Comparison of artificial neural networks and autoregressive model for inflows forecasting of Roseires Reservoir for better prediction of irrigation water supply in Sudan

The Blue Nile River is utilized in Sudan as the main source of irrigation water. However, the river has a long, dry, low-flow season (October–May), which necessitates the use of regulations and rules to manage its water use during this period. This depends on the use of accurate lead time forecasts of inflows to the reservoirs built along the river. Thus a reliable and tested forecasting tool is needed to provide inflow forecast, with sufficient lead time. In the present study, artificial neural network (ANN) is used to model the recession curve of the flow hydrograph at El-Deim gauging station, which subsequently is used as inflows to the Roseires Reservoir on the Blue Nile River. Different scenarios of ANN have been tested to forecast 23 10-day mean discharges during the recession period and their performances were assessed. Results from the optimal ANN model were compared to those simulated with an autoregressive (AR1) model to check their accuracy. Modelling results showed that the ANN model developed is capable of accurately forecasting the inflows to the Roseires Reservoir and outperforms the AR1 model. It has then proposed for use in operation of the reservoir for purposes of predicting irrigation water supply