Dear Sheng Dong and coauthors,

Thank you for your detailed reviewer responses. At this stage I encourage you to implement the changes you have detailed in your response. I am unclear whether you intend to implement the methodology changes suggested by Reviewer 1 (RC1) in their comment "L185-192: Identifying subglacial water using clustering analysis". Please implement these changes or justify why you do not consider this necessary.

Please also make changes or further justify your response to suggestion (ii) provided by Reviewer 2 (RC2) "ii) Instead of the clustering, the authors can identify where the currently known subglacial lakes are located in the latent space (i.e., plot these samples in Figure 3a)." I appreciate your response but the reviewer suggestion is a sensible one and I'm not sure your one trace justification is adequate justification not to follow the advice given.

Finally, thank you again for your high quality submission and responses. I look forward to receiving your revised manuscript.

Sincerely, Huw Horgan

Dear Sheng Dong and coauthors,

Thank you for the submission of your revised manuscript. I have reviewed your submission and appreciate the time you have spent implementing the reviewer's recommendations.

It is my opinion that your manuscript still requires some changes prior to acceptance. You have included the relevant content but there are now some structural issues as much of the requested information has been placed in the discussion section while it is more suited to methods and results. Specifically I am referring to the method used to select the number of clusters used in your K-means cluster determination. You have done a good amount of work demonstrating your reasoning and the impact of your selection.

Specific changes I think should be implemented include:
-Include the details on the number of clusters selection (currently paragraph 2 of your discussion) in your Data and Methods section.
- Include details on changing number of subset reflectors in cluster analysis in Data and Methods.
-Shift the figures that show the sensitivity to number of clusters from discussion to results.
-Include the K-means elbow analysis in the main body of the manuscript (Data and Methods).
-The presentation of the spatial distribution of latent space distance also seems better suited to Results.

Smaller technical changes:
L3 'in the' > 'from the'
L6 'are used' > 'are then used'
L7 (and elsewhere) 'known lakes inventory' > 'known-lakes inventory'
L23 Paden and Arnold references are not the correct references here.
L37-L46 consider combining these short paragraphs with sentence linking existing basal picks to deep learning suitability.
L50 'Using this' > 'Using these'
L62 'We will introduce' > 'we introduce'
Figure 1. Should include the spatial and power filtering applied
L69 'lake inventory' > 'lake inventories' Correctly bracket references.
L84 'which maintain' > 'which maintains'
L85 'length of 128' > 'length of 128 samples'
L97 Include a short definition of 'latent space' so as not to lose the non-experts
L106 'easier presenting' meaning unclear, please clarify
L120 'VAE' > to 'the VAE' (twice)
L127 'low dimensional bottleneck' briefly define
L138 '2-length' > 2*1
L144 'intact dataset' meaning unclear, clarify
L154 '...amount of data used.' citation needed.
L158 'which based' > 'which is based'
L166 '(Figure S2)' present in main body.
L168 'ultimately selected K = 15' justify.
L186 'continuously distributed' > 'spatially continuous'
Section 2.4 include subheadings 'Spatial continuity' 'Interpolation' 'Depth-dependent echo power filtering' or similar.
L190 'In further applications of observational data' Meaning unclear, probably not needed. > 'It has been...'
L191 'radar signals.' citation needed.
L209 'Considering....' Hard to parse this sentence. Please clarify.
L224 '...similar in the ...' > '...similar at the length-scales covered by the radar image.' Or similar.
L254 '...than the established thresholds.' > '...than the thresholds used here.'
L244 'long-distance' > 'spatially extensive'
Figure 9 caption 'subglacial lakes or lakes' and 'regional subglacial lakes' meaning unclear, please clarify.
L291 'distinct phases' meaning unclear. Clarify.
L297 'Besides, ' > 'Also, '
L302 'submerged' > 'obscured'
L303 'Besides, the' > 'The.....can also provide'
L304 'are essential' > 'will help'
L305 'miss-detection' > 'false positives'
Discussion. Much of this first paragraph belongs in Methods and Results.
L353 'Compared with' > 'Compared to'
L372 'automated' > 'mostly automated'
L381-382 odd tangent to end on
Conclusions. These are very brief.

With best wishes and thanks for your interesting contribution,

Huw Horgan

Dear Sheng Dong and Coauthors,

Thank you for your revised submission "Deep Clustering in Subglacial Radar Reflectance Reveals Subglacial Lakes". I now consider your manuscript to be ready for copy-editing and subsequent publication in The Cryosphere. Congratulations on making a useful contribution to the field.

With best wishes, Huw Horgan