Snow Cover Characterization Using L-band PolSAR Data in Parts of the Himalaya

Geo-spatial variation of snow physical parameters can be gathered through remote sensing data. This information can provide users an important spatial decision-making capability in various applications, e.g. monitoring of avalanche hazard susceptible regions, snow load calculation on avalanche slopes and snow hydrology. The detailed investigations of Himalayan snow-covered regions using geospatial data are become extremely important to plan various activities. Therefore, it is essential to have precise snow cover characterized maps of such complex terrain. In this study, a new methodology adopted using ALOS PALSAR 2 (ALPS-2) L-band Full-pol. data for characterization of snow cover into dry and moist/wet snow in mountainous regions. Two data sets are analyzed for snow-accumulation period and snow-melt period, respectively. Yamaguchi decomposition was used to generate various polarimetric descriptors. Further, a supervised classification scheme using Support Vector Machine (SVM) in-conjunction with polarimetric descriptors and combined with training samples is used to classify snow cover into dry and moist/wet snow. In addition, optical multispectral Sentinel-2A data of nearby dates has been analyzed using Decision Tree (DT) classification method for comparison of the results [1] [2]. The results comprehended with the unsupervised Wishart H-α classifier for understanding the scattering process happening in metamorphosed snow cover with time. The L-band SAR data shows its capability for snow cover type characterization and will further enhance its applicability with upcoming NISAR mission.