I did plots for both GISS station data and GISS gridcell data on the same basis as USHCN.
First Try at GISS Adjusted
Here’s a plot for 1765 GISS U.S. stations for trends after 1900. When I inspected the cold anomalies, they turned out to be sites with very incomplete records – some values in the 1960s and some values in the 1980s. Some of the records had insufficient informatino to create a 1961-1990 normal. What does Hansen do in such cases? Only the Shadow knows. How does such patchy network meet any QC standards? (I guess the “high quality” of USHCN data means that it’s not grossly patchy, not that they’ve actually QC’ed the site.
I then restricted the populations to stations having at least 1000 measurements and got the following. The regionalization is similar in gross terms to USHCN but the smoothing is obviously much greater. Does this eliminate the impact of bad data? Obviously not – it just spreads it out so it’s harder to isolate.
Then I did the same thing for GISS gridded data – this has sort of the same geography, but everything is smoothed even further. The cooling in the southeast has become a slight warming (I apologize for the scale on the legends, but I don’t know how to specify the legend labeling right now.) The most notable warming on this map is actually in the nearshore oceans, which draws form a different data set altogether (and it looks like a much bigger job to tie down the ocean data than it is to tie down the land data.)
If you compare the GISS gridded version back to the USHCN data that underpins at least the “better quality” portion of this, it would be quite a bit of work to see exactly how one gets from A to B, but you can certainly tell Dorothy that we’re not in Kansas any more.