The artificial neural networks (ANNs) have been applied to various hydrologic problems in recently. In this paper, the artificial neural network (ANN) model is employed in the application of rainfall-runoff process on a semi-arid catchment, namely the Kurukavak catchment. The Kurukavak catchment, a sub-basin of the Sakarya basin in NW Turkey, has a drainage area of 4.25 km 2. The performance of the developed neural network based model was compared with multiple linear regression based model using the same observed data. It was found that the neural network model consistently gives good predictions. The conclusion is drawn that the ANN model can be used for prediction of flow for small semi-arid catchments. © Springer-Verlag Berlin/Heidelberg 2006.