the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 24 Feb 2025
The Greenland Ice-Marginal Lake Inventory Series from 2016 to 2023
Abstract. The Greenland Ice Sheet and its surrounding peripheral glaciers and ice caps are projected to be the largest cryospheric contributor to sea level rise in the next century. While glacial meltwater is typically assumed to flow directly into the ocean, ice-marginal lakes temporarily store a portion of this runoff, influencing glacier dynamics, lacustrine-driven ablation, ecosystems, and downstream hydrology. The size, abundance and dynamics of ice-marginal lakes are expected to change in the future. However, they remain under-represented in projections of sea level change and glacier mass loss. Here, we provide eight annual records across Greenland of lake abundance, lake surface extents, and surface water temperature estimates from 2016 to 2023. The dataset catalogs 2918 automatically classified ice-marginal lakes and reveals their evolving conditions over time. Our dataset fills critical gaps in understanding Greenland’s terrestrial water storage and its implications for sea level change projections, providing a first step toward quantifying meltwater storage at ice margins. Equally important, it supports assessments of ice sheet and glacier dynamics, such as lacustrine-driven ablation, and Arctic ecological studies of lake changes impacting ecosystems. The inventory series will also aid environmental management and hydropower planning aligned with Greenland’s proposed commitments under the Paris Agreement. The inventory series is openly accessible on the GEUS Dataverse (https://doi.org/10.22008/FK2/MBKW9N) with full metadata, documentation, and a reproducible processing workflow (How et al., 2025).
- Preprint
(12327 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on essd-2025-18', Anonymous Referee #1, 28 Mar 2025
This paper presents an annual ice-marginal lake inventory from 2016 to 2023, classified using an established remote sensing approach. The paper is short. Figures and Tables are nicely drafted. The dataset is certainly valuable, but the authors also report that their automated processing are underdetecting the lakes but apparently do not take any means to improve it. I miss more information on validation and more discussion on how improvements could be made. I also question why there are so many authors (I counted 12 incl 3 'managers') on this rather short paper not involving any field investigations.
Minor comments
I have checked the readme.txt, and downloaded one of the lake files. Seems fine
7 ‘The dataset catalogs 2918 automatically classified ice-marginal lakes and reveals their evolving conditions over time.’ Does not the number of lakes vary through time?
26 do you have a reference for this sentence?
30-41 I found this a bit detailed, is all needed? I suggest shortening this part, do it a bit wider in terms of authors cited and more general before zooming in on Greenland.
55 I am not sure that you in your paper defend the statement ‘ and assess the impact of these changes on future sea level projections.’ There is no reference for the statement and I would be careful with it.
63-65 here you use both ice-marginal and ice-contact lakes- you define ice-contact lakes, but not the other. I suggest you define both and check the use throughout.
68./103/ can you here or elsewhere mention if there are any previous lake inventories based on Landsat that you use for reference or if this is the first?
Can you specify that 2016 is the first year due to launch of Sentinel. It is implicit but not explicitly stated.
105 remove –?
3.1.1./3.1.2./3.1.3/3.2/5.1/throughout where your work are described
Usually work done in methods are written using past tense. If you did the work, use past tense. This makes it easier differencing published work (present tense) from what has been done for this paper/dataset.’
163+ will this number of lakes sharing margin not differ with time, this can change for year to year, which year of the dataset are you referring to? Suggest to rewrite this. You mention below that lake number vary from year to year, so how come one number is static while other differs.
209 past tense, was?
243 can you elaborate a bit more on how this was done and explain this better, it is a huge difference. Were lakes then included annually based on this effort? What was the follow up from the results you found. Did you try to improve the mapping method or include manual digitisation? Not clear to me. The last sentence ‘ ..no manual lake delineations are included’ seems to be a strange follow up if the underdetection is so substantial. I am not sure if any of the statements in 6.1 is valid if the dataset is missing so many lakes. Could the methods have been improved? Or could manual updated help?
The result of the validation is completely missing from conclusion and abstract, data uncertainty and accuracy is important part of an ESSD paper. I would rewrite the abstract to be less general and more direct on results and uncertainties. same with conclusion.
I miss a figure showing validation of the method from one of the newer years that is previously not published.
Citation: https://doi.org/10.5194/essd-2025-18-RC1 -
RC2: 'Comment on essd-2025-18', Adrian Dye, 04 Apr 2025
The manuscript provides an overview of the multi-temporal high resolution (spatial and thermal) development of ice marginal lake inventories of Greenland from 2016 to 2023. This represents a substantial advance in the science of ice marginal lake evolution at the regional scale across very remote. However, I do have some concerns over some of the parameters and lack of boundary error estimation within the manuscript. These could all be relatively easily address to provide a high quality inventory that will provide a substantial advancement in the study of ice-marginal lakes relationship with climate change in order to understand their role as a resource but also a hazard in some areas.
The combination of the multi-sensor remote sensing approach for lake detection following How et al., (2021) provides a relatively robust approach for a complex dynamic problem over a very large and mostly very remote area. The accurate detection of water bodies at these locations are not only important glaciologically, but also for hydrology and ecology downstream. With this in mind I would highly recommend that lakes are not eliminated from the inventory once they lose contact with the ice front, but instead are classed as ‘non-contact’ or perhaps archived in a sub-inventory. The presence of these lakes will further modify any glacial meltwater, as they can be a substantial sediment sink (Vowels et al., 2025) as well as thermal modification and also implications for passage of GLOFs. All of which have substantial importance downstream.
The detection of water bodies at these ice marginal lake sites across Greenland is an unenviable task – given the presence of snow and ice as well as frequent cloud cover. The detection of ‘lake margins’ becomes a critical problem in analysing the spatial evolution. Some of these lake margins will have substantial snow and ice banks throughout most years. Consequently uncertainty assessment of lake margin boundaries becomes complicated, but still important – yet this does not appear in the manuscript? The combined areal uncertainty from How et al., (2021) could be referred to. This could be dealt with through standard remote sensing approaches to lake boundary errors or there is potential for classifying or flagging lake margins that have substantial snow/ice cover. This will effect the thermal conditions in the lake. It is also important for understanding some of the large lake decreases that your report – are these from increased snow/ice cover? Drainage/lowering? Or glacier terminus ‘advance’ from increased glacier velocities? If the latter then they are very important to examine further (King et al., 2018).
Unfortunately the Landsat Level 2 thermal product is resampled to 30m, which creates problems for eliminating the original 100m pixels that would be a combination of water and non-water. As well as the added uncertainty with defining lake boundaries and also any uncertainty with the alignment/correlation between the Landsat thermal sensor and infrared/visible sensors. In order to reduce pixels with thermal contamination in the dataset from boundary issues I would strongly recommend setting a buffer of 100m minimum around the lake margins. I think this will reduce some of the noise in the dataset, particularly for smaller lakes and those with peninsulas/rock islands etc.
Detailed comments
Abstract – it may be beyond the scope of this style of paper but it could have more results in it? Some of the big lake decreases? Thermal results? Currently reads more like a short research proposal/rationale. Also would like to see ‘1km buffer’ and 0.05 km2 in the abstract.
Lines
20 – Greenland ‘glacial’ lakes were 21% of Zhang et al., (2024) global inventory…
20- St Pierre et al. (2019) argued proglacial lakes could be substantial CO2 sink – this inventory provides a big step to monitoring suspended sediment concentrations of them (contact and non-contact)
21 – as well as hydrological modification (higher temperature and lower SSC).
31 – Warren and Kirkbride (2003) needs to be cited here. Haresign and Warren (2005) and Roehl (2006) should be really as well.
40 – and hydrological modification – key to know characteristics of water feeding into one of the most delicate parts of the thermohaline circulation…
53 – Can these be reclassified rather than retired? They’re still important (see above).
60 – And glacial lake response to climate?
Table 1 - Landsat 8/9 – filter – It should be 20% rather than 30%?
Table 2 – Could you also add the following to improve useability further; i. centroid (could be from 2017) ii. Contact or non-contact iii. Temp_time – time of Landsat image -> images in PM will likely capture daytime surface warming of very near surface layers (especially with high SSC)
Adding dam type into the inventory would be desirable but not essential and clearly a huge amount of work that would be a whole different project in itself – probably requiring citizen science ground validation?
103 – I would like to see ‘1km buffer’ and ‘0.05km2’ in the text here and also in the Abstract as they are key defining parameters of the inventory.
128 – Nice strategy – this has been a key problem for a while especially with differing geology and consequently reflectance spectral signature from the lake SSC
130 – ‘best quality’ – is this user defined? Or a class of product?
140 – 30 metre spatial resolution is incorrect – the original thermal pixels are 100m – so some pixels could easily be a combination of land (10 to 20 C) and lake water (4 C) with substantial thermal contamination. Either more needs explaining regarding the resampling method or stick with 100m to be on the safe side – there is plenty of data so can afford to lose some pixels.
142 – There needs to be more detail (Is it the NCEP reanalysis? 6 hour? 1 degree? ) on the atmospheric data and correction used in this product given the nature of the study area and the Landsat L2 thermal product struggles in coastal zones (Dyba et al., 2022). I would imagine it handles the lower topography/stabler climate of SW Greenland better than the higher topography and dynamic climate and microclimates of E Greenland… (although I would still like to see how the regional average LSWT compared – see below)
Did you eliminate temperatures below 0oC ? Needs adding in Methods.
145 – Lakes in Poland are very different to glacial lakes in Greenland… (More stable atmosphere and less water input as well as lower SSC) So the comments on developing a calibration factor for Greenland in the Discussion are well founded. Using the GEE script from Ermida et al. 2020 would be more robust though? The validation data for SW Greenland does look very good though – which I think proves sufficient robustness.
159 – I think the simplest way to deal with the thermal contamination at lake boundaries is to increase the buffer to 100m
163 – Add ‘… that have existed between 2016 and 2023’ ?
173 – Yes I think the variability in abundance is impossible to study further at this scale – could be variations in meltwater flux, permeability of substrate (a large number may be permafrost underlain/ground ice – could potentially develop taliks…) dam porosity etc.
192 – CE looks to be low variability in area in Fig. 3a? NO looks to be second largest variability in area on those results.
195 – I think these declines in area are an important result. This should either be explored a bit further (maybe not appropriate in this paper) or flagged for future research etc (see comments above)
201 – Yes I can see the glacier remnant on the NW side of the terminus. Some of the ‘margin lines’ on the South side look to be possibly from lake ice?
203 – Important observations for 5c and 5d – worth flagging I think – are these a topic of ongoing research?
Section 4.3
This section currently is under explored/reported.
209 – Is this the average for all pixels? Or Sum of all lake averages divided by number of lakes? (some of the large lakes could skew this)
211 – of all pixels?
212 – Replace ‘falling’ with ‘being lower’ as these are snapshots
214 – Replace ‘rising’ with ‘being higher’
There is more room for exploration of the thermal results – either here or signposted to a future publication. If the thermal contamination issues are resolved, the average lake temperature by region should be shown. Also potential for average temperature by different lake size classes etc. With higher confidence in the data the lower temperatures in 2018 could be explored further too. At the moment the thermal data feels a bit like a ‘bolt on’ component.
224 – Again can these detach lakes be reclassed as non-contact or put in a separate archive?
247 – Boundary error/uncertainty estimations of some kind need including here.
258 – Agreed estimates are reliable. The validation data has an interesting cluster of points around the lake sensor temperature of 4 oC – could be afternoon warming of surface water? Time of day of image capture is important.
265 – Yes this is a really important inventory for assessing how glacial lakes form part of a deglaciating land system and how their response to climate affects hydrology and ecology downstream. At the moment this section reads too glaciology focused – there is much wider scope for assessing the glacial lake evolution. How will lake development affect downstream sediment budgets? How will temperature changes affect ecology? (Fellman et al., 2014)
292 – Are there any TanDEMX products that would help? (Is access through ESA possible?)
Figures
Figure 1 – More thermal results need to be added
Figure 3 – NE lake area was high for 2018
Figure 4 – Currently difficult to distinguish between the ocean and ice sheet. I suggest having ocean in a shade of blue.
Figure 5a – See above. I would like to query the lake margin for 2017 in Figure 5a – which looks to have a pattern suggesting lake ice at the margins?
Figure 6 – data filtered below 0C? (If so please add this in the caption) Looks like lots of noise > 8 C
This is a huge body of work that will have a transformative impact on glacial lake science (after a few modifications)!
Citation: https://doi.org/10.5194/essd-2025-18-RC2
Data sets
Greenland Ice-Marginal Lake Inventory annual time-series Edition 1 Penelope How, Dorthe Petersen, Nanna B. Karlsson, Kristian K. Kjeldsen, Katrine Raundrup, Alexandra Messerli, Anja Rutishauser, Jonathan L. Carrivick, James M. Lea, Robert S. Fausto, Andreas P. Ahlstrøm, and Signe B. Andersen https://doi.org/10.22008/FK2/MBKW9N
Model code and software
GrIML v0.1.0 Penelope How https://doi.org/10.5281/zenodo.6498006
GrIML code repository Penelope How https://github.com/GEUS-Glaciology-and-Climate/GrIML
GrIML code documentation Penelope How https://griml.readthedocs.io
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 237 | 54 | 6 | 297 | 8 | 10 |
- HTML: 237
- PDF: 54
- XML: 6
- Total: 297
- BibTeX: 8
- EndNote: 10
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1