# The Current Crop Of Computer Models

…are close to useless:

…here we see a major problem with IAM-based climate policy analysis: The modeler has a great deal of freedom in choosing functional forms, parameter values, and other inputs, and different choices can give wildly different estimates of the SCC and the optimal amount of abatement. You might think that some input choices are more reasonable or defensible than others, but no, “reasonable” is very much in the eye of the modeler. Thus these models can be used to obtain almost any result one desires.

How (politically) convenient.

As I retweeted, they’re worse than useless to the degree that people trust them for policy decisions.

## 17 thoughts on “The Current Crop Of Computer Models”

1. George Turner

I like to think of the models as using unconstrained equations, which are the most versatile kind of equations because you can make them whatever your heart desires.

Last month Jo Nova had an interesting piece about how the climate models produce quite different outputs when run on different hardware platforms or with different compilers. As one commenter there said, “In any numerical (finite math) simulation run long enough, the rounding errors become the result.”

2. Paul Milenkovic

I think a person needs to be a little more focused in their critique than his article.

Non-linear dynamical systems are known to exhibit a property known as chaos. One quantification of this property is something known as the Lyapunov Exponent http://en.wikipedia.org/wiki/Lyapunov_exponent.

What this means is that you run an exact rendering of the dynamical system but with two slightly different starting conditions. In an amount of time equal to the reciprocal of that exponent, the conditions in these two “runs” of the system will differ by an amount equal in magnitude to the thing you are trying to measure. In other words, the “disturbance” grows exponentially as e^{\alpha t} where \alpha is the Lyapunov exponent and that \alpha t = 1 gives you the Lyapunov time t.

This small disturbance in initial conditions resulting in large differences over time is often called the “butterfly wings” effect — the microscopic influence in airflow from a butterfly flapping its wings grows exponentially so they cannot predict the weather more than a day or two out.

The computer model is not the underlying dynamical system for a whole host of reasons, but a computer model with some semblance to a chaotic natural system will also be chaotic, and yes, the near infinitesimal differences at the start grow to large differences at the end of a simulation run.

Now the airflow in a jet engine is also chaotic, and you probably cannot predict the exact pattern of air current swirls for more than a millisecond. But your computational fluid dynamic model (CFD) does not have to predict the exact air currents, rather, it needs to make predictions of average flows, pressures, and temperatures in the bulk system even though the exact flow pattern is chaotic and un-modelable in that sense. CFD makes accurate enough predictions of average properties that it has been thoroughly useful in designing more fuel efficient and more powerful jet engines.

The claim is that climate is not weather and that the climate models will not predict the same conditions twice at the end of a run on different hardware. They cannot even tell you how warm or cold by the end of the week, but even the weather models are not predicting freezing temperatures in July in Minnesota or 80 deg-F temperatures in January.

But, this is not to say that the climate models have the underlying physics right and are hence truthful. I imagine that a lot of work went into validating the CFD models in the aviation industry against wind tunnels or engine test stands, and the Earth’s climate system is not amenable to such checking. But I think we weaken our case when we latch on to “climate is chaotic and hence the climate models are worthless.” There is a little bit more to it than that.

1. George Turner

I don’t think the problem is in the nature of chaos (or CFD), it’s that what they’re studying is heavily dependent on what’s usually left in the round off errors of most other types of studies. In aviation, for example, you don’t really care about a 1% change in humidity, much less how it might vary due to the different plant responses over time, or vagaries in ocean/atmosphere mixing.

So you have a bunch of parameters that we haven’t measured, and often ones we’re not really sure how to measure. What’s the effect of aerosols at different altitudes? Well, just make a guess. What kind of aerosols are there, and how do they change in response to lower altitude wind patterns? Well, just make a guess. There are so many guesses involved that the simple equation that would relate CO2 to temperature is the product of a bunch of guesses, and the model runs are just dynamic runs of the fictionalized world that embodies all those guesses. It’s like a flight simulator written by people who didn’t really know the equations of lift and drag and just cobbled together a reasonable looking response profile for the vehicle, like early computer games where you fly an X-wing to attack the Death Star.

3. Larry J

Oh, they’re not useless at all. Billions of dollars have been redirected to particular cronies and constituencies as a result of those models. Al Gore became vastly rich as have quite a few others. For them, the models have been extremely useful. Others have received many years of employment and research grants to produce and run those models.

Now from a scientific perspective, that’s another matter.

1. Al

I expected to find that one of the Ferengi Rules was something like “Never explain the scam to the mark”, but I didn’t see it. However, most all of the others do seem to apply quite well to climate science.

1. Peterh

Not all the Rules of Acquisition are known. I expect “Never explain the scam to the mark” is in there somewhere, possibly alongside “Don’t let the mark know they’ve been scammed.”

2. Paul Milenkovic

The Ferengi fancied themselves ruthless capitalists in the manner of the Klingons styling themselves as ruthless warriors. Whereas neither had a code of ethics according to Federation/Earth/Western standards, Worf for Klingons and Quark for Ferengi established that each in their sphere had a code of honor.

Didn’t Quark, ruefully, take a large financial loss, maybe a Casablanca-esque shutting down of his bar in taking a moral stand?

1. Al

Which just shows how far the term ‘capitalism’ has debased into the concept of unrestricted-near-violent-scamming-R-us.

The Liaden view of Melanti and fair trading is much closer in my own not-so-humble opinion to what capitalism -should- mean. Fair (no coercion, no falsehoods, no omissions) trading can make wealth. Not just ‘making a buck’, but increasing the pie.

4. Leland

Slight off topic, but on topic of useless computer codes…

1. Gregg

I get the same thing. It causes the loading of the page to drag on forever.
And it never speeds up until I’m given the option to kill the script.

2. Bart

Every so often, I get a totally mutant screen where comments are not even nested. Running FireFox.

1. Gregg

Bart,

That has to do with the system using the mobile app for some reason. I *think* it’s unrelated to the script issue.