Snow Cover
Last updated
Last updated
We use satellite imagery in order to detect snow on the earth's surface, which enables us to provide a continuous snow cover map for all parts of the world that not lie within the polar shadow, like Greenland during winter on the Northern Hemisphere. For areas of special interest, like the European Alps, we use satellite images with a higher spatial resolution of 20 meters (compared to 500 meters globally), meaning a single pixel covers an area of 20 by 20 meters on the ground. This is a unique level of detail in this domain and special know-how is required to successfully detect snow in different landscapes (forests or crop fields) and regions (sunny or shadow sides of mountains).
To achieve high spatial accuracy and daily updates, data of nine different optical satellites is digested by our system (ESA Sentinel-2 A/B, ESA Sentinel-3 A/B (OLCI + SLSTR), NASA Landsat 8/9, NASA VIIRS (SNPP) and NASA MODIS Aqua/Terra). Unfortunately, clear sight for the sensors can not always be achieved due to cloud cover. To overcome this limitation, machine learning algorithms are deployed to intelligently model snow lines, topographic dependencies and land cover influences to fill the gaps and accurately estimate snow cover probabilities. Furthermore, estimations of previous days are obviously very helpful as well.
It takes five to twenty hours for satellite data for satellite data to be accessible for us on earth, so our daily snow cover mask best represents the conditions of the respective previous day. For prolonged cloud cover situations, this time lag is further increased and documented for every single pixel in our product "Last Direct Measurement".
The snow information layers are updated every day.
Artificial intelligence algorithms are deployed to map snow cover over clouded regions.
The machine learning algorithm integrates previous observations to better estimate current conditions.
Currently, global results are based on NASA MODIS Aqua/Terra satellites and offer a spatial resolution of 500 m per pixel.
The high resolution 20 m products are based on ESA Sentinel-2 A/B and NASA Landsat 7/8 and are available for areas of special interest.
