**5. Discussions**

General observations can be deduced from the results of this initial study. The present study shows that displacement of snow's internal structures during compaction can be investigated from the micro-CT data coupled with the DVC and porosity analysis. Embodied in this study are the vital observations that (1) the majority of snow grains have deep hollow core, (2) displacement of snow grains due to grains breakage and recrystallization, and (3) vertical heterogeneity in porosity distribution of the snow sample in response to compaction.

The growth of ice crystals depends on the temperature and humidity [27]. Shimada and Ohtake [28], Wergin et al. [29] and Riche et al. [30] reported that ice crystals grow as columns and needles around –5 ◦C and as plates and dendrites around –15 ◦C. This may explain the complex structure of the investigated dry natural snow as colder temperatures higher up in the atmosphere produces dendritic snow and needles are formed during the precipitation close to ground.

The complex microstructure of the investigated snow volume shows the inter-granular structural changes due to compaction. Compaction of snow by applying micro-mechanical uniaxial load forces the individual ice crystals to move closer via grain sintering and bond breakage. This process is known as pressure sintering [31], which results in rounded and less complex microstructure of snow. Wilkinson [32] detailed the influence of pressure sintering on snow microstructure and its mechanical properties. The sintering process in a snow pack was observed previously [33,34], where snow coarsening was forced due to loading and liquid water content. A common observation of the snow compaction between the tomography data and our previous works, is that the majority of displacement of snow grains occur at the near-surface layers. Gubler [35], Szabo and Schneebeli [36] reported that they observed similar behavior of snow microstructure during compaction with respect to temperature and time. Moreover, Schleef et al. [16] observed the sintering effect during the relaxation phase of a snow sample, which has similar microstructural characteristics as the one presented in this study. Major observations from their work are based on variations in strain rates and microstructural properties during the relaxation stage. They also discussed about a critical point during higher compression rates where a snow structure can be collapsed and rearranges into more compact structure.

The tomography data presented in this initial study was useful in terms of higher resolution measurements, observation of hollow core structure and analysis of porosity distribution. However the study is also limited by the long measurement time, lack of stress relaxation measurements, correlation window size for DVC analysis and the sample size. In principle, this initial study can be further extended in future to analyze different snow types under the influence of loading and isothermal metamorphism conditions. In addition, the selected VOI (refer to Section 2.2) must consider the whole volume to observe the edge effects. Figure 7 shows that some of ice crystals move out of the scan volume during compaction due to the finite size of the VOI, where the relevant information about the re-distribution of these ice crystals is missing.

Furthermore, the sample preparation appears to play a major role especially when it comes to tomography measurements. Ebner et al. [12], Zermatten et al. [37] performed structural analysis of snow samples by exposing to an advective airflow, and further observed porosity distribution, coarsening rate and settling of snow. In addition, Ebner et al. [12] reported that there is no correlation between the porosity distribution and settling of snow. However, a clear correlation is observed in this study where, porosity distribution at 0.3 MPa load stage appeared to be due to the snow settlement, see Figure 8.

The elastic characteristics of the described hollow core feature can be further investigated in future experiments. The observations from this study to understand the microstructural changes in snow grains and properties in response to compression tests, may be helpful to investigate the snow mechanics. Further experiments considering the limitations, can be advantageous in various fields of snow research. In specific, the observations in this study can be useful to understand snow mechanics for winter tire testing, ski track preparation and avalanche prediction. When snow is compressed, friction between the snow and tire decreases, which can be investigated further based on the presented study. For the ski tracks and avalanche, the investigation in this study helps to understand the snow quality and snow pack stability.
