This paper presents varied procedures to identify tropospheric ozone depletion events (ODE) in springtime at Utqiaġvik, Alaska during 23 years, although the authors are focused on two specific years, 2012 and 2021, and four procedures. Three types of criteria were compared: traditional methods, variability-based methods, and machine learning methods. Moreover, the influence of meteorological conditions is investigated. The analysed variables include wind speed and direction, 2-m temperature, and surface pressure. Moreover, the influence of halogen species is suggested. The paper is quite complete. However, some minor changes should be considered prior to its final acceptance.

Since the analysis is made on a specific site, the interested readers could wonder about the result robustness, i.e., the authors should comment about limitations of the result extension to different sites.

In paragraph between lines 245 and 254, the authors discuss about the suitable number of ODE hours. They should indicate a reason for such suitable number or when a number could be excessive.

Finally, the references used for discussion are unevenly distributed. For instance, they are frequent for the relationship with meteorological parameters, but less frequent in previous result sections. References should be the link between this research and previous studies.

Advantages and disadvantages of the presented procedures should be highlighted in the conclusions.

Minor remarks.

Line 46. Replace “1999),.” By “1999).”

Since curves are superposed in Fig. 2. Perhaps, additional information could be obtained if each criterion is represented by its mean and standard deviation and all of them are ordered following their means.

This paper compares three criteria on identifying springtime ODEs at Utqiagvik, Arctic, and tracking the declining trend of ODEs from 2000 to 2022, which is possibly due to the climate change and Arctic Sea ice melting. Furthermore, authors investigate the relationship of ODEs and meteorological factors, including wind, temperature, and pressure, which give a certain explanation of ODEs characteristics.

However, the paper deserves more work before publication. Major revisions are needed to make a nice addition to the literature. I request that the authors consider the following points as they revise this manuscript.

Major comments:

1. The discussion is a bit superficial. Especially in the subsequent impact of different criteria on ODEs event identification (i.e., health impact, climate change, etc.). Besides, as mentioned in the final "Conclusions and Future Work”, authors should involve more observation data (such as BrO) to further evidence the assumption or the attribution in the discussion.
2. The trend discussed in Section 3.1 and Section 3.3 are opposite, the discussion throughout the paper should be more rigorous and unified.
3. In section 3.2, the authors should add all the other tests’ results in SI, other than the TM1, TM4, VM, and IF presented in main text.
4. A map of station location is necessary, especially when the author discuss the relationship between meteorological condition (such as wind directory) and ODEs event.
5. The font in all figures should be of same size, there are some figures having too-small font, such as Figure 3, 5 and 6 etc... I suggest authors to replot all figures, all of which are not clear enough, and the styles are more like a report, not paper.

Specific comments:

1. Line 20, the reason why the ODE only happen in spring should be explained in the Introduction part.
2. I suggest the author to move the Figure 1 to SI, and present the original O3 concentration trend since 2010 instead.
3. The Lines in Figure 2 is not easy to distinguish, I suggest authors to classified them into two panels, such as traditional results in one, and the rest methods in another one. The color of lines should reflect the method clusters. The current color setting is too hard to follow.
4. Line 200, the IF curve and VM cuvre are not alike at all in Figure 2. How do the authors have this conclusion: “it is seen that after the year 2014, the IF curve behaves similarly to those of the TM methods, while before 2014, the IF curve’s trend is more like the VM method’s trend”?
5. Line 235. Add results in 2013 and 2022 in SI.
6. Line 265, the conclusion of “the machine learning approach exhibits a limitation in accurately identifying ODE hours in years characterized by a high frequency of ODE occurrences” is too arbitrary, since the authors only tried one ML method.
7. How is regression done in Figure 5?