Temperature reconstructions indicate that the Pliocene was approximately 3°C warmer globally than today, and several recent reconstructions of Pliocene atmospheric CO2 indicate that it was above pre-industrial levels and similar to those likely to be seen this century. However, many of these reconstructions have been of relatively low temporal resolution, meaning that these records may have failed to capture variations associated with the 41 kyr glacial–interglacial cycles thought to have operated in the Pliocene. Here we present a new, high temporal resolution alkenone carbon isotope-based record of pCO2 spanning 3.3–2.8 Ma from Ocean Drilling Program Site 999. Our record is of high enough resolution (approx. 19 kyr) to resolve glacial–interglacial changes beyond the intrinsic uncertainty of the proxy method. The record suggests that Pliocene CO2 levels were relatively stable, exhibiting variation less than 55 ppm. We perform sensitivity studies to investigate the possible effect of changing sea surface temperature (SST), which highlights the importance of accurate and precise SST reconstructions for alkenone palaeobarometry, but demonstrate that these uncertainties do not affect our conclusions of relatively stable pCO2 levels during this interval.
The Pliocene was the most recent epoch in the Earth's history that had global temperatures greater than today; this coupled with the similar continental positions and vegetation cover to the present has led to interest in the Pliocene as a possible analogue for the warmth expected by the end of the century . Research to constrain the global temperatures of the Pliocene has been ongoing for some time, including the significant contributions of the Pliocene Research, Interpretation and Synoptic Mapping (PRISM) project and successors (e.g. [2,3,4]), with the consensus that the Pliocene was globally approximately 3°C warmer than today. Similarly, many studies suggest that Pliocene pCO2 was also higher than pre-industrial levels [5,6,7,8].
Despite these similarities, a significant difference between the Pliocene and the present day is the magnitude and pacing of Pliocene glacial–interglacial changes. Based on the foraminiferal carbonate δ18O record , Pliocene glacial–interglacial cycles were less pronounced than those of the late Quaternary, and they were 41 kyr long in contrast to the 100 kyr cycles of the last 0.7 Ma. Variations in pCO2 greater than approximately 100 ppm fluctuations of the Pleistocene are, therefore, unexpected in a Pliocene world, especially if, as suggested by Pagani et al. , Pliocene Earth-system sensitivity was probably greater than 3°C for a CO2 doubling. Consistent with this, alkenone-based pCO2 reconstructions have shown very little glacial–interglacial variation, especially prior to the intensification of Northern Hemisphere glaciation at 2.8 Ma. These records have been, in part, confirmed by boron isotope-based reconstructions . However, owing to their low temporal resolution, it is possible that they fail to capture higher frequency variability in pCO2 and represent neither the mean state of the climate system nor its variability on glacial time scales well. A recent reconstruction using boron isotopes  has increased the temporal resolution, and in fact targeted specific glacial and interglacial peaks and troughs to attempt to resolve this. Their record exhibits fluctuations in Pliocene pCO2 that are in fact larger (almost one and a half times) than those observed in Pleistocene ice core records [10,11,12].
To address this apparent discrepancy, we reconstruct pCO2 at similarly high resolution from 3.3 to 2.8 Ma using alkenone δ13C values. We apply our approach to Ocean Drilling Program (ODP) Site 999 in the Caribbean because this is the same site used by Bartoli et al.  and Seki et al. , allowing direct comparison of our records. Additionally, we also conduct a sensitivity analysis of our reconstructed pCO2 levels, allowing us to constrain their potential range during this time.
2. Approach and methods
(a) Alkenone palaeobarometry
The isotopic fractionation between dissolved inorganic carbon (DIC) and marine organic matter during photosynthesis (εp) is controlled by, among other factors, the concentration of CO2 in the water in which the organism is photosynthesizing ([CO2(aq)]). This is ultimately controlled by the concentration of CO2 in the overlying atmosphere with which the ocean is in equilibrium. Other factors that can affect εp values include physiological factors, such as cell geometry  and membrane permeability , and environmental factors, such as nutrient and light availability and their impact on carbon demand (i.e. growth rate) and carbon assimilation mechanisms [15–18].
In order to constrain the physiological factors, biomarkers derived from a narrow taxonomic range can be used rather than bulk organic matter. This approach also prevents terrestrial or non-photosynthetically produced organic matter from biasing the marine organic matter isotopic signature . Long-chain ketones (alkenones) containing 37 carbons are produced only by a restricted group of haptophyte organisms, for instance Gephyrocapsaceae coccolithophores . Thus, work over the past 20 years has focused specifically on the alkenone palaeobarometer as a tool to reconstruct ancient atmospheric pCO2, so long as other contributing factors (growth rate, light regime) can be constrained. In order to determine εp values, the isotopic composition of both the DIC pool and organic biomass must be known. The isotopic composition of the organic biomass (δ13Corg) is calculated from the alkenone δ13C (δ13C37:2), corrected for a fractionation between alkenone and haptophyte biomass by assuming a constant fractionation of 4.2‰ [13,16] 2.1The isotopic composition of DIC is estimated by measuring the δ13C value of planktic foraminifera, assuming the experimentally determined temperature-dependent fractionation between calcite and CO2(g) (εcalcite−CO2(g)) shown in the following equation : 2.2where T is sea surface temperature (SST; in degrees Celsius). This fractionation factor can then be used to calculate the carbon isotopic composition of CO2(g) (δ13CCO2(g)), 2.3From this, the carbon isotopic composition of CO2(aq) (δ13CCO2(aq)) can be obtained using the experimentally determined relationship of Mook et al.  as shown in the following equations: 2.4and 2.5Photosynthetic fractionation (εp) can then be calculated from the determined and δ13Corg as follows: 2.6and this is then used to calculate [CO2(aq)] according to the following equation: 2.7where εf represents the isotopic fractionation during carbon fixation, assumed here to be constant and 25‰ . The ‘b’ term represents the summation of physiological factors, such as cell size and growth rate. In the modern ocean, this term shows a close correlation with [PO43−], allowing estimation of ‘b’ by assuming that past [PO43−] was similar to that present at the site today (0.2 μM; [6,16,23]). Finally, from [CO2(aq)], atmospheric pCO2 can be calculated using Henry's law (the following equation) and assuming equilibrium between the surface water and overlying atmosphere: 2.8The solubility coefficient (KH) is salinity and temperature dependent and calculated following the parametrization of Weiss [24,25]. The assumptions inherent in the above treatment are discussed further in §4c.
In this study, analytical determinations of εp values were conducted similar to those of previous alkenone palaeo-pCO2 studies (e.g. [6,7,23,26,27,28]) from ODP Site 999 (12°44.639′ N, 78°44.360′ W, 2838 m water depth). Site 999 is slightly out of equilibrium in the modern ocean, with surface water oversaturated in CO2 relative to the atmosphere, providing a small (less than 10 gCm2 yr−1; ) net source of CO2 to the atmosphere. However, the site has been shown to be capable of recording past changes in pCO2, and the air–sea equilibrium is not thought to have changed significantly from the Pliocene to today (see discussion in ). Specifically, 27 samples were freeze dried, ground by hand and solvent extracted either by Soxhlet apparatus or ultrasonically. Soxhlet extractions were performed using a dichloromethane (DCM)/methanol (MeOH) azeotrope (2:1, v/v), refluxing for 24 h. Ultrasonic extractions were performed with, sequentially, DCM, DCM/MeOH (1 : 1, v/v) and MeOH, repeated three times for each solvent with each extraction taking 15 min in an ultrasonic bath with approximately 15 ml of solvent each time. Supernatants were removed and combined before reduction by rotary evaporation and finally evaporated to dryness under a stream of N2. Following elution through small (4 cm) sodium sulfate columns to remove excess water, total lipid extracts were divided into apolar and polar fractions by means of alumina flash column chromatography using four column volumes of n-hexane/DCM (9:1, v/v) and three column volumes of MeOH, respectively. Alkenone concentrations were quantified by gas chromatography (GC) flame ionization detector (Hewlett Packard 5890 Series II) following trimethylsilyl derivatization. The GC oven was programmed to increase in temperature from 70°C to 130°C at 20°C min−1, then to 300°C at 4°C min−1 and finally being held isothermal for 25 min. The column was a CPSil-5CB (dimethylpolysiloxane equivalent), 0.12 μm film thickness, approximately 50 m length and 0.32 mm internal diameter with a H2 carrier gas. Alkenone identification was confirmed by GC mass spectrometry (ThermoQuest Trace MS, He carrier gas). Absolute compound concentrations were quantified by reference to an internal standard (hexadecan-2-ol) added prior to column chromatography.
SST was reconstructed using the alkenone unsaturation index (; [30,31]), 2.9where C37:2 is the di-unsaturated methyl alkenone and C37:3 is the tri-unsaturated compound. SSTs were then calculated using the calibration of Müller et al.  2.10Concerns have been raised about the use of as the index approaches 1 . This would be of particular concern at Site 999 as over the studied interval is greater than 0.9. However, the challenge of calibrating SSTs towards the upper limit of seems to be a problem largely restricted to sediment trap-based calibrations. For core tops, a linear calibration seems to hold true, and in fact the updated core-top calibration of Conte et al.  is essentially identical to that of Müller et al.  (which is more widely used and therefore our preferred parametrization). They are especially similar towards the top end of the scale. As we are dealing here with alkenones which have made it to the sea floor, a core-top calibration seems the most appropriate.
Alkenone isotope analyses were performed on a ThermoFisher Delta V connected via a GC isolink and ConFlo IV to a Trace GC. The GC oven was programmed to increase in temperature from 70°C to 200°C at 20°C min−1, then to 300°C at 6°C min−1 and finally held isothermal for 25 min. Conversion to the VPDB scale was calculated by reference to a laboratory standard gas tank of known δ13C. Instrument performance was monitored using an in-house fatty acid methyl ester standard and long-term precision is approximately 0.3‰.
Between 10 and 15 specimens of the planktic foraminifera Globigerinoides ruber were picked from the 300–350 μm fraction for δ13C analysis. This was determined with a Finnigan MAT 251 with an online automatic carbonate preparation device at the Alfred Wegener Institute for Polar and Marine Research, Bremerhaven, Germany. Calibration to the VPDB scale was performed using the international NBS19 standard. Reproducibility is better than ±0.06‰ over a 1 year period based on repeat measurements of a laboratory standard.
The age model for ODP Site 999 is as discussed in Seki et al. . Uncertainty propagation on our alkenone-derived CO2 estimates was performed by Monte Carlo modelling (n=25 000). Uncertainties of 2°C and 0.1‰ were applied to temperature and foraminiferal calcite δ13C (normal probability density function (pdf), 2 s.d. error), and 2 and 0.1 to salinity and [PO43−], respectively (2 s.d.; uniform pdf). Two s.d. errors on alkenone δ13C were estimated from replicate runs, calcite δ13C from repeat runs of an internal standard, estimated integrated analytical and calibration error for -based temperatures  and conservative estimates of likely variation for salinity and [PO43−]. An 11% error on the slope of b=a[PO4]+c was assumed .
Alkenone and G. ruber δ13C values (figure 1b) were used to calculate εp values (figure 1a). These εp values are fairly stable throughout the study interval, varying between 12.2‰ and 9.4‰. This is close to the values reported by Seki et al.  for the same site over this interval (12.2–10.9‰).
Using modern [PO43−] for the Caribbean Sea, εp values can be converted to [CO2(aq)] (equation (2.7); figure 1b). Using our SSTs derived from indices and assuming air–sea equilibrium, [CO2(aq)] can then be used to determine atmospheric pCO2 (equation (2.8)). indices range from 0.90 to 0.99 (close to the maximum recordable value for ), resulting in SSTs at Site 999 of approximately 28°C that show a slight decrease over the 500 kyr of our record (figure 2b). These are approximately 2°C higher than the planktic foraminifer (Globigerinoides sacculifer) Mg/Ca-based SST record of Groeneveld  from the same site, less than 1°C lower than the SSTs estimated by Bartoli et al.  based on a seawater Mg/Ca correction of these same data and very similar to modern SSTs that range from 26.7°C to 28.2°C .
Our resulting pCO2 reconstruction (figure 2a) reveals relatively stable pCO2 values that are within the range of previously published alkenone records from ODP Site 999 (without the lith size correction of Seki et al. ) and elsewhere . All of our reconstructed pCO2 levels (250–300 ppm) are similar to or slightly higher than the 240–290 ppm for Pleistocene interglacials reconstructed from ice cores [10,11,12] and are consistent with glacial–interglacial variability of at most 40 ppm. In fact, the entire range of determined pCO2 values for the end of the Pliocene is less than the 80 ppm difference between the Holocene and the Last Glacial Maximum . There is some variability outside of uncertainty in the εp values in the younger part of the record, hinting to some variability after 3 Ma; however, once the full propagation of uncertainties is taken through to the CO2 reconstruction, the variation is no longer significant. Below, we discuss the pCO2 estimates, their variations with respect to Pliocene glacial–interglacial cycles and the potential range of pCO2 given our assumptions of growth rate and SST.
(a) Glacial–interglacial pCO2 variations
We estimate absolute pCO2 to be approximately 270 ppm for much of the period studied here, based on our most likely temperature, cell geometry and growth rate assumptions (see subsequent sections for sensitivity analysis of these parameters). This is similar to pre-industrial levels and around the peak level of the Pleistocene ice core records (298.6 ppm; [10,11,12]). Our record is within the range of estimates given by Pagani et al.  (figure 3a), although it should be noted that these authors report a broad range of absolute CO2 owing to differences between the sites. Our record is below the ‘CO2slope’ reported in Pagani et al. , i.e. their extrapolated trend from the early Pliocene to the present day.
Estimating absolute pCO2 from a single site is complicated by uncertainty as to whether the site has been in equilibrium with the atmosphere over the period of interest. As highlighted by Pagani et al. , different sites can exhibit very different estimates for atmospheric pCO2, as not all of the surface ocean is in equilibrium with the atmosphere . The surface ocean at Site 999 is close to equilibrium today  and reconstructed alkenone-based pCO2 values are similar to ice core records where they overlap in the Pleistocene , suggesting that the site was in equilibrium through much of this time. It is difficult to know whether this remained so in the Pliocene with different circulation in the Caribbean, so, as with all single-site records, our absolute pCO2 should be treated with some care.
Our absolute pCO2 is similar to the alkenone-based record without secondary corrections of Seki et al.  from the same site, although somewhat lower than both the cell size-corrected alkenone record and boron isotope-based records of Seki et al.  (figure 3a). Bartoli et al.  report a broad range of pCO2 (170–400 ppm; figure 3b), and our record is within that range. The difference between our record and the cell size-corrected record of Seki et al.  (figure 4) highlights the importance of secondary corrections, particularly on alkenone-based methods, and we explore this further below. Given the potential difficulty of assessing absolute pCO2 levels from single-site records, we now focus on pCO2 variability during this interval in the Pliocene.
Previous alkenone-based palaeobarometry has been at a relatively low temporal resolution, and given the 41 kyr glacial–interglacial variability in the Pliocene world, it is possible that these records do not capture rapid changes in pCO2 [6,7]. Our new record increases the resolution of the alkenone-based records, but unlike the boron record of Bartioli et al.  shows virtually invariant pCO2 within the precision of the alkenone palaeobarometer (figure 4). The differences between these two records cannot be due to differences in ocean–atmosphere equilibrium, as both are based on Site 999. The magnitude of variability in our record is similar to that seen in previous, low-resolution records (i.e. the boron and alkenone records of Seki et al.  and the alkenone records of Pagani et al. ), which may suggest that these records have captured pCO2 variability despite their lower resolution, and the small estimated range of Pliocene pCO2 is a feature of Pliocene climate dynamics rather than a sampling artefact. An alternative hypothesis is that the alkenone palaeobarometer underestimates variability for an as yet unknown reason.
It should perhaps not be surprising that Pliocene pCO2 appears to be relatively stable; the large, 100 kyr glacial–interglacial cycles of the Pleistocene are associated with approximately 100 ppm of change in pCO2 [10–12], and it is likely that the smaller amplitude variations in the Pliocene would be associated with significantly smaller pCO2 changes. The large-amplitude changes of Bartoli et al.  are therefore somewhat surprising. Given that the full uncertainty envelope for our alkenone pCO2 records is approximately ±40 ppm, it is plausible that smaller Pliocene variations in pCO2 would be below the detection limit of our methods.
(b) Cell size and productivity corrections
Atmospheric pCO2 reconstructions from alkenone isotopes can be affected by cell size and productivity, i.e. growth rate variations. Seki et al.  applied a conceptual cell size correction to the alkenone data from Site 999, based on the low-resolution lith size record of Kameo & Bralower . However, recent high-resolution data show no evidence of changes in coccolith size over the time interval of interest . Crucially, there are no changes in coccolith size—and, thus, inferred coccolithophorid cell size—on glacial–interglacial time scales, and hence it is unlikely that they could account for the low variability observed here. It remains possible, however, that they could account for the relatively low absolute pCO2 values determined for the Pliocene at Site 999, compared with δ11B-based estimates (approximately 400 ppm; [7,8]).
There is not yet a consensus approach to the application of a cell geometry correction (see discussion in [7,41]); however, attempts have been made to correct for cell size changes by adjusting the ‘b’ term in equation (2.7). Hendericks & Pagani  adjusted the ‘b’ term based on the ratio of ‘fossil’ haptophyte cell
volume/surface area (V : SAfossil) to that of the modern Emiliania huxleyi (V : SAE . hux) used in culture studies 4.1Popp et al.  determined V : SAE . hux to be 0.9±0.1 μm and the value of V : SAfossil can be estimated using the relationship between cell diameter (Dcell) and Reticulofenestra coccolith length (Lcoccolith; ) 4.2Reticulofenestrids (Noelaerhabdaceae) are thought to be important alkenone producers in the past , although there is some evidence that this may not be the case for some earlier parts of the Neogene .
The cell size correction results in a linear correction to pCO2, the gradient of which is temperature dependent (figure 5), where a larger coccolith length results in higher reconstructed CO2. This effect is increased at higher SSTs. Approximately 1 μm of change in Lcoccolith would be required to alter pCO2 beyond our uncertainty envelope. This represents a size change of approximately 25%, and there is no evidence for such a change on glacial–interglacial time scales at Site 999 .
Similarly, there is no evidence that growth rate changed on glacial–interglacial time scales at Site 999 through the interval studied. Seki et al.  noted that there could have been changes in the oceanography of the site as the shoaling of the Panama isthmus isolated the Caribbean from the Pacific. O'Dea et al.  had argued that these changes could have influenced the nutrient regime at Site 999; however, the closure of the strait to deep water is thought to have been complete by 4.6 Ma . Moreover, alkenone and other biomarker concentrations and mass accumulation rates are low and relatively invariant over the studied interval . Other indicators of productivity (organic carbon mass accumulation rates) have not shown evidence of significant changes in productivity . We have improved the resolution of these analyses to 16 kyr, and no systematic variation is apparent, suggesting that on glacial–interglacial time scales no correction to our pCO2 reconstruction is justified.
No proxy exists to directly reconstruct the growth rate conditions at the site during the Pliocene. Our approach is to use the relationship between ‘b’ and [PO43−], which has been calibrated globally [6,23,45] and appears to be a proxy for growth-limiting nutrients . Our favoured assumption is to use a modern-day value for surface water [PO43−] at Site 999 , but here we explore the possibility that this assumption is incorrect. To this end, we have performed a sensitivity test and applied a [PO43−] that represents an oligotrophic site (0.05 μM), a high-nutrient area similar to present-day eastern equatorial Pacific (0.6 μM) and an extreme case representative of an active upwelling region (0.9 μM) . The resulting pCO2 reconstructions vary from approximately 230 ppm for the oligotrophic model, approximately 390 ppm for the eastern equatorial Pacific model and approximately 480 ppm for the extreme case (figure 6). The sensitivity tests demonstrate that our favoured assumptions that result in pCO2 of approximately 270 ppm may be a lower bound, and if the nutrient regime was significantly different in the Pliocene, then our record may be an underestimate. However, as discussed earlier, we have no direct evidence for such a change, and, critically, no evidence of changes on glacial–interglacial time scales that may explain the low variability of our record.
(c) The importance of accurate and precise temperature determinations
Critical to the validity of alkenone isotope pCO2 reconstructions is the accurate and precise determination of SST. While some of the previously discussed parameters (such as an evolutionary change in haptophyte cell size or significant changes in oceanographic regime leading to changes in growth rate) are likely to be stable over short periods of time, SSTs can change on glacial–interglacial time scales and affect pCO2 estimates. SST is involved three times in the reconstruction of pCO2 from alkenone δ13C values, in the conversion of δ13Ccalcite to values (equations (2.2) and (2.3)), δ13CCO2(g) to δ13CCO2(aq) values (equations (2.4) and (2.5)) and the calculation of the solubility coefficient (equation (2.8)). This results in a nonlinear temperature dependence of the δ13Calkenone–pCO2 relationship (figure 7). The size of this effect can be important in terms of both the accuracy and precision of alkenone-based reconstructions. Proxies that show potential to reconstruct SSTs suitable for alkenone palaeobarometry include those based on alkenone unsaturation indices [30,31] and planktic foraminiferal Mg/Ca ratios . Estimates of uncertainty in the measurement of SST using either Mg/Ca or alkenone unsaturation suggests a combined analytical and correlation error of approximately 2°C (2 s.d.; [48,49]).
The relationships between alkenone δ13C, εp, [CO2(aq)] and pCO2, and the effects of SST are shown in figure 7. Higher reconstructed SST results in higher apparent εp for a given δ13C value (figure 7a) and higher pCO2 for a given [CO2(aq)] (figure 7b). Higher reconstructed SST, therefore, results in higher apparent pCO2 for a given δ13C value by integrating these two effects (figure 7d). The magnitude of this effect is more pronounced at higher pCO2 and more negative alkenone δ13C values (figure 7d), and also as εp approaches εf (see the discussion in Pagani et al. ).
For example, for an alkenone δ13C value of −25‰ (which gives a representative pCO2 of 300 ppm in this sensitivity test), the 2°C analytical and calibration error in Mg/Ca or SST estimates would result in an error of approximately 23 ppm in pCO2; at a more negative δ13C value of −28‰ (pCO2=400 ppm), the same error in SST results in an error of approximately 34 ppm in pCO2 (figure 7d). One result of this is that either an incorrect or overestimated decline in SST can lead to an artificial apparent decline in pCO2. This requires careful consideration if estimating climate or the Earth-system sensitivity from coupled alkenone pCO2 and SST records. To apply this directly to the data presented here, SSTs 2°C cooler than our data suggest (i.e. at the edge of the quoted uncertainty for alkenone unsaturation-based temperatures) would result in an average reconstructed pCO2 over the interval studied of 255 ppm, a 15 ppm reduction. Conversely, SSTs 2°C warmer would give an average reconstructed pCO2 20 ppm higher, at 290 ppm.
For the Pliocene, the choice of SST record is important, given the uncertainty as to the possible effects of changing Mg/Casw on Mg/Ca palaeothermometry. Recent reconstructions of Pliocene Mg/Casw suggest that the Pliocene value could have been more than 1 mol mol−1 lower , which could change the reconstructed SST by as much as 6°C. Our preferred temperatures for Site 999 lie between the Mg/Casw corrected and uncorrected records for the same time period [8,35]. The uncorrected SST estimates of Groeneveld  are approximately 3.8°C lower and would result in pCO2 approximately 29 ppm lower if applied to our records, whereas the up to 2.5°C higher temperatures of Bartoli et al.  would increase our estimates by approximately 24 ppm. At Site 999 today SSTs vary between 26.7°C and 28.2°C, whereas temperatures at the habitat depth likely for G. sacculifer are 24.2–26.6°C . The cooler temperatures of Groeneveld  may therefore be due to differences in the depth habitat of the recording organism. Crucially, however, none of the SST records for Site 999 indicates glacial–interglacial variability that would result in pCO2 variations with a magnitude similar to those of the Pleistocene, or as recorded by Bartoli et al. .
We reconstruct atmospheric pCO2 for the Pliocene (3.3–2.8 Ma) of approximately 270±40 ppm (2 s.d.) similar to Pleistocene interglacials. We record little or no variability, suggesting that pCO2 was persistently at about Pleistocene interglacial values. Only at the outer bounds of our uncertainty envelope would we record Pleistocene glacial levels of pCO2. Uncertainty in our assumptions for productivity, SST and cell size all result in a broad uncertainty envelope around our preferred parametrization, with our best estimate suggesting that pCO2 was between approximately 230 and 300 ppm. These absolute values are lower than those derived from other approaches, and this could reflect a combination of local palaeoceanographic conditions and the impact of secondary effects on alkenone δ13C values. However, we see no evidence that such secondary effects were varying on glacial–interglacial time scales, and, consequently, our data collectively indicate that pCO2 at this point in the Pliocene was relatively stable. We see no evidence for glacial–interglacial changes larger than the fundamental precision of the proxy method of 40 ppm, and our record supports the idea that minimal pCO2 variability was associated with the small glacial–interglacial climate variability of the Pliocene. However, further work is needed to improve the precision of all proxy methods and to reconcile differences between records of Pliocene pCO2.
M.P.S.B. was funded by NERC grant no. NE/H006273/1, D.N.S. by a Royal Society URF. R.D.P. acknowledges the Royal Society Wolfson Research Merit Award.
This research used samples and data provided by the Ocean Drilling Program, which is sponsored by the US National Science Foundation and participating countries under the management of Joint Oceanic Institutions. We thank Alex Hull, Gemma Bowler and Marilyn Potts for laboratory work, Lisa Schönborn and Günter Meyer for technical assistance and Alison Kuhl and Ian Bull of the NERC Life Sciences Mass Spectrometry facility for research support. We thank two anonymous reviewers and the editor, Dan Lunt, for careful comments which greatly improved the manuscript.
One contribution of 11 to a Discussion Meeting Issue ‘Warm climates of the past—a lesson for the future?’.
- © 2013 The Author(s) Published by the Royal Society. All rights reserved.