Climate Model Uncertainty: Part I

Nearly all---the exceptions to this are rarer than sober Paul Krugman columns---statistical models, and many physical models, are checked against the data that was used to fit, or create them. Since it is an elementary theorem that any model may be made to fit perfectly---not just closely, perfectly---to any set of historical data, to claim that your model is good because it fits old data well is a hollow boast.

Now, it is the great shame in the field of statistics—both in the dismal way it is taught and the worse way it is practiced by most—that (a) is nearly always is interpreted to mean “what happened in the past.”

Let us put this another way, we have said so before and will say so again:

Unless the climate model can start two hundred million years ago with a set of initial conditions and when subsequently run without intermediate input repeat the oscillatory history of the climate over that span, it is totally bogus.

That is the same as what the author is saying above. Start with the given state, the parameters, and feedbacks, and run the damn thing. If it shows the history, over the entire vacillating span, then you might have something.

Any algebra student can make a model that fits a limited data set, even using variables that have nothing whatsoever to do with the mechanisms at hand. But apply that over the extended, irregular, variable behavior of the system and then, and only then might there be something worth looking at. That is the regime where chaos theory is relevant.

The global climate modelers, at the highest levels of "academia" on the planet, have NOTHING, NADA. Even worse than ZILCH actually because they presume with hubris, they do have something and we are to pay for it.

Like Marxism and Harvard University the ignorance is an embarrassment to the human intellect.

Like a friend says, "the models are nothing."

Johnny Suntrade

"the science is in," - Barack Obama

Yours is a counsel of perfection and is not realistic. For me, it would be great if modelers could build climate models that accurately predict key future climate parameters within normal margins of error and over meaningful time horizons that allow for verification.

Since global climate is variable on a regional basis, the models must predict on a regional basis without being tuned to the specific regions.

Modelers are free to define their own terms but those terms must leave room for genuine predictions to be made, i.e., saying that the Sahara will remain hot and dry is not a meaningful prediction, and within timespans that are practical for policy making purposes. Predicting NorthEastern US temperature average in 5 years is far more credible than predicting global average temperature in 30 years. If the modelers cannot do it yet, then they should talk among themselves until they can and stop trying to secure additional funding with stories of doom and catastrophe.

“Modelers are free to define their own terms but those terms must leave room for genuine predictions to be made, i.e., saying that the Sahara will remain hot and dry is not a meaningful prediction, and within timespans that are practical for policy making purposes. Predicting NorthEastern US temperature average in 5 years is far more credible than predicting global average temperature in 30 years. If the modelers cannot do it yet, then they should talk among themselves until they can and stop trying to secure additional funding with stories of doom and catastrophe.”

I agree with you with one caveat. On the data from 1970 to 1998, I could make really good ruler predictions. Lay a ruler on the graph and then predict that temperatures would keep going up about the same amount. It did about as well as the climate models. Not very impressive because all that is happening is that we are predicting temperature increases in a short portion of a time period that has been characterized by temperature increases since the early 19th century.

Then, both sets of models (climate models and ruler models) broke down after 1998 when temperatures stopped warming. Both sets of models said that couldn’t be happening.

So with time series data, I’m not really interested in models that haven’t been shown to be able to predict trend changes. Lots of people lose money in the stock market this way. Had you built models on data from the 1990-1998 time frame, your models would predict stocks would always trend up. They would have missed the 1999 crash entirely. A stock model that has predicted both the up and down trends and the changes in trend prospectively is much more strongly validated than one that only has to predict up-trends. It’s really not much different than what ought to be being done with climate model validation.

The other thing that is so aggravating about climate modelers is that they do not publish standard errors for their predictions. This makes it almost impossible to validate the models on their own terms.

I agree. Models should incorporate actual meaningful mechanisms or processes. The past is a good predictor of the future if you have a convincing explanation about why the past is a good predictor of the future. If past climate data indicates anything it is that there are significant cycles and cycles within cycles. These needed to be explained a priori and not just quantitatively modeled or explained ex post.

“The past is a good predictor of the future if you have a convincing explanation about why the past is a good predictor of the future.”

That’s another problem with the AGW hypothesis. They reverse your procedure. They build models with CO2 as a non-linear affector of climate. They they say, “see, the models match the climate data. Therefore, CO2 is a non-linear predictor of climate.” It’s all backward. Even if researchers were producing properly validated models that actually predict the future well within published confidence intervals, they still cannot properly infer that causal conclusion from the models.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.