Application of a Bayesian snow water equivalent reconstruction technique to a mountainous basin in the Sierra Nevada

Water from seasonal snowmelt is one of the major components of the hydrologic cycle in many parts of the world, especially in mountainous regions where in-situ measurements of snow properties are often impractical. The accurate estimation of snow water equivalent (SWE) in such regions requires an integrated approach combining snow mass/energy balance modeling and remote sensing techniques. In this study we aim to apply a recently developed Bayesian remotely sensed snow covered area (SCA) data assimilation scheme to reconstruct seasonal SWE in a 625 km2 mountainous domain in the Sierra Nevada Mountains. The system is designed to use readily available forcing information from the North American Land Data Assimilation System (NLDAS) and includes a downscaling scheme for reference-level air temperature and humidity, pressure and incoming radiation (based on topographic information) so that the model can be applied at high resolution. The NLDAS precipitation product is used as an a priori estimate of the model pixel snowfall; where an a posteriori snow accumulation estimate is then obtained via assimilation of SCA measurements. Preliminary results from a series of synthetic experiments indicate the method should work for the case of deep snowpacks seen in the Sierras. Assimilation experiments are performed using fractional SCA estimates from both the standard MODIS product and the advanced MODSCAG product to assess differences in performance in the estimation of spatial SWE fields. The posterior estimates are compared to a suite of in-situ data to assess their validity and the potential for extension to a wider domain. Finally, the method is applied over several years to examine the interannual variability and persistence of spatial patterns of SWE.