Sunday
Jul032011
by 
Bishop Hill
Bishop Hill Material World on climate models
Jul 3, 2011 Climate: Models The BBC's Material World programme interviewed Prof Paul Valdes, a climate modeller. The message appears to be that climate models are very bad at reconstructing major climate shifts in the geological record and are probably bad at predicting future ones too.
The conclusion of the interview appears to be that it's worse than we thought. This struck me as slightly odd given that the rest of the interview appeared to revolve around the fact that the models don't tell us anything very useful.
Eduardo Zorita has further thoughts at Klimazwiebel.
Reader Comments (80)
Climate modellers are worse than we thought!
"The conclusion of the interview appears to be that it's worse than we thought. This struck me as slightly odd given that the rest of the interview appeared to revolve around the fact that the models don't tell us anything very useful."
Odd? Its the BBC.
The models don't recapitulate geological climate history because the models assume the wrong paradigm. He is right, in geological terms, there is something to worry about, but nothing to do with CO2 emissions. That something is the next full glacial cycle, recalling that, in general, each successive Quaternary cycle appears to be increasingly severe for the Northern Hemisphere cold temporate land-dominated high latitudes.
Stop Press: June's temperature in central England was the 254th warmest in the last 353 years.
http://junksciencearchive.com/MSU_Temps/Warming_Look.html
The warmest June on record was 1846. If only we had a time machine we could send Mr. Huhne back to 1845.
I think this is quite common.
Lindzen talks about the scientists he knows just getting on with their work, studying various aspects of the climate, publishing their results. A quick nod of the head to AGW is all that is required whilst "keeping their science straight" he calls it.
Its an interesting survival tactic - as long as in the end you say;
or, as in this case, if you work out that the models are a bit of a joke, then obviously it can be
or better than we thought(?) - I'd stick with worse, it'll get published, you won't be denounced as a denier, you'll keep your funding and there will be less hate mail.
Tip your hat to AGW and you can pretty much get anything through peer review. climate science "peer review" is more about screening out inconvenient papers than anything to do with the quality, as they claim.
So Valdes says, "we can't rely on what the models say - BTW that's really bad"
- Ok, wink ;), gotcha.
The article by Paul Valdes in Nature Geoscience is about the difficulty in modelling abrupt climate change. This in no was undercuts the value of models used to understand the likely impact of GHG forcing used by the IPCC and others. Replacing the term 'abrupt change change' with 'major climate shift' is a dishonest attempt to distort the implications of Valdes' work.
http://www.nature.com/ngeo/journal/v4/n7/full/ngeo1200.html
"Leap of faith" - yes, a religious metaphor just about sums it up.
@Mike
What's your point?
According to the article, 'abrupt' climate change of 20C at high latitudes ocurred over 'a few thousand years' during the PETM. Let's assume that was 3,000 years, that's 1c per 150 years. OMG we're in an 'abrupt' climate change since the LIA! The models can't handle it.
Paul was actually my superviser when I studied at Reading, a really nice guy. Another way of looking at 25 sudden warming events of up to 8 degrees in the last 120,000 years though is that life went on during all of those (during an ice age) and nothing seem to go catastrophically wrong during all those events. Perhaps, even if we did have a 'sudden warming' event then we might be able to cope. Also as pointed out in the article the models are useless for predicting the future anyway.
Mike
Please feel free to criticise, but could you avoid piling straight in with accusations of dishonesty as this is likely to lead to a food fight.
You speak as if we are talking about two different sets of models, but from the paper we are not - Valdes says "I argue that climate models of the current generation, as used in the latest assessment of the Intergovernmental Panel on Climate Change (IPCC), have not proved their ability to simulate abrupt change when a critical threshold is crossed."
You might argue that although the models can't capture abrupt climate change, they can still capture slower changes. But as Billy Liar has pointed out, this appears to be an abrupt climate change. It seems to me that the models have a problem.
In other words, the models have less intelligence than a blancmange and aren't able to predict the awful things we'd like them to, so very likely things will turn out even worse than that. I've heard about "arguing from ignorance," but I've never before seen arguing from ignorance of one's stupidity.
off topic - but should be interesting...
Mark Lynas commits green sacriledge - in 2 ways at once..
Writing in the DAILY MAIL.. is bad enough
Writing this in the Daily Mail, should start a green twitter frenzy.....
Daily Mail:
You mustn't believe the lies of the Green zealots. And I should know - I was one -
Mark Lynas, 4th July 2011
http://www.dailymail.co.uk/news/article-2010981/You-mustnt-believe-lies-Green-zealots-And-I-know--I-one.html
I wonder what Caroline Lucas & co will make of it..
To be honest Barry, I don't think Caroline Lucas reads the Daily Mail. Besides, that newspaper is schizophrenic on the subject of AGW.
I note that the Mail article has the usual photograph of power station cooling towers and the implication in the caption that these are polluting the atmosphere.
I suppose on the basis that water vapour is the major greenhouse gas and CO2 is considered to be a pollutant they could be said to be correct.
Sort of.
Lynas' article is actually very well reasoned (he can't be blamed for the sub-editors' obsession with cooling towers) but as usual he reaches the right conclusion for the wrong reasons. Nuclear power is an obvious choice but wind power is a waste for all the reasons we have all rehearsed before. (And see Booker in the ST yesterday; Some of his commenters are really quite intelligent.)
His demolition of the greenies' manic opposition to anything that smacks of technological development is well worth a read and will certainly give La Lucas and her acolytes a series of hissy fits!
That's because to them technocracy is the problem, not simply our energy source. But energy being the foundation of modern industrial society, is a very good target to attack. This goes back to the 60's, and possibly before then (Konrad Lorenz, for example).
"it's worse than we thought"
Which is the default position of the BBC, unfortunately. They're a news organisation and bad news sells.
Are "green hissy fits" an as yet, untapped form of renewable energy?
Can they go into implosive melt down?
Would any sane person be bovvered?
These models, when I knew a little more about them, were largely extrapolators of things that already exist - real things when they are used for weather forecasting, virtual things when they are used to illustrate hypotheses. To use an analogy, spotting an Atlantic depression heading for the UK is a bit like spotting a somewhat erratic bus heading along a road network towards you. The models help forecast whether and when it will arrive, and what sort of shape it will be in. They know where they tend to go, and they can look around at things that guide them such as jetstreams, and things that fuel them, mostly notably supplies of warm moist air. They do not take into account punctures, engine failure, traffic accidents, suddenly blocked roads, or the driver deciding on the spur of the moment to go somewhere completely different. On the bigger picture, they are also not any good at predicting the invention of the bus, or other vagaries of human and Mother nature. And that's now one well-mixed metaphor.
Barry Woods
Off-off-topic: the same subject, with the same Lynas (he has a book coming out you know) is rehearsed in the throbbing heart of warmism at
http://www.guardian.co.uk/environment/2011/jul/02/green-movement-lost-its-way
where greenery is rescued from the clutches of the left (where the subsidies are) and handed over to the right (where the profits are). Or something.
The fun is in the comments, where the Greens can be seen tearing each other apart, to the joy of one or two sceptics who haven’t yet been banned.
I'm reminded of the comment that anyone who is certain his horse will win the Derby is mistaken — even if it wins! I hope that's not too abstruse. I think I understand the idea!
golfcharley, I reckon if you can find a way to tap green hissy fits as an energy source you've got it made. They're just so easily renewable. All you need do is repeat the key phrases, "nuclear power", "GM crops", "Mark Lynas" ...
Models are merely faintly instructive toys, as useful as a clockwork model that may hang from a nursery ceiling that demonstrate our planets revolving about the sun, which when one has learnt from it that the planets orbit our sun we realise that's where that particular model's usefulness ends.
'Scientific' models of dubious utility are part of a religious belief system and have little to do with grown-up science. Childish things should be left behind with childhood, but the urge to play with ever more complicated toys seems to linger on into what should be adulthood.
John @11:36 AM, your metaphor is an interesting one, and raises some interesting issues. The difficulty is that there are different degrees of difficulty in coming up with models for different problems. Science has a pretty good track record of coming up with successful models of some things. For example, we can model planetary motion well enough to know when solar eclipses are going to happen - indeed, that is allegedly a successful prediction made way back in the early days of science and philosophy. There are lots of other even more complicated things that we model quite well. Yet there are other things where models seem to work a lot less successfully - e.g. for the share price of Northern Rock (click on the 'Max' button below the graph). So where does climate modelling fit in with this? The article and radio program that are the subject of this post suggest that it is closer to the share price example on long timescales. And sceptics would actually go further, and point out that the success of models in predicting (as opposed to hindcasting) climate on the decade time scale is not that great either.
The conclusion drawn by Valdes is that because abrupt changes of climate have occurred in the past, apparently in response to quite small stimuli, and because radiative transfer physics suggest that CO2 does have some influence on climate (both things that many lukewarmers and sceptics accept), then things might even take a nastier turn in coming decades than models predict. That sort of conclusion is one that you commonly hear coming from the consensus side when you point out that uncertainty about climate is large. We had it from dana1981 on the recent thread about John Cook. The truth is, it is not a completely worthless argument to make. Sceptics or lukewarmers nevertheless tend not to believe this way of thinking, and, if pressed, also try to argue that they are not being inconsistent when claiming that we don't understand the climate well enough to model it, and also that the climate is unlikely to change significantly. I think that is because we have an opposing "model": "Next year and next decade are likely to be much like this year in climate terms". As a model, it has a lot going for it (cheaper to run - no Met Office supercomputer needed, good track record of getting things broadly right, based on historical records), but it is not perfect by any means. E.g., as the paper by Valdes points out, it is a model that might have got you into trouble in the Sahara some few thousand years ago. So the interesting question that crops up is, given the imperfections of the climate models, sophisticated as they are, do we definitely trust the seat of the pants instinct that things are likely to be OK more than we trust the models? My view is yes.
The interviewer makes a truly ignorant statement early on. He comments that climate models are typically tested against the last one hundred years - largely true, although very selectively tested - but then goes on to insist that this is typical across all areas of scientific research.
Nope. Not in my area of research. We always test models against out-of-sample data. Testing in a hindcast mode will almost always give misleading results for a number of reasons.
I wish they had a scientist from another research area - they would have put him straight on that one very quickly.
A final observation - we already have a good explanation for the geological scale variability in climate; Hurst-Kolmogorov dynamics. Of course, H-K dynamics does not fit the IPCC narrative so doesn't even get a mention in Paul's article.
Spence_UK, in my field too nobody would take seriously the agreement of a model with data that is all in-sample - any fool knows that goodness of fit can be just a product of diligence in fitting, and with multiparameter non-linear feedback models (as in climate science) you don't even need to work very hard to get a good fit. Models of this kind are only really tested on out-of-sample data, which is why future projections are so interesting, and why the disappointing agreement of 1990s models with 21st century climate data is such a scandal.
Computer models of climate are mainly useful as a propaganda tool to further the CAGW scam.
"Computer models of climate are mainly useful as a propaganda tool to further the CAGW scam."
Jul 4, 2011 at 2:20 PM | cosmic
Oh really? Perhaps you would talk me through how you would monitor global climate without using modelling?
Yes, really.
It isn't monitoring, it's the predictions far into the future, these models are claimed to be able to make.
@ spence and nicholas
I do not believe the practice you discuss would ever be tolerated in climate psyence.
First, because the psyence is already settled, and second, because "a scientist from another research area" would by definition not be competent to contribute to, much less critique, climate psyence and would be denied permission to speak.
"Yes, really.
It isn't monitoring, it's the predictions far into the future, these models are claimed to be able to make."
Jul 4, 2011 at 2:39 PM | cosmic
What a surprise - someone on this messageboard who can't answer a question when the untenability of their position would be exposed.
Why don't you actually answer, and explain how you'd monitor climate without using modelling?
What do you mean by monitor? Explain why this monitoring would need a computer model? Explain how the existing models are useful in this modelling with the assumptions they are making.
I think you are simply someone who plays with words.
"The fun is in the comments, where the Greens can be seen tearing each other apart, to the joy of one or two sceptics who haven’t yet been banned."
Jul 4, 2011 at 11:47 AM | geoffchambers
Can I suggest your obsession with the Graun's messageboards isn't very healthy? It's almost all you seem to comment about, mentioning more often than not that you've been banned.
Also, on the article you link to, the extreme comments seem to be those from Hilly Billy types, which in this example. are childish, disingenous and often hysterical. As well as frequently just thuggishly unpleasant.
Z, you monitor the climate by measuring it with thermomemters. No models required.
"What do you mean by monitor? Explain why this monitoring would need a computer model? Explain how the existing models are useful in this modelling with the assumptions they are making.
I think you are simply someone who plays with words."
Jul 4, 2011 at 2:47 PM | cosmic
Ah - the good old 'answer a question with a question' technique employed by those who are frantically wriggling on the hook.
Let's keep it really simple, and you could explain to me how you'd be able to tell whether the planet's climate was getting hotter or colder over time, without using models?
"Z, you monitor the climate by measuring it with thermomemters. No models required."
Jul 4, 2011 at 2:51 PM | PaulM
You really, really don't know much about this subject do you.
monitor — vt to act as a monitor to; to check (e.g. for radioactivity); to track, to control, to watch, check, or supervise.
Why would you need a model to monitor something? You would monitor climate (or anything else) by observation. Even you should be able to work that out, I would have thought.
Models are only of use as a means of testing a hypothesis and then only to a very limited extent since they are totally dependent on the assumptions that have been made when programming them.
Models are very poor at providing any useful information about systems which are largely chaotic.
See Jeremy Harvey's post above for more information on the subject. He seems to know what he's talking about.
ZedsDeadBed
Sorry Zed, I'm curious as well about what you mean about "monitoring" - do you mean weather forecasts?
More than you Z. And at least I know the meaning of the word 'monitor'.
Come on then Hilly Billies - you claim to be sceptical types. I've just exposed a rich seam of real ignorance amongst some of the people who post here. Let's see if any of you are genuinely sceptical enough to put your own house in order before looking elsewhere.
"Let's keep it really simple, and you could explain to me how you'd be able to tell whether the planet's climate was getting hotter or colder over time, without using models?"
Statistics.
Z, the only ignorance you have exposed is yours of the English language.
"More than you Z. And at least I know the meaning of the word 'monitor'".
Jul 4, 2011 at 2:57 PM | PaulM
What is it they say? There's none so knowledgable as those that know nothing about a subject.
"Let's keep it really simple, and you could explain to me how you'd be able to tell whether the planet's climate was getting hotter or colder over time, without using models?"
Statistics."
Jul 4, 2011 at 3:01 PM | cosmic
And that would give you enough information, and be able to process it all would it?
@zed
I'm curious as well. Please explain how and why monitoring of the climate could be done with models? The term monitoring implies observation of current conditions - you cannot monitor something that does not yet exist, and if you monitor a computer simulation, you are monitoring the computer simulation, not the reality it purports to simulate. I, like everyone else here, am talking about GCM's used in climate prediction by organisations like GISS and the Hadley Centre, not computational tools used to interpret and compile temperature readings from various sources. Of course, you're intelligent enough to already know this, and you wouldn't just post a meaningless and opaque question (albeit one with a sheen of credibility in the use of climatological jargon) to make a totally redundant rhetorical point.
And no, it's perfectly reasonable to reply to a question with a question if that reply is a request for clarification, otherwise how could your question be answered? I don't see anyone changing the subject here, just people confused as to what exactly you are talking about.
Jul 4, 2011 at 3:31 PM | JJB MKI
There are people on this board so utterly clueless about climate science, that they claim models aren't necessary for studying the climate. No hind or forecasting, just passively assessing it.
This is of course completely impossible. But the silence of people not rushing in to correct them is deafening, and speaks once again of the almost total lack of true scepticism of people who post on this website.
This may be a bit o/t, but I was wondering about the modeling that produced the original graph used by the IPCC, (before the hockey stick was invented), of temperature through the last 1000 years. This showed the MWP and LIA quite distinctly.
I have a book, (Lost Villages of Britain, Richard Muir), which uses this graph, and attributes it to M.L.Parry, who’s career is:
Until September 2008 he was Co-Chair of Working Group II (Impacts, Adaptation and Vulnerability), of the Intergovernmental Panel on Climate Change (IPCC) based at the Hadley Centre for Climate Prediction and Research, U.K. Meteorological Office. Previously he was Director of the Jackson Environment Institute (JEI), and Professor of Environmental Science at the University of East Anglia (1999-2002); Director of the JEI and Professor of Environmental Management at University College London (1994-99), foundation Director of the Environmental Change Institute and Professor of Geography at the University of Oxford (1991-94), and Professor of Geography at the University of Birmingham (1989-91).
He looks like a prominent warmist, but I haven’t seen his name on any climate blogs. Unfortunately, Muir’s book didn’t give any hard reference to where the graph came from.
Does anybody have any idea?
Also, has anybody asked a medieval historian their opinion on the hockey stick?
To return to some more interesting issues, I recommend that anyone interesting in model performance take a look at this post on Science of Doom. As Spence and Nicholas point out, its best to test a model by comparing its output with data that was not used in developing the model. In many fields, this can be done by making predictions ahead of time - which in this field would mean predicting what the global average temperature will be in 20 years time. Hansen famously made some predictions back in 1990 or so that are not looking so great now that we have been able to monitor the actual temperatures. Though there is room for (some) reasonable arguing about which of his three predictions should be taken most seriously, and how exactly it should be compared to actual outcomes. From about 2000, there were many more models and you start to get things like multi-model means. As people like Lucia delight in showing, the actual temperatures seem to be deviating quite a bit from the model means also.
OK, but if we want to know NOW how well the models work, having to wait another 10 or 20 years to see how it pans out is not ideal. The science of doom page I linked to above shows that we can assess the models in other ways, by looking at how well they reproduce other properties than global average temperature. Even in hindcasting mode, it will be harder to tune models to reproduce variables such as the difference in mean temperature, at a whole set of places around the world, between the hottest and coldest month (standard deviation of temperature – “over the climatological monthly mean annual cycle”), or the modelled vs. experimental albedo of a section of the earth's atmosphere (Reflected Solar Radiation), or annual rainfall, and so on. If you look, you'll see that if you're not feeling so generous, the models are typically not very good for such properties. That has to raise questions as to how well the models perform even for the last few decades, and suggests that the agreement for global temperatures may reflect tuning rather than model "skill".
Err Zeds, I know you're just a troll but I think you misunderstand what a model actually is: it's a conceptual representation of a real entity. This is all fine until you realise that it's only as valid as the concepts it encapsulates. That is to say, for example, that a model of atmosphere might not be valid unless a something such as turbulence is modelled. Or the behaviour might be substantially different at different resolutions. Or if you don't understand enough about clouds, or the oceans, or the Sun, you might not be able to adequately incorporate their effects into your model.
So a model might be ok for testing certain scenarios, but it only predicts what you programmed it to assume in the first place. This is why the word "prediction" and "model" should not be used together. The word "theory" or "hypothesis" and "prediction" can be joined, of course, but that is not the same thing.
In conclusion, model output is not empirical evidence.
"In conclusion, model output is not empirical evidence."
Jul 4, 2011 at 4:04 PM | Robinson
Actually Robinson, although that is technically correct, I think it might be you that doesn't fully understand it.
Some people here are claiming that you don't need modelling for looking at climate and noting any changes over time. This is incorrect. As soon as you do introduce a model, then one is moving away from empirical evidence in the strictest sense, but what a lot of the real heid-bangers who inhabit blogs like this one fail to realise, is that just saying 'thermometers' is useless for climate analysis. There's a lot more to it than that.
I am still chuckling about Zed stating that it has found a rich seam of ignorance here when it does not know the meaning of the term 'Monitoring'. Monitoring can be done in many ways, Zed. I monitor my weekly spending by checking on what I have spent and the items I spent it on. I used to monitor the progress of my students on an almost continuous basis. I monitor my own bodyweight every week. I monitor the weather using a max/min thermometer, a rain guage and a barometer. I monitor the daily BBC weather forecast for accuracy in my post code by charting the forecasts and assessing them against my own readings. Despite owning and using two computers, I have never found a use for maths and/or computer-based modelling except for 'predicting' the marks of my former high school students, a practice riven with error and uninformed supposition which frequently seemed as foolish as the UK Met Office's former Long Range Forecasts.
So please enlighten me, Zed, considering that you regard me as mired in this rich stream of ignorance, please lay out for me a dictionary definition of Monitoring. I eagerly await my induction into the throngs of the non-ignorant.