General comments

This paper presents an ensemble of simulations of ice flow in Antarctica over a scale of several centuries. The ensemble members are obtained by varying two parameters, one controlling the sensitivity of sub-ice shelf melt to ocean thermal forcing, the other relating to basal friction near the glacier grounding line. Overall I find it to be a good exercise in determining the sensitivity of the types of models currently in common use to unknown parameters and I recommend publication with minor revisions. I have a few suggestions but again I think the paper achieves the goal the authors had in mind and makes a valuable contribution.

My biggest concern is with the low mass loss from the Amundsen sector that arose as a consequence of the thermal forcing correction. The separate experiment just for the Amundsen sector without the thermal forcing correction felt a little like an ad hoc way to make up for the deficiences of the spin-up process. It doesn't detract from the conclusions of the paper as an exercise in probing the sensitivity of the system to its parameters around a particular reference state, with the understanding that this reference state is not identical to the modern. Stating this shortcoming more explicitly in the conclusion would help readers who aren't modeling experts from interpreting more than what these interesting results actually say.

Finally, the code repository on github consists of the inputs and outputs as a bunch of NetCDF files, but it could be helpful to have some more code and scripts to aid in reproducing the workflow itself.

Specific comments

132: The formulation for the effective pressure in equation 4 with p close to 1 is used often in the literature. It assumes that the ocean is the primary determiner of subglacial hydrology, and one of its implications is that there's no basal water when the glacier bed is above sea level. Now it would be crazy talk for me to suggest that you to run all this with a subglacial hydrology model too, but we know that those assumptions are incorrect -- there's appreciable basal water far upstream of the Siple Coast ice streams that has nothing to do with seawater infiltration. I think it should be said somewhere that this is a bit of a necessary hack that we've all kind of accepted for the time being but that the principled thing would be to use either a hydrology model or some other independent means of specifying the effective pressure.

155: The fact that CISM achieved similar results using a 4km resolution as other ice flow models in these intercomparison experiments doesn't necessarily guarantee that this is an adequate resolution for this particular experiment. Can you provide some other assurances that this resolution is adequate? Standard practice in numerical PDE would be to run the simulation with, say, degree-1 and degree-2 finite elements and checking where the two disagree. It's well-known that grounding line evolution is sensitive to resolution near the grounding line, see e.g. Goldberg et al (2009), Grounding line movement and ice shelf buttressing in marine ice sheets.

Technical corrections

64: This is annoyingly nitpicky but it's a pet peeve of mine when people say "modulate" unless they're referring to frequency or amplitude modulation of a periodic signal. The word I'd use here is "control" or "determine" because the γ parameter directly controls the melt rate.

115: "non-local and non-local" I'm guessing this should be "local and non-local"?

Figure 4: It looks like the grounded ice mass is sampled coarser in time near the start of the simulation than the total ice mass; is this a mistake?

Berdahl et al. presents an ensemble of ice sheet simulations, exploring centennial-scale impacts of model initialisation, focusing on two parameters that impact ice sheet sensitivity to climate forcing: a scaling factor for the basal ice shelf melt rate (ϒ₀) and the effective pressure near the grounding line (p). They then run forward experiments with climate forcing from 13 CMIP6 models under a high emissions SSP585 scenario. They run additional experiments focused on the Amundsen Sea Embayment (ASE), which show ocean thermal forcing anomalies above 1.5°C increase the likelihood of ASE mass loss. This is well-written paper with a comprehensive analysis that should be of interest for The Cryosphere, but I have a few comments that I hope the authors can address.

General comments:

Spin-up procedure:  In the 10,000 year spin-up runs, what type of climate forcing is used (e.g. paleoclimate, modern)? Figures 1 and 2 show good agreement with modern ice thickness and velocity observations, but it is not clear if the models produce reasonable mass loss under a “historical” climate, and details of the control run are lacking. How do SMB, BMB and calving rates compare to observations? The authors acknowledge that a steady-state assumption might not be valid and perform additional sensitivity experiments targeting mass loss in the ASE, but I also wish this was explained more clearly in Section 2.2.

Fixed calving front: It is my understanding that the calving front is held fixed in its current location for the forward simulations. This is unrealistic and I’m wondering to what extent it impacts the study findings. For example, a retreat of the calving front would reduce ice shelf area / buttressing, and could thereby reduce sensitivity to ϒ₀, and increase sensitivity to p. One option to address this would be to use the approach of ISMIP6 and include some ice shelf collapse experiments.

Glacio-isostatic adjustment: The authors do not explicitly state if the model includes glacio-isostatic adjustment. Larour et al. (2019) show that the elastic response of the underlying lithosphere is important in ice sheet projections of the ASE, reducing the sea level contribution by 20-40% over 250 years. Given the study’s focus on this region and that the simulations run for 500 years, the authors should consider this effect. Incorporating the solid earth response in the models would likely decrease the overall spread by limiting grounding line retreat and dynamic mass loss of the high melt/low friction simulations, and also increase the TF anomaly needed to trigger mass loss. At the very least, the authors should discuss this as a limitation of the study.

Specific comments:

Line 70: “size of the region” is not clear to me.

Line 74: Instead of “baked in sensitivities”, maybe “committed responses” is more appropriate.ϒ

Fig 1: Could include the relative error

Fig 8: There is a clear outlier with the control forcing (i.e. low ϒ₀, mid-range p). Can the authors explain this?

Line 337: See above general comment on spin-up procedure. I think this should be included in the methodology section 2.2.

Line 417: With the caveat that solid earth feedbacks would slow Thwaites retreat / collapse.

Fig A9: This is a useful figure. I suggest adding this to the main text.

Please find my comments in the attachment.