Two coastal heavy rain events occurred in 4-5 and 21-22 August 2010 were caused by the subtropical high. For several kinds of numerical products that are used widely, verification is conducted mainly from the aspects of circulation situation, influencing systems, and precipitation elements. The results show that these numerical products have certain prediction capability for heavy rainfall quantitatively, but predicted rainfall levels are generally smaller, and the prediction of the strong precipitation center is poor in stability; the position forecasts of EC and T639 models are more accurate, but the intensity forecasts are smaller than the actual situation. There are certain differences between numerical models and the actual situation in the position and intensity of the weather systems producing the heavy rainfall. As for the West Pacific subtropical high, the 72 h forecast of the T639 model is weaker than the actual storm, and the stability of EC model is higher, but there is certain deviation in upper trough and shear line forecasts. In torrential rain forecasting, it is necessary, on the basis of numerical forecast products, to combine strong convective products with such auxiliary tools as intensive observation data, physical fields, similar examples, experienced empirical extrapolation, etc., to improve the forecast accuracy of heavy rain forecasts.