Climate forcings

With the brouhaha over the forcings, and editors being purged and all, this is a good time to post something that hadn't bubbled up to the top of the list before.  Remember, the editor resigned because the paper he published that disputed the positive forcings was so riddled with errors and whatnot.

That sort of thing would never happen with the IPCC.  Oh, wait:

As most people know, I am a lukewarmer -- somebody who accepts carbon dioxide's full greenhouse potential, but does not accept the much more dubious evidence for net positive feedbacks on top, and who therefore thinks that a temperatuire rise of more than 2C in this century is unlikely.

This view just got a strong boost. Nic Lewis, the indefatigable mathematical sleuth who helped expose the mistakes in a paper about Antarctic temperature trends has been looking at how the IPCC estimates climate sensitivity -- that is, the warming expected for a doubling of CO2. He finds that the one study that estimated sensitivity entirely from experimental data -- Forster and Gregory 2006 -- was distoted by the IPCC when it came to present their results. The distortion was the imposition of a Bayesian "uniform prior" in a way that statisticians say is wholly inappropriate, because it effectively assumes a priori that strong warming is more probable than it is. Yet you don't even have to know that the use is inappropriate to know that it's inappropriate to take a published result and alter the graph from it, adding an obscure footnote to say you have done so. A published result is a published result.

He shows how badly this changed the results, with graphs.  Obviously under the new rules, the editor of the IPCC will resign any day now ...

This is why it was so important to get Michael Mann's source code that produced the Hockey Stick graph.  Statistics are subtle, and as the saying goes if you torture the data long enough, it will confess to anything.  For example, you can make a statistically valid case that butter production in Bangladesh predicts the S&P 500.  Careful selection of input data (as Mann did) and careful selection of statistical transforms (as the IPCC did) will fit all sorts of data to all sorts of things.

The solution, of course, is full transparency: publishing all data and source code.  Odd how the folks claiming that we're facing the end of civilization (women and minorities hardest hit™) keep resisting this, even in the face of Freedom Of Information Act requests.

You might almost wonder if they were hiding something.  Almost.

If you want an excellent - and accessible - introduction to the science, MIT's Jim Lindzen has a great overview.  Err, or you could just click on Clint Eastwood in the upper right of this blog.