Deciphering Precambrian Litho-structural settings and barite-associated mineralization through integrated hyperspectral and geophysical datasets

Effective lithological mapping and mineral exploration remain challenging in geologically complex and underexplored Precambrian terrains. This is particularly relevant to the Arabian–Nubian Shield (ANS), a Neoproterozoic orogenic belt extending across northeastern Africa and Arabia. This study integrates Landsat 8/9, Sentinel-2, ASTER, PRecursore IperSpettrale della Missione Applicativa (PRISMA) hyperspectral imagery, gravity-derived structural analysis, supervised machine learning, and field observations to characterize lithology, alteration, and structural controls relevant to barite exploration in the Tehilla–Abu Samar area, Red Sea Hills, northeastern Sudan. Band ratios, principal component analysis, false-colour composites, Minimum Noise Fraction transformation, and Sequential Maximum Angle Convex Cone analysis were used to enhance lithological and mineral-sensitive spectral variations. Five classifiers—support vector machine (SVM), artificial neural network, random forest, k-nearest neighbors, and eXtreme Gradient Boosting (XGBoost)—were evaluated for lithological classification. XGBoost achieved the highest overall accuracy (81.49%) and Kappa coefficient (0.792), closely followed by SVM (81.40%; κ = 0.791), while SVM yielded a marginally higher macro-F1 score. Gravity anomaly and edge-enhancement analyses delineated major NE–SW and NW–SE structural trends and density contrasts associated with intrusive bodies and deformation zones. Integration of independently derived lithological, spectral, structural, and field evidence established felsic metavolcanic rocks as the principal barite-hosting lithology and provided a coherent framework for prioritizing barite-associated geological targets in this structurally complex Precambrian terrane.