In the previous newsletter (Newsletter #8, https://ukesm.ac.uk/portfolio-item/coupling-greenland-and-antarctic-ice-sheets/), we briefly described two UKESM1 variations 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). Here we discuss in a bit more detail the latter. UKESM-is2 has been developed to allow coupling between the BISICLES ice sheet and UKESM1 models via land-ice interaction (over Greenland & Antarctic) and ocean-ice interaction (Antarctica only).

In the aforementioned article, we mentioned that we were working on a UKESM-is2 configuration with a model resolution of N96-ORCA025 (a higher ocean model resolution than the standard UKESM1). Recently, we decided to deliver two UKESM1-is2 configurations, i.e. in addition to the N96-ORCA025 resolution, we also implement an N96-ORCA1 (ocean model resolution the same as in UKESM1). Like the ocean model component of UKESM1, both of these configurations use GO6 settings (Storkey et al. 2018) in the global ocean. The traceability hierarchy in GO6 requires that the lower resolution ORCA1 use higher horizontal viscosity, as well as the Gent-McWilliams (GM) parameterisation describing mixing due to unresolved ocean eddies. Extending the ocean into the cavities under ice shelves in ORCA1 is a challenging task. For stability reasons, we reduced the horizontal viscosities and switched off the GM parameterisation under ice shelves.

Basal melt rates under ice shelves are calculated explicitly, using the three-equation model of Jenkins (1991). The coupling frequency between BISICLES and the NEMO ocean in the fully coupled run is 1 year, therefore after the coupled model runs for 1 year, the resulting mean basal melt rate and the surface mass balance (SMB) are fed into BISICLES which is then run to update the ice-sheet geometry. The coupling becomes two-way (full) when the new ice sheet geometry is passed back into the coupled model for a subsequent 1-year period of run. While numerical stability during the geometry change has been maintained, tracer conservation issues related to this geometry change are not yet covered by this work.

We run both model configurations for 40 years under present-day climate forcing. The BISICLES ice-sheet model, with mesh refinement down to 1.2km, was initialised using an Antarctic ice-sheet geometry obtained from a standalone BISICLES spin-up run forced by a similar present-day climate. The initial temperature and salinity in the global ocean are taken from EN4 data, which are then extrapolated from Antarctic coasts into the cavities.

Figure 1: Melt rate (in m/year) in four Antarctic regions. The middle panel is the observation data, whereas the left and right panel are from the UKESM-is2 runs.

One of the most important ocean-ice sheet interactions to capture in a coupled model is basal melting. Figure 1 shows the spatial pattern of the Antarctic basal melt rate in the two configurations, compared to estimates from observational data (Rignot et al., 2013). We divide the continent into 4 well-known regions: West Antarctica (covering ice shelves near the Amundsen & Bellinghausen Sea, known as warm shelves), the East Antarctica (cold shelves), and the two large ice shelves Ross and Ronne-Filchner (in the Ross & Weddell sea respectively). The regional pattern of melt in both N96-ORCA1 & N96-ORCA025 configurations follows that seen in the observations in the way that melting (red colours) and freezing (blue colours) occur in largely the same areas and that larger the major melt rates also occur in almost the same areas.

The main difference between the observations and the two configurations is the colour intensity, where the observations looks darker, indicating total melt rate is higher in the observations. This is confirmed in Figure 2a, which displays total basal melt rate across the continent, where the solid red line and the two red dashed lines are the mean, upper and lower limit of observations respectively. In the last 30 years of the run, both the N96-ORCA1 (blue) and N96-ORCA025 (green) are below the observation limit.

Pine Island Glacier (PIG) is a location of major importance as it is among the fastest melting glaciers in Antarctica and may therefore give an indication of how warm West Antarctica is becoming. Its average temperature and total melt flux are shown in Figure 2b and 2c, both figures show similar time-series patterns as the total melt flux.

Figure 2: (a) Total basal melt flux (Gton/year) across the Antarctic continent, (b) total basal melt flux (Gton/year) from the Pine Island Glacier, (c) Average temperature in the Pine Island Glacier cavity.

Despite the lower melt rate in the two UKESM1-IS2 configurations, they show different behaviours in the last 25 years of the simulations. The N96-ORCA1 melt flux is very steady and it is only slightly lower (50-100 GTon/year) than the lower limit of observation. With further tuning and approximations available in the 3-equation model, we expect an improvement of the N96-ORCA1 results. The opposite is the case in the N96-ORCA025 simulation, where the time series melt flux shows a decreasing pattern, falling to less than half the observed melt flux. The lower average PIG temperature gives an indication that warm deep water does not reach the continental shelf in this model run. This is confirmed in Figure 3 (the last 10 years mean of 300-1000m depth averaged temperature), where the N96-ORCA025 Amundsen Sea average temperature decreases rapidly from the open ocean to the continental slope. In the same figure, the average temperature anomaly also shows that a cold bias in the West Antarctic coast is higher in N96-O25.

Further analysis shows the Drake Passage transport in UKESM-is2 N96-ORCA1 and N96-ORCA025 configurations is similar to that seen in the HadGEM3- GC3.1 N96-ORCA1 and N216-ORCA025 simulations (Kuhlbrodt, T., et al, 2018), where the transport in the higher resolution ocean drops quickly below 100 Sv whereas the lower resolution stays above 140 Sv (much closer to observational estimates). This gives us a hint that coupled models with ORCA025 ocean model (regardless of the resolution of the atmospheric model) has significant deficiencies in the Southern Ocean. Further investigations we carried out suggest almost all available ORCA025 ocean models (coupled as well as standalone) have a tendency to expand the Ross gyre further east as far as into the Bellinghausen Sea. This is generally not found in UKESM1’s N96-ORCA1 ocean models or many other coupled or standalone ocean models that use ORCA1.

Figure 3: Depth averaged (300-1000m) temperature (average over the last 10 years of the simulation) in (a) N96-ORCA1, (b) N96-ORCA025. The lower panel shows the depth averaged temperature bias (minus EN4 data) in (c) N96-ORCA1, (d) N96-ORCA025.

These preliminary analyses have given us some details that the UKESM1-is2 with lower ocean resolution (N96-ORCA1) looks promising and hence its challenging developments are worth exploring and testing. As for the UKESM1-is2 N96-ORCA025, with its higher computational cost, before a fully interactive coupling is continued it will be important to improve the simulation of the main Southern Ocean thermohaline and circulation features.

References

• Storkey, D., et al.: UK Global Ocean GO6 and GO7: a traceable hierarchy of model resolutions. Geoscientific Model Development Discussions, 11, 3187–3213, 2018
• Jenkins, A.: A one-dimensional model of ice shelf-ocean interaction, Geophys. Res.-Oceans, 96, 20671–20677, 1991
• Rignot, E., et al.: Ice-shelf melting around Antarctica, Science, 341, 266–270, 2013
• Kuhlbrodt, T., et al. : The Low-Resolution Version of HadGEM3 GC3.1: Development and Evaluation for Global Climate. Journal Of Advances In Modeling Earth Systems, 10, 2865–2888, 2018