Polar-to-midlatitude teleconnections in a warming world: Statistical relationships from large ensembles

How are polar-to-midlatitude teleconnections represented in recent large ensembles of coupled climate model simulations? And how do they evolve with global warming? Using the rich information on internal variability available from large ensembles, we investigate the relationship between sea ice amount and atmospheric circulation for both Arctic and Antarctic sea ice variability in CESM2 and ACCESS-ESM1-5, using a composite analysis. We find that the links between sea ice and sea level pressure (SLP), the midlatitude jet stream and temperature depend on the region in which sea ice varies, for instance with low Barents-Kara sea ice in January being associated with a positive North Atlantic Oscillation SLP pattern and high pressure over Northern Eurasia. These circulation patterns persist with increased levels of global warming, until around 3 or 4°C when they start to evolve in some cases, as sea ice starts to disappear. Surface air temperatures are anomalously high around the region of sea ice retreat with varying patterns of remote cooling elsewhere. Lagged analysis shows that sea-ice circulation relationships when the atmosphere leads sea ice are very similar to the instantaneous relationships, suggesting that the latter largely reflects the atmospheric patterns leading to reduced sea ice. For positive lags (sea ice leading the atmosphere), for some regions the SLP teleconnections persist in a weakened state for subsequent months, whilst for others they evolve, e.g. into a negative Arctic Oscillation response for Barents-Kara sea ice reduction. However, results for positive lags differ between the two models examined. SLP relationships with Antarctic sea ice are model dependent, but feature a negative Southern Annular Mode pattern in ACCESS-ESM1-5. In CESM2, we find a less zonally symmetric pattern which also consists of high pressure over the pole in Autumn and Winter.