This manuscript describes an extended evaluation of recent improvements to the radiance assimilation system of ozone observations from the IASI instrument, and compares the results to those from a state-of-the art chemistry transport model (GEOS-CCM) and data assimilation product (CAMSRA). While the authors don’t present new data-assimilation developments, the evaluations presented in manuscript are very useful to benchmark the current configuration, both compared to observations, and to other products. 

The manuscript is well written, and I have only small comments. the only weak point I find is the slightly limited scope of the analyses and evaluations presented here. It could be useful to extend the scope of the evaluation and benchmarking exercise, e.g. in the directions as suggested below.

1. The authors now mostly focus on evaluations of zonal mean concentration at five latitude bands (Figs. 2.3), or as zonal mean (Figure 5). In addition it would be interesting to see the performance of the assimilation system for more regional variations, e.g. by plotting presenting evaluations as in Figs 2/3 specific for certain regions (e.g. Europe), or presenting the data as in Figure 5, but (for instance) showing a lot/lat plot averaged over DJF and JJA, at selected altitude levels. 
2. In Figure 5, in addition to showing the results of IASI-r, why not also present plots for the IASI-a product for reference?
3. Can the authors comment on the differences in computational costs for the L1 and L2 assimilation configurations, and discuss the reasons for differences, and potential means to improve on this? Also, it’s good to put these differences in costs into perspective, particularly in regard to their comment in the conclusion in the case “IR measurements are already assimilated”, (p12, l16)

Furthermore, considering the data-thinning approach (pp 7, line 3), in the current study only a single observation is used in every 2x2 grid box. Can the authors explain a bit more about their data-thinning approach, e.g. do you make any additional check on selecting representative observations (apart from the dynamic filter)?

Technical comments

General: please don’t refer to “C-IFS”, which is outdated terminology, but plainly to ‘IFS’ instead, or plainly to “CAMS”.  Also, please make more distinction between “model” and “(data-assimilation / model) product”, as indicated also below.

p1, l7 “models”-> “products”

p1, l8 “C-IFS” -> CAMS Reanalysis

l11: C-IFS reanalysis-> CAMS Reanalysis

l15 “concentration” -> evolution”

l17 ‘ENSO’ : explain acronym

p2, l7 remove “assess”

l24 add ‘a’ in “found a positive impact”

p3, l16 “and to the CAMSRA chemical reanalysis (Inness et al)” suggest to change to “and to the CAMS atmospheric composition reanalysis (CAMSRA, Inness et al)” 

l18 “satellites assimilation” -> satellite data assimilation

l23 reanalysis experiments -> reanalysis datasets

l25 “chemical”->”chemistry”

l29 “experiments” -> “datasets”

p4, l7 “ACMMIP”->“ACCMIP”

l16 “aerosol schemes(C-IFS, Flemming et al.(2015))” ->“aerosol schemes(Flemming et al.(2015))” 

l16 “Within C-IFS” -> “within IFS”

p5, l 23 and 26 “ozonesondes” -> “Ozonesonde” ?

p6, l10 “limit” -> “”limitations

l12 “satellites” -> “satellite retrievals”

p7, l12: please check units (incl upper case, and use of ‘-1’ and ‘/’ in mW/m2/sr/cm-1

p8, l13 “chemical reanalyses” -> “chemistry model and reanalysis”

p11 l32 “modelling experiments” -> “modeling and data assimilation products”

Summary

The paper titled ‘Impact of IASI thermal infrared measurements on global ozone reanalyses’ by Emili and Aabaribaoune aims to answer the question: what benefit does using the IASI-r O3 reanalysis have over using the IASI-a O3 reanalysis and how well does it reproduce measurements compared against alternative approaches?

I recommend publication once the following minor revisions have been addressed.

General Comments

A well planned methodology, clear results and valuable discussion make for a high quality paper that is suitable for GMD.  

Specific Comments

page 2 lines 15-20: the uninterrupted nature of IASI measurements versus the interrupted nature of TES measurements is a key motivation for the paper.  However because the reader is not told what measurements Inness et al. (2019) use, this motivation is not so clear.

page 6 line 18: it would be beneficial to quantify ‘very small’, even just a headline number to give the reader a feel for the significance of changing from 4D- to 3D-Var

page 6 line 19: as pointed out by the authors the radiative transfer model changes between IASI-a and IASI-r, but there is no comment on the effect this has on resulting O3 fields.  The reader would appreciate the distinction between the effect of changing IASI product level and the effect of changing radiative transfer model.

page 7 line 3: could the reason for using a dynamical filter to reject pixels that differ from the model by more than 12 % be given?

Technical Comments

page 2 line 7: should assess be assessing?

page 2 line 16: I don’t understand what Huijnen et al. (2020) compared - was it comparing Inness et al. (2019) against Miyazaki et al. (2015)?  Please make clearer what is being compared here.

page 2 line 17: does ‘similar accuracies’ mean the accuracy when compared against radio-soundings as a reference?  If so, please make clearer what the meaning of accuracy is in this section.

page 4 line 5: change ‘took part to’ to ‘took part in’

page 4 line 10: change ‘for the 2010’ to ‘for 2010’

page 4 line 27: ‘heritage’ makes IASI-a sound like an old legacy method, perhaps ‘product’ would be more suitable

page 4 line 29: change ‘In the recent years’ to ‘In recent years’

page 5 line 23: I cannot see a definition of the OMI-MLS acronym, nor the separate definitions of OMI and MLS

page 7 line 12: change the second ‘ii’ to ‘iii’

page 8 line 26: change ‘This type of statistics are’ to ‘This type of statistic is’

page 10 line 17: is the intention here to reference the empirical bias correction of both Emili et al. (2014) and Peiro et al. (2018)?  If so, please replace the semicolon with ‘and’

Figure 4: please make legend labels consistent in spelling with caption and main text