Cristian Proistosescu, David S. Battisti, Kyle C. Armour, Gerard H Roe

Improved projections of local temperature change over the 21st century are essential for evaluating impacts and setting policy targets (Lehner and Stocker 2015, Senevirante et al 2016). Uncertainty in these projections is due to two approximately equal factors: uncertainty in greenhouse gas emissions, and uncertainty in the response of climate to those emissions (Hawkins and Sutton 2009). For the latter, it is well known that climate models all have a similar global pattern of warming (Knutti and Sedlacek 2013). Here, we show that differences among projections of warming also share a common pattern of variability. Specifically, the leading empirical orthogonal functon(EOF) of inter-model differences strongly resembles the ensemble-mean response itself. This pattern explains 60% of the total variance in projected regional warming, with higher fractions of variance explained over tropics and midlatitudes. When projected onto the model residuals, it is strongly correlated (r2 = 0.9) with Equilibrium Climate Sensitivity (ECS), and more-weakly correlated with the Transient Climate Response (TCR, r2 = 0.6). We show how this strong correlation between equilibrium global warming and transient regional warming uncertainty arises due to the large scales over which the atmosphere mixes energetic perturbations. The dominant source of variance in both ECS and 21st-century warming are low-latitude feedbacks, whereas TCR is more sensitive to uncertainties in CO2 radiative forcing and ocean heat uptake. The results imply that throughout the tropics and midlatitudes, and especially over land, most of the uncertainty in local temperature projections.