Estimation and forecasting of daily suspended sediment data using wavelet-neural networks

Accurate prediction of the suspended sediment loads in streams is important for water resources engineering. Suspended sediments are a determining factor of the service life of hydraulic structures, like dams. Although the use of artificial neural networks (ANNs) has allowed significant progress in the estimation of suspended loads, there is still a need for more sensitive estimation methods. This study aims to estimate and predict the suspended sediment load in rivers by a combined wavelet-ANN method. Measured data were decomposed into wavelet components via discrete wavelet transform, and the new wavelet series, consisting of the sum of selected wavelet components, was used as input for the ANN model. The wavelet-ANN model provides a good fit to observed data for the testing period. The first part of this study deals with the prediction of suspended sediment using past sediment data. In the second part of the study, estimation of the sediment load is studied using daily river flow data. The results are compared with those of the conventional ANN method and of the sediment rating curve (SRC) method. Wavelet-ANN model predictions are shown to be significantly superior to the ones obtained by the conventional ANN model and the SRC model in terms of conventional performance evaluation criteria. The peak sediment values are approximated more closely by the wavelet-ANN method. The number of inaccurate sediment estimations decreased significantly and the cumulative sediment sum is closely approximated with the wavelet-ANN method. (c) 2008 Elsevier B.V. All rights reserved.