Andrew Rivkin writes about climate change for the New York Times. One of the stolen emails says:
At 17:07 27/10/2009, Michael Mann wrote:
Hi Phil,
…
p.s. be a bit careful about what information you send to Andy and what emails you copy him in on. He’s not as predictable as we’d like
In other words: Most reporters are predictable. Meaning they repeat what they are told instead of thinking for themselves. Otherwise there would be no need to say this.
Think about it. Michael Mann, a respected climate scientist, thinks that whatever line he and Phil Jones, another respected climate scientist, are pushing is so poorly supported by the evidence that they worry about a New York Times reporter finding holes in it! Independent thinking, even by someone without technical training, worries them! Really, it’s hard to avoid concluding that these guys are clowns, propped up by all sorts of people (journalists, Al Gore, many others) who benefit from a false certainty about this stuff.
Please, someone tell me: Why should I believe climate models? Have their predictions (not their fits) been compared to what actually happened?
I am agnostic about the proportion of human influence on climate change, and think it is plausible that much of what comes out of the IPCC is political and of dubious quality. It is true however, that we are having a significant influence on the composition of the atmosphere — CO2 is at its highest level in over 600,000 years, and given China is adding one coal-fired power plant a week to their electricity generation as combustion engines are adopted by developing countries, we are going to massively overshoot changes in composition that would occur through other means.
I share your concern about climate scientists’ credibility, but do you think it is reasonable to apply the precautionary principle here, and act to limit actions whose outcomes are unknown, but potentially catastrophic, for future generations? How do you think we should we treat such unknowns? The decision-making tools we apply for prediction are horribly unreliable, as our financial markets, foreign policy and weather-predicting abilities show. Humans tend to systematically underestimate risk by excessively discounting the future relative to the present, and I think multiple speculative financial market collapses attest to that tendency. Climate change seems a similar example, but the environment may not rebound as quickly as a market.
If it should prove to be the case that we cannot rely on climate models, do you have any thoughts on how we should proceed with policy? The consequences are almost all for future generations I think, so how we assess these issues depends on how we value their interests relative to our own and I would be interested to hear your take on that as well.
A lot of science reporting gets the science wrong, quotes scientists out of context & misrepresents their views, etc. What makes you think that they’re worried about Rivkin finding holes in the science, rather than this kind of stuff?
I tried to post this link in response to your previous climate post to a brief list of some ways in which climate models have been tested, but it didn’t go through. If you google “climate models are unproven” then the first link (to grist) will give you the page.
Why do I think they are worried about Rivkin finding holes rather than making mistakes? Because the term “error-prone” or “inaccurate” is used to describe people who make a lot of mistakes. “Not as predictable as we’d like” is used to describe people who think for themselves. Note the word “we”.
Thanks for the link about testing. Apparently James Hansen had not tested any predictions of his model before he presented it to the world in 1988. If he had, he doesn’t mention it in an essay about why we should believe his model.
Mike, I think a desire to reduce air and water pollution, and a need to not run out of fuel, and a need to avoid economic stagnation are plenty of reasons to reduce our dependence on fossil fuels. By a lot. Scaring us inappropriately is just one more way that other people take power away from us. Stagnation in American transportation is much older than worries about catastrophic climate change. The first hybrid cars are from Japan! That says it all.
Policy: I think government decision makers should grasp the concept of economic stagnation — lack of development of new ways of doing things. And fight it in all areas. All stagnation is a problem, in the sense that problems build up unsolved. The stagnation in health care is far more damaging to all of us than the stagnation in development of other sources of energy. As many have noted, the stagnation in health care is so bad it was a big reason GM went broke. It is starting to cripple the rest of the American economy. Not to mention the huge burden of depression, obesity, diabetes, etc. They are big problems right now.
It’s always very hard to deal with economic stagnation because any solution involves taking power away from the powerful (those who benefit from the status quo) and giving it to the less powerful (those who benefit from innovation).
“Please, someone tell me: Why should I believe climate models? Have their predictions (not their fits) been compared to what actually happened?”
James Hansen of Goddard Institute for Space Studies. I won’t link to the Wikipedia article since it will throw my post into moderation. But if you go there, scroll down to “Climate model development and projections”. I have a better article from Technology Review that I can send you if you are interested.
Weather models are constantly compared to what has actually happend. Models are run with old observation data to see if their outputs correspond to what actually happened with the weather. You can attempt to validate climate models in the manner described in the Hansen article.
I know you’re skeptical and I value your fresh ideas, but really did it not occur to you that people are constantly looking for ways to validate and improve computer models (not just climate models). Weather and climate models are among the most advanced because modeling has been used in that field much longer than it has been in other fields, and because there is such a huge body of observational data to draw on in order to perform these experiments.
Sorry I skimmed over the previous response. I’ll send you a copy of the TR article I mentioned.
I’d like to see that article from Technology Review, thanks.
“People are constantly looking for ways to validate and improve computer models.” That’s awfully vague. In psychology, modelers have spend the last 50 years pretending that fitting their model to data constituted testing it. All their work: worthless. I wonder if the same thing is happening in climate science. Sure, people like to improve stuff but they also like professional advancement, status, and attention. Sometimes the two types of goals (improvement and advancement) are in conflict.
Seth asks why we should believe climate models, and that’s a good question. In many areas, especially in science, a certain deference to authority is warranted unless we ourselves can take the time to truly understand something about the field. Climatology is one of those complex areas. But what Climategate shows, IMO,is that quite a few of the main climatologists were not doing disinterested science, but were political advocates, and thus they can’t be trusted. They are no longer authorities that we ca defer to.
I found this email more revealing than the rest. Sure, the consensus is overstated. Sure, they try to stifle dissent. But who would have guessed that these guys are afraid of a newspaper reporter? Perhaps we will learn that when they give talks at high schools, they refuse to answer questions.
I agree with your comments about economic stagnation and think you can add institutional structures with perverse incentives on top of that. Many of the problems in automotives and health care for instance are because the corporations and other institutions driving these industries lack incentives to pursue the best solutions for health or transportation problems, and are instead pursuing personal rewards, often based on share price or career progression. Until the institutions are reformed we will predictably see these significantly sub-optimal outcomes.
With health as an example, the patent system provides incentive to pursue pharmaceutical treatments for health problems, and to ignore preventive solutions or non-patentable treatments like nutrition. They then have an incentive to market these solutions, paying media companies who then have a disincentive to criticize them as it effects their own profitability, and the result is antidepressants and statins. ADs don’t show clinically significant results for any but the most severely depressed yet are widely prescribed, and statins don’t reduce all-cause mortality yet are the biggest selling drugs in the world. I think this is a predictable outcome of the institutional structures.
I share your doubts about climate science, but don’t find the email you cited a convincing example of insecurity in the quality of research on behalf of a climate scientist. There are other plausible explanations for Michael Mann’s caution to Phil Jones, such as concern that Rivkin was already partisan on the issue, or simply not a great science reporter who might misrepresent their findings due to incompetence. That doesn’t mean they feel their work is of low quality, just that they want it to be represented accurately, as they see it.
Similarly, I can imagine if a journalist didn’t understand the Shangri-La diet and misrepresented it in their writing, you might be cautious dealing with them in the future. We don’t know what Mann meant by “predictable” or what history he had with Rivkin that drove this comment.
Seth,
Two quotes for climate change part of the topic:
1. “Do not disturb a complex system since we do not know the consequences of our actions owing to complicated causal webs. … “leave the planet the way we got it”” Nassim Taleb
2. ” Don’t pee in the pool too much because soon we are going to notice” Old mentor of mine
Walter
The most ridiculous aspect of the global warming, I mean climate change, scam has been the prescriptive one. That alone should have been enough to alert sceptics to what was going-on.
I mean the idea that we humans know how to control the earth’s climate.
The idea, endorsed at the highest levels of international conferences, that we can choose a global temperature, and decide to hold the planet at that temperature by means of economic policy (think of those serious debates about whether ‘we’ should ‘allow’ the average temperature to rise by two degrees, maybe three? ).
Add to it this that we have a narrow window of time in which strong action can – luckily! – ‘save the planet’.
Well, wasn’t that fortunate! – that we discovered a climate process just in the nick of time, and that although the process has been supposedly going for some hundred plus years, our discovery allows that intervention can be effective in a political time frame of about five or ten years; just right for electing the next government, but not so long as to allow for any testing of computer simulations against observations…
If – in the space of just a few years – governments can be so persuaded of their omnipotence that they openly claim to have precise control of the global termperature, then I fear that anything is possible.
This puts King Canute into perspective. Stopping the tides by sitting in a throne with your hand raised is nothing compared with fixing the global climate at whatever temperature you desire by fiddling with economic policies.
Check out the actual programming comments in the code leak. This is bigger than watergate. I mean it. It’s more than bullying, it’s down right fraud and data manipulation. Pull up that old blog post you did on “Cargo Cult Science” and Biology you did a long time ago — YOU WERE RIGHT!
Thanks, Ted. This is my post on Cargo Cult Science and Biology:
https://sethroberts.org/2009/10/08/modern-biology-cargo-cult-science/
You mean I was right that the extent to which the public is deceived by scientists about scientific stuff is really large?
Seth, I e-mailed the article to your address on the website “contact” page. I had to straighten out something with my TR subscription before I could send it.
Mr. Bowerman’s two comments seem most directly on point to my concerns. I, too, agree that we have other reasons for moving from carbon-based energy, as Seth suggests. But in the end, we can have no doubt, can we, that we’re putting an incredible amount of carbon into our atmosphere. We (humanity) are like a kid with a chemistry set and no idea of what combination of chemicals may blow us to kingdom come. Climate is a complex system, and to think that we completely understand it or can control it amounts to folly. However, even in the face of ignorance, not to chose is to chose: and the safest choice remains significant reduction of carbon and other warming gases.
CCS wrote:
>Weather models are constantly compared to what has actually happend. Models are run with old observation data to see if their outputs correspond to what actually happened with the weather. You can attempt to validate climate models in the manner described in the Hansen article.
That is not science: that is curve fitting.
There is a general theorem that any finite set of data can always be fit with enough free parameters. Indeed, this can be done in multiple ways (I like Lagrange interpolation myself — the future extrapolations are just so wildly insane!).
To engage in real science, they need to make firm, unambiguous, statistically significant predictions of future data that they are willing to stand by. I.e., if the data does not meet their predictions, they have to admit their science was wrong, not just say that they will fiddle the free parameters a bit to get a better fit.
Anyone — i.e., the dominant in-group in climate modeling — who does not understand this is simply ignorant of science.
Dave Miller in Sacramento
Mike Bowerman wrote:
> I share your concern about climate scientists’ credibility, but do you think it is reasonable to apply the precautionary principle here, and act to limit actions whose outcomes are unknown, but potentially catastrophic, for future generations? How do you think we should we treat such unknowns?
We need to be careful with the word “catastrophic.”
The extinction of the human race or the destruction of the biosphere would indeed be catastrophic, to be avoided at almost any cost.
But, a slow rise in sea level, over a hundred years or more (and this is the sort of scenario even many of the warming alarmists foresee) is not necessarily a “catastrophe.”
After all, this has in fact actually happened over the last ten thousand years due to the end of the Ice Age.
The Little Ice Age from the 1600s to the 1800s also had a very real impact on human life, but it has not usually been viewed as a “catastrophe.” Nor has the quite substantial global warming from 1800 to the present (due largely to natural causes — i.e., the end of the Little Ice Age) been generally thought a “catastrophe.”
I’m not suggesting that future global warming, if it occurs (based on current evidence, I think that the big global warming was in the last 200 years, not the next 200 years), will not be a major inconvenience.
But the prospect needs to be evaluated rationally, based on standard probability techniques, standard cost-benefit analysis, etc.
There is not and cannot be a rigid, iron-clad “precautionary principle” or we would all have to live fifty feet underground to avoid being hit by meteorites!
Dave Miller in Sacramento
Dave, yes, I agree the only way to test a model is to make predictions with it, not fit it to data. The climate modelers don’t seem to understand this.
After posting an earlier comment, I realized that my statement
>There is a general theorem that any finite set of data can always be fit with enough free parameters.
would probably be misinterpreted.
A more accurate way of making the point would be to say that there are theorems that prove that it is possible to fit any amount of finite data by introducing enough free parameters of a very simple sort — for example, polynomial coefficients.
In practice, the distinction between my more careful statement and my more careless initial statement does not actually matter: anyone with much experience in simulations knows that if you introduce enough free parameters, you can indeed fit pretty much any data, even data generated to be simply random noise.
There are statistical techniques to guard against falling into this hole. There are also simple simulation techniques to avoid this: Feed randomly generated data into your parameter-fitting program. If the program can adjust the parameters to fit what is in fact random data, then it is not really modeling anything real at all: you just have a curve-fitting program of no scientific interest.
Alas, the dominant group in the global-climate-modeling community, based on both their published results and the leaked CRU e-mails, seem rather unaware of all this. Indeed, whenever their predictions turn out wrong, they seem rather pleased with the ease with which they adjust their models to the “new” data.
That is a profound problem. A model that can model anything really models nothing at all: no matter how wildly wrong its predictions turn out to be, the model can be adjusted to fit the new data that disagrees with the older predictions.
Since it can be adjusted to any new data, no matter how wildly that data differs from earlier predictions, the predictive value of the model is nil.
This is the fundamental problem with the global-warming work — CRU, the GCMs, etc.
Unfortunately, it is very hard to get this point across to non-scientists: if a model can model anything, this actually seems to many non-scientists a positive feature!
It isn’t. But how can that be explained to scientific illiterates?
Seth, you study how humans think, right? Any suggestions?
Dave Miller in Sacramento
Dave, no I don’t study how humans think. My mainstream psychology work is with rats.
Skeptics have reasonably focused on the data rather than on the models. If the data are scary enough you don’t need a model; and if the data are unscary enough, you don’t need a model then, either. I would like to first figure out what the data are before I worry a lot about models.