Snow observation by satellite: A review

Publisher Website
Google Scholar
Abstract
Surface snow cover is able to be detected within the spectral and thermal wavelengths by a range of satellite sensors and the area and frequency of observation is a function of both spatial and temporal resolution. The Landsat, NOAA and GOES satellite are primarily employed for routine snow mapping, although each sensor has specific limitations. Snow observation is inhibited by sensor saturation problems, and also cloud cover which both obscures the snow surface and exhibits some spectral overlap with snow. A number of developed techniques allow snow/cloud discrimination, with varying degrees of success, although the most promising of these include the middle infrared (1.6 μm) channel in the analysis. The close correspondence of the distribution of snow cover with terrain has also enabled the interpolation of snow cover into cloud obscured regions. Shadows from terrain generally confuse the location of snow covered pixels and procedures correcting for the variation in illumination have been generated. Vegetation cover, particularly conifer forest, also reduces the reflectance of snow covered surfaces and prevents reliable calibration of pixel intensity to snow depth or percentage snow cover. Many studies have therefore developed techniques to identify snow in vegetation covers. The detection of the snow/no snow boundary and subsequent estimation of snow area has been achieved by using a variety of approaches ranging from interactive delineation and planimetry or thresholding to multi‐temporal analysis and more sophisticated grid‐ding or digital techniques.