Examining Antarctic sea ice bias sensitivity in the multi-variate parameter space using a global coupled climate modelling system

Abstract

Coupled global numerical climate models (GCMs) typically underestimate mean Antarctic sea ice area and extent, particularly during the austral summer months, contributing to uncertainties in climate prediction. This study examines the climatological behaviour of Antarctic sea ice in a coupled GCM in the multivariate sea ice model parameter space. Individual parameters dominate the ice response in different seasons and regions, with a compensatory effect in some parameter combinations and an amplified effect in others; however, certain parameter combinations are found to improve aspects of Antarctic sea ice climatology well beyond the limitations of a univariate approach. For example, the disparity between observed and simulated summer sea ice extent and area is halved, and the tendency towards very low-concentration ice (<15 %) reduced in favour of a more compact summer and autumn ice pack. Regardless, clear limitations in the extent to which a coupled GCM can be calibrated with sea-ice model parameters also emerge. Relatively unconsolidated winter ice cover persists and, in some experiments, becomes looser still, exacerbating the already overestimated maximum sea ice extent. Furthermore, the seasonal evolution of sea ice and the exaggerated asymmetry of the seasonal cycle, with the onset of ice advance too slow and maximum sea ice reached too late, sees negligible improvements. We note that, even with the large gains under certain parameter combinations, bias and other deficiencies still remain. Using coupled data assimilation to optimise parameters in both sea-ice and ocean models will likely assist in contributing to further model improvements.