Equilibrium Climate Sensitivity

Abstract

Equilibrium climate sensitivity (ECS) is usually defined as the global mean equilibrium temperature response to a doubling CO2 concentration of its preindustrial level in the atmosphere [Hegerl et al., 2007; Meehl et al., 2007]. The pre-industrial concentration of CO2 is generally accepted to be 280×10 −6 and the doubled concentration was considered to be 560×10 accordingly in early studies. New studies tend to use 600×10 instead to stand for the doubled concentration of CO2 around 1900 AD. The initial values of ECS estimated by professionals [Charney et al., 1979], including those reported in IPCC FAR [Mitchell et al., 1990], SAR [Kattenberg et al., 1996] and TAR [Cubasch et al., 2001], were in the range of (3±1.5)◦C or 1.5–4.5◦C. Since IPCC TAR, great many efforts have been made to re-examine the ranges of ECS. As a result, IPCC AR4 [Meehl et al., 2007] offered a comprehensive discussions on this issue and proposed the range of ECS to be 2.0–4.5◦C, which was slightly narrower than previous estimations. Figure 1, derived from IPCC AR4, shows the probability density function of ECS from 10 studies, among which the first 7 results [Forster and Gregory, 2006; Frame et al., 2005; Knutti et al., 2002; Andronova and Schlesinger, 2001; Forest et al., 2002; 2006; Gregory et al., 2002] were from observation-constrained simulations and the other 3 results [Hegerl et al., 2006; von Schneider et al., 2006; Annan et al., 2005] came from Last Glacial Maximum(LGM) and millennia-scale proxy data constrained simulations. It is evident in Figure 1 that the spread of the probability density functions differs widely with the peaks of the curves ranging from around 0.2 to above 0.6 and the broadness of the curves also varies greatly while the mean values of ECS are at 2.0–3.5◦C, which are often greater than the most likely values of ECS (corresponding to the peaks of the probability curves). If should be noted that the uncertainty in estimated ECS mainly comes from the different sources of data that used to constrain the models. The observations with centennial length and LGM proxys were