Forecasting and modeling of monthly runoff using ANN model in Mazandaran province

Document Type : Original Article


1 Assistant Professor in Engineering Hydrology, Rangeland and Watershed Management Department, Faculty of Agriculture & Natural Resources, University of Gonbad Kavous

2 Bsc Student in Water Engineering, Faculty of Agriculture & Natural Resources, University of Gonbad Kavous



Nowadays, water flow modeling and forecasting plays a key role in flood hazard reduction, reservoir optimization, and Water resource management. These models are mostly developed and applied for simulation and prediction. In this research, to model and forecast the monthly runoff, the data of 2 hydrometric stations of Siahrood and Talar in Mazandaran province were used in a period of 20 years (2002-2021). The time series homogeneity was examined using the Chow`s method. Since, monthly runoff data are time-dependent, these data were first arranged as time series. After sorting the data, artificial neural network (ANN) model was used to forecast the monthly runoff in selected hydrometric stations. Data entry into the model (ANN) was done using the forward algorithm. After modeling the (ANN), monthly runoff changes were forecasted in selected hydrometric stations for the next 12 months with SPSS 25 software. Lastly, based on the forecasted values and using MAD, RMSE, MAPE and R2 indices, the accuracy and precision of (ANN) model was evaluated. The results showed that the artificial neural network model performed very well for predicting monthly runoff values for both Siahrood (R2=0.9945) and Talar (R2=0.9864) hydrometric stations. Also, consecutive overestimation and underestimation, which increases the error and decreases the performance of the models, was not observed for (ANN) model.