Abstract: An integrated approach of absolute principle components score-multivariate linear regression (APCS-MLR) and structural equation modeling (SEM) were developed to understand the influence of water chemistry variables on chlorophyll a (Chl a) in Lake Dianchi. The SEM result was further validated with the artificial neural networks (ANN) model. It proved that there was a good agreement on the results of the various models. The model results demonstrated that, among the water chemistry factors, physical factors (T > DO > SD > pH) had the greatest influence on Chl a; whereas nutrients had little influence. In severely polluted water with chronically high nitrogen (N) and phosphorus (P) concentrations like Lake Dianchi, the change of nutrients concentrations will not significantly influence on Chl a, while the sensibility of N is higher than P. Therefore, nitrogen load reduction should be put in priority for eutrophication control in Lake Dianchi.

Key words: eutrophication, structural equation modeling, artificial neural networks, chlorophyll a, Lake Dianchi

摘要： 应用基于主成分绝对得分的源解析模型(APCS-MLR)和结构方程模型(SEM)识别滇池富营养化的关键影响因子, 定量描述叶绿素a (Chl a)浓度与关键影响因子的关系, 并与神经网络模型(ANN)分析结果进行对比, 检验此结果的可靠性。模型结果表明, 影响滇池富营养化发生的最关键影响因子为物理因子(T > DO > SD > pH), 其次为营养物质(NH₃-N); 在当前观测的高氮高磷水环境下, 营养物质的浓度变化对Chl a浓度的影响并不显著, 但相对而言, 控氮比控磷可能更为有效。

关键词: 富营养化, 结构方程模型, 神经网络, 叶绿素a, 滇池