There’s a website called surfacestations.org which claims to be leading the effort to conduct an objective “survey” of surface stations which contribute data to the U.S. Historical Climate Network (USHCN). USHCN data are added to the Global Historical Climate Network (GHCN), and form part of the data used by NASA GISS to estimate the global average temperature anomaly.
It’s abundantly clear that the purveyors of surfacestations.org doubt the validity of the surface temperature data. They’re surveying the stations by organizing efforts to take photographs, and identifying what they believe to be problems with the location of temperature (and other weather) data collection equipment. They have photos of over 200 USHCN stations. Never mind that taking pictures isn’t really a good way to investigate the quality of the data, or that 200 is only a small fraction of the USHCN stations, or that the U.S. is only about 1.5 % of the surface area of the globe, or that many other evidences (including glacial retreat, ocean warming, species migration, sea level rise, sea ice extent, and satellite estimates of lower-troposphere temperature) corroborate global warming over the last 30 years. Never mind that the good people at NASA GISS have worked very hard to identify any problems that exist in the source data, so that they can be corrected when possible and discarded when not. The surfacestations.org people are clearly convinced that the evidence for global warming can’t be trusted, and that the warming in the surface temperature record from thermometers is mainly due to bad siting of surface stations.
If you visit their site, the first pictures you’ll see — featured prominently on the home page — will be two stations in California, Orland and Marysville. They call Orland a “well maintained and well sited” location and Marysville as a “not-so-well maintained or well sited USHCN station”. Here are the pictures:
Note also the temperature graphs overlayed on the photos; the Orland data show a sizeable cooling while the Marysville data show warming. The data are from NASA GISS, for the two stations in question. The implication is unavoidable, and in my opinion deliberate, that Marysville data show warming because the data are bad, while Orland data show cooling because the data are good.
Never mind that two stations do not a planet make. How accurate is the impression given by these graphs?
First, let’s take a closer look:
Something is odd here. The Orland plot starts just after 1880, but the Marysville plot starts after 1900. Also, the y-axis on the Orland plot has a range of 6 deg.C while the y-axis on the Marysville plot only covers 5 deg.C; when comparing the graphs, that would tend to exaggerate the Marysville trend relative to Orland.
NASA GISS station data are available for download, so I retrieved the data for Orland and Marysville, CA. Since GISS corrects for identifiable biasing factors like station moves, instrument changes, time-of-observation bias, and yes, urban heating, I retrieved the corrected GISS data. From the raw monthly data, I computed yearly averages, keeping only those years that had data for all 12 months; plotting them together gives this:
Suddenly the two stations don’t seem quite so different as the impression given by the graphs on surfacestations.org. Much of the different impression is due to the fact that the Orland data start decades before the Marysville data, when the Orland time series is cooling strongly. Also, GISS has corrected for known biasing factors. Lo and behold, all of a sudden the time series aren’t nearly so different!
I have often stated that the period of “modern global warming” is 1975 to the present. Others may choose a slightly different starting point, but objective mathematical analysis indicates 1975 as the “turning point” for modern warming, and that is the time interval I have always referred to. Hence the really important question to ask about these data sets is, “what’s happened since 1975?” Here’s a graph of the annual averages for complete years since 1975, together with trend lines fit to both data sets:
Not only are the graphs strikingly similar, the trend rates are nearly identical. Analyzing the monthly, rather than annual, data for greater precision, and accounting for the effect of autocorrelation, the indicated trends with their error ranges are:
Marysville: 3.77 +/- 1.88 deg.C/century
Orland: 4.06 +/- 1.98 deg.C/century
The trend rates are nearly the same! The difference between the computed trends is much smaller than the uncertainty in the calculations. So, for all intents and purposes, these two stations indicate the same trend during the modern global warming era.
Surfacestations.org calls the Orland station a “well-sited” station and Marysville a “not-so-well sited” station, and displays them prominently to give the impression that “good” data indicate cooling while “bad” data indicate warming. It’s ironic that when you look at the actual data, during the modern global warming era Orland indicates more warming than Marysville.
UPDATE UPDATE UPDATE
Both Eli Rabett (in reader comments) and Gavin Schmidt (via email) have reminded me that the last step of NASA GISS adjustments — the correction for urban heating — uses data from nearby rural stations (like Orland) to apply a correction to non-rural stations (like Marysville). Hence in part, the urban heating correction applied to Marysville depends on the trend at Orland. Therefore these two data sets are not completely independent, so it’s not a complete surprise that they give similar trend rates.
So, I retrieved the data for Marysville before the urban-heating-adjustment was applied, and recomputed the results. The final graph now looks like this:
The trend rates are now:
Marysville: 5.06 +/- 1.88 deg.C/century
Orland: 4.06 +/- 1.98 deg.C/century
The Orland trend is unchanged, since it’s a rural station so it had no urban heating adjustment to remove.
The two trend rates, Orland and Marysville-without-urban-heating-correction, are still within each other’s error limits. And of course, the Orland data (labelled “good” by the surfacestations.org people) still show a large warming rate 1975-present (twice the global average, in fact). But my statement that Orland shows more warming than Marysville depends on the urban-heating correction applied to Marysville, which in turn depends in part on Orland data itself.