The forecast of a hydrological time series has been one of the most complicated tasks due to wide range of data, uncertainties in the parameters influencing it and lack of adequate data. Feed forward multilayer neural networks are widely used as predictors in aquatic fields of applications. The paper demonstrates the potential of artificial neural network (ANN) models for simulating and forecasting daily river stages at Pandu, Pancharatna and Dhubri gauging sites of the river Brahmaputra. Back propagation method has been chosen. The main purpose has been the modelling of the river Brahmaputra. in àmulti gauging station catchment and real time forecast of daily stage downstream. Model efficiency of all the networks override the conceptual idea of use of the ANN has been clearly articulated.