Application of LS-SVM, ANN, WNN and GEP in Rainfall- Runoff Modeling of Kiyav-Chay River

Streamflow forecasting is necessary for water resources management and planning in rivers, lakes, reservoirs and protection of river banks during flood time. In this study, different soft computing models including artificial neural networks (ANN), the hybrid of wavelet-artificial neural networks (WANN), gene expression programming (GEP) and least square-support vector machines (LS-SVM) was utilized for river flow estimation of Khiav-Chay. Statistical measures and ANOVA test was used for evaluation of applied models. The results indicated that WANN model was the best one with the highest correlation coefficient (R=0.877) and lowest root mean squared error (RMSE=0.696) and Nash Sutcliff coefficient (NS=0.767) in validation phase. The results of ANOVA test was in agreement with statistical criteria values and WANN model with the lowest F statistic (F=0.11) and highest significant resultant (0.75) was selected as the best model. Furthermore, in estimation of maximum discharge, WANN with mean relative error equals to 30.19% has the minimum error of estimation compared to other models.