Sea surface temperature evolution of the North Atlantic Ocean across the Eocene-Oligocene Transition

A major step in the long-term Cenozoic evolution toward a glacially-driven climate occurred at the Eocene Oligocene Transition (EOT), ~34.44 to 33.65 million years ago (Ma). Evidence for high latitude cooling and increased latitudinal temperature gradients across the EOT has been found in a range of marine and terrestrial environments. However, the timing and magnitude of temperature change in the North Atlantic remains highly unconstrained. Here, we use two independent organic geochemical paleo-thermometers to reconstruct sea surface temperatures (SSTs) from the southern Labrador Sea (Ocean Drilling Program – ODP Site 647) across the EOT. We find a permanent cooling step of ~3 °C (from 27 to 24 °C), between 34.9 Ma and 34.3 Ma, which is ~500 kyr prior to Antarctic glaciation. This step in SST values is asynchronous across Atlantic sites, signifiying considerable spatiotemporal variability in SST evolution. However, it is part of an overall cooling observed across sites in the North Atlantic (NA) in the 5 million years bracketing the EOT. Such cooling is unexpected in light of proxy and modelling studies suggesting the startup or strengething of the Atlantic Meridional Overturning Circulation (AMOC) before the EOT, which would warm the NA, although parallel Eocene CO2 decline on the decent into the Oligocene icehouse might counter this feedback. Here we show, using a published modelling study, that a reduction in atmospheric CO2 from 800 to 400 ppm is not sufficient to produce the observed cooling, if combined with NA warming from an AMOC startup, simulated here through Arctic closure from the Atlantic. Possible explanations of the apparent discrepancy are discussed and include uncertainty in the SST data, paleogeography and atmospheric CO2 boundary conditions, model weaknesses, and an earlier AMOC startup that just strengthened at the EOT. The results highlight the remaining uncertainty in many aspects of data and modelling results which need to be improved before we can draw robust conclusions of the processes acting before and across the EOT.