SMITE: A NOVEL METHOD FOR ESTIMATING SEA SURFACE TEMPERATURES USING SCLERACTINIAN CORALS

Scleractinian coral skeletons are a premier archive for reconstructing tropical climate variability from seasonal to millennial timescales. Most coral-based paleotemperature studies rely on a calibration of a single geochemical proxy to sea surface temperature (SST) using modern observations. However, it is well-known that these geochemical SST proxies can be influenced by metabolic processes (‘vital effects’) and environmental signals not directly related to SST. Here we present a novel multivariate model for estimating SST while simultaneously constraining vital effects from an array of proxies with varying degrees of temperature sensitivity (δ¹¹B, B/Ca, Ba/Ca, Li/Ca, Li/Mg, Mg/Ca, Sr/Ca, U/Ca). Our Scleractinian Multivariate Isotope and Trace Element (SMITE) model utilizes a unique combination of spectral and multivariate statistical methods to leverage the covariance of SST across multiple, potentially noisy geochemical proxies. Once the temperature signal is accounted for, vital effects are assessed as a secondary axis of common variability. The SMITE model is applied to two different coral species (Porites asteroides and Siderastraea siderea) from two different locations in the Tropical Western Atlantic (Bermuda and Belize, respectively). Both corals were sampled via laser ablation inductively coupled mass spectrometry (LA-ICP-MS) at sub-millimeter resolution. Total number of months sampled for P. asteroides and S. siderea was 36 and 16, respectively. Estimated SST derived from the SMITE model significantly outperform those derived from the best single-proxy estimators (r²_SMITE = 0.92, r²_Sr/Ca ≤ 0.84, n = 52). Sr/Ca, U/Ca, Li/Mg, and B/Ca all display strong loadings (> 0.25) onto the SST model. δ¹¹B, Li/Ca, and Mg/Ca display strong loadings onto the suspected vital effect component, suggesting that these proxies are influenced by changes to pH or pH-induced calcification rates. However, SMITE reconstruction skill is highly sensitive to the length of the calibration period and accuracy of the age model. Future work will focus on (1) how SMITE model parameters change among corals of different species and genera across ocean basins; and (2) the impact of sampling and instrumentation methods (e.g., LA-ICP-MS vs. micromilling and solution analyses).