Snow cover extraction in mountain areas using RadarSat-2 polarimetrie SAR data

Abstract

Optical remote sensing data provide an effective way of mapping snow cover but limited by solar illumination conditions, whereas polarimetric decomposition technology offers the ability to monitor snow cover in all weathers. In the present study, a support vector machine (SVM) method for extracting snow cover based on RadarSat-2 Polarimetric SAR data in rugged mountain terrain is introduced. In this method, backscattering coefficient images of RadarSat-2 are analyzed, using snow-covered and snow-free areas obtained from GF-1 satellite observations as the “ground truth.” The analysis' results indicate that the backscattering coefficient in four polarizations is clearly correlated with the underlying surface type and local incidence angle, and there is a slight difference in backscattering coefficient between snow-free areas and snow-covered areas in the snow-accumulation period, and the backscattering coefficient of snow-covered areas is 3~10 dB smaller than snow-free areas in the snow-melt period. Then local incidence angle, underlying surface type, training samples from GF-1 wide field viewer (WFV) data combined with the optical polarimetric feature combination obtained from polarimetric feature decomposition were used to build a SVM classifier. The classification results demonstrate that snow cover extraction using this method can achieve mean accuracies of 73.6% and 82.7% for snow-accumulation and snow-melt periods, respectively.