Review of Lecavalier et al. “A history-matching analysis of the Antarctic Ice Sheet since the last interglacial – Part 1: Ice sheet evolution”

Summary

The work of Lecavalier & Tarasov uses a new numerical model, the Glacial Systems Model (GSM), to reconstruct the evolution of the Antarctic Ice Sheet (AIS) since the Last Interglacial through a very large ensemble of simulations constrained by likely the most comprehensive paleo-ice sheet constraint database up to date (AntICE2). The GSM is not just an ice sheet model but presents coupled modules and parameterisations allowing it to consider several different processes that affect ice sheet dynamics. As a result, a total of 38 model parameters can be varied throughout the ensemble to assess different types of uncertainty inherent to (paleo) ice sheet modelling. The parameter space covered for each parameter was refined through what the authors call “waves”, i.e., different iterations of their large ensemble run, which were then used to train an emulator that, in combination with MCMC sampling, helped determine which range of values the model was most sensitive to, and at the same time produced results that best matched the AntICE2 constraints within the standard deviation boundaries allowed by the authors. Once the final, most refined, ensemble of runs was obtained, the AIS evolution was assessed for key periods: the Last Glacial Maximum and Meltwater Pulse 1A. A proper evaluation of the Last Interglacial was hindered by the lack of suitable paleo constraints.

Apart from refinements in terms of assessing the AIS volume at the LGM and its contribution to Meltwater Pulse 1A, I believe this manuscript’s greatest contribution is the presentation of a novel and much more comprehensive methodology for constraining ice sheet history from large ensemble modelling, and a numerical model that integrates several processes in a way that allows for an unprecedented number of (uncertain) parameters to be explored. For this reason, both these key points need to be described very clearly and able to be reproduced, which is the main point where this manuscript needs improvement.

This work definitely brings enough innovation to merit publication and will be of wide interest to the numerical modelling and (paleo)glaciology communities, but there are substantial issues which need to be addressed before I can recommend it to be published. Apart from a reorganisation of the introduction and some rewriting or clarification of some sections, for which my comments and suggestions are listed below, the model-description paper cited as Tarasov et al., in prep needs to be available in some way for a proper appreciation of this work by both reviewers and readers. I assume, since it is “in prep”, that there is no pre-print available. I expect that a publicly available pre-print would provide enough information to solve this main issue, both for me as a reviewer and for the potential readers that will come across this paper before the model description is published in its final version. As mentioned, my comments regarding this paper are provided below, since I believe the study merits being published eventually (and to expedite the process), but not before details on GSM are available, and that the results presented here can be reproducible.

General comments

1.My greatest concern with the manuscript is the fact that it uses an unpublished model, for which there is no available manuscript (referred to as “Tarasov et al., in prep”) or reference to “data and code availability”. This poses a serious issue for reproducibility and for a proper assessment of several parts of the methodology, as I believe these are not in this manuscript because they will be properly described in the cited in-prep manuscript. Considering the model presents several improvements and more complex parameterisations/couplings than the current ice sheet models, including its older version (Briggs et al., 2013; TC), e.g., in basal sliding dynamics, sub-grid pinning point parameterisations, different surface boundary conditions, basal melting of floating and grounded ice, it is of extreme importance that proper descriptions of all these innovations are available. Similarly, the way the three forcing schemes are described (L243) is rather vague, and without a thorough description, it is hard to assess how they differ from each other. In a similar fashion, the way how the data-model scoring is done should be more explicit, as it is hard to evaluate from the descriptions given for each data type. Without seeing the in-prep manuscript, I believe this fits better in this manuscript than in the in-prep paper (L370).

2.The description of model initialisation/spinup is not clear. There are some hints of part of it throughout section 3, e.g., it seems to start from PD geometry, but there are no details on how the internal ice thermal structure was initialised, or how the spatial distribution of basal drag was obtained apart that it depends on elevation. The description given in L291-302 is not enough to properly evaluate how this was done. I assume a better description will be available in the in-prep paper, but this is not available for a proper understanding and fair assessment of the methodology.

3.The assessment of Section 6 is hindered due to the fact that Fig. 11 is completely absent from the manuscript. From the text, I have the feeling that this figure is key to understand the discrepancies between the much larger ice volume simulated during the LGM – an important piece of this work.

4.The introduction needs a thorough rearrangement of paragraphs. The content is good, it indeed introduces all concepts and sets up the problem that the paper wants to address. However, it does so in a rather erratic way, and there is little connection between the paragraphs. For example, I would expect the paragraph starting on L47 (which sets up the problem and paper aims) to come at the end of the introduction. Other concepts that need to be introduced, such as the key periods targeted by the paper (MWP1a, LIG, LGM), data-related issues (types of data for ice sheet reconstructions, their availability, and their use in modelling), and outstanding research questions have been presented, but all come after what reads as the closing introduction paragraph. Similarly, current model limitations (paragraph at L36) should come much later in the introduction, after discussing the points mentioned above. Therefore, I strongly recommend the contents of the introduction to be revised and rearranged, with some rewriting to better link paragraphs and improve the flow of reading.

5.Why did the authors opt to use TraCE-21ka to reconstruct their forcing? Considering the transient signal is reconstructed with a glacial index, instead of using TraCE throughout, this choice needs further justification.

6.It would be worth expanding briefly on how the classification of data as tier1 to tier3 is done (suggestion: a summarised version of the 1^st paragraph of section 3.1 in Lecavalier et al., 2023; ESSD). Even though the methodology is detailed in the original paper, given the strong relevance to the assessment performed in this manuscript, it is worth expanding a bit more on it.

7.The concept of “waves” should be more clearly described (likely in the paragraph starting in L382), because the amount of work put into such an “ensemble of ensembles” really cannot be properly appreciated until much further down the manuscript.

Specific comments

L196: How are the ice flow approximations coupled in the hybrid scheme? Are they just superimposed, or is there a (e.g., weighted) transition between them?

L242: The mention of the glacial index here does not read well. You first mention about the three forcing schemes, then mention the glacial index, and then proceed to present the forcing schemes. I suggest mentioning the glacial index first, and that it is used by (seemingly) two of the three schemes, before describing them.

L243: The three forcing schemes could benefit from a more thorough description that would help in evaluating and reproducing the results.

L251: Please make it clear whether the positive-degree-day and positive-temperature-insolation surface-melt schemes are used for all forcing schemes or not.

L264: What is the reasoning for grouping Dronning Maud Land with Wilkes and Victoria Land? The former is rather far apart from the latter two and is under a different oceanographic regime. This needs to be either properly justified or they need to be separated.

L271: It is not clear how the far-field global sea-level forcing affects the ice sheet in GSM. I assume through their solid-Earth model, and that it would be better described in the model-description paper, but this needs some sort of description here.

L273: This paragraph could use some rewriting. Not only do the sentences not properly link well, but it is not clear how the basal forcing conditions are computed. Furthermore, it is not clear how the two different fields are blended. Is the “ensemble parameter” some sort of weight between two fields? This is only made explicit way further down, in L525, and should also be clearly stated when first describing the method. Finally, the first sentence could be better formulated: wouldn’t basal topography and thickness already yield ice-surface elevation? Or do the authors mean “bedrock topography”? I can see this being argued and explained in different ways (i.e., among bedrock topography, ice-base topography, ice thickness, and ice-surface topography), but as it stands it is not consistent.

L355: Please clearly describe how “a quadrature was calculated”.

L396: This paragraph is very helpful in explaining the types of uncertainties that have to be dealt with in the history-matching approach. It should be moved further up, before or when the term “structural uncertainty” is used for the first time (Section 4.1).

L415: I find it problematic to use a method not fully described anywhere that is deferred to a future publication. At least some level of detail should be given, even if it is an appendix to the paper.

L444: This sentence could be rewritten in a much clearer way. Suggestion: “Simulations beyond the thresholds for each data type (Table S1) were then ruled out from the history-matching analysis (i.e., sieved out).”

L508-513: I can follow the reasoning, but the writing feels a bit disconnected. It would be good to rewrite. For example, stating that you are talking about vertical advection – this might not be clear considering the mention of “higher velocities at warm beds drawing in ice”, which I believe is a reference to horizontal movement.

L539: Why are the ice stream profiles relegated to a lower tier? I had understood that data tiers were linked to their quality, but this phrasing suggests it is by how well the data is captured by the ensemble. Perhaps it is just bad wording, but please clarify.

L550: I am not sure I followed the reasoning in this sentence. Do you mean that the subsequent “ensemble waves” aimed to improve this mismatch? Or what kind of analysis/prioritisation was done? Or do you mean that you will focus on the more inland temperature profiles? Please clarify.

L808: The figures mentioned are quite heavy in information, and it is not trivial to find exactly which parameters the authors refer to from a glance. It would be helpful to name which parameters are considered to induce the greatest sensitivity instead of just referring to the figures.

L813-814: This is a very important statement and should be used as an opportunity to guide further field efforts rather than just acknowledge a limitation to hopefully be addressed in the future. I think the authors should explicitly state/suggest what kind of “novel constraints” would be helpful, and which regions should be prioritised for data collection (e.g., in which data-poor region was the ensemble range the widest?).

L835: Please mention which parameter combinations you refer to when mentioning Fig. S10, as it is not trivial to hunt for different parameter pairs in such a large heatmap.

Technical corrections

L99: there is an extra ‘(‘ in the citations.

L109: Greenland Ice Sheets? It is not clear why it is in plural.

L120: I don’t think the paper DOI should be listed in the citation.

L184: Terms such as “paleoExt” and “paleoH” should be clearly defined at their first mention. This only happens in L363 and L357, respectively.

L267: TRACE here is all capitalised, please change for consistency.

L274: “Poorly observed”, as there is no hyphenation for adverbs

L278: “an envelope of model realisations”, for clarity

L286: either “across Antarctica” or “across the Antarctic”

L345: I believe a comma separating these two sentences works better than a full stop

L483: “metrics of interest”

L513: Please remove the comma

L551: temperature profiles

L659-660: There should be no commas in this sentence

L728: Please remove “ice shelves”, as this is already clear from how the sentence begins.

L732: This sentence needs rewriting for better readability. Suggestion: “Slow-moving ice is usually present at inland locations where the PD elevation is above this threshold.”

L733-735: These two sentences can be merged. Suggestion: “Conversely, faster-moving marginal ice is expected for areas below the surface elevation threshold, which include the aforementioned ice streams/shelves.”

L855: Please replace the comma for “that”, and “is” for “was”

L907: “This study presents”

L909: Please remove the comma.

L930: There’s a comma missing between “Balco” and “Claus”

Figures

Figure 1: Please add subpanels where the main locations mentioned in the text are marked. I appreciate that adding all names to the already existing figure would not work, as there would be too much clutter, hence the suggestion to add the most important regions discussed as subpanels.

Figure 10: Please make the x axes have the same maxima and minima for each row, so the different ensembles can be more easily compared, and increase the font size of the legend in each plot.

Figure S1: Please reorder the panels according to the order that they are mentioned in the text and refer to the respective subpanel when mentioning Fig. S1 in the main text.

Please increase the figure resolution for figures S2, S11, S12.

General comments

In this paper, the authors summarized the results of nearly 30,000 ensemble experiments based on the dataset established by the authors, reconstructing the history of the Antarctic Ice Sheet from the Last Interglacial to the present. The papers about the model itself and GIA are published separately, so it is difficult to make a judgement on these. But, I believe this series of studies will make a significant contribution to this research field.  After all related papers published, a concise summary paper would make a greater contribution to the community.