When confirmation bias affects global warming analysis

November 17, 2008

The report about an enormous goof in collating global temperature data  in Britain’s Telegraph newspaper – “The world has never seen such freezing heat” – has generated it’s own storm system of hot and cold air, with global warming skeptics declaring it “Another dagger in the heart of global warming” and environmentalists responding that one screw up does not a trend undermine.

NASA’s Goddard Institute for Space Studies (GISS) had declared that October was the warmest on record, despite evidence that some parts of the world were experiencing record or near-record shifts in the opposite direction.  In the U.S., as the Telegraph reported, “the National Oceanic and Atmospheric Administration registered 63 local snowfall records and 115 lowest-ever temperatures for the month, and ranked it as only the 70th-warmest October in 114 years.”

So what accounted for this startling discrepancy? Here’s how the Telegraph’s Christopher Booker put it:

GISS’s computerised temperature maps seemed to show readings across a large part of Russia had been up to 10 degrees higher than normal. But when expert readers of the two leading warming-sceptic blogs, Watts Up With That and Climate Audit, began detailed analysis of the GISS data they made an astonishing discovery. The reason for the freak figures was that scores of temperature records from Russia and elsewhere were not based on October readings at all. Figures from the previous month had simply been carried over and repeated two months running.

The error was so glaring that when it was reported on the two blogs – run by the US meteorologist Anthony Watts and Steve McIntyre, the Canadian computer analyst who won fame for his expert debunking of the notorious “hockey stick” graph – GISS began hastily revising its figures. This only made the confusion worse because, to compensate for the lowered temperatures in Russia, GISS claimed to have discovered a new “hotspot” in the Arctic – in a month when satellite images were showing Arctic sea-ice recovering so fast from its summer melt that three weeks ago it was 30 per cent more extensive than at the same time last year.

A GISS spokesman lamely explained that the reason for the error in the Russian figures was that they were obtained from another body, and that GISS did not have resources to exercise proper quality control over the data it was supplied with. This is an astonishing admission: the figures published by Dr Hansen’s institute are not only one of the four data sets that the UN’s Intergovernmental Panel on Climate Change (IPCC) relies on to promote its case for global warming, but they are the most widely quoted, since they consistently show higher temperatures than the others.

A colder than usual fall does not mean that global warming is not happening, nor does one or more errant sets of data suggest that it’s all a bunch of hooey; but the admission that there isn’t “proper quality control” over how this data is collected should be seen as alarming – as should the failure to spot the anomalous findings until critics began speaking up.

What it suggests is a bad case of confirmation bias: Goddard’s researchers are so focused on confirming that global warming is getting worse that they were overly disposed to accepting data which confirmed their worst fears and under disposed to double check its veracity. This is how science gets skewed.

This may well be a singular mistake, and one should perhaps not suddenly embrace global warming skepticism; but the skeptics have demonstrated that the data cannot be taken for granted, and that, at the very least, the Goddard Institute would benefit from less consensus among its staff.


Get every new post delivered to your Inbox.

Join 39 other followers