The misrepresentation of Tropical SSTs in climate models

Recent 50-yr trends of surface temperature and precipitation around the globe have been strongly influenced by the warming of the tropical oceans. The current generation of fully coupled climate models with prescribed radiative forcing changes generally fail to capture the geographical variations of these trends. On the other hand, even uncoupled atmospheric models without the prescribed radiative forcing changes, but with the observed SST changes specified only in the tropics, are more successful in this regard. The tropical oceanic warming pattern is poorly represented in the coupled simulations. This error needs to be reduced to increase confidence in regional climate change projections around the globe.

We provide important new evidence that a primary source of this error lies in the way an SST change in one tropical region affects and is affected by a change in another region in the models. Specifically, we estimated SST feedbacks among 8 geographically localized regions in the tropical Pacific, Indian, and Atlantic oceans from 3 observational datasets and 76 IPCC model simulations of the 20th century. The 8-dimensional SST feedback matrix in each case was obtained from a linear inverse model (LIM) constructed using observed or simulated monthly SST variations over the 100-yr period. The simulated feedback matrices were found to differ in several key respects from the observed matrices, and also from one another. For instance, the influence of the eastern Pacific ENSO region on other regions, and of the other regions on the ENSO region, was found to vary considerably from model to model. The varied representations of interactions with the Indo-Pacific Warm Pool region were also striking. Because the relatively short-term SST interactions investigated here occur primarily through the atmosphere, these results highlight the need to improve the representation of not only the local radiative-thermodynamic but also the remote atmospheric teleconnective feedbacks on tropical SSTs in climate models.