Extending CMIP5 projections of global mean temperature change and sea level rise due to thermal expansion using a physically-based emulator

We present a physically-based emulator approach to extending 21st century CMIP5 model simulations of global mean surface temperature (GMST) and global thermal expansion (TE) to 2300. A two-layer energy balance model that has been tuned to emulate the CO2 response of individual CMIP5 models is combined with model-specific radiative forcings to generate an emulated ensemble to 2300 for RCP2.6, RCP4.5 and RCP8.5. Errors in the emulated time series are quantified using a subset of CMIP5 models with data available to 2300 and factored into the ensemble uncertainty. The resulting projections show good agreement with 21st century ensemble projections reported in IPCC AR5 and also compare favourably with individual CMIP5 model simulations post-2100. There is a tendency for the two-layer model simulations to overestimate both GMST rise and TE under RCP2.6, which is suggestive of a systematic error in the applied radiative forcings. Overall, the framework shows promise as a basis for extending process-based projections of global sea level rise beyond the 21st century time horizon that typifies CMIP5 simulations. The results also serve to illustrate the differing responses of GMST and Earth's energy imbalance (EEI) to reductions in greenhouse gas emissions. GMST responds relatively quickly to changes in emissions, leading to a negative trend post-2100 for RCP2.6, although temperature remains substantially elevated compared to present day at 2300. In contrast, EEI remains positive under all RCPs, and results in ongoing sea level rise from TE.