Distributed vs. semi-distributed simulations of snowpack dynamics in alpine areas: case study in the upper Arve catchment, French Alps, 1989–2015

Abstract. We evaluated distributed and semi-distributed modeling approaches to simulating the spatial and temporal evolution of snow and ice over an extended mountain catchment, using the Crocus snowpack model. The distributed approach simulated the snowpack dynamics on a 250-m grid, enabling inclusion of terrain shadowing effects. The semi-distributed approach simulated the snowpack dynamics for discrete topographic classes characterized by elevation range, aspect, and slope. This provided a categorical simulation that was subsequently spatially re-projected over the 250-m grid used for the distributed simulations. The study area (the upper Arve catchment, western Alps, France) is characterized by complex topography, including steep slopes, an extensive glaciated area, and snow cover throughout the year. Simulations were carried out for the period 1989–2015 using the SAFRAN meteorological forcing system. The simulations were compared using four observation datasets including point snow depth measurements, seasonal and annual glacier surface mass balance, snow covered area evolution based on optical satellite sensors, and the annual equilibrium-line altitude of glacier zones, derived from satellite images. The results showed that in both approaches the Crocus snowpack model effectively reproduced the snowpack distribution over the study period. Slightly better results were obtained using the distributed approach because it included the effects of shadows and terrain characteristics.