Divergent estimates of Miocene to Pleistocene upper ocean temperatures in the South Atlantic Ocean from alkenone and coccolith clumped isotope proxies

Abstract

Estimates of surface ocean temperatures in the past are essential for evaluating the sensitivity of Earth's surface temperature to higher atmospheric CO 2 concentrations such as those characterizing the Miocene. However, in the higher latitude regions, many proxy-based temperature estimates suggest extreme warmth, which imply much lower latitudinal temperature gradients than can be simulated by most coupled general circulation climate models under enhanced greenhouse gas forcing. This discrepancy implies either systematic biases in temperature proxy interpretation or the absence of key feedback processes in models. Here, we use a new approach to estimate high southern latitude surface ocean temperatures using clumped isotope thermometry in coccoliths – calcite plates precipitated in the surface ocean by the calcifying phytoplankton group coccolithophores. We present new determinations of the clumped isotope ratio in well-preserved coccoliths spanning the last 15 million years, extracted and purified from a sediment core located just south of the modern subtropical front (Ocean Drilling Program Site 1088, 41° S). Coccolith clumped isotopes reveal a 10 °C decline in temperatures at this location over the last 15 million years, and over the last 11 Ma of overlapping records the magnitude of cooling is similar to that estimated from the degree of undersaturation of alkenone biomarkers. However, the temperatures derived from coccolith clumped isotopes are 8–12 °C cooler than those estimated from alkenones, even though both are biosynthesised by the same organisms and therefore must reflect an identical production depth and season. This implies that some of the model-proxy mismatch may be due to unresolved issues in proxy interpretation. We propose that at this site, calibration biases lead to alkenone sea surface temperature estimates up to 5 °C too warm, whereas coccoliths reflect temperatures at the production depth which is several degrees cooler than the sea surface. The influence of secondary diagenetic carbonate precipitation at the seafloor is constrained to contribute a cold bias of 2 °C or less on the clumped isotope temperature for most samples.