Transient Middle Eocene Atmospheric CO2 and Temperature Variations

See allHide authors and affiliations

Science  05 Nov 2010:
Vol. 330, Issue 6005, pp. 819-821
DOI: 10.1126/science.1193654

The Dependable Warmer

During the middle of the Eocene, about 40 million years ago, a transient warming event interrupted the long-term cooling trend that had been in progress for the previous 10 million years. Bijl et al. (p. 819; see the Perspective by Pearson) constructed records of sea surface temperature and atmospheric CO2 concentrations across the warming period. It appears that vast amounts of CO2 were injected into the atmosphere, and a sea surface temperature increase of as much a 6°C accompanied the atmospheric CO2 rise.


The long-term warmth of the Eocene (~56 to 34 million years ago) is commonly associated with elevated partial pressure of atmospheric carbon dioxide (pCO2). However, a direct relationship between the two has not been established for short-term climate perturbations. We reconstructed changes in both pCO2 and temperature over an episode of transient global warming called the Middle Eocene Climatic Optimum (MECO; ~40 million years ago). Organic molecular paleothermometry indicates a warming of southwest Pacific sea surface temperatures (SSTs) by 3° to 6°C. Reconstructions of pCO2 indicate a concomitant increase by a factor of 2 to 3. The marked consistency between SST and pCO2 trends during the MECO suggests that elevated pCO2 played a major role in global warming during the MECO.

The Middle Eocene Climatic Optimum [MECO; ~40 million years ago (Ma)] (1) interrupts a long-term middle Eocene cooling trend (2), with a globally uniform 4° to 6°C warming of both surface and deep oceans within ~400,000 years, as derived from foraminiferal stable oxygen isotope records (3). A decrease in carbonate mass accumulation rates during the MECO argues for ocean acidification induced by a rise in pCO2 (3). Application of paleo-pCO2 proxies across the MECO has yet to confirm whether pCO2 changes are indeed associated with this interval of transient warming.

We investigated a sedimentary succession spanning the MECO recovered from the East Tasman Plateau at Ocean Drilling Program (ODP) Site 1172, which at that time was situated on the shelf (~65°S paleolatitude; Fig. 1 and figs. S1 and S2) (4, 5). To fully capture the magnitude of the sea surface temperature (SST) change associated with the MECO at this site, we applied two independent temperature proxies: the alkenone unsaturation index (UK37) (6) and the index of tetraethers consisting of 86 carbon atoms (TEX86) (5, 7) (fig. S3). At the onset of the MECO, UK37 and TEX86 indicate a rise in SST of 3°C and 6°C, respectively, which, also at this location, stands out as an interruption of long-term middle Eocene cooling (Fig. 2). Bulk carbonate oxygen isotope values (δ18O) decrease by 1.0 to 1.2 per mil (‰), which, if controlled by SST only, also indicate a SST rise of ~4° to 5°C (5).

Fig. 1

Paleogeographic configuration of the southern high latitudes during the middle Eocene (~49 to 37 Ma; map was obtained from and ocean surface current configurations inferred from general circulation model experiments (13). The orange star indicates the paleogeographic location of ODP Site 1172 at 65°S in the southwest Pacific Ocean (24), under the influence of the Antarctic-derived Tasman Current (TC).

Fig. 2

Geochemical and palynological results across the MECO at ODP Site 1172, Hole A, cores 42X to 47X. The MECO is identified by integrating magnetostratigraphy (25), biostratigraphy (25), and chemostratigraphy (3) (see fig. S2). The UK37 (purple), TEX86 (red), and bulk δ18O (green) SST reconstructions show a warming of 3° to 6°C. The yellow shaded area delimits the MECO interval. The increase in percentage of low-latitude dinocysts (in orange) at the expense of endemic dinocysts during the MECO illustrates biotic response to warming. We estimated pCO2 from carbon isotopic fractionation during carbon fixation (εp; black) by haptophyte algae with phosphate concentrations between 0 and 1 μmol liter−1 (light gray band). We further constrained phosphate estimates (dark blue line) by allowing phosphate concentrations to vary between 0.1 ± 0.1 μm liter−1 and 0.9 ± 0.1 μm liter−1 as a function of the ratio of peridinioid over gonyaulacoid dinocysts (P/G ratio; light blue line). This results in further constrained pCO2 estimates (dark gray band). TEX86, UK37, oxygen isotope, and pCO2 data are plotted with a 3-point running mean (solid orange, purple, green, and red lines, respectively). The error bars on TEX86 and UK37 represent analytical error. On εp the errors bars represent the difference between the use of TEX86 and UK37 to determine εp.

Additional evidence of warming is derived from assemblages of hypnozygotic organic cysts of surface-dwelling dinoflagellates (dinocysts) (5). Whereas the middle Eocene dinocyst record at ODP Site 1172 is dominated by taxa that are endemic to the Southern Ocean (8), an incursion of low-latitude dinocyst taxa characterizes the MECO (Fig. 2 and fig. S4). A SST increase of 3° to 6°C is consistent with inferences from benthic foraminiferal and fine-fraction carbonate oxygen isotope records at other sites (1, 3). The UK37 and TEX86 proxies are independent of seawater δ18O. Hence, the consistent magnitude of warming between the proxies suggests that the carbonate δ18O records were not affected by a change in δ18O of seawater, and that global ice volume did not change considerably during the MECO.

Absolute SSTs as indicated by UK37 and TEX86 are consistent, with 26°C or 24°C just below the onset of the MECO for the two proxies, respectively, and peak MECO SSTs exceeding 28°C. These SSTs are much (~10°C) higher than those derived from fine-fraction carbonate oxygen isotope measurements from elsewhere in the Southern Ocean (1, 3). At least part of this large discrepancy is most likely the result of diagenetic alteration of calcite (9).

We assessed pCO2 changes by determining the stable carbon isotopic composition (δ13C) of alkenones, long-chained ketones exclusively synthesized by specific haptophyte algae. Carbon isotopic fractionation during carbon fixation (εp) by haptophyte algae varies as a function of dissolved CO2 [CO2(aq)] (10, 11), specific cell physiological parameters (which show good correspondence to the surface-water concentrations of soluble phosphate), and other environmental parameters, primarily light intensity (5). The carbon isotopic composition of diunsaturated alkenones (δ13CC37:2) ranges between –32.5 and –35.5‰ (fig. S3). We used bulk carbonate δ13C to estimate the δ13C value of the dissolved inorganic carbon (DIC) pool in seawater (5) to determine εp. The data show background εp values of 21 to 22‰ rising up to 24.5‰ during MECO (Figs. 2 and 3 and fig. S3). The relationship between εp and pCO2 is exponential, which results in a relatively large uncertainty in reconstructed pCO2 levels with high εp values (Fig. 3). Temperature variations, however, play a minor role in the range of temperatures indicated by TEX86 and UK37 (Fig. 2) and cannot explain the high εp values (Fig. 3). It seems unlikely that changes in light intensity (12) influenced εp substantially at ODP Site 1172 (5). The soluble phosphate concentration exerts a strong influence on the relation between εp and pCO2, particularly if εp values are high (5).

Fig. 3

Relationship between εp [versus VPDB (Vienna pee dee belemnite) standard] and pCO2. The phosphate concentration ranges plotted are those from the present-day surface ocean, and the SST ranges (22° to 30°C) are those inferred from TEX86 and UK37 data presented in Fig. 2 and a suite of Southern Ocean sites (22, 2628). The gray vertical bar indicates the values for εp as reconstructed for the MECO interval at ODP Site 1172, with use of TEX86-based and UK37-based SST reconstructions for these same levels, and bulk carbonate δ13C measurements on the same section. Black horizontal line represents present-day pCO2. Figure is modified from Pagani et al. (10).

To evaluate all possible absolute pCO2 estimates from our record, we applied the full range of present-day surface-water phosphate concentrations. These vary between 0 μmol liter−1 in the oligotrophic gyres to >2 μmol liter−1 in the Southern Ocean (5) (fig. S5). Yet even when phosphate concentrations of 0 μmol liter−1 are assumed, εp values between 21.2‰ and 24.5‰ yield pCO2 estimates between 600 parts per million by volume (ppmv) before the MECO and 6400 ppmv during the MECO (Figs. 2 and 3). Hence, elevated levels of pCO2 must in part be responsible for the high εp values, with middle Eocene pCO2 being more than twice the pre-industrial value. When we assume maximal phosphate concentrations of 2 μmol liter−1, pCO2 ranges between ~2500 and ~24,000 ppmv (Figs. 2 and 3). The marginal marine Eocene East Tasman Plateau (4) likely experienced phosphate concentrations that were higher than 0 μmol liter−1 but lower than 2 μmol liter−1, because closed oceanic gateways during the Eocene (13) prevented mixing associated with the Antarctic Circumpolar Current (ACC) that causes high phosphate levels in the present-day Southern Ocean. Eocene southwest Pacific surface-water phosphate concentrations were unlikely to have exceeded 1 μmol liter−1, which implies maximum pCO2 estimates of 1600 ppmv just before the MECO and 15,000 ppmv during the MECO (Figs. 2 and 3). With a realistic range of phosphate concentrations, pCO2 values were between 600 and 1600 ppmv just before the MECO, which is in line with previous estimates of middle Eocene pCO2 values using the same proxy (14), and rose to between 6400 and 15,000 ppmv during the MECO (Fig. 2, light gray band). MECO values exceed any previous Eocene alkenone-based estimate even when we assume phosphate concentrations of 0 μmol liter−1 (5). Despite uncertainties regarding absolute pCO2 values, we note that the trends in pCO2 follow those in the SST records remarkably well (Fig. 2).

Surface-water phosphate concentrations, however, may have varied in a marginal marine setting at ODP Site 1172. A tool for the reconstruction of phosphate concentration uses the remains of dinoflagellates (dinocysts), which are known to be extremely sensitive to surface-water nutrient availability changes (5). The ratio between peridinioid and gonyaulacoid dinocyst groups (the P/G ratio) is often used to reconstruct changes in relative nutrient abundance (15). The MECO at ODP Site 1172 shows a major decrease in the P/G ratio (16) (Fig. 2), suggesting a decrease in nutrient concentrations (16). As an experiment, we allowed phosphate to linearly vary as a function of the P/G ratio (Fig. 2) (5). The most prominent shift toward low P/G ratios occurs at MECO warming, resulting in lower phosphate concentration estimates. Also, when this drop in phosphate is taken into account, pCO2 rises during the MECO and follows the SST trends (Fig. 2, dark gray band). Hence, regardless of the constraints on phosphate concentrations and other environmental parameters, pCO2 levels must have been substantially higher during the MECO relative to the middle Eocene background.

One outstanding issue is the source of carbon responsible for the increase in middle Eocene atmospheric CO2. The rise in pCO2 by 2000 to 3000 ppmv emerging from our data requires a carbon source capable of injecting vast amounts of carbon into the atmosphere. Moreover, the absence of a prominent negative carbon isotope excursion excludes reservoirs with δ13C signatures below that of marine DIC (3). One mechanism capable of emanating carbon with such a geochemical signature is the metamorphic alteration of carbonates (decarbonation) (1). Massive decarbonation occurred until the late Eocene, with the subduction of vast amounts of Tethyan Ocean pelagic carbonates under Asia as India drifted northward (1719). However, the flux of carbon required to increase pCO2 by 2000 to 3000 ppmv within ~400,000 years appears too high to invoke metamorphic (volcanic) outgassing as the sole mechanism.

Our pCO2 and SST reconstructions allow for a tentative assessment of high-latitude climate sensitivity to CO2 forcing on ~100,000-year time scales, assuming that all MECO warming was caused by pCO2 and associated feedbacks. With an average 5°C SST increase and a factor of 2 to 3 increase in pCO2, we arrive at a climate sensitivity of ~2° to 5°C per pCO2 doubling. When we consider the pCO2 decline from the Middle Eocene (~2000 ppmv; this study) to the latest Eocene (~1000 ppmv) (20) and the coeval high-latitude temperature decline (~3.5°C) (21, 22), we derive similar values. Thus, long-term climate sensitivity to pCO2 forcing in a world without the amplifying effects of ice-albedo feedbacks (23) may have been larger than previously anticipated.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S5

Tables S1 to S3

References and Notes

  1. See supporting material on Science Online.
  2. Surface ocean CO2(aq) originates from atmospheric CO2 and deep waters, of which the latter is of major importance in marginal marine upwelling areas. Changes in upwelling through time may substantially change CO2(aq) and hence would skew the reconstructed pCO2 record. At Site 1172 we argue that a change in upwelling is not responsible for the signal we recorded in the alkenones, because that would have led to prominent shifts in the bulk carbonate carbon isotope profile (figs. S2 and S3).
  3. This research used samples and data provided by the Ocean Drilling Program (ODP) sponsored by NSF and participating countries under the management of Joint Oceanographic Institutions Inc. We thank Utrecht University and the LPP foundation (P.K.B.), Statoil (A.J.P.H.), and the Netherlands Organization for Scientific Research (Vici grant to S.S.; Veni grant 863.07.001 to A.S.) for financial support. A.S. acknowledges the European Research Council under the European Community’s Seventh Framework Program for ERC Starting Grant 259627. Groundwork for this research was provided at the Urbino Summer School of Paleoclimatology. We thank N. Welters, J. van Tongeren, J. Ossebaar, E. Hopmans, M. Kienhuis, G. Nobbe, E. van Bentum, E. Speelman, and E. van Soelen for technical support and M. Pagani, C. Stickley, J. Zachos, and two anonymous reviewers for invaluable discussions and comments.
View Abstract

Navigate This Article