Thanks to reader “bbttxu” for this idea. He pointed out a discrepancy between the NOAA year to date temperature anomaly map and the NCDC year to date anomaly number.
http://www.hprcc.unl.edu/products/maps/acis/YearTDeptUS.png
When I numerically integrated the NOAA map above, I came up with a YTD anomaly of 0.38ºF. But NCDC shows a much larger anomaly of 1 .03ºF for the current year, as seen below.
http://climvis.ncdc.noaa.gov/cgi-bin/cag3/hr-display3.pl
So why the discrepancy?
The NCDC data uses USHCN adjustments, which add almost 0.6ºF on to recent temperatures, and subtract from older temperatures. This accounts for much of the claimed upwards temperature trend, and accounts for the entire discrepancy between the NOAA map and the NCDC graph.
http://www.ncdc.noaa.gov/img/climate/research/ushcn/ts.ushcn_anom25_diffs_urb-raw_pg.gif
So there you have it. The thermometers show temperatures close to normal, but the people in charge of the NCDC data adjust the temperatures upwards. Here is their rationale.
http://www.ncdc.noaa.gov/oa/climate/research/ushcn/ushcn.html#QUAL