Estimate of equilibrium climate sensitivity and uncertainties by using observations and CMIP6 global coupled climate models

Energy budget estimates of the effective climate sensitivity (effCS) are derived based on estimates of the historical forcing and of observations of the sea surface temperature variations and the ocean heat uptake. Recent revisions to Greenhouse gas forcing and aerosol forcing estimates are included and the data is extended to 2018. We consider two different approaches to derive the effCS from the energy budget: 1) a difference of the energy budget between the recent period 2005-2018 and a base period 1861-1880 (following Sherwood 2020) and 2)  a regression of the differential form of energy budget over the period 1955-2017 (following Gregory et al. 2020). These estimates of the effCS over the historical period are representative of the climate feedback experienced by the climate during the historical period. When accounting for the uncertainty in the forcing, the surface temperature and the ocean heat uptake estimates plus the uncertainty due to the internal variability we find a range of effCS of [1.0;9.7] (at the 95%CL) with a median of 2.0 K with approach (1) and [1.2;2.7] with a median of 1.7 K with approch (2). We find that the lower and the upper tail of the distribution in effCS arise dominantly from the uncertainty in the historical forcing, particularly for the regression method, and at a lower extent for the difference method. This is consistent with previous studies (e.g. Lewis and Curry 2018 and Sherwood et al. 2020).

Using the same approach based on historical observations but accounting for the pattern effect and the temperature dependence of the feedback estimated with climate model simulations, we derive new estimates of the effECS that should encompass the equilibrium climate sensitivity (assuming that climate model simulate properly the pattern effect and the temperature dependence of feedback). We find that adding the pattern effect and the temperature dependence of the feedbacks shifts upwards the median of the effECS and increases significantly the uncertainty range. For the difference method, the median is now 2.5 K and the uncertainty range [1.1;17.2]. For the regression method the median is now 2.0 K and the uncertainty range is [1.2;4.7] K ((5-95%). On the overall, we find that the regression method performs better to constraint the equilibrium climate sensitivity and that the major source of uncertainties comes from the differences in the simulation of the pattern effect among climate models rather than the uncertainties on the historical forcing.