Supraglacial lakes serve as proxies for ice dynamics on the Antarctic and Greenland Ice Sheets, so recent research has employed a variety of techniques to derive lake depth and volume. This paper evaluates three platforms (ICESat-2, ArcticDEM, Sentinel-2) to estimate lake depth for five lakes on the Greenland Ice Sheet. A particular focus is given to the Sentinel-2 lake depths with a sensitivity analysis based on Sentinel-2 bands (green or red) and radiative transfer equation parameters. The authors found that ICESat-2 and ArcticDEM depths are generally in agreement for the five lakes, though both are limited in spatiotemporal coverage. Sentinel-2 is shown to be a more desirable platform to survey more lakes, but both the red and green bands struggle to agree with ICESat-2 and ArcticDEM. The authors conclude with a recommendation to improve calculation of lake bottom reflectance (A_d), which is shown to have a large influence on optically-derived lake depths.

Overall, I think this was an interesting study, and although the lake inventory is small compared to other studies, there is enough here to warrant publication in the TC. I especially appreciate the sensitivity analysis on the RTE parameters, and the use of ArcticDEM was unique.

Before I recommend final publication, there are a few issues that need to be addressed. They probably will not need too much extra analysis, but they warrant further discussion.

Major Comments

First, the conclusions suggest using ICESat-2 to constrain or correct the RTE methods. Although they do not use machine learning, Datta et al., (2021) uses ICESat-2 lake depths to constrain empirically-derived depths from Landsat-8, among other imagery sources. They also do not use the RTE equation used here, but I still think it is worth elaborating on how this study builds upon theirs (or how results from both could benefit the community).

On the subject of ICESat-2, its value for this study is not immediately clear to me. I assume that it serves as high-accuracy validation for the RTE and ArcticDEM methods, but this is not mentioned explicitly in the text. Also, ICESat-2 is barely discussed after Section 3.1. Between these issues and the sparse coverage over all five lakes (especially Lakes 4/5), the manuscript needs to provide more justification on why the ICESat-2 depths are useful here.

Minor Comments

Page 2, Line 32: For a general audience, I recommend mentioning why lakes prefer draining over refreezing (or vice versa).

Page 2, Lines 33-36: Hydrofracturing is formally defined on Line 36, so I would refrain from mentioning it until then.

Page 2, Lines 37-39: I suggest making this sentence more concise, or splitting it into two.

Page 2, Line 45: I would like to see a reference or two here demonstrating the importance of melt lakes for models, if possible.

Page 2, Line 51: Replace semi-colon with colon.

Figure 1: Nice figure! For panels 1-5, I would specify in the caption where the imagery is from.

Page 4, Lines 92-93: This sentence doesn’t add much justification – the following sentence is enough.

Page 6, first paragraph: This is overall well-written, but it is also getting a bit in the weeds. I suggest condensing it to something like: “Previous studies assumed that  K_u≈ 1-2.5K_d, with Brodsky et al., (2022) suggesting that higher K_u values (and therefore higher g values) lead to more accurate lake depths. Here, we use an average of the above range and take  K_u = 1.75K_d, or  g = 2.75K_d.”

Page 7, Line 183: Outdated reference. Refer instead to Markus et al., (2017) – see “New References” section below for full citation.

Section 2.4, first paragraph: I suggest noting that the spacing between ICESat-2 beam pairs is 3.3 km, which limits the coverage of individual lakes.

Page 7, Line 186: Cite Neumann et al., (2019) with the mention of ATL03 usage.

Page 9, Lines 208-209: Not sure if I follow the logic here. Why not keep ICESat-2 at its native resolution, if the other datasets are resampled to 0.7 m?

Table 1: Given Figure 3, I don’t think this table adds much to the paper. I think a table showcasing the maximum depth for each lake and method would be more useful.

Figure 4: This is a really nice figure, and I think it merits more discussion. In particular, I notice DEM/RTE differences that seem to be related to depth. Also, the green band has expectedly large underestimations in a few spots for Lakes 4 and 5. I would like to see some speculation in the Results or Discussion on these points.

Page 15, Lines 269-270: For visual reference, I suggest pointing out the “plateau depths” in Figure 5, using a dotted line or marker(s).

Page 15, Line 275: “the agreement between the red band RTE depths and the ArcticDEM depths decreases [as a consequence of red band saturation.]”

Figure 6 caption, first sentence: Suggest rephrasing to “Sensitivity analysis of RTE parameters, with plausible values given for each.”

Lines 298-300: It might be a bit much to call this a flaw in the method. Water reflectance is very low (~0.1 in the red band) unless specular reflection is observed, so you would need a very dirty lake bottom to achieve negative lake depths. This could be a more feasible issue for the green band, but I would imagine that it is still very uncommon.

Page 17, Lines 318-320: Just curious, what is a saturation depth for the green band? A ballpark number is sufficient.

Page 17, Lines 326-328: This sentence has redundant wording. I suggest revising to something like, “We determined that use of the green band RTE can lead to lake volume overestimations of more than 150% relative to ArcticDEM, with similar overestimations expected at larger scales.”

Page 18, Line 339: “…and consistent overestimations in the green band.”

Page 18, Lines 344-347: Long sentence, needs to be more concise (or split into two sentences).

Page 20, Lines 406-407: Given that only five lakes were observed, this is a rather strong conclusion to make.

Appendix A, Section 1: “Characteristics” ---> “Criteria”

Table A1: The ICESat-2 data used for this study is out of date – Version 006 was released in June. As a sanity check, I would see if there’s any significant differences in the ICESat-2 V003/V006 data over the five lakes.

New References

Markus, T., Neumann, T., Martino, A. et al., (2017). The Ice, Cloud, and land Elevation Satellite-2 (ICESat-2): Science requirements, concept, and implementation. Remote Sensing of Environment, 190, 260-273. https://doi.org/10.1016/j.rse.2016.12.029.

Neumann, T., Martino, A., Markus, T. et al., (2019). The Ice, Cloud and Land Elevation Satellite-2 mission: A global geolocated photon product derived from the Advanced Topographic Laser Altimeter System. Remote Sensing of Environment, 233, 111325. https://doi.org/10.1016/j.rse.2019.111325.