Tuesday, August 14, 2007

NASA Climate Calculation Error

My brother sent me this week a copy of an article in the NY Sun that was hard for me to believe. It seemed incredible, and I figured there must be some mistake. So I looked it up. And, there appears to be no mistake.

First, the article, titled An Inconvenient Truth. Seems that NASA climate research department publishes temperature data, and has been for the last few years pointing out that each year in the 2000's has been very hot in the US in comparison to the last millennium. This was big news in January of this year, when they reported that 2006 had been the hottest year in the US of the previous 112 years (see, for example, this Washington Post article).

Turns out that the numbers were wrong. Earlier this month, Steve McIntyre, a climate researcher in Toronto, got suspicious about the numbers because of some peculiar anomalies that showed up out of the blue in 2000. He did some checking, some reverse engineering, and determined that the US numbers had been significantly off since 2000 (which, in some way that I don't understand, affects previous numbers as well). About a week later, NASA acknowledged the change and quietly adjusted their numbers.

I've been digging into this this morning to find more detailed information, and I'll pass along what I've discovered.

Anthony Watts has the best technical summary of the situation I've found. (Steve McIntyre's site is down, possibly because of a DOS attack.)

Roger Pielke also has a writeup, and the comments section has quite a bit of opinion and links to other information.

If you're interested in the raw data, here's an archived copy of the data before the update, and a direct link to the updated data. Note that you won't find any mention of the update on the NASA site itself.

The RealClimate blog has a post from Gavin Schmidt, a climate scientist at NASA, arguing that the differences are statistically insignificant and not worth mentioning.

So, what do I think?

Well, in the first place, it seems that the bulk of the comments on the Watts and Pielke blogs are taking this form: "See, global warming is disproven, because temperatures aren't rising! No hockey stick!" Well, not really. I mean, it's not like before this adjustment, NASA was saying that 1998 or 2006 were huge amounts hotter than 1934. They were pretty much in a dead heat even then. So, this doesn't seem to "disprove" anything.

On the other hand, there a number of responses that take this form: "If they can't see this type of problem, how reliable is the rest of their data"? I think there's some validity there. This isn't the first time that McIntyre has found data and calculation problems in widely-accepted (and widely propagandized) climate numbers. Every time something like this happens, my confidence in the reliability of the scientific research drops a bit.

I'm not convinced by Gavin Schmidt. I agree with him that the change in 1998 is statistically insignificant. However, it seems to me that the changes from 2000-2006 are much more noticeable. Quoting Gavin: "The net effect of the change was to reduce mean US anomalies by about 0.15 ÂșC for the years 2000-2006." Over a tenth of a degree is extremely significant for short-term climate change. Those sorts of changes have been used very recently (as I pointed out in my Washington Post example) to raise concerns of accelerating changes and impending doom. More than a tenth of a degree difference? That changes the picture in a big way.

It is worth mentioning that these are only US numbers, and they don't seem to make any significant difference in the global numbers.

But I would feel quite a bit better about the scientific honesty of the NASA scientists if they would publicly and prominently acknowledge this flaw in their numbers, and try to explain how they could make such a ridiculous mistake and not catch it for 7 years. Their dismissive attitude makes sense to me only from a political perspective, trying to control the media presentation of the information... and that does not make me confident about the impartiality of their science.

It also surprises me that the purportedly ultra-effective Bush administration politics-overriding-science machine wasn't able to catch this one and use it. Strangely, it seems that NASA (which, I believe, is part of the Bush administration) isn't taking the correct conservative political line here. Interesting.



Kevin said...


I'm impressed that Steven McIntyre figured this out, curious why the actual processing materials such as program source code weren't provided to him, fascinated that it took 7 years for someone to identify a jump in data that I imagine is the basis for numerous studies and judgements, and left wondering what other, even less obvious flaws might exist (in US data or worldwide).

It looks like the problem originated in 2000 with a shift from using already adjusted data (that statistically accounted for location, time of day, etc.) to using raw data. I'd guess that the apparent effects on data before the year 2000 shift are due to some sort of statistical smoothing.

Your link to Anthony Watts's post also mentions the effort to consider the placement of the U.S. temperature stations (e.g. next to an air conditioner?), which may be even worse globally, and its peripheral relationship to this discovery.

I've read that the U.S. data is probably more accurate than the rest of the world and also that "the U.S. history has a rather minimal (warming) trend if any since the 1930s, while the ROW [rest of the world] has a very pronounced trend since the 1930s".

I think it may be difficult to say just how significant this change is, since it's probably dependent upon the particular model the data is being used for. But at the very least, considering the public and political attention Global Warming is receiving, this change affects talking points.

In any case, I agree that NASA should have publicly announced this error so that people know that the data has changed. Their handling of this event is somewhat disconcerting, as you note.

By the way, Steve McIntyre's Climate Audit site is back up.


Kevin said...

On the more general topic of and surrounding Global Warming, here's a few obliquely related Volokh posts I found interesting:

(1) Walking to the Shops 'Damages Planet More Than Going By Car'

(2) Forget Corn, Conserve Forests Instead

(3) No Reason to Follow the E.U.'s Climate Path argues against carbon trading schemes and for "a carbon tax that would be tied to the mean temperature of the tropical troposphere (a region of the atmosphere that is believed to be particularly susceptible to greenhouse gas-induced warming). If the temperature rises, the tax should rise; if it falls the tax should fall."

Also, The Sun Has a Dimmer Switch.

Kevin said...

SurfaceStations.org is the project I referred to previously that documents weather stations in the USA and the world.

MarkC said...


While we're posting random environment-related links, I read an interesting article in the Guardian today that confirms many of my suspicions about the carbon-trading industry.

The inconvenient truth about the carbon offset industry

Also, I read the "Walking to the shops is worse than driving" post on VC, and found myself much more convinced by the comments showing the analysis to be deeply flawed. It's a neat soundbite, but I don't think it holds up to careful scrutiny.


Kevin said...

Thanks, Mark. The damage done by tree planting services was new to me.

Yeah, in general, carbon offsets seem like a big scam to me. Even the carbon tax I mentioned has its flaws, and I wonder at fixating upon a direct and limited causality between corporate release of CO2 and temperature.

Thanks for pointing out about "walking vs. driving". I agree. I think the post still provides an interesting perspective in even making the comparison, given that CO2 is now the focal pollutant.

In that vein, David Smith who dropped by here previously had a brief post on ethanol which makes a similar argument to the other Volokh post -- that it might not be worth the tradeoff.