Wednesday, November 25, 2009

COTD: Code

After a day or two looking through the code (I’m a programmer, not a climatologist) and getting some good responses from gavin at realClimate I am surprised at a few things.

1) I really thought code like this would be super complicated — obviously simulating weather long term is a very tough problem. However in the code it comes down to, get the measurements, apply arbitrary weighting/changes to the end set of data (blade portion), and then weight in hotter measured temps for the blade only (non linearly). One could argue that the numbers used to reduce proxy signals in the 40’s and amplify them in the blade portion were the result of hundreds of inputs and deep thought, but seeing -.03 and 2.6 there doesn’t inspire confidence. That’s like the ultimate answer being 42. The comments above the lines don’t help either, things like “****** APPLIES A VERY ARTIFICIAL CORRECTION FOR DECLINE ******”.