UCAR scientist Kevin Trenberth was quoted by Environmental Science & Technology as saying that newcomers to the climate field sometimes do “incredibly stupid” things. I don’t necessarily disagree with this and we have intentionally kept our published comments to very narrow matters that have been resistant to refutation attempts to date. The concern about over-reaching is one reason why we did not attempt to present our "own" reconstruction.
As all of you know, I am very interested and attentive to issues of autocorrelation as they affect climate issues. I recently stumbled on a 1984 article by Trenberth, entitled “Some effects of finite sample size and persistence on meteorological statistics. Part I: Autocorrelation”. The article contains observations which pertain almost directly to MBH98 short-segment centering. I will leave it to you to decide whether, in light of Trenberth , Trenberth’s recent comments should have been addressed towards Mann himself, who arguably in 1997-1998, was a relative “newcomer” to the climate field, raising at least the possibility that he might have done something “incredibly stupid” within Trenberth’s definition.
Trenberth  says:
climatic noise and the persistence, along with the finite size of the samples, must be taken into account when computing statistics of the circulation or the resulting statistics may be significantly biased. This is especially true for computing variances, covariances and correlations and has not been properly taken into account in many meteorological analyses’
Music to my ears. Trenberth goes on to point out potential problems in this respect with several then current studies. He then proceeds to a detailed discussion of problems involved in calculating a sample variance and sample autocorrelation, when the data are not independent. Remarkably, he even presents results for AR1 red noise series, reporting that the estimated autocorrelations have a large negative bias so that the estimated time to de-correlation is “grossly underestimated”. Trenberth identifies the problem as being due (in this case) to not basing the estimate of the process mean àÅ½à⺠on all available data. He says:
We have shown that the method used to compute certain statistics, such as autocorrelations, can lead to seriously biased results. The main lesson to be learned from the above is that for stationary time series the best possible estimate of àÅ½à⺠[the series mean] should be made using all available data … for determining departures from the mean in any subsequent analysis.”
Trenberth then proceeds to show how various studies were unsatisfactory in this respect, criticizing some of them for “not very clearly stating the statistical methodology used”. Later in his conclusion, he states:
A clear statement of the statistical methodology used is essential for others to be able to understand and interpret the results;
Back to MBH98: in addition to criticizing the inaccurate and misleading description of methodology, we strongly criticized its centering on short segments prior to principal components calculation. The MBH98 procedure was obviously inconsistent with the admonition of Trenberth  to use the best possible estimate of the series mean. Failing to observe the well-documented procedures of Trenberth  in respect to the calculation of series means and variances would no doubt seem to him as being “incredibly stupid”.
We have corresponded with Trenberth in an attempt to determine what he had in mind in using the term “incredibly stupid”. First, he purported to be unaware of the ES&T article, saying that he talked to lots of reporters. After some prodding, he then referred to a 2003 EOS criticism of Soon and Baliunas, as though it had something to do with us. After we pointed out that this had nothing to do with us, he cited some realclimate disinformation criticizing “our” reconstruction and challenged us to respond to that. When we pointed out various responses on climateaudit and summarized them, he failed to reply.
At present, Trenberth has failed to identify any “errors” by us, let alone ones which are “incredibly stupid”. Since Mann was in 1997-1998 a relative newcomer to climate science and probably unaware of Trenberth , we’d like to think that perhaps Trenberth was subconsciously thinking of the erroneous short-segment centering in MBH98 as the “incredibly stupid” error by a newcomer when he made his comments to ES&T.
Trenberth, K. , Some effects of finite sample size and persistence on meteorological statistics. Part I: Autocorrelation. Monthly Weather Review, 112, 2359-2368