[The very last thing I wrote for Earth Island Journal.]
Useless Arithmetic: Why Environmental Scientists Can’t Predict the Future
by Orrin H. Pilkey and Linda Pilkey-Jarvis 256 pages, cloth, $29.50 2007, Columbia University Press, New York
What do fisheries collapse, beach loss, invasions by destructive exotic species, and sea level rise have in common with nuclear waste disposal and open-pit mine toxic leakage? As Orrin Pilkey and Linda Pilkey Jarvis point out in Useless Arithmetic, each of the above environmental issues provides an example of the failure of one of the most important items in the environmental scientist’s toolbox: the mathematical model.
Do your eyes glaze over at the mention of mathematical models? You’re not alone. In fact, once the modelers get involved in an environmental issue, grass-roots citizen involvement gets shoved aside. When the models get brought in, those of us without an extensive grounding in math above the basic algebra level might feel a bit out of our depth.
And that would be just fine, if a bit intimidating, if the mathematical models accurately described the world. There is a place in the world for experts, after all. But those models don’t often describe the real world with any accuracy. In fact, mathematical models are often so divorced from the real world they are intended to describe, and the figures on which the models are based often chosen so arbitrarily, that basing decisions on those models can actually hasten the rate of damage done to the environment. At a minimum, modelers often incorporate “fudge factors” into their models to achieve the results they were looking for in the first place.
Take Yucca Mountain, the US’s proposed permanent high-level nuclear waste storage facility in the Mojave desert in Nevada. Faced with the prospect of assuring the public that the dump would remain leak-proof for 10,000 years into the future – later expanded by court decree to one million years – environmental scientists, geologists, hydrologists and engineers took it upon themselves to render mathematically the entire mountain, its structure and surroundings, the flow of water around and through it, and a host of other climatic, biotic, and geologic factors in order to determine whether highly radioactive waste placed in storage within the mountain would stay there for the requisite number of millennia. Imagine writing the math to cover every layer of rock in the mountain, unseen except through sampling and remote sensing, and trying to account for every hairline crack. Then, using that data, predict how the mountain will change – down to that same hairline crack level – after 10,000 years of earthquakes in the seismically active Mojave.
The Yucca Mountain modelers’ chief concern wasn’t the cracks in the mountain, though, so much as the water that would leach the wastes through those cracks. The local water table is 2,000 feet below the surface, and the mountain is in a basin with no outlet to the sea. But 12,000 years ago there were gigantic freshwater lakes in the Mojave, and not too long before that water flowed from what we now call Death Valley, a similar basin only a few miles west of Yucca Mountan, into the Pacific Ocean. What will the climate be like in the Mojave in another 10,000 years? Any claim that one can model local conditions that far in advance is just ludicrous.
Pilkey and Pilkey-Jarvis are no Luddites: they concede that modeling has its place, as long as each model’s assumptions, and limitations, are stated up front. Their readily accessible book should be read by any activist who’s ever had to face off against the opposition’s engineers.
=v= Sounds like an interesting topic. Modeling from the Club of Rome to Ken Caldeira could use some scrutiny.
A longstanding worry of mine is the public’s lack of grounding in statistics. Anything with lots of variables—and that includes all of ecology and all of society—is easy to distort, but good statistical math can detect that. Unfortunately, most people just throw up their hands, grumble about the dishonesty of “statistics,” and believe whatever anecdotes they were going to believe in the first place.
I suspect that, after enough similar flops, the public will react to mathematical models with the same sort of cynicism and nihilism. What can we do?
I just concluded another frustratingly fruitless online search for a very interesting book that I read some 25 years ago (and whose title I’ve forgotten) that spoke to the roles of “hard” and “soft” sciences. (And in this context biology and geology were “soft”.)
I think it was written by a couple of geologists and covered a number of the usual scientific issues (I recall Kelvin/Age of Earth and Wegener/Continental Drift) where “harder” but incomplete mathematical models and considerations proved less accurate (and had a chilling effect on progress) than well-argued geological and biological points of view. They even formulated a (or mentioned an established) “law” along the lines of “Hard science tends to drive out soft, even when the soft is more appropriate.” I do wish I could re-locate it.
In my experience, even when the explicit assumptions and limitations of mathematical models are acknowledged, people tend to fall down on the implicit ones, such as you illustrate. The tension between the two approaches is undoubtedly healthy, what isn’t is when one perspective is granted an unwarranted privileged position.
While I take your point (and that of the book), I think the case is just a bit overstated. Yes, models can be ‘tweaked’, but scientists don’t routinely put in fudge factors to “get the results they want”.
When the models are published in a scientific journal, the assumptions are explicitly stated, as are the sources of data that have been plugged into the model. In subsequent popular reporting, things can be a mess. Scientific journalism these days is, for the most part, crap.
Trying to model what might happen at Yucca Mtn. for tens of thousands of years is clearly absurd. The point of the model-building exercise might even be to point out the absurdity of generating wastes that will be toxic for those spans of time.
There are models and models, and certainly some overreach. But I think “Useless Arithmetic” is not a fair characterization. It depends so much on the scope of what’s being modeled. Tomorrow’s weather vs. the weather next January 12th. But just going through the exercise of building (and testing it against real life!) is important in getting to know what parameters really influence complex systems, and feedbacks among them.
That said, I will end with a sort of quote, whose author I forget. Something like…
“There are systems of sufficient complexity such that the only predictive model for that system is the system itself.”
- sgage
It’s at moments like these when i wonder about the value of models at all. I don’t doubt that they’re helpful in thinking through certain kinds of highly defined, limited-variable situations, but…
So much of science - hell, toss in policy and culture too - these days seems embedded in the notion that you don’t have to go out and interact with anything real - engaging with a virtual version is not only sufficient, but superior.
It really worries me, this idea that there’s a generation of people growing up thinking that the world is a theme park or a video game, because the actual physical world is something they don’t see as any more real than the limited experiences of their everyday, comfortable lives. It’s like we’ve decided to hide in a cocoon with our fingers in our ears and our eyes shut, pretending that if we wish hard enough, the world will go away - only now we’re getting at the stage where we’re in the cocoon and we’ve forgotten that there is a world out there, a world with its own dynamics, a world that doesn’t play along with our tidy little fantasies.
You close with the authors’ comment that modeling does have its place. I’m a hydrogeologist, and I’ve seen many ground-water flow, contaminant-transport, and other models that have been quite useful. Obviously, how well a model works depends on how well it accounts for the factors involved in governing the situation being modeled. For instance, a model that assumes aquifer isotropy and homogeneity might fail significantly in describing certain fractured-rock or karst settings (perhaps scale-dependent). Done poorly, modeling can be worse than useless; for some situations, modeling by itself is insufficient. Nevertheless, modeling remains a useful tool for scientists and engineers.
It’s at moments like these when i wonder about the value of models at all.
Rana, don’t tell these guys. They’ll be deeply offended ;-)
While I haven’t read the book yet, I did read the New York Times review, and am familiar enough with Pilkey’s work to know that this thesis is the natural progression of a career of critical analyses regarding modeling of coastal erosion and dynamics along the southeastern seaboard in particular. And that the frequent target of his criticism is the Army Corps of Engineers. Pilkey’s main point is that the dynamics of sediment transport and wave energy in a transport-limited, highly variable environment is susceptible to so much non-linear response to minor perturbations that even highly robust models have a tendency to fail to predict major trends as often as not, and that we’d be just as well off just dumping a bunch of sand offshore and seeing what happens. As I recall this was done recently at Ocean Beach in SF with relative success, and a great deal of equivocation on the part of the planners, who weren’t sure what was going to happen. Pilkey is an outstanding maverick for bucking the trend toward ever-more-detailed modeling of processes that that are insanely complex and subject to the whimsy of nature. His West Coast counterpart is undoubtedly Jeff Mount at UC Davis, who is also noted for being highly critical of Corps of Engineers and state-generated plans for confining rivers and the Delta behind levees, and thinking that you can model your way out of the coming catastrophe. His point, as well as Pilkeys’s, I believe, is that ultimately “Nature bats last”, and its better to be a realist about the nature and extent of the “problem” (induced by human assumptions that the environment can be molded to our desires), than to think that fine-tuning model parameters will provide our engineering works the factors of safety necessary to weather the unforeseeable outcomes of high-magnitude, low-frequency events. Add in shoddy execution related to politics and corruption (synonymous?) a la Katrina, and all the elements of failure and recrimination are in place. Far better, say Pilkey and Mount, to place oneself AWAY FROM THE HAZARDS (novel concept), and allow the areas with the greatest dynamics (coastal barrier islands; floodplains) to adjust and buffer impacts as they have designed themselves to do, and we would be be in a far better place. The likelihood of governments and engineers adopting this philosophy is, however sadly, close to zero though, I think.
I read the comments above by Fred, and they are exactly right on. The problem, though, in response to his observation:
“The likelihood of governments and engineers adopting this philosophy is, however sadly, close to zero though, I think.”
is not that this philosophy is not adopted by government planners and engineers—they are right there with us—rather, it is that policy is dictated by politicians who respond to their constituents (read “contributors”). These constituents (ahem) are developers, who look for short-term profits, much like politicians only care about the next election.
The other institutional problem is insurance. When the hurricane or flood comes, and the development that should not have been there in the first place is removed during Nature’s last at-bat, everybody gets paid and they rebuild to do it all over again!
This is insane, and yet we keep on doing it, over and over again. Why? Because some people like it this way, because they make a lot of money.
Grrr.