Matheus Cardoso Pavon, MSc
M.C. Pavon1, E. Gloaguen1, V.S. dos Santos1
1Water Earth Environment Center, National Institute of Scientific Research, Québec, Québec, Canada
In the initial phase of mineral exploration, one key source of information are geological maps, which help locate potential prospection targets. They are usually produced through ground truth observations and, when that is not available, by extrapolating the geological data through conceptual or numerical methods (e.g. defining lithological boundaries by interpreting aeromagnetic maps or through indicator kriging). They usually rely on one or few data sources or assume linearity between variables, which may limit their capability to faithfully model the complexity of real geological data. Artificial intelligence offers a promising tool to address this issue. In particular, deep learning algorithms can process multivariate geospatial sources of information and infer non-linear relationships hidden in complex datasets. The SCB-Net, a recently developed deep learning algorithm using convolutional neural networks, has demonstrated success in generating predictive lithological maps by integrating field samples with remote sensing and geophysical data. It leverages publicly accessible multidisciplinary datasets from Google Earth Engine and from Québec Spatial Reference Geomining Information System, SIGÉOM. However, the accuracy of its predictions in regions with limited geoscientific data, which is often the case of mineral exploration frontiers, still requires investigation. In order to test the algorithm’s predictive capability in data-sparse regions, this study aims in training the SCB-Net in a location with high-density data and apply that model in an understudied area. The training dataset will be the well-documented area of the Grenvillian Central Metasedimentary Belt. The model will then be tested in the metasedimentary Wakeham Group, in the allochthonous northeastern section of the Grenville Province. Despite limited available data, many field campaigns documented its enrichment in REE and potential to host IOCG-type and skarn-type mineral deposits with Co, Cu, Pb and Zn, strategic elements of surging demand due to the ongoing global energetic and digital transition.