Tracking Cu‐Fertile Sediment Sources via Multivariate Petrochronological Mixture Modeling of Detrital Zircons

Whereas the ability to acquire petrochronological data from detrital minerals has exploded, development of tools to analyze and interpret the multivariate data sets has not kept pace. Herein, we present a case study, which applies the recently developed non‐negative Tucker‐1 decomposition (NNT1) method to a multivariate detrital zircon data set from till samples collected above the Cu‐bearing Guichon Creek Batholith in southern British Columbia, Canada. Zircon composition variables that we consider include age, Ce anomaly, Ce N /Nd N , Dy N /Yb N , ΔFMQ, Eu anomaly, ΣHREE/ΣMREE, Hf, Th/U, Ti temperature, and Yb N /Gd N . The NNT1 approach successfully deconvolves the multivariate data set into two endmembers, which are consistent with derivation either from non‐oxidized and relatively anhydrous (i.e., low Cu‐ore potential, Source 1) or oxidized and hydrous (i.e., potential Cu‐ore bodies, Source 2) igneous rocks. Furthermore, we attribute each of the zircon grains to either the Source 1 or 2 endmember based on maximization of the likelihood that their measured multivariate geochemistry was drawn from one or the other of the learned multivariate endmembers. Finally, we demonstrate that the proportions of the Source 2 endmember decrease with increasing distance from the ore bodies, as expected due to down‐ice or off‐axis zircon mixing and dilution. We conclude that the NNT1 approach provides insight into geologically meaningful sediment transport processes and multivariate sediment sources even when those sources are unknown. It thus provides a basis for future petrochronological interpretations with applied and pure geoscience applications.