A hybrid form of rainfall-runoff models integrating artificial neural networks(ANNs)with conceptual models is proposed.The integrated model is a semi-distributed form of conceptual rainfall-runoff models

in consideration of the spatial variation of rainfall

the heterogeneity of watershed characteristics and their impacts on runoff.Genetic algorithm is used to optimize the parameters of the conceptual model and GIS software and DEM data are used to divide the whole catchment into sub-catchments based on the spatial distribution of rain-gage stations.As a result

in each sub-catchment

runoff generation is simulated in consideration of the spatially distributed model parameters and rainfall inputs.In runoff routing

instead of a linear superposition of routed runoff from all sub-catchments as traditionally performed in a semi-distributed form of conceptual models

artificial neural networks as an effective tool in nonlinear mapping are employed to estimate runoff.The feasibility of the new approach is demonstrated in Dapoling watershed

the upper tributary of Huaihe River basin

and the results of coupling model are compared with those of the Xinanjiang model.Verified results of the model indicate that the approach integrating artificial neural networks with conceptual models presented in this paper can achieve the promising results with acceptable accuracy for flood event simulation and forecast.