(Supporting Information) 14Cage plateau tuning of local Atlantic sediment cores

Modeling studies predict that changes in radiocarbon (14C) reservoir ages of surface waters during the last deglacial episode will reflect changes in both atmospheric 14C concentration and ocean circulation including the Atlantic Meridional Overturning Circulation. Tests of these models require the availability of accurate 14C reservoir ages in well-dated late Quaternary time series. We here test two models using plateau-tuned 14C time series in multiple well-placed sediment core age-depth sequences throughout the lower latitudes of the Atlantic Ocean. 14C age plateau tuning in glacial and deglacial sequences provides accurate calendar year ages that differ by as much as 500-2500 years from those based on assumed global reservoir ages around 400 years. This study demonstrates increases in local Atlantic surface reservoir ages of up to 1000 years during the Last Glacial Maximum, ages that reflect stronger trades off Benguela and summer winds off southern Brazil. By contrast, surface water reservoir ages remained close to zero in the Cariaco Basin in the southern Caribbean due to lagoon-style isolation and persistently strong atmospheric CO2 exchange. Later, during the early deglacial (16 ka) reservoir ages decreased to a minimum of 170-420 14C years throughout the South Atlantic, likely in response to the rapid rise in atmospheric pCO2 and Antarctic temperatures occurring then. Changes in magnitude and geographic distribution of 14C reservoir ages of peak glacial and deglacial surface waters deviate from the results of Franke et al. (2008) but are generally consistent with those of the more advanced ocean circulation model of Butzin et al. (2012).

Revised depth scale, splice tie points and chron boundaries of ODP Hole 151-907A and Site 162-907

The magnetic polarity stratigraphy at Site 907 obtained from the shipboard pass-through magnetometer and from discrete samples is readily interpretable back to the onset of the Gilbert Chron (5.89 Ma). From this level to the base of the section at ~14 Ma, the interpretation is corroborated by silicoflagellate datums with predictable correlation to polarity chrons. The resulting magnetostratigraphic interpretation differs from those proposed in the Leg 151 (Hole 907A) and 162 (Holes 907B and 907C) Initial Reports volumes. An important hiatus in the 7-10 Ma interval at Site 907 caused sedimentation to slow or cease for ~2.7 m.y. We have revised the shipboard correlation among the three holes at Site 907, resulting in a new composite section splice and recalculation of composite depths. For Site 985, magnetostratigraphic interpretation is possible down to ~150 meters below seafloor (mbsf) (C3An/C3Ar) at ~6 Ma. There are no useful biostratigraphic datums from Site 985 to support this interpretation; however, the interpretation is supported by the correlation of Sites 985 and 907 using natural gamma data from the shipboard multisensor track. Below ~150 mbsf at Site 985, drilling-related deformation at the onset of extended core barrel drilling precluded magnetostratigraphic interpretation.

(Table 1) Gas tie points used to correlate NorthGRIP and Vostol records

North Atlantic sediment records (MD95-2042), Greenland (Greenland Ice Core Project (GRIP)) and Antarctica (Byrd and Vostok) ice core climate records have been synchronized over marine isotopic stage 3 (MIS 3) (64 to 24 kyr B.P.) (Shackleton et al., 2000). The resulting common timescale suggested that MD95-2042 d18Obenthic fluctuations were synchronous with temperature changes in Antarctica (dDice or d18Oice records). In order to assess the persistency of this result we have used here the recent Greenland NorthGRIP ice core covering the last glacial inception. We transfer the Antarctic Vostok GT4 timescale to NorthGRIP d18Oice and MD95-2042 d18Oplanktonic records and precisely quantify all the relative timing uncertainties. During the rapid warming of Dansgaard-Oeschger 24, MD95-2042 d18Obenthic decrease is in phase with d18Oplanktonic decrease and therefore with NorthGRIP temperature increase, but it takes place 1700 ± 1100 years after the Antarctic warming. Thus the present study reveals that the results obtained previously for MIS 3 cannot be generalized and demonstrates the need to improve common chronologies for marine and polar archives.

(Table 1) Gas tie points used to correlate the ice cores NGRIP and EDML

Water isotope records from the EPICA Dronning Maud Land (EDML) and the NorthGRIP ice cores have revealed a one to one coupling between Antarctic Isotope Maxima (AIM) and Greenland Dansgaard-Oeschger (DO) events back to 50 kyr. In order to explore if this north-south coupling is persistent over Marine Isotopic Stage 5 (MIS 5), a common timescale must first be constructed. Here, we present new records of d18O of O2 (d18Oatm) and methane (CH4) measured in the air trapped in ice from the EDML (68-147 kyr) and NorthGRIP (70-123 kyr) ice cores. We demonstrate that, through the period of interest, CH4 records alone are not sufficient to construct a common gas timescale between the two cores. Millennial-scale variations of d18Oatm are evidenced over MIS 5 both on the Antarctic and Greenland ice cores and are coupled to CH4 profiles to synchronise the NorthGRIP and EDML records. They are shown to be a precious tool for ice core synchronisation. With this new dating strategy, we produce the first continuous and accurate sequence of the north-south climatic dynamics on a common ice timescale for the last glacial inception and the first DO events of MIS 5, reducing relative dating uncertainties to an accuracy of a few centuries at the onset of DO events 24 to 20. This EDML-NorthGRIP synchronisation provides new firm evidence that the bipolar seesaw is a pervasive pattern from the beginning of the glacial period. The relationship between Antarctic warming amplitudes and their concurrent Greenland stadial duration highlights the particularity of DO event 21 and its Antarctic counterpart. Our results suggest a smaller Southern Ocean warming rate for this long DO event compared to DO events of MIS 3.