Quantification of Global Warming: Critical Evaluation of GCMs

We examine decadal time scale predictability of global climate using two approaches: calculations conducted using an empirical model of global climate (EM-GC) developed by our research group (Canty et al., ACPD, 2013), and analysis of archived output from the Coupled Model Intercomparison Project Phase 5 (CMIP5) general circulation models (GCMs). Our focus is on developing a probabilistic forecast of the rise in global mean surface temperature (GMST) likely to occur over the next several decades.  We begin by evaluating attributable anthropogenic warming rate (AAWR) over 1979 to 2010, equilibrium climate sensitivity, and transient response of climate to cumulative carbon emission (TCRE), using our EM-GC and from GCMs. We show AAWR inferred from the climate record over the past 32 years using our EM-GC is about a factor of two less than the GCM-based multi-model-mean for AAWR. The CMIP5 GCMs tend to warm about a factor of two too quickly, on average, most likely because most models represent climate feedback in a manner that too strongly amplifies the radiative forcing of climate due to greenhouse gases. We extend these comparisons to ECS and TCRE, taking into consideration the impact of uncertainty in ocean heat content on the analysis.  Finally, we compare probabilistic projections of the rise in GMST, out to 2060, from the GCMs to calculations conducted using our EM-GC approach. The projections are used to assess the reduction in the emissions of GHGs that will be needed to achieve the goal of the Paris Climate Agreement (target of 1.5 °C warming; upper limit of 2.0 °C warming) based on the CMIP5 GCMs and our empirical model of global climate.