Prediction to chlorophyll-a concentration of impoundment process in Xiangxi Bay of Three Gorges Reservoir

Abstract

The support vector machines (SVM) model was established to predict chlorophyll-a concentration of impoundment process in Xiangxi Bay of Three Gorges Reservoir. In surveys, 10 stations have been investigated and 191 samples were collected from September 25 to October 14 in 2007. Using stepwise multiple linear regression (MLR) method, six important environmental factors (water temperature, dissolved oxygen, pH, phosphate, total nitrogen and ammonium nitrogen) were selected as independent variables in SVM model. The optimal parameters of the SVM model was determined based on leave one out cross validation (LOOCV). For the LOOCV test, the cross validated squared correlation coefficient Q 2 for optimal SVM was 0.7428. Compared with stepwise MLR model, the SVM model has more powerful predictive capacity with the squared correlation coefficient R 2 of 0.8768 for the test set.