It is now well recognized that tropical biases in coupled general circulation models (CGCMs) are seriously degrading our ability to predict sub-seasonal-to-interannual climate variability and our confidence in climate change projections. Documenting tropical biases via model comparison studies are well-known in the current literature, the focus of a series of workshops (see http://www.iges.org/ctbp/ ), and were a key component of the recent World Climate Research Program (WCRP) model biases meeting in San Francisco California. The literature and the workshops are quite clear with respect to the dominant tropical biases. For example, the so-called “double ITCZ” problem remains a major source of error in simulating the annual cycle in the tropics. The so-called equatorial Pacific cold tongue is too narrowly confined to the equator and extends to far into the western Pacific. Along the equator the models fail to adequately capture the zonal SST gradient and typically have thermoclines that are far too diffuse. Most CGCMs fail to capture the meridional extent of the SST anomalies in the eastern Pacific and tend to produce SST anomalies that extent too far into the western tropical Pacific. Most, but not all, CGCMs produce ENSO variability that occurs on time scales considerably faster than observed, although this is improving to some degree, but the mechanisms for this improvement are unclear. The conventional wisdom is that westward extension of the ENSO events and the fast periodicity is due to the cold tongue mean state errors noted above. Simply, errors in the mean state are the cause for the errors in the anomalies.

While there have been a number of notable successes in trying to reduce some of these errors, the “fixes” rarely translate from one model to another. In fact, resolution increases seems to improve some models, but not others, which is not well understood. Here we propose a different approach for understanding these systematic errors and why promising sensitivities fail to translate from one model to the next. We suggest that the errors in the mean state are, at least in part, due to errors in the local air-sea coupling feedbacks in the western Pacific. In other words, if there are large errors in the simulation of the tropical weather statistics and how weather events impact the ocean, then the large-scale climate simulation is seriously degraded. Moreover, we argue hypothesize that the changes (or lack there of) in the weather statistics can explain the large differences in model sensitivity. In this talk we described the theoretical basis for this hypothesis and demonstrate how it can be used to improve air-sea feedbacks and the large scale simulation through a series of novel state-of-the-art CGCM simulations.