Greenland data by using something other than surface data, such as upper-air temperatures, though you'd have fewer years available. My other points relate to Tom's comments. First, the early instrumental temperature data seem to have been a little high for summer and a little low for winter, relatively speaking. As far as I can tell from nineteenth-century papers, this is fairly well documented for Europe, though less so for North America. Second, uncertainties in the corrections to instrumental SSTs may make a difference of about 0.1° C in estimates of nineteenth-century temperature. I suspect that the lack of spatial coverage then is a greater problem, however.
JONES: The important thing is that the difference between summer and winter in the 1850s and 1860s is markedly different from what it has been for most of the twentieth century.
KARL: David, from your work would you say that the cooling Phil showed in the 1880s is real, or might it be the result of the switch to using Stevenson screens?
PARKER: Perhaps both; you'd have to compare the changes with atmospheric circulation data from the same part of the world. The change of three-or four-tenths of a degree around 1880 seems a bit much to be simply instrumental.
KARL: The point I'd like to bring out is that some of the global analyses begin calculating trends and changes from 1880, so we may be starting at a low point.
RASMUSSON: Speaking of data problems, I recently saw that the Climate Analysis Center people found that in their data set the famous World War II SST average jump occurred between December 1941 and January 1942, which seems suspiciously coincidental.
JONES: David tells me that the data receipts prior to December 8, 1941 came from various nations' ships, whereas after that date they came mainly from U.S. ships.
ROOTH: When we look at this business of correlation lengths, we shouldn't forget that the regional meteorology can be used to help us interpret the data. For instance, Fennoscandia is very much influenced by intermittent invasions of Atlantic maritime air. A lot of the low-frequency variability there may reflect those maritime invasions as well as variability over longer time scales. The invasions don't really reach the Urals, and you see different frequency characteristics there.