The Equilibrium Climate Sensitivity (ECS) is an important metric in climate models in evaluating the global climate state, defined as the warming for a doubling of atmospheric CO2 concentration relative to preindustrial climate. Most studies select the 150 years simulation (whole stage) of the climate model, via the linear relationship between the global average surface air temperature (SAT) and the net radiation flux at the top of the atmosphere. The model climate behaves nonlinearly due to the mean-state biases and the inter-model spread in feedback evolution. We divide 150 years into fast stage (1-20 yrs) and slow stage (21-150 yrs) .The choice of year 20 is based on the maximum standard deviation of the 150-year global mean SAT in the abrupt4×CO2 experiment relative to the piControl run of multi-models that participated in Coupled Model Intercomparison Project phase 5 and phase 6 (CMIP5 and CMIP6). The standard deviation of CMIP6 is larger than CMIP5 models, due to its warming curve branch. Some models’SAT is higher (about 1K) in CMIP6 than in CMIP5, which is related to the changes of global mean precipitation and the temperature changes in the Arctic via inter-model spread. The study shows that prominent model spread in ECS results from uncertainty in different feedbacks: the cloud shortwave feedback is closely related to the simulation of the temperature and convective precipitation; the albedo feedback is closely related to the simulation of Polar region causing the higher multi-model mean ECS in CMIP6 than that in CMIP5.

ECS,climate model,feedback