It appears NOAA is finally making a better algorithm to account for the urban heat island effect on instruments. This has caused 2006 to be changed in rank in the new data set, to below 1998 in warmth.
From the article:
An improved data set being developed at NCDC and scheduled for release in 2007 exploits recent scientific advances that better address uncertainties in the instrumental record. Because different algorithms were used in making adjustments to the station data which comprise both data sets, there are small differences in annual average temperatures between the two data sets. These small differences in average temperatures result in minor changes in annual rankings for some years. Athough undergoing final testing and development, this new data set also shows 2006 and 1998 to be the two warmest years on record for the contiguous U.S., but with 2006 slightly cooler than 1998.
Below is the actual data they had on the page, comparing version 1 and 2.
