Towards an Aassimilation of MODIS VIS/NIR reflectance into the detailed snow model SURFEX/ISBA-Crocus

Abstract:

SURFEX/ISBA-Crocus is a physically based multi-layer snowpack model used for numerous scientific and operational applications such as avalanche risk forecast. Although some snowpack models simulations usually performed reasonably well, differences with real snowpack still exist and may be due to various origins such as weather forcing input. Yet, no snow observations are assimilated into the snow model SURFEX/ISBA-Crocus so that the simulation error is accumulated over the winter season. Some efforts will be done to assimilate data from visible and near-infrared imagers into the snowpack model to improve the snowpack simulations. The new optical scheme of SURFEX/ISBA-Crocus, called TARTES, allows the use of reflectance as diagnostic variables of the model. These reflectance are sensitive to snow properties such as specific surface area (SSA) and impurity content. They are measured by the MODIS spectroradiometer and can thus be used in an assimilation framework to account for the high spatial and temporal variability of the snow cover in mountainous areas. Prior to assimilation, we used ensemble methods to find the best assimilation scheme to be implemented. The distribution of model errors is investigated together with the relationship between simulated reflectance and model prognostic variables (density, SSA, …). First tests of reflectance assimilation were then carried out using a particle filter and MODIS measurements at Col du Lautaret (French Alps). The impact of the assimilation has been evaluated in terms of simulated snow properties.