摘要
以太湖2005年的监测资料为基础,运用多元统计回归和BP人工神经网络方法构建模型,探求叶绿素a与水深、水温、营养盐等10项环境因子之间的关系,通过验证发现BP模型对叶绿素a浓度的拟合值与叶绿素a浓度的实测值之间的均方误差为220.3059,优于统计回归模型的235.4569;此外对两种模型进行了灵敏度测试,结果都显示总磷不是太湖叶绿素a浓度的限制因子,而水深、水温、总氮的变化对叶绿素a浓度影响显著。本研究对太湖叶绿素a浓度预测模型的建立是十分有意义的。
This study deals with the relation between chlorophyll - a and 10 environmental factors such as water depth ( D), water temperature (T), nutritive salt and etc. based on the monitoring data of Taihu lake in 2005. We constructed two kinds of model using the multiple statistical regression method and the back propagation artifical neutral network methed. Through the validation we find that the BP model works better than that statistical regression one, which the mean squared deviation of fitted value and measured value by BP model is 220. 3059, while is 235.4569 by statistical regression. The results of both models showed that phosphorus was not the limiting factor in the lake, while water depth, water temperature and TN had a strong influence on chlorophyll - a. This preliminary study can be helpful to the construction of chlorophyll - a prediction model in Taihu lake.
出处
《环境保护科学》
CAS
2009年第4期46-49,共4页
Environmental Protection Science
基金
国家重点基础研究发展计划资助(2008CB418003)
江苏省自然科学基金项目(BK2007151)
教育部博士点基金(20060284011)
关键词
太湖
叶绿素A
预测
统计回归
神经网络
Taihu Lake Chlorophyll- a Prediction Statistical Regression Neutral Network