The November 17th issue of Science has an interesting exchange of letters between Christy and Spencer; Mears and Wentz; and Sherwood and Lanzante (ref here; subs required for substance). The context of this discussion is the tropospheric temperature record; see Et tu LT and The tropical lapse rate quandary for two RC posts that discuss the issue, and in particular three papers in the August 11th issue of Science.
The first part of the Christy and Spencer letter simply admits the error that Mears et al. found back in August. Nothing new there (so why is it being published in Science?); though they are somewhat coy about the nature of their error: but judging from the rather blunter reply from Mears and Wentz, it was a simple sign error. M+W point out that C+S first introduced their error in 1998, in response to a Wentz and Schnabel paper, which pointed out a previous C+S error, related to orbital decay. Christy has used the erroneous data in testimony to the Senate trying to cast doubt on the quality of climate models; it will be interesting to see if this is now corrected. C+S attempt to argue that their new data has a trend “that is within the +/- 0.05″ error margin of the previous trend. But this is specious: those error bounds
are the standard statistical margins for line-fitting; they [See note] aren’t supposed to represent the possibility of errors in the dataset construction.
C+S continue by arguing that (contrary to the Sherwood paper from August) there may be some spurious warming trends in the radiosonde data; and they give one example. They point out a possible problem with the Day-Minus-Night adjusted data showing too large trends for 1958-1997 for the region south of 30N. Sherwood and Lanzante disagree, and call this a “misleading statistic”, on the grounds that while the tropics are well sampled the region south of 30S isn’t, which leads to erratic trends from radiosondes averaged over that region. C+S attempt to argue that data from US radiosonde validates the C+S satellite dataset: S+L retort rather sharply that C+S have claimed this agreement for the previous version of their data, so shouldn’t the new revised data now disagree? (an alternative is that the agreement with the US radiosondes is simply a poor test of trends in the satellite dataset).
S+L conclude with a rather ironic line: “…it is hard to believe that Christy and Spencer would argue that a data set showing the ‘wrong’ amount of warming must therefore be flawed. If that were a valid argument, their own satellite analysis would have been discarded years ago.”
[Update: the US CCS program report Temperature Trends in the Lower Atmosphere is now available as a public draft; its 300+ pages though so we haven’t read it yet!]
[Another update: an anonymous contributor (see comment 8) points out that the 0.05 bounds are supposed to represent structural uncertainty, not line-fitting as I thought – W]