High-Resolution and Rapid Estimation of P2O5 and MgO in Sedimentary Phosphate Cores Using Infrared Hyperspectral Imaging

Summary Recent advancements in sensor technologies and machine learning have transformed geological exploration, particularly through hyperspectral infrared core scanning, which enables rapid, cost-effective mineralogical and geochemical analysis. This study explores the potential of infrared hyperspectral core imaging combined with portable XRF (p-XRF) to predict key elements like P2O5 and MgO in a 65-meter sedimentary phosphate drill core from Ben Guerir, Morocco. Using spectral and geochemical data from 100 points, stacking models with various regression algorithms were trained and optimized through feature selection. The models were evaluated on 330 data points and used to generate geochemical maps, demonstrating a fast, non-destructive, and efficient approach to trace lateral and vertical geochemical variations at a millimetric scale.