On the relationship between global mean temperature in historical runs of Earth system models and equilibrium climate sensitivity

Abstract

Equilibrium climate sensitivity (ECS) measures the long-term global surface temperature response due to a doubling of CO2 in the atmosphere. Estimates of ECS is not well constrained, 1.5-4.5C [Pachauri et al., 2014], and there is a wide spread between different Earth system models (ESMs). Recently it has been suggested that ECS can be constrained using an observed relationship between the statistical properties of the unforced temperature fluctuations extracted from historical runs of energy system models (ESMs), and the Gregory estimates of ECS in these models [Cox et al., 2018]. In this thesis I derive general fluctuation-response relations for linear stochastic climate models and investigate the claimed relation over an ensemble of ESMs. My findings are consistent with the existence of a fluctuation-response relation, but uncertainties are large, and I find it unlikely that they can be used to constrain ECS. My conclusion is that the time period 1850-present is too short for estimation of ECS, and that we ultimately have to rely on longer reconstructed temperature time series or satellite measurements of Earth’s energy imbalance.