We had a panel discussion about the high ECS suggested by some CMIP6 models.
Thorsten Mauritsen (Stockholm Univ.): We can rule out ECS > 5 K
Kyle Armour (Univ. of Washington): We can't rule out ECS > 5 K
Each made a short presentation and then we had a discussion.
Text chat:
11:09:03 From Schmittner, Andreas : What are the different color symbols?
11:09:34 From stephenklein : Blue is satellite, orange is CMIP5/CMIP6 models
11:26:49 From Maria Hakuba : When I look at the histogram it seems a bimodal pdf is a better fit. Please comment.
11:29:06 From Robert Wood : Kyle highlights the useful information in the paleoclimate simulations. What do we know about the aerosol forcing in the LGM?
11:29:20 From Gavin Schmidt : To credibly argue that the high ECS CMIP6 models are not inconsistent with paleo, you need to show that they have a strong asymmetry btw cooling and warming. I’m not aware of any such evidence.
11:31:14 From Schmittner, Andreas : Robert: more dust in the LGM caused more aerosol forcing, but it is quite uncertain. I’m not sure, but my guess would be 0-1 W/m2.
11:32:41 From jkay : @Gavin - How asymmetric would the warming and cooling have to be? I have seen evidence for more 2xCO2 global warming than 0.5xCO2 global cooling in experiments we are doing with CESM1.
11:33:13 From Gavin Schmidt : CMIP6 models don’t include anomalous freshwater from Antarctica - links to S Ocean cooling (Rye et al, 2020)
11:34:45 From Jonah Bloch-Johnson (he/him) : @Gavin we also have a paper out recently finding that for 9 of 10 CMIP6 models, ECS increases with CO2 concentration (abrupt0.5x, 2x, 4x)
11:35:59 From Gavin Schmidt : Asymmetry needs to be ~50% (see the Zhu results for LGM)
11:36:43 From Isaac Held : for Kyle,
11:37:34 From Isaac Held : sorry, hit send unintentionally
11:37:52 From Maria Rugenstein : What about feedback temperature dependence? Bloch-Johnson et al. 2020 showed that they kick in around 3-4K warming (in the GCMs, independent of the pattern effect). They could dramatically increase ECS, independent of the historically observed warming. Literature on estimating this in the paleo record are scarce.
11:39:50 From Peter Watson : Do we need to worry about high ECS if actual warming seems to depend on cumulative emissions and TCRE i.e. warming doesn't reach the value implied by ECS in practice?
11:42:21 From Martin Renoult : @Gavin and @jkay: I think only CESM2 has performed both Pliocene and LGM for CMIP6 / PMIP4. IPSLCM5A performed both at CMIP5 / PMIP3 and seemed to have a strong LGM cooling, but average Pliocene warming based on my numbers
11:43:26 From Gavin Schmidt : Not really! If a model doesn’t include dust changes and it’s still a factor of two too cold?
11:43:41 From Benjamin Sanderson : @peter - yes, but if the real world had large slow timescale feedbacks related to deep ocean warming - then it could allow for hysteresis in the TCRE relationship. This would effectively give you a larger Zero Emissions Commitment
11:44:22 From Jiang Zhu : To clarify, we didn’t impose aerosol/vegetation forcing in the CESM2 LGM simulation; they may make the simulated LGM even colder.
11:44:49 From Kyle Armour : Interesting, thanks Jiang!
11:45:11 From Robert Wood : I would be amazed if ECS is not temperature dependent given the Clausius-Clapeyron equation.
11:45:21 From Greg C : Kyle, you mentioned that models may all miss a process that make them biased low in terms of ECS . But the opposite could be true as well. What if the future SST pattern is different from what current GCMs predict and actually weakens the cloud feedback. Or what if all the models now underestimate the mixed-phase cloud feedback, which contributes to decreasing ECS
11:45:29 From Peter Watson : @ben interesting point re ZEC - so warming would resume some long time after emissions cease? Do you know of any studies that have looked at that?
11:45:37 From Andrew Williams (he/him) : I think recent work by Seeley+Jeevanjee has made a pretty convincing case for temperature dependent ECS
11:46:06 From Andrew Williams (he/him) : (and Koll+Cronin, re closing of the water vapour window)
11:46:25 From Schmittner, Andreas : It is clear that the ECS is temperature dependent. If you imagine going to cold temperatures, at some point the ice-albedo feedback takes over and you get a transition to Snowball Earth, so essential an infinite ECS.
11:47:11 From Gavin Schmidt : The issue is not the existence of temperature or climate state dependence, but how large it is.
11:48:30 From Benjamin Sanderson : @peter - yeah - the toy model relationship between ZEC and slow timescale feedbacks was the main point in this paper https://esd.copernicus.org/articles/11/563/2020/
11:48:36 From Peter Caldwell : Can’t we approximate the Sherwood+Webb (2020) 95% bound as being entirely set by paleo since Bayesian inference cuts off the posterior PDF whenever any single line of evidence says something is impossible. Given this, are we *really* sure that the Sherwood+Webb (2020) paleo bounds are correct?
11:48:54 From jkay : Thanks @Gavin - I totally agree. The question of state dependence has been in the literature for a long time. A 50% asymmetry between 2xCO2 warming and 0.5x cooling is more than I have seen…
11:49:14 From Gavin Schmidt : @jkay exactly
11:49:21 From Jonah Bloch-Johnson (he/him) : @gavin most models have much less increase than 50% between abrupt2x and 4x, but temperature dependence can greatly change the sensitivity on that order depending on how sensitive you are to start with. The coupled runs we used for CESM2 had a ~50% estimated change between 2x and 4x, but slab ocean runs of CESM2 had much less
11:50:12 From Peter Watson : @ben thanks, will look at that
11:50:26 From Gavin Schmidt : @Jonah thanks.
11:50:30 From Kyle Armour : @Greg C: yes, that’s definitely a possibility that the models all miss a process or share a bias in the other direction! More research is needed! We are investigating just the historical pattern effect piece of it, which seems to suggest a low bias in ECS estimates. Other things may bias it high, though I’m not aware of work on that.
11:50:43 From Jonah Bloch-Johnson (he/him) : @gavin (For CESM2, we have ECS from abrupt2x ~3.5K, ECS from abrupt4x ~6.3K)
11:51:31 From Isaac Held : We have to be careful what we wish for — a low sensitivity world with large Pacific pattern effect will presumably have much more dramatic precipitation changes in the tropics and stronger changes in stationary waves propagating into midlatitudes
11:58:10 From stephenklein : @Ben: The WCRP assessment places 1/3 weight on process evidence (based on inter annual variability), 1/3 weight on the historical change, and 1/3 weight on paleo changes — that’s arbitrary weighting of course, but we felt that those represent 3 “relatively” independent lines of evidence, and they sample 3 different time-scales of adjustment.
12:02:18 From Mark Webb : In the WCRP assessment we removed paleo and historical lines of evidence in turn in the robustness tests. Hence one cloud argue that the robust range takes into account the possibility that paleo and historical evidence are dependent.
12:03:12 From Benjamin Sanderson : @steve - sure, that's reasonably argued in the paper, but my question is (1) whether the actual degrees of freedom (Effective climate sensitivity + pattern effect parameter) bake in some interdendency between feedbacks on different timescales - given that EffCS is some combination of slow and fast feedbacks. and (2) whether the assumption of an independent pattern effect parameter for the paleo evidence and the recent warming evidence is justified. In short =does the implicit simple model used in the statistical model impact the result due to underlying assumptions of interdependencies of feedbacks on different timescales?
12:04:01 From Andrew Williams (he/him) : Thanks all! Great talks
12:04:05 From Marcus Sarofim (US EPA) : Thank you all.
12:04:14 From Gavin Foster : Thanks all - great talks and discussion
12:04:15 From Florent : thanks. bye.
12:04:18 From Benjamin Sanderson : Cheers for a nice discussion all
12:04:45 From Gavin Schmidt : You could combine zoom with slack?
12:04:46 From Peter Watson : thanks all!
12:05:41 From stephenklein : @Ben regarding a), see Section 6.2. of WCRP report “Transfer function/SST pattern error” - Mark is now talking about this
12:05:44 From Laura Suarez : This was a great discussion, and great presentations Kyle & Thorsten! Thanks everyone
12:05:59 From stephenklein : But there is some uncertainty in that calculation in Section 6.2
12:09:30 From Daniel Koll : Arguably, TCS is more relevant to humans than ECS. If paleo is an important piece of evidence in constraining ECS, does that mean that TCS is inherently more difficult to constrain?
12:11:11 From stephenklein : @Casey: Constraining deep convective cloud feedbacks would help - it’s one of the largest existing uncertainties. You can run the Webb code to test different values of tropical deep convective cloud feedbacks on ECS
12:11:23 From Marcus Sarofim (US EPA) : Your discussion of using the full century information would be something like a Chris Forest approach, using varying ECS, ocean uptake, and aerosol forcing in historical EMIC runs, and using the distance between model & observation (for multiple observational metrics, e.g. latitudinal surface temperature and ocean heat) to weight the parameter combinations...
12:12:06 From Mark Webb : I need to go. Thanks for the great talks and discussion!
12:12:13 From Kyle Armour : Thanks, Mark!
12:13:14 From Mark Zelinka : Thanks Kyle and Thorsten! This was great
12:13:21 From Gavin Schmidt : Thanks all!
12:13:26 From Casey Wall : Thanks!