Implications of NASA’s Climate Data Errors

In a matter of just a couple of weeks, a single, seemingly minor error in climate data calculations has spawned a heated discussion on the issue, leading to retractions of alarmist claims and corrections of global warming data by NASA and its prominent global warming alarmist James Hansen.

A prominent scientist in the global warming debate, Steve McIntyre has discovered that James Hansen’s NASA office made a mistake in how they applied a mathematical filter to several years of surface station temperature data from the USHCN data network between the years 2000-2006. This mistake had the effect of erroneously exaggerating surface temperature averages for the US during the period. Not only did that alter the data for those years, but also the analysis of surrounding years was forced into revision. The earlier widespread claims that 1998 was the hottest year on record for the US are now incorrect (it was 1934).

This issue has received very little mainstream media attention. A google news search for the terms “Global Warming GISS Error McIntyre” yields only 7 results as of 08/13/07. Ironically, the issue may be getting more play in McIntyre’s hometown of Toronto, Canada, as published in the Toronto Star today (Red Faces as NASA over Climate Change Blunder):

In the United States, the calendar year 1998 ranked as the hottest of them all – until someone checked the math.

After a Toronto skeptic tipped NASA this month to one flaw in its climate calculations, the U.S. agency ordered a full data review.

Days later, it put out a revised list of all-time hottest years. The Dust Bowl year of 1934 now ranks as hottest ever in the U.S. – not 1998.

More significantly, the agency reduced the mean U.S. “temperature anomalies” for the years 2000 to 2006 by 0.15 degrees Celsius.

NASA officials have dismissed the changes as trivial. Even the Canadian [McIntyre] who spotted the original flaw says the revisions are “not necessarily material to climate policy.”

You may remember Steve McIntyre as the one who called into question the reliability and robustness of the infamous, disputed Mann “Hockey Stick” global warming graph which I discussed earlier (The Impossibility of Prediction).

In discovering this surface station data error, Steve McIntyre has, in my view, shed light on a bigger question: how reliable is this data after all? If such an error could so easily have been overlooked, what other errors exist? What other implications are there to the changing environments of these data collecting stations?

Background

Early in 2007, the website surfacestations.org was born. The mission of surfacestations.org, as it states, is:

Via collaboration, creation of a publicly available photographic database of weather stations and weather station metadata.

Unstated in its mission, but nonetheless, a significant driving factor for the volunteers and editor of this database project, the website has brought to the attention of climatologists and meteorologists how questionable the environmental placement of the USHCN surface weather stations may be. From moving temperature sensors near airport runways an asphalt parking lots, to placing them within feet of massive air conditioner exhaust units, the changes in environmental conditions in the immediate vicinity of these data collection stations over the past century is evidenced by photographs of the stations. Check out some of these particularly odd and irregular placements. The eventual goal is to get as many stations as possible chronicled in the online database (it currently stands at around 23% as of 08/13/07). Not only is the current placement of such stations brought into question, the changes that have been made to the immediate environments of these stations during the period of time that data has been recorded from each site is also being analyzed.

(Note: In the months following the debut of surfacestations.org, NOAA abruptly and without notice removed a great deal of the station location information (latitude/longitude coordinates) from the once publicly-available database, hindering the work of surfacestation.org contributors who are seeking to document the environment and condition of these stations. NOAA has apparently done so under the guise of protecting the privacy of the volunteers who house these data collecting stations. If they are volunteering to collect and provide the data, they also volunteer that location information for public scientific scrutiny, in my view. Otherwise, this reeks of yet another attempt to silence the skeptics.)

An Error Discovered

A couple of weeks ago, the latest addition to the online surfacestations.org database was a surface observing station from Detroit Lakes, Minnesota. The graph of temperature data from this station showed a curious jump in the late 1990s:

Along with the graph of the temperature data, a photograph of the station was added to the database, showing the proximity of the station to two large air conditioner exhaust units.

Steve McIntyre and those involved in the surfacestations.org project recognized the instrument placement problems immediately. But this is nothing new - several other stations are placed in even more questionable locations. Many surfacestations.org readers became curious about that jump in the temperature record. Was it due to a change in the placement of such influencing factors as the AC units? A review of maintenance records from the site indicated the temperature jump did not coincide with the AC placement, so McIntyre delved deeper.

The temperature data is only valuable to the end that it correctly evidenced the discrete temperature of that exact location on earth during that point in time. This discussion is not about instrument error, it is about an error in analysis and application. Any analysis of the trends of such temperature data over time is worthless. There are no mathematical models, no computer simulation, no data extrapolation exercises, no manipulation what so ever that would allow one to statistically negate the effects of the changes in the immediate station environment as to allow the “true” trends of temperature data (independent of effects by such changes as the AC units) to be shown.

As discussed in several posts on McIntyre’s own blog, the NASA department responsible for the manipulation of surface temperature data does not make their methods publicly available. Surface station data must be adjusted to allow for analysis between stations and over time. Such changes include accounting for the time of day of each observation. If one station reports at 13:53 and another at 14:00 each day, those data points are not directly comparable without some sort of manipulation to negate that time difference. As such, McIntyre was forced to reverse-engineer NASA’s data correction methods. After doing so, he realized that the GISS office has erroneously applied their formula to the data from 2000-2006. The correction of this error resulted in a decrease in the calculated temperature anomalies of about 0.15 C. This may not sound like much, but in the world of climate analysis, just one tenth of a degree is very significant.

After contacting NASA about the error, Steve McIntyre discussed this process further in several blog postings. Over just a few days’ time, NASA quietly corrected the erroneous data online with no statement of correction applied to each data set. But these corrections had some significant impact on the global warming “leader-board”: The list of the hottest years in the US since records have been kept (a fraction of 1% of the time the US has existed, mind you), namely that 1998 is no longer considered the hottest year in the history of records. Other changes to the leader-board also took place (See the corrected list of temperature anomalies via NASA: Contiguous 48 U.S. Surface Air Temperature Anomoly (C))

James Hansen did take the time to respond directly to Steve McIntyre in a derogatory email (A Light On Upstairs?) but has made no other public statement on the issue.

There is no evidence that any attempt was made by NASA to contact those scientists who may have used such data in scientific analyses in the past. McIntyre discusses the changes to the data (absent any public statement on the issue) in Will the Real USHCN Data Please Stand Up). One is left to wonder how many previous climate change studies have been influenced by the erroneous data.

The Big Picture

This event brings several important points into the forefront:

1. Why have such surface stations been granted validity in the scientific community when it comes to statistical analysis of climate data? A main tenet of even middle school science is that you can only change one variable at a time. Presumably, temperature is the variable being examined…one must not then change the environmental conditions immediately surrounding the equipment if one hopes to complete a robust analysis of large-area trends (admittedly, data will be applicable to an examination of the small-scale changes to the localized environment…but that is not what global climate change is about).
2. How are such “correction” schemes applied to the data by NASA’s USHCN data set and others?
3. Should such (unknown) data manipulation schemes be granted the status of “correcting” the climate data in a way that allows for detailed statistical analysis?
4. If such statistical errors exist on this level, is it appropriate for social, political, and governmental policies to be shaped around such data?