the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Dec 2025
A nine-year record of slush on the Greenland Ice Sheet
Abstract. Surface melt on the Greenland Ice Sheet (GrIS) has intensified in recent decades, accelerating GrIS mass loss. Slush (i.e., water-saturated firn or snow) is a key component of this system, yet its areal extent and behaviour remains poorly constrained. Here, we present the first GrIS-wide classification of slush, using over 300,000 Sentinel-2 images and a supervised Random Forest classifier in Google Earth Engine. We generate a nine-year (2016–2024), high spatial resolution (10 m) dataset, which we then use to perform an ice sheet wide systematic assessment of slush distribution across all six major drainage basins. On average, slush covers ~2.9% (~50,400 km²) of the entire GrIS each summer, with around 40% (~29,300 km²) of this area in regions with known low-permeability subsurface structures (e.g., ice slabs and firn aquifers). Slush shows marked interannual variability that mirrors variability in melt intensity, ranging from a maximum slush areal extent of 1.2% (20,500 km²) of the ice sheet in 2018, the lowest melt year of the time series, to a maximum of 5.2% (90,300 km²) in 2019, the highest melt year. When our results are evaluated alongside those from other studies, we find that slush is the dominant meltwater feature in terms of spatial coverage on the GrIS each melt season: the maximum slush area was nine and four times greater than the combined area of supraglacial lakes and streams in 2019 and 2018, respectively. As climate change drives more frequent and prolonged extreme melt seasons on the GrIS, slush is likely to increase in area. Given the influence of slush on lateral meltwater transport, firn-air depletion, and melt–albedo feedback, its incorporation into energy balance and hydrological models will help better constrain ice-sheet mass balance projections.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(6765 KB)
-
Supplement
(7571 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(6765 KB) - Metadata XML
-
Supplement
(7571 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-5159', Pete Tuckett, 17 Feb 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5159/egusphere-2025-5159-RC1-supplement.pdfCitation: https://doi.org/
10.5194/egusphere-2025-5159-RC1 -
AC2: 'Reply on RC1', Emily Glen, 02 Apr 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5159/egusphere-2025-5159-AC2-supplement.pdf
-
AC2: 'Reply on RC1', Emily Glen, 02 Apr 2026
-
RC2: 'Comment on egusphere-2025-5159', Baptiste Vandecrux, 19 Feb 2026
Review of "A nine-year record of slush on the Greenland Ice Sheet" by E. Glen and co-authors
Baptiste Vandecrux, Geological Survey of Denmark and Greenland (GEUS)
The manuscript presents and analyses nine years of slush mapping based on annual and monthly (May to September) Sentinel-2 mosaics. The resulting dataset is very valuable, and the manuscript is clearly written, well illustrated, and suitable for publication in The Cryosphere. My only major concerns are as follows.
1. Although the uncertainty of the classification algorithm is currently provided, it is unclear how this uncertainty scales when the algorithm is applied to the entire ice sheet. Beyond the uncertainty quantification itself, all results are presented without uncertainty ranges, making it difficult to assess whether reported trends and interannual differences exceed the algorithm’s intrinsic uncertainty. I wonder whether the random forest (RF) classifier outputs class probabilities that could be used to derive upper and lower bounds on slush extent, and potentially an uncertainty metric to be distributed with the dataset. I am also not fully convinced by the use of a K-means clustering algorithm to segment slush and non-slush areas during the training and validation process. To my understanding, unsupervised classification groups pixels based on spectral characteristics rather than physical meaning. Depending on the scene, the sharpness of the spectral difference (and transition) between slush and non-slush, and the image-specific number of classes, there is no guarantee that clusters will align with physical boundaries. I assume that the manual labeling step also revealed some ambiguity, with some slush clusters matching the visible slush extent better than others depending on the image. These ambiguities in the labeling and the manual curation process that follows therefore appear insufficiently documented: for example, where is the boundary drawn between slush and channels within slush areas? This combination of unsupervised classification and undocumented manual correction currently limits confidence in the uncertainty assessment. Providing illustrative examples of the workflow, including class maps and labeled images in the supplement, would help build that confidence. Finally, comparing expert delineations of “uncertain slush” areas with the RF-derived uncertainty metric would help demonstrate that the proposed uncertainty quantification is realistic. These points are intended as suggestions, and alternative solutions consistent with the authors’ methodology are of course welcome.
2. The analysis of the slush maps is limited to their spatio-temporal patterns (with some repetition between regional and ice-sheet-wide behavior) and rather superficial comparison with RACMO estimates of melt and air temperature. Deriving additional insights on the formation of slush and/or impact of the slush on the energy balance, would greatly increase the manuscript's relevance. For instance, a potentially valuable analysis would be to study the role of surface topography and annual snowfall in controlling slush formation and its persistence from year to year. It could also support more informed discussion of future evolution: for example, whether slush can form everywhere on the ice sheet, and whether increased melt could shift its occurrence to higher elevations. Additionally, although the impact of slush on surface albedo and energy uptake is not directly quantified, it could be discussed using previously published work scaled to the slush extents derived here.
Addressing these points, together with the minor comments below, would improve the robustness and depth of the study and make it a valuable contribution to The Cryosphere.
---
l.19 "maximum" -> "minimum"?
l.46 missing parenthesis after "Machguth et al., 2022"
l.47 missing period after "laterally"
l.127–130 How does the use of the "high-end" ice slab extent impact the statistics presented later on? Why use this instead of a more reasonable mid-range estimate?
l.253 "trend": do you mean temporal trend? Please check for other instances where "temporal" might be missing before "trend".
l.319–320 This statement seems counterintuitive. One would expect the slush area to be located at lower elevation in the northern regions. Could you elaborate?
l.365–370 SW is also the region where ice flow, melt rates, and surface topography lead to the longest bare ice duration, which favors bare ice darkening (Feng et al., 2024). Do these same settings also shape the slush area above the bare ice zone? Reading this paragraph, I am also wondering about areas that may be mapped as slush early in the melt season and later turn into bare ice after depletion of the snowpack. This could be briefly discussed, as it may introduce noise in the slush extent versus elevation relationship: when slush reaches its maximum elevation, low-elevation slush areas may already have transitioned to bare ice.
l.374 "30% of Greenland's slush area lies above ice slabs" Here, as in many other places, I miss the uncertainty range associated with this value, both from your slush maps and from the ice slab maps.
l.378 "regions lacking underlying low-permeability layers" There are several issues with this framing. (1) There is substantial slush mapped in the bare ice area in Fig. 3. These should be counted separately and definitely have an underlying low-permeability layer: the ice surface. (2) Firn aquifer regions are known for deep meltwater infiltration enabled by temperate firn and the absence of refreezing. Even if these areas do present ice layers (e.g., Miller et al., 2020, Fig. 6), they should not be grouped with ice slabs, which are defined by their low permeability. Finally, there is evidence of physical limits to meltwater infiltration in Greenland firn in the absence of massive ice layers, driven instead by contrasts in grain size and density (Humphrey et al., 2021). By documenting slush outside bare ice and ice slab areas, you may be providing the first observation of meltwater ponding due to minor ice lenses combined with density and grain-size contrasts, as studied experimentally by Humphrey et al. (2021).
l.382–384 Could this paragraph be more quantitative? If there is topographic control on slush formation, slush areas should be largely persistent for years with similar melt. I find non-persistent slush areas in unexpected zones such as the firn aquifers suspicious at first glance. It would be helpful to provide either an RF-derived uncertainty estimate for these cases or at least an expert assessment of the reliability of these non-persistent slush events.
References:Feng, S., Cook, J. M., Naegeli, K., Anesio, A. M., Benning, L. G., & Tranter, M. (2024). The impact of bare ice duration and geo-topographical factors on the darkening of the Greenland Ice Sheet. Geophysical Research Letters, 51, e2023GL104894. https://doi.org/10.1029/2023GL104894
Humphrey, Neil F., Joel T. Harper, and Toby W. Meierbachtol. “Physical Limits to Meltwater Penetration in Firn.” Journal of Glaciology 67, no. 265 (2021): 952–60. https://doi.org/10.1017/jog.2021.44.
Miller, O., Solomon, D. K., Miège, C., Koenig, L., Forster, R., Schmerr, N., et al. (2020). Hydrology of a perennial firn aquifer in southeast Greenland: An overview driven by field data. Water Resources Research, 56, e2019WR026348. https://doi.org/10.1029/2019WR026348Citation: https://doi.org/10.5194/egusphere-2025-5159-RC2 -
AC1: 'Reply on RC2', Emily Glen, 02 Apr 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5159/egusphere-2025-5159-AC1-supplement.pdf
-
AC1: 'Reply on RC2', Emily Glen, 02 Apr 2026
Peer review completion
Post-review adjustments
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-5159', Pete Tuckett, 17 Feb 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5159/egusphere-2025-5159-RC1-supplement.pdf
-
AC2: 'Reply on RC1', Emily Glen, 02 Apr 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5159/egusphere-2025-5159-AC2-supplement.pdf
-
AC2: 'Reply on RC1', Emily Glen, 02 Apr 2026
-
RC2: 'Comment on egusphere-2025-5159', Baptiste Vandecrux, 19 Feb 2026
Review of "A nine-year record of slush on the Greenland Ice Sheet" by E. Glen and co-authors
Baptiste Vandecrux, Geological Survey of Denmark and Greenland (GEUS)
The manuscript presents and analyses nine years of slush mapping based on annual and monthly (May to September) Sentinel-2 mosaics. The resulting dataset is very valuable, and the manuscript is clearly written, well illustrated, and suitable for publication in The Cryosphere. My only major concerns are as follows.
1. Although the uncertainty of the classification algorithm is currently provided, it is unclear how this uncertainty scales when the algorithm is applied to the entire ice sheet. Beyond the uncertainty quantification itself, all results are presented without uncertainty ranges, making it difficult to assess whether reported trends and interannual differences exceed the algorithm’s intrinsic uncertainty. I wonder whether the random forest (RF) classifier outputs class probabilities that could be used to derive upper and lower bounds on slush extent, and potentially an uncertainty metric to be distributed with the dataset. I am also not fully convinced by the use of a K-means clustering algorithm to segment slush and non-slush areas during the training and validation process. To my understanding, unsupervised classification groups pixels based on spectral characteristics rather than physical meaning. Depending on the scene, the sharpness of the spectral difference (and transition) between slush and non-slush, and the image-specific number of classes, there is no guarantee that clusters will align with physical boundaries. I assume that the manual labeling step also revealed some ambiguity, with some slush clusters matching the visible slush extent better than others depending on the image. These ambiguities in the labeling and the manual curation process that follows therefore appear insufficiently documented: for example, where is the boundary drawn between slush and channels within slush areas? This combination of unsupervised classification and undocumented manual correction currently limits confidence in the uncertainty assessment. Providing illustrative examples of the workflow, including class maps and labeled images in the supplement, would help build that confidence. Finally, comparing expert delineations of “uncertain slush” areas with the RF-derived uncertainty metric would help demonstrate that the proposed uncertainty quantification is realistic. These points are intended as suggestions, and alternative solutions consistent with the authors’ methodology are of course welcome.
2. The analysis of the slush maps is limited to their spatio-temporal patterns (with some repetition between regional and ice-sheet-wide behavior) and rather superficial comparison with RACMO estimates of melt and air temperature. Deriving additional insights on the formation of slush and/or impact of the slush on the energy balance, would greatly increase the manuscript's relevance. For instance, a potentially valuable analysis would be to study the role of surface topography and annual snowfall in controlling slush formation and its persistence from year to year. It could also support more informed discussion of future evolution: for example, whether slush can form everywhere on the ice sheet, and whether increased melt could shift its occurrence to higher elevations. Additionally, although the impact of slush on surface albedo and energy uptake is not directly quantified, it could be discussed using previously published work scaled to the slush extents derived here.
Addressing these points, together with the minor comments below, would improve the robustness and depth of the study and make it a valuable contribution to The Cryosphere.
---
l.19 "maximum" -> "minimum"?
l.46 missing parenthesis after "Machguth et al., 2022"
l.47 missing period after "laterally"
l.127–130 How does the use of the "high-end" ice slab extent impact the statistics presented later on? Why use this instead of a more reasonable mid-range estimate?
l.253 "trend": do you mean temporal trend? Please check for other instances where "temporal" might be missing before "trend".
l.319–320 This statement seems counterintuitive. One would expect the slush area to be located at lower elevation in the northern regions. Could you elaborate?
l.365–370 SW is also the region where ice flow, melt rates, and surface topography lead to the longest bare ice duration, which favors bare ice darkening (Feng et al., 2024). Do these same settings also shape the slush area above the bare ice zone? Reading this paragraph, I am also wondering about areas that may be mapped as slush early in the melt season and later turn into bare ice after depletion of the snowpack. This could be briefly discussed, as it may introduce noise in the slush extent versus elevation relationship: when slush reaches its maximum elevation, low-elevation slush areas may already have transitioned to bare ice.
l.374 "30% of Greenland's slush area lies above ice slabs" Here, as in many other places, I miss the uncertainty range associated with this value, both from your slush maps and from the ice slab maps.
l.378 "regions lacking underlying low-permeability layers" There are several issues with this framing. (1) There is substantial slush mapped in the bare ice area in Fig. 3. These should be counted separately and definitely have an underlying low-permeability layer: the ice surface. (2) Firn aquifer regions are known for deep meltwater infiltration enabled by temperate firn and the absence of refreezing. Even if these areas do present ice layers (e.g., Miller et al., 2020, Fig. 6), they should not be grouped with ice slabs, which are defined by their low permeability. Finally, there is evidence of physical limits to meltwater infiltration in Greenland firn in the absence of massive ice layers, driven instead by contrasts in grain size and density (Humphrey et al., 2021). By documenting slush outside bare ice and ice slab areas, you may be providing the first observation of meltwater ponding due to minor ice lenses combined with density and grain-size contrasts, as studied experimentally by Humphrey et al. (2021).
l.382–384 Could this paragraph be more quantitative? If there is topographic control on slush formation, slush areas should be largely persistent for years with similar melt. I find non-persistent slush areas in unexpected zones such as the firn aquifers suspicious at first glance. It would be helpful to provide either an RF-derived uncertainty estimate for these cases or at least an expert assessment of the reliability of these non-persistent slush events.
References:Feng, S., Cook, J. M., Naegeli, K., Anesio, A. M., Benning, L. G., & Tranter, M. (2024). The impact of bare ice duration and geo-topographical factors on the darkening of the Greenland Ice Sheet. Geophysical Research Letters, 51, e2023GL104894. https://doi.org/10.1029/2023GL104894
Humphrey, Neil F., Joel T. Harper, and Toby W. Meierbachtol. “Physical Limits to Meltwater Penetration in Firn.” Journal of Glaciology 67, no. 265 (2021): 952–60. https://doi.org/10.1017/jog.2021.44.
Miller, O., Solomon, D. K., Miège, C., Koenig, L., Forster, R., Schmerr, N., et al. (2020). Hydrology of a perennial firn aquifer in southeast Greenland: An overview driven by field data. Water Resources Research, 56, e2019WR026348. https://doi.org/10.1029/2019WR026348Citation: https://doi.org/10.5194/egusphere-2025-5159-RC2 -
AC1: 'Reply on RC2', Emily Glen, 02 Apr 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5159/egusphere-2025-5159-AC1-supplement.pdf
-
AC1: 'Reply on RC2', Emily Glen, 02 Apr 2026
Peer review completion
Post-review adjustments
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 2,626 | 1,616 | 182 | 4,424 | 443 | 196 | 176 |
- HTML: 2,626
- PDF: 1,616
- XML: 182
- Total: 4,424
- Supplement: 443
- BibTeX: 196
- EndNote: 176
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Emily Glen
Alison F. Banwell
Katie E. Miles
Amber A. Leeson
Rebecca L. Dell
Malcolm McMillan
Jennifer Maddalena
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(6765 KB) - Metadata XML
-
Supplement
(7571 KB) - BibTeX
- EndNote
- Final revised paper