Application of extended artificial neural network method in a watershed for estimation of monthly discharge sub-basins (Case study: Beheshtabad watershed)

Finding hydrologic criteria for regions with insufficient or no data is important in water resources studies. In basins with lack of streamflow data, it is preferred to use black-box methods to relate inputs (meteorologic and physiographic parameters) and output (river discharge) without the analysis of the process. Artificial Neural Networks model is one of the methods which can be used in such situation. In the present research, MLP artificial neural network with back-propagation algorithm is used to estimate monthly discharge of Beheshtabad watershed. The results showed that last-month's discharge and precipitation of two months ago were the main input parameters. The best model obtained for the Beheshtabad station was used for Babahaidar and Kuhe-sokhteh sub-basins and the measured and estimated discharges were compared. The results showed that the calibrated model obtained for the main hydrometry station could be applied in sub-basins.