APPLICATION OF REMOTE SENSING AND DEEP LEARNING FOR GOLD DEPOSIT IDENTIFICATION: A CASE STUDY OF KANKOSSA, MALI

Gold exploration in the Birimian terranes of West Africa continues to face significant challenges due to geological complexity and the limitations of conventional exploration methods. This study evaluates the contribution of remote sensing and artificial intelligence to the identification of gold deposits in the Kankossa region of Mali. Multispectral satellite data from Lands at 8 OLI and Sentinel-2 MSI, combined with geophysical, geochemical, and topographic datasets, were analyzed using machine learning and deep learning algorithms, including Random Forest, Convolutional Neural Networks (CNNs), and unsupervised clustering techniques. The results demonstrate a significant improvement in lithological classification and the detection of major tectonic structures controlling gold mineralization, achieving an overall classification accuracy exceeding 89%. The integration of multi-source datasets enabled the production of a gold prospectivity map highlighting several high-potential target areas that are consistent with known gold occurrences. This integrated approach provides an effective decision-support tool for mineral exploration, contributing to cost optimization and reducing uncertainty in geologically complex environments.