Adaptive Thresholod Method for Snow Covered Area Detection from LANDSAT Imagery

Abstract

This paper is concerned with a new detection method of snow covered area in LANDSAT imagery. For the purpose, A degaard's method has been well-known, in which if brightness of a pixel is greater than a threshold value, its domain is classified as snow covered. The threshold value is specified as a separating value between the brightness distributions of snow covered and snow free areas. The method, however, is not necessarily effective in mountainous areas because snows in shady areas or in sunny areas but not directed to the sun are not so bright.
In the new method, by refering to a digital elevation model and the sun position at image collection, domains of each pixel are classified into three categories, i.e., (1). sunny domain directed to the sun, (2). sunny domain but undirected to the sun, (3).completely shady domain. By applying three threshold values adaptively due to domain categories, snow covered areas can be detected by following to φ degaard's method. Results of detection for LANDSAT-MSS data of various seasons, the new method efficiently detect snow covered area, especially for the data when the elevation is low.