Forecasting hydrological variables are suitable tools for water resources management. For this aim, time series models as efficient methods have been considered by the hydrologists. Also, in this study, time series modeling has been considered and seasonal time series models on the total flow, base flow and run-off data of hydrometric station Arazkuseh with 35-years period have been fitted. Afterwards, condition and accuracy of fitted models in forecasting future discharges were considered using Akaike Information Criterion (AIC), Root Mean Squared Error (RMSE) and other parameters.The results show that, ARIMA(1,0,0)(0,1,1) is the best model for forecasting the total flow and base flow, respectively and ARIMA (1,0,1) (0,1,1) is the best model for forecasting the run-off. For assessing the accuracy of models, 5 years data from 2007 to 2012 were used for estimation of total flow, base flow and run-off. RMSE and R2 for forecasting the total flow is 3.46 and 0.47, for base flow 2.61 and 0.48 and run-off 0.83 and 0.23 were obtained. The results revealed that if the total discharge was obtained by summation of values obtained from base flow and run off time series model, the more accurate results could be obtained.