I’ve regularly written about the Equilibrium Climate Sensitivity (ECS), in particular estimates by Nic Lewis and why they are probably a bit too low. For balance I should probably now mention a new paper called Observational constraints on mixed-phase clouds imply higher climate sensitivity by Tan, Storelvmo & Zelinka.
You can probably get the key result from the description, which finishes with
Tan et al. used satellite observations to constrain the radiative impact of mixed phase clouds. They conclude that ECS could be between 5.0° and 5.3°C—higher than suggested by most global climate models.
If you want a rather alarmist take on it, you can read the Guardian’s article. There is a more measured response by Chris Mooney.
The key point seems to be that they considered clouds feedbacks and suggest that cloud feedbacks might be much more positive than we currently think. They
show that the ECS can be up to 1.3°C higher in simulations where mixed-phase clouds consisting of ice crystals and supercooled liquid droplets are constrained by global satellite observations.
So, what’s the basic issue. As James Annan points out, an ECS of 5.3C is pretty hard to reconcile with our understanding of past climate changes. Bear in mind that the ECS is the equilibrium response considering fast feedbacks only, and so the equilibrium system sensitivity (ESS) – which includes slow feedbacks – could be higher. However, this study is considering clouds, which are fast feedbacks.
Also, even if energy balance estimates might be on the low side, they’re still a reasonable way to get a ballpark figure. If they’re wrong by a factor of two or more, it either means that natural variability is masking a lot of attributable warming, or the response is highly non-linear. Both are possible, but it seems unlikely that the difference can be quite this large.
A number of fairly high-profile scientists also seem rather skeptical; Kevin Trenberth in the Guardian article and Gavin Schmidt on Twitter and in Chris Mooney’s article. It’s clear that clouds are one of the major uncertainties with respect to climate sensitivity. However, just as some of the energy balance models seem to produce results that appear a bit too low, this study seems to be suggesting results that are somewhat on the high side. Also, promoting it as they have is – I would argue – somewhat sub-optimal. It’s one thing to present controversial results, but claiming that
global climate models have significantly underestimated how much the Earth’s surface temperature will rise if greenhouse gas emissions continue to increase as expected
is much stronger than is justified.
...and Then There's Physics 
As well as nonlinear response and internal variability, it could be something to do with the efficacy of different forcings.
I haven’t read the paper in detail yet but I’m looking forwards to it. For those who do, a common issue with these emergent constraints is that they might miss other important factors that go the opposite way. For example, models might get cloud phase wrong in a way that means they underestimate ECS, but they might also miscalculate the organisation of convection in a warmer climate which leads to overestimated ECS. Keep this in mind when reading the paper: the strength of the conclusions will depend on how this is addressed.
ATTP:
Alos see:
Climate Models May Overstate Clouds’ Cooling Power, Research Says by John Schwartz, New York Times, Apr 7, 2016
Schwartz quotes Gavin Schmidt and Kevin Trenberth as well as auhors of the paper.
There are a variety of studies into various cloud processes finding those that match the detailed process have a higher climate sensitivity. And different process are involved – from memory something about the ITCZ, and something about the height at which certain clouds form. Do we add up all the increases in climate sensitivity for each different process and get really really high sensitivity? Are all the processes connected so that fixing one fixed all the others? (eg if we fix the super cooled liquid, we also make clouds form correctly over the ITCZ, and at the correct height). Or are there lots of different processes, some which push sensitivity up, and some down?
And when the models are corrected for this process, what then? Current models have a good match to temperature history. Will the corrected model still have the same match to temperature history? Is constraining models according to which models match observations of global temperature over the last century a less valid technique then constraining models according to which models best match observations of super cooled liquid in high cloud? Is it possible that the corrected model would predict the same rise for the last 100 years as the lower sensitivity models, yet predict a higher rise over the next 100 years?
With 30+ years of climate studies, I think it’s prudent not to jump to conclusions based on a single novel paper about something as complex as clouds. With time it will take its appropriate place in the literature and maybe in climate models, but I think much of the early hype is definitely overblown.
MarkR and Michael Hauber twigged to the same issue I have. Tan et al. only looks at one GCM, NCAR CAM5.1, which they say is “among the GCMs that severely underestimate SLFs”. So the obvious questions are: (1) what if they looked at a GCM that doesn’t underestimate SLFs as much, and (2) how does the overall skill of their constrained CAM5.1 compare to the unconstrained CAM5.1?
There seems to be an assumption that a higher ECS for a constrained CAM5.1 implies a higher ECS for CAM5.1 with a more correct SLF model.
MarkR,
Yes, a good point about efficacy. Also, you’re right, matching one emergent constraint doesn’t mean you’ve matched them all.
JH,
Thanks.
Michael,
Yes, I’m not sure how well these models match our observed warming if they correct for these cloud processes. That would seem like a pretty standard test, and I don’t think they did it.
Dan Riley,
If I remember correctly, underestimating supercooled liquid in clouds is pretty much universal among the CMIP5 GCMs. Of course, some are more severe than others – Greg Cesana has done some nice work testing cloud phase in models with CALIPSO and linking this to the different microphysics schemes they use.
My current gut feeling based on putting together different papers, conference presentations etc is that we’re going to eventually find two offsetting effects. I think Tan’s paper is just one piece showing that doing mixed phase clouds properly will lead to a higher climate sensitivity. Other work has shown that changes in cloud cover (perhaps related to convective aggregation) might go the other way. See Mauritsen and Stevens:
http://dx.doi.org/10.1038/ngeo2414
or Kerry Emanuel gave a talk on convective aggregation at the last AGU meeting:
https://agu.confex.com/agu/fm15/webprogram/Paper62237.html
This is speculation, but it’s a way that two of the current big issues might both cause big changes to our models without really affecting the ECS if they combine the right way.
We also have semi-independent estimates that are also broadly consistent with the model estimates.
What James Annan said. I’ll stick with palaeo evidence and about 3C per doubling.
IMO there is an obvious conclusion from two facts:
1. The paper shows that the “supercooled liquid fraction” (SLF) in clouds are grossly underestimeated in GCM: “Global satellite observations of cloud thermodynamic phases have enabled us to show that unrealistically low SLFs common to a multitude of GCMs lead to a cloud-phase feedback that is too negative. This has important ramifications for ECS estimates. Should the low-SLF bias be eliminated in GCMs, the most likely range of ECS should shift to higher values.”
2. The consideration of the observed SLF in otherwise unchanged GCM lead to unrealistic high values for ECS. ( See James Annan)
The conclusion is: Something is very wrong with GCM?
Perhaps, but is any of this objectively Bayesian?
BBD,
As am I. Something I’ve been pondering today: suppose Tam et al. are more correct than not that most state of the art GCMs underestimate cloud feedbacks, but more wrong than not about what that implies for ECS. frankclimate concludes by query, Something is very wrong with GCM?
I ask it more in the form of what else is also wrong with GCMs that causes them to get the “right” answer? It’s beyond my ability to figure out except to say that it seems plausible that other parameters have been unrealistically tuned to compensate. Perhaps one of the literati here have more concrete ideas?
ps, I belatedly see that MarkR makes a similar speculation.
It’s always worth remembering that the contrarian emphasis on ‘the models’ is just a rhetorical trick. The models aren’t the primary source of knowledge. Model issues (if such there really are) don’t affect the actual physics of climate or our understanding of the most likely value for ECS.
That said, there is a real possibility that the issues [lie] more with the study than with the majority of models. Only the modelling community is qualified to speak to that, and I await any expert response with great interest.
@frankclimate
You should probably note Del Genio’s comment, taken from Chris Mooney’s article:
In other words, Del Genio considers that Tan et al have interpreted the satellite observations, and that interpretation might possibly be wrong
frankclimate, Given the uncertainties about clouds–which this paper does little to resolve definitively–I think it would be a little rash to throw out the climate models, which have been fairly well validated, albeit to varying extents.
This is interesting initial research that may (hopefully) open up more refinements in the future. It is interesting in its own right, independent of its implications about ECS.
Sorry, in my post above, I managed to mangle my HTML, The quote from the Mooney article ends with the words “is the same”. The sentence beginning “In other words” is my own. Apologies … [Mod: fixed]
Frank,
As others have already indicated, this is probably wrong. GCMs are scientific tools that allow us to probe how our climate will respond to various changes. They’re not perfect and will continue to improve and change. That, however, does not mean that there is something very wrong with them.
Pingback: Pole position - Ocasapiens - Blog - Repubblica.it
BBD,
Sometimes, it’s enough for the word “model” to appear in literature, even in reference to, say, a statistical model. Providence forfend a paper with the temerity to so much as hint at the words “assumed”, “projected”, “interpolated”, etc. Immediate disqualification.
It’s models all the way down, constrained to gobs more empirical evidence. Somehow, magically, they know the real truth without accepting most of it.
/rant
They have issues, they’re models. I gots specifics, and I’m sure you do too. However, I understand what you mean, so I won’t. 🙂
/pedant mode
Or what Anders said in his response to Frank.
Hear hear. Seconded.
Brandon G
Yes, sloppy writing on my part. I should have said ‘if Tan et al. is correct’ etc. Phil’s comment above suggests that some experts are already questioning that.
Pingback: Storelvmo et al. by proxy – Stoat