Investigating the Impact of Wavelet Network on the Efficiency of Artificial Neural Network in Predicting Flood Sediments

Asignificant amount of the damages caused by floods is usually due to suspended sediments in the flood and dredging costs due to their subsidence in natural, residential and industrial areas. Therefore, sediment monitoring is very important when the water discharges. On the other hand, the non-linear nature of sediment data have made it difficult to predict this parameter. Wavelet theory is one of the pre-processing methods that can help we lead to a better resolution of the internal relationships of non-linear data by breaking down the main time series into sub-signals. I this research, the sediment data values in two stations of Abnama and Minab from Hormozgan River watershed were broken through wavelet conversion into sub-signals, and then the prediction process was carried out by the artificial neural network. Moreover, in order to investigate the impact of wavelet transform on the performance of the neural network model, the results obtained from this combined model were compared with the results obtained from the single neural network model, and their efficiency was evaluated using multi-part validation method, correlation, and root-mean-square error. The results showed that the artificial neural network in the two studied stations is able to simulate the sediment discharge with a correlation of 0.89 and 0.68 as well as the wavelet neural network with a correlation of 0.9 and 0.8. Moreover, the normalized root-mean-square error statistics were 0.104 and 0.35 in artificial neural networks and 0.124 and 0.18 in combined networks, respectively. The results showed that the impact of the wavelet on identifying sub-signals and thus improving the performance of the model compared to individual neural networks on predicting the amount of sediments in floods is clearly significant.