The East Antarctic Ice Sheet (EAIS) is the largest potential contributor to sea-level rise. However, efforts to predict the future evolution of the EAIS are hindered by uncertainty in how it responded to past warm periods, for example, during the Pliocene epoch (5.3 to 2.6 million years ago), when atmospheric carbon dioxide concentrations were last higher than 400 parts per million. Geological evidence indicates that some marine-based portions of the EAIS and the West Antarctic Ice Sheet retreated during parts of the Pliocene1,2, but it remains unclear whether ice grounded above sea level also experienced retreat. This uncertainty persists because global sea-level estimates for the Pliocene have large uncertainties and cannot be used to rule out substantial terrestrial ice loss3, and also because direct geological evidence bearing on past ice retreat on land is lacking. Here we show that land-based sectors of the EAIS that drain into the Ross Sea have been stable throughout the past eight million years. We base this conclusion on the extremely low concentrations of cosmogenic 10Be and 26Al isotopes found in quartz sand extracted from a land-proximal marine sediment core. This sediment had been eroded from the continent, and its low levels of cosmogenic nuclides indicate that it experienced only minimal exposure to cosmic radiation, suggesting that the sediment source regions were covered in ice. These findings indicate that atmospheric warming during the past eight million years was insufficient to cause widespread or long-lasting meltback of the EAIS margin onto land. We suggest that variations in Antarctic ice volume in response to the range of global temperatures experienced over this period—up to 2–3 degrees Celsius above preindustrial temperatures4, corresponding to future scenarios involving carbon dioxide concentrations of between 400 and 500 parts per million—were instead driven mostly by the retreat of marine ice margins, in agreement with the latest models5,6.
Access optionsAccess options
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank the Antarctic Research Facility for AND-1B samples, and J. X. Mitrovica for his help in performing the glacial isostatic adjustment modelling. This research was supported by National Science Foundation (NSF) grant ARC-1023191 (to P.R.B. and L.B.C.); Boston College start-up funds (to J.D.S.); Vermont Established Program to Stimulate Competitive Research (EPSCoR) grants EPS-1101317 and NSF OIA 1556770 (to K.U. and D.M.R.); NSF grant EAR-1153689 (to M.W.C.); and the New Zealand Ministry of Business Innovation and Employment contract C05X1001 (to T.N. and N.R.G.). This is Lawrence Livermore National Laboratory project LLNL-JRNL-735619.Reviewer information
Nature thanks J. Gosse, E. Gasson, J. Willenbring and the other anonymous reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
a–d, Simulated erosion potential under the Antarctic Ice Sheet, calculated from modelled driving stress and basal velocity fields for several uniform (atmosphere and ocean) warming scenarios of: 4 °C (a), 8 °C (b), 12 °C (c) and 15 °C (d)55. The location of the AND-1B core is shown by the yellow dot. We note that erosive zones tend to extend towards the continental interior with warming. dT, temperature anomaly from present; dV, ice-volume anomaly from present, in sea-level equivalent (s.l.e.).
a–d, Antarctic land above sea level (yellow) 0 kyr (a), 5 kyr (b), 10 kyr (c), and 15 kyr (d) after a near-instantaneous (1-kyr) collapse of all marine-based ice-sheet sectors, in two different models of mantle viscosity26. Model 1 is from ref. 56, and model 2 (our model) has the following parameters: lithosphere thickness, 96 km; upper-mantle viscosity, 5 × 1020 Pa s−1; and lower-mantle viscosity, 1022 Pa s−1. The location of the AND-1B core is shown by the star.
a, b, Cumulative exceedance probabilities of measured (that is, not blank-corrected) 10Be (a) and 26Al (b) nuclide abundances in AND-1B samples (blue) and in all blanks run by the same operator in the same low-level fume hood (red), with 1σ uncertainties. These plots display the fraction of measurements that exceed a given nuclide abundance. Note that probabilities are generally higher for the samples than the blanks; in other words, a random draw from the samples is more likely to be above a random draw from the blanks, suggesting that they are separable populations.
Shaded intervals surrounding the blue line show 1σ uncertainties, while shaded intervals not surrounding the blue line show the possible range of decay-corrected concentrations in samples that are below the detection limit. The dashed black line simulates the 26Al concentration in non-eroding material at 2,000 metres above sea level (m asl) that was originally saturated at 14 Ma and subsequently decayed under cold-based, non-erosive ice. The fact that several AND-1B samples have higher concentrations than those in this extreme scenario (which is the most favourable to having nuclides persist to the present) suggests that the AND-1B nuclides were produced after the expansion of the EAIS in the mid-Miocene.
Extended Data Fig. 5 Modelled concentrations of cosmogenic nuclides for various durations of interglacial exposure and glacial erosion rates.
a–d, Simulated 10Be (a, b) and 26Al (c, d) concentrations in material sourced from sea level and from 2,000 m asl in Antarctica as a function of the fraction of time for which land is exposed, during 40-kyr glacial cycles. (Results are nearly identical if the cycles are instead 100-kyr long.) Erosion rates were assumed to be 0 m per Myr during ice-free conditions, on the basis of geologic evidence for negligible late Cenozoic erosion in ice-free areas of the TAMs9,10. Black arrows next to the scale bars show the range of decay-corrected nuclide concentrations in AND-1B samples. The model was initialized with zero nuclides at 8 Ma (representative of conditions suggested by AND-1B sample H); the model also assumes instantaneous transport of eroded sediment to the ocean with no mixing, and continuous radioactive decay. Concentrations shown are the Pliocene (5 Ma to 3 Ma) average. Comparison of these simulations with AND-1B nuclide concentrations suggests that land exposure in sediment source regions was probably quite limited in duration or extent through the Plio-Pleistocene.
a–d, Each panel shows actual AND-1B decay-corrected 10Be concentrations with 1σ uncertainty (green), as well as simulated 10Be concentrations assuming a single 10-kyr (a), 50-kyr (b), 100-kyr (c) and 200-kyr (d) exposure of a bedrock column in the mid-Pliocene. The exposure event was chosen to start at 3.6 Ma and extend for up to 200 kyr in duration on the basis of the presence of a 60-m-thick diatomite unit in the AND-1B core, thought to reflect warm interglacial conditions from 3.6 Ma to 3.4 Ma1. Simulated records are driven by production at sea level (grey) or at 2,000 m asl (black), and are subjected to continuous radioactive decay and continuous erosion at rates of 0 m per Myr (solid lines), 20 m per Myr (dashed lines), and 100 m per Myr (dotted lines). The model assumes that the sediment source was initially devoid of nuclides and that sediments are transported instantaneously to the sea floor. The synthetic time series have been binned to the same resolution as the AND-1B data.
Extended Data Fig. 7 Modelling a mid-Pliocene exposure event with eroded bedrock mixed through a deformable bed.
The figure shows AND-1B decay-corrected 10Be concentrations with 1σ uncertainties (green). It also depicts simulated 10Be concentrations, assuming a single exposure event from 3.6 Ma to 3.4 Ma and routing of eroded bedrock through a well mixed deformable bed, for various bed thicknesses and erosion rates. Material eroded from the bedrock profile is instantaneously mixed throughout the deformable bed in each time step, and an equal amount of material is removed from the bed, keeping its thickness constant. Sediment mixing in the deformable bed dilutes the surface 10Be signal of the exposure event but extends its longevity through time in comparison with the bedrock simulations shown in Extended Data Fig. 6. Simulated records are driven by production at sea level, and subjected to continuous radioactive decay and continuous erosion. The model assumes that the bedrock and deformable bed were initially devoid of nuclides and that sediments eroded from the deformable bed are transported instantaneously to the sea floor. The synthetic time series have been binned to the same resolution as the AND-1B data.
Extended Data Fig. 8 Conceptual diagram showing the outcomes of Bayesian one-group t-tests and their interpretation.
a, Nuclides are credibly present above background: that is, the sample value is greater than the mean of the blanks (defined at the mode of the posterior distribution), and the region of uncertainty surrounding the sample value fully excludes the 90% credible interval (C.I.) on the posterior distribution of the mean of the blanks. The grey shaded regions give the uncertainty range in the sample nuclide concentration. b, Nuclides are not credibly present above background: the sample value is less than or equal to the blank mean. c, Nuclides are not credibly present above background: although the sample value is greater than the blank mean, the region of uncertainty surrounding the sample value does not fully exclude the 90% C.I.
This file contains AND-1B sediment processing data, AND-1B cosmogenic nuclide data and process blank cosmogenic nuclide data.