Answers are grounded in this paper and cite their sources.
Shows the effectiveness of machine learning methods for mapping REE-bearing bauxite, indicating improved mineral exploration efficiency.
Modeling mineral deposits is a crucial step in assessing and managing natural resources. Surficial features are essential guides to prospect supergene mineralization in large areas, thus making mineral exploration surveys less expensive and time-consuming. In this context, Remote Sensing has proven to be a powerful tool for mapping supergene ores, especially when supported by other data. This study investigates the potential of supervised classification and machine learning techniques, namely the Spectral Angle Mapper (SAM) and the Gradient Tree Booster (GTB) algorithm, to map and model rare earth element (REE)-bearing bauxite. We aim to develop an accurate model for identifying and mapping REE-bearing bauxite occurrences developed over Archean granulitic rocks from the São Francisco Craton, Brazil. The research comprises multisource data: Landsat-8 orbital images, topographic, gamma-spectrometric (airborne and in samples), spectroradiometric, and field data. The applied methodological approach includes field sample collection, spectral data measuring and processing, calculation of spectral indices such as the Lateritic and the Regolithic indices, and integration into supervised classification and machine learning models, SAM and GTB. The analyses were performed in Google Earth Engine. Complementary outcomes were achieved with SAM and GTB, with the former more effective for residual and the latter for colluvial bauxite. The GTB results indicate that all analyzed random combinations exhibit similar distributions, suggesting consistently high model accuracy for different tested configurations, with a median above 96% accuracy in each evaluated group, regardless of data variation. The test accuracy increases to a maximum of five analyzed features and stabilizes thereafter, with minimal variability, defining the five main features that contributed most to predicting bauxite areas and providing valuable insights into the most influential variables in mapping. The results indicate that integrating multisource data with advanced machine learning techniques enables precise identification of areas with potential for REE-bearing bauxite. The Gradient Tree Booster model demonstrated high predictive capability, standing out as an effective tool for mineral mapping and prospecting.
No takes yet. Share an insight, caveat, or question.
Torres et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: