象山港海水养殖区贝叶斯网络水质评价及预测

水质评价是水环境保护与管理的重要环节,传统的评价方法在处理评价中的不确定性、大量信息处理等方面存在局限性。贝叶斯网络可以有效地表达和分析不确定性问题,实现定性分析与定量分析的有机结合。以近10a来象山港海水养殖区的水质监测数据为样本数据,采用贝叶斯网络技术,建立反映各水质指标及水质级别之间相互关系和相互影响强度的贝叶斯网络模型。模型结构表明直接影响水质级别的水质指标为氨氮、化学需氧量、硝酸盐、无机磷和叶绿素a,而其他亚硝酸盐、无机氮等4个水质指标与水质级别存在间接的因果关系。对200条监测数据进行模型精度检验,结果表明,其预测精度达94.8%,Kappa指数为0.892,这说明采用贝叶斯网络技术对水质进行评价及预测是可行的。

Assessment and Prediction of Water Quality in Mariculture Zone of Xiangshan Harbor Based on Bayesian Network

Water quality assessment plays an important role in water environmental protection and management.Traditional methods have some limitations in dealing with the uncertainty in assessment and massive information.A Bayesian network model can effectively express and analyze uncertain problems and combine qualitative analysis with quantitative analysis.Based on the mornitoring data for nearly ten years in Xiang-shan harbor of Ningbo City,a Bayesian network model expressing the relationships and interactions between different water quality indexes and water quality levels was constructed by Bayesian network approach.The model structure indicated that ammonia nitrogen,COD,inorganic phosphorus,nitrate,and chlorophyll had direct effects on water quality level,whereas there was an indirect causality between other water quality indexes such as nitrite and inorganic nitrogen and water quality level.The results of the model validation using 200 monitoring data showed that the predictive precision reached 94.8% and the Kappa was 0.892,which suggests the Bayesian network is feasible for assessment and prediction of water quality.