Approach

In “AI4Snow”, we propose a new way of facilitating Articial Intelligence (AI) methods to achieve three improvements crucial for the application of remote sensing-based snow cover products for hydrological modeling, forecasts, and climate change studies:

• harmonization of the various snow cover parameters originating from different sensors
• filling of gaps in the data caused by e.g. cloud cover
• (down-)scaling of all input products to a 100 m grid with daily availability

The AI methods required to help predict these three developments will be trained based on a high-quality, physical-based snow process model. The AI will then be deployed on a data cube consisting of snow cover products derived from various optical satellite data and SAR including Sentinel-1 (wet/dry snow), Sentinel-2 (FSC, 20 m), Sentinel-3 (FSC, 300 m), Landsat-8/9 (FSC, 30 m), land cover information (Forest/Fractional Forest cover, Glaciers), a Digital Elevation Model (DEM), and potentially also gridded meteorological data.