The following review is for the manuscript submitted to EGU, entitled: “Uncertainty in the projected Antarctic contribution to sea level due to internal climate variability”, by J. Caillet and others.

The submitted manuscript describes a model investigation of the uncertainties in medium-range projections of the Antarctica Ice Sheet’s contribution to sea level.  The study focuses on uncertainties related to internal climate variability of climate forcing, both ocean and atmospheric.  First, the authors evaluate CMIP6 models, and choose a subset on which to conduct their analysis.  They then present historical diagnostics on the available model ensemble for each climate model, extracting the temporal variability of various ocean and atmospheric related variables. Finally, they choose a subset of ensemble members of the SSP2-4.5 scenario for each climate model and use them to force an Elmer/Ice continental Antarctica, resulting in an ensemble of projections for Antarctic Ice Sheet sea-level contribution through year 2100.  Results suggest that internal climate variability can affect sea-level contribution, ranging in magnitude from 45-93%, but most of that uncertainty is dominated by atmospheric forcing over ocean forcing.  The authors conclude that internal climate variability varies among the climate models, especially for the ocean forcing; therefore, they suggest a strategy for choosing ensemble members that most realistically represent the dominant climate modes of the Antarctic region.  They also make recommendations for how to best consider internal climate variability in ice sheet model projections.  In general, the methods are well-described and the figures are adequately presented.  The analyses and science results are of high quality, and the discussion and conclusion bring up intriguing and relevant points for the ice sheet modeling community.

Overall, I find that this is an interesting study, with important results comparing the effect of internal climate variability due to the ocean and the atmosphere on ice sheet model projections.  While results presenting ice sheet modeling projections by themselves could constitute their own manuscript, the authors present much more analysis, including a list of metrics for choosing appropriate model ensemble members to capture internal climate variability.  While interesting, these metrics are not the ones used for choosing members for the ensemble results presented.  In addition, the authors do not show outcomes that illustrate/quantify the consequences resulting from an ice sheet model using all the suggested updates to their projection procedures.  As a result, I find that the addition of these extra results leads to a manuscript that lacks focus. For instance, I think it would benefit the manuscript if some of the secondary analyses were moved to a supplement.  In this way, the main manuscript could be dedicated to presenting results specifically on the quantification of the uncertainty in Antarctica’s sea level contribution due to internal climate variability. If the authors feel as if the new metrics should be highlighted instead, then a new organization and general story built around those results would benefit the manuscript.

Due to the extensive modification needed, I suggest that major revisions be required before the manuscript is accepted. If the authors work on better organization of the results and on improving the clarity of their language (per suggestions outlined below), I am confident that this manuscript could result in a valuable scientific contribution to the community.  Please see my comments/suggestions/questions below with regards to my major and minor concerns with the current version of the paper. 

General comments/questions:

• Results show that the ocean internal variability has a minimal effect on the projections, as compared to atmospheric variability or choice of model. If this is the case, why do the designed ensemble metrics mostly focus on evaluating the ocean forcing of ensemble members?  More specifically, why is it pertinent to choose members that capture ocean internal variability well if this variability is less important?  Do the authors suggest that the climate model ocean representation of internal climate variability lacks skill to the point that using an entire ensemble of forcing does not offer a realistic projection spread? It would improve the manuscript if these questions were considered in the text/discussion/overall story of your paper.  It would be even more beneficial to the paper if the authors could support the answers with analysis or results, expanding upon the plots that are already included in the paper.

• As discussed in your manuscript, climate model ensembles are typically used to represent the spread of model internal climate variability. Forcing the ice sheet model with a large subset of members allows for the propagation of uncertainty due to this variability into projections of sea-level contribution.  Here, it is suggested that this might not be appropriate, and that filtering for members that exhibit more realistic variability (“in phase with observed”) could be an adopted strategy. Do the authors anticipate that selecting for members would introduce bias into the interpretation of projection uncertainty due to internal climate variability?  Is it possible to make runs, or use the runs already completed, to answer this question?  (See further questions/comments on this below.)

• In the conclusion paragraph, there are four listed recommendations for ice sheet model projections. While these are all pertinent discussion points, it is not clear to me why they are included as conclusions of the presented study.  More specifically, these four statements - though they may be valid suggestions - are not directly justified by the results shown.  It may be that the authors believe that they are, and in this case, please rework this section, so that is clear to the reader how the results map to each of these statements; or perhaps additional figures that better illustrate the connection can be included in the manuscript revision.

Specific comments/questions -

Lines 12-14:  Please rephrase this sentence.  It is awkward and unclear.  Also please specify the type of convection you refer to.

Line 15: Please rephrase to something like: “We recommend based on our results that ice sheet model projections consider …”  or something similar.

Line 54: “than” -> “rather than”

Lines 58-64: It is difficult to read this list organized in the current configuration.  Is there a way to simplify this so it would be easier to digest for a reader, like in a table for instance?

Line 66:  Please clarify what is meant by “best” here?  Can this be quantified?

Line 71: “have some kind of”, this wording is very informal and difficult to understand.  Does it mean that there is a tuning included for the historical?  Please articulate this more clearly for the reader.

Line 74:  Please expand upon this in your text, including a summary of the period and in what way they compare well to ERA5.

Line 76: Please summarize how these are the best, or add more quantitative language, i.e. the best with reference to what? 

Figure 1:  This figure might be better suited for a supplement, since contains more supportive information, based analysis of the climate model runs.

Line 82:  Please include a reference for the friction law.

Line 85: Please clarify in the text what is meant by curvatures here?

Line 104: Please explain in the text more about how this is done.  Are there numerical techniques used (inverted?), or is there a procedure designed determine the right correction?

Line 107: “correct” -> pleased rephrase this, as this term is not appropriate to describe model results, and if I am reading the rates right, the WAIS trend is still technically outside of the error bounds.  

Lines 109-119:  Awkward – please rephrase this last sentence of the paragraph.

Lines 112-113:  Please specify that this statement is for an Antarctic Ice Sheet model CMIP simulation.

Line 114: This phrasing is confusing for a reader because the sentence before already implies that you can add them together.  I think the point is that we can attribute dynamic ice loss to ocean-forced changes, because the SMB driven dynamics is trivial.  Please rephrase.

Line 120: Please give more specifics on how these ensembles were chosen.  Even though “see next section” is included here as a reference, it is unclear where in the next section this information is included.  If so, please note the specific section number for clarity.

Line 123: If this statement refers to both types of forcing (atmospheric and ocean), please specify that here, as the sentence is currently vague.

Lines 129-135: I do not think this detailed explanation is really needed here.  A sentence explaining that MAR output was not available would likely suffice for justification.

Line 173: Awkward sentence, please rephrase.

Line 184: “instead” -> something like “as well as” or “in addition to”

Line 187: Since one of the conclusions of this paper is related to this point, it would be beneficial to add the not shown plots to the manuscript, instead of just describing them (i.e. for 60 years). Perhaps in a supplement, so that the point can better be made that 60 would be an improvement over 20 years.

Line 197: Please note here that this therefore infers more precipitation. 

Line 198: Please rephrase, e.g., “We therefore focus on variability in SMB components such as precipitation and air temperature.”

Line 202: consistently -> consistent

Line 203: Awkward, please rephrase this sentence.

Line 216: Please make these statements more specific, e.g., … explained by “grounding line change and dynamic response” of these glaciers.

Lines 226-227: relative -> related?  (both instances in the sentence)

Line 252: although -> even though

Lines 265-267: The wording here is confusing, please rephrase.

Line 269: It would be helpful to also note in the text that for the 20 years in question, the Antarctic Ice Sheet of reality appeared to be in a stable state, as opposed to its state under the random phase of a climate model you mention that would occur during these same “years” in a climate model.

Line 270: Please clarify here what is meant here by high variability of 20-year means.

Line 279: “applying” -> “to apply”

Line 290: The previous paragraphs justify why this would make sense to do, but please add some discussion about how one would go from that previous logic to each of the outlined decisions here.  More specifically, based on this list of assessments, as a reader I can gather that one should choose the members that have the best representation of important modes of variability known to affect the ocean and atmosphere in/around Antarctica.  Please walk the reader through why these key internal variability metrics are chosen, i.e. what it is in the results that can point to each of these as a justified criterion for ensemble members.

Line 344:  Please remind the reader here that this is only for the IPSL members.

Line 354: Would the results (projections and spread) change if one were to follow the newly created rankings and chose a different group of members for the analysis?  I realize it might not be possible to do new runs to completely answer this question, but are there subsets of runs that are already completed that can be used (i.e. redo analysis of runs with top half vs. bottom half of members?).  Or even still, in your discussion, is it possible to use the results presented and extrapolate the results to suggest how much it would matter to the projections to use the ranked members instead? 

Line 367: Please specify what is meant by convection here, for example where, and at what depth?

Line 368: Please rephrase, i.e. changing -> “using the following practices to remove the effects of internal climate variability on projection results…”

Lines 369-372: Here, metrics to find ensemble members that are in phase with observed variability are described.  Please comment on the consequences of filtering these members.  In a way, this could defeat the purpose of using a large ensemble, because it discounts variability that is intrinsic within a climate model (even if it is a variability that is not realistic or wrong).  Is doing a filtering justified because results suggest that the ocean models are so wrong (i.e. the various members are all over the place in terms of variability and not at all "realistic"), that they no longer offer trustworthy projections?  Is it important to rank climate models themselves based on metrics of variability we can derive from their ensemble, in order to tell the community which ones might have completely unrealistic internal variability? Adding these concepts to the discussion would broaden and enhance the scope and impact of the presented work.

Line 374: Please add some text on why this is so, i.e., add a “because …”.  It would be helpful to clearly state your reasoning here.  For example, 1) climate models show important modes longer than 20 years, and/or 2) the 20 years in an observational period that may seem appropriate in “reality” may not align with an appropriate period within a climate model.

Line 377: “but not in the observations”, following this, please add something like “and therefore introduce a different source of bias.”

Line 382: “Ice-sheet models should” -> something like “To capture the full uncertainty due to internal climate variability, ice sheet models would ideally be …”

Please see the supplement.