Multiscale tomography of the seasonal evolution of the snow microstructure

Due to very active metamorphism, snow on the ground exhibits a wide range of microstructural patterns. Indeed, snow is a very porous material and it exists on Earth close to its melting point, which promotes its structural evolution through vapor transport and melting-refreezing processes. State-of-the-art detailed snowpack models such as SURFEX/Crocus still represent this microstructure in a very rough way. This representation is based on manual observations from the 1990’s using magnification lenses, where the snow grain shape were classified into different types. The descriptors derived from this classification, such as sphericity or grain size, are not based on a sound physical background and cannot be measured, which limits any further improvement of the existing parameterizations. Nowadays, tomography has become a standard technique to capture the 3D snow microstructure at a micrometrical scale in laboratory conditions. Besides, homogenization methods can now numerically estimate several essential but difficult-to-measure snow properties such as thermal conductivity or mechanical viscosity from tomographic images and the ice and air properties. To overcome the limitations of existing snowpack models and to benefit from the wealth of data provided by tomography and numerical homogenization, a new generation of snow models with an explicit and objective representation of the snow microstructure is currently under development. To develop and evaluate these new models, characterization of the snow microstructure evolving in the field is required. The objective of the presented work is to develop a measurement and data processing protocol to be able to conduct these measurements. This represents a challenge because, to date, tomography was mainly limited to small volumes of snow mostly harvested in laboratory conditions. First, we installed a tomograph directly at our snow field site, Col de Porte, 1325 m a.s.l., french Alps. Second, we designed a specific snow cutter to sample snow cores without destroying their very fragile microstructure. Cutters equipped with a sharp hole saw and with an inner diameter of 44 mm and a height of 100 mm are sufficiently large to prevent sample failure and small enough to conduct partial tomography at a very high resolution. Last, we combined two types of tomographic scan in order to capture a high-order approximation of the snow microstructure while maintaining the scanning time short enough. In particular, on each snow core, we scanned a sub-volume (15 mm diameter, 15 mm height, 48 min scan duration) at an effective resolution of 10 microns and the whole sample column (25 mm diameter, 100 mm height, 20 min scan duration) at a resolution of 42 microns. Based on a modified two-point correlation function which applies directly to the greyscale tomographic images and the combination of the two scans, we were able to recover physically-based proxies of the snow microstructure of the full core in a reasonable measurement duration. This includes density, specific surface area and mean curvature.