Spread in the magnitude of climate model interdecadal global temperature variability traced to disagreements over high-latitude oceans.

Unforced variability in global mean surface air temperature can obscure or exaggerate global warming on interdecadal timescales, thus understanding both the magnitude and generating mechanisms of such variability is of critical importance for both attribution studies as well as decadal climate prediction. Coupled atmosphere-ocean general circulation models (climate models) simulate a wide range of magnitudes of unforced interdecadal variability in global mean surface air temperature (UIT), hampering efforts to quantify the influence of UIT on contemporary global temperature trends. Recently, a preliminary consensus has emerged that unforced interdecadal variability in local surface temperatures (UIT) over the tropical Pacific Ocean are particularly influential on UIT. Therefore, a reasonable hypothesis might be that the large spread in the magnitude of UIT across climate models can be explained by the spread in the magnitude of simulated tropical Pacific UIT. Here we show that this hypothesis is mostly false. Instead, the spread in the magnitude of UIT is linked much more strongly to the spread in the magnitude of UIT over high-latitude regions characterized by significant variability in oceanic convection, sea ice concentration, and energy flux at both the surface and the top of the atmosphere (TOA). Thus, efforts to constrain the climate model produced range of UIT magnitude would be best served by focusing on the simulation of air-sea interaction at high latitudes.