the Air Vent

Because the world needs another opinion

Robust – Not so much

Posted by Jeff Id on May 9, 2010

A reader sent me a link to a new paper by Berger, analyzing the coherence (covariance) of various popular proxy reconstructions (hockeysticks).  The paper can be accessed for free here. Similarly to global temperature metrics from the same data, such reconstructions are often declared to support each other confirming the accuracy of the methods.  Of course, it’s my opinion that they are bad data crammed together with bad statistical methods that are ‘designed’ to force a balance of noise in the calibration period which creates a correlation to temperature (the blade) out of basically — nothing.  The handles are basically noise and deamplified signal, which averages toward zero.

Burger’s analysis looked at the coherence (of the handles) of various reconstructions with each other, finding clusters of the most like of 10 reconstructions and at which significance thresholds their likeness represents. Burger writes:

By avoiding the (calibrating) instrumental period, and by using a fairly robust spectral
measure for low-frequency performance, the above coherence analysis has uncovered
several inconsistencies among the group of millennial reconstructions that figured
prominently in the latest IPCC report and elsewhere. An immediate lesson from this
10 is that simple visual inspection of smoothed time series, grouped and overlaid into a
single graph, can be very misleading. For example, the two reconstructions Ma99 and
Ma08L, which have previously been described to be in “striking agreement” (cf. Mann
et al., 2008), turned out to be the most incoherent of all in our analysis.

It’s an interesting paper, in several aspects, including the fact that despite claims by RC scientists that the reconstructions support each other, mutual comparisons of the reconstructions have not been carried out.  Yes the reconstructions all look like hockey sticks, but if they are temperature, we should have coherence.   Still, five groups of reconstructions were created from the 10 which which had a highly significant coherence >99%.  However, it’s not until we drop below 90% that the five groups begin to match each other.  An interesting result, considering that the papers often use the exact same source data.

Similarly, the 95% significance level (d=0.61) yields identical clusters, while under the 90% level (d=0.65) clusters {Br00, Es02, dA06} and {MJ03, Mo05} merge.

The graph above can be interpreted as the quality of match on the vertical scale having different reconstructions on the horizontal scale.  Where individual reconstructions are joined with a horizontal line is the quality of match between them, as indicated by the value d.   From this plot, our favorite whipping boy, Mann has the least similar of all the reconstructions not joining to the rest until d=0.77ish.

Table 2 is another way to look at the comparison:

There are a few notable quotes in the papers discussion section.

Using inconsistent reconstructions to approximate the temperature curve has one par25
ticular visual consequence. Whether overlaying them in one figure or forming an average,
the result tends to be a cancellation of larger amplitudes as inconsistency means
here to be indistinguishable from random covariations. Together with the mentioned
synchronization through the instrumental calibration period, such “synthesis” figures
automatically resemble a hockey-stick.

This is the point made in the hockey stick posts in the header bar above.  I don’t know Burger’s other work, but basically Mann08 IS exactly what he recommends against.  The noise in the pre-calibration period is essentially random as he has shown and forced syncrhonization in the calibration period — automatically creates a hockey stick.  In this case he’s just discussing a merging of the finished reconstructions which already have the temperature variance well matched since 1850.

The paper does have one minor issue though in the following statement.

The analysis is constrained to reconstructed data prior to 1850, to ensure that the estimated coherence is not 5 inflated by calibrating effects from the instrumental period.

Popular calibration methods other than decentered PCA which include CPS and various regression analyses, affect the reconstructions well prior to the calibration timeframe.  I’ve described regression style hockey stick math here as shock and recovery.  The frequency of the noise in the proxy determines the recovery rate and from my messing with M08, I would have suggested doing the analysis on pre-1650 instead of pre-1850, data, not that it would make that much difference.

It seems that Burger is far less skeptical than myself of proxies to represent temperature.  In his concluding recommendations he makes this point:

It was shown that the target area plays only a minor role. Furthermore, if type and
processing of proxies are sufficiently even, coherent reconstructions are produced. If
5 that is true in general, the main source of reconstruction inconsistency is the use of
mixed types of proxies (“multi-proxies”), and their role for temperature reconstruction
should be revised. One should systematically check whether “uni”-proxy reconstructions
tend to be more coherent than multi-proxy reconstructions, and if so, which types
of proxies actually create the inconsistencies.

By uni-proxy reconstruction Burger is referring to the types of proxy used.  While this is an interesting point, I wonder why he didn’t take the comparison analyisis one step further and look at the proxies themselves.  My belief is that they are highly, highly (two times) inconsistent and if there is a temperature signal in them, the statistical methods of paleoclimatology have no chance of recovering it.

Anyway, this statistical analysis is now usable as a reply to climate science which often makes the claim that independent studies came to the same conclusion.  You can point them to Burger 2010 and say, check again.. I’m actually surprised that there wasn’t better coherence between the reconstructions, especially considering they use much of the same data.

The paper can also be found at this site, which allows for interactive discussion that is currently open, it will be interesting to see if anyone replies.

40 Responses to “Robust – Not so much”

  1. steven Mosher said

    Thanks for this. The mess of spagetti has always bothered me and I’d never read of anyone trying to do clustering analysis on it.

    Since it takes a value of 90% to get Mann into line with everybody else.. you can already hear the arguments forming.. besides if you average bad data with good data, it gets better. plus you have more N which is always a good thing.

  2. Tony Hansen said

    Mr. Mosher,
    Would that be a cluster analysis of a cluster ….?

  3. Kenneth Fritsch said

    I cannot site the exact references, but I am certain that Steve M, and perhaps Ross M, have noted these discrepancies in proxies in public criticism of them with regards to stating that they are same and reinforcing one another. The IPCC review comes to mind first. I think it is rather easy to see these discrepancies in proxy comparisons if one simply eyeballs the sometimes out of phase hills and valleys of two proxies.

    If you had proxies that were nothing but red and white noise with the up swinging instrumental data tacked onto the end, I think you would see what Gerd Burger (and Steve M) found. One or both of the matched proxies have to be wrong and that paraphrases what Steve M said in an IPCC review, as I recall.
    As an aside, my approach to analyzing the three main surface temperature data sets and the two satellite sets has been to look for statistically significant differences -differences which do exist by the way. I look forward to reading Burger’s paper.

    I think some climate scientists see what could be noise in the pre-instrumental era with instrumental tacked on, and since that agrees with the “settled” science of current unprecedented warming, go no further in their analyses- nor are they motivated to do so.

  4. Steve Fitzpatrick said

    The same author has been noting serious problems with the multi-proxy reconstructions for a while: http://www.agu.org/pubs/crossref/2005/2005GL024155.shtml

    I’m betting he is not real popular with the Team.

  5. steven Mosher said

    RE3

    kenneth, I seem to recall the same thing. My thought with proxies is that you wanted to test for some sort of coherence between them before plotting them all on the same graph and then just waving your arms about “agreement”
    I’ve never done any clustering analysis..but have skimmed the literature when I was faced with similar analysis problems. As this result fits our priors about Mann, it probably merits some further investigation.

    Jeff has an interesting note about using 1650 as a cut off.

    I’d like to see how the cluster analysis changes as a function of that date so.. same analysis running from 1650, 1750, 1850, 1900 ( yes into the calibration period) I dont know why guys dont explore the sensitivities of their analytical decisions. write an effing loop.

  6. j ferguson said

    Kenneth Fritsch said above:
    “I think some climate scientists see what could be noise in the pre-instrumental era with instrumental tacked on, and since that agrees with the “settled” science of current unprecedented warming, go no further in their analyses- nor are they motivated to do so.”

    Mr. Fritsch, are you saying they are seeing (imagining) signal in the noise and tacked on instrumental record?

  7. AMac said

    Meanwhile, there’s an active, ongoing project being funded by NOAA to tackle the very questions that Burger is addressing. It’s the “Paleoclimate Reconstruction Challenge.” Introduction here. From 2009, there’s a presentation that describes the project. 13 powerpoint slides.

    The vision:

    Many areas of uncertainty in current climate can be illuminated by studying the past climate record. Document highlights areas of interest and potential targets for data synthesis and modelling exercises.

    Forward modelling of proxy data whereby the proxy data is explicitly modelled directly by climate models, is of fundamental importance to further improving model-paleodata comparisons.

    Reducing uncertainties in proxy reconstructions for improving targets for climate modelling and in better understanding the intrinsic variability and forced response of the climate system.

    [snip]

    Paleo-data can potentially illuminate past behaviour of these
    systems and provide a test bed for model predictability.

    Membership:

    Valerie Masson-Delmotte
    Gavin Schmidt
    Caspar Ammann
    Juerg Beer
    Keith Briffa
    Kim Cobb
    Elsa Cortijo
    Eystein Jansen
    Michael E Mann
    Andreas Schmittner
    Paul Valdes
    Anna Pirani
    Thorsten Kiefer

    Four names I recognize as taking the AGW Consensus position, aggressively. Most aren’t familiar. Burger’s not among them. Prof Mann is certainly aware of Burger’s work in the area, having cited him in a 2007 review.

    The sense I get is that this consortium’s consensus view is that paleoclimate reconstructions are robust.

    If so, it would make sense to use them for “hindcasting” exercises with GCMs, as a way to improve those models.

    But if not: The project will be an exercise in “assuming the conclusion.” GCMs will be “improved” alright, but on a “garbage-in, garbage-out” basis.

    One might think that the directors of the “Paleoclimate Reconstruction Challenge” would really, really want Gerd Burger to join their effort, and provide insights and challenges from a critical point of view.

    In March, I wrote one of the coordinators, inquiring about the relevance of the use by Mann08 of the uncalibratable, upside-down Tiljander proxies: could that affect the validity of the resultant chart of paleotemperature anomalies? The response I received was polite but uninformative.

  8. AMac said

    Re: Steve Fitzpatrick #4 (May 9 13:37),

    The 2005 Burger & Cubasch paper you reference, Are multiproxy climate reconstructions robust?, was the one that was briefly discussed and cited by Prof. Mann in his 2007 review, “Climate Over the Past Two Millennia” (PDF).

  9. Steve Fitzpatrick said

    AMAc #8,

    Behind a paywall?

    Unless you are at a university and have free access, $30 – $60 a pop to see what someone, who you already think is an arrogant SOB, says is a bit too much. Sorry, not going to spend the money.

  10. Nick Stokes said

    Re: Steve Fitzpatrick (May 9 19:08),
    Steve, try here.

  11. Harry said

    But who is going to believe anything said by the “RC-scientists”?

    For me personally, “RC” is sufficient to open the bin and discard anything they are balking about. What a useless bunch. Yawn.

    They overstreched far beyound their power (which was minimal, to start with)
    Go Gavin, Go! I hope you know the direction.

  12. Steve Fitzpatrick said

    Nick Stokes #10,

    Thanks.

  13. Steve Fitzpatrick said

    AMac #8, Nick Stokes #10,

    Mann dismisses Burger & Cubasch (2005) and Burger et al (2006) with one sentence and a wave of the arms…. how surprising for Dr. Mann!

  14. Robert E. Phelan said

    7.AMac said May 9, 2010 at 2:19 pm

    A quick and dirty search suggests that ALL of them are warmists, some like, Kim Cobb, are active climategate deniers (you know, scientists under attack reacting badly but nothing to change the fundamental soundness of the scient) some, like Paul Valdes, have a vested financial interest in AGW:

    http://www.greenstonecarbon.com/board/35/prof-paul-valdes/

    like the IPCC, they are being tasked to critique their own work. Integrity you can believe in.

  15. AMac said

    Re: Nick Stokes (May 9 19:25),

    Thanks, Nick. Sorry SteveF, I didn’t realize that was behind a paywall.

  16. steven Mosher said

    Amac. that group and the challenge is discussed on CA.

    bringing the proxies up to date and actually archiving the data would be a good thing.

  17. Nick Stokes said

    Something odd in Burger’s paper. If you look at his Table 3, he has a mix of NH and global, and of summer and all season temps. So a lack of coherence is not surprising – they aren’t measuring the same thing.

    But if you look at the Fig 6.10 of the AR4 that Burger references, they have a different set of 12 papers, and these are NH only.

    I wouldn’t be surprised if they showed some lack of coherence too – there is a mix of methods. I think the main point of the AR4 diagram is to show that, however you do it, temperatures over the last millenium don’t show that much variation.

    But Burger seems to be just testing the wrong thing.

  18. steven Mosher said

    That’s a good point Nick. So really two questions here.

    1. is the methodology sound (clustering)
    2. is the selection of data sound.

    Instead of claiming that they dont show that much variation, it would be good to subject them to a clustering analysis?

    Maybe burger posted his code and people can just go test that?

  19. steven Mosher said

    BURGER

    Cluster 1 ALL NH& Extra tropics, all land.
    Br00 Briffa, 2000 Summer trees
    dA06 d’Arrigo et al., 2006 Annual trees
    Es02 Esper et al., 2002 Annual trees

    Cluster 2: NH Land and SEA
    MJ03 Mann and Jones, 2003 Annual multi-proxy
    Mo05 Moberg et al., 2005 Annual multi-proxy

    Cluster 3 NH&Extratropics land and sea
    CL00 Crowley and Lower y, 2000 Annual multi-proxy
    Jo98 Jones et al., 1998 Summer multi-proxy

    Cluster 4 NH Land and Sea
    Ma08L Mann et al., 2008 Annual multi-proxy
    Ma08 Mann et al., 2008 Annual multi-proxy

    Cluster 5 NH land and Sea

    Ma99 Mann et al., 1999 Annual multi-proxy

    Hmm.

    coherence:

    Tables Figures
    NH NH&Extratropics
    land + sea 0.93 0.89
    land 0.90 0.91

    Hmm.

    NONE are global. 2 are summer

    None are global nick.

    Now AR4; from table 6.1 the 12 different papers? I dunna, it seems to refernce table 6.1
    Cluster 1 matches
    ECS2002 831–1992 Annual Land, 20°N–90°N ◢ ◢ □ □ Esper et al., 2002; recalibrated by Cooket al., 2004a
    B2000 1–1993 Summer Land, 20°N–90°N ◢ □ □ □ Briffa, 2000; calibrated by Briffa et al., 2004
    DWJ2006 713–1995 Annual Land, 20°N–90°N ■ ◢ □ □ D’Arrigo et al., 2006

    CLUSTER 2 matches
    MSH..2005 1–1979 Annual Land + marine, 0–90°N ◢ ◢ ◢ ◢ Moberg et al., 2005

    MJ2003 200–1980 Annual Land + marine, 0–90°N ◢ ◢ □ □ Mann and Jones, 2003

    Cluster 3

    JBB..1998 1000–1991 Summer Land, 20°N–90°N ◢ ◢ □ □ Jones et al., 1998; calibrated by Joneset al., 2001
    CROWLEY MISSING
    CLuster 4
    MANN 08 missing from AR4

    Cluster 5
    MBH1999 1000–1980 Annual Land + marine, 0–90°N ■ ■ ◢ ◢ Mann et al., 1999

    Maybe I’m reading the tables wrong.. crap.

  20. steven Mosher said

    checking on nicks idea that the AR4 was NH only

    http://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch6s6-6.html#table-6-1

    and Burger used 2 M2008. kinda hard to have those in Ar4..

    here is Burgers list again and his rationale

    Such questions may arise when investigating the modeling – or reconstructing – of
    20past millennial Nor thern hemispheric (NH) temperature. They arose in me, at least,
    in an attempt to understand the reconstructions of the latest IPCC repor t (Jansen et
    al., 2007; Fig. 6.10), of which the following extend back to the year 1000: (Jones et
    al., 1998; Mann et al., 1999; Briffa, 2000; Esper et al., 2002; Mann and Jones, 2003;
    Moberg et al., 2005; d’Arrigo et al., 2006). The figure in that repor t displays an overlap
    25
    of the 1σ and 2σ uncer tainty bands of the reconstructions, weighted accordingly, that
    660
    CPD
    6, 659–679, 2010
    approximates the “most likely” temperature for any given year. In the present study
    I am going to define a notion of “consistency” that is suited to these reconstructions,
    based on time series coherence. I have for completeness also included the recon-
    structions (Crowley and Lower y, 2000) and (Mann et al., 2008), making a total of ten
    reconstructions listed in Table 1.5

    Mann2008 was published after AR4 Crowly is added for obvious reasons when you look at what he is trying to do with a
    collection of proxies that are combinations of Annual/Seasonal/NH and NH & extratropics.land.land&sea.
    Like AR4 regions from 0-90, annual and summer, land and sea.

    mann of course sticks out:

    In this display, Ma08L appears as the most “excentric” reconstruction, followed
    20by CL00, Ma99, and Mo05. Ma08L and Ma99 show the greatest distance, that is, of all
    pairs they are maximally inconsistent.

    But lets just recap what burger looked at relative to AR4

    BURGER
    Abstract Introduction
    Conclusions References
    Tables Figures
    􏰂 􏰁
    􏰂 􏰁
    Br00 Briffa, 2000 Summer trees : IN AR4
    dA06 d’Arrigo et al., 2006 Annual trees IN AR4
    Es02 Esper et al., 2002 Annual trees IN AR4
    MJ03 Mann and Jones, 2003 Annual multi-proxy IN AR4
    Mo05 Moberg et al., 2005 Annual multi-proxy: IN AR4
    CL00 Crowley and Lower y, 2000 Annual multi-proxy: ADDED TO GIVE NHExtratropics land and sea
    Jo98 Jones et al., 1998 Summer multi-proxy: IN AR4 ( NH EXtra tropics land and sea)
    Ma08L Mann et al., 2008 Annual multi-proxy PUBLISHED AFTER AR4? not in table 6.1
    Ma08 Mann et al., 2008 Annual multi-proxy PUBLISHED AFTER
    Ma99 Mann et al., 1999 Annual multi-proxy In AR4

    I dunna. Nick am I reading that table 6.1 right? I suppose this argues that M08 is incoherent with Ar4?

  21. Nick Stokes said

    Agreed, Steve, I misread Burger’s Table 3. He was distinguishing between NH and NH extratropics, not NH and global.

    But your breakdown is interesting. It does seem to show clustered results are those with similar properties from this still diverse range.

    Again I don’t think the results are all that surprising. The AR4 didn’t claim this diverse group of NH results would be coherent. They just give a similar broad picture of the temperature pattern.

  22. Jeff Id said

    “They just give a similar broad picture of the temperature pattern.”

    Ya, gotta be kidding. Please say you’re kidding. Are you telling me that Nick Stokes, doesn’t see the variance mashing in hockey sticks? Or that the raw data has no appreciable correlation to temp.

    These scribbles are as related to temperature as spaghetti noodles are.

    The reason for the clustering, can VERY likely be found in the raw data. — It’s probably the exact same data. — i.e. summer vs annual using the same data.

  23. mikep said

    Kenneth Fritsch is right that Steve McI looked at similar issues, though largely within the multi-proxy reconstructions rather than between them. Here is one reference

    http://climateaudit.org/2008/08/03/conflict-and-confidence-mbh99/

    Brown and Sundberg seem to be the experts. SteveM and Ross Mc also referred to this work inthier PNAS contribution, and Mann ignored it.

  24. Nick Stokes said

    Re: Jeff Id (May 10 04:26),
    Well, Jeff, I’ll settle for your summary that they “all look like hockey sticks”.
    But not the next bit “but if they are temperature, we should have coherence”
    Temperature, what of? That was my point above. OK, they are all NH, but some are just extratropical, some land-only etc.
    And if I understand Steve’s analysis rightly, the coherence is much better between datasets which are measuring more or less the same thing.

  25. timetochooseagain said

    24-What physical reason could there be for the hemisphere versus extra-tropics, summer versus annual, to be incoherent with one another? Should not their topologies be very similar given how closely related the are? The same forcings impact the summer as the winter, the hemisphere as the extra-tropics, more or less right? So why wouldn’t they have the same shapes?

  26. steven Mosher said

    As i understand it Burger has used a GCM to show the coherence between NH and NH& Extratropics. I suppose his argument being that the model shows coherence, therefore the proxies should. ( see his table on coherence)

    Further, in some of the dendro studies I have seen the dendros test for correlation where ever they can find it.
    Testing for a seasonal correlation or annual. Taking the best of course. WRT to reconstructing climate they also take
    the liberty of reconstructing the local grid or teleconnecting to some place else, or using a pick two strategy.

    I guess my point would be this. If we want to object ( I think we should) to an analysis that tests for coherence between
    say NH land only proxies and NHextratropics LandSEA, Then I think that

    A. we should be critical of studies that “mine” for correlation until they find one.
    B we should object to charts like that in AR4 which mix them together.

    I think the spagetti type charts are potentially misleading, economical but not what I would do.

  27. steven Mosher said

    RE25:

    Polar amplification would lead to an weakening of the coherence at certain time scales I would think. to the extent that you have a northern bias in proxies ( see arctic treelines) you might have an issue

    WRT season. Annual is ‘coherent with seasonal, but I suspect that if you try to recon the seasonal from the tree and then the annual from the season you’ll have mush by the end.

  28. timetochooseagain said

    27-“Polar amplification would lead to an weakening of the coherence at certain time scales I would think. to the extent that you have a northern bias in proxies ( see arctic treelines) you might have an issue”

    Polar amplification should just be a ratio. If we are comparing the shape of proxies, as opposed to their magnitudes, everything should be hunky dory IMAO.

  29. steven Mosher said

    Looking at burgers approach, You have

    NH, NH&extratrop,
    Land/landsea
    Seasonal/annual.

    With full factorial youve got what 8? NHLandAnnual, NHLandSeasonal,NHLandSeaAnnual,NHLandSeaSeasonal
    and 4 for the cases with the extratropics added.

    I think our expectation is that the proxies should be coherent within these combinations of factors. is that right?
    And if a proxy is incoherent with the other proxies that reconstruct that combination of factors, then what?
    who gets to stay on the island?

    Might be interesting to see how coherent observations are in the temp record..

  30. steven Mosher said

    RE 28:

    Thats why I added the “time scale” qualifier. The ratio changes over time. I have no clue of the magnitude or the time scale.. I’m just saying there are reasons to NOT expect the coherence to be constant over time.

  31. John F. Pittman said

    S Mosher, JeffID: Shouldn’t the real point be as Burger has indirectly stated it? The claim has been that the proxy reconstructions have been shown to be good regional predictors, “robust’, “teleconnected.” If nothing else doesn’t this call into question the use of proxies for anything other than local/similar. In which case, Jeff’s point that one ends up mashing the results towards the mean is shown to be true by the analysis. Do you agree?

  32. Kenneth Fritsch said

    They are not imagining, they see the HS and since it fits the consensus position, for science and advocacy, they do little analyses beyond that. There is also some advocacy leaking through where the paper results of the consensus tend always to agree with the advocacy position. I think Nick Stokes represents this position well in his comments in this thread. The expected reply from the consensus scientists will be something like, well of course they fail consistency testing, but since the proxies are not exactly a replication of one another we would expect that. Now that position to me shows a real lack of scientific curiosity and the effects of advocacy. Surely the inconsistency finding must lead to further digging and explanations.

    I think the link to the Steve M thread at CA given by Mikep @ Post #23 has some excellent points to discuss here. For example, the sharp difference between the instrumental and the pre-instrumental periods in consistency testing between proxies. Steve M asks what does that inconsistency mean for CIs pre-instrumental.
    By the way, the consistency tests have some bases in statistics while the spaghetti graphs and the instrumental/calibration/verification period forced HS is more a visual aid for marketing an advocacy position.

  33. Kenneth Fritsch said

    I left out the comment to which I was replying in my post above.

    Mr. Fritsch, are you saying they are seeing (imagining) signal in the noise and tacked on instrumental record?

    They are not imagining, they see the HS and since it fits the consensus position, for science and advocacy, they do little analyses beyond that. There is also some advocacy leaking through where the paper results of the consensus tend always to agree with the advocacy position. I think Nick Stokes represents this position well in his comments in this thread. The expected reply from the consensus scientists will be something like, well of course they fail consistency testing, but since the proxies are not exactly a replication of one another we would expect that. Now that position to me shows a real lack of scientific curiosity and the effects of advocacy. Surely the inconsistency finding must lead to further digging and explanations.
    I think the link to the Steve M thread at CA given by Mikep @ Post #23 has some excellent points to discuss here. For example, the sharp difference between the instrumental and the pre-instrumental periods in consistency testing between proxies. Steve M asks what does that inconsistency mean for CIs pre-instrumental.
    By the way, the consistency tests have some bases in statistics while the spaghetti graphs and the instrumental/calibration/verification period forced HS is more a visual aid for marketing an advocacy position.

  34. Kenneth Fritsch said

    Read RomanM’s comment in Post #1 in that linked CA thread in Post #23 – we assume the proxy follows temperature consistently throughout the period of interest and if it does not in a given sub period we have reason to doubt the proxy validity overall.

  35. Layman Lurker said

    I’m actually surprised that there wasn’t better coherence between the reconstructions, especially considering they use much of the same data.

    At the proxy level, it would be interesting to compare a sorting of “coherence” of the shaft vs “insturmental correlation” of the blade

  36. kdk33 said

    “Meanwhile, there’s an active, ongoing project being funded by NOAA to tackle the very questions that Burger is addressing. It’s the “Paleoclimate Reconstruction Challenge.”

    So, sloppy government funded science begets more government funded science.

    Nice. And mirrors the cause of all this nonesense: government funded scientists advocating for more government to fund more science.

  37. steven Mosher said

    re31.

    proxies are in someway untestable. Note I say in some way.

    Given all the assumptions in the model a proxy gives you nothing more than an estimate. That estimate ( say of temp in 1150) is untestable. First you select tree rings. If we allow you the luxury of selecting only those tree that correlate well with temp, then you can build a model. You cannot directly assess the error due to spurious correlation. Then you can screen the proxies through a verification period, this is getting close to testing the model you built in the calibration phase. Then you can model the temperature beyond the verification period… back to infinity and before ( hehe) that period is essentially untestable.

    That doesnt mean we know nothing. That means we have estimates with a whole host of assumptions, some of them quantitative. Using standard methods ( i think roman will chime in) the CIs on these prior periods are floor to ceiling.
    theoretically we should be able to improve on this.. perhaps using forward modelling of tree growth. dunno. But the methods used to date, dont inspire a bunch of confidence. I’ll put it bluntly. You tell me you can resolve the NH temp with
    3 three or 4 or 1 dozen tree ring series, and resolve it to with .5C, and I’m gunna wanna see all that math. And further, I’d argue that given what we know about sampling with thermometers that this claim is false on its face. I dont even have to check the math. its that insight that got Mcintyre interested.

    Tim Osborn I believe had this issue with Mann. Arguing that mann was claiming MORE accuracy with tree rings, than Jones was with thermometers in 1850.

  38. j ferguson said

    #32 Thanks for working with me on this Mr. Fritsch. Let me put it another way. Did what you wrote imply that they looked at noise and saw a(n) hs? I apologize for being so slow here – and maybe for writing with insufficient clarity.

    Again thanks.

  39. Mark T said

    34.Kenneth Fritsch said
    May 10, 2010 at 12:22 pm

    Read RomanM’s comment in Post #1 in that linked CA thread in Post #23 – we assume the proxy follows temperature consistently throughout the period of interest and if it does not in a given sub period we have reason to doubt the proxy validity overall.

    What RomanM is referring to, albeit indirectly, is two possible problems: stationarity and linearity. If the statistics of the “noise” are not stationary, and particularly if they differ from one another over time, then any type of variance based decomposition won’t work properly. If the “signals” contained within the proxies change over time, under time-varying conditions, then likely the true input sources are not linearly combined (or their otherwise linear combination is also varies with time) and linear extraction methods won’t work properly. A better word to use than properly is reliably, i.e., such methods won’t work reliably under either of these conditions.

    That is, of course, if there exists a clear definition of “signal” and “noise” in the first place, which, to my knowledge, has not been presented.

    There’s a reason Amman had to go out to 5 decimal places to get a non-zero digit in r2 100 years prior to the calibration period…

    Mark

  40. AMac said

    Re: steven Mosher #37 (May 12 15:32),

    > Tim Osborn I believe had this issue with Mann. Arguing that mann was claiming MORE accuracy with tree rings, than Jones was with thermometers in 1850.

    Let me show you the most remarkable year from the Tiljander varve series, 1326. It’s bracketed with the 5 years before and after. The preceding and following decades bounce around pretty much like 1321-1325 and 1327-1331. Tiljander-Korttajarvi-varves.xls

    MM – Mineral matter thickness, mm
    OM – Organic matter thickness, mm
    XRD – X-Ray Density, arbitrary grayscale

    Year MM OM XRD

    1321 0.21 0.53 72
    1322 0.39 0.53 108
    1323 0.22 0.37 95
    1324 0.57 0.96 94
    1325 0.56 1.14 84
    1326 6.14 12.86 173
    1327 0.52 0.58 120
    1328 0.47 0.72 101
    1329 0.36 0.53 104
    1330 0.45 0.59 111
    1331 0.61 0.74 115

    What happened in 1326? Was the temperature ~20 standard deviations less than normal (MM)? Or was it ~40 SDs greater than normal (OM)? Or was it about 3 SDs less than normal (XRD)?

    Mann08 solved one problem by inverting MM and XRD. So now the question would be, is temperature higher by about 20 SDs, or by ~40 SDs, or by only 3 or so SDs?

    You can reduce the glare of the issue by performing a rolling average–and Mann08 does.

    All this of course elides the question: What happened to the Lake Korttajarvi lakebed in 1326? And, whatever it was, how does that relate to the weather of that year, specifically temperature?

    OK, so the answer to this conundrum is intuitively obvious. These three measurements of the sediment deposited in 1326 aren’t about paleotemperature, they’re an echo of some other event. A forest fire, a mini-Tunguska meteor strike, Vikings rampaging through the forest, a mudslide rolling into the lake…

    My guess would be that Mann’s group has come to see these as number problems. The temperature correlation is in there. Solve the puzzle with the right computational and statistical tools, and the climate record of that year will be made visible.

    But they’re not just strings of numbers. They are representations of physical properties of actual things — tree rings, or, in this case, lakebed mud.

    The climate information in some of these records may not be extractable with today’s tools. Or, perhaps ever. That’s too bitter of a pill for the AGW Consensus to swallow. There must be a way!

    So one is devised.

    Common sense tells me to be very careful about trusting the interpretations of people who have confidence in extreme makeovers of data like these.

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: