Resource information
Irregular shape of terrain causes variable illumination angles and diverse reflectance values within same land cover type in optical remote sensing image. It causes problems in image segmentation and misclassification (snow with other land cover). This perception leads to develop an empirical-statistical based topographic correction (ESbTC) algorithm for reflectance modeling after compared with existed topographic correction methods like Cosine correction, C-correction, Minnaert correction, sun–canopy–senor with c-correction (SCS + C) and slope matching, in the context of snow reflectance. An image based atmospheric correction has used in present study included dark-object subtraction (DOS) and effect of Rayleigh scattering on the transmissivity in different spectral bands of AWiFS and MODIS image data. The performance of different models is evaluated using (1) visual analysis, (2) change in snow reflectance on sunny and shady slopes after the corrections, (3) validation with in situ observations and (4) graphical analysis. Further snow cover area (SCA) has been estimated with normalized difference snow index (NDSI) and validated with support vector machine (SVM), a supervised classification technique. The result shows that the proposed algorithm (ESbTC) and slope-matching technique could eliminate most of the shadowing effects in Himalayan rugged terrain and correctly estimate snow reflectance from AWiFS and MODIS imagery as compared with in situ observations whereas other methods significantly underestimate reflectance values after the corrections.
