The coupled atmospheric -hydrological modeling system is constructed by WRF model, WRF -3DVar system, Hebei rainfall-runo￡f model and ARMA model in the Daqinghe basin. Three flood events occurred in 2012 and 2013 were chosen to test the system. The results show that assimilating radar reflectivity and GTS data can effectively improve the rainfall forecasts, which can reduce the flood forecasting errors of the system. The accuracy of the flood forecast can be further improved by the application of the ARMA model. Although the flood forecasting accuracy is reduced with the extension of the forecasting time, the application effect of the system performs well within 6h forecasting time. With the double correction mode (data assimilation and real -time adjustment method), the coupled atmospheric -hydrological modeling system can not only extend the forecasting time, but also has high accuracy of flood forecasting. The system will have a certain application in flood warning and forecasting.