Comparing the physics of ocean-atmosphere coupling and estimates of extratropical predictability in models and reality

The reliability of estimates of seasonal and longer-timescale predictability of extratropical phenomena rely crucially on the accuracy of the simulation of ocean-atmosphere interactions. Forcing can be in both directions and the overall interaction is a balance between the two. How can we be sure that models represent this balance correctly? If the atmospheric part of a coupled model is too sensitive to the underlying SST field, then seasonal predictability may be greatly over-estimated. Conversly, predictability could be under-estimated. Agreement of simultaneous correlations between oceanic and atmospheric variables in a model and reality is not sufficient to guarantee that coupling is properly represented. Also, the analysis of ensembles of coupled simulations does not in itself say anything about predictability in reality. A further test has been devised based on lagged correlations to compare the physics of coupling in models against reality and to compare model estimates of predictability with estimates from reality. Several coupled models have been tested in this way. Results have important implications for the predictability of the NAO, PNA and stormtracks.