We are developing the social individualist meta-context for the future. From the very serious to the extremely frivolous... lets see what is on the mind of the Samizdata people.

Samizdata, derived from Samizdat /n. - a system of clandestine publication of banned literature in the USSR [Russ.,= self-publishing house]

A few thoughts on Climategate.

A few weeks ago, we were having one of many conversations on this blog about the subject of climate change. In the comments, I said the following

The climate is clearly changing. There is nothing unusual about this. The climate is always changing. I’m happy to concede that the trend in recent decades has been to hotter temperatures. Again, nothing unprecedented about that. The world has hot periods and cold periods. The trend seems to have slowed or reversed over the last few years. This is not a short enough period of time to prove anything, but it does make you wonder how strong the trend is. Some of the data analysis that purports to show the trend has been presented in ways that deliberately or otherwise state the data in such ways that appear to indicate the trend is stronger than it is, and/or choose starting points and data series lengths that appear to show the trend as more abnormal than it is, in my opinion.

Again, with the impact of human activity, I am happy to concede an impact exists. There is a lot of human activity – it must have some impact on the climate. Whether it is a significant impact is another question.

Having those two thoughts, you look for a correlation, and find one between CO2 in the atmosphere and average temperature. One can be found, although it is not clear whether it is a causal relationship (CO2 levels vary historically before significant human activity existed, and a lot of the time CO2 increases seem to trail temperature changes rather than the other way round).

So how much are higher temperatures caused by higher CO2 levels, and how much of the increased CO2 level caused by human activity? The answer to the last question is clearly “quite a lot”, but that is not an answer to the question “How much?” Is it “70%? 90%? 100%? 120%? To be able to come up with a meaningful model, we have to have a good numerical answer, and we don’t remotely.

As to what impact increased CO2 levels have on average temperatures, there is much greater uncertainty. Basically you have to enter a fudge factor into your model, see how well it models the past, and hope you can then model the future successfully. A few people have created models that can just about model the past, but that doesn’t mean you have the mechanism right – it just means you have found a mathematical function that fits the points on your curve.

As it is, we have a few extremely crude mathematical / computer models that suppose mechanisms that go from human activity to CO2 release to global warming. They don’t agree with one another, and they are incredibly crude. (The Earth’s atmosphere is an extremely complex system. These models only have a tiny fraction of its complexity). They have a poor record of predicting the future.

The science of global warming ultimately boils down to saying that “The level of warming is unprecedented”. “Human releases of CO2 into the atmosphere are unprecedented”. “Therefore, the second causes the first”. This isn’t an inherently ridiculous thing to say. If climate change really is unprecedented then we would look for other unprecedented things as likely causes and human activity would be the likely one. We could then look for mechanisms and solutions, but we would largely be doing so with our eyes closed.

I will listen to somebody who more or less says this and that the risks of global warming are so great that we must do something about them, but somebody who simply states that the science is settled and beyond discussion is frankly not even worth arguing with.

In response, I received a mocking reply from a true believer, saying more or less that if I knew so much about it, why didn’t I publish papers in a refereed journal myself, and he was sure that a Nobel Prize would be beckoning. There was no attempt to address anything I said – merely an observation that what I was saying did not have the approval of the clique controlling the argument.

In a way this was odd, because I was not actually claiming to know anything about the workings of the climate: only about the likely limitations of the methodology of climate scientists.

As it happens, once, in another life, I was a research scientist. At least, I did a Ph.D. in a field not a million miles away from climate science before departing for other parts of the economy. I have written one or two computer models of physical systems. And as it happens, they are hard. It is possible to use all the computational power you have in simply modelling something tiny: the vortices around the tip of an aircraft wing, say. As the systems you have become larger and larger you make more and more approximations and more and more assumptions that particular terms in equations are small and will be small in the future because they have been small in the past. There comes a point where models more from theoretical to empirical. You end up basically extrapolating from the recent past to the near future. In systems containing a lot of nonlinearity, factors that have not had macroscopic impacts in the past can suddenly flare up and become dominant in the future. Sometimes it is possible to figure out just when this will happen and add these effects to your models at the right time. Sometimes it isn’t. The earth and its climate is a huge system. In places it is highly nonlinear. It is horribly hard to model.

None of this is to say that well constructed models of such systems cannot be useful. However, they are inevitably uncertain. They are inevitably approximate. Everyone who has ever worked with one knows this.

A second thing one learns from working as a research scientist is that people in research labs resemble people in other workplaces. Petty fiefdoms exist. People stab one another in the back. Some people do better work than others. People will have different levels of respect for the work of other researchers. Some people rise to the top through doing good work. Others rise to the top through playing good politics (good researchers generally hate such people, but they none the less manage it). People at opposite ends of the corridor hate one another. If one is going to work in a particular team, one must work within the culture and beliefs of that team. In one’s work, you often have to start with whatever the person before you left behind.

In scientific research involving computer modelling and data analysis, this often leads to computer models consisting of layer on layer of code crufted on top of lower layers that are not well (or at all) understood. Data does get lost, or assumed to be correct because the previous person used it and there is no real way to verify it. Supposedly impartial journals do become captive of a particular point of view. People’s whole careers do become dependent on a particular interpretation of the results, and it then becomes very hard for them to back down. People become more and more certain of their results when the personal cost of abandoning them gets greater and greater.

However, once again as in any other workplace, good work still happens amongst all this. If there are six different cliques in different places, they will compete with one another until the truth comes out. If there are six different journals, then they won’t all become captive to the same clique, and eventually the one with the best and most meaningful results will become the most prestigious. There will be enough ability to move between teams that younger scientists will not necessarily be caught in a particular viewpoint because of who they work with. Politics will be horrible. Much bad work will be done. It will be a messy process, but it will generally be understood who does the best work, and the truth will come out. Really good researchers will be able to figure out what in the crufty codebase is good and what is not, and get meaningful results anyway.

This, fundamentally, has been my problem with the science of global warming – the denial of the messiness of it all. We have been told that “The Science is Settled” by men in white coats in ivory towers, and that we are “denialists” and unworthy of being listened to, if we dare to question the process or to state the obvious – that science is a messy and uncertain process and that as a consequence of being a very hard problem, modelling the climate is going to give answers with huge margins of error and huge unpredictability. (Nicholas Naseem Taleb would say it’s a system highly susceptible to Black Swans, and he would be right).

Which was why, when I was cc’ed on an e-mail last Thursday stating that there was a huge leak of data from CRU at the University of East Anglia, I pretty much knew what it contained and I haven’t been remotely surprised by anything we have learned. There is lots of politics, lots of bad work, lots of crufty code, and lots of uncertainty and disagreement.

The scandal here, is the pretence that this was ever not so. The careerist political side of this unit, mixed in with an unholy political alliance of Greens, Luddites, politicians with hidden agendas or at least vested interests in Climate Change being real, managed to create an environment in which the normal competition and disagreement between teams of scientists has not been allowed to take place. To even suggest that climate scientists behave like other scientists and to ask them to fully explain their work, has been to be opposed to their noble efforts to save the planet. “The Science is Settled” means that this is not necessary.

We have learned little in the last week that we should not have been already aware of, but perhaps now that it is out in the open, we can have a proper debate. This should hopefully be a relief, and if we are going to discuss policy for the whole world, it is required to be entirely out in the open.

Almost certainly, there are good researchers and good work being done in climate science, even at
CRU. Hopefully we now have the opportunity to identify it amongst all the cruft. Once it is out in the open that the science is not settled, we can do some good science. Good science in this field gives results with large margins of error. Remember that. If the researchers will not admit this right up front, then they are probably not worth listening to.

There is of course a second global warming related assumption that many of us have been called names for questioning. This is that a Kyoto Protocol type solution is the appropriate response to it. Bjorn Lomborg has bravely attacked this line, but much more time needs to be spent on properly quantifying the benefits of warmer temperatures as well as the negatives, and the economic costs of fighting climate change by slowing the speed of our technological civilization. Technical fixes down the line (when and if necessary) may well be a better strategy. This is particularly so if we acknowledge that we are very uncertain as to what is happening. And we need to say this, over and over.

37 comments to A few thoughts on Climategate.

  • chuck

    I am in general agreement here, but I think you understate just how poor the modelling skills of the big shots at the CRU appear to be. They use a limited set of techniques that basically assume time invariance and linearity and then when there are problems fitting the data, set about getting the data changed. Hence the divergence problem, the 40′s bump, the need to limit the medieval warming. These aren’t “problems”, they are indications that the model sucks. But these guys don’t have the skills to try anything else. I also think it worth pointing out that the only good temperature data set at this point is that from satellites; the land/sea records have been so corrupted by adjustments and poor quality control that they are pretty much useless.

  • That’s possible. I was going to say that the nature of the leak was what I expected, although the models were significantly cruder than I was expecting, but I think I left that sentence out.

    (Actually I have just gone back and put it in. No great harm in doing this, I think).

  • Vinegar Joe

    Personally, I’d much rather the climate get warmer than colder. Civilization doesn’t fair too well during an Ice Age.

  • steve

    I am getting sick of “gates”. It is probably too late for climategate, but I hope someone comes up with a better name for the next scandal.

  • Nuke Gray

    I don’t suppose you’d be happy to call the current storm of -gate names namegate? I thought not.

  • “I also think it worth pointing out that the only good temperature data set at this point is that from satellites…”

    Agreed – in fact, since the 2008 data became available it really hasn’t looked good for the AGW hypothesis at all. This is the axe I used whenever the topic came up in the last year or so.

  • K

    if we are going to discuss policy for the whole world, it is required to be entirely out in the open.

    I agree. At the very least what should come out of this episode is total transparency for climate research. All codes, notes, data and calculations must be published. The reasoning behind all “fudge factors” in all computational physics work should be explained. The record of hypothesis prediction and verification should be explicitly made public.

  • The Ambling Dutchman

    Well written article Michael.

    I welcome the opportunity for everyone to gather around and make some real science happen. Look at the data, create models, and kick the crap out of these models to see if they look like anything that happens in reality.

    I have this hope that the old “peer review” model will go the way of the dodo, and be replaced by a more “open source” model of scientific review, where subject matter experts do their research and their best thinking, and then make their data sets, their methods, and their complete computer programs available for all the world to see, so that e.g. a software developer expert can look at the code and review it, and a mathematician can look at the methods and review those, a philosopher can help them cope with their thinking and a politician can council them on their presentation of the facts.

    And all out in the open. We won’t completely get rid of hidden agendas, but at least science can be saved.

    What a wonderful world it would be.

    –GJ–

  • “..and a politician can council them on their presentation of the facts.”

    Mmmm… what exactly would be the point of that?

  • Maz

    Michael – excellent and lucid article, thank you. I see that the critic you referred to is now likening AGW “deniers” to Nazis…

    http://www.selfdestructivebastards.com/2009/11/don-be-good-german.html

    As usual, the science (not to mention the “fix” and the moral issues) is portrayed as being settled.

  • John B

    It seems MSM is trying very, very hard to ignore the information released by the emails.
    How about picketing Copenhagen??
    With slogans such as:
    Hockey lied, millions died
    Hockey lied, now you’ll be tried
    Al Gaia is a laia
    It’s getting colder as the earth gets older!!
    Light my fire!
    ??

  • Johnathan Pearce

    Maz, quite. Debating such people is pointless. One might as well ignore them or shoot them (well, in my dreams, anyway).

    Michael, very good post, as usual.

  • Michael

    Please tell us more about this adjective ‘crufty’ .

    Good piece.

  • Mike

    I had a sudden thought last night with regards to the “mainstream” response, particularly the self-serving response from UEA itself:

    it is rather like claiming that Lysenko was just a rogue element within Soviet biology, and in any case his findings are supported by the overwhelming majority of Soviet biologists working in many places around the Soviet union and its satellite states, so the Lamarckian consensus within Soviet science remains intact.

  • Alice

    Mr. Jennings — Gold Star for your post. One of the best, most reasonable, most accessible discussions of the topic I have ever seen. Please send it to the Guardian & the New York Times, and suggest they publish it. Seriously!

  • Frank S

    A very good article indeed.

    Part of the problem is that there are people who really want there to be a climate crisis linked to CO2. Such people, not least a Mr. M Strong, set up the IPCC in the first place as a political initiative to make the case for CO2 driving climate. The handful of scientists involved in the last stages of this UN-funded agitprop outfit number less than 60, some say less than 10. There has been no debate. There has instead been a clever and extremely effective hijacking of science for nefarious political ends. The debate should begin, but we need many enquiries. One I would like to see is into the leadership of the Royal Society and of the UK Met Office.

  • Michael has posted well in many ways. There is a lot of reasonableness and moderation in what he wrote. We are all human; nothing is perfect. It’s not easy to disagree with those sentiments, and I don’t. However …

    The issue here is really, IMHO, one of extent. In this, it really is close to the whole issue of Catastrophic Anthropogenic Global Warming (CAGW). Global warming has occured (though perhaps it has now stopped or reversed); there is theory to support an anthropogenic component, based on CO2 emissions. However, what sceptics like me doubt (and that is surely all sceptics do: doubt) is: (i) whether the extent of warming is (or will be) on balance catastrophic (and some warming in some places will surely actually be beneficial); (ii) how much of the warming is down to CO2, rather than other possible explanations; (iii) how much (including possibly none) of the CO2 increase is anthropogenic (especially as there is some evidence that warming causes CO2 release from the oceans, with an 800-year time lag rather than or in addition to CO2 causing the warming).

    The above are my primary questions; they are not at all easy to answer. There are a couple of others.

    Firstly, how much warming is actually occuring. We have measurements going back a long time for land-based meteorological stations. They are useful but do not tell us everything; much more energy is stored in the oceans than is stored in the atmosphere. Only since around 1978, from satellites, do we have more reliable and widespread information on solar irradiance and on surface and atmospheric temperatures. We then find worrying things like accusations of tampering with the evidence, such as this recent paper about New Zealand’s temperature records (h/t Bishop Hill, WUWT, and others); its fairly short: do have a read. Now, I don’t know if that report is correct, but I reckon it must be fairly easy to refute if it is not. Such undenied alleged tampering with the raw data must surely be a concern for us: not just for the accuracy of the record, but for all those involved in the alleged tampering, all those who fund and supervise them, and for all their motives. On this, I just struggle to see how we can be in a position where raw data and the immediate processing of it (for example to correct downwards any heat island effect from urban expansion to form the record for the surrounding non-urban area) is still open to doubt (but it is), and is not fully in the public domain (but it is not).

    Secondly, what can and should be done concerning lower-CO2 energy generation and economics. I’ll skip more on this, as my knowledge and useful contribution is more limited that on the scientific method and the computer modelling.

    We now get to the computer modelling. Like Michael, I have done a bit of computer modelling though not on climate, but with a different take on the acceptability of what seems to have been happening.

    Computer modelling, IMO, is quite difficult. In even small physical systems, the complexity can exceed our knowledge and separately exceed our ability to compute using the most sophisticated theoretical models; large systems are even more difficult. There will therefore be some modelling algorithms that have an excellent theoretical basis and some that are more ad hoc. On the ad hoc side, we come to parametrisation of (perhaps simpler) models as part of the whole. Care is needed to differentiate between sensible parametrisation of ad hoc models and more brutal and unscientific approaches, which might be properly labelled “fudge factors”. I have a view on this.

    Modelling by parametrisation of simpler real-world models requires a great deal of care, not to be fooled (even though one is a competent scientist) into believing the models show more than they do.

    Concerning climate models, what I think should happen is that raw data up to some time point (say year 19XX) is used to parametrise the ad hoc parts of the overall model. Then the accuracy of the composite parametrised model is checked by running the model to predict from year 19XX+1 to 19XX+some. That model’s output (extrapolating past 19XX) is then compared to the actual record of climate (all of it, not just temperature, and in varying and appropriate levels of detail). If the model’s predictions match well enough to the later actuality, the model is good. Of course, the meaning of a good match must be stated. Also, given extrapolation, one might expect the longer past 19XX, the worse the match. However, the error must be small; otherwise extrapolating into the future (which must surely have larger error) will be less useful.

    Now, all of this checking (often known as model validation) is often done in various stages. Care must also be taken, because validation using any period 19XX to 19YY can be viewed as informing the modelling process for future algorithmic improvements to the models, such that really validation of any improved model should only use years after 19YY. Given we are only at 2009, this does present problems. However, we can at least ensure that any model tuning (ie of the parameters) is done using data up to some year 19ZZ and model validation is only done using years after 19ZZ. This is sometimes called the requirement for strict separation of tuning and validation data.

    I ask, does this happen; show me.

    True: it’s not really practical for so much detail to appear totally in one or even several scientific journal papers. But there must surely be extensive tuning and validation reports, supported by input data and output data, that the individual scientific groups must have produced for their own internal validation purposes. Cannot these be made available? If there are requirements to protect ‘intellectual property’ (and sod knows why if it is paid for from the public purse), cannot (key parts of) the source code be kept private and subject only to independent review. Cannot all the input and output data be published, in detail, and a proper detailed report of the tuning and validation be published on the rest?

    And, if this separation of tuning and validation data is not practical for climate modelling, as some might claim, on what basis are we to place confidence in the predictions from such models: if they cannot predict what has already happened without ‘knowing’ in advance.

    I think I’m still waiting; have I missed something?

    Best regards

  • “I am getting sick of “gates”. It is probably too late for climategate, but I hope someone comes up with a better name for the next scandal.”

    I heard “Warmaquiddick” yesterday. :o)

  • Exactly! I agree in pretty much every particular, and with quite of lot of Nigel Sedgwick’s comment too. Indeed Lucia at Rank Exploits has shown that models that ran in 1999/2000 have done a particularly bad job in predicting the years since, so much so that in most if not all of them, the 2008 temperatures are now outside (below) the 95% confidence limits of the models.

    It does seem to me that we see a lot of cargo cult science (c) R Feynman in this area, and the related problemof confirmation bias in that results that prove warming are given less critical examination than those that don’t.

  • In scientific research involving computer modelling and data analysis, this often leads to computer models consisting of layer on layer of code crufted on top of lower layers that are not well (or at all) understood.

    This happens in commercial software development, too. I’m always having to deal with other people’s code and figure out what is a dodgy hack that covers up symptoms without fixing the underlying cause. The newly leaked CRU code is worse than any I have seen.

    But now matter how good or bad it is, I’m *amazed* that this source code isn’t completely out in the open as a matter of course. The kind of work the hockey team and the modelers are doing is entirely about the source code. Their results are meaningless without reference to it.

    There are people who wouldn’t run an operating system on their desktop PC without being able to see the source code (or at least know that people they trust with more expertise and more pressing reasons to be sure about such things *can* see it), never mind turn the world economy upside-down on the basis of the output of proprietary, closed-source software.

  • la marquise

    Some public-spirited person might like to translate Michael Jennings’ post and send it to the Figaro – which today has an article (the first on this subject) of mind-bogglingly obtuse mendacity (possibly that should be mendacious obtuseness). My mind is too boggled and I have difficulty in getting my Maw off the floor.

  • la marquise

    that should be ‘jaw’, obviously – though ‘maw’ would probably do at a pinch…

  • Sam Duncan

    The Ambling Dutchman:

    I have this hope that the old “peer review” model will go the way of the dodo, and be replaced by a more “open source” model of scientific review

    The irony being that the open-source model of software development has its origins in the scientific method.

    Rob Fisher:

    But now matter how good or bad it is, I’m *amazed* that this source code isn’t completely out in the open as a matter of course. The kind of work the hockey team and the modelers are doing is entirely about the source code. Their results are meaningless without reference to it.

    There are people who wouldn’t run an operating system on their desktop PC without being able to see the source code (or at least know that people they trust with more expertise and more pressing reasons to be sure about such things *can* see it), never mind turn the world economy upside-down on the basis of the output of proprietary, closed-source software.

    Me included.

    As far as I can see, the CRU’s work can’t be said to have been peer-reviewed in any meaningful sense at all. I’m not too clear on the details, but either the source was only made available on a nod and a wink to people they knew would review favourably, or it was never released at all. I’m no scientist, but I know bad science when I see it. My guess is that this is not standard practice, but if it is it has to change.

  • The Ambling Dutchman

    ..and a politician can council them on their presentation of the facts.”

    Mmmm… what exactly would be the point of that?

    Humour value?

    –GJ–

  • What Michael said, pretty much.

    I suspect it is happening and I know some people who do this for a living whom I trust who are pretty convinced.

    Whatever.

    Burning hydrocarbons for energy that we get from our “friends” in politically unacceptable areas is insane however you look at it.

    We need a LOT more nuclear power, we need some renewables, where it makes sense, we need new battery/capacitor technologies and a bunch of other technical stuff whether or not atmospheric based CO2 is a part of the problem.

    It seems that both sides of the climate debate are missing this.

  • Stonyground

    It has to be said that short term weather forecasting is now very impressive. If you check out the morning weather forecast for the day it is almost always spot on. The forecasters can even say when a particular band of rain, or whatever, is going to hit us and get it right to within an amazingly small margin of error.

    The problems start when attempts are made to predict the weather more than a day or two in advance. The predictions become less and less accurate the further ahead we go.

    I do understand that climate and weather are not the same thing, but surely the main differences are the size of the variations and the size of the timescales involved. It makes sense that the climate can be predicted further ahead due to changes happening more slowly over a longer timescale, but surely the same problems arise due to the fact that the subject that we are trying to predict is inherently unpredictable.

    Therefore I would say that there is far too much uncertainty involved to go around confidently predicting apocolyptic, biblical scale mayhem that we must all take drastic action to avoid.

  • Thanks Michael – Happy Thanksgiviong

  • Brad

    I have little doubt that even with all the models, well or poorly understood by those radicals preaching AGW, the prevailing attitude is driven by superstition. I read a transcript years ago of an “expert” on ozone depletion (AGW’s still born cousin) who after much testimony in front of some government panel or other boiled all the scientific chatter into one over-soul comment “what IF we’re right?”. It was still conditional. There was no hard proof REALLY. Just dream up an awful scenerio and ask “well, what IF we are right?” All the code, good and bad, and all the conferences, and all the press releases, and all the alarmist articles all boil down to a belief. A belief in a Hell waiting around the corner and it is that belief, that superstistion, that drives the politics. For all the egg-headery it is a superstition driven political machine. All the bible and verse of the thing is merely measuring stick of proving piety. For some it is reenforcing, others may just mouthe the words for gain. It functions just like any other religion over-grown and bureaucratized. But at its root it is a superstition heaven/hell dynamic. The code and models are merely the liturgy.

  • In response, I received a mocking reply from a true believer, saying more or less that if I knew so much about it, why didn’t I publish papers in a refereed journal myself, and he was sure that a Nobel Prize would be beckoning. There was no attempt to address anything I said – merely an observation that what I was saying did not have the approval of the clique controlling the argument.

    Because, as the leaked emails explicitly show, the controlling clique used intimidation and coercion to prevent any dissenting papers from being published?

    It’s the classic catch-22. “There’s no published dissent from our position.” Here’s my dissent that I would like to publish. “You can’t publish that, its unfounded.” Why is it unfounded? “There’s no published dissent from our position to support it.”

  • Phelps et al

    A while back I wrote (in this) about (among other more ancient things) a nasty little Bolshevik book which denounced all talk about the crimes of Stalin as mere gossip, because none of this “gossip” was in writing. Meanwhile, the government of the USSR shifted heaven and earth to prevent anything getting into writing, and destroyed anything they found in writing, documenting the all too real crimes of Stalin.

    Same trick here. For “crimes of Stalin” read “doubts about AGW”. For “in writing” read “peer reviewed”.

  • Don

    “Crime’ate data” is a bureaucratic/political issue apparently that won’t go away until government comes clean on the hoax. The BBC’s “The Great Global Warming Swindle” Documentary took a lot of anthropogenic data material to prove their position. I haven’t heard of any direct attacks on research data coming from “experts” suggesting that this transparent data in the documentry was “fixed” ! Let alone hidden.

    No counter argument means not having to prove your own data. The pro Gore element is, in my opinion, guilty of fraud.

  • John Q. Public

    Climategate

    Let’s see:

    a) subverting the peer review process
    b) stacking the UN IPCC
    c) obstruction of the Freedom on Information Act
    d) breach of university and state ethics codes

    … and we haven’t even talked about the data yet.

    Climate Science – the new Ponzi scheme!

    p.s. – Is this what Science is all about? Meet the new boss (science), same as the old boss (religion). When are they issuing funny hats to scientists?
    p.p.s. – Who needs Wall Street when you have Science?