Comparison of general circulation model outputs and ensemble assessment of climate change using a Bayesian approach

A number of general circulation models (GCMs) have been developed to project future global climate change. Unfortunately, projected results are different and it is not known which set of GCM outputs are more creditable than the others. The objective of this work is to present a Bayesian approach to compare GCM outputs and make an ensemble assessment of climate change. This method is applied to Cannonsville Reservoir watershed, New York, USA. The GCM outputs under the 20C3M scenario for a historical time period of 1981-2000 are used to calculate posterior probabilities, and the outputs under the scenarios (A1B, A2 and B1) for the future time period of 2084-2100 are then processed using the Bayesian modeling averaging (BMA) which is a statistical procedure that infers a consensus prediction by weighing individual predictions based on the posterior probabilities, with the better performing predictions receiving higher weights. The obtained results reveal that the posterior probabilities are slightly different for four variables including average, maximum and minimum temperatures, and shortwave radiation, implying that the GCM outputs are qualitatively different for these four variables, but the distributions of posterior probabilities are flat for precipitation and wind speed, suggesting that the GCM outputs are qualitatively similar for these two variables. The results also show that no one set of GCM data are the best for all six meteorological variables. Furthermore, the results indicate that the projected changes are for regional warming, but the changes in precipitation, wind speed, and shortwave radiation depend on the emission scenarios and seasons. The application of the method demonstrates that the Bayesian approach is useful for the comparison of GCM outputs and making ensemble assessments of climate change.