Revisiting causality using stochastics

The Proceedings of the Royal Society A has just published two papers by Koutsoyiannis et al. on revisiting causality using stochatics, the first being the theory paper and the second presenting some case studies. One of the case studies in the second paper considered Atmospheric temperature and carbon dioxide concentration, with their results suggesting that

the results in figure 14 suggest a (mono-directional) potentially causal system with T as the cause and [CO2] as the effect.

They then concluded that

[h]ence, the common perception that increasing [CO2] causes increased T can be excluded as it violates the necessary condition for this causality direction

As regular readers will probably realise, this is clearly nonsense. There is overwhelming evidence that the increase in atmospheric CO2 is due to human-caused CO2 emissions, and that this increase is the dominant cause of the increase in global surface temperatures.

There probably isn’t much point in going through all the reasons why what is suggested in this paper is almost certainly wrong. I thought I might simply highlight that I started a PubPeer thread about this paper and Gavin Cawley has already posted a couple of useful comments. A PubPeer thread about the Zharkova et al. paper produced quite an extensive comment thread and probably played a role in it being retracted.

You might argue that a paper shouldn’t be retracted just because one of the case studies produces results that are almost certainly wrong. On the other hand, you might also argue that if one of their case studies produces such results that it rather undermines their whole method. You might also argue that it’s rather embarassing that one of the Royal Society’s journals could publish a paper with what is, these days, a very obviously wrong result.

Advertisement
This entry was posted in Greenhouse effect, Research, Science and tagged , , . Bookmark the permalink.

41 Responses to Revisiting causality using stochastics

  1. dikranmarsupial says:

    “On the other hand, you might also argue that if one of their case studies produces such results that it rather undermines their whole method.”

    I think it is more that they applied their own method badly, and didn’t consider the consequences of their design choices (particularly differencing the time series). And didn’t consider the known physics/chemistry of this question (which has over 50 years of solid research behind it) rather than just dismiss it as “common perception”. The fact that the paper begins with a Chinese proverb:

    “Deep doubts, deep wisdom; shallow doubts, shallow wisdom”.

    becomes rather ironic. If you are making a controversial claim, using a new untested method, then it doesn’t take especially deep wisdom to see that deep doubts about your result are warranted.

    “You might also argue that it’s rather embarassing that one of the Royal Society’s journals could publish a paper with what is, these days, a very obviously wrong result.”

    Their response to the error being pointed out was also somewhat underwhelming. Many thanks for setting up the Pub Peer thread, it would be nice if it resulted in an erratum, but I think that is somewhat unlikely.

  2. Dikran,

    I think it is more that they applied their own method badly, and didn’t consider the consequences of their design choices (particularly differencing the time series). And didn’t consider the known physics/chemistry of this question (which has over 50 years of solid research behind it) rather than just dismiss it as “common perception”.

    Good point. That is a better way to put it.

    Their response to the error being pointed out was also somewhat underwhelming.

    I wasn’t sure if you wanted this mentioned, but – yes – “underwhelming” is a polite way to describe their response.

  3. dikranmarsupial says:

    I always try (but sometimes fail) to be polite ;o)

    I think there is a good reason why people say things like “X ‘Grainger-causes’ Y” rather than just “x causes Y” when using Grainger’s method. Statistical causality really doesn’t mean by “causes” what we normally think of as “causes” (and people not rushing to look up the definition can be one of those feature-bug things ;o)

  4. RickA says:

    Doesn’t the ice core data show CO2 levels could have lagged behind rising global temperatures by as much as 1,400 years? Doesn’t that evidence support this example?

  5. RickA,
    No, firstly the lag is probably quite small. Plus, it probably is the case that initial increases in temperature triggered the increase in atmospheric CO2, but it’s this increase in atmospheric CO2 that then drove further increases in temperature. This post discusses this in more detail. The updates at the end, and some of the comments, are well worth reading.

  6. dikranmarsupial says:

    RickA They look at ice core data and the modern data separately. In the case of the ice-core data, CO2 is acting as a feedback, so it is “Egg or Hen” causation (both directions), temperature causes degassing of the oceans but also CO2 causes temperature via the greenhouse effect. For the modern data, the rise is primarily due to anthropogenic carbon emissions, so it is acting as a forcing rather than a feedback, so the greenhouse effect is the dominant causal direction. What degassing of the oceans there has been has limited the natural environment’s ability to oppose the rise caused by anthropogenic emissions, rather than actually causing any of the increase in atmospheric CO2. Their contentious claim is about the modern period.

  7. RickA says:

    Isn’t this example you are saying is obviously wrong merely saying that delta T causes delta CO2? How can that be obviously wrong and at the same time “it probably is the case that initial increases in temperature triggered the increase in atmospheric CO2 . . .”? Isn’t this example trying to determine started the feedback loop? Not look at an intermediate feedback variable?

    Couldn’t the temperature rise from the LIA (little ice age) have caused the CO2 to rise – which then increased the temperature further, which caused more CO2 to be added to the atmosphere, and so on?

    It just seems to me that the ice core data, showing CO2 lagging T (no matter how small the lag) plus the temperature rise out of LIA causing CO2 to rise, both support the example in this paper. That doesn’t mean the paper is correct or the example is correct – but it also doesn’t mean delta T causing delta CO2 is “obviously incorrect” either.

  8. RickA,

    Couldn’t the temperature rise from the LIA (little ice age) have caused the CO2 to rise – which then increased the temperature further, which caused more CO2 to be added to the atmosphere, and so on?

    No, for a number of reasons. One is illustrated by the second figure in this post. We can estimate how much atmospheric CO2 would increase if we warmed by about 1C. It’s about 10ppm. So, the warming cannot explain the increase in atmospheric CO2 from 280ppm to 410ppm.

    The other point is that the amount of CO2 emitted by humans is more than twice as large as the increase in atmospheric CO2. Where has this gone if the rise was a consequence of increasing temperatures? As Dikran mentions, the increase in temperatures has influenced atmospheric CO2, but primarily through influencing how the sinks have taken up some of our emissions.

    Also, if you read this Skeptical Science article it present numerous other lines of evidence indicating that the increase in atmospheric CO2 is predominantly due to human emissions.

    Finally, we’re also confident that the increase in atmospheric CO2 is enhancing the planetary greenhouse effect, which is causing the planet to warm.

    It really isn’t the other way around.

  9. dikranmarsupial says:

    RickA We know that the natural carbon cycle is opposing the rise, it is taking more CO2 out of the atmosphere each year than it puts in. How can the recovery from the little ice age be causing CO2 levels to rise when the net action of the natural carbon cycle is in the other direction?

    We know this from the mass balance argument, which has been discussed here many times (and again on the Pub Peer thread).

    “Isn’t this example you are saying is obviously wrong merely saying that delta T causes delta CO2?”

    They are saying a lot more than that, they are saying this is the dominant direction of causality, which is inconsistent with many lines of evidence, including the mass balance argument. You can’t seriously study the carbon cycle without knowing these basic facts. The authors should have done the groundwork before making such a claim.

  10. RickA says:

    dikranmarsupial:

    I thought the evidence was that 50% of the human added CO2 was absorbed by the ocean and plants and so forth – not that MORE CO2 came out of the atmosphere than went in. Otherwise, the CO2 level wouldn’t be rising – but falling. So I think you may have worded something wrong.

    ATTP:

    As I said above – 1/2 of whatever humans emit is sucked back out of the atmosphere, but the level still rises because 1/2 is left in (over and above the non-human natural cycle). So there is delta CO2 which could be cause by delta T.

  11. RickA,

    As I said above – 1/2 of whatever humans emit is sucked back out of the atmosphere, but the level still rises because 1/2 is left in (over and above the non-human natural cycle). So there is delta CO2 which could be cause by delta T.

    Can you explain this again, because this doesn’t seem to make sense. The natural sinks take up just over 50% of what humans have emitted, leaving just under 50% in the atmosphere. There isn’t space for delta T to be causing some of the rise. As I mentioned above, the increase in temperature probably means that the increase in atmosphere CO2 is slightly greater than it otherwise would be. However, this effect is relatively small (~10ppm) and mostly means that the natural sinks have taken up slightly smaller fraction of our emissions than they would have done had temperatures not also increased.

  12. RickA says:

    Or you could look at the average global temperature when CO2 was 280 and the temperature now at CO2 of 416 and conclude that temperature difference caused the 136 ppm of CO2 increase?

    Plus, the effects of CO2 is logarithmic – meaning a 10 ppm change at 280 makes a bigger difference than a 10 ppm change at 416. So a small temperature increase which first caused a small co2 increase back at 280 would have bigger effects than the same co2 increase today.

  13. RickA,

    Or you could look at the average global temperature when CO2 was 280 and the temperature now at CO2 of 416 and conclude that temperature difference caused the 136 ppm of CO2 increase?

    No, because this is inconsistent with the glacial cycles and with the carbonate chemistry of seawater.

    Plus, the effects of CO2 is logarithmic – meaning a 10 ppm change at 280 makes a bigger difference than a 10 ppm change at 416. So a small temperature increase which first caused a small co2 increase back at 280 would have bigger effects than the same co2 increase today.

    It’s the warming that depends logarithmically on changes in atmospheric CO2, not the other way around.

    Seriously, the increase in atmospheric CO2 is entirely due to human emissions of CO2 through fossil fuel use and land use changes.

  14. RickA says:

    “Seriously, the increase in atmospheric CO2 is entirely due to human emissions of CO2 through fossil fuel use and land use changes.”

    Maybe. Or 50% of the warming could be caused by human CO2 emissions and 50% could be caused by prior natural delta T. We don’t really know the answer and that is what we are trying to figure out. I am not so sure it is absolutely clear that all CO2 increase is due to humans or that all temperature increase is due to humans. I suspect it is partly natural and partly human caused.

    I guess more science will shed light on this knotty question.

  15. RickA,
    It is now “unequivocal” that the warming since the mid-1800s is due to human emissions of CO2. There is virtually no doubt. Similarly, it is virtually certain that the increase in atmospheric CO2 is entirely due to human emissions of CO2. I know what you suspect. I suspect (which is an understatement) that you’re simply wrong.

  16. dikranmarsupial says:

    RickA “I thought the evidence was that 50% of the human added CO2 was absorbed by the ocean and plants and so forth”

    The natural environment has to be a net carbon sink for that to happen. If it is a net sink, it is opposing the rise, not causing it.

    “– not that MORE CO2 came out of the atmosphere than went in. “”

    Nobody is suggesting that.

  17. dikranmarsupial says:

    “I guess more science will shed light on this knotty question.”

    There has been more that 50 years of intensive research on the subject. A lot of the journal papers are freely accessible, if you think more science is needed on this topic, it is because you haven’t read the science that has already been published.

    You can start with the mass balance argument, which I give briefly in the PubPeer thread. If you can identify a flaw in the logic, I’d be happy to discuss it here.

  18. b fagan says:

    I couldn’t get to the paper (no real need), but is there a reason they chose a rare time in planetary history where glacial cycles prevailed, rather than, for example, periods where large igneous provinces and CO2/temperature increases ensued with entirely different feedback loops? Or the episodes where rapid depletion of atmospheric CO2 caused rapid cooling?

    RickA, to have temperature increase explain any meaningful fraction of temperature increase during the Industrial Age, you would have to explain a few difficult things:
    – why didn’t CO2 shoot up similarly high in other centuries of warming? Ice cores would show that in over the last hundreds of thousands of years.
    – fossil industry records document a good deal of how much carbon has been bought and burned. It produces CO2 in quantities that are like current atmospheric increase + plant growth + ocean and lake uptake.
    – the fossil carbon is changing the isotopic signature of surface CO2 and plants. Warming alone doesn’t surface carbon with those isotopic ratios, but mining and drilling does.

    But the big question I want to know about the papers, since the Supplemental Link mentioned it – have they “solved” the hen-or-egg question to their satisfaction? Which one did they pick?

  19. b fagan says:

    And forgive my typo in the bit to RickA:
    “to have temperature increase explain any meaningful fraction of CO2 increase during the Industrial Age”

  20. Have I just fallen through a Wormhole. I thought the sceptics had moved onto solution denial!?

  21. dikranmarsupial says:

    @Richard I think this (temp causing rise in CO2) is one of the first zombie myths, it was mentioned in the first or second IPCC report, it will never die 😦

    Probably shouldn’t still be appearing in non-predatory journals in 2022 though!

  22. This is what I added to PubPeer

    “This is nuts in that they need to have a plausible physical explanation for why a less than 1C change in ocean temperatures in the last 100 years would cause the atmospheric CO2 concentration to jump from 280 PPM to above 400 PPM. The activation energy would need to be at least 3 eV, which is very steep for a simple phase transition of a dissolved gas. I think the heat of vaporization of water is like 0.42 eV, a much shallower activation barrier — think of how a 1C change in water temperature isn’t going to change the relative humidity of a closed volume by 50% !

    One really needs to perform a physics sanity check before publishing a paper like this.”

    If I have the details of this wrong I will retract, just like they probably should retract their article if they have it wrong.

  23. Paul,
    I don’t know if the exact details are right, or wrong, but suggesting that a ~1C increase in temperature could increase atmospheric CO2 by > 100ppm goes against our understanding of the carbonate chemistry of seawater and evidence for glacial cycles.

  24. dikranmarsupial says:

    It is ironic that they analyse both the ice core data and the modern data, where the major causal difference between the two is anthropogenic carbon emissions and conclude that modern causality is dominantly from temperature to CO2.

  25. russellseitz says:

    It is ironic that they analyse both the ice core data and the modern data, where the major causal difference between the two is anthropogenic carbon emissions and conclude that modern causality is dominantly from temperature to CO2.

    Koutsoylannis controversial conclusion should earn a Fox TV interview by Tucker Carlson, who has taken to curating scientists and ecomodernists worthy of misrepresentation across a broad range of policy concerns, from molecular virology to energy policy.

    https://vvattsupwiththat.blogspot.com/2022/07/he-also-doesnt-believe-in.html

  26. mrkenfabian says:

    I expected better from the Royal Society. RickA, not so much.

  27. izen says:

    Have Koutsoyiannis et al. not argued themselves into a corner ?
    If the rise in temperature has caused the rise in CO2, then unless they have some good reason for claiming that rising CO2 has NO influence on temperature, they are predicting a continued warming from the interaction between the two effects.
    And no way to prevent it.

  28. izen,
    I think they’re claiming that the causality is that changes in T causes changes in CO2 and not the other way around (I.e., I think they’re disputing that changes in CO2 cause changes in CO2).

  29. izen says:

    @-ATTP
    “I think they’re claiming that the causality is that changes in T causes changes in CO2 and not the other way around ”

    That would imply that they refuse to accept there is any warming effect from any atmospheric component, not just CO2. Warming also results in greater water vapour content and rises in methane emissions.
    I can kind of conceive of them denying CO2 causes warming BECAUSE warming causes CO2 emissions and causation can only be one-way. However such dogmatic beliefs that feedback processes are impossible would seem to fly in the face of empirical evidence. But denying any ‘greenhouse’ warming strikes me as extreme and unscientific.

  30. angech says:

    ATTP
    “. We can estimate how much atmospheric CO2 would increase if we warmed by about 1C. It’s about 10ppm. So, the warming cannot explain the increase in atmospheric CO2 from 280ppm to 410ppm.”

    If 10 ppm is associated with 1C warming , your words, then 280 to 290 ppm would have given this result.
    Going to 410 would cause a 13 degree C rise.
    Since this has not happened your assumption that a warming of 1C would only cause a 10 ppm rise is wrong.

  31. angech,
    No, 1C of warming is associated with an increase in atmospheric CO2 of about 10ppm. This does not mean that an increase of atmospheric CO2 of 10ppm would produce 1C of warming.

  32. izen says:

    @-ANG
    “If 10 ppm is associated with 1C warming , your words, then 280 to 290 ppm would have given this result.
    Going to 410 would cause a 13 degree C rise.”

    The sensitivity of temperature to CO2 is estimated to be about 2.5-3degC for a doubling of CO2.
    The sensitivity of CO2 to temperature is estimated to be about 10ppm for a 1degC rise.

    This is one reason why although there is a reciprocal causal relationship between the two, it cannot lead to a runaway warming.
    Thankfully.

  33. Dave_Geologist says:

    Strikes me this is a textbook example of Not Even Wrong (thanks for the PubPeer explanation dikran).

    They’ve chosen a methodology which makes it impossible to detect causality from long-term CO2 emissions to temperature, and concluded that there is no causality from long-term CO2 emissions to temperature.

    Face-palm!

  34. Magma says:

    “[h]ence, the common perception that increasing [CO2] causes increased T can be excluded as it violates the necessary condition for this causality direction”

    Oof. Seems to be in the same spirit as disproving the existence of owls or hats by reasoning from first principles. Perhaps the journal editors were on holiday.

  35. Magma,
    Well, if they were on holiday, they decided to double down when they were emailed to point out this rather egregious error.

  36. dikranmarsupial says:

    Dismissing decades of careful research as “common perception” is something that really shouldn’t get past peer review. It is clearly either rhetoric or very poor scholarship (there is a section in the IPCC WG1 report listing the evidence that shows the rise is not a natural phenomenon). All reviewers have off-days occasionally, but I suspect that none of the reviewers had expertise on the carbon cycle (or even climate change generally).

  37. Koutsoyiannis is highly cited in the hydrology literature with respect to rainfall distributions and extremes. We cited him in our book Mathematical Geoenergy as examples of how to model fat-tail statistical distributions. He has always had a beef with AGW, stemming from his assertion that warm periods in history are also fat-tail, fractal, or Hurst occurences.

  38. dikranmarsupial says:

    Unfortunately this is not the first obviously wrong paper they have written. Another example is

    D. KOUTSOYIANNIS , A. EFSTRATIADIS , N. MAMASSIS & A.
    CHRISTOFIDES (2008) On the credibility of climate predictions, Hydrological Sciences Journal,
    53:4, 671-684, DOI: 10.1623/hysj.53.4.671

    Where they compare station data with co-located grid boxes from single runs of climate models, and in doing so showing they don’t understand how climate models work (they simulate climate, they don’t predict weather, so the internal variability in the model run cannot be expected to be coherent with the internal variability in the historical observations). There is a good thread on this at RealClimate.
    So again, they are being critical without doing their homework and understanding the science that they are contradicting.

    As hydrologists, they should know that this is not going to work, as hydrologists use statistical downscaling to predict the local effects of climate change, and those methods do not use single grid points, they use synoptic scale climate model output (for example).

    Note that the paper received a comment paper:

    Citation Huard, D. (2011) A black eye for the Hydrological Sciences Journal. Discussion of ‘A comparison of local and aggregated climate model outputs with observed data’ by G.G. Anagnostopoulos et al. (2010, Hydrol. Sci. J. 55 (7), 1094–1110). Hydrol. Sci. J. 56(7), 1330–1333.

    and naturally a reply:

    Koutsoyiannis, D., Christofides, A., Efstratiadis, A., Anagnostopoulos, G.G. and Mamassis, N. (2011) Scientific dialogue on climate: is it giving black eyes or opening closed eyes? Reply to “A black eye for the Hydrological Sciences Journal” by D. Huard, Hydrol. Sci. J. 56(7), 1334–1339.

    … in which they demonstrate that they didn’t understand the criticism. Why would you test whether climate models can predict station level data, when we know a-priori that this is an unreasonable expectation, and nobody uses climate models that way?

  39. dikranmarsupial says:

    That should be “… in which they demonstrate that they didn’t understand the criticism. Why would you test whether single grid boxes from single climate model runs can predict station level data, when we know a-priori that this is an unreasonable expectation, and nobody uses climate models that way?”

    Obviously climate models can make projections for station data, that is the point of statistical downscaling.

  40. dikranmarsupial says:

    As to being highly cited, I should point out that the paper in question has 185 Google scholar citations, despite being obviously wrong-headed. If you write a lot of papers (Koutsoyiannis has published 903 according to Google Scholar) you will pick up a fair few citations, sometimes even for questionable work. Koutsoyiannis’ mean citations per paper (18.77) is rather worse than mine (48.05) and I wouldn’t say I was particularly eminent in my field.

    Metrics should be taken with more than a pinch of salt.

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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.