Unsupervised machine learning-based prospectivity analysis of NW and NE India for carbonatite-alkaline complex-related REE deposits

Abstract

Rare Earth Elements (REE) form critical raw materials in environment-friendly, high-technology devices and components, and therefore have been classified as “critical minerals and metals” by most countries. About 62% of the global resources of REE occur associated with carbonatite-alkaline complexes; however, the entire production of REE in India currently comes from secondary deposits, even though India contains a variety of REE-enriched primary source rocks, particularly carbonatites and alkaline complexes. There is, therefore, a significant potential in the county for new REE deposit discoveries associated with carbonatite-alkaline complexes. This research attempts to identify exploration targets for REE associated with carbonatite-alkaline complexes in northern India utilising a Self-Organising Maps (SOM)-driven workflow. This unsupervised machine-learning-based workflow eliminates the hand-crafting of input predictor features. The algorithm creates clusters of features directly from primary gridded geophysical and topographical datasets. The obtained clusters are then analysed based on available geological knowledge and empirical spatial associations with known occurrences in the study areas to identify prospective clusters and generate prospectivity maps. Nine new targets are identified across the Shillong plateau in northeastern and Western Rajasthan in northwestern India. These new targets, in addition to the known carbonatite-alkaline complexes, are recommended for further data collection and follow-up exploration. It is noteworthy that these targets conform to the targets identified by Aranha et al. (2022a, 2022b) using mineral systems-guided fuzzy inference systems.