On the limitations of prescribing sea surface temperatures in atmospheric General Circulation Model experiments

Abstract

It is widely accepted that extratropical sea surface temperature (SST) variability arises primarily in response to fluctuations in surface heat fluxes driven by atmospheric variability. The back-interaction exerted by SST anomalies on the atmosphere in the real world is fundamentally different, therefore, from the one-way forcing that takes place in atmospheric general circulation model (GCM) experiments with prescribed SSTs, even if those SSTs are time dependent. Several recent studies have shown, for instance, that a primary effect of coupling is to attenuate the thermal damping exerted by the infinite heat capacity ocean in prescribed SST integrations. This also suggests that atmospheric low-frequency variability will be underestimated in atmospheric GCM experiments with prescribed SSTs.

We examine these issues using a consistent approach; namely, we compare the simulated climate from a coupled ocean-atmosphere GCM to that from an atmospheric GCM, where the latter is forced with the sea ice and SST fields internally generated by the former. This has been done using the third generation of the Hadley Centre GCM in coupled (HadCM3) and uncoupled (HadAM3) mode. The coupled data are from a 150-year subset of a longer integration, and the uncoupled data consist of a 3-member (150-year) ensemble. The results confirm that coupling enhances lower tropospheric thermal variance and reduces net surface energy flux, but the effects are smaller than most previous studies indicate. Coupling can also slightly alter the amplitude (and persistence) of dominant modes of natural middle latitude variability such as the North Atlantic Oscillation. Overall, however, the detrimental effect of specifying middle latitude SST is small, consistent with a weak forced response relative to internal variability.