Abstract ID: 3.11773 | Accepted as Talk | Talk/Oral | TBA | TBA

Spatially-distributed snow albedo data can be derived from multispectral satellite data, including MODIS, Sentinel-2, and Landsat platforms. While a number of snow albedo algorithms exist, a recent MODIS-focused analysis shows that approaches that use spectral mixture analysis (SMA) result in the most accurate snow albedo products. SMA uses spectral libraries to solve for the snow fraction, grain size, and the impact of light absorbing particles (LAP) in each satellite pixel; snow albedo is then estimated by combining the grain size with darkening due to LAP. The accuracy of SMA-derived snow albedo estimations increases when more spectral information is available. The upcoming NASA Surface Biology and Geology (SBG) satellite mission will provide hyperspectral data at approximately 30 m spatial resolution, which can improve snow albedo estimates compared to multispectral satellite data. However, the planned orbit of SBG will result in temporal data gaps up several weeks depending on the region of interest. These relatively sparse temporal observations will miss important snow albedo changes on daily to weekly timescales, especially large changes from early season accumulation or late season dust-on-snow events. In this presentation, we demonstrate the potential of a data fusion approach to produce daily, SMA-derived snow albedo data at high spatial resolutions using multispectral and hyperspectral imagery. Our model fuses snow albedo products instead of surface reflectance measurements, to leverage the ability of SMA to calculate pure snow albedo even in partially snow-covered pixels. We use a random forest model trained on airborne snow albedo surveys from both AVIRIS-NG and the Airborne Snow Observatory (ASO) at 50 m spatial resolution. Predictor variables include daily, 500 m MODIS snow albedo generated using SMA, as well as terrain characteristics and solar illumination. The fused product takes advantage of the more accurate and finer resolution hyperspectral data while maintaining the daily temporal resolution of multispectral MODIS imagery. This flexible approach demonstrates a potential way forward for a combined VIIRS/SBG snow albedo product in the future.