Loading VESL Page loader
Sea level rise resulting from removing ice in Antarctica and/or Greenland.

Relative ice volume above flotation for Antarctica basins.


Simulation Controls

Movie Controls

Plot of projected Sea Level Rise

Extra Controls

Camera Controls

Camera Frame Controls



We present results of ISSM simulations from present day to 500 years into the future. These simulations capture sea-level rise around the world as one of Antarctica's main glaciers, Thwaites Glacier, retreats inland. The first simulation shows how the grounding line (line where the grounded ice of Antarctica comes afloat) retreats when the bedrock under the ice does not move (green line). The second simulation includes a new state of the art model where the bedrock is now considered elastic, and rebounds as the ice retreats inland (red line). As time evolves and Thwaites Glacier starts thinning, both grounding lines retreat. However, the red grounding line is slower to retreat. This in turns results in a smaller contribution to sea-level rise. Indeed, the rebound of the bedrock acts as a negative feedbafk, or stabilizer, slowing down the thinning of Thwaites Glacier.

You can find more details on our methods, datasets and results in our publication at: Slowdown in Antarctic Mass Loss from Solid Earth and Sea-Level Feedbacks, Science, 2018 . Thanks for Eric Steig for his perspective on our study, which you can find at How fast will the Antarctic ice sheet retreat, Science 2018.?.


For each city, sea-level is modeled for 500 years as a result of both the thinning of Antarctica (in particular Thwaites Glacier) but also as a result of the melting of ice everywhere in the world, and steric effects where sea-level increases as the ocean warms. The simulation is fully interactive in Antarctica, but for the rest of the world, ice and ocean processes are not modeled, rather extrapolated in time using observations from present-day satellites (GRACE in particular). These simulations therefore are not fully representative of all the processes at play in sea-level rise, but they definitely showcase the impact of elastic bedrock rebound in Antarctica on projections of sea-level rise.

Datasets Used for the Simulation


  • Atmospheric forcing: we apply a constant yearly surface mass balance corresponding to an average between 1979 and 2010, from the Regional Climate Model RACMO2.1 (Lenaerts et al, 2012).
  • Ocean forcing: we use a melt-rate parameterization calibrated against the MITgcm ocean circulation model (Seroussi et al, 2017) and generalize it to all Antarctica ice shelves using already available computed melt rates (Schodlok et al, 2016) and existing knowledge of sub-ice-shelf cavity shapes.
  • Boundary conditions at the ice/bed interface: basal friction is inverted for usin InSAR surface velocities (Rignot et al, 2014).
  • Bedrock topography: we cover the model domain over the entire AIS as observed today, and interpolate its geometry from the Bedmap2 dataset with specific exception to the Amundsen Sea sector, Recovery Ice Stream, and Totten Glacier. In these areas, we combine ice thickness data from ice penetrating radar, satellite derived ice velocities and surface mass balance from a regional climate model to map the bed at 150 m spatial resolution using the mass conservation (MC) approach (Morlighem et al, 2011) previously adapted to the Amundsen Sea Sector.
  • Surface topography: we rely on a dataset consistent with the MC approach (Morlighem et al, 2011)

Global Sea Level

  • World coastlines: crude GSHHG coastline (Global Self-consistent, Hierarchical, High-resolution Geography Database,c-L1 coastlines, ( Wessel et al, 1996 ).
  • World ice mass balance: for all other glaciated areas, such as Greenland or Alaska, mass transport is constrained to match GRACE mass trends from 2003 to 2016, and extrapolated in time for every year starting model year 2000.

Model Settings

  • Antarctica mesh: resolution ranging from 1 km for the grounding line of Thwaites Glacier to 50 km inland. The Antarctica mesh comprises a total of 221,266 elements.
  • Antartica model: for the Antarctic ice sheet, we rely on the ice module of ISSM, which captures mass transport, stress balance ( using the order 2D SSA approximation MacAyeal 1989 ) and high-resolution sub-element grounding-line dynamics (Seroussi et al, 2017). The ice-sheet model is initialized using an instantaneous spin-up of a 2D SSA modele where basal friction is inverted for from InSAR surface velocities. This spin-up is carried out to match the configuration of Antarctica in terms of surface velocity and ice thickness in year 2000. The thermal regime of the ice sheet is assumed steady-sate and the rheology of the ice dependent on an Arhenius type law. For ice shelves, where the rheology is poorly known, an inversion using observed velocities is carried out. The model itself is run in 2D, because 3D higher-order models such as the Blatter/Pattyn ( Pattyn, 2003) are too computationnally expensive for a 500 year 1-50 km resolution model. To account for the numerical shock in the instantaneous spin-up, we carry out a quick relaxation of 1 year before letting the transient model fully evolve in time.
  • Global sea-level mesh: resolution ranging from 1 km for the grounding line of Thwaites Glacier to 1000 km in the middle of the Pacicific. Both the global mesh and the Antarctica mesh coincide. The global mesh comprises 412,364 elements (from which 375,192 are in Antarctica).
  • Sea-level is computed over 500 years, at 5 year time intervals, according to the Farrel et al, 1976 sea-level computation theory. Our model implementation is described in the SESAW module of the Ice Sheet System Model (ISSM) framework. Both the RSL and ice-sheet model are coupled on the same unstructured mesh. Because mesh resolution and computational requirements are significant (each model run takes 4 days on 180 cpus of the NASA AMES Pleaides cluster), the RSL model is only run at time intervals of 10 years. The ice-sheet model is run on Antarctica at a time interval of 14 d. Every 10 years, ice thickness changes for every glaciated area of he Earth are used to compute a new relative sea-level. In turn, this new relative sea-level is transferred to the glaciated areas for use as boundary conditions to the ice-flow model, at the calving front (sea-level) and at the ice/bed interface to update the bedrock topography. Here, only Antarctica is computed interactively on a basin by basin basis. For all other glaciated areas, such as Greenland or Alaska, mass transport is constrained to match GRACE mass trends from 2003 to 2016, and extrapolated in time for every year starting model year 2000 (Larour et al, 2017). No attempt is made at modeling the stress balance, calving front and grounding line dynamics of these areas.

How to visualize our results

Click Let's go, and you will see a movie (500 year long, 10 year time steps) of the thinning of Antarctica, focusing in particular on the region of Thwaites Glacier, along with grounding line positions retreating. In the controls section, you can decide to display only the new model (see our publication in Science), the old model (similar to models carried out within the ISMIP6 or SeaRISE benchmark activities where solid-Earth is not taken into account), or both models' grounding lines. The new model includes elastic rebound of the Earth, while old models do not. You can also choose to display the grounding line for the entire Antarctica continent, or just the Thwaites Glacier (default choice). At any point in time, you can stop the movie, and slide the cursor to see a snapshot in time. Feel free to rotate the movie in 3D and do some exploring around. You will also see a plot in the controls section showing sea-level rise for the next 500 years for a specific city (selectable in a drop-down menu). The red curve corresponds to the new model, the green curve to the old models. Feel free to change the city at will. Finally, in the controls section, you can control the display of the grounding line to either follow the bedrock (true position, but rather hard to visualize) or be at sea level (easier to follow).



If you have any questions or feedback, please send us an email.



Green : grounding line position
(old models)
Red : grounding line position
(new model with elastic rebound)