General comments

The manuscript presents the first experimental data on the air-hydrate geometrical properties in the EDML ice core, below the bubbles-to-hydrates transition, which were obtained by automated image analysis of microphotographs of thick sections of ice taken in the field a few days after core retrieval. The resulting data set represents a significant addition to the existing experimental data on the geometric properties of air hydrates in polar ice sheets, which until recently were obtained only from GRIP, Vostok, and Dome Fuji ice cores. Specific conditions of the EDML site on the Antarctic Plateau (relatively warm temperature of ice and enhanced accumulation, location on the ice divide) allowed authors to reveal some new peculiarities of hydrate properties such as weakening of their correlation with the climate (isotope content of ice) on small time scales, and a distinctive change in air hydrate properties associated with changes in the deformation regime of ice reflected in its microstructure. The most important achievement of this work is a carefully developed technique of using automatic analysis of high-quality images of thick sections of ice in order to obtain various geometric characteristics of air-hydrate inclusions in polar ice. This technique was, for the first time, successfully applied to quantitatively describe air hydrates in an Antarctic ice core. Its further refinement will provide a reliable tool for the study of future ice cores, including those that will be obtained in Antarctica as part of the Oldest Ice project. To summarize, I believe that this work deserves to be published in TC after the small corrections and clarifications suggested below are made by the authors.

Concerning the presentation of the material, I would like to encourage the authors to make an additional effort to improve the structure of the text in order to make it easier to read and understand.To me, the most obvious idea in this regard would be to move sections 4.1 and 4.2 of the ms from Discussion to Results, and to merge them with sections 3.1 and 3.2, respectively. Some of my specific comments below also reflect the difficulties I encountered in reading the manuscript and therefore, hopefully, may help to improve its presentation.

Specific comments

L15-17: The BHTZ depth range depends on the temperature in the ice sheet (e.g., 500-1250 m at the cold Vostok site, but 1000-1500 m at the GRIP drilling site, which is warmer). What site does the 500-1500 m depth interval refer to? I did not find in the ms clear information about the BHTZ depths at the EDML site, which is important for interpreting the bubble and hydrate data. At L249 one reads: “Note that LGM-ice at EDML coincides with the top of the BHTZ”, which is a bit misleading, since the LGM ice at EDML is buried at a depth of about 1000 m. According to Bendel et al. (2013), the BHTZ at EDML is between 700 and 1200 m and therefore LGM ice coincides with the middle of the transition zone, which disrupts the correlation of bubbles’ properties with climate. Please state clearly the BHTZ depths at EDML in the introduction. It would also be useful to give in the introduction the maximal depth (from the ice core log) of the borehole when it reached the bedrock (at 2774.15 m?).

L28-30: To the best of my knowledge, the most recent review on this topic is given in Lipenkov V. Y. How air bubbles form in polar ice, Earth’s Cryosphere, 2018, 22, 16–28.

L186-188: For hydrate inclusions with non-isometric shape and preferred orientation in space, it is also important to choose the right plane for the thick section analysis (i.e. image plane) that would allow proper assessment of inclusions’ dimensions and aspect ratios. Have you tried to solve this problem, and if so, how?

L196 “Surprisingly, no air hydrates were found in the lowermost sample at about 2774 m depth”: Any comment on this: distance of this sample from the bedrock, measured air content in it, accreted ice?

L68-69 “For this study, mosaic images located about 1.5 mm below the sample surface were analyzed…”; L179 “We convert the measured air hydrate counts per sample to Nah using the measured ice-sample thickness…”; L180-182 “Air hydrate counts per sample and Nah have a good linear correlation (Fig. 4) and we conclude that the observed volume (i.e. hydrate mapping depth; Fig. 1) is consistent for all samples”; L253-254 “… the hydrate-mapping depth in our microscopic set-up might not correspond to the physical sample thickness (Fig. 1). In other words, the microphotographs from one focus plane might not display the entire sample volume”. The construction of the narrative is such that it is not until L253-255 that the reader begins to realize that you could not have obtained Nah (in cm-3) because you did not know the thickness of the ice layer (the hydrate mapping depth) within which the hydrate counts were made. Why not say this at the very beginning, and treat the hydrate count data as a reliable but relative metric of Nah (assuming the observed volume was the same in all samples), and then scale that metric with the measured air content of ice, and use the thus derived Nah in further consideration? (Such a relative metric of Nah can be the density of hydrates’ projections (cm-2) observed over an 80 x 30 mm area of the mosaic image of the sample).

Figure 5 caption: “Error-band for b) is explained in section 2.4”. I could not find an estimate of the resulting Nah error in Section 2.4. In any case, I now suspect that this resulting error shown in Fig. 5 does not account for the difference between the sample thickness that was used to calculate Nah and the hydrate mapping depth.

L226 “Naturally, the TACh and Vc signals show the same pattern (Fig. 6b)”. Not sure that this statement matters when both TACh and Vc are shown by a single graph. Delete the sentence?

L237-241: It would be appropriate here (or elsewhere) to provide information on the inclination of the borehole from which the studied ice core was obtained.

L247-248: In Fig. 3 from Bendel et al. (2013), the average Nab value in LGM ice appears to be closer to 400 cm-3.

L250-251: Number concentration of air bubbles and that of succeeding hydrates (if we assume one-to-one conversion) depends on the ice formation conditions (accumulation rate, firn temperature and surface snow density), so the properties of the Vostok ice cannot be simply projected onto the EDML ice core. Using a simple model, as described in Lipenkov (2018), one can estimate Nab ~400 cm-3 for present-day conditions at EDML (-44.5 °C; 6.4 g/cm2 yr; 0.38 g/cm3) and Nab ~475 cm-3 in the LGM ice (assuming: -54 °C; 3 g/cm2 yr; 0.38 g/cm3). The model estimate for Holocene ice is close to Bendel’s et al. data while that for the LGM ice is slightly higher than their experimental data (because LGM ice coincides with the transition zone), and lower than the Vostok-based value used in the ms (the extremely low LGM temperature at Vostok led to a very small grain size at close-off and hence a great number of bubbles forming).

L259: It appears that the CFA system used by Ruth et al. (2004) does not provide accurate absolute values of air content (it was designed primarily to document high-resolution relative TAC variations). For that reason, the obtained average value (0.0815 cm3/g) is lower than the TAC measured with absolute methods in the Holocene ice cores from the most elevated drilling sites in Antarctica (Dome F, Vostok, EDC). For the EDML elevation (2892 masl) and atmospheric pressure (~700 mb) we can expect the TAC to be slightly above 0.09 cm3/g (Martinerie et al., 1992) in Holocene ice and even higher in LGM ice. This prediction has been confirmed by the measurements made with an absolute barometric method in LGGE/IGE, Grenoble, which gave TAC=0.0906±0.0025 cm3/g in Holocene ice, 0.0970±0.0020 cm3/g in LGM ice, and a mean value equal to 0.0924±0.0031 cm3/g (unpublished data). I suggest that the authors use estimates based on Martinerie, P., Raynaud, D., Etheridge, D.V., Barnola, J.-M., Mazaudier, D. Physical and climatic parameters which influence the air content in polar ice. Earth Planet. Sci. Lett. 1992, 112: 1-113, rather than the experimental data from Ruth et al. (2004).

L260: It may be noted here that uncertainty in hydrate size determination can also be the cause of the observed mismatch between calculated and measured air content (Vc is very sensitive to even small errors in hydrate linear dimensions), but in this particular case the greatest contribution to this mismatch comes, of course, from the uncertainty in Nah.

L267: Replace “air hydrates dissociating” with “air hydrates dissolving”.

L269-273: This reasoning seems a bit odd to me, because: 1) below the BHTZ, almost all of the air trapped by ice (98% or so) is stored in air hydrates. Therefore, whatever evolution the hydrate system undergoes (growth of larger hydrates at the expense of smaller ones, coalescence, etc.), the geometric properties of the hydrate ensemble, if accurately measured, should allow us to correctly estimate the TACh; 2) if the cage occupancy in Eq. (9) is erroneously overestimated, the calculated TACh must also be overestimated, but it seems to show the opposite trend?

L286-289: The geometric properties of air hydrates (inherited from the properties of bubbles) depend on both the temperature and accumulation rate prevailing during ice formation. The correlation between accumulation and temperature (isotope content of ice) on small time scales is not as obvious as in the case of global glacial-interglacial climate changes, and may not exist at all. This may also be the reason for the weakened correlation between Nah and δ18O in the case of smaller-scale climatic fluctuations.

L299-300: I would rewrite as “Due to the greater surface curvature, smaller particles are more soluble than larger particles…”

L310-312: With the exception of two data points, I do not see particularly large mean hydrate diameters in region 2 in Fig. 5c. It would be good to say here how the inability of your image analysis to distinguish individual air-hydrate crystals within their clusters may affect the count of hydrate number? It seems like it should lead to an underestimation of the number of individual hydrates (although the hydrate number concentration shows anomalously high values in zone 2).

L312-313: I cannot agree with this explanation for the abnormal volume concentration and TACh. Whatever the mechanism of hydrate coalescence, this process cannot change the initial (pre-coalescence) volume concentration of hydrates (and TACh), since this property of ice is as constant as its air content. As for the quantitative assessment of the geometrical properties of all hydrate inclusions (without discrimination into clusters and individual hydrate crystals), one would expect an underestimation of the size of hydrate clusters rather than an incorrect count of the total number of hydrate inclusions (individual crystals and their clusters), which should lead to an underestimation of the calculated total volume concentration of hydrates, but not the other way around. The underestimation of cluster size in image analysis could be quite expected due to the partial overlap of the projections of two or more individual hydrate crystals included in the cluster (see Fig. 7c) and the difficulty in assessing the cluster size in the direction normal to the image plane.

L326-327: It is not quite clear what phenomenon the authors have in mind, but if it is the disappearance of the climate-related variations in the hydrate properties in the course of Ostwald ripening, then the fundamental reason for this is that the hydrate growth rates are inversely proportional to the square of the mean initial hydrate size, which leads to dumping of the climatically induced variations in the hydrates’ geometrical properties (Salamatin et al., 2003).

Technical corrections

The caption to Fig 5 needs some explanation of the zones 1 and 2 shown in the figure.

Is the subscript 0 at n0 really needed in Eq. (9)?

L207: Reference to Fig. 8 is given before references to Figs. 6 и 7. Consider changing the order of the figures.

In Fig. 6, the blue and green symbols are not well distinguishable on my screen. Consider changing the colors.

Publisher’s note: this comment was edited on 8 May 2025. The following text is not identical to the original comment, but the adjustments were minor without effect on the scientific meaning.

The authors present a new method for analyzing clathrate hydrates using a mosaic-style automated image processing method in "thick sections" of ice-core ice...specifically in their study using the EDML core. They note that this technique allows for more automation and analysis of multiple clathrates while reducing (often tedious and time-consuming) manual counting. The authors used images that were taken almost immediately after core recovery, thereby reducing the risk of clathrate decomposition as a factor in their measurements/analyses.

One of the biggest questions I had was regarding the interpretation of calthrate number-density from a 2D image. The authors do note that they used the recorded thicknesses of the "thick sections" to calculate appropriate volumes for each sample, but that they also used mosaic images from a specific 2D plane in the samples (1.5 mm). This would seem to present some stereological considerations. For example, once clathrate counts are completed for a sample, when determining the appropriate volume for that sample, are cut corrections factored in? Should it be assumed that, similar with ice core bubbles in "thick sections", when sections are planed (or microtomed) for imaging, that some of the clathrates are cut at the surfaces? Given this, an additional amount of "volume" needs to be added back in to account for the portion of the clathrate (or bubble) that was removed. Martinerie and others (1999) as well as Saltykov (1976) note this and base the correction on both the shape of the sample as well as the mean size of the feature being measured in that sample. The end result usually means slightly lower number-densities than first measured.  This interpretation of thick sections also leads to considerations when looking at shapes or other orientation data on features as it's not always clear which axis the observer is looking down. Fegyveresi and others (2019) note this to some extent, and made some notable assumptions.

Because the number of clathrates is ostensibly determined by the number of bubbles, which in turn are determined by the number of grains, and ultimately the site temperature and accumulation rate, it may be beneficial to show a comparison of Nah to bubble number-density with depth. Visually it would also be interesting to see the crossover of bubble decrease and clathrate increase in the transition zone. It would also give some preliminary estimates of bubble-to-clathrate ratios.

In the calculations/estimations of Total Air content, (ln ~260-275), are there any possible surface elevation implications that could be interpreted from changes in TAC, or perhaps changes in impurity content?

Are there any other ice-flow or deformational considerations that need to be made with respect to the site and observed measurements? Has the ice divide migrated over time? has the ice flow direction changed? What about borehole inclination or azimuth uncertainty of thick sections with respect to ice flow? Was ice-flow azimuth of the cores even recorded during recovery?

Figure 5, panel 1, are the d18O data different colors because they come from two sources? (Also, the caption the Meyer et al. has no date.) While they are referenced in the text, the shaded bands should also be noted in caption. It's also not entirely clear what "a pronounced deviation of climatic influence" means here. Is this specific to the d18O data?

Figure 6 - I had some issues with the two colors being so similar...maybe just swap for something different? Perhaps noting in the graphs that the green dots are "Scaled values" instead of "adjusted". ...or maybe Nah-derived values (w 1.4 scaling applied). Either way, it should be clearer which color goes with which axis.

Figure 7 the eigenvalue lambda label should also be labeled on the axis or identified in the caption for the reader. Also, what are the shaded bands 1 and 2?

Figure 7 also shows a single max c-axis at ~2230, yet the text says it happens at 2030m.



Figure 8 - can you re-identify the stars? I know they are noted in previous plots, but it would be helpful to identify here as they are so noteworthy.

250 Is there a physical significance to the 1.4 scaling, or was it simply the number that allowed for the Nah values to correlated with measured TAC? Also why was this not shown in the graphs?

330  Authors note: For the EDML ice core, the age of the ice deeper than 2415 m is not known. However, the sudden increase in ice grain size by one to two magnitudes to up to 0.5 - 1 m for the deepest samples (Faria et al., 2010a; Weikusat et al., 2017) and the relatively warm in-situ ice temperature of at about 2550 m (Figure 7d) indicate favorable conditions for air hydrate crystal growth. Thus we interpret our data to mean that from this depth, the climate signal is dampened by considerable air hydrate crystal growth." - So should it be assumed then that there is a trend that would need to be accounted for at any depths being influenced by warming from the bed? 

196 - "No air hydrates were found in lowermost sample". Is this related to the proximity of the bed? Basal temperatures? etc.?