Wang Jian-Ping, Cheng Sheng-Tong, Jia Hai-Feng
Department of Environmental Science and Engineering, Tsinghua University, Beijing 100084, China.
Huan Jing Ke Xue. 2006 Jan;27(1):24-30.
Parameter identification plays an important role in environmental model application. Markov Chain Monte Carlo method was introduced to estimate parameter uncertainty, since usual Bayes discrete methods were not applicable to produce posterior distribution of complicated environmental model due to the limit of computation. In order to study the performance and efficiency of MCMC, two case studies were used. Results indicate that, either sampling performance or sampling efficiency, MCMC method both has its special advantages in producing posterior distribution. Moreover, results of Gelman convergence diagnostics indicate that sampling sequence can converge to a stationary distribution. A key finding was that the MCMC scheme presented herein provided a powerful means of parameter identification and uncertainty analysis.
参数识别在环境模型应用中起着重要作用。由于常规贝叶斯离散方法因计算限制不适用于生成复杂环境模型的后验分布，因此引入了马尔可夫链蒙特卡罗方法来估计参数不确定性。为了研究马尔可夫链蒙特卡罗方法（MCMC）的性能和效率，使用了两个案例研究。结果表明，无论是抽样性能还是抽样效率，MCMC方法在生成后验分布方面都有其独特优势。此外，Gelman收敛诊断结果表明抽样序列可以收敛到平稳分布。一个关键发现是，本文提出的MCMC方案提供了一种强大的参数识别和不确定性分析方法。
