Review of « Vertical structure of a springtime smoky and humid troposphere over the Southeast Atlantic from aircraft and reanalysis ».

I find that this is a good and well written that can be published after only minor corrections.

Page 5 line 23 : What does « incorporated » mean in this context? Is it assimilation ?

Page 8 : Was there some space-time interpolation needed to get the reanalysis on the same measurement locations as the ORACLES ? If so, what was it ?  The co-location part of the method is a bit unclear to me.

Page 8,  line 20 : reference needed for underestimation of CO by CAMS (e.g. Inness et al 2019) or other

Fig2 caption:  change scatters to scatterplots

Table 1 : What accounts for the improvement in water vapor correlation from 2016-2018 in CAMS  but at the same time, a decrease in CO correlation ?  The q vs q correlation suggest that the dynamics and physics are working well in ERA5/CAMS but something has gone wrong with the CAMS CO. The correlation R^2 has fallen and is possibly not even significant. I think you need some significance test on these correlations. What is the explanation for this fall and for the low correlation (0.55). It there a problem with the long-range transport in CAMS. Maybe a contingency table would be helpful to see if the plumes were captured but at the wrong location or the wrong time, or where nothing was seen at all with CAMS. 

Page 17, line 18 : Least-> lowest

Page 20, line 2 : the most humid profiles (q1 and q2) are the most frequent

Page 20, Line32 : it would be clearer to write :  the air at the altitude of the jet appears to originate from higher altitudes to the north of the jet more frequently that from the continent  directly to the east.

Page 22 line 20 : The backgound/smokeless case is probably not 100% smokeless and therefore you are capturing a real increase at altitude due to the sampling of long-range transport of plumes.  In the free troposphere there can be several layers with higher or lower mixing ratios of CO. 

Page 22, line 27 : with only a slight decrease in magnitude – does « magnitude » here refer to the magnitude of the discrepancy ?

Page 22 line 29, I thought profile 6 was showing an increase in CO with altitude.  Isn’t it unlikely that it is a constant-CO atmospheric structure so there are always small increases throughout the free troposphere  depending on the transport of different airmasses in different layers. 

Page 23, Line 24 : how would the frequency of the humidity profiles have any effect on the amount of smoke available for transport ?

Page 23, Line 25 : The biomass burning plume, (which one) ? Or biomass burning plumes (in general) ?

Page 24, Line 25 : Please rewrite this sentence: This does not mean that this more remote region is not influenced by the biomass burning plume (indeed, this sector contains St Helena Island, whose smoke and humidity vertical structure are studied by Adebiyi et al. (2015) and others, and the Ascension Island observatory is located to the west of the northwesternmost sector we consider, i.e. even more removed from the source); but the smoke there is more episodic.  

Page 28 : I do not see the link between the first paragraph of this discussion and the results presented in the article. You need to integrate or link the discussion with the results section to make this part relevant to what we have just been reading about, otherwise it seems like an afterthought.

Review on preprint “Vertical structure of a springtime smoky and humid troposphere over the Southeast Atlantic from aircraft and reanalysis” by Pistone et al.

This manuscript proposes a climatological characterization of water vapor and biomass burning plumes in the south-east Atlantic area, with the overarching aim of gaining a better understanding of the radiative impact of aerosols. Therefore, the authors rely on model reanalyses, constrained by data assimilation (ERA5 and CAMS). As observations and simulations of aerosol concentrations are difficult to exploit due to higher uncertainties, the diversity of the compounds making up aerosols and the complexity of the physical parameters influencing AOD (species involved, optical properties but also aerosol size distribution, mixing state…), the authors use CO concentrations as a proxy for the signature of fire plumes.

To support this choice, the authors first demonstrate the ability of the models used to reproduce water vapor, as well as the general structures of CO and AOD, and then demonstrate the good correlation between AOD and the total CO column in observations and simulations. Using k-means clustering of the CO and q profiles mapped onto their 2 principal components, the authors then show that the variability is well represented by 6 clusters of q and 6 clusters of CO. Finally, the number of profiles in each cluster for each month of the studied time period is analyzed.

This manuscript reports on a milestone in a wider study, since it does not go as far as analyzing the radiative impact of aerosols for the identified typical situations. Nevertheless, the work is original, rigorous and well presented. Most of the figures are very clear and the text well written.

I recommend publication, and suggest only minor revisions, listed below.

Minor comments:

Abstract.: it would be helpful to have a summary of the main features identified in the clusters.

Section 2:

• The different sections should include some discussion on the uncertainties for each dataset (with numbers).
• For the reanalyses: I think CAMS also assimilates IASI CO observations. I would also be important to detail the biomass burning emission inventory used in the simulations, as well as the potential corrections. I think that the GFAS inventory is used. In the reference paper, Kaiser et al. (2012) recommend applying a factor of 3.4 on BB emissions for aerosols. Is it the case in the reanalyses used here? This could partly explain why the underestimate in AOD is lower than the underestimate in CO mentioned in Section 3.1.

Kaiser, J. W., Heil, A., Andreae, M. O., Benedetti, A., Chubarova, N., Jones, L., Morcrette, J.-J., Razinger, M., Schultz, M. G., Suttie, M., and van der Werf, G. R.: Biomass burning emissions estimated with a global fire assimilation system based on observed fire radiative power, Biogeosciences, 9, 527–554, https://doi.org/10.5194/bg-9-527-2012, 2012.

Section 3:

• For model evaluations, it would be helpful to provide the statistics of the comparisons in terms of mean relative bias for q and CO in Table 1 (as for AOD in Table 2).
• From Figures 2-4, it is not clear to me that CAMS only underestimates large CO values (authors mention values > 300ppbv). Showing the mean simulated and observed profiles, and the corresponding mean bias, would be interesting. Also, models tend to simulate wider plumes due to too large dispersion. Could this explain an underestimate of maximum values (in addition to the dilution due to the resolution), and an overestimate of background values in some areas? A general discussion of the CAMS model performance in simulating the LRT of BB plumes in the literature would be helpful (could also be in section 2).

Section 3.1:

• Would it be possible to calculate the relative aerosol mass below the aircraft to evaluate the possible overestimate in AOD if the full column is considered?
• Among the uncertainties in the simulated AOD, the authors could also mention the size distribution and the aerosol mixing state…
• Regarding the impact of hygroscopicity, what would be the associated variation in AOD? As this been evaluated in CAMS?

Section 4:

• The figures in the supplementary material really help to understand the method, but the legends need to be revised as they are far too vague and not self-explanatory (especially figures S3-S8). It seems to me that the principal components themselves are not shown (cf. l.12, p.17).
• I am surprised that the shape of CO profile is very similar for all clusters, unlike that of water vapor, and I was wondering if similar profile shapes were observed during the ORACLES campaign. Could this shape be due to the data assimilation using satellite instruments that lack vertical sensitivity?
• Do the CO profiles corresponding to the cluster q4 also exhibit a signature from a transport of air masses from another area (linked to the dry intrusion above the PBL)?
• p.22, the authors discuss the cluster C6 and a possible overestimate in background values: cf previous comment for Section 2.
• p.23: l. 1-2: the authors state that the profiles from the k-means is consistent with the observations. I may have missed the information but I think it would be useful to include more precise comparisons in terms of profile shape, for CO in particular since uncertainties in the simulations are larger. This could be essential for later use in a radiative transfer model assessing the radiative impact of aerosols.
• The authors mention that clusters corresponding to lower CO concentrations include both background profiles and more episodic transport (p.23). Would the number of clusters help differentiate profiles with no BB signature and those with a smaller plume?
• The discussion of the temporal variability of the number of profiles in each cluster could be linked more explicitly to the variability in fire activity.

Discussion:

p.28, l. 23-24: The authors state that the case studies of Cochrane et al. 2022 are included in the climatology presented in the paper. To what clusters do they correspond?

Technical corrections:

• p.5, l. 6: scattering aerosol extinction (no need for a capital A)
• Section 3: only a subsection 3.1 for aerosols. Maybe include another subsection for CO comparisons?
• p.8, l.24: missing verb?
• Supplemental figures need more precise legends.