Two-Stage Polsar Scattering Model-Based Classification Scheme for Improved Glacier Facies Mapping

Variations in glacier facies may signify the glacier’s response to the surrounding climate, and continuous monitoring of glacier facies can reveal a lot about the glacier’s behavior and stability. The swift development of remote sensing and the handiness of polarimetric SAR data has gained popularity for monitoring glaciers and their dynamics. We used ALOS-1/PALSAR-1 L-band data over the Siachen glacier in the Karakoram Himalayan region for this study. For glacier facies/zones classification, we employed a two-stage scattering model-based SVM classification scheme for improved glacier facies mapping. Results showed that two-stage classification using 6SD-SVM is effective, with a kappa coefficient of 0.82 and an overall accuracy of 87.58%. Integration of scattering-based polarimetric information extends a new dimension in glaciated terrain classification, and generates enhanced accuracy in classified images. Even though the employed technique produces satisfactory results, but classes for mid- & low-percolation and debris cover are misclassified. To further clear up any ambiguity between the aforementioned classes, the probability difference between surface and volume backscattering has been added as a second step in the second stage of the classification process. In comparison, 6SD-SVM outperforms the backscatter [T]-SVM classification and the overall accuracy is enhanced by 7%.