Climate Modeling, Redux

Gregory Young over at American Thinker wrote a great post about climate modeling. I can’t complain about most of it. It’s the same stuff I’ve screamed time and time again.

I do take offense at one issue. Here’s the quote that sums the issue:

Like all modeling, one attempts to study the past through scientific observation, accurately and unbiasedly collect the data, and then fit the data to a dynamic computer model that is meant to predict, to some degree of accuracy, some measure of tomorrow.  In this way scientists hope to discover trends that not only document the past, but could forecast the future. 

Computer modeling does not try to predict the future. Scientists and engineers are not fortune tellers predicting that someone tall dark and handsome will enter your life after a long trip.

Computer modeling is nothing more than an approximation of an output based on a series of inputs. And the inputs are nothing more than assumptions  based on biases, approximations, pseudo-science, bad science, junk science, science, guesses, inaccurate measurements, accurate measurements, estimates, statistics, plain thin air, and the color Al Gore’s underpants last Tuesday (orange and hot pink plaid, if you really must know – and don’t ask how I know… I am a woman of mystery).

Computer modeling output is a scenario dependant upon a very narrow and precise set of circumstances. It’s not predicting the future. We should stay away from that language. It makes the scientific and engineering community look like spoon benders at the circus.

To the average Joe, the climate change community already appears to be a cult with a doomsday Apocalyptic message. Scientists and engineers should take care that they are not tainted by the Church of Climate Change’s overly zealous claims and predictions.


More Climate Modeling

See? I told you so. Climate computer models are worth shit if the assumptions aren’t right. When you assume in the models that global warming is occurring, then the model will show that global warming is occurring. Because, you know, that was the premise upon which you based the model in the first place.

So now that the most recent data shows that warming isn’t happening, all those modelers scratch their heads in confusion. They predicted a warming and they got a cooling. What to do? I will much enjoy the hand wringing from the climate change modelers that is sure to follow.

This is what happens when eager scientists and engineers get their hands on computer programs when they don’t understand the basic concepts of 1) the program in particular and 2) computer modeling in general. PhD or no, when you don’t understand the above, the results are meaningless.

But even though there’s plenty of reasonable doubt regarding global warming (and the premise that man causes it (hubris, I say!)), plenty of climate change believers are going to still believe that global warming is occurring because of man.

[Via Ace.]

Climate Modeling

Climate modeling is not proof of climate change.  And no, math is not passe.

I’ve never made any climate computer models. So I am not in any way qualified to talk about the specifics of climate modeling. But I have made plenty of stormwater computer models (of various types, with various programs). So I am somewhat qualified to talk about computer modeling in general.

There are many problems associated with predicting anything with computer models (especially anything concerning nature). The main problem is that conditions have to be estimated and assigned a value by an experienced, knowledgeable modeler.

All computer models take real-life conditions and parse them down into some numerical format which can then be used in prognosticative calculations. Which means that modelers take stabs at assigning number values to existing conditions to estimate future conditions. [For example, in stormwater modeling, we have to estimate ground conditions to find out how much rain will become runoff and how much will be intercepted by the ground. The ground conditions must take into account the permeability of the ground, the infiltration rate of the ground, the ground cover type, the ground cover quantity, the ground cover quality, the retentive ability of the ground, and a few other minutiae details concerning dirt. And in the most commonly used formula (good old, Q=CiA), all of those issues with ground cover have to be estimated with a single number, C. There are guides to help modelers estimate C, but they’re meant to point modelers in the right direction. They’re not a solution to the problem of finding C itself.]

The point is that computer modeling isn’t just math. It’s not as easy as, say, solving for the hypotenuse. There are assumptions involved. And the model itself may be a work of art, but if a modeler doesn’t know what they’re doing and if the assumptions are bad, the thing isn’t worth shit.

Now, a good modeler will try to use their assumptions to predict current conditions before applying those assumptions to predict future conditions. This works great in calibrating the assumptions. [For example, in stormwater modeling, if I build a model and the predicted current conditions in the model don’t match the current conditions at the site, I can go back and fudge with the assumptions until my predicted current conditions match the real-life current conditions.] From what I’ve been able to tell about the climate change models, they don’t do this. But then, the climate change modelers have been very tight-lipped about the whole modeling process, so I can’t be sure. [Apparently, they have good reason to be so evasive.]

Also, a good modeler will run a sensitivity analysis. All that means is that they run many different models with varying differences in the assumptions to see how sensitive the model is to small and large differences in the assumptions. [If your model results vary enormously when you make small changes to your assumptions, you’ve possibly got an overly sensitive model. That’s never a good thing. It means that you’ve got to be extra careful with your assumptions because the smallest mistake in your assumptions could mean a large mistake in your results. That’s not a good position to be in, having to explain to your boss -or the world – why you predicted something huge which turned out to be a mistake all because of a tiny error. Embarrassing.] Again, because of the evasiveness of climate modelers, I have no idea if their models have been so analyzed.

So, what would make me feel better about climate change models? Well, if the modelers published the results of a sensitivity analysis, I’d feel a little better. If they could predict current conditions with their models, I might think that they actually cared enough to calibrate their models. If they published their assumptions, made them open for peer review among the general population and not just among their own fan club, I might think that they weren’t trying to hide something. And if they made their models available for peer review among modelers other than their own fan club, I might think that they were honestly genuine in wanting to predict future climate conditions.

As it is, all I’m left with is the feeling that they use the results from their models to try to bring about social, political, economic, and developmental change without having to justify their position. All they have to do is point to a model that no one understands (and no one has reviewed externally) and say that it predicts a dire future for us all if the changes they want are not met.

I guess the whole point of this is: numbers can be manipulated. And models are nothing but numbers – a great many of those numbers are assumptions that depend heavily on the beliefs of the modeler. So they’re not proof of climate change. They’re proof that someone somewhere knows how to manipulate numbers such that the results predict the same future the modeler assumes will happen.