the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Aug 2025
Evolution of the Antarctic Ice Sheet from 2000–2300 and beyond: model sensitivity and uncertainty analysis using MPAS-Albany Land Ice
Abstract. We present a description of the Antarctic Ice Sheet model configuration submitted to the ISMIP6-Antarctica-2300 experiment using the MPAS-Albany Land Ice model, along with three new sets of simulations: (1) a set of extended simulations to 2500 for three forced experiments and to 2775 for the control experiment; (2) a sensitivity analysis of our model configuration to parameters controlling basal sliding and sub-shelf melt, and to model structural choices including the choice of the energy and stress balances; and (3) a 72-member ensemble run on graphics processing units (GPUs) and analysis of variance to determine the primary sources of uncertainty in our ice-sheet model projections. Our extended simulations predict rapid retreat beginning after 2300 for SSP1-2.6 forcing and after 2500 for present-day (control) forcing, primarily in the Amundsen Sea Embayment. Our parameter sensitivity experiments reveal only moderate sensitivity to the value of the sub-shelf melt parameter, ranging from the 5th to 95th percentile values. The Amundsen Sea Embayment region displays a strongly non-linear dependence of mass loss on the sliding law, with no discernible relationship between the sliding law exponent and the mass loss by 2300, while the sectors feeding the Ross and Filchner-Ronne ice shelves exhibit more mass loss with a more-plastic sliding law and vice versa. Our model fidelity sensitivity experiments indicate a modest sensitivity to the choice of stress balance approximation and a very strong sensitivity to thermomechanical coupling versus an uncoupled configuration. Our 72-member ensemble and analysis of variance show that the uncertainty in long-term projections is dominated by the choice of Earth system model forcing and the presence or absence of hydrofracture forcing, rather than uncertainty in sliding and sub-shelf melt parameters. We hypothesize that initial condition uncertainty could account for much of the inter-model spread in the ISMIP6 ensembles.
-
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
(73247 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(73247 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-3942', Anonymous Referee #1, 02 Oct 2025
- AC2: 'Reply on RC1', Trevor Hillebrand, 04 Dec 2025
-
RC2: 'Comment on egusphere-2025-3942', Anonymous Referee #2, 03 Oct 2025
Hillebrand et al. present an analysis of the ISMIP6 AIS 2300 simulations (Seroussi et al., 2024) they performed with MALI as well as some additional sensitivity experiments. This is a useful study to help understanding the large spread of sea level projections for Antarctica.
The manuscript is well written and clear, however lacks some explanations and motivations. I currently do not understand why exactly those parameters, values, and modelling choices were tested, and not others. Furthermore, more explanation on the “non-linearity” of the sliding parameter sensitivity is required, and I cannot follow the argument that the melt parameter choice is less relevant. The argument about initialisation uncertainty in ISMIP6 AIS 2300 being the dominant driver for uncertainty is not clear to me.
I suggest minor comments concerning the framing of the results which should be addressed.
Comments:
Abstract
Line 8: “only moderate sensitivity”. I disagree with this point in contrast to claiming higher sensitivity for other parameters later on. Looking at Figures 7a (melt parameterisation parameters) and Figure 9a and 11a, the melt parameter induces differences >50cm SLE in 2300 for each experiment, which is similar to changes induces by the energy balance solver (which is claimed to induce “very strong sensitivity” in the abstract) and the sliding law exponent (except for the q=1 case for the HadGEM forcing; claimed to induce strongly non-linear dependence). I do hence not see how the authors say that melting has “only moderate sensitivity” in contrast to the other parameters, as the overall influence on the sea level results is at the same order?
Line 9: “ranging from the 5th to the 95th percentile values” – unclear where these come from, think about adding “of the AntMean ISMIP6 calibration” or something similar. I suspect that the choice of your range of parameters has a strong influence on your sensitivity results, and if you had included other values, you would get a different answer (as you mention yourself in the discussion). The statement you can make is that using this parameter tuning, you find a less important contribution in the ANOVA results.
Line 12: not clear what you mean with “vice versa”. Do you mean that FRIS and Ross have less mass loss with a less plastic sliding behaviour? In that case, maybe simply remove it.
Introduction
General: The introduction is missing a discussion of previous results on sensitivity analysis of ISMIP6 simulations. It might be worth checking all ISMIP publications. For example, I could find this study TC - ISMIP6-based Antarctic projections to 2100: simulations with the BISICLES ice sheet model which seems to do something very similar.
Moreover, I cannot find a motivation as to why exactly the parameters/schemes tested were tested, at the moment their choice appears arbitrary. Make sure to motivate your choices!
Adding these two points might make the introduction longer, you could consider removing information that is not directly relevant for your study (e.g., on MICI, calving, flow law exponents, etc).
Line 26: “Onset of such runaway retreat leads to……” – there is a subtly here. To my knowledge DeConto & Pollard, De Conto et al. do not check for the retreat being MISI driven in their results, hence it is not clear if the MISI-driven retreat causes the multiple metres of sea level rise in their simulations. MICI is then added as an afterthought, as if it was to enhance the already strong results. I suggest you reformulate.
Line 58 “regardless of their skill in reproducing historical ice-sheet behaviour”. I suggest reformulating, as it sounds like this was tested/required by ISMIP6, however, reproducing historical behaviour was not an aim of ISMIP6, and the trend analysed by Aschwanden et al. was never meant to be a representation of historical behaviour.
Line 152 “As MISI is likely already under way” the reference is quite old, and more recently, it has been more specified that “MISI being under way” needs to be tested by more than just finding continued retreat in the Amundsen Sea. You do not need MISI here, simply that the ASE is changing is sufficient.
Line 169: How much do your deltaT values differ from the original ones? Note that this might impact how sensitive your melt reacts to changes in the ocean forcing, see for example Lambert & Burgard 2025.
Methods
Section 2.3.3 & 2.3.4. The choice of the parameter ranges for the sub-shelf melt and the sliding parameters appears a bit arbitrary. If they are not sampled equally broadly, how can you claim that one factor has a higher influence than the other?
Results
Section 3.3 & 3.4 How do the control runs look with changed melt or sliding parameters? Is the drift again small enough to be ignored in your plots, or how much influence does this have on your results (especially with the “non-linearities” in the sliding exponent experiments)?
Line 289 “indicate only moderate sensitivity” as stated above, I disagree with this as your difference is still 50-60cm between the low and high end, which is more than a doubling of the contribution for the CCSM4 forcing.
Line 295 “This indicates… the calibrated range of gamma for this single paraeterization using the AntMean calibration is fairly well constrained.” This appears grammatically odd, reformulate.
Section 3.4: How much of your non-linear behaviour comes from the following sources: (1) drift in the control state if you simply switch the sliding exponent and re-scale basal slipperiness, (2) the fact that the Amundsen Sea basin is simply “empty” at some point, (3) how much your choice of the sliding exponent determines if Pine Island also disappears (under control conditions)?
Line 306: How much of the mass gain, and “non-linear behaviour” for the simulations comes from the drift in the control runs when you simply change parameters? Especially the q=1 case with net mass gain might hint at some model drift generated by switching parameter values. Your figure 2 shows mass gains in that case until 2015, what happens if you extend this case? You might want to consider analysing results relative to the drift.
Line 309: see comment before.
Line 325: “Under …, all three values <=1/3 yield almost the same SLC” is that not because around 1m, the Amundsen Sea basin is empty?
Section 3.5
Line 340 “strong effect” how do you see this is stronger than for the melt sensitivity in the magnitude and patterns?
Line 349: What do you think the effect of your coarse vertical resolution of only 5 layers is on this result?
Section 3.6
What happens in the control runs for the two different solvers?
Section 3.7
General: Are results analysed relative to the model drift?
Figure 20: the two blues are hardly discernible, please update one.
Line 403: “Despite the wide range of melt parameter sampled”, arguably, using the AntMean does not necessarily sample a wide range.
Comparing with Seroussi et al., 2024, Figure 15, your hydrofracture and climate terms reach similar values of around 60cm variance by 2300. However, the “ice model” term in Seroussi is much higher in variance than your “sliding exponent” and “melt parameter” terms, with overall variance of the Seroussi ensemble being larger (1.6m instead of 1m in your ensemble). It appears that the variance you find from hydrofracture and climate forcing is in line with the variance in the full ISMIP6 2300 ensemble. However, you sample less variance due to parameter choices than in Seroussi (where everything goes into the “ice model” term), which explains why overall the relative contributions of hydrofracture and climate are much higher in your case.
Line 400 “Likely reflecting the importance of each of these processes on their own”. Not sure about this, as their individual importance should be in the single contributions? Please explain.
Discussion
Line 431: Feldmann & Levermann 2015 do not make a statement about modern day melt rates causing retreat, Favier et al. 2014 is on Pine Island not Thwaites.
Line 433: Feldmann does not show that reducing melt rates would halt retreat. They show that after kicking the Amundsen Sea with a strong melt water pulse for a short while, reversing to current melt rates is not sufficient to stop long term retreat.
Line 444: “Our .. remarkable insensitivity” see previous comments.
Section 4.3: How big do you think is the influence of vertical resolution on your finding?
Line 525: Do you mean with grounding line parameterization a scaling of basal friction around the grounding line with the grounded fraction or do you mean a schoof-type assumption on grounding line flux?
Line 547: “We take this as further support…” I think this is the first time this hypothesis is raised in the paper. Moreover, I do not understand the jump from discussing model fidelity to the initial conditions.
Line 565: “historical behaviour” I would call this trends in the initial conditions or so, as there was not experiment with historical forcing to test for their behaviour.
Line 566: “This further supports”.. again, the hypothesis has not been stated anywhere clearly and argued for. The variance in your ensemble is overall smaller than in the original ISMIP6 AIS 2300 simulations, and as you say, the reasons could be multiple, ranging from the initial condition uncertainty to you not sampling parametric uncertainty as widely. I do not see how you can exclude all other factors based on your experiments?
Line 571 and following: This is an interesting observation.
Conclusions
Line 673: See comment on Favier and Feldmann before. An interesting study here is also Bett et al., 2024 (TC).Line 675: See comments before on claims of melt sensitivity being less important.
Line 692: How generalisable are your results? Do other ISMIP6 AIS 2300 spin-off studies report similar findings? Otherwise, your findings might be model-specific?
Line 698: Again, I do not follow your argument about initialisation uncertainty.
Appendix
Figures C1-C4: I do not think they were referred to in the main manuscript. Explain please why you include them. Figure C3 something went wrong with the colour bar labels.
Citation: https://doi.org/10.5194/egusphere-2025-3942-RC2 - AC1: 'Reply on RC2', Trevor Hillebrand, 04 Dec 2025
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-3942', Anonymous Referee #1, 02 Oct 2025
General comment:
In this paper, the authors present their contribution to ISMIP6-Antarctica-2300 using the ice-sheet model MALI. They extend the standard ISMIP6 experiments with additional simulations, including time extensions of some runs and a sensitivity analysis of key model parameters and structural choices. They also perform an analysis of variance on a 72-member ensemble to identify the main sources of uncertainty in their projections.
Their results tend to confirm trends suggested by previous studies: a long-term retreat in the Amundsen Sea Embayment under constant present-day forcing, a non-linear sensitivity of the ASE to the sliding parameter, and a sensitivity of the timing of retreat to the choice of basal stress approximation. While these behaviours are not entirely new, they remain interesting as they remind us of the influence of modelling choices and lower model fidelity on the projected mass loss. From my perspective, perhaps the most interesting result suggested by these sensitivity analyses with MALI is the strong sensitivity to thermomechanical coupling, as a result of sub-shelf melting. This highlights the drawbacks of prescribing fixed temperature fields, which is still relatively common in ice-sheet modelling.
The overall goal of the manuscript is to understand, with a single model, how modelling choices affect projections of sea-level change. I believe the authors did this quite well by systematically exploring some key sources of uncertainty. As they write themselves in their conclusion, while perhaps less scientifically ‘exciting’ than multi-model ensembles, such a paper is a valuable documentation of the impact of common assumptions in ice-sheet models. I do not have many comments, as the authors anticipate and discuss most of the limitations I would think of in a maybe somewhat lengthy but thoughtful discussion.
In addition, the study identifies points of attention that will be relevant for other model simulations and, perhaps more importantly, for future community efforts such as ISMIP7. The authors explicitly highlight these aspects and provide constructive guidance for the design of future intercomparison projects. I particularly appreciated their discussion of model fidelity and the sources of uncertainty in the ISMIP6 ensemble, which adds a useful perspective to some of the ISMIP6 results and provides some insights into the possible influence of model-specific parametric uncertainty.
The manuscript is long but clearly written, and I appreciate the effort made to ensure reproducibility of the experiments through detailed methodological descriptions. I found it interesting and enjoyable to read, and I believe it raises some important questions of interest to the community.
I would therefore recommend it for a publication in The Cryosphere, provided the minor comments below are addressed.
Specific comments:
l.63: Consider adding references to the datasets mentioned
l.135-142: A figure illustrating the modifications to the bedrock dataset would be very helpful. Perhaps this could be included as supplementary material.
l.144: This is the first mention of “sectors.” Since multiple basin delineations exist, it would be helpful to specify which sectors you are referring to.
l.168-169: Is the melting of grounded marine termini a significant process in Antarctica?
l.169-171: Do I understand correctly that you do not use the deltaT values provided by Jourdain et al., 2020, but rather calculated your own?
l.184: This was also shown in Coulon et al. (2024), where significant retreat in the ASE is found under constant present-day forcing, using a different initialisation procedure than van den Akker et al. and while accounting for parametric and climate uncertainty.
l.185-186: It would be useful to add a short explanation of what motivated the choice of forcings to be extended. I assume that the goal was to sample a wide range of forcings?
l.272-273: I believe fig.2 from Seroussi et al. (2024) only shows ice thickness and velocity RMSE, not the historical mass change trend.
l.274-275: Maybe I am looking at the wrong figure, but from Figure 4 in Seroussi et al. (2024), I don't find that DOE_MALI contributions are closer to NCAR_CISM than to other models (for example VUW_PISM). Also, I am not sure what to take away from this information.
l.294-296: I’m not sure I understand what you are trying to say here. Could you clarify?
Figures 16 & 18: These figures would benefit from error bars or whiskers, as the overlapping shaded regions are difficult to distinguish.
l.376: I believe the reference should be to Fig. 16b, not 17b.
l.379-381: Consider splitting this sentence into two for clarity.
l.406-407: Could this interaction term be explained by the fact that some ESMs (e.g., CESM) generate more surface melt, making them more susceptible to triggering hydrofracture?
l.422-423: This could support my earlier comment regarding the e–h interaction term.
l.438: I could not find the 1.5 m value in Stokes et al. (2025). Are you referring to the 134 cm reported in their Table 2?
l.441: I agree with your point, but it may also be worth noting that the extended experiments rely on a single model configuration. The onset time of retreat is also likely to be strongly influenced by structural and parametric uncertainties.
l.455-456: This is an interesting result. Could you provide a tentative explanation for the different behaviors of ASE versus Ross and FRIS? Also, did I understand correctly that all simulations with different q values start from the same initial state, with the basal friction field rescaled? If so, please specify this explicitly to rule out influences from the initialisation.
l.517-578: This is an important point, but I think it would be worth noting that Willams et al. expect the effects of model resolution on the upper tail to be model dependent.
l.518: Is it the case for MALI as well? This is what I think I understand from section 2.2, although it is not specifically mentioned.
l.594: One possible impact of surface meltwater on ice dynamics could be its influence on the temperature profile of ice shelves, potentially leading to warmer conditions than those shown in Fig. 13 using the temperature solver.
Figure C3: There is something strange with the colorbar.
Minor corrections:
l.51: ‘can further’ → ‘can be further’
l.566: ‘hypothesize’ → ‘hypothesis’
References
Coulon, V., Klose, A. K., Kittel, C., Edwards, T., Turner, F., Winkelmann, R., and Pattyn, F.: Disentangling the drivers of future Antarctic ice loss with a historically calibrated ice-sheet model, The Cryosphere, 18, 653–681, https://doi.org/10.5194/tc-18-653-2024, 2024.Citation: https://doi.org/10.5194/egusphere-2025-3942-RC1 - AC2: 'Reply on RC1', Trevor Hillebrand, 04 Dec 2025
-
RC2: 'Comment on egusphere-2025-3942', Anonymous Referee #2, 03 Oct 2025
Hillebrand et al. present an analysis of the ISMIP6 AIS 2300 simulations (Seroussi et al., 2024) they performed with MALI as well as some additional sensitivity experiments. This is a useful study to help understanding the large spread of sea level projections for Antarctica.
The manuscript is well written and clear, however lacks some explanations and motivations. I currently do not understand why exactly those parameters, values, and modelling choices were tested, and not others. Furthermore, more explanation on the “non-linearity” of the sliding parameter sensitivity is required, and I cannot follow the argument that the melt parameter choice is less relevant. The argument about initialisation uncertainty in ISMIP6 AIS 2300 being the dominant driver for uncertainty is not clear to me.
I suggest minor comments concerning the framing of the results which should be addressed.
Comments:
Abstract
Line 8: “only moderate sensitivity”. I disagree with this point in contrast to claiming higher sensitivity for other parameters later on. Looking at Figures 7a (melt parameterisation parameters) and Figure 9a and 11a, the melt parameter induces differences >50cm SLE in 2300 for each experiment, which is similar to changes induces by the energy balance solver (which is claimed to induce “very strong sensitivity” in the abstract) and the sliding law exponent (except for the q=1 case for the HadGEM forcing; claimed to induce strongly non-linear dependence). I do hence not see how the authors say that melting has “only moderate sensitivity” in contrast to the other parameters, as the overall influence on the sea level results is at the same order?
Line 9: “ranging from the 5th to the 95th percentile values” – unclear where these come from, think about adding “of the AntMean ISMIP6 calibration” or something similar. I suspect that the choice of your range of parameters has a strong influence on your sensitivity results, and if you had included other values, you would get a different answer (as you mention yourself in the discussion). The statement you can make is that using this parameter tuning, you find a less important contribution in the ANOVA results.
Line 12: not clear what you mean with “vice versa”. Do you mean that FRIS and Ross have less mass loss with a less plastic sliding behaviour? In that case, maybe simply remove it.
Introduction
General: The introduction is missing a discussion of previous results on sensitivity analysis of ISMIP6 simulations. It might be worth checking all ISMIP publications. For example, I could find this study TC - ISMIP6-based Antarctic projections to 2100: simulations with the BISICLES ice sheet model which seems to do something very similar.
Moreover, I cannot find a motivation as to why exactly the parameters/schemes tested were tested, at the moment their choice appears arbitrary. Make sure to motivate your choices!
Adding these two points might make the introduction longer, you could consider removing information that is not directly relevant for your study (e.g., on MICI, calving, flow law exponents, etc).
Line 26: “Onset of such runaway retreat leads to……” – there is a subtly here. To my knowledge DeConto & Pollard, De Conto et al. do not check for the retreat being MISI driven in their results, hence it is not clear if the MISI-driven retreat causes the multiple metres of sea level rise in their simulations. MICI is then added as an afterthought, as if it was to enhance the already strong results. I suggest you reformulate.
Line 58 “regardless of their skill in reproducing historical ice-sheet behaviour”. I suggest reformulating, as it sounds like this was tested/required by ISMIP6, however, reproducing historical behaviour was not an aim of ISMIP6, and the trend analysed by Aschwanden et al. was never meant to be a representation of historical behaviour.
Line 152 “As MISI is likely already under way” the reference is quite old, and more recently, it has been more specified that “MISI being under way” needs to be tested by more than just finding continued retreat in the Amundsen Sea. You do not need MISI here, simply that the ASE is changing is sufficient.
Line 169: How much do your deltaT values differ from the original ones? Note that this might impact how sensitive your melt reacts to changes in the ocean forcing, see for example Lambert & Burgard 2025.
Methods
Section 2.3.3 & 2.3.4. The choice of the parameter ranges for the sub-shelf melt and the sliding parameters appears a bit arbitrary. If they are not sampled equally broadly, how can you claim that one factor has a higher influence than the other?
Results
Section 3.3 & 3.4 How do the control runs look with changed melt or sliding parameters? Is the drift again small enough to be ignored in your plots, or how much influence does this have on your results (especially with the “non-linearities” in the sliding exponent experiments)?
Line 289 “indicate only moderate sensitivity” as stated above, I disagree with this as your difference is still 50-60cm between the low and high end, which is more than a doubling of the contribution for the CCSM4 forcing.
Line 295 “This indicates… the calibrated range of gamma for this single paraeterization using the AntMean calibration is fairly well constrained.” This appears grammatically odd, reformulate.
Section 3.4: How much of your non-linear behaviour comes from the following sources: (1) drift in the control state if you simply switch the sliding exponent and re-scale basal slipperiness, (2) the fact that the Amundsen Sea basin is simply “empty” at some point, (3) how much your choice of the sliding exponent determines if Pine Island also disappears (under control conditions)?
Line 306: How much of the mass gain, and “non-linear behaviour” for the simulations comes from the drift in the control runs when you simply change parameters? Especially the q=1 case with net mass gain might hint at some model drift generated by switching parameter values. Your figure 2 shows mass gains in that case until 2015, what happens if you extend this case? You might want to consider analysing results relative to the drift.
Line 309: see comment before.
Line 325: “Under …, all three values <=1/3 yield almost the same SLC” is that not because around 1m, the Amundsen Sea basin is empty?
Section 3.5
Line 340 “strong effect” how do you see this is stronger than for the melt sensitivity in the magnitude and patterns?
Line 349: What do you think the effect of your coarse vertical resolution of only 5 layers is on this result?
Section 3.6
What happens in the control runs for the two different solvers?
Section 3.7
General: Are results analysed relative to the model drift?
Figure 20: the two blues are hardly discernible, please update one.
Line 403: “Despite the wide range of melt parameter sampled”, arguably, using the AntMean does not necessarily sample a wide range.
Comparing with Seroussi et al., 2024, Figure 15, your hydrofracture and climate terms reach similar values of around 60cm variance by 2300. However, the “ice model” term in Seroussi is much higher in variance than your “sliding exponent” and “melt parameter” terms, with overall variance of the Seroussi ensemble being larger (1.6m instead of 1m in your ensemble). It appears that the variance you find from hydrofracture and climate forcing is in line with the variance in the full ISMIP6 2300 ensemble. However, you sample less variance due to parameter choices than in Seroussi (where everything goes into the “ice model” term), which explains why overall the relative contributions of hydrofracture and climate are much higher in your case.
Line 400 “Likely reflecting the importance of each of these processes on their own”. Not sure about this, as their individual importance should be in the single contributions? Please explain.
Discussion
Line 431: Feldmann & Levermann 2015 do not make a statement about modern day melt rates causing retreat, Favier et al. 2014 is on Pine Island not Thwaites.
Line 433: Feldmann does not show that reducing melt rates would halt retreat. They show that after kicking the Amundsen Sea with a strong melt water pulse for a short while, reversing to current melt rates is not sufficient to stop long term retreat.
Line 444: “Our .. remarkable insensitivity” see previous comments.
Section 4.3: How big do you think is the influence of vertical resolution on your finding?
Line 525: Do you mean with grounding line parameterization a scaling of basal friction around the grounding line with the grounded fraction or do you mean a schoof-type assumption on grounding line flux?
Line 547: “We take this as further support…” I think this is the first time this hypothesis is raised in the paper. Moreover, I do not understand the jump from discussing model fidelity to the initial conditions.
Line 565: “historical behaviour” I would call this trends in the initial conditions or so, as there was not experiment with historical forcing to test for their behaviour.
Line 566: “This further supports”.. again, the hypothesis has not been stated anywhere clearly and argued for. The variance in your ensemble is overall smaller than in the original ISMIP6 AIS 2300 simulations, and as you say, the reasons could be multiple, ranging from the initial condition uncertainty to you not sampling parametric uncertainty as widely. I do not see how you can exclude all other factors based on your experiments?
Line 571 and following: This is an interesting observation.
Conclusions
Line 673: See comment on Favier and Feldmann before. An interesting study here is also Bett et al., 2024 (TC).Line 675: See comments before on claims of melt sensitivity being less important.
Line 692: How generalisable are your results? Do other ISMIP6 AIS 2300 spin-off studies report similar findings? Otherwise, your findings might be model-specific?
Line 698: Again, I do not follow your argument about initialisation uncertainty.
Appendix
Figures C1-C4: I do not think they were referred to in the main manuscript. Explain please why you include them. Figure C3 something went wrong with the colour bar labels.
Citation: https://doi.org/10.5194/egusphere-2025-3942-RC2 - AC1: 'Reply on RC2', Trevor Hillebrand, 04 Dec 2025
Peer review completion
Journal article(s) based on this preprint
Data sets
MPAS- Albany Land Ice simulations of the Antarctic Ice Sheet through 2300: Exp02–05 flux fields T. Hillebrand et al. https://doi.org/10.5281/zenodo.16803637
MPAS- Albany Land Ice simulations of the Antarctic Ice Sheet through 2300: Exp02–05 state fields T. Hillebrand et al. https://doi.org/10.5281/zenodo.16798097
MPAS- Albany Land Ice simulations of the Antarctic Ice Sheet through 2300: Exp11–14 flux fields T. Hillebrand et al. https://doi.org/10.5281/zenodo.16805185
MPAS- Albany Land Ice simulations of the Antarctic Ice Sheet through 2300: Exp11–14 state fields T. Hillebrand et al. https://doi.org/10.5281/zenodo.16805033
MPAS-Albany Land Ice simulations of the Antarctic Ice Sheet through 2300: time series T. Hillebrand et al. https://doi.org/10.5281/zenodo.16805241
Model code and software
Energy Exascale Earth System Model E3SM Team https://doi.org/10.5281/zenodo.16809723
Interactive computing environment
mali-ismip6-ais-2300-anova analysis scripts Matthew Hoffman https://doi.org/10.5281/zenodo.16813658
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 14,605 | 1,492 | 202 | 16,299 | 405 | 403 |
- HTML: 14,605
- PDF: 1,492
- XML: 202
- Total: 16,299
- BibTeX: 405
- EndNote: 403
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Matthew J. Hoffman
Holly K. Han
Mauro Perego
Alexander O. Hager
Andrew Nolan
Xylar Asay-Davis
Stephen F. Price
Jerry Watkins
Max Carlson
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(73247 KB) - Metadata XML
General comment:
In this paper, the authors present their contribution to ISMIP6-Antarctica-2300 using the ice-sheet model MALI. They extend the standard ISMIP6 experiments with additional simulations, including time extensions of some runs and a sensitivity analysis of key model parameters and structural choices. They also perform an analysis of variance on a 72-member ensemble to identify the main sources of uncertainty in their projections.
Their results tend to confirm trends suggested by previous studies: a long-term retreat in the Amundsen Sea Embayment under constant present-day forcing, a non-linear sensitivity of the ASE to the sliding parameter, and a sensitivity of the timing of retreat to the choice of basal stress approximation. While these behaviours are not entirely new, they remain interesting as they remind us of the influence of modelling choices and lower model fidelity on the projected mass loss. From my perspective, perhaps the most interesting result suggested by these sensitivity analyses with MALI is the strong sensitivity to thermomechanical coupling, as a result of sub-shelf melting. This highlights the drawbacks of prescribing fixed temperature fields, which is still relatively common in ice-sheet modelling.
The overall goal of the manuscript is to understand, with a single model, how modelling choices affect projections of sea-level change. I believe the authors did this quite well by systematically exploring some key sources of uncertainty. As they write themselves in their conclusion, while perhaps less scientifically ‘exciting’ than multi-model ensembles, such a paper is a valuable documentation of the impact of common assumptions in ice-sheet models. I do not have many comments, as the authors anticipate and discuss most of the limitations I would think of in a maybe somewhat lengthy but thoughtful discussion.
In addition, the study identifies points of attention that will be relevant for other model simulations and, perhaps more importantly, for future community efforts such as ISMIP7. The authors explicitly highlight these aspects and provide constructive guidance for the design of future intercomparison projects. I particularly appreciated their discussion of model fidelity and the sources of uncertainty in the ISMIP6 ensemble, which adds a useful perspective to some of the ISMIP6 results and provides some insights into the possible influence of model-specific parametric uncertainty.
The manuscript is long but clearly written, and I appreciate the effort made to ensure reproducibility of the experiments through detailed methodological descriptions. I found it interesting and enjoyable to read, and I believe it raises some important questions of interest to the community.
I would therefore recommend it for a publication in The Cryosphere, provided the minor comments below are addressed.
Specific comments:
l.63: Consider adding references to the datasets mentioned
l.135-142: A figure illustrating the modifications to the bedrock dataset would be very helpful. Perhaps this could be included as supplementary material.
l.144: This is the first mention of “sectors.” Since multiple basin delineations exist, it would be helpful to specify which sectors you are referring to.
l.168-169: Is the melting of grounded marine termini a significant process in Antarctica?
l.169-171: Do I understand correctly that you do not use the deltaT values provided by Jourdain et al., 2020, but rather calculated your own?
l.184: This was also shown in Coulon et al. (2024), where significant retreat in the ASE is found under constant present-day forcing, using a different initialisation procedure than van den Akker et al. and while accounting for parametric and climate uncertainty.
l.185-186: It would be useful to add a short explanation of what motivated the choice of forcings to be extended. I assume that the goal was to sample a wide range of forcings?
l.272-273: I believe fig.2 from Seroussi et al. (2024) only shows ice thickness and velocity RMSE, not the historical mass change trend.
l.274-275: Maybe I am looking at the wrong figure, but from Figure 4 in Seroussi et al. (2024), I don't find that DOE_MALI contributions are closer to NCAR_CISM than to other models (for example VUW_PISM). Also, I am not sure what to take away from this information.
l.294-296: I’m not sure I understand what you are trying to say here. Could you clarify?
Figures 16 & 18: These figures would benefit from error bars or whiskers, as the overlapping shaded regions are difficult to distinguish.
l.376: I believe the reference should be to Fig. 16b, not 17b.
l.379-381: Consider splitting this sentence into two for clarity.
l.406-407: Could this interaction term be explained by the fact that some ESMs (e.g., CESM) generate more surface melt, making them more susceptible to triggering hydrofracture?
l.422-423: This could support my earlier comment regarding the e–h interaction term.
l.438: I could not find the 1.5 m value in Stokes et al. (2025). Are you referring to the 134 cm reported in their Table 2?
l.441: I agree with your point, but it may also be worth noting that the extended experiments rely on a single model configuration. The onset time of retreat is also likely to be strongly influenced by structural and parametric uncertainties.
l.455-456: This is an interesting result. Could you provide a tentative explanation for the different behaviors of ASE versus Ross and FRIS? Also, did I understand correctly that all simulations with different q values start from the same initial state, with the basal friction field rescaled? If so, please specify this explicitly to rule out influences from the initialisation.
l.517-578: This is an important point, but I think it would be worth noting that Willams et al. expect the effects of model resolution on the upper tail to be model dependent.
l.518: Is it the case for MALI as well? This is what I think I understand from section 2.2, although it is not specifically mentioned.
l.594: One possible impact of surface meltwater on ice dynamics could be its influence on the temperature profile of ice shelves, potentially leading to warmer conditions than those shown in Fig. 13 using the temperature solver.
Figure C3: There is something strange with the colorbar.
Minor corrections:
l.51: ‘can further’ → ‘can be further’
l.566: ‘hypothesize’ → ‘hypothesis’
References
Coulon, V., Klose, A. K., Kittel, C., Edwards, T., Turner, F., Winkelmann, R., and Pattyn, F.: Disentangling the drivers of future Antarctic ice loss with a historically calibrated ice-sheet model, The Cryosphere, 18, 653–681, https://doi.org/10.5194/tc-18-653-2024, 2024.