Stations are selected for the USHCN dataset based on the length of period of record, the percentage of missing data in the station’s record, the total number of station moves and other station changes that may affect the homogeneity of the site’s data, and, finally, how the site contributes to the spatial coverage of the USHCN network.
The measurement stations must meet criteria established by the National Weather Service. According to Davey/Pielke most stations fail to meet the criteria.
The USHCN sites with good temperature exposure characteristics (i.e., meet all or almost all of the WMO standards) are in the minority in the set discussed in this paper. If the majority of observing sites elsewhere have similar problems to those in eastern Colorado, a significant number will have nonrepresentative exposure features.
What most don't realize is that the margin of error in global tempurature measurement is actually larger that any global variances measured. Ever!
According to Davey/Pielke it is very possible that "a significant number will have nonrepresentative exposure features". Based on this I submit that the margin of error is even higher than previously thought.