General comments

This paper conducted cold chamber experiments for continuous observation of snow metamorphism under the strong temperature gradient approximately 100 K/m and clarified the heterogeneous grain growth. As the authors say, there have been several previous studies in which snow layer was continuously observed in a low-temperature chamber using X-ray CT, but the novelty of this paper lies in the experiment that the vertical structure of the approximately 10 cm snow layer was observed precisely and continuously. By applying a strong temperature gradient to a thick snow layer, the different regimes of crystal growth within a single snow layer were achieved. The authors also mentioned a decrease in the density at the bottom of the snow layer, and the formation of hard depth hoar. These results will provide fundamental data for snow transformation modeling and snow stability prediction.

Specific comments

1. The authors mention that different regimes of crystal growth were observed depending on the temperature range. Figure 11 shows in the height direction for each temperature zone, but the columns and plates are very difficult to see from the 2D image. Photographs of the particles or a 3D surface rendering would be appreciated.

2. Sturm and Johnson (1991) reported that the depth hoar near the bottom of the natural snowpack in Alaska has a C-axis that is oriented almost horizontally in some places and is growing horizontally. (same figure of Fig. 2 of Sturm and Benson (1997)).

In this range of prismatic growth, did the prismatic face of the snow with the C-axis oriented horizontally grow vertically and the horizontal basal growth was not observed?

3. The experiment with large temperature gradients at very low temperatures is similar to that of Kamata and Sato (2007). However, Kamata and Sato's experiment lasted 5.5 days, whereas the authors observed for about a month. The authors would mention what differences they found over a longer period of time, although there is a description of a large change in the initial period.

4. The differences in temperature ranges appearing in the long vertical samples lead to interesting results. However, since the number of experiments was only two, it is hoped to increase the number of experiments to obtain a data set in future work. Kamata and Sato also have a few experiments, so these experiments will provide valuable data.

Technical corrections

Line 77

“Those evolution” is “Those evolutions”

Line 306

“can not explained” is “can not explain”

Line 590

“Yosida: “ is “Yosida, Z.”

General comments

The paper by Bouvet et al. discusses two laboratory experiments where snow is put under a temperature gradient typical for arctic conditions. The changes in snow structure are mainly analysed by micro-tomography. Similar experiments have been conducted since 2004 by other authors. Using humidity sensors, an attempt was made to measure relative humidity inside the samples. The presentation of the data is sufficient. However, the data can not be publicly assessed, and the deposition of the data in a repository is now a scientific standard.

As observed by others, snow sublimates at the warm side of temperature gradient experiments, forming an air gap (already observed by Nakaya in the 1950'ies). Such an air gap immediately caused the thermal conductivity to be reduced to the one of air, and the initially vertical and parallel heat flux became distorted as the samples were surrounded by higher conducting plastic. As much as the reviewer could see, this fact was not taken into account (e.g. by numerical simulations) for the interpretation of the structural evolution of the snowpack.

The interpretation of the temperature and humidity profiles is consequently misleading. Without considering the non-vertical heat fluxes, no valid conclusions are possible. The sections "Results" and "Conclusion" must be rewritten and re-interpreted, considering a heterogenous temperature gradient and heat flux.

Specific comments

No details and data are given on how the humidity sensors are calibrated at below zero-degree conditions. Calibration before and after the experiment would have been necessary to have valid data.

The authors state that the initial density is almost constant. Their figs. 5 and 9 clearly show density fluctuations of up to about 30% at a distance of a few millimetres. Such density variations strongly affect thermal gradients and, therefore, snow metamorphism. A detailed interpretation of thermal conductivity and temperature gradients is necessary.

The mean covariance length (which should probably read "mean correlation length") is given without a directional index, and no formula or precise reference is given for its calculation. Is the mean covariance length averaged in the horizontal and vertical directions?