Machine learning in subsurface physical properties and lithofacies prediction in a mining context
To address energy challenges linked to decarbonization, geosciences are focusing more on advancing mineral resource exploration and exploitation. This study uses the Iberian Pyrite Belt, one of the world’s largest metallogenic provinces, as a test site to: (a) develop predictive models of physical p...
| Autores: | , , , , , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2025 |
| País: | España |
| Institución: | Consejo Superior de Investigaciones Científicas (CSIC) |
| Repositorio: | DIGITAL.CSIC. Repositorio Institucional del CSIC |
| OAI Identifier: | oai:digital.csic.es:10261/410010 |
| Acceso en línea: | http://hdl.handle.net/10261/410010 |
| Access Level: | acceso abierto |
| Palabra clave: | Machine learning Rock properties prediction Lithofacies classification Massive sulfide deposits Exploration mining and Iberian pyrite belt |
| Sumario: | To address energy challenges linked to decarbonization, geosciences are focusing more on advancing mineral resource exploration and exploitation. This study uses the Iberian Pyrite Belt, one of the world’s largest metallogenic provinces, as a test site to: (a) develop predictive models of physical properties of rock (PPR) and (b) classify lithological units based on these PPR. Over 1,000 surface rock samples and six boreholes from the Riotinto mine were analyzed, providing a comprehensive dataset. A robust quality control process, aided by Machine Learning models (ML), refined the PPR data, ensuring accuracy and reliability. Both traditional statistical models and cuttingedge ML models were used for PPR prediction and lithofacies classification. Our study revealed that geological evolution can lead to significant overlaps in PPR across different lithologies, making traditional models insufficient for accurate predictions. However, ML models, such as Random Forest, XGBoost, k-Nearest Neighbors, and Support Vector Regression, demonstrated over 80% accuracy in predicting PPR and classifying lithofacies. This approach redefines how lithofacies are identified and establishes an innovative methodology for subsurface lithological characterization. Results highlight the potential of ML models in mining and geology, paving the way for more accurate 3D characterization of lithological units through integrating geophysical data and direct measurements. |
|---|