基于贝叶斯方法的水文地质参数识别  

Identification of hydrogeological parameters based on the Bayesian method

作　　者：陆乐[1] 吴吉春[1] 陈景雅[2] 

机构地区：[1]南京大学水科学系,南京210093 [2]江苏省交通规划设计院,南京210005

出　　处：《水文地质工程地质》2008年第5期58-63,共6页Hydrogeology & Engineering Geology

基　　金：国家自然科学基金资助项目(40725010,40672160)

摘　　要：文章基于马尔可夫链(MCMC)采样法,耦合地下水数值模拟的MODFLOW软件,提出了贝叶斯方法,用于水文地质参数识别这一复杂的非线性优化问题。采取单分量自适应Metropolis(SCAM)采样算法并加入了权重因子,从而大大提高了该方法解决高维参数问题的效率。通过两个实例研究,验证了该方法具有独特的寻优性能和效率。结果表明,贝叶斯方法适用于复杂地下水问题的参数识别,并体现出了出色的全局寻优性。The Bayesian method is introduced into the field of groundwater to identify hydrogeological parameters, which is usually a complicated nonlinear problem. A Bayesian hydrogeological parameter identification algorithm is proposed based on the MCMC sampler which is eoupled with the groundwater modeling software, Modflow. The efficiency of identifying high dimensional parameters is greatly improved by using the SCAM sampler and a weighting factor. Two examples are presented to illustrate the distinctive optimization performance of this algorithm. The results indicate that the proposed algorithm is capable for complex groundwater problem and shows outstanding advantage of global optimization.

关 键 词：贝叶斯方法 MCMC方法 SCAM算法 水文地质参数 优化 参数识别 

分 类 号：P641.2[天文地球—地质矿产勘探]