Item

Abstract

The North Atlantic sea surface temperature (SST) cold bias is a remarkable feature of general circulation models (GCMs). This bias is a primary concern in climate science because they directly affect the skill of predictions and the confidence in projections on the North Hemisphere climate. Here the characterization and causes of the cold bias are investigated by combining Atmospheric, Ocean, and Coupled Model Inter-comparison Project (AMIP6/OMIP6/CMIP6) simulations with observations. It is found that the cold North Atlantic SST bias in CMIP6 is primarily caused by weak heat transport in ocean models induced by weak Gulf Stream currents. The weak atmospheric fluctuations induced shallow mixed layer depth also plays a role. These biases are much reduced by increasing the atmospheric and oceanic model resolution to around 1º and 0.25º in HadGEM3-GC31-MM, respectively. The radiation heat flux bias in AMIP6 and CMIP6 model are small. The large turbulent heat flux bias linked to SST bias in CMIP6 models is not shown in AMIP6 models, which implies that the SST bias dominates the turbulent heat flux bias in coupled models. Our results suggest that to reduce the North Atlantic SST bias in the coupled models, representing unresolved oceanic and atmospheric processes is crucial.