Towards the assimilation of MODIS reflectance into the detailed snowpack model SURFEX/ISBA-Crocus.

Abstract:

Numerical simulations of snow on the ground are used for numerous scientific and operational applications such as avalanche hazard forecasting. Although the chain of models used in French mountain ranges for meteorological analysis and forecast (SAFRAN) and detailed snowpack modeling (SURFEX/ISBA-Crocus) usually perform reasonably well, significant differences with snowpack observations are common and are primarily attributed to the uncertainties in meteorological input and to the heterogeneity of snowpack conditions at all scales.

So far, no snow observation is assimilated into this model chain, so that simulation errors can accumulate over the winter season. Current efforts are devoted to the assimilation of data from visible and near-infrared imagers into the snowpack model. These efforts rely on the recently developed “TARTES” optical scheme that computes reflectances at various wavelengths using the vertical profile of the physical properties of snow predicted by the snowpack model.

In a first step, we performed ensemble simulations by perturbing the atmospheric forcing consistently with its estimated uncertainty. These experiments showed that the simulated snowpack evolution is extremely sensitive to this uncertainty, and that the assimilation of observations can greatly improve model results.

In a second step, we performed assimilation experiments using synthetic imager observations and a particle filter. The experiments were carried out for the location of Col du Lautaret area (French Alps) over 5 hydrologic seasons. They provide a good insight about the potential and limitations of assimilating imager data to improve the representation of the snowpack.