Bayesian Calibration of the Mg/Ca Paleothermometer in Planktic Foraminifera

The Mg/Ca ratio of planktic foraminifera is a widely used proxy for sea-surface temperature but is also sensitive to other environmental factors. Previous work has relied on correcting Mg/Ca for nonthermal influences. Here, we develop a set of Bayesian models for Mg/Ca in four major planktic groups—Globigerinoides ruber (including both pink and white chromotypes), Trilobatus sacculifer, Globigerina bulloides, and Neogloboquadrina pachyderma (including N. incompta)—that account for the multivariate influences on this proxy in an integrated framework. We use a hierarchical model design that leverages information from both laboratory culture studies and globally distributed core top data, allowing us to include environmental sensitivities that are poorly constrained by core top observations alone. For applications over longer geological timescales, we develop a version of the model that incorporates changes in the Mg/Ca ratio of seawater. We test our models—collectively referred to as BAYMAG—on sediment trap data and on representative paleoclimate time series and demonstrate good agreement with observations and independent sea-surface temperature proxies. BAYMAG provides probabilistic estimates of past temperatures that can accommodate uncertainties in other environmental influences, enhancing our ability to interpret signals encoded in Mg/Ca.