Robin Smith: NCAS and UKESM core group

One of the most novel developments in the UKESM project will be the incorporation of interactive models of continental-scale ice sheets. Ice sheets have long been treated by climate modellers as static boundary conditions that do not evolve significantly over the decadal to centennial timescales of addressed with general circulation models. However, this approach ignores the climate-ice feedbacks that play a fundamental role in the Earth’s climate and does not consider how the current Earth System will change in the coming centuries as Greenland’s outlet glaciers speed up and the risk of Antarctic ice shelf collapse increases, both potentially leading to significant loss of mass from grounded ice sheets with large impact on global sea level. UKESM1 will include the option of ice sheets that interact in a physically consistent way both with the atmosphere above and also the ocean underneath the ice shelves, making it a world-leading model in this area of science.

The ice sheet model to be used in UKESM1 is BISICLES (Cornford et al., 2013), a vertically integrated “higher order” model that can model both the slower-moving bulk of the ice as well as the dynamics of fast ice-streams and floating shelves without the significant computational expense of solving the full Stokes flow equations. BISICLES uses an adaptive grid mesh that saves resources by only putting the necessary high resolution where it is required, for example, at the grounding line where the ice shelf floats clear of the ocean floor. BISICLES is therefore one of the computationally less expensive components of UKESM1 and should scale well over the large number of processors required in UKESM1 simulations.

Successfully coupling BISICLES to the Global Coupled (GC) 3 configuration of the Met Office Unified Model (UM) in a physically consistent way brings a number of major scientific and technical challenges. The resolutions required by atmosphere and ice models are different. Modelling most atmospheric processes requires timescales of minutes but can be represented on spatial scales of several kilometres. Key areas of the ice must be modelled with spatial scales of a few metres but changes occur over much slower timescales of months or longer. Ice sheets integrate the details of climate they see over many decades, therefore small but persistent model errors in the polar climate can easily become major problems for the ice. Ice sheet simulations are very sensitive to their initial state, which has evolved over many thousands of years and also to their bottom boundary conditions. The latter can be buried under kilometres of ice and are poorly understood. To further complicate matters, flowing ice not only alters the active domain of the ice model, but changing sea-level turns land into ocean and back again, something that the current UM is structurally not designed to do.

Figure 1
Figure 1. Computed ice velocity for Antarctica from BISICLES (right), and meshing and grounding line location for the Pine Island Glacier (left) (source: SL Cornford).

Finding solutions to these issues has motivated parallel developments in the UM, which are being tested in UKESM prototypes. JULES, the land surface component of UKESM1, will use multi-layer snow on special sub-grid scale tiles at different elevations to improve the spatial resolution of the surface forcing provided to the ice sheets. NEMO, the ocean component of UKESM1, has been modified to allow circulation underneath the solid surface of the ice shelf and includes parameterisations of sub-shelf melting. Moving the land/ocean boundaries as the ice volume and extent change remains a largely unresolved issue however and the underlying spatial resolution of the UM still brings a number of restrictions to the ice simulation. For example, the narrow fjords that control many of Greenland’s outlet glaciers are simply not resolved in a global ocean model.

Due to the scale of these technical challenges additional development is required to make the UM technically ice sheet compatible. Tuning may also be required to ensure that significant biases are not introduced into all components by the additional degrees of freedom introduced by a fully interactive treatment of land ice. It is therefore planned to introduce the interactive BISICLES ice sheet component in a later, separate release of UKESM, UKESM1-IS in mid-2017. This is the version that will be used for the coupled ice sheet-climate simulations to be submitted under ISMIP6 to CMIP6.

References

Cornford, SL, Martin, DF, Graves, DT, Ranken, DF, Le Brocq, AM, Gladstone, RM, Payne, AJ, Ng, EG & Lipscomb, WH, 2013, ‘Adaptive mesh, finite volume modeling of marine ice sheets’. Journal of Computational Physics, vol 232., pp. 529-549

Download this article as a PDF (114KB)