In the manuscript, CAMS reanalysis tropospheric ozone profiles are evaluated during folding events using ozonesonde and IAGOS aircraft data in Europe. A control run without data assimilation is then also used to understand the differences, especially in the upper troposphere, between the modelled and observed ozone concentrations.

The manuscript is well written and gives a nice overview of the current knowledge about STT and tropopause folding. The scope of the manuscript is very focused, and the methodology very clear, although the knowledge of earlier studies by the authors is almost a must in the description. The results are interesting, but I had the feeling at several places that the authors could go more in depth. The authors observe, but do not interpret their findings enough. I will give examples here below.

1. In the introduction, you might refer to the work by Zhao et al. above Asia as well https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2020JD033955 , https://doi.org/10.1016/j.atmosres.2020.105158, and Luo et al. (https://www.hindawi.com/journals/amete/2019/4375123/).
2. Why did you restrict your evaluation of the CAMS reanalysis ozone during tropopause folds to Europe? This is clearly not the region with the highest number of tropopause fold events, so I guess the availability of the ozonesonde and IAGOS profile data has driven the choice of the study area. Please clarify your choice for Europe in the introduction (lines 38-44 are not that convincing for the current focus of the manuscript).
3. It would also be nice to give some additional climatological information on tropopause frequency, spatial and temporal variability of STT events over your study domain (Fig. 1) in Europe.
4. It is not clear to me which selection criteria have been used for the ozonesonde and IAGOS sites. For instance, the Prague ozonesonde dataset, which you have been using in an earlier study, is absent. If the data criterion for using ozonesonde time series is number of observations available throughout the 2003-2018 time period (line 71), I do not understand why, for instance, Observatoire Haute Provence (OHP), Sodankylä, Valentia (?) data were not selected. The same question arises for the IAGOS airports: what is the data temporal coverage (line 77) criterion used to include data from an airport or not?
5. As you included a control run without data assimilation, please specify the sources (which satellites? which products? during which period of the 2003-2018 time frame) of partial column and profile ozone retrievals that are assimilated in CAMS reanalysis.
6. In section 2.3, I would expect more details on the fold detection algorithm. Now, the summary is very limited. Major clarifications: how is the stratospheric source of air identified, what is the weight of the specific humidity content, and are the ozone concentrations used in the detection (I guess not, but please confirm clearly).
7. The selection of STT events (section 2.4) in ozonesonde and IAGOS profile data seems to be rather indirect , based on your database of STT events detected in CAMSRA (with 3D-labeling algorithm). I assume this algorithm is not directly applicable to ozonesonde and IAGOS “2D” data? Please specify. However, algorithms exist to detect tropopause folds in ozonesounding data as well (e.g. Van Haver et al., https://doi.org/10.1029/96GL00956), so how did you confirm the presence of an STT event independently from CAMSRA in the ozonesonde and IAGOS data? Details are missing how a fold is found in the ozonesonde profile (line 126) and in the IAGOS profiles (lines 131-135).
8. At the end of section 2.4, it would be nice to include some statistical information: how many of the CAMSRA STT events are also identified in the coincident ozonesonde and IAGOS profiles? Are the detection rates site dependent? Dataset (ozonesonde vs. IAGOS) dependent? What are the relative ratios between the STT events and rest of events? Again: site or dataset specific? Consistent?
9. In section 3.1 and 3.2, you mention some spatial differences between the CAMSRA and observational ozone profile differences, e.g. in lines 145-148, in lines 157-159. But, you do not give any explanation why features in the differences arise at some sites, and not at other sites. Is this related to data quality issues at some sites, instruments used, differences in the spatial representativeness of the tropospheric ozone observations at some sites compared to others, etc? Some discussion and/or thoughts would be helpful here.
10. My previous comment was on the spatial fingerprint of the CAMSRA ozone evaluation. But what about the temporal fingerprint? Is there a temporal evolution in the figures 3 to 10 that is smeared out by considering only the full 2003-2018 period? There might be a temporal component due to a change of the data quality of the observations, change of data assimilation source data, etc. Can you comment if you detected a temporal fingerprint?
11. There is no clear explanation given in section 3.3 why chemical data assimilation deteriorates the comparison with the observations above 350 hPa. Only in the conclusions, a list of possible improvements in the data assimilation is given, which might explain the larger inconsistencies between model and observations at those pressure levels.
12. Please reformulate the expression “with a bias increase close to O3 increases closes to the upper troposphere”; its meaning is not fully clear to me.

Review for “A process-oriented evaluation of CAMS reanalysis ozone during tropopause folds over Europe for the period 2003–2018” by Akritidis et al.

The authors present an evaluation of climatological tropopause folds in CAMSRA reanalysis against WOUDC ozonesonde and IAGOS measurements over Europe. A proven labeling algorithm is used to identify tropopause folds in 2003-2018. The overestimation of upper tropospheric O3 in CAMSRA results from the chemical data assimilation. The study presents useful validation and the manuscript is well-written, which merit its publication in ACP. However, there are some improvements that are required both in the writing and discussion. Some of the statements require detailed explanation to support.

Major comments:

1. Please be more specific about the locations of folding hot spots. Why you are especially interested over Europe? I suggest to add some additional references on the global STT over different hot spots of fold activities.
2. How did you perform the control simulation of IFS without the use of chemical data assimilation? Since you claimed that the chemical data assimilation is the key factor resulting in most of the biases, you need to be more specific about your experiment design.
3. I have concerns about the great differences between O₃ and O₃s in the upper troposphere (300-400hPa) as shown in your Figure 4 and 5. Green profiles differ largely away from the red ones at 300 hPa. Will they merge near the tropopause level? If not, I am very worried about the quality of the upper tropospheric O₃ as well as the stratospheric ozone tracer in CAMSRA. can you please show a horizontal map of both the climatologies of O₃ and O₃s at 100 hPa and 850 hPa, respectively? Additionally, there have been various methods when calculating the O3s tracer among different models. Please add more details about the stratospheric ozone tracer (O₃s) in CAMSRA. What is the choice of tropopause in defining the tropopause? I’m curious about it because the diagnostics of STT might be sensitive to the choice of tropopause definition.
4. Please add more details about the two evaluation metrics (FGE vs MNMB) chosen in your study. Why you chose these two statistics and what are differences between the two? What does the score mean, respectively?
5. Section 3.1 and 3.2, what leads to the different biases across stations?

Minor comments:

Lines 30-35: please be more specific about the hot spots of fold activities. Please add some references for transport of VSLS to the lower stratosphere

Line 39, a comma is missing before “resulting”

Line 68: can you add a few sentences about the differences between the ecc and the Brewer Mast ozonesondes?

Line 90: can you be more specific about “CY42R1” and “4D-VAR”?

Line 92: NO₂

Line 99: what is the vertical resolution in the upper troposphere?

Line 130: the rest of

Line 142: I suggest adding the longitude and latitude information of each station to Figure 3.

Line 166: Figure