Preliminary characterization of Alpine snow using SnowMicroPen

Objective and accurate observations of snowpack layering and properties are essential for avalanche warning. Human observations are helpful, but they are often highly subjective and therefore inconsistent making them difficult to interpret. In this paper we address this problem by applying a high-resolution penetrometer to determine snowpack properties. We develop an algorithm to characterize the different snow microstructure classes using the signal obtained from the penetrometer measurements. For this purpose a database consisting of various snow profiles of Alpine snow was created, which contains the records of the mean, standard deviation and coefficient of variation of penetration resistance over 1 mm depth. The algorithm is able to characterize the major snow classes, namely, new snow, faceted snow, depth hoar, rounded grains and melt-freeze with an acceptable accuracy for warning purposes. We find that new snow, depth hoar and melt-freeze snow layers are characterized better than the rounded grains and faceted snow layers. However, knowledge-based rules, formulated from the experience of field researchers, can further improve the automated snow profiling procedure. For a comprehensive prediction of snow classes with better accuracy, it is proposed in future to include more expert rules and to enlarge the database of measurements. This preliminary approach could be helpful in validating numerical simulation models of snowpacks and subsequently numerical avalanche forecasting. The potential application of snow class information derived from the SMP to infer the specific surface area of snow layers is also discussed.