The Pliocene and Early Pleistocene, between 5.3 and 0.8 million years ago, span a transition from a global climate state that was 2-3• C warmer than present with limited ice sheets in the Northern Hemisphere to one that was characterized by continental-scale glaciations at both poles. Growth and decay of these ice sheets was paced by variations in the Earth's orbit around the Sun. However, the nature of the influence of orbital forcing on the ice sheets is unclear, particularly in light of the absence of a strong 20,000-year precession signal in geologic records of global ice volume and sea level. Here we present a record of the rate of accumulation of iceberg-rafted debris o shore from the East Antarctic ice sheet, adjacent to the Wilkes Subglacial Basin, between 4.3 and 2.2 million years ago. We infer that maximum iceberg debris accumulation is associated with the enhanced calving of icebergs during ice-sheet margin retreat. In the warmer part of the record, between 4.3 and 3.5 million years ago, spectral analyses show a dominant periodicity of about 40,000 years. Subsequently, the powers of the 100,000-year and 20,000-year signals strengthen. We suggest that, as the Southern Ocean cooled between 3.5 and 2.5 million years ago, the development of a perennial sea-ice field limited the oceanic forcing of the ice sheet. After this threshold was crossed, substantial retreat of the East Antarctic ice sheet occurred only during austral summer insolation maxima, as controlled by the precession cycle.

We interpret this erosion to be associated with retreat of the ice sheet margin several hundreds of kilometres inland and conclude that the East Antarctic ice sheet was sensitive to climatic warmth during the Pliocene.Recent satellite observations reveal that the Greenland and West Antarctic ice sheets are losing mass in response to climatic warming 6 . Basal melting of ice shelves by warmer ocean temperatures is proposed as one of the key mechanisms facilitating mass loss of the marine-based West Antarctic ice sheet 7 . Although thinning of ice shelves and acceleration of glaciers has been described

1] The Integrated Ocean Drilling Program Expedition 318 to the Wilkes Land margin of Antarctica recovered a sedimentary succession ranging in age from lower Eocene to the Holocene. Excellent stratigraphic control is key to understanding the timing of paleoceanographic events through critical climate intervals. Drill sites recovered the lower and middle Eocene, nearly the entire Oligocene, the Miocene from about 17 Ma, the entire Pliocene and much of the Pleistocene. The paleomagnetic properties are generally suitable for magnetostratigraphic interpretation, with well-behaved demagnetization diagrams, uniform distribution of declinations, and a clear separation into two inclination modes. Although the sequences were discontinuously recovered with many gaps due to coring, and there are hiatuses from sedimentary and tectonic processes, the magnetostratigraphic patterns are in general readily interpretable. Our interpretations are integrated with the diatom, radiolarian, calcareous nannofossils and dinoflagellate cyst (dinocyst) biostratigraphy. The magnetostratigraphy significantly improves the resolution of the chronostratigraphy, particularly in intervals with poor biostratigraphic control. However, Southern Ocean records with reliable magnetostratigraphies are notably scarce, and the data reported here provide an opportunity for improved calibration of the

Antarctic glaciation was abruptly established during the EoceneOligocene transition (EOT) in two ∼200-kyr-spaced phases between 34.0 Myr and 33.5 Myr ago, as recorded by the oxygen isotope composition of marine biogenic calcite 3,4,9 (δ 18 O). The first shift (EOT-1) is believed to represent a transient glaciation [10][11][12] , later followed by the establishment of a continental-scale ice sheet across the Oligocene isotope event-1 33.7 Myr ago; ref. 4). This is consistent with Northern Hemisphere ocean-sediment cores, which indicate a 60 ± 20 m relative sea level (rsl) fall across the EOT (refs 13-15). Under isostatic equilibrium conditions, the observed regression nearly corresponds to the rsl drop expected from glacioeustasy 6 . Along the Antarctic margins, however, the rsl changes accompanying the glaciation are expected to strongly deviate from the eustatic, because of large crustal and gravitational perturbations induced by the ice sheet on the deformable Earth 6 . Furthermore, strong regional rsl change gradients would be maintained long after the ice-sheet stabilization, by the flexure of lithosphere 16 . This necessitates self-consistent physical models for rsl change to compare near-field sedimentary sequences with far-field ice-sheet volume estimates.Here we evaluate the regionally varying rsl changes in response to glacial expansion and their effects on glaciomarine facies 17 around East Antarctica with a numerical model for glacial-hydro isostatic adjustment (GIA). Our model is based on the solution of the gravitationally self-consistent sea-level equation 7,8 for a prescribed Antarctic ice-sheet chronology 18 and a linear viscoelastic rheology for the solid Earth (Methods and Supplementary Information). We compare the model results with sedimentary records from the Wilkes Land Margin (Fig. 1a) The ice-sheet model employed in our GIA computations is characterized by a 2.2 Myr growth phase caused by a combination of decreasing CO 2 and orbital forcing that drives summer temperatures below the threshold for glaciation 18 but uses a new reconstruction of Antarctic topography 24 at EOT time (Methods). The ice-sheet volume at the glacial maximum corresponds to ∼69.0 m of equivalent sea level in this model, ∼14.0 m more than previous modelling results 18 , probably owing to larger Antarctic land surface 24 .We run a reference simulation (Fig. 1a-d) for an Earth model defined by an elastic lithosphere thickness (LT) of 100 km, and by a viscosity profile (RVP) that is discretized into a lower mantle (LM), a transition zone (TZ) and an upper mantle (UM) and is characterized by viscosities of 1.0 × 10 22 , 5.0 × 10 20 and 2.5 × 10 20 Pa s respectively (RVP-100 km-LT simulation). To evaluate how the GIA signal varies according to the mantle viscosity, we perform simulations for an ensemble of viscosity profiles (EVPs) characterized by viscosities varying in the range of