Abstract
Climate sensitivity is the change in global-mean surface temperature required to restore radiative equilibrium in response to a CO2 doubling. It is the most widely used metric to quantify the susceptibility of climate to an externally forced change. When evaluating the effect of CO2 changes on Earth’s climate, it is widely assumed that variances in climate sensitivity arise from differences in radiative feedbacks. The observed interannual variation provides a useful constraint on long-term cloud feedback. Although the differences in radiative feedbacks are widely attributed to a time-dependence of feedbacks, our results indicate a temperature-dependence of radiative feedbacks is concealed in the time-dependence of feedbacks with roughly equal contribution as that from pattern effect. Most of the state-dependence comes from a temperature-dependence of shortwave cloud feedback. This study also suggests that the role of radiative forcing in determining climate sensitivity has been vastly underrated. Compared to the widely held assumption that instantaneous radiative forcing (IRF) from a given CO2 concentration perturbation is constant, our results show the IRF2×CO2 is not constant, but also depends on the climatological base-state, increasing by ~25% for every doubling of CO2, implying a proportionate increase in climate sensitivity. This base-state dependence also explains about half of the inter-model spread in the IRF2×CO2. A comprehensive evaluation of simulated radiative forcing from climate models, reveals a continuing increase in the simulated radiative forcing. The stronger IRF in the most recent generation of climate models resulting from a colder stratosphere causes higher radiative forcing. Combined with the finding that the underestimate of radiative forcing is the cause of the underestimated climate sensitivity by the most widely used method for estimating climate sensitivity, our results underscore the urgency for accurate radiative forcing diagnoses.