Abstract: Based on the global ensemble forecast data of EC and NCEP in TIGGE database from January 1, 2016 to December 31, 2017, the 2 m air temperature and 24 h accumulated rain forecasts within 10 days of the two systems in southwest China are verified and analyzed. The 2 m air temperature verification results show that the root mean square error of control forecast and ensemble mean of EC_GEPS and NCEP_GEPS are generally very high, especially higher for EC_GEPS. For each forecast lead time, the root mean square error of ensemble mean forecast is basically equal to that of the control forecast, and the ensemble spread is significantly low. The Talagrand distributions all showed a "J" distribution characteristics. The outlier score is generally high, and the outlier score of EC_GEPS is significantly lower than that of NCEP_GEPS. From the minimum ensemble forecast to the maximum ensemble forecast, with the increase of the ensemble percentiles, the root mean square error of each forecast lead time decreases gradually, and the root mean square error of the maximum ensemble forecast is the lowest. The 2 m air temperature verification results show that the Talagrand distribution of 24 h accumulated rain for each forecast lead time of EC_GEPS and NCEP_GEPS generally show a certain "L" distribution characteristics, which is more obvious for NCEP_GEPS. The outlier scores of NCEP_GEPS for each forecast lead time are generally high, and much higher than those of EC_GEPS. The rainfall ensemble prediction skill of EC_GEPS is much better than that of NCEP_GEPS. The prediction skill of 70% ensemble percentile forecast of EC_GEPS is optimal among all ensemble percentile forecasts, and the prediction skill of 80% ensemble percentile forecast of NCEP_GEPS is optimal among all ensemble percentile forecasts. Both EC_GEPS and NCEP_GEPS have some systematic errors in the 2 m air temperature forecast and precipitation forecast in Southwest China. Therefore, conducting corresponding systematic error correction for the ensemble forecast should improve the forecast technique.