A Numbers Game

I've been writing about global warming and the zealousness by which those claiming that global warming is man-made go after those who think the science is far from settled and that the situation is nowhere near as dire as claimed.

Well, imagine my surprise when I read that the main database for US climate data has to be recalculated because the statistical analysis used was flawed.

What happened? I'll let Coyote Blog run down the deliciousness of the situation.

One of the interesting aspects of these temperature data bases is that they do not just use the raw temperature measurements from each station. Both the NOAA (which maintains the USHCN stations) and the GISS apply many layers of adjustments, which I discussed here. One of the purposes of Watt's project is to help educate climate scientists that many of the adjustments they make to the data back in the office does not necessarily represent the true condition of the temperature stations. In particular, GISS adjustments imply instrument sitings are in more natural settings than they were in say 1905, an outrageous assumption on its face that is totally in conflict to the condition of the stations in Watt's data base. Basically, surface temperature measurements have a low signal to noise ratio, and climate scientists have been overly casual about how they try to tease out the signal.

Anyway, McIntyre suspected that one of these adjustments had a bug, and had had this bug for years. Unfortunately, it was hard to prove. Why? Well, that highlights one of the great travesties of climate science. Government scientists using taxpayer money to develop the GISS temperature data base at taxpayer expense refuse to publicly release their temperature adjustment algorithms or software (In much the same way Michael Mann refused to release the details for scrutiny of his methodology behind the hockey stick). Using the data, though, McIntyre made a compelling case that the GISS data base had systematic discontinuities that bore all the hallmarks of a software bug.

Today, the GISS admitted that McIntyre was correct, and has started to republish its data with the bug fixed. And the numbers are changing a lot. Before today, GISS would have said 1998 was the hottest year on record (Mann, remember, said with up to 99% certainty it was the hottest year in 1000 years) and that 2006 was the second hottest. Well, no more. Here are the new rankings for the 10 hottest years in the US, starting with #1:

1934, 1998, 1921, 2006, 1931, 1999, 1953, 1990, 1938, 1939

Three of the top 10 are in the last decade. Four of the top ten are in the 1930's, before either the IPCC or the GISS really think man had any discernible impact on temperatures.

Okay, that wasn't the truly delicious part of this. James Hansen's involvement is.

Who is James Hansen? He's just the guy who has been pumping Al Gore with all those numbers and he was responsible not only for the flawed statistical analysis, but refused to allow other scientists access to the formulas used in converting the data. It took another scientist to reverse engineer the data to find the flaw and GISS had to admit that the data was faulty.

The flaw that was discovered? It was a Y2K bug.

This doesn't mean that the science is settled or that the earth hasn't experienced warming in the past century. However, it throws many of the assumptions made by global warming proponents into doubt, including using past data to predict future temperature patterns.

Others blogging: Ace, Hot Air, Tigerhawk, IMAO, American Pundit, Pajamas Media, and Junkyard Blog.