Abstract:Based on the highresolution numerical forecast precipitation products from the ECMWF, JMA, T639, GERMAN, Hunan local fine grid HNGRIB from May to December of 2016, the forecast effectiveness of the models on the Hunan reservoir area rainfall is tested from the aspects of forecast lead time, areas, precipitation intensity grades, influence systems and forecast biases. The results show that the accuracy of various fine mesh models for area rainfall decreases generally with lead time, while the forecast effectiveness of ECMWF and JMA is good and stable, and the stability of the T639 is the worst. The area rainfall forecast effectiveness in nonflood season (October to December) is better than that in flood season (May to September). With the increase of the rainfall level, the prediction capability of each model decreases, especially the areal rainfall over 30 mm, is poor. The prediction capability of each model for vortex precipitation is the worst, but the best for upper trough precipitation. The area rainfall forecast of ECMWF, JMA and HNGRIB is smaller in various lead times, but those of GERMAN and T639 increase with the lead time.