Steig’s “Hockey Stick”

I’m writing a response to Gavin Schmidt’s rant about Yamal, which I should finish by tomorrow. Schmidt’s rant does not refute anything in my Yamal post. Indeed, Schmidt barely touches on the actual content of my post. Most of his post has nothing to do with Yamal.

In the present post, I’ll deal with the following spitball:

Similarly, McIntyre recently accused Eric Steig of suppressing ‘inconvenient’ results from an ice core record from Siple Dome (Antarctica). Examination of the record in question actually demonstrates that it has exceptionally high values in the late 20th Century (reflecting the highest temperatures in at least the last 700 years, Mayewski et al.), exactly counter to McIntyre’s theory.

Schmidt’s complaint arose from a CA post earlier this year, in which I criticised Neukom for their failure to archive proxy data in a study prominently cited in IPCC AR5 First Order Draft. Despite self-serving assertions that archiving has improved (e.g. Maibach et al 2012), little has really changed.

In my post on Neukom, I observed that one of the long South Hemisphere series of Neukom et al had been developed by Eric Steig in the mid-1990s and was not “new” since AR4. I also observed that it was unarchived. Based on the fact that this series had not been archived, I speculated (cynically, I’ll admit) that the series would not have a HS-shape.

Steig turned up to comment. He first claimed that the data had been archived and that I had simply not been “adept” enough to locate it. Subsequently, Steig conceded recognized that the data had not been archived and (commendably) placed the data online at his website. The comments ar worth re-reading.

Schmidt claimed that my cynical surmise had been refuted in this case because Steig’s dD record “has exceptionally high values in the late 20th century”:

Examination of the record in question actually demonstrates that it has exceptionally high values in the late 20th Century

Unfortunately, as too often, Schmidt’s statement is not only untrue, but easily seen to be untrue. Below is a plot of the 20th century portion of Steig’s dD record. There isn’t a HS shape in the 20th century – if anything, there is a slightly downward trend. Nor does the record have “exceptionally high values in the late 20th century”, as shown below.


Figure 1. Steig dD from Siple Dome – 20th Century

Update: Schmidt posted up a 1000-year graph in a form similar to the graph shown in comments below, describing the graphic as showing that “the 20th Century is warmer than the long term mean in both dD and d18O”. The archived d18O series is relatively short. If Schmidt had originally said that “the 20th Century is warmer than the long term mean in dD”, I would not have taken exception to his comment. But that’s not what he said. He said that the unarchived record “has exceptionally high values in the late 20th century”. It didn’t. As I originally observed, in my opinion, if this series had had a big HS, I’m convinced that it would have been publicized and archived years ago. As events turned out, it was archived approximately 18 years after being collected, and only after an issue was made about it at Climate Audit. I didn’t use the word “suppress” – that’s Schmidt’s word. In this case, it’s more that the results seemed “uninteresting” and were neglected. One situation is not necessarily the same as another.

83 Comments

  1. Steve McIntyre
    Posted May 12, 2012 at 3:57 PM | Permalink
    	loc="http://faculty.washington.edu/steig/sipledome/SipleDomeCores_AB.txt"
    	download.file(loc,"temp")
    	work=read.table("temp",skip=3,header=TRUE)
    	names(work)=c("year","dD","O18")
    	work[work==999999]=NA
    	steig=work
    
       trim=function(x) window(x, start= min(time(x)[!is.na(x)]),end=max(time(x)[!is.na(x)]) )
    	dD=trim(ts(rev(steig$dD),start=995))
    	x=window(ts(scale(dD), start=tsp(dD)[1]),1880)
    
    #png(file="d:/climate/images/2012/proxy/siple_dD_20cent.png",h=420,w=600)
    	plot(x,ylab="SD Units" ,xlim=c(1880,2000),ylim=c(-3,3))
    	title("Steig's dD: 20th Century")
    	abline(h=0,lty=3)
    #dev.off()
    
    
  2. Salamano
    Posted May 12, 2012 at 4:32 PM | Permalink

    Are there any other ways to filter, scale, adjust, etc. this data so that a graph of it shows what Gavin was talking about? I suspect he wouldn’t have made his statement up out of thin air, os perhaps his graph looks different somehow?

    • Posted May 12, 2012 at 4:54 PM | Permalink

      The most recent record has a value of 999999 as a placeholder for no data. If you actually graphed that value there would be a great big blade.

    • Steve McIntyre
      Posted May 12, 2012 at 5:05 PM | Permalink

      Here’s a plot of the millennium of Siple dD against Yamal. Siple dD is somewhat elevated in late 19th century and 20th century, but nothing more than that. Not like Yamal with a nearly 6-sigma (smooth) close.

      Update: For comparison, here is Schmidt’s rendering of the same data:

      Schmidt’s editorial comment was:

      The data from the archived file are plotted below, annual data and loess smooth (span ~ 30 years). The 20th Century is warmer than the long term mean in both dD and d18O (as with the figure in Mayewski et al), and the late 20th century appears somewhat exceptional in the d18O record. It’s worth pointing out that this data got McIntyre’s attention because they were used in a reconstruction by Neukom, not really something one would expect if people were suppressing the data.

      • Jerome
        Posted May 12, 2012 at 9:47 PM | Permalink

        On Gavin’s graph, the only exceptional thing i see is the coldness in 1780-1810.

        The rest also seems very normal, even the final uptick is exactly the same as the one in n 1870.

        • Posted May 12, 2012 at 10:05 PM | Permalink

          Gavin’s in a bind, because the d18O data he is plotting is not the same as was used for the Mayewski, Steig et al (04) plot that was the basis for his earlier statement.

      • Kenneth Fritsch
        Posted May 13, 2012 at 10:23 AM | Permalink

        When Schmidt can make the following observation about the temperature proxies in the graph excerpted by SteveM above, I would have to say that we are definitely not getting a science based one. On the one hand, he says the late O18 proxy is somewhat exceptional in the late 20th century and then on the other hand admonishes his strawmen for thinking there was any motivations for repressing data because Gavin evidently is implying here that the proxy data goes against consensus thinking on the modern warm period. (Is it just me or have my recent readings of Gavin’s comments at RC, that lead me to opine his statements are a bit cloudy on the issues, been noted by others and made it difficult to determine exactly what he means.)

        Also is it not telling about these proxies under discussion here, and proxies in general, that when one can scramble to obtain proxy series that show very different responses that:

        (1) Reconstructions are very sensitive to the selection process – and selection processes that are not made a priori.

        (2) An important issue here would seem to me to be whether one accepts these large proxy variations with near proximity as actually representing temperature difference or the series merely showing the meanderings of a proxy primarily responding to or being affected by something other than temperature.

        (3) Also would not large variations in close spaced proxies imply huge confidence intervals for reconstructions that can practically only sparsely sample major regions or the globe?

        Gavin’s Response: “The data from the archived file are plotted below, annual data and loess smooth (span ~ 30 years). The 20th Century is warmer than the long term mean in both dD and d18O (as with the figure in Mayewski et al), and the late 20th century appears somewhat exceptional in the d18O record. It’s worth pointing out that this data got McIntyre’s attention because they were used in a reconstruction by Neukom, not really something one would expect if people were suppressing the data.”

  3. Don Keiller
    Posted May 12, 2012 at 4:32 PM | Permalink

    Steve, “facts” don’t matter to Climate “Scientists” like Gavin.

    Unfortunately what Gavin says will be taken as gospel by his acolytes and more importantly the IPCC and then green-tax raising Governments.

  4. Vorlath
    Posted May 12, 2012 at 5:02 PM | Permalink

    I find it funny that he can’t mention your name without a negative adjective or qualification.

    Here is my summary of gavin’s post using his own words.

    unjustified and false accusations of misconduct and deception, he demeans both himself and his contributions”
    “The idea that scientists should be bullied into doing analyses McIntyre wants”
    “By rights we should be outraged and appalled … once again initiated by Steve McIntyre… and unfailingly and uncritically promoted by the usual supporters”
    “we are no longer shocked nor surprised that misinformation based on nothing but prior assumptions gains an easy toehold on the contrarian blogs”
    “making a few observations that undermine the narrative that McIntyre and company are trying to put out”
    “So McIntyre’s appeals have tried to insinuate
    “unless of course one greatly exaggerates the importance of a single proxy record”
    Worse, McIntyre has claimed”
    “Furthermore, McIntyre is using the fact that Briffa and colleagues responded online to his last deceptive claims about Yamal”
    “The history of this oddity (and it is odd) dates back to McIntyre’s early obsession with a reconstruction called the “Polar Urals””
    “McIntyre got the erroneous idea that studies were being done”
    “This is of course a classic conspiracy theory and one that can’t be easily disproved.” (Gotta have the nutjob angle in there)
    “Why McIntyre thought …[whole bunch of stuff]… remains mysterious.”
    “since apparently university professors have nothing more pressing to do than than respond instantly to McIntyre’s requests”
    “In short, you have to give McIntyre what he wants within 48 hours or he will publicly attack your integrity.”
    “Unsurprisingly, no apology for that unjustified smear has been forthcoming.”
    “McIntyre expressed great exasperation at this point, which is odd” (Again, he really finds you odd. For wanting information in a scientific endeavor. That is odd in of itself, no?)
    “immediately prior to his accusations (again false)”
    “McIntyre’s claim of deception comes from a strained reading of the MR submission ”
    “In response to extended (and yet again false) accusations from Ross McKitrick in the Financial Post:” (Sorry Ross, you’re not getting away either)
    “This is clearly a response to McKitrick’s unjustified accusations, and in using the reference to the 2008 paper is a little contradictory to the paragraphs above”
    “and accusing people of deliberate deception is a huge overreach.”
    “McIntyre conflates the situation at the time of the 2000 paper with the very different situation around 2008 in order to paint a imaginary picture of perfidy.” (Ooohhh, now we have a treasonous act. The highest crime of the land.)
    “McIntyre then quotes an email from Osborn […], but that is again a very strained reading.” (See Steve, you just don’t know how to interpret not just climate data, but also emails).
    “McIntyre would have been (rightly) scornful.” (Ohhh, bring out the claws gavin.)
    “All of [Montford’s] claims [of various nefarious plots] are simply products of his imagination” (Montford doesn’t get a pass either)
    “he imagines even more ‘deceptions’ than McIntyre. Since he assumes the worst of the people involved, everything he sees is twisted to conform to his prior assumptions” (That entire quote could be bolded)
    “As for Watts, … and attempts a twitter campaign of harassment [against Mann]” (gavin can’t resist going after Watts. Surprised he waited until the end).
    “Watts is clearly a cheerleader for the ‘Blame Mike First’ campaign, so maybe his next post will be on why Mike is responsible for the Greek bank default” (double body slam on Watts, or so he thinks)
    “and suggestions that academics have to work on issues according to a timetable dictated by hostile and abusive commentators”

    That’s basically his entire post. Gavin is a source of constant amusement.

  5. bernie1815
    Posted May 12, 2012 at 5:04 PM | Permalink

    Steve:
    I just re-read your Neukom post and Eric Steig’s comments. First, I am not sure Steig conceded anything. You simply got him to acknowledge that the sets of data at issue had not been archived. Second, there is something similar in the tone of Steig’s parting comment/shot that reminded me of Gavin’s recent Yamal post. I would not be surprised if Gavin’s rant is a “team” effort.

    Steve: you’re reading seems right: it’s not that he “conceded” that he’d mis-spoken; he remedied the situation without admitting the point.

  6. Mark McNeil
    Posted May 12, 2012 at 5:04 PM | Permalink

    Make sure you make a copy of the data, a another black hole might find the data.

  7. Posted May 12, 2012 at 5:05 PM | Permalink

    I wondered why he threw in that apparently irrelevant addendum to his poor and illogical argument. Thanks for clearing up the fact that, yet again, RC is into misinformation. When will this charade ever stop? Don’t these people realise that conmsistent, basic dishonesty undermines their “brand”? Sorry for the rhetorical question. No need for answers.

  8. Posted May 12, 2012 at 5:46 PM | Permalink

    At
    http://faculty.washington.du/steig/SipleDomeCores_AB.txt , Steig gives dD (d Deuterium) for Siple Core A for 1000-1993AD, but also d18O data for Siple Core B for 1654-1994. I’ve plotted the latter at http://climateaudit.files.wordpress.com/2012/05/bd18o.png .

    In the d18O series, 1993 is the “warmest” since 1654. This could be the source of Gavin’s claim. However, the last 10 years as a whole aren’t particularly “warm”. 1873 and 1874 are very close 2nd and 3rd place, while 1819 is #4, so it’s hard to claim that the second half of the 20th century is exceptional.

    In fact, Gavin’s reference is not to the dD series per se, but to a paper giving a 700-year reconstruction based on this data by Mayewski, Steig et al in Annals of Glaciology 2004. I haven’t looked at that paper, but perhaps it forms a composite of the two series, which has an all time high in 1993.

    It’s not clear why if there are two cores, Steig doesn’t give dD and d18O for both cores. Also, if one core only goes back to 1654 and the other goes back at least to 1000AD, I would expect a composite either to go back to 1654 only, or else to go back to 1000. Again I haven’t looked at Mayewski04, but it’s odd that they instead go back to about 1300 but no farther.

    Here I’ll attempt to imbed the d18O plot:

    • Steve McIntyre
      Posted May 12, 2012 at 8:18 PM | Permalink

      Hu, it was the dD series that was at issue in connection with Neukom. The O18 series had been archived for a number of years.

      • Posted May 12, 2012 at 9:24 PM | Permalink

        The Mayewski, Steig et al (2004) article cited by Gavin shows only d18O data, back to at least 1200, with no mention of dD. Steig’s new d18O series only goes back to 1654, and looks somewhat different when plotted the same way, with 25 year running means, and so must be a different series. More below.

        Where is the previously arhived d18O data?

  9. Laws of Nature
    Posted May 12, 2012 at 5:52 PM | Permalink

    Hi Steve,

    please continue this investigation, no matter how strenuous this seems.. I think you do an important job!

    My comment over at RealClimate was moved to the borehole, not that it was that important (just a remark on Schmitts reply in comment #1, that the “judgement calls” might be what you are really after rather than some data),
    but I let me realize, that beside all these blocking of comments over at RC (which seems to reach a new peak, showing you hit a mark), there is a change.. they actually to mention your name :)

    Also, whenever a commenter insults or intenionally misspells your name over there ..since it is a heavily moderated site, it really means, that not only Schmitt, but each one of the moderators over there is accepting this kind of behavior as being okay for their blog!
    Just mentioning it here for all the RC people coming here to check the facts.. :)

    All the best,
    LoN

    • Posted May 13, 2012 at 5:47 PM | Permalink

      there is a change.. they actually to mention your name

      Not just Steve’s name, Anthony Watts and Andrew Montford too… plus hyperlinks for all of them… wonders will never cease.

  10. Posted May 12, 2012 at 6:04 PM | Permalink

    In Steve’s plot of Siple A dD vs Yamal above (both smoothed), it appears that the 13th and 20th centuries are the two “warmest” in the record. A composite of Siple A dD and Siple B d18O running back to 1000 AD would then likely find that the same two centuries were comparable.

    Could this be why Mayewski et al, as cited by Gavin, only go back to the 14th c?

    Here is my plot of the raw (unsmoothed) Siple A dD data:

    1881 is clearly the “warmest” of the millenium!

    • Posted May 12, 2012 at 6:09 PM | Permalink

      For some reason my imbed image didn’t work. The URL is

      I’ll try again:

      • Gdn
        Posted May 18, 2012 at 11:07 AM | Permalink

        Hu,

        That section from about 1750 to 1850 really stands out as lacking much variability. I don’t know what that means, but it seems odd.

        • Posted May 20, 2012 at 10:17 AM | Permalink

          Gdn — Agreed. I don’t know what it means either, but my guess is there’s something wrong with those data points.

          I’m working on a plot of the annual and 25-year mean dD together, with CIs on the means.

        • RomanM
          Posted May 20, 2012 at 11:21 AM | Permalink

          Hu, if you plot the first and second differences of the series (diff() and diff(diff()) in R), it really highlights the effect of how the character of the series has changed since the late 1600s:

          Siple B Differences

          [Note: The graph has been updated to correct for some typos]

      • Posted May 21, 2012 at 7:40 AM | Permalink

        Roman —
        Yes, something is really strange about Siple A dD circa 1700-1800. Perhaps Steig will clarify, but unfortunately he (ordinarily) studiously avoids even reading CA.

  11. Posted May 12, 2012 at 9:13 PM | Permalink

    Gavin has posted a plot: http://www.realclimate.org/?comments_popup=11699#comment-235433

    Gavin seems to be trying to soften his original statement, with the word ‘somewhat’ before ‘exceptional’.

  12. Posted May 12, 2012 at 9:46 PM | Permalink

    Figure 2 in the Mayewski article cited by Gavin shows a Siple d18O 25 year running average series since 1200AD that has a pronounced HS, with the blade turning up around 1800.

    However, the Steig data file Gavin links only give d18O for Core B back to 1654. 25 year averages are similar to the Mayewski graph, but are clearly not the same series:

    Circa 1970 shows as the warmest in the last 350 years, though the most recent quarter century is unexceptional. Perhaps Mayewski was using d18O for Core A?

    Similar 25 year running averages of the dD series from Core A show that the early 13th century was the warmest 1/4 century in the last 1000 years (though the 20th c squeaks out the 13c on the basis of 100 year averages.):

    The Mayewski article also shows d18O from the Law dome. This has a clear HS shape, but the blade points down, starting around 1700!

    In fact, Mayewski’s abstract summarizes:

    Late 20th-century West Antarctic [Siple] coastal temperatures are the highest in the record period [generally last 700 years], and East Antarctic [Law Dome] temperatures close to the lowest.

    So Gavin was quoting the “highest temperatures in at least the last 700 years” out of context, and even that was not supported by the Steig data file he linked.

  13. sam
    Posted May 12, 2012 at 11:23 PM | Permalink

    What happened to not posting as much Steve? :-)

    Nah, the world needs you, Godspeed.

  14. Posted May 13, 2012 at 3:48 AM | Permalink

    Following Law of Nature’s comment above, I went to RC and read the “Borehole” posts for the last few days.

    An extraordinary paradox emerges – where the “mainstream” thread is mainly a series of science-free rants, pile-ons and ad homs while the comments in the “rejected” thread generally add up to a serious attempt to define and argue the issue in polite and rational terms.

    Reading the two threads in parallel is an instructive primer in the current state of the climate debate and the ethics of Real Climate’s position.

  15. nono
    Posted May 13, 2012 at 12:21 PM | Permalink

    Steve, what is your rebuttal of this?

    “””
    Another little appreciated fact: When McIntyre started to get interested in this, he asked Briffa for the underlying measurement data from Yamal and two other locations whose reconstructions were used in Osborn and Briffa (2006). In May 2006, Briffa politely replied:

    Steve these data were produced by Swedish and Russian colleagues – will pass on your message to them
    cheers, Keith

    Briffa was conforming to the standard protocol that directs people to the originators of data series for access to the underlying data, as opposed to the reconstructions which had been archived with the paper. McIntyre expressed great exasperation at this point, which is odd because in email 1548, McIntyre is quoted (from Sep 26, 2009 (and note the divergence in post URL and actual title)):

    A few days ago, I became aware that the long-sought Yamal measurement data url had materialized at Briffa’s website – after many years of effort on my part and nearly 10 years after its original use in Briffa (2000).

    To which Rashit Hanterminov responds:

    Steve has an amnesia. I had sent him these data at February 2, 2004 on his demand.

    Thus at the time McIntyre was haranguing Briffa and Osborn, McIntyre had actually had the raw Yamal data for over 2 years (again, unmentioned on Climate Audit), and he had had them for over 5 years when he declared that he had finally got them in 2009 (immediately prior to his accusations (again false) against Briffa of inappropriate selection of trees in his Yamal chronology).
    “””

    • Kenneth Fritsch
      Posted May 13, 2012 at 2:56 PM | Permalink

      Nono, no, no.

      You got your facts wrong here. Briffa and Russians were never specific on identifying the proxy data as one and the same. It was left for later developments for SteveM to determine that the proxies were one and the same.

      Beter to the point, nono, what do you think of SteveMs sensitivity test dealing with Yamal and proxy selections.

      • nono
        Posted May 14, 2012 at 1:54 PM | Permalink

        I find this hard to follow. The Russians sent the data Steve asked Briffa, but it was not clear these were one and the same?

    • John M
      Posted May 13, 2012 at 3:11 PM | Permalink

      My comments have been diasppearing, but let me try responding to nono.

      That issues has been discussed over at Bishop Hill.

      http://www.bishop-hill.net/blog/2012/5/11/realclimate-on-yamal.html

      The comments in particularly provide more information, since other RC acolytes were trying a similar line of inquiry (small world).

      Apparently, because Briffa refused to provide the list of sites used in the regional chronology, it wasn’t obvious in 2006 that it was the same (actually, a subset apparently) data.

  16. Paul Penrose
    Posted May 13, 2012 at 12:41 PM | Permalink

    Maybe Gavin has a different meaning for the word “exceptional”, but I sure didn’t see anything using the accepted definition of the word.

  17. Posted May 13, 2012 at 12:57 PM | Permalink

    The peak year for the dD data in the period from 1840 to 1994 was 1881. I thought I would look at CRUTEM3 Southern Hemisphere to see if 1881 was exceptionally warm. And there is some confirmation of ~1881 being really, really warm. These are the ones that jumped out at me (graphs are at my blog).

    The 2nd warmest August in the Southern Hemisphere was 1884.

    The 3rd warmest December was 1881. (First was 1850, 2nd was 1977)

    The 3rd warmest September was 1878.

    The 5th warmest July was 1877.

    The 2nd warmest June was 1873.

    http://sunshinehours.wordpress.com/2012/05/13/steigs-hockey-stick-and-crutem3-southern-hemisphere/

  18. pdtillman
    Posted May 13, 2012 at 1:36 PM | Permalink

    Steve, happy to see you active again. It certainly appears that Gavin has caught some of Jim Hansen’s semi-religious CAGW activism. At least he hasn’t been arrested!

    I really think the house of cards these fellows have built is tottering. Your pushes help.

    Thanks much, Peter D. Tillman

  19. Posted May 14, 2012 at 7:04 AM | Permalink

    “late 20th century appears somewhat exceptional in the d18O record”, but the filter is non-causal and the end-points are often exceptional:

    • Posted May 14, 2012 at 8:35 AM | Permalink

      UC — It’s great to see you back and in good form!

      Gavin’s LOESS (local polynomial) filter has a symmetrical effect away from the endpoints, but near the ends it starts to get asymmetrical, and puts a lot of weight on the last point(s). This isn’t quite as bad as Mann’s double flip end-pegging procedure, but as you show it likewise tends to make the last point look extreme — until it is dampened by the availability of subsequent observations.

      • Steve McIntyre
        Posted May 14, 2012 at 10:35 AM | Permalink

        UC, nice to hear from you as well.

        Schmidt’s introduction of the O18 series into the discussion is a diversion. It was the dD data that was in discussion and archived only this year after CA commentary.

        • Posted May 14, 2012 at 11:05 AM | Permalink

          True, but the Mayewski paper Gavin cited really does show a good Siple d18O HS, even if the dD data does not (even when enhanced at the end with a LOESS filter). So I think it’s interesting to find out why it doesn’t match d18O data Steig has now put on line, even over its the post-1654 period.

          Examination of the record in question actually demonstrates that it has exceptionally high values in the late 20th Century (reflecting the highest temperatures in at least the last 700 years, Mayewski et al.), exactly counter to McIntyre’s theory.

          The “Steig HS” Gavin had in mind is clearly the d18O one in Mayewski, Steig et al. Maybe there’s data for Siple Core A d18O archived somewhere that I just didn’t look. Perhaps the Mayewski paper plots it, or perhaps the average of both cores.

        • Posted May 14, 2012 at 11:56 AM | Permalink

          This paper: http://faculty.washington.edu/steig/papers/recent/SchneiderSteigGRL2006.pdf

          Fig 3 shows a reconstruction with a massive decline in temperature starting about 1990. But the peak was pretty similar to ones at 1940 and 1825.

          “Other
          evidence for long-term warming includes ice borehole
          temperature inversions at Siple Dome [Engelhardt, 2004]
          in West Antarctica and Law Dome [Dahl-Jensen et al.,
          1999] in East Antarctica. At Law Dome, the minimum
          (1850) to maximum (1990) temperature difference was
          quantified as 0.7C, which agrees with the difference of
          0.6C that is evident in a smoothed version of our reconstruction
          between the low in the 1800s to the maximum
          around 1990 (Figure 3c). Some lower resolution isotopic
          records, at Siple Dome [Mayewski et al., 2004] and Dronning
          Maud Land [Isaksson et al., 1996], also suggest 20thcentury
          warming.”

      • Posted May 16, 2012 at 1:54 PM | Permalink

        This isn’t quite as bad as Mann’s double flip end-pegging procedure, but as you show it likewise tends to make the last point look extreme — until it is dampened by the availability of subsequent observations.

        Yes, and you can get lots of ‘exceptional’ results. Assume that there is no data after 1820. Then the del18O makes a clear record (warmest ever):

        sd1
        If there is no data after 1850, the del18O is a record (coldest ever), but delD makes a record as well (warmest ever):

        SD2

        Video colors are odd with some browsers, using full screen helps.

        • Steve McIntyre
          Posted May 16, 2012 at 3:37 PM | Permalink

          UC, what f-value did Steig use in his lowess?

          Also, Siple Dome A (not B) was used for O18 in Mayewski et al 2004. It has a very very low last value, which can affect lowess smooths with low f-values.

        • Posted May 16, 2012 at 3:56 PM | Permalink

          Steve —
          It was Gavin, not Steig, who used a lowess smoother. He said the span is “~30 years”, which I guess means that it uses the nearest 30 years to fit a polynomial. These would be centered except at the ends, where its asymmetry enables it to reach for end-values. He didn’t say what degree of polynomial he was fitting, but a webpage I visited said that quadratic is common. Lowess typically also weights the observations by distance, so that’s another factor that would need to be specified to get an exact match.

          The Mayewski-Steig paper Gavin cited just used a 25 year MA, which isn’t as fancy, but at least is easy to replicate.

        • Brandon Shollenberger
          Posted May 16, 2012 at 5:01 PM | Permalink

          It’s worth pointing out Gavin said he used LOESS, not LOWESS. The two are related, but there are a number of differences. Not the least of which is the fact LOWESS uses a linear polynomial while LOESS (typically) uses a quadratic. As for weighting, the standard weighting in both is tri-cube weighting, so that is almost certainly what Gavin used.

          As for his f value (in LOESS, it’s actually called span), there is a thousand years of annual data. This means to get a smoothing window of ~30 years, Gavin would have had to pick an f value of ~0.03.

          Steve: as I pointed out at Lucia’s, the Siple Dome A O18 data with a lowess smooth with low f-values goes down at the end. Doubtless overinfluenced by a a downspike, but nonetheless the result. Siple B is an odd data set for Gavin’s graph. There’s no published data scale (that I can locate anyway.)

        • Brandon Shollenberger
          Posted May 16, 2012 at 6:20 PM | Permalink

          Steve, that’s true, but because LOESS fits a quadratic polynomial instead of a linear polynomial like LOWESS does, it can get dramatically different results even if the same smoothing window is used. This means you cannot directly compare the results you showed to the results Gavin showed.

          Then again, when a smooth gives a point of inflection for every ten data points (in very noisy data, no less), you’re overfitting so much I wouldn’t put any stock in the graphs generated (by you or Gavin).

        • Posted May 18, 2012 at 1:24 PM | Permalink

          Matlab smooth(data,28,’loess’) gives quite good fit, but not exact

          Anyway, if something is declared unprecedented, smoothing effects near the end-points should be explained well. With error bars or bootstrapping etc.

        • Posted May 19, 2012 at 12:41 PM | Permalink

          UC —
          Thanks for pointing me to Matlab smooth(x,span,’loess’), where loess indicates quadratic. It turns out that if span is even, it sets the effective span to span-1, so as to make a symmetric centered filter. So if Gavin set span=30 (and if he was using Matlab), he should have gotten the same results with span=29. Does 28 (= 27) replicate his results better than 29 or 30?
          In any event, with span = 29, the weights away from the end points are symmetrical and negative near each end, so that the quadratic smoother reaches up into any concavity or down into any convexity in x. The weight on the center point is about 11. But near the end points they get asymmetrical, and at the very ends, the endpoint gets weight .35. This isn’t as bad as endpegging, but it does exaggerate the end behavior. “lowess”, which does linear smoothing instead, only places a .19 weight on the endpoints (and .06 on the center points away fom the ends) with span = 29.
          Since the weights start to change near the ends, the first and last (span-1)/2 smoothed points should either be discarded, or else plotted with a different type of line (eg thinner or dashed), to emphasize that they are computed differently.
          Although smooth(x,span,’loess’) “uses” span points, in fact the tricubic (?) weighting system gives 0 weight to the farthest points, so that it actually only assigns nonzero weight to span-2 points when it is more than (span-1)/2 points from the ends. But near the ends, it “uses” the nearest span points, stretching the weights out to “include” all of these points, and so in fact assigns nonzero weight to span-1 points.
          Oddly, Matlab smooth(x,span), which computes a centered MA(span) away from the ends, behaves differently near the ends. When it is closer than (span+1)/2 from the ends, it simply reduces the effective span to just reach the endpoint. Thus at the endpoints, it actually fits an MA(1) that exactly endpegs the first and last observations. (Matlab filter( ) will also compute a (lagging) MA, but it instead prepads with zeros.)

        • RomanM
          Posted May 19, 2012 at 2:41 PM | Permalink

          Hu, this web page seems to indicate the span parameter is a value between 0 and 1 when the loess options are used. This would be in line with the equivalent function loess in R except that R also allows values of span greater than 1.

          The integer-valued spans seem to be used with other method options.

        • Brandon Shollenberger
          Posted May 19, 2012 at 6:34 PM | Permalink

          Hu McCulloch, it’s important to remember the span value is a percentage. If you’re using a thousand years of data, 30% would be 300 years. To get 30 years, you’d want 3%, or a span of .03.

        • Posted May 20, 2012 at 8:24 AM | Permalink

          Roman and Brandon —
          Matlab smooth(x, span, ‘loess’) (or ‘lowess’) interprets span as a fraction of the total length of x if it is less than 1, but takes it as the actual span if it is an integer greater than 1, so you can use it either way. The fact that Gavin gave his span as “~30 yrs” suggests that perhaps he input a fraction that would give approximately this value.

          “rloess” and “rlowess” are “robust” options that omit values that would be greater than 6 standard deviations from the curve. This would be relevant for the last observation (1995) in the long-archived d18O data, since for some reason it came out extremely “cold”. Although Mayewski was just using a 25 yr moving average, he may have done something like this, since the end of his smoothed series is not responding to this extreme last point.

        • Brandon Shollenberger
          Posted May 20, 2012 at 10:41 AM | Permalink

          Hu McCulloch, are you sure about that? If so, I think it is different in R. Or maybe I’m just misremembering because I only used values lower than one.

        • Posted May 20, 2012 at 11:01 AM | Permalink

          Brandon —
          I tried Matlab smooth with span = 7 and 29, and it worked. It also works with span as a fraction.

  20. Posted May 14, 2012 at 9:23 AM | Permalink

    In Eric Steig’s 2004 article with Mayewski et al that Gavin cites, Figure 2 shows a pretty good HS for “Siple d18O” from 1200AD to the present, even without the end-point exaggerating LOESS filter that UC (aka uc00) analyzes above. The core it is based on is not identified.

    However, this is not the data Steig has recently put on line and which Gavin plotted, since the new d18O data for Core B only goes back to 1654, and since the data for Core A going back to 1000 AD is dD, not d18O.

    Over on RC, (reply to comment #81), Steig says that all his Siple data either is newly online or has been online for over 10 years:

    Why the heck would I have hidden these data when the are actually identical to those on line (check out the low resolution Siple Dome A core data, easily found on line and available for at least the last decade, and compare them with the data McI is complaining about), and when in fact they do show that the 20th century was more elevated than most other decades in the last 1000 years, if ‘suppressing non-hockey stick graphs’ were the goal?

    So perhaps the Mayewski HS series actually is online somewhere.

    I tried ftp://ftp.ncdc.noaa.gov/pub/data/paleo/icecore/antarctica , but with no success. There is a Siple subdirectory there with a file siple2005dd.txt, but it only has data for 8.63-57.11 kyr BP.

    In the antarctica folder itself, there is a file antarctica2011iso.tx that has Siple (dD-10)/8 and normalized Siple D18O for 9000-21000 BP1950 plus other data for Byrd and Law Dome in 20 year steps both by age and by depth with age. (Does anyone know why one would tabulate (dD-10)/8 rather than dD itself?)

    File west-antarctica-iso2008.txt uses Siple Dome A as well as Siple Station (which is very far from Siple Dome and not to be confused with it), plus 6 other sites. However, it only shows the average of the 8 cores, and then only for 1900-2001AD. This average was evidently reported in Schneider and Steig, PNAS 2008.

    File antarctica-temp2006.txt gives output data plus some input data for a 200-year reconstruction by Schneider, Steig et al GRL 2006. The reconstruction used Siple Station (not Siple Dome) data back 1800AD, but merely says this record is “available from Dr. Ellen Mosley-Thompson of the Byrd Polar Research Center at The Ohio State University.”

    http://www.faculty.washington.edu/steig/sipledome/SipleDomeCores_AB.txt links to http://icereader.org/icereader/, but the data there says it is “only for viewing” and not for use. It in turn links to http://nsidc.org/data/waiscores/iso.html, but the link to Siple dome 1994 just gives metadata and no isotope data. Under WAISCORES, it has alink to Siple B, but this only gives ions, and no isotopes. It mentions White stable isotopes for Siple A and B, but says these are available to WAISCORES PIs only.

    Nevertheless, the nsidc site links to ftp://sidads.colorado.edu/pub/DATASETS/AGDE/waiscores/white/SDM-A_all_half_20meter.txt, which gives Siple Core A mu18O (= d18O)? in 1/2 meter steps down to 1km with no dates. The file states that higher resolution data exists for the top 393m but doesn’t give this. File SDM-B20Core.txt gives “preliminary” (and therefore unofficial) Siple Core B “muD” and “mu18O” in 5 cm steps down to 100 m, but with no inferred dates.

    Maybe I’m just not looking in the right place, but I can’t find the Mayewski2004 d18O data back to 1200, with the HS Gavin originally called attention to, anywhere, despite Steig’s claim that it is online.

    • Posted May 14, 2012 at 10:54 AM | Permalink

      Correction — the URL for the White Colorado WAISCORES data is
      ftp://sidads.colorado.edu/pub/DATASETS/AGDC/waiscores/white/SDM-A_all_half_20meter.txt

    • Steve McIntyre
      Posted May 14, 2012 at 11:44 AM | Permalink

      Re: Hu McCulloch (May 14 09:23),

      An age-scale for SIple is located in the script below.

      
         #siple age: core A 
      	loc="ftp://sidads.colorado.edu/pub/DATASETS/AGDC/waiscores/timescales/SDMA_0-514_Nov2003.txt"
      	download.file(loc,"temp")
      	age=read.csv("temp",skip=18,header=FALSE)
      	names(age)=c("depth","bp")
      	apply(age,2,range)
      #	       depth   bp
      #[1,]   0.270  -45
      # [2,] 514.764 8609
      	siplef=approxfun(age$depth,age$bp)
      
      	loc="ftp://sidads.colorado.edu/pub/DATASETS/AGDC/waiscores/white/SDM-A_all_half_20meter.txt"
      	download.file(loc,"temp")
      	work=read.table("temp",skip=8,sep="\t",fill=TRUE)
      	names(work)=c("top","bottom","O18") #depth in meters
      	work$mean=apply(work[,1:2],1,mean)
      	apply(work,2,range,na.rm=T)
      	  #         top   bottom    O18     mean
      	#[1,]    0.000    0.500 -38.77    0.255
      	# [2,] 1003.604 1003.829 -23.17 1003.716
      	work$bp=siplef(work$mean)
      	work$bp[1]=1994
      
      • Posted May 14, 2012 at 12:12 PM | Permalink

        It looks like you’ve generated the Siple A d18O series by age as above at http://climateaudit.org/2012/01/19/neukom-and-the-steig-overunder/#comment-322735 , so this may answer my question. However, the reason it doesn’t agree with the new d18O series is that the new series is for Core B. I’ll try plotting 25-year means tonight to see if they match Mayewski.

        Steve: Hmmm. I haven’t seen an age-scale for Core B. I guess that it’s in the “not archived” archive.

    • Steve McIntyre
      Posted May 14, 2012 at 11:48 AM | Permalink

      Hu, the Mayewski paper shows a Law Dome series which has a very elevated medieval O18. I’ve spent lots of time on this series over the years, some of which are recorded in Climategate emails. There’s an amusing exchange where I ask IPCC to show this proxy in AR4. They knew that it had an elevated MWP and refused.

      • Posted May 16, 2012 at 10:12 AM | Permalink

        While it is true that the Mayewski plot of the Law Dome d18O has an inverted HS shape (flat up to 1700 and then trending distinctly downwards to the present), the scale is twice as large as that on the accompanying Siple A d18O graph, so that on the same scale it would have only half the magnitude of the upward HS in the Siple A d18O data.

        But still, AR4 should have included it.

        I’ve finally gotten graphs of Siple A d18O to work and will post them soon for comparison to Steig’s new Siple A dD and Siple B d18O data.

    • Kenneth Fritsch
      Posted May 20, 2012 at 11:24 AM | Permalink

      Hu, I really appreciate seeing all these various proxy series graphed individually and particularly when the proxies originate from a proximate location. I have attempted to track the discussions on this thread and which proxy series is which. It is my impression that the differences we see in the individual proxies and particularly the most recent times are real as opposed to being a matter of processing the proxy data or presentation. Is it possible to provide links to the individual proxy data and/or show the raw series data (times series of O18 or D2) together in a graph (not spaghetti but in single graphs one atop the other)?

      • Posted May 20, 2012 at 2:24 PM | Permalink

        Ken —
        I’ve tried to do that in the graphs below that I plotted earlier. If you search down, there’s a plot of Dome A and B d18O superimposed, and then in the next comment (it was supposed to be in the same comment but I got the image links wrong), there’s Dome A d18O along with dD/8-2 (scaled to make the vertical scales roughly equal per the 8:1 rule of thumb that is often used and shifted to separate them a little).

        The raw data for Dome A d18O and the timescale are in separate files at ftp://sidads.colorado.edu/pub/DATASETS/AGDC/waiscores/white/SDM-A_all_half_20meter.txt and ftp://sidads.colorado.edu/pub/DATASETS/AGDC/waiscores/timescales/SDMA_0-514_Nov2003.txt

        Steig’s new Dome A dD and Dome B d18O series are at http://faculty.washington.du/steig/SipleDomeCores_AB.txt .

        Eric also sent me an unarchived file that has Dome A d18O directly by date, which he gave me permission to circulate. Steve has put it online at http://www.climateaudit.info/data/ice/siple_dome_steig_20120515.txt . I haven’t had time to plot it out yet, but I’m assuming it looks very much like the archived series, except perhaps for a small timing discrepancy that Eric is concerned about.

        • Kenneth Fritsch
          Posted May 20, 2012 at 7:57 PM | Permalink

          Hu, I still have to sort this out since you were working with grsphs in real time on this thread. The link to Steig’s new Dome A dD and Dome B d18O series did not work for me.

          What is most surprising to me is that I could not go to Steig’s website or any website for that matter and obtain a time series of these O18 or H2 proxies. Did you and SteveM have to covert the depth data to time before displaying the graphs?

        • Posted May 20, 2012 at 9:02 PM | Permalink

          Ken —
          My bad — I got Steig’s URL wrong. It should be

          http://faculty.washington.edu/steig/sipledome/SipleDomeCores_AB.txt

          Yes, Steve and I used both archived files to construct d18O by age. Steve’s code is above at http://climateaudit.org/2012/05/12/steigs-hockey-stick/#comment-333554 . I’m not sufficiently conversant with R to see what he’s doing, but I simply assigned each year the average d18O value of the half-meter it fell in. Not super precise, but adequate for present purposes.

        • Kenneth Fritsch
          Posted May 20, 2012 at 9:22 PM | Permalink

          Hu, before I read your last post I went back to the original CA discussion and got the same link you have. Steig had provided it in the original thread. I did my calculations and plots in R and the link below shows the graphs. I think our plots look reasonably close.

          Now I need to go back and find the data SteveM used for his plots which are from data from a different core I am assuming. I will then add these proxies to a file I have of closely located proxies.

        • Posted May 21, 2012 at 7:54 AM | Permalink

          Ken —
          It looks like we’re on the same page with regard to Siple Dome Core A dD and Core B d18O.
          Steve did earlier plot the archived data for Core A d18O as the red line in

          http://climateaudit.org/2012/01/19/neukom-and-the-steig-overunder/#comment-322735

          However, it’s hard to see it with the superimposed black line from Core B. By itself, it does show a fair HS, as noted by Gavin and plotted below, but unlike the only now released dD data.

        • Kenneth Fritsch
          Posted May 21, 2012 at 10:25 AM | Permalink

          Below is the R code used for calculating the trends from 1900-1994 and 1956-1994 for D2TS (del D Series) and O18TS (del O18 Series) and the results. Note the series are from the above linked plots that I made and a core designation that remains to be identified and differentiated from other core data shown here. The series used here are the annual series and not those with a 25 year centered MA.

          The important points are that, using the starting date of 1956 which coincides with the Steig (08) Antarctica article in Nature starting point, these data shows no significant trend – although del D series almost shows a significant negative trend. Those trends are counter to those found in Steig (08).

          The trends for del D and del O18 show no significant trend for the period 1900 to 1994. Those findings are counter to global trends.

          I do not know whether the authors of these del O18 and del D series from ice cores, reported here and elsewhere, make any specific claims about the origins of the water found in the precipitated snow analyzed. One could visualize that water coming from considerable distances away and thus making the temperature correlation less local and more regional.

          W1=window(D2TS,start=1900,end=1994)
          W2= window(O18TS,start=1900,end=1994)
          W3= window(D2TS,start=1956,end=1994)
          W4= window(O18TS,start=1956,end=1994)
          Lm1=lm(W1~time(W1))
          Lm2=lm(W2~time(W2))
          Lm3=lm(W3~time(W3))
          Lm4=lm(W4~time(W4))
          summary(Lm1)
          summary(Lm2)
          summary(Lm3)
          summary(Lm4)

          Coefficients:
          Estimate Std. Error t value Pr(>|t|)
          (Intercept) -1.840e+02 7.369e+01 -2.497 0.0143 *
          time(W1) -8.227e-03 3.785e-02 -0.217 0.8284

          Coefficients:
          Estimate Std. Error t value Pr(>|t|)
          (Intercept) -35.686366 9.887451 -3.609 0.000497 ***
          time(W2) 0.005037 0.005078 0.992 0.323751

          Coefficients:
          Estimate Std. Error t value Pr(>|t|)
          (Intercept) 312.4357 272.0614 1.148 0.2582
          time(W3) -0.2589 0.1378 -1.879 0.0681 .

          Coefficients:
          Estimate Std. Error t value Pr(>|t|)
          (Intercept) 1.47862 43.44788 0.034 0.973
          time(W4) -0.01366 0.02200 -0.621 0.538

        • Kenneth Fritsch
          Posted May 21, 2012 at 10:55 AM | Permalink

          Hu, I am of notion after comparing all the series here that we have Steig’s old data for del D and del O18 and then the new data that he provided you in the link and comment here:

          “Eric also sent me an unarchived file that has Dome A d18O directly by date, which he gave me permission to circulate. Steve has put it online at

          http://www.climateaudit.info/data/ice/siple_dome_steig_20120515.txt .

          I haven’t had time to plot it out yet, but I’m assuming it looks very much like the archived series, except perhaps for a small timing discrepancy that Eric is concerned about.”

          It would appear that these are the only Steig del D del O18 series we have and there is no third set. Is that correc?

          I am going to plot the “new” series from the link above.

        • Kenneth Fritsch
          Posted May 21, 2012 at 12:31 PM | Permalink

          I have plotted the “new” Steig Siple Dome delD and delO18 series back to year 1000 and linked them below. The new series agree very (too?) well with one another. I thought I might have made a mistake in downloading or selecting data to obtain that exceptional agreement but on checking I did not find any problem.

          While both series indicate an elevated plateau late in the series, the calculated trends indicate no significant trends from 1850 to 1994 and significant negative trends from 1956 to 1994. The results are summarized directly below. Not much here to support AGW or Steig (09). Note in the post above I incorrectly referred to this Nature article coauthored by Eric Steig as Steig(08).

          Period 1850 to 1994:

          delD Slope=-0.0046 SE Slope=0.0271 Prob= 0.86
          delO18 Slope=-0.00055 SE Slope=0.0034 Prob=0.87

          Period 1956 to 1994:

          delD Slope=-0.3637 SE Slope=0.155 Prob= 0.029
          delO18 Slope=-0.0458 SE Slope=0.0193 Prob=0.027

        • Steve McIntyre
          Posted May 21, 2012 at 1:06 PM | Permalink

          Hmmm, if we does a regression of delD against del18O in the “new” Steig data, the r-squared is 0.9999. While there is a relationship between the two, I don’t recall seeing this sort of relationship in other data sets: I’ll check sometime.

          A=read.table("http://www.climateaudit.info/data/ice/siple_dome_steig_20120515.txt",sep="\t",header=TRUE)
          	A$depth=apply(A[,1:2],1,mean,na.rm=T)
          	A$year=1950-siplef(A$depth)
          	A=A[A$year>=1000,]
          	fm=lm(delD~del18O,A)
          	fm1=update(fm,delD~del18O-1)
          #Multiple R-squared: 0.9999,     Adjusted R-squared: 0.9999 
          #F-statistic: 4.583e+06 on 1 and 246 DF,  p-value: < 2.2e-16 
          
        • Posted May 21, 2012 at 4:00 PM | Permalink

          Ken —
          It looks like Steig slipped up somehow when he generated the new unofficial dD/d18O by date file at http://www.climateaudit.info/data/ice/siple_dome_steig_20120515.txt . The d18O matches the old waiscore files well enough, but from your graph it looks like the dD series is just a linear transform of the d18O. We should stick instead with the semi-official (not exactly archived, but on Steig’s UW server) dD data at http://faculty.washington.edu/steig/sipledome/SipleDomeCores_AB.txt .
          I’ll try checking it out myself and then e-mail him.

        • Kenneth Fritsch
          Posted May 21, 2012 at 4:36 PM | Permalink

          Hu, just so I can keep these cores/proxies straight I am assuming the first link in your post above is to the “new” core data and the second link is to the “old” core data. Obviously the first del O18 goes back only to 1600 something while the second goes back past 82000 BC for both del D and del O18.

          I also have to look up the article on ice core sampling in the Antarctica that Steig coauthored before Steig (09). I was pouzzled since it was not references in Steig (09). The core finds as I recall did not agree with Steig (09) and showed a cyclical climate pattern. I do not know at this point if those core(s) are different than those we are currently looking plotting.

        • Kenneth Fritsch
          Posted May 22, 2012 at 8:38 AM | Permalink

          In the links below I have listed the cores that were used in composite representations of the individual cores and in turn appear in plots in the articles. Both papers are coauthored by Eric Steig before 2009 and only the second linked paper about 1940s Antarctica warmth paper is referenced in Steig (09). I am wondering who exactly has the data from these cores.

          The composites found in these two linked articles produce different responses in the 1960- forward periods. In the second linked article below the composite shows the unprecedented warming in the 1940s in the Antarctica and a correlation with the warming in the regional SSTs. The 1960-forward part of the composite shows a rather straight trajectory of an upward trend.

          In the first linked article below the 1960-forward part of the composite shows a trending up into the 1980s and then downward with no overall trend appearing – on eyeballing the plot. The plot shows the 1940s warming but not to the extent of the plot in the second linked article. This first linked article shows the undulating and cyclically appearing nature of the proxy response over the last 200 years and shows why selecting endpoints for a trend over this period could affect greatly the trend produced.

          If all these ice core data were all available a more interesting comparison of the individual cores could be made.

          http://faculty.washington.edu/steig/papers/recent/SchneiderSteigGRL2006.pdf

          Antarctic temperatures over the past two centuries from ice cores, by
          David P. Schneider, Eric J. Steig, Tas D. van Ommen, Daniel A. Dixon,
          Paul A. Mayewski, Julie M. Jones, and Cecilia M. Bitz

          Law Dome 2000 (stack)
          Siple Station
          Dronning Maud Land (several sites)
          US ITASE 2000-1
          US ITASE 2000-5

          http://www.pnas.org/content/105/34/12154.full

          Ice cores record significant 1940s Antarctic warmth related to tropical climate variability, by David P. Schneider and Eric J. Steig

          Dyer Plateau
          Siple Station
          01–5 ITASE
          00–5 ITASE
          01–2 ITASE
          01–3 ITASE
          00–1 ITASE
          Siple Dome A core

        • Kenneth Fritsch
          Posted May 22, 2012 at 8:46 AM | Permalink

          Looking at the “new” Steig Siple Dome ice cores I have noticed that the year’s corresponding to the del O18 and del D measurements are not in whole years but rather a year plus a fraction and the years are not consecutive. I was under the impression that ice cores were sampled by years at least in the most recent period before compaction obliterated annual resolution.

        • Kenneth Fritsch
          Posted May 22, 2012 at 9:24 PM | Permalink

          I found the data for the ice cores used in the 2 Steig authored papers above and will link the data and plots of the individual results as soon as I have time.

          Hu and Steve, I am thinking that the “new” Steig data where we obtained an unbelievable correlation between del O18 and del D was actually meant to be either only a del D or del O18 data and either the del O18 or del D was meant to be used in a standardized graph were Del D and del O18 data were both used.

        • Kenneth Fritsch
          Posted May 26, 2012 at 10:31 AM | Permalink

          I am posting here what I said I would in my previous post by showing all the composite and individual ice cores for the Antarctica that have stable isotope measurements with annual resolution to near present time. The link directly below list ice core data where specifically data from the Antarctica can be retrieved. What I quickly found on searching the list was that many of these Antarctica ice cores were either not measuring stable isotopes (used as temperature proxies) or stable isotopes measurements for time periods ending too far back in time for my purposes here. What I did found were 3 composites and 7 individual ice cores. In my plots I also included the 2 “old Steig” and the 1 “new Steig” Siple Dome ice cores. I am assuming that although the new Steig ice core data showed both del D and del O18 measurements that only one was unique (del D) and the other was correlated for producing an del D and del O18 intermixed plot.

          http://hurricane.ncdc.noaa.gov/pls/paleox/f?p=517:1:859780108795203:::APP:PROXYDATASETLIST:7:

          Dyer Plateau −70.66 −64.5 2,002 del O18 1989
          Siple Station −75.92 −84.15 1,054 del O18 1983
          01–5 ITASE −77.06 −89.14 1,239 del O18 1999
          00–5 ITASE −77.68 −123.99 1,828 del D 1999
          01–2 ITASE −77.84 −102.91 1,336 del O18 1999
          01–3 ITASE −78.12 −95.65 1,620 del D 1999
          00–1 ITASE −79.38 −111.23 1,791 del D 1999
          Siple Dome A core −81.65 −148.81 615 del D 1993

          Of the eight cores in the table above 5 were individual measurements and 3 (Siple Station and Dome and Dyer Plateau) were a composite. The 5 ITASE ice cores are also plotted together as a composite as are all 8 together. These 8 cores were used in the Steig paper showing an unprecedented 1940s warming and is linked above. Why the data for the composite 3 ice cores were not listed separately is beyond me at this point. The data for these 5 individual and the composite of 3 cores were found in NOAA link are specifically at the link:

          ftp://ftp.ncdc.noaa.gov/pub/data/paleo/icecore/antarctica/west-antarctic-iso2008.txt

          The 2 other individual cores found at the NOAA site are listed below with the link to the data. The 3 cores in the new and old Steig data are not listed at the NOAA site and the data were found at the sites noted above in this thread.

          Gomez Ice Core d18O Data
          Gomez, SW Antarctic Peninsula: 73.59°S, 70.36°W, 1400 m

          ftp://ftp.ncdc.noaa.gov/pub/data/paleo/icecore/antarctica/gomez/gomez2009d18o.txt

          Talos Dome Ice Core Deuterium Data
          TD core: 72°48´S; 159°06´E; 2316 m. elev.

          ftp://ftp.ncdc.noaa.gov/pub/data/paleo/icecore/antarctica/talos/talos_dd.txt

          I have plotted (1900 to approximately 2000) the 3 composites in the first link below , the 5 ITASE cores in the second link and Gomez, Talos and the 3 Steig cores in the third link. While some of the proxy series structure is similar there are definite differences. It is easy to see differences in the 1940s temperature relative to the series and the differing trends in the series. The trends and AR1s were calculated for the two periods of time. The first was based on the time period from 1900 to 1995 and the second from 1956 to 1995. The trends were tested for significance at p=0.05 with the formula by Quenouille for AR1 adjustment for degrees of freedom.

          1900 to 1995:

          Old Steig del O18: Positive trend not significant
          Old Steig del D: Negative trend not significant
          New Steig del D: Negative trend not significant
          ITASE 00-1 del O18: Positive trend not significant
          ITASE 00-5 del D: Negative trend not significant
          ITASE 01-2 del O18: Negative trend not significant
          ITASE 01-3 del D: Positive trend not significant
          ITASE 01-5 del O18: Positive trend significant
          Gomez del O18: Positive trend significant
          Talos del D: Negative trend not significant
          Composite of Siple Dome, Siple Station and Dyer Plateau: Positive trend not significant

          1956 to 1995:

          Old Steig del O18: Negative trend not significant
          Old Steig del D: Negative trend not significant
          New Steig del D: Negative trend significant
          ITASE 00-1 del O18: Positive trend significant
          ITASE 00-5 del D: Negative trend not significant
          ITASE 01-2 del O18: Positive trend not significant
          ITASE 01-3 del D: Positive trend not significant
          ITASE 01-5 del O18: Positive trend significant
          Gomez del O18: Positive trend significant
          Talos del D: Negative trend not significant
          Composite of Siple Dome, Siple Station and Dyer Plateau: Positive trend significant

          I see differences in trends from one series to another even though the individual trend series do not always have significant trends. Del D tends to produce more negative trends while del O18 produces more positive ones. My question remains that if we cannot know with any certainty how localized the water was that evaporated originally (producing the del factor for the stable isotopes) and subsequently fell in the ice core area we cannot know how different these ice core proxies would be if the effect was related to local temperatures only. A second question is how different would we expect these Antarctica regional temperatures to be, i.e. are we looking at local variability or differences in proxy response for near same temperature changes. We know that the Peninsula has warmed faster than the remainder of Antarctica and East Antarctica has probably little of no warming over the periods measured for trends here. Most of the proxies here are from West Antarctica near the Peninsula and/or Southern Ocean and one in the Peninsula.

          The 8 ice cores from the Steig paper, showing a warm 1940s decade as a composite, consist of 7 core locations spread out over West Antarctica and near the Southern Ocean and one core from the Peninsula (Dyer Plateau). The Steig paper uses a composite of all 8 and indicates an association with the SST of the Southern Ocean. I am not at all sure why one would simply show a composite and not the individual ice core measurements. We have Siple Dome as an individual core(s) from Steig in previously presented (old) and recently presented (new) cores but we do not have Dyer Plateau and Siple Station as individual cores. Dyer Plateau would be of interest since it is on the fast warming Peninsula. Gomez core (not in the Steig paper) is on the SW Peninsula and it does show significant warming in both the 1900-1995 and 1956-1995 periods. Gomez actually extends to 2005 and shows significant warming to there from both 1900 and 1956. I wonder if the Dyer Plateau is overwhelming the effects of Siple Dome and Siple Station in Steig’s composite of 3. The Talos ice core is located, I guess, technically in East Antarctica, but is near both West Antarctica and the Southern Ocean. It shows cooling in both periods measured, but the data from that core is so noisy that the cooling does not show significance.

          Next I might look at differences series of ice cores to determine whether the difference series will show significant trends – as I suspect a number of differences will.

          I did find 6 high resolution ice cores with stable isotope measurements versus depth from Siple Dome at the link below. I do not have an algorithm (if that is indeed what it takes to convert depth to age) to plot stable isotope ratios to time, but this link does show that more ice cores are available than I was previously aware. Lacking depth to time conversion for comparison of these closely located ice cores, I can compare the cores based on equal depths as part of a future investigation.

          http://nsidc.org/data/waiscores/iso.html

        • Kenneth Fritsch
          Posted Jun 3, 2012 at 11:25 AM | Permalink

          This thread appears to have ended even through questions remain concerning the ice cores very high correlation between the del D and O18 measurements for a given core and what data belongs to which cores.

          Here I want to record six plots from six ice cores from Siple Dome data that are linked below (first link). Of interest to me generally in this matter is analysis of temperature proxies that are closely located.

          I processed the raw ice core data using the R code provided above by SteveM in this thread. Five cores covered the 1901-1994 time period and one core (Core C) covered 1837-1994. For better comparison I show only the 1901-1994 periods for all six cores. The del D and O18 data had a very high correlation within in each core and thus I show only the del D plots. The plots are shown in the last link below in this post.

          The cores were resolved to greater than annual resolution and thus when I averaged the readings for the resultant annual mean I could have included some differences year to year depending on the portion of each season captured in the mean. The annual data was an average of at least 3 evenly spaced readings which should make the seasonal effect minimal, but since I handled all cores essentially the same the comparisons are unaffected.

          The calculated trends for the 1901-1994 and 1956-1994 periods for all cores had no statistically significance for being different than 0. Viewing the plots shows obvious differences from core to core, but without better meta data it is not possible to know the potential origins of those differences.

          http://nsidc.org/data/waiscores/iso.html

        • Posted Jun 3, 2012 at 12:06 PM | Permalink

          Ken — Thanks for the plots of Siple Dome cores C-H. They clearly do not show much of 20c HS, contrary to the Core A d18O record cited by Gavin. However, these are shallow cores that don’t go back very far, and hence can’t be used to make inferences back to 1000 AD.

          I’m still puzzled by the two versions of Core A dD from Steig — they don’t seem to be the same data, though maybe I just haven’t graphed them right yet, since the chronology isn’t presented the same way. The newer version that Steig sent me and that Steve has put up on CA is almost a linear transformation of the archived d18O, but not close enough to have been generated that way for illustrative purposes. But the earlier version that Steig had recently put on his website looks much more different than the d18O series.

  21. Gras Albert
    Posted May 14, 2012 at 1:09 PM | Permalink

    An eyeball of the graph might suggest that Gavin is astigmatic, certainly exceptionally high values statement implies more UnSkepSci syndrome issues…

  22. Posted May 16, 2012 at 5:28 PM | Permalink

    I finally got the long-archived Siple Dome Core A d18O to plot out. A small complication is that the d18O data and time scale were done by different groups, and are in different files, at ftp://sidads.colorado.edu/pub/DATASETS/AGDC/waiscores/white/SDM-A_all_half_20meter.txt and ftp://sidads.colorado.edu/pub/DATASETS/AGDC/waiscores/timescales/SDMA_0-514_Nov2003.txt . Eric Steig sent me a nice e-mail helping point me to the files and apologizing for the inconvenience of not having them together in one place. He was a PI on the d18O data, but at the time was low on the totem pole and had nothing to do with the archiving.

    The d18O file is in half-meter increments, each of which corresponds to a few years of data. I just assigned each year in each half-meter the average value for that increment. Higher resolution data once existed, but this was good enough for present purposes.

    Here is the Core A d18O data by depth:

    The end of the last ice age is clearly visible between 600 and 700 meters. However, the dating only goes back to 515 meters, at 8609 years BP, so they did not attempt to date this. There is also an intriguing even colder period at the bottom of the core (bedrock) at about 1000m, but again they did not date this.

    Here is the d18O data by date, back to 1000AD, together with 25-year running means as in the Mayewski article cited by Gavin:

    This matches Mayewski’s graph of 25-year averages well, except for the final downturn, which is about 3 times deeper than Mayewski’s. The final 25-year point is heavily influenced by the final multi-year point, which was the “coldest” in the series since the last ice age. However, it’s conceivable that Mayewski impartially rejected points that were several standard deviations from their nearest neighbors,and hence discarded this, so there is not necessarily any foul play here. If they had wanted to “hide the decline”, they could have just eliminated the final dip altogether.

    This graph does represent a fair HS, as claimed by Gavin, in that the 20th (and late 19th) c contain distinctly warmer periods than any earlier period, including the MWP (which does show up here in the late 12c and 13c).

    Here is a comparison of the long-archived Core A d18O and the new Core A dD data which Eric has just put on his website. In Eric’s e-mail, he notes that d18O in precipitation varies almost 1-for-1 with dD/8, so I have plotted d18O together with dD/8, with a 3 point offset in dD/8 to separate the two graphs:

    Indeed the principal wiggles match very closely. However, there is a, distinct phase shift in the older data, which suggests that perhaps a different time scale was used in the two data sets. Eric noticed this discrepancy and is concerned about it, but the dating is close enough for present purposes.

    The big problem for Gavin’s claim that the Siple data shows an unbiguous HS is that the dD doesn’t have much of a “warming” in the last 150 years, and in fact never reaches the heights of 1200AD. Whatever the motive for not releasing the dD data earlier, Steve’s conjecture that it would not show a clear HS is borne out.

    Finally, here is a comparison of the old d18O data for Core A with Eric’s new d18O data for Core B (which only goes back to 1654AD):

    The two cores are clearly picking up many of the same wiggles, in particular some events in the 18c. However, since the new B data doesn’t go back very far, it’s hard to say whether it is or is not a HS. It does at least show a stronger LIA than Core A did.

    • Posted May 16, 2012 at 5:45 PM | Permalink

      Two of my graphs didn’t come through right. Here’s another try at Core A d18O by date:

      and Core A dD vs d18O:

  23. Posted May 20, 2012 at 11:16 AM | Permalink

    Here’s the new dD data, comparing the annual data to 25 year running means, with a 95% CI on the 25 year means. I assumed the errors have equal variance throughout, despite the mysterious quiet spell circa 1800.

    While the last 125 years have been relatively “warm”, they aren’t significantly warmer than most of the 13th c.

    The “warmest” year in the millenium is in this recent period, but it was back in 1881, before most of the anthropogenic CO2. The last 50 years don’t stand out as being any warmer than the rest of the last 125 years.

    So whatever Eric’s motive for not releasing this dD data earlier, Steve’s prediction that it would not show a HS is borne out.

  24. Steve McIntyre
    Posted May 13, 2012 at 9:51 AM | Permalink

    Ron Broberg has an excellent quantitative blog and comments as follows:

    Steve McIntyre has a post about Steig’s Siple dD series in which he posts a chart of the 20th century to undermine Schmidt’s claim that there are exceptionally high values in late 20th Century. And, indeed, the 20th century appears to be relatively trendless when eyeballed on that chart. McIntyre included his code.

    Broberg also included his code for several lowess smooths: visit his site for his graphic.

  25. Not Sure
    Posted May 13, 2012 at 1:57 PM | Permalink

    It’s awfully big of you to say nice things about Mr. Broberg, Steve. In a previous post, he coins the therm “mcintyretude” which is immediately defined as “…scorn, derision, insults, and the questioning of motives.”

    This wisecrack got him mentioned previously on “Neverending Audit”. This is a wonderful site full of gems like this one “You don’t Go to Canada for a Fair Inquiry”.

    Apparently, you have single-handedly managed to impair the analytical abilities of 34 million people. How does it feel to have that kind of power?

    Sorry if this is off-topic. I don’t really expect it to survive moderation or sniping. I couldn’t help myself.

  26. Posted May 13, 2012 at 5:59 PM | Permalink

    Broberg: Steig Siple dD: 11th – 20th Century with Lowess

2 Trackbacks

  1. […] McIntyre has a post about Steig’s Siple dD series in which he posts a chart of the 20th century to undermine […]

  2. […] McIntyre has a post up looking at Siple Dome ice core data and Gavin Schmidt’s claim that the data […]

Follow

Get every new post delivered to your Inbox.

Join 3,416 other followers

%d bloggers like this: