Mean-state dependence of future tropical-Atlantic sector climate projections

Mean state and internal variability in the tropics are crucially linked to air-sea interactions. State-of-the-art climate models exhibit long-standing problems not only in simulating tropical mean climate, such as too cold sea surface temperature (SST) over the central tropical Pacific and too warm SST over the eastern tropical Pacific and Atlantic, but also with respect to seasonal and longer variability. These biases question the credibility of future climate projections with the models, and it has not been shown to date whether or how such SST biases affect the projections. Here we focus on the tropical Atlantic (TA) and investigate how the mean state influences climate projections over the region.

We use two versions of the Kiel Climate Model (KCM) in global warming simulations, in which only atmosphere model resolution differs: one version carries ECHAM5 with a horizontal resolution of T42 (~2.8°) and 31 vertical levels, and the other ECHAM5 with a horizontal resolution of T255 (~0.47°) and 62 levels. Although only the atmospheric resolutions differ, the two KCM versions exhibit very different mean states over the tropical TA, with the higher-resolution version, among others, featuring much reduced warm SST bias over the eastern basin.

The response to increasing atmospheric carbon dioxide levels is found to be sensitive to the mean state. The model employing high atmospheric resolution and featuring a small SST bias projects an eastward-amplified SST warming over the TA, consistent with the pattern of interannual SST variability simulated under present-day conditions and in line with the observed SST trends since the mid-20^th century. The model employing low-resolution and exhibiting a large SST bias projects more uniform SST change. Atmospheric changes also vastly differ among the two model versions.

Analysis of models participating in the Coupled Model Intercomparison Project Phase 5 (CMIP5) support the KCM’s results: models with small SST bias project stronger warming over the eastern TA, while models with large SST bias either project uniform warming across the equator or largest warming in the west. This study suggests that reducing model bias may enhance global warming projections over the TA sector.