Sunday, April 15, 2012

Misunderstanding Trends


A couple of posts back (see Maunderings on Trend Projection) I talked about the difficulties inherent in attempting to project complex trends into the future.  As a student of history (and a fan of Asimov's "Foundation" series, which is based on the conceit that it may some day be possible to develop a science of long-term historical trend projection) it's a topic that's near to my heart.  As I put it in that post, when trying to figure out how a complex trend may evolve into the future, "The best we can do is work from a comprehensive knowledge of the past, and make reasonable, parsimonious, short-term predictions based on every last scrap of data we can muster."

Sometimes it's difficult, though, to illustrate clearly the problem with attempting to project trends into the future.  The biggest problem lies in trying to project trends that are influenced by factors that you don't understand and therefore haven't taken into consideration.

Then I found some pictures.  Back in 1985, Ebony magazine published an "artist's rendition" of what Michael Jackson, that year's hottest pop star, would likely look like 15 years hence.  I think the artist was influenced a little too heavily by Billy Dee Williams and Lionel Ritchie, but whatever; Mike was a good-looking kid, and, ceteris paribus, there was every reason to suspect that he'd grow into a good looking forty-year-old.

The problem with trend projection is that ceteris is never paribus.

Figure 1 - Michael Jackson, from 1985-2000: Trend Projection vs. Empirical Reality

The gulf between what Ebony thought would happen, and what really happened, could hardly be more startling.  Without casting aspersions on the character and conduct of a chap who, in his youth, was unquestionably a great artist, let's just say that there were probably a few variables in the Jackson equation that the editors of Ebony either didn't understand, or didn't want to incorporate into their equations for fear of upsetting a multi-million-dollar applecart.  Ignorance obviously played a factor - the only question is whether the ignorance was unwilling, or wilful.

Sound familiar?  Well, it does to me.  Here's another example.

Figure 2 - Global temperature, climate model projections vs. empirical reality

That chart (obtained from Steve Goddard's inimitable RealScience blog) is taken from James Hansen's late-1980s prediction of what was likely to happen to global temperatures based on human greenhouse gas emission trends.  Scenario A - the highest black dotted line - was what Hansen called "business as usual", based on a 1.5% per annum increase in atmospheric CO2, and a 3% per annum increase in aerosols.  Scenario C - the lowest black dotted line - was "zero emissions", i.e. no increase in CO2 emissions beyond the prediction date.  The red line is the temperature trend.  The red dot is the current average global temperature for 2012.

In other words, since Hansen made his climate model-based projection, the global temperature trend has been consistently well below the projected temperature for the zero-emissions scenario.  Problem is, human CO2 emissions have continued to skyrocket.  In fact they are pretty close to Scenario A.  This chart is the climate science equivalent of predicting that Michael Jackson would evolve into Billy Dee Williams, and instead watching in horror as he gradually transforms into an extra from the "Thriller" video.  And for the same reason, too: there was something wrong with the model.  Through either ignorance or wilful dismissal of obvious controlling variables (the Sun in the case of climate, and complete, lip-flipping, python-snuggling, skeleton-purchasing, looney-tunes insanity in the case of Michael Jackson), the models failed catastrophically to accurately project the outcome of the trend.

Bottom line: you can't model a trend if you don't understand the variables that influence it, and the test of any hypothesis - and that's all that any model is, an hypothesis - is how well predictions match empirical observations.  If your projections end up looking NOTHING AT ALL like measured data, then it's time to revisit your model.  One wonders whether the staff at Ebony ever sat back in 2000 and, marvelling at Michael Jackson's attempt to gradually polymorph himself into Carol Channing, thought to themselves, "Boy, did we get THAT one wrong." 

James Hansen has taken a slightly different approach.  Instead of questioning his models, he simply uses his influence as head of NASA GISS to modify the historical temperature record, revising past temperatures downwards and recent temperatures upwards in order to prove that he was always right.  In other words, instead of revisiting the models to try to come up with projections that more closely match the data, he changes the data.

Hansen's approach - like Stalin's - is apparently to convince us that Michael Jackson has ALWAYS looked like one of Stephanie Meyer's gaunt, pasty-faced, sparkly vampires.  But he didn't; he once looked pretty normal.  Similarly, most of the warming in the US temperature record appears to be largely an artefact of these sorts of adjustments to the temperature record, rather than the measured temperature record itself.  This is one of the reasons that I personally do not subscribe to the notion that everyone ought to agree that the Earth's average temperature has risen over the past century.  The fact is, we don't know whether it has, because the records are no longer reliable.  That's not "conspiracy-theorist" contrarianism, it's simple historical skepticism.  The University of East Anglia Climate Research Unit has already admitted that it no longer maintains the unaltered, original, measured temperature data.  And it doesn't help when people like Michael Mann go to extraordinary lengths to design models that allegedly "prove" things that we KNOW, from empirical records, really happened - like the Medieval Warm Period and the Little Ice Age - didn't exist.  There is no substantive or ethical difference between rewriting history to disappear the MWP, and airbrushing photos to "disappear" Leon Trotsky.

And besides, how can we trust the temperature record when it is owned, maintained, and subject to regular revision by individuals who stand to profit from altering it?  That's like writing an historical treatise on the Napoleonic Wars based solely on Bonaparte's autobiography, or trying to understand the Caesar's invasion of Gaul using no primary sources other than De Bello Gallica.  These are not disinterested individuals.

Come on, folks - would you believe Al Capone's book-keeper, or would you want an independent auditor to look the numbers over?  Especially if Al wanted you to spend a trillion dollars on something he told you was absolutely necessary, based on his accounting?

You can't model a trend if you don't understand it, but you can certainly alter data records to make your model look more accurate.  Maybe the editors of Ebony should take a leaf from Hansen's book and spend all their time altering the contrast on all of the pre-2000 photographs of Michael Jackson and photoshopping in that epic chin-cleft. 

Could take a while, and be really expensive; there are an awful lot of pics of the Unigloved One out there.  And Ebony's editors don't have Hansen's US government sinecure, a guaranteed six-figure include, or endless time on their hands.  Besides, they don't have a motive; after all, their entire weltanschauung, and their funding, doesn't depend on whether they can convince the world of the accuracy of their model for predicting the evolution of pop star physiognomy.

But Hansen's does.