American Journal of Environmental Sciences

Evaluation of Seasonal Autoregressive Integrated Moving Average Models for River Flow Forecasting

Kassahun Birhanu Tadesse, Megersa Olumana Dinka, Tena Alamirew and Semu Ayalew Moges

DOI : 10.3844/ajessp.2017.378.387

American Journal of Environmental Sciences

Volume 13, Issue 5

Pages 378-387


Reservoir operation policies cannot be functional in instant decision making without forecasting the future reservoir inflows. For forecasting inflows into reservoirs with only hydrological data is available like Koga irrigation dam, multivariate forecasting models cannot be used to generate accurate river flow information. As a result, an evaluation of univariate Seasonal Autoregressive Integrated Moving Average (SARIMA) models was done for forecasting monthly Koga River flow with Gnu Regression, Econometrics and Time-series Library (GRETL) software. The stationarity of historical river flow sequence was checked by Augmented Dickey-Fuller (ADF) unit root analysis. Then, seasonality was removed from the river flow time series by seasonal differencing. Using seasonally differenced correlogram characteristics various SARIMA models were identified and evaluated, their parameters were optimized and diagnostic checks of forecasts were performed using residual correlograms and Ljung-Box tests. Finally, based on minimum Akaike Information criteria, SARIMA (1, 0, 1) (3, 1, 3)12 model was selected for Koga River flow forecasting. The stationarity test of the forecasted values of this model has proved the similarity of forecast values and patterns with those of the historical ones. Thus, irrigation managers could use this model and forecast information for optimal irrigation planning and development of reservoir operation strategies in order to protect farmers and downstream environment from water shortages. Moreover, the use of stationarity test of forecast flow patterns is useful and applicable in selecting best forecast model during forecasting of any river flows.


© 2017 Kassahun Birhanu Tadesse, Megersa Olumana Dinka, Tena Alamirew and Semu Ayalew Moges. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.