I'm glad you asked this, because that comment made no sense at all to me. Here's what I heard: The old data where we couldn't verify it very well shows that it is good, but the new data where we can verify better it shows it is bad, so therefore it must be good.

I must have misunderstood something there.

Absolutely. I'm not very good at finding information in scientific papers, but here's the first few I found with Google. If those don't satisfy you, I'll see what I can find when I get home from work.

http://www.ldeo.columbia.edu/res/fac/trl/downloads/Publicati... (PDF) is a paper that specifically addresses a lot of the concerns about temperature prediction using tree rings. In particular, I'd like to quote the following section as it pertains directly to your question:

"A number of tree-ring series indicate a divergence between tree growth and temperature at some northern sites in recent decades (e.g. Briffa et al. 1995, Jacoby and D'Arrigo 1995, Briffa et al. 1998, Vaganov et al. 1999, Barber et al. 2000). Theories for the cause (s) of this observed divergence, which may vary from site to site, include decreased temperature sensitivity due to warmer temperatures, drought stress, increased winter snowmelt and ozone effects. This divergence needs to be considered to avoid bias in dendroclimatic reconstructions; however it is not present everywhere. For example, temperature-sensitive elevational treeline sites in Mongolia and the European Alps exhibit dramatic growth increases in recent decades (D'Arrigo et al. 2001, Buntgen et al. 2005)."

http://www.ldeo.columbia.edu/res/fac/trl/downloads/Publicati... (PDF) is an earlier paper referenced in the above that also discusses the problem.

http://icecap.us/images/uploads/Loehle_Divergence_CC.pdf (PDF) is a paper by someone that's more skeptical, it goes into a fair bit of detail about the possible causes, mostly to argue against them.

Of course, we're getting a bit away from the actual topic. We know about the post-1960ish divergence. We know that it's a good idea to exclude that data from any kind of analysis, which isn't a problem because we have accurate modern measurements from that period to use instead. There may be other periods of inaccuracy - but if there are, they haven't been discovered, and there is a substantial amount of overlap between these reconstructions and our actual instrumental record during the calibration period. In other words, it's the best reconstruction we have at the moment, and like any scientific theory, it's subject to change if better evidence were to come along. Unfortunately, the skeptics seem to be more interested in politicizing the debate rather than finding that better evidence themselves.

None of this really changes what the comment says, which was basically "Don't use this to plot past 1960, because we know that part needs corrections". I don't see an issue with that. The wording could be better, but it's a useful note to have, to make sure that the data isn't misused. I don't think it would be necessary to say "Oh yeah, the rest is just a theory, it could be wrong too" - everyone who was supposed to have access to the source knew that. It comes with the territory. The problem is when internal stuff is leaked and that territory changes - a random person reads that comment, completely misses the tentative scientific context and thinks that the programmer is asserting the absolute correctness of everything he doesn't single out as wrong.

Edited to add in a (PDF) tag that I forgot.

I appreciate it. The first link is coming up 404, but I'll check the other two. I think I'm superficially familiar with both, but I haven't actually read them in their entirety. Will do if I have time.

there is a substantial amount of overlap between these reconstructions and our actual instrumental record during the calibration period

From memory, I don't think this is quite the right conclusion. Rather, the instrumental record is used to calibrate the reconstruction, so the overlap approaches 100%. Worse, the accusations are that the correlation is even better because any trees not matching the temperature record have been eliminated as non-performing! I don't think either of these papers shows accuracy of reconstruction on an out-of-sample set for which there exists an outside measurement. I'll recheck, though.

None of this really changes what the comment says, which was basically "Don't use this to plot past 1960, because we know that part needs corrections".

I don't disagree here. In the absence of some overriding context, that seems like a valid reading of the comment.

Hmm, it looks like the first link got eaten by HN's linkerator due to the %27 (single quote) in the URL. Sorry about that, try http://preview.tinyurl.com/ydb8h28.

As for the calibration period, you're right, there are definitely some issues there, and that's a problem with the entire temperature reconstruction approach. We only have reliable instrumental records back to the middle of the 19th century, so it's hard to test any kind of reconstruction on a large scale.

Personally, I don't put much faith in the reconstructions themselves - I just think that it's important for people to understand the context around these experiments, since there's so much blatantly false information going around. It's hard enough for those of us who aren't climatologists to keep up with the information without people screaming bloody murder over a few lines of comments that they don't even understand.

When I was a kid I remember hearing about how acid rain was hurting forests in many areas, including large parts of Canada and the USA.

Factors like that are certainly going to have an impact on the recent tree record.