Dynamics of ~100-kyr glacial cycles during the early Miocene

Here, we present high-resolution stable isotope records from ODP Site 1264 in the 
South-Eastern Atlantic Ocean, which resolve the latest Oligocene to early Miocene 
(23.7–18.9 Ma) climate changes. Using an inverse modelling technique, we decom- 
posed the oxygen isotope record into temperature and ice volume and found that the 
Antarctic ice sheet expanded during distinct episodes (e.g., Mi zones) of low short-term 
(∼100-kyr) eccentricity forcing, which occur two to four long-term (400-kyr) eccentricity 
cycles apart. We argue that a non-linear mechanism, such as the merging of (several) 
large East Antarctic ice sheets, caused the build-up of a larger ice sheet. During the 
termination phases of these larger ice sheets, on the contrary, we find a more linear response 
of ice-sheet variability to orbital forcing and climate became highly sensitive to 
the ∼100-kyr eccentricity cycle. At the Oligocene-Miocene transition the model output 
indicates a decrease in Northern Hemisphere temperatures such that a small ice cap 
could develop on Greenland. This supports the hypothesis of a threshold response for 
the development of Northern Hemisphere land ice to decreasing pCO2.


Introduction
Earth's climate has gradually cooled during the past 50 million years in conjunction with declining atmospheric pCO 2 conditions (Zachos et al., 2008). Following the cooling and rapid expansion of Antarctic continental ice-sheets in the earliest Oligocene, deep-sea oxygen isotope (δ 18  (Oi-1, Oi-2, Mi-1-Mi-6) were described (Miller et al., 1991), followed by Mi-1a, Mi-1b, Mi-7, Mi-1aa (Wright and Miller, 1992), Oi-2b.1, Mi-1.1 (Billups et al., 2002) and another, yet unnamed, zone (Paul et al., 2000) for the latest Oligocene and early Miocene. It has long been suspected that the large-scale changes in Antarctic ice volume are coupled to long-term eccentricity (2.0-2.6 myr) and obliquity (∼1.2 myr) modulations 5 of the Earth's orbit and axial tilt (Miller et al., 1991;Beaufort, 1994;Lourens and Hilgen, 1997). This theory could only recently be investigated with the generation of high-resolution (≤10 kyr) oxygen isotope records (Zachos et al., 2001b;Pälike et al., 2006a, b;Billups et al., 2002). Here we will assess the long-term orbital pacing theory of the late Oligocene to early Miocene (∼19-24 Ma) time inter-10 val by presenting a new high-resolution (<3 kyr) and continuous stable isotope record based on the benthic foraminiferal species Cibicidoides mundulus from Ocean Drilling Program (ODP) Site 1264 situated at a water depth of 2505 m on the northern flank of the Walvis Ridge (29 • S) in the Southern Atlantic Ocean (Zachos et al., 2004). We will compare our isotope results with those of ODP Site 926 Hole B (3 • N) at 3598 m water 15 depth and ODP Site 929 Hole A (6 • N) at 4358 m water depth, both from Ceara Rise in the Equatorial Western Atlantic (Zachos et al., 1997(Zachos et al., , 2001bPaul et al., 2000;Pälike et al., 2006a;Shackleton et al., 2000), and the composite record of ODP Site 1090, based on Holes D and E, at 3699 m water depth from the Agulhas Ridge (43 • S) in the Atlantic section of the Southern Ocean (Billups et al., 2002(Billups et al., , 2004 (Zachos et al., 2004). Six sites were drilled along a depth-transect of which two sites, Site 1264 (2505 m) and 1265 (3083 m), are used in 5 this study. Both sites are -and were throughout the entire Neogene -situated above the level of the present day lysocline and CCD (Zachos et al., 2004). The extended late Oligocene and early Miocene section of Site 1264 consists of foraminifer bearing nannofossil ooze (Zachos et al., 2004). Site 1264 is uniquely situated (Fig. 1) to record major changes in regional and/or global ocean carbon chemistry, ocean circulation and 10 intermediate bottom water chemistry and circulation (Zachos et al., 2004). Site 1265 was applied to provide the magnetic inclination record that Site 1264 lacks.

Age model
Because Site 1264 lacks a good magnetostratigraphy, we transferred the magnetostratigraphic data from the nearby ODP Site 1265 by pattern matching the magnetic 15 susceptibility (MS) and colour reflectance (CR, 600/450 nm) records ( Fig. 2

Methodology and stable isotope chemistry
Samples of approximately 10 g of sediment were taken every 2-2.5 cm from the latest Oligocene and early Miocene part of the Site 1264. The samples were freeze dried, washed and sieved to obtain the larger than 37, 65 and 125 µm fractions for 5 foraminiferal analysis. Mostly single specimen samples of the benthic foraminifer species Cibicidoides mundulus were analysed. On every sample stable oxygen and carbon isotope ratios (δ 18 O and δ 13 C, respectively) were measured and the δ 18 O values were corrected for disequilibrium with seawater by adding 0.64‰ (Shackleton, 1974;Zachos et al., 2001a). Approximately 80% of the samples were measured at 10 the Faculty of Geosciences of Utrecht University (UU) where foraminiferal tests were dissolved in a Finnigan MAT Kiel III automated preparation system. Isotopic ratios of purified CO 2 gas were then measured on-line with a Finnigan MAT 253 mass spectrometer and compared to an internal gas standard. The remaining part was measured at the Department of Geological Sciences of the University of Florida (UF) on two inter-15 calibrated devices. Of the samples with plentiful specimens, foraminiferal calcite was reacted using a common acid bath of orthophosphoric acid at 90 • C using a Micromass Isocarb preparation system. Isotope ratios of purified CO 2 gas were measured online using a Micromass Prism mass spectrometer. Of the samples with few Cibicidoides mundulus specimens, foraminiferal tests were dissolved using a Finnigan MAT 20 Kiel III automated preparation system coupled to a Finnigan MAT 252 mass spectrometer to measure the isotopic ratios of purified CO 2 gas. The standard NBS-19 and the in-house (at UU) standard "Naxos" were used to calibrate to Vienna Pee Dee Belemnite (VPDB). Reproducibility (same sample on the same device) is 0.19‰ for δ 18 O and 0.13‰ for δ 13 C (Supplement Fig. 1). Between universities an unexplained offset 25 of 0.30‰ in δ 18 O is found between analyses of foraminifera from the same samples (Supplement Fig. 1 correction has been applied for this offset because a lower resolution (step size 100kyr) record, spanning the interval of this study, measured entirely at UF shows no offset (B. D. A. Naafs, personal communication, 2010). Outliers were defined by an upper and lower boundary of 2 standard deviations (of the entire time series) added or subtracted from a 13-point moving average. Because the stable isotope analysis is 5 paired, outliers defined in δ 13 C or in δ 18 O were removed from both records (Fig. 4, Supplement Fig. 2). Where possible, outliers were re-measured. After outlier-removal the stable isotope records of Site 1264 contain 1754 data points.

Stable isotope stratigraphy
The δ  Power spectral results indicate the dominance of the long-term (400-kyr) eccentricity cycle (Fig. 4). Additional smaller peaks are found at the short (95 and 125-kyr) eccentricity periods and to a lesser degree at the obliquity (41-kyr) period. No clear precession-related peaks are detected in the power spectra. Cross spectral analysis revealed that both records are highly coherent at the eccentricity periodicities with 5 the δ 13 C record lagging δ 18 O by 36±8, 0±3 and 5±3 kyr for the 400, 125 and 95kyr periods, respectively (Fig. 4). Almost similar results were found for the δ 13 C and δ 18 O records of Ceara Rise, indicating a strong coupling between climate states and changes in the oceanic carbon reservoir (Zachos et al., 1997(Zachos et al., , 2001bPaul et al., 2000). The dominance of 400-kyr eccentricity-related variability in both records is confirmed by

1-Dimensional inverse modelling output
A set of 1-D ice sheet models for West and East Antarctica, Greenland, North America and Eurasia in combination with an inverse routine was applied to deconvolve the δ 18 O signal into separate temperature (δ T ) and ice volume (δ w ) components (De Boer 20 et al., 2010) and followed the procedures as previously described (De Boer et al., 2010;Bintanja et al., 2005) (Fig. 6, Supplement Fig. 3 is equivalent to the amount of land-ice storage in Antarctica and on the Northern Hemisphere (mainly Greenland). On average, δ w leads δ T by ∼7 kyr, implying that ice-sheet growth precedes polar cooling, in contrast to previous findings (Bintanja and Van de Wal, 2008). The outcome of our ice-sheet model simulations show that changes in δ 18 O are accompanied by large shifts in ∆T NH (Fig. 6, Supplement Fig. 3). Minimum 5 NH air temperatures of ∼4 • C were reached during Mi-1 (∼23.1 Ma) and caused a small ice sheet to develop on Greenland, which may have contributed to a global sea level lowering in the order of 20 to 30 cm (Fig. 6). These findings are in agreement with the expected pCO 2 threshold described for NH land-ice at ∼280 ppmv (DeConto et al., 2008) and the reconstructed drop in atmospheric CO 2 levels approaching that value 10 within error (Kürschner et al., 2008;Pagani et al., 1999). An extra sensitivity experiment has been applied to test the significance of the Northern Hemisphere contribution to sea level lowering across the Oligocene-Miocene boundary.  , respectively (Fig. 7). The build-up phases (large-scale Antarctic ice-sheet expansions) seem to occur as a result of a weak imprint of the ∼100-kyr eccentricity-related climate cycles. Accord-10 ing to our age model, these intervals are not perfectly aligned to either a node in the ∼2.4 myr modulation of eccentricity nor to a node in the ∼1.2 myr modulation of obliquity (Fig. 4). They do appear, however, during short intervals in which the power of the ∼100-kyr eccentricity cycle is significantly suppressed (e.g. the 400-kyr minima at ∼23.1, ∼22.3, ∼21.4, and ∼19.8 Ma, black arrows in Fig. 4). Since the timing of these 15 ice-sheet build-up phases occur two and four 400-kyr eccentricity cycles apart, this suggest that a long-term non-linear mechanism of some kind may have modulated the dominantly eccentricity-paced Antarctic ice-sheet expansions and retreat during the late Oligocene and early Miocene. An example of a non-linear mechanism could be the merging of (several) large East Antarctic ice-sheets (DeConto and Pollard, 2003), 20 which may have resulted in an ice cap of significant proportions that survived the relatively warm global climate conditions during a weak eccentricity maximum; analogous to the mechanism proposed for the Mid Pleistocene Transition where the Laurentide and Corderillan ice sheets merged (Bintanja and Van de Wal, 2008). It is still possible that this non-linear mechanism derived from a pre-conditioning of the Antarctic ice 25 sheets by changes in the Earth's obliquity, since this cycle strongly determines insolation variations at high latitudes. Although the imprint of obliquity on the δ 18 O record of Site 1264 is weaker, its clear presence in the Ceara Rise records indicates that CPD 6,2010 Dynamics of ∼100-kyr glacial cycles during the early Miocene the Antarctic ice fluctuations are at least partly controlled by high-latitude insolation changes (Zachos et al., 1997(Zachos et al., , 2001bPaul et al., 2000;Pälike et al., 2006a). Given our age constraints, the link between the long-term (∼1.2 myr) obliquity modulation and the ice-sheet build-up phases are as yet too inconsistent too suggest any causal relationship between them (Fig. 4). Our findings thus imply a greater non-linearity in the early 5 Miocene climate system to internal ice-sheet dynamics, which may have caused the (Mi-) episodes of large Antarctic ice-sheet expansion. These episodes were then followed by enhanced climate variability on ∼100-kyr timescales during their termination phases. Thus, although the long-term (two or four 400-kyr cycles) pacing of ice-sheet build-up phases is probably controlled by (a) non-linear mechanism(s), their accompa-10 nying termination phases are, for a relatively briefer period of time (≤400-kyr), characterised by a greater linear climatic response to (weak or strong) ∼100-kyr eccentricity cyclicity (Lisiecki, 2010). The more symmetrical terminations of the individual ∼100-kyr cycles during the Oligocene-Miocene transition clearly illustrate this enhanced linearity of the Antarctic ice-sheet (Fig. 4). The consecutive ice-sheet build-up phases followed  where both records are coherent (panel D) represent the coupling between climate states and the changes in the oceanic carbon reservoir which has also been described in other records (Zachos et al., 1997(Zachos et al., , 2001bPaul et al., 2000).   with Site 1264. Ages of Sites 563, 608 (Berggren et al., 1995) and magnetostratigraphy of site 747 (Oslick et al., 1994) have been recalculated to the ATNTS2004 (Lourens et al., 2004). Sites 929 and 1090 are plotted on the Walvis Ridge Site 1264 age model. The ∼100-kyr "worlds" described in this study are close within the age estimates of the previously described Mi-1, Mi-1a and Mi-1aa zones or episodes.