Abstract
Abstract Granites are widely distributed, phaneritic igneous rocks renowned for their high compressive strength and durability, making them a premier choice for dimension stone applications. This study aims to detect and map granitic complexes using geospatial data and spectral algorithms within the arid and semi-arid environment of northwestern Parwan Province, northeastern Afghanistan. Also, to establish an effective and optimized supervised classification approach specifically tailored for identifying granitic complexes in similar terrains. This study utilizes FCC imagery to highlight lithology and employs various machine learning algorithms, including ML, MD, SVM, and SAM, for mapping granitic complexes within the study area. Training and test data were collected from field observations, Google Earth imagery, and geological maps. Our analysis identified two primary granitic complexes within the study area, measuring approximately 19 km × 13 km2 and 7 km × 3 km2, respectively. Ground truth data validation yielded an accuracy of 75%, indicating a positive correlation between the predicted and observed distributions. This enhanced understanding of granite distribution can serve as a valuable guide for future exploration endeavors targeting metallic and non-metallic resources, including aluminum, iron, manganese, rare earth elements, tungsten, etc.