Paul's CSALT model is similar to what everyone here has been talking about, so I do not see what is the problem with his approach.

I am skeptic in using the linear regression for these sorts of data we have here, but save that his approach is IMHO is sound.

Basically Paul models an atmospheric phenomenon with some math and physics theory, then manually or via software fits a differential equation and then solves it, this the foundation of all known physical theories these days.