Thread by @mzelinka

We just published a paper on climate sensitivity in #CMIP6 models: agupubs.onlinelibrary.wiley.com/doi/abs/10.102…

Figured it was time for my first twitter thread!

We found that effective climate sensitivity (#ECS), when averaged across all #CMIP6 models* (orange), was larger than in #CMIP5 (blue) by about 0.6 deg C (about 1 deg F).

*that were available by the end of 2019

This is mostly (but not entirely) due to stronger positive feedbacks from changes in low-level clouds. That is, most models predict that as Earth warms, cloud properties change so as to allow more sunlight to be absorbed, causing additional heating & ultimately more warming.

This amplifying cloud feedback has strengthened in the latest models, particularly in middle and high latitude regions. There, fractional coverage and albedo of low-level clouds decrease more w/warming in #CMIP6 than in #CMIP5:

We expected this, to some extent. Models have been trying to fix a persistent error in how bright S. Ocean clouds are. One way to do so is to increase cloud liquid at T<0C, in better agreement w/obs. This has previously been shown to lead to higher #ECS: tinyurl.com/yzz5cpal

Of course improving this aspect could be unmasking some other model flaw that leads to cloud feedbacks that are too large. That remains to be established.

But it remains the case that an ever-growing body of evidence (from theory, hi-res modeling, and observations) points to positive overall cloud feedback, in agreement with nearly all climate models: nature.com/articles/nclim…

2 other interesting nuggets:

(1) #ECS values span 1.8 - 5.6 deg C across these 27 #CMIP6 models, a wider range than we previously saw across 28 #CMIP5 models (2.1 - 4.7 deg C)

(2) 10 models have #ECS exceeding 4.5 deg C, which is outside the #IPCC “likely” range of 1.5 - 4.5 deg C.

These 10 models (filled orange symbols) have an unprecedented combination of weak overall negative feedback parameter (weak radiative damping of planetary warming) AND moderately strong effective radiative forcing from 2xCO2 (ERF):

In contrast, if a #CMIP5 model had a weak negative feedback, it also tended to have weak forcing (top left corner of figure). This was also discussed in @Tim_AndrewsUK's recent paper: agupubs.onlinelibrary.wiley.com/doi/epdf/10.10…

Why care? Higher #ECS implies greater warming for a given amount of CO2 increase. It is closely tied to the severity of climate change, &, if higher, means it is harder to avoid crossing dangerous T thresholds, requiring larger emissions reductions, e.g.: preview.tinyurl.com/yg966v63

Does this mean our previous #ECS estimates were too low? Not necessarily. Climate models are but one of several tools we have for estimating Earth’s true climate sensitivity.

Other methods include inferring it from observations of past warming and ocean heat uptake (+ model estimates of radiative forcing) and from the paleoclimate record (e.g., the last ice age): agupubs.onlinelibrary.wiley.com/doi/full/10.10…

Many of these lines of evidence do not support the highest climate sensitivities at the upper range of the current models.

And several of the high #ECS models do a ...shall-we-say... less than perfect job capturing the global temperature evolution of the past few decades.

Not a death knell, but certainly not inspiring confidence in their sensitivity to CO2. Others probably have more to say on this @Knutti_ETH @kasia_tokarska @ClimateOfGavin @FemkeNijsse @coxypm @theFosterlab

Looking forward, the community will continue to thoroughly investigate #CMIP6 models to try and understand where these high sensitivities come from and interrogate them against observations, paleoclimate, etc.

They are a piece of evidence that must be weighed against other evidence in our ever-evolving understanding of #ECS.

Nerdy details: We diagnosed effective (not equilibrium) climate sensitivity via Gregory regression over 150-year abrupt-4xCO2 simulations; used radiative kernels to break down the feedbacks; used nerdier techniques including #APRP to further break down the cloud feedbacks

Big thanks to my awesome coauthors. Most of them aren't on twitter, but @D_McCoy_Atmos and @PauloCeppi are.