Mapping of Radar Glacier Surface Facies Using Supervised Algorithms

The main objective of this study is to implement and evaluate five different models namely Support Vector Machine using radial basis function (SVM-RBF), Support Vector Machine using polynomial (SVM-P), Support Vector Machine using sigmoid (SVM-S), Minimum Distance (MD), and Parallelepiped (PP) classifier for glacier surface facies mapping of Samudra Tapu glacier using data from ALOS-1 PALSAR satellite. Since the quantitative and qualitative assessment of monitoring glaciers is one of the most efficient means to observe the glaciated terrain, there is a prerequisite to examining the accuracy of different algorithms for glaciated features classification to identify the best classifier for the selected study region. Our evaluation showed that with optimal tuning parameters, the SVM-RBF classifier with penalty parameter 100 yielded the highest overall accuracy (OA) of 80%, performing better compared to other methods.