Snowpack Modeling in the French Mountains Driven by Short-Range High-Resolution Weather Forecasts

Numerical Weather Prediction (NWP) systems operating at kilometer scale in mountainous terrain offer great opportunities for forecasting snowpack-related issues such as mountain hydrology or avalanche hazard. In this study, daily forecasts of the NWP system AROME at 2.5-km grid spacing over the French Alps and the Pyrenees were considered for four consecutive winters (2010/11 to 2013/14). They were used to drive the detailed snowpack model Crocus and simulate the snowpack evolution over these regions. The evaluation was performed through comparisons to ground-based measurements of snow depth, snow water equivalent (SWE) and to snow extent derived from satellite data. Results of AROME-Crocus were also compared to snowpack simulations driven by the meteorological analysis system SAFRAN, specially developed for alpine terrain. SAFRAN assimilates observations from the manual and automatic meteorological stations deployed in the French Alps and the Pyrenees. It includes in particular a precipitation analysis.

When evaluated locally at the experimental site of Col de Porte (1340 m, French Alps), AROME-Crocus and SAFRAN-Crocus show good agreement with measurements of snow depth and SWE. At the scale of the French Alps and the Pyrenees, snow depth simulated by AROME-Crocus exhibits an overall positive bias with strong spatial patterns. Differences in snow depth simulated by AROME-Crocus and SAFRAN-Crocus are mainly related to differences between AROME and SAFRAN seasonal snowfall. Comparison between the two reveals that high elevations areas and leeward and windward side of some mountain ranges exhibit especially significant differences. The simulation of mesoscale orographic effects by AROME allows to capture a realistic regional snowpack variability, unlike SAFRAN–Crocus as illustrated by winter 2011/2012 in the Pyrenees. Finally, a categorical study of daily snow depth variations is carried out to provide a more detailed analysis of model results during accumulation and ablation phases. It reveals in particular that both systems underestimate strong snow depth ablation, mainly because of non-simulated wind-induced erosion and underestimated strong melting rates.

This study constitutes the first step towards the development of a distributed snowpack forecasting system using AROME forecasts.