Streamflow forecasting at ungaged sites using support vector machines

Abstract

Developing reliable estimates of streamow prediction are crucial for water resources management and ood forecasting purposes. The objectives of this study are to investigate the potential of support vector machines (SVM) model for streamow forecasting at ungaged sites, and to compare its performance with other statistical method of multiple linear regression (MLR). Three quantitative standard statistical indices such as mean absolute error (MAE), root mean square error (RMSE) and NashSutcli_e coe_cient of e_ciency (CE) are employed to validate both models. The performances of both models are assessed by forecasting annual maximum ow series from 88 water level stations in Peninsular Malaysia. Based on these results, it was found that the SVM model outperforms the prediction ability of the traditional MLR model under all of the designated return periods.