Covariability of aspects of north American climate with global sea surface temperatures on interannual to interdecadal timescales

Rotated canonical correlation analysis between seasonal- and longer-mean global SSTs and either U.S. surface temperatures or 700-hPa heights in the Pacific-North America region have led to decompositions into three distinct signals. One of these represents the interannual variability of ENSO and a second is related to the North Atlantic oscillation and exhibits considerable variability on interdecadal timescales. In contrast the temporal behavior of the third, which is referred to here as the global signal, is mostly characterized by a steady trend since the late 1960s. The robustness of this time series to variations in the analyses, as well as the robustness of the spatial structure of the SST pattern accompanying it, suggests that the decomposition represents a successful separation of the climate signal from the climate noise. When viewed in the context of other recent work, the global signal cannot be discounted as a "fingerprint" of global warming. Finally, calculations that exploit ensemble mean output from prescribed-SST GCM runs reveal notable systematic errors in the simulation of the features of all three signals.