Arctic sea ice has decreased substantially and is projected to reach a seasonally ice-free state in the coming decades. Little is known about whether dwindling Arctic sea ice is capable of influencing the occurrence of strong El Niño, a prominent mode of climate variability with global impacts. Based on time slice coupled model experiments, here we show that no significant change in the occurrence of strong El Niño is found in response to moderate Arctic sea-ice loss that is consistent with satellite observations to date. However, as the ice loss continues and the Arctic becomes seasonally ice-free, the frequency of strong El Niño events increases by more than one third, as defined by gradient-based indices that remove mean tropical Pacific warming induced by the seasonally ice-free Arctic. By comparing our time slice experiments with greenhouse warming experiments, we conclude that at least 37–48% of the increase of strong El Niño near the end of the 21st century is associated specifically with Arctic sea-ice loss. Further separation of Arctic sea-ice loss and greenhouse gas forcing only experiments implies that the seasonally ice-free Arctic might play a key role in driving significantly more frequent strong El Niño events.

The Antarctic sea ice has undergone dramatic changes in recent years, with the highest recorded sea ice extent in 2014 and the lowest in 2017. We investigated the impacts of the observed changes in these two extremes of Antarctic sea ice conditions on the atmospheric circulation in the Southern Hemisphere. We conducted three numerical simulations with different seasonal cycles of Antarctic sea ice forcings using the Community Atmosphere Model Version 5: the maximum sea ice extent in 2014 (ICE_14), the minimum sea ice extent in 2017 (ICE_17), and the average sea ice extent between 1981 and 2010 (ICE_clm, reference simulation). Our results suggest that the atmospheric response in the Southern Hemisphere showed strong seasonal variations and the atmospheric circulation in winter was more sensitive to the decreased Antarctic sea ice in 2017 than the increased sea ice in 2014. In ICE_14, the westerlies over the polar region were enhanced in summer, but there was no significant change in the zonal-averaged wind in winter. In contrast, in ICE_17, there was a clear equatorward shift in the subtropical jet in winter, but no significant change in summer. The temperature responses were limited to the Antarctic coast, where there were changes in the sea ice in ICE_14 and ICE_17. The warming on the coast of the Amundsen Sea in summer led to a slight increase in precipitation in both simulations.

Associations between the autumn Arctic sea ice concentrations (SICs) and North American winter precipitation were examined using singular value decomposition. The results show that a reduced SIC in the majority of the Arctic is accompanied by dry conditions over the Great Plains, the southern United States, Mexico, eastern Alaska, and southeastern Greenland, and by wet conditions over the majority of Canada, the northeastern United States, and the majority of Greenland. Atmospheric circulation anomalies associated with the SIC variability show a wave train structure that is persistent from autumn to winter and is responsible for the covariability between the autumn Arctic SICs and North American winter precipitation. This relationship suggests a potential long-term out-look for the North American winter precipitation.