Abstract: Forecast products, including rainfall, 2 m-temperature, temperature at 850 hPa, wind at 850 hPa and geopotential height at 500 hPa from WRF, which is a operational numerical model in central China, and GRAPES, which is the operational numerical model in China, from January to December of 2013, are examined using operational statistical verification methods. Results show that the prediction of all 24-hour forecast variables is better than those of the 48-hour. For rain/no-rain forecast, the TS score of GRAPES model was higher than that of WRF model; however, as precipitation intensity increases, the TS scores of WRF model become almost always higher than that of GRAPES. At the same time, the precipitation prediction range of WRF is significantly greater than the observed rainfall. The verification on 2 m-temperature found that the prediction by the WRF is better than that of GRAPES, and it is better in warm season than in cold season. The analysis of the atmospheric circulation field shows that the WRF model has a considerable advantage in the temperature and wind fields at 850 hPa. But there are some seasonal characteristics in the forecast errors of 500 hPa height field of WRF. The forecast of geopotential height by WRF is better for the summer half year than for the winter half year; the opposite is true for GRAPES model. In general, the operational numerical model in central China has some referential value for the weather forecast.