Can climate sensitivity be really high?

The answer to the question in my post title is – unfortunately – yes. The generally accepted likely range for equilibrium climate sensitivity (ECS) is 2oC – 4.5oC. This doesn’t mean that it has to fall within this range, it means that it probably falls within this range. There is still a chance that it could be below 2oC and a chance that it could be greater than 4.5oC. However, there is a difference between it possibly being higher than 4.5o and this being likely.

There’s been quite a lot of recent coverage of studies suggesting that the ECS may be higher than 5oC. My understanding is that one reason for this increase in the ECS in some climate models is an enhanced short-wavelength cloud feedback in these models. There are also some indications that these high-sensitivity models do a better job of representing some of the cloud processes than was the case for the earlier generation of models.

However, there are also indications that the high-sensitivity models struggle to fit the historical temperature record, and that lower sensitivity models (at least in terms of the transient climate responses) better match some observational constraints. As I understand it, it’s also difficult to reconcile these very high climate sensitivity estimates with paleoclimatological constraints.

So, I think it is interesting, and somewhat concerning, that some of the newest generation of climate models are suggesting that the equilibrium climate sensitivity (ECS) could be higher than 5oC. That these newer models seem to also represent some relevant processes better than the previous generation does provide some indications that it could indeed be that high. However, it’s also possible that these models are poorly representing some other processes that may be unrealistically inflating their ECS values.

Given that there are also other lines of evidence suggesting that the ECS is unlikely to be as high as 5oC makes me think that we should be cautious of accepting these high ECS estimates just yet. I do think it’s worth being aware that it could be this high, but I don’t think it’s yet time to change that the ECS is likely to fall between 2oC and 4.5oC, with it probably lying somewhere near 3oC.


Climate worst-case scenarios may not go far enough, cloud data shows – Guardian article about the new high climate sensitivity studies.
Short-term tests validate long-term estimates of climate change – Nature article about a recent study that tested one of these high climate sensitivity models.
CMIP6 – some of my recent posts about the newest generation of climate models.

This entry was posted in Climate change, Climate sensitivity, Research, Science, The philosophy of science and tagged , , , , . Bookmark the permalink.

36 Responses to Can climate sensitivity be really high?

  1. JCH says:

    I thought Nic ruled 3 ℃ out. On to herd immunity.

  2. RK Upadhya says:

    Total novice at climate-science here: are these climate sensitivity models linear? That is, if we assume that climate sensitivity is 5 C, would the model imply then that we are currently locked in for ~1.7 C of temperature rise? Or, are there more developed models that break down the climate sensitivities with respect to different atmospheric CO2 concentrations, i.e. one for 350-400, another for 400-450, etc. Well, actually since I have seen these non-linear models, I guess my question is specifically how these are included or related to these general discussions of one single climate sensitivity.

  3. RK,
    Good question. Bit tricky to answer. We expect it to be roughly linear with respect to the change in forcing. The relationship between forcing and concentration is

    \Delta F = 5.35 \ln \left(\dfrac{C}{C_0} \right),

    where \Delta F is the change in forcing, C is the current atmospheric CO2 concentration, and C_0 is the initial atmospheric CO2 concentration (typically taken to be about 280 ppm).

    The change in forcing if we double atmospheric CO2 (which is how climate sensitivity is defined – how much will we warm if we double atmospheric CO2) is 3.7 W/m^2. The change in forcing from pre-industrial to today (taking into account all emissions, including aerosols) is about 2.5W/m^2. So, if the ECS is 3C and we were able to keep conditions the same as they are today, then we’d eventually warm to about 2C (2.5/3.7 x 3). If the ECS is 5C and we were able to hold conditions the same as they are today, we’d eventually warm to over 3C.

    However, there is another complication. There are really 2 climate sensitivities – the equilibrium climate sensitivity (ECS) and the transient climate response (TCR). The ECS is what we will eventually warm to if we double atmospheric CO2. The TCR is the warming at the time when atmospheric CO2 is doubled. Since the oceans have a very large thermal inertia, it takes a long time (centuries) to warm to equilibrium. So, in some sense the TCR (which is lower than the ECS) is more relevant. That’s expected to be around 2C.

    A further complication is that when we get our emissions to ~zero, the oceans will continue to take up some of what we’ve emitted, so that atmospheric CO2 will actually drop slighty.

    So, how much warming we might expect depends on what scenariop you’re considering. Continued emissions, reduced emissions, or zero emissions will have different warming committments.

    Hope that makes some sense.

  4. My gut tells me ECS is going to fall in the high end of the ECS range, probably in the 4 degree range, so no reason for alarm. If it turns out to be higher than that, we will have some extra work to do to manage the extra warmth. I read through your post and get a sense that your mood about this matches my own. No reason to get all worked up about it at this particular moment.
    Thanks for keeping a level head and providing analysis and narrative that helps us keep the alarmists at the Guardian and Nature.

  5. Dave_Geologist says:

    Going back to past threads which discussed the PALAEOSENS compilation ATTP, the paleoclimatological constraints were reconciled with the CMIP ECS range by removing the longer-term Earth System Sensitivity (ESS) studies. Pagani et al. is fair enough, because they explicitly say in their paper that they’re setting out to determine climate sensitivity on a longer time-frame than the CMIP models, incorporating geological and biogeographical feedbacks that are not in those models (but are in palaeo models run in much coarser time-steps and at lower resolution). Although it’s worrying that it represents the last time climate was where we’re sending it.

    I’m a bit less sanguine about discounting the high values around the Last Glacial Maximum, on the basis IIRC that there was a strong albedo feedback from the loss of Northern Hemisphere continental ice sheets and you can only lose them once. Once they’re gone, that feedback is also gone. They’re ESS because it takes a long time to build or melt ice sheets. And on that note, we’ll be worrying about the sea-level impact anyway before we have to worry about the ESS impact. But OTOH loss of thin snow cover, albedo change in melting tundra/permafrost etc. may be lurking out there as a fast feedback that was not available in a world of ice sheets with dry mammoth steppe to the south.

    The PETM is the other high-ECS/ESS outlier, but since there’s pretty much a consensus that a large input of methane was responsible, through positive feedback and release of stored methane, that’s almost by definition fast because of the short atmospheric residence time of methane. If Antarctic tundra/permafrost was the source we can discount that one, because you can only build the organic carbon reservoir in a greenhouse world, and again once that’s gone, it’s gone. Most of what’s under the Antarctic ice sheet nowadays is bare rock. But we may have our own organic carbon reservoirs which were not there in the PETM, arctic tundra/permafrost again and marine clathrates.

    I’m not saying any of them is likely, any more than I’d say Nic Lewis’s lowball estimates are likely, but we shouldn’t discount either end.

    We might also be concerned about state-dependency, as per the Köhler et al. paper: A State-Dependent Quantification of Climate Sensitivity Based on Paleodata of the Last 2.1 Million Years. Our calibration data has a lot of glacial-interglacial transitions built in because you have repeated events which you can stack up and be impressed by how similar all the answers are. But they’re all asking an in-sample question, and may be inappropriate for an out-of-sample question about a world we haven’t seen for millions of years.

  6. David B Benson says:

    Earlier comments on Real Climate suggested looking at the global temperature about 3.5 million years ago when the continents were close to the same positions as now and the carbon dioxide concentration was 350+ ppm. This resulted in a global temperature product of 2–3 K warmer than ‘now’, where the last is a bit tricky. This gives an Earth System Climate Sensitivity considerably greater than 3 K, but I don’t remember just what Hansen and coworkers obtained.

    The main point, I suppose, is that ECS is irrelevant in the face of melting ice sheets as it doesn’t take the effects thereof into account.

  7. David,
    IIRC, Hansen got 5-6C for the ESS, but that was from a glacial state. I think that we’d expect it to be a little lower in an inter-g;acial, given that the ice sheets are smaller.

  8. David B Benson says:

    aTTP, that’s not how I remember it, but it is easy enough for those who still have academic connections to check the global temperature and CO2 concentration from 3.5 million years ago to define yet another sensitivity, one that applies for a world with very little ice, yet still there was the climate controlling circumpolar Antarctic winds, not impeded as the Drake Passage had opened.

    This is the actual climate sensitivity going forward to some sort of equilibrium. Probably easier just to read “The Long Thaw” by David Archer.

  9. David B Benson says:

    Willard, my comment fell into moderation, again for no obvious reason. Thanks.

  10. David,
    Maybe I’m missing your point, but I think that if we were to fix concentrations at today’s level (~400 ppm) and let the system go to equilibrium, then we might expect 2C from fast feedbacks and another ~1C from slow feedbacks on millenial timescales. So, 2-3C in a similar world to ours in the past seems consistent with current estimates for the ECS.

  11. David B Benson says:

    aTTP — No, that’s too small given the data from 3.5 million years ago. More like 3–4 K from fast feedbacks and 2 K from the slow ones for 400 ppm of carbon dioxide.

  12. Mitch says:

    There is a big difference between the Pliocene (3.5 million years ago) and today which makes it inappropriate to make a simple ECS estimate by comparing the two.
    Abyssal ocean waters were about 3 deg C warmer than modern (roughly 5 deg C), and the world was slowly cooling down, rather than rapidly warming up as today. This huge heat reservoir of the Pliocene tended to prevent cool climate transients from expanding to cool the global climate. It tended to keep both poles relatively warm.

    Today we have a cold deep ocean that minimizes temperature change especially in Antarctica and insulates the earth from warm transients. The warming ECS we should expect should be lower than what one might think comparing the Pliocene to the modern.

  13. David B Benson says:

    Mitch — ECS has a peculiar and narrow definition which, in my opinion, renders it an inappropriate measure for what will actually happen at quasi – equilibrium. So the simplest is to ask how warm will it be when the ice sheets melt and the oceans warm? Voila, we have an answer from the mid-Pliocene with a CO2 concentration of but 350 ppm: about 3 K warmer than now.

    This seems to me to be straightforward physics plus geology and far more to the point than the usual calculations of the limited ECS.

  14. David B Benson says:

    Willard, my comment went into moderation once again.

  15. Jon Kirwan says:

    To: Dave_Geologist.

    I’m just a hobbyist in these things. (My specialties have more to do with computer architecture — worked on chipsets at Intel — and, with respect to physics, about pyrometry and phosphor thermometry — 20+ years of active design work there as an engineer. And none of this as a scientist; instead only being blessed by finding work allowing me to work with them from time to time — for example, on the first re-writable CD.)

    Almost every time I read your writing I don’t just learn something, I also see in your thoughts a push towards a more comprehensive view, as well. Something I’ve learned over the years to respect and value. Once again, I enjoyed the sweep and grasp of your thinking. It helps keep me also thinking more broadly and comprehensively and helping me to avoid falling into thinking-ruts (that are all too easily had.)

    Here, I think your comments, when added to those from …and Then There’s Physics, provide a much improved balancing. I also felt that …and Then There’s Physics failed, alone, good as it was still failed to provide that needed extra bit. So, good to see your addition. (Not that there aren’t still more needed additions…)

    Once in a while I think it is a good idea to let someone know that you appreciate their thoughts and contributions. Many of us want that once in a while, though I suspect serious scientists engaged in doing more fundamental work usually must learn how to do without. 😉

    I’m glad I have some modest access to your thoughts from time to time. Just wanted to say so, as it matters to me. I don’t have anything to add to it except my thanks.


  16. Dave_Geologist says:

    Thanks for the compliment Jon. I tend towards a “don’t forget the ESS – the world doesn’t end in 2100” side of the argument. But I appreciate people struggle to act for their grandchildren’s benefit, let alone subsequent generations, and that Rome wasn’t built in a day. If we wreck our civilisation by 2100, it’s probably a bit moot to be worrying about 2500. Or 3000. There is one of those long-term papers with slow feedbacks and Milankovitch cycles included which says we have 1000 years to get CO2 down to pre-industrial, otherwise we’ve irrevocably cancelled the next Ice Age. You get into one of those state changes where so much albedo has been lost that the Milankovitch cycle isn’t strong enough to trigger the level of feedback required for a glaciation. Of course we may not want to uncancel the next Ice Age – it would be rather inconvenient.

    I guess one of the reasons why 1.5-2°C is considered “Mostly Harmless” is that it keeps us in the in-sample range of stadial-interstadial cycles. If we get to Pliocene temperatures we’re calibrating to a world with different oceanography, as Mitch points out. We don’t know how fast we’ll get there, and whether there will be tipping points which have a large surface impact (for example, a change that impacts surface waters much faster than the thermal-inertia-controlled changes to deep waters), on a shorter time scale than ESS. Actually checking back to PALAEOSENS I see the very high Pleistocene ECS is from the Charney 1979 NAS report, of the same era as Hansen’s 5-6°C and possibly the same calculation. So you might argue that it’s been superseded by subsequent studies of the same era. The main reason it’s discounted though is that it’s ESS, so not relevant for the 2100 timeframe. Even with accelerated ice streams and calving, I just don’t see that it’s possible to lose the Greenland and most of the Antarctic ice sheets in a century. Too much latent heat capacity locked in. We can of course melt enough to make life very uncomfortable for coastal cities, and those rebuilding them will have to take account of the fact that the melting by 2100 is only the first instalment, even if we’ve stabilised surface temperature by then.

  17. I agree that the ESS is worth considering. However, there is this interesting relationship between emissions and warming which indicate that the relevant metric is the transient response to cumulative emissions (TCRE). Essentially, if we ignore the slow feedbacks, then our long-term warming committment will essentially be the TCR at the time when emissions cease. This is because the oceans will continue to take up some of our emissions which will roughly cancel how much we would need to warm to equilibrium when emissions ceased (i.e., atmospheric CO2 will drop slightly when we get emissions to zero).

    I’ve often wondered if something similar might apply to the ESS. In other words, on very long timescales, how will the continued take up of our emissions compensate for the warming due to the slow feedbacks? I don’t know the answer.

  18. Dave_Geologist says:

    FWIW, these are the time-frames (in years) used in a recent (2018) review of ECS/ESS, with Köhler as a co-author: Comparing Climate Sensitivity, Past and Present. It appears to be open access. I’ll try to post some figures later if it’s not blocked.

    Direct responses, including vapor and cloud feedbacks: 1
    Aerosol and land-surface feedbacks: 10
    Snow and sea-ice albedo feedback: 20
    Carbon-cycle feedbacks: 150
    Surface-ocean temperature equilibration: 150
    Vegetation albedo: 500
    Continental ice albedo feedback: 750
    Deep-ocean temperature equilibration: 1,500
    Carbonate compensation: 10,000
    Weathering: 200,000

  19. Dave_Geologist says:

    Our comments crossed ATTP, but you might infer from their Fig 5 that it’s fairly flat for 1,000 years, doubles over the next 10,000 years, then really takes off. I see a rise in GHG then but I assume that’s the environment becoming a net emitter rather than a net absorber. For example deepening of the CCD would release CO2 currently locked up in seafloor sediments, which would be released to the atmosphere on the time scale of deep-ocean overturning. Since it’s a climate sensitivity study I presume the anthropogenic forcing is just an abrupt or 70-year doubling. On their timescale it wouldn’t matter which. It’s encouraging that a study with slow feedbacks included reaches essentially the same conclusions as CMIP models forced with a sudden change to net zero emissions on a specified date: that the feedbacks roughly cancel out for centuries, not just decades. Not for millennia though. IOW our temperature for the next 1000 years is entirely in our own hands, although what we do in that timeframe will strongly influence subsequent millennia (although you might argue that a civilisation thousands of years in the future would have geo-engineering capability we can’t even dream of).

    That would be consistent with the last non-geological feedback (deep ocean warming) petering out after 1,500 years and continental ice albedo change roughly cancelled out by opposing feedbacks on a similar timescale, and also with the 1,000-years-to-save-the-next-Ice-Age paper.

  20. Dave_Geologist says:

    One thing to bear in mind is that if it’s down to clouds and aerosols, an increase in ECS may not translate to an increase in ESS, to the extent that ESS is geologically calibrated to an end-point without reference to how it got there.

  21. Stephan Harrison says:

    Hi Dave-Geologist
    Could you DM me as it would be good to discuss these issues.

  22. David B Benson says:

    Dave_Geologist — On the contrary, from the mid-Pliocene to the present the configuration of the oceans remains essentially the same; closure of the Isthmus of Panama and the opening of Drake Passage with Antarctica at the South Pole. So today’s ocean currents will evolve toward that prior state as the ice sheets melt.

    Back to the Future!

  23. jamesannan says:


    More on this in a month or 6, perhaps…..

  24. Dave_Geologist says:

    David, the different oceanography I was commenting on was to do with the deep/shallow water circulation, not the east-west circulation. And to the starting point in terms of stored heat which may mean that even if you have a closed hysteresis loop, the trajectory may be fast-slow on one arm and slow-fast on the other. Or slow-fast on both. That could impact what falls into the ECS time-frame, even if it’s not formally an ECS feedback. Unless there’s a bifurcation the geological analogues tend not to help with that, just giving us the start-point and the end-point. And as the PLAEOSENS and follow-up papers said, there is in practice a blurring of time-frames between fast and slow. Nature doesn’t compartmentalise for our convenience.

    Stephan, don’t know how, and I prefer to share as a way of thinking out loud, as I’m not an expert. “Thinking by talking” is my style, as I was told by a pop-psychologist on a work training course, as opposed to thinking then talking. Sometimes, just talking 😉 .

  25. Stephan,
    Would be interesting to hear you views on the likely impact of slow feedbacks on long-term anthropogenically-driven warming.

  26. David B Benson says:

    Dave_Geologist —
    is the science. That state of the world ocean is what varies over time.

    Yes, there is hysteresis. The path back to the past state is not simply the reverse of the path forward.

  27. Jon Kirwan says:

    Dave_Geologist — “Even with accelerated ice streams and calving, I just don’t see that it’s possible to lose the Greenland and most of the Antarctic ice sheets in a century.”

    Something newly published here:

    “It’s within the realm of possibility that we could get 1 to 3 meters of sea-level rise from West Antarctica within our lifetimes,” Minchew says.

    Just FYI. And, of course, enjoying reading here.


  28. Greg Robie says:

    I forget when the last time was that the import of the Inuit observations regarding seasonally increased refraction was functionally ignored/dismissed here on ATTP, but I think it was before the paper that associated Greenland ice sheet melt with high pressure. To me, that association of a higher tropopause to melt is more confirmation that the seasonal rise in the tropopause, and the increased refraction is a difference that the models don’t get right when it comes to the latent heat of ice being lost sooner than predicted. The same with the permafrost.

    To the degree my insight/hypothesis that anthropogenic combustion of fossil carbon, at the outset of the Industrial Revolution, introduced a new forcing (a positive feedback of heat gain by conduction in the upper troposphere within the Arctic twilight), which the Inuit observations of increased refraction document as possible … and which modeling conventions do not include, isn’t this assertion about 3° C a great example of motivated reasoning?

    It seems reasonable to me that until models can get the crysophere a bit closer to what is being observed [at least] the upper end of the range should be given more credence than the lower numbers.


  29. Dave_Geologist says:

    Indeed Jon, and thanks for the reference. But it would need to be 20-50 times that to strip the continent down to bare rock. Say, 750 years.

    I agree that those models which do incorporate ice-sheet melting underestimate the rate we’ll see due to things like surges and removal of barriers. In which case we may get an overshoot as albedo change comes faster than we think, then a small decline, then the long-term geological feedbacks push temperatures up again. And even that is overshoot, because eventually weathering will pull them back down. I think sea level is a bigger worry from that though. The temperature impact will be a small increment on top of existing warming, but it will make a huge difference to adaptation whether 30m of SLR happens over a century or over a millennium.

  30. Greg Robie says:

    Dave_G, the early loss of annual frost, and then the loss of it period, is mot inconsequential. Where I live in the Mid-Hudson Valley in New York, this year is an example of why. For the second year running we had multiple ice outs on the ponds I’ve made here on this hillside … in winter. The frost never really got into the ground. I could have tapped my maple trees on the first weekend in January.

    By the end of January, the soft maples (red and silver) were blooming this is at least six weeks too soon given how theses species are evolved to live in the [former?!?] climate zone. I decided to forego sugaring this year so as not to stress the trees beyond the hammering they were experiencing. The freezes if February killed all the sugar maple flower buds. Net result is no maple seeds this year.

    We then had a hard freeze on Mother’s Day weekend. This took the oak tree blossoms. No acorns this year. The freeze variously stunted the leaves and/or caused the trees to rebud. Things almost look “normal” now, but I’m calling thisbthe year of the tiny leaves.

    Healthy leaves can rebud without too much problem (every now and then – like when the invasive gypsy moth defoliated these forests about 50 years ago when it got introduced). This climate shift induced rebudding, like any rebudding, weakens the tree making is more susceptible to disease. Part of the reason I got back into sugaring is so I’d be paying attention as the sugar maple species no longer can grow here. It’s range’s southern boundary is/was ~150 miles to the south.

    Furthermore, I read a paper this winter that said the frost free zone may be about 150 miles south of here by 2050. As the [former] ‘winter’ winds move across unfrozen ground they are warmed. As the solar insolation increases with spring and the leaves are not yet out, the duff layer of the forest floor warms. Because the latent heat of ice is significant, isn’t it foolish to not see the terrestrial shifts as a both/and when it comes to adaptation. Gardening has been a particular challenge this year.

    And the dif’rent ‘winter’ took its toll on bee colonies. The two apiarists I know lost all their hives. Between low pollination and the Mother’s Day weekend freeze, none of the species of fruit trees I have set fruit – & I have no idea how to apportion the cause of this. For a society to adapt to sea level rise it has to be fed …

    Like this morning, I looked out my kitchen window on the unfrozen and warming soil in February’s two Arctic air masses as I made my coffee. I thought about the cold that bothwasnt there and wasn’t going deeper into the soil. I thought about how that cold would not be there to moderate spring temperatures … at least down wind/on a global scale.

    Less frost, delayed leafing, early ice outs, tiny leaves. Those whose careers are lived in models, but have to basically omit the crysophere when drawing conclusions from those models (because the models don’t capture what is being observed), wouldn’t it be sapient to get out from behind ones screen and look afresh at trusted assumptions?


    sNAILmALEnotHAIL …but pace’n myself

    life is for learning so all my failures must mean that I’m wicked smart


  31. Greg Robie says:

    My emailed replies are not posting. The first one I edited and reposted here – thinking it could have been moderated (so the original can be ignored). But the last one hasn’t showed up either, so perhaps it can be released?


  32. Chubbs says:

    The strongest non-model ECS evidence for me are the observational constraints based on recent temperature trends mentioned in the blog article. With the recent warming spike, the CMIP5 mean matches the warming ramp since 1970 well: +0.9C for a forcing increase (GISS, BEST, HADCRUTCW) of roughly 0.5 X double CO2, supporting a 2 to 4.5C ECS range. I’d watch the warming spike. If it persists long enough to increase the temperature trend by another 10% or so, i.e. longer than the hiatus, then the high ECS models could be on to something.

  33. angech says:

    Peter Webster has a new book out and a write up at JC.
    He has added a controversial touch in a recent reply which fits in with both this topic and past topics on climate model paradigms.
    I’m sure ATTP is aware of it, is it worth a post?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.