General comments

In the paper the authors present results from simulations of a coupled climate-ice sheet model exploring the interaction between ice sheets and the ocean during glacial times. In particular they apply different scaling factors to the basal melt rate and investigate the resulting ice sheet model response. The coupled climate model further allows them to explore the subsequent impact of the ice sheet meltwater input on the ocean state and its feedback on ice sheets. They show that scaling the basal melt rate generally results in a decrease in ice sheet volume, which is more pronounced for the Fennoscandian and Greenland/Iceland ice sheets than for the Laurentide ice sheet. The freshwater input results in an AMOC weakening, sea ice expansion and cooling and provides an negative feedback on ice sheet retreat.

The paper is generally well written and clearly organized. However, I have some comments that should be addressed before the paper is suitable for publication.

Major comments

I’m missing some more guidance in the interpretation of the experimental setup and how the experiments are related to the real world. What are the simulations trying to represent? Is it trying to explain the climate-ice sheet response to something like DO or Heinrich events? What is the increase in basal melt intended to represent? Subsurface warming during DO stadials? Why is basal melt increased instantaneously?

A map showing more clearly the ice shelf thickness in the control run would be useful to interpret the results. Similarly, the 2D field of basal melt rate would provide useful information the spatial distribution of reference melt rate. Both ice shelf thickness and basal melt could be conveniently integrated as additional panels into Fig. 1.

Why is the freshwater flux increase after the application of the basal melt scaling so large initially and decreases so quickly? Is the shelf ice generally so thin that it melts away in a few years? Or is the calving process responsible for this behaviour?

To what extent are the results expected to depend on the ice shelf area in the control run? And what determines how far the floating ice extends into the ocean in the model? Is it mainly controlled by calving? This should be discussed, as the initial ice shelf area is probably relevant for the strength of the response of the model to the applied basal melt perturbation.

The rationale behind the basal melt scaling approach is not very clear to me. If the intention is to investigate the response to subsurface warming, from a physical point of view it would make more sense to apply different temperature offsets instead and then derive basal melt from eq. (1). I guess that results would be quite different if a uniform temperature increase would be applied instead of the melt scaling. The main reason why the melt scaling approach is not very appropriate is because where basal melt is very small in the control run, due to e.g. low water temperatures, even a scaling by a factor 100 or more will not necessarily result in a substantial increase of melt. In other words, a given scaling factor will result in a very different change in basal melt for different water conditions. This probably explains much of the differences seen in the response of the different ice shelfs to the perturbation experiments presented in the paper. This is acknowledged in section 4.1, but it remains unclear why the authors opted for the scaling approach in the first place.

Some more details on the freshwater balance of the climate-ice sheet system are needed. Some additional information is provided in the SI, but I don’t quite understand how the freshwater flux from the ice sheet can be ignored without resulting in a drift in ocean salinity. Water is evaporating from the ocean, will fall as precipitation over the ice sheets and, if ice sheets are in steady state, the same amount of water that is precipitating over the ice sheets has to reach the ocean as surface runoff, basal melt or calving in order to conserve water. If these freshwater fluxes to the ocean are ignored, this will lead to a drift in ocean volume and salinity. How is this prevented in the noFWF experiments, including the spinup?

Minor comments

L. 2: Please clarify that the forcing that is applied is actually not ‘subsurface warming’ but scaling of basal melt, which can produce very different results.

L. 5: ‘destabilisation’ of what?

L. 26-27: should be kyr BP and not years BP

L. 53: again, this can be misleading and doesn’t reflect what is done in the experiments

L. 72-73: Does this imply that ice shelves can not grow back after their thickness falls below 250 m?

L. 82: Does the model resolve the diurnal cycle?

L. 86-87: And what about the latent heat associated with basal melt? Is that not accounted for?

L. 93: Is the depth dependence of the freezing temperature not considered? As mentioned also in the introduction, this dependence can be important.

L. 96: What is the target of the calibration procedure?

L. 101-102: How is that done in the free-surface ocean model?

L. 106: ‘relatively stable conditions regarding June insolation at 65°N.’. What does that mean in the context of constant 40 ka orbital configuration?

ky, k.y.,  k.y. B.P., ky B.P. are used interchangeably, causing some confusion. Please consistently use the same throughout the paper.

L. 107-108: What does ‘The ice sheet model is forced with sea level reconstruction’ practically mean? Does sea level affect the bathymetry and land-sea mask? But the bathymetry in the climate model is set to LGM, so how are the two approaches combined?

L. 107-108: Since constant values are used for GHGs and sea level, it would be clearer to state those values in the text.

L. 125-127: This needs some clarification. Does that mean that during the spinup phase the freshwater fluxes (runoff, basal melt and calving) from the ice sheet are ignored? I understand that the total ice volume change is small at equilibrium, but that doesn’t imply that the freshwater fluxes are zero, just that total ice accumulation and ablation balance each other. I would think that it is important that the ocean ‘sees’ the freshwater fluxes from the ice sheets in the equilibrium initial model state.

L. 141-142: Not fully clear what is meant here. I don’t think that a new equilibrium can be reached in only 500 years.

L. 177-178: Would be interesting to see the separate contributions of basal melt and calving. This should also be discussed in some more detailed in relation with the calving law that is applied in the ice sheet model, which I guess could result in thick shelf ice being cut instantaneously if basal melt lowers the ice thickness below 250 m. How realistic is that?

L. 178: The equivalent number in Sv units would be useful, as that is more commonly used when it comes to freshwater forcing of the ocean. Moreover, this number should depend on the magnitude of the basal melt scaling factor, no?

L. 204-206: This sounds very speculative. Is it not more likely that the warming is a result of the transition from perennial ice shelf to seasonal sea ice cover?

L. 260-262: Do you really account for the difference in temperature of the meltwater relative to the ocean water temperature? Or is the described cooling due to the latent heat extracted from the ocean to melt the ice?

L. 334: Why ‘contradiction’?

Fig. 6: I guess that 5 m depth is equivalent to the surface? Or how thick is the top ocean model layer? Panels g-i seem to show depth of convection rather than density, as stated in the caption.

Specific comments

L. 48: increases -> increase

L. 284: others -> other

The study uses coupled-climate ice sheet modelling to assess the interactions between ocean subsurface warming and ice sheet dynamics during the last glacial period (Northern Hemisphere ice), with a particular focus on basal melt rates and freshwater release. The authors describe the impact of the basal melt of coastal ice on ice sheet geometry, and explain how the response of regional sea-ice to freshwater reduces local oceanic convection, with consequences for AMOC. The overall effect of the freshwater, is a negative feedback to ice mass loss in Greenland and Eurasia, although the authors find the North American ice sheet to be stable anyway.

The language used is clear to follow as, for the most part, are the figures. I enjoyed reading it. However, there are some parts of the framing that need tightening up before publication, to properly situate the work, making sure it’s relevance/importance can be appropriately contextualised by the reader and thus built on, by future work in the community. There is a bit too much assumed contextual knowledge/perspective. I think with some improvements in the presentation of the work (as suggested more fully below), the work will add some value to the field.

In the whole framing of the study, the relation between the experiment design (and results) and the past glacial cycle is somewhat vague; it is not very clear specifically what, beyond a thought experiment for understanding this coupled-models behaviour, is being tested. The final paragraph on page 2 explains the specific aims, but this needs relating back to the real world. Discussion section 4.3 has some relation to real past events, but it is short, and vague, and comes only at this late stage in the manuscript.

In this vein, I find the final sentence in the abstract hard to understand: what is it specifically that the presented results show, and what here is new, or how precisely does the study feed into an existing body of evidence? What is meant by ‘more accurately constrained mode results’? Constrained how and by what? And how/what would more accurate model constraints help with? Which ‘past changes’ or ‘future ice sheet behaviour’ specifically are being referenced here? I’m not questing the value of the study, but the framing needs clearer precision so that I understand precisely what message(s) to take away. That ice-ocean interactions are complex are reasonably well established with a very wide body of precise literature already.

Specific points:

Introduction:

Is too short. It lacks detailed information on what is known about the processes of ice-ocean interactions, and at different scales, or what debates/known-unknowns/limitations exist within that topic. What is included is at a very large macro-level, but what are the important dynamics in grounding line-processes, and what about ice sheet cavity-scale processes in the ocean?

Methods:

Is GRISLI an appropriate tool for addressing marine ice sheet interactions?  From my background knowledge, I believe it is, but this is not clear from the manuscript. Further justification/evidence of this is needed.

Experiment design needs more justification/explanation, e.g.:

-  Initial Condition section:

(a) Why 40 ka BP? Not enough to say it is relatively stable and precedes HS4 when no background is given on why this is important. There are other stable periods. What is HS4? Is LGM bathymetry appropriate? What is meant by ‘sea level forcing’?

(b) Figure 1: shades of gray are hard to see

(c) Line 117: ‘corresponds rather well to the extent of the reconstructed ice sheet.’ Not in my opinion, not for North American or Eurasia. Greenland also looks too large. I see big differences, and your total ice volume is (at most) almost twice as large as the reconstructed estimate. A more thorough and transparent assessment of agreement is needed here, AND a thorough and transparent discussion of what the reconstruction actually is – how certain/uncertain is that (where, when and why).

- Perturbation experiment:

(a) why did you drop the acceleration and move to 1:1 run speed for both models?

(b) What does the applied range in basal melt amplification factor correspond to in a non-technical (model) sense, i.e. in the real world? How was the range determined? How is it justified?

(c) Are there implications for applying the subsurface warming perturbation as a basal melt amplification factor for the response of the ocean to the resultant freshwater flux? i.e. the thermal profile of the ocean does not have subsurface warming, initially, so what does this mean for the ocean structure/stability (in terms of buoyancy profile and perturbation), both in terms of the initial ocean condition (including convection) and the response of the ocean to ice sheet freshwater fluxes? This also needs revisiting in the results/discussion.

Results:

- How well are the ice streams depicted in the model? This may also be a question about model ability, or resolution. The ice velocity maps show a wide splurge of acceleration in ice velocity, but I’m struggling to pick out many streams (with the few exceptions mentioned at the bottom of page 6). Please assess and comment (critically) on this in the revised manuscript.

- page 8 (around line 175) – yes, but we come back to the question of what ‘X >= 100’ actually means in real-world terms. What does this equate to in terms of subsurface warming/ocean structure change to drive the ice sheet change? This needs relating back to real conditions, rather than simply technical model components, to offer genuine insight for understanding the ocean-ice interaction.

- Line 181: ice sheet stabilizing mechanisms such as…? Please diagnose and include these, and then relate them to the real world.

- Figure 5: sparks the question, what should the conditions be for these depicted metrics (based on the real, past world)? Needs showing/discussing as part of analysis. Also, can difference in timings/temporal evolution of the different variables plotted be highlighted and explained.

Line 204-206: can this be tested? What would you need to do to verify?

Discussion:

- Section 4.1: is your model too stable? Or is your model correct? How do you make this assessment? How might inaccuracies in your simulation affect this result? Is It limitations with dynamical ice sheet process simulation that make the model too stable, biases in simulated climate (atmosphere or ocean), uncertainty in other boundary conditions/forcings, differences in palaeogeographies, issues of resolution, or are the palaeo data simply misinterpreted or overly generalised?

- Line 302: ‘sensibly’, how and why (i.e. caused by what)?

- Line 310: doesn’t quite follow. What results in the greater anomaly? Simply that the model started with a smaller ice sheet? But that doesn’t mean it will lose the same amount, it might lose less because less ice is in a vulnerable state with respect to your forcing. Line 311 doesn’t make sense (in terms of English) – how can you compare LGM bathymetry with AMOC change? This all needs expanding, explicitly, the whole paragraph; too much is brushed over.

- Section 4.3: what about atmospheric dynamics? What are the model limitations? What is shown in the existing literature? How might your model representation affect your results? Is any of this relevant (and if not, why not)? We also come back to the comment above on model performance (point c on experiment design) – is that a relevant feature for this discussion section?

Conclusion:

I very much like this section, tight and punchy. You may want to revise/add to it following revisions in response to some of the comments above.

Grammatical points/typos:

Line 93: ‘Were’ -> ‘Where’

Line 106: ‘present’ -> ‘presents’

Line 124 ‘experiment’ -> ‘experiments’

Line 144: ‘Eurasian ice sheet …’ or ‘Eurasia and Greenland…regain ice mass’.

Line 144: ‘North American ice volume’

Line 153: ‘Upstream from the’

Line 163: ‘thickness decrease’ -> ‘thinning’?

Figure 6: update units on panels to match caption

Line 202: ‘subsists’ -> not sure this is the correct term to use here.

Line 213: remove first three ‘the’s (leave the fourth)

Line 229: ‘allows us to highlight’

Line 252: ‘there is less loss’ or ‘there are fewer losses’

Line 273: ‘pointed out’

Line 292: needs rewording to improve the English

Line 298: ‘terminating’

Line 314: ‘at the beginning’

Figure 10: Is there a problem with the rendering of the LGM line? I can’t see it before ~250 yrs in the plot.