Projections about the global climate remind me of large-scale macro models, in that there are all sorts of interactions, feedback effects, that frustrate any theory's implications. in the 70's macroeconomists were making the models bigger and bigger, and after Chris Sims pointed out they did no better than simple ARMA models, that thread just died. I remember my old professor, future Fed member Laurence Meyer, who ran a private forecasting firm based on macro models, said it was great because no one was getting into the field any more (this was around 1987). That should have told him it was a bad field, in that young people, with nothing invested, chose not to waste their time. Of course, Meyer was able to be successful, because you can always sell such advice to old buyers, in the same way you can sell Lincoln Towncars to senior citizens.

But back to global warming. Some economists at Columbia estimate that

yields of three major crops in the United States are predicted to decrease by 25-44% under the slowest warming scenario and 60-79% under the most rapid warming scenario in our preferred model.

Yikes! But read the study and you find it is predicated on extrapolating a 6th-degree polynomial. I personally am wary of squared terms, but a 6th degree polynomial is crazy. Of course the extrapolations outside the sample will be extreme.

I remember thinking getting to be a professor at Columbia was very difficult, but this seems a pretty absurd bit of research. Have you ever seen a 6th-order polynomial in a decent academic journal?