On the origin of tropical Atlantic SST biases in coupled general circulation models

Coupled ocean-atmosphere general circulation models typically exhibit serious biases in the mean state of the tropical Atlantic. Most conspicuous among those is the reversal of the zonal SST gradient along the equator in the annual mean. While these biases have drawn considerable attention a comprehensive understanding of the problem remains elusive. In this study we analyze output from models of the Intergovernmental Panel on Climate Change (IPCC) as well as the Atmospheric Model Intercomparison Project (AMIP) with a focus on the climatological annual cycle in the Climate of the 20th Century experiments.

All the models examined feature a warm SST bias in the eastern equatorial Atlantic that typically peaks around July with maximum amplitudes in the models ranging from 2 to 6 K. This warm bias is preceded by a westerly bias in equatorial surface winds during spring that deepens the thermocline. In the following months, as the cross-equatorial surface winds intensify due to the developing West African monsoon, the unrealistically deep thermocline in the models reduces the cooling effect from equatorial upwelling and thus contributes to the warm bias in the simulations. The AMIP counterparts of the IPCC models exhibit a similar albeit weaker bias in surface winds. This indicates that the problem originates in the atmospheric components and is amplified by air-sea interaction in the coupled system. Thus a significant potential for improvement of coupled models in the tropical Atlantic might be achieved by a detailed analysis of the performance of their uncoupled atmospheric components in spring.