A significant portion of the interannual to decadal variations of seasonal rainfall over Africa is driven by anomalous global sea surface temperature (SST) patterns. Hence, if this rainfall variability is to be reliably predicted, and its mechanisms well understood, then it is crucial that global coupled ocean-atmosphere models can adequately simulate these SST anomalies and their teleconnections to Africa. This is assessed using control data from models in the CMIP3 archive and from recent Met Office Hadley Centre models. This will later be extended with data from the CMIP5 archive.

By comparing observed and simulated SST-Rainfall teleconnections, for a number of large homogeneous regions of Sub-Saharan Africa, and within an appropriate statistical framework, a number of findings are revealed. First, some teleconnections are clearly easier to model than others, in particular those from the Indian Ocean to East Africa and Southwest Africa (during the OND and DJF seasons, respectively). Whereas other teleconnections appear to be harder to represent well, with a majority of models providing poor links between the Indian Ocean and the Sahel (JAS), ENSO and East Africa (OND), and (perhaps surprisingly) the Equatorial Atlantic and the southern coast of West Africa (JAS). Second, some models consistently perform better than others, but even these are nevertheless unable to properly simulate all the investigated teleconnections, and so require further improvement. Third, by analyzing atmosphere-only simulations, forced by observed historical SSTs, it is found that errors in the coupled model teleconnections arise both from errors in their SST climatology and variability and from errors in their representation of the atmospheric component of the teleconnections to Africa. Further analysis also examines the roles of model resolution, of model errors in the SST variance and model errors in rainfall climatology, and whether these play an obvious role in determining the quality of its teleconnections.