A statistical interpretation of surface ocean temperature trends during the Paleocene-Eocene Thermal Maximum

often considered a geologic analog to the modern, rapid addition of carbon dioxide to the atmosphere. The ability to assess the sensitivity of many rate-dependent climate change impacts, based on the PETM analogy, hinges critically on the PETM warming rate. Current estimates of the PETM warming rate face considerable problems, for example, due to the nontrivial statistical challenges in accounting for the key known uncertainties in a mathematically sound way. Here we introduce a new method for quantifying the PETM warming rate recorded by the oxygen isotopes of single-specimen surface-dwelling foraminifera at Southern Ocean ODP Site 690. We use Monte Carlo samples from the Bayesian predictive distribution of reconstructed climate histories to estimate the probability distributions of millennial temperature trends. Our approach produces probabilistically sound hindcasts of the observations and the implied rates of change that comprehensively account for the combined effects of uncertainties in (i) proxy measurement, (ii) age estimates, (iii) long term rates of temperature change, (iv) the autocorrelated natural temperature variability, and (v) the conversion from proxy to temperature. We test the sensitivity of the reconstruction to the choice of chronology by comparing the results derived from two independent chronologies, based on orbital cyclostratigraphy and extraterrestrial helium-3 (3 He) accumulation. We estimate the peak millennial-scale warming rate (at the onset of the PETM) as 1.4 °C/kyr (with a 95% credible interval from-2.3 to 5.4 °C/kyr) using orbital chronology and 1.1 °C/kyr (95% credible interval from-2.5 to 4.9 °C/kyr) using 3 He age control. By comparison, a suite of IPCC AR4 AOGCM model projections for the 21 st century predict ocean surface warming rates at the Southern Ocean site ranging from-4 to 14 °C/kyr, when extrapolated from century to millennial scale rates, with half of the models projecting rates of less than 3 °C/kyr. Although the paleo-reconstruction method cannot resolve century scale changes, we conclude that Site 690 data support a PETM peak warming rate of a magnitude potentially comparable to (possibly only 2-3 times slower than) the rate of warming projected for the next century. An unresolved caveat is whether the onset of the PETM at Site 690 is truncated due to a brief period of dissolution, leading to a spuriously high apparent rate of warming.