Based on the low resolution version of RMAPS-NM forecast system and WRFDA-FSO diagnostic tool, this paper evaluates the impact of existing radiosonde and surface observation on RMAPS-NM forecast system in July 2021. This method is relatively cheap in computation, and allows the observation impact to be divided to the subset of observation according to observation variables, observation types, barometric levels, geographical regions, etc. The cost function is the difference between the prediction error of the background field and the analysis field measured by the total dry energy. The results show that the total sum of observation impact is negative, and observation plays a positive role in prediction. The observation that contributes most to the reduction of 12 h prediction error comes from the dynamic variables (U, V wind components) of radiosonde observation. However, the contribution of radiosonde observation to the average observation impact per unit quantity of a single time is about 1/2 of that of surface observation. The radiosonde observation has a positive contribution to the reduction of 12 h prediction error from the near surface layer to the top of model layer, and there are two maximum zones in the middle and lower troposphere and in the troposphere upper jet layer. The positive contribution of surface observation is obvious in the lower layer below 850 hPa. The radiosonde observation, when assimilated by the assimilation system, has a favorable influence overall, which also reflects the characteristics of stable and high-quality characteristics to radiosonde observation. The zone with the most times of positive contribution for surface observation to the reduction of 12 h prediction error is particularly significant in the Hetao Region. At the same time, the problem that the assimilation rate of surface observation data should be further improved is discussed.