贝叶斯方法在水环境系统不确定性分析中的应用述评

A Review of Bayesian Methods and Their Application in Uncertainty Analysis of Water Environmental System

贝叶斯方法是解决不确定问题的新思路,评述了以贝叶斯公式、贝叶斯统计推断及贝叶斯网络为基础的贝叶斯方法在水质评价、水环境模型参数识别、水环境管理及风险决策方面的应用。贝叶斯公式可很好地解决水质评价中监测数据、水质级别、水质标准等蕴含的不确定信息。贝叶斯统计推断耦合水环境模型为模型参数识别提供新方法,其应用难点为贝叶斯后验分布的计算。介绍了后验分布的离散贝叶斯算法、传统及改进MCMC算法。贝叶斯网络在水质评价、模型预测、水环境管理及风险决策中可同时考虑多个变量的综合作用,得到影响管理决策各因素的不确定性信息,为水环境的管理决策提供科学依据。

Bayesian methods provide new ideas for solving uncertainty problems in Water environmental system. Several Bayesian methods, such as Bayesian formula, Bayesian statistical inference and Bayesian networks, are commented on applying to water quality evaluation, parameters identification of water environment model, water environment management and risk decision making. Bayesian formula can solve uncertain information of monitoring data, water quality grade and standard in water quality evaluation. Bayesian statistical inference coupling the water environmental model provides a new approach for model parameter identification. The posterior distribution calculation is the key of application of Bayes Jan statistical inference. The Bayesian discrete algorithms based posterior distribution, the traditional and improved MC- MC algorithms are introduced. The application of Bayesian networks to water quality assessment, model prediction, water environment management and risk decision making can take multiple variable into account simultaneously. Then the uncertain information of factors influencing on management decision making is obtained, which provides the scientific basis for water environmental management decision making.