General comments

The study utilizes a novel data fusion of satellite data from ICESat, ICESat-2, and REMA strips, providing convincing evidence for the presence of active subglacial lakes in the coastal DML (Dronning Maud Land) region. Additionally, the authors integrate a stochastic analysis for subglacial water routing adding another novel analysis to demonstrate the uncertainty in water routing predictions. The discovery of active subglacial lakes in coastal DML and stochastic water routing adds new insights to the understanding of subglacial hydrology.

Weaknesses include the limited spatial scope of the study, undescribed methods regarding the stochastic water routing analysis (and unreleased code), and numerous technical mistakes outlined in the Technical Corrections section.

Specific comments

Throughout

• You continually reiterate that height changes are inferred as subglacial volume changes; I think this repetition is unnecessary and clutters your manuscript; this method of inference is commonly used in other papers and does not need to be repeated ad nauseam

1 Introduction

• You do not explain the difference between active and stable volume lakes despite discussing both in your manuscript. Using a sentence or two to do so will help your reader.

3.3 Subglacial water flow

• Did you use a uniform melt rate or spatially variable melt rate (if spatially variable, which one?)? The methodology is under described and quite important to interpreting your results.

Figs. 2, 3, 4

• The secondary colormap you select for ICESat and REMA both have blue colors for the positive height anomalies (this is typical for subglacial lake and mass gain/loss papers); however, the colormap you select for ICESat-2 uses blue for negative height anomalies; this discrepancy makes the figures challenging to interpret; I suggest you reverse the colormap for the ICESat-2 data so that positive height anomalies correspond to the blue colors of red-blue colormap
• Using a non-diverging colormap for ICESat and REMA height changes makes it unclear what color corresponds to zero, making height anomalies challenging to interpret. I understand the desire to have separate colormaps for each instrument (ICESat vs. ICESat-2), but this is accomplished by using the annotated ICESat/ICESat-2 track/RGT numbers. I would suggest a diverging colormap for both data sources (ICESat and REMA) and use annotation for ICESat tracks and ICESat-2 RGTs and the figure caption to communicate the gridded height changes are from REMA. 
• (Fig. 2, but could be added to Figs. 3 and 4 too): In the figure caption, you should refer to ICESat-2 tracks as reference ground tracks (RGT) (vs. tracks) since you use ‘RGT’ in the figures’ annotations; this will additionally help readers know which track belongs to which instrument.

Technical corrections

Fig. 3

• a) Missing X-X’ annotations to correspond with Fig. 3b
• b) and e) you say profiles are relative to cycle (cyc) 3; however, you do not plot cyc 4; perhaps use a caption or methods to explain why this cycle is omitted from your analysis/plotting; perhaps use methods to explain why cyc 3 is the reference cycle (vs. something you might expect like cyc 1)?

Line specific

10: “previously-identified lakes” should be “previously identified lakes” as compound adjectives of adverb and past participle are not hyphenated.

10-11: “Most previously-identified lakes have been found upstream (>100 km) of fast-flowing glaciers in West Antarctica”

• Consider changing to ‘fast-flowing ice’ or fast-flowing ice streams’ as much of the fast-flowing ice in West Antarctica are ice streams, not glaciers

24: “Hydrologically-active subglacial lakes…” "hydrologically active" does not need to be hyphenated because "hydrologically" is an adverb modifying the adjective "active." Adverbs ending in "-ly" and the adjectives they modify are typically not hyphenated.

30: “...lakes can range from ~5 km2 to tens of square kilometres”

• This statement is inaccurate as there are many active subglacial lakes in the hundreds or even thousands of square kilometers; consider some examples: Byrd_2 is 725 km^2 and the largest lake, Nimrod_2, is 1257.9 km^2 (Siegfried & Fricker, 2018)

32-33: “Downstream subglacial water flow has been linked to cascading lake drainage events which transport excess water episodically towards the grounding line”

• I think there are earlier pub’s to cite for this point: Flament and others, 2014, Siegfried & Fricker, 2018

39-40: “Over the past two decades, 140 active subglacial lakes have been detected…” This number is from the Livingstone and others, 2022 review paper, which does not include lakes from the Neckel and others, 2021 paper you cite elsewhere, which argues they have found more active lakes, so wouldn’t you say the detected number of active lakes is greater than 140?

45-46: “Few active subglacial lakes have yet been reported beneath much of the grounded ice close to the Antarctic Ice Sheet margin (Livingstone et al., 2022).”

• Perhaps ‘coastal’ is a better word instead of ‘margin’; active subglacial lakes are often described as being in the marginal fast-flowing ice regions vs. stable volume lakes that are the continental interior

55-57: “We further estimate subglacial stream probability using water routing analyses derived from stochastic simulation (Shackleton et al., 2023) to assess upstream drainage basins and potential downstream impacts of the newly observed subglacial lakes.”

• I read your manuscript as assessing drainage pathways (not basins as in the boundaries between basins or quantifying the areas/spatial extent of basins); I would change ‘basins’ to ‘pathways’ to reflect this

58: “previously-unreported active subglacial lakes” 

• See comment for line 10

71: “subglacial lakes 40 km or further inland” For American English this should be ‘farther’ since you are referring to a physical distance but I understand British English is more lax with with the farther/further distinction

72: “(7.2-16.2° E)” longitude reference doesn’t seem useful to me

119: Described methodology contradicts Fig. 3 where you omit cyc4

122-23: Your citation for this sentence (Zwally and others, 2002) says footprints are 60 m, not ~65 m as you state; why the discrepancy? Is ~65 m estimate perhaps from post-mission launch analysis? If so, you choose a different citation or add a citation for that

126-7: “...at the point between successive ascending and descending passes over the same location” 

• would be clearer if you said point of “overlap” or “intersection”; “between” implies a gap between the ascending and descending passes 

128: how is error lower than the range you report that includes flat surfaces?

136-8: “We further neglected potential long-term elevation changes due to surface mass balance and large-scale ice dynamics in the plane fitting as these are generally small in the study region and could interfere with changes due to subglacial lake activity.” A citation would be useful here especially since earlier in the manuscript you contradictorily say this region “has recorded significant ice-sheet thickening in DML over the last two decades (Smith et al. 2020) due to high snowfall rates (e.g. Boening et al., 2012).” (78-9)

150: Why is “Differencing” capitalized?

150-151: these two sentences seem contradictory: did you use ICESat elevation anomalies to select REMA strips or not? First sentence suggests not while the second sentence suggests yes.

171-174: I applaud you specifying your methods to delineate the lakes; most papers skip this important detail

601-2: I applaud JA and GM for releasing your ICESat/ICESat-2 analysis code; however, your team (CS and KM) could do more by releasing the code used to conduct the subglacial water routing stochastic simulations, especially considering this analysis relies on an open-source tool, GSatSim (Mackie and others, 2023); why not contribute more use cases to this project by releasing your code? 

596-599

• Data availability section lists a data repository DOI for the DML lake outlines; however, trying to download the outlines results in the following error: “Failed to authorize the download request.” (it’s not clear if the file embargoed; perhaps adding this to the dataset description would help those trying to access the files)
• A similar data repository DOI is listed for the subglacial routing pathways however the DOI results in a ‘404 not found’ error
• Thus the data sets produced from this work are currently inaccessible and The Cryosphere policy requires stating when they will be available and how they can be accessed until publicly available (“If the data are not publicly accessible at the time of final publication, the data statement should describe where and when they will appear, and provide information on how readers can obtain the data until then.” from https://www.the-cryosphere.net/policies/data_policy.html)

382-83: “Ice thickness above these three lakes” in reference to Takahe Lakes (TL) below Haynes Glacier detailed in Hoffman and others, 2020 should be “four” lakes (See Fig. 1a or Supplement Fig. 1 to see four lakes)

391-3: This sentence could be clearer by stating that many marginal regions are predicted to have cold beds and selected a different citation to make this point:

• Pattyn and others, 2010 (doi:10.1016/j.epsl.2010.04.025) may be a better citation here as it has figure a figure of geothermal heat flow and likely warm/cold beds that would better illustrate your points related to geothermal heat flow and basal temperature and their Fig. 3c of mean basal melt rate that would bolster your point that the marginal DML region has a low probability of basal melt from modeling

406: “su1ggested” is misspelled

425-7: You could cite subglacial sediment probability paper (Li and others, 2022, doi:10.1038/s41561-022-00992-5) to bolster argument that bed is likely permeable and you would not see seawater intrusion at these length scales here

434: “quiescence (filling)”: parenthetical “filling” doesn’t make sense because quiescence is not filling or draining

435: “known active lakes” some would argue that the only active lakes we ‘know’ are those we’ve drilled to for in situ sampling; perhaps it’s better you stick with your previous terminology of “previously identified active lakes”

Summary

This study provides an updated account of active subglacial lakes in Dronning Maud Land from various satellite observations, with seven new lakes identified near the grounding line. Stochastic bed simulations are used to probabilistically model subglacial flowpaths and connect the lake drainage with ice shelf channels. Overall, this manuscript is nicely written with a sound methodology. The results are novel and will be of broad interest to TC readers.

I was excited to see geostatistical simulation being used here, but the methods require further elaboration. I made up the concept of using clustering to divide spatial domains into different variogram regions in order to account for non-stationary when using sequential Gaussian simulation (MacKie 2023), so this is by no means a standard and widely recognized method. While section 2.6 makes perfect sense to me, I suspect this description won’t make much sense to most readers - even those familiar with geostatistics. As such, I recommend including a more detailed description of the method. Additionally: Which variogram model type was used? Why 12 clusters? Showing a map of the cluster boundaries might be helpful. Was the clustering based solely on spatial coordinates (lat, lon or x, y?), or were other variables used as well? It would also be nice to include an image of a topographic realization, or several realizations in the supplement. 

There was some interpretation and discussion in the results section that should be moved to the discussion section.

Figures

Overall, the figures are complete and aesthetically pleasing. 

I was a bit confused by the flight lines in Figure 1, which shows flight lines at latitudes above 72 degrees in part a, but shows <100 m bed uncertainty flight line shapes below 72 in part b. What do the lines in part a mean, if there was more radar data used in the interpolation? In the legend in part a, “Existing airborne radar data” is denoted by a light gray bar. However, the radar lines in the figure appear much darker. The active and radar detected lakes in the Antarctic map in the upper corners needs a legend. The Goeller et al. lakes are purple in the large panel, but turquoise in the Antarctic map, which is confusing.

In figures 2 and 3, it’s difficult to visually compare the ICESat and ICESat-2 colors because one uses a diverging colorbar and the other is a sequential colorbar. I’m also confused by the numbers on the colorbars. They are centered at 0, but aren’t scaled the same way on either side? 

There are visual inconsistencies between the figures. For example, sometimes lake outlines are solid red, but in other figures they are dashed black lines. Dashed black lines are used to denote lake outlines, ice shelf channels, and grounding lines in different figures. In some figures, radar lines are purple, but in others they are grey. Sometimes the grounding line is solid yellow, sometimes it is dashed black, and sometimes it is solid black. I know this is sometimes unavoidable, but it would make the figures much more interpretable to have more continuity throughout the paper. 

Sometimes figure parts are shown as a without parentheses and sometimes it’s shown as (a). Make sure that figure parts are labeled according to the journal figure formatting requirements.

Line comments

L45: “Few active subglacial lakes have yet been reported beneath much of the grounded ice close to the Antarctic Ice Sheet margin”

Can you specify what you mean by “close”? I thought that active lakes were generally close to the grounding line.

L189: “...based on an ensemble of water routing analyses following the approach of Shackleton et al., (2023).”

May I humbly suggest that you cite MacKie et al. 2020 or 2021? These were the original studies to perform an ensemble geostatistical water routing analysis.

L196: “...  calculated the experimental variogram…”

I assume the SciKit-GStat package was used to do this, which should be cited:

Mälicke, M. (2022). SciKit-GStat 1.0: a SciPy-flavored geostatistical variogram estimation toolbox written in Python. Geoscientific Model Development, 15(6), 2505-2532.

L458: “Also, subglacial channels in these regions could also be ephemeral and only form during lake drainage events (Smith et al., 2017)”

Remove one “also”