Present-day capabilities of numerical and statistical models for atmospheric extratropical seasonal simulation and prediction

A statistical model and extended ensemble integrations of two atmospheric g eneral circulation models (GCMs) are used to simulate the extratropical atm ospheric response to forcing by observed SSTs for the years 1980 through 19 88. The simulations are compared to observations using the anomaly correlat ion and root-mean-square error of the 700-hPa height field over a region en compassing the extratropical North Pacific Ocean and most of North America. On average, the statistical model is found to produce considerably better simulations than either numerical model, even when simple statistical corre ctions are used to remove systematic errors from the numerical model simula tions. In the mean, the simulation skill is low, but there are some individ ual seasons for which all three models produce simulations with good skill. An approximate upper bound to the simulation skill that could be expected f rom a GCM ensemble, if the model's response to SST forcing is assumed to be perfect, is computed. This perfect model predictability allows one to make some rough extrapolations about the skill that could be expected if one co uld greatly improve the mean response of the GCMs without significantly imp acting the variance of the ensemble. These perfect model predictability ski lls are better than the statistical model simulations during the summer, bu t for the winter, present-day statistical forecasts already have skill, tha t is as high as the upper bound for the GCMs. Simultaneous improvements to the GCM mean response and reduction in the GCM ensemble variance would be r equired for these GCMs to do significantly better than the statistical mode l in winter. This does not preclude the possibility that, as is presently t he case, a statistical blend of GCM and statistical predictions could produ ce a simulation better than either alone. Because of the primitive state of coupled ocean-atmosphere GCMs, the vast m ajority of seasonal predictions currently produced by GCMs are performed us ing a two-tiered approach in which SSTs are first predicted and then used t o force an atmospheric model; this motivates the examination of the simulat ion problem. However, it is straightforward to use the statistical model to produce true forecasts by changing its predictors from simultaneous to pre cursor SSTs. An examination of the decrease in skill of the statistical mod el when changed from simulation to prediction mode is extrapolated to draw conclusions about the skill to be expected from good coupled GCM prediction s.