the Air Vent

Because the world needs another opinion

Delayed Oscillator Fast Reply

Posted by Jeff Id on November 2, 2009

Back in my college days on a whim, a friend and I made an electronic scale from a quad strain gauge taken from a small yet very high precision printing press. The guage was attached to a 0.20 thick piece of spring stainless. Everything except the electronics was taken from junk of one kind or another including a gorgeous precision analog meter from a device I don’t remember. Anyway, during calibration experiments it could weigh a 0.062″ diameter piece of writing paper with a 20% scale deflection so it was very sensitive. Small breezes in the room caused expected problems on the signal by deflecting the beam so the whole thing was intended to go in a breeze tight box.

By the way, the Quad strain guage configuration is good for thermal and side strain corrections so it was ideal and my friend and I spent a week on it. However after it was finished we found an imperceptibly small gust of wind on the room temperature very low power electronics could cause a 5% deflection. As planned, the whole project went into a wind protecting box to guard from any breeze yet because of room temperature drift, the intended accuracy was impossible. We had balanced the electronics plus and minus to account for temp but you know, even our best guess wasn’t enough. – wrong again.

There is no substitute for confirmation of theory.

Today my venting elicited a response from my post on Briffa’s Yamal. DO, is apparently a climatologist of some kind (he’s not saying but if you google DO you can understand) so his knowledge of the detail of the subject of climatology is far better than my own. He’s been quite honest in his email and other replies which is more than I can say for most of the AGW blogs so I’m happy to link.

DO does not understand how a non-climatologist could be frustrated with the state of dendroclimatology and he reacts strongly against the tone of my posts here. I’m not going to apologize for my tone yet am always happy to learn. To explain our (mine and your) collective angst to DO further, I’m forced again to quote Ryan O – “Guage R & R” every dendro should read up on it.

I don’t want to be overly kind so there are several points he claims were made which are misunderstandings of tAV own points. The temptation would be to call them straw men but I think they are honest misunderstandings and reaction to the Air Vents IMO appropriate tone. For instance –

Finally, what of the claim that combining the mean-detrended series demonstrates that the RCS method is invalid?

The careful readers here know that while my comments are often extreme sounding to those who don’t follow, they are constrained slightly better than that. I don’t recall claiming RCS is invalid, nor has anyone else I recall. What’ Ive claimed is that exponential decay RCS does not allow for increased growth in later years as demonstrated by many series and that process is invalid. There are a few other points made which I’ll work on later.

Beyond that, there isn’t much to do except put my own vents and work Link 1 & 2 for reference and Delayed’s new criticisms in link 3 below. Enjoy.

Here’s tAV which you’ve read before:

Hockeystickization Revisited

Fixing Briffa’s Latest

Below is DO’s latest. Check it out, leave comments and questions. Unlike RC, Delayed Oscillator has been quite open to questions and minor criticisms. Let me know if you’re not getting through.

Yamal Reply DO

Those of us who aren’t dendro’s may not know trees to the same degree but we know signal processing pretty well. We know it from a confirmation of experience – of which I have MANY. Our job’s depend on proper metrology and the eyes closed methods of dendro’s would get us driven into downtown by our coworkers and dropped off in a dark alley.

The reply is appreciated but like so many have said in different ways– Confirmation rules physics. We’ll see if I can replicate and explain the difference between his post and mine.


49 Responses to “Delayed Oscillator Fast Reply”

  1. delayedoscillator said

    Hi Jeff,

    Thanks — I think your post sets the stage nicely for useful conversation on this issues. I look forward to it.

    My intent is not to create strawmen — I don’t think anything useful would come of that. Here is the statement I was reacting to:

    ‘none of this is any better than a simple mean.’

    … hmmm, and another I can’t seem to locate now from an even earlier post. I’ve understood you to ‘mean’ that the simple mean detrending is more accurate than RCS. Following this logically, the increase in growth in the RCS chronology would presumably be inaccurate. Is that a more complete representation?

    I notice above a post of yours I had not seen before now that also deals with these issues (I’m afraid that, after the initial post of yours that I linked to, I didn’t really have much hope for useful or civil conversation). I’m interested in the source of our analytical differences, however. I don’t have a lot of time to dedicate to this (my ‘fast’ response was already half in the can, I admit), but let’s see what we develop in this exercise.

  2. curious said

    DO – if you are still around. Can you give me a “fast response” on the worth of the “time varying spline”? I might be asking a dumb question but so far noone has picked it up. Putting a spline through the ringwidths is just a curve fit right? So is an exponential although that has the additional “feature” of a functianal form to drive it. Potentially a functional form that one could support through a causal relation (but I haven’t seen this claimed).

    The spline on the other hand is just a fit within data points – right? Depending on the stiffness of the spline relative to the variance in the data the “anomaly” for any given ring relative to the curve will be defined. So varying the spline stiffness for a given set of data will adjust the “anomalies” of that data set? Ok, so given we are interested in the values of these anomalies as we believe they can proxy for a variable of interest the stiffness of the spline becomes a factor in our results? So what are the selection criteria for the stiffness of the spline? I’ve seen reference on this at CA which was along the lines of a minimisation of the anomaly sum – sorry to be vague – and I sort of ticked it off as “explainable” but without accepting it as a good idea. To my mind this is basically saying you remove as much variability as your smooth can and then look at the remainder. This still seems to have the problem that you don’t know what you are looking at?

    Then on top of that a further arbitrary choice seems to be put into the mix of “time varying stiffness”. So now the spline can be tuned to follow or not the data points. Again affecting the anomolies and our results? On your page you refer to a 67% spline – what is this? I’m guessing it is a stiffness factor for the whole spline as you mentioned you haven’t built the time varying step into your code yet. If so how and why was 67% chosen? If not what is it? How would 60% or 70% etc. affect results?

    If we now go back to the variable stiffness set up – are the same stiffness selection criteria extended to cover the individual portions of the spline? (bender posted up a FORTRAN snippet for a spline which appeared to have them being read in from an external file but I could be wrong). So, if I’m following this correctly, in the end we get results which are arbritrarily biased by the method? In the case of Jeff’s argument over late growth possibilities I see on one of yours you’ve cut the series beyond a certain date and gone for a straight line – an infinitely stiff spline? Pleased to hear your comments and to be put right if I’ve got this all wrong.

  3. Jeff Id said

    #1 Regarding simple mean de-trending being a-priori more accurate I wouldn’t agree with that statement. The simple mean should be used as a sanity check though and not much I’ve seen proves increased accuracy. However, there is nothing I see wrong with RCS in general. The pitfalls are different trees in different regions should probably be corrected separately and whatever method you use for correction should be flexible to allow for general growth or general shrinkage of tree ring width over the chronology. The exponential decay of Briffa deserves to be kicked for that – especially after denying data and methods to the field for 9 years.

    The final and larges pitfall is assumption that the result is temp.

  4. michel said

    DO, the reason many lay people have just stopped believing in Dendro or indeed reconstructions at all, is the inexplicable behavior of the RC people. This may not be a scientifically definitive reason, but its what we have to go on, and its sort of what we use when we assess our investments. We have the refusals to publish data and algorithms, the outright loss of raw data as by CRU, the revelations about for instance Yamal when the data is finally published, the misuse of statistics as in Mann’s PCA, and its defenses, as in Tamino. At some point we say to ourselves, this is smelling like Enron or like Madoff. No, its not proof of any sort, and its not proof were it proof that something is wrong, of how it went wrong.

    But it smells bad, and so it contaminates all dendro for us, all reconstructions, and it then goes on to contaminate the whole movement. My sincere advice to the AGW movement would be, begin by cleaning up RC, admit the errors of the Hockey Team and purge them, and then maybe you guys could start to recover some credibility. You don’t do that, get ready to have disbelief mount. Your choice. A choice that matters, and especially matters if AGW really is a pending urgent catastrophe. Which right now, it smells decidedly of fish. But that could be just because there is one bad barrel in the hold. Or it could be the whole lot is rotten. We cannot know till you get it all out on deck. But in the interim, we ain’t buying.

  5. Ryan O said

    Dendros have problems other than the behavior of RC. 😉

    Failure to understand metrology is but one of them. I’ve personally set up two metrology labs, and the claims of accuracy/precision in climate literature (dendro stuff especially) is laughable.

  6. delayedoscillator said

    #2, Curious,

    Sorry, quick response🙂 I’m not familiar with the time-varying spline itself. The paper you’d be interested is here, and includes a FORTRAN snippet:

    http://www.cru.uea.ac.uk/cru/people/melvin/Melvin2007/Melvin2007.pdf

    In general, however, the spline is fit to track some frequency (say, equivalent to 67% of the series length) with a certain amplitude of the response at any given frequency. So, in my example, the spline is designed to preserve 50% of the amplitude of the original series at periods equivalent to 67% of the series length.
    As I think you’re getting at, and as many thing in life, the spline fit is a compromise between how close it is to the original data points and how smooth it is.

    Now, the real question is empirical — how sensitive the result is to, as you say, 50% vs 75% segment length (or a fixed cut off, etc). In my post, I’ve shown that in developing the regional curve, stiffer fits to the data (the negative exponential) doesn’t give a much different answer when you follow it through to the mean chronology. In this case, most (stiff) frequency response functions yield the same answer.

    One place the spline can be useful is in forest environments with growth suppression and release.

    #3, Jeff,
    Checking the influence of any one detrending method is good practice — you see it quite a bit in the literature these days and I do it myself. The problem with detrending trees individually is the ‘segment length curse’, which becomes a problem if your signal has a longer characteristic time scale than your individual tree ring series. Hence RCS. I’m not sure I understand your dislike for the negative exponential curve fitting — it emerges theoretically from tree biology and geometry and is quite appropriate in certain environments. And, as I’ve shown, doesn’t strongly influence the final mean chronology when used in developing the regional curve in this data set.

    #4, Michel
    If you can’t stick to the science or the topic, I’ll be out of here pretty quick. Nearly all of what you say is incorrect in my experience, training, interactions with the public and with the larger scientific community. Moreover, if you’re willing to throw away all the fundamental research in physics, chemistry, biology, and geology that informs our understanding of human influence on the ocean-atmosphere system because more websites have made you not like me and my colleagues, there is not much I can do.

  7. delayedoscillator said

    #5, Ryan,

    Do you consider that a productive, content-rich comment?

  8. Kondealer said

    Delayed Oscillator, many thanks for agreeing to respond to questions.
    I have a few questions/comments that you may wish to reply to:

    Jeff (and others- e.g. Steve McIntyre at “Climate Audit” have shown
    1) By changing the curve fit to the data, you change the outcome of the reconstruction. i.e. the conclusions are related to the method of analysis = “non-robust”
    2) One only gets the 3 sigma events (uptick) as the number of replicates fall below 20 = “non-robust”
    3) The reconstruction is highly sensitive to the choice of the end point = “non-robust”

    Speaking now as a plant physiologist (my speciality) I note that old trees make up most, if not all, of the uptick. Old trees, by definition, are survivors, perhaps the ones most genetically able to make the most of a small amelioration in climate. Be it temperature, CO2 fertilisation, lack of competition from younger trees that died in the “Little Ice Age”, whatever? You tell me- how does Briffa KNOW it is temperature?-He can’t and what is more he can’t prove it. Correlation is not proof of cause and effect.
    The fact of the matter is that these “survivors” are atypical and not representative of the earlier parts of the chronology = “non-robust”.

  9. Layman Lurker said

    #6

    I believe that negative exponential RCS is a potentially useful model of tree growth. As you say it makes sense from a theoretical perspective. However, one of the critical preconditions for a robust RCS is population homogeneity. If a chronology mistakenly includes and interprets the growth characteristics of a sub population as signal then the result is biased.

    Another critical issue with RCS is that it has no way to distinguish true signal from other – non temp – environmental signals at the end points. Noise interpreted as signal in this manner will again create bias. Is this type of uncertainty accounted for?

    One of the frustrations that I see is that many of these questions could potentially be answered if there was a traceable path from sample protocols, to sample archive and metadata, to data screening, to method.

    DO, I would welcome your comments on these matters.

  10. Mark T said

    Another critical issue with RCS is that it has no way to distinguish true signal from other – non temp – environmental signals at the end points.

    That’s a “problem” with any blind source separation algorithm. The algorithms don’t put little post-it notes on each extracted “signal” to identify their corresponding sources.

    Mark

  11. John F. Pittman said

    RCS as implemented has problems. Kondealer hit on some. I would like to expand this to include the problems with non-stationary. The first inclusion is something that should be included in the theoretical model. It is known from past work, that trees do not grow nor live forever. The first item to mind is that if RCS was a method that was robust to the stationary problem, then a population study of massive numbers of the specie(s) in question would have an accepted curve. Then changes with respect to that could be examined from a basis of what is explainable. One of the problems with engineers, and biologists have with this and other methodologies, is the lack of this accepted curve. To show the relevance, consider the age differences of the trees. Do trees that die young, or die old differ from the standard curve, and in what a priori manner do they do so? From the practical point, leaf growth is also influenced by rain, and nutrients. With timerline tree loss (changes) documented in the peer reveiwed literature, how many years, and how much will the increased nutrients and increased water supply from the rotting roots affect the basic theoretical model, and how can this be implemented in actual studies? Does this specie(s) have a symbiotic fungal relationship that behaves in a logarithmic or semi-log relationship to nutrients or growth of the tree? They are questions related to the theoretical model, but these are good examples of documented conditions and responses that would challenge an assumption on stationary and would make explanation of non-staytionary phenomena difficult if not impossible without large meta data possibly on the order of centuries without some independent measurements of some physical biological expression by the specie(s).

  12. Paul said

    I think that the real problem with treating trees as thermometers is at the level of biology, not what particular function to fit to the tree rings. Until it can be shown, which I doubt, that tree ring widths have a simple relation to temperature, all this work is meaningless. You’re just doing a nonsense fit of a one dimensional function to a multi-dimensional nonlinear system that depends on water, CO2, nutrients, sunlight, disease, …, and marginally on temperature. And don’t forget genetics. Trees reproduce sexually so there is no guarantee that any one tree responds the same as another in the same environment, just like children in a family.

    The case would be a lot more convincing if a bunch of trees were grown under controlled conditions and all these variables worked out before hand instead of just ad hoc assuming that rings are thermometers. But experiment doesn’t seem to fit into the “climate science” paradigm.

  13. curious said

    DO at 6 – many thanks for the informative reply and the link to the paper – at this stage I’m afraid I still don’t see the fundamental step of a causal relation driving the curve fit. The paper linked strikes me as using a mathematical “shuffle and bend” equivalent to using a manual flexible curve to produce a “best fit”. As to the implications for study results and conclusions I can’t go further without crunching the numbers, and there are plenty here who are way ahead on that front. Hope you pick up the dialogue. bw C

  14. Tony Hansen said

    As I see it, the response from DO to Michel is the same ‘textbook’ response that so many have heard so often over the years.
    Because DO has had no problems then the problems others have found can’t possibly exist.
    And if Michel won’t back up/off then DO will pack up and leave (‘I’ll be out of here pretty quick’).
    And for DO to say ‘..if you are willing to throw away all the fundamental research in physics, biology,chemistry and geology..’ is, in no way, an honorable interpretation of Michels comment.
    DO, you want Michel to stay on topic but then go out of your way to create your own strawman.
    Jeff, snip me at will if you wish.

  15. Jeff Id said

    #14 I don’t snip here typically. Probably less than 10 times in 9600 comments and 600,000 views now (about a weeks worth at WUWT), I’ve edited a bunch of cussing out and put TCO in the moderation bin which ultimately drove him away but people are generally free to make as big an idiot of themselves as the host. I do agree that he’s awful quick to call an ultimatum like that but what can we do. He can moderate how he likes but I see it as a form of censorship and prefer open real time posting rather than an agreement blog which is what moderation usually turns into. DO will need to have thicker skin IMHO to mess with the public.

    By the way I’m rather proud to have a growing group of adult professional readers who don’t require constant babying and despite all that open air, tAV doesn’t have as bad a tone as RC or Tamino.

  16. Carrick said

    #13 Curious:

    DO at 6 – many thanks for the informative reply and the link to the paper – at this stage I’m afraid I still don’t see the fundamental step of a causal relation driving the curve

    Spline fits assume that the the underlying curve is locally differentiable. Since we know that most physical systems give solutions that are at least C2-differentiable, it is not an unreasonable semi-parametric model to use to constrain the data. This is probably better than e.g. running means, which don’t make any such guarantee (as far as I know).

    The biggest problem I have with tree cores is the fact they aren’t really concentric circles.

    See for example this annotated image of a tree ring (original is here).

    The red dashed lines are presumably the rays along which the core samples will be obtained, since by construction they pass through the geometric center of the tree (as defined by its external perimeter). The green lines show the actual trajectories that are perpendicular to each successive tree ring. The black lines illustrate the actual tree ring shape for three years, the red dash lines are for reference to show how the tree ring becomes more progressively out of round as you move closer to its origin date.

    You can see from this example just how noisy a few tree cores (or even one core, as is sometimes used in dendrochronology) would be at measuring the real growth of the tree over time. I just don’t get how somebody could rely on just 12 trees in the calibration period for purposes of reconstructing climate:

    If you look at the tree, and follow the lines of constant growth (the green lines), clearly the main source of changes along a given line over time is mainly controlled by changes in the ellipticity of the tree, rather than climate or any other facts, especially as you approach the “young tree” part of the growth patterns.

    It is pretty clear that there should be pretty substantial errors in tree-ring measurement if you are relying on cores rather than using a cross-section. If you note the “young part” of the tree ring growth is generally more significantly influenced because it is more “out of round” than the older part.

    I am itching to have enough time to sit down with a good dendrology text book and see how well they handle some of these issues. (Maybe over Christmas break if I’m lucky).

    To paraphrase Jeff, there’s nothing like getting your hands dirty on real data.

    The sad thing (to me) is how much work needs to be done to establish whether tree rings should be used in paleoclimate temperature reconstructions at all, and I’m like Jeff, I’m guessing it’s a waste of time.

    I’m figuring unless you have really big N, the red noise associated with normal changes in tree ring growth over time are going to very much dominate over the much smaller climate forcings, especially if you are relying on tree ring core samples rather than tree ring cross-sections.

  17. Carrick said

    DO:

    If you can’t stick to the science or the topic, I’ll be out of here pretty quick

    Sounds like a cop-out to me. Politics and science have always been enmeshed.

    Since when do scientists stay “on topic” on anything? Or not argue over politics? You haven’t been to enough conferences yet if you think otherwise.

  18. curious said

    Carrick – thanks, good stuff and I share your (and others) reservations on tree rings, their geometric properties, sample type, orientation, collection and meta data issues. On one thread at CA someone flagged up what sounded like a good standard dendro text which I checked out but it was a £65 amazon job so I didn’t go any further. Sorry I don’t recall the poster or the author – if you put a request on unthreaded I dare say it’d get picked up. Now I’ve seen a bit of the scanning approach to tree sections I think an image processing approach to dendro would be worth pursuing and I’d be surprised if it’s not out there. All sorts of possibilities for area based assessments and geometric quality measures come to mind. If the samples exist it could be a useful exercise to take an existing “cored” or rw based study and see what else could be learned from it (either about the tree growth itself of the validity of the rw approach) by using area evaluations.

    On the spline business – sorry but I’ll have ask you to slow down. Please can you briefly expand on what you mean by C2 differentiable and how this fits with life, esp. growth, processes? A quick google skim suggests it’s a harmonic function? My spline knowledge is from the drafting angle and I always view it as essentially a polynomial which you can pull about or splice to suit your (aesthetic) needs. I don’t view it as having any data integrity like a generating/causal function could have – any frequency characteristics would be imposed on it by the knot point choices and therefore are arbitrary IMO. I’d value your comments or point me to a ref. I can check out. Thanks.

    Re: Xmas – yes, do you know what I fancy? Ryan and Jeff et al’s paper, a good couple of texts, peace and quiet and a good cup of tea on a “Which Antarctic do you want coaster?”. OMGIWTIT – what has the internet done to me??🙂

  19. David Jay said

    Carrick #16

    But it’s not even 12 trees, its “The One Tree”: YAD061

  20. Jeff Id said

    #18, You asked DO about splines and got a partial answer so I left it alone.

    Splines are a parametric fit of polynomial equations. Cubics are the most common. This is because in fitting multiple ax^3+bx^2+cx+d together. The cubics are transposed into time based rather than coordinate based math – t instead of x which allows them to wrap back on themselves (parameterization) my x based equation cannot wrap back. I don’t know your math background so if it doesn’t make sense don’t worry. The splines are multiple sets of cubic equations with the tangent (slopes) of the mesh points set equal. The ‘knots’ determine different things depending on the form of spline, in most cases. In your graphics the knots are the points you drag and in different splines the knots can sometimes on the line or not.

    What got me to finally answer all this though was your perfect description of the arbitrariness of spline characteristics.

    any frequency characteristics would be imposed on it by the knot point choices and therefore are arbitrary IMO.

    Arbitrary frequency isn’t the worst thing in the world, consider what happens with filtering choices. If you look at the esper quote on CA and SteveM’s latest comment he points out that not only is the frequency arbitrary but the frequency is chosen to not diminish the older trees. The math is basically the same as though knots are moved or in some cases eliminated in older tree data.

    This choice is VERY VERY highly suspect IMHO. I almost ranted about it again. Consider that the oldest trees make the HS, the oldest trees in my brief experience have shown an upslope after 200 years and these guys are finding hand waiving explanations for why the old signal should not be corrected to an equal degree as the younger.

    I don’t know, there’s a lot to write about with little reward. In the meantime DO could shed some light on it. Maybe we’re missing something..

  21. Jeff, did you notice Craig Loehle’s mention at CA of Michael Huston’s reporting of bimodal distributions in sizes from trees with random spatial distribution. He’s going to send me pdfs. It looks intriguing – the emergence of “alpha trees” is something that sticks out in the stats.

    Also it seems to me that it’s not as though any individual tree follows a neg exp curve very well – though a population can. Sometimes it seems like trees grow strongly (but consistent) and then switch to a slow growth mode and then die. Different trees seem to “retire” from fast growth at different ages.

  22. Carrick said

    Curious, Jeff ID has mostly explained the character of splines. The one thing I wanted to follow up on is a C2-differentiable function is just mathese for a function whose second derivative is continuous (or equivalently the curvature of the function is a continuous function). That’s pretty much guaranteed in most physical systems, so C2-differentiability is sort of a minimalist requirement for physicality of any system.

    With 3rd order (I think I have that right) cubic spline of the form that Jeff described, you will get a continuous function, a continuous first derivative (slope) and second derivative (curvature). Jeff is technically correct that splines don’t limit the frequency content of the fitted spline. In practice, this rarely is a significant problem, and I would suspect that the loess spline is in general a better choice than the exponential decay model (which actually looks to be a flawed model based on my very limited exposure to dendro data).

    If you know a priori that your signal is band-pass limited though, you are better off building that information into your optimization function than to not do so. The more (accurate) constraints you can place on your fitted model of the data, the better you will be able to separate the signal from the noise.

  23. delayed.oscillator said

    Jeff et al.,

    There is a lot going on in the comments here and it goes in a lot of different directions, some of them interesting and genuine, some less so, naturally. If any of y’all have particular comments or genuine interest (I’ve had some great questions lately) on the analysis I’ve done, you’re welcome at my blog (the ‘lion’s den!’ as Jeff’s conversation with Steve Mosher at Climate Audit so poetically puts it) to engage in scientific discussions.

  24. Jeff Id said

    #23, My new Leonid friend, don’t fear. We boys (and girls) are pretty sharp despite some of the snark and growling. Ursa’s cave is not so dangerous. Wolf meat for all to share.

    All kidding aside, several reasonable questions were asked. I’ll make sure the ‘tone’ doesn’t go too far, you might find an easier moderation method from it. Remember, not many in the cave are paid to study climatology day in and out. We study other sciences and blogging is a form of entertainment for the non brain dead, you will find reasonable points will be accepted with surprising ease.

    Tell the group that tiljander is ok upside down and fur will fly🙂 Tell them it’s wrong and it doesn’t make much difference in Mann 08 and you’ll get a bunch of fans.

  25. michel said

    Yes, the problem is, we have to make up our minds about this, because we are voters. Alas, we are not climate scientists, but that doesn’t take the decision away from us. I am not a heart specialist either, but I had to decide whether to have Maze in addition to the other procedure during my recent heart surgery. No getting away from it, very pleasant young man sitting by the bed, is it yes or no tomorrow? I understand well that in DO’s view, he has given a fully adequate response that should settle the matter for me and people like me. And in a way it does. I also understand that he doesn’t want to discuss these issues here, and that’s just fine. I wasn’t hoping to discuss, just to say to a genuine climate scientist what the social climate is like in the city where I live. Thanks for the reply.

  26. Jeff Id said

    DO asked me why I would be interested in a field which has so many problems on his blog. My reply was partially snipped although I didn’t consider it impolite. Paragraphs 2 and 3 below were removed but I’ll put it here for people following along, it makes a bit more sense when seen in full.

    I should mention that DO disagrees with my point about Yamal and RCS and says his recent curve is the same thing – red link above. It doesn’t look like the same goalie stick to me. Anyway I didn’t get any time last night as we went to a friends house and watched the Red Wings shut out the Bruins. Go wings!

    ————-

    DO,

    One thing I learned many years ago is that no matter how deeply you believe something you can always be wrong.

    Typically, the responses from Mann, RC and other dendros are off topic, censored and often unable to address reasonable criticism. – as paraphrased from Mann PNAS- the claim that a proxy is used upside down is bizarre as multivariate methods are insensitive to sign.

    An honest person has got to admit that’s a pretty big whiff, apparently completely missing the point while everyone knows exactly what happened. Mann messed up in an obvious yet likely immaterial way and instead of simply saying yup, I’ll fix it we get that goofy reply. This kind of nonsense is what leads to disdain.

    So in short, I’m curious about the methods of reasonable scientists on this topic. When we see that the original yamal with its huge blade is a result of RCS and an assumed rather than proven growth curve – that’s a problem and it’s frustrating.

    Also, I’m interested in different proxies as the dendro stuff is not terribly convincing at this point. I am open to being convinced but there are some very difficult questions to answer. Linearity, divergence and correlated overlayed signals. These are not problems that can be answered with a hand waive.

    Can you explain why your exponential fit didn’t replicate the original Yamal in the last graph?

    The last plot has a fairly similar look to the results I got including all of Briffa’s data from his sensitivity test and the Schweingruber set SteveM chose.

  27. Mark T said

    DO apparently does not realize, either, that there are people in here (and certainly over at CA) that don’t just use these various extraction methods, but have spent a significant amount of their time developing such methods. When the obvious mistakes are made, such people tend to take notice. When their criticisms go unheeded and they get labeled “amateur” or “incompetent,” they have a legitimate reason to show disdain toward those defending such practices.

    Mark

  28. Tony Hansen said

    So DO censored the paragraph with ‘censored’ in it?

  29. Jeff Id said

    #28, Irony abounds. He asked why the disdain and should have expected an explanation.

  30. Kenneth Fritsch said

    If any of y’all have particular comments or genuine interest (I’ve had some great questions lately) on the analysis I’ve done, you’re welcome at my blog (the ‘lion’s den!’ as Jeff’s conversation with Steve Mosher at Climate Audit so poetically puts it) to engage in scientific discussions.

    DO, to be blunt, I base my reply to your offer on what you bring to the party here with your visit. I respectfully decline and chose to devote my valuable time reading and posting on scientific issues here and at CA –and, of course, the occasional venting that is allowed here at tAV.

  31. Carrick said

    Jeff ID:

    DO asked me why I would be interested in a field which has so many problems on his blog. My reply was partially snipped although I didn’t consider it impolite. Paragraphs 2 and 3 below were removed but I’ll put it here for people following along, it makes a bit more sense when seen in full.

    This is exaclty why I’m not bothering with Delayed Oscillator or taking him seriously.

    He can stay on his own blog with his private echo chamber, all the while pretending that his hyper-controlled comment threads encourage open scientific discourse all that he wants.

  32. curious said

    Jeff, Carrick – thanks for the follow up. My maths is rusty so I can’t put in the legwork as I’d like to – the investment of time to get back up to speed and fluency is not possible at the moment and, as noted above, it’s clear there there are plenty here and at CA who are well “on the case”. I’d also have to pick up new stuff as signal processing seems very relevant and this wasn’t included in my syllabus, though I have reservations over the signal/noise distinction in these recons. There was some basic stats including a long forgotten bit on stochastics.

    On splines, yes I recalled them as parametized polynomials (typically cubic) and I did a quick wikipedia read before posting the first comment to check I wasn’t completely out of touch. However I’m not confident I understand how they are being applied to the ring width standardisation problem. What I am confident of is the arbitrary nature of the amplitude and frequency a spline has and hence alarm bells ring when I hear it discussed for use as a correction procedure without there being analytic criteria for determing its “fit”. As far as the natural process bit goes my understanding is typically growth functions follow exponential relations and hence I could see this being supportable as a ring width correction factor if it weren’t for the later life growth uptick.

    Other cyclic time varying physical processes I could imagine to follow a harmonic function but IMO to correct for this in a growth series means you are eliminating/imposing that which you seek. As ever happy to be called on this if I’m off track. And interesting to see Steve’s reference to “Alpha trees” – I’m not sure if this a genetic reference but IMO this must surely be a factor? I have seen other posters suggest this in the past but get quite short shrift?

    As far as use of splines for data analysis goes I thought I’d look into loess splines as Carrick mentioned them. These are new to me so a quick Google pulled this up:

    http://www.who.int/whosis/mort/20080306mtg_Present_Day1_Session6_DrSilverwood.pdf

    which I found interesting for several reasons:

    1) It seems that they are using a linear spline – I wonder whether this would be enough for dendro? Given the number of data points I could see that cubic could be an overkill?
    2) There are definite selection criteria for the tuning parameters
    3) The tuning parameters have significant effects on the results
    4) Uncertainty in method and sampling are discussed
    5) The error bands are potentially very wide

    Interested in other views on the above – apologies if it is irrelevant or ground already covered.

    Re: bigger picture – I think the videos are brilliant. Graphic reminder that the Earth’s climate system is a dynamic beast and, as the heat flows around it, one area warms as another cools – I think this is relevant to the signal vs. noise issue. Looking forwards to looking at them and the new arctic post properly.

  33. Mark T said

    curious said
    November 4, 2009 at 7:50 pm

    I’d also have to pick up new stuff as signal processing seems very relevant and this wasn’t included in my syllabus, though I have reservations over the signal/noise distinction in these recons. There was some basic stats including a long forgotten bit on stochastics.

    Unfortunately, the viewpoint offered by statistics via signal processing (my education) is somewhat different than that used by folks like RomanM, Hu McCulloch, et al., and requires a certain “filter” to translate between the two. There are at least a couple others I won’t specifically name that do walk this line rather well (not me – it takes effort to see things from the stats point of view). I think the biggest difference is that signal processing folks start with rather well known signal and noise expectations, even if neither signal nor noise has explicitly known characteristics a priori.

    Mark

  34. Carrick said

    Mark T:

    . I think the biggest difference is that signal processing folks start with rather well known signal and noise expectations, even if neither signal nor noise has explicitly known characteristics a priori

    I’m not really a signal processing guy by trade, but I do a lot of it, especially related to calibration (in particular) and characterization (more generally) of physical systems.

    I have noticed that there aren’t that many people in the climate field who have signal processing experience (with a few notable exceptions). On thing I certainly would do differently in calibrating dendro data is use the frequency domain, since it’s quite an assumption that the amplitude and phase response of a tree to a relatively high-frequency signal is the same as its response to a low frequency one.

  35. Mark T said

    I’m the other side of the fence: detection. I suppose a big part of detection, at least in communication systems, is channel estimation, which is similar to characterization.

    Yes, btw, that is quite an assumption they make. I can think of a few others that are similarly unfathomable.

    Mark

  36. Kondealer said

    Steve McIntyre spotted this paper “Tree growth and inferred temperature variability at the North American Arctic treeline” (http://www.st-andrews.ac.uk/~rjsw/all%20pdfs/D'Arrigoetal2009b.pdf) by D’Arrigo et al

    Its conclusions are a bombshell to “dendroclimatology”.

    “Our analyses appear to support the hypothesis that drought stress
    can be an important factor contributing to recent loss of temperature
    sensitivity at some northern tree-ring sites (e.g. Fig. 7, Jacoby and
    D’Arrigo, 1995; Barber et al., 2000; D’Arrigo et al., 2008). In a previous
    paper (D’Arrigo et al., 2004), we identified a temperature threshold to
    explain decline of tree growth at an elevational treeline site in
    northwestern Canada. Results showed that this threshold level had
    been consistently exceeded since the 1960s due to warming,
    demonstrating that even under treeline conditions trees can be
    negatively affected when temperatures warm beyond a physiological
    threshold. Precipitation may have actually increased in some regions
    of north central Canada in recent decades”

    Mind you none of this is news to a plant physiologist!

    1)Drought can overide temperature changes but can be missinterpreted as a “temperature” signal
    2) Higher temperatures can lead to lower growth- which has been missinterpreted as lower temperatures in reconstructions
    3)Recent precipitation has increased = more growth- which has been missinterpreted as higher temperatures in reconstructions

    And this is before we get into the complications of increased CO2 increasing growth.

    Would DO like to comment on the implications of this paper to previous reconstructions that use tree rings as a temperature proxy?

  37. delayedoscillator said

    17, Carrick
    Heh.

    24, Jeff:
    Wait, which place is the lion’s den, again?🙂 I thought my place was the lion’s den?

    25, Michel:
    My point was that it doesn’t make sense (to me, at least) to reject the entire volume of work from physics, chemistry, biology, and geology that support the pathways through which human influence the climate system (writ large) because you’ve come to believe (again, in my opinion, incorrectly) that temperature reconstructions from tree rings are of no worth. Of course you (and all of us) have to make decisions (in this case, political, social, economic, and personal), but I also presume you want to make the best decisions, as in your analogy, and speaking only for myself, that means holistically understandings and weighing all the data.

    36, Kondealer:
    I think you’ll find (maybe to your surprise?) if you read my blog that I have repeatedly emphasized that I think divergence is a big deal and an important area of research for temperature reconstructions, although thus far no single cause can explain it everywhere it has been ‘detected’. I actually favor a combination of climate (change in limiting factors at some site) and statistical issues (including ‘detection’, see recent papers by Jan Esper), as the cause, but there is much work to be done.

  38. Jeff Id said

    #37 All kidding aside, I got a chance to run some of the data last night. My initial results for trees dead before 1900 are quite similar to your own. Unfortunately there just aren’t very many trees but we can only work with what we have. It’s taking time but I should be able to post something tonight.

  39. Harold Vance said

    I’m fascinated by the degree of confidence that dendros have in their ability to find the most temperature-sensitive trees (a priori no less) and to read and analyze the rings and thence convert the widths into temperatures with decent confidence intervals. And the analysis takes place without benefit of the actual data for the vectors (affecting growth) for the periods in question, some of which span hundreds of years.

    To me, this is one of the most interesting stories playing out in the blogs right now.

  40. Mark T said

    delayedoscillator said
    November 5, 2009 at 12:34 pm

    My point was that it doesn’t make sense (to me, at least) to reject the entire volume of work from physics, chemistry, biology, and geology that support the pathways through which human influence the climate system (writ large) because you’ve come to believe (again, in my opinion, incorrectly) that temperature reconstructions from tree rings are of no worth.

    There are two strawmen arguments here, DO. First, nobody is rejecting any of these things not the least of which is man’s ability to influence the climate. Second, neither are we saying temperature reconstructions from tree rings are of no worth. Certainly you can’t get that from Michel’s statement in 25. In general, from what I’ve seen at least, most people that would fall under the skeptic side of the argument simply think that both claims, i.e., man’s influence on the climate and temperature reconstructions derived from tree rings, are overstated.

    Of course you (and all of us) have to make decisions (in this case, political, social, economic, and personal), but I also presume you want to make the best decisions, as in your analogy, and speaking only for myself, that means holistically understandings and weighing all the data.

    Indeed, we do need to make the best decisions, but it is arrogant of you (anyone, for that matter) to imply (or assume) that we might not be weighing all the data.

    Mark

  41. delayed.oscillator said

    40, Mark:

    In my arrogance, I can only refer you to the second paragraph of Comment 6 (Michel). He quite clearly says that if we don’t ‘clean up’ and ‘purge’ ourselves of certain members of our community that he has identified, that we should ‘get ready to have disbelief mount. Your choice. A choice that matters, and especially matters if AGW really is a pending urgent catastrophe.’ As to whether Michel is considering all the evidence of anthropogenic climate change: ‘But in the interim, we ain’t buying.’

    You can have the last word.

    39, Harold:
    Can I presume you have the same ‘admiration’ for all retrospective sciences (geology, evolutionary biology)?

  42. Jeff Id said

    #41, I was planning to watch the hockey game at home. It looks like I’ll be doing it with friends so there is another potential delay, I’ll mess around with it some I’m sure but it might not get finished. Hockey is way more important than climate though. Go Wings!!

  43. delayedoscillator said

    42, This we can agree on🙂 Go Wings!

  44. Mark T said

    40, Mark:

    In my arrogance, I can only refer you to the second paragraph of Comment 6 (Michel).

    Which, as I noted, has nothing to do with any of the points you made.

    He quite clearly says that if we don’t ‘clean up’ and ‘purge’ ourselves of certain members of our community that he has identified, that we should ‘get ready to have disbelief mount.

    He is correct: alarmism and dishonest dealings are working against the cause of science. Every time one of these data manipulations (or related issues) reaches the public ear, the public becomes increasingly distrustful of even honest scientific work. Perhaps more honest scientists should make a stand against the nonsense?

    Neither of your follow-up comments addresses your strawmen, btw.

    Your choice.

    I chose to make an argument without fallacy.

    A choice that matters, and especially matters if AGW really is a pending urgent catastrophe.

    Another red herring. I said nothing regarding this.

    As to whether Michel is considering all the evidence of anthropogenic climate change: ‘But in the interim, we ain’t buying.’

    I’m curious how you can conclude he is not considering all the evidence by that statement? Because he does not “believe” as you do, he obviously is ignoring evidence? That’s the arrogance I was referring to.

    You can have the last word.

    Perhaps if you actually addressed my points, and withdrew your fallacious arguments, I wouldn’t have to. Instead, you follow up with more fallacious arguments.

    Mark

  45. Mark T said

    In my previous post, the “as I noted” should be removed. It should simply say “Which has nothing to do with any of the points you made.” Referring, of course, to the comments I was originally addressing in #40.

    Mark

  46. Harold Vance said

    41, DO:

    If I had to choose geology, evolutionary biology or dendrochronology for further study or for vocation, I’d probably pick the latter. Second choice would be evolutionary biology, especially if it involved genetics.

    Is not the flip side to my original comment that the field has a bunch of interesting and challenging problems remaining to be solved?

  47. Kondealer said

    DO, if you think “divergence is a big deal”, then it must by logical extension invalidate the all the reconstructions of past temperatures that use tree ring proxies.
    The point being is that if we can see divergence now, how do we know it didn’t occur in the past?
    More to the point, for the science to be “robust”, how can we PROVE it didn’t occur in the past?
    Also who is to say that trees like Yad061 do not show a “divergence” problem either?
    Its recent growth does not reflect local temperatures, yet it has a wholly disproportionate effect on Kaufman’s reconstruction.

    I’m still waiting a scientific explanation for these “inconvenient truths”.

  48. “Divergence” was very much on the table during the NAS panel presentations. Cuffey was very interested in the issue and the answers by D’Arrigo, Mann etc were nonsensical and didn’t appear to satisfy Cuffey or anyone else. But in their report, they completely dodged the issue – shamefully relying on a cargo cult explanation also referred to by IPCC.

  49. Jeff Id said

    Jeffinition for Dendros. Divergence occurs when an allegedly dependent variable demands independence.

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

 
%d bloggers like this: