Marked Decline in Atmospheric Carbon Dioxide Concentrations During the Paleogene

See allHide authors and affiliations

Science  22 Jul 2005:
Vol. 309, Issue 5734, pp. 600-603
DOI: 10.1126/science.1110063


The relation between the partial pressure of atmospheric carbon dioxide (pCO2) and Paleogene climate is poorly resolved. We used stable carbon isotopic values of di-unsaturated alkenones extracted from deep sea cores to reconstruct pCO2 fromthe middle Eocene to the late Oligocene (∼45 to 25 million years ago). Our results demonstrate that pCO2 ranged between 1000 to 1500 parts per million by volume in the middle to late Eocene, then decreased in several steps during the Oligocene, and reached modern levels by the latest Oligocene. The fall in pCO2 likely allowed for a critical expansion of ice sheets on Antarctica and promoted conditions that forced the onset of terrestrial C4 photosynthesis.

The early Eocene [∼52 to 55 million years ago (Ma)] climate was the warmest of the past 65 million years. Mean annual continental temperatures were considerably elevated relative to those of today, and high latitudes were ice-free, with polar winter temperatures ∼10°C warmer than at present (13). After this climatic optimum, surface- and bottom-water temperatures steadily cooled over ∼20 million of years (4, 5), interrupted by at least one major ephemeral warming in the late middle Eocene (6). High-latitude cooling eventually sustained small Antarctic ice sheets by the late Eocene (7), culminating in a striking climate shift across the Eocene/Oligocene boundary (E/O) at 33.7 Ma. The E/O climate transition, Earth's first clear step into “icehouse” conditions during the Cenozoic, is associated with a rapid expansion of large continental ice sheets on Antarctica (8, 9) in less than ∼350,000 years (10, 11).

Changes in the partial pressure of atmospheric carbon dioxide (pCO2) are largely credited for the evolution of global climates during the Cenozoic (1214). However, the relation between pCO2 and the extraordinary climate history of the Paleogene is poorly constrained. Initial attempts to estimate early Paleogene pCO2 have provided conflicting results, with both high (15) and low (i.e., similar to modern) (16) estimates of pCO2. This deficiency in our understanding of the history of pCO2 is critical, because the role of CO2 in forcing long-term climate change during some intervals of Earth's history is equivocal. For example, Miocene pCO2 records (∼25 to 5 Ma) argue for a decoupling between global climate and CO2 (1517). These records suggest that Miocene pCO2 was rather low and invariant across periods of both inferred global warming and high-latitude cooling (17). Clearly, a more complete understanding of the relation between pCO2 and climate change requires the extension of paleo-pCO2 records back into periods when Earth was substantially warmer and ice-free.

Paleoatmospheric CO2 concentrations can be estimated from the stable carbon isotopic compositions of sedimentary organic molecules known as alkenones. Alkenones are long-chained (C37-C39) unsaturated ethyl and methyl ketones produced by a few species of Haptophyte algae in the modern ocean (18). Alkenone-based pCO2 estimates derive from records of the carbon isotopic fractionation that occurred during marine photosynthetic carbon fixation (ϵp). Chemostat experiments conducted under nitrate-limited conditions indicate that alkenone-based ϵp values (ϵp37:2) vary as a function of the concentration of aqueous CO2 ([CO2aq]) and specific growth rate (1921). These experiments also provide evidence that cell geometry accounts for differences in ϵp among marine microalgae cultured under similar conditions (21). In contrast, results from dilute batch cultures conducted under nutrient-replete conditions yield substantially lower ϵp37:2 values, a different relation for ϵp versus μ/CO2aq (where μ = algal growth rate), and a minimal response to [CO2aq] (22). Thus, comparison of the available culture data suggests that different growth and environmental conditions potentially trigger different carbon uptake pathways and carbon isotopic responses (23). A recent evaluation of the efficacy of the alkenone-CO2 approach, using sedimentary alkenones in the natural environment, supported the capacity of the technique to resolve relatively small differences in water column [CO2aq] across a variety of marine environments when phosphate concentrations and temperatures are constrained (24).

In our study, we extended records of the carbon isotopic composition of sedimentary alkenones (δ13C37:2) from the middle Eocene to the late Oligocene and established a record of pCO2 for the past ∼45 million years. Samples from Deep Sea Drilling Project sites 516, 511, 513, and 612 and Ocean Drilling Program site 803 (Fig. 1) were used to reconstruct δ13C37:2 and ϵp37:2 records ranging from the middle Eocene to the late Oligocene (∼25 to 45 Ma). These sites presumably represent a range of oceanic environments with a variety of surface-water nutrient and algal-growth conditions and thus reflect a set of environmental and physiological factors affecting both δ13C37:2 and ϵp37:2 values.

Fig. 1.

Site location map. Sites 612, 516, 803, 511, and 513 were used to reconstruct Eocene and Oligocene ϵp37:2 values. Sites 588, 608, 730, and 516 were used to reconstruct Miocene ϵp37:2 values.

These data are presented as a composite record, in large part because the measurable concentration of di-unsaturated alkenones varied both spatially and temporally. Moreover, continuous alkenone records spanning the entire Eocene and Oligocene from individual sites were not recovered. As a consequence, most of the Oligocene record is represented at site 516, whereas the majority of the Eocene is represented at site 612 (Fig. 2A). Age models for each site were developed by linearly interpolating between biostratigraphic datums (2531), calibrated to the Geomagnetic Polarity Time Scale (32).

Fig. 2.

(A) Stable carbon isotopic composition of di-unsaturated alkenones. Each data point represents one measurement or an average of multiple measurements, with error bars bracketing the range of values for each sample. PDB, Pee Dee belemnite standard. (B) Compilation of the carbon isotopic composition of di-unsaturated alkenones from this study and Pagani et al. (17, 42, 53). (C) Paleogene ϵp37:2 values. ϵp37:2 is calculated from the δ13 C of di-unsaturated alkenones as follows: ϵp37:2 = [(δd + 1000/δp + 1000) – 1] × 103, where δd is the carbon isotopic composition of CO2aq calculated from mixed-layer carbonates and δp is the carbon isotopic composition of haptophyte organic matter enriched by 4.2‰ relative to alkenone δ13C (54). Carbon isotopic compositions of mixed-layer carbonates were used to calculate δd by assuming equilibrium conditions and applying temperature-dependent isotope equations (55, 56). Mixed-layer temperatures were calculated from the δ18O of planktonic foraminifera (57) as follows: site 612, Acarinina spp.; site 513, Subbotina spp. and Chiloguembelina cubensis; and site 511, Subbotina spp. Temperatures for sites 516 and 803 were estimated from the δ18O compositions of the <60-μm carbonate fraction, assuming that the δ18O composition of seawater changed from –0.75‰ during the Eocene to –0.5‰ during the Oligocene. Error bars reflect the range of ϵp37:2 values calculated by applying the maxima and minima of both the measured δ13C of di-unsaturated alkenones and calculated temperatures. (D) Compilation of ϵp37:2 values from this study and Pagani et al. (17, 42, 53). Dashed horizontal lines bracket the range of ϵp37:2 values from surface waters of modern oceans. In general, higher and lower ϵp37:2 values come from oligotrophic and eutrophic environments, respectively. The shaded box represents the range of ϵp37:2 values from oligotrophic sites where [PO43–] ranges between 0.0 to 0.2 μmol/liter.

Eocene δ13C37:2 values range from ∼–30 to –35 per mil (‰), with the most negative values (sites 511 and 513) occurring near the E/O boundary. δ13C37:2 values increase substantially through the Oligocene with maximum values of ∼–27‰ by ∼25.5 Ma. This trend is briefly reversed near the end of the Oligocene as δ13C37:2 values become more negative, reaching ∼–32‰ by 25 Ma (Fig. 2A). The overall pattern of 13C enrichment continues into the Miocene, establishing a clear secular trend from the middle Eocene to the middle Miocene (Fig. 2B). These isotopic trends do not mirror changes in the δ13C of dissolved inorganic carbon (δ13CDIC) because δ13C records of bulk carbonate (33) and benthic foraminifera (10) indicate small changes in δ13CDIC for the Eocene to Oligocene relative to the change in δ13C37:2. Nonetheless, interpretations of long-term trends in δ13C37:2 are enhanced when δ13C37:2 values are converted to ϵp37:2 (34), thus eliminating the influence of δ13CDIC.

The temporal pattern of ϵp37:2 is similar to that of δ13C37:2 (Fig. 2, C and D), consistent with other studies (17). Higher values of ϵp37:2 (∼19.5 to 21.5‰) characterize the Eocene and then decrease through the Oligocene. The ϵp37:2 values recorded for the Eocene and earliest Oligocene are higher than any recorded for the modern ocean (Fig. 2D). Given our present understanding of the controls on ϵp37:2, the decrease in ϵp37:2 from the Eocene through the Oligocene could be driven by a consistent change in the cell dimensions of alkenone-producing algae over time, a secular increase in growth rates of alkenone-producing algae, or a long-term decrease in [CO2aq] and/or increased utilization of bicarbonate ([HCO3-]) as a result of low [CO2aq] (35). At present, evolutionary changes in algal cell geometries are poorly constrained. If the long-term decrease in ϵp37:2 were driven solely by changes in algal cell dimensions, it would require a pattern of increasing ratios of cell volume to surface area with time. If ϵp scales linearly with the ratio of cell volume to surface area (21), the observed change in ϵp37:2 values would require an ∼60% increase in the cell diameters of alkenone-producing algae from the Eocene to the Miocene (i.e., sites 516 and 612). Further, given that Miocene and Modern ϵp37:2 values are similar, Eocene coccolithophores would have to have been ∼60% smaller than modern alkenone producers, such as Emiliania huxleyi, with cell diameters of ∼5 μm (21). However, the available data suggest that placoliths from probable alkenone producers, specifically species within the genus Reticulofenestra, were substantially larger than modern species and then decreased through the Oligocene and early Miocene (36, 37). If we reasonably assume that placolith geometry scales to cell geometry (38), then cell diameters decreased during the late Paleogene. A trend of decreasing cell diameters should lead to an increase in ϵp37:2 values (21), which is contrary to our measurements. Thus, although a long-term change in cell geometry might have influenced the relative magnitude of Paleogene ϵp37:2 values, it was not responsible for the pattern observed in our record.

Alternatively, variations in ϵp37:2 could be ascribed to variations in the specific growth rates of alkenone-producing algae (μalk), with higher μalk values associated with lower ϵp37:2 values. Under this scenario, Eocene and early Oligocene ϵp37:2 values must reflect substantially lower μalk than modern μalk found in oligotrophic waters where [PO43-] is ∼0 μmol/liter (Fig. 2D). That is, algal growth rates during the Paleogene from both eutrophic and oligotrophic environments would have to be lower than the lowest growth rates found in the modern oligotrophic ocean. Further, if growth rates were indeed the first-order control on ϵp37:2 values, the lowest Miocene ϵp37:2 values would require substantially higher algal growth rates in oligotrophic settings, comparable to those of the highly productive Peru upwelling margin (Fig. 2D). Therefore, we conclude that rather extraordinary changes in μalk are required to explain the temporal pattern of ϵp37:2 values and thus are not the primary cause for the observed long-term trends. Instead, we contend that the Cenozoic evolution of ϵp37:2 was forced primarily by changes in [CO2aq] and pCO2. Accordingly, these records would qualitatively reflect high surface-water [CO2aq] during the middle to late Eocene, a pattern of decreasing [CO2aq] through the Oligocene, and near-modern levels during the Neogene. If the change in ϵp37:2 values during the Paleogene was brought about by an increased utilization of HCO3- over CO2aq, then it implies that [CO2aq] became increasingly limiting to algal growth in both oligotrophic and eutrophic environments. Although this would compromise quantitative estimates of atmospheric pCO2, it would still support a scenario of decreasing pCO2 with time. Until evidence emerges to the contrary, we must assume that the physiological processes responsible for ϵp37:2 in the past were similar to those operating in modern surface waters (19, 24) and use these data to estimate both [CO2aq] and pCO2 over the past 45 million years.

The conversion of ϵp37:2 values to pCO2 requires an estimate of surface-water [PO43–] (39) and temperature for each site. For this study, we assumed that modern surface-water distributions of [PO43–] at each site between 0 and 100 m encompassed the probable range at any given time, and we applied temperatures derived from the oxygen isotope composition of coeval carbonates in order to convert estimates of [CO2aq] to pCO2. This approach assumes relative air-sea equilibrium, which may not be valid for every site. However, although disequilibrium could lead to overestimates of pCO2, our treatment of the data ultimately yields a range of CO2 concentrations that reflects the uncertainty associated with this effect. On a broad scale, our results indicate that CO2 concentrations during the middle to late Eocene ranged between 1000 and 1500 parts per million by volume (ppmv) (40) and then rapidly decreased during the Oligocene, reaching modern levels by the latest Oligocene (Fig. 3A). In detail, a trend toward lower CO2 concentrations is evident from the middle to late Eocene, reaching levels by the E/O boundary that could have triggered the rapid expansion of ice on east Antarctica (2). An episode of higher pCO2 in the latest Oligocene occurs concomitantly with a ∼2-million-year low in the mean δ18O composition of benthic foraminifera (Fig. 3B), indicating that global climate and the carbon cycle were linked from the Eocene to the late Oligocene. This association weakens in the Neogene, when long-term patterns of climate and pCO2 appear to be decoupled (17).

Fig. 3.

(A) pCO2 estimates calculated from ϵp37:2. ϵp = ϵfb/[CO2aq] (39), where b = {(118.52[PO43–]) + 84.07}/(25 – ϵp37:2), calculated from the geometric mean regression of all available data (19, 20, 23, 58, 59). [CO2aq] values were calculated using mean ϵp37:2 values and a range of [PO43–] values for each site. [PO43–] ranges applied for individual sites were as follows: site 612, 0.5 to 0.3 μmol/liter; site 516, 0.4 to 0.2 μmol/liter; sites 511 and 513, 1.10 to 0.8 μmol/liter; site 803, 0.3 to 0.1 μmol/liter; and site 588, 0.3 to 0.2 μmol/liter. Values of CO2aq were converted to pCO2 by applying Henry's Law (60), calculated assuming a salinity of 35 and surface-water temperatures derived from δ18O of marine carbonates. Maximum pCO2 estimates were calculated using maximum temperatures (61) for each sample and maximum [PO43–] for each site. Intermediate and minimum (dashed line) pCO2 estimates were calculated using intermediate and minimum temperatures for each sample and minimum [PO43–] for each site. An analytical treatment of error propagation suggests that relative uncertainties in reconstructed CO2 values are ∼20% for the Miocene data and approach 30 to 40% for Paleogene samples with higher (20 to 24‰) ϵp37:2 values (62). (B) Global compilation of benthic oxygen isotope records (5).

In addition to climate, the change in CO2 implied by our record would have substantially affected the growth characteristics of terrestrial flora. In particular, the expansion of C4 grasses has received considerable attention as an indicator of environmental change (41, 42). The C4 pathway concentrates CO2 at the site of carboxylation and enhances rates of photosynthesis by eliminating the effects of photorespiration under low CO2 concentrations (43). Moreover, higher rates of carbon assimilation can be maintained under water-stressed conditions. This results in a water-use efficiency (water loss per unit of carbon assimilated) in C4 plants that is twice that of C3 plants at ∼25°C (44). Given our understanding of the environmental parameters affecting C3 and C4 plants, a prevalent supposition has emerged that C4 photosynthesis originated as a response to stresses associated with photorespiration (41, 45). The CO2 threshold below which C4 photosynthesis is favored over C3 flora is estimated at ∼500 ppmv (41), a level that is breached during the Oligocene. Molecular phylogenies (46, 47) and isotopic data (48) place the origin of C4 grasses before the Miocene between 25 to 32 Ma (46, 47, 49), the interval when CO2 concentrations approached modern levels. This confluence strongly suggests that C4 photosynthesis evolved as a response to increased photorespiration rates forced by a substantial drop in pCO2 during the Oligocene. Near-global expansion of C4 ecosystems ensued later in the Miocene (41), possibly driven by drier climates and/or changes in patterns of precipitation (42).

References and Notes

View Abstract

Navigate This Article