Application of Seasonal Time Series Models for Prediction of Monthly Inflow to Yamchi and Sabalan Reservoirs in Qarasu Catchment, Ardabil

Predicting volume of water stored in reservoirs in the future periods plays an important role in planning and managing the optimal use of water resources systems. In this study, time series analysis method was used to predict the monthly inflow to Yamchi and Sabalan reservoirs in Ardabil province. The monthly flow data measured at hydrometric stations, located at the dam's entrance for 21 years (1994 to 2015) were used to build and test an appropriate model. The structures of the seasonal models were identified according to the auto-correlation charts (ACF) and partial auto-correlation (PACF), and then the appropriate model for each hydrometric station was selected based on the Akaike Information Criterion (AIC), Akaike Information Criterion Correction (AICC) and Bayesian Information Criterion (BIC). By fiting the model to the observational data, the parameters of each model were determined and the adequacy of the selected models was also examined by diagnostic tests. The results showed that ARIMA (1,0,0)(0,1,1)12 and ARIMA (1,1,1)(0,1,1)12 models, respectively for the monthly flow data of Yamchi and Arbabkandi stations have the lowest root mean square error (RMSE) and mean absolute error (MAE) and the highest determination coefficient. The values of these indicators in the model related to Yamchi hydrometric station were 1.04, 0.606 and 0.63, respectively, and for Arbabkandi hydrometric station were 1.35, 0.8 and 0.74, respectively. Therefore, the selected models accurately predict the monthly inflows to Yamchi and Sabalan reservoirs. Comparing the predicted results with the observational data showed that the selected models are not very accurate in predicting high discharge values.