The manuscript assesses the improvement in tropospheric ozone achieved through the assimilation of satellite observations of ozone and some of its precursors (NO2 and CO) in five reanalyses products compared to ozonesonde observations and satellite-based tropospheric ozone estimates. The study also distinguishes between the effects of directly assimilating ozone observations and assimilating only precursor observations. The main conclusion is that assimilation improved the consistency among the reanalyses products, though the impact varies among systems because of differences in the forecast model and assimilation configurations.

The study makes a significant contribution to the field by demonstrating both the capabilities and limitations of the current observing system in constraining the distribution of tropospheric ozone. By examining multiple reanalysis products, it goes beyond most previous studies that have often considered just a single product. The results are presented in a balanced manner, figures and tables are used where necessary, and the text is well-written. 

The following are a few aspects that could be discussed further to help improve the manuscript:

i) It would be helpful to clarify why none of the recently developed satellite-based tropospheric column products (eg. IASI-GOME2 product) were used for evaluation. While these products aren’t available for 2010 (the study’s evaluation period), the reanalysis datasets extend to more recent years.

ii) The analysis of the comparison of the reanalysis products with the OMI-MLS tropospheric column data could be expanded. In particular, it is interesting that assimilation introduces a bias in the models that was absent in the control runs without data assimilation (Figure 3). Also, the high biases seen in the comparison with OMI-MLS are not seen in the comparison with ozonesonde measurements.

iii) It would be useful to discuss how the loss of instruments like MLS, OMI, and MOPITT upon decommissioning of Aura and Terra might affect the quality of tropospheric ozone reanalysis datasets in the future.

iv) It would also be useful to discuss whether new types of observations (e.g. profiles in the UTLS) are needed to better constrain tropospheric ozone. Would continuing the current observing system be sufficient, provided that the information from these observations is fully utilized?

In “Assessing the relative impacts of satellite ozone and its precursor observations to improve global tropospheric ozone analysis using multiple chemical reanalysis systems,” Sekiya and coauthors compare 5 different global chemical reanalysis systems and the ozone in those systems against ground, ozonesonde, and space based observations. They evaluate assimilating total column ozone, ozone profiles, tropospheric ozone, and precursors of ozone, including total column CO, profiles of CO, and tropospheric NO₂ columns. For one system, TCR-2, authors test the impacts of different combinations of assimilated observations on the reanalysis. Authors conclude that UTLS ozone was most improved by assimilating total column and tropospheric ozone, while middle troposphere ozone was most improved by assimilating precursor observations. Lower troposphere ozone was improved by assimilating precursor observations. Across all reanalysis systems, the influence of assimilation on surface ozone was small, and in some cases made biases worse. Authors note that comparing across systems with different settings and assimilated species introduces uncertainties.

Overall, results will be of interest to the readers of ACP and the conclusions authors draw from their analysis will be informative for future reanalyses. The consistent evaluation against observations is a useful contribution. Some of the conclusions are not very clear, making it difficult to find the primary contributions of the work. The different settings in the reanalysis systems makes it difficult to draw specific conclusions in some cases, but this is probably unavoidable.

General comments

The structure of section 2.1 and its subsections makes it difficult to follow and therefore difficult to understand the differences between the reanalysis systems. Could authors adjust this section so that the same key words are used across the sub sections for key components? For example, it would be helpful to clearly mention the forward model, a priori emissions, and simulations settings used in each system using the same key words.

In the discussion, authors say that “Furthermore, the present study shows that the spread of data assimilation impacts among multiple systems can be used to evaluate whether the observing system impacts are dependent on the reanalysis system,” and that “These findings should lead to more robust assessments of the observing system impacts.” It would be helpful to clearly summarize and/or enumerate what those findings are, and specifically how the current study shows whether the observing system impacts are dependent on the reanalysis system.

Ozone exhibits seasonal biases in many modeling systems, but here authors present annual mean results only. It would be ideal to present seasonal results to understand the impact of different assimilated observations in different seasons, but that may be out of the scope of this analysis. At a minimum, authors should mention this.

In all figures besides 6 and 7, the fonts in the colorbars and legends are too small and are unreadable. Please make the fonts larger. In the ozonesonde plots, the lines are too small to distinguish colors and line types. Please make the lines larger and more easy to distinguish.

Specific comments

Line 4-5: This sentence is hard to follow, please adjust for clarity.

Line 10: What values are these percentages referring to?

Lines 32-34: This is a very general statement that I’m not sure can be attributed only to the review paper cited here. Consider rewording.

Line 226: How was this regridding done? It seems import enough to mention this earlier and briefly describe how it was done.

Section 3.1: It seems worth noting that GEOS-Chem control minus reanalysis is essentially zero, different from other systems, and describing why there is essentially no change in the GEOS-Chem reanalysis.

Line 267: “differences in procedures for computing TOC” – how different are the procedures between the models, and roughly how much difference do you expect this to contribute?

Line 268-271: The discussion about positive bias in or against OMI-MLS profiles is confusing and hard to follow, please consider rewording.

Lines 349-351: This sentence is difficult to follow, what are the numbers referring to?

Line 413: Wouldn’t chemical systems be stiff, and more likely to be numerically “unstable” (rather than “stable” as the authors write)? Perhaps I am misunderstanding, and authors can clarify in that case

Line 442-443: It is unclear what exactly is being reduced and by how much, please adjust for clarity

Line 444-445: My understanding is that Turnock et al. is referring to global climate chemistry models which can have more simple chemistry than the global models used here, so the same biases may not be present. Also, according to Figure 5, there is not a positive bias across the board. Can authors please adjust their wording or correct me if I am wrong on this point?

Lines 448-449: But the GEOS-Chem analysis showed almost zero influence on ozone, how does that square with this conclusion? As authors say, the impact of the measurements varied widely, so it’s not convincing to simply state it was important without any qualifying statements.

Technical corrections

Line 191: typo in “difficultiy”

Data availability statement has typo.