Antarctic sea ice plays a critical role in modulating global climate, influencing surface albedo, air-sea carbon fluxes and the global ocean overturning circulation. Despite global warming, overall Antarctic sea ice extent increased during 1979–2013, however the majority of Coupled Model Intercomparison Project phase five (CMIP5) models simulate a decline, and mechanisms causing this discrepancy remained elusive. Here I show that weaker westerly-wind jet intensification trends simulated by CMIP5 models may contribute to the disparity. During austral summer a strengthened jet increases upwelling of cooler subsurface water and strengthens equatorward transport, conducive to increased sea ice. This cooling process is underestimated in the majority of models and is insufficient to offset global warming. Through the sea ice-albedo feedback, models produce a high-latitude surface warming and sea-ice decline, contrasting observations. A realistic simulation of observed wind changes may be crucial for reproducing observed sea ice trends.

A strengthened Amundsen Sea Low has also been shown to largely explain the recent sea ice increase in the Ross Sea and decrease in the Bellingshausen Sea. While these changes are not generally seen in CMIP5 simulations, I show they can be reproduced in simulations of two independent coupled climate models constrained by observed tropical variability. I further show that the phase change in the Interdecadal Pacific Oscillation from positive to negative over 1979–2013 likely contributed to the strengthened Amundsen Sea Low and pattern of sea ice trends, highlighting the importance of accounting for teleconnections from low to high latitudes.

The Southern Ocean surface freshened in recent decades, yet CMIP5 models underestimate this. I demonstrate that imposing a broad-scale surface freshening to the Southern Ocean in global coupled climate model experiments causes a surface cooling and sea-ice increase, due to reduced ocean convection and weakened entrainment of warm subsurface waters to the surface. Additional experiments with surface salinity restoration applied to capture observed regional salinity trends accurately represent the spatial pattern of surface temperature and sea-ice trends around Antarctica. These results highlight the importance of accurately simulating changes in Southern Ocean precipitation, meltwater and salinity to capture changes in ocean circulation, surface temperature and sea ice.