S. Gascoin

Moutain snow cover is highly variable both spatially and temporally and has a tremendous impact on ecosystems and human activities. Numerical models provide continuous estimates of the variability of snow cover properties in time and space. However, they suffer from large uncertainties, for instance originating from errors in the meteorological inputs. Here, we show that the snow depth variability at 250 m spatial resolution can be well simulated by assimilating snow depth maps from satellite photogrammetry in a detailed snowpack model. The assimilation of a single snow depth map per snow season using a particle filter is sufficient to improve the simulated snow depth and its spatial variability, originally poorly represented due to missing physical processes and errors in the precipitation inputs. Assimilation of snow depth only is nevertheless not sufficient for both compensating for strong bias in precipitation and for selecting the most appropriate representation of the physical processes in the snow model. Regarding this limitation, combined assimilation of snow depths maps and other snow observations, such as snow cover area, surface temperature or reflectance, is a promising avenue for accurate simulations of mountain snow cover.