Today’s tide at surfacestations.org brought in an eastern site,Stamford Ct, courtesy of Kevin Green. A couple of interesting features, including something really weird with the GISS adjustments.
First, it shows modern values greater than the 1930s. There was no observable trend through the mid-1980s. The documentation is inconsistent as to whether there was a station mvoe in 1985. In the CDIAC information, the station history shows that it was located at Norwalk from 1892 to 1985, at the Norwalk Gas Plant from 1956 on. At the new NOAA MI3 website, the Stamford 5N station information shows the station at Stamford from 1955 to 1985, but also shows a Norwalk station from 1955 to 1985. My guess is that CDIAC information is primary for the older information and the MI3 information is incorrect.
At the time of the station move from (presumably Norwalk to Stamford) in 1985, the time of readings was changed from 8 am to 5 pm. This change in observation time is plausibly related to slightly warmer readings (the time of observation adjustment), and a downward adjustment of 0.5 deg C is made effective 1985 (see top row). Other than this, no adjustment is made for the change in station location, although one would presume that the characteristics of the present Stamford site differ from the previous Norwalk site.
TGhe next figure shows the GISS adjustments in stages relative to the three USHCN series. The top row shows that the GISS dset=0 series (“GISS raw” in my usage) is equal to the USHCN raw series after 2000, while the third row shows that the GISS dset =0(raw) series is equal to the USHCN adjusted prior to 2000. This switch in source (which is not reported in any GISS publications to my knowledge) results in about a 0.5 degree upward jump (bias) in the GISS version from 2000 on. I’ve noticed this in other series and it is extremely hard to think of a justification. It looks like a programming error of some kind (though doubtless, if it is, Hansen will say that it doesn’t “matter”). And it might not be a pervasive problem, as I’ve not noticed a consistent pattern. But from a programming point of view, it’s hard to understand why it would happen in some series and not others.
The big problem here seems to be the change in location plus observing time in 1985, coincident with a quasi-jump in the temperature series. I don’t see how you can disentangle these changes such that you can estimate differences between the 1930s and the present from this particular dataset with any security, which is too bad.