Antarctic near-surface winds are important for blowing snow, precipitation and ice shelf stability. This article presents projections of Antarctic winds for the end of the century under a high-emissions scenario, using the MAR model with a suite of driving global models from CMIP6. A comprehensive overview of changes in winds is presented, with meticulous evaluation of the added value of dynamical downscaling. The authors also decompose the momentum budget in the regional model projections. Although a previous paper on future winds (Bintanja et al., 2014) did estimate two of these terms, to my knowledge this is the first to provide a full budget decomposition of projected winds from regional model output. The results are novel and interesting and I would support this paper’s publication subject to revisions. The main points I raise in this review are regarding the application of the budget:

Major comments:

1. It’s not quite clear from this analysis how the momentum budget decomposition was applied for each of the model configurations. Was your choice of parameters for diagnosing theta0 the same as in Davrinche et al. (2024)? Is there a way to test how robust your results are to the choice of Hmin?
2. The regional analysis is quite long and has a large figure count – you may be able to significantly improve readability by reducing the number of figure panels shown in Section 3.5, which focuses on a very specific question (the drivers of decreases). The vast majority of the changes shown in those regional panels are already visible in Figure 5.
3. A time-correlation analysis between the wind speed and budget terms would really help understand the role that surface forcing plays, instead of just looking at the change. How much variance in the July monthly mean between 1980 and 2100 do we explain with just VLSC from Equation 5, then how much when we add other terms?

Specific comments:

• L30: A “thermal wind that acts to replenish the pressure low created by the downslope displacement of air”. Could you describe what you mean by this in more detail – a thermal wind operates at large scales in a baroclinic atmosphere. Do you mean the local thermal wind acceleration term or the thermal wind relationship used to calculate lsc?
• Intro: please briefly review Bintanja et al. (2014) and consider the advancements made here relative to that research. I see you reference them later – it’s worth signposting early on.
• L54: I’m not sure if dynamical downscaling alone ensures a physically realistic simulation of boundary-layer dynamics. Rephrase perhaps?
• L100: Why select July and not the more usual climatological season of JJA?
• The supplement is very large and there’s a lot of flipping back and forth between the main text and the supplement. If there is a way you can reduce the size of the supplement it would improve the flow.
• S1 and S2 are one equation I believe?
• S1.1: is the relative uncertainty the standard error?
• L124: why calculate the metrics for December and the annual mean if we are only focused on July?
• Supplement L16: check the reference to ‘Figure 5 in the manuscript’.
• Supplement: in my PDF Figure S3 shows after Figure S4.
• Figure S3: I am a bit confused by the (b) panel colour bars. What is the left and right coloured bar showing? Maybe it would be simpler to show each individual TPS on the grid and e.g. hatch the gridcells which pass the threshold?
• L203 Is this strictly speaking the boundary layer? You imply here that the height at which the vertically integrated temperature deficit becomes zero is the top of the boundary layer. In East Antarctica however the temperature deficit can extend to ~4km height (see e.g. Figure 3 in van den Broeke and van Lipzig, 2003). This is much deeper than the top of the stable boundary layer, which vdB and vL say is ‘poorly constrained’. Is it not more correct to say that it’s just the vertical integral of the temperature deficit? My understanding is that the temperature deficit can extend far above the boundary layer, which over the plateau may be e.g. 10-150m at Dome C, Pietroni et al., 2012: https://doi.org/10.1007/s10546-011-9675-4
• L226 does the residual here also encompass any errors from closing the budget (e.g. finite difference approximations) or is it directly output from the model?
• Figure 3a – what is the x-axis here
• Section 3.1 no need to restate this first para, or move to the introduction.
• L270 figure panel reference needed
• Figure 3: I think I missed what the collocation method is? Are you using nearest neighbour or bilinear? Is MAR regridded to the same grid as the ESMs? If not it would be useful to do this as an additional analysis to just check if the added value comes from being able to collocate a gridpoint closer to the location of the AWS in MAR.
• Figure 4: (v) not quite able to tell but it looks like this is not the Ronne ice shelf? It may be worth checking – in my understanding the Ronne hugs the peninsula and the Filchner ice shelf is east of that.
• Table 3: in my PDF this appears below Figure 4 (but referred to beforehand).
• L357: my understanding is this ˆ is the vertical integral of the deficit rather than the depth of the layer
• L357: please specify where in Figure S7 you are referring to for the coastline
• L359: in some regions (e.g. offshore of Adelie land) the thermal wind is a positive forcing term and does not oppose the katabatic wind so it doesn’t necessarily increase wind speed if you reduce it.
• L398: where are these regions where ‘surface forcing can also contribute to significant wind speed increase’?
• PIG -> Amundsen embayment region?
• L454 regional specifics would be helpful here as this compensating effect only applies where the katabatic winds are active
• L455 you imply here that some regions have an increased wind speed due to surface forcing, and it’s true that the surface forcing does increase (kat+thw) in some regions but I don’t see these mapping onto obvious increases in wind speed.
• Section 4: I may have missed it but I think the added value of dynamical downscaling is an important result to mention here too?

The manuscript focuses on projections of surface wind speeds over Anarctica, which is comparatively under researched. The approach is very comprehesive (e.g., four downscaled GCMs, as well as a budget analysis) and the figures are well made - and clearly a lot of effort has gone into this work.  Additionally, in general much of the results are well explained, with a good level/balance of detail.  So there is much merit to the paper.

However, unfortunately the manuscript is let down by other instances of poor writing and organisation - and indeed I would even go as far as suggesting that the manuscript in its present state was not ready for submission.  These concerns are especially evident in the Introduction and Methods section, which come across as rather muddled, disorgansied, and disjointed. I am sure this is not reflective of the authors abilities and knowledge, so this really has to be remedied before the manuscript can be considered for publication.  I would really suggest a very thorough rewrite/revision of many of the section is necessary, with all authors contributing.   

Major comments: 

+ Many parts of the manuscript come across as rather unpolished and the writing disjointed.  This really needs to be improved.  For example, many of the sentences in the Introduction claim something but do not include a citation for evidence.  So sentences such as ‘On the one hand, the greenhouse warming causes an increase of the incoming longwave radiation.’ and ‘Although there is a consensus on the reduction of surface forcing in climate projections’.  There are also incidences of repetition, such as in the Introduction with something along the lines of ‘which is proportionate to the strength of the temperature inversion’ mentioned twice, and in the methods and Introduction which both mention something along the lines of ‘Because of their resolution, GCMs are not expected to perform well in locations with complex topography.’  Other instances are the preambles/motivation before the results, which just say in a slightly different fashion what was said before. Please remove all repetition, and remember that your audience/readers only need to be told something once.  Also there are typos, such as ‘(e.g. north of Ross and Amery ice shelves and north of the Peninsula’ in the Introduction (so no closure of parentheses). Mistakes such as AWS defined, and the phrase automatic weather station still used. Very random / ad hoc approaches such as using m/s in one sentence and km/hr in the following sentence (methods).  These give the feel of a rushed writing process, and of a manuscript submitted before it was really ready. There are also parts of it which are disorgansied, such in section 2.1 mentioning ERA5, and then ERA5 not being explained until later (also it’s not explained in a logical fashion from the methods that ERA5 is being used to select the GCMs.). Poor paragraph structure such as section 2.1.2.  Finally, some odd sentences such as ‘We focus on the Antarctic continent, which is the source region of the katabatic forcing’ in the final paragraph of the Introduction. 

+  Methods: Out of the blue it is mentioned that the subset of AWSs are selected based on their ability to represent ERA5.  This is not justified. Additionally, this seems a rather strange choice, as ERA5 would also struggle to represent steep coastal gradients, so also do poorly representing katabatic winds.  So justification is clearly required.   The correction to the AWS dataset is also poorly explained (Equations 1 and 2) – its not even clear what is being corrected, and what 1-3 and 1-6 refers to.

+ Selection criteria for GCMs: This seems to state that their performance in the Arctic is also taken into account, which is completely unjustified.

+ There are a lot of locations mentioned, but I don’t think they are always shown on a map.

Minor comments (this is just a selection as there are a lot of ‘minor’ concerns that need to be addressed by a very thorough revision of the paper)

+ Abstract: Not clear what the distinction between katabatic and thermally driven winds is. I think some explanation of the term 'thermally driven' is necessary here, as otherwise the reader is lost.

+ The introduction mentions one mode of variability, the SAM.  But what about the Amundsen Sea Low?

+ The katabatic winds are also dependent on the size of the slope.

+ Section 2.1.3: Not sure why the comment on the length of observations in the summer season is necessary.  And shouldn’t the number of AntAWS stations mentioned here, actually be mentioned in section 2.1.1. Comes across as disorganised.

+ Section 2.1.4: At least the third time that GCMs issues over representing complex orography has been mentioned.  Repetition. Makes the manuscript look extremely disorganised and amateurish.

+ Line 118. Typo. Grill -> Grid

+ Methods: Its not clear what the term ‘Implausibility’ is being used here for. 

+ Section 2.3.2: Poor paragraph structure.

+ Section 3.1: The preamble here is inappropriate / repetition.  This material should be in the Introduction or Methods, not repeated at the beginning of the results section. This weakens the paper and makes it look disorganised.    

+ Section 3.3: Similar comment to above, no need for the preamble.

+ Line 319: SAM defined again

+ Section 3.3.1: Huge amount of repetition on how SAM will change.