Bishop Hill WaPo on climate models
Apr 7, 2010 Climate: Models Climate: RC The Washington Post has an interesting article on climate models which features Gavin Schmidt making a robust defence of their usefulness:
Put in the conditions on Earth more than 20,000 years ago: they produce an Ice Age, NASA's Schmidt said. Put in the conditions from 1991, when a volcanic eruption filled the earth's atmosphere with a sun-shade of dust. The models produce cooling temperatures and shifts in wind patterns, Schmidt said, just like the real world did.
If the models are as flawed as critics say, Schmidt said, "You have to ask yourself, 'How come they work?' "
Now last time I heard, the models could get into an ice age but couldn't get out again, so I'm not sure whether Gavin is being entirely straight with us here. Perhaps the models have moved on though, although one could still wonder if they could move so quickly from not being able to get out of an ice age to being useful.
Reader Comments (30)
I don't understand Gavin Schmidt's snark ""You have to ask yourself, 'How come they work?' "
Work? In what sense? They reproduce past climate (or rather weather) change because they are tuned to do so.
But the future? Is Gavin going to bet his pension on one of these model scenarios over even one year?
Of course not.
The trouble with models is the modellers have a lot of latitude when it comes to deciding whether they 'work'. For example, we only have data for a couple points on earth during the ice ages. This means they can make up all of the other data for things like ocean temps et. al. and claim the models works as long as the temps for the two points at the poles are close.
Early start today, yer Grace?
The models are built out of past events. And hey look everyone - they can predict those past events!
So I guess they do "work" - in the same way that the Bayeux Tapestry does "work".
Wow!! Climate model takes into account Ice Age! Ergo the sealevels will rise 20 feet by three weeks come next Tuesday.
Little Gavin has taken his first steps...we'd better enter him for the London Marathon next year. And give him the Olympic Gold 100m. Maybe a nice Nobel prize too?
Gavin dear heart, you have still a lot to learn. When you've finished tinkering with your playbox, try thinking about using your 'climate skills' on making some predictions that can be tested...not just reproducing the past. It used to be that science worked on theory ...prediction...experiment...confirmation. Only climate 'science' seems to have forgotten about the last three.
Of course their models 'work' - hindcasting is an exact science. But they tell you nothing about the future.
I wish they did; I spent many years betting on football matches, using ever-more complex models based on past results. A new model, new hope, better hindcasting, same outcome i.e. hopeless.
The Pinatubo eruption example is particularly poor because it only deals with one specific, limited, short-term causal relationship in the climate system. How does that tells us anything about the models' ability to predict other relationships, including the effects of greenhouse gases?
The models "predict" the temperature increase in the second half of the 20th century, but not the recent lack of warming, nor the peak around 1930-1940.
When someone presents evidence the first question is "What does it mean?". The second question should be "What haven't they revealed?". In looking at climate models a lot of time has been spent on the first question, looking at what modellers present, very little on what they don't show. The IPCC TAR has only two small diagrams (1/8 page) showing the accuracy of modelling temperature (9.5a) and precipitation (9.18a) anomalies. Have you ever wondered why they do not present temperatures as C relative to zero and precipitation as annual total in mm? The answer is this shows the models have not performed well.
In terms of temperature leading models differ from each other by almost 2 C. Have a look at:
www.climatedata.info/Temperature/Temperature/simulations.html
and in particular at figure 6.
In terms of precipitation models differ from each other by more then 150mm and from observed precipitation by an even larger margin.
www.climatedata.info/Precipitation/Precipitation/global.html
In both cases the implied difference in energy is much larger than the assumed anthropogenic forcing.
I suspect that there are so many adjustable parameters in these models that you could put in the price of potatoes around the world, the magnitude and location of climate-related research grants, plus the carbon-profits feedback loop, and still get an ice age out with a little bit of ingenuity. They may have some merit as illustrative devices, but predictive powers they have not. To argue otherwise is to be intent on deception.
In fact, as far as I know, there has not been a single prediction from the climate alarmists that has been verified. I ignore their fitting their models and/or their newspeak to existing trends or events. There have been, on the other hand, a great many contradictions of their models and their scare stories.
Give me any temperature curve you like and I'll be able to find a polynomial to fit it almost exactly. In fact, I can find an infinite number of polynomials. So there you would have a universe of excellent hindcasting models. But looking ahead they would likely be absolutely useless. There are an infinite number of models that will match past data pretty closely, but diverge wildly in predictive capability. How would Gavin know how to select from this infinite set? Is he a prophet? It is this type of basic philosophical problem that climatologists seem to ignore. Fitting the past is easy: predicting the future is uncertain. Claiming that because models fit the past then they must have predictive capability is as nonsensical as a historian claiming to be able to be a good futurologist because he has a good grasp of history.
As usual, SFT has hit the nail on the head. What fascinates me is whether Gavin really thinks that because his models have been tuned via an array of adjustable parameters to reflect the past, or at least the past as described by MIchael Mann and others' fiddled models, they "work", or alternatively he is just bluffing us. In either case what Gavin is doing is no different from all the risk managers and traders in the investment banks, with their wonderfully accurate models of the past, or the loss modellers in the insurance industry, with their constant revision to parameters to retrofit away their inaccuracy, and its output is no more reliable.
What do other people think? Is he trying to pull the wool over our eyes, or does he really believe the models work, just because they predict the past, having been calibrated to do so?
I second SFT and David S. Exactly what I thought when I read Gavin... though their comments were more precise than mine would have been.
The problem is that if you spend years of your life working on a sophisticated climate model, solving all the pde's, implementing state of the art advanced numerical schemes, factoring in more and more of the physical effects in a way which still allows you to solve the equations, scaling up the number of machines that run the model, writing papers about the model and its implementation etc... you quickly get a sense of wonder of the whole complexity of your creation. You basically fall in love with it because it is the embodiment of all of your greatest thoughts and efforts. And when you get to that point, you are no longer a scientist.
20/20 hindsight. How remarkable.
David S
Absolutely right on the financial models. A great quote from the August 2007 breakdown of the financial system concerned markets sustaining 1 in 1000 year events 10 trading sessions in a row. Sensible people might say that these are not, in fact, 1 in 1000 year events at all....
"With four parameters I can fit an elephant, and with five I can make him wiggle his trunk."
...attributed to John von Neumann by Enrico Fermi.
I found some interesting writings by Jerome Ravetz, father of Post-Normal Science, when I did my post on Climate Change and the Death of Science last October
http://buythetruth.wordpress.com/2009/10/31/climate-change-and-the-death-of-science/
Ravetz let the cat out of the bag: he and the likes of Prof. Mike Hulme, his acolyte, know that the models are simply metaphors, don't have predictive capability, and are designed to seduce those who create them to serve a wider political aim. He writes:
" …climate change models are a form of “seduction”…advocates of the models…recruit possible supporters, and then keep them on board when the inadequacy of the models becomes apparent. This is what is understood as “seduction”; but it should be observed that the process may well be directed even more to the modelers themselves, to maintain their own sense of worth in the face of disillusioning experience.
…but if they are not predictors, then what on earth are they? The models can be rescued only by being explained as having a metaphorical function, designed to teach us about ourselves and our perspectives under the guise of describing or predicting the future states of the planet…A general recognition of models as metaphors will not come easily. As metaphors, computer models are too subtle…for easy detection. And those who created them may well have been prevented…from being aware of their essential character."
Gavin is seduced and a seducer, and as Ravetz details, the use of models to describe or predict the future states of the planet is merely a cover for a political purpose - they are destined to fail, anyway, but it's important to prevent the modellers and the seduced scientists from perceiving this.
(1) "They reproduce past climate (or rather weather) change because they are tuned to do so." Mind you, the past climate has also been tuned.
(2) Nevertheless, it's news to me that they do "reproduce past climate": if so, the success must be recent. When I first took an interest in this whole farrago, they were not capable of it. So much so that various sorts of special pleading were advanced to explain why it was an unreasonable demand.
"You have to ask yourself, 'How come they work?' "
Even a stopped clock is right twice a day.
Does Gavin mean 'they work' in the sense that the that Jones hid the decline:
That they make the proper sales prop?
dearieme
Good point. If your retrodiction doesn't work, either you can tinker with the parameters or you can ask a friend to fix the historical data.
In the same way, if an IKEA part doesn't fit, either you can sand it to size or you can belt it with a hammer.
I'll believe the models when they put up temperature maps of the entire earth and predict every season correctly for the next ten years (in pdf format so they can't cheat and continually revise). Correct prediction means that they agree with the satellite temperatures, not the junk that's produced by the surface stations. Predicting the global anomaly is hoohey since you could have alligators at the poles and polar bears at the equator and still get it right.
Dearime
Have you read Gavan's piece done the Edge.org website in June last year where he talked about the models and mentioned Pinatubo and paleo-climate simulations?
http://www.edge.org/3rd_culture/schmidt09/schmidt09_index.html?
I wonder if the timing of the WaPo article is related to the guest post by Hiroshi L. Tanaka of the University of the University of Tsukuba on Roger Pielke Snr's blog on 30 March?
Back to the Edge and Gavin. This paragraph (3rd from end) is an indicator of the confidence the 'Team' had in their work and communications strategy before the November miracle posting.
"Over the past five years, I have spent a lot of time building up resources. We spend a lot of time building background for journalists, staffers, and for science advisors of various kinds. We're building up resources that people can use so that they can tell what is a good argument and what is a bad argument. And there has been a shift. There has been a shift in the media; there has been a shift in the majority of people who advise policymakers; there has been a shift in policymakers. This kind of effort — and not just by me, but also by other equally concerned people — has had the effect of elevating the conversation."
SFT beat me to what I would have said, had I not been asleep in California. However, he said it much better than I would have. Very nicely said SFT!
Models are useless unless they PREDICT what will happen. Haven't seen it this winter. My winter climate index, the Snow Shoveling Index (SSI) is at 5 for this winter. For the last 8 winters is was either 0 or 1, never higher. As for what happened in the east coast of the US and most of Europe, it looked more like a mini-ice age. Global Warming?
So just how does his marvelous model work?
Oops typo! Gavan should be Gavin.
The question is not whether the model can produce 1991 outcome when fed with 1991 input. The question is whether the model can produce 2001 outcome, or 2100 outcome when fed with 1991 input. Is the model good enough that when fed 2009 input, the 2010 output is so accurate that it can be fed back into the model to give 2011 output, still so accurate that when fed once again into the model it can give 2012, etc etc.
I think it's a typo. and a tone of voice thing. Should have been:
If the models are as flawed as critics say, Schmidt said,
"You have to ask yourself, 'How can they work?' "
Gavin's moment on the Road to Damascus?
There is no way the climate models work for the ice ages or for the Mount Pinatubo eruption.
For the Ice Ages, they consistently under-estimate the impact all that snow and ice had on the Earth's Albedo. Hansen builds in something like a 0.308 Albedo (versus today of 0.298) for the ice ages when it is clearly as high as 0.330. This is done so that they can keep the Temp C Change per watt/m2 up at 0.75C/watt/m2 so that the greenhouse forcing can still get to +3.0C per doubling for today's warming prediction.
It is fudged completely.
For the Pinatuba eruption, they build in a fake 'efficacy of forcing" factor so that they can get somewhere close to the actual temperature impact. Hansen estimates Pinatubo reduced solar forcing by 2.9 watt/m2 but the temperatures only declined by 0.4C (or just 0.14C/watt/m2).
Again, it is fudged completely.
You can program a model to do any damn thing you like. The question is, can they accurately (or, hell, even semi-accurately) reproduce known past climate from known initial conditions?
And the current answer is "no". As it's likely to remain.
The question to ask the modelers is "Does your model predict the medieval warm period - and be careful how you answer that?"