Comparison of the performance of SWAT, IHACRES and artificial neural networks models in rainfall-runoff simulation (case study: Kan watershed, Iran)

The catchment area is essentially a heterogeneous dynamic, time and space hydrological system, and so the process of rainfall-runoff transmission in the catchment area is a very complex phenomenon. The temporal and spatial changes in the catchment characteristics, uncertainties in rainfall patterns, and a large number of parameters that alter the rainfall in to runoff, are the main sources of complexity in such relationships. Hydrological models are vital and exigent tools for water resources and environmental planning and management. In present study three models of SWAT, IHACRES and ANN were used on a daily, monthly and annual basis in the Kan watershed, which located in the west part of Tehran, Iran. The results showed that the performance of the three considered models are generally suitable for rainfall-runoff process simulation, however, ANN model showed a better performance for daily, monthly, and annual flow simulations compared with other two models (NSE= 0.86, R-2= 0.87, RMSE= 2.2, MBE= 0.08), and particularly for the simulation of maximum and minimum flow values. In addition, the performance of SWAT model (NSE= 0.65, R-2= 0.68, RMSE= 3.3, MBE= -0.168) was better than the IHACRES model (NSE= 0.57, R-2= 0.58, RMSE= 3.7, MBE= 0.049). However, the results of the IHACRES model were still acceptable.