Abstract
The GISS general circulation model (GCM) is used to investigate the influence of a positive sea surface temperature (SST) anomaly in the subtropical North Pacific on the Northern Hemisphere wintertime circulation. As the set of model data is small, the signal-to-noise ratio is low, and statistical significance is difficult to establish. Only local effects are detected with the univariate t-test, but a multivariate analysis based on hypothesis testing shows that the SST anomaly induces a small signal in the middle and upper troposphere. The same technique is applied to investigate whether the GCM response can be predicted in part from a specification of the SST anomaly, using a simple linear quasi-geostrophic equivalent barotropic model. The model is forced by a vorticity sink proportional to the SST anomaly, as the latter induces condensation heating and upper level divergence. When the barotropic model is linearized about the zonally symmetric part of the mean GCM state in the control runs, no satisfactory prediction of the GCM response is obtained. However, when the zonal variations of the GCM basic control state are taken into account, the linear model prediction is statistically significant, even if only a small fraction of the GCM signal is accounted for. The agreement with the GCM data is best when the equivalent barotropic level is in the upper troposphere, and it depends little on the amount of dissipation. On the other hand, it is very sensitive to the details of the mean flow waviness.
