Satellite retrievals contain cloud cover, biases e.t.c which is the main challenge to prepare the data for further analysis. After all, how we can efficiently calculate the snow cover and its temporal change in any language base platform from satellite data?
Identification of snow in satellite images is done using multiwavelength measurements and including ancillary information. Snow obviously has far more reflectivity compared to barren soil, water or vegetation and any edge identification algorithm on visible channel images will identify their extent. The biggest trouble comes from clouds which can appear as bright as snow in visible channel images. Satellite measurements in the infrared region of spectrum help in the discriminate snow vs cloud. Infrared emissions of an object depend on temperature and clouds being at a higher altitude has far less temperature compared to snow. The algorithm may use other information also, for example, day to day variability. Clouds have a very short lifetime and display large variability whereas snow/ice disappears far slowly compared to cloud.