Machine Learning and Singularity Analysis Reveal Zircon Fertility and Magmatic Intensity: Implications for Porphyry Copper Potential

Yuanzhi Zhou,Zhenjie Zhang,Jie Yang,Yunzhao Ge,Qiuming Cheng

DOI: https://doi.org/10.1007/s11053-022-10122-y

IF: 5.609

2022-10-04

Natural Resources Research

Abstract:The chemical compositions of zircons have been used as magma fertility indicators for porphyry copper mineralization potential. Here, we trained and tested two machine learning (ML) models: support vector machines and random forest to classify the metallogenic fertility (fertile vs. unfertile) based on igneous zircon chemistry. Both models achieved good classification accuracy (~ 93%) against the testing dataset, suggesting the feasibility and effectiveness of using the ML methods to discriminate whether zircons are from fertile or unfertile magmas. The feature importance analysis of the RF classifiers identified that the most important parameters of fertile magma are Eu, Ce, and Dy, further validating previous studies of the genesis of porphyry copper deposit-related magma. As a case study, we predicted the metallogenic fertility of the detrital zircons in the eastern Gangdese belt in southern Tibet by applying the ML models trained by igneous zircon labeled data. Our ML models suggest that the magmas that formed after 50 Ma, especially from 30 to 10 Ma, have higher fertile percentages. Taking this together with the local singularity analysis of detrital zircon U–Pb ages, we suggest that magmatic flare-ups with high FP and singularity index values are more favorable for porphyry copper formation. Our ML models and defined FP may help reveal temporal variation of the favorability for porphyry copper formation.

geosciences, multidisciplinary