This paper addresses important issues for understanding the evolution of ozone pollution over China and how it is responding to air quality regulations. This is a complicated problem so although this study does not resolve discrepancies between data sets, it is a careful analysis and represents an advancement in our understanding of Chinese ozone pollution. I recommend publication once minor issues are addressed.

Specific comments:

p5 line 8: I could not find references at https://quotsoft.net/air/ regarding the accuracy and precision of this data. Please provide more information and/or other references that use the data.

p7 line 1: Zheng et al., 2018 uses the MEIC inventory – I assume that also used here? This should be stated explicitly since there are other bottom-up inventories for China, e.g., as mentioned in Zheng, B., et al., (2018), Rapid decline in carbon monoxide emissions and export from East Asia between years 2005 and 2016, Environ. Res. Lett., 13(4), 044007, doi:10.1088/1748-9326/aab2b3.

p8 line 12: Could the lower bias and degraded correlation with IASI-IAGOS compared to IASI-ozonesondes also be related to the assumption of a vertical profile at the lat/lon of the mid-point in a slanted profile? This might explain the low bias in the lowest layer since those measurements would be taken closer to urban areas near the airport.

Fig. A1: In addition to the 2014-2017 mean and statistics, it would be useful to see the trends in ground station observations compared to the model surface ozone.

Wording change suggestions:

p2 line 15: “have been applying” to “have been enacted”

p3 line 3: “In this study, we question the ability…” to “In this study, we examine the ability…”

p3 line 8: “Results are also discussed in light with the TOAR outcomes” to “Results are also contrasted with the Gaudel et al., (2018) TOAR outcomes”

p4 line 28: Since validation references are not given in this section, I suggest adding a line: “Initial validation of the KOPRAFIT IASI ozone retrievals with ozonesonde and IAGOS data is presented in Section 3.”

p8 line 14: “worser” to “lower”

p9 line 23: “less than 300 data” to “less than 300 profiles”?

p11 line 19: “desertic” to “desert”

p11 line 21: “This region should not be considered here.” to “This region is not considered here.”

p11 line 24: “This translates even stronger to the model” Not quite sure what is meant by this– maybe “This feature becomes even stronger in the model when AKs are applied”?

p17 line 2: “nuding” to “nudging”

p19 line 20: “These results seem to comfort the consistency” to “These results corroborate the consistency”

p19 line 25: “and the caution to take to not overinterpret the results” to “and the need for caution to prevent overinterpreting the results.

p20 line 12: “Some individual studies exist but once again they do not allow one to conclude” Conclude what?

p21 line 23: “leads to reduce” to “further reduces”

This paper reports ozone trends in the free troposphere over China using the IASI satellite instrument and interprets them using an atmospheric transport model. I have a few comments that the authors should consider when they revise their manuscript. One major comment is associated with the authors not using averaging kernels when they compare their model to the IASI data.

Second paragraph: this area of the world is attracting a lot of attention. Already since this paper was submitted there are a few more papers that need citing in the introduction, e.g. https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2021GL092816?af=R

Page 4, Line 17: Three different a priori and constraints….ambiguous.

American spelling: use of centre for ECMWF and elsewhere. Vapor instead of vapour. 

Section 2.3. There is little in the way of data description for the surface measurements. What is this website? How are the measurements collected? Uncertainties? What are criteria pollutants? 

Page 5, Line 16: ...the larger the amplitude of the diurnal ozone cycle… This is not immediately obvious to me. Surely the boundary layer diurnal variation has a large role to play? This reviewer would have used a short-lived gas or particulate that would be relevant to the distance from anthropogenic activity.

Section 3.1. Here, I would focus on providing the validation that is pertinent to your study. A more detailed validation is always available in another paper.

Minor point: page 7, line 9: shorter than +/- 6 hours. 

Page 9, line 2: We select IAGOS profiles with top measurements not lower than 500 hPa? This statement is unclear.

Page 9, line 7: most sensitive to what?

Page 9, line 12: why does the column degrade re correlation? It is reverting to the prior?

Page 9, line 15: last sentence is unclear. 

Page 9, line 21. Model has a positive bias of 12% compared with the surface stations over China?

Page 11, line 11. Here and elsewhere I strongly disagree with the absence of averaging kernels when the model is compared to the satellite data. It seems as though the default for this study is to ignore them and place the comparisons that do use them in an appendix. It should be the other way around, if anything. You cannot legitimately compare the model with IASI without taking into account the vertical sensitivity of IASI to changes in ozone. 

Page 13, line 11: confirm this corresponds to two sigma. Explain to the readers why you’ve opted for that measure. 

Page 16, line 16: typo. 

Page 17, line 2: typo or at least I think so!

Table 4: the sensitivity calculations appear to be a progressive (cumulative) degradation of the reference case. In that case, won’t the sensitivity calculations become more non-linearly different from the reference case? 

Page 18: while discussing the impact of various sensitivity calculations it would be useful to link them back to the model and observed comparisons. For instance, the influence of the biomass burning in the export region looks impressive in the model (Figure 6) but the model doesn’t look particularly favourable against the data in Figure 5. BTW, Figure 5 appears to be cropped.

Section 5.2 was interesting but I think it would benefit from being more accurate in describing differences between TOAR and this study, even if the authors run the risk of repeating themselves, e.g. differences in time period. 

Page 19: length of period. Sensitivity to perturbations can be removed by using, for example, the Theil-Sen estimator. Using that approach may also reduce the sensitivity of your results to end points.

Page 21: impact of sampling. To check consistency would it be useful for the authors to sample the model coincident with IASI and then again with IAGOS and compare the trends using the two sets of sampled data?

For Table 6, I suggest also include the number of points in each calculation since this value will change a lot for the rows and columns.