Spatio-temporal copper prospectivity in the American Cordillera predicted by positive-unlabeled machine learning

Plain Language Summary

Copper is essential for modern technology — from electrical wiring to electric vehicles and renewable energy systems. Most of the world’s copper comes from porphyry copper deposits, which form in volcanic arcs above subduction zones where one tectonic plate dives beneath another. However, predicting where undiscovered deposits might be hiding remains a major challenge because the geological factors that control their formation are complex and poorly understood.

This study applies a novel machine learning approach called positive-unlabeled (PU) learning to predict where porphyry copper deposits are most likely to be found along the American Cordillera — the mountain chain stretching from Alaska to Patagonia. Unlike traditional methods that require examples of both deposit and non-deposit locations, PU learning only needs the known deposit locations, making it well suited to mineral exploration where “absence of discovery” does not mean “absence of deposit.”

The resulting prospectivity maps highlight several areas with high predicted potential for copper mineralization that have no known deposits — presenting new targets for exploration. The analysis also reveals that thick arc crust, fast plate convergence, and a supply of volatile fluids into the subduction system are the primary ingredients for forming porphyry copper deposits, though the relative importance of these factors varies between North and South America.