Comparison of Intelligent Models Efficiency for Routing of River Daily Flow (Case Study: Baleqlu-Chay River, Ardabil Province)

Routing of river flow is one of the most important of water resources management topics to adopt appropriate decision in occurrence time of flood or droughts. In this study, artificial neural networks (ANN), gene expression programming (GEP), wavelet- neural network (WNN) and least square support vector machine (LS-SVM) models were used for routing of Baleqlu-Chay river daily flow, located at Dareh-Roud watershed. Daily river discharge data of two consecutive hydrometric stations located at the mentioned river for the period of 1997-2013. The statistics indices including, root mean square error (RMSE), correlation coefficient (R), Nash–Sutcliffe efficiency coefficient (NS) and Bias were used to evaluate the precision of the models. Comparison of results demonstrate that the LS-SVM with RMSE=1.540 m3/s, R=0.894, NS=0.713 and Bias=0.013 had the best performance in the test period. But in estimating of peak discharge values, the WNN model with average relative error equal to 34.21% had the least error. It should be mentioned that all of the models tended to underestimate the discharge values.