Atmospheric CO2 records for the centennial scale cooling event 8200 years ago (8.2 ka event) may help us understand climate-carbon cycle feedbacks under interglacial conditions, which are important for understanding future climate, but existing records do not provide enough detail. Here we present a new CO2 record from the Siple Dome ice core, Antarctica, that covers 7.4–9.0 ka with 8 to 16 year resolution. We observe a small, about 1–2 ppm, increase of atmospheric CO2 during the 8.2 ka event. The increase is not significant when compared to other centennial variations in the Holocene that are not linked to large temperature changes. Our results do not agree with leaf stomata records that suggest a CO2 decrease of up to ~25 ppm and imply that the sensitivity of atmospheric CO2 to the primarily Northern Hemisphere cooling of the 8.2 ka event was limited.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Accurate understanding of climate-carbon cycle feedbacks is important to better constrain the future climate under continued anthropogenic CO2 forcing. Model-based estimates for climate sensitivity of the global carbon cycle suggest that climate change will be amplified by carbon cycle feedback, although the magnitude of these feedbacks remains poorly constrained [Friedlingstein et al., 2006]. Although climate models produce quantitative results with plausible mechanisms, tests of the models are limited due to the brevity of the instrumental record. Paleoclimate records, on the other hand, provide long records of natural experiments on variety of time scales and under a variety of climate boundary conditions. In particular, the relationship between abrupt climate change and atmospheric CO2 is of great interest. This topic has been relatively well studied on multimillennial time scales for the last glacial period and deglacial conditions [Monnin et al., 2001; Ahn et al., 2004; Ahn and Brook, 2007; Ahn and Brook, 2008; Ahn et al., 2012a; Bereiter et al., 2012; Schmitt et al., 2012]. However, the relationship between CO2 and abrupt climate change on submillennial time scales or under interglacial climate boundary conditions remains poorly investigated.
The 8.2 ka event is a centennial cooling and drying event, which is the most abrupt climate feature of the Holocene [Alley et al., 1997; Alley and Agustdottir, 2005; Rohling and Paelike, 2005]. The event is pronounced around the circum-North Atlantic region. Well-dated Greenland ice cores show that the cooling event lasted for 160 years [Thomas et al., 2007] and was accompanied by a sharp atmospheric CH4 decrease, indicating that the event was probably at least hemispheric in its extent [Kobashi et al., 2007]. It is likely that the 8.2 ka event was related to a reduction of the Atlantic meridional overturning circulation (AMOC) by freshwater discharge from proglacial lakes Agassiz and Ojibway into the Labrador Sea [von Grafenstein et al., 1998; Barber et al., 1999; Teller et al., 2002; Ellison et al., 2006; Kleiven et al., 2008; Pratoreous et al., 2008; Carlson et al., 2009; Hoffman et al., 2012] and was associated with a total eustatic sea level rise of 0.8–2.2 m [Li et al., 2012; Törnqvist and Hijma, 2012]. The freshwater discharge may have reduced northward heat transport by the AMOC, causing a cooling in the circum-North Atlantic [Alley et al., 1997; Barber et al., 1999]. Model studies indicate that freshwater discharge may have caused the abrupt cooling event [LeGrande et al., 2006; Wiersma and Renssen, 2006]. Compared to glacial conditions, climate boundary conditions of the 8.2 ka event (e.g., sea level and ice sheet extent) were not very different from modern ones, although hemispheric seasonality of orbital forcing was different and there still was a small ice sheet in North America [Carlson et al., 2008; Renssen et al., 2009]. Thus, the 8.2 ka event provides an important target for testing climate-carbon cycle models for AMOC disturbance and associated climate change under interglacial conditions.
Ice cores permit us to directly analyze CO2 mixing ratios in ancient air with an analytical precision of ~1 ppm [Ahn et al., 2009; Bereiter et al., 2012]. Previous ice core studies did not directly address the CO2 change at the 8.2 ka event, mostly because of the low sampling resolution of 100–200 years [Indermühle et al., 1999; Flückiger et al., 2002] and smoothing effects in the firn layer by diffusion and gradual bubble close-off [Spahni et al., 2003]. In addition, insufficient chronological control has hampered exact comparison of atmospheric CO2 change, which is normally measured in Antarctic ice cores, with the timing of the 8.2 ka event, which is expressed in Greenland ice cores.
As an alternative to ice core records, past CO2 levels can be reconstructed from the density of stomata in fossil leaves [e.g., Beerling et al., 1995; Wagner et al., 2004]. A high-resolution stomata-based CO2 record for the 8.2 ka event obtained from fossil leaves of European tree birches indicates a CO2 decrease of up to ~25 ppm during 8.4–8.1 ka that was associated with the 8.2 ka event [Wagner et al., 2002]. However, the mean sampling resolution of ~110 years and the mean uncertainty of CO2 estimates of ±10 ppm suggest that this record should be cautiously interpreted, particularly given the relatively short span of the 8.2 ka event [van Hoof et al., 2005]. The fact that the estimated CO2 concentration from the stomata record is higher than that from ice cores by 35–40 ppm is also a concern, given that such large interhemispheric gradients in the background preindustrial atmosphere are very unlikely [IPCC, 2007].
To resolve centennial atmospheric CO2 variability during the 8.2 ka event, we analyzed the Siple Dome ice core at a very high mean sampling resolution of 8–16 years. The modern Siple Dome site has a high snow accumulation rate of 12.4 cm we/yr (water equivalent/year) [Brook et al., 2005], higher than those of the Taylor Dome (6.4 cm we/yr) and Dome C (2.5 cm we/yr) cores, which were previously used for Holocene CO2 reconstructions [Indermühle et al., 1999; Flückiger et al., 2002]. The high snow accumulation rate results in a relatively small gas age distribution width [e.g., Schwander et al., 1997; Buizert et al., 2012] (Table S1 of the supporting information). For these reasons, Siple Dome gas records have experienced only minimal smoothing by diffusion and gradual bubble close-off; the width of the age distribution is 42 years (full width at half maximum), compared to ~137 years at Dome C (Table S1 of the supporting information).
Details of the methods for CO2 analysis are given by Ahn et al.  and in the supporting information. We analyzed Siple Dome ice samples from 107 depth intervals that correspond to ages including the 8.2 ka event. For most depth intervals, we analyzed two samples from a mean depth range of 7 cm (corresponding to 2.6 years), and we report the mean depths of the replicates (Figure 1). The CO2 mixing ratios follow the NOAA WMOX2007 CO2 mole fraction scale.
The timing of the 8.2 ka event in our Siple Dome gas records is well constrained by comparison of our new Siple Dome CH4 record with the NGRIP (North Greenland Ice Core Project) ice core δ18Oice record, which is a proxy for Greenlandic climate [North Greenland Ice Core Project members, 2004] (Figure 1) because the abrupt cooling event in Greenland is essentially synchronous with a sharp CH4 decrease within ±4 years [Kobashi et al., 2007]. The synchronized ages are on the GICC05 (Greenland Ice Core Chronology 2005) time scale, which is constructed by annual layer counting and has been widely used for paleoproxy data [Rasmussen et al., 2006; see also supporting information] (Figure 1). We also utilized a high-resolution CH4 record from the GISP2 (Greenland Ice Sheet Project 2) ice core and correlated it with the Siple Dome record (Figure 1). We find that the CH4 change in the Siple Dome ice is similar to that of the GISP2 ice with an almost constant offset, supporting the contention that Siple Dome gas records are well preserved and have experienced only minimal firn smoothing. The offset in CH4 between the GISP2 and Siple Dome records is attributed to the interhemispheric CH4 gradient caused by the predominance of Northern Hemisphere CH4 sources [e.g., Etheridge et al., 1998; Dlugokencky et al., 2005].
Our new CO2 record from the Siple Dome core covers 9.0–7.3 ka with sampling resolutions of 8 and 16 years for the 8.2 ka event and the other age intervals, respectively (Figure 1). Our new CH4 records from the Siple Dome and GISP2 ice cores clearly define the timing of the 8.2 ka event, and we observe a slight increase of CO2 of about 1–2 ppm (Gaussian filter, 1σ = 10–40 years) during that event compared to the preceding century. We also observe atmospheric CO2 background variability of 2–4 ppm on multidecadal to centennial time scales during 9.0–7.3 ka. Therefore, the CO2 increase during the 8.2 ka event is similar to multidecadal to centennial CO2 variability observed in other parts of the record (Figure 1), which are not associated with large cooling. Our results are not consistent with the ~25 ppm decrease of CO2 constructed from leaf stomata [Wagner et al., 2002].
Smoothing due to diffusion and gradual bubble close-off in the firn column is an important consideration when interpreting ice core gas records. At some sites, this process might be expected to remove an abrupt response at the 8.2 ka event. For example, the CH4 drop at the 8.2 ka-cooling event is estimated to be attenuated by 34~59% in the European Project for Ice Coring in Antarctica Dome C (EDC) core [Spahni et al., 2003], a low-accumulation site (Table S1 of the supporting information). At sites like Siple Dome (Table S1 of the supporting information), the influence of smoothing should be significantly less. In order to quantitatively estimate the smoothing effect in the Siple Dome CO2 record, we modeled gas diffusion and bubble close-off in the firn layer (Figure S3 of the supporting information) to obtain gas age distributions (Figure S4 of the supporting information). We used a 1-D firn air transport model [Buizert et al., 2012], and details of the modeling are given in the supporting information. We applied the Siple Dome gas age distributions to imaginary square pulses of atmospheric CO2 with various durations of 50, 100, and 200 years, which are comparable to the ~160 year duration of the 8.2 ka event (Figure 2). We find that the magnitude of the pulse with duration of 50 years is decreased by ~25%, and pulses with durations of 100 and 200 years are recorded in the ice with their full magnitude (orange curves). In good agreement with earlier studies [Spahni et al., 2003], we find that at EDC a 100 year event would be recorded in the ice with only half the true atmospheric magnitude (blue curves). This analysis confirms that, given our sampling resolution, any centennial scale CO2 response to the 8.2 ka cooling event would be visible in our records with essentially unattenuated amplitude. The sharp decrease of the Siple Dome CH4 (Figure 1) confirms that the smoothing of gas records in the Siple Dome ice is minimal on centennial time scales and supports our estimation of gas smoothing as discussed above. We also convolved the stomatal records from the work of Wagner et al.  with the Siple Dome gas age distribution (Figure 3). If the stomatal CO2 reconstruction represented the true atmospheric CO2 variations, then the convolved record shows how it would be recorded in the ice core. We observe that the smoothing would not significantly change the magnitude of the putative CO2 decrease at the 8.2 ka event derived from the stomatal data (Figure 3). The Antarctic ice cores and the stomatal records represent opposite hemispheres, yet the offset is too large to be attributed to the interhemispheric gradient, which is currently ~2.5 ppm and would have been smaller during the preanthropogenic period [IPCC, 2007, p. 517].
Given the small difference among the three different ice core CO2 records and our assessment of firn smoothing, we conclude that the elevated CO2 levels and high amplitude variability in the stomatal CO2 reconstruction [Wagner et al., 2002] do not represent variations in the background atmosphere and are more likely representative of other factors that impact stomatal density and/or very local variations in CO2.
4 Implications for Climate-Carbon Cycle Links
The atmospheric CO2 concentration changes through carbon exchanges with ocean and land. Many of the processes that control the carbon fluxes are sensitive to temperature. Ice core records for the late Holocene show positive correlations between CO2 and climate proxies from the Northern Hemisphere on centennial time scales [MacFarling Meure et al., 2006; Ahn et al., 2012b] probably because carbon emissions from land [MacFarling Meure et al., 2006; Gerber et al., 2003] or from both land and oceans [Ahn et al., 2012b] are positively linked with surface temperatures. An ensemble reconstruction by Frank et al.  finds a positive global carbon cycle sensitivity of 7.7 ppm (with a range of 1.7–21.4) CO2 per °C for the late Holocene.
Although the average global temperature change at the 8.2 ka event is not well known, paleoclimatic data and model results indicate that the magnitude of cooling in the Northern Hemisphere is greater than that of warming in the Southern Hemisphere [Wiersma and Renssen, 2006], and therefore, we might expect a small decrease in CO2 associated with the event due to surface temperature forcing by reduced carbon emissions from land and increased dissolution of atmospheric CO2 in ocean as we observe for the late Holocene [MacFarling Meure et al., 2006; Ahn et al., 2012b; Gerber et al., 2003].
Previous studies for the last glacial and deglacial periods indicate that long Greenlandic stadials (cold periods) and associated major Antarctic warmings accompanied reductions in AMOC and CO2 increases of up to 20~40 ppm on multimillennial time scales [Monnin et al., 2001; Ahn and Brook, 2008; Ahn et al., 2012a; Bereiter et al., 2012; Indermühle et al., 2000; Marchitto et al., 2007]. These CO2 increases are often attributed to upwelling of CO2-rich deep water in the Southern Ocean [Marchitto et al., 2007; Anderson et al., 2009]. The small increase of 1–2 ppm during the 8.2 ka event could perhaps be attributed to a weaker version of mechanism operating during the deglaciation and glacial period, where CO2 rises during stadial periods. However, direct comparison is difficult due to different durations of the cold events (millennial versus centennial) and different boundary conditions (glacial versus interglacial), and as we point out above, the 2 ppm fluctuation does not appear to be anomalous with respect to times without large cooling.
Multiple model simulations for land and oceanic carbon during a reduced AMOC under late Holocene climate boundary conditions show that the terrestrial CO2 source overcompensates the oceanic sink, resulting in increased CO2 [Obata, 2007; Menviel et al., 2008; Bozbiyik et al., 2011]. However, another simulation indicates that terrestrial carbon processes could act as a sink [Köhler et al., 2005]. We also note that the change in atmospheric CO2 in the models depends on the shape (e.g., continual versus abrupt change) and amplitude of freshwater discharge. The magnitude and timing of the freshwater discharge flux associated with the 8.2 ka event is not well constrained [Törnqvist and Hijma, 2012; Carlson and Clark, 2012]. However, estimates based on sea level rise at Louisiana [Li et al., 2012] and the Netherlands [Hijma and Cohen, 2010] suggest that the magnitude was 0.09–0.25 sverdrup (Sv) (1 Sv = 106 m3 s−1) for 100 years and 0.06–0.15 Sv for 300 years, respectively [Törnqvist and Hijma, 2012; Carlson and Clark, 2012], comparable to the flux ranges in the modeling of Köhler et al.  and Bozbiyik et al.  that did not impact atmospheric CO2 for 100–200 years. Thus, the weak and/or gradual freshwater discharge might have not been sufficient to change atmospheric CO2 during the 8.2 ka event.
Understanding the response of the global carbon cycle to climate change is important for predictions of future warming. Our results provide a target for climate-carbon cycle models with a new important constraint on Holocene climate-carbon cycle feedbacks during abrupt climate change associated with ocean circulation changes as might be expected in the future associated with a weakened AMOC [IPCC, 2007].
We thank Michael Kalk, James Lee, and Logan Mitchell for analytical assistance and the staff of the National Ice Core Lab for ice sampling and curation. We also thank Andreas Schmittner for helpful discussions. Financial support was provided by National Science Foundation grant OPP 0944764-ANT to E.B. and the NOAA Climate and Global Change Fellowship Program to C.B., administered by the University Corporation for Atmospheric Research. This work was also supported by Basic Science Research Program through the National Research Foundation of Korea, funded by the Ministry of Education, Science and Technology (2011-0025242), and Korea Meteorological Administration Research and Development Program under grant CATER 2012-7030 to J.A.
The Editor thanks Andrei Demekhov and an anonymous reviewer for their assistance in evaluating this paper.