the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Apr 2026
From surface processes to Marine Ice-Sheet Instability: The Collapse of the Barents–Kara Ice Sheet during the last deglaciation
Abstract. During the last deglaciation, the Barents-Kara Ice Sheet (BKIS), a marine-based sector of the Eurasian Ice Sheet, was subject to a drastic retreat over only a few centuries. While the timing of the BKIS deglaciation is well documented, the mechanisms driving the ice-sheet retreat remain debated. Using the GRISLI2.0 ice sheet model, we investigate the behavior of BKIS during this period and identify the marine ice sheet instability (MISI) as the primary driver of the BKIS collapse. Contrary to current interpretations found in the literature, which suggest that a MISI is primarily initiated by ocean-induced basal melting, our results suggest that surface processes, particularly atmospheric warming, can directly trigger such a dynamic instability. Our results highlight the combined roles of atmospheric and oceanic forcings, with atmospheric warming triggering the initial retreat at the onset of the deglaciation and oceanic processes subsequently controlling its dynamics. We therefore encourage future studies on marine ice sheets instability, to give a better consideration to variations in atmospheric conditions on their impact on ice sheet destabilization.
- Preprint
(1685 KB) - Metadata XML
-
Supplement
(2581 KB) - BibTeX
- EndNote
Status: open (extended)
- RC1: 'Comment on egusphere-2026-1137', Tijn Berends, 06 Jun 2026 reply
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 234 | 80 | 22 | 336 | 45 | 21 | 23 |
- HTML: 234
- PDF: 80
- XML: 22
- Total: 336
- Supplement: 45
- BibTeX: 21
- EndNote: 23
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Summary
The Barents-Kara Ice Sheet (BKIS) was a mostly marine-based remnant of the larger Eurasian ice sheet, that existed during the last deglaciation. Geomorphological evidence shows that this ice sheet disintegrated rapidly around 14,000 years ago, with worrisome implications for the future of the West Antarctic ice sheet.
The authors present results of a number of simulations of the deglacial evolution of the BKIS, performed with the GRISLI2.0 numerical ice-sheet model. By including/excluding atmospheric forcing, oceanic forcing, and ice shelves, they investigate the relative importance of these different processes in driving the retreat of the BKIS. Based on their findings, they argue that sudden atmospheric warming provided the initial trigger for the retreat, which was then amplified by a combination of ocean-induced basal melt, and self-reinforcing retreat of the grounding line along the retrograde marine bed of the (isostatically depressed) Barents-Kara Sea, which the authors attribute to the Marine Ice-Sheet Instability (MISI).
General comments
I find the topic of this study interesting. Paleoglaciological evidence for rapid ice-sheet retreat can be very valuable for studying the future evolution of the Greenland and Antarctic ice sheet. Process-based studies like this one can be particularly useful.
Generally, the paper is well-written and easy enough to follow. There are some instances where the technical details of the methodology are not clear, but nothing that I think cannot be easily remedied.
However, I do have some concerns about the numerical model that the authors used, and its applicability to this particular study.
GRISLI2.0 was presented in 2019 by Quiquet et al. That study does not include any of the idealized-geometry experiments that have, in the last decade or so, become de facto benchmarks for numerical ice-sheet models (e.g. EISMINT, ISMIP-HOM, MISMIP, MISMIP+, MISOMIP, ABUMIP). The current study revolves around the concept of MISI, which is now known to be particularly difficult to represent in numerical models. While the model described by Quiquet et al., having a 20-km square grid and a grounding-line flux condition to circumvent the inaccurate velocity solution, should in principle be able to produce reasonably accurate grounding-line dynamics, they do not provide any evidence of this. Having spent the last few years working on ice-sheet model development, I am only too painfully aware of how easy it is to make mistakes when building something so complicated, and how important it therefore is to perform these benchmarks regularly. Without evidence of how GRISLI performs relative to other ice-sheet models in these standardized experiments, I find it difficult to assess the results presented in this manuscript.
The same holds for the basal melt parameterization. In lines 677-680, the authors state that “ocean temperatures were artificially increased by 10 °C, […] producing basal melt rates at the grounding line […] on the order of 30 m yr⁻¹.” In Antarctica, basal melt rates of over 50 m/yr are observed in the Amundsen Sea region, with ocean temperature that are nowhere near as warm as what you’re describing. This suggests that the basal melt parameterisation in GRISLI is not tuned correctly, and produces values that are too low. One of the main conclusions of this manuscript is that ocean-induced basal melt plays only a secondary role in the collapse of the BKIS, which would be undermined if there is indeed an error in this parameterization.
Specific comments
Line 17: “MISI is primarily initiated by ocean-induced basal melting” The way I’ve always understood it is that the instability is an inherent property of an unbuttressed ice sheet grounded below sea level on a retrograde slope. This produces an unstable equilibrium, where any perturbation, be it from changes in the surface mass balance, basal melt, or the thermal regime of the ice (as in the MISMIP experiments that studied MISI), can result in self-sustaining retreat. A lot of recent literature about MISI focuses on West Antarctica, where the causal chain of oceanic warming, increased sub-shelf melt, thinning shelves, and loss of buttressing is the prime suspect for causing grounding-line retreat in the future – but that does not mean that MISI is, by itself, an ocean-induced process.
Line 47: Please include here a reference to Seroussi et al. (2024), the most recent and comprehensive model ensemble for the future evolution of the Antarctic ice sheet.
Line 48: The article by Mercer (1970) doesn’t contain anything related to MISI. The first publication to delve into this is Weertman (1974).
Line 50: “positively correlated” suggests a relation deduced from statistics, please rephrase.
Line 53: “can only be tempered by ice-shelf buttressing” Well, or until the grounding line reaches a prograde slope, or the friction underneath the grounded ice can change.
Line 66: There have been other studies that simulated the evolution of the entire Eurasian ice sheet complex during the last glacial cycle, using models that did include proper grounding-line dynamics (e.g. Scherrenberg et al., 2024), although they did not focus specifically on the Barents-Kara sector.
Line 69: Please also cite Schoof (2007), who derived the first semi-analytical solution to the grounding line flux, and Pollard and DeConto (2009), who were the first to implement it into a numerical model.
Line 121: “we impose a linear basal friction law” Please discuss why you chose this rather unusual sliding law; most ISMIP models currently use either an n=3 (or equal to whatever exponent they use for their constitutive equation) power-law, a constant-friction Coulomb law, or a ‘pseudo-plastic’ hybrid between the two.
Line 132: “to prevent excessively fast ice flow” Since this parameter is crucial to some of your experiments, I’d like to see some more discussion here. Please define what constitutes “excessively fast” ice flow, also considering observed velocities of extant ice streams in Greenland and Antarctica. The current phrasing suggests that this is purely a ‘numerical’ parameter, intended to reduce artefacts in the velocity solution. However, later on, you interpret it as a ‘physical’ parameter, which you use to study the sensitivity of your modelled ice sheet to basal sliding.
Line 138: I am confused. Tsai (2015) derived their grounding-line flux solution using a Coulomb sliding law; Quiquet et al. (2018) state that, when GRISLI uses a power-law sliding law, it uses Schoof’s (2007) grounding-line flux solution. The combination you describe in the text here (power law + Tsai solution) is not consistent; is this what you did, or is your text not correct?
Line 151-156: Does this mean that your model set-up does not include any large-scale feedback of ice geometry changes onto the climate, but only a local lapse rate-based adjustment? And no adjustment of the precipitation?
Line 174-177: “small-scale processes not explicitly resolved by the ice-sheet model” What processes are these? Other models that use this approach (usually called a ‘partial melt parameterisation’ after Leguy et al., 2021) usually justify it from numerical considerations.
Line 233-234: “the ice-sheet extent remains fixed at LGM conditions throughout the simulation” This seems quite unrealistic, producing a strong overestimation of the albedo in the region and consequently biasing the GCM to colder temperatures. I wonder if this can (partially) explain the large difference in temperature between the two GCMS, which you discuss in lines 290-305.
Lines 264-265: “the index was constrained between −2 and 2” This suggests that the two GCMs predict temperatures that are, at some point during the deglaciation, either colder than T_LGM - deltaT_LGM, or warmer than T_PI + deltaT_LGM. Is this so? Do you have an explanation for this?
Lines 330-331: “consistently with the subsurface temperatures calculated using the TRACE21k index” I do not see any similarity between the sediment core data, and the two indices. If you want to argue there is one, please substantiate this with some sort of statistics.
Figure 2: What’s the difference between panels a and c? The only difference I can find in the caption is that c shows ‘anomalies’, but with respect to what? And shouldn’t that just result in the same curve, but with an offset? Also, what’s going on with the shallow ocean temperature from TRACE21k between 17ky and 14.5ky in panel d? Did their model freeze?
Lines 352-355: You describe a steady-state initialization, which you achieve by prescribing a fixed LGM climate for 100 kyr. However, the LGM climate only existed for a few thousand years at the most, shorter than the response time of the ice sheet, which implies that the ice sheet would not yet have been in equilibrium with this cold climate before it started warming again. I know there have been several studies that demonstrated this (only the ones by Stap et al., 2022/2024 come to mind right now). I’m not sure what (if anything) you did to tune your model to get an ice sheet at the end of these spin-ups that looks reasonably LGM-like (you do not mention any tuning in the manuscript), but assuming equilibrium between the ice sheet and the climate at LGM bias this tuning towards an ice sheets that is too small.
Lines 358-359: “sea level according to the reconstructions of Waelbroeck et al.” I assume this means your ELRA model does not include the gravitational effects on sea level? There’s a few nice papers from Natalya Gomez that show how this stabilizes the ice sheet, preventing MISI’s self-sustaining retreat on mild retrograde slopes. Please discuss the relevance of these findings for your work.
Lines 372-373: “we set the basal melting to zero and removed the analytical parameterization of grounding line flux”. The flux condition is, essentially, a trick to reduce the huge numerical errors that arise from the discontinuity in basal friction at the grounding line. Without it, your spatial resolution of 20km would produce numerical errors in grounding-line position and migration rate that far exceed the signal you would get from oceanic or atmospheric forcing. As such, I don’t think this experiment can be meaningfully interpreted.
Lines 383-385: “in which grounding line migration was entirely cancelled by imposing an artificially low sea level of -1000 m relative to present-day” Well, I mean, it’s true that you cannot have grounding-line migration when there is no grounding line… But I really struggle to see the point of this experiment. I am also surprised that the ice extent in this experiment (Fig. 5) is not larger than in the other ones. In the absence of basal melt and calving, the ice extent in this experiment should be entirely dictated by the surface mass balance, implying that you already have enough surface melt to prevent any further advance south into mainland Siberia; is this indeed the case?
Lines 394-396: “following the DEGLA-ATM-LGM design, but with amplified basal melting at the grounding line” What do you mean by this? Earlier on, you mention that you use the PMP scheme for sub-grid basal melt at the grounding line. Do you still use this, but with a higher melt rate before scaling? Also, this again seems more like a study into numerical errors to me, rather than looking at physical processes.
Line 405: “3.1 Ice sheet geometry at the LGM” Please make it clear from the start that you describe here some reconstructions that you use to compare your model results against.
Figure 3: The vertical axes of the panels all show integrated ice volume (erroneously called ‘ice thickness’ in panels a,b) and ice area as a percentage of a value at 21 ky. The fact that they all start at exactly 100% suggests to me that you scaled all the curves by their own respective value at 21 ky, is that correct? If so, I find this misleading, as this approach obscures any differences in absolute ice-sheet volume or area between different model simulations and data-based reconstructions. Please show the absolute values as well. Also, in panels c-d, please improve the legend – right now, the black outlines around the boxes make it very difficult for me to distinguish the grey and black markers.
Line 449: “the initial states we built (see Fig. S3) overestimate the ice extent in this region” Which is why you need to show the absolute ice volume/area in Fig. 3.
Line 505: “when the ice flux at the grounding line is ignored” it’s not ignored, it’s just modelled very inaccurately.
Figure 6: is the “atmospheric index” in panel A the spatial average of I_T_iL? Averaged over the entire region, or only over the ice area? And in panel B, is the dashed line the “zero lapse rate” experiment? If so, please correct the legend.
Figure 7: This is an interesting way to analyse the relative importance of mass balance changes vs. ice-dynamical processes. The ‘SMBB’ and ice flux divergence panels are integrated over the 500-yr period, right? In that case, use an integral sign in the title, not a summation sign. Also, do the panels show the ice mask at 14 ky or 14.5 ky? I think the latter option would be the most informative. Lastly, how can it be that the integrated surface melt in the NOSL experiment is so much larger? You mention that the ice sheet in this experiment is larger and thicker than in the baseline experiment, which implies that the surface slope at the margins would also be steeper, right? That would imply that an increase in surface temperature would result in a smaller increase in ablation zone area, than in the baseline experiment (due to the steeper slope), so I would expect this ice sheet to be more ‘resilient’ to atmospheric warming. Or am I missing something here?
Lines 615-616: “he stabilized experiments… …demonstrated that the BKIS retreat is both abrupt and irreversible” They do not; you didn’t do any ‘reverse’ experiment. ‘Irreversible’ means the retreat will go on even when the forcing is reversed, i.e. the climate is set back to LGM conditions. The original MISMIP experiments (Pattyn et al., 2012) demonstrate irreversibility; even when the ice viscosity is reversed to its original value, the grounding line does not advance back to its original position (i.e. hysteresis). The fact that your ice sheet continues to shrink when you ‘freeze’ the forcing only demonstrates that your system has a long response time.
Lines 631-639: For readers who, like me, are not familiar with your earlier paper, please provide a very brief (2-3 lines) summary of what you did there.
Lines 647-648: “ice sheet reconstructions (DATED-1, ICE-6G_C, and GLAC-1D) indicate that confined ice shelves were indeed very limited at the LGM” DATED-1 and ICE-6G cannot represent shelves; they were not created with ice-dynamical models, but rather derived by inversion from observed uplift rates, combined with geomorphological evidence. Only GLAC-1D can be used for a comparison like this.
Lines 684-692: your “minimum friction coefficient” looks to me like a numerical parameter, not something physical. Changing it does not tell you anything about the uncertainty in the bed roughness beneath the BKIS; it only tells you something about the importance of numerical artefacts in your model.
Lines 696-697: “an ice sheet equilibrated under an LGM climate tends to develop a geometry fully adjusted to cold climatic conditions” See my earlier comment. You do not show how your modelled LGM ice volume compares to reconstructions (see my earlier comments on Fig. 3); this potential bias is another reason why this comparison needs to be made.
Lines 697-698: “a more pronounced glacio-isostatic depression of the bedrock” Good point, I hadn’t even thought of the GIA bias yet. Indeed, this would also affect the grounding-line stability during the deglaciation.
Lines 711-720: True, a more viscous mantle would delay the uplift during the deglaciation. On the other hand, since the really cold part of the glacial cycle didn’t last very long, the Eurasian ice sheet also didn’t keep its maximum extent for a very long time, so that the depression likely hadn’t reached its equilibrium depth yet when the deglaciation started. A more viscous mantle would then also lead to a less deep depression, and consequently a more stable grounding line. I’m not sure which effect would be stronger.
Lines 722-736: “Brendryen et al. (2020) conclude…” Brendryen et al. looked at (newly-calibrated) geomorphological evidence of ice margin/grounding-line evolution. The statements they make about the causes of the ice sheet’s retreat are not based on their own work.
References
Pattyn et al., 2012.: Results of the Marine Ice Sheet Model Intercomparison Project, MISMIP, The Cryosphere 6, 573-588.
Pollard and DeConto, 2009: Modelling West Antarctic ice sheet growth and collapse through the past five million years, Nature, 458, 329–332 doi:10.1038/nature07809,
Scherrenberg et al., 2024: Late Pleistocene glacial terminations accelerated by proglacial lakes, Climate of the Past 20, 1761--1784.
Schoof, 2007: Ice sheet grounding line dynamics: steady states, stability and hysteresis, J. Geophys. Res., 112, F03S28, doi:10.1029/2006JF000664
Seroussi et al., 2024: Evolution of the Antarctic Ice Sheet Over the Next Three Centuries From an ISMIP6 Model Ensemble, Earth's Future 12, e2024EF004561.
Stap et al., 2022: Net effect of ice-sheet--atmosphere interactions reduces simulated transient Miocene Antarctic ice-sheet variability, The Cryosphere 16, 1315—1332.
Stap et al., 2024: Miocene Antarctic Ice Sheet area adapts significantly faster than volume to CO2-induced climate change, Climate of the Past 20, 257--266.
Weertman, 1974: Stability of the junction of an ice sheet and an ice shelf, J. Glaciol., 13, 3–11.