Prediction of sediment resuspension in Lake Taihu using support vector regression considering cumulative effect of wind speed

Abstract Sediment resuspension is critical to ecosystem function in shallow lakes. Turbidity is one of the most commonly used indicators of sediment resuspension and has proven to be strongly related with wind. However, it is still difficult to predict sediment resuspension due to its complicated mechanisms. In this study, a support vector regression (SVR) model considering the cumulative effect of wind speed was trained to predict sediment resuspension based on intensified field observations at two sites in Lake Taihu. The accuracy of the SVR model was evaluated, and the initial turbidity was introduced to the model to illustrate its contribution to sediment resuspension. The critical wind speed was also evaluated based on this model. The results indicate that the SVR model considering the cumulative effect of wind speed can increase the accuracy of prediction in comparison with traditional fitting methods. The root-mean-square error (RMSE) of the predicted turbidity dropped to 11.36 NTU at one site and 16.78 NTU at the other, and the maximal information coefficient ( c i m a x ) for the relationship between wind speed and turbidity increased to 0.56 for both observation sites. The introduction of initial turbidity significantly improved the performance of the SVR model, with an RMSE value lower than 8.00 NTU and a c i m a x value higher than 0.95. Analysis of the critical wind speed using the SVR model shows that the critical wind speed generally increased with the rise of initial turbidity. The critical wind speeds at initial turbidities of 30, 40, 50, and 60 NTU were 5, 6, 7, and 7 m/s, respectively.