From Jones/Briffa et al "High-resolution palaeoclimatology of the last millennium: a review of current status and future prospects" (pdf) - a very interesting paper.
Recent divergence between tree-ring growth and temperature
The final aspect of tree-ring studies that needs to be highlighted is what has become known as the 'divergence' issue. This refers to the apparent failure of some (established as temperature-responsive) tree-ring data to follow the trend in instrumental temperatures observed over the latter part of the twentieth century. Chronology time series that vary largely in parallel with changing temperature in earlier periods progressively fail to show the increasing trends that would represent a continuing positive response to the strong warming observed during recent decades. Originally this was noted primarily in certain northern high-latitude areas for ring-width data in Alaska (Jacoby and D'Arrigo, 1995) and in ring-width and particularly ring-density data, in more extensive regions of northern Europe and Russia (Briffa et al., 1998). In the earlier work, it was suggested that the cause of the North American observations was a shift from a direct dominant temperature control on tree growth to one where lack of available moisture becomes increasingly influential, possibly to an extent where the sign of the temperature influence becomes negative rather than positive (Jacoby and D'Arrigo, 1995; D'Arrigo et al., 2004).
Subsequently, various studies focused mainly on recent treegrowth in Alaska and Canada support the idea that current tree growth may no longer be responding positively to increased warming (Barber et al., 2000; Lloyd and Fastie, 2002; Davi et al., 2003; Wilmking et al., 2004; Driscoll et al., 2005; Pisaric et al., 2007). Other suggestions have been offered as the cause of the widely observed loss of temperature response over northern Eurasia. The increasing influence of drought has also been suggested as the cause (Jacoby et al., 2000), though other suggestions include possible reduced atmospheric clarity, localized persistence of spring snow cover and seasonal changes in ozone-related surface UV concentrations (Briffa et al., 1998, 2004; Vaganov et al., 1999; D'Arrigo et al., 2008).Go read the whole thing.
The IPCC recently laid particular stress on this issue, pointing out that any significant shift in the recent growth response of trees would invalidate the assumptions that underlie the simple regression-based approach to reconstructing past temperature changes. This would imply an inability to recognize potential underestimates of the degree of warmth in earlier periods of reconstructions (Jansen et al., 2007). It is important to stress that not all high-latitude regions display this apparent decoupling between observed and dendroclimatically estimated temperatures (Briffa et al., 2007; Wilson et al., 2007).
However, the issue remains a crucial one. Unfortunately, a comparative scarcity of recent (ie, post-1980) tree-ring data remains a major obstacle to further exploration of the extent and causes. Hence we stress the vital requirement for widespread updating of major treering networks, as well as for the acquisition of data for new regions.
[emphasis added]
-
The smoking gun is here:
WMO STATEMENT ON THE STATUS OF THE GLOBAL CLIMATE IN 1999
In simplest terms- anywhere you see Briffa(1999), you are looking a tree stump study wherein the temps after 1960 rapidly declined. The real data looks like this.
Jones and Mann decided to lop off the data after 1960 and replace it with thermometer data without indicating they had done so. That is the decline that was hidden via 'Mike's Nature trick'.
Michael Mann, Dec 2004
"No researchers in this field have ever, to our knowledge, “grafted the thermometer record onto” any reconstruction. It is somewhat disappointing to find this specious claim (which we usually find originating from industry-funded climate disinformation websites) appearing in this forum [realclimate]."
Phil Jones, Nov 1999
"I’ve just completed Mike’s Nature trick of adding in the real temps to each series for the last 20 years (ie from 1981 onwards) and from 1961 for Keith’s to hide the decline."
from ClimateAudit
Is there any way not to conclude from the above that Michael Mann was lying?
The tree-ring data has never been "established as temperature-responsive". It is a spurious correlation and it isn't present in most of the tree ring series. It is only the strip bark pines and the Yamal series that have this correlation. Correlation does not equal causation. The paper goes on to say:
This should be the first clue that the correlation is spurious. It seems these guys have to much invested in tree rings to let it go.
Steve McIntyre has a related post.
Some context for this paper is likely useful: Tree ring studies are all about estimating 'proxy' temperature data for past periods for which we have no instrument data. That proxy temperature data will in turn be used to calibrate or test a climate model. If the proxy temperatures generated are bogus, that will in turn impact the usefulness of any climate model calibrated using it.
The mapping of raw proxy data to estimated temperature is itself a model, and subject to being evaluated as such. The first thing I note is that the treering -> temperature mappings are all brute force statistical fits (linear regressions, multiple regressions, principle components analysis). There's no actual simulation of a mechanism of tree growth, which would be presumably include things like rainfall, insolation, shading, etc. (This is inherent in the choice of tree rings as a proxy, since we don't have measurements of those independent variables any more than we have temperature data for the times in question.)
So the actual system is something like Ring_Growth = f(temp, rain, insol, shading). (There will be time integration internal to the function, e.g., degree-days.) Now it's an obvious hand-wave to try to back temp out of that function without knowing the other inputs. The tree-ring advocates attempt to get around that by finding tree samples for which there is a very strong correlation between observed growth and known temperatures for a recorded, modern period. There are a lot of arguments regarding this, and it's at the heart of the heated discussions about 'stands' and 'cherrypicking'. AFAIK there's no experimentally verified mechanism and model for why this correlation works in some places, not in others, and why or why not the correlation should hold over time.
So there's one big question hovering over the methodology to begin with. There are also the two usual risks with purely statistical models: that you will 'over fit' the data with parameters that will not work outside the observations, or that your sampling period is inadequate to represent the actual system dynamics. "Hide the decline" speaks to all three of these. Evidently when running the tree ring proxy models 'forward' for the period 1960 forward, the output deviates strongly downward from actual temperature observations. This failure to validate forward casts grave doubts on the validity of tree ring proxies in the past, for any or all three of the reasons above.
The selection methodology for proxy trees is reminiscent of a common error in statistics. Someone (for instance) will go out and do a big survey, take the results and run about a hundred cross-correlations, and come up with a few significant at (say) p < .05 levels. Well, duh! That level actually says you'll observe any particular correlation by chance one time in twenty, so if you run a 100 tests over (even random) data, guess what happens? Substitute all the tree rings measures for the survey, and matching observed temperature to a given significance level, and you've got one possible explanation for the 'decline' failure? If the correlation to temperature (alone) is by chance in the first place, the likelihood of it holding in the past, and continuing to hold in the future, is negligible.
This whole thing leaves me wondering why people are bothering with tree rings that seem bound to be flaky proxies in the first place. There are some very long ice core proxy series over at Foresight - is there something wrong with them? What data hole is the tree ring business supposed to fill? Is there just good grant money in it? Can someone who's followed the debate at more length enlighten us?
Ice cores have their own problems. Intrinsically, they have the fatal problem of only giving you information in a very small percentage of the Earth (places where there are ice cores available). Assuming those areas reflect what is happening everywhere else on Earth is... problematic.
As i've been saying, there actually aren't very many ways to measure historical temperatures and a lot of reasons to doubt the accuracy of what we do have. Particularly when people like Mann are busily playing games with the data.
mark: The WUWT post you link is a cross-post of the Foresight one that I linked. I'm trying to see what's the problem with the data, granting of course the inherent limit on geographic diversity.
I can see the problem that the ice proxy data poses for some warming models. If you look at the longest time scale it looks like cyclic behavior of some sort, which tends to rebut the notion of positive feedbacks and runaways. It also looks like we might be near the top of one of those cycles. I assume someone out there has shoved the original data through a Fourier transform to check the eyeball periodicity?
Sorry about that Tim!
This is disturbing. Raw vs adjusted data in the US since 1880.
If adjustments are what they say they are, they should net out to be about neutral (some stations move to colder spots, some to warmer.) If anything they should adjust downward, because certainly a station is more likely to be seeing a reduction in local flora and increase in people as opposed to the opposite. There is no reason I can think of that temps should have been manually adjusted upwards so much more often than downward to create this gap.
And when I followed your link to WUWT and looked at the main page I found this -- it's just one station, but there doesn't be any sense to the adjustment, which appears to be washing out the Depression-era warm spell.
Everybody's getting in on the act.
This guy did Antarctica
This guy did Illinois
You can argue with any of this stuff, but why does it seem the adjusted data on any kind of scale is always going up?
There's good overview of ice core techniques, locations, and proxy issue here - and doing a power spectrum of the data is just as obvious as I'd thought. The earliest citation I found on a quick search is 1997, and there's one outside a paywall here - you can get to the parent article from that page's menu bar.
it was suggested that the cause of the North American observations was a shift from a direct dominant temperature control on tree growth to one where lack of available moisture becomes increasingly influential, possibly to an extent where the sign of the temperature influence becomes negative rather than positive
Well that dog don't hunt. From the '75 through 2005 was the rainiest thirty years on record in California, where the bristlecone pines are located. I imagine the same holds true for Colorado, where the other stand of bristlecones live.
Also something you might be interested in, trees by their nature maintain an internal temperature conducive to photosynthesis. This applies from the Yukon, all the way to the tropics, for all species.
(I'm confused by your instructions on linking, so I'll just leave you some search terms for Bing. "Helliker & Richter". The first return will be _ Subtropical to boreal convergence of tree-leaf temperatures. Brent R. Helliker 1 & Suzanna L. Richter 2_ from Nature magazine.) It's good enough.
Trees aren't thermometers.