Throughout Earth's history, CO2 is thought to have exerted a fundamental control on environmental change. Here we review and revise CO2 reconstructions from boron isotopes in carbonates and carbon isotopes in organic matter over the major climate transition of the past 66 million years. We find close coupling between CO2 and climate throughout the Cenozoic, with peak CO2 levels of ∼1,500 ppm in the Eocene greenhouse, decreasing to ∼550 ppm in the Miocene, and falling further into ice age world of the Plio–Pleistocene. Around two-thirds of Cenozoic CO2 drawdown is explained by an increase in the ratio of alkalinity to dissolved inorganic carbon, likely linked to a change in the balance of weathering to outgassing, with the remaining one-third due to changing ocean temperature and major ion composition. Earth system climate sensitivity is explored and may vary between different time intervals. The Cenozoic CO2 record highlights the truly geological scale of anthropogenic CO2 change: Current CO2 levels were last seen around 3 million years ago, and major cuts in emissions are required to prevent a return to the CO2 levels of the Miocene or Eocene in the coming century. ▪ CO2 reconstructions over the past 66 Myr from boron isotopes and alkenones are reviewed and re-evaluated. ▪ CO2 estimates from the different proxies show close agreement, yielding a consistent picture of the evolution of the ocean-atmosphere CO2 system over the Cenozoic. ▪ CO2 and climate are coupled throughout the past 66 Myr, providing broad constraints on Earth system climate sensitivity. ▪ Twenty-first-century carbon emissions have the potential to return CO2 to levels not seen since the much warmer climates of Earth's distant past. Expected final online publication date for the Annual Review of Earth and Planetary Sciences, Volume 49 is May 2021. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Sea level and deep-sea temperature variations are key indicators of global climate changes. For continuous records over millions of years, deep-sea carbonate microfossil–based δ18O (δc) records are indispensable because they reflect changes in both deep-sea temperature and seawater δ18O (δw); the latter are related to ice volume and, thus, to sea level changes. Deep-sea temperature is usually resolved using elemental ratios in the same benthic microfossil shells used for δc, with linear scaling of residual δw to sea level changes. Uncertainties are large and the linear-scaling assumption remains untested. Here, we present a new process-based approach to assess relationships between changes in sea level, mean ice sheet δ18O, and both deep-sea δw and temperature and find distinct nonlinearity between sea level and δw changes. Application to δc records over the past 40 million years suggests that Earth’s climate system has complex dynamical behavior, with threshold-like adjustments (critical transitions) that separate quasi-stable deep-sea temperature and ice-volume states.

The Middle Eocene Climatic Optimum (MECO) was a gradual warming event and carbon cycle perturbation that occurred between 40.5 and 40.1 Ma. A number of characteristics, including greater-than-expected deep-sea carbonate dissolution, a lack of globally coherent negative δ 13 C excursion in marine carbonates, a duration longer than the characteristic timescale of carbon cycle recovery, and the absence of a clear trigger mechanism, challenge our current understanding of the Earth system and its regulatory feedbacks. This makes the MECO one of the most enigmatic events in the Cenozoic, dubbed a middle Eocene "carbon cycle conundrum." Here we use boron isotopes in planktic foraminifera to better constrain pCO 2 changes over the event. Over the MECO itself, we find that pCO 2 rose by only 0.55-0.75 doublings, thus requiring a much more modest carbon injection than previously indicated by the alkenone δ 13 C-pCO 2 proxy. In addition, this rise in pCO 2 was focused around the peak of the 400 kyr warming trend. Before this, considerable global carbonate δ 18 O change was asynchronous with any coherent ocean pH (and hence pCO 2 ) excursion. This finding suggests that middle Eocene climate (and perhaps a nascent cryosphere) was highly sensitive to small changes in radiative forcing.Plain Language Summary Geoscientists often look to periods of global warming in the geological past to understand how the Earth responds to input of atmospheric CO 2 . However, during the Middle Eocene Climatic Optimum (or MECO) 40 million years ago, the Earth did not respond in the way one would expect, given what we know from these earlier warming events. The MECO poses a number of puzzles for geoscientists relating to what caused it and why the Earth system responded in the way it did. Before we can hope to answer these questions, however, we need to know what atmospheric CO 2 levels were in the middle Eocene and how much they changed over the MECO event. Here we use boron isotope ratios in fossil plankton shells to tell us how ocean pH (which predominantly reflects CO 2 levels) changed over the MECO. We show that relatively little change in CO 2 at this time were associated with large-scale changes in climate. This suggests that during the Eocene, when CO 2 levels were similar to those likely to be reached by the end of this century, the Earth's climate (and possibly ice sheets) was very sensitive to minor disturbances.