Direct linking of Greenland and Antarctic ice cores at the Toba eruption (74 kyr BP)

The Toba eruption that occurred some 74 kyr ago in Sumatra, Indonesia, is among the largest volcanic events on Earth over the last 2 million years. Tephra from this eruption has been spread over vast areas in Asia where it constitutes a major time marker close to the Marine Isotope Stage 4/5 boundary. As yet, no tephra associated with 5 Toba has been identiﬁed in Greenland or Antarctic ice cores. Based on new accurate dating of Toba tephra from Malaysia and on accurately dated European stalagmites the Toba event is known to occur between the onsets of Greenland Interstadials (GI) 19 and 20. Furthermore, the existing linking of Greenland and Antarctic ice cores by gas records and by the bipolar seesaw hypothesis suggests that the Antarctic counterpart 10 is situated between Antarctic Isotope Maxima (AIM) 19 and 20. In this work we suggest a direct synchronization of Greenland (NGRIP) and Antarctic (EDML) ice cores at the Toba eruption based on matching of a pattern of bipolar volcanic spikes. Annual layer counting between volcanic spikes in both cores allows for a unique match. We ﬁrst demonstrate this bipolar matching technique at the al- 15 ready synchronized Laschamp geomagnetic excursion (41 kyr BP) before we apply it to the suggested Toba interval. The Toba synchronization pattern covers some 2000 yr in GI-20 and AIM 19/20 and includes nine acidity peaks that are recognized in both ice cores. The suggested bipolar Toba synchronization has decadal precision. It thus allows 20 a determination of the exact phasing of inter-hemispheric climate in Our direct bipolar Toba synchronization gives very strong support for the bipolar seesaw hypothesis acting with no or little time lag in the MIS4/5 boundary region. At the EDML site in Dronning Maud Land the climatic pattern di ﬀ ers somewhat from that of the East Antarctic Plateau suggesting that local e ﬀ ects such as sea ice cover may play an important role in that region.

Toba has been identified in Greenland or Antarctic ice cores. Based on new accurate dating of Toba tephra from Malaysia and on accurately dated European stalagmites the Toba event is known to occur between the onsets of Greenland Interstadials (GI) 19 and 20. Furthermore, the existing linking of Greenland and Antarctic ice cores by gas records and by the bipolar seesaw hypothesis suggests that the Antarctic counterpart 10 is situated between Antarctic Isotope Maxima (AIM) 19 and 20.
In this work we suggest a direct synchronization of Greenland (NGRIP) and Antarctic (EDML) ice cores at the Toba eruption based on matching of a pattern of bipolar volcanic spikes. Annual layer counting between volcanic spikes in both cores allows for a unique match. We first demonstrate this bipolar matching technique at the al- 15 ready synchronized Laschamp geomagnetic excursion (41 kyr BP) before we apply it to the suggested Toba interval. The Toba synchronization pattern covers some 2000 yr in GI-20 and AIM 19/20 and includes nine acidity peaks that are recognized in both ice cores.
The suggested bipolar Toba synchronization has decadal precision. It thus allows 20 a determination of the exact phasing of inter-hemispheric climate in a time interval of poorly constrained ice core records, and it allows for a discussion of the climatic impact of the Toba eruption in a global perspective. Furthermore, our bipolar match provides a way to place paleo-environmental records other than ice cores into a precise climatic context.

Linking of Greenland and Antarctic ice core records
Synchronization of paleo-environmental records in ice cores from Greenland and Antarctica has long been a topic of great interest. In Greenland, glacial climate is associated with a number of abrupt climate shifts, Dansgaard-Oeschger (DO) events (Dansgaard et al., 1993), whereas Antarctica experienced a number of more gradual climate variations, Antarctic Isotope Maxima (AIM) (EPICA community members, 2006). A precise linking of ice cores from the two hemispheres is essential to determine the exact sequence of events and to reveal the dynamics related to these climate changes. Furthermore, because many paleoenvironmental archives can be linked to polar ice cores 10 (Blockley et al., 2012) a North-South synchronization of polar ice cores provides a way to place those archives into a global climatic context. A number of different approaches have been taken to synchronize ice cores from the two hemispheres. Bender et al. (1994) used the global signal of oxygen isotopes of air trapped in the Greenland GISP2 and the Antarctic Vostok ice cores to link the two 15 climatic records during the last glacial period. Blunier et al. (1998) and later Blunier and Brook (2001) applied the atmospheric methane concentration records extracted from the Greenland GISP2 and GRIP cores and the Antarctic Vostok and Byrd cores to construct a relative bipolar ice core chronology for the last 90 kyr. Methane records were also applied to link the Greenland NGRIP (North Greenland Ice Core Project members, Introduction impurity content of the ice (Goujon et al., 2003;Hörhold et al., 2012). Therefore, uncertainties of the order of 100-1000 yr remain in synchronizing ice-core climatic signals using gas-based matching (Blunier et al., 2007). A different approach of interhemispheric ice core linking was taken by Raisbeck et al. (2007) who applied the global signal of cosmogenic 10 Be in ice cores to match 5 the Greenland GRIP and the Antarctic EDC ice cores at the Laschamp geomagnetic excursion that occurred about 41 kyr ago. The 10 Be approach does not involve a delta-age as it directly links horizons in the ice phase of the cores in question. At the Laschamp excursion that is characterized by a characteristic double spike in 10 Be records, the Greenland and Antarctic ice cores could be linked with decadal preci- 10 sion The Laschamp excursion also provides a constraint for the deltaage estimation (Loulergue et al., 2007). Both gas and 10 Be linking was applied to tie the Antarctic EDML  and EDC (Parrenin et al., 2007) ice cores to the Greenland NGRIP ice core whereby the unified Greenland Ice Core Chronology 2005 (GICC05) (Rasmussen et al., 2006;An-15 dersen et al., 2006) could be transferred to Antarctica back to 41 kyr BP. This approach was recently refined and extended by Lemieux-Dudon et al. (2010) who applied an inverse model to combine the ice core time scales for NGRIP, EDC, EDML, and Vostok and the stratigraphic constraints among those cores to obtain a unified and consistent set of age scales for both ice and gas records. 20 The linking of Greenland and Antarctic ice cores paved the road for the bipolar seesaw hypothesis that suggests a mechanism to link the millennium scale climate variability of the last glacial period between the hemispheres (Broecker, 1998;Stocker and Johnsen, 2003) The bipolar seesaw hypothesis suggests that the Greenland DOevents and the Antarctic AIM-events are linked through the Atlantic Meridional Over- 25 turning Circulation (AMOC) implying that a warming in Greenland causes a cooling in Antarctica and vice versa. Blunier et al. (1998) and consecutive publications show that the picture is more complex: A rapid warming in Greenland ends a slow warming in Antarctica which conceptually can be explained by adding a heat buffer in the Southern Ocean (Stocker and Johnsen, 2003). The EDML ice core is thought to have a strong expression of the bipolar seesaw due to its location in the Atlantic sector of Antarctica. Within the last 50 kyr the Greenland-Antarctic ice-core linking is well established and the bipolar seesaw hypothesis appears to be confirmed by the relative timing of climate events (EPICA community members, 2006;Lemieux-Dudon et al., 2010;Stenni et al., 5 2010;Pedro et al., 2011). Beyond this time window, the relative timing of N-S records is somewhat uncertain, but the bipolar seesaw hypothesis appears to be active throughout the last glacial period (Blunier and Brook, 2001;Jouzel et al., 2007). During the earliest part of the last glacial the bipolar seesaw may have had somewhat different characteristics (Capron et al., 2010a), but in this period the ice core chronologies are 10 still being adjusted Vallelonga et al., 2012). Recently, the Greenland NGRIP and Antarctic EDML ice cores have been linked over the last 17 kyr through layer counting and matching of bipolar volcanic markers identified in both ice cores (Vinther et al., 2012): the two ice cores were dated independently by annual layer counting, NGRIP using the existing GICC05 chronology (Vinther 15 et al., 2006;Rasmussen et al., 2006) and EDML by application of a multi-parameter high resolution Continuous Flow analysis (CFA) dataset. The ice cores were then synchronized using major volcanic events that are unambiguously identified in both cores and thereby allowing for a highly precise transfer of the Greenland GICC05 timescale to EDML ice core (Vinther et al., 2012).

The Toba eruption occurring around 74 kyr BP
The Toba caldera is located close to the equator in Northern Sumatra, Indonesia. Over the last 1.3 million years the volcano has had several very large eruptions of which the most recent that occurred some 74 kyr ago is considered one of the largest volcanic eruptions of the Quaternary (Chesner, 2012). Tephra from this eruption known 25 as Younger Toba Tuff (YTT) is spread over a vast geographical region in Southeastern Asia on both sides of the equator and has been identified at many locations in the surrounding oceans and continents (e.g. Williams, 2012 (Westaway et al., 2011) and age estimates of the YTT are provided in several studies by application of various techniques (eg. Chesner et al., 1991). Most recently, well-characterized Toba tephra sampled in Malaysia has been precisely dated by Ar-Ar to 73.88 ± 0.32 kyr BP (1-sigma) by Storey et al. (2012). So far tephra originating from Toba has not been identified in Greenland (Abbott 5 et al., 2011) or in Antarctic ice cores. Already in 1996, however, Zielinski et al. (1996) suggested that the Toba eruption could be associated with a major sulfuric spike identified in the Greenland GISP2 ice core at the transition from Greenland  to Greenland  occurring close to the Marine Isotope Stage 4/5 (MIS 4/5) boundary. A few years later, Toba tephra was identified in marine sediment cores from the Arabian sea, which show a glacial climate variability comparable to that of the Greenland ice cores (Schulz et al., 1998). The position of the tephra in those and several other marine records seems to confirm the timing of the Toba eruption at the GI-20 to GS-20 transition (Schulz et al., 2002;Kudrass et al., 2001;Huang et al., 2001). 15 Being a much larger eruption than any historical eruption and probably being among the largest volcanic eruptions of the Quaternary the environmental and climatic effects of the 74 kyr Toba eruption are topics of great interest and debate. The suggested climatic impact of the Toba eruption ranges from very little impact (Haslam and Petraglia, 2010;Oppenheimer, 2002;Chesner and Luhr, 2010) to severe impact of "Volcanic win-20 ter and accelerated glaciation" (Rampino and Self, 1992). The effect of the eruption on regional vegetation, humans, and mammals is discussed in a number of papers (e.g. Petraglia et al., 2007;Louys, 2012) and the topic is still controversial (Williams, 2012). For archaeology, the Toba eruption is of particular interest because it may have occurred close to the time when Homo sapiens migrated out of Africa and into Eurasia 25  and it is speculated that the eruption may have caused a "human population bottleneck" (Ambrose, 1998

Outline of the present study
Inter-hemispheric volcanic linking of ice cores is quite different from volcanic ice core synchronization within Greenland or within Antarctica. In Greenland the volcanic record is dominated by northern hemispheric eruptions and major eruptions are generally present in all deep ice cores despite a large spatial variability of volcanic deposition 5 on the ice sheet (Clausen et al., 1997;Robock, 2000). Greenland ice cores have been synchronized in the last glacial period at annualto-decadal precision by major volcanic reference horizons . Due to its large geographical extent the spatial variability of volcanic deposition is even more pronounced in Antarctica, but still ice cores can be linked accurately by volcanic tie points Parrenin et al., 2012). For globally dispersed volcanic events the magnitude of Greenland and Antarctic acidic signatures will generally not be comparable as they depend strongly on several factors, such as latitude of the eruption, atmospheric circulation and the amount of sulfur injected into the stratosphere (Zielinski, 2000;Gao et al., 2008). Therefore, it is not straightforward to establish a Greenland-Antarctic volcanic ice-core synchroniza-15 tion from acidity spikes alone and in general additional evidence is needed in order to establish a bipolar volcanic link (Bay et al., 2006). In this study, we take a similar approach to that applied for the last 17 kyr by Vinther et al. (2012) of matching Greenland and Antarctic ice cores by annual layer counting and linking of bipolar volcanic reference horizons. The independent dating of Green-20 land and Antarctic ice cores by annual layer counting permits to match sequences of globally dispersed volcanic markers. To demonstrate the feasibility of this technique in the last glacial period we first establish a volcanic link of the Greenland NGRIP and Antarctic EDML ice cores at the Laschamp geomagnetic excursion (41 kyr BP) where the N-S linking is already well constrained. In this period we apply the existing 25 layer counted GICC05 chronology for NGRIP and count annual layers in EDML using a high resolution impurity dataset. We then extend this approach to a 2000 yr period at around 74 kyr BP where prominent acidity spikes that are most likely related to the Toba eruption can be identified in both NGRIP and EDML. Annual layer counting is performed independently in NGRIP and EDML and based on the relative chronologies an interhemispheric match is obtained by identifying a sequence of volcanic markers that are recognized in both cores. Based on published and new tie points we then expand the N-S Toba link to other major Greenland and Antarctic ice cores.

2 Datasets and methods
For annual layer counting in NGRIP we apply the high resolution CFA impurity dataset (Bigler, 2004) that provides continuous records of among others sulfate, sodium, calcium, electrolytic conductivity of melt water ("conductivity"), and the amount of insoluble dust particles ("dust") (Ruth et al., 2003) together with the visual stratigraphy (VS) grey scale profile (Svensson et al., 2005). Those records were previously applied to construct the glacial part of the GICC05 time scale (Rasmussen et al., 2006;Svensson et al., 2008). For the EDML layer counting we apply a CFA high resolution dataset that provides continuous records of sodium, ammonium, calcium, dust, and conductivity (Kaufmann et al., 2010) as well as the VS grey scale profile (Faria et al., 2010). This

15
CFA dataset was applied to establish the layer counted EDML time scale for last 17 kyr (Vinther et al., 2012). All of the records are available in 1 mm depth resolution, but due to varying degrees of profile smoothing in the firn and ice and during ice core analysis, the minimum resolvable features differ from record to record. The CFA records can typically resolve annual layers down to a few cm thickness whereas the VS grayscale 20 profiles can easily resolve annual layers of 1 cm thickness.
To identify volcanic markers in NGRIP we apply the CFA sulfate and conductivity records, the Electrical Conductivity Measurement (ECM) (Dahl-Jensen et al., 2002), and the Dielectric Profile (DEP) (Wilhelms et al., 1998). NGRIP ECM and DEP do not usually resolve annual layers in the glacial period, but they are good indicators of 25 volcanic events in particular during interstadials For EDML volcanic markers were identified in the CFA conductivity, in DEP and ECM, and in sulfate, all of which are robust Introduction indicators of volcanic events that have previously been applied to match Antarctic ice core records (Udisti et al., 2004;Parrenin et al., 2012;.

Bipolar synchronization at the Laschamp geomagnetic excursion
At the Laschamp geomagnetic excursion that occurred close to 41 kyr b2k (Singer et al., 2009), a bipolar synchronization is already established based on two 10 Be spikes 5 identified in both Greenland and Antarctic ice cores (Raisbeck et al., 2007;Yiou et al., 1997). The two 10 Be peaks essentially bracket the Greenland GI-10 event that is associated with the Antarctic AIM 9 and 10 events. In NGRIP the youngest of the 10 Be peaks occurs very close the onset of GS-10 and the oldest peak is located towards the end of GS-11 In NGRIP the two peaks are separated by some 17.5 m corresponding to 10 an interval of about 850 yr duration The precision of the 10 Be NGRIP-EDML synchronization is decadal to centennial (Raisbeck et al., 2007;Loulergue et al., 2007). The Laschamp section of the NGRIP ice core has been dated by annual layer counting for the GICC05 chronology which we adapt here (Andersen et al., 2006). We date the Laschamp section of the EDML ice core by layer counting in the CFA and VS datasets following the same principles as applied for NGRIP. We identify "certain" and "uncertain" annual layers that are counted as 1.0 ± 0.0 and 0.5 ± 0.5 yr, respectively. The accumulated uncertainty of the uncertain annual layers provides the maximum counting error estimate of the dating following the approach outlined in Andersen et al. (2006). The EDML mean annual layer thickness for the Laschamp section of 20 around two cm is well resolved in the VS record, fairly well resolved in dust, calcium, and ammonium but only partially resolved in conductivity ( Fig. 1) and not resolved in sodium (not shown). The resulting relative time scale is presented in Table 1. Figure 2 shows the 10 Be matched section of the two ice cores around the Laschamp event with indication of the two 10  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 1997). During interstadials volcanic spikes are usually quite distinct and easy to identify whereas in the coldest periods the volcanic spikes are often difficult to separate from the generally higher and noisier background level (Fig. 2). We note that because prominent sulfate peaks normally dominate the conductivity of the ice, the sulfate and the conductivity records are both robust indicators of major volcanic events. 5 Between the two Laschamp 10 Be peaks the NGRIP records show evidence of 3-5 larger and several smaller volcanic events whereas EDML has expression of 3-4 significant volcanic markers and some smaller events. Based on the 10 Be matching and on the annual layer counting of the two ice core sections we suggest that three of those events (referred to as L1-L3) are bipolar and can be applied to synchronize the ice 10 cores ( Fig. 2 and Table 1). Because we have no tephra evidence there is no way to prove that the suggested match points are truly global markers and potentially one or more of them are regional events that are just lining up by coincidence. Considering however the few volcanic events over the 850 yr period of interest and the tight correspondence in the timing of events a coincidental overlap of all of the events appears 15 unlikely.

Bipolar volcanic linking at the 7kyr Toba eruption
Because no Toba tephra has been identified in polar ice cores we do not have any proven knowledge of the location of the Toba volcanic signature neither in Greenland nor in Antarctic ice cores. The approach taken here to establish the bipolar Toba link 20 takes two steps: first, we provide an estimate for the position of the Toba eruption in the NGRIP and EDML ice cores based on recent independent dating of the event and on the bipolar seesaw hypothesis. Second, we perform layer counting in a 2000 yr long section of the two cores around the proposed Toba location and show that a bipolar volcanic synchronization can be established. Having done this, we still have not proven to 25 have identified Toba in the ice cores but we have increased the likelihood substantially and we have provided a bipolar volcanic link.  Zielinski et al. (1996) were the first to suggest the position of the Toba volcanic marker in Greenland at the GI-20 to GS-20 transition, and further evidence for this location was provided by the finding of the Toba tephra in the Arabian Sea (Schulz et al., 1998). A new well-constrained age for the Toba eruption of 73.88 ± 0.32 kyr BP is provided by Ar-Ar dating of tephra from Malaysia (Storey et al., 2012). Recently, 5 a record of several stalagmites from caves in the Alps (NALPS), showing a climatic pattern very similar to that of the Greenland ice cores, has been precisely dated by U-Th (Boch et al., 2011). Although the NALPS stalagmite record does not contain the onset of GS-20 it unambiguously brackets the Malaysian tephra age between the onsets of GI-19 and GI-2. The new tephra age and the stalagmites thus provide additional 10 and independent evidence for the location of the Toba event close to the onset of GS-20 in Greenland. The new tephra age also falls close to the Greenland GICC05modelext ice core time scale age of 74.1 kyr b2k for the onset of GS-20 (Wolff et al., 2010), but the ice core age is much less wellconstrained than that of the NALPS record due to the cumulative error of ice-core annual layer counting. 15 According to the linking of Greenland and Antarctic ice cores (Blunier and Brook, 2001;Jouzel et al., 2007;Capron et al., 2010b) and consistent with the bipolar seesaw hypothesis (Stocker and Johnsen, 2003), the onset of Greenland GS-20 is associated with the isotopic minimum between Antarctic AIM-19 and AIM-20 (Parrenin et al., 2012). We therefore, examine this region of the NGRIP and EDML ice cores for signif-20 icant volcanic markers (Fig. 3). In NGRIP the well-known pattern of elevated impurity content in the stadial periods is apparent. The major acidity spikes in the GI-20 and GS-20 intervals appear however within GI-20 close to the isotopic transition into GS-20 and the most significant spike in sulfate and conductivity appears at the depth of 2548.01 m right on the transition. In the interstadial acidity spikes are also visible in ECM and DEP, 25 whereas those records vanish in stadials where the ice is slightly alkaline (Wolff et al., 1997). In EDML, the most prominent conductivity, DEP, and ECM spikes in the AIM-19 and AIM-20 intervals appear around the isotopic minimum between the two and the most significant spike occurs at a depth of 1867.56 m (Fig. 3). For layer counting in the potential Toba region we apply the same NGRIP and EDML datasets that were utilized for the Laschamp matching. Compared to Laschamp, we are now some 500 m deeper in both cores and annual layers are correspondingly thinner. At NGRIP annual layers in GI-20 have a thickness of 1.0-1.5 cm that is well resolved in the VS record but only marginally resolved in the CFA data (Fig. 4). The approximately 5 2 kyr long section in GI-20 is dated in a similar manner as it was done at shallower depths for sections of GICC05 (Svensson et al., 2008). In the subsequent stadial GS-20, however, annual layer thicknesses are below 1 cm and this period cannot be reliably dated using existing datasets. The obtained NGRIP layer thickness profile is in very good agreement with that of the modeled "ss09sea" time scale (not shown). In EDML annual layers have comparable layer thicknesses of 1.0-1.5 cm which are again well resolved in VS and marginally resolved in CFA (Fig. 5). The EDML layer thickness profile obtained from layer counting is in very good agreement with the time scale of Lemieux-Dudon et al. (2010) and deviates somewhat from the EDML1 time scale of   (Fig. 6). 15 We now synchronize the two ice cores by identifying a sequential pattern of acidity tie points ( Fig. 7 and Table 1). The proposed match has nine match points (referred to as T1-T9) of which the major NGRIP and EDML acidity spike (T2) is one. 20 The 74 kyr Toba eruption is generally regarded as a single event (Westgate et al., 1998;Zielinski, 2000;Chesner, 2012). In the ice core records investigated in the present study there are, however, four bipolar acidity spikes (T1-T4) occurring within centuries of the GS-20 onset that are all potential Toba eruption candidates (Fig. 7). Based on the ice core data alone we have no means of determining how many of those events Introduction In the bipolar match presented here the T2 match point corresponds to the major acidity spikes in both the NGRIP and EDML ice cores (Fig. 7), but, interestingly, this is generally not the case for other ice cores (Fig. 8) In the GISP2 core the major sulfate spike identified by Zielinski et al. (1996) at 2591.1 m depth thus corresponds to the match point T1 of the present study (Table 1). Smaller spikes are also present in this 10 depth interval of the GISP2 sulfate record but the peak corresponding to T1 is definitely dominating. In high depth resolution the GISP2 sulfate concentrations at T1 are reaching almost 2000 ppb (Yang et al., 1996) in comparison to 1200 ppb for NGRIP at T2. In the nearby GRIP ice core the overall behavior of the ECM record is very similar to that of the GISP2 ECM record, but the T1 peak is strongly attenuated in GRIP as compared 15 to GISP2. In the Antarctic EDC ice core the major acidity spike in the Toba interval is associated with match point T3 of the present study, although the acidity spikes corresponding to T1 and T2 are also prominent. In the Dome Fuji (DF) ECM records the spike associated with T4 is the strongest T2 is significant and T3 is very weak. In the Vostok ice core ECM profile the T1 peak is the strongest followed by slightly weaker 20 T2 and T3 peaks (Parrenin et al., 2012). Based on ice core data there are thus several possible candidates for Toba eruption(s) and it cannot be excluded that several of them are related to Toba.

Age and timing of the Toba event
Assuming that the major Toba eruption can be associated with one or several of the 25 T1-T4 acidity spikes, the best Greenland ice core age estimate of the events are those of the GICC05modelext time scale (Wolff et al., 2010). Those ages fall in the interval of 74.1-74.5 kyr b2k (Table 1) source area" mode. As the d-excess mode jump is asynchronous with the Toba candidates we consider the timing of the NGRIP source area temperature change and the Toba events to be unrelated.

Implication for the bipolar seesaw hypothesis
After the bipolar 10 Be synchronization around GI-10 and the recent bipolar synchro- synchronization allows for a very precise comparison of interhemispheric climate coupling. A detailed comparison of the oxygen isotope curves demonstrates an extremely close coupling between Greenland (NGRIP) and Antarctic (DF and EDC) climates ( Fig. 9). At the GS-20 onset, Antarctic warming starts within a century of Greenland reaching the cold stadial and, likewise, at the onset of GI-19 and GI-20 DF and EDC 5 cooling starts immediately after Greenland has reached its isotopic maximum. Our results thus provide very strong support for an active bipolar seesaw close to the MIS4/5 boundary similar to what is observed in other parts of the last glacial (EPICA community members, 2006;Capron et al., 2010b;Pedro et al., 2011). The phasing between the Greenland NGRIP and the Antarctic DF and EDC records 10 reflects the classical imprint expected from an extended Bipolar Seesaw Mechanism as described by Stocker and Johnsen (2003) very well, where an Atlantic temperature seesaw is connected to a large Southern Ocean heat reservoir (warming in Antarctica when Greenland is in a cold state connected to a lowered Atlantic Meridional Overturning Circulation (AMOC) and cooling in Antarctica when Greenland is warm and 15 the AMOC is increased). The EDML ice core shows a somewhat different behavior. In the EDML record we find a much faster warming at the onset of GS21 and 20 which then tends to level out for the rest of these stadials. This pattern is also recognized for younger AIM during MIS3 (EPICA community members, 2006;Stenni et al., 2010). Taking the bipolar seesaw mechanism at face value this would imply that the ocean heat 20 pool relevant for the Atlantic sector of East Antarctica (EDML), which is warmed by the northward heat flux deficit across the Atlantic during the stadials, was much smaller than that relevant for the temperature response in the Indian Ocean sector (EDC). Since the Antarctic Circumpolar Current, however, is mixed rapidly within decades, this is highly unlikely and it appears difficult to maintain the required strong longitudinal 25 temperature gradient in the Southern Ocean between the Atlantic and Indian Ocean sector. Accordingly, the Southern Ocean heat pool should be the same for both cases.
In contrast the lower latitude South Atlantic should respond much faster to changes in the northward heat transport by the AMOC and should be essentially in antiphase CPD 8,2012

Linking of Greenland and Antarctic ice cores
A. Svensson et al.  (Stenni et al., 2010), shows essen-5 tially the same temporal behavior during AIM as without correction, i.e. a relatively fast temperature increase at the onset of the AIM. Accordingly, we need some other factor that can amplify the atmospheric warming signal in the δ 18 O record at the beginning of the AIM at EDML. This could potentially come about by a threshold in sea ice coverage and here especially in the extended 10 summer sea ice coverage in the Atlantic sector during cold conditions. We speculate that a warming induced by the bipolar seesaw mechanism in the Atlantic sector of the Southern Ocean crosses such a threshold early during the AIM. A significantly reduced sea ice coverage then leads to a stronger atmospheric warming signal in this region. In fact the EDML sodium record, which has been used as sea ice indicator, shows minima 15 during the early stage of the AIM in MIS3  pointing to quickly declining sea ice coverage at that time. In contrast no such minima are observed in the EDC record . In this region of the Southern Ocean summer sea ice was not significantly expanded even for peak glacial conditions (Gersonde et al., 2005) and thus no such threshold for the existence of summer sea ice exists. Accordingly, 20 a somewhat different temporal evolution in the atmospheric temperature record of the EDML and EDC ice cores can be accommodated in the Bipolar Seesaw concept, if we take a sea ice amplification effect in the Atlantic sector into account.

On the climatic impact of the Toba eruption
As pointed out in several studies (e.g. Zielinski, 2000), it is intriguing that the Toba Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | event in this part of the last glacial period (Wang et al., 2008). It seems highly unlikely that a volcanic eruption should have initiated GS-20 when the initiation of all of the other DO-events was independent of major volcanic events. Still, there is the possibility that the eruption somehow contributed to make the GS-20 event a more extreme stadial period in the Northern Hemisphere, although it is difficult to imagine a mechanism that 5 can sustain the extremely cold temperatures over more than a millennium. In a global perspective, however, the Toba eruption did not initiate a long term cold period (Fig. 9). In contrast, except for a dip in the isotopes right after the T2 peak, Antarctica experiences a major warming shortly after the Toba event(s). Thus our results do not support the idea of a major global climatic cooling impact of the Toba event. 10 The isotopic dip occurring right after T2 in EDML and DF lasts for about a century and is possibly a cooling event related to Toba that has no significant influence on the long term climate. The rather short duration of this cold spell appears in good agreement with recent model simulations of the climatic impact of the Toba event (Robock et al., 2009;Jones et al., 2005;Timmreck et al., 2010).

Conclusions
By application of high resolution impurity records from the Greenland NGRIP and the Antarctic EDML ice cores and by applying existing bipolar ice core synchronizations, it has been possible to obtain bipolar volcanic matches at the Laschamp geomagnetic excursion (ca. 41 kyr b2k) and at the Toba mega-eruption (ca. 74 kyr b2k). 20 Using constraints from precise Ar-Ar dating of Malaysian Toba ash and from precise U-Th dating of European stalagmites we are certain that the Toba event occurs close to the onset of GS-20 in Greenland and between AIM-19 and AIM-20 in Antarctica. The ice core records suggest that the Toba event may consist of up to four individual eruptions occurring within four centuries, but we have no means to decide how many Our direct bipolar Toba synchronization gives very strong support for the bipolar seesaw hypothesis acting with no or little time lag in the MIS4/5 boundary region. At the EDML site in Dronning Maud Land the climatic pattern differs somewhat from that of the East Antarctic Plateau suggesting that local effects such as sea ice cover may play an important role in that region.
Greenland and Antarctic temperature proxies suggest that there may be a century long cooling episode associated with the Toba eruption(s) but they do exclude the possibility of a longer term global cooling impact of the eruption(s). In contrast, Antarctica warms up rapidly after the eruption(s).
The approach taken here to synchronize Greenland and Antarctic ice cores by vol-  Dahl-Jensen, D., Gundestrup, N., Miller, H., Watanabe, O., Johnsen, S. J., et al.: The North-GRIP deep drilling program, Ann. Glaciol., 35, 1-4, 2002. Dansgaard, W., Johnsen, S. J., Clausen, H. B., Dahl-Jensen, D., Gundestrup, N. S., et al.: Evidence for general instability of past climate froma 250-kyr ice-core record, Nature, 364, 218-220, 1993. 5 EPICA community members: Eight glacial cycles from an Antarctic ice core, Nature, 429, 623-628, 2004. EPICA community members: One-to-one coupling of glacial climate variability in Greeland and Antartctica, Nature, 444, 195-198, 2006. Faria, S. H., Freitag, J., and Kipfstuhl, S.: Polar ice structure and the integrity of ice-core pale-10 oclimate records, Quaternary Sci. Rev., 29, 338-351, doi:10.1016/j.quascirev.2009.10.016, 2010 Reconstruction of millennial changes in dust emission, transport and regional sea ice coverage using the deep EPICA ice cores from the Atlantic and Indian Ocean sector of Antarctica, Earth Planet. Sci. Lett.,      (Watanabe et al., 2003;Kawamura et al., 2007) isotopic profiles around GI-20 based on the Toba volcanic match. All records show δ 18 O except EDC that is δD. The EDML and EDC age models are modified from the Lemieux-Dudon et al. (2010) time scales that are shifted in time and linearly stretched by a few percent to fit the GICC05modelext ages of match points T2 and T9. The Dome Fuji age scale is treated in a similar way but based on the time scale of Kawamura et al. (2007) For EDML and Dome Fuji the thin step curve shows measured high resolution isotopes and the thick curves are running averages.