Monday, December 7, 2009

Mike's Nature trick

There seems to be lot of interest, and also a lot of confusion, in trying to interpret what the stolen CRU e-mails really mean. In some of the following posts we will try to place some of these emails in their proper background and include some explanations for those interested. I will try to do it in a neutral way, and let the reader make his/her own judgement.




The email alluding to the 'Mike's Nature trick' has been perhaps the one most frequently quoted.

Background: temperatures during the past few centuries and millennia are reconstructed from the so-called climate proxy data, of which the rings of old or fossil trees are the best example. In some places, at high altitudes or high latitudes, the growth of trees is severely constrained by cold temperatures, so that in warmer than normal summers, trees tend to grow thicker rings or build wood of higher density, for instance. By applying statistical methods, the treering width can be interpreted in terms of past temperatures being above or below normal on those locations. The statistical methods themselves are subject to considerable debate, but this is not what this particular 'trick' is about. It happens that some trees - how many and where is also debated - show a 'divergence' from their local temperature since 1980 or so: they do not reflect the warming experience in some of these locations in the last 20-30 years. The reasons for this divergence are so far not well known, but several hypothesis concerning other environmental factors, such as air pollution, have been put forward. To explain this divergence is important, because as long as it remains unexplained, it can be suspected that this divergence may have happened also in the past, thus increasing the uncertainty in the reconstructions of past temperatures.

The 'trick' was to not show this mismatch between treering records and temperatures in the late 20th century. Instead of showing the treering records drifting away from the instrumental temperatures, the instrumental temperatures were substituted for them, 'hiding the decline' of the treerings (and not of the temperatures). In doing so, a potential problem of temperature reconstructions was ensconced and an artificial sense of robustness of these reconstructions was conveyed. An important consequence is that claims about the record level of warmth of particular recent years, say 1998, against the backdrop of the past millennium are not really meaningful, because data of different nature are being compared: instrumental versus treerings.

By the way, this divergence problem was and is well known by dendroclimatologist. Some accept this 'trick' as as a makeshift solution until the real explanation for the divergence is found. But the current situation would have been avoided if, from the very beginning, these difficulties had been openly presented and discussed.

30 comments:

Anton said...

Dr. Zorita,

I am a career academic researcher, but not in a field related to climate studies.

I came across your name when reading the CRU emails. My immediate impression on reading your email to CRU investigators was of an honest scientist trying his best to be helpful and colleagial, but still frank and uncompromising regarding data analysis and interpretation. I then Googled your name and found this site. Please accept my appreciation and respect for your scientific integrity.

Anonymous said...

This is a very clear explication of the problem and its implications. Thank you.

Unknown said...

It is worth noting that the specific instance of note would be the Mann et al. 1998 "Hockeystick" chart.

Briffa's data - which diverged - was truncated.

The Jones & Mann data - which doesn't diverge as dramatically - was not truncated.

Additionally: Briffa's data is smoothed data being carried to the end of the range. The observed smooth seems quite odd relative to either reflected data or using the actual-but-cropped data for smoothing.

Anonymous said...

I cannot cut an paste into the comment section of this blog.

Other comment sections on other blogs do not have this problem.

It may be a specific problem with Blogger.com

Joshua Corning
stendec@gmail.com

Anonymous said...

Thank you. That is the most concise non technical description I have seen.

Olaus Petri said...

Great! We need some balance in this Mann-heated debate, not the least to put some shackles on the man(n) from Penitentiary State. ;-)

Keep up your excellent work!

Jean S said...

Eduardo, just a minor (most readers can skip this) technical note. Mike's trick is not for nothing called "Mike's Nature trick". It was already used in MBH98, see comment #316 in "Mike's Nature trick" at CA for an illustration.

And thanks a lot for starting this blog!

Bohemond said...

" reconstructed from the so-called climate proxy data, of which the rings of old or fossil trees are the best example."

I would say "best-known example." As temperature proxies, tree-rings are dubious as best, especially compared to proxies resulting from physical rather than biological processes like glacial cores, ice-rafted debris etc.

In fact one of the cornerstones of dendrochronology, as used by archeologists, is the very fact that ring patterns are local, so that we can say that, e.g., a Viking ship dug up in Denmark was built from trees felled near Dublin. The notion that one or a handful of trees from a single Siberian hillside is a global proxy is risible.

Anonymous said...

”If tree-rings no longer reflect temperatures from 1960 forward, what reason is there to suppose they reflect temperatures from 1860, or 1730, or 1200?”””
Tree rings are proxy and need to be scaled by temperature observations. That is done by choosing convenient observational record to do correlation and scaling. The recent decades where despite tree ring decline we had warming trend are not taken into account..

Unfortunately reconstructions we know (Mann et al etc etc; hockey
stick) are mainly based on tree rings.

Tree ring crowd hijacked science of paleoclimatic reconstructions.!!!!!!!!!

William M. Connolley said...

At http://coast.gkss.de/staff/zorita/myview.html, you used to say "Why I think that Michael Mann, Phil Jones and Stefan Rahmstorf should be barred from the IPCC process..." and so on.

Should your removal of that text be taken to imply that you no longer hold that view? That you still hold it, but no longer wish to publically assert it? Or something else?

bigcitylib said...

If it is well-known to dendroclimatologist than how is it being hidden?

Anonymous said...

"If it is well-known to dendroclimatologist than how is it being hidden?"


To state the obvious: it is being hidden from policy makers and the general public, as well as other scientists who are not experts in the field.

Anonymous said...

Belette, Eduardo still has that statement emblazened across the top of his home page (did you think to look there?!) so that would suggest he still believes it. My guess is that has been bullied into removing it from his official university web page.

I look forward to reading more of this blog, from what appears to be an honest climate scientist

bigcitylib said...

Anon, if I google "divergence problem" I find links to several dozen papers, a wiki article, and etc. If they are trying to hide something they're doing a terrible job. Note that even objections to the way the graph is presented are in the public domain (for example IPCC reports comments); I think McI has been going on about this for some years before this latest.

MH said...

'I look forward to reading more of this blog, from what appears to be an honest climate scientist'

This blog is run by two scientists.

Anonymous said...

MH, oops, sorry, I mean, of course, "two honest climate scientists".

big, as pointed out by the other Anon, a policy-maker reading the document would not know to google divergence problem, would not have time to look through thousands of IPCC comments, and would not have heard of McI.

William M. Connolley said...

"Belette, Eduardo still has that statement emblazened across the top of his home page (did you think to look there?!)" - good point. He does indeed have the *headline* still on his page - however, "CRU files: Why I think that Michael Mann, Phil Jones and Stefan Rahmstorf should be barred from the IPCC process" is now really a very strange headline, because it links to a page that provides no explanation at all (which, in fact, you could argue was true of the prior text too). I think a more plausible reason is that Z has neglected to remove that banner headline.

William M. Connolley said...

>> 'I look forward to reading more of this blog, from what appears to be an honest climate scientist'

> This blog is run by two scientists.

Let us hope that it was not an accidental snark :-)

bigcitylib said...

Anon, it strikes me that you hypothetical policy maker is far dumber than the real thing.

Jim Prall said...

A point to keep in mind in looking at this debate is that the divergence of the tree rings raises doubts about the usefulness of tree ring proxies, but none about the thermometer records. Since the start of the "intstumental period" - the mid-1800's or so - we have widespread use of thermometers and record keeping of their readings. These are obviously far more reliable and direct indications of temperature trends, and this email says nothing about anyone trying to mis-represent the thermometer readings. There is no 'decline' to 'hide' in the instrumental record, only in the recent tree rings.
So to give a single overview of temperatures for the past millenium, we are compelled to rely on far fewer and very indirect proxy indicators like tree rings, statactite and coral rings, ice core isotopes (del-O18) and borehole readings (the only ones done with thermometers, but requiring clever inverse calculations to convert temp-depth profiles into what past surface temps would have yielded that profile.)
The farther back we want to look, the fewer sources we have to compare and the smaller the data sets. Where the graphs include uncertainty ranges around the central estimates of past temperatures, you see that those ranges get wider as you move further into the past. Still, there is enough data to give us a general picture of temperature changes over the past 1000+ years.
So the mention of wanting to 'hide the decline' was about choosing to leave off the divergent down trend in tree rings since the 1960s, and leaving in the far, far more robust direct thermometer readings.
There is indeed an academic question of why there is that divergence, but that's all it is, and in building a single graph to represent temperature from 1000 years ago to now, the only reasonable choice is to let the thermometers take precedence where we have them.

Anonymous said...

re:Is it just an academic question? (re: divergence).
I agree that thermometer record takes precedence and relying on it rather than on reconstructions is tryvial.
However, the problem is that tree ring index requires applying procedure which allows converting it to temperature. When you look at all these comparisons of recent record of temperature with period before the observation started , such comparisons are mostly irrelevant. It is comparing apples and oranges . Statements about some recent decade being hottest in the last 2 milleniums or 1 millenium can not be validated with good certainty. We compare physical record of temperature withsome 'proxy' reconstructed temperature in middle ages. Scaling is the problem for tree ring method. These tree ring data are useful only if used with extreme care. Let say ' if I was to use latest decades tree ring indexes and correlate them with good temperature records such correlation being negative would produce a reverse history in comparison with IPCC ones. Therefore, divergence is not just an academic argument in my opinion. Unless we explain divergence problem, we can not use proxy tree ring index with certainty as a tool to find out what the real temperatures in the past were.
Jack

Anonymous said...

Birdbrainscan wrote:
"There is indeed an academic question of why there is that divergence, but that's all it is, and in building a single graph to represent temperature from 1000 years ago to now, the only reasonable choice is to let the thermometers take precedence where we have them."

1. The problem of divergence is more than an academic question. It raises significant doubts whether tree rings can be used as an accurate proxy for temperatures. At the very least, the divergence would reduce the confidence one can have in the proxies. You got it correct when you wrote: "the divergence of the tree rings raises doubts about the usefulness of tree ring proxies."

2) The most reasonable approach is to show the temperature record, along with the divergence in the proxies. Such a graph would be simple and accurately convey the data and would require no complex explanations as to how it was generated.

Anonymous said...

BigCity Lib wrote:
"Anon, it strikes me that you hypothetical policy maker is far dumber than the real thing."

Perhaps you meant to write that, "the scientists at CRU must have assumed that policy makers are far dumber than the real thing."

Hans Erren said...

bigcitylib said...
If it is well-known to dendroclimatologist than how is it being hidden?
Before the trick (original data):
http://members.casema.nl/errenwijlens/co2/hidethedecline1.jpg
After the trick (WMO graph)
http://members.casema.nl/errenwijlens/co2/hidethedecline2.jpg

Anonymous said...

"the only reasonable choice is to let the thermometers take precedence where we have them."

Yeah, it would be pretty cool if we had a temperature record that is unaffected by UHI, the inevitable judgement calls of Phil Jones, and other effects. Maybe something like those ocean bouys, except on land, placed far from cities and even roads.

I mention roads because last winter it was twenty six F below one night, according to my car thermometer. Except on the highway it was only ten below and as I drove over a significantly elevated bridge maybe seventy five to one hundreed feet, it got up to minus five. Then, as I pulled onto a side road that leads to my neighborhood, where there was far less traffic, the temp dropped to minuse 18, then as I pulled up to my house, at the end of a seven hundred foot driveway through a small wood, the temp read minus twenty six.

My inference from this experience is that on very cold nights, the layer of extremely cold air is extremely thin, so that driving over an elevated bridge, you get out of it, and that the turbulance of traffic mixes the warmer air from above with the surface, the heavier the traffic, the more intense the effect. So thermometers, if you really want to compare temps from horse and buggy days, need to be located well away from roads. Just because a site is "rural, no lights", doesn't mean that it is unchanged by development, even if the site itself in its physical configuration, does not change over time, and the only change is the increase in traffic. Like the cars buzzing by Stonehenge on the busy highway, just a few feet from the structure

There is another shoe that dropped, outside of "Climategate" on the surface temps, which was the paper by a professor Wang, I believe, from State University of New York at Albany has been shown to be either extremely sloppy, or fraudulent. This paper had been used to rule out UHI as a significant factor in the temperature rise over the past century.

Not to mention that the re analysis of the raw temperature data that forced NASA to concede that the thirties were the warmest years of the US instrumental record is currently not possible on the other temperature data sets which have been "lost."

Given that

- Nobody understands the "divergence" problem.

- Other proxies show the present warming is not unprecidented.

- There is only one temperature data set that has been subject to hostile review.

All we can really say is that the thirties were the warmest years, and the only counterargument seems to be "trust Phil Jones"

- Moptop

Anonymous said...

Last Anonymous: that was the best summary of the situation that I have read so far!

Anonymous said...

Apologies for being the "late comer" to this, but I wanted to ask a question about the "divergence problem". I have no particular axe to grind, merely curiosity, and in looking about the web, this would seem the most appropriate forum to ask it.

As a total non-scientist, I would assume that dendroclimatologists would use a point of departure from what is known. By that I mean they would take recent temperature readings, and correlate these against corresponding tree ring width/density, from where they would extrapolate past temperatures. To the extent that these temperature readings "diverge", I would assume that they correct their past temperature regressions. So by definition, no present "divergence" would be possible.

I can only think of 2 possible ways in which this might not be the case, as follows:

1. Similar width/density ring measurements have emerged in diametrically opposed temperature environments, negating the ability to regress back to past temperatures. This would call into question at the very least, the maturity of the discipline.

2. The modern average temperatures are somehow faked, and the modern tree ring data are accurately extrapolated against the true modern temperature data.

As I said, I have no particular axe to grind: my position could be described as Cartesian. The risks associated with not fixing the possibility of AGW outweigh the benefits of not doing so.

So my question is: am I barking up the wrong tree?

eduardo said...

The divergence is possible because one would need two different regressions to calibrate the treerings, one for the pre-divergence period an another one for the divergence period. Ideally, only one regression would be needed. The divergence indicates that either other factors are affecting recent tree growth or that tree growth is more complex than a simple linear regression to temperature. Both possibilities introduce further uncertainties when reconstructing past temperatures. It has to be said that not all trees display a diverengence. Perhaps you may want to read this for more information http://www.wsl.ch/staff/jan.esper/publications/Wilson_2007_JGR.pdf

TCO said...

If the divergence is something everyone knows, why hide it? The obvious answer...not everyone does know it and they want to prevent that people see the issue. If someone pulled this crap in solid state physics, PRL would cut their nuts off.

TCO said...

And yes....they discussed the divergence later in text...but was CYA and more oblique.

Thos guys sexed that graph up. If you pulled that shit in chemistry or physics...you would get butt-whipped!