Towards an improved Antarctic sea-ice representation in HadGEM3-GC5

The historical runs of CMIP6-era coupled climate models generally exhibit negative biases in Antarctic sea ice, as identified across a range of models during the CMIP6 simulations (Roach et al. 2020). The UK's national coupled climate model, HadGEM3, has been no exception to this. The CMIP6 version, HadGEM3-GC3, underestimated Antarctic sea ice in historical simulations owing to a Southern Ocean warm bias (Andrews et al. 2020). As part of the DEFIANT (Drivers and Effects of Fluctuations in sea Ice in the ANTarctic) project, in this research we perform an analysis of the representation of sea ice in HadGEM3-GC5, a more recent version of the coupled model. Analysis of existing HadGEM3-GC5 simulations has identified unrealistic convection events associated with open water polynyas. We have started to perform a suite of sensitivity experiments to investigate the importance of the freshwater budget and ocean mixing parameterisation scheme on these convection events, and subsequently on pan-Antarctic sea ice. These initial experiments take the form of short simulations with constant year-2000 forcings, incorporating various parameter perturbations and modifications to freshwater input. We present evidence of the improved characterisation of pan-Antarctic sea ice in HadGEM3-GC5 compared to HadGEM3-GC3, and the preliminary analysis of perturbation simulations aimed at understanding and addressing the remaining challenges in the model coupled climate system.