The sampling technique.
“determined by interpolation”
Also advocated for the census....”Census Nominee Rules Out Statistical Sampling in 2010”
http://online.wsj.com/article/SB124241977657124963.html
*************************EXCERPT***********************************
All graphs can be clicked to enlarge in a new window.
So what do we see in the Pacific region? We see that the region was warmer than today around the first World War, and about the same as today around World War II (+.2 vs +.4, which is insignificant and in the error bars). We also see massive station drop outs in 1997, after which the temperature finally breaks out of the standard deviation range and starts rising in true hockey stick fashion. The number of stations in this region began at around 30, peaked at around 710, and then dropped back to around 120-160.
Looking at this region alone it there is no clear indication of global warming, human made or otherwise. Lets move onto South America: