To predict the future contributions of the Antarctic ice sheets to sea-level rise, numerical models use reconstructions of past ice-sheet retreat after the Last Glacial Maximum to tune model parameters1. Reconstructions of the West Antarctic Ice Sheet have assumed that it retreated progressively throughout the Holocene epoch (the past 11,500 years or so)2,3,4. Here we show, however, that over this period the grounding line of the West Antarctic Ice Sheet (which marks the point at which it is no longer in contact with the ground and becomes a floating ice shelf) retreated several hundred kilometres inland of today’s grounding line, before isostatic rebound caused it to re-advance to its present position. Our evidence includes, first, radiocarbon dating of sediment cores recovered from beneath the ice streams of the Ross Sea sector, indicating widespread Holocene marine exposure; and second, ice-penetrating radar observations of englacial structure in the Weddell Sea sector, indicating ice-shelf grounding. We explore the implications of these findings with an ice-sheet model. Modelled re-advance of the grounding line in the Holocene requires ice-shelf grounding caused by isostatic rebound. Our findings overturn the assumption of progressive retreat of the grounding line during the Holocene in West Antarctica, and corroborate previous suggestions of ice-sheet re-advance5. Rebound-driven stabilizing processes were apparently able to halt and reverse climate-initiated ice loss. Whether these processes can reverse present-day ice loss6 on millennial timescales will depend on bedrock topography and mantle viscosity—parameters that are difficult to measure and to incorporate into ice-sheet models.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
J.K. and the Weddell Sea fieldwork were funded by Natural Environmental Research Council (NERC) grant NE/J008087/1, led by R. Hindmarsh. Logistical support was provided by many members of the British Antarctic Survey’s air unit and field operations team. We particularly thank I. Rudkin and S. Webster for assistance in the field. We also thank H. Pritchard for supplying bed elevation data and Schlumberger Limited for a software donation. PISM development is supported by NASA grants NNX13AM16G and NNX13AK27G. T.A. is supported by the Deutsche Forschungsgemeinschaft (DFG) in the framework of the priority program ‘Antarctic Research with comparative investigations in Arctic ice areas’ through grants LE1448/6-1 and LE1448/7-1.We acknowledge the European Regional Development Fund, the German Federal Ministry of Education and Research, and the Land Brandenburg for providing high-performance computer resources at the Potsdam Institute for Climate Impact Research. We also thank the Gauss Centre for Supercomputing e.V. (http://www.gauss-centre.eu) for providing computing time on the GCS Supercomputer SuperMUC at Leibniz Supercomputing Centre (http://www.lrz.de; project code pr94ga). We thank C. Buizert for providing ice-core temperature reconstructions; D. Peltier for access to eustatic sea-level reconstructions; J. Lenaerts for surface mass balance data from the RACMO climate model; and S. Jamieson for providing the RAISED consortium’s grounding-line reconstructions. R.P.S., J.C., R.D.P. and S.T. were funded by National Science Foundation (NSF) WISSARD Project grants ANT-0839107, ANT-0839142, ANT-0838947 and ANT-0839059. Collection of subglacial sediment samples at Subglacial Lake Whillans and the Whillans Grounding Zone was facilitated by the US Antarctic Program and the efforts of multiple field support teams, including the drilling team from the University of Nebraska–Lincoln and WISSARD traverse personnel, as well as by Air National Guard and Kenn Borek Air who provided air support. WIS, KIS and BIS samples were recovered by B. Kamb’s program at the California Institute of Technology (1988–2001), which included R.P.S. and S.T.; samples from the US Antarctic Program’s Ross Ice Shelf Project (1977–1979) cores were made available for study by the US Antarctic Sediment Core Repository, Florida State University. P.L.W. is funded by a NERC Independent Research Fellowship (NE/K009958/1). This research is a contribution to the Scientific Committee on Antarctic Research (SCAR) Solid Earth Response and Influence on Cryosphere Evolution (SERCE) program. We thank R. Arthern, R. Bell, R. Hindmarsh, C. Martín, J. Southon and K. Tinto for discussions that contributed to this study. We particularly thank D. Pollard for sharing ideas and unpublished Penn State model outputs for discussion.Reviewer information
Nature thanks R. Drews, J. Smith and the other anonymous reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
a, Radargram, aligned perpendicular to the divide ridge (inset shows the location). One undulating isochrone is delineated with colours showing normalized elevation. b, c, Close-up views of the boxed regions indicated in a. In both close-up panels, diffractors (hyperbolic reflectors) are interpreted as expressions of relic crevasses (data are unmigrated). The red vertical dashed line is the present-day grounding line31. d–f, Radargrams aligned approximately perpendicular to northern relic crevasses (d and e show migrated data). In c (6 km ≤ x ≤ 8 km) and f (0.3 km ≤ x ≤ 1.4 km) isochrones intercepting the bed are evident. g, Three relic crevasses mapped across several radar lines over a Radarsat Antarctic Mapping Project (RAMP) image94. The inset is an oblique, three-dimensional view of the features over an interpolated surface, showing the bed elevation zb (see Methods). Crevasse spacing in these areas ranges between approximately 200 m and 600 m. The arrow indicates the view direction of the oblique view.
a, RAMP94 image showing the surface expression of ice-shelf crevasses in synthetic aperture radar data. Light areas indicate high backscatter from (near-)surface reflectors, interpreted to be surface crevasses. Crevasses form over and immediately downstream of Doake Ice Rumples. We hypothesize that crevasses once formed in a similar manner over the topographic high beneath the northern tip of HIR. b, Close-up view of the crevasses (the black box in a shows the location), whose spacing (100–300 m), orientation (perpendicular to the flow of the ice shelf) and lateral extent (roughly 10 km) are similar to the steeply dipping reflectors discovered near the bed of the northern tip of HIR (for example, Extended Data Fig. 2g) in the region of a topographic high. Yellow curves are flow lines computed from satellite-derived surface velocities30. Flow is from bottom to top. Polar stereographic coordinates are in km. The present-day grounding line31 is in red.
Cross-sections along transects through the Weddell (left) and Ross (right) Sea sectors, at 5-kyr intervals (for transect locations, see Fig. 3). The horizontal axis shows the distance from the present-day grounding line. The vertical blue dashed line shows the position of maximum grounding-line retreat. a, b, 15 kyr bp, with the grounding line close to the continental shelf edge. c, d, 10 kyr bp, with the grounding line having retreated to approximately its minimum, most retreated location. e, f, 5 kyr bp, with both ice shelves grounded on sub-ice-shelf bathymetric highs owing to seafloor uplift. g, h, Present day, with the grounding line having re-advanced to roughly the present-day configuration in response to the grounding of the ice shelf and uplift at the grounding line. The Crary, Bungenstock and Henry ice rises (CIR, BIR and HIR) are labelled in g and h. The Whillans Ice Stream (WIS) and Subglacial Lake Whillans (SLW) sediment-core locations are labelled in d. Blue dotted lines show the observed present-day ice-sheet bed, ocean floor and ice surface29, remapped on to the 15-km grid of the ice-sheet model.
a, b, Results from four simulations (the reference simulation, and three additional experiments) designed to examine the cause of re-advance in the Weddell Sea (a) and Ross Sea (b) sectors (Methods). The most inland grounding-line location in the reference simulation, at around 10 kyr bp, is in blue. The colour map shows the flow buttressing number95 at 10 kyr bp in the ‘No uplift, grounding of ice rises’ experiment. The ice-front position is in grey. Background images over the grounded ice sheet are from MOA32. c, Basal topography and bathymetry in the Weddell Sea sector (with the grounding line in red) according to a 1-km-resolution dataset, constrained by geophysical observations (Bedmap 2; ref. 29). d, Conservative remapping of these data to 15-km resolution. Remapping substantially lowers the apparent maximum bed elevations beneath ice rises in the Weddell Sea sector: 135 m at KIR, 112 m at HIR and 36 m at BIR.
The 11 grey lines show exponential 14C-decay curves connecting the 14C/12C ratios (scale on the left) measured on acid-insoluble inorganics (AIOs) from our subglacial sediment samples to the apparent radiocarbon ages calculated from these measurements. The latter calculation assumes that the initial 14C/12C ratios in AIOs were equal to the modern ratio in radiocarbon dating standards. As discussed in the text and Methods, organic matter in Antarctic glacigenic sediments frequently contains an admixture of old 14C-dead material41,42. The record of oxygen isotopes in water ice from the WAIS Divide ice core (green line, with scale on the right) provides climatic context for the period between now and 35 kyr bp (ref. 96). Three key climatic periods are labelled: WAIS LGM39, Antarctic cold reversal (ACR) and Holocene.
In the middle and right panels are time series of grounding-line position along transects, showing model sensitivity in the Weddell Sea (middle panels) and Ross Sea (right panels) sectors to: a, different sea-level reconstructions69,78,79,80; b, different scalings of the sea-level forcing to mimic self-gravitational effects; c, different surface-temperature forcings; and d, different accumulation forcings. In the left panels are: a, four alternative sea-level reconstructions; b, three alternative scalings of the reference-simulation sea-level forcing and a version that has been uniformly shifted 2,000 years earlier; c, temperature reconstructions from two ice cores, WAIS Divide and EPICA Dome C (EDC), and a reconstruction from the Last Interglacial (from EDC data); and d, four alternative accumulation histories. The constant LGM accumulation uses the EPICA Dome C core83 and a scaling of 2% per degree. Temperature and accumulation are expressed relative to the present day. Grounding-line positions are relative to the present-day position (vertical dashed line) along the transacts shown in Fig. 3. In all simulations, the grounding line is in its most advanced position, up to 1,000 km beyond its present-day position, before MWP1a (14.4 kyr bp; horizontal dotted line). During the Holocene the grounding line retreats up to 500 km upstream of its present location, and usually re-advances towards its present-day position. Grey shading indicates the spread of grounding-line responses, and grey curves show the mean of each sensitivity experiment. In each case the violet curve shows the reference simulation. Grounding-line positions (based on marine and terrestrial geological evidence) from the RAISED reconstruction with associated uncertainties are shown in black2.
Time series of grounding-line position along transects, showing model sensitivity in the Weddell and Ross Sea sectors to: a, mantle viscosity, μ, and the flexural rigidity of the lithosphere, D; b, c, enhancement factors ESSA and ESIA; d, the sliding-law exponent, q; e, the till water decay rate, T, and till effective pressure fraction, N; f, minimum till friction angle and the method used to derive the friction angle (see Methods); g PICO ocean model parameters for overturning strength, C, and heat exchange, g; and h, the dependence of calving rate on the ice-shelf spreading rate, K (Extended Data Table 2). In each panel, the violet curve shows the reference simulation. Grounding-line positions (based on marine and terrestrial geological evidence) from the RAISED reconstruction, with associated uncertainties, are shown in black2.
The results of three simulations using different grid resolutions: a, 15 km (reference simulation; identical to Fig. 3); b, 10 km; and c, 7 km. Owing to computational limitations, the two higher-resolution simulations cover only the past 20 kyr, so they lack a higher-resolution spin-up period. These higher-resolution simulations display similar Holocene retreat and re-advance driven by isostatic rebound to the reference simulation, but the LGM extent and grounding-line re-advance in the Weddell Sea are much less. A full exploration of the resolution dependence of the model requires using higher resolution during entire simulations for all ensemble members. This is limited at present by computing resources. Background shading shows basal topography and bathymetry29.
Top-left panel shows the vertically averaged ice speed (logarithmic colour scale). Lower-left panel shows the bedrock rebound rate. Notice the rapid rebound following deglaciation in the Weddell and Ross Sea sectors. Upper-right panel shows modelled surface elevation relative to present day on grounded ice and basal melt rates beneath ice shelves. Lower-right panel shows time series of the sea-level forcing (left vertical axis), the temperature forcing (right vertical axis) and the above-flotation volume of the ice sheet expressed in sea-level equivalent, assuming a constant ocean area of 3.61×1014 m2. See Methods for details of forcings and parameterisations.