Lamay

Nicely said. I agree with your #3; this seems like the main (perhaps only) scientific content issue. McI's analysis was very far from thorough but was nonetheless swallowed whole by his supporters, and then spun into a story about AGW fraud.

The data acquisition issue seems more administrative. For McI, it's a bit embarrassing that he had the data but didn't know - even if you buy his excuses, why didn't he say so in the first place?

P.S. Is your lost link to Delayed Oscillator? (http://delayedoscillator.wordpress.com/)

[I now think it was http://bigcitylib.blogspot.com/2009/10/briffa-ascendant.html -W]

Hi William,

To clarify (I've seen this misunderstanding elsewhere, too), dendrochronologists don't reject or downweight individual tree ring series based on correlation with the instrumental record. To quote Keith Briffa from your link above:

'We do not select tree-core samples based on comparison with climate data. Chronologies are constructed independently and are subsequently compared with climate data to measure the association and quantify the reliability of using the tree-ring data as a proxy for temperature variations.'

That's the way it is done. I'm working on a post on the nuts and bolts of field and laboratory dendrochronology, but real work calls more strongly at the moment.

[Hmm, would like more info on this. I certainly found while preparing this (but have now of course lost) a paper saying that one method of combining the various tree proxies into a series is to weight by correlation with the training record -W]

Yeah but while the science shifts hardly at all this becomes fodder for another round in the political wars, where the fact that McL gets five days of media coverage from all the usual suspects is profoundly unhelpful, especially since the additional fact that all his accusations re witholding data have collapsed (he got Briffa's dataset years ago from the Russians)will get absolutely no coverage. For another week bullshit ruled the airwaves, a disturbing development if you have hopes for Waxman-Markey or anything as mundane as that.

PS. When is your "I was right about arctic ice" superiority dance scheduled?

[Soon. I'm waiting for the Penguin -W]

Here is the problem with what you say. Since we don't have an instrumental record for the sub fossil trees to compare to, we need independent tests (independent of the modern instrumental record) for good tree-mometers. If McI's trees pass those tests, the fact that they diverge suggests a problem with the methodology. If you can't depend on the methodology for living trees, how can you depend on it for the sub fossil trees.

Of course that's all irrelevant if there is some reason to believe that McI's tree don't pass independent tests.

The relevant question is not whether Yamal warmed recently - it did - it is whether tree-mometry is on solid footing.

[We have an instrumental record for 18??-now; as I understand it (which is not very far), the divergence problem kicks in around 1960 (though my eye on Briffa ?2000? says earlier). So you still have most of a century for training. I'm not sure what time period you're thinking of for the sub-fossil trees. Also I'm not sure whether you mean tree-mometry in general, or tree-mometry of this site. People seem fairly convinced by it in general. It is possible that is has been over-extended though -W]

William, CIP,

[1] William, I'd be happy to address your questions further, if you can point me in the direction of the reference you're thinking of -- I can tell you that the standard procedure is exactly as stated by Keith Briffa and quoted by me above. If you're interested, Fritts 1976 'Tree Rings and Climate' and Fritts 1979 'Reconstructing Large-Scale Climatic Patterns from Tree-Ring Data: A Diagnostic Analysis' are standard references, as is Cook and Kairiukstis 1990 'Methods Of Dendrochronology'.

[2] CIP, 'McIntyre's tree's' aren't McIntyre's trees (they are Schweingruber's) and they aren't subfossil -- they are young and they were living. And please, can we refrain from the 'treemometers' silliness?

Well, Eli agrees with your first and last paragraphs. The problem is that you have to stand up to the caca storm lest it roll over the landscape. The Rabett strategy is to wedge out the weakest issue and start there until the Ents arrive.

Speaking of storms in teacups, if you are feeling excessively upbeat and cheerful today, try doing a google search on the phrase:

Antarctic snowmelt 30

I would warn you to move your keyboard back a bit and put a cushion on your desk instead first, this will avoid any injury when you compulsively bang your head down.

{You mean http://www.wnd.com/index.php?fa=PAGE.view&pageId=112227 ? That is not so bad. It was the "Dancing from Genesis" blog link that I liked -W]

Hi,

It's more the way that someone scans a single graph from a paper and it appears in about 100 assorted skeptic blogs within a day.. clearly a sign that climate skeptics are being suppressed by the Big Mean Moinbot Brigade.

First I think you need to decide if this is to be discussed as a particular interesting scientific issue, or whether discussing it necessarily has to include all the immense surrounding baggage. My preference is to look at it as a specific thing.

[Agreed; that is what we shall do here -W]

It is quite clear to me that it isn't appropriate to select trees for their correlation to the temperature record, and I don't think that any of the scientists involved are saying that it happened in this case. They Russian scientists apparently selected trees from the entire multi-thousand year series based on their overall sensitivity to climatic changes. This is measured by the differences in their ring sizes. I haven't found anything that says whether this is a good idea or not, but it seems neutral to me.

[Not sure this is right. "They Russian scientists apparently selected trees from the entire multi-thousand year series based on their overall sensitivity to climatic changes." - err, but how can you possibly know this, unless you've tested that against a known record? "This is measured by the differences in their ring sizes" - are you suggesting they have picked trees with the most variable rings? But how would they know that those reflect temperature rather than anything else? -W]

Yamal is used in a lot of studies. The impact probably varies based on how regional those studies are.

The fundamental issue here is obviously the comparison of the modern period with the MWP, and whether that is even possible with this type of study. It is important to remember that this series would be compared with a single gridded cell of the temperature network, and I'm not sure what that specific cell would show over the instrumental record.

[I'm not sure it is (assuming we are indeed discarding all the baggage). That is definitely the "headline" for the newspapers. But the real issue is long-term cliamte variability -W]

I think the question of current temperatures compared to the MWP is interesting. Was the MWP local or global is another interesting question. There have been recent papers that go either way.

[Are there recent papers that say it was global (as opposed to, oh we found this in this site)? -W]

Not sure this is right. "They Russian scientists apparently selected trees from the entire multi-thousand year series based on their overall sensitivity to climatic changes." - err, but how can you possibly know this, unless you've tested that against a known record?

The idea is location, location, location, and it's based on research into the growth characteristics of trees in various habitats.

You can find a discussion of the rationale here.

[Yes thanks. So, based on that, is Yamal suitbale? Another significant bit is Exploratory analyses identify the strongest temperature relationships between tree growth and a wide range of monthly, seasonal or annual temperature variables. However, that article says nothing about how the series are cmbined. See-also the update -W]

I'll post a bit below:

Classically, open-grown latitudinal and altitudinal treeline sites are mostsensitive to temperature variations whereas lower treeline and/or forest border sites are moisture sensitive.

From my personal experience I can attest that precipitation isn't what causes the treeline to be at 6,000 or so feet on my local mountain (Mt. Hood) - it's wet from top to bottom. I think it's pretty much agreed upon that treelines are sensitive to temperature, and apparently it's the summer temperature that's the key (even conifers largely shut down in winter).
And more:

However, growth at many treeline sites in arid mountains is sensitive to both precipitation andtemperatures (e.g. Bristlecone pines). The key to dendroclimate research is careful site selection as thelimiting factor to growth can also vary based on microsite considerations. Dendroclimate studies furtherassume that the growth-limiting factor at a site does not change over time and therefore contemporarytree-ring climate relationships can be used to reconstruct past climate conditions

Nicolas says:

They Russian scientists apparently selected trees from the entire multi-thousand year series based on their overall sensitivity to climatic changes. This is measured by the differences in their ring sizes. I haven't found anything that says whether this is a good idea or not, but it seems neutral to me.

My (limited, through some skimming via google) is that the idea is those trees showing the strongest differences in inter-annual growth ring sizes are less likely to having their growth being limited by another factor. A seed falls in a crack in a rock and its root system doesn't have great access to nutrients, perhaps, so its ability to speed its growth rate if temps are warmer is limited. That kind of thing.

I don't think the researchers in this field are just pulling stuff out of their ass - there's a whole lot of research on what makes trees tick, given their economic value ...

W,

Thanks for your comments.

Here is the quote from H+S 2002. I am assuming that sensitive here, means sensitive to climate. Later they show that in the instrumental period there is a good correlation to local summer temperatures.

"In one approach to constructing a mean chronology, 224 individual series of subfossil larches were selected. These were the longest and most sensitive series, where sensitivity is measured by the magnitude of interannual variability. These data were supplemented by the addition of 17 ring-width series, from 200â400 year old living larches."

[My initial inclination is that "where sensitivity is measured by the magnitude
of interannual variability" by itself isn't good enough. You have no idea what they are sensitive too -W]

In this particular paper they didn't attempt to compare temperatures in different periods. Instead they compared the amount of climate variability in different periods. It seems like the main reason that the cored a small set of living trees was to create an anchor for the time series, not to create a large set of current temperature proxies.

[Hold on, suddenly you've jumped to "climate variabilty". Where does that come from? -W]

This recent paper in nature looks at SST. They specifically measure the "Indo Pacific Warm Pool" IPWP over a two thousand year period. The paper states that the IPWP "represents a major heat reservoir that both influences global atmospheric circulation and responds to remote northern high-latitude forcings."

[So that's an example of one from a particular region then :-). It would be nice to see their graphs to consider "from about ad 1000 to ad 1250, towards the end of the Medieval Warm Period" - W]

http://www.cartoonstock.com/lowres/amc0043l.jpg

My initial inclination is that "where sensitivity is measured by the magnitude
of interannual variability" by itself isn't good enough. You have no idea what they are sensitive too

Let me repeat:

The key to dendroclimate research is careful site selection as thelimiting factor to growth can also vary based on microsite considerations.

The paper I linked to is fairly comprehensive.

Also Liebig's law of the minimum.

They're looking for sites where temperature is the factor limiting growth for a population of trees, which, again, is why there's focus on trees at high latitudes.

Oh, and this, too, again:

However, growth at many treeline sites in arid mountains is sensitive to both precipitation and temperatures (e.g. Bristlecone pines).

Yamal appears to be wet, so they can eliminate that as a limitation of growth. I assume they believe that their selection procedure eliminates nutrient-stressed individuals, too.

W,

I agree with dhogaza on the rationale on sensitivity. As I mentioned they test this by comparing the portion that overlaps the modern period with temperature measurements. Whether conditions over the multi-thousand year period were similar enough to the present to make the assumption over the whole record is probably open to debate.

Here is what I meant about climate variability although stated differently.

"The width of the annual growth rings in many tree species depends, among other things, on the width of one or more pre- vious rings (manifest as statistically measurable autocorrelation). This is due to the degree of biological persistence that follows from extended physiological processes such as needle formation and longevity, and storage of materials (Fritts, 1976). Rather than using a lagged regression model, incorporating predictors from years prior to (and sometimes following) the predictand climate year (e.g., Briffa et al., 1983), the chronology was instead statisti- cally prewhitened and the residuals from a general autoregression
model were used for estimating past climate variability. This means that the resulting reconstructions are representative of interannual to multidecadal timescales only and will not show century- to millennial-scale changes. These are explored later using the evidence of tree-line changes."

Thus using the H+S methodology century long temperature records can't be compared. For example 20th century to 12th century. I haven't finished reading Briffa's paper, but I believe he uses a different method which he believes allows for this type of comparison.

On the subject of the Indo Pacific Warm Pool. This is more than a regional temperature. The oceans cover two thirds of the planet, and this is an important region. It seems to me unlikely that over century long periods the IPWP would be very warm while the rest of the SST weren't. I think this why the author's think that the results are especially significant.

Here is the key chart.

a, ERSSTv38 mean annual (red line) and JAS (green line) SST reconstructions based on the instrumental record for the grid box containing the BJ8 core sites. Blue line, Mg/Ca-based SST estimates using a published calibration16. Crosses, Mg/Ca-based SST estimates. Lines are three-point running means. b, Downcore SST, and c, delta18Osw reconstructions (31MC, blue crosses; MD60, red crosses; 34GGC, green crosses; 32GGC black circles). Colour-coded lines are three-point running means. Upper and lower horizontal lines in a and b are modern (1997â2007) mean annual and JAS SST8 at the BJ8 core sites, respectively. Colour-coded triangles in b denote radiocarbon age control, except for the most recent red triangle, which denotes the Mt Tambora ash, tentatively identified in MD60 (Supplementary Notes). delta18Osw values are relative to Vienna Standard Mean Ocean Water (VSMOW).

The reference is as follows (I believe this is fair use).

2,000-year-long temperature and hydrology reconstructions from the Indo-Pacific warm pool

Delia W. Oppo, Yair Rosenthal & Braddock K. Linsley

Nature 460, 1113-1116(27 August 2009)

doi:10.1038/nature08233

"It is quite clear to me that it isn't appropriate to select trees for their correlation to the temperature record"<>

I've seen a few people say this but it don't follow. Isn't it likely that tree's ring width temperature dependance will show some consistentency over time - if they show a higher degree of correlation during the instrumental period their correlation will tend to be higher in the pre-instrumental period. Hence choosing a subset based on the instrumental record would give you a more accurate pre-instrumental temperature estimate?

Obviously it depends on having an overlap with the instrumental record, so you can't do with (long) dead trees and - but if all your set overlaps it seems sensible - maybe dendrochronologists have estimated the gains and they aren't worthwhile?

andrewt,

If you had single trees that were alive during the entire period then it might make sense to select them for correlation. But this is not true in the studies we are discussing.

What will happen if you do that is that the reconstruction will look like the instrumental temperature record during the period where you have measurements. This is true since you have specifically selected the trees that meet that requirement. Before that period you have no way of selecting the "correct" trees so if the trees that are highly correlated with temperature are a subset of all the trees then all earlier periods will just be noise.

The whole concept depends on the entire population of trees, as selected without reference to the instrumental record, being correlated with temperature.

Sorry Nicholas, that has not illuminated me. Obviously if trees don't overlap the instrumental record you can't use it to select a subset. And if you use the instrumental record as an input to your temperature reconstruction it may be hard to infer anything about the instrumental period, which doesn't seem important - who needs tree rings if you have thermometers.

If you have a mix of overlapping and non-overlapping trees selecting a subset of the overlapping tree might create statistical problems - but these aren't immediately obvious to me - but presumably dendrochronologists have this figured out.

You could use an incremental process selecting first a subset of the trees that overlap the instrumental records and then using the trees to select a subset of tree that overlap them but not the instrumental record, and so on. This might require an impractical number of trees to get any useful improvement or have other practical issues, but in principle it seems OK.

This might require an impractical number of trees to get any useful improvement or have other practical issues, but in principle it seems OK.

Actually you do because there is an overlap period, and in many cases a substantial one. In that case the problem become whether the younger trees growth had the same behavior in the earlier period, but, absent a time machine, you are stuck with it.

andrewt,

Well perhaps it is silly to discuss because this isn't how it is done.

In any event I considered the idea of using the overlap period, but if you imagine the error propagation you can see the problem.

The trees growing during the instrumental period have perhaps a .7 correlation with summer temperatures in their area. In the first generation back what does the correlation with the pre-fossilized trees have to be in order to select or deselect? Over several generations it isn't clear at all that this doesn't just produce noise.

Like I said if you read the literature the fundamental assumption in these studies is that the population of trees serve as the proxy, not a subset selected for correlation to the instrumental record.

However this is different than some of the multi-proxy studies which do select entire proxy records based on their correlation with local temperatures. It seems to me that the effect of that is similar in that it produces a close fit in the instrumental period, and the possibility of noise in earlier periods, but it is a separate issue.

>"[Soon. I'm waiting for the Penguin -W]"

Penguin has posted now and it appears to me there is lots to complain about ;)

Sorry Nicholas I still can't see any conceptual problem with using information from the instrumental record to choose (or weight) trees which overlap the record or even with using information from these trees to select (or weight) trees which don't overlap the instrumental record. Yes, the information you have for selections/weights would decline as you go back in time but this should mean your subset size approaches the entire set (or weights approach uniformity).

There may be practical issues which make it useless, but I was worried there was some logical reason not to do this that I couldn't see.

andrewt,
the way I understand it, the problem is the following. If you select trees based on their correlation to the (recent) temperature record, you will get two types of trees: (a) those that truly are a good indicator of temperature, but (b) also those that are not and show a correlation only by chance. These (b)-class trees will then naturally reduce any other climate signal in the record, and bias any comparison between then and now towards a higher signal in the now.

how do they differentiate growth caused by temperature and growth caused by increased CO2 availability?

[This is one of the many problems with dendroclimatology; the answer is "you'll have to read the papers" I think; I haven't -W]

paz, these are more things I've seen people say that don't make sense ,

I would expect temperature dependence like most things be roughly gaussian within a species, I don't understand why people suggest a dichotomy.

Yes, if you use the instrumental record as an input and try to propagate this information back in time, you would have increasing uncertainty as you go back in time - a common situation I would have thought when reconstructing the past - and not bias in the sense I think I remember from undergrad stats.

andrewt,
I haven't modelled it so I can't say for sure, but the same holds of course when the situation is less dichotomous. On the one hand, you select trees that recently show a specific shape (warming), but in the past are allowed to vary freely. This means that in the now you just pick the upper end of the gaussian, but more from the middle in the past.

Re the MWP flattening, or not. My understanding, which may be wrong, is that we would prefer it to be flatter (eg not as warm) as that means the climate sensitivity is lower, yes?

[All else being equal, a smaller response during the MWP would mean a smaller cl sens, yes -W]

sorry for asking stupid questions, but this is climate sensitivity thing is something I never completely understood. My understanding was that climate sensitivity was always calculated relative to the specific forcing. So a larger MWP would mean higher sensitivity only when there was concurrent rise in forcing, correct? Or am I not understanding this correctly?

"All else being equal, a smaller response during the MWP would mean a smaller cl sens, yes"

Many thanks. It's one of those things I've been going around "knowing", fairly sure I read it somewhere, but then can't find where I found it, and suddenly not sure it's actually true. I find a lot of things like that (why I'm so bad in "discussions"). I need a better filing system. :)

"So a larger MWP would mean higher sensitivity only when there was concurrent rise in forcing,"

My understanding is that you couldn't (too strong?) have the former without the latter. Eg you need a change in forcing to create a climatologically warmer/colder period.

"My understanding is that you couldn't (too strong?) have the former without the latter. Eg you need a change in forcing to create a climatologically warmer/colder period."
interesting. But isn't climate sensitivity calculated separately for different forcings, i.e. solar, CO2, etc.? To do this, wouldn't you need a record of concurrent forcings during the MWP?

Again, as I understand it, the sensitivity for different forcings will be different as in amount - what you calculate, but the general sensitivity is what determines those calculations.

So, to completely make up some numbers for illustration, if the CO2 St is 2.8C, and the Solar Ss is 0.4C for "general sensitivity" X, then 2X would mean St = 5.6C and Ss = 0.8C. Though the relationship may not be linear.

But I could be way off. It's why I like people asking "stupid" questions as that's where my level of understanding is and it's a good test of what I think I know. Obviously, it's more useful when I'm wrong. :)

> differentiate

here, for example, the devil is in the details:

http://biogeo.botanik.uni-greifswald.de/fileadmin/user_upload/_temp_/Ma….

It's interesting to examine McIntyre's selectivity. For example, he paid considerable attention to the paper linked by NN in #11, but seems to have missed this contemporary one entirely:

'Last nine-thousand years of temperature variability in Northern Europe

'Abstract. The threat of future global warming has generated a major interest in quantifying past climate variability on centennial and millennial time-scales. However, palaeoclimatological records are often noisy and arguments about past variability are only possible if they are based on reproducible features in several reliably dated datasets. Here we focus on the last 9000 years, explore the results of 36 Holocene pollen-based July mean and annual mean temperature reconstructions from Northern Europe by stacking them to create summary curves, and compare them with a high-resolution, summary chironomid-based temperature record and other independent palaeoclimate records. The stacked records show that the "Holocene Thermal Maximum" in the region dates to 8000 to 4800 cal yr BP and that the "8.2 event" and the "Little Ice Age" at 500â100 cal yr BP are the clearest cold episodes during the Holocene. In addition, a more detailed analysis of the last 5000 years pinpoints centennial-scale climate variability with cold anomalies at 3800â3000 and 500â100 cal yr BP, a long, warmer period around 2000 cal yr BP, and a marked warming since the mid 19th century. The colder (warmer) anomalies are associated with increased (decreased) humidity over the northern European mainland, consistent with the modern high correlation between cold (warm) and humid (dry) modes of summer weather in the region. A comparison with the key proxy records reflecting the main forcing factors does not support the hypothesis that solar variability is the cause of the late-Holocene centennial-scale temperature changes. We suggest that the reconstructed anomalies are typical of Northern Europe and their occurrence may be related to the oceanic and atmospheric circulation variability in the North Atlantic â North-European region.'

One conclusion is that northern Europe did not experience a distinct MWP.

Imagine that.

SB,

Two things.

First I'm not aware of SM paying attention to the paper I linked to. Do you have a reference for that?

Second the paper you chose hardly makes your point.

"It is noteworthy that the MWP cannot be clearly observed in the stacked pollen-based record...Thesefeaturessupport many earlier investigations according to which the MWP is not reflected as a clear peak in Northern Europe, but rather represents the final centuries of a longer warm period before the onset of cooling at 1000â800 cal yr BP towards the lower temperatures during the LIA (Bradley et al., 2003; Bjune et al., 2009)."

and

"On the basis of their pollen-based reconstructions from 11 sites in the Fennoscandian tree-line region, Bjune et al. (2009) argued that during the 20th century summers were warmest since about 1000 cal yr BP."

In other words the MWP isn't a clear peak because it was part of a long warm period. And in this study, in this small region, it was as warm as the present 1,000 years ago.

paz puts the general point re: selection in #26, and Nicolas in #23.

However, there is a difficulty in asserting that this is the only problem. It is not. The Yamal data set has now been exposed, and is open for inspection. It relies on <13 trees since 1990, and <6 since 1995. My understanding of the methodology is that that small a number of trees is an unacceptably low number.

"Briffa was a leading proponent on RCS standardization and his writings all stated the need for large populations very clearly."

oops, an unintentinal omissions through use of "less than" signs.

The Yamal data set has now been exposed, and is open for inspection. It relies on less than 13 trees from 1990 onwards, and less than 6 from 1995 onwards. It is my understanding that the statistical methodology used (RCS) needs larger sample sizes than 5 trees at a time.

per

Well, gosh, per, you'll be glad to hear that H&S, Briffa's russian colleagues, have greatly extended the dataset and are submitting their latest reconstruction for publication (or perhaps it's already accepted and in the pipeline, their e-mail wasn't clear).

They've already stated that their reconstruction using a larger sample size comes up with essentially the same hockey stick as Briffa 2000.

dear dhogaza

I would be grateful if you can answer the question.
If Briffa has indeed repeatedly emphasised the need for large numbers of trees in an RCS analysis, and we are talking about an rcs analysis with only 5 or 10 trees over a time period, surely that is a severe criticism of this particular period of the rcs analysis ?

is that not a simple question ?

If someone else has indeed done a different, and perhaps better analysis, that will doubtless be a good thing. But how does that detract from whether or not Yamal has severe problems ?

per

If someone else has indeed done a different, and perhaps better analysis, that will doubtless be a good thing. But how does that detract from whether or not Yamal has severe problems ?

1. The people who actually do this stuff for a living say no, there's nothing wrong with Briffa's work (he's perhaps the world's leading authority on this stuff, after all - not McI, no matter how much you may fervently believe in McI's godlike powers).

2. Briffa was using data collected by the Russians (H&S). Getting more data isn't like taking a walk in the park, it's expensive. Now apparently the Russians have more data, and as I said above get substantially the same result. Wanna bet the error bars have narrowed, to?

3. McI and his fanboys, like you, are always looking in the past. Mann99! Briffa00! Science - and scientists - work forward. Rather than say "paper X has issues, therefore all of climate science is a fraud!", scientists work to get more data, more proxies, improve methods, and work forward. The fact that the result is confirmation of previous supposedly "McI debunked!" work should tell you something. The repeated "hockey stick is dead!" proclamations should tell you something.