We have performed some experiments with a thermodynamic model of the type used by Klaus Hasselmann. With this simple model you can transform an SST time series that is representative of the active region into a time series that looks a lot like the DOE time series for the tropical land stations. So I believe the relationship between the interannual temperature variabilities in different parts of the tropics is much simpler than those for other variables like precipitation and cloudiness.

GROISMAN: I wanted to say that I completely agree that there has to be more than one decay time, and neither the initial decay time nor a sustained signal is an acceptable approximation of the long decay times. And second, don't forget how much global temperature variance differs by season. And sometimes just a change in the Y-axis scale will make a trend look good.

WALLACE: It's only the shorter-term variability—the "noise"— that's larger in winter; the interdecadal component appears to be the same for both. That's why the scatter plots for the summer months look better.