This paper presents the results of k-means cluster analysis on the particle number size distributions (PNSDs) measured at 4 different Antarctic research stations. It is an extension of the work in Lachlan-Cope et al., ACP, 2020 which applied this analysis to the same year of PNSD data from one of the stations (Halley). Applying cluster analysis to the combined data is an interesting idea but it is not clear what new insights are gained. In addition, a lot of the material in the discussion and conclusion sections seems more like introductory material.  I would suggest restructuring the paper to move all introductory material to the introduction and to make clearer in the discussion what new information is gained from the combined cluster analysis. Specific comments are:

Major comments:

1. Please be specific about new insights gained from this analysis. Is it confirmation of the cluster types observed in Lachlan-Cope et al. (2020). That paper found 8 clusters and this one finds 6, so a comparison of the two sets of clusters would be helpful. Is there new information about the source of a particular cluster type? If so, please add this to the discussion.
2. The paragraph from line 601 to line 625, the dangling sentence on lines 626 to 628, and the paragraphs from line 662 to line 691 do not seem like material that is about interpreting the results in this paper. Rather this seems like broad overview material that should be combined with the introduction.
3. I don’t understand what Figure 7 adds to the discussion. Everything is the same between summer and winter except with slightly more snow and sea ice in the winter. Perhaps the goal was to show that boundary layer NPF does not contribute as much in winter, but the symbol is really hard to see. Why are the panels slanted? Is the sun shining in winter or is that supposed to be the moon?
4. I find Figure 5 really hard to interpret with backtrajectories from all of the stations overlapping. I would move Figure S9 to the main body of the paper and skip Figure 5.

Minor comments:

1. The paragraph in lines 196 to 205 is confusing. The first part talks about wintertime PNSDs at Halley and the second part talks about NMR analysis from a summertime cruise. How are they related?
2. In the paragraph about the Halley data (lines 238 to 244), please state that this same cluster analysis has already been applied to the Halley data and published in Lachlan-Cope et al. (2020).
3. Please define the parameters in Equation 1.
4. Please list the cluster types in the same order in the text and the figures. For example, they are listed in one order in lines 313-314, in a different order in the paragraphs in lines 320 to 348, and in yet another order in Figure 4.
5. I find the name Bursting confusing because I think of bubble bursting but you are using it to mean new particles that fail to grow. In fact, you use bursting in both senses in the paragraph on lines 485 to 497. Is it possible to change the name of this cluster?
6. I would refer to Figures S4 and S5 at the end of line 357.
7. Lines 378-379: Is the increased contribution of Nucleation and Bursting in September due to the sun returning?
8. In lines 107 to114, you talk about the debate between upper troposphere NPF and boundary layer NPF. On lines 429-430, you note that trajectories corresponding to Nucleation have a lower than average height. Does this provide an answer to the debate?  You don’t come back to this observation in the discussion.
9. Line 489: Can you explain briefly what the Hoppel minimum is?
10. Figure 1. Why not show the ice extent in 2015 rather than the 30-year average?
11. Figure 3a: I would like to see a comparison of the Halley cluster distribution from this work with the one derived in Lachlan-Cope et al. (2020), maybe in a figure in the SI. I’m curious whether the distribution of clusters changes when you include data from the other sites.
12. Figure 4. Is panel a) useful? Since Dome C is entirely over Land, it seems like averaging all 4 sites together just dampens any pattern among the different clusters. It might make more sense to show the average across clusters for each site since you discuss that in the text, i.e., add a 7^th column for the average for each of the 4 sites. Can you use the same names in the legend as in the text, e.g., “Sea ice” rather than “Consolidated pack ice”? Can you shift the names of the clusters slightly left so they are under the tick marks?
13. Figure 6. I would use more distinct colors for the different sites. Weddell is spelled with two l’s.
14. Figure S6. A sentence is duplicated in the caption.
15. References: Heinrichs is out of order and the two Humphries references are in the list twice.

Review for the article: “Multiple eco-regions contribute to the seasonal cycle of Antarctic aerosol size distributions” By Brean et al., 2024.

In this study, the authors, have studied the particle number size distribution collected for one year in several stations in the Antarctic region plus some additional campaigns/measurements. They have found that the four different stations have very different size distribution possibly highlighting different sources. All the data are analysed with the k-means cluster analysis, a very common method often used by these authors. With this method they have identified six main categories (Nucleation, Bursting, Aitken high, Aitken low, Bimodal and Pristine).

The data analysis is well done and interesting however is not clear from the data analysis how they could get to their conclusions (Schematic Figure 7). Such as: “We provide evidence that both primary and secondary components from pelagic and sympagic regions strongly contribute to the annual seasonal cycle of Antarctic aerosols which add insight on the possible sources of aerosol production/activity in the whole Antarctic region.” I find it hard to connect the final schematic with the data analysis done in the study. I do understand that PSND can’t give you too much information on biological processes and chemistry. Finally, it is not really clear the advancement of this study compared to Lachlan-Cope et al., (2020). 

My concern are clearly not major and probably with some change in the text and some “toning down” of the conclusions/results should be fine. I therefore think that it is a very good paper and deserve to be published in this journal. 

Minor comments:

I have a question about the K-means clustering. How does it perform when the instruments have different cut-off (as the one shown in this study)? Especially in the nucleation mode where a small change in the cut-off can have a big impact on the particle number. 

Specify acronym such as VOC, DMS and so on.

It feels that the discussion related to new particle formation around line 600-625 belongs more to the introduction then the results. I would consider of moving that part there.

I’m a bit puzzle by this statement in the code and data availability section: “The code and data used to produce all non-illustrative 700 figures are available from the corresponding authors under reasonable request.” What does it mean reasonable request? Shouldn’t all the data and code be in a repository freely accessible. I think it’s important to have these very important data (and possibly code) open access.