This post follows on from the recent remarks by John Bates about Tom Karl having his “thumb on the scale” and the apparent instability of the GHCN adjustment algorithm, raised by Bates and discussed in a series of blog posts by Peter O’Neill and a blog post here.
As well as errors in the adjustment algorithm, there are also errors in the raw input data. One example of an error of omission is for Iceland, where for Teigarhorn and other sites there are chunks of data missing, even though the data is available from the Iceland Met Office.
Peter O’Neill has a new blog post showing an example, Cork Airport, where the input data used by GHCN is simply wrong. It disagrees with the data from the Irish Met Office, and it’s “obviously wrong” in the sense that the same number, 10.4, or 1040 in hundredths of a degree, keeps coming up in the GHCN unadjusted file:
An average temperature of 10.4C seems suspiciously warm for January 2015, quite apart from the unlikeliness of January 2013 and 2012 and three Decembers having the same value, and Peter’s post shows that according to the Irish Met Office it should be 5.4C.
Is this just an isolated problem for one site? No, the same problem occurs for another site, Shannon airport, where the mysteriously repeating temperature value is 11.8C.
Note that as well as the dodgy repeating numbers, the files also include the number -9999 in some places. This indicates that the number is “missing”. But at his blog, Peter has cunningly located the numbers that GHCN seem to be unable to find, by looking them up at the site of the Ireland Meteorological Service.
Notice that the GHCN files include some letters alongside the numbers. These are explained here. The W after a number indicates that the computer has noticed that this number is the same as the previous one. The M and the C indicate the source of the dodgy data.
So where does the dubious data come from? The M and C indicate that it’s from “Monthly Climatic Data of the World (MCDW)”, which is another branch of NOAA. The MCDW file for January 2015 is located at https://www1.ncdc.noaa.gov/pub/data/mcdw/mcdw1501.pdf, and you can find the data for other months by changing the file name in the obvious way. You can see the incorrect data in there. In fact, all of the numbers for Ireland for January 2015 look suspiciously high.
Who is responsible for this incorrect data? Whose thumb is on the scale? Well, at the front page of each of these MCDW files, there’s a signature: