Global warming under old and new scenarios using IPCC climate sensitivity range estimates

Climate projections for the fourth assessment report1 (AR4) of the Intergovernmental Panel on Climate Change (IPCC) were based on scenarios from the Special Report on Emissions Scenarios2 (SRES) and simulations of the third phase of the Coupled Model Intercomparison Project3 (CMIP3). Since then, a new set of four scenarios (the representative concentration pathways or RCPs) was designed4. Climate projections in the IPCC fifth assessment report (AR5) will be based on the fifth phase of the Coupled Model Intercomparison Project5 (CMIP5), which incorporates the latest versions of climate models and focuses on RCPs. This implies that by AR5 both models and scenarios will have changed, making a comparison with earlier literature challenging. To facilitate this comparison, we provide probabilistic climate projections of both SRES scenarios and RCPs in a single consistent framework. These estimates are based on a model set-up that probabilistically takes into account the overall consensus understanding of climate sensitivity uncertainty, synthesizes the understanding of climate system and carbon-cycle behaviour, and is at the same time constrained by the observed historical warming. A thorough comparison of SRES scenarios and RCPs would ideally be based on results computed by the exact same set of models. Running the new RCPs with the full suite of CMIP3 atmosphere–ocean general circulation models (AOGCM) is unrealistic because many models are now obsolete or unmaintained, and also because the computational cost is prohibitive. The latter restriction also applies to running all SRES scenarios with present versions of AOGCMs. We therefore use a reduced-complexity carbon-cycle and climate model MAGICC (ref. 6) version 6 to compare SRES scenarios and RCPs. The MAGICC model closely emulates7 the global and annual mean behaviour of significantly more complex AOGCM and C4MIP carbon-cycle models. We use historical constraints and calculate probabilistic time-evolving temperature projections for both sets of scenarios (see Methods). We derive a climate sensitivity distribution starting from the overall consensus understanding of climate sensitivity uncertainties—and then re-sample the joint distribution of climate model parameters such that historically observed ocean’s surface and land’s air temperatures in both hemispheres8, as well as ocean heat uptake observations9, are matched. The resulting model set-up closely reflects the uncertainties in radiative forcing, carbon-cycle and climate sensitivity from the AR4 (seeMethods and ref. 10). Equilibrium climate sensitivity (ECS)—the change in global mean surface temperature at equilibrium following a doubling of atmospheric carbon dioxide (CO2) concentrations—remains a critical source of uncertainty in long-term temperature projections1. It is not a physical quantity that can be measured directly through observations, but can be estimated with different indirect methods