Streamflow prediction has a great significance in hydrology, water resources planning and magement studies. Either long term or short term predictions of streamflow are necessary to optimize the operation of water resources systems. An artificial neural network (ANN) is a nonlinear black-box modelling approach which has the capability to model complex nonlinear hydrological processes without physical expression. It is an altertive modelling approach that is inspired by brain and nervous systems to conventiol hydrological models. This paper describes implementation of an ANN to predict catchment flows in a snow domited mountainous basin med Karasu Basin, which is in the headwater of the Euphrates Basin in Turkey. Due to the non-availability of accurate snow data, catchment flows were predicted using only basic meteorological data, and a best meteorological data set was investigated to achieve best performance of the model. The ANN model was calibrated with recorded runoff data available for a period of twelve years. Calibrated model was further used for a period of another ten years. To achieve model accuracy, the model was simulated with both the whole-year (annual) data and part-year (seasol) data. It was found that model accuracy increased significantly when it was used for part-year (seasol) data. The model results are encouraging and model accuracy can be further increased when the simulations include snow depth data and/or snow water equivalent data.
Unless otherwise indicated, works by Griffith University Scholars are © Griffith University. For further details please refer to the University Intellectual Property Policy.