LETTERS Eocene/Oligocene ocean de-acidification linked to Antarctic glaciation by sea-level fall Agostino Merico 1,2 , Toby Tyrrell 1 & Paul A. Wilson 1 One of the most dramatic perturbations to the Earth system dur- ing the past 100 million years was the rapid onset of Antarctic glaciation near the Eocene/Oligocene epoch boundary 1–3 (,34 million years ago). This climate transition was accompanied 3 by a deepening of the calcite compensation depth—the ocean depth at which the rate of calcium carbonate input from surface waters equals the rate of dissolution. Changes in the global carbon cycle 4 , rather than changes in continental configuration 5 , have recently been proposed as the most likely root cause of Antarctic glaciation, but the mechanism linking glaciation to the deepening of calcite compensation depth remains unclear. Here we use a global bio- geochemical box model to test competing hypotheses put forward to explain the Eocene/Oligocene transition. We find that, of the candidate hypotheses, only shelf to deep sea carbonate partition- ing is capable of explaining the observed changes in both carbon isotope composition and calcium carbonate accumulation at the sea floor. In our simulations, glacioeustatic sea-level fall associated with the growth of Antarctic ice sheets permanently reduces global calcium carbonate accumulation on the continental shelves, lead- ing to an increase in pelagic burial via permanent deepening of the calcite compensation depth. At the same time, fresh limestones are exposed to erosion, thus temporarily increasing global river inputs of dissolved carbonate and increasing seawater d 13 C. Our work sheds new light on the mechanisms linking glaciation and ocean acidity change across arguably the most important climate transi- tion of the Cenozoic era. Evidence for Eocene ice rafting is compelling 6 but, despite specu- lation to the contrary 7 , high-resolution oxygen isotope records 3,8 suggest that Cenozoic ice sheets approaching their modern size were not initiated on Antarctica until the Eocene/Oligocene boundary. Explanations for Antarctic glaciation at the Eocene/Oligocene boundary fall into two main categories: those invoking changes in ocean circulation through tectonic opening of Southern Ocean gate- ways, and those invoking changes in the global carbon cycle, with recent studies supporting the latter 4,9 . Proxy records of atmospheric CO 2 over the past 50 million years (Myr) 10 suggest an overall shift from high levels during the Eocene (,1,000 to ,2,000 parts per million by volume, p.p.m.v.) to lower levels thereafter. Support for a fall in CO 2 levels at the Eocene/Oligocene boundary comes from recent Ocean Drilling Program records 3 . These data show that dee- pening of the calcite compensation depth (CCD) across the Eocene/ Oligocene boundary was rapid and took place in two jumps, in lock- step with the growth of Antarctic ice-sheets and carbon cycle per- turbation—as tracked by changes in benthic foraminiferal calcite d 18 O and d 13 C, respectively. Yet the mechanism teleconnecting the onset of major Antarctic glaciation and CCD deepening is a subject of debate. At least four hypotheses have been invoked to link d 18 O increase (glaciation), d 13 C increase and CCD deepening (carbon cycle perturbation) across this key interval. These are H1, an increase in organic carbon burial rates 2,11–13 ; H2, an increase in weathering of silicate rocks 13–15 ; H3, an increase in global siliceous (versus calcare- ous) plankton export production 3 ; and H4, a shift of global CaCO 3 sedimentation from shelf to deep ocean basins 3,16,17 . To test the power of these hypotheses to explain the Eocene/ Oligocene transition, we have developed a biogeochemical box model of the global ocean (Fig. 1) with long-term control on the carbon cycle given by weathering processes 18 . A full suite of time series are output, including nutrient and phytoplankton concentrations, alkalinity, dis- solved inorganic carbon (DIC), pH, ½CO 2{ 3  for sea water, CCD, d 13 C of burial fluxes plus atmospheric CO 2 . In Fig. 2 we present the para- meters best compared to the high-resolution data sets available: simu- lated CCD and d 13 C in benthic foraminiferal calcite. In some ways H1 represents the obverse scenario to present-day fossil fuel emissions—a period of rapid burial of organic carbon (carbon sequestration) leading to CO 2 extraction from the atmo- sphere–ocean system. Eocene/Oligocene boundary sections in the Southern Ocean 11,19 show increased accumulation rates of opal and benthic foraminifera suggesting that elevated marine organic carbon burial rates contributed to ice-sheet growth via acceleration of the biological pump and CO 2 removal 2,11 . Higher pH associated with CO 2 drawdown would have increased ½CO 2{ 3  and deepened the CCD. At least three different processes have been proposed to elevate organic carbon burial at the Eocene/Oligocene boundary. Arguably the simplest (H1a) is increased nutrient supply via more vigorous stirring in a cooler climate with a steeper meridional temperature gradient and a large Antarctic ice sheet 5,13,20 . We test H1a by a per- manent 3.5-fold increase in vertical mixing rates in the model (mix- ing between surface and middle boxes, and between middle and deep boxes, KSM and KMD, respectively, in Fig. 1). The simulation fails (Fig. 2a and b, red lines) to reproduce the proxy records by predicting a permanent change in benthic d 13 C and only a temporary change in the CCD (opposite to the observations). The latter reflects rapid depletion of the finite ocean nutrient inventory. The increase in ventilation also produces a pronounced temporary increase (by about 400 p.p.m.v.) in atmospheric p CO2 (Supplementary Fig. 1)— at odds with the onset of Antarctic glaciation. A second process (H1b) is increased efficiency of organic carbon burial at the sea floor 2 , perhaps attributable to slower bacterial acti- vity in a cooler climate 12 . We test H1b by employing a permanent 50% increase in the fraction of surface-produced organic carbon (and phosphorus) that survives to be buried. The simulation fails (Fig. 2a and b, green lines) by predicting a permanent change in d 13 C, whereas observations show a recovery. Using a temporary version of the same perturbation fails by predicting a temporary change in both d 13 C and CCD. 1 National Oceanography Centre, Southampton, European Way, Southampton SO14 3ZH, UK. 2 GKSS-Forschungszentrum, Institute for Coastal Research, Max Planck Straße 1, 21502, Geesthacht, Germany. Vol 452 | 24 April 2008 | doi:10.1038/nature06853 979 Nature Publishing Group ©2008