Comparison of data-driven modelling techniques for river flow forecasting

Abstract

Accurate forecasting of streamflow is essential forthe efficient operation of water resources systems. Thestreamflow process is complex and highly nonlinear.Therefore, researchers try to devise alternativetechniques to forecast streamflow with relative easeand reasonable accuracy, although traditionaldeterministic and conceptual models are available. Thepresent work uses three data-driven techniques, namelyartificial neural networks (ANN), genetic programming(GP) and model trees (MT) to forecast river flow oneday in advance at two stations in the Narmada catchmentof India, and the results are compared. All the modelsperformed reasonably well as far as accuracy ofprediction is concerned. It was found that the ANN andMT techniques performed almost equally well, but GPperformed better than both these techniques, althoughonly marginally in terms of prediction accuracy innormal and extreme events.