That would be true if what the theorists depended on was entirely statistical inference. But that is not the case entirely. Increasingly the climatological modelers are using physical prediction rather than statistical prediction alone.

What's the difference?

Let's say you were going to develop a theory involving energy, mass, and velocity. A statistical approach might be to obtain the mass of various objects and smash them into some sort of measuring target at various velocities. The result would be a 2-d graph of velocity and mass on the two axes and plots depicting the relationship between them. (I'm betting it would be a parabolic curve of some sort). Then we might run a regression analysis checking different models against the data and one might derive a model that says that y = 1/21(mv^1.99). Or maybe y = 1/1.97(mv^2.00001) etc. And maybe you'd get an r^2 of 0.9998.

Is this the "correct" model for the relationship between mass and velocity? Statistically we might say that there's a strong curvilinear relationship.

A theoretician might look at the lab scientists model output and give it all sorts of deep thinking and say, "the ideal model depicting the relationship between mass and velocity is y = 1/2 mV^2."

The same sort of two-pronged approach is what is happening in the climate sciences concerned with predicting climate changes. Some egghead develops a model and runs it against conditions in say, 1899 Vermont. Then the egghead checks actual conditions to validate his model.

Since accurate data doesn't extend very far into history, climate scientists must rely on surrogate data. What is surrogate data? Here's an example. There are many, many more.

Trees grow fast under favorable conditions and slow under unfavorable conditions. Many temperate trees grow by producing sleeves of tissue, one over the other, through time. Likes stacked paper cups. If we cut a cross-section through a tree wee see rings, which are merely a 2-d sample slivce through the stacks of cups. Under favorable conditions, these rings are thick. Under unfavorable conditions they are thin. So by looking at the relative thicknesses of concentric rings one can infer with proven accuracy, some aspects of the environment when those rings were established. By sampling a cohort in a given area and over a certain time, a dendrochronologist can develop a fairly accurate local model of the climatic conditions during the year in which all of the various trees rings were established. The University of Arizona has been accumulating a huge library of modern, historic, pre-historic and fossil tree ring sequences that extend back many thousands of years.

Analogous procedures have been developed for many periodic phenomena including seashell growth lines, sediment layers, and ice deposition in glaciers. Cross-analysis of these data demonstrate strong correlations among them during times and in locations where their data overlap. Some of these data extend back hundreds of thousands or even hundreds of million years.

In my opinion, none of the analysis that I have seen has thoroughly convinced me that we are in an anthropogenic warming phase. However, I am pretty firmly convinced that we are in a warming trend of unknown cause and unknown duration. I believe there is strong evidence supporting a human impact or influence on that warming trend, but I have seen nothing that I would accept as definitive proof.


B