Capturing equatorial Pacific variability with multivariate Sr‐U coral thermometry

Abstract

Sr-U, a coral-based paleothermometer, corrects for the effects of Rayleigh Fractionation on Sr/Ca by regressing multiple, paired U/Ca and Sr/Ca values. Prior applications of Sr-U captured mean annual sea surface temperatures (SSTs), inter-annual variability, and long-term trends. However, because many Sr/Ca-U/Ca pairs are needed for a single Sr-U value as originally formulated, the temporal resolution of the proxy is typically limited to 1 year. Here, we address this limitation by applying laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS) to three Porites colonies from Jarvis and Nikumaroro Islands in the central equatorial Pacific (CEP), generating ∼25 Sr/Ca-U/Ca pairs per month of skeletal growth. Both Sr/Ca and U/Ca vary significantly over small (sub-mm) length scales and support the calculation of Sr-U values using the original regression method. Over the represented temperature range of 24–31°C, the Sr/Ca-U/Ca-SST relationships are nonlinear, a finding consistent with predictions of the Rayleigh model. To reflect this non-linearity, we developed a calibration using multivariate nonlinear regression. The multivariate, three-coral calibration was applied to 20 years of monthly resolved Sr/Ca and U/Ca of a coral interval not included in the calibration, yielding RMSE = 0.73°C and r2 = 0.85 (p < 0.05; df = 256). The multivariate calibration performed significantly better than Sr/Ca alone (r2 = 0.28). Applying the new calibration to a subfossil Porites from Kiritimati Atoll, CEP (2200 Before Present) yields equivalent phase and amplitude of interannual variability, but water temperatures ∼1.6°C cooler than they are in this region today.