Assessment of Snow Cover Fraction Parameterizations for High Resolution Snowpack Reanalyses

ABSTRACT Recent advances in spatially‐distributed snow water equivalent (SWE) estimations have benefited from the assimilation of remotely sensed snow‐covered area data, especially in mountain regions where precipitation variability is poorly known. Such data assimilation frameworks depend on a snow cover fraction (SCF) parameterization, that is, an observation operator which maps variables from the model space (e.g., snow depth, SWE) to the observation space (snow‐covered fraction of a model grid cell area). This parameterization has been subject to many studies with coarse resolution models (> 1 km) but the conclusions are not applicable to snowpack models of a few hundred meters since this parameterization is known to be scale‐dependent. Here, we assess the performance of several SCF parameterizations within a 100 m resolution data assimilation framework in a mountainous region with Mediterranean climate (Tuolumne river basin, USA). We design three experiments to assess these parameterizations, taking advantage of the accurate lidar snow depth products acquired by the Airborne Snow Observatory (ASO) to reduce model forcing uncertainties using the particle batch smoother. Despite its simplicity, an asymptotic parameterization performs the best with an SCF RMSE of 16%. Then we assimilate SCF maps derived from the ASO observations and the asymptotic function still performs the best with a snow depth RMSE of 0.40 m. Finally, we assimilate 500 m SCF from MODIS products and find that the asymptotic parameterization and the parameterization accounting for pre‐melt subgrid SWE variability yielded similar results in terms of posterior snow depth with an RMSE of 0.33 m. Both parameterizations are well suited to improve high resolution SWE or snow depth estimates through the assimilation of remote sensing SCF in such mountainous and climatic context.