太湖水体Chl-a预测模型ARIMA的构建及应用优化  

Construction and Application Optimization of the Chl-a Forecast Model ARIMA for Lake Taihu

作　　者：李娜 李勇[1,2] 冯家成 单雅洁 钱佳宁 LI Na;LI Yong;FENG Jia-cheng;SHAN Ya-jie;QIAN Jia-ning(Key Laboratory of Integrated Regulation and Resources Development on Shallow Lakes,Ministry of Education,Hohai University,Nanjing 210098,China;College of Environment,Hohai University,Nanjing 210098,China)

机构地区：[1]河海大学浅水湖泊综合治理与资源开发教育部重点实验室,南京210098 [2]河海大学环境学院,南京210098

出　　处：《环境科学》2021年第5期2223-2231,共9页Environmental Science

基　　金：国家重点研发计划项目(2018YFC0407906);国家自然科学基金项目(51879081,51579074)。

摘　　要：叶绿素a(Chl-a)是湖泊浮游植物生物量的重要指标,其含量能反映水中浮游植物的丰度和变化规律.以1999年12月~2019年8月太湖水体Chl-a和环境因子的逐月监测数据为基础,运用主成分分析方法探讨了Chl-a与环境因子的关系,据此建立了Chl-a与主要环境因素之间的多元线性逐步回归模型及自回归综合移动平均模型(ARIMA).结果表明:(1)太湖Chl-a浓度存在着明显的季节变化,且总体处于上升趋势.总磷(TP)、高锰酸盐指数、月均气温(MAT)和月度降雨量(MR)与Chl-a浓度存在较好的变化同步性,总氮(TN)和氨氮(NH_(4)^(+)-N)则表现出明显的滞后性.(2)主成分分析结果表明,太湖水体藻类暴发条件不仅仅是基于N和P等限制性因素,而是发展为TN、NH_(4)^(+)-N、TP和高锰酸盐指数、MR和MAT等多元因素的综合影响.(3)两种模型经验证比较,基于1999~2019年逐月资料建立的Chl-a浓度的ARIMA模型模拟效果和预测精度明显优于所建立的多元线性逐步回归模型,特别是在考虑主要环境因素作为自变量及优化自变量取值情况下其预测效果得到进一步提升.建立的ARIMA(0,1,1)(0,1,1)模型将有助于太湖藻类暴发的预报和预警,并为及时有效地安排水资源调度及调控等水环境管理措施提供依据.As an important indicator of phytoplankton biomass in lakes,the chlorophyll-a( Chl-a) concentration reflects the abundance and variation of phytoplankton in the water. Based on the monthly monitoring data of Chl-a and environmental factors in Lake Taihu from December 1999 to August 2019,key environmental factors related to Chl-a and their relationships were found using the principal component analysis( PCA) method. A multiple linear stepwise regression model and an auto-regressive integrated moving average( ARIMA) model were developed to predict the monthly Chl-a concentrations. The results showed that the Chl-a concentrations in Lake Taihu exhibited clear seasonal change characteristics and an overall trend of a gradual increase. The changes in total phosphorus( TP),the permanganate index,monthly average temperature( MAT),and monthly rainfall( MR) matched the Chl-a concentrations relatively well,whereas the changes in total nitrogen( TN) and ammonium nitrogen( NH_(4)^(+)-N) lagged significantly. The PCA results showed that the increased phytoplankton biomass and consequent algae outbreaks in Lake Taihu were not limited to the effect of a single factor such as TN or TP,but were comprehensively affected by multiple factors such as TN,NH_(4)^(+)-N,TP,the permanganate index,MR,and MAT. Through further validation,the ARIMA model of Chl-a concentrations was proved to be significantly better than the multiple linear stepwise regression model,especially when considering the key environmental factors as independent variables and optimizing their values. The established ARIMA( 0,1,1)( 0,1,1) model would be helpful for forecasting algae blooms in Lake Taihu and provide useful suggestions for water environmental management,such as water resources dispatch and regulation.

关 键 词：太湖 叶绿素A 环境因子 多元线性逐步回归 ARIMA模型 

分 类 号：X524[环境科学与工程—环境工程]