Performance of Neural Network models with Kalman learning rule for flow routing in a river system

River flow routing provides basic information on a wide range of problems related to the management, design and operation of river systems. In this paper, three layer Cascade Correlation Artificial Neural Network (CCANN) models with Kalman learning rule have been developed to forecast the one day ahead daily flow at Ilarionas station on the Aliakmon river, in Northern Greece. Three multipurpose reservoirs, downstream of the Ilarion station, are currently in operation along the Aliakmon river route for irrigation, water supply and power generation. One day ahead forecasts are useful in decision making for the operation and the management of the three reservoirs. The tested network topology is using multiple inputs, which include the time lagged daily flow values from different upstream stations and a single output, which are the daily current flow values at Ilarionas station. Kalman's learning rule was used to modify the artificial neural network weights. The results show a good performance of the CCANN models for forecasting the one day ahead daily flow values, at Ilarionas station and demonstrate their adequacy and potential for river flow routing. The performance of the proposed models becomes worse as the lead time increases. The results also show that the CCANN models with the suggested Cascade Correlation algorithm in which Kalman's learning rule is embedded performed better than the back propagation ANNs. The CCANN models introduced in this study are sufficiently general and have great potential to be applicable to many hydrological and environmental applications.