Schmidt Responds To "Problem" In GISTEMP U.S. Temperature Data

Gavin Schmidt, a climate modeller at the Goddard Institute for Space Studies, and a regular contributor to the Real Climate blog, has given a quick response to Steve McIntyre's discovery of issues with the Global Historical Climatology Network's data. As noted here, this discovery has led the GHCN to re-organize its U.S. temperature rankings so that a number of recent years come out cooler than originally recorded, while several years in the 1930s come out hotter. What effect does this adjustment have on the overall case for AGW? Mr. Schmidt writes in the comments:

McIntyre noticed that there was an odd offset in the GISTEMP analysis in 2000 which turned out to be related to the transition between USHCN data to the GHCN data. The offset occurred because the USHCN corrections (for Time of Observation bias mainly) affect the more recent values in USHCN but not GHCN (as opposed to only affecting earlier values). Once notified of the problem, GISS investigated immediately, found the error, and added an extra step to the analysis to remove any jump at the transition. This only affected the US temperatures (reducing the mean by about 0.15 ºC in 2000-2006), but since the US is such a small part of the world, it doesn’t effect the global temperatures. Note that this wasn’t a problem with the USHCN data - rather in how the different data sources are melded. It also had nothing to do with any micro-site issues. - gavin]

Furthermore, when a commentator stated:

With this error, it isn’t a small percentage of the world’s temperature that is the issue. It is that the US data is a large percentage of the data available.

Schmidt replied:

[Response: But the global average isn’t simply an average of all the stations divided by the number of stations. You need to adjust for area so that a high concentration of stations in one spot doesn’t bias the mean. If the US is 2% of the area, then a 0.15 ºC correction there implies only a 0.15*0.02=0.003 ºC correction to the global mean (though it’s actually a little higher because of incomplete global coverage). These things should obviously be fixed, but the implications need to be kept in perspective. - gavin]

I suspect that Mr. Schmidt is correct in his assessment, and that the adjustment to U.S. data will only marginally effect the global mean, making the "blade" of the hockey-stick graph infinitesimally less steep.

Which means that Steve Ms work is less a problem for the Science of Global Warming and more a problem for the Politics and PR end of the issue. And here it is a fairly serious concern, because who knows how many more months of delay and inaction it might give rise to? The most obvious argument his work will inspire is: how can we trust ANY of the data, now that we have discovered this particular flaw in in this particular data-set (you might notice how people are already trying to conflate McIntyre's research with the "study" of microsite issues being undertaken by Anthony Watts and the gang at Surface Stations). Secondly, while the U.S. temperature data is only a small slice of the global pie, the fact that U.S. temperatures have not risen as much as those elsewhere may well inspire an "I'm alright Jack" response from American Pols and citizens. Who cares, in other words, if its only Asians and Africans that get fried?

And of course that would be a bad thing.