General comments:

Comparing and Evaluating reanalyses for atmospheric composition is an effort which, to my best knowledge,  has not been done so far on such a comprehensive data basis. The authors pick up tropospheric ozone, probably the species of broadest interest for air quality and climate.  While the complex interplay between atmospheric chemistry and transport, together with uncertainties in sources and sinks, makes it difficult to model this species adequately on all scales, different kind of satellite observations can be assimilated to relax the model results to observational evidence. Here, three global reanalyses and two regional reanalyses are intercompared and evaluated with independent observations from surface stations and ozonesondes. The study is generally well written, touches on a topic highly relevant for ACP and should be published with minor modifications. In the following, I suggest to take a few points into consideration:

• The underlying model basically differ in their chemical mechanisms, their emissions, and in their spatial resolution and extent. Some information is still missing or not well referenced in the model descriptions. It will certainly be helpful for the reader if such model basics can be summarized in a table, together with data assimilation specifics as assimilated species and satellite instruments considered.
• It would be intriguing to know more about the influence of the different chemical mechanisms and emission inventories. A few plots showing the most important precursor emissions (or some statistics) could shed light on the uncertainties introduced here. Ideally, one would also compare model results without data assimilation, but this will need additional model simulations which may be out of the scope of this manuscript.
• Unfortunately, the regional reanalyses provide only surface data. It should be feasible to include the necessary data for model evaluation also on 500 and 250 hPa for these models.

Specific comments:

Lines 17-20: Please state here explicitly that you are using global as well as regional reanalyses.

Lines 34-41: You should spend a few more sentences on the processes regulating tropospheric ozone on different scales, particularly the complex interplay between VOC and NOx emission for ozone production, major sinks of ozone, and the role of long-range transport and Stratosphere-Troposphere exchange. Not all of these factors can be constrained by observations but need to be addressed by the chemical mechanisms and transport peculiarities of the models used. For further reading, you could refer to, e.g., Monks et al., 2015.

Line 39: Reference “Chang et al., 2017” is missing.

Line 43: Reference “EPA, 2024” is missing.

Lines 66-72: The influence of STE on tropospheric ozone is still under debate and variates considerably between models, as stated in, e.g., Young et al. (2018) and Griffiths et al. (2021). Please reflect here latest reviews on the topic and mention other potential sources in the upper troposphere as lightning NOx.

Line 73: “PBL” is not used furtheron in the manuscript and can be omitted.

Lines 79-80: Give some references here, like Inness et al. (2013).

Lines 84-86: All five references are missing.

Line 91: Reference “Jiang et al., 2017” is missing. Should it be “Jiang et al., 2018”?

Line 92: Reference “Cady-Pereira at al., 2017” is missing.

Line 99: You could refer here to https://igacproject.org/activities/TOAR/TOAR-II .

Line 100 and Table 1: CONUS is first mentioned here and needs to be explained (instead of line 173).

Line 158: You could refer to https://gmao.gsfc.nasa.gov/reanalysis/MERRA-2/ for MERRA-2.

Line 159: Please give numbers for species and reactions of the chemical mechanism.

Lines 163-169: Interestingly, GEOS-Chem assimilates NO2 only. Can you give a few details on the decision not to assimilate further species?

Line 178: Please give numbers for species and reactions of the chemical mechanism.

Line 179: You could refer to https://www.epa.gov/air-emissions-inventories/2014-national-emissions-inventory-nei-data .

Line 180: Citation for CMAQ-SMOKE: de Almeida Albuquerque et al., (2018).

Line 183: Please give a reference for BEIS.

Lines 184-190: Again, it will be interesting to get information on the choice of assimilation species.

Line 203: I don’t understand the term “inflation factor” in the context of data assimilation. Please explain shortly its relevance or omit.

Line 230: Replace “northern India” by “Tibetan plateau”.

Line 243: The NH maximum extends over large areas of subtropics and mid latitudes.

Line 244: The fact that only surface fields are available from the regional reanalyses is important for the soundness of the study and should be stated much earlier. What is the reason for not providing at least two additional levels at 500 and 250 hPa from these data sets?

Lines 246-247: I see differences of more than 5 nmol mol-1 for both reanalyses, GEOS-Chem and TCR2. Please revise. Replace “mean differences” in line 247 by “major differences”?

Lines 278-281: Be more careful in your description of Fig. 5. TCR2 is higher over South America than GEOS-Chem in all months except July. CMAQ is higher over the US than GEOS-Chem from November to May. Also, TCR2 is higher than GEOS-Chem for parts of the year.

Lines 283-286: I see discrepancies in all tropical regions. Please revise description of Fig. 5.

Lines 289-290: Same holds true for China.

Lines 306-308: I can’t see from Fig. 6 that the variability of GEOS-Chem is much different from that of the other reanalyses. There are some shifts between the curves and also differences in amplitude and variability from month to month, but without any preference for one model. How can you say that GEOS-Chem fails to reproduce year-to-year variability without knowing the ground truth?

Lines 309-316: Are the reported trends considered to be significant in a statistical sense? If yes, please mark the respective areas in Fig. 7. If not, please omit the whole paragraph including Fig. 7.

Line 325: Can you add a Figure or Table providing information on the surface stations, their location and observation statistics?

Line 326: I would not use “global bias” here, as the surface observation do not have global coverage. Rather use “average bias” here.

Line 335: How many ozonesondes are used for each latitude band? Are all seasons well covered everywhere?

Lines 338-339: There is no further discussion of single ozonesonde stations. The sentence as well as Figures S6 and S7 could be omitted.

Line 384: Reference “Sekiya et al., 2021” is missing.

Lines 406-408: I guess all that what is currently possible to produce background error covariances is already being done. How do you envisage to produce better background statistics without having more or better observational data?

Line 418: Can you give a reference for TEMPO?

Line 420: Both references are missing.

Lines 440-441: Can you give an explanation for the shift in maximum ozone over the US between the models?

Line 459: Don’t forget to mention the different emission inventories here.

Line 475: Please point here to the exact data location, which is: https://ads.atmosphere.copernicus.eu/datasets/cams-global-reanalysis-eac4

Line 651: Reference “Miyazaki et al., 2021” is not used in the manuscript and should be omitted.

Figures 5, 6, 11: Use of  “mixing ratio” will be more precise than “concentration”.

Fig. 8 / Fig. S5: Please move the information on the regridding from Fig. S6 to Fig. 8.

Technical comments:

Line 54: “)” is missing after references.

Line 175: Replace “reanalyses” by “reanalysis”.

Line 176: Replace “… is based the …” by “… is based on the …”.

Line 238: I guess you mean Fig. 1b here instead of Fig. 2.

Line 266: Replace “Flemming et al., 2018” by “Fleming et al., 2018”.

Lines 322-323: Replace “… observations TOAR-1 …” by “… observations in the TOAR-1 …”.

Line 401: Replace “… compared jointly …” by “… compared to jointly …”.

Line 409: Replace “… factor influences …” by “… factor which influences …”.

Line 477: Link is broken.

Figures 10, 11: Please use “nmol mol^-1 “ as units.

References:

de Almeida Albuquerque, T.T., de Fátima Andrade, M., Ynoue, R.Y. et al. WRF-SMOKE-CMAQ modeling system for air quality evaluation in São Paulo megacity with a 2008 experimental campaign data. Environ Sci Pollut Res 25, 36555–36569 (2018). https://doi.org/10.1007/s11356-018-3583-9

Griffiths, P. T., Murray, L. T., Zeng, G., Shin, Y. M., Abraham, N. L., Archibald, A. T., Deushi, M., Emmons, L. K., Galbally, I. E., Hassler, B., Horowitz, L. W., Keeble, J., Liu, J., Moeini, O., Naik, V., O'Connor, F. M., Oshima, N., Tarasick, D., Tilmes, S., Turnock, S. T., Wild, O., Young, P. J., and Zanis, P.: Tropospheric ozone in CMIP6 simulations, Atmos. Chem. Phys., 21, 4187–4218, https://doi.org/10.5194/acp-21-4187-2021, 2021

Monks, P. S., Archibald, A. T., Colette, A., Cooper, O., Coyle, M., Derwent, R., Fowler, D., Granier, C., Law, K. S., Mills, G. E., Stevenson, D. S., Tarasova, O., Thouret, V., von Schneidemesser, E., Sommariva, R., Wild, O., and Williams, M. L.: Tropospheric ozone and its precursors from the urban to the global scale from air quality to short-lived climate forcer, Atmos. Chem. Phys., 15, 8889–8973, https://doi.org/10.5194/acp-15-8889-2015, 2015.

Young, P. J., Naik, V., Fiore, A. M., Gaudel, A., Guo, J., Lin, M. Y., Neu, J. L., Parrish, D. D., Rieder, H. E., Schnell, J. L., Tilmes, S., Wild, O., Zhang, L., Ziemke, J. R., Brandt, J., Delcloo, A., Doherty, R. M., Geels, C., Hegglin, M. I., Hu, L., Im, U., Kumar, R., Luhar, A., Murray, L., Plummer, D., Rodriguez, J., Saiz-Lopez, A., Schultz, M. G., Woodhouse, M. T., and Zeng, G.: Tropospheric Ozone Assessment Report: Assessment of global-scale model performance for global and regional ozone distributions, variability, and trends, Elementa, 6, 10, https://doi.org/10.1525/elementa.265, 2018.

General comments

The manuscript provides an evaluation of the potential utility of 5 chemical reanalyses (3 global, 2 regional) for quantifying regional and interannual variations in tropospheric ozone. Therefore, those reanalyses are intercompared and compared with surface ozone and free-tropospheric ozone measurements in terms of their climatological ozone distribution and regional ozone variations.

The manuscript certainly provides very useful information for a reader who wants to have an idea about the current state of tropospheric ozone representation by chemical reanalyses. The manuscript gives a nice overview of the differences between the chemical reanalyses and independent observations, but, unfortunately, the reasons and explanations behind those differences are often not mentioned. A lot of observations are being made, but much less interpretations are shared. When reading the manuscript, I often asked myself “why?”, but in many cases, I did not find the answers in the manuscript. The authors should give more insight in those findings, or simply state that the reason is unclear or not known.

Re-reading the summary, you’ll find that only in lines 451-455, an interpretation of the results is given, together with some general statements in lines 458-459.

Specific comments

• Line 54: add “)” after Cuesta et al., 2013
• Lines 58-74: very nice introduction of the current state of atmospheric chemistry models in terms of representing the tropospheric ozone field. I would expect that the current manuscript provides more insight on the contribution of the different listed “model issues” that explain tropospheric ozone discrepancies between models and observations. A part of this introduction has been retaken in the Discussion section, but rather in (too) general terms. Were you not able to shed more light on the “model issues” responsible for discrepancies between models and with observations with the current study?
• Table 1: I would include two additional columns here: one describing the emission databases, and one mentioning all the assimilated species (+satellites?) in the reanalysis system. It should also be clearly mentioned in this table that the latter two only provide surface ozone output, and not tropospheric ozone.
• Line 221: regridded instead of regrided.
• Lines 230-241 (description of Fig. 1b): WHY? And also, why are the GEOS-Chem biases w.r.t. the ensemble mean higher over the continents than over the oceans? What explains the CAMSRA high bias band over the high latitudes w.r.t. the ensemble mean?
• Lines 243-244: This “note” should have already been mentioned in Section 2!
• Line 246: “the reanalyses are all closer to the mean in the middle troposphere”. Closer compared to what? Compared to the situation at the surface (Fig. 1a)? Wouldn’t it be recommended to use relative comparisons (percentages of mean biases) instead?
• A comment for the entire manuscript: have the authors not considered to also report relative biases (percentage differences), complementary to the absolute biases, in their comparisons between models and with observations? Why not?
• Lines 248-250: what is reason why TCR-2 and GEOS-Chem are exhibiting an opposite pattern of differences w.r.t. the ensemble mean in the middle tropospheric mean ozone distribution?
• Lines 260-264: explanations? Here, or in the discussion…
• Define MDA8 here again (it has been defined earlier in line 70, but only in the context of a study cited in the introduction, so it does not harm to define it again here). And also better argue why you find it important to consider this ozone metric in the comparison as well, next to the mean surface ozone metric.
• Line 275: how have those nine regions been defined? Based on which criteria?
• 3.2: looking at the plots of the seasonal variation, I would start the discussion with describing if the different reanalyses show a similar shape of the seasonal cycle, rather than with the differences in the monthly ozone mixing ratios (e.g. lines 276-281), which is (at least partly) covered by the previous figures 1-4 and S1-S3. To me, the most striking feature from Fig 5 is the quite different shape in seasonal cycle at most regions of GEOS-Chem compared to the other reanalyses. Explain this.
• Lines 291-292: Finally, some explanation, but only of some specific features (Lines 289-290). I would expect more of these explanations throughout the text.
• Lines 293-301: The discussion of the time series of the regional mean ozone concentrations in Fig. 6 again starts with comparing the mean ozone concentrations, whereas I found especially the differences in the amplitudes of the seasonal cycle, for especially the surface ozone levels, most striking. Maybe you can first highlight this and try to come up with an explanation for this feature.
• Lines 301-308: Why is the worst agreement found over Northern Africa (Fig. 6b and 6c)? Why does TCR-2 significantly overestimate the ozone minima? Explain why the simulated variability is most different in GEOS-Chem.
• Lines 309-316: the authors might add which (sign of) trends are expected, based on observations, in the different regions, rather than summing up the trend differences calculated from the different reanalyses for the different regions.
• Section 3.3: should the discussion in this section not benefit from including the relative (percentage) mean differences, especially if you compare the mean biases between different seasons (e.g. lines 330-334), and at different latitudes (e.g. lines 342-351).
• Fig. 9: perhaps add the upper and lower bounds of the found correlation coefficients for the different latitude bands and reanalysis, so that the reader gets an idea about their variability.
• Section 3.4: it is a very nice add-on that you also look at the surface NO2 distribution, but rather it should be better explained what the figures 10 and 11 can learn us (or not) to interpret the surface ozone differences between the reanalyses, as shown in figures 1 and 5.
• Section 4 (Discussion): very general explanations (model resolution, different chemistry, assimilation configuration) are given, often citing from literature, to explain (surface) ozone differences between the different reanalyses. A direct linking to many of the in great detail reported discrepancies between the reanalyses earlier in the paper is missing (there and in this discussion section). This is somewhat disappointing for the reader. It might be that the authors do not have the information or knowledge to dig deeper into those discrepancies, because many contributing factors intervene here. If so, please state this and be clearer on that.
• Section 5 (Summary): as already mentioned at the end of my general comments, there is very little interpretation included in the summary.
• Data availability section: please, also mention where the TOAR-I surface ozone data and the used ozonesonde can be obtained.
• Supplementary Figures S6: What time series are represented here? Monthly means? Or have the reanalyses been sampled to the ozonesonde observations? Please specify! How have the ozonesonde graphs been ordered? Alphabetically? Why not e.g. with decreasing latitude? It also seems that there are some outliers in the ozonesonde observations (e.g. at Fiji, Lauder, Paramaribo, Resolute, Samao, Scoresbysund, Sodankyla, South Pole) that possibly might be removed to make the comparisons clearer.