Bob Tisdale of Climate Observations is one of the more careful climate bloggers, and he highlights today a severe problem with the Hadley/CRUT temperature record: a bad splice in 1998 due to a change of the core data sets. This chart is one of four he constructed to show the problem is endemic across the various data sets. The splice increased the reported temperature by around 1 degree, a huge amount given the general trendline has temperature increasing less than one degree per century. Bob concludes:
If the Global HADCRUT data is in record territory in 2010, it is in large part the product of poor source dataset splicing of the HADSST2 portion of the HADCRUT data. The period bias in their SST data varies from 0.06 to 0.12 deg C, depending on the dataset one uses for comparison.