Dear authors,

Overall, the paper is of good quality. There are some interesting new insights into the region presented and utilizes unique methodological techniques (synthetic simulations for waveform comparison.

There are several inconsistencies in the interpretations which I discuss in more detail in the following comments. To summarize, there is an issue in figure 6 and 9 with the polarity of the observed/synthetic waveforms and their interpretation. The glacier-bed interface appears to be a negative polarity to my eyes but is interpreted as positive. Further, figure 9a compares glacier-lake synthetics and real data, but the polarity of the first arrival here appears normal, not reverse. Providing a definition of how you are defining positive vs. negative may help here. This leads into the next key issue which is the use of reflection coefficient as the color axes for the figures. It is my understanding that you are migrating, normalizing, and stacking, and therefore without correction for geometry or attenuation/path effects, this is not representative of the reflection coefficient at the base. The absolute polarity is correct, but the value is not the coefficient. This brings me to my third issue, which is the assumption of liquid water at the base. I understand there is abundant evidence without seismic data for drainage and refilling of this region, but I feel that more can be done to interpret the subglacial structure via reflectivity analysis rather than assuming structure for your synthetic calculations. This would also likely help explain some of the amplitude variability seen across the lines. Similarly, I wonder why the authors choose to use published values for firn/ice velocity when these are things that can be calculated from their data. Note, I did not find an attached supplement. If there is one that I missed, my apologies. If not, I think a supplementary file which answers some of the questions about both the interpretation but also the data processing as well.

I recommend this paper is accepted with some necessary revisions. This is a very rich dataset that produces nice subsurface images. I believe there are a few further steps to be taken in the analysis and interpretation such that the conclusions the authors make are more strongly supported by direct observation.

Overall, I found the paper to be well written and engaging, and the results to be significant and in line with the objectives of The Cryosphere. Thank you very much for the opportunity to review.

Specific Comments:

• I wonder if the extensive processing scheme is fully necessary, given reflections are quite strong in the raw data. For example, two rounds of curvelet filtering and spatial filtering via interpolation seem redundant. At the very least, could you provide more information about some of the key filtering parameters? Specifically, to remove ground roll?
• In figures, be a bit more specific about geometry. Add a 21X’ on one end for example to see which direction the record sections are. You can figure it out, but it might make them a bit clearer
• Figure 4 and 6: the colorbar is not appropriate. My understanding of what’s plotted is the normalized seismic section between -1 and 1, which would be equal (or rather proportional, equal after accounting for geometric spreading, source, attenuation etc) to the reflection coefficient only at normal incidence. I think the colorbar should either be labelled as normalized stacked energy or actual units given.
• On this point, I wonder why not try to calculate an absolute reflectivity curve (AVO Horgan et al. 2021; Peters et al. 2008) or relative reflectivity curve, to confirm you are truly seeing water at the base rather than some kind of wet sediment layer, which would also result in a negative polarity PP reflection. Even if you just do this for one of the lines, the coherence with the analytical reflectivity curve would make the argument much stronger.
• Figure 6e: I’m a bit confused on the interpretation here. The glacier-bed ghost (4) appears to be normal polarity, while the direct reflection appears to be reverse polarity? This seems opposite to the discussion in the analysis. I’m wondering if I am seeing something wrong, or perhaps the color scale is flipped for this inset?
• Scours: Really interesting interpretation, but I wonder how you can be sure these features are from depths rather than near surface or englacial crevassing? They look similar to observations from stations near shallow faults which show significant scattering or spurious arrivals from interactions with reflected phases. I also wonder if you could use cross correlation/autocorrelation with the direct P to increase the coherence of the reflected arrivals and to better resolve the shape of these features? Further, while I see that the scour features interfere significantly in the synthetic, the observed amplitude is still comparable to the direct reflections, which is again not true in the data. I wonder if instead, the near surface scattering argument is invoked, would you expect significant defocusing and thus lower amplitude?
• I am wondering about the firn velocity model used: If I do a linear interpolation, I see that the average firn velocity in the upper 25 meters is around 1800 m/s, consistent with what is stated in the text. However, a consistent feature of the firn is the steep velocity gradient/density near the surface which shallows with depth, which is very nonlinear. I believe this effect is probably small, but in terms of the raytracing, the incident angles of the incoming waves in the simulations of a linear vs nonlinear firn will be very different, which affects reflectivity. If you calculate your own Vs firn model, or if you use an empirical relation/calibration for your firn model, you could solve for a firn to ice transition depth which would be another interesting finding of the paper.
• I wonder if the full processing scheme need be applied to the synthetic data. Given the extensive processing combined with the lack of ambient noise/crevasse scattering, the data should basically be clean enough to examine on its own. I understand for consistency’s sake why the same processing is applied, but here, I would emphasize that the synthetics before and after processing remain pretty much the same. If they do not, that could point to artifacts being introduced in your processing scheme.
• I think Figure 9a also has a polarity issue. This is showing the synthetic and observed seismic data shot gathers with arrivals corresponding to the glacier lake interface, but the first arrival in both synthetic and observed appear to be normal polarity as opposed to the reverse polarity that is stated. Am I interpreting these figures wrong? IN addition, back to the glacier-bed interface, the synthetic phases 3, 4 representing the glacier-bed and ghost have normal and reverse polarity respectively, while the field observations seem to have reverse and normal polarity respectively. Is this a color bar issue, or am I interpreting these images wrong?
• On figure 9, again, the same issue with reflection coefficient used as the colorbar. Neither the migrated sections nor synthetics are showing reflection coefficient, but rather normalized amplitude. I think this gets into a slight issue I have with the results, which is that apart from travel time calculations, they are mainly qualitative. The velocity of the subsurface is assumed (water, bed) rather than estimated via something like AVO to give you actual reflectivity. This means that while yes, the synthetics have the same shape, their amplitudes are not consistent with each other both laterally and in time (later arrivals are significantly weaker). This is likely due to a combination of attenuation and structure, but it was unclear whether this is considered in the modeling. I do feel that this comparison is only relevant for the geometrical features such as the scour surfaces or the steep bed slopes.
• Additionally, why are the timings of the synthetics and field data off? The synthetics arrive consistently earlier than the observations. Does this mean the depth used is too shallow? Or the firn is not fully accounted for?
• Generally, I am not exactly sure how the results compare to the previous estimates of the D2 boundaries. From my understanding, the seismic interpretation is that the lake is on the opposite side of the marked boundary in orange on Figure 3. This could be me confusing the direction of the cross sections, but I think that emphasis the need for more clarity about the geometry of the features described.
• Figure 10: Could you provide a reference for the resolution of both the IPR and seismic data? Perhaps on the plot, add points for the individual datapoints for the IPR line, which looks sparser? Additionally, why is the seismic prediction almost always slightly thinner, on the order of tens of meters? I wonder if including the uncertainty bars would help show the agreement between the two.
• I think the use of published values for glacial firn, ice, water is fine for a starting model, but there is significant variation across estimates of these values (Yang et al. 2024; Picotti, Carcione, and Pavan 2024; Agnew et al. 2023). Also, these values or at least estimates of them should be visible during the velocity analysis stage. You could get a P wave velocity from the PP observation on the velocity/depth section and use that for the ice velocity in your model rather than a published one. Perhaps they will agree, but it’s worth checking. Also, firn velocities for P wave can dip below 1 km/s (Yang et al. 2024), perhaps this can help explain the timing discrepancy you observe? Again, I wonder if you can get an estimate for the shallow velocity from the observed surface waves or pseudoacoustics.

My suggestions for revisions are as follows:

• Data presentation and Interpretation: Provide a figure either in the main text or supplement that shows the clearest example of the reverse polarity reflection, perhaps accompanied by a synthetic. I think that part of the confusion is the difficulty in making out features on the record section as well as the slight inconsistencies in terminology. If you can also show the normal polarity reflection from the glacier-bed clearly, perhaps showing the actual waveform as done in 9a, the difference between the two regions would be immediately clear.
• Modeling: Calculate Vp, Vs in firn and ice. It doesn’t have to be in depth, as I understand that is not the point of the paper, but an average value or upper/lower bounds to at least confirm that your simple model is appropriate would help strengthen the interpretation and perhaps resolve some of the discrepancies between synthetics and observations. If you don’t want to derive these values from the data, I suggest trying different basal models (wet sediment/till) to confirm that your simple 2d model matches the data best.
• Supplementary information: Providing supplementary information, particularly on the data processing and interpretation, will clear up a lot of uncertainty about the signals that are observed. For example, the scour features; while you mention removal of surface waves and surface scattering, the detail given is not sufficient to explain exactly what the resultant signal we are seeing here is. Showing the raw data, or a panel with the data for each or important processing steps could help clarify your certainty in interpreting these features as structural rather than artifacts from the processing. I believe that at minimum a table which describes the processing parameters (anomalous amplitude attenuation, curvelet filtering, surface consistent amplitude compensation) is necessary for the sake of repeatability. Supplementary information can also contain a more thorough exploration of the model space.

Line Specific Comments within attached PDF.

Agnew, Ronan S., Roger A. Clark, Adam D. Booth, Alex M. Brisbourne, and Andrew M. Smith. 2023. “Measuring Seismic Attenuation in Polar Firn: Method and Application to Korff Ice Rise, West Antarctica.” Journal of Glaciology 69 (278): 2075–86. https://doi.org/10.1017/jog.2023.82.

Horgan, Huw J., Laurine van Haastrecht, Richard B. Alley, Sridhar Anandakrishnan, Lucas H. Beem, Knut Christianson, Atsuhiro Muto, and Matthew R. Siegfried. 2021. “Grounding Zone Subglacial Properties from Calibrated Active-Source Seismic Methods.” The Cryosphere 15 (4): 1863–80. https://doi.org/10.5194/tc-15-1863-2021.

Peters, L. E., S. Anandakrishnan, C. W. Holland, H. J. Horgan, D. D. Blankenship, and D. E. Voigt. 2008. “Seismic Detection of a Subglacial Lake near the South Pole, Antarctica.” Geophysical Research Letters 35 (23). https://doi.org/10.1029/2008GL035704.

Picotti, Stefano, José M. Carcione, and Mauro Pavan. 2024. “Seismic Attenuation in Antarctic Firn.” The Cryosphere 18 (1): 169–86. https://doi.org/10.5194/tc-18-169-2024.

Yang, Yan, Zhongwen Zhan, Martin Karrenbach, Auden Reid-McLaughlin, Ettore Biondi, Douglas A. Wiens, and Richard C. Aster. 2024. “Characterizing South Pole Firn Structure With Fiber Optic Sensing.” Geophysical Research Letters 51 (13): e2024GL109183. https://doi.org/10.1029/2024GL109183.