Mapping sequences and mineral deposits in poorly exposed lithologies of inaccessible regions in Azad Jammu and Kashmir using SVM with ASTER satellite data

Exploring minerals, lithological sequences, and geological formations remained challenging in territorial armed conflict and environmentally hazardous zones. Satellite-based remote sensing is appropriate when direct studies are cumbersome due to boundary problems or morphological strains over large inaccessible regions. Therefore, the main objective here is to map different minerals and identify poorly exposed lithology in district Poonch of the Pakistani territory of Azad Jammu and Kashmir, which has a variety of economic deposits and rock sequences. We used support vector machine (SVM) and maximum likelihood classification (MLC) techniques with ASTER and LANDSAT 8 OLI imagery. To validate our results, we conducted ground surveys using GPS, a geological hammer, and a digital camera. Results reveal several mineral deposits, including clay (65%), carbonates (15%), quartz (10%), and ferrous silicate (16%) in the study area. About 45% of minerals show mineral alterations, particularly the clay and quartz minerals. Several formations from the recent Pleistocene age are observed, including the surficial deposits, Kamlial, Murree, Patala, and Abbottabad formations. With ASTER imagery, the accuracy of the SVM classifier is better than MLC to obtain lithological classes with overall kappa statistics (0.86 Versus 0.72), respectively. Overall, the SVM classifier outperformed when used with ASTER imagery. Separate rock samples are tested in the laboratory to validate the minerals mapped from remote sensing. We obtained 90 to 95% accuracy for the mapped minerals. The present study presents a simple approach for mapping poorly exposed lithology in inaccessible regions.