Performance Of DifferentArtificial Neural Networks In Monthly StreamflowForecasting For DiyalaAnd Adhim Rivers Northern Iraq.

Streamflowforecasting is needed for proper water resources planning and management. Since The most challenging task for water resources engineers and managers is a streamflow forecasting. In this study a brief application and comparison of artificial neural networks approaches are employed for two case studies which were Diyala River .and Adhim River northern Iraq. Different training algorithms and different artificial neural networks such as LevenburgMarqudat LMNN , Scaled conjugate gradient SCGNN , radial basis function networks RBNN and generalized regression networks GRNN were selected in modelling and generation of synthetic streamflow for the mentioned case studies. The performance of applied networks were determined according to well known test parameters R2, Enash, Rbias ,MAPE, MAE. It has been found in this study that LevenburgMarqudat is faster and have better performance than Scaled conjugate gradient algorithm in training operation while the radial basis networks and generalized regression networks presented the best performance among all kinds of networks. [ABSTRACT FROM AUTHOR]