It is time to stop quivering in our boots in pointless fear of the future and just roll up our sleeves and build it.
- Ray Pierrehumbert

Friday, June 4, 2010

Reality vs Model

Many readers will have seen the scary movie showing the Gulf oil spreading into the Atlantic.

I ask you to compare the movie against this widget at the New York Times which shows actual historical behavior of the spill.

What you will see is that the loop current is continually pulling a filament of thin oil out to sea (comparable with the UCAR simulation movie) but that the filament is continually dissipating. In the simulation, that phenomenon is not visible at all.

My understanding of what is going on is that they initialized the standard NCAR ocean model with observations, continually injected passive tracer at the surface, and drove it with climatological winds.

I'm only slightly less sure that the biggest problem with the simulation is that they neglected the oxidization of the hydrocarbons by various processes.

If you actually watch the oil slick pattern, it's clear that the parts that grabbed by the loop current get thinned out enough to pretty much vanish a day or two later.

I think it would have been better (and just as easy, though anything with these models is not easy, running these codes is like going back in time thirty years as far as software ease of use goes) to have a simple half-life on the tracer. That would also be wrong, I would guess, but not as wrong. I don't think treating the oil as a passive tracer yields a good picture.

My wife would not like it if I made a substantial bet, but I'd happily bet that there will be no impact of any consequence anywhere in the Atlantic, say, north of Key Largo.

This is not to say that I am happy with the situation in the Gulf. But I'm starting to see the problem that many people have with others' eagerness toward predictions of catastrophe. So far there is no sign in the real world of anything like the model's scenario happening.


manuel "moe" g said...

> But I'm starting to see the problem that many people have with others' eagerness toward predictions of catastrophe.

I wish there was an easy way to talk about and visualize a hierarchy of concerns. Capable humans naturally understand a hierarchy of concerns, where lower concerns only come into play once higher concerns are reasonably satisfied (like my concern for hydration is below my concern for respiration, and both of these are above my concern for oil-free aquatic bird life). And sibling concerns may be substitutable or complementary and have a ranking or command a fraction of a whole.

The mainstream media handles concerns like a pre-schooler on a sugar rush - one big shiny new concern can push out all others, irrationally and ineffectively, rendering the news consumer incapable.

Maybe a little web widget that displays the hierarchy of concern, interactively. A hierarchy, running from the mundane but individualistic corporally viscerally urgent all the way down to the global political abstract. And is easily updatable with new information. And tools to meaningfully compare between hierarchies, of different people or the same at different times.

And tools to draw attention, to yourself (and embarrassingly to others), that inconsistencies exist (I am much more interested in finding out inconsistencies and conflicts between my concerns than I am in maintaining the illusion of perfect consistency from yesterday to today and the illusion of omniscience). Because mundane actions have global political abstract consequences, and the global political abstract imply certain mundane actions, and it is tempting to conveniently compartmentalize these two extremes, and conveniently ignore contradictions and inconsistencies.

Getting back to lacking models leading to overstating the potential for catastrophe -- it is hard to put appropriate demands on models without the context of this hierarchy of concern. Because the actionable decisions based on predictions of the models is a function of the demands put on the model during construction and manipulation and evaluation.

ac said...

anything with these models is not easy, running these codes is like going back in time thirty years as far as software ease of use goes

And CCSM is one of the better ones imo.

In the UCAR sim movie it's clear that the oil that ends up out of the gulf is right at the bottom of the dilution scale. It would be interesting to know the cutoff below which no tracer was rendered.

EliRabett said...

wrt to this, there is the old chemists saw, dilution is the solution to pollution. The risk is where the concentration is high, not where it is spread out.

wagelaborer said...

Prediction of worse catastrophes, like assignment of blame, is a distraction from the obvious conclusion that should be drawn from this ongoing catastrophe - that we need to switch from single person oil-fueled automobile transportation, with its attendant urban sprawl and massive on-land death toll.
We need to go back to the public transportation system that we had in the US before GM, Standard Oil and Firestone got together with the Federal Government to destroy it,and turn to the private automobile as a means of people transportation and trucking as a way to transport goods.