This manuscript presents a reconstruction of the accumulation rate at site S27 in the Allan Hills Blue Ice Area in Antarctica during the last interglacial period. The method is based on the inference of the ice-air age difference (Delta-age) in the ice samples, which is reconstructed by an ice-air cross synchronization onto the EDC ice core. It is found that Delta-age almost vanishes at ~128 ka BP during the peak of the interglacial. By using an Herron-Langway firn model in an inverse mode, the authors infer the accumulation rate and found that it should have been very high compared to background level, with an increase of more than an order of magnitude. Then they discuss some possible explanations of this increased accumulation rate, due to local changes in sea ice, the Ross ice shelf and the West Antarctic Ice Sheet.

This is an interesting manuscript and I enjoyed reading it. The subject is an important one. The manuscript is well written.

The main comment I have is regarding the treatment of the uncertainty of the EDC chronology. In the supplement, it is written: "The Δage uncertainty in the EDC timescale is not reported explicitly. Instead, the reported uncertainty in the final AICC2012 chronology for each depth is the maximum of the gas and ice uncertainty (Bazin et al, 2013). Because there is no feasible way to evaluate the EDC Δage uncertainty, our focus here becomes the relative errors of S27 to the EDC chronologies." AICC2012 was built using the probabilistic model Datice. I then developed a new tool with the same basic principles, called IceChrono, and tested it with the AICC2012 dating experiment (Parrenin et al., GMD, 2015). IceChrono outputs the Delta-age uncertainty, so I can provide that easily. Better than that, IceChrono could provide the relative uncertainty of a pair of gas depth / ice depth (with different depths). This would require a little extra work because it is not something which is currently outputed, but this should not be too difficult. I think it would strengthen the manuscript to account for the true AICC2012 uncertainty.

Other minor comments:

- l. 25 : "the peak in S27..."

- l. 428-429 : Are you sure 3 ka is enough to re-form the WAIS and/or Ross ice shelf? This could be discussed.

General comments:

Yan et al. present a reconstruction of the accumulation rate for the S27 ice core in the Allan Hills Blue Ice Area over the Last Interglacial and investigate the moisture evolution and implication on West Antarctic Ice Sheet during this time period. New gas measurements in S27 ice core are presented and combined with previous data from Spaulding et al. (2013). A refined gas chronology has been developped using δ¹⁸O_atm synchronization with the EDC record and improved using greenhouse gases concentration measurements. Together with the gas and ice ages, the authors used firn densification inverse modeling to estimate the accumulation rate.

Based on this record indicating exceptionally high snow accumulation over the early LIG, the authors proposed several hypotheses to explain this peculiar event as well as associated consequences on the stability of the WAIS. The paper is well-written and easy to read. This study is interesting and worth to be published in Climate of the Past. However, I do have comments which I hope can help improve the manuscript.

The authors made the hypothesis that gas loss affects the δ¹⁸O_atm in the same proportion in both fractured and non-fractured ice. The gas loss correction uses then the same values over the entire record. It is however not clear if such a difference exists or not in the record. For example in the Figure S4 the Δδ¹⁸O_grav vs Δ δO₂/N₂ relationship is shown but we don’t know if the data are for fractured, non-fractured ice or both. The slope could then maybe be differenciated in two differents slopes if the two zones are represented and if a difference between them is observed. In the results part, differences between EDC and S27 δ¹⁸O_atm are observed for samples older than 145 ka (and especially between 202-210 ka) and the core quality is proposed as explanation. This is supported by the younger ice age than the gas age at similar depth interval suggesting large gas loss. Could a different slope for gas loss correction for the deeper part of the core reduce these differences?

The discussion of possible hypotheses to explain the increase in moisture during the early LIG is convicing but would deserve further comparison with proxy records used to infer changes in moisture source and shift of evaporative regions (in relation with ice-sheet extent, shift of atmosperic/oceanic pattern…) such as d-excess, T_site or T_source. Such data exist for other Antarctic ice cores. A comparison to these records could maybe support the proposed hypotheses.

I appreciated the uncertainties part in the Supplementary but I missed some values about the final uncertainty associated with the new gas chronology (even if they are given in the Supplementary Data Table). Few lines could be added to explain the final gas age uncertainty, i.e. ± x years, how it varies along the record and maybe compare it to the AICC2012 gas age uncertainties (and also refer to it in the main text). Similar comment can be done for the ice age uncertainties. The uncertainties of EDC are assigned to S27 age points during the synchronization of the ice chronology without any estimation or discussion of the range of associated uncertainties and tie-points previously used to develop the ice chronology. This could be developped a bit more.

Other comments:

Line 19: Write "during the LIG maximum".

Line 25: Remove the s in “insS27”.

Line 43: Specify that it is for both past and future simulations.

Figure 1: Add WAIS and Taylor Dome to the map.

Line 57: Remove “that of”.

Line 59: Detail the time period covered by the new record.

Line 71: Figure S1 instead of S2.

Line 72: Rephrase the sentence. I guess the missing peak in δ¹⁸O_atm is only because no measurements have been done at these depths.

Line 73: It is not clear what the δ¹⁸O_atm sampling strategy was. Improve the resolution? Complete missing intervals?

Line 75: Specify that new CO₂ and CH₄ measurements are used to improve the gas chronology between 105-147 ka.

Line 79: Change to “circulation changes and ice shelf / ice-sheet stability during the LIG”.

Line 88: Figure S2 instead of S1.

Line 102: Add to the end “to prevent contamination from exchange with ambient air”.

Lines 105 and 141: Give the temperature in °C (for consistency).

Lines 110-113: Use “δ¹⁸O of O₂”.

Line 113: Give the equation for gravitational fractionation correction.

Line 114: Remove “paleo”.

Line 145: This sentence suggests that there are also fractures in the ice above 151 m. Are they numerous? Is there an influence on the δ¹⁸O_atm?

Line 150: Specify between 115-255 ka.

Line 198: Remove the extra parenthesis for δD_ice.

Line 205: Wrong units, kg.m^-3.

Line 228: Give the value for h_diff.

Line 242: Correct “samples”.

Figure 4: Change “per mil” into “‰”. You also compare in the main text the δ¹⁸O_atm variations to orbital variations. Maybe add the insolation curve on the figure.

Figure 5: Add the tie-points and anchor points used for the chronology on the figure. In the caption, precise that CH₄ data are from EDC, the CO₂ is a composite record and the timescale is AICC2012.

Line 260: Remove “at this peak differ from by 2 ppm”.

Line 271: To conclude this part on the gas chronology I missed a sentence on the total uncertainty associated with this new chronology. How much is it?

Line 281: “Figure 2” not usefull here.

Lines 321-324: I don’t know if we can say that the accumulation rate at S27 is comparable to Vostok and EDC. The trend is similar yes but the absolute value not. And how is the 0.02 m.yr^-1 value defined?

Lines 338-339: Could you support this hypothesis using model comparison?

Line 341: Remove the values of the accumulation rate, not usefull.

Line 344: Delete “apparently”.

Lines 345-350: TALDICE’s accumulation rate starts to increase earlier than S27 (and is more similar to Vostok and EDC). As for the magnitude, it is much larger for S27 than for TALDICE. The S27 site is already pretty coastal so I would rather say that the peak in accumulation rate at 128 ka reflects more open-ocean conditions than a transition into a coastal site.

Figure 9: I would have removed the Greenland temperature record and drawn instead the variation in mean ocean temperature from Shackleton et al. (2020). It could also be good to add an insolation curve to have an orbital context to refer to in the discussion. Change “(g) Relative sea-level vs present day”.

Line 414: Remove “(~80 km)”.

Line 422: Correct “with a high sea-level stand”.

In the supplementary:

Line 16: Remove “age”.

Exchange the Figure S1 and the Figure S2 to match the order of citation in the main text.

Figure S2: Remove “paleo”

Figure S3: The δO₂/N₂ equation for depth > 148 m has to be corrected in δO₂/N₂ = -0.205*depth(m) + 24.26. In the caption, give the exact number 0.0067 for the slope.

Figure S4: It is not clear if the data presented here are only for non-fractured ice or for both non-fractured and fractured ice. Please indicate if this is non-fractured ice or differentiate the data with two regression lines for the two zones.

Figure S6: We don’t see the peak in the C.I. modelled δ¹⁵N at 128 ka. Adjust the y-axis.

In Figures S2, S3, S4, S6, change the “per mil” to “‰” for consistency with the main text.