Review of the manuscript: “High-resolution imaging of supraglacial hydrological features on the Greenland Ice Sheet with NASA’s Airborne Topographic Mapper (ATM) instrument suite” by Studinger, Manizade, Linkswiler and Yungel.

General Comments

This study combines lidar data from NASA’s Airborne Topographic Mapper instrument with high-resolution optical imagery from Operation IceBridge to develop methods for mapping and calculating depths of supraglacial hydrology over the Greenland Ice Sheet. The authors argue that, although, airborne surveys have limited temporal and spatial coverage they offer a level of detail which lower resolution, spaceborne instruments cannot.

The data and code provided will allow for replications of the methods/results and the appendices provide context to the main manuscript. The introduction and references provide a solid overview of the available literature, while the title accurately describes the contents of the paper.

The manuscript is generally clear and easy to follow; however, some sentences and sections are difficult to read and are highlighted in Technical Corrections. To the best of my knowledge the methods presented are novel with a sound conclusion. I have some specific comments and additional technical corrections which are listed below.

Specific Comments

Although the authors state: ‘The robustness of the hydrological feature identification is suitable for the purpose of this paper, since the goal is not to perfectly delineate hydrological features, but to identify lidar granules for analysis.‘ I feel the study would benefit from some validation/accuracy metrics for the approaches to determine ‘snow and ice’ from ‘water’ and the water depths presented in Figures 7, 8 and 9. The conclusion could then be improved by including the results of these computed metrics.

Figure B3 summarises the methods coherently and should be considered for transfer to main manuscript.

Whether Section 3 is part of the methods section is unclear. The manuscript would be improved with a paragraph (at the beginning of the section) describing the methods, as in Sections 2 and 5. Additionally/alternatively the clarity of Section 4 could be improved with sub-headings. Similarly, Section 5.2 could benefit from further subheadings.

Technical Corrections

Line 22: insert comma between ‘months’ and ‘seasonal’.

Line 34: insert comma between ‘scale’ and ‘the network’.

Line 46: insert comma after ‘cryoconite’.

Line 47: insert comma after ‘crevasses’.

Lines 48-50: suggest rewrite of the sentence beginning ‘Yang and Smith’.

Line 50: suggest rewrite to ‘However, the recent increase in availability of…’

Line 51: insert comma after ‘streams’.

Line 52: insert comma after ‘wide’.

Lines 65-67: suggest rewrite of the sentence beginning ‘Airborne’.

Line 74: insert comma after ‘strategy’.

Line 76: insert comma after ‘Spring campaigns’. Should ‘summer’ be capitalised as ‘Spring’ is, for consistency? Should ‘flow’ be ‘flown’.

Lines 80-83: suggest rewrite of the sentence beginning ‘We analyze’.

Line 93: should ‘angle’ be ‘angles’?

Line 114 & 140: Section headings are confusing. Are these sections both ‘Methods’? If so, suggest re-label as 3.1 and 3.2. Methods section could then have an introduction as other sections do.

Figure 3: ‘[ ]’ is confusing following ‘NDWI_ice’.

Line 134: replace ‘expense’ with ‘expensive’.

Line 139: suggest remove ‘the purpose of’.

Line 157: suggest replace ‘Over’ with ‘For’ and ‘in estimating’ with ‘to estimate’.

Line 158: insert comma after ‘cases’.

Lines 161-163: suggest rewrite of sentencing beginning ‘For some’.

Line 178: suggest insert ‘the’ before ‘ground test’.

Lines 257-259: suggest rewrite of the sentence beginning ‘The gaps…’.

Line 266: suggest replace ‘reaching maximum’ with ‘reach the maximum’.

Line 297: suggest add ‘the’ before ‘recording’.

Line 300: insert comma after ‘lake’, before ‘there is a’.

Line 309: replace ‘shows’ with ‘show’.

Line 312: suggest replace ‘part’ with ‘parts’.

Line 316: suggest remove ‘back’.

Line 317: suggest insert ‘those’ between ‘as’ and ‘found’.

Line 338: suggest insert ‘as’ between ‘acts’ and ‘a specular’.

Appendix:

Figure A2: Y-axis labels should contain a unit within ‘[ ]’.

Line 380: In Figure A2  caption the sentence ‘For T6 28084 waveforms were stacked and for T7 48703’ should be rewritten to be more clear.

Line 423: suggest insert comma after ‘In pressurized aircraft’.

Line 441: The line ‘https://doi.org/10.5281/zenodo.6248436 (Studinger et al., 2022).’ appears out of place, with multiple blank lines above it.

Line 457: suggest insert comma after ‘altimetry’.

Line 480: Should ‘Sensitivity analysis’ be presented as a heading like ‘Range bias’, ‘Gaussian fit’ etc.? If so, it should be formatted as sections above and the colon removed.

This paper describes as workflow for combining high-resolution optical aerial imagery and airborne lidar measurements to identify supraglacial water features and estimate their depth. It then shows examples of the obtained results over three different water features with different characteristics and discusses the capabilities and limitations of the method in each of these environments. In particular, the authors argue that airborne data can provide a high-resolution data at large spatial scales that fill a gap between spaceborne and field-based measurements.

This work is technically correct to the best of my knowledge and provides a useful for studying supraglacial bathymetry from the existing OIB ATM data record. The authors have published their codes, which makes the method easily reproducible. However, the manuscript lacks a robust quantification of uncertainty in the bathymetric measurements and could benefit from some discussion of where these methods might break down.   

Major Comments:

[1] If I have understood the paper correctly, the primary contribution is methods for retrieving the bathymetry of supraglacial lakes and streams at high resolution from airborne lidar measurements. The introduction provides a strong review of existing methods for estimating lake and stream depth and their limitations. However, the motivation for this study would be would be strengthened by a little bit more discussion of the science that would be enabled by these more accurate or higher resolution bathymetric measurements. Would this improve our understanding of total water budget and volume of melt impounded in the surface network? Energy balance controls on channel morphology and evolution? Evidence for past prior hydrofracture events in flooded closed basins? A brief discussion of why bathymetry in particular is worth improving would be useful.

[2] The study makes some general statements about “methods to study supraglacial hydrological features” (see for example lines 69-70). The contribution really seems to be to studying the bathymetry of these features and I would encourage the authors to be precise about this throughout the paper.

[3] I would encourage the authors to move Figure B3 to the main text early in the paper (perhaps as a new Figure 3), as it would provide a very effective roadmap for the methods and might clarify the overall structure of the paper and the approach for the reader early on.

[4] Is it possible to characterize the uncertainty in the final bathymetric estimates based on the uncertainty in the various calibration and correction steps, the SNR, etc? What is the typical error or bias introduced by picking the peaks of the Gaussian fit, rather than the direct waveform peaks? In general, the paper would be strengthened by a dedicated section discussing possible sources of uncertainty in the bathymetric estimates and how they might be identified, mitigated, or quantified. Figure B2 may partially address this question, but it did not get much discussion or explanation in the text.

[5] The authors argue that errors in the NDWI classification are not problematic, because the lidar analysis will look for surface and bottom returns and so even areas that might be misclassified as water ultimately will not be assigned a depth estimate because there will only be a surface return in those areas. However, the authors later go on to discuss how they estimate the location of the surface return when only a bottom return is visible in the lidar waveform. This raises the question of how they classify as single return as being a bottom return rather than a surface return and whether that classification relies on the assumption that the NDWI classification is accurate. Similarly, I wonder if there would be cases where the lidar waveform might have multiple peaks caused by something other than penetration through water (could complicated crevasse networks or ice damage cause similar returns?). Overall, the paper might benefit from some discussion of when and where (what types of physical environments) the authors would expect the assumptions inherent in their workflow to breakdown. Similarly, it would be good to mention whether this workflow is applicable to every generation of the ATM data currently published, or if different seasons would require any tuning of the methods.

Minor Comments:

Line 85 – I suppose it does not matter that much, but it is not clear why the fall campaign needs to be discussed since none of the data contributed to the methods development or analysis presented in this paper.

Line 119 – How reliable/consistent is the 0.05 NDWI threshold? How radiometrically stable are the CAMBOT images? Would this threshold require tuning for data from different seasons or flight dates?

Line 210 – The flight lines in southern Greenland largely overly areas that are either above the typical visible surface runoff limit or where firn aquifers have been identified (see Miege, et al (2016) “Spatial extent and temporal variability of Greenland firn aquifers detected by ground and airborne radars”), so it does not seem particularly surprising that there would not be surface lake detections in these areas, particularly in May.

Line 355 – Do you have a sense for the theoretical penetration depth that should be possible with the instrument, or is that too dependent on variable environmental parameters?

Line 480 – Presumably it is also a function of the relative amplitude difference between the two peaks? Given the shape of a Gaussian pulse, you would need greater separation between two pulses with very different amplitudes to separate the peaks, compared to two pulses of approximately equal amplitude. Can this be quantified and how much of a concern might it be for bathymetry measurements where there seems to be significant variation in the relative amplitude of the surface and bottom returns?

Line 480 – An in-text discussion of Figure B2 and how those calculations were carried out would be helpful. I do not fully understand where the 250,000 non-linear regressions come from. Is this from fitting stochastic realizations of the waveform with different noise levels?