Abstract: Based on global numerical forecast model, 2170 surface automatic weather stations in China were assimilated by using Grid point statistical system (GSI) and a new model initial field was established. The characteristics of the three main meteorological elements in background field and initial field were compared with ERA5 reanalysis data over China, respectively. The results show that the mean deviation, root-mean-square error (RMSE) and cost function of the model are significantly reduced after the assimilation of surface automatic weather station observations in China, which indicates that the assimilation of ground observation station data can effectively reduce the errors of meteorological elements such as temperature, air pressure, and wind speed in the background field of the model, so that the initial analysis field of the model is more consistent with the observation field. In China, the prediction effect of atmospheric pressure is the best among the three meteorological elements. The correlation coefficients of atmospheric pressure in seven regions are all above 0.94, and the correlation coefficients of seven regions are improved after assimilation, and the correlation coefficients of Northeast and East China are as high as 0.99. The correlation coefficient of temperature in different regions increased slightly while the RMSE decreased after assimilation, especially in South China. Compared with atmospheric pressure and air temperature, the meridional wind and zonal wind improved less after assimilation, and the correlation coefficient with the reanalysis data is small while the RMSE is large. The correlation coefficient of meridional wind speed in East China, Southwest China and Central China is lower than 0.5 while the RMSE of meridional wind and zonal wind speed in northeast China are both greater than 5 m s⁻¹.