Anthony Watts has posted up a quality assessment of the USHCN stations here. John V presented some graphics in the comments thread here and below is my first pass – this comment is not intended to exhaust all possible cross-cuts of the data is merely the first thing that I looked at.
I compared the USHCN TOBS versions for the CRN=1 stations and CRN=5 stations, converting all series to 1961-1990 anomalies and then doing a simple average. Another cut will weight them regionally, but my guess is that such results will not vary much from the ones below. The first figure shows the annual averages for sites classified by quality: CRN1 = “good”; CRN5 = “worse”.
The next figure shows the difference between the two series. At a first look, there is a material difference between the two versions, with the main difference between the CRN1 and CRN5 series arising in the 1950. For comparisons between the 1930s and 2000s, the differences are material, but for comparisons over the past 30 years in the U.S., the differences are less.
If this result holds up, then one could conclude that there is an actual bias differences between CRN1 and CRN5 quality and that the cooling bias from trees and shrubs accounting for one class of QC failure is insufficient to offset the warming bias from QC failures resulting from asphalt, urbanization etc. A simple comparison like this does not say whether the bias is arising from microsite issues or more general urbanization.
UPDATE: For completeness, here are a few more comparisons: USHCN “raw”: so the change in the 1950s is not due to changing TOBS adjustment as one reader wondered:
NExt here is the USHCN adjusted version, showing a little attenuation of the difference:
Here is the GISS adjusted version from the collation of GISS adjusted data into USHCN sites. This is not from the September 10 NASA adjustment vintage – I’m not sure offhand whether it’s from the pre-Y2K NASA adjustments or the post-Y2K adjustments. It’s hard to keep up with the dizzying pace of Hansen adjustments: I would have annotated my downloads a little differently had I realized that the adjustments would be so frequent. In any event, there is still an effect in whatever NASA version this is, but it’s considerably attenuated. In fairness to Hansen, for his U.S. data, he makes an effort to adjust for urban effects – whether his lit-unlit criterion works as well as it might is a different story. On the other hand, as far as we know, Jones makes no attempt whatever to adjust and the problems noted here are going to be more applicable to Jones than to Hansen.
And as I’ve said repeatedly, the real issue in all of this is the ROW (where’s Waldo?) The ROW trend is much different than the US trends: the most interesting result of this will (in my opinion) be, not so much a major revision of US temperature history where one already has pretty warm 1930s (but there will be an effect there), but the information on variations in trends resulting from site quality differences than need to be included in ROW calculations and confidence interval calculations.