As a first check on this hypothesis, we compare the atmospheric circulation trends simulated by atmospheric GCMs coupled with an ocean or with imposed SSTs (AMIP experiment from CMIP5); the latter being in theory able to reproduce changes caused by natural SST variability. While coupled models simulate in aggregate trends that project on the SAM structure, strongest in summer, the AMIP simulations add in the winter season a pronounced Amundsen Sea Low signature (and a PNA signature in the northern hemisphere) both consistent with a Niña-like trend in the tropical Pacific.
We then use a specific coupled GCM setup, in which the surface winds over the tropical Pacific are strongly nudged towards the observed ones, variability. The model then reproduces the Amundsen-Sea low trend in winter, as well as a dipole in sea-ice cover (but no global increase). Further analysis shows that the sea-ice dipole is forced by surface heat flux anomalies in early winter, then maintained by ice-albedo feedbacks.