Matching of Vertical Snow Profiles

Stratigraphic vertical profiles of snow properties measured a few meters from each other are usually different. In particular, slope-parallel variations of the thickness of one or several layers can create an apparent spatial variability, which masks possible common stratigraphic sequences in these profiles. Besides the referencing in depth of the layer position, it is thus necessary to match the stratigraphic sequence between the different profiles. On a few profiles measured with a low vertical resolution, this time-consuming stratigraphic matching can be done manually. Here, we propose an automated matching method that applies to large data sets. The proposed matching algorithm enables to produce a single profile representative of the study site and provides a partitioning of the apparent spatial variability between property variability and depth variability. In our method, the thicknesses of individual layers are taken as free parameters to be adjusted in a least square optimization procedure that minimizes a certain metric between the profiles. The method efficiently synthesizes several vertical profiles of hardness such as the ones measured by the SnowMicroPen into one representative profile. Combined with clustering methods, it successfully groups numerous simulated specific surface area profiles into representative snowpack classes at the massif scale.