学与工程版）》2010年第6期1275-1279,共5页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：国家自然科学基金项目(批准号:50609005);霍英东青年教师基金项目(批准号:101075);中国博士后科学基金项目(批准号:2009451116);黑龙江省博士后科研基金项目(批准号:LBH-Z08225);黑龙江省教育厅科学技术研究项目(批准号:11451022);东北农业大学科学研究基金项目(批准号:190152)资助

摘　　要：贝叶斯理论不仅可与任一确定性水文模型协同工作实现概率洪水预报,而且还可以通过综合参数的先验信息和样本信息来研究指定参数的不确定性.由于难于求得参数的贝叶斯后验密度解析式,采用改进的基于自适应采样的马尔可夫链蒙特卡罗算法(AM-MCMC)求其数值解的方法.经实例应用,得到了长江三峡地区沿渡河流域Nash模型参数k,n的后验分布,实现了该流域的概率洪水预报,同时给出了各时刻洪水流量的均值和方差的预报值.Bayesian theory not only can work with any certain hydrological model to realize probabilistic flood forecasting,but also can be used to study uncertainty of certain parameters through integrating prior and posterior information about parameters.Because it is hard to get an analytic expression of posterior distribution of parameter,the Markov Chain Monte Carlo algorithm based on Adaptive Metropolis（AM-MCMC） was used to obtain its numerical solution.Through the case study of Yanduhe basin in the Three Gorges of the Yangtze River,posterior distributions of Nash model were gotten,and then probabilistic flood forecasting for the basin was realized.At the same time,predictive values of mean and variance of flood discharge at each time were given,which can be used to estimate risks of flood control decision.

关 键 词：贝叶斯理论 AM-MCMC Nash模型 参数不确定性 概率洪水预报 

分 类 号：TV124[水利工程—水文学及水资源]