An empirical model of global climate – Part 2: Implications for future temperature

Abstract. IPCC (2007) has shown that atmosphere-ocean general circulation models (GCMs) from various research centers simulate the rise in global mean surface temperature over the past century rather well, yet provide divergent estimates of temperature for the upcoming decades. We use an empirical model of global climate based on a multiple linear regression (MLR) analysis of the past global surface temperature anomalies (ΔT) to explore why GCMs might provide divergent estimates of future temperature. Our focus is on the interplay of three factors: net anthropogenic aerosol radiative forcing (NAA RF), climate feedback (water vapor, clouds, surface albedo) in response to greenhouse gas radiative forcing (GHG RF), and ocean heat export (OHE). Our model is predicated on two key assumptions: whatever climate feedback is needed to account for past temperature rise will persist into the future and whatever fraction of anthropogenic RF (GHG RF + NAA RF) is exported to the oceans to match the observed rise in ocean heat content will also persist. Even with these assumptions, modeled future ΔT mimics the behavior of GCMs because the ~110 record of global surface temperature can not distinguish between two possibilities. If anthropogenic aerosols presently exert small cooling on global climate, feedback must be weak and the future rise in global average surface temperature in 2053, the time CO₂ is projected to double according to RCP 8.5, could be moderate. If aerosols presently exert large cooling of global climate, feedback must be large and future ΔT when CO₂ doubles could be substantial. Reduced uncertainty for climate projection requires observationally based constraints that can narrow the uncertainties that presently exist for net anthropogenic aerosol radiative forcing as well as the totality of feedbacks that occur in response to a GHG RF perturbation. GCMs are often compared by evaluating the equilibrium response to a doubling of CO₂, termed climate sensitivity. In our model framework, ΔT at the time CO₂ doubles is nearly independent of OHE, because climate feedback must be adjusted to properly simulate observed temperature. Our simulations show that if a small fraction of anthropogenic RF is exported to the ocean, equilibrium climate sensitivity closely represents the modeled ΔT at the time CO₂ doubles. Conversely, if this fraction is large, ΔT when CO₂ doubles is much less than the equilibrium climate sensitivity (i.e. the model is now far from equilibrium). Similar behavior likely occurs within GCMs. We therefore suggest the dependence of climate sensitivity on OHE be factored into analyses that use this metric to compare and evaluate GCMs.