Snow cover algorithms:

The Unit “Land, Vegetation and Snow” provides two different snow cover products of medium and high resolution respectively.
The snowcover map based on MODIS data has a medium resolution of 250m. The input data are the atmospherically corrected reflectances of MODIS MOD09GQ , MOD09GA and the MODIS level 1 product MOD021KM. The processing steps include converting the map projection (Sinusoidal to UTM WGS 84) and a topographic correction based on a digital elevation model (DEM). Then snow and clouds are classified using a decision tree algorithm. Snow under forest is detected by comparing the Normalized Difference Vegetation Index (NDVI) of a winter image and a reference image from summer season. To reduce the cloud covered area in the classification the snowcover maps derived from data of both MODIS sensors onboard of Terra and Aqua for the same date can be also combined.
The snowcover map of high resolution (30m) is based on Landsat TM / ETM+ images. The underlying algorithm starts with converting band 2,3,4,5 from DN to radiance and reflectance in order to calculate the Normalized Difference Snow Index (NDSI) and the Normalized Difference Vegetation Index (NDVI). Temperature (T) is retrieved from the thermal band 6 and a threshold of T<283K for possible snowcover is defined. An initial snow mask is obtained by the thresholds NDSI>0.40 and NIR>0.11. For shadowed areas, which are identified by a solar radiation map calculated in ArcGIS, slightly adapted thresholds are applied. To better distinguish between snow and clouds, the the “Automatic Cloud Cover Assessment” algorithm introduced by Irish, 2000 (Irish, R. R., 2000. Landsat 7 Automatic Cloud Cover Assessment. Electronic document: http://landsathandbook.gsfc.nasa.gov/handbook/pdfs/ACCA_SPIE_paper.pdf - July 2009)  was introduced and further modified. As a final step all resulting data are put in a decision tree classifier to obtain the final snow cover map with the classes: snow, no snow, clouds and no data.