http://www.nature.com/articles/s41558-019-0551-4.epdf?author_access_token=lev7cCLvJaqZgqEL2fkrCtRgN0jAjWel9jnR3ZoTv0OkPw7OWz-ctumf1Sllaa-sNqBW8kixcwl-ojSoyKUmUfvHbXmZ3llXRmZN-HO_pmKRWEHLwCuZqZkuv4bolog-ehQV4R4jg8i93P7ntZX4_w%3D%3D
Observations show that reduced regional sea-ice cover is coincident with cold mid-latitude winters on interannual timescales.
However, it remains unclear whether these observed links are causal, and model experiments suggest that they might not be.
Here we apply two independent approaches to infer causality from observations and climate models and to reconcile these
sources of data. Models capture the observed correlations between reduced sea ice and cold mid-latitude winters, but only
when reduced sea ice coincides with anomalous heat transfer from the atmosphere to the ocean, implying that the atmosphere
is driving the loss. Causal inference from the physics-based approach is corroborated by a lead–lag analysis, showing that
circulation-driven temperature anomalies precede, but do not follow, reduced sea ice. Furthermore, no mid-latitude cooling is
found in modelling experiments with imposed future sea-ice loss. Our results show robust support for anomalous atmospheric
circulation simultaneously driving cold mid-latitude winters and mild Arctic conditions, and reduced sea ice having a minimal
influence on severe mid-latitude winters.