Prediction of daily river flows in Firat-Dicle basin using different artificial intelligence methods

The forecasting and modeling of river flow in hydrological processes is quite important to deliver the sustainable use and effective planning and management of the water resources. The river flow process in any cross section of river system can be characterized as the function of various variables such as, spatial and temporal distribution of rainfall, catchment and river physical characteristics. In order to estimate hydrological processes such as runoff and change of water level using existing methods, parameters such as the physical properties of the catchment and river network and detailed observation data are necessary. The artificial neural network (ANN) has been successfully used in the hydrological sciences during recent years. The recent studies indicated that the ANN offers a promising results in the field of water resources and hydrology, such as streamflow estimation, rainfall?runoff modeling; reservoir inflow forecasting; reservoir operation; longitudinal dispersion in the natural channels; and suspended sediment estimation.The main aim of this study is to develop a suitable ANN and Wavelet Transform models for river flow forecasting in the Firat-Dicle Catchment, Turkey. In this study, daily river flow at the time period 1968-2006 was used for modeling. The performance of the ANN and Wavelet Transform models were compared with multi-linear regression models. Root mean square errors (RMSE), mean absolute errors (MAE) and deterministic coefficient (R2) statistics were used for the evaluation of the models? performances. Comparison results indicated that the neural computing techniques could be employed successfully in modeling river forecasting.Key Words: Wavelet Transform, Artificial Neural Networks, Radial Basis Neural Networks, River Flow Forecasting