Prediction of Monthly Inflow to Karkhe Reservoir Using ARIMA model

Forecasting future river flow is a critical aspect in efficiently managing water resources, particularly in meeting the diverse downstream requirements of reservoir dams. The significance of predicting inflow to the dam is amplified due to its role in addressing its downstream needs. The present study focuses on predicting the monthly inflow to the Karkheh Reservoir Dam through the utilization of integrated autocorrelated moving average (ARIMA) models, including the seasonal variant (SARIMA). The development of these models involved analyzing 57 years of monthly flow data into the Karkheh dam reservoir. Of this dataset, 47 years were designated for model training, while the remaining 10 years were used for model testing. The determination of optimal ARIMA model parameters involved assessing various combinations of (p, d, q), with selection based on the Akaike information evaluation criterion. Results indicate that the ARIMA model with parameters (8,0,7) yields the lowest Akaike information evaluation criterion. Additionally, recognizing the seasonality in the data, a SARIMA model was constructed and employed for predicting monthly flow into the Karkheh dam reservoir. A comparison of the root mean squared error between the ARIMA and SARIMA methods reveals superior accuracy in predicting monthly flow to the Karkheh dam reservoir with the ARIMA model.