Real Climate’s Bold Bet

May 9th, 2008

Posted by: Roger Pielke, Jr.

The Real Climate guys have offered odds on future temperature changes, which is great because it gives us a sense of their confidence in predictions of future global average temperatures. Unfortunately, RCs foray into laying odds is not as useful as it might be.

The motivation for this bet is the recent Keenlyside et al. paper that has caused a set of mixed reactions among the commenters in the blogosphere. Some commenters here have stridently argued that the predictions in the Keelyside et al. paper are perfectly consistent with predictions of climate models in the IPCC. However, when one such commenter here was asked to show a single IPCC climate model run showing no temperature increase for the 2 decades following the late 1990s he submitted an irrelevant link and disappeared. Others have argued that the Keenlyside et al. projections (and this includes Keenlyside) are inconsistent with the IPCC predictions. Real Climate apparently falls into this latter camp.

The Real Climate Bet (and there is also one for a later period) is that the period 1994-2004 will have a higher average temperature than the period 2000-2010. Since the periods have in common 2000-2004, we can throw those out as irrelevant. Thus, the bet is really about whether the period 1994-1998 will be warmer than the period 2005-2010. And since we know the temperatures for 2005 to present, the bet is really about what will happen in 2009 and 2010. (Using UKMET temps here.)

It is strange to see the Real Climate guys wagering on 2-year climate trends when they already taught us a lesson that 8 years is far to short for trends to be meaningful. But perhaps there is some other reason why they offer this bet. That reason is that they are playing with a stacked deck, which is what you do when looking for suckers. The following figure shows why.

RcsBold.jpg

For the Real Climate guys to lose the bet global average temperatures for 2009 and 2010 would have to fall by about 0.30 from the period 2005-present (and I’ve assumed Jan-Mar as the 2008 value, 2008 obviously could wind up higher or lower). Real Climate has boldly offered 50-50 odds that this will happen. This is a bit like giving 50-50 odds that Wigan will come back from a 3-0 halftime deficit to Manchester United. Who would take that bet?

Another interpretation of the odds provided by RC is that they actually believe that there is a 50% chance that global temperatures will decrease by more than 0.30 over the next few years. Since I don’t think they actually believe that, it is safe to conclude that they’ve offered a suckers bet. Too bad. When Real Climate wants to offer a 50-50 bet in which the bettor gets to pick which side to take in the bet (i.e., the definition of 50-50) then we’ll know that they are serious.

24 Responses to “Real Climate’s Bold Bet”

    1
  1. John V Says:

    The dates for the bet were not made up by RealClimate. The dates are the exact prediction made by Keenlyside et al.

    As I read it, the same paper that you promoted last week predicts that 2000-2010 will be cooler than observed temperatures from 1994-2004.

  2. 2
  3. John V Says:

    One more thing — I was impressed by the way you took a bet based on 10-year overlapping averages and converted it to “wagering on a 2-year climate trend”.

  4. 3
  5. legion Says:

    Remember, these are the guys who are “revising” the averages and creating a moving target. They can make the “anomaly” anything they want it to be. You’d have to be a sucker’s sucker to take that idiot’s bet.

  6. 4
  7. jmrsudbury Says:

    Their bet is that 1994-1999 will have a *higher* average than 2005-2010? That seems to be against their pro-AGW position they have held for years. Have they switched sides? Their bet is not consistent with the ensemble of models used by the IPCC AR4! — John M Reynolds

  8. 5
  9. terry Says:

    that may be true john V but these are the same people who chastised Roger for taking a look at an 8-year time frame in terms of warming.

    Roger’s very right to be skeptical this time.

  10. 6
  11. Sylvain Says:

    I’m curious about a certain number of thing:

    Do they have preselected to which data the model will be compare?

    Would they bet on their own prediction?

    It is easy to say that one model is wrong the odds are largely against them.

  12. 7
  13. Jon Says:

    “stridently argued that the predictions in the Keelyside et al. paper are perfectly consistent with predictions of climate models in the IPCC.”

    Are you comparing forecasting runs with projections of the forced component of climate over 20 year periods?

    “However, when one such commenter here was asked to show a single IPCC climate model run showing no temperature increase for the 2 decades following the late 1990s”

    Can you please point out in the Keenlyside paper where they “show no temperature increase for the 2 decades following the late 1990s”? The paper I’m reading shows all future temperatures as warmer than the ‘late 1990s’ (1995-1999) for their forecast.

    Also, when you say “IPCC climate model run”, what do you mean? If you are referring to Fig 10.5, as I’ve said before, that is a projection of the forced component of climate where multiple runs are averaged to smooth out the very internal variability that Keenlyside et al are talking about. If you want an example of climate model that the IPCC used that is capable of producing what Keenlyside et al did, the answer is ECHAM5/MPI-OM, the model Keenlyside et al used. Whether or not one of the individual runs of the ECHAM5 model was initialized in the same way as Keenlyside 2008, I couldn’t tell you, but the fact that Keenlyside et al used the same model as the AR4 makes that question irrelevant, doesn’t?

    “It is strange to see the Real Climate guys wagering on 2-year climate trends when they already taught us a lesson that 8 years is far to short for trends to be meaningful.”

    They aren’t claiming a two-year period is “meaningful”. What point are you trying to make? I can’t seem to see what value this statement has other than as a non-sequitur jab at RealClimate.

  14. 8
  15. Jon Says:

    “they are playing with a stacked deck, which is what you do when looking for suckers.”

    Keenlyside et al are the ones that provided “the deck” via their own forecasts over their own chosen periods. If you want to call it stacked, then its own authors are the ones that stacked it.

    The suckers are those that are treating Keenlyside 2008 as though it is anything more than a short term forecast that many others don’t consider likely. I suspect that Keenlyside et al will not take the bet as they also realize that their forecast is not likely to be realized.

    This has the intended consequence of showing all those in the delay/denialosphere that its own authors aren’t even treating it as a solid prediction, so anyone using it to bolster their claims that the warming stopped/it’s cooling will appear as idiotic to everyone else as they do to those that know what is going on.

  16. 9
  17. jmrsudbury Says:

    Jon. It is tiresome to have to go back through the post and previous comments to get the context that you miss quoting. Your points would come through much clearer if you stopped fisking. — John M Reynolds

  18. 10
  19. Jon Says:

    The paper I’m reading shows all future temperatures as warmer than the ‘late 1990s’ (1995-1999) *of* their forecast.

    That says it a bit more clearly.

  20. 11
  21. Roger Pielke, Jr. Says:

    Jon-

    Some replies.

    1. “Are you comparing forecasting runs with projections of the forced component of climate over 20 year periods?”

    This has been asked and answered. Again, no, trend with trend.

    2. “Can you please point out in the Keenlyside paper where they “show no temperature increase for the 2 decades following the late 1990s”?”

    Sure, in their Figure 4, compare the red (observed) curve with the green (forecast). The late 1990s temps are not exceeded for about 2 decades.

    3. “If you want an example of climate model that the IPCC used that is capable of producing what Keenlyside et al did . . .”

    Um, no, but nice dodge. If you want to argue that 20 years of no temp increase is consistent with the projections of the models used by the IPCC, then you simply need to show one run with such an outcome. I don’t think you can. (And of course I don’t mean a model with the same name, I mean the specific model’s used by the IPCC in AR4, who are you trying to fool?)

    4. “They aren’t claiming a two-year period is “meaningful”.”

    Well, if they are making a bet on which the outcome depends only on a future 2-year period, then they sure are claiming it is meaningful.

    5. “delay/denialosphere”

    Yawn;-)

  22. 12
  23. Jon Says:

    “Again, no, trend with trend.”

    A forecast isn’t in and of itself necessarily a ‘trend’. And the forecast diverging from the ‘trend’ illustrated by the projections doesn’t mean much, which I’ll get to in a bit.

    “Sure, in their Figure 4, compare the red (observed) curve with the green (forecast). The late 1990s temps are not exceeded for about 2 decades.”

    Keenlyside’s forecast run shows no temperature lower than the ‘late 1990s’. The only way to say it forecasts lower temps is to compare observations with their forecast, which are already higher than the forecast itself. The Keenlyside ‘forecast’ run (in green) shows warming from the late 1990s on. The authors believe that their hindcasts are skillfull enough to make a prediction that future temps will more closely correlate with their forecast, which implies a cooling. The forecast itself is one warming from the late 1990s for the next two decades.

    Some may believe this to be an argument of semantics, but it cuts to the heart of not only the silliness of the confusion that this paper caused among the media, but it illustrates in part why RealClimate and others are right to doubt the forecast- it doesn’t match reality now, and it would take a lot of cooling to make observations match their predictions. Not likely.

    “Um, no, but nice dodge. If you want to argue that 20 years of no temp increase is consistent with the projections of the models used by the IPCC, then you simply need to show one run with such an outcome.”

    Roger, I am going to say this for what I hope to be the last time. The projections were created in such a way as to smooth out internal variability to show the forced component of the climate. The projections were never intended to capture the variability illustrated by the Keenlyside paper- quite the opposite. “[O]ne run with such an outcome” would be Keenlyside’s. Same model is used in the AR4. The only difference is the initialization. And as I said, it really is irrelevant whether or not one of the runs that went into the ECHAM5 mean showed a similar cooling as it would have been averaged out when creating the projections. The fact is that it obviously could have, it is the same model run under the same emissions scenario. The specific initialization is key the difference. Why is this a sticking point? If I went through the individual runs of all of the models and found one or even a dozen that were similar to the Keenlyside paper, it wouldn’t prove the point anymore than what I am saying now.

    This is like claiming that because a single roll of a die was a 1 and the mean of many rolls was 3.5, that a 1 is inconsistent with the die. The runs under different initializations were averaged. What in your own words do you believe it would prove if I did find a single run out of all of the runs that went into the ensemble projections that was similar to Keenlyside’s forecast? What meaning would that have? I say none.

    It confirms what we already know- the model is capable of such a run. But we already know that because Keenlyside uses one of those very models. All you can really hope to find out is whether or not one of the initializations of one of the model runs was similar enough to Keenlyside’s to produce a similar forecast.

    Think about it this way- what do you believe would change in the IPCC projections if the Keenlyside run was added to the ECHAM5 runs? Nothing. It would change the ECHAM5 mean by such a small amount that it would *barely* nudge the ensemble mean if at all. That is why I am saying that whole line of thought is meaningless.

    “Well, if they are making a bet on which the outcome depends only on a future 2-year period, then they sure are claiming it is meaningful.”

    You can’t possibly believe this can you? The authors of the paper stacked the deck against themselves. The people at RC are placing no particular emphasis on the importance of the next two years as any sort of significant trend in and of itself. The forecast itself is unlikely. It is unlikely not because any two year period has any significant meaning.

    “Yawn;-) ”

    You can “yawn” all you want. RealClimate is making its point quite well from my perspective, and neatly trapping those that make ridiculous assertions based upon a single paper- the ’suckers’ you mentioned earlier.

  24. 13
  25. Roger Pielke, Jr. Says:

    Jon-

    Thanks for your continued comments. I think we’ve come full circle. A few replies.

    1. Lets remember where this began. I asserted that nothing could be observed in the behavior of the global climate system that would be inconsistent with the predictions from models used by the IPCC.

    You now say:

    “The projections were never intended to capture the variability illustrated by the Keenlyside paper.”

    You ask:

    “what do you believe would change in the IPCC projections if the Keenlyside run was added to the ECHAM5 runs? Nothing.”

    You are incorrect, it would increase the distribution of outcomes defined as “consistent with”. If you do not believe this have a look at the Real Climate critique of Douglass et al., in which they use individual model realizations to define an appropriate spread to compare against observations. If the Keenlyside run were added to the ECHAM5 ensemble it would not change the mean by much at all, you are correct, but it would broaden the distribution of outcomes, meaning that even more observations would be”consistent with” this models projections.

    Again, back to my original point, the Keenlyside paper means that even more future observations can now be claimed to be “consistent with” climate models, including 2 decades of no warming from the late 1990s.

    I’ve played along with your questions for a while now, why don’t you answer my original question? What observations of the climate system to 2020 would be inconsistent (lets say at the 95% level of certainty) with the climate model projections of the IPCC AR4? It is a simple question. use global average surface temps from UKMET as the variable of interest if you’d like, since that is what we’ve been discussing, or use a different one.

    2. “Some may believe this to be an argument of semantics”

    Bingo.

    If today’s temperature is actually 90 degrees outside, but my local TV station predicted that it would have instead been 80 degrees, and that same station predicts 89 degrees midweek, one would have to have serious difficulties distinguishing between a model and the real world to conclude that the TV station is predicting a “warming trend”.

    Sorry, but real world observations are the baseline I use to project into the future. Keenlyside et al. forecast temperatures for around 2015 to be no higher than those observed in the 1990s. You can semanticize all you want on this point.

    3. If Real Climate is using this exercise to score political points in the blogosphere than their proposed wager is worse than I thought.

    You can just clear all of this up by answering my original question:

    What observations of the climate system to 2020 would be inconsistent (lets say at the 95% level of certainty) with the climate model projections of the IPCC AR4? It is a simple question. use global average surface temps from UKMET as the variable of interest if you’d like, since that is what we’ve been discussing, or use a different one.

    I’ll be happy to post up your answer as a main entry of the blog.

  26. 14
  27. Frank Bi Says:

    This is disingenious, Roger. You asked for a falsifiable prediction, and you got one, and now you’re giving excuses to ignore it.

    While we’re at this “falsifiability” game, I’d also like to ask: what observations would convince you that climatologists (on either or both sides of the “debate”) are not In It For The Money?

    http://tinyurl.com/42eutp

    – bi, International Journal of Inactivism, http://frankbi.wordpress.com/

  28. 15
  29. Frank Bi Says:

    This is disingenious, Roger. You asked for a falsifiable prediction, and you got one, and now you’re giving excuses to ignore it.

    While we’re at this “falsifiability” game, I’d also like to ask: what observations would convince you that climatologists (on either or both sides of the “debate”) are not In It For The Money?

    http://tinyurl.com/42eutp

    – bi, International Journal of Inactivism, http://frankbi.wordpress.com/

  30. 16
  31. Roger Pielke, Jr. Says:

    Frank Bi-

    What was that falsifiable prediction again?

    Thanks for the link, where you call me a crank and a conspiracy theorist. You’ve got me there ;-)

  32. 17
  33. Martin Says:

    1) I did the calculation with HadCRUT3 anomalies through Mar 2008, and came up with 0.17 degree C, but that is the same basic idea.

    2) If (i.) someone were a “temperature is a random walk” skeptic (on a monthly time frame) and the last two anomalies were the same as January (0.056), then it would be a good bet to take.

    3) And on the theme of 2) what would be nice is if someone would play Las Vegas between the, shall we say, differing opinions and offer up actuarially unfair bets with the following catch: If position A is believed, bet bet is a much better than a 50/50 proposition. And if position B is believe, the opposite bet is a better than a 50/50 proposition. Somebody did this a while back for regarding the speed of various horses, and well — the they say — the rest is history. Today’s bookie/epistemologist would no doubt offer this a Peer-testing, verification methodology for an NSF grant, but probably wouldn’t be allowed to keep the vig.

  34. 18
  35. Lupo Says:

    The bet is the paper’s 1994-2004 compared to the other two 10 year periods, is it not?

    See comment #3, it was already known to be a fair impossibility.

    http://www.realclimate.org/index.php/archives/2008/05/global-cooling-wanna-bet/#comment-86110

    In replies Stefan Rahmstorf makes at Real Climate it clarifies it all.

    “The green points, despite being connected by a line, do not represent one model run. Rather, each green point is an individual forecast starting from somewhere near the red line. That is why in the paper they say they predict a slight cooling relative to 1994-2004. You are right that already their prediction (or better to call it hindcast) for the 1994-2004 period was too cold. Otherwise, if you compare the green and black curves from 1999-2010, their evolution is the same, they are simply offset. So if you just took the relative change since 1999, not the absolute numbers as compared to the red curve, their new model would predict the same warming as a standard scenario run (i.e. the black one), which would hardly have been a reason to go to the worldwide media with a “pause in warming” prediction.”

    “We are absolutely not proposing this on the assumption that the authors are “denier sympathizers”. The authors are very good and respected colleagues, and this post is entirely about the ability to predict natural climate variability a decade ahead. It is not about anthropogenic warming, a topic on which we completely agree with Keenlyside et al. The short time scale of this prediction makes it amenable to have some fun with a bet, because the outcome will be seen in a reasonable time frame. If for some a bet is not “serious” enough, we will follow it with a serious discussion of the scientific issues shortly. We think framing this as a bet with specified conditions will help to clarify what exactly it is that the authors are predicting – after reading the paper at first this was not entirely clear to us, and it clearly is not entirely clear to many of the journalists reporting on it either.”

    “I think you understand the point we are trying to make. This supposed pause in global warming has been reported widely as if it were almost a fact, not a forecast, and as if this was widely supported by the climate science community, almost on a par with IPCC reports. Some articles framed it as if this new forecast now revises IPCC forecasts. If the prediction turns out to be wrong (which is what we think, and quite a few other climate scientists I have spoken to), this will damage the credibility of the whole community. This bet is supposed to signal to the public: on this decadal forecast the climate science community is not in wide agreement. In contrast to the global warming issue, where we have a wide agreement.”

    “We bet against their forecast as they made it in Nature. We did not pick the years.”

    “Actually, they made a forecast and took it to the media. We proposed this bet because we want to see how confident they are about it.”

    Gavin Schmidt – who is not listed as in the bet – got a bit of it wrong himself.

    “Oops. I was fooled by the green line in figure 4. That joins up different predictions but is not a trajectory itself.”

    He is quite clear what this is about.

    “Hint – this wager has nothing to do with who’s the better scientist, it is all to do with how new research results play out in the public perception. The forecast for the first decade has almost no chance of being correct, and the one for the second decade implies a degree of natural variability that is significantly larger than their model generates on their own or is seen in the obs. Yet, the forecast was seized upon all over the media as a likely, nay probable, truth.”

    Fear not, the next post is up.

    http://www.realclimate.org/index.php/archives/2008/05/the-global-cooling-bet-part-2/

  36. 19
  37. Frank Bi Says:

    I said,

    “You asked for a falsifiable prediction, and you got one, and now you’re giving excuses to ignore it.”

    Roger asks,

    “What was that falsifiable prediction again?”

    You know, you just wrote an entire blog post about one such falsifiable prediction… right above this very comment.

    “Thanks for the link, where you call me a crank and a conspiracy theorist. You’ve got me there ;-)

    So that’s what you have to say about the falsifiability of _your_ theory…

    – bi, http://frankbi.wordpress.com/

  38. 20
  39. Roger Pielke, Jr. Says:

    Frank- If you expect to have a serious discussion with someone, it is advisable not to start out with insults and silly accusations.

  40. 21
  41. Frank Bi Says:

    Roger,

    So falsifiability of theories is only an issue when other people aren’t saying unflattering things about you?

    I’d thought that you’d have given some thought on how to falsify your “climatologists artificially prolong the debate to get funding” theory, when you proposed it way back in _2003_. Were you waiting for someone to call you a “crank” so that falsifiability no longer matters?

    – bi, http://frankbi.wordpress.com/

  42. 22
  43. Roger Pielke, Jr. Says:

    Frank- Inventing a quote that I did not write and then attributing it to me is pretty poor form. If you want to have a serious discussion it won’t happen on these terms.

  44. 23
  45. Frank Bi Says:

    Roger, here are the facts: you proposed a theory about climatologists ( http://www.webcitation.org/5Xjh5U1sL ) _back_ _in_ _2003_, and even now (2008) you haven’t even bothered lay out any potential observations that can falsify it, and now you’re giving all sorts of excuses to avoid talking about that.

    So, what were you doing with your pet theory all these years? Were you thinking about how to test your theory, how to falsify it? Or were you waiting all these years for someone to label you, or misquote you, or quote you out of context, or whatever the latest excuse is, so that you can avoid talking about falsifiability?

    You can talk to yourself now, Roger.

    – bi, http://frankbi.wordpress.com/

  46. 24
  47. Roger Pielke, Jr. Says:

    Frank-

    Your reference is to this paper:

    Pielke, Jr., R. A. and D. Sarewitz, 2003. Wanted: Scientific Leadership on Climate, Issues in Science and Technology, Winter, pp. 27-30.
    http://sciencepolicy.colorado.edu/admin/publication_files/2003.01.pdf

    In it we asserted that:

    “Our position, based on the experience
    of the past 13 years, is
    that although the current and proposed
    climate research agenda has
    little potential to meet the information
    needs of decisionmakers,
    it has a significant potential to reinforce
    a political situation characterized,
    above all, by continued
    lack of action.”

    I am comfortable that this situation has persisted.

    We further assert:

    “The situation persists
    not only because the current
    research-based approach supports
    those happy with the present political
    gridlock, but more uncomfortably,
    because the primary beneficiaries of this situation include
    scientists themselves. Things are
    unlikely to change for the better
    unless the climate research community
    adopts a leadership role
    that places societal responsibility
    above professional self-interest.”

    What would such a leadership role look like?

    We answer this as follows:

    “Politicians
    are able to substitute research for
    other action because a large portion
    of the climate science community,
    particularly those focused
    on global and regional climate
    modeling and the earth and spacebased
    platforms to provide data for
    such models, continue to claim that
    more of their research will indeed
    lead to the reduced uncertainties
    allegedly necessary for policymaking.
    If scientists blew the whistle
    on this claim, its political viability
    would vanish.”

    Hey I wrote about that just this morning;-)

    We also said:

    “We support a robust, wellfunded
    basic research effort on the
    global earth system to provide a
    long-term base of understanding.
    But there should be no pretending
    that such research will be directly
    policy relevant, except in the obstructionist
    sense demonstrated
    during the past decade.”

    We say more, it is a good paper, thanks for highlighting it!