Sensitivity Analysis of River Chlorophyll-a to Environmental Factors Based on Support Vector Machine Regression Algorithm in Northern China

Abstract

The outbreak of planktonic algae seriously affects the water environmental quality of the artificially reformed rivers in the north of China, and it is difficult to control. In this study, taking a typical small river dam area in the north of China—Qingshui River dam area in Zhangjiakou City as an example, the spatiotemporal variation characteristics of environmental factors were analyzed firstly. On this basis, the Chlorophyll-a (Chl-a) prediction model was established by using the Support Vector Regression (SVR). Further, the sensitivity of Chl-a was analyzed, moreover, we confirmed the main limiting factors of Chl-a. The results showed that in 2018, the average content of Chl-a was 126.25ug/L, while the content of DO and pH was also totally high. The maximum content of TN was 16.68mg/L and high in the whole year; the maximum value of CODmn was 50.00mg/L, which was slightly higher in summer; the average value of NH4+-N and TP was only 0.78mg/L and 0.18mg/L. The eutrophication evaluation showed that the proportion of areas with hyper eutrophication is 73%, and the most severe eutrophication was in summer. We used the RBF kernel function SVR model and 10-fold cross-validation method to optimize the parameters. The penalty parameter c is 1.4142, the kernel function parameter g is 1, and the training and verification errors were only 0.032 and 0.067. The application effect was good. Based on the sensitivity analysis of SVR prediction model, the sensitivity coefficients of Chl-a were TP, WT, DO, pH, TN, NH+ 4-N in turn. According to the analysis of the present water pollution situation, TP is the limiting factor of Chl-a in Qingshui River with 0.571 sensitive coefficient and 33% contribution rate, and it is also the main prevention and control factor of planktonic algae outbreak.