The Influence of CMIP5 Tropical SST Biases on Simulations of Regional Climate

This paper reports on an investigation of the influence of climatological sea surface temperature (SST) biases in CMIP5 models on the simulation of regional climates around the globe, with emphasis on SST biases in the tropical oceans. We first identify, and quantify, common biases in climatological SSTs across a set of 31 CMIP5 models as a function of month through the annual cycle. These biases are typically associated with major upwelling regions, both coastal and equatorial and locally can be of substantial magnitude (e.g., as compared to seasonal/interannual variations in SST). We then impose these monthly, climatological model biases on top of observed SSTs for the period 1979-2005 and use this combined SST field to force two atmospheric general circulation models (AGCMs). A set of 16 ensemble members covering the period 1979-2005 are generated for each model using slightly different atmospheric initial conditions and seasonal climatologies of precipitation, winds and other fields from these "bias" runs are then compared with control runs for the respective models. For precipitation, the results reveal that major differences between the bias and control runs generally occur over ocean areas, although there are important influences in several land locations that will be discussed. In addition, the model SST biases result in atmospheric circulation changes that include spurious changes in vertical wind shear across the tropical North Atlantic ocean in the main development region for hurricanes. The anomalous wind shear resulting from the SST bias is seasonally varying, but in some months of the hurricane season are of the same magnitude as that associated with ENSO. Some implications for climate change projections will also be discussed.