Unsupervised statistical classification of the Antarctic marginal ice zone.

This work presents a new method for quantifying the Antarctic marginal ice zone (MIZ). The MIZ acts as an interface between the open Southern Ocean and the consolidated inner pack, and is generally described as area of sea ice affected by ocean surface waves. We use standalone CICE6, which includes a floe size distribution with atmospheric, oceanic and wave forcing, to simulate the evolution of Antarctic sea ice from 2010 – 2020. CICE output variables were categorised as static (sea ice concentration, age, thickness, etc.), thermodynamic, and dynamic. Unsupervised statistical methods were used to classify distinct sea ice regions and then to identify the dominant processes which contribute to the spatial and temporal variance of Antarctic sea ice cover. The unsupervised sea ice classification agrees with recent MIZ extent estimations using altimetry observations of wave attenuation. The addition of floe size information enhanced our MIZ classification to include high concentration pancake fields (which are promoted by waves). These results support the inclusion of floe size within sea ice modelling, and the importance of multi-variate approaches to describe sea ice.