Progress coupling Greenland and Antarctic ice sheets into UKESM1

Robin Smith1*, Vicky Lee2* and Antony Siahaan3*

1 National Centre for Atmospheric Science (NCAS), 2 Centre for Polar Observation and Modelling (CPOM), 3 British Antarctic Survey (BAS). * UKESM core group member

As regular readers of this newsletter will know, the main version of UKESM1, being used for CMIP6 simulations, is not the only version of our Earth system model. Alongside other variations, we are developing versions of UKESM with direct, physical two-way couplings to models of the Greenland and Antarctic ice sheets – see “An overview of the Land Ice in UKESM1” (Newsletter #2, and “Progress towards interactive Ice Sheets in UKESM1” (Newsletter #6, We now have two separate configurations that have interactive ice sheets: one with only Greenland coupled (referred to as UKESM1-is1), and a prototype model where both Greenland and Antarctica are coupled (referred to as UKESM-is2). Coupling Greenland generally requires only atmosphere-ice coupling at the surface of the ice sheet. Due to Antarctica’s climate and a fringe of floating ice shelves means coupled ocean-ice interactions also need to be simulated.

ISMIP6 (Nowicki et al., 2016) is the project within CMIP6 coordinating projections of ice sheet mass loss under future climate change. Most of the work in ISMIP6 will use standalone ice sheet models, but some coupled climate-ice sheet simulations, with only Greenland coupled, are planned. Results need to be available in 2019 if they are to be considered by IPCC AR6. We have therefore frozen our most stable, Greenland-only UKESM-is configuration – an N96 atmosphere, ORCA1 ocean coupled to the ice sheet model BISICLES, with mesh refinement down to 1.2km as required to simulate details of the Greenland ice sheet (see figures 1-2). Simulations for ISMIP6 were started at the end of 2018. In keeping with the ISMIP6 design we first concentrate on idealised abrupt 4xCO2 and 1%/yr CO2 increase experiments, both relative to a control run representative of recent decades.

Figure 1: A snapshot of the coupled climate mass balance and glacier flow on Greenland simulated by the UKESM1-is model run for ISMIP6. The flow speeds and climate components all compare well with results from more specialised modelling studies.

There is potential for a significant “coupling shock” to occur when two previously uncoupled models are made to interact with each other and can respond to, and influence, biases in the other model. Directly coupling climate and ice sheet models, as we are doing, is a new and active area of research. Reliable techniques to guarantee a stable and realistic coupled state are therefore not yet available. Nevertheless, the climate and flow fields for Greenland in the component models of UKESM1-is have so far shown themselves to be compatible, and terms contributing to the mass balance of the Greenland ice sheet (e.g. snowfall, melt, runoff and calving) compare well with simulations by more highly-tuned regional models, with  little drift apparent in the control simulation (figure 2).

Figure 2: Accumulation and runoff components of the Greenland surface mass balance in a UKESM-is present day control run are stable and compare reasonably with results from a highly tuned regional model, RACMO (van Angelen et al. 2012). There are no observations for these quantities covering the entire ice sheet. The net positive surface mass term is balanced by calving of icebergs at the edges of Greenland.

As noted above, coupling Antarctica, as well as Greenland, is both technically and scientifically more challenging, requiring a good simulation of the ocean circulation under the ice shelves, as well as the of the surface climate of the ice sheet and ice dynamics and the ability for the models to pass information in all directions and update boundary conditions accordingly. We have made enough progress on the technical side of the UKESM ice coupling that we can now run a configuration (UKESM-is2) that has both Greenland and Antarctic ice sheets interactively coupled. As far as we are aware, no other global climate model exists with this capability.

Figure 3: Pine Island Glacier is observed to be thinning and retreating at up to 20 m/yr as the warm ocean melts up to 100m/yr from the underside of the ice shelf (Wingham et al. 2009, Jenkins et al. 2010). Our first simulation achieves reasonable melt rates in the ocean model, but the ice state is not yet compatible with these, and in our first simulation the ice flow from upstream means that thinning is patchy at Pine Island, while the Thwaites shelf thickens. Black lines show the limit of the grounded ice sheet, blue lines the limit of the floating ice. BISICLES data is shown on a coarser grid than actually used in the model.

To resolve the ocean under the ice shelves UKESM-is2 uses a higher resolution ocean model (ORCA025, with ~15km gridboxes at the Antarctic ice-ocean interface) than the Greenland-only UKESM1-is1. We still use the N96-resolution UKESM1 atmosphere to maintain an acceptable computational cost. To achieve a realistic Antarctic ocean-ice simulation, we require the water mass properties under the ice shelves to be reasonably accurate (there is significant regional variation in the waters around Antarctica, and some shelves are characterised as “warm”, and some as “cold”, e.g. Dinniman et al. (2016). We further require that the temperature and velocity of water below the ice shelves produces accurate melt rates at the ice fronts and, for the location and size of those melt rates to balance where the ice sheet model flows out from the grounded ice upstream. In addition to all this, we need to smoothly and realistically evolve these features as the ocean domain itself grows (or shrinks) as the ice shelves retreat (or advance). The coupling shock, outlined above, has potential to be particularly severe when simulating Antarctic ice as we need to balance the atmosphere, ice sheet and ocean models, all in the presence of positive feedback processes at the retreating ice fronts.

It is an extremely challenging problem, but our results so far are encouraging. Our first simulation with UKESM1-is2 shows a basically correct distribution of warm and cold shelves around Antarctica, and we find the highest melt rates under shelves in the Amundensen Bay, matching observations (figure 3). The ice and ocean are not yet fully compatible here, however, and our melting does not give us the observed rates of shelf retreat as flow from the grounded ice upstream balances the loss of ice to the sea. We do not plan to run a full suite of ISMIP6 climate change runs with UKESM-is+, but we will exploit our unique modelling capability by conducting at least one global ice sheet-climate sensitivity experiment in 2019. Watch this space…


  • Dinniman M., Asay-Davis, X., Galton-Fenzi, B., Holland, P., Jenkins, A. and Timmermann, R., Modeling Ice Shelf/Ocean Interaction in Antarctica: A Review, Oceanography, 2016
  • Jenkins, A., Dutrieux, P., Jacobs, S., McPhail, D., Perrett, J., Webb, A. and White, D.. Observations beneath Pine Island Glacier in West Antarctica and implications for its retreat Geosci., 2010
  • Nowicki S., Payne, A., Larour, E., Seroussi, H., Goelzer, H., Lipscomb, W., Gregory, J., Abe-Ouchi, A. and Shepherd, A.,Ice Sheet Model Intercomparison Project (ISMIP6) contribution to CMIP6, Model Dev. 2016
  • van Angelen, J., van den Broeke, M., Wouters, B. and Lenaerts, J.,Contemporary (1960-2012) evolution of the climate and surface mass balance of the Greenland Ice Sheet, Surveys of Geophys, 2012
  • Wingham, D., Wallis, D. and Shepherd, A., Spatial and temporal evolution of Pine Island Glacier Thinning, 1995-2006, Res. Lett., 2009

Download this Newsletter as a PDF (2015KB)