the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 Jun 2026
Fast approximation of Antarctica’s GIA response to future ice melt with 3-D Earth structure
Abstract. Projections of Antarctic ice mass loss and associated sea level contributions over the coming centuries are intrinsically linked to glacial isostatic adjustment (GIA), a process by which changing ice sheets deform the solid Earth and sea surface. Altering bedrock topography and sea levels at the grounding line, GIA exerts a strong control on marine ice sheet dynamics, especially on the multi-century timescales to be considered in ISMIP7 (Ice Sheet Model Intercomparison Project for the Coupled Model Intercomparison Project - Phase 7). To accurately capture bedrock and sea level changes, ice sheet models must be coupled with GIA models that include realistic spatial variations in solid Earth structure. However, GIA models that incorporate 3-D variations in Earth structure are computationally expensive, limiting their use in ice sheet modelling. Consequently, most ice sheet models still assume rigid bedrock topography or rely on simple GIA models that neglect realistic lateral variations in Earth structure. Here, we assess the performance of FastIsostasy, a computationally-efficient regional 2-D GIA model, relative to Seakon, a state-of-the-art 3-D GIA model, in iteratively coupled ice sheet – GIA simulations of Antarctic Ice Sheet evolution over the next five centuries. Coupled simulations that employ FastIsotasy produce GIA, ice thickness, and grounding line predictions that closely match those from simulations using Seakon, and, more specifically, perform better than ice sheet simulations that rely on overly simplified GIA models. With the protocols for ISMIP7 under active development, FastIsostasy offers a viable approach for ice sheet modellers to accurately and efficiently capture solid Earth – ice sheet feedbacks, permitting improved projections of Antarctic ice mass change and associated sea level contributions over the coming centuries.
- Preprint
(30457 KB) - Metadata XML
-
Supplement
(20360 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2026-2640', Holly Han, 07 Jul 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-2640/egusphere-2026-2640-CC1-supplement.pdfCitation: https://doi.org/
10.5194/egusphere-2026-2640-CC1 -
RC1: 'Comment on egusphere-2026-2640', Matt King, 09 Jul 2026
The authors present an analysis of the FastIsostacy code against a fully featured 3D GIA code, Seakon, in the context of an iteratively-coupled ice sheet/GIA model. Such a code would provide practical benefit to sea level projection realism and is a very worthy topic with potential for significant downstream impact. The work finds that these two models are in pretty close agreement, happily so, and certainly much better agreement than with ELRA style models. The work is well suited to TC and will be of wide interest to those working with ice sheet models, especially given the speed of the FastIsostacy model and its potential for coupling to ice sheet models. The paper is clearly written and makes good use of figures.
I don't have any major concerns, although I think there are things that do need revising.
Section 2.3.2 outlines the weak earth model which reaches to just below 10^19 Pa s in regions where the geodetic data have suggested more like 3x10^18. I think this section and 2.3.1 should summarise the range of upper mantle viscosities and here or in the discussion note that the model doesn't go as low as the obs may suggest and what issue that may cause. This is important for the ASE, especially since most of the ice loss will occur there this century. That is, the testing in this paper doesn't cover the viscosities most relevant to the near future. Out of interest, is this a technical restriction in Seakon or FastIsostacy or both that prevented going to lower viscosities?
Section 3.1 . The focus in the text is mean differences, but I think max abs(diff) needs reporting also. These are much larger. Figure 4 could have 2 further panels to show maximum differences. This applies to all subsequent discussions. I presume the means are spatial means computed over the common grounded ice sheet? Further, the regions where the differences are largest is also worth mentioning. They are all at the grounding line. This difference may affect different types of ice sheet models in different ways., notably those that resolve grounding line regions differently.
The quite bold statement on L284 that more complex GIA models than Fastisostacy probably needs the caveat that it applies to the class of ice sheet model used here and within the range of mantle viscosities tested.
As a separate note, I wonder if the bedrock topography offshore, which controls warm water pathways, is also little different between the two approaches. This isn't relevant to the ice-GIA coupling, but it is for the ocean-ice-GIA system. That may be something to comment on in discussion if it is trivial to compute.
L58 benchmarks are referred to lacks an explicit reference to Swierczek-Jereczek et al., 2024 and I think it should be cited again in this sentence to avoid confusion. I think some quantification of "performed well in benchmarks" would be appropriate for the introduction.
More minor remarks:
L148 it may not be necessary to modify the example of Wilkes Subglacial Basin, but I note the paper by Hansen and Emry suggests there could be lower-viscosity mantle in this region. https://www.nature.com/articles/s43247-025-02140-4
L212 re RCP8.5 - it is appropriate to say that this is no longer regarded a realistic pathway due to climate mitigation and hence is an extreme upper bound for comparing the models. https://gmd.copernicus.org/articles/19/2627/2026/
L244 references Fig 2a, but I think 2c,f is meant.
the discussion of Figure 7 would be improved by referencing the individual panels of the figure in the text.
Conclusions. Perhaps some comment could be made about the coupling interface and how this may or may not suit other ice models.
Citation: https://doi.org/10.5194/egusphere-2026-2640-RC1 -
RC2: 'Comment on egusphere-2026-2640', Ingo Sasgen, 12 Aug 2026
Dear Jan Swierczek-Jereczek and co-authors,
The manuscript “Fast approximation of Antarctica’s GIA response to future ice melt with 3D Earth structure” (egusphere-2026-2640) presents coupled ice sheet/GIA simulations for the next five centuries. Its aim is to evaluate the performance of a regional 2D GIA representation compared to a full 3D GIA model solution, particularly in the context of the stark lateral differences in Earth structure present in Antarctica. The model setup consists of the 3D model Seakon (used as reference) and FastIsostasy, a computationally far less expensive approximation, both coupled to the thermomechanical ice sheet / ice shelf model PSUICE3D. The simulations follow ISMIP6-type experiments. Earth model parameters for Antarctica are derived from a published regional seismic velocity solution, converted to viscosity using two different established methods to construct a weak and a strong Earth structure model. The approximation’s performance is evaluated against Seakon using several metrics, including ice thickness, grounding line position, vertical displacement, and sea-surface elevation, at various stages of the future evolution. The authors claim that FastIsostasy approximates the GIA response sufficiently well, and better than even simpler realizations, opening a pathway for improved ice sheet projections in ISMIP7-like experiments.
The manuscript is clear, well-written, and pleasant to read. The figures are of high quality, even though they do not always present the results in the best way (see comments below). The study is novel and timely, in the sense that the implementation of GIA responses in ice sheet models is still far from common. The study is carefully designed to answer the targeted questions; the models are state-of-the-art, and the nesting of the Antarctic Earth structure within the global models is nicely done. On a personal note, I think the improvement of FastIsostasy over the ELRA models is not as large as expected. And I think the authors could highlight the advantages of FastIsostasy more clearly (e.g., sea-level representation). However, I would like to see an improvement to Figures 2 and 3 (see major comments), and would also like the authors to consider my suggestions for the other figures. Finally, the authors should be explicitely mentioned that evaluation is specific to this time framge, range of viscosities and not the least the ice sheet model / glacial history.
I consider these changes minor, since they require no new computations or additional checks. After these changes are made, I think the paper is acceptable for publication in The Cryosphere.
Best regards,
Ingo SasgenMajor comments
[L255ff] Figures 2 and 3. It is recognized that this paper focusses on the difference between regional 2D versus 3D coupled simulations. However, without a map of the absolute changes in ice thickness, or a relative metric, the figures become far less useful; the reader can't judge whether the 2D approximation is justified, given the magnitude of load changes. Second, it is impossible to judge whether 2D really fails for the MICI scenario or whether it is just a large load change overall. Also here, a relative metric could help for the intercomparison. Helpful would also be to zoom in on the Amundsen Sea Embayment and the Wilkes Land coastal sector. Also, do you really need the Zwally basins, in particular without labeling…? Further, there is a performance degradation for FastIso and weak Earth structures & MICI concerning the grounding line position (Fig. 4). This is invisible in Fig. 2; so I think it is safe to remove all grounding line positions from Fig. 2 and 3. Also, is this difference significant; if not, the conclusion would be that the ELRA approximations would be passable as well… The very critical perspective one might take on the results is that, for these projections, FastIso does not significantly outperform ELRA500.
[L285ff] Figures 5 and 6. It would be helpful to show relative sea-level as well. After all, this is the relevant quantity for the grounding line dynamics. And it would nicely show how vertical displacement controls sea-level fall and acts as a stabilization for the weak Earth model. Thus it also shows that, although FastIso may fail to represent the sea surface well, it represents the relative sea-level well. Finally, it would also be great to show rates as well, particularly for displacement, in order to put them in context with uplift rates at GNSS stations, possibly selecting one or two locations.
[L310ff] The introduction mentions the stabilizing effect of rapid rebound on ice-dynamic retreat. Yet, the barystatic curve for the strong Earth model appears to be lower in 2500 than the one for the weak Earth model. Please comment on this. I am pretty sure this is simply due to the limited projection period (one problem of the study design in general that the interesting effects like stabilization and tipping will only appear on longer time scales). In addition, FastIso seems to systematically perform worse for the low-viscosity cases (e.g. Figure 4, grounding line; Figure 6, local sea surface height) – can you explain why? A final comment: the improvement with FastIso may be considered not that significant compared to a tuned ELRA model. Considering all the loose ends in ice sheet modelling and the uncertainties in the forcing, I am afraid that modelers will stick with ELRA even if this means that sea level cannot be represented properly. However, making your code available will certainly help to improve coupled runs.
Minor and Technical Comments
[L3] GIA control much weaker than ocean forcing or bedrock parameters. Maybe consider writing „important“.
[L6] “include realistic spatial variations in solid Earth structure ”. And gravitational and rotational variations?
[L8] Actually, when checking I was surprised to see that the majority of ISMIP6 models use a rigid bed approximation; nonetheless, you should introduce the more advanced approximations like ELRA you use later as intermediate step between a rigid Earth and a 1D GIA model.
[L13] simpler – "overly simplified" is subjective and is part of the results of this paper.
[L19] I think it would be fair to include some climatic drivers here, which exert the dominant control and are responsible for a large part of the uncertainty.
[L25] Across → underneath; also, please mention values for this statement already here: 10^17 Pa·s to 10^22 Pa·s?
[L26] Absolute values would be more helpful.
[L32] Please add: https://doi.org/10.1038/s43247-026-03738-y
[L44] Maybe add: „although computationally less expensive“, or similar.
[L49] models → modelling
[L56] and radially? You explicitly mentioned this before. Also please very briefly describe how the approximation works.
[L60] It might be useful to mention that you use the PennState ice sheet model already here, to highlight that that you are really coupling two numerical models.
[L73] Please mention that you start from a quasi-equilibrium state of the GIA model in 1950, and that the deformational response from earlier retreat is neglected (if this is the case).
[L73] Changes in bedrock à vertical deformation and sea surface elevation – the quantities you actually show in the figures.
[L83] I know the effect won't be large – but can you mention how you deal with Greenland Ice Sheet mass loss, or give a justification for neglecting it?
[L109] border à domain margin?
[L116] What grid spacing are you choosing for the FastIso runs? 10 km, as for the reference 3D GIA model and PSUICE3D?
[L129ff] It is a bit awkward to state the elastic lithosphere thickness explicitly and then fall back to seismic velocity anomalies for the upper mantle structure. Please resolve.
[L132] Could you please state the extent of these “localized regions”? Are these small patches of super-low viscosity within the rift system of West Antarctica? I think this is important for the GNSS interpretation.
[L136ff] Are these average upper-mantle viscosities? Could there even be a lower-viscosity asthenosphere layer?
[L143] time → times
[L144] … fascinating and a nightmare from a modelling standpoint … (comment)
[L149] Please add: Konrad et al., 2015
[L150-156] Please shorten to 30-50%. A bit generic and repetitive.
[L156] Please mention already here that you consider these models as plausible upper and lower bounds.
[L157] I think it would be very helpful to make the 3D Earth structure models available, and add them to the data sets of plausible viscosity models for Antarctica.
[L164] GLAD-M25 → global seismic shear-wave speed model GLAD-M25
[Table 1] Please state, for completeness, what happens below 800 km.
[L206] I think the approach of performing the lumping based on minimizing deviations from the 3D reference is a practical one; using the minimum viscosity within 150 km of the LAB is somewhat physically justified. However, it was based on the specific weak and strong Earth models. Do you think this is transferable to other viscosity profiles and simulations?
[Figure 1] Would it be possible to compress the color scale at high values for lithosphere thickness, to allow a finer color grading for the thinner values? It would be interesting to see more detail for West Antarctica and along the continental margin.
[L215] Do any of these simulations map onto ISMIP6 experiments? It would be impactful to place the results within the wider modelling community's efforts.
[L216] A bit difficult to understand, maybe: “…producing a spatially large and complex GIA response that is more difficult to approximate by a 2-D regional GIA model; an extreme case which specifically tests the performance of FastIsostasy versus Seakon.” You could also give some more details on why FastIsostasy may fail in the MICI case (forebulge, domain boundary, rotation, etc.).
[L233] Please add information on the initialization of the solid Earth models.
[L236] Add: Antarctic → future Antarctic
[L237] Predictions → evolution.
[L239] Predicted → represented.
[L242] “Average differences in ice thickness” Please explain how this metric is calculated.
Citation: https://doi.org/10.5194/egusphere-2026-2640-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 140 | 70 | 14 | 224 | 43 | 11 | 15 |
- HTML: 140
- PDF: 70
- XML: 14
- Total: 224
- Supplement: 43
- BibTeX: 11
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1