BAYSPLINE: A New Calibration for the Alkenone Paleothermometer

Abstract

The alkenone-based U-37(K') proxy is a cornerstone of paleoclimatology, providing insight into the temperature history of the Earth's surface ocean. Although the relationship between U-37(K') and sea surface temperatures (SSTs) is robust and well supported by experimental data, there remain outstanding issues regarding the seasonality of production of alkenones and the response of U-37(K') at very warm and cold SSTs. Using a data set of over 1,300 core-top U-37(K') measurements, we find compelling evidence of seasonal production in the North Atlantic, North Pacific, and Mediterranean Oceans. We also find significant attenuation of the U-37(K') response to SST at warm temperatures (> 24 degrees C), with the slope reduced by nearly 50% as U-37(K') approaches unity. To account for these observations in a calibration, we develop a new Bayesian B-spline regression model, BAYSPLINE, for the U-37(K') paleothermometer. BAYSPLINE produces similar estimates as previous calibrations below similar to 24 degrees, but above this point it predicts larger SST changes, in accordance with the attenuation of the U-37(K') response. Example applications of BAYSPLINE demonstrate that its treatment of seasonality and slope attenuation improves paleoclimatic interpretations, with important consequences for the inference of SSTs in the tropical oceans. BAYSPLINE facilitates a probabilistic approach to paleoclimate, building upon growing efforts to develop more formalized statistical frameworks for paleoceanographic reconstruction. Plain Language Summary "Alkenones" are lipids (fats) made by marine phytoplankton. The plankton alter the degree of unsaturation in these lipids in response to sea surface temperature (SST), producing more unsaturated compounds in colder water. These lipids are well preserved in marine sediments, such that paleoclimatologists can measure the unsaturation and determine what SSTs were in the past. In this manuscript, we review the calibration of this powerful "paleothermometer" and propose a new model that uses spline fits and Bayesian regression. We find that the new model improves our ability to estimate past SSTs, especially in the tropical oceans.