Abstract
Abstract. The Earth's equilibrium climate sensitivity (ECS) to a doubling of atmospheric CO2, along with the transient climate response (TCR) and greenhouse gas emissions pathways, determines the amount of future warming. Coupled climate models have in the past been important tools to estimate and understand ECS. ECS estimated from Coupled Model Intercomparison Project Phase 5 (CMIP5) models lies between 2.0 and 4.7 K (mean of 3.2 K), whereas in the latest CMIP6 the spread has increased to 1.8–5.5 K (mean of 3.7 K), with 5 out of 25 models exceeding 5 K. It is thus pertinent to understand the causes underlying this shift. Here we compare the CMIP5 and CMIP6 model ensembles and find a systematic shift between CMIP eras to be unexplained as a process of random sampling from modeled forcing and feedback distributions. Instead, shortwave feedbacks shift towards more positive values, in particular over the Southern Ocean, driving the shift towards larger ECS values in many of the models. These results suggest that changes in model treatment of mixed-phase cloud processes and changes to Antarctic sea ice representation are likely causes of the shift towards larger ECS. Somewhat surprisingly, CMIP6 models exhibit less historical warming than CMIP5 models, despite an increase in TCR between CMIP eras (mean TCR increased from 1.7 to 1.9 K). The evolution of the warming suggests, however, that several of the CMIP6 models apply too strong aerosol cooling, resulting in too weak mid-20th century warming compared to the instrumental record.
Highlights
The equilibrium climate sensitivity (ECS) is defined as the long-term globally averaged amount of surface temperature increase in response to a doubling of atmospheric carbon dioxide (CO2) relative to pre-industrial levels
A potential complication exists when applying such standard methods to compare mean ECS: statistical tests for independence of means such as t tests usually rely on an assumption of a Gaussian or approximately Gaussian underlying distribution and may not be appropriate for samples with skewed distributions, such as ECS (Roe and Baker, 2007). One might view such generational ensembles as small random samples taken from some generic modeling activities that are subject to noise. How likely is it that we obtain the CMIP6 ensemble mean ECS increase by chance? In other words, do the high CMIP6 climate sensitivities represent a statistically significant shift in an envisioned underlying probability distribution based on modeling, or are they encapsulated by the uncertainty of climate modeling? We address the question of statistical significance by assuming the underlying ECS distribution is well described by Eq (1)
The ECS and total feedback parameter values were computed with the Gregory method, and we found that both the ensemble mean ECS and the spread in ECS values has increased between Coupled Model Intercomparison Project Phase 5 (CMIP5) and CMIP6
Summary
The equilibrium climate sensitivity (ECS) is defined as the long-term globally averaged amount of surface temperature increase in response to a doubling of atmospheric carbon dioxide (CO2) relative to pre-industrial levels. Recent community efforts to improve on this stalemate on bounding ECS instead focuses entirely on basic process understanding, historical warming, and paleoclimate evidence (Stevens et al, 2016). This may be viewed as scientists abandoning climate models as evidence for ECS, but this is not true. Recent studies of several individual CMIP6 models, including CNRM-CM6-1 (Voldoire et al, 2019), CESM2 (Gettelman et al, 2019), E3SMv1 (Golaz et al, 2019), and HadGEM3-GC3.1 (Bodas-Salcedo et al, 2019; Andrews et al, accepted), each with an ECS of about 5 K or greater, have pointed to model parameterization changes that increased the positive shortwave cloud feedbacks or added aerosol–cloud interactions as having driven up their ECS values. The results allude to excessive aerosol cooling in early historical warming in a majority of the models
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
