This is the first review of the manuscript “Trends in polar ozone loss since 1989: First signs of recovery in Arctic ozone column” submitted by Pazmiño et al. to the EGUsphere journal.

The paper discusses ozone loss rates over the polar regions detected from the SAOZ and MSR2 datasets and uses passive tracer from the TOMCAT/SLIMCAT model to test the indicators of ozone recovery within the polar vortex. Authors utilize three metrics to analyze ozone recovery over the Antarctic and Artic polar regions in the 2000-2021 period. The detection of ozone recovery in the Antarctic vortex is supported by two metrics that indicate negative trends in ozone depletion parameters that are statistically significant. However, in the Arctic, the detection of trends is complicated by large dynamical variability and difficulty to establish threshold value. The first signs of stratospheric ozone recovery (represented as negative residuals with respect to the sunlit volume of the polar stratospheric clouds) in the Northern polar vortex are detected significant at 1 standard deviation error.

This paper is well written and the results of the findings are attributed to the dynamical and chemical processes governing ozone loss in the polar stratosphere. The discussion of large ozone anomalies in the individual years are linked to the SSW and vortex stability. This manuscript has figures that support discussion and conclusion.

I recommend publication of the paper after several questions/comments are answered.

Clarifying qestions and comments.

Line 128. What is the definition of the "overpass" criteria?

Line 143. Please provide station and satellite/model matching criteria. The satellite grid is at 0.5 and SLIMCAT model is at 2.8 degrees. Are there additional matching/averaging is done to reduce sampling biases? Also, it might be useful to provide the number of observations for all stations inside of the vortex during the analyzed period.

Line 149. Please provide additional details of datasets merger, i.e. temporary and special matching, treatment of missing data, weighting of the MSR2 and SAOZ data in the combined record.

Line 163. WHat is the reason for not selecting Match to normalize all years of SAOZ dat? This could make normalization consistent through the entire analysed record.

Line 185. Please clarify what you mean by "diurnal differences".

Line 215-216. Can you please meantion ozone variability in 2019 that was also an anomalous year in the Antarctic ozone depletion?  It clearly deviates from other years.

Line 228. Fig. 4 caption. I would not say that 2002/2011 winters are unusual anymore sine we had similar anomalies in recent years. Do you agree?

Lines 325-328. Is there a known reason for the offset between observations and the model since 2003?

Lines 375-377. Please provide uncertainty of the linear and the parabolic fit for the sunlit PSC area and ozone. What does the SLIMCAT data fit show? Do data and a model fit agree? Can you add a plot that shows the change in the sunlit VPSC as function of time? This could provide a reference of climate change over polar regions.

Lines 396-402. If uncertainty of the ozone/PSC fit is taken into account, would the trend of the residuals be significant?

Lines 465, acknowledgements need to be made for the NDACC data

“The data used in this publication were obtained from “NDACC PI name” as part of the Network for the Detection of Atmospheric Composition Change (NDACC) and are available through the NDACC website www.ndacc.org.”

Line 449. Please provide the link to the ERA5 data.

Referee report on "Trends in polar ozone loss since 1989: First signs of recovery in Arctic column" by Pazmino et al., considered for publication (in ACP).

*** General

This analysis of metrics related to ozone depletion and recovery in the polar regions is well presented, overall. The analyses are well thought out and they are described with enough detail, for the most part, although (for the reader) some of this requires some knowledge from past references and studies. The results appear to be robust enough, in many cases, but I do have some reservations when it comes to the Arctic and the real robustness of those results using only one of three metrics, and at a one sigma leval of significance (at best), given the lack of sufficient sensitivity studies as well. To claim a "first detection" regarding ozone recovery in the Arctic seems overdone, and I would not recommend the use of such language, especially since other words in the manuscript are mush more cautious as well.

Apart from this, the work does provide new results, tied to good datasets and models. I hope that these comments are taken into account for a revised manuscript, without much work required, in reality, unless one wants to try to claim that a robust detection of positive ozone recovery rates can be made for the Arctic, based on this analysis; from this work, I do see potential signs of a positive rate of recovery for total ozone columns over the Arctic, but I would not try to overstate the conclusions based on the evidence presented here, despite the overall good quality of the work. Even if a statement such as a positive rate with a 1 sigma level of significance is technically correct, using the word "detection" is an overstatement, in my view - and care is needed for such a word not to be misinterpreted by a casual reader. There are some signs... but no robust "detection of ozone recovery" over the Arctic has been made, in my opinion, based on these results; in fact, the authors seem to agree - so there is no need to potentially confuse those interested in this topic. Staying positive about this manuscript, my caution and comments should really not affect the vast majority of the analyses, wording, and conclusions. Thus, I consider that only minor revisions are really needed, even if a non-minor comment exists regarding the wording and potential misinterpretation (for the Arctic). 

*** More specifics regarding the more major issues/comments 

- L34 and related discussion in the manuscript: I would argue that a 1 sigma "detection" is not really a detection with enough significance; it is a likely detection as opposed to a robust detection (at 2 sigma or more), and some scientists in various disciplines would argue for even stronger significance levels, in addition to the fact that one often cannot or does not include all possible error bars in the analyses. In this case, the assumption of a linear relation between ozone loss amounts and VPSC is just that, an assumption taken as "truth" and any departures from this "truth" signify something related to ozone trends or recovery. In my view (and hopefully in the views of others in this field), this is just an indirect method at suggesting there may be "signs of recovery" (per the title of your manuscript), which is a better language than a "first quantitative detection". I understand that there is often a desire to show a "first" in research, but this can be overdone, and science progress is usually obtained via multiple analyses over time, especially for inferring trends, and the Arctic can change enough that adding or subtracting years can make substantial differences in the results. Rather than using a bold assessment like "first quantitative detection", I would urge the authors to use a more cautious statement. Error bars here are a lower limit, especially since the same sort of analysis for the Antarctic region yields error bars that are significantly smaller than other metrics results, so this is somewhat suspicious to me just on this basis, in addition to the fact that this method is more indirect than the other two metrics. Please counter this argument if you feel that there is a strong reason to declare victory on the Arctic recovery signal based on just one indirect metric and at the 1 sigma level (at best). I am not convinced, at this stage, and I feel that more metrics and years are needed for such a bold statement (including a broader community assessment, such as another WMO report, for example). I would not try to argue the validity of line 33 too strongly, as long as line 34 gets deleted, or replaced by something like "We argue that this points to the first signs of ozone recovery in the Arctic springtime lower stratosphere." [Although I personally would probably say this "may point", being a cautious person on such matters.] Alternatively, please make the case regarding such a bold statement by performing more error analyses - but this will typically increase the error bars, so the case will just become even weaker, I predict. Moreover, given that the ozone recovery path depends on both ODS and greenhouse gas effects, it is also difficult to provide a robust attribution of slightly positive trends to one effect or the other, without more detailed analyses; there are not enough model results to compare to, in terms of what a model would predict for one effect versus a combination of effects, in general, with comparisons to any of the observations shown in this work. I am therefore going to remain skeptical of broad sweeping conclusions for the Arctic, especially (although some caution is also recommended for Antarctic ozone studies). In fact, your own words at the end of the manuscript show more restraint and caution (with a pointer to another reference as well), so I imagine you actually agree with my words of caution. I think this shows nice results, whether one wishes to claim a "first" or not, and this is what should be the more important conclusion, a good set of analyses with hopefully reasonable error bars, and without overstating the possible conclusions.

- As a related comment regarding Fig. 11, if one wants to claim enough robustness in the result and error bars, one should try to use 2 or 3 years less (or more) at the beginning or end of the series, to see how this affects the results and error bars. I think it is best, again, to be cautious in terms of a "robust detection" comment, unless the analysis is at least significant at the 2 sigma level, with enough sensitivity analyses as well.

- In addition, why not show what a polar-focused model would predict for such a metric, if one could add some credibility to the conclusions (in the Arctic especially) in terms of consistency with expectations.

- Also, of some interest, is there not a model-derived ozone loss onset date curve that can be compared to the Figure 9 results? What does this (or would this) show? Would this not be a useful comparison for wht might be expected? This is not directly tied to ozone loss (but I do understand its connection to this). Any comments about this (in the text or at least as part of a reply) would be appreciated, since this might be worth considering as an added comparison (although not necessarily in this particular work).

*** More minor corrections and suggestions

- L22: Add "column" between "cumulative" and "loss" for clarity.

- L23, Abstract: since there are somewhat complicated calculations that involve more than just linear trends, as stated on line 27 (parabolic), this seems somewhat simplified of a description, even if the Abstract has to be simplified and short.

Maybe consider the following wording: "Three metrics are used in trend analyses that aim to assess the ozone recovery rate over both polar regions: ..."

- L28: I think you mean (or should use) standard deviation as part of the error analysis (see other comments above), instead of standard error, or justify the use of this terminology better.

- L29: you should be less vague and specify what threshold refers to here, what quantity (ozone column), instead of making the reader guess (if/as the manuscript has not been fully read at this point).

- L29: I would suggest "all of them showing a time delay as a function of year, in terms of when the threshold is reached."

- L32/33: "the difficulty in finding a threshold value in enough of the winters."

- L63: wildfires events --> wildfire events

- L72: change "concentrations/columns(?)" to just "columns".

- L120: SAOZ ozone data could be more specific SAOZ total column ozone data?

- L133: delete the period after "used".

- L173: delete parenthesis after "merged dataset"; also change "bias" to "biases".

- L176, Figure 1 caption: please specify the year in the caption also (position of the 2021/2022 vortex edge...").

- Figure 4: Please try to plot the thin black lines last, so they can be better seen on both panels. Consider making them slightly thicker as well. Should the y-axis title have a space between O3 and loss (maybe not, if you are referring to a specific variable instead of actual words). Also change "Day of the Year" in both panels (x-axis label) to "Day of Year" (as used in Figs. 5,6,7).

- L235: For Figure 4, please specify in the text at which day of year the maximum amplitude of ozone loss between recent winters occurs (for both hemispheres). This will help the reader.

- Figure 7: vortex edge as defined in Pazmino, but does that follow from Nash et al. (could specify here as well, if so, in addition to the text). Also correct the typo in NH y-axis (Gradiant --> Gradient).

- Figs. 5,6,7: please change the thin black lines so they are plotted on top, for better visibility (and consider making them slightly thicker as well).

- L253: and so was the vortex stability (rather than "and as well as the vortex stability").

- L254: August, linked to a wavenumber 1 event,...

- L255: upper levels, with an associated decrease in size.

- L262: heat flux increases rapidly at the end

- L264, slowing down rapidly thereafter.

- L268: event, of a magnitude similar to the anomalies related to the Calbuco...

- L271: mean T anomaly of -10.1 +/- 4.5 K, as in 2018, but with a much larger sunlit VPSC than in 2018.

- L272: The persistently low temperatures [or The persistently cold lower stratosphere...]

- L273: led to an acceleration of the October ozone loss...

- L276: (Fig. 6), and the strength...

- L280: The sunlit VPSC values are similar...

- L283: Fig. 4), which lies within the climatology.

- L293: Do you mean "The strength of the vortex edge exhibited values larger than climatology...?

- L294: vortex led to moderate ozone loss; also please specify again (if need be) where the ozone loss error bar values come from (or refer back to that discussion)., and if these represent one standard deviation (presumably not two).

- L300: ozone loss of the 2019 warm winter

- L304: temperature anomalies at the 475 K level... and the mean anomaly value for the whole winter reached -5.3 K, as in 2018.

- L306.307: final warming mode, also shown by the low values...

- L308: persistent low temperatures less than ...

- L319: a possible recovery path of total ozone... [or recovery rate]

- Figure 8: please make the grey legend stand out more; for example, use larger fonts for the legends and say % dec-1 to shorten the units and legend length, and use a bolder font if needed.

- L344: are positive (1.0 +/- 2.2 %dec-1)

- L348: values, as we might expect a later onset... in relation to lower...

- L350: what is the ozone loss time dataset? Is this not the same as the ozone loss onset days "are used instead of total ozone columns"... (?)

- L371: You say a "3rd order polynomial..." and also mention a parabolic fit; to me, a parabolic fit is a 2nd order polynomial (i.e., a quadratic). Please clarify.

- L369-L371, I would say "dynamic range" or just "range" really; dynamical could appear to refer to something atmospheric...

- Figure 10: It might be interesting to try a linear fit after 2000 for the SH; not necessary for this paper, just a thought (how would that affect the results?).

- L419: was calculated, but it is not significant.

- L420-422: It would flow better if the sentence "Regarding the SH..." was placed one sentence above, directly after the SH comment. Then one could just say "This metric appear sensitive..."

- L433: Note that this trend is similar to the SH trend.

- L436: add a comma after "datasets".