Wang et al. simulated CO₂ mole fractions over Western Europe in the summer of 2018 using the WRF-Chem model combined with three different CO₂ emission inventories. The simulations were evaluated by comparing with ground-based in situ and column observations. They showed, by taking into account the sector-specific vertical profiles of emissions, the agreement between the simulations and the observations was significantly improved for sites near large emission sources.

The topic of this manuscript is important and relevant to the scope of Atmospheric Chemistry and Physics. In addition, the analysis method is appropriate, and the writing structure is well organized. However, it seems to lack novelty and contains few new scientific findings. I recommend clarifying what is novel and addressing the following concerns and questions.

Specific comments

L56: Please provide a clear description of whether the signal is positive or negative.

L72: Do the multi-source observational data refer to ICOS and TCCON data? If so, I do not think they were used for simulating CO₂ mole fractions.

L242–247: Please add a discussion of how the overestimation in inland areas and the underestimation in coastal regions affected the mean bias error in wind speed.

L253: It would be easier to read if the expressions “between the observations and the simulations” and “between the simulations and the observations” were made consistent throughout this manuscript.

Figures 8 and 9: The differences between the WRF-GHG simulations and TCCON data in Figure 9 appear to be larger than those in Figure 8. Are the differences due to the fact that Figure 9 does not take into account smoothing using the column averaging kernel?

338–347: If the data period for Orleans was matched with that for Paris, would similar results be obtained? In other words, would the difference between Paris (an urban area) and Orleans (a suburban area) be reflected in the observed XCO₂?

L378: Please add an explanation of SNAP.

L382–383: In Figure 10, it appears that there are also large emission sources near sites other than KIT (i.e., CBW and SAC). What degree of closeness does “near” represent?

L419: including -> includes

L431–434: The simulated XCO₂ values in July and August 2018 were higher than the observed XCO₂. What caused this overestimation by the model? In years other than 2018 when no drought occurred, will the simulated XCO₂ values be lower than the observed values?

L454: analyze -> analyzing

L486–488: Please revise the sentence by adding a conjunction.

General comments:

This manuscript simulates CO₂ concentrations in Belgium and surrounding areas and conducts sensitivity tests using different emission inventories and in situ and remote-sensing CO₂ observations. The topic is interesting and meaningful, but many statements and explanations in the manuscripts are not rigorous enough. I suggest more modifications and improvements before acceptance.

Special comments:

• The title includes Western Europe, but this study only focuses on Belgium and the surrounding areas. Is it reasonable to use Western Europe in the title?
• This paper also mentioned that the drought could increase CO₂ I think it is necessary to include precipitation in the research period and compare it with the year before and after.
• I think Figures 1 and 2 can be combined.
• Lines 130-135: I suggest consistency in the parameters used in the equation and the text. For example, “T_scale” in Eq and “Tscale” in text. Maybe “T_scale” is more suitable. Also, other parameters such as Wscale, Pscale, Ts, Tmin, Tmax, …
• Lines 158-159, Here are two downscaling methods used. What are the differences between them?
• Table 1. I suggest adding some words in the column of the aberrative. Also, the CBW attitude is 0?
• Figures, the figures in the main text are not clear as the Figures in the Appendix.
• Figure 3. What are the sunrise and sunset times in these ICOS stations? The highest temperature occurred at 18:00 local time, and the PBL reached its maximum height between 15:00-17:00 (line 319). It seems unreasonable.
• Figure 5. I think it is better to keep the y-axis of CO₂ concentrations the same across different emission inventories at the same height.
• Figure 6. It seems there is no contribution from biomass burning in these figures. Why are biogenic contributions nearly negative from 10:00- 23:00 at all sites? Why are biogenic contributions only positive around 10:00?
• Figure 7. It is better to give the slope and correlation coefficient in Figure 7c for the three TCCON sites.
• Figure 8. Although the STD and RMSE increase from S to P, MBE becomes large. Which parameters are more critical to evaluate the model's sensitivity?
• Line 378. What does SNAP mean in this paper?
• Figure 11. There is a large gap between SNAP 1 for CAMS and EDGAR. Figure 10d also showed more emission sources than Figure 1a. Why did this gap occur? Were the emissions included in other emission sectors in CAMS?
• Figure 12. It seems that in late July and August, the land system was also active as a carbon source (Figure 12c), but anthropogenic emissions nearly disappeared from Figure 12b. Usually, drought can increase temperatures and the electricity demand for air conditioning, hence the anthropogenic emissions could increase in this period.
• Lines 438-439, please add ° before N and E for the GPS location. Also, add this to the GPS location in Table A3. What does “acid fen” mean here for FR_LGt?