De Rydt et al provide a framework for a modelling intercomparison exercise on ice sheet-ocean interactions in Antarctica. The intercomparison focuses on two key regions of Antarctica: the Amundsen Sea and the Weddell Sea. These two regions are representative of warm and cold ocean conditions and understanding their future evolution is key to predict sea level rise and changes in ocean circulation. The use of realistic configurations as well as the implementation of ice-sheet-ocean coupled models are important steps forward compared to previous intercomparison projects. Very important is also the development of “MIPkit” that allows comparison between model output and oceanographic/glaciological observations (from satellites and in situ measurements). I fully support this modelling exercise, especially because it will allow to look at features/processes in the models that can help explain observed changes and to test the sensitivity of different models to prescribed perturbations based on what we could expect over the next decades.

 I have one main comment on the modelling setup and a couple of minor comments.

Main comment

I understand that requiring fixed lateral boundaries is a strong request to the different groups at this stage, especially because it is important to include enough group in the intercomparison project and it is also useful to look at both regional and global models. However, I feel that by having very different lateral boundaries, this could affect interpretation of the model response to external forcing and also internal variability/feedback are likely to change. Maybe some further discussion about this can be useful.

Minor Comments 

• Figure 1: would it be possible to distinguish the location of CTD from moorings?
• Define CDW and mCDW in the manuscript.

Review of experimental design paper for MISOMIP2 by De Rydt et al

This paper documents the design for a series of model intercomparison experiments that aim to probe the realism of and agreement between the models that are used for predicting ocean-driven ice loss from the Antarctic Ice Sheet. The experiments involve simulations under realistic forcing and geometries, and perturbed (but still realistic, i.e. not idealised) forcings and geometries.

I have a very positive overall impression of both the experimental design and its writeup in this GMD paper.  The authors have thought carefully about the purpose of their experiments and have clearly specified the many details required.  One particularly nice aspect is that the observational data that will be used to test the models has been provided as part of the experimental file sharing.  I have no major reservations about the experiments or the paper.

Major comment

Given the goals of this MIP, when validating the Weddell Sea ocean model results it seems a significant missed opportunity not to use the borehole mooring and CTD observations from beneath FRIS that have been collected over the years.

Specific comments

L16: ‘with a marine termination’ seemed like a strange phrase to me.

L33: can you cite an ‘in prep’ paper?  It seems unlikely it can be citeable before the present paper is published?  So this sentence needs to stand alone, somehow.

L40: ditto

L49 and 50: I was confused by the word ‘coupled’ here, as I think this is just discussing ice-shelf—ocean simulations.

L52: I don’t know if it is worth highlighting that the melting parameterisation used in these ocean models (the 3 equations) is itself both highly tuned and structurally uncertain, so anything arising from these models will inherit those uncertainties.

L59: I would add citations to 10.5194/egusphere-2023-1587 and 10.5194/tc-2023-77 because I think those studies represent advances in the state of the art that are relevant to the setup of this MIP.

L86: ‘emission-based’ -> ‘scenario-based’

L94: Mouginot et al is about ice speed, not grounding lines

L98: reference typo

L102: It would only be a good indication of model robustness if the models were not re-tuned or structurally different between the two locations, but using a common configuration in both regions is not required in the protocol, I think.

Section 2.1: I found the naming of the experiments a bit confusing. I feel using numbers is sub-optimal because it gives no clue toward the aim of the experiment.  I would also find it easier to have a short name for each experiment. Would the authors consider dropping the numbers and renaming the experiments?  To illustrate the point with an example, it could be Ocean-*h, *w, *c, *p, *f or something? (hindcast, warm, control, present geometry, future geometry).

L135, and below, e.g. section 2.5: It would save a little work if the participants used the common geometry in their *1p runs, meaning they could skip *2p and having to spin up the model twice.  Is it worth highlighting this clearly and recommending participants do this unless they have a good reason not to?  Also, if participants do this, could the protocol recommend how they go about it, i.e., do they upload the results twice, or skip one of *1p or *2p, and if so which one, and how do they indicate this is what they are doing?

L140: I would imagine coupled models will be likely to be worse than ocean-only models over the historical period, so I would reword this sentence to refer to a ‘change in bias’.

L158: Cosgrove does not seem essential to me

L159: I feel this needs to be more specific.  Is the requirement that the models include the shelf break so that they include all of the shelf?  In that case 71S is fine.  Or is the requirement that the models include shelf break processes, i.e. undercurrent, in which case I would imagine some decent area of ocean north of the shelf break is needed – 65S?  Also, why is there this limitation in the first place?  Why can’t modelling groups just model whatever parts of the Amundsen Sea they want to?  Same comments apply to Weddell Sea domain and shelf break also.

L168: Why stop in 2022?  1979-present?

L179: What observations of sea ice fluxes are available?  My personal opinion is that this is a sufficiently different approach that it should be precluded.  Or if this is permissible, the choice of surface forcing should be completely CAYA, e.g. including manually created forcings, climatologies, ‘normal year’ forcing, climate model output, paleoclimate reconstructions, etc etc.  I would personally require that the use of a reanalysis and dynamic sea ice model is essential.

L218: Generating absolute forcings from disparate data can be risky in terms of creating unphysical fields, such as negative shortwave, shortwave in winter, relative humidities <0 or >100%, etc.  Do users need to be advised on such considerations?

L231: Naughten paper is now published.

L260: Is ‘early stages’ accurate for geometries representing 200y and 300y of strong ocean warming?

Figure 3 caption: final sentence needs correcting.

L305: Is it worth explicitly stating that all geographical restrictions on the ocean domain are now dropped?

Following on from above part 1: So this causes a bit of a disconnect between the Ocean and IceOcean experiments, in that models may contribute to the latter without having completed the former? 

Following on from above part 2: In general, are there any restrictions on which simulations the participants must perform, or dependencies between experiments? All I understood from the paper was that if participants are completing the *1vw experiments they must also perform the *1p-ext experiments, but even that was not actually explicitly stated.  If the participants can perform whatever experiments they like, should that be explicitly stated?

General: it would be good to state explicitly whether participants are permitted to upload multiple experiments, and if there is any limit to this (there must be!) or guiding philosophy about what variations might be of interest.

L369: I didn’t understand why the ice elevation change and velocities are provided.  1) Is this for model evaluation or initialisation, or what is it for?  2) This seems to go against the CAYA approach, in that participants may prefer different products from those provided. 3) Why host separate online versions of these products when they are already online elsewhere with proper DOIs, version controlled and documented by the data originators?

L372: James Clark Ross should be spelt out fully, analogous to Jan De Rydt.

L389: Mark the sections on Figure 1?

L402: Important sea ice?

L406: Any moorings in the Dotson Trough?

L502: ‘original model’ -> ‘participating model’

L514: What does ‘we prefer seeing dynamical features’ mean?

Review of "Experimental design for the marine ice sheet and ocean model intercomparison project - phase 2 (MISOMIP2)" by Jan De Rydt et al.

General comments

I thank the authors for their hard work in creating this protocol.

This manuscript describes the protocol for MISOMIP2. This is a follow-up to previous model intercomparison efforts focusing on ice sheet and ocean processes in Antarctica. The goal of MISOMIP2 is to investigate the performance of ocean-only and coupled ice-sheet-ocean models by comparing realistic simulations against observations. The Amundsen and Weddell Seas have been selected to represent warm and cold ocean conditions, respectively. The protocol welcomes a “come as you are” approach with no constraints regarding the model domain, grid resolution, physical parameters, and forcing. This exercise is crucial for testing ice-sheet-ocean models and improving our ability to forecast sea level rise and understand ice-sheet-climate system interactions. 

The manuscript is well-written and delivers the necessary information for ocean and ice-sheet modelers to conduct the proposed experiments. My only major comment is regarding possible complications due to the lack of constraints in the forcing of the experiments. For example, in the JRA55-do dataset (Tsujino et al., 2018 - DOI: 10.1016/j.ocemod.2018.07.002), both liquid (river flux of water) and frozen (land ice calving flux) runoff are prescribed around Antarctica and vary with time. In contrast, the CORE2 forcing (Large and Yeager, 2009 - DOI: 10.1007/s00382-008-0441-3) only gives an annual mean river runoff climatology but lacks values around Antarctica. Similar discrepancies may exist in other datasets (MARv3.9.3, ERAint, etc). Differences in runoff forcing can lead to significant changes in, for example, sea-ice concentration (e.g., Tsujino et al., 2020 - DOI: 10.5194/gmd-13-3643-2020), which in turn may affect the ocean and ice-sheet responses. A similar argument can be used for regional models using different datasets as open boundary conditions. Why not, at the very least, restrict the forcing to be employed in these experiments?  

Specific comments

Line 33: “Asay-Davis et al., in prep.” - it would have been preferable if the outcomes of ISOMIP+ and MISOMIP1 were published before the MISOMIP2 protocol was released. The lack of documentation of the previous MIP's results may discourage involvement in the current MIP.

Lines 214 - 216: how about perturbations in runoff?

Lines 270 - 272: The reference densities for seawater and ice might differ among models. Could this cause significant changes in the geometry of ice-shelf cavities when the ice-shelf draft and bathymetry are imposed?

Page 24, Table A1: it might be helpful to include the water mass flux due to liquid and frozen runoff as ocean variables. 

Editorial/Typos

Line 33: “Asay-Davis et al., in prep.” -  this citation is missing from the bibliography.

Line 40: “Smith et al., in prep.” -  this citation is missing from the bibliography.

Line 98: "e.g.,[" - please correct the LaTeX code.

Figure 4: please specify which panels correspond to (a) and (b) or use (left) and (right).