Eccentricity pacing and rapid termination of the early Antarctic ice ages

Abstract

Earth's obliquity and eccentricity cycles are strongly imprinted on Earth's climate and widely used to measure geological time. However, the record of these imprints on the oxygen isotope record in deep-sea benthic foraminifera (δ¹⁸O_b) shows contradictory signals that violate isotopic principles and cause controversy over climate-ice sheet interactions. Here, we present a δ¹⁸O_b record of high fidelity from International Ocean Drilling Program (IODP) Site U1406 in the northwest Atlantic Ocean. We compare our record to other records for the time interval between 28 and 20 million years ago, when Earth was warmer than today, and only Antarctic ice sheets existed. The imprint of eccentricity on δ¹⁸O_b is remarkably consistent globally whereas the obliquity signal is inconsistent between sites, indicating that eccentricity was the primary pacemaker of land ice volume. The larger eccentricity-paced early Antarctic ice ages were vulnerable to rapid termination. These findings imply that the self-stabilizing hysteresis effects of large land-based early Antarctic ice sheets were strong enough to maintain ice growth despite consecutive insolation-induced polar warming episodes. However, rapid ice age terminations indicate that resistance to melting was weaker than simulated by numerical models and regularly overpowered, sometimes abruptly.

Fig. 1. Contrasting indications of the astronomical pacemaker of Antarctic glacial-interglacial cycles during the high CO₂ mid-Cenozoic unipolar icehouse.

a atmospheric carbon dioxide (pCO₂) reconstructions with 2σ uncertainties, from phytoplankton (cyan triangles) and boron isotopes (blue circles). b marine benthic foraminiferal oxygen isotope (δ¹⁸O_b) composite record CENOGRID with 10-point (black) LOESS smoothing, relevant climate events and glacial states shown. c contrasting astronomical pacing of two marine δ¹⁸O_b records (ODP sites 926 and 1264, refs. ^,) 26−18 Ma. d power spectra for the data shown in c. Vertical black lines denote bandwidths used throughout this paper. e Antarctic (ice-proximal) records AND-2A, CRP-2/2A, and DSDP Site 270 with colours representing simplified lithology from refs. ^,,: light grey mudstone (often including ice-rafted debris), dark grey sandstone, green diamictite, brown sandstone, and sinusoidal lines disconformities. The question mark in DSDP Site 270 denotes uncertain lithological age. Site locations shown in Figure S1.

Fig. 2. Our δ¹⁸O_b record from IODP Site U1406 compared to the records from sites 926 and 1264 and the astronomical series.

a δ¹⁸O_b from Site U1406 (bright red, this study). EMI (Early Miocene Interval), OMT (Oligocene-Miocene Transition), LOW (Late Oligocene Warming), and MOGI (Mid-Oligocene Glacial Interval). Arrows indicate abrupt decreases in δ¹⁸O_b at Site U1406. b and c δ¹⁸O_b at sites 926 (dark red) and 1264 (yellow), respectively. d and e La2004 astronomical solutions for obliquity and eccentricity, respectively, with obliquity instantaneous amplitude and long-term eccentricity modulations of ~405 kyr and ~2.4 Myr. Site locations shown in Fig. S1.

Fig. 3. A globally coherent response to eccentricity and inconsistent response to obliquity in the Oligo-Miocene δ¹⁸O_b record.

a δ¹⁸O_b of ODP/IODP sites U1406 (bright red; presented here), 926 (dark red), 1090 (purple), 1218 (blue), and 1264 (yellow) during EMI (Early Miocene Interval), OMT (Oligocene-Miocene Transition), LOW (Late Oligocene Warming), and MOGI (Mid-Oligocene Glacial Interval). Blue shading indicates colder intervals (roughly equivalent to the MOGI/LOW transition and OMT interval) with high-amplitude variability in δ¹⁸O_b on frequencies corresponding to eccentricity. b eccentricity variance (V_Ecc) of the δ¹⁸O_b records shown in a), the composite δ¹⁸O_b record^, and the La2004 eccentricity solution shown in (c) with long-term eccentricity modulations of ~405 kyr and ~2.4 Myr offset for clarity. Grey line denotes maximum V_Ecc values in the 5.0−1.2 Ma interval of the LR04 δ¹⁸O_b stack (see Fig. S3). d obliquity variance (V_Obl) of the δ¹⁸O_b records shown in a), the composite δ¹⁸O_b record^,, and La2004 obliquity solution (La2004, dashed) shown in (e) with amplitude modulation. Grey line denotes maximum V_Obl values in the 5.0−3.1 Ma interval of the LR04 δ¹⁸O_b stack (see Fig. S3). Note that the y-axes of V_Obl and V_Ecc are scaled differently to allow site-to-site comparisons (see Fig. S2b for plot with identical y-axes). Site locations shown in Fig. S1.