« Learning Russian | Main | Bad Science »

July 12, 2007

Timmy Elsewhere

Something at the ASI. There is at least a claim that the scientific forecasts from the IPCC violate all the rules about how you're supposed to make scientific forecasts.

July 12, 2007 in The Blogger Himself | Permalink

TrackBack

TrackBack URL for this entry:
https://www.typepad.com/services/trackback/6a00d8341c2d3e53ef00e0099198e78833

Listed below are links to weblogs that reference Timmy Elsewhere:

Comments

Tim:

If you're a betting man and have the cash to spare, then you'd do well to consider putting up a £10K sidebet that neither set of forecast proves correct to agreed measurement of accuracy.

The central problem in this whole debate comes down to the fact that while the core science of climate change is sound, i.e. greenhouse gases do cause warming, as does fluctuations in solar activity - and both can be demonstrated in simple terms under lab conditions - trying to apply that science to a large scale chaotic system like the global climate and derive accurate forecasts is a near impossibility due to the overall complexity of the system you're dealing with.

It all comes back to the possibly apocryphal story of Werner Heisenberg being asked what questions he'd ask god, to which he reputedly replied that he'd ask 'why quantum mechanics?' and 'why turbulence?' before noting that the he thought he might get an answer to the first one.

Even the most sophisticated climate models - or any other forecasting models for that matter - cannot handle the non-linear maths necessary for making accurate long-term predictions in such a chaotic system as the global climate.

A 10K sidebet would give you 2:1 odds on the money and, at worst, a even money chance of taking the pot, which is not a bad set of odds for a bet of this kind.

Posted by: Unity | Jul 12, 2007 11:47:19 AM

Unity, I doubt whether the fundamental problem is the nonlinearity; I think it's the ignorance. In two senses: first, the system is complex, and depends on relationships which are unknown, or, when known, that involve parameters whose values are unknown. Secondly, the practitioners can't learn much by confronting model predictions with experiments, nor even with observations, principally because the predictions they care about concern times far in the future. I'd describe the whole enterprise as risible, except that I think it's far worse - it's despicable.

Posted by: dearieme | Jul 12, 2007 2:27:54 PM

No, seriously, it is the nonlinearity.

Even if one can come to understand all the relationships and derive accurate values for the many unknown parameters, the sheer mass of nonlinear equations you'd have to deal with would render your efforts to accurately model global climate systems pretty much useless.

Complexity in modelling is relatively easy, its just a matter of having enough processing heft to generate results in a reasonable amount of time.

Non-linear and Chaotic systems is where it gets hard - we can't currently handle the maths for even fairly simple turbulance problems, never mind for 20-50 year climate projections.

Posted by: Unity | Jul 12, 2007 3:16:48 PM

We're just going to have to agree to differ, Unity.

Posted by: dearieme | Jul 12, 2007 10:55:38 PM

...except that hindcasting has shown the models match empirical climate data.

Posted by: StuartA | Jul 15, 2007 4:03:27 PM