Estimating climate sensitivity from an ensemble of GCM configurations optimized to outgoing TOA radiation

Abstract:

Multiple parameters controlling subgrid-scale physical processes in the version 3 of the Hadley Centre Atmosphere Model (HadAM3) were optimized to global mean outgoing longwave radiation and reflected shortwave radiation estimates from CERES and ERBE satellite observations. An ensemble of sucessfully-optimized model configurations were identified. Equilibrium climate sensitivity of each model configuration was estimated using a statistical emulator, which had been trained on a 14,001-member ensemble of HadAM3 slab-ocean experiment simulated over the distributed computing platform of climateprediction.net. To produce an uncertainty estimate for the equilibrium climate sensitivity, likelihood for each model configuration was calculated by combining model-observational discrepancy arising from satellite measurements, observational radiation imbalance, total solar irradiance, radiative forcing, natural aerosol, internal climate variability, and the prescribed sea-surface temperature and that arising from parameters that were not varied. Combining different prior probabilities for model configurations with the likelihood for each configuration and taking account of uncertainty in the emulated climate sensitivity gives, for the HadAM3 model, a 2.5%-97.5% range for climate sensitivity of 2.7-4.2 K if the CERES observations are correct. If the ERBE observations are correct, then they suggest a larger range, for HadAM3, of 2.7-5.6 K. Amplifying the CERES observational error estimate by a factor of 20 brings CERES and ERBE estimates into agreement. In this case the climate sensitivity range is 2.7-5.4 K. The results rule out, at the 2.5% level for HadAM3 and several different prior assumptions, climate sensitivity greater than 5.6 K.