This is a well-written and interesting manuscript comparing long-term ice cover histories derived from multiple cosmogenic nuclides measured in slowly eroding Antarctic bedrock surfaces with two ice-sheet model parameterizations. The manuscript suggests a novel way of testing these ice-sheet model parameterizations by use of empirical data. The paper is of wide interest and convincingly written. My comments below mainly contain suggestions to improve/clarify the figures and figure captions.

Figures

Fig. 2:

• The box in panel (a) is not explained in caption, I assume it is outlining the extend of boxes in (d) and (e), but if so, the blue curve in (d) is lacking an initial and partially a final ‘plateau’ (horizontal part) similar to the red curve in (e). Even though it is a conceptual figure, I think this way of exaggerating the difference is misleading.
• Are you sure the blue curve in panel (b) is reflecting the blue curve in panel (a) correctly? Comparing the curves in panel (a), the blue and red curve shapes are similar – the blue curve has lower amplitude, but still quite steep slope around zero crossings. I would therefore expect the blue curve in (b) to be bimodal, but with less sharp peaks located closer to zero compared to the red line (approximately +- 0.25).
• Scalebar for panel (b) is missing a number (-0.5) or a white box is partially covering it.
• Avoid rainbow color palette for panels (f) and (g), when the figure is printed in grey both ends of the scale has the same color.

Fig. 3:

• y-labels for panels (b) and (c) missing
• Note that the histograms in panel (b) reflect a composite of two quite different responses before and after the Quaternary. I think this makes the blue/desensitized curve look ‘artificially’ unimodal because of the remarkable stability prior to the Quaternary, while the Quaternary period shows a bimodality very similar to the red/sensitized scenario but with a smaller amplitude.
• Panels (d) and (e): why did you pick these two time slices that both predate the max time (5 Ma) shown in panel a? Are these (7.68 Ma, 9.42 Ma) even within the modelled time range?
• Check formatting of ‘d18O’ in caption

Fig. 4:

• Panel (a): you state in the caption that “The sensitized model (red) displays larger variation in ice thickness and is more likely to occupy extreme values, whereas the desensitized model (blue) is more likely to occupy intermediate values". Too me it looks like the truth of this statement depend on what period you look at. For the last million year for example (most important for cosmogenic nuclides) both curves appear to occupy extreme values most of the time. Between 1-3 Ma, the blue curve seems to stabilize at an intermediate thick stage before thinning again in most glacial cycles, but the transition between stages still appears to be rapid. I wonder how sensitive the distinction between curves in (b) and (c) is to the choice of timescale.
• Caption: you refer to grey shading in (b) which is not there.
• You refer to dashed black line in (b), which is really shown in (c). You appear to have two definitions for this line, see two last sentences. It is not obvious to me what you mean by "modern ice thickness… is chosen to approximately align the range of ice thickness in the model simulation with that inferred from geologic evidence" (last sentence). How can you choose a modern ice thickness?

Fig. 5:

• I expected a brief explanation in the caption of the difference between the top and bottom scenario. I can see that D varies between the two but consider spelling out why these two examples have been chosen.

Fig. 7:

• The green dots represent sites with >1 Ma histories, is that also the case for the blue dots?
• Label for colorbar is missing

Fig. 8

• What nuclides are these apparent ages calculated from? Apparent ages depend in part on the half-life of the measured nuclide (e.g., 14C vs 21Ne in the same sample could yield wildly different app. ages). Are there any patterns in what nuclides are measured at different elevations/distances from coast? I would guess not but this may be worth addressing if you are mixing ages derived from nuclides with different half-lives.

Figs. 9-13

• Would it be worth explaining the double y-axes and specifying what ‘h’ is?
• The legends vary between the different figures – I suggest you add nuclide name to all (missing in fig. 9 and 12).
• You state in the paper that “26Al data can provide no information about events prior to ~3 Ma”, so why did you choose to show 26Al/10Be and 26Al/21Ne ratios in the left panels (5 Ma to present)?

Fig. 14

• There is a discrepancy between white and green dots in legend and caption. Legend has white dots as ‘Any age >1 Ma’ whereas caption describes white dots as ‘long-exposed bedrock surfaces are not likely to exist’. Conversely green dots are ‘All ages < 50 ka’ in figure legend, but caption has them as ‘long-exposed bedrock surfaces are likely to exist’ and ‘ages >1 Ma have been observed’.
• I find it hard to distinguish the white dots on the inset panels, can they be made bigger?

Manuscript

• l. 52 and 55: Since subglacial data has been gathered in recent years, perhaps state that this type of data is (yet) too sparse rather than saying that ‘it is not possible at the moment’ and ‘In contrast to subglacial basins, it is possible…’.
• l. 121: patterns in plural
• l. 213-217: The described difference between the red and blue curve is not representative for the last 1 Ma. For this period, the blue curve also appears to switch abruptly between states, although the endmembers are closer together. As cosmogenic nuclides are increasingly sensitive to the most recent period due to decay (and erosion, although not considered here), I think you should comment on why that is the case and whether it has an impact on your interpretations.
• l. 240-241: You state that 'the desensitized model is more likely to occupy intermediate values near 1200 m’', however, it looks to me like the blue curve spends relatively little time near 1200 m within the last 1-2 Ma.
• l. 313: should this only refer to Lower Beardmore (Fig. 7i) since Upper Beardmore fail this criterion?
• l. 370: Would it be worth also mentioning that the exposed bedrock would need to contain minerals where production rates are well-calibrated for nuclides with half-lives that cover the relevant timescales?
• l. 415: Do you need a reference for the ICE-D:Antarctica database according to journal guidelines?
• l. 425-430: consider citing data references in this and the following sections, I see them in the figure captions, but not in the main text.
• l. 466: specify 40-km resolution model
• l. 498: change ‘there do exist rock outcrops’ to ‘rock outcrops do exist’
• The approach in this paper regarding constraining long-term ice-sheet cover based on an elevation transects of cosmogenic nuclides seems comparable to the one in “Jones, R. S., Norton, K. P., Mackintosh, A. N., Anderson, J. T. H., Kubik, P., Vockenhuber, C., ... & McKay, R. (2017). Cosmogenic nuclides constrain surface fluctuations of an East Antarctic outlet glacier since the Pliocene. Earth and Planetary Science Letters, 480, 75-86.” Would it be worth a citation?
• Consider spelling out MPWP since you only use the abbreviation a few times.

Comment attached as a supplement in PDF format.