A BAYESIAN APPROACH TO ZIRCON AGE INTERPRETATION

The common accessory mineral zircon is often used to constrain the timing of igneous processes such as magma crystallization or eruption. However, zircon U/Pb ages strictly date zircon crystallization, which is not an instantaneous process. Zircon saturation calculations provide a link between zircon crystallization, temperature, and melt fraction, allowing for the estimation of temporal zircon crystallization distributions. Such distributions provide an informative prior for Bayesian estimates of magma eruption time, and allow us to compare the relative accuracy of common weighted-mean and youngest-zircon age interpretations through forward zircon crystallization simulations.

The accuracy of weighted-mean and youngest-zircon estimates of eruption age varies as a function of magmatic zircon crystallization timescale relative to analytical uncertainty (ΔT/σ) and number of zircons analyzed (N). However, both interpretations carry a risk of substantially underestimating the uncertainty in eruption age; this risk is generally lower at low N for ΔT/σ < 5. The spread in observed zircon ages as characterized by statistics such as the mean square of weighted deviates (MSWD) provides some insight into ΔT/σ, but a low MSWD does not necessarily guarantee the accuracy of weighted mean uncertainties, particularly at high N. Consequently, analytical advances resulting in increased data volume and/or decreased analytical uncertainty provide better resolution of magmatic processes, but require care to avoid underestimating the uncertainty of interpreted eruption and crystallization ages.