Dear Lucas and co-authors,

your study analyses the seasonal and decadal evolution of the supraglacial lakes on eight neighboring debris-covered glacier in the Khumbu Region, Nepal. Therefore, the satellite images of PlanetScope, Sentinel-2, and Landsat 5-9 were used. The high temporal resolution of PlanetScope enabled the monitoring of the seasonal evolution and Landsat of the decadal evolution since 1988. For the classiciation of the supraglacial lakes, the Normalized-Different Water Index with manually optimized threshold was applied. The study gives an interesting insight into the lake evolution in the Khumbu Region. Unfortunately, the manuscript needs major revisions (see below). The findings must be better elaborated and the results must be specified with exact numbers like lake area or number. If the maximum lake area is End of April, which is differently to other studies, the statement should be underlined with better explanations or evidences and hence to strengthen your key findings. I hope my comments are helpful and I am looking forward to reading the revised manuscript.

All the best,

Anna Wendleder

Major corrections:

1. The chapter “Methods” is very comprehensive and detailed processing steps like downloading the data or manual check are unnecessary. It would be good to revise this chapter, focus only on the important steps, and shift all important information about the three sensors in the new chapter “Materials”.
2. What are your key findings regarding the method and the classification results? Please elaborate both points and underline your results more with exact numbers. You derived the supraglacial lakes for 35 years but without mentioning number, maximum area, or date of maximum area. A table with all these parameters would improve your study immense. Furthermore, it would be interesting to analyze why the supraglacial lakes on the eight neighboring glaciers behave differently.
3. Reduce the figures - few meaningful figures would underline your core message.
4. PlanetScope gives insight into the seasonal evolution and Landsat into the decadal evolution. What is the benefit of using Sentinel-2 in your study? If you would fuse the classification results, then it would make sense to use Sentinel-2 as well. But in this case, the results of Sentinel-2 have no context to the overall results.

Minor corrections:

Line 6: Please, indicate a concrete number for the supraglacial lake area and define the period of pre and post-monsoon.

Line 37: Correct to “and Synthetic Aperture Radar”.

Line 51: Change to “Therefore, we integrate PlanetScope…”.

Line 63: Do you have a quantity specification for the precipitation?

Line 71: What exactly is novel about your approach? Could you define it in 1-2 sentences because applying NDWI is not at all novel? Why are you using the NDWI and which advantages does it has compared to other methods?

Line 73: Change to “observations (3 m spatial resolution)”.

Line 85: Was the exposed glacier ice excluded manually?

Line 100: It is an important point that you have used PlanetScope data harmonized to Sentinel-2. In case of Landsat, there exist the sen2like data. Have you used them as well?

Line 109: Could this limited in time, for example during winter (November to March)?

Line 116 ff.: How were the thresholds (0.10, -0.15, and 0.02) defined? Empirically?

Line 119: Would the red band be better for shadow detection compared to the RGB?

Line 121: What is the advantage of using DN here in comparison with the surface reflectance?

Line 165: How were the thresholds defined?

Line 179: Please delete the sentence “calculated after scaling images to surface reflectance” as optical data are mostly

Line 198: Please write out the abbreviation.

Line 199: “…0.04 m pixel spacing.”

Line 211: What do you mean with physically realistic?

Line 218: Please indicate a concrete number of your results. Have you used the same acquisition time as Watson et al. (2016?).

Line 220: How much was these large differences?

Line 263: How do you define larger and smaller lakes and do you have a concrete number of how many lakes are stable?

Line 264: At which distance from the terminus? How large is the annual maximum extent?

Line 280: Is the high-resolution DEM acquired by UAV or do you refer to the NASA DEM? I am very sorry, but I do not see the link between the derived area-volume relationship and the seasonal evolution of the lakes. It does not fit to the context. What was your intention here?

Line 290: Could you indicate the slope of the positive trend?

Line 297 ff: Could you indicate exact numbers for the trends? Do you have an explanation for the significant increasing trend at Khumbu, Lhotse, and Ambulapcha – it would be really interesting to understand this phenomenon.

Line 318 ff: Are the differences mainly due to a different seasonal evolution? The explanation of the coarser resolution does not agree with my study as I used Planet Scope with a spatial resolution of 3 m as well (the classification result was resampled to 10 m afterwards). I think that the different seasonal evolution is more due to climatic conditions. Have you checked precipitation and temperature of your study site? I am also wondering how lakes in winter could be detected when they are snow covered.

Line 319: Please specify the definition of smaller lakes.

Line 322: Could you please define more precisely which months you are talking about here?

Line 326: Repetition in one sentence (“higher-resolution imagery” / “high-resolution images”).

Line 329: What do you mean with “without controlling for seasonality”?

Line 333: How does the same pattern look like?

Line 334: Could you please explain the discrepancy in detail?

Line 361: How much increased the SGLs?

Line 385: If you are defining the thresholds empirically, I would not use the wording “automated approach” as it is more a semi-automatic approach.

Line 388: Could you specify “poor” in numbers?

Line 392: Please change to “… imagery allows to remove …”

Line 392: How large was the noise? Please be more exactly with your statement.

Line 394: Do you have an accuracy for the better geolocation?

Line 397: Perhaps it is possible to identify ice with Sentinel-2 or Landsat using other bands than visible and NIR?

Line 400 ff: Please describe more the other studies (which data have been used, which methods, acquisition date of the satellite images). For a better comparison, it would be better to indicate number and lake area of the other studies.

Line 420: Please, indicate maximum and minimum area in exact numbers.

Figure 1: Please add the coordinate frame. If the illustrations are aligned to the North, there is no need for a north arrow anymore.

Figure 2: Your workflow is divided into different boxes which are indicated with a), b), etc. Perhaps it would be better to use the chapter number to make it easier to read? Please, include the intermediate steps and the queries of the thresholds as well into the workflow.

Figure 3: Please add the coordinate frame and a white glacier boundary would be better visible.

Figure 4 upper part: Please indicate a legend and a coordinate frame. A north arrow is not needed here. It would be better to enlarge the classification results and to shift them together I assume that the colors indicate the water frequency – do you think that the higher temporal resolution of PlanetScope leads to a higher water frequency?

Figure 4 lower part: Please add more x-tic (every 3 months) and indicate the monsoon period in greyish (c.f. Figure 5). If you shift the legend into the diagram, you could enlarge the diagram to make it easier to read. How do you explain the differences between the sensors?

Figure 5: It would make sense to put all your results (number and area of the lakes) into a table to better compare the lakes on the different glaciers.

Figure 6: Do you mean vertical instead of horizontal lines? Please correct to “with each point representing…”.

Figure 7: Why did you only use PlanetScope when you have all three sensors available? Please indicate the period of the monsoon.

Figure 10: Correct to “…individual Landsat images,…”. Could you specify the slope of the linear regression in the diagram?

Figure 11: Does the frequency depend on the frequency of the acquisitions as well? I am sure, there were not that many acquisitions in the early 1990ies as it is nowadays. Please add the legend.

Table 1: If PlanetScope data are harmonized to Sentinel-2, shouldn't they have the same threshold? I would delete the table as you indicated the threshold in Figure 5 and in the text.

This study brings together a range of satellite image sources to characterise the dynamics of surface lakes (or ponds) on debris-covered glaciers in the Everest-region of Nepal. The analysis is neatly divided between looking at the seasonal dynamics of these ponds using daily Planet imagery, and the long-term (decadal) patterns captured in the Sentinel and Landsat archives. The result is a considerable dataset that builds on previous work on the same glaciers and adds an element of detail with the availability of the finer-resolution imagery. I support its publication, but I do think more could be made of the analysis/data presentation so that future studies, which look to build on this one in coming years, can readily use it as a baseline for comparison.

In particular, the authors should consider:

1. Adding summary statistics to the main text (or, if you prefer, in a table). What are the mean, max and min lake areas for example, per glacier per year, as well as mean, max and min number of lakes? Which glacier hosts the most lakes/greatest area? Is that the same every year? How many lakes are ephemeral vs permanent? How much of the overall lake area do the ephemeral lakes account for (and therefore how important are they, relatively speaking?). These sorts of stats help the reader to interpret the patterns you talk about in the text in general terms, as well as providing concrete values for future studies to use as comparison.
2. Providing more information on the life-cycle of these smaller lakes that appear to be responsible for the seasonal patterns you show. For example, and since you have already gone to the trouble of correcting for surface displacement, can you elaborate on how frequently lakes appear and then drain, how long they last (more or less than a single season?), how often they coalesce, and whether it is the same ones that reappear each time, or new ones that emerge? Kneib et al., 2021 do a nice job of this for ice cliffs as an example. This will tell us more about the processes that are driving the surface changes on these glaciers and add significant value to your manuscript.
3. Whether the inclusion of the UAV data is necessary – I’m not sure at present it adds anything to the key story – if anything it detracts from it. Consider re-packaging it as a ground validation dataset for the Planet imagery (see below)?
4. Reducing the number of figures overall (including in Supplementary) and condensing the text where possible (particularly methods)
5. Being more explicit about the % errors on your lake areas, rather than presenting it as a proportion of the debris-covered area.

More minor comments:

Line 11: I interpret ‘annual’ variation to mean from one year to the next, but I think you’re meaning from one season to the next here?

Line 20: present -> presents

Line 50: can you add a sentence here to underline the importance of understanding these seasonal cycles that your paper characterises?

Line 52: I’m not sure if it is ‘in-situ’ or ‘in situ’ but be consistent

Line 66 (Section 3): this is a very long section! Can you be a bit more concise and focus the text on key points?

Line 68-69: remove sentence starting ‘Each individual source…’ – it’s superfluous

Line 75: Unmanned -> Uncrewed

Line 75-76: if that’s what your UAV data helped with, I’m not sure it is evident in this paper? Consider revising?

Line 114: maybe spell out the NDWI using mathematical notation?

Line 116-117: does this methodological step not fall down for larger lakes given all pixels within that 150 m buffer will have similar NDWI values? Can you clarify?

Line 190 (Section 3.3): I’m not convinced that the inclusion of these field data adds value to the manuscript. How about using them instead to assess the uncertainty in using your PlanetScope imagery as the ‘truth’ for the other coarser resolution datasets? At 3 m spatial resolution there is still some ambiguity as to exactly where the lake margins lie – and with your drone imagery you can put a figure on that, which will be of use to anyone using Planet as a validation dataset going forward?

Line 200: can you be sure that with an error estimate of +/- 1 pixel these manual delineations are more robust than the semi-automatic results? Normally one would use a finer-resolution dataset than that being evaluated to produce these values…? See point above.

Line 213: can you also state the % error on the lake areas? It’s useful to know within the context of the whole debris-covered area, but probably more important is what it means for the data you present in the plots (and ideally, these would have error bars on too)

Line 228-230: same point as immediately above

Line 255-256: this same point has been made three times in quick succession. Maybe remove this sentence?

Line 278-285: this feels a bit odd in the context of the satellite-based observations.

Line 309-317: Not discussing your results in comparison to Watson et al., 2016, who used similarly fine-resolution imagery along with Landsat to look at seasonal patterns, seems like a bit of an obvious omission here?

Figure 11: needs a colour scale to show frequency values

Line 348: can you explain (in the text) why there would be increased solar radiation absorption during winter months?

Line 371: I agree the topographic characteristics of Lhotse are unique within this suite of glaciers, but is the predominant expansion of ponds at Khumbu not also a couple of km from the glacier terminus?

Line 372: Figure S17 takes me to Lhotse Shar, not Lhotse

Line 399 (Section 6.5): I’m not sure this is packaged up in a fair way – the other papers/inventories you refer to here didn’t attempt to delineate small ponds on the glacier surfaces as you have here – they all set a minimum pond area for detection, and were largely focussing on what may otherwise be termed as a lake (i.e. much larger than a pond) because of their much broader spatial coverage. It’s a bit like comparing apples with oranges in my mind.

Figure S12: is it worth pointing our somewhere that the negative trend you identify on Lhotse Shar and Imja is at least partly a consequence of the glacier area shrinking (and Imja Tsho expanding) over the period of observation?

Figures S13-S19: these all need a legend to give meaning to the shades of red.

Figures S13-S19: There are some suspicious areas in the uppermost part of your debris-cover boundary that look to be misclassification, rather than lakes. They are particularly apparent in the Lhotse Nup (2002-2005), Ama Dablam (1998-2001) and Ambulapcha (1998-2001) figures – could they be areas of wet snow? Do they translate through to your data presentation in the main text? Or do you believe them to be genuine…?

References

Kneib, M., Miles, E. S., Buri, P., Molnar, P., McCarthy, M., Fugger, S., & Pellicciotti, F. (2021). Interannual dynamics of ice cliff populations on debris-covered glaciers from remote sensing observations and stochastic modeling. Journal of Geophysical Research: Earth Surface, 126(10), e2021JF006179. https://doi.org/10.1029/2021JF006179

Watson, C. S., Quincey, D. J., Carrivick, J. L., & Smith, M.W. (2016). The dynamics of supraglacial ponds in the Everest region, central Himalaya. Global and Planetary Change, 142, 14–27. https://doi.org/10.1016/j.gloplacha.2016.04.008