Armstrong has just offered Al Gore a $US20,000 ($23,000) bet that the Intergovernmental Panel on Climate Change temperature forecasts are wrong.
This passage intrigued me. What counts as “being wrong”? Here’s what I learned at the bet’s official website:
Scott Armstrong of the Wharton School challenges Al Gore $20,000 that he will be able to make more accurate forecasts of annual mean temperatures than those that can be produced by climate models. Scott Armstrong’s forecasts will be based on the naive (no-change) model; that is, the forecasts would be the same as the most recent year prior to the forecasts…
The charity designated by the winner will receive the total value in the fund when the official award is made at the annual International Symposium on Forecasting in 2018. (emphasis mine)
If I understand this bet correctly, Armstrong wins if 49% of the global warming predicted by Gore occurs. So while the bet is far from rigged, it’s less than ideal. If the issue is whether global warming will happen at all, the obvious bet is “if annual mean temperature goes up, you win; if it goes down, I win.”
Or if you prefer a smarter bet: If it goes up by more than 2 standard deviations, you win; if it goes down by more than 2 standard deviations, I win; otherwise the bet is off.
Even so, I’m impressed; as far as I can tell, no climate change skeptic has done more to put his money where his mouth is than Armstrong. My question for Armstrong: Will you bet with anyone other than Gore? Last time I checked, there was a queue of believers in climate change who couldn’t find anyone to bet with.
READER COMMENTS
Matt
Jul 9 2007 at 4:37pm
“Eight of the 10 warmest years since 1860 have occurred within the last decade.”
According to BBC reporting.
So Scott Armstrong would likely pick a global warming scenario if he used recent years as his model. If the recent trend is due to something like El Nino, as opposed to CO2, then Scott would pick temps warmer than the climate models, for I would suspect climate models consider El Nino to be temperary.
Shakespeare's Fool
Jul 9 2007 at 9:42pm
Bryan,
Since the long term average temperature of the earth is warmer than the current average, betting against global warming is betting against reversion to the mean. Why would any sensible person do that?
John
George
Jul 9 2007 at 10:39pm
Armstrong isn’t betting that there is no global warming, he is betting the the climate models are overestimating it.
TGGP
Jul 9 2007 at 11:34pm
Shakespeare’s Fool, although I haven’t read the actual terms of the bet, it would seem sensible that he is not make a prediction about long-term temperature on anything like the scale of the earth’s lifetime, but rather a more recent period where local effects can predominate.
Matt
Jul 10 2007 at 2:04am
“Since the long term average temperature of the earth is warmer than the current average,…”
Dunno how this statement got here, but according to my ice core data:
http://www.daviesand.com/Choices/Precautionary_Planning/New_Data/
The long term average, over the past half a million years is 4 deg cooler than the current.
I can tell this quote is partially made up because anyone interested would know that we have been in the hottest part of the glacial cycle for 10,000 years.
Bob Hawkins
Jul 10 2007 at 9:05am
Matt: your “long term” covers the current ice age (since it relies on ice cores). “Long term,” as in the whole history of the earth, can’t be derived from ice cores, since most of the time the only ice on the planet is at the tops of mountains.
Dan Weber
Jul 10 2007 at 10:38am
If “long-term” includes the times when the Earth was still a molten lump of rock, then, yes, we are unexpectedly cold.
But that’s just silly.
We are approaching the top of an oscillating temperature curve going back a half-million years or so. The questions are:
– will the top be a lot higher this time?
– if so, how much of that is because of humans?
– when will that peak occur?
(Graph of CO2 levels here; CO2 is at least correlated with temperature.)
Gorgasal
Jul 10 2007 at 11:18am
Armstrong is betting that a simpler model (the mean) will beat a more complicated one – this is why he would let Gore choose any non-black-box model.
That simpler forecasting models often beat complex ones, independent of the process being modeled, is standard knowledge in the forecasting community. Armstrong is trying to draw attention to the fact that no forecasters as methodological experts are involved in climate modeling and forecasting and that some quite basic principles of forecasting apparently are not taken into account. This is quite a valid point.
I attended Armstrong’s talk at this year’s International Symposium on Forecasting where he announced the bet officially. Kevin Trenberth, who had given a keynote speech on climate forecasting, tried to counter Armstrong’s arguments. Trenberth didn’t convince me, and my impression was that the ~80 professional forecasters assembled were more on Armstrong’s than on Trenberth’s side.
William Newman
Jul 10 2007 at 11:36am
I know hardly anything about climate models: anything involving fluid mechanics is pretty much a closed book to me. But once upon a time I did a lot of quantum mechanical modelling. Even when one’s main interest is one observable (like a will-it-be-stable question based on total chemical energy, or a policy question based on regional average temperature), in my experience a detailed model tends to have a lot of detailed spinoff predictions about other observables. In chemistry the detailed predictions might be things like chemical bond angles or ionization energies or magnetic moments. In climate models I’d expect lots of details too, dunno what, but at a guess, perhaps statistical patterns in fluctuations and correlations of expected temperature, humidity, and wind as a function of position, altitude, and time of day and year.
If a model has lots of predictive power (compared to simpler alternatives like “our best guess is that it will be the same as last year”), it might not be evident in a year or two of global temperature averages, but it could be much, much clearer in detailed properties like correlations in local time sequences. Does anyone know whether any of the bet-offerers are proposing bets on a big basket of properties like that? Or are the bettors just targeting the big politically charged numbers?
Shakespeare's Fool
Jul 10 2007 at 4:37pm
Give the earth a billion years to cool off from its “molten lump” state and 400,000 years is still one eight-thousandths of its history.
Going back to the end of the ice core tell us that it was warmer than the temperature that made ice caps possible at least long enough to remove the ice caps. (Unless there was, at that time, a shift in the way the earth rotated so the poles changed places.)
Comments are closed.