the Air Vent

Because the world needs another opinion

Hockeystickization Revisited

Posted by Jeff Id on October 28, 2009

I’ve got plenty of posts right now to work on but today Steve McIntyre called our attention to a couple of acerbic replies from Keith Briffa to Steve’s discovery that the Briffa Yamal temperature data which has a HUGE hockey stick blade was actually the EXACT same data as the orignal H&S Yamal which shows NONE. The difference is in the RCS “standardization” method (AKA – hockeysickization) used to correct for tree ring widths. If I don’t miss my guess, we’ll be seeing more of RCS with improper exponentialesque curve standardizations as in my opinion it’s a near guarantee of a hockey stick blade and after Mann and now Briffa I don’t trust these paloclimate ‘scientists’ any farther than I can throw them.

I’ve got to say again Briffa’s original Yamal is a disgusting piece of garbage work and the sooner paleo’s drop the P.O.S. the better. It’s got an unreasonable blade created from RCS with NO science or verification to prefer the ‘accidentally’ chosen exponential curve that is ENTIRELY RESPONSIBLE for the big evil bullcrap blade. See one of my posts on this HERE.

Well I’m not happy with Briffa or anyone who chooses to defend this garbage work so I’ve decided to dig through his reply to Steve a bit and make a little trouble. In particular, I want to focus on this little gem by Briffa himself.

We would never select or manipulate data in order to arrive at some preconceived or regionally unrepresentative result. However, as we will show here, the fact that we did not incorporate the KHAD data has no serious implications for the general validity of our published work.

My emphasis. But those of us who follow climate know that is not the case. So I went looking for a few examples from his own PUBLISHED work. Here is an article from Briffa. although Osbourne was the lead author. I assure you this is standard in paleo science where proxy’s are often assumed to be temperature without any verification other than mathematical correlation. There are many papers which have similar statements, Mann08 is no exception. I could write a whole post on the differences between clear and obvious selection or the ‘partial selection’ created by multivariate methods where non-correlating proxies are deweighted rather than deleted resulting in a less transparent selection of data based on preconceived conclusions.


The 387 density chronologies were screened according to their correlation with the “local” grid-box April-to-September temperature, already computed by Briffa et al. (2002a; their Figure 5a) over the period 1881–1994. The actual period over which these correlations were computed is often shorter than 1881–1994, because either some temperature observations are missing (see section 2.2) or the trees were sampled before 1994 (Figure 2a). Almost half of the chronologies exhibit a correlation of 0.5 or above, and almost 90% exhibit a correlation of 0.22 or above (the latter threshold indicates that we can be 95% confident of a true association between MXD and April-to-September temperature, given a sample of 81 years, which is the average overlap between chronology and temperature data). The 46 chronologies that do not meet this criterion were excluded from further use;

Oh yeah, Steve McIntyre is the bad guy. They would NEVER select data to arrive at some preconceived result. No chance. IT’S RIGHT THERE IN PRINT!!

A reasoned mind might ask the obvious question which dendroclimatology has never answered clearly. – What about the fact that trees might not ACTUALLY be good linear thermometers?

Briffa et al. (2002a) have already identified many of the climate signals recorded in the treering data, and concluded that the tree-ring density variability is dominated by the temperature during the growing season.

Well that about settles it then. Briffa has determined that MXD (latewood density) data IS temperature so we don’t have to WORRY OUR PRETTY LITTLE HEADS about that anymore. But wait, in the same PEERREVIEWEDLITERACHURE there’s this.

A number of factors were taken into account when selecting the most appropriate periods for calibration and verification of the gridded density data against observed temperatures. The most important factor is the identification by Briffa et al. (1998a) of a recent downward trend in the highlatitude tree-ring density data, relative to (and apparently unrelated to) warm-season temperature. This density decline becomes large enough to impair the calibration after about 1960. For this reason, both Briffa et al. (2001) and Briffa et al. (2002a) used only pre-1961 data for calibration of their subcontinental, regional temperature reconstructions. This is a reasonable choice, provided that it is explicitly stated that this approach assumes the apparent recent density decline is due to some anthropogenic factor and that similar behaviour is assumed, therefore, not to have occurred earlier in the reconstruction period – which would otherwise introduce bias in the reconstructed temperatures. At present, no satisfactory explanation of the relative MXD decline has been identified, and further work must dictate whether this assumption will be supported or rejected (Briffa et al., 1998a, 2003, and Vaganov et al., 1999, discuss and investigate possible causes).

MY GOD, it’s time for dendros to stop the charade. I know they’ve spent years trying to extract temperature from noisy data but scrapping data for PRECONCEIVED RESULTS is REGULARLY done!! Equally invalid are reweighting methods that reduce the importance of lower correlation data- its the same damn thing in a more confusing approach.

But wait, why is Jeff so wound up today?!! Because Briffa responded specifically to the criticisms of too few proxies ignoring the criticisms of the method and repeated the same MESS with RCS and more trees. See the new improved YAMALINATOR at the link in the quote below!

We have now undertaken a more extensive sensitivity test, using the RCS approach, to examine the relative growth rates of trees at each of the 3 original locations removed by McIntyre, as well as the KHAD site. We have also taken the opportunity to acquire and incorporate additional data from the 3 original sites, in this analysis.

Individual site RCS using only recent data

Figure C This Figure shows the mean RCS indices plotted by site over the period 1801 to the end of each chronology (1996 for YAD, 1994 for POR, 1991 for JAH, and 1990 for KHAD). In each case the chronology was constructed using a RCS growth/age model based only on the tree-ring data for that site (shown in the upper panel). The "signal-free" implementation of RCS is employed so as to reduce bias in the RCS arising from co-incident climate signal in any near-equal-aged trees when aligned by life cycle. The upper panel reveals the differences in the magnitude of expected ring-width as a function of tree age at different sites. The lower panel shows the differences in the temporal trends of chronology indices between sites. The comparatively low growth at the KHAD site after 1970 is clear.

We don’t have the actual math used for this yet but if you see the RCS curves they came up with, they are functionally no different than the original exponential curves of Briffa Yamal the original.

Next are these new datasets with Yamal included.

Individual site RCS incorporating sub-fossil data

Figure D This Figure is equivalent to Figure C except that at each site the RCS curve, and resulting site indices, are calculated including the Yamal_SF data as well as the measurement data from living trees. The dominance of the common sub-fossil measurements produces a very similar RCS curve in each case (upper panel). The indices produced exhibit a similar picture of recent growth trends varying between sites, as that seen in Figure C, with mean tree-ring index trends higher for the POR and YAD sites, lowest (even negative after 1970 with respect to the long-term mean) at KHAD, and at an intermediate level at JAH. The black line in the lower panel represents the chronology (from 1750) produced using all of the data, Yamal_All, standardised with a Yamal_All RCS curve.

For some reason EVERY RCS CORRECTION Briffa can conceive of refuses to turn upward to fit the ACTUAL data. This lack of flexibility in the RCS curves is what creates the HS. We ARE going to see more of this, mark my words! Below is what the Yamal original data looked like when averaged per year.

Yamal mean ring width per year

Original Yamal Tree Ring Width Data Averaged Per Year

Old trees grow faster in their end years, the RCS curves used by Briffa in the new improved Yamal don’t match the data again!! It’s the same exact thing as the original Briffa Yamal.

I’m tired again. The insanity of paleoclimatology is getting to be too much. Mann’s doubletalk on flipping proxies upside down, RCS creating hockey sticks out of nothing. People creating one obviously bad result after another. Comments by people on the internet which are confused by this ignorant crap. There is just too much bad work for my head to absorb. WTF, none of this is any better than a simple mean. NONE!! Data IS thrown away for predetermined conclusions I’m SICK TO DEATH of the lying and posturing. NONE of this is reasonable.

I have email conversations with Dr. Christy which do nothing but make sense and then am assaulted with this IDOCY.

Understanding Keith Briffa will make you dumber!!! It will suck the brains right out of your ears – don’t read it…. ever!!

In my opinion, this is not honest science.


27 Responses to “Hockeystickization Revisited”

  1. timetochooseagain said

    Did it never occur to anyone to ask why if trees are reliably temperature sensitive, 46 would show no significant sensitivity to temperature???

  2. Layman Lurker said

    I thought you learned your lesson Jeff. We can’t intermingle signal with the growth function. ;)

  3. Gordon Ford said

    As a wooden bowl turner I love sensitive trees. Sensitive trees have rings that tell a story. (Some have stories that would make a Playboy centerfold blush. I know, they talk to me on the lathe) Even in well adjusted trees (with boring grain) tree rings are neither concentric nor even. I must admit though, tree rings do provide excellent entertainment.

  4. When you say we will be seeing more use of RCS, I think you are right. Take a look at this.

    http://gotw.nerc.ac.uk/list_full.asp?pcode=NE%2FG018863%2F1

  5. hang in there Jeff, I won’t say calm down but I will suggest that passion is a great driver for real good science as well as scientific justice, if you can completely channel that energy into targeted fencing. Which is what I see Steve McI do. You’re halfway there.

  6. You got the same temper as Harry Potter. There’s a great model for you. Put into a “magical” reality so people can’t insist it’s about “real” life. Which of course it it.

  7. Steve Fitzpatrick said

    Jeff,

    I think Briffa’s post is a step in the right direction. First and foremost, he lays out the raw data in detail (which should of course have been done many years ago), and explains the approach taken in analyzing the data. Briffa even admits the most recent portion of the reconstruction (the last 8-10 years, apparently everything after 1988 or 1989) must be treated with “caution”, since apparently even Briffa understands this part of the trend is distorted by cherry-picked living trees that show rapid recent growth. A fair question: Why would Briffa not at least have advised readers of his earlier papers that this most recent part of the reconstructions is very doubtful due to cherry-picked trees?

    Despite his claim that “We would never select or manipulate data in order to arrive at some preconceived or regionally unrepresentative result.”, it is not altogether clear from Briffa’s post why he did not include in his earlier Yamal analysis the Yamal data set that Steve McIntyre did (KAH trees) which show absolutely nothing happening in the 20th century. He also does not explicitly state that the same selection methods were used for both living and dead (sub-fossil) trees, leaving me with the impression that some cherry-picking may still contaminate the Yamal data for the living tree samples. Briffa should have explicitly and clearly described what selection criteria were applied, and should have confirmed that EXACTLY the same criteria were applied to both dead and living trees. He did not do this. The in variance in annual growth in the data before the 20th century is much lower than the 20th century variance in the two data sets that show 20th century warming. This made my cherry-picking-BS-antenna signal go off scale. It is simply not credible that there was so much less variance in the pre-20th century part of the record, and similarly lower 20th century variance in the two data sets (out of four) that show nothing extreme happening in the 20th century.

    It is most interesting that once “all the Yamal data” was included in Briffa’s new reconstruction, and once you discount the very doubtful final decade (due to Briffa’s admission of hand selection of rapidly growing trees), there was not much of a trend at all in the 20th century. How can growth in the 1920’s (when instrument temperature data shows the world was cooler) be higher than the 1980’s? The only rational conclusions I can draw are: 1) Yamal trees make useless thermometers, 2) the gross increase in variance in some data sets for the 20th century suggest the possibility of distortion due to cherry-picking, and 3) you can infer virtually nothing about historical temperatures from Briffa’s Yamal reconstruction.

    One prediction you can make with confidence is that Briffa will be a lot more careful in the future about disclosing data and methods. Let’s hope others in the paleoclimate field do the same.

  8. jeff id said

    #4 That’s 378,000 USD!!!! To Briffa and Osborn which will most likely result in another RCS style reconstruction. Jeezuz, I don’t know why I work so hard during the day.

  9. hmmm said

    “A number of factors were taken into account when selecting the most appropriate periods for calibration and verification of the gridded density data against observed temperatures. The most important factor is the identification by Briffa et al. (1998a) of a recent downward trend in the highlatitude tree-ring density data, relative to (and apparently unrelated to) warm-season temperature. This density decline becomes large enough to impair the calibration after about 1960. For this reason, both Briffa et al. (2001) and Briffa et al. (2002a) used only pre-1961 data for calibration of their subcontinental, regional temperature reconstructions. This is a reasonable choice, provided that it is explicitly stated that this approach assumes the apparent recent density decline is due to some anthropogenic factor and that similar behaviour is assumed, therefore, not to have occurred earlier in the reconstruction period – which would otherwise introduce bias in the reconstructed temperatures. At present, no satisfactory explanation of the relative MXD decline has been identified, and further work must dictate whether this assumption will be supported or rejected (Briffa et al., 1998a, 2003, and Vaganov et al., 1999, discuss and investigate possible causes).”

    Wow. I am floored by this paragraph. Of course the hypothesis they come up with is that the lost correlation after 1961 is some sort of purely modern and anthropogenic anomaly. They don’t provide any clue, reasoning, or proof for this assumption and they don’t explore a single alternative. Therefore I am left with the conclusion that these people are truely clueless (lacking a clue). What the **** happened in 1961 exactly?

    And why aren’t alternative theories explored ore even MENTIONED??? For instance, this divergence may simply mean that trees are not always good temperature data loggers.

    Correlation with temperature can clearly be lost. So if you get rid of post 1961 calibration because it doesn’t correlate, you’d also have to get rid of the reconstruction itself because we don’t know what the pre-instrumental correlation values are at all. The post 1961 data is direct proof that you can’t assume the pre-instrumental correlation to be as good as the pre 1961 instrumental correlation.

    A much more “robust” alternative reconstruction would be to calibrate to the entire instrumental period and provide the applicable correlation/confidence/error data based on the entire instrumental period (and assume your pre-instrumental correlations were at THESE levels). Why do I have the lurking feeling this approach wouldn’t provide the hockey stick impact we are so used to?

    Post 1961 is swept under the rug, we are told only to be careful. Here’s an idea, why don’t YOU guys be careful and explore this a little further? Why don’t YOU guys create a reconstruction calibrated against the entire modern period so we can all see how much it “matters”?

    If you’re allowed to make assumptions like this and ignore the alternatives you can make any data you want.

  10. Layman Lurker said

    #4

    Interesting. New innovative methods to determine OMGIWTWT? ;)

  11. #9. This very paragraph made so little sense to me that it prompted me to investigate the reconstructions, starting with the most prominent one (MBH98-99). What is worse is that in the spaghetti graphs, including the IPCC spaghetti graphs in 2001 and 2007, the post-1960 of the MXD reconstruction is DELETED, making the reconstructions look more attuned than they are. As an AR4 reviewer, I objected vociferously to this deletion urging the authors to disclose the discrepancy and explain it as best they could. They flatly refused, saying that this disclosure would be “inappropriate”. Covered in various CA posts.

  12. jnorv said

    Why does nobody use the California Redwood trees in their studies? Some of the trees are 2000 years old and still growing. They are both costal and high altitude.

    Jim N

  13. Jeff Id said

    I was thinking last night that Briffa’s writing style is really amazing. He regularly points out huge flaws in the records and methods just slamming home the fact that it isn’t working. He did it in RCS also, where he points out the exact problems I and others have discussed. Then he follows it up with an arm waive and an over-certain hockey stick conclusion. The paragraph in 9 is a great example of his style. Strong absolute paper killing comment, unfounded statement of reasonableness for chopping data — conclusion.

    Now he’s taken the additional step of getting mad that someone suggested he removed data. HE DOES IT ALL THE TIME!!

    Getting dumber…must stop reading.

  14. Mark T said

    There’s definitely an IQ tax to be paid in making any attempt to understand these guys.

    Mark

  15. Kondealer said

    I’ve emailed Briffa to ask him what is going on.
    No response (so far) to either email.

    Dear Professor Briffa, I am pleased to hear that you appear to have recovered from your recent illness sufficiently to post a response to the controversy surrounding the use of the Yamal chronology; (http://www.cru.uea.ac.uk/cru/people/briffa/yamal2009/cautious/cautious.htm)
    and the chronology itself;
    (http://www.cru.uea.ac.uk/cru/people/briffa/yamal2009/)

    Unfortunately I find your explanations lacking in scientific rigour and I am more inclined to believe the analysis of McIntyre
    (http://www.climateaudit.org/?p=7588)

    Can I have a straightforward answer to the following questions
    1) Are the reconstructions sensitive to the removal of either the Yamal data and Strip pine bristlecones, either when present singly, or in combination?
    2) Why these series, when incorporated with white noise as a background, can still produce a Hockey-Stick shaped graph if they have, as you suggest, a low individual weighting?

    And once you have done this, please do me the courtesy of answering my initial email.

    Dr. XXXXXXXX

    —–Original Message—–
    From: XXXXXXXXXX
    Sent: 02 October 2009 10:34
    To: ‘k.briffa@uea.ac.uk’
    Cc: ‘p.jones@uea.ac.uk’
    Subject: Yamal and paleoclimatology

    Dear Professor Briffa, my apologies for contacting you directly, particularly since I hear that you are unwell.
    However the recent release of tree ring data by CRU has prompted much discussion and indeed disquiet about the methodology and conclusions of a number of key papers by you and co-workers.

    As an environmental plant physiologist, I have followed the long debate starting with Mann et al (1998) and through to Kaufman et al (2009).
    As time has progressed I have found myself more concerned with the whole scientific basis of dendroclimatology. In particular;
    1) The appropriateness of the statistical analyses employed
    2) The reliance on the same small datasets in these multiple studies
    3) The concept of “teleconnection” by which certain trees respond to the “Global Temperature Field”, rather than local climate
    4) The assumption that tree ring width and density are related to temperature in a linear manner.

    Whilst I would not describe myself as an expert statistician, I do use inferential statistics routinely for both research and teaching and find difficulty in understanding the statistical rationale in these papers.

    As a plant physiologist I can say without hesitation that points 3 and 4 do not agree with the accepted science.

    There is a saying that “extraordinary claims require extraordinary proof”.

    Given the scientific, political and economic importance of these papers, further detailed explanation is urgently required.

    Yours sincerely,
    Dr. XXXXXXXXX.

    (XXXXXXXX I have deleted my name for the purposes of this blog)

  16. Steve Fitzpatrick said

    #13. “Getting dumber…must stop reading.”

    I think you are smart enough to be not too damaged by Briffa’s dissembling style!

    What is clear in reading Briffa’s post a second time is that he has very carefully chosen his words to perhaps not disclose everything about certain key points (like selection criteria, uniformity of the application of those criteria, reasons for limiting calibration to pre-1961, etc.), while still giving himself a semantic escape if someone should point out clear flaws or errors. Reminds me a bit of a certain former US President testifying under oath.

    With all the data now available in an easily usable format, I expect there will be additional valid critiques of Briffa 2000 and 2008 which show the papers draw conclusions that are largely or wholly unsupported by the data. You can expect Briffa and the rest of the Team to try their best to maintain enough of a fig leaf that all the studies using Briffa 2000 and 2008 aren’t damaged as well.

  17. stumpy said

    As a modellor I find the logic of paleoclimatologists bizzare. If a tree ring growth record does not match the local thermometor record, this must be explained not ignored! There may be other factors at play that need to be accounted for.

    If I looked at some flow data and simply threw it out as it didnt match my model I would be chastised!

    If only we could assume things are ok and pick and choose the ones that look the we think they should!

    Essentially they are doing the following:

    – Assume tree ring width relates to temperature cos some one else said so
    – Pick the records that look the way we think they should
    – Put them all toghether and form a hocky stick
    – Proclaim that because temperature rose in a few isolated parts of the world that man must be causing global warming
    – Fail to provide any evidence for the claim, other than the IPCC said co2 from humans causes warming
    – Fail to compare the results with observed data to confirm the results
    – With hold data and give obscure defense to avoid detection

  18. MartyH said

    To me the logical thing to do would be assess the trees that do correlate well with temperature and note their particular properties (eg. southwest slopes between 10 and 20% gradient, between 500 and 800m altitude, soil PH6.2, not less than 2km from large water course… whatever it is that makes them special and is unique to the well correlated trees) then go back and sample as many trees with the same or similar local conditions – then see if you’ve still got any correlation. If there is still correlation, hooray, you’ve worked out how to choose temperature sensitive trees. Otherwise I can’t see that it’s anything other than picking noisy squiggles that match other noisy squiggles and deleting ones that don’t.

  19. timetochooseagain said

    MartyH-The response you’ll get to such a thing (incidentally, a great many proxies need this done just to bring them up to the present-see CA “Bring the Proxies up to date!: http://www.climateaudit.org/index.php?p=89 ) is that it’s hard to go out to the “ass end of nowhere”.

  20. Tony Hansen said

    re #9 Hmmm,
    ‘What the **** happened in 1961 exactly?
    Lots of things actually… and Tsar Bomba in October.
    Chance correlation?:)

  21. william said

    Why couldn’t tree growth be a function of trees reacting to increased levels of CO2 in the atmosphere? Perhaps CO2 levels hit a “threshold” post 1961 that ramped up their growth levels.

  22. hmmm said

    #20, the loss of correlation was a gradual trend per Briffa, not a singularity and has not been recovering since to my knowledge. Jeff, did the loss of correlation referred to above by Briffa occur just in Russia or is this a global divergence at the same date? I doubt Briffa would invoke the bomb theory for support.

    #21, I certainly think that’s possible on some levels but there are studies on how increased CO2 increases many plants’ growth and I’ve never seen any sort of threshold that corresponds to 1961 CO2 levels (if there were, Briffa would have used it for support). I’m no expert though.

  23. Ryan O said

    I miss these classic Jeff rants. :D

  24. DrDweeb said

    Dang! William – my heretical thought exactly! I seem to recall reading that there is some evidence to support this idea.

    Tangentially, there are two sides to the AGW hypothesis, and I wonder how ell the CO2 proxies hold up? I have not investigated, and they are likely better as they will likely rely on hard science (chemistry and/or physics), however I was wondering what the state of play was that score.

    Dweeb

  25. DrDweeb said

    Jeff, You realise that Biffra has just arranged to have someone pay the Team 1/3 million dollars to investigate themselves and their own methodologies?

    A quick read of the abstract suggests that this research is definitely what is needed to help sort out this mess. However I word suggest that Briffa et.al. are the least likely candidates to successfully and objectively carry out this research. They cannot lose. No matter what result they come up with, they can spin it to advantage. Think through the possible results and scenarios and imagine the Team position.

    Beer time.

  26. […] a comment » Via Jeff at his blog The Air Vent, in the midst of a post that includes a nontrivial quantity of ALL CAPS and multiple exclamation points (no <blink> […]

  27. MikeN said

    Even Wikipedia’s representative form the Hockey Team has recognized the problem of hockey stickization.

    http://scienceblogs.com/stoat/2009/11/tiljander_again.php

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

 
Follow

Get every new post delivered to your Inbox.

Join 148 other followers

%d bloggers like this: