Building a Knowledge-based and Machine Learning Framework to analyze Critical Mineral Systems

Critical mineral systems evolve through complex processes that are driven by gravity and/or heat and stress gradients in various geotectonic and geological settings. We are developing a framework that combines knowledge modeling and machine learning to characterize computationally the essential geochemical and geological/tectonic processes in different settings inside and on the Earth crust. The Critical Minerals Ontology (CMO), designed based on the mineral system paradigm, extends the top- and mid-level Basic Formal Ontology (BFO) and Common Core Ontologies (CCO). The CMO semantically represents the rock and fluid sources, pathways that allow transport of metal-carrying fluids, forcings that drive the fluids, changes in the properties of the material that participate as input in the critical mineral forming geological processes, and the output of these processes as ore deposits. As a reference ontology, CMO will significantly shorten the time and effort for building interoperable domain and application ontologies from scratch by geologists. The domain and application ontologies that extend CMO will be able to seamlessly integrate, share, and combine critical mineral data from different mineral systems. CMO will help build large-scale computable analyses and facilitate the use of technologies like machine learning/AI to the field of critical mineral exploration. We will instantiate and evaluate CMO with data from rare-earth elements in kaolin, regolith, and heavy mineral sands in the Piedmont and Coastal Plain provinces of Georgia and South Carolina. The integrated data from these and other areas, served by the Critical Minerals Mapping Initiative (CMMI) of the U.S. Geological Survey, Geoscience Australia, and Geological Survey of Canada, will be used as input to apply knowledge-based machine learning algorithms to identify potential occurrences of critical mineral deposits in unexplored areas throughout the U.S.