Crowley and Lowery (2000) still cited quite often purported to show that the MWP was a dog’s breakfast of odds and ends – very different from the Modern Warm Period. The “proof” was the presentation of a hodge podge of proxies, which supposedly did not show a MWP, but did show a Modern Warm Period. Central Colorado” was one of the series, citing an early Lamarche paper.
As noted elsewhere, Crowley lost his collation of original series and couldn’t remember where he got the digital data from (but acknowledges Jones). Crowley’s “Central Colorado” series is very likely a transformation of Lamarche’s original chronology (Crowley used some really OLD versions) – Crowley’s transformation standardizing proxies to a rank range between 0 and 1. Here’s a comparison of our extended Almagre chronology and the smoothed Crowley version (which he did manage to locate). While the two versions track one another more or less, the updated version has reduced values in the mid-20th century and ends at pretty much the long-term median.
In this case, one can argue that the MWP was as elevated as the Modern Warm Period (although a more likely interpretation is that the data is not a thermometer.)
97 Comments
Interesting to see this graph, like the temp from the GISP2 core, shows peaks about 210-220 years after dips, which is probably caused by volcanos. Is this from somewhere close to Greenland? So we (or this location) still have a few years to reach the max after the last VEI 7 volcano, Tambora in 1815.
Yeah, it’s teleconnected to Greenland. 🙂
That’s one heck of an offset around 800 (depending on the actual scale, which is a little confusing).
#2, “Yeah, its teleconnected to Greenland.” LOL
Is teleconnection the next hockey stick to be band-sawn and spoke-shaved in the labs and libraries?
I really don’t get the resistence to the idea of teleconnections. What is ridiculous is to believe that you can know them for certain at some time in the distant past. But, what if there were more hurricanes in the Gulf of Mexico and the Carribean? Upwelling would cool the waters feeding the Gulf Stream. This cooling would end up in Europe, probably greatly attenuated, for sure. Many reasonable examples can be though of. La Nina causes wind shear which reduces hurricanes in the Atlantic, which deprives the southwest of an important source of moisture. BCP’s produce narower rings. Of course creating plausible scenarios like the above is nothing like proving that, even were we to describe the nodes and connections that exist today through use of satellite technology, that we could then project them into the past.
#5: All you have described is connection through a chain of cause and effect events. Just because you find this to be true on several occasions, does not mean that it is always true.
Off-topic, guys. I’m guessing there will soon be a teleconnections thread, JEG willing.
#5 Yorick
There is no objection to the idea of teleconnections. The problem is that it is used in post-hoc fashion to justify a desired result. In the absence of compelling physical reasoning and solid correlations it should not be invoked. The chance for spurious correlation increases substantially the further the effects are from one another.
#5, Yorick
Perusing past posts I suspect at least 50% of the resistance by newcomers to the concept is, ab initio, to the word teleconnections itself. To my American ear the word sounds not scientific but flaky, new-agey, or the name of a dating service.
re 9… it’s likely due to the closeness to telekenesis
RE: #3 – personally, I think that “fault scarp” type features in reputed reconstructions and reputed instrumental records are a red flag, saying “audit me to the nth degree!” But what do I know, I’m just a dumb old high tech manager with a geophysics degree, MSEE and more than a couple decades of serious hard knocks thrown into the mix.
#10
Certainly could be for some. Or telepathy. It’s also the overused word connections which is used in all sorts of strained arguments or loose associations.
Could also be an intuitive resistance to a contrived word that mixes Greek and Latin roots.
Opening post:
By “Crowley style” do you mean:
(1) it’s “bonehead simple”, or
(2) you’re just “wingin’ it”?
“Just asking, coach.”
On “teleconnection”: #8 Tom C wins a prize. (Someone actually reads this stuff?)
A distinction needs to made between teleconnection and teleconnection patterns.
If it is hot in Perth, then chances are it’s cold in Melbourne, because of our predominant ‘high in the Bight’ weather pattern. That is teleconnection.
If Perth has a warming trend and Melbourne has a cooling trend then that’s a teleconnection pattern. However, if Perth has a warming trend and Melbourne also has a warming trend or drying trend, then that’s also a teleconnection pattern at least according to Mann.
And that is my problem with teleconnection (patterns). It’s a concept for all seasons (pun intended), capable of explaining everything in the climate and therefore explaining nothing.
Evolving teleconnections 🙂 i.e. time-dependent correlation approach..
What if Steve’s curve actually represents the temperature?
May be there is no real warming in 20th century, mostly urban warming. Then MWP would be quite reasonable when looking at historical records.
Larry Huldén
Finnish Museum of Natural History
I don’t have any problem with teleconnection, either. I have a problem with using a tree in Melbourne that’s measuring the temperature in Perth, however, since I want an average between Melbourne and Perth and all I’m getting is two Perths. What good is the Melbourne tree at that point?
Mark
The basic problem that I (and I suspect many others with the slightest knowledge of plant physiology) have with the concept of “teleconnections”, Mannian style, is that plants (and trees) respond to local microclimate conditions. Invoking a connection with temperatures halfway across the World, or indeed, with mean Global temperature (the “sweet spot” in Mann-speak) is, quite frankly ludicrous. Anyone, even me, could eventually find some spurious correlation with a tree ring chonology if I did enough post hoc comparisons.
The chief research forester in Newfoundland mentioned some years before on a CBC radio broadcast that dendroclimatology was useless. There is no way to tell from tree ring growth what an average temperature would be for that year. Too many variables and using normal ecological principles (eg that which is in the least supply will limit growth)mean annual temperature is never one of them. Just looking at the ecosystem where bristlecone pine live, the last thing one would expect to limit tree growth would be temperature. Tree growth is not a thermometer.
If nothing else, this shows how desparately thin the statistical evidence is for placing current climatic conditions into any proper historical context.
The simple process of updating a proxy for one decade or data completely changes the millenial picture.
Tree ring data will give one answer and one answer only and that is the age of the tree at the height where the sample was taken (although if you hit rot then all bets are off) It will not give any other information with any degree of accuracy. It won’t indicate mean annual temperature since trees (especially where these samples were taken)only put on growth for 6 months or so. Doesn’t tell you a thing about the other six months. It won’t tell reliably about rainfall, disease and insect infestations, damage from abiotic means, available nutients, etc. etc. etc. It most certainly won’t tell anything about anywhere else in time or space. To use tree ring growth as a proxy for climate is delusional.
My grand uncle is a professional forester in the SO of France of pines forests planted under Napoleon to drain the Landes swamps. As a boy, I remember being told on the field that tree rings reflect precipitation, sometimes pest infestation, sometimes forest fires, somestimes big storms which clears weakest or oldest pines and let the remaining ones thrive. NOT temperatures.
But hey, my grand uncle is just a tree murderer, he doesn’t have the talent of Mann to extract confession from trees.
Sorry folks, but *treeline* trees do tell you something about local temperature. Just not very much. It’s beyond all statistical probability that treeline trees of many species around the world correlate (weakly, mind you) with temperature, without there being a causal relationship. All things being equal, longer growing seasons do lead to wider rings.
Granted, not all things are equal. Drought, pests, fire, these things badly muddy up the temperature signal. But the outright dismissal of treeline trees as a potential source of information on past climate borders on denial. If you want to go there, you had better be ready to defend yourself. My advice is don’t go there.
Paul (#20):
I agree. It seems that proxy records — and particularly the discredited ones used to construct the HS — are unlikely to provide much usable information about millennial global temperatures.
The instrumental record, though plausibly corresponding more closely on average than the proxy data to global temperatures, is also plagued by problems [see Anthony Watts here]. These problems are particularly worrisome because they coincided with GHG increases in a way that complicates sorting out the causes of observed trends in the temperature time series.
Perhaps the satellite record can provide reliable temperature data. However, satellite data are recent, which limits the precision of inferences that might be drawn.
These data issues are of specific concern in light of Kiehl’s recent paper (discussed here), which makes clear that the GCMs are in fact tuned to historical temperature data.
Is there another, more rigorous, line of attack? Maybe. A while back, SteveM (here) asked CA readers — including some prominent climate scientists — for a physically-based approach for quantifying climate sensitivity. At the time, that sounded promising. However, AFAIK, no one has yet put forward a model based on first principles.
It would be ironic if the reason no has done so is because such models cannot be reconciled with the historical data. 😉
In addition to trees seemingly being lousy thermometers, I am suspect of the differences between the Crowley series and the updated one. Crowley shows a cooler LIA and a warmer modern period. This gives a very noticable increase over the last 500 years. The updated one is warmer in the LIA and cooler in the modern period, with neglibible “warming” over the last 500ish years.
They cannot both be right. Who’s wrong?
#20 Paul & #24 TAC, where he agrees with #20:
I disagree. Look carefully at what #20 said.
But Steve M has shown in the other Almagre thread that Almagre bcp is *not* a temperature proxy. i.e. It was not a “proxy” that was updated; it was a mere bcp chronology. If OTOH you were to update a *valid* temperature proxy, then the millienial picture would be robust and would not overturn so easily.
One should be careful not to be too dismissive. [I am detecting a bias in these threads toward being overly dismissive. Thus far, only Steve M has been careful enough not to cross that line.]
MattN & any others who think they have the skill required to interpret these graphs: read the blog! This is not a plot of temperature, dude. It’s a plot of ring width. (It would help fs the y-axis label wasn’t cut off; but of course Steve M is busy preparing for AGU.) In the last sentence Steve M is speaking tongue-in-cheek about what it would imply if this WERE a valid temperature proxy. But it’s a moot point because it ain’t; it’s an Almagre bcp chronology, nothing more. [Irony and tongue-in-cheek are part of Steve M’s writing style. You have to know when he’s using it, or you risk taking him literally at the wrong moment.]
Sorry to be so harsh, but the misunderstanding is spreading like wild fire, and only YOU can stop it. THINK! Read before you write!
bender (#26): Note that I agreed only with the first sentence of #20 — the part about the thinness of the statistical evidence — which I quoted. I agree with you on the point you’re making about the second sentence of #20, and I also share your view that “Almagre bcp is *not* a temperature proxy.”
Moving on, I am curious as to whether you disagree with my comment that:
#28 TAC: Understood.
Re: blockquote. “Disagree” and “agree” are, unfortunately too categorical for what is really a quantitative question. My reasoned opinion is that the true error bars on these reconstructions will turn out to be massive (UC-like), meaning that the recons will not be useful – not in the short term – as an aid to sound policy making. In the long term, maybe. As a scientific tool, maybe. But not now, and not for policy-making purposes.
But what, really, do the proxies give you, anyways, policy-wise? The ability to write a single sentence as to whether CWP is “unprecedented” compared to MWP? That’s a steep price tag for such a vanilla conclusion. Is it worth it?
Bottom line: I don’t think the tree rings will tell us any time soon if CWP is “unprecedented”. Treelines, maybe. Glacier lines, maybe. But so what. It’s the future that matters, not the past. And that is all about the GCMs. (And there better not be any tuning of the GCMs to the paleo record!)
#29 bender: Thanks for the quick response. I agree with most of what you wrote. However, I think an accurate record of millennial global temperatures would serve purposes beyond the “ability to write a single sentence as to whether CWP is ‘unprecedented’ compared to MWP.” Specifically, it would give us a record against which we could test the models. For example, do GCMs reproduce the statistical structure, in particular the spectrum, of the natural system? We need a long record in order to investigate that question.
Conversely, inaccurate climate reconstructions that are perceived to be accurate may cause harm by falsely discrediting uncalibrated models and discouraging their creation. That was my point in #24:
bender,
Teleconnected or not, North-America, Greenland and Europe are all by the North-Atlantic and climatically much under its influence. What I would like to know is whether the climate of Colorado is more ruled by the Pacific or by the Atlantic. I know they can have some awful cold winds from the north during the winter, but I assume thats from NW and not NE.
If the main driver is the Pacific then the MWP should at least be seen over most of high lat NH. Is there a good high-resolution proxy from “the other side of” NH, northern pacific? I would love to check that for temperature peaks after high volcanic activity, it’s great fun:
http://virakkraft.com/almagre.ppt
http://virakkraft.com/volcanotemp.xls
#26 Bender.
I am sorry. My understanding was that Crowley et al maintain that it is, hence the original paper. This though clearly shows that it is not. I am no sure what your issue is here.
But you do point to the elephant in the room, namely, how on earth can we bestow the title “millenial temperature proxy” on any time series given the severe data limitations that we have. In my field (economics/econometrics/finance) there is an old joke about 3 sigma events (e.g one in a thousand years). Nobody actually knows what a one in a hthousand year event looks like in that context. Same with temperature.
Worse, we have to deal with the “identification problem” inherint in the form of partial least squares approach that is meployed here. Inverting a multivariate function is a tricky thing to do, especially when one cannot control for the full range of variables that matter. Any tree ring is a proxy for all natural conditions that combine to produce growth and affect the rate of growth. Not only do we have to control for those variables when testing our ptemperature proxy hypothesis, but we then need to incorporate historical data alongside that of the proxy into any hindcasting of the target variable.
Then we add the issue of statistical validation of any hypothesis claiming the efficacy of any particular proxy. With maybe a few decades of reliable target variable data (temp.) that leaves us ille equipped to make any statistically robust statement about the relationship between temperature and the proxy over millenial timescales. This specific exmaple is just an extremely good example, notwithstanding that there are fundamental priors that raise questions about its use as a proxy.
#32 Paul
AFAICT paleoclimatologists have been pursuing the statistical robustness question with some (minimal, IMO) energy – the amount of energy required to keep the scientific boat afloat, but insufficient to withstand, say, a court challenge. As Steve M has indicated, though, it’s not the guys who do the canonical science that are at fault; they are just doing their jobs. The problem is the promoters who don’t understand what they’re promoting. They don’t need no stinkin’ confidence intervals.
Trees can be used a thermometers, provide you know with a fair bit of accuracy, all of the other factors that affect growth rates.
Unfortunately, this requirement is nearly impossible to meet.
For example, scientists have only recently discovered several decades long climate cycles. There are hints of longer cycles in the records.
How do these cycles affect things like clouds and rainfall? We don’t know. We’ve only just started studying these cycles.
If solar cycles do indeed affect cosmic rays, and cosmic rays do indeed affect clouds, then the cycles of the sun will impact plant growth. There is growing evidence that the sun has cycles that measure in the hundreds of years, maybe even thousands.
How do these cycles affect climate, and hence growth rates. That’s another unknown. How do these cycles interact with the earthly climate cycles.
Unknown squared.
According to physicists, worm holes and time travel are theoretically possible.
If somebody claims that he has created a worm hole, I’m going to want extraordinary proof before I invest.
The same goes with using trees as thermometers. It’s theoretically possible, but given the huge practical problems, I’m going to demand a high level of proof before I put any confidence in any such reconstruction.
The fundamental problem with the entire subject of using trees as temperature proxies, however, is summed up in this one thought.
A scientific way of finding a tree-ring/temperature proxy would involve deciding criteria and picking trees prior to coring them.
The current method is – pick a tree, core it, test it for matching local temperature.
IOW: Why is this tree not a temperature proxy? Answer: because it doesn’t provide a sufficient correlation to temperature. That’s completely nuts.
Al, I’m not going to spend all day fighting tree-ring denialism. You have your concept of the way the science works. Great. At some point Rob Wilson is going to step in here and clean your clocks. Until then … lurk/on
RE: #23 – Treeline trees in humid climates. That I’ll buy. Not buying if we are talking about BCPs.
My limited understanding was, that the paleo record would be important to help closing the many gaps in the knowledge of how the natural variability and external, non-anthropogenic forcings interact. If in deed there was a MWP, if it was global and if it did reach or even exceeded today’s temperatures, then there would be some sort of not yet understood mechanism/feedback/forcing or whatever that caused these past changes in climate and would have to be included into the GCMs accordingly, wouldn’t it?
#34 Henning. In principle, yes, you’ve nailed it. In practice, however, your ability to make robust inferences is limited by the amount of true uncertainty in the data. Steve M is actively exploring precisely that question, and right now it does not look encouraging. But we need to be patient yet. It is too soon to judge. Unfortunately, there are many people who don’t even want the question to be asked. At this point, Steve M is one of the few saying it ought to be asked, if it is going to be used as an entire chapter (#6) in IPCC AR4.
IMO, teleconnection derived from proxy data can only be justified if the proxy is representative of its local environment. To say that the proxy doesn’t correlate with local temps but is valid because of teleconnection is preposterous. The tree didn’t grow in Greenland so any lack of correlation with the local environment just shows that the particular proxy is not a good indicator for the chosen quantity (temperature, soil moisture, etc.)
In the absence of a lecture from JEG …
If the high-frequency local response/correlation (at ‘A’) is attenuated but the long-range low-frequency response/correlation at ‘B’ is amplified (because the high frequency component in the climate input is absent from the dominant forcing node at ‘B’, or because trees integrate the climatic input signal over many years), then the (local) proxy relationship (at ‘A’) may be even STRONGER than what the long-range teleconnection correlation (‘A’ vs ‘B’) suggests. (It is just that the response is hidden due to poor data or a mis-specified model.)
This is a caveat I did not include in my simple, later explanation of teleconnection, but did include in my earlier explanation.
That is my proof that the idea in not “preposterous”.
Ask Steve M when he gets back from AGU if this idea makes sense to him.
My (layman’s) take on the teleconnections issue:
It is at least conceivable that there could be a valid use of teleconnections in some cases. For example, I wondered for a while if there might be a legitimate teleconnection to Mann’s use of the California bristlecone pines in the Hockey Stick. These series had zero (actually -0.02) correlation to local temperatures in the instrumental period, but some (0.35 IIRC) correlation to global mean temperature. Could these then be a valid proxy for global temperature?
One possible mechanism to explain this would be if El Nino events are the dominant factor in global temperature, as in 1998. Strong El Ninos mean a lot of rain in California, which could increase bristlecone growth. So maybe it shouldn’t be dismissed out of hand. However, Willis mapped the correlations of precipitation to global temperature here some time back (can’t find the link now) and showed that the area of the Bristlecones was drier in higher global temperatures. Hard to get a legitimate teleconnection out of that.
I agree with above posters that one should be very, very wary of any assertion like this, and that a strong burden should be on the asserter to demonstrate the physical and statistical validity of such a connection. The chance for spurious correlation is just too strong. Demonstrating statistical significance would have to include an analysis of the number of possibble teleconnections (determined a priori), which I’m almost certain is not presently being done.
I don’t believe it. Teleconnection makes sense to me but only to the extent that the local conditions are modified by some phenomena originating in another locale. So for instance, if we have a volcano eruption that eventually sends particulates to the atmosphere in the vicinity of the tree in question, then that will potentially have some effect on the tree’s growth patterns. But at any moment in time, the area around the tree will have some state at time t, soil moisture, soil nutrients, solar radiation, land use, air temps etc. and this state is the only influence on the tree at that time. How can it be otherwise? Now, I can believe that the “system” can change a non temperature related parameter so that the tree responds to the change in this non temperature related parameter in a way that de-correlates its growth patterns from local temps but the tree is still responding to its local environment.
Bender #42
I’m doing my level best to be open minded regarding this teleconnection business, but but please indulge me for a moment as I put on my skepticlal philosopher’s hat:
Do you view the teleconnection theory as an extraordinary claim (not quite preposterous) requiring extraordinary proof (virtually indisputable data indicating the existence of teleconnections)?
Or do you view it as an ordinary claim that merely requires a plausible logico-scientific explanation (i.e. something just North of handwaving)?
Or do you see it somewhere inbetween, or something different entirely?
Thanks.
#44 Jaye, I understand perfectly your argument. You do not understand mine. But I do not ask you to believe my argument. I will agree that my scenario is abstract enough that it requires a demonstration to be convincing. Even then, I will concede that it might be far-fetched. But I do not think it is preposterous.
Last post on teleconnection.
Ok. last.
Think about the “teleconnection” correlation as an index – a possible indicator of the correlation you *might* potentially be able to detect at a local scale *if*, at that location, you had good data and a good model to interpret it. But because you have *bad* local data and a *bad* model, the *local* connection appears weaker than the *long-range* connection.
i.e. It is all about *appearances* (and what you choose to infer based on their interpretation).
I will let JEG answer #45 as squarely as he is willing.
If its not the above then it can’t be correct. We are not talking about some weird quantum phenom. These are trees that get their water, nutrients, solar, atmospheric conditions, etc from the immediate vicinity. If some far away event changes those local things, then so be it…teleconnection. If it doesn’t change the environment in some volume around the tree (don’t know how big that is but it is probably finite) then it can’t have an effect on the growth of the tree.
This I believe. This more a function of the limitations of the available measurements/models etc. than about what is really going on in the physical system…which is likely not something one can describe perfectly.
Not only that, but once it is responding to some other non-temperature related parameter, it is no longer a temperature proxy. If the goal is to find some global average, then that tree no longer contributes spatially to the data.
Mark
Bender,
I’m feeling lucky.
I’ll grant that treeline trees are temerature limited and may well record local temperatures in their rings. But if they are treeline now – how can one be certain they were treeline in the past? That is a problem – they can only be a proxy for locally small changes in temperature. If there are any significant shifts in temperature they won’t be treeline anymore and they won’t be temperature limited. An important question might be how large a variation in temperature they can record before some other factor in the past becomes relevant. Given the information on the extent of the treeline in mediaeval times that Steve has occassionally mentioned I would be inclined to think that this limits their utility for millenial reconstructions.
But feel free to set me straight on this.
(Economists have an assumption, ‘ceteris paribus’ that is almost never true. Dendroclimatologists appear to be making similar assumptions when they assume that current conditions prevailed in the past – to justify their ability to make inferences about temperature – as the basis for reaching a conclusions that current conditions did not prevail in the past.)
#49 Jaye
Good. This is precisely what I was saying earlier, by the way. There are multiple layers to the teleconnection proposition: you have input, output, an integrator, a model to interpret the integrator’s behavior, and data streams at two locations. You must contemplate them all jointly. There’s a lot of slop in that one little idea! [Sorry for bandwidth wastage, folks. Sometimes you have to say things two different ways to get the message through.]
#51 Mark T.
You have it right. There is no reason to believe the sensitivity parameters are stationary in time. You are trying to estimate a global (in the temporal/statistical sense) response, but all you have are local (in the temporal/statistical sense) data. I agree completely and have said so on previous occasions. The net result is that your proxy model is an approximation, possibly quite a poor one. The more numerous the potential suite of limiting factors, the more problematic the non-stationarity becomes. No question.
52
Seems like something that one would have to proceed with very carefully before declaring teleconnection as the driving phenomena in a given analysis. A solid understanding of the physical processes going on (like particulates due to a volcano eruption) would seem to be required before teleconnection could be derived via manipulation of the time series.
#47
Fair enough. And thanks for going one post more.
If JEG ever returns, whatever he says is bound to be entertaining. The hard part will be determining if it is “sophisticated” or “sophistry”.
#54 Jaye. Agreed. Glad to see you’ve back-pedalled to a far more defensible position. Hopefully that closes the discussion. It’s archived. People can be referred here whenever the teleconnection disbelief emerges.
Mike B,
A fun game would be to hide this discussion from JEG, see what he says independent of this, and then compare notes. I think that’s what Kenneth Fritsch wants: to make the guys with the bold claims stand up and fend for themselves.
Now I can’t stop with a comment like that…therefore
Last word begins now. 😉
Maybe if we didn’t make fun of JEG on every possible occasion he might just show up here again to actually comment on some of the posts.
Me as a layman, to the best of my knowledge, weren’t impressed by JEGs comments but, nevertheless, enjoyed the debate (on Craig Loehle’s paper) which explained some interesting stuff and really brought out the best from some of the skilled regulars here. (For me to learn from…)
It would seem that to find a “Teleconnection” you would first subtract the local conditions then any difference would be defined as a “Teleconnection”. Or asume that the local condition is the result of all necessary variables + “Teleconnection”
Re #42
Yeah, Bender, something like that. What is needed, in my mind, however, is an application and explanation for a teleconnection of this type for a specific instance where it was used in a temperature reconstruction. Without being tied directly to a real example, I am afraid that one opens the flood gates for a lot of high powered and wordy arm waving about theories (conjectures).
Re: #59
Rob, I think your view is a common one that some people make about academic types who are some times stereotyped as being innocent and childlike and easily have there feelings hurt — delicate in nature, if you will. That may be the case for a very few who have been able to remain sheltered from the real world but I strongly doubt that that includes but a tiny fraction in today’s world. In JEG’s case, it is quite obvious that he is a man of the world (go to his web site) and has been thoroughly exposed to the real world. When someone parades onto this blog with a JEG attitude it makes an astute observer question how much he is “making a caricature of himself” just to have some fun with the rest of us. When Lucia made the Hercule Poirot connection it was like an epiphany for me and made me appreciate the extra curriculum part of JEG even more. I thoroughly enjoy the Poirot character for his personality and deductive abilities. If JEG has anything to offer this forum in the way of climate science he will return.
#61. JEG has sent me a cordial invitation to make a presentation at Ga Tech in Feb – which I’ve accepted.
I always liked JEG.
Thanks for your reply, Bender. I’m beginning to understand the importance of CA when it comes to the scientific (rather than merely political) consequences of things like MBH. What I still don’t get about the entire treering approach is this: if I understand correctly (being just a numbercruncher working in the industry), our current warming is mainly northern hemisphere mild winters. However the treering approach focuses on the growth aspect which is mainly affected by the conditions in early summer. Assuming that previous warm periods were also determined by mild, short winters rather than hot summers (which would fit the MWP descriptions in the literature) – wouldn’t a proxy have to focus on winter conditions? I browsed around somewhat and found a number of possible effects, mainly around the tree’s ability to chemically repair cell damage caused by freezing. I apologize if this has been discussed before – I couldn’t find it.
Rainfall in a particular area might be dependant on the temperature of a nearby ocean. Higher temperatures mean more evaporation, and hence more rainfall. Or changes in the water temperature could change the positon of the jet stream.
The problem is, how confident that the relationship between water temperature and rainfall in a particular area is constant over time.
We have recently discovered that things like El Nino affect hurrican formation.
Now it appears that the phase of the PDO (which we didn’t even know existed 10 years ago) affects the El Nino/hurricane relationship.
How much confidence do we have that there aren’t other cycles that also play a roll, cycles that haven’t been discovered yet because their frequency is longer than the data record?
Proving that a particular tree, or a particular stand of trees is temperature limited in recent times, only proves that in a climate situation like the present these trees are temperature limited. We are checking these trees to tell us how different the climate was in the past. The problem is, if the climate is different in the past, then by definition, it is different from the present. If it is different from the present, then how can we be certain that the trees are still temperature limited in the past.
The very changes that you are looking for, invalidates the data that you are trying to collect.
#64 You are correct, henning. I’ve made this point before. The dendros will readily concede that their tree ring models and reconstructions are approximations. The problem is that if the tree is sampling only one-half the year, then the reconstructed signal will correlate with the actual annual signal by ~0.5 – quite a poor approximation. Multiply that by the proxy’s reconstruction skill for the months in which it functions, say 0.3, and you’re starting to get some small numbers. All in all, that leaves a fair amount of room for error (cool bias) during a period such as MWP.
JEG is a good egg. Glad he stopped by.
Re: #66
Bender, as I recall in reading some of Rob Wilson’s papers, the published correlations of TRW/MXD could be for time spans as short as 2 or 3 months and sometimes using maximum (or minimum?) temperatures in place of a mean temperature. Claims made for historical climates/temperatures usually are commonly, however, limited to a statement like “comparing summer months maximum temperatures we see..” The selection of months and min/max temps appears to me to be based as much on obtaining a better regression fit than any good a prior reasons.
In more recent papers, I see more use made of MDX in temperature reconstructions which with the weighting of the TRW and MXD parts of the regression without good a prior reasons appears to me to provide more opportunity to over fit the models.
I also recall reading an explanation of a poor fit of a TRW/MXD reconstruction with either another reconstruction of a different kind or the instrumental record as being due to a change in the diurnal temperature cycle – and again I think it was from a Rob Wilson paper.
I would be curious to see what would be derived if one looked at the correlations of temperature anomalies for summer months (maximum, minimum and mean) temperature to the mean over the instrumental record for the US lower 48 and local regions of that area. I have these data in form to do this analysis and should do some of these calculations if for no other reason than to satisfy my own curiosity.
Rob Wilson does make the effort to show the TRW/MXD reactions in his temperature reconstructions to major volcanic events of the past. In my mind that simply demonstrates that TRW/MXD react to a majority of these events, but the question remains as to how well that reaction is calibrated to temperature and temperature changes.
#68 Ken, yes, I was being generous. I almost wrote “3 months”, but figured someone would complain about the potential for warming in the shoulders of the growing season, April, September in the odd year. (Would rather be accused of warm bias here than the opposite at other sites.)
Re: bender’s post on A and B, attenuation and amplitude
Do you know of any numerical simulations that show this? I’m having a hard time wrapping my head around what a teleconnection is. In my cynical moments it seems like sifting thru climate-linked metrics until you find a hit, i.e., a “good” correlation, and then proceeding to explain why this is so. In the end, the existence of a teleconnection seems plausible but I’ve never come across any study where the authors had a hypothesis about what should teleconnect to what and then did the study afterwards sans snooping. Again, I’d love to be corrected and I’d love to see some numerical experiments.
Re #62 – Now we are going places!!
Re teleconnections
The real problem is that the teleconnections would be manifest in at least two phenomena, and the probability that they would both be linear is nil. That is why physical understanding of the effects would be necessary – in order to develop what the non-linear responses are. No one has come close to doing any such thing, all hand waving so far.
Bender #66. An undeniable, physical attribute of choosing high altitude trees. I do not remember reading of how this was explained (unless it is “teleconnection”). To me not only should using 3 months for a yearly phenomena be avoided, the point, if I remember correctly, that it is the increase in minimums rather than maximums that RC and the Team claim (Is my memory correct?) that determines the global warming increase. If so, it is indeed strange that for most of the minimum, their proxies have little to no response.
That is great. Is there any possibility (with JEGs permission, of course) of setting up a video camera at the back of the room, and have you wired with a microphone?
What I would really look forward to is listening to the Q&A afterwards with you and his students. Just like what occurs here on CA, the back and forth interchange is always the most illuminating and informational.
I would like to thank JEG for his contributions here. His articulate and passionate arguments always make for thought-provoking reading.
#72 TomC. So if temperature in Timbuktu were teleconnected to tree ring growth in Toledo, there would need to be a physical (climatic, or other)connection between the two locations, but over time, the temperature in Timbuktu would be variably connected to the physical connection, which would in turn be variably connected to the tree ring growth in Toledo? That’s a lot of variables. I don’t see how those kind of Proxies could be useful.
Any Georgia Tech readers, could we have a top 5 arguments for AGW theory, and top 5 against, before and after? Would be good. Post anonymously. CA readers, please don’t comment on Georgia Techies till after (and even then be nice).
Re #75
MarkR,
I think it is deemed sufficient if those kinds of proxies are novel and sophisticated.
.
.
.
OK, what is a sarcasm tag exactly? 😉
If you can, stop by JEGs page and thank him.
MarkR, moshpit is caged. Dont worry about him.
I was at JEG’s blogspot to see if he had posted anything new on teleconnections (he hasn’t) and ran across this comment from JEG. I will not include the TCO comments on Steve M because most of us here have had to endure them being repeated ad infinitum, but here is JEGs reply and in it his thoughts on inviting Steve M to GT.
MarkR, I think just Steve showing up, talking to the students, and then answering any questions (with any followups at a later date to those questions that can’t be answered on the spot) would prove that Steve is not some demented anti-AGW lunatic that eats babies and drowns kittens.
What more could you ask for?
🙂
Seriously, I think just talking with them would be illuminating and helpful in this.
Wish I was a student there, to watch it all.
While I find JEG entertaining, he is a might too dismissive in tone for my tastes in a teacher — and a young one at that. Let us see how Steve M is treated at GT and how well the engagement stays on topic.
#75 MarkR
You are right. While it might be theoretically possible to use a teleconnected proxy the practical difficulties in establishing a well understood correlation are overwhelming, to the point of making the whole thing impossible.
Many commenting on teleconnections have not read what I’ve written. Their purpose, IMO, is to prevent the investigator from incorrectly tossing out a potential proxy just because the data are noisy or the model ill-specified. There’s *nothing* wrong with the concept. The issue is how it’s used. For constructing working hypotheses, fine. For making robust inferences about past temperatures, not fine. Don’t confuse the tool with how it’s used (or abused). When I’m making cracks about California being teleconnected to Greenland, it’s not the teleconnection that’s the joke; it’s the investigator who doesn’t understand post hoc ergo propter hoc or doesn’t care about its consequences.
bender #84: Perhaps I have not read all you’ve written, but it seems what you’re saying is that the concept of teleconnections is not absolutely preposterous. OK. That’s hard to argue against; at our current level of understanding, I suppose it would be unwise to rule out anything.
But, seriously now, do you really believe this stuff? To be specific, do you find it plausible that teleconnections justify use of California (strip-bark) BCPs as valid global temperature proxies? Sure, it’s a clever hypothesis, and damnably hard to falsify given its vagueness, but Is there any statistically valid evidence to suggest it might actually be true?
A statement by JEG, quoted in #80:
So JEG still considers it to be an open question whether or not Steve has “anything of substance to say” — even after everything Steve has done to date?
My final point on teleconnections.
Bender is really bending backwards to be accommodative of the princple. I tend to reamin on the skeptical foot until I see at the very least some proper treatment of these hypotheses. First necessary condition is full specification of the system to identify the form of the connection. It has to be a physical process somewhere; this is supposed to be teleconnections, not telconesis. Then we have some priors with which to choose appropriate model structures, parameters to be estimated or tested.
Again, Econometricians like myself just stare at these claims with disbelief. I could attempt to build a macroeconometric model of the world. We know that all variables will be “teleconnected” to each other in some way. To simply pick these at random or try and find them through niave correlation analysis would find you a laughing stock.
No. (Remove “strip-bark”, and the answer is still “no”.) The principle is valid. However, in practice, the strength of apparent teleconnection correlation may not be indicative of the true local proxy response to temperature.
Tempted to reply: ‘that’s what benders do: bend things’. But I’m not bending backward to accommodate the principle of teleconnection. I am bending backward to accommodate the use of a teleconnective proxy response correlation in lieu of a local proxy response correlation. In the absence of a better explanation by a JEG or a Rob Wilson, I am trying to provide the best reasonable argument a dendro could give in defense of the practise. I clearly distinguish between the immature phase of hunting for correlative patterns, vs. the mature phases of proxy research where you know what the causal relationship is betwen proxy object and its environment.
One would hope that trillion dollar global policies would not be based on weak preliminary guesses. But that’s not always how the push-pull science-policy interface works.
I cannot imagine how a teleconnection signal can be obtained from ring widths alone. Widths, in conjunction with synoptic isotopic and other chemical analyses might be useful if there is a physiochemical model to support it. Even though I respect the utility of statistics (and only understand them at a High School level), multiple lines of evidence are clearly needed to avoid spurious correlations.
Yes, Howard, this is the problem. The signal may indeed be in the ring widths, but extracting it requires several other pieces of knowledge. First, the non-stationary behavior of the ring-width to it’s environment must be understood. What, if any, transfer function drives ring widths over time? This also presents as a non-linearity problem (i.e. there may be more than one combination of inputs to create a given output: the response may not be invertible). Next, you need to understand the correlations between inputs. This also affects linearity/stationarity. When one input gets to a certain point, it may limit the effect of one or more of the others. Without knowledge of these confounding factors, error bars quickly become “floor to ceiling” so to speak.
Mark
And you seemed like such a nice young boy. 😉
In my judgment Bender’s position on teleconnections avoids the lectures and wasted bandwidth such as we received from JEG revealing his misconception of the general comprehension of the term teleconnections expressed at CA. Teleconnections in climate exist for every- and anyone who agrees that the ENSO originating in the southern Pacific affects climate/weather in the US.
I do not want someone of JEGs climate science persuasion avoiding the pertinent question, which for our purposes is not explaining teleconnections in climates past and present, as he has done in peer reviewed papers, but rather how does one explain the use a prior of a teleconnected temperature that can be separated from the local temperature. And do it with an example or two by showing how this was validly applied to temperature reconstructions.
In JEGs case we have reason to believe from reading at his blog that he is a very politically sensitive climate scientist who has stated he wants to educate and change the world. Combine that with his personality and we have a climate scientist whose little grey cells can impart some insights on the science (or at least give us a glimpse into what they actually contain on this matter) or digress into quips about denialists behavior.
If JEG does return to discuss teleconnections in temperature proxies I would defer to an initial one on one conversation between Bender and JEG that could be opened to general discussion later. We might learn something from Steve Ms visit to GT and how they handle what could be a politically charged environment if their preconceptions of Steve M are as a political enemy and not the retired mathematician/statistician/puzzle solver with an in depth knowledge of a variety of areas of climate science that we are more familiar with.
Steve:
Would you clarify where data for your “update” prior to 1240’s comes from? I was not aware you had any cores going back to the 900’s.
Is it a “fresh” manipulation of Graybill or LaMarche?
Atlantic-Pacific climate teleconnections at millennial time scales
Click to access marchitto_2003.pdf
So that’s what it’s all about, climate events with global impact.
Like the Dansgaard-Oeschger events all over, also through the holocene, just on a smaller scale
http://virakkraft.com/greenland_curves.html
Actually, we are in one right now.
Out of curiosity I took my USCHN Urban data set and made two histograms to show the relationship between summer maximum and annual mean temperature for approximately 1/3 of the 1221 USCHN stations. I did these calculations with the discussion here in mind about correlating TRW growth that occurs mainly in the summer months and responds mainly to the daytime (high) temperatures to an annual mean temperature.
The first histogram below was constructed using the correlation of summer maximum to annual mean temperatures for the approximate period 1894-2006 for approximately 1/3 of the USCHN stations. There are missing data and particularly in the earlier part of period. The correlation function ignores data where one or both of the summer maximum and annual mean temperatures were missing. Note that the frequency of occurrence is given for r^2 (and not r). The station to station spread in correlations somewhat surprised me.
The second histogram compared the differences in station temperature trends for the summer maximum and annual mean and is reported in degrees C per century for the period 1920-2006. I used only stations with complete data for this period and as a consequence reduced the number of stations compared to the number used in the first histogram by 15%. The spread in the trend differences probably surprised me more than the R^2 spreads.
Sorry, but the first graph (r^2 regression of summer maximum vs annual mean temperatures) in my preceding post should have been the graph given below.
Satisfying my curiosity further concerning correlating TRW growth that occurs mainly in the summer months and responds mainly to the daytime (high) temperatures to an annual mean temperature, I compared the differences in all qualifying station temperature trends for the summer maximum and annual mean in degrees C per century for the period 1920-2005. The qualifier for using a station was that less than 5 data points were missing on an annual basis and resulted in using 1135 stations in the analysis. I again used the USCHN Urban data set and constructed the histogram presented below.
The histogram shows that the range and distribution was similar to that presented previously for approximately 1/3 of the stations. A chi square goodness of fit test indicated that the distribution would not fit a normal distribution, and therefore in the following analysis I used median, quartiles and range to specify some sub groups in the sample.
I deemed further analysis necessary since the TRW measurements come more frequently from higher elevations. To this end I divided the station sample of summer maximum trend annual mean trend into somewhat equal parts by elevation. Those results are presented in the table below and are given again in degrees centigrade per century. While at the extremes of elevations one might discern a difference, that difference is small.