Advanced CNN-based remote sensing for mineral mapping of porphyry systems in the Gilgit region

Purpose. This study aims to improve mineral identification in porphyry hydrothermal alteration zones, particularly in the challenging terrain of the Gilgit area, by combining remote sensing data with Convolutional Neural Networks (CNNs). Methods. Landsat 8 Collection 2 Level 1 imagery from the United States Geological Survey (USGS) was processed using ENVI 5.3 software. Spectral Angle Mapping (SAM) classification was applied to identify alteration minerals. The dataset was then augmented, normalized, and split into training (75%), validation (15%), and testing (15%) sets. A CNN model incorporating convolutional, pooling, and fully connected layers was developed to perform binary classification of mineral compositions. Findings. Advanced CNN-based remote sensing techniques have demonstrated significant potential in mapping porphyry systems. The CNN model achieved over 90% classification accuracy for minerals like feldspar and chalcocite, based on their spectral properties and dominant color features. This approach is beneficial in challenging terrains like Gilgit, where traditio-nal methods can be difficult and expensive. Originality. This study demonstrates the successful integration of remote sensing data with CNN-based algorithms for accurate mineral classification, providing a novel approach to overcoming the limitations of conventional field-based methods in challenging terrains. Practical implications. The approach provides a practical and efficient solution for remote mineral exploration, particularly in regions with limited accessibility, supporting more accurate and faster geological assessments in the Gilgit area.