SEARCH

SEARCH BY CITATION

Keywords:

  • carbon dioxide;
  • carbon isotopes;
  • foraminifera

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

[1] The carbon isotope ratio (δ13C) of plant material is commonly used to reconstruct the relative distribution of C3 and C4 plants in ancient ecosystems. However, such estimates depend on the δ13C of atmospheric CO2 (δ13CCO2) at the time, which likely varied throughout Earth history. For this study, we use benthic and planktonic δ13C and δ18O records to reconstruct a long-term record of Cenozoic δ13CCO2. Confidence intervals for δ13CCO2 values are assigned after careful consideration of equilibrium and non-equilibrium isotope effects and processes, as well as resolution of the data. We find that benthic foraminifera better constrain δ13CCO2 compared to planktonic foraminiferal records, which are influenced by photosymbiotes, depth of production, seasonal variability, and preservation. Furthermore, sensitivity analyses designed to quantify the effects of temperature uncertainty and diagenesis on benthic foraminifera δ13C and δ18O values indicate that these factors act to offset one another. Our reconstruction suggests that Cenozoic δ13CCO2 averaged −6.1 ± 0.6‰ (1σ), while only 11.2 million of the last 65.5 million years correspond to the pre-Industrial value of −6.5‰ (with 90% confidence). Here δ13CCO2 also displays significant variations throughout the record, at times departing from the pre-Industrial value by more than 2‰. Thus, the observed variability in δ13CCO2 should be considered in isotopic reconstructions of ancient terrestrial-plant ecosystems, especially during the Late and Middle Miocene, times of presumed C4 grassland expansion.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

[2] The δ13C value of atmospheric CO2 (δ13CCO2), along with isotopic fractionations associated with carbon fixation and a plant's specific water use efficiency, determine terrestrial plant carbon isotopic compositions [Farquhar et al., 1989]. Precise measurements of δ13CCO2 and plant δ13C values are readily available for modern canopy and leaf-scale research [Lai et al., 2006; Pypker et al., 2008], and allow the relative abundance of C3 and C4 plants in various environments to be accurately determined by stable carbon isotope analysis of sedimentary organic carbon [Follett et al., 2009; Wang et al., 2010]. Ice core measurements of δ13CCO2 are only available for the last 50 kyr [Elsig et al., 2009; Friedli et al., 1986; Leuenberger et al., 1992; Smith et al., 1999]. Consequently, the value of δ13CCO2 must be assumed for earlier periods of Earth history in order to establish C3 and C4 distributions using the 13C/12C compositions of organic carbon.

[3] Carbon fluxes into and out of ocean-atmosphere reservoirs can vary overtime [Berner, 1998, 2006; Kump and Arthur, 1999], and alter the δ13C value of individual carbon reservoirs [Katz et al., 2005; Zachos et al., 2001] and the partial pressure of atmospheric carbon dioxide (pCO2) [Pagani et al., 2005; Royer et al., 2004]. On human time-scales, δ13CCO2 has changed by ∼−1.5‰ concurrent with the increase in anthropogenic CO2 since the Industrial Revolution [Marino et al., 1992]. While the δ13CCO2 value has almost assuredly evolved over longer time intervals, terrestrial paleoecologic reconstructions often apply a pre-Industrial δ13CCO2 value of −6.5‰ for geochemical reconstructions of plant distributions, because δ13CCO2 before the late Quaternary is not precisely constrained [Bibi, 2007; Boisserie et al., 2005; Feakins et al., 2005; Feakins et al., 2007; Fox and Koch, 2003, 2004; Hopley et al., 2007; Levin et al., 2004; Ségalen et al., 2006; Zazzo et al., 2000].

[4] Reconstructions of δ13CCO2 have been developed using δ13C records of ancient plant matter [Gröcke, 2002; Jahren et al., 2001; Marino et al., 1992] and foraminifera [Passey et al., 2009, 2002; Smith et al., 2007]. Long-term (multimillion year) δ13CCO2 reconstructions are often based on the δ13C of ancient woody material that was subjected to different levels of molecular degradation and thermal maturation [van Bergen and Poole, 2002]. As a consequence, interpretations of δ13CCO2 from fossil-wood δ13C values require a detailed understanding of diagenetic histories – a condition rarely satisfied. Moreover, environmental conditions strongly influence carbon-isotope compositions of terrestrial plants, contributing as much as ±5‰ variability [Deines, 1980], adding to the uncertainty to δ13CCO2 reconstructions [Bump et al., 2007].

[5] Marine carbonates are abundant and better preserved in comparison to terrestrial organic matter. It is well established that the carbon-isotopic composition of foraminiferal calcite (δ13Ccc) is primarily controlled by the δ13C of dissolved inorganic carbon (δ13CDIC), species-specific biological (‘vital’) effects, temperature, and pH [Spero and Lea, 1993]. Consequently, planktonic and benthic δ13Ccc records are useful in inferring surface productivity [e.g., Stott et al., 2000], depth of calcification [e.g., Holsten et al., 2004], and ancient ocean circulation [e.g., Wright and Miller, 1993]. If the factors affecting δ13CDIC are carefully considered, δ13Ccc records can provide the basis for determining the long-term history of δ13CCO2. Passey et al. [2002] evaluated δ13CCO2 for the past 20 million years from existing planktonic δ13C records by applying a constant isotopic offset determined from seven planktonic foraminiferal species and near-modern atmosphere δ13CCO2. A similar approach was used to re-evaluate Miocene δ13CCO2 from benthic foraminiferal δ13C records using a constant isotopic offset between benthic foraminifera and atmospheric CO2 determined by comparing Pleistocene benthic foraminifera and pre-Industrial δ13CCO2 [Passey et al., 2009]. In general, these studies do not take into account the temporal sampling resolution of foraminiferal δ13C variability, associations with photosymbiotes [Spero, 1992; Spero and Lea, 1993], downwelling rates and the strength of the biological pump [Kroopnick, 1985] — factors that increase the uncertainty in δ13CCO2 reconstructions.

[6] Here, we present a meta-analysis of existing planktonic and benthic δ13Ccc records and develop two proxy records of Cenozoic δ13CCO2 values. We test the veracity of our approach by direct comparison with ice core δ13CCO2 measurements spanning the last 50,000 years, and present a sensitivity analysis for uncertainties associated with diagenesis and temperature. The resulting δ13CCO2 reconstructions of the past 65 million years are compared against each other and to other methodologies to reconstruct δ13CCO2 values.

2. Summary of Methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

2.1. Benthic and Planktonic Foraminifera δ13C and δ18O Records

[7] We used 2,128 planktonic and 2,576 benthic foraminifera stable carbon (δ13Ccc) and oxygen (δ18Occ) isotope measurements from 35 Deep Sea Drilling Project (DSDP), Ocean Drilling Program (ODP), and continental sites (Figure 1). Planktonic data was compiled from published articles (Table 1), whereas the benthic records are a subset of those compiled by Zachos et al. [2001]. The compiled records span the past 65 Myr and provide sample densities that resolve major paleoclimatic and paleoceanographic variation seen in the larger data sets of Miller and Katz [1987] and Zachos et al. [2001] (Figure 2). The restricted use of data reflects specific choices regarding foraminifera species and locations as discussed below.

image

Figure 1. Marine core locations. Cores from which planktonic, benthic, and both planktonic and benthic δ13Ccc records were compiled are shown with open circles, black circles, and circles with dots, respectively.

Download figure to PowerPoint

image

Figure 2. Raw foraminifera δ13Ccc (dots) and reconstructed δ13CCO2 (open symbols) over the past 65 Myr. Benthic foraminifera δ13Ccc measurements (blue dots) and benthic-based δ13CCO2 (blue circles) are shown with planktonic δ13Ccc measurements (red dots) and planktonic-based δ13CCO2 (red diamonds). Benthic foraminifera δ13Ccc values reflect offsets used by Zachos et al. [2001]. Dashed line marks the pre-industrial δ13CCO2 average of −6.5‰ [Marino et al., 1992].

Download figure to PowerPoint

Table 1. Cruise, Site, and Location Information for Published Foraminiferal Isotope Data Sets Used
DSDP/ODP LegDSDP/ODP SiteLatitudeLongitudeBasinLocationPlanktonicaBenthicaSourcesb
  • a

    Crosses mark if data are planktonic and/or benthic.

  • b

    Data sources for each data set are provided in Text S1.

  • c

    Continental Tanzania Drilling Project.

  • d

    NA, not applicable.

64419.31−169.02Central PacificHorizon RidgeX 1
64732.45157.71North PacificShatsky PlateauX 1
6559.30142.54Central PacificCaroline RidgeX 1
22213–10.2193.90IndianNinetyeast RidgeX 2
22215–8.1286.79IndianNinetyeast RidgeX 2
72516–30.28–35.29South AtlanticRio Grande RiseX 1
73524–29.483.51Central AtlanticWalvis RidgeX 3
74525–29.072.99Central AtlanticWalvis RidgeX 4
74526–30.123.14South AtlanticWalvis RidgeX 4
74527–28.041.76Central AtlanticWalvis RidgeX 4
74528–28.522.32Central AtlanticWalvis RidgeX 4
74529–28.932.77Central AtlanticWalvis RidgeX 4
8155356.09–23.34North AtlanticRockall Plateau X5
8255837.77–37.35North AtlanticMid-Atlantic Ridge X6
8256333.64–43.77North AtlanticMid-Atlantic Ridge X7
855721.43–113.84Central PacificClipperton Fracture ZoneX 8
855755.85–135.04Central PacificClipperton Fracture ZoneX 9
90590–31.17163.36South PacificLord Howe RiseX 10
9460741.00–32.96North AtlanticMid-Atlantic Ridge X10
9460842.84–23.09North AtlanticMid-Atlantic RidgeXX1, 7
113689–64.523.10SouthernMaud RiseXX11, 12, 13, 14
113690–65.171.22SouthernMaud RiseXX11, 12, 15
114702–49.87–40.85South AtlanticIsla Orcadas RiseX 2
11773017.7357.69IndianOman MarginX 1
119738–62.7282.78SouthernKerguelen PlateauXX16
119744–61.2380.59SouthernKerguelen Plateau X17
120748–58.4478.98SouthernKerguelen Plateau X14, 18
121757–17.0288.18IndianNinetyeast RidgeX 2
1217585.3890.35IndianNinetyeast RidgeX 2
143865–18.43–17.55Central AtlanticAllison GuyotX 19
198120932.65158.50North PacificShatsky RiseX 20
19912188.89–135.37Central PacificClipperton Fracture ZoneX 21
TDPcNAd–8.8639.46Indian X 22

[8] We limit our analysis to benthic genera Cibicidoides and Nuttallides [Zachos et al., 2008, 2001] from the North Atlantic and Southern Oceans. Subtle differences between seawater and individual species δ13C values increase the range of uncertainty of δ13CCO2 produced from δ13C records derived from multiple foraminiferal species. However, on multimillion year time scales, genera-specific adjustments are appropriate considering the limited geologic occurrence of individual species [Hilting et al., 2008]. We adjusted Nuttallides δ13C values in order to directly compare them to Cibicidoides values following the assumptions of Katz et al. [2003], where

  • equation image

To account for disequilibrium related to vital effects, Cibicidoides and Nuttallides δ18O values were adjusted by +0.6‰ and +0.4‰, respectively [Shackleton et al., 1984; Zachos et al., 2001].

[9] Both Cibicidoides and Nuttallides are thought to be epifaunal genera [Katz et al., 2003; Thomas and Shackleton, 1996]. However, recent work from Sexton et al. [2006a] suggests that both could have been shallow infaunal organisms. This slight infaunal behavior potentially protects these organisms from dissolution and better represents the isotopic character of bottom waters [Katz et al., 2003; Sexton et al., 2006a]. Overall, the available data indicate that Cibicidoides and Nuttallides genera provide a reliable proxy for the isotopic chemistry of ancient bottom waters.

[10] In our analysis, planktonic records are limited to surface-dwelling genera Globigerinoides, Acarinina, and Morozovella. Ancient Globigerinoides, Acarinina, and Morozovella are characterized by δ13Ccc and δ18Occ values that are consistent with a surface dwelling habit [Boersma et al., 1979; Sexton et al., 2006b]. Extant Globigerinoides, Acarinina, and Morozovella species are associated with photosymbiotes, with an inverse relationship between test size and δ13C value [D'Hondt and Zachos, 1993; D'Hondt et al., 1994; Pearson et al., 1993]. For this work, we do not attempt to correct for size effects in our planktonic foraminifera compilation, given that the isotopic effects are small (∼0.2‰) and consistent size fraction sampling in older records is often lacking.

[11] We reviewed all existing age-depth relationship, updated age-depth relationships as needed, and placed all sites on an internally consistent age model using available magnetostratigraphic and biostratigraphic datums. All core age models are relative to the standard geomagnetic polarity time scale [Cande and Kent, 1995] and biostratigraphic datums for the Cenozoic [Berggren et al., 1995].

2.2. Controls on DIC and Foraminiferal δ13C Values

[12] Foraminiferal carbonate (δ13Ccc) is in approximate isotopic equilibrium with dissolved inorganic carbon (δ13CDIC), roughly composed of 91% bicarbonate (HCO3), 8% carbonate (CO32−) and 1% carbonic acid (H2CO3) [Zeebe and Wolf-Gladrow, 2001]. Changes in the proportion of the dissolved carbonate species will influence foraminiferal δ13Ccc values, and such changes are expected over time due to changes in the saturation state of calcium carbonate. If ocean pH changed during the Cenozoic from 7.6 to 8.2 in concert with changes in the ocean Mg/Ca ratio and the saturation state of CaCO3 [Tyrrell and Zeebe, 2004], we estimate an isotopic effect of <−0.7‰ for δ13CCO2 over tens of millions of years. Therefore, for simplicity, we assume that the proportion of each carbonate species is fixed over the Cenozoic.

[13] The balance between photosynthesis and air-sea mixing [Gruber et al., 1999; Lynch-Stieglitz et al., 1995] governs surface water δ13CDIC values. Photosynthesis and export productivity preferentially remove 12C and lead to 13C-enriched surface water DIC. In addition, cold surface water interactions with atmospheric CO2 in high-latitude regions decrease the 13C/12C ratio of DIC [Broecker and Maier-Reimer, 1992]. These processes produce relatively stable spatial surface water δ13CDIC distributions that are offset from chemical equilibrium [Gruber et al., 1999; Lynch-Stieglitz et al., 1995]. For example, North Atlantic and Southern Ocean surface water δ13CDIC values are 1.0 ± 0.2‰ (1σ) more negative than those predicted by equilibrium calculations (Table 2).

Table 2. Coupled Surface and Bottom Water δ13CDIC Measurements, Air-Sea Surface Disequilibrium and Biologic Pump Offsets
 Measured Surface δ13CDICExpected Surface δ13CDICMeasured Bottom δ13CDICAir-Sea Surface DisequilibriumBiologic PumpSourcesa
Mean (n)2.0 (25)3.0 (23)0.8 (18)1.01.223, 24
Standard Deviation0.20.20.40.20.4 

[14] Export productivity and remineralization of organic carbon leads to water column δ13CDIC gradients [Kroopnick, 1985; Lynch-Stieglitz et al., 1995]. North Atlantic and Southern Ocean deep-water δ13CDIC values proximal to modern downwelling zones are 1.2 ± 0.4‰ (1σ) more negative than surface water δ13CDIC (Table 2), and then become increasingly more negative with advection and age. Consequently, knowledge of past ocean circulation is required in order to restrict our evaluation of δ13CCO2 to downwelling ocean sites with “young” 13C-enriched deep waters. Early Cenozoic deep waters were primarily sourced from the Southern Ocean [Via and Thomas, 2006; Wright and Miller, 1993], with the gradual increase of northern-sourced deep waters from 33 Ma to present [Davies et al., 2001; Via and Thomas, 2006]. Accordingly, we use Southern Ocean benthic foraminiferal δ13C values between 65 and 33 Ma, and Northern Atlantic Ocean records between 33 Ma-present.

[15] Finally, the average infaunal benthic foraminifera δ13Ccc values are 0.6 ± 0.3‰ (1σ) more negative than bottom water δ13CDIC values (Table 3). This mean value and its associated uncertainty are utilized in our long-term δ13CCO2 reconstruction.

Table 3. Coupled Extant Cibicidoides Species δ13CCC and Bottom Water δ13CDIC Measurements
 Measured Cibicidoides spp. δ13CCCMeasured Bottom δ13CDICΔδ13C (Remineralization/Oxidation)Sourcesa
Mean (n)0.3 (23)0.9 (23)0.625, 26, 27, 28, 29
Standard Deviation0.20.10.3 

2.3. Reconstruction of δ13CCO2 From Benthic Foraminifera δ13C

[16] The carbon isotopic composition of carbon dioxide in equilibrium with dissolved inorganic carbon is expressed as

  • equation image

where ɛDIC-CO2(g) represents the empirical temperature-dependent fractionation factor between dissolved inorganic carbon and CO2(g) [Zhang et al., 1995]:

  • equation image

where

  • equation image
  • equation image

Carbon isotope compositions and ɛ values are expressed in per mil notation relative to the Pee Dee Belemnite (PDB) standard.

[17] Temperature is estimated from the oxygen isotopic composition of foraminiferal calcite following Erez and Luz [1983]:

  • equation image

where T is the calcification temperature, δ18Ow is the oxygen isotopic composition of seawater, and δ18Occ the isotopic composition of the benthic foraminifera. High-latitude surface and deep-water temperatures are similar in downwelling zones [Millero, 2006], thus we assume calculated calcification temperatures from benthic species are similar to surface temperatures. Reconstructions of continental ice volumes [Miller et al., 2005] were applied to estimate δ18Ow through time. Our nominal solution uses δ18Ow values of 0.0‰, −0.5‰, and −1.2‰ from the Present-5 Ma, 5–33 Ma, and 33–65 Ma, respectively. To evaluate the uncertainty in reconstructed deep-water temperature associated with (1) our model for δ18Ow changes, and (2) potential diagenetic alteration of δ18Occ, we present a series of sensitivity analyses in section 4.2.

[18] It should be noted that calculated δ13CCO2 values are expected to be relatively insensitive to uncertainties related to estimated water temperatures. A ± 1‰ δ18Occ uncertainty reflects ∼±4°C (from equation 6), corresponding to < ± 0.5‰ variation in δ13CCO2. Considering that the entire Cenozoic range of δ18O change is less than 6‰, our assumption of a three-step change in δ18Ow makes a negligible difference in the calculation of δ13CCO2. To account for surface water δ13CDIC distributions, export productivity, and the bacterial oxidation of organic carbon [Corliss, 1985], δ13C values were increased by 1.0‰, 1.2‰, (Table 2), and 0.6‰, respectively. The above considerations yield the following expression for our δ13CCO2 reconstruction from published benthic foraminifera [Zachos et al., 2001]:

  • equation image

where A represents the sum total of disequilibrium effects from Tables 2 and 3 (surface water DIC disequilibrium, biological pump, and subsurface bacterial oxidation of organic material) related to biogenic carbonate production, equivalent to 2.8‰ for North Atlantic Ocean and Southern Ocean benthic foraminifera. The carbon isotope fractionation between calcium carbonate (calcite) and dissolved inorganic carbon is independent of temperature [Romanek et al., 1992]:

  • equation image

We recognize that the magnitude of isotopic offsets applied could have been somewhat different in the past due to changes in ocean circulation and water mass production rates, but the range of uncertainty on δ13CCO2 attributable to these effects is relatively small (see further discussion in section 4.2).

2.4. Reconstruction of δ13CCO2 From Planktonic Foraminifera δ13C

[19] δ13CCO2 can also be estimated from planktonic foraminifera δ13C and δ18O. For this work, no adjustments for export productivity or bacterial oxidation of organic carbon were used for the surface-dwelling species. Given the highly localized nature of the ecosystems and microhabitats of planktonic foraminifera, it is difficult to constrain such factors in our global data set. The following relationship was used to estimate δ13CCO2:

  • equation image

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

3.1. Benthic Foraminifera-Based δ13CCO2 Reconstruction

[20] Both planktic and benthic δ13Ccc varied significantly over the Cenozoic, and it is likely δ13CCO2 did so also (Figures 2 and 4 and auxiliary material). Measured δ13CCO2 values from ice cores over 50,000 years average −6.7‰ (range: −6.3 and −7.7‰) [Elsig et al., 2009; Friedli et al., 1986; Leuenberger et al., 1992; Smith et al., 1999], in accord with our benthic-based δ13CCO2 results that average −6.8‰ (range: −6.1 and −7.4‰) (Figure 3). Furthermore, the benthic foraminifera-based δ13CCO2 estimates accurately reconstruct measured δ13CCO2 values for approximately 95% of the record (Figure 3). This is in good agreement with our statistical procedure, as we should expect that ∼5% of the time the true δ13CCO2 is not captured by the 95% confidence interval. Some of this disagreement may also be attributable to the fact that North Atlantic benthic foraminifera form their carbonate shells in bottom waters that were in contact with the atmosphere 50–400 years previously [Broecker and Peng, 1982; Broecker et al., 1985]. Regardless, the observed accuracy across this time of profound climatic and oceanographic changes provides confidence that the assumptions we apply in our model are robust and applicable to more ancient geologic time intervals.

image

Figure 3. Given is δ13CCO2 over the last 50 kyr. Direct measurements of δ13CCO2 from CO2 trapped in ice-core bubbles (dotted line) and benthic-based (blue) δ13CCO2 estimates are shown with 95% confidence envelopes. Gray boxes surrounding the benthic-based δ13CCO2 reconstructions greater than 10 kyr in age indicate a nominal temporal uncertainty of +/−1 kyr. The pre-industrial δ13CCO2 average of −6.5‰ is shown with a dashed line [Marino et al., 1992].

Download figure to PowerPoint

[21] Moving average δ13CCO2 values for the entire Cenozoic are shown with 90% confidence intervals that incorporate the uncertainty associated with non-equilibrium effects and mean δ13Ccc estimates (Figure 4 and Appendix A). Due to the widely variable sampling resolution of the benthic δ13Ccc data, we dynamically adjust the number of points in the moving window in order to maintain a fixed temporal duration of 3 Myr. This approach contrasts with the typical n-point moving average, and is advantageous in this study because it more faithfully represents the true uncertainty associated with long-term changes in the average δ13Ccc; where data resolution is good, the 3 Myr uncertainty estimates are smaller because the mean value is well constrained. This approach provides a more geologically useful proxy for terrestrial paleoclimate and paleoecology applications (auxiliary material).

image

Figure 4. Cenozoic reconstructions of δ13CCO2 from planktonic and benthic foraminiferal records. A 3 Myr-moving average benthic-based δ13CCO2 reconstruction is shown in blue with 90% confidence intervals, and a 3 Myr-moving average planktonic-based δ13CCO2 reconstruction is shown in red. No confidence intervals are assigned to the planktonic record because non-equilibrium factors affecting individual planktonic species cannot be rigorously quantified. The pre-industrial δ13CCO2 average of −6.5‰ is shown with a dashed line [Marino et al., 1992].

Download figure to PowerPoint

[22] The benthic-based δ13CCO2 record averages −6.1 ± 0.6‰ (1σ) for the entire Cenozoic, similar to the commonly used pre-Industrial value of −6.5‰ (Figure 4). However, δ13CCO2 also displays significant variations throughout the record, at times departing from the pre-Industrial value by more than 2‰. The most notable 13C-enrichment (reaching −4‰) occurs between 61.5 to 52.8 Ma. Between 52.8 and 49.6 Ma, δ13CCO2 was more positive by ∼1‰ and returned to the Cenozoic average value (−6.1‰) for the following ∼32 Myr. The second largest enrichment, up to −5.3‰, occurs during peak warming of the Middle Miocene Climatic Optimum (17.3 to 14.5 Ma). δ13CCO2 returned to −6.0 ± 0.5‰ between 12.9 to 7.9 Ma. From 7.9 to 5.3 Ma δ13CCO2 values were ∼0.5‰ more negative, and from 7.9 to 1.5 Ma δ13CCO2 is characterized by a general decrease, achieving its lightest values of −6.6‰ in the most recent portion of the reconstruction. The Middle Miocene episode of 13C-enrichment is particularly important for studies of ancient terrestrial-plant ecosystems, because it overlaps with the presumed time of C4 grassland development (see section 4.3).

3.2. Planktonic Foraminifera-Based δ13CCO2 Proxy Record

[23] Planktonic δ13Ccc are more variable than the benthic δ13Ccc record (Figure 2) and the 3 Ma moving average planktonic-based δ13CCO2 time series is shown in Figure 4 (data provided within the auxiliary material). Confidence interval estimates are not provided because non-equilibrium factors affecting individual planktonic species cannot be rigorously quantified in a global compilation of this type. Planktonic δ13Ccc captures similar trends as the benthic δ13Ccc record with an average +1‰ offset. Planktonic-based δ13CCO2 reconstructions average −6.9 ± 1.2‰ (1σ), ∼0.4‰ more negative than the pre-Industrial value of −6.5‰ and 0.8‰ more negative than the benthic foraminifera-based δ13CCO2 record (Figure 4). From 26.4 to 30.0 Ma a notably large offset (1–2‰) between planktonic- and benthic-based δ13CCO2 reconstructions is observed (Figures 2 and 4). The planktonic-based δ13CCO2 estimates during this interval are from a single species, Globoquadrina venezuelana at an equatorial Pacific locality (ODP 1218). While few studies of depth habitat have been performed for Oligocene species, limited data suggests G. venezuelana may live in the upper thermocline resulting in slightly more positive δ18Occ and more negative δ13Ccc values than true surface-dwelling species [Wade and Pälike, 2004]. Nonetheless we include this data as it covers a large gap in the planktonic record.

4. Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

4.1. Comparison Between Benthic and Planktonic Foraminifera-Based δ13CCO2

[24] Shallow-dwelling planktonic species are closely associated with the atmospheric CO2 reservoir, but are subject to a variety of biologic, diagenetic, and temperature effects that vary in time and space and are difficult to constrain. Further, the planktonic δ13C records in our complication have more limited temporal resolution and more isotopic variability (Figure 2) that add uncertainty to δ13CCO2 estimates. For example, calculations of δ13CCO2 performed for the Paleocene-Eocene Thermal Maximum use planktonic isotope records from ODP Sites 690 and 1209 [Smith et al., 2007], and result in differences of 1.5–2.5‰ in the δ13CCO2 value. Additional variability from Site 1209 is evident when δ13CCO2 is estimated from both δ18Occ and Mg/Ca temperature estimates, yielding a difference of 1.0‰ in reconstructed δ13CCO2 values. Further, temperature differences between planktonic foraminifera δ18Occ and the molecular temperature proxies U37K′ and TEX86 suggest surface water temperature differences of 3–5°C for the latest Eocene [Liu et al., 2009], resulting in 0.4–0.6‰ variations in δ13CCO2 depending on which temperature estimate is applied.

[25] The influence of photosymbionts, depth of production, and seasonal variability also affect the isotopic composition of planktonic foraminifera. Spero [1992] demonstrated that planktonic foraminifera grown under high-irradiance are as much as 3.7‰ more 13C-enriched than the same species grown under ambient light. Seasonal variability and depth of production are difficult to constrain and lead to isotopic scatter in planktonic δ13Ccc and δ18Occ data (Figure 2). Furthermore, the shallowest-dwelling species are not always available or easily identified in marine cores, and isotopic variability of shallow assemblages changes through time as new species emerge and others become extinct [Hilting et al., 2008]. Since benthic species are not associated with photosymbionts and live exclusively in bottom waters they are free of these confounding effects.

[26] Finally, δ18Occ values of planktonic foraminifera can be altered due to recrystallization and incorporation of secondary calcite after burial. Diagenetic recrystallization and secondary cements increase the δ18Occ value of planktonic foraminifera, yielding lower apparent temperatures [Schrag, 1999]. While these diagenetic processes affect both planktonic and benthic foraminifera carbonate, they do not impact benthic isotope values to the same degree as benthic foraminifera are in near equilibrium with bottom water conditions.

[27] Given the large variability observed in the raw planktonic δ13Ccc values (Figure 2) and the lack of a means to quantify the precise causes of this variability over large time intervals, we cannot assign reliable confidence intervals to these δ13CCO2 estimates. We conclude that the benthic foraminiferal record provides a better estimate of ancient δ13CCO2 over multimillion year timescales.

4.2. Sensitivity Analysis of Bottom Water Temperatures and Diagenesis

[28] To evaluate the impact of uncertainty in bottom water temperature on δ13CCO2 estimates, we present two sensitivity analyses. The first of these sensitivity analyses utilizes a constant δ18Ow value of 0.0‰. If the extreme assumption of a static δ18Ow (0.0‰) is correct, δ13CCO2 values are more 13C-enriched by 0.2‰ between 5 and 33 Ma, and 0.5‰ between 33 and 65 Ma.

[29] In our nominal solution for δ13CCO2 values, we assume that temperatures calculated from benthic foraminifera δ18Occ values are not subject to post-burial alteration, however, a general cooling trend is apparent across the Cenozoic. Thus, Paleogene-aged benthic foraminifera could have interacted with cooler Neogene bottom waters, promoting secondary alteration. Our second sensitivity analysis evaluates the influence of δ18Occ diagenesis by decreasing the bottom water temperatures by 2.5°C and increasing the measured δ18Occ values by 0.5‰ from 5 to 65 Ma. A 2.5°C increase in bottom water temperature decreases the nominal solution by 0.2‰ from 5 to 65 Ma (Figure 5). This sensitivity analysis does not include the last 5 Myr, which was the coldest of the Cenozoic Era, and thus bottom water temperatures would have had little effect on altering original δ18Occ values.

image

Figure 5. Sensitivity analysis of δ18Occ alteration and source water δ18Ow variations on benthic-based δ13CCO2 reconstructions. Nominal solution for δ13CCO2 from Figure 4 shown with blue circles, while solutions for δ18Occ alteration and source water δ18Ow variations are shown with red and grey circles respectively. The 90% confidence intervals for δ18Occ alteration and source water δ18Ow solutions are shown with beige and blue envelopes. Brown regions indicate where both alteration and source water confidence intervals overlap.

Download figure to PowerPoint

[30] The modeled variations in δ18Ow and δ18Occ diagenesis result in opposing effects in the calculation of δ13CCO2. As shown in Figure 5, the nominal solution is intermediate between the δ18Ow and δ18Occ diagenesis sensitivity analyses results. The entire δ13CCO2 range spanned by the 90% confidence intervals clearly excludes a pre-Industrial δ13CCO2 value of −6.5‰ for much of the record. Furthermore, these results reveal that the general trends observed in the nominal solution are robust.

4.3. Impact of Atmospheric δ13CCO2 Variations on Terrestrial Plant Reconstructions

[31] The rise of C4 grassland in the Late Miocene/Pliocene, inferred from carbon isotope compositions of paleosols, soil carbonates, and fossil ungulate teeth [Cerling et al., 1997; Fox and Koch, 2003, 2004; Kingston et al., 1994; Kleinert and Strecker, 2001; Morgan et al., 1994; Passey et al., 2009, 2002], is arguably the most important terrestrial ecological event of the Neogene, with terrestrial C4 expansion driven by a combination of regional climatic controls preconditioned by low pCO2 levels [Pagani et al., 1999; Tipple and Pagani, 2007]. Quantitative estimates of C4 plant abundances are based on the carbon isotopic composition of terrestrial carbonates and/or organic matter, and require knowledge of the average C3 and C4 plant carbon-isotope compositions at any particular time. Our record indicates that δ13CCO2 was −6.0‰ between 8 and 10 Ma and −5.2‰ during the Middle Miocene, roughly 0.5‰ and 1.3‰ more positive than the pre-Industrial value. Accordingly, average C3 and C4 plant carbon isotope compositions during these times were similarly 13C-enriched.

[32] The presence of C4 plants during the Early and Middle Miocene has been inferred from molecular clock data [Christin et al., 2007; Vicentini et al., 2008], with the Middle Miocene suggested to have been a major interval of C4 diversification [Vicentini et al., 2008]. Our benthic-based δ13CCO2 reconstruction indicates that most published paleoecologic records covering these time periods are biased toward greater C4 distribution because changes in δ13CCO2 were not accounted for (Figure 4). For example, if a pre-Industrial δ13CCO2 value is used to calculate C4 contribution during the Middle Miocene, soil carbonate records from East Africa indicate between 10 and 36% C4 biomass [Kingston et al., 1994]. However, if the reconstructed δ13CCO2 value of −5.2‰ is used for this interval, then C4 contribution decreases to 0–18%. While these records continue to indicate C4 plants were present during the Middle Miocene, new δ13CCO2 reconstructions suggest C4 grasses were less abundant than initially argued.

[33] Important changes in δ13CCO2 appear to occur during the past ∼10 Ma (Figures 2 and 4). The observed negative trend in δ13CCO2 during this time has been previously interpreted as reflecting weathering of organic-rich rocks [Shackleton, 1987] — a supposition further supported by 187Os/186Os records [Ravizza, 1993]. However, other processes, such as a global expansion of terrestrial C4 grasses and marine C4-like pathways would have also impacted ocean δ13C and δ13CCO2 records [Katz et al., 2005; Kump and Arthur, 1999]. Whatever the origin of the decrease in δ13Ccc and δ13CCO2, our data suggests that terrestrial ecologic reconstructions using carbon isotope ratios should include changes in δ13CCO2 value in their discussions of plant community change.

5. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

[34] Two Cenozoic reconstructions of δ13CCO2 are established from a meta-analysis of planktonic and benthic foraminiferal δ13C and δ18O records. We find that planktonic records have limited utility in the reconstruction of δ13CCO2 across long geologic timescales due to lower temporal resolution and higher isotopic variability evident in modern species. High-resolution benthic foraminifera δ13C and δ18O records constrain δ13CCO2 values using a systematic treatment of environmental, diagenetic and species-specific effects, along with improved age models. Our results indicate that the Cenozoic was characterized by an average δ13CCO2 value of −6.1‰ (∼0.4‰ more positive than Pleistocene and pre-Industrial values). The resulting δ13CCO2 record indicates pre-Industrial δ13CCO2 and plant δ13C values should not be assumed for most of the Cenozoic.

Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

[35] Atmospheric δ13CCO2 values and associated uncertainty are constrained by estimating mean values and 90% confidence intervals associated with four distinct variables: (1) benthic foraminiferal δ13C (δ13Ccc), (2) the offset between δ13Ccc and deep-water δ13C, attributable to the incorporation DIC from early diagenetic organic carbon remineralization (δ13Ccc-deep), (3) the surface-deep water offset associated with the biological pump (δ13Csurf-deep), and (4) the offset due to air-surface water disequilibrium (δ13Cair-surf).

[36] Three million year moving averages are calculated (Figure 4), following the application of species-specific corrections to benthic foraminiferal δ13C measurements (Figure 2). In the algorithm utilized here, the number of points in the moving window is adjusted in order to maintain a fixed temporal duration. This approach is advantageous in this study, because the δ13Ccc data set has widely variable sampling resolution. The 90% confidence interval for each 3 million year average is estimated as

  • equation image
  • equation image

where equation image is the 3 million year average, ucc is the uncertainty, s is the standard deviation, n is the number of data points in the window, and t* is the upper 0.05 critical value associated with the t(n-1) distribution.

[37] The 90% confidence intervals for equation image, equation image, and equation imageare estimated in a similar fashion, using published δ13C data (see Tables 2 and 3). The corresponding uncertainty estimates are designated ucc-deep, usurf-deep and uair-surf. To obtain a total uncertainty (utotal), the individual uncertainties are added in quadrature [Taylor, 1982]:

  • equation image

This total uncertainty is utilized to in equation 7 to estimate δ13CCO2 uncertainty from the measured benthic foraminiferal δ13C data. Benthic temperature estimates determined in equation 6 utilize δ18Occ from this same set of benthic foraminifera, and are also smoothed with the 3 Myr moving average procedure.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

[38] We wish to thank M. Hren, K. Turekian, and D. Zinniker for helpful discussions. In addition, we would like to acknowledge P. Koch, J. Hayes, L. Kump, G. Dickens, B. Passey, and two anonymous reviewers for their thoughtful and constructive comments that greatly improved this work.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Summary of Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Appendix A:: Atmospheric δ13CCO2 Confidence Interval Estimation From Benthic Foraminifera δ13C
  9. Acknowledgments
  10. References
  11. Supporting Information

Auxiliary material for this article contains a data set, one figure, and a text file.

Auxiliary material files may require downloading to a local drive depending on platform, browser, configuration, and size. To open auxiliary materials in a browser, click on the label. To download, Right-click and select “Save Target As…” (PC) or CTRL-click and select “Download Link to Disk” (Mac).

See Plugins for a list of applications and supported file formats.

Additional file information is provided in the readme.txt.

FilenameFormatSizeDescription
palo1616-sup-0001-readme.txtplain text document2KReadme.txt
palo1616-sup-0002-ds01.txtplain text document201KData Set S1. The data set gives δ13CCO2 values reconstructed from planktonic and benthic foraminifera.
palo1616-sup-0003-fs01.tiffTIFF image17315KFigure S1. A 0.5 Myr moving average reconstruction of δ13CCO2 from planktonic and benthic foraminiferal records.
palo1616-sup-0004-txts01.txtplain text document6KText S1. Text S1 provides data sources and citations.
palo1616-sup-0005-t01.txtplain text document2KTab-delimited Table 1.
palo1616-sup-0006-t02.txtplain text document0KTab-delimited Table 2.
palo1616-sup-0007-t03.txtplain text document0KTab-delimited Table 3.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.