Models, Computer Models, and Fantasies (Updated)

If the report from the Daily Mall is true, it may comprise one of the most outrageous abuses of computer modeling in my knowledge:

Britain’s airspace was closed under false pretences, with satellite images revealing there was no doomsday volcanic ash cloud over the entire country.

Skies fell quiet for six days, leaving as many as 500,000 Britons stranded overseas and costing airlines hundreds of millions of pounds.

Estimates put the number of Britons still stuck abroad at 35,000.

However, new evidence shows there was no all-encompassing cloud and, where dust was present, it was often so thin that it posed no risk.

The satellite images demonstrate that the skies were largely clear, which will not surprise the millions who enjoyed the fine, hot weather during the flight ban.

Hat tip: Amba

Normally, since it’s overseas news this would be fodder for a post of mine over on OTB but, since we’ve been discussing modeling here for the last couple of months, I thought I’d post it here.

I don’t oppose mathematical models. I’m not opposed to computer models. As I’ve written before I took independent studies in computer and mathematical modeling and developed my first computer model going on fifty years ago. But models have limits.

They must be based on observation, rooted in a knowledge of the underlying principles involved, and calibrated and re-calibrated with real, live, sampled data. And taken alone they do not prove that something exists; only that it is possible that something exists.

My questions about the stimulus package are based on the apparent unwillingness of the economist supporters of the tactic to provide real data. Here, for example, is an apparently data-based study that finds a weak stimulus effect for deficit spending for fiscal stimulus. I have seen other data-based studies that find no stimulus effect whatever. I have seen no data-based study that supports the idea of a strong stimulus effect from deficit spending. I have, however, seen lots of models.

I also think that one of the problems that fomented the fiscal crisis was an excessive use of computer models without the moderating effect of common sense. Another thing I’ve commented on before: how the use of computer models, operating in lock step, can exaggerate a real life problem. And the incentives support operating in lock step.

Update

Former metallurgical engineer, present private equity firm partner, and frequent commenter Drew makes this helpful contribution:

Jet engines have combustion and then cooling sections. The combustion chamber is of sufficient temperature to liquify silica based ash. It then cools and precipitates (in globs) in cooler sections disrupting the air flow. Second, if still solid, these particulates travelling through the engine are moving so fast that they act like a sand blaster, destroying (like right now) the surfaces they come into contact with. This all with apparent “clear” skys.

As I understand it the bottom line is that you can’t determine the quantity, nature, danger, or even the presence of the ash in the air from the ground by visual inspection or from looking at satellite photos (at least using natural light). Clear skies don’t necessarily mean anything.

However, models don’t necessarily mean anything, either. To determine the actual state of affairs you’ve got to take real measurements and use them to calibrate your models. Back when I was a student I have great confidence that most of the engineering students in the country and many of the practicing engineers would be busily trying to figure out how to measure the amount, composition, direction, and so on of the ash from the Icelandic volcano. I guess things have changed.