Multi-platform, multi-sensor snow surface properties for energy balance and model validation

Snow cover, snow albedo, and the impact of dust are important properties used to estimate snow water equivalent (SWE) and snowmelt. Near-real-time satellite observations can be used to guide forecasts. Historical records can be used to refine forecasting methods and to improve our understanding of the cyrosphere and interactions with the biosphere. While multiple platforms and sensors exist, the trade-off between spatial and temporal resolution of available satellite data for estimating these snow surface properties remains a challenge. Daily observations of snow cover and snow albedo are available from MODIS and VIIRS at a resolution of 500 m and 1 km, respectively, but snow properties vary at much finer scales. Landsat observations at 30 m have been used for validation and calibration of MODIS and VIIRS algorithms, but only at 16-day intervals. At 10 m spatial resolution, the Sentinel-2 satellites (a+b) provide more frequent observations (10-day repeat for 1 satellite, 5-day repeat for 2) and harmonized Landsat and Sentinel-2 are nearly frequent enough to track snow accumulation and melt. However, currently no sensor at such a fine spatial resolution provides daily images of the whole terrestrial surface. Using surface reflectance inputs from VIIRS, MODIS, Landsat, and Sentinel, we use a physically-based spectral mixture analysis and spectral differencing methods for mapping snow cover, snow albedo, and the impact of dust. Using project examples funded by NASA, NOAA, and the California Department of Fish and Wildlife, we demonstrate the utility of the data. For example, we fuse these multi-sensor data using random forests and find fused data exhibits good skill in cross-validation experiments. We assess the snow cover mapping from with data from the Airborne Snow Observatory and commercial data from Digital Globe and Planet. Snow cover and albedo data are assimilated into the NOAA National Water Model and used for validation in the NOAA GFDL model. Our SWE products drive snowmelt in the Army Corp of Engineer's HEC-HMS model. These physically-based snow property retrievals are critical in global energy budgets and to validate global and regional models.