Sofia Mantilla Salas, PhD
S. Mantilla Salas1, P. Mejia2, J. Klockner1, A. Asadi1, D. Zhen Yin1, J. Caers1
1Mineral-X, Stanford University, Stanford, CA, USA
2Ero Copper Corp, Vancouver, BC, Canada
This study demonstrates how geophysical data from a single known commercial deposit can be leveraged to efficiently identify similar deposits in unexplored regions. The method automatically pinpoints target exploration polygons whose patterns closely resemble those of the known deposit by analyzing multivariate geophysical signatures from an existing commercial deposit. This approach offers a fast, automated solution that significantly accelerates mineral exploration while remaining highly efficient and scalable.
Using public domain magnetic data (Total Magnetic Intensity, TMI) from the Carajás region in Brazil—a highly prospective area for mineral deposits—we developed a novel workflow that applies Spatial Mixed Principal Component Analysis (PCA) to sliding windows of varying sizes and orientations. This enables the extraction of key spatial patterns, such as the distribution of magnetic anomalies or the orientation of geological structures. The similarity of these patterns is then measured across the region to highlight areas (target polygons) that exhibit geophysical characteristics akin to the known deposit—demonstrated here with the Paulo Afonso Ni-Cu deposit.
This study makes several key contributions. First, it requires minimal training data, relying on only a single known economic deposit to detect additional deposits. Second, it replaces manual, time-intensive interpretation with an automated process, enabling rapid and resource-efficient exploration. Third, the method is highly scalable, allowing application over vast areas and extension to various spatial data types, including lithological and structural data. Lastly, the workflow’s predictive power is demonstrated by successfully identifying other known deposits in the Carajás region that were excluded from training.
In contrast to traditional manual approaches, which can take months, and machine learning methods, which typically require large datasets, this technique offers a scalable and efficient alternative. Streamlining the discovery of mineral deposits holds significant promise for advancing exploration strategies and meeting critical resource demands tied to the global energy transition.