Study of early flood warning based on postprocessed predicted precipitation and Xinanjiang model

Precipitation is the most common cause of flood. Accurate precipitation prediction is therefore important for flood forecasting and can be a key factor that increases lead time and the accuracy of early flood warning. In this study, the reforecast precipitation of the global ensemble forecast system (GEFS) was postprocessed using CSG EMOS (censored and shifted gamma distribution-based ensemble model output statistics) method to improve its reliability, and then used as the forcing data for Xinanjiang model to increase the lead time of flood forecasts in order to provide a more effective early flood warning based on an empirically water level–discharge curve at Wangjiaba section, Huaihe River basin, China. Three scenarios were set to demonstrate the importance of precipitation prediction in flood forecasting. The results showed that predicted precipitation became more reliable after postprocessing and this improvement increased as lead time expanded. It is also demonstrated that the postprocessed predicted precipitation brings the improvement for flood forecasting, then leads to the gain of early flood warning. However, this improvement becomes less significant by the increase of lead time and fades away when lead time reaches 7 d. In addition, the results of flood forecast and early warning in predicted precipitation scenario were not as good as those in observed precipitation scenario, indicating that substitution of predicted rainfall for observation requires further refinement in future.