Briffa and Sodankyla Church

The old Sodankyla church was built in 1689. Nine cores from beams at this church have been measured and archived. Briffa 2008 used one of 9 cores. Why only one of 9? Your guess is as good as mine.

Proving that Briffa only used one of 10 cores isn’t entirely trivial, though I think that it can be proved quite convincingly. At the request of Phil Trans B, Briffa archived his Finnish and Swedish measurement data, but, just to make things interesting, he did so without any metadata saying where the data is from. I sort of like decoding these little puzzles set by the team and I’ve been working at this one and have posted up some reference materials for interested parties.

First, I collated the measurement data archived by Briffa in “Tucson format” (an obsolete 80 column punch-card format) into an R-table. The original data is here. You can download my collation (as an R data frame) as follows:

download.file(“http://data.climateaudit.org/data/briffa.2008/TornFin.rwl.tab”,”temp.dat”,mode=”wb”)
load(“temp.dat”)
briffa=tree; dim(briffa) # [1] 188618 4

It’s always handy to make an information file for start and end dates of the cores. This can be done as follows:

url=”http://www.cru.uea.ac.uk/cru/people/melvin/PhilTrans2008/TornFinADring.raw”
x=read.fwf(url,widths=8);
briffa.info=unique(x);
names(briffa.info)=”id”
briffa.info$id=as.character(briffa.info$id)
briffa.info$id=gsub(” +$”,””,briffa.info$id)
row.names(briffa.info)=1:1022
briffa.info$start=tapply(briffa$year,factor(briffa$id,levels=briffa.info$id),min)
briffa.info$end=tapply(briffa$year,factor(briffa$id,levels=briffa.info$id),max)
briffa.info

Squinting as this information table, the first 431 entries are sorted by decreasing end date; the next 591 are sorted by increasing start date. The original article says that there are 430 samples from Finland and 587 from Sweden. So it appears highly likely that the first 431 entries are from Finland and the last 591 are from Sweden, with a slight inaccuracy either in the table in Briffa et al 2008 or in the SI now provided. The nomenclature is different as well with the first 431 having alphabetical starts, while most of the Swedish series have numerical starts.

The primary references for the Finnish data are Helama et al 2002 (Holocene) and Eronen (Briffa) et al 2002 (Holocene).

The Swedish Tornetrask data (Grudd et al. 2002) and Finnish–Lapland data (Eronen et al. 2002; Helama et al. 2002), for pine (Pinus sylvestris), were combined to create a single Fennoscandia regional chronology.

Eronen et al 2002 mentions IDs passim that have a similar appearance to the Briffa IDs. Some Finnish data pertaining to Helama et al 2002/Eronen et al 2002 has been archived at WDCP/ITRDB, but the archive is unfortunately spotty. The Swedish data (Grudd et al Holocene 2002) is entirely absent from ITRDB, though earlier Schweingruber data from Tornetrask is archived (and there might be some overlap).

I spent some time trying to match IDs numbers and found a few matches and report today on an interesting match for the old Sodankyla church (which is mentioned in Eronen (Briffa) et al 2002), which states:

Samples from the old Sodankyla parish church, nine beams in all, were also used (Lindholm et al., 1999).

Measurement data for Sodankyla church is archived at ITRDB. You can download my collation as follows, which shows that all the series have NGC in their ID format (and lists 9 beams (2 ores each), as in the article (though one ID (0008) is missing).

download.file(“http://data.climateaudit.org/data/briffa.2008/finl048.rwl.tab”,”temp.dat”,mode=”wb”)
load(“temp.dat”)
unique(tree$id)
# NGC0001A NGC0001B NGC0002A NGC0002B NGC0003A NGC0003B NGC0004A NGC0004B NGC0005A NGC0005B NGC0006A NGC0007A NGC0007B NGC0009A NGC0009B NGC0010A NGC0010B

Now let’s look up this code, NGC, in the Briffa information file. This yields one series NGC0003X ending in 1687 (just before 1689 construction).

briffa.info[grep("NGC",briffa.info$id),]
# 100 NGC0003X 1497 1687

It looked like Briffa (Eronen) et al might have averaged the two cores NGC0003A and NGC0003B and this indeed proved to be the case. The following script yielded the figure shown below:

temp=(briffa$id==briffa.info$id[100])
test1=ts(briffa$rw[temp],start=min(briffa$year[temp]))/10
temp=(tree$id==”NGC0003A” )
test2=ts(tree$rw[temp],start=min(tree$year[temp]))/10
temp=(tree$id==”NGC0003B” )
test3=ts(tree$rw[temp],start=min(tree$year[temp]))/10
y=ts.union(test1,test2,test3)
plot(c(time(y)),y[,1],type=”l”,xlab=””,ylab=””,main=”NGC0003 (Sodankyla)”)
points(c(time(y)),apply(y[,2:3],1,mean,na.rm=T), pch=19,col=2,cex=.5)


Figure 1. NGC0003

Helama et al refer to selecting a subset of living trees as follows:

The number of living trees used here to construct the ‘modern’ section of the chronology was limited to 50, in order that the recent era should not be over-sampled.

But there is no mention of “limiting” the number of beams from Sodankyla; quite the opposite. They say that 9 beams were used, but it appears that only one was used. So what happened to the other 8 cores? Not that I understand the need to “limit” the number of modern trees, given that an average is later taken. Surely this creates potential selection bias. How were the 50 series selected? And what was the effect?


73 Comments

  1. Hank
    Posted Dec 31, 2008 at 12:39 PM | Permalink

    How’s the saying go? …. “You can’t make cherry pie without picking cherries.”

  2. Gary
    Posted Dec 31, 2008 at 12:47 PM | Permalink

    One answer: nine beams were sampled and eight were inadequate (too short, bad angle, difficulty in determining rings, etc.). Maybe this is a subtle way of suggesting great effort (we searched for everything we could find, but alas the samples were just faulty and we recovered all that we could). There’s ambiguity in the word “used” – does it mean ‘used in the suite of samples’ or ‘used in construction of the chronology’? Whatever the case, it should have been explained more precisely.

  3. John Andrews
    Posted Dec 31, 2008 at 12:58 PM | Permalink

    Steve, can you clarify the number of beams and the number of samples. It isn’t clear from your exposition.

  4. Posted Dec 31, 2008 at 1:12 PM | Permalink

    Was the church considered to be built on a temperature or moisture limited site?

    I wonder how the dendros handle that one.

  5. Soronel Haetir
    Posted Dec 31, 2008 at 1:13 PM | Permalink

    The grad student typing in the measurements got tired and quit after the first core.

  6. Posted Dec 31, 2008 at 1:15 PM | Permalink

    What did the data from the other beams look like?

  7. EW
    Posted Dec 31, 2008 at 1:17 PM | Permalink

    nine beams were sampled and eight were inadequate

    If that were the case, why not to say so? If some of my herbarium specimens give only some information, I just write the reason for limited amount of data gained from that specimen in the Methods section.

  8. Hu McCulloch
    Posted Dec 31, 2008 at 1:25 PM | Permalink

    If all 9 of these series end by 1689, well before the standard instrumental series, what use would any of them be as a temperature proxy? How could they possibly be calibrated to temperature?

  9. henry
    Posted Dec 31, 2008 at 1:28 PM | Permalink

    Re: #4, Jeff Id

    Was the church considered to be built on a temperature or moisture limited site?

    I wonder how the dendros handle that one.

    Or I wonder if they were able to plot a rough location as to where the trees originally stood. See if any other trees from that original stand were sampled.

    Its possible that the trees used for the beams (requiring larger trees) may have come from some distance away.

  10. Jeff Alberts
    Posted Dec 31, 2008 at 1:40 PM | Permalink

    Steve, your blockquote at the end isn’t closed properly.

  11. EW
    Posted Dec 31, 2008 at 2:27 PM | Permalink

    Seems that Sodankyla is a closely studied location:

    It is probably not primarily water-limited – it’s a taiga. Here they say that “The Sodankylä-Pallas site is a typical representative of Eurasian taiga belt characterized by a mosaic of sparse conifer-dominated forests and open/forested bogs. The landscape is generally relatively flat or gently rolling although small mountain regions (fjelds) are typical.” Another site description here.

    In the abstract of Macias et al.(Eronen is among the et al.) they say about Fennoscandian woods: “Current July temperature and, to a lesser degree, May precipitation are the main growth limiting factors in the whole area of study.”
    Another Finnish local climate project studies pines at Laanila, close to northern treeline. They say, that they didn’t find any Mannian stick under these trees…

  12. Craig Loehle
    Posted Dec 31, 2008 at 2:39 PM | Permalink

    What is that saying about removing the beam from your own eye before removing the mote from your neighbor’s? In this case, ouch.

  13. RomanM
    Posted Dec 31, 2008 at 2:48 PM | Permalink

    To see what the rest of the trees from Sodankyla looked like I ran the following (after first running all of Steve Mc’s nicely done data gathering script) to generate some graphs:

    #form sodankyla tree ring ts
    nam = unique(tree$id)
    temp=(tree$id==nam[1])
    sod = ts(tree$rw[temp],start=min(tree$year[temp]))/10
    for (i in 2:length(nam)) {
    temp=(tree$id==nam[i])
    sod = ts.union(sod, ts(tree$rw[temp],start=min(tree$year[temp]))/10)}
    colnames(sod) = nam
    sod = ts(sod,start = min(tree$year))

    #plot trees
    par( mfrow = c(3,1))
    matplot(time(sod),sod[,5:6], xlab= “Year”, ylab = “Ring Width”, main = “NGC0003″, ylim=c(0,350),type = “l”)
    matplot(time(sod),sod[,-(5:8)], xlab= “Year”, ylab = “Ring Width”,main = “Excluding NGC0003 and NGC004″, ylim=c(0,350),type = “l”)
    matplot(time(sod),sod[,7:8],xlab= “Year”, ylab = “Ring Width”, main = “NGC0004″, ylim=c(0,350),type = “l”)

    which gave the result (all plotted on the same vertical scale):

    Look here if the graph looks small of your screen.

    There were a toatal of nine trees with all but one of them having two cores each. The results for the cores looked pretty consistent with each other except for the two trees graphed separately. The tree used by Briffa does not show the considerably higher values at the beginning of the series evident for most of the other cores. I thought it might be an artifact of the age of the wood, but all the cores appear to start at an age of one (I assume that the variable tree$age does indeed represent the actual age of the core). As well, there is still something strange about NGC0004 so further examination of the data is needed to figure out why.

    Re: Hu McCulloch (#8),

    If all 9 of these series end by 1689, well before the standard instrumental series, what use would any of them be as a temperature proxy? How could they possibly be calibrated to temperature?

    Theoretically, the data could be “calibrated” to the temperature by calibrating it to a temperature reconstruction created from the other series (preferably from the same geographical region) which do extend into the measured temperature era. I don’t know how Briffa would do this, but I have been looking at the problem for a while and I have been developing some methods for doing this using the approach of least squares inverse regression. Intuitively, you would be matching the wiggles in your data to the wiggles in the reconstruction for the period of overlap.

    • MC
      Posted Jan 1, 2009 at 11:52 AM | Permalink

      Re: RomanM (#13), Putting aside the questions for the use of the dubious ‘robust’ temperature reconstructions for the last 2000 years and assuming that there is a ‘linear’ relationship of some sort for tree ring width and temperature (I know Craig, its just an assumption for the sake of argument), if the tree ring data was averaged and errors put in as 2SD what would this correspond to in terms of temperature anomalies i.e. how many tree ring widths does it take to get 0.1 °C? 50? 100? Did anybody do a graph showing estimated anomalies for each tree and the average with errors included.
      Part of my new years resolution list is to get an R package so I can do this myself

      • RomanM
        Posted Jan 1, 2009 at 12:48 PM | Permalink

        Re: MC (#34),

        how many tree ring widths does it take to get 0.1 °C? 50? 100?

        There is no absolute answer for the question. Even given the validity of all of the assumptions, it would still depend on the physical characteristics of the type of tree and its physical environment. It could be estimted by regressing the tree ring data on the temperature record to determine the amount of increase or decrease in tree ring width due to a specific change in tempreature. The slope coefficient of the regression corresponds to a change of one unit of temperature.

        R is easy to install from http://www.r-project.org/ and the price is right. Go for it!

        • Mark T.
          Posted Jan 2, 2009 at 1:13 PM | Permalink

          Re: RomanM (#36),

          It could be estimted by regressing the tree ring data on the temperature record to determine the amount of increase or decrease in tree ring width due to a specific change in tempreature.

          Unfortunately, this method only works if the relationship is relatively linear (stationary) over time.

          Mark

        • RomanM
          Posted Jan 2, 2009 at 2:47 PM | Permalink

          Re: Mark T. (#43),

          I realize that. You left out a relevant phrase in my post: “Even given the validity of all of the assumptions…”.

          In general, unless there is stationarity of the relationship in time, the currently used methods of reconstruction cannot be expected to produce valid results, so I see that as an automatically required assumption in any paleoclimate situation. I have some ideas for checking such assumptions, but I suspect mentioning these is OT in this thread.

        • Posted Jan 2, 2009 at 3:53 PM | Permalink

          Re: Mark T. (#43),

          This method doesn’t work unless the data has a low noise level. Something which isn’t found in tree ring data. If you apply it to a large group, the noise tends to reduce (flatten) the historic signal average compared to the calibration range.

        • Posted Jan 3, 2009 at 5:58 AM | Permalink

          Re: RomanM (#36),

          It could be estimted by regressing the tree ring data on the temperature record to determine the amount of increase or decrease in tree ring width due to a specific change in tempreature. The slope coefficient of the regression corresponds to a change of one unit of temperature.

          More common dendro approach seems to be regressing temperature on proxies. See this post ,

          quite different CIs one obtains with these two different methods. But you know, in this field uncertainties are taken into account as fully as possible

        • MC
          Posted Jan 3, 2009 at 6:56 AM | Permalink

          Re: UC (#54), This is what I was talking about. Its fairly obvious that the confidence intervals are not 0.1 °C but it demonstrates that assumptions and method can be applied with transparency and then see where it leads. This shows that regression in this manner is not recommended. However this could also be determined without doing any maths as the imposed assumptions about calibration are reaching. Still its a good demonstration.

          With regards to Mark T I agree that there are issues with the nature of the regression and that a lot of the time realisable statistics cannot be inferred or obtained from a bounded data measurement i.e. measurement ± uncertainty, unless the percentage uncertainty is very low. This is something that happens frequently when method is applied to an empirical data set such as these tree rings. There is room for both non-linear, non-stationary, linear, stationary, periodic relationships if the measurement errors or CI are large enough. Then it comes down to conceptual application. The variation of one relationship by adjusting a parameter in one type of equation is the same as adjusting another in a linear equation. They cannot be distinguished in the data set so the application of one relationship as opposed to the other in the ‘noise’ can become troublesome. There has to then be a theoretical reason why one method is favoured.
          I also agree though that this means that the data is not as useful but that’s science. There is a danger of imposing a relationship for the sake of a result without clearly showing the limits of assumption and method. I can’t overemphasise to people that science is not just about the result, its about method and assumptions too and if the concept fails in this stage then this must be applied to the results.
          This doesn’t happen a lot and hence we get into a whole discussion about maths and statistics that is unnecessary apart from the bare bones auditing to try and repeat what someone has done with the data they used (FOI requests notwithstanding). I also appreciate that this holistic approach does not appear to be as clear in the various ‘robust’ reconstructions from other reporters.
          I guess that everytime I read more about reconstructions I get more depressed that a lack of clarity in scientific method, a ‘sloppiness’ if you will, passes though without proper rebuttal.

  14. Scott Brim
    Posted Dec 31, 2008 at 3:14 PM | Permalink

    Hu McCulloch #8: If all 9 of these series end by 1689, well before the standard instrumental series, what use would any of them be as a temperature proxy? How could they possibly be calibrated to temperature?

    RomanM #13: ….. Theoretically, the data could be “calibrated” to the temperature by calibrating it to a temperature reconstruction created from the other series (preferably from the same geographical region) which do extend into the measured temperature era …..

    A commonly accepted time-temperature relationship graph for the last two-thousand years is available within the peer reviewed climate science literature which should allow a reconstruction of the necessary data.

    • Geoff Sherrington
      Posted Dec 31, 2008 at 8:02 PM | Permalink

      Re: Scott Brim (#14),

      A commonly accepted time-temperature relationship graph for the last two-thousand years is available within the peer reviewed climate science literature which should allow a reconstruction of the necessary data.

      It is precisely this relationship that needs auditing because there is so little belief in it. A purpose of threads like this is to find ways to validate and strengthen and refine that relationship to make it valuable, not to simply accept it. So far, it has been shown that people can count and correlate tree rings from various localities with acceptable agreement. Most of us can count and correlate. But, the link to temperature remains replete with unsolved problems. Please read past CA posts more deeply to see some of the problems.

      • Scott Brim
        Posted Dec 31, 2008 at 9:09 PM | Permalink

        Re: Geoff Sherrington (#27)

        Actually, I was repeating word-for-word a comment I made about a month ago in another thread. So in other words, it’s deja vu all over again, and by design.

        The goal was to see if Craig Loehle would pick up on the true import of the comment, in relationship to its actual context, just as he did the first time around.

        Let’s see if anyone else but Craig remembers which topical thread it was in. However, under no circumstances will I cave in to pressure from other CA readers and reveal which thread it’s hidden in.

        You Geoff Sherrington have missed the boat in deciphering my ironic remark. May you therefore be condemned to working for mining engineers for the rest of your career and be assigned the task of writing strange doggeral concerning desert sand and the Sphinx.

  15. Steve McIntyre
    Posted Dec 31, 2008 at 3:28 PM | Permalink

    Roman, I assigned the first available year as age=1. This is what dendros do, tho heartrot is a problem with old trees and it may not be true. So there is an real potential error in the age. It’s the sort of statistical issue that is very poorly handled in dendro literature (the statistical issues in making a tree ring chronology being sort of interesting. I’ll try to post up something on this in this cycle.)

  16. Jean S
    Posted Dec 31, 2008 at 3:43 PM | Permalink

    I think those Sodankylä measurements are available from SAIMA (here).

    AFAIK those Finnish dendro people are very helpful, so I’m pretty sure you get a nice respond if you need any assistant with those data sets. Some of the authors of the papers mentioned have a nice web site, LUSTIA, which I recommend also for CA readers to check out. The main maintainer of the site, Mauri Timonen, commented also here in CA a awhile back, so I think he will read this post :) Hi Mauri!

    This will be my last comment this year (it’s quarter to midnight here ;) ), so happy new year to Steve and all CA readers!

  17. EW
    Posted Dec 31, 2008 at 3:49 PM | Permalink

    Jean S, I have yet 1 h 15 min to midnight. Yes, I just found some more articles from the Lustia site you mentioned. Apparently, the Finns have now over 7000 year chronology of pine. They would like to go more back in time, to find out about that swift cooling drop 1-5 C global mean some 8000 yes ago.

  18. Posted Dec 31, 2008 at 3:54 PM | Permalink

    #16 make sure you don’t forget the leap second! :) 1/f in Earth rotation makes it difficult to predict those. Happy new year dudes!

  19. Steve McIntyre
    Posted Dec 31, 2008 at 4:01 PM | Permalink

    #16. Jean S, Happy New Year to you as well. Thanks for the link. That looks much more complete than the ITRDB version.

  20. RomanM
    Posted Dec 31, 2008 at 4:16 PM | Permalink

    We still have some hours to go, but Happy New Year to everyone on CA. May the next year be even more fun on this blog than it has been so far!

  21. Fred
    Posted Dec 31, 2008 at 4:32 PM | Permalink

    “Was the church considered to be built on a temperature or moisture limited site?”

    And maybe the used kiln dried lumber :)

  22. Posted Dec 31, 2008 at 5:54 PM | Permalink

    It still baffles me how tree ring growth is singularly linked to temperature. Considering that as the trees diameter gets larger each year , more material has to be grown to just cover the added area via diameter changes, just as you need more jean when you grow from a 28″ waist to a 32″ waist, and then to pull some linearality in added width out of it, seems quite extraordinary. It would seem to me that moisture and soil nutrient levels would have much more bearing on it.. Maybe I need some insight from the dendros lurking.

  23. Brooks Hurd
    Posted Dec 31, 2008 at 7:33 PM | Permalink

    Still 6 1/2 hours to go in PST.
    Happy New Year!

  24. David L. Hagen
    Posted Dec 31, 2008 at 7:35 PM | Permalink

    Steve
    In Norway, the oldest wood church is the Urnes Stave Church built in 1150 AD in Luster, beside the Sognefjord.
    See: Stave Churches

    The Borgund church was built in 1180 AD.

    At 1689 AD, you might say that the Sodankyla Church is quite “modern”!

    You may find some comparative cores in some of the following links:
    Norwegian conifer chronologies constructed to date historical timber Terje Thun. (Department of Biology, Norwegian University of Science and Technology, NTNU, NO-7491 Trondheim, Norway) Dendrochronologia Volume 23, Issue 2, 8 December 2005, Pages 63-74

    Tree-ring chronologies have been constructed for Scots pine (Pinus sylvestris L.) from central Norway (Trøndelag), west Norway and south-east Norway back to 552, 765 and 871, respectively. The purpose was to enable the dendrochronological dating of medieval and post-medieval timber constructions and other wooden artefacts, and this proved to be possible. Absolute dates of floating chronologies based on medieval building material from urban excavations in Trondheim, Bergen and Oslo were also obtained. In northern Norway, samples taken from recently living Scots pines in Troms have enabled the construction of a chronology back to 1294.

    Chronologies based on samples taken from timber constructions built of Norway spruce (Picea abies (L.) Karst.) have been constructed for central Norway (Trøndelag) and outer Oslofjord back to 1351 and 1466, respectively.

    Historical events have influenced the availability of building timber. Although much material is available from the early and post-medieval periods, hardly any constructions have been found dating from an approximately 60-year period after the Black Death in 1349–1350.

    The various regional chronologies based on Scots pine have a matching tree-ring pattern with high t values and sign tests at a 99.9% level of significance. The match between Norwegian chronologies and chronologies constructed in Sweden also give high t values and sign tests at a 99.9% level of significance.

    Dendrochronological Dating of Wooden Artefacts using Photography Terje Thun; Haringkan Hytteborn, Norwegian Archaeological Review, Volume 40, Issue 2 2007 , pages 179 – 186

    Bibliography, Terje Thun

  25. Larry Huldén
    Posted Jan 1, 2009 at 6:38 AM | Permalink

    Henry #9 said: “Or I wonder if they were able to plot a rough location as to where the trees originally stood. See if any other trees from that original stand were sampled.
    Its possible that the trees used for the beams (requiring larger trees) may have come from some distance away.”

    We can say for sure that the trees used for the church have come from upstreams along the river Kittinen within 0-60 km from Sodankylä. No trees can have come from downstreams.
    This area is climatologically quite uniform with continental climate.
    Upstreams from Sodankylä we find the village of Pokka where temperatures down to -50-51C degrees have been measured.

  26. Ron Cram
    Posted Jan 1, 2009 at 6:41 AM | Permalink

    Steve, it is possible for you to archive at the appropriate data center the solution to these little puzzles, as you call them? If Briffa will not provide the metadata at the archive, will they accept metadata from a third party if you state the uncertainties? I am not sure exactly how these things work. It seems to me at a minimum this would provide a service to the scientific community and it may also embarrass Briffa and others into providing more metadata.

  27. Tony
    Posted Jan 1, 2009 at 6:59 AM | Permalink

    #13 RomanM:
    It looks to my eye like NGC0004 should start 75 years earlier (roughly).

    And NGC0003 got used? It’s (almost) the only one from the nine not showing strong ring growth before 1500. How very peculiar indeed…

    (Been watching too much Lord of the Rings lately…)
    McIntyre: One that is here yet nine there were set out from Sodankyla. Tell me where are the others? For I much desire to see them.

    Phil Trans B: They have fallen into shadow.

    McIntyre: Then it is true. The flame of Briffa has taken them.

    • RomanM
      Posted Jan 1, 2009 at 10:26 AM | Permalink

      Re: Tony (#30),

      That was my initial reaction as well since the initial pattern for NGC0004 looks very similar to that of all the others except NGC0003. However, without knowing whether there is a physical effect due to the age of the tree (intuitively, it makes some sense to me that rings formed early in the tree’s life might be wider than later due to smaller diameter, initial higher growth rate, etc.), I didn’t hazard a guess that it might be due to possible shift in the dating process.

      It does appear that Briffa’s choice of NGC0003 could have been made on the basis of a high correlation with the effect desired in the reconstruction… ;)

  28. Steve McIntyre
    Posted Jan 1, 2009 at 9:00 AM | Permalink

    Points about tree rings and temperatures have been made hundreds of times on many threads. Let’s stipulate that all relevant points have been made and only deal with Finnish matters here.

  29. Steve McIntyre
    Posted Jan 1, 2009 at 9:12 AM | Permalink

    I spent some time looking at the SAIMA data here . As far as I can tell, they have also archived the data in mirror at ITRDB/WDCP (which I had already picked up.) Thus the Sodankyla data at SAIMA matches what I’d already used.

    Here’s the problem.

    As far as I can tell, Eronen et al 2002 uses many sites that are not included in the SAIMA archive. It looks like Timonen (FFRI) has archived data, but it looks like another group (Helsinki: Eronen?) hasn’t. For example, Fiugre 1 of Eronen et al 2002 (HOl) shows sites by Ailigasjarvi Lake, none of which appear to have been archived.

    Also the Timonen archive is incomplete, with more sites being referred to than are available e.g. Leppäjärvi, Enontekiö and all sites without a red DB.

    • Jean S
      Posted Jan 1, 2009 at 2:42 PM | Permalink

      Re: Steve McIntyre (#32),
      Eronen’s data might be available somewhere else (e-mail him if you need it). The groups are now co-operating under LUSTIA in order to build the Finnish Tree-Ring Data Base, which I think you’ll appreciate. According to Timonen that should be available latest by the WorldDendro2010 meeting (June 2010).

      Re: Dennis Wingo (#35),
      Well, I think they won’t become popular in those circles in near future … especially after this poster (covered also here). Especially Timonen’s views are very unorthodox as you can read from his latest interview (in Finnish, Google translation): he is arguing that there is a real chance that the world will be cooling in the coming decades and that it is dangerous for the society to prepare only for the warming option!

  30. Posted Jan 1, 2009 at 12:16 PM | Permalink

    EW 16, Steve

    I am amazed at the Finnish work on the long term temperature reconstructions. Has Mike Mann included the Finnish Tree line data into his studies? The Medieval Warm Period and before is extremely well represented in this study as well as other warm and cool periods. The only pause that I have was that the 1783 Laki explosion and the known reduction in Northern European temps is not represented. The Tambora event is though.

    The big question that I have is why is this Finnish data being ignored by the wider IPCC and other communities interested in the anthropogenic global warming issue?

    Yea, I know, but I would like to hear what others think.

  31. Hu McCulloch
    Posted Jan 1, 2009 at 7:55 PM | Permalink

    Re Geoff S #26, Scott Brim #14, 27, I had assumed that Scott was referring tongue-in-cheek to the HS as the the definitively peer-reviewed thermometer for the past 2000 years. But perhaps Craig remembers his earlier post? (I can’t get the search engine to work at the moment.)

    • Geoff Sherrington
      Posted Jan 1, 2009 at 10:17 PM | Permalink

      Re: Hu McCulloch (#38),

      I think I had the mickey taken out of me for poor memory recall. Can’t remember the context of the original to Craig either. Please excuse me for slipping into Geoffrey Lecture mode, it’s a weakness that I exploit in myself.

    • OldUnixHead
      Posted Jan 2, 2009 at 7:13 AM | Permalink

      Re: Hu McCulloch (#38), and Re: Geoff Sherrington (#39), etc.

      I think you’re looking for this and CL’s reply here (although, at the time, it seemed a bit more cryptic to me than ScottB alluded, above). Even after it’s explained, it’s still gets me chuckling.

      • Geoff Sherrington
        Posted Jan 2, 2009 at 5:30 PM | Permalink

        Re: OldUnixHead (#41),

        Yes, I swallowed the bait of Scott Brim #14 hook line and sinker. For some wrong reason I though he was a newcomer.

        Re: Mark T. (#43),

        One cannot go earlier than the written record and instrumental periods without some assumptions about stationarity/linearity/uniformitarianism. One might have confidence in topics like orbital geometry and resulting solar irradiance on Earth, but even now advancements are bring made, like the properties (size, mass) of some lesser orbiting bodies.

        Of techniques in general, delta O-18 isotope climate dating methods have a similar fundamental objection to the CO2 climate sensitivity equation if it exists). Both start with qualitative assumptions that turn into precise quantitative relations.

        Re: Jeff Id (#46),

        Might I rspectfully suggest that more important than a low noise level, is a high confidence that the cause and effect are quantitatively linked by equations that are “robust”? Mere repetition of techniques whose bias has not been eliminated leads to Mark Twain’s “chloroform in print” and no improvement in quality. Such repetitions should be placed in experimental texts and papers and not used in mature Summaries for Policy Makers.

  32. andy
    Posted Jan 2, 2009 at 6:12 AM | Permalink

    The slide 10 of http://lustiag.pp.fi/advance10K.pdf locates the FIL6201 serie to Lake Ailigas. Briffa seems to have several FIL6*** series in his SI as well.

  33. Scott Brim
    Posted Jan 2, 2009 at 8:46 AM | Permalink

    Hu … Re Geoff S #26, Scott Brim #14, 27,
    I had assumed that Scott was referring tongue-in-cheek to the HS as the the definitively peer-reviewed thermometer for the past 2000 years.

    Many of the warmers do view the hockey stick as being the definitively peer-reviewed thermometer for the past 2000 years.

    In the world of nuclear safety analysis, this kind of circular reasoning would never be allowed to contaminate the formalized analytical processes.

    Be that as it may, the hockey stick is often an indispensable tool for wrapping up loose ends, when loose ends need to be wrapped up. Especially for those who are assessing various climate change impacts.

    Am I correct in observing that its popularity among climate scientists for that sort of purpose is growing with every passing year? And in spite of M&M and the Wegman report?

  34. Mark T.
    Posted Jan 2, 2009 at 3:43 PM | Permalink

    Sorry RomanM, I wasn’t trying to correct you, merely add to your description an explicit statement (meant for MC) the number one assumption that must be satisfied, i.e., linearity/stationarity (the two technically go hand in hand), for regression to work.

    I know you understand that, just not everyone else does, strangely not even the climatologists.

    Perhaps one of these days we’ll get Steve to generate a thread specifically for discussing means for checking such an assumption?

    Mark

    • RomanM
      Posted Jan 2, 2009 at 4:49 PM | Permalink

      Re: Mark T. (#45),

      No offence taken, Mark.

      Although wide participation is useful for initially determining the viability of an idea, the problem with such threads is that the material is pretty technical so that it limits the participation to those with a suitably technical background – mathematical derivations, non-linear regression, multivariate statistical methods, etc. Such concepts are better discussed in smaller groups.

      Re: Jeff Id (#46),

      I think of the “signal” as a deterministic sequence of values (they have already been generated) which need to be estimated. In this context, I tend to agree with Mark that it is a problem of extracting the information which is improved by increasing the amount of information.

  35. Mark T.
    Posted Jan 2, 2009 at 4:18 PM | Permalink

    Define low, Jeff… ;) Noise is immaterial when the governing equations are time varying anyway. The way around “high” noise is to increase the number of series in the regression, but they would all require linear/stationary responses. Not possible last time I checked.

    Mark

    • Posted Jan 2, 2009 at 5:39 PM | Permalink

      Re: Mark T. (#47),

      I think low means very low, I’m used to nice clean(ish) engineering electronic signals not tree rings. From my experimenting I found any noise level caused trouble. What Roman says is correct, high sampling improves the noise level but there are several problems in that the noise may not be random but rather moisture related. Also, applying the same defective process repeatedly and averaging creates a high quality r fit to the calibration range but the result cannot extract a historic signal. I tried all different random ‘noise’ levels and found it only affected the magnitude of the distortion. Good fit on the calibration, bad fit for historic signal.

      I think Geoff said it better than me.

  36. Mark T.
    Posted Jan 2, 2009 at 5:01 PM | Permalink

    True, but you’ll notice the more technically themed threads tend to devolve to a few interested posters. Heck, there are even a few threads in which UC is the only one left posting. :)

    Mark

  37. Mark T.
    Posted Jan 2, 2009 at 5:38 PM | Permalink

    Agreed Geoff, but in the absence of evidence of stationarity/linearity/uniformitarianism, one must assume the null hypothesis: a lack thereof. In fact, the evidence actually indicates the opposite, i.e., non-linear, non-stationary and not uniform). I know, the crowd would then whine “but what are we supposed to do?” To which I answer “find another means as this line of evidence is not valid.” Oh well, them’s the breaks, right? Bad science is no replacement for no science.

    Mark

  38. Larry Huldén
    Posted Jan 3, 2009 at 2:18 AM | Permalink

    One thing should be pointed out. It was mentioned that the Sodankylä tree rings end in 1689. It is not completely sure that the church was build in 1689. It was in use in that year but it is more probable that it was build already late in 1688.
    Constructions work was usually done in summer or autumn. Trees were always felled during winter and dried for one season (year) before used. The transportations were always in the summer season.
    Final decision about the building place was made in 27 Febr. 1688. At that date we may expect that the trees were already available at the place and probably transported there in the previous year. That is because the local inhabitants had offered this place already in 1687. The trees were possibly felled in the winter 1686-87.
    The construction of the church could then be done within 2-3 months in 1688(or 1689).

  39. Scott Brim
    Posted Jan 3, 2009 at 10:46 AM | Permalink

    Geoff Sherrington: Yes, I swallowed the bait of Scott Brim #14 hook line and sinker. For some wrong reason I though he was a newcomer.

    For several years I posted as “Scott-in-WA.” But I decided recently that people whom I converse with off-line about the topic of AGW and its public policy impacts needed to know that it was me who was offering various commentaries concerning the poor quality of today’s climate science — both in the adequacy of its analytical processes and in the quality of its analytical products — relative to what is commonly expected from the application of the Scientific Method within the nuclear industry.

    In any case, my thoroughly dry sense of humor continues as before, at least to the extent Steve McIntyre allows it.

    MC: …. I also agree though that this means that the data is not as useful but that’s science. There is a danger of imposing a relationship for the sake of a result without clearly showing the limits of assumption and method. I can’t overemphasise to people that science is not just about the result, its about method and assumptions too and if the concept fails in this stage then this must be applied to the results….

    In the world of nuclear, it’s all one thing. A quality assurance failure in the process is fully equivalent to a quality assurance failure in the final product. The potential impacts of poor quality work on public safety demand this be so.

    As of yet, nothing done so far in the world of climate science has had any significant impact on the life styles of those might be most affected by the outcome of carbon-restrictive public policy decisions.

    This situation is about to change. What we see here on Climate Audit is a possible harbinger of the kind of informative scientific debate which might ensue if carbon-restrictive public policies actually begin seriously affecting the average person on the street.

  40. MC
    Posted Jan 3, 2009 at 10:59 AM | Permalink

    Back to the church before we all get OT, if not already. Was there a reason why other trees in the area were not sampled, or do the trees used as beams offer a cut-off in the past? Are other trees in the area in the Finnish database?

  41. Mark T
    Posted Jan 3, 2009 at 11:10 AM | Permalink

    The whole debate about tree-ring proxies should have been over at “prove the relationship is stationary.” Period. But alas, here we are bickering about what most of us know to be true, while the addicts refuse to let it go.

    Mark

  42. Steve McIntyre
    Posted Jan 3, 2009 at 11:11 AM | Permalink

    I sent an inquiry to Mauri Timonen about the Finnish data used in Briffa and received a very prompt, cordial and informative reply (sending greetings to CA as well), which I’ll discuss in a forthcoming post on the Briffa data.

  43. theduke
    Posted Jan 3, 2009 at 11:17 AM | Permalink

    Larry Hulden, #53: I think this church took more than three months to build. It is a log church, the walls are solid logs. The following link says that dendrochonrologists determined that of ten logs tested, most logs were cut down in 1687, one in 1688 and two in 1689. It wouldn’t surprise me to learn that construction was going on all through that period. The logging alone would have taken two years, if you take into account the number of logs in the walls. It’s a “log pillar” church, which means the center of the longitudal wall is braced with log pillars cross-laid. I’m also not convinced carpenters at the time dried their wood for a year. And I believe they probably cut the trees and sledded them out in winter because it was easier to move them on snow and ice. As for the location of the trees and the distance of transport, I’d guess tha the trees were all local, probably cut down within a mile of the building site, since pinus sylvestris is ubiquitous in the area.

    http://books.google.com/books?id=PjPWrlFpi-cC&pg=PA132&lpg=PA132&dq=Sodankyla+Church&source=web&ots=XQP7yBNRmF&sig=5zITT5ppzIrRvPyTA3OdVkqdaiI&hl=en&sa=X&oi=book_result&resnum=19&ct=result#PPA131,M1

    • Andy
      Posted Jan 3, 2009 at 2:13 PM | Permalink

      Re: theduke (#60),

      Logging could have taken two years wether there had been only one or two men working. But from a group of skillful craftsmen of that time, it had been rather a couple of months. Anyway, via a system called taxation, local farmers or nomads or whatsoever were building the church, amount of workforce was not the issue.

      And I assume the original reason why the logs or beams were sampled, was exactly the timing issue. When the building yesr of the church is known from other sources, it’s straight forward to assume that at least most of the trees should have cutted down on that year or closely before.

      Major issue should be instead, arfe the Briffa samples averaging well the spatial and timing distributions. Eg, are there samples from the colder side of the current tree line from trees grown there at the MWP, and samples from the warmer side of the current treeline from the LIA period, and then claimed that the trees show steady growing during the last thousand years?

  44. Larry Huldén
    Posted Jan 3, 2009 at 2:21 PM | Permalink

    “I think this church took more than three months to build.”
    According to historical documents the workers (they were specialists on such buildings) promised to start the work in the autumn 1688 and finish the work before Christmas. This is described in the history of Suur-Sodankylä (1995)(in Finnish).
    There were other people preparing the logs before the construction work, this took of course a longer time.

  45. tty
    Posted Jan 3, 2009 at 7:17 PM | Permalink

    Three months is quite realistic if the timber was already prepared. Timber buildings in “knuttimmer” technique are still being built in Scandinavia and such buildings can be erected quite quickly by experienced workmen. They are also simple to dismantle, move and re-erect. A surprisingly large proportion of old buildings of this type have been moved, often more than once, which can cause problem with provenience. However that should not be the case with a church.
    Also, yes the logs would probably have been dried for at least a year before being used. If the logs are hewn square while still fresh the uneven subsequent drying will lead to serious splitting, and it is doubtful if such logs would last for 300 years. I’ve seen timber building where the pine logs had not been dried properly, and have deteriorated badly in less than 100 years.

  46. Larry Huldén
    Posted Jan 4, 2009 at 2:23 AM | Permalink

    theduke said:
    “As for the location of the trees and the distance of transport, I’d guess that the trees were all local, probably cut down within a mile of the building site, since pinus sylvestris is ubiquitous in the area.”
    The logs were transported floating in the stream. Pinus silvestris trees are not growing in dense stands. Logs had to be carefully selected for such buildings. The local trees were certainly already used, and transport on land was burdensome. As a consequence, high quality logs were sought for upstream along the river. A transport of logs downstream 40-80 km was a simple task with long traditions.
    The logs were first transported to the destination place before drying. Otherwise dried and prepared logs could have been damaged in the stream during the transport.

  47. Jeff Norman
    Posted Jan 4, 2009 at 11:24 AM | Permalink

    Larry Huldén: January 3rd, 2009 at 2:21 pm

    According to historical documents the workers (they were specialists on such buildings) promised to start the work in the autumn 1688 and finish the work before Christmas. This is described in the history of Suur-Sodankylä (1995)(in Finnish).

    How many people here have had or have had a friend who had a contractor promise they “will finished the work by Christmas”. Another historical trend that demonstrates no significant changes.

  48. theduke
    Posted Jan 4, 2009 at 12:13 PM | Permalink

    Larry, #64: sounds like you have a lot of knowledge of the area and its history. I have some knowledge of old timber structures, although not necessarily of log structures. The link I provided asserts that two of the logs tested were cut down in 1689, which, if accurate, suggests the building needed extra trees during construction in 1689. Jeff Norman also makes a valid point in #65.

    Regardless, the subject is not necessarily germane to the topic at hand, so if you or anyone else wants to continue discussing it, you can email me at doug(at)eatonbuilding(dot)com.

  49. Graeme Rodaughan
    Posted Jan 5, 2009 at 1:28 AM | Permalink

    Reminds me of finding useful car parts from a car wrecking yard, you have to serach for a while to find the actual useful bit that “fits” what your looking for…

    Kinda presupposes the result preceding the search.

  50. Larry Huldén
    Posted Jan 5, 2009 at 1:44 AM | Permalink

    Jeff Norman said:
    “How many people here have had or have had a friend who had a contractor promise they “will finished the work by Christmas”. Another historical trend that demonstrates no significant changes.”

    In this special case the Swedish king simply offered extra money when the decision about the building place was settled. I think the workers were eager to get the work quickly done.

    I was thinking of a possible misinterpretation about when the trees were cut. Trees from “1689” could have been cut in 1688 (Jan-Febr) and the other accordingly one year earlier than expected. On the other hand they might have made some corrections in 1690 or so (using trees from 1689). We don’t have detailed information to be sure about that.

  51. tty
    Posted Jan 5, 2009 at 8:04 AM | Permalink

    This is getting more and more OT, however this is my take on how the timber would have been procured, based on the church being built in late 1689, and how these things have traditionally been done until quite recently in Sweden and Finland.

    The timber would have been felled during the winter 1687/88. This was almost always done in winter for two main reasons, the timber was much easier to get out of the forest by sledge on the snow, and there was less conflict with farming work. Winter is still the preferred season for logging in Scandinavia.

    The timber was sledged down to the river during the late winter and put in the river during the spring flood in 1688 and floated down to the building place during the summer.

    It was then stacked for drying during the late summer/autumn of 1688. A year later, after the summer season 1689 the timber would have dried, and could be prepared (hewed to shape), ready for building the church in late 1689.

  52. Håkan B
    Posted Jan 6, 2009 at 11:38 AM | Permalink

    I grew up in central Sweden during the fifties where old people still knew how to build with those methods.

    One thing I learned was the importance of selecting the right trees. Size and shape weren’t the only criteria to consider when selecting a tree, it was very important to find trees that had been groving slow, those were more resistant to weather. Old folks knew how to select those trees. So my conclusion is that samples from trees used for buildings are not representative the general growing conditions for an area.

    Just an idea!

    Steve: Nice observation. I’ll pass this on to Rob Wilson.

  53. theduke
    Posted Jan 7, 2009 at 11:00 AM | Permalink

    I’ll extend the invitation again: if anyone wants to discuss the particulars of the construction of this building, I’d be interested to know what you know. While I am not a member of the following group

    http://www.tfguild.org/ttrag.html

    I have built timber structures with several of them. I’ve been to seminars and taken courses from others and read the books of those who are authors, most notably Abbott Lowell Cummings and Jack Sobon.

    doug@eatonbuilding.com

  54. Posted Jan 14, 2009 at 1:11 PM | Permalink

    It looks to my eye like NGC0004 should start 75 years earlier (roughly).

  55. curious
    Posted Apr 3, 2009 at 8:36 PM | Permalink

    Is this the same Briffa?

    http://users.ictp.it/~smr1972/Briffa_Cook.pdf

    (from Kenneth at #23 “Ammann’s April Fool’s Joke” : http://users.ictp.it/~smr1972/ )

One Trackback

  1. [...] Briffa and Sodankyla Church [...]

Follow

Get every new post delivered to your Inbox.

Join 3,299 other followers

%d bloggers like this: