The revised manuscript articulated reviewer’s concerns adequately in most places. The scope of the paper is narrowed so that the arguments in the revised form is more justifiable in terms of the methods used here. Below, I list some additional issues that should be addressed before the publication, and two reviewers also suggested some changes. Therefore, I request a minor revision and will review the next manuscript by myself for the final judgement. In the next round, please make sure that the marked-up manuscript includes correct reference information (they are mostly shown with ?? in the marked-up manuscript due to a Tex error) and submit a response letter.

1. Include radargram showing 18 radar reflectors used for the modeling and the bed. The current manuscript doesn’t include any radar data/image, so readers cannot really evaluate the quality of the data. Radar profiles A-A’ and B-B’ are probably appropriate for this purpose, as A-A’ goes close to the drilling site (label radar reflectors with their ages if possible), and both profiles include large accumulation variations (see my technical comment). This point is brought by the third reviewer too.

2. Depth uncertainties are listed in Table 1 but its basis is not well presented. Depth uncertainty relative to their depths decreases in terms of depth; it is about 0.6% for the isochrone #1 at ~310 m, and 0.28% for the isochrones #4 at 1080 m. Please describe (1) radio-wave propagation speed (and firn corrections) used in this work, and (2) source of the depth uncertainty. Separately (3) add the reference for age and age uncertainty in the figure caption. I understand that the full information is given in Cavitte et al. (2016, JGlac) but this manuscript should include key information for completeness.

3. There are numerous cross-over points of the radar profiles, and I assume that radar reflectors along crossing profiles do not match to each other perfectly. I guess that the authors used the 18 reflectors that are distinct from the others (so little chance to see incorrect correspondence), and used many closed loops to minimize the uncertainty of reflector correspondence. Please articulate this analysis and (if possible) show depth uncertainty associated with the reflector correspondence in Section 2.2. Cavitte et al. (2016 JGlac) analyzed the data in a larger area but cross-over analysis of this dense grid is not presented in that work.

4. Regarding reviewer #1’s comment, I also struggled to read Section 2.3. Please carefully revise this section and make sure that the necessary (but only necessary) model setup is presented more clearly so that readers can understand how the model is used to interpret the data and what’s the limitation of this modeling procedure. I agree with this reviewer that the point presented in Fig. 6 should be better articulated. First, I can see a good agreement (visually in the map) between the curvature and the anomalous accumulation only in limited areas, so please specify the geographical area where you want to argue the good agreement. Second, I think that the main argument associated with this figure is that surface curvature affects snow accumulation in this scale and the surface curvature is largely related to the bed topography. The secondary argument is that, because the surface topography (in small scales) is affected by the bed topography, this snow accumulation pattern didn’t change in the past. Is my understanding correct? Please present these two points more clearly and more logically. If possible, conduct statistical analysis as suggested by the reviewer.

5. Regarding reviewer #3’s comment, I agree with this reviewer that the manuscript should include radargram(s). See my own suggestion #1 above.

Technical/editorial suggestions

P5L139: change to “R(t) is obtained from accumulation variations inferred from the AICC2012 chronology (REF).”

P5L144: Fig. 3 is cited before Fig. 2 is cited. Please update the figure numbers (and include radargram as requested above).

P7L215ff: State the role of this dataset in this work at the beginning of this section. The authors say that ECMWF is “independent dataset” but more important is that it shows the present-day precipitation, whereas the radar data show past surface mass balance. So, these are independent but not suitable for apple-to-apple comparison. I think that the author’s main point is that the SMB averaged in the past century is similar to that of the present day.

P9L261-265 and Fig. 2: I am not sure whether accumulation is more variable along A-A’ than B-B’. The youngest layers show 3 mm_weq/a difference over 150 km along A-A’. My reading from Fig. 2 shows that the difference is larger, 3-5 mm_weq/a over 100 km, along B-B’. Please consider scaling these two panels so the apparent gradients of the curves can show spatial gradients of the accumulation clearly. If the authors don’t want to make this change, show the value of the gradients in the text to clarify the argument.

P11L339ff: Please consider shortening the discussion. Issues are repeatedly discussed in a slightly different context so these paragraphs can be shortened to show the essence of the discussion in a more efficient way. If the authors don’t agree, please consider developing subsections under 4. Discussion.

P13L420: LDCM, not LCDM.

P18 Fig. 1: Add the second labels showing the latitudes. I can see only one label of 76S and one label of 74S (longitudes are shown with a pair of labels).

P19 Fig. 2: See my comment on scaling. Also, the colorbar is not necessary. It is probably better if all curves are labeled with their ages.

P20 Fig. 3 caption: Show the period when the time-averaged accumulation rates are calculated. Also, what’s the dashed rectangle (coverage of the new bed topography data?).

P12 Fig. 4: please consider using the same colorbar for all panels, or at least assign the colorbar to the same magnitude of the span (i.e. the full range of the colorbar show 10 mm_we /a variations). Or show the relative variations to the regional mean of the corresponding period in percent and show the regional mean value in each panel. I assume that the temporal variation of the spatial pattern is the main point that the authors want to present in this figure. In any case, please make sure that the key point is clearly shown (i.e. pattern, magnitudes, or what)?

P22 Fig. 5 caption: Is a_100yrs a typo (1000 years instead of 100 years)? Anyway, it cannot be said “Holocene average”.

P23 Fig. 6: change the first colorbar title to “Anomalous accumulation” or such. Also the caption can start with “Spatial anomaly of paleoaccumulations ….”

Supplement: Clarify that all figures in the supplement have names such as S1, S2, …

Supplement P6: change the colorbar title to “accumulation uncertainty”. Why don’t you show this relative to the accumulation, i.e. percentage?

Dear authors,

I am happy to accept this manuscript as it is. You responded to the comments satisfactory and the revised manuscript meets the journal criteria.

Congratulations.

Kenny Matsuoka
TC/TCD Editor