Prediction of chlorophyll-a in the Daning River of Three Gorges Reservoir by principal component scores in multiple linear regression models

Abstract

After the impoundment of the Three Gorges Reservoir (TGR) since 2003, eutrophication has occurred and has become severe in Daning River. To predict chlorophyll-a (Chl-a) levels, the relationships between Chl-a and 11/13 routine monitoring data on water quality and hydrodynamics in Daning River were studied by principal component scores in the multiple linear regression model (principal component regression (PCR) model). In order to determine the hydrodynamic effect on simulated accuracy, two 0-day ahead prediction models were established: model A without hydrodynamic factors as variables, and model B with hydrodynamic factors (surface water velocity and water residence time) as variables. Based on the results of correlation analysis, score 1 and 2 with significant loads of phosphorus and nitrogen nutrients were omitted in developing model A (R(2) = 0.355); while score 2 with significant loads of nitrogen was omitted in developing model B (R(2) = 0.777). The results of validation using a new dataset showed that model B achieved a better fitted relationship between the predicted and observed values of Chl-a. It indicated hydrodynamics play an important role in limiting algal growth. The results suggested that a PCR model incorporating hydrodynamics processes has been suitable for the Chl-a concentration simulation and algal blooming prediction in Daning River of TGR.