Comparing the Performance of WetSpa Hydrological Model, Artificial Neural Network and Adaptive Neuro-Fuzzy System for Simulating River Flow Discharge (Case Study: BalukhluchayWatershed, Ardabil Province)

Generated runoff has significant effects on quality and quantity of the surface and groundwater resources, soil erosion and even vegetation cover in watersheds. In recent years, different computer models have been applied for estimating and forecasting the runoff and its effects. In this study, the performances of WetSpa hydrological model and ANN and ANFIS intelligent models were evaluated in simulation of the rainfall-runoff and estimation of the daily discharge in Balukhluchay watershed of Ardabil province. The required data including the needed information for the digital elevation model, land use and soil maps and also the climatic and hydrological data of the daily precipitation, temperature, reference evapotranspiration and discharge time series were prepared. The data of 2007 to 2010 were used for calibration and training and the data of 2010 to 2012 were used for validation and test of the models. Different evaluating coefficients such as Nash-Sutcliffe, Correlation Coefficient and Root Mean Square Error (0.457, 0.696 and 1.719 for WetSpa, 0.724, 0.865 and 1.232 for ANN and 0.289, 0.603 and 1.968 for ANFIS, respectively) showed that the highest accuracy among the used models belonged to the ANN, WetSpa and ANFIS, respectively. Despite high accuracy of the ANN, since it is a black box model, its usage for investigating the land use and topography effects on the runoff has some limitations in comparison with the completely distributed Wetspa model.