A new method for weekly sub-kilometer mapping of deep snow in mountainous regions using ICESat-2 and Sentinel-1

The following abstract is based on a master thesis project in Geography from University of Copenhagen. It relies on freely available satellite data and uses Google Earth Engine and python for large scale analysis of deep snow in the Southern Scandinavian Mountains of Norway.

Knowledge about seasonal snow accumulation is key for managing water resources, especially in mountainous regions. However, accurate measurements of snow depth or SWE at a high spatiotemporal resolution are sparse. In this study, we investigate the effectiveness of a multi-satellite approach to mapping the depth of large-scale deep snow in the Southern Scandinavian Mountains of Norway. First, snow depths are measured using geolocated photons from the ICESat-2 satellite. These snow depths are matched spatio-temporally with the nearest Sentinel-1 scene, where an index based on the ratio between VV and VH polarization has been proven to be correlating partially with snow depth. Using a simple regression analysis, we model this relationship using a new sampling method, to further investigate the relationship between Sentinel-1 index and snow depths. The model is used to predict snow depths at 500m resolution every 6/12 days. When compared to in situ measurements from weather stations within the study area, our model has an RMSE of 36 cm.