Kaufman's BSi Selection

Below is a plot comparing sediment BSi (biological silica) to depth (cm) from two of Kaufman’s lakes (done by different students). I’ve shown it by depth (rather than ascribed age) since the dating of these sediment series is not without some hairiness. I’ve shown equal lengths for each lake, both covering at least 800AD-present on their assigned dates. The dotted red line shows where 1963 is assigned in each study (1963 is a key date in radiogenic sediment testing.)

Kaufman et al 2009 used only one of these series (and I’m sure that the reasons are impeccable.) Similarly only one of the two series is archived in the NCDC paleo archive (I’ve extracted the other data from an online M.Sc. thesis). Again I’m sure that the reasons for only placing one of the two series in the NCDC paleo archive are impeccable.

The question for CA readers: which of the two series is used in Kaufman et al 2009? No prize for guessing the correct answer.

BTW I wish that site reports for climate proxies were as well presented as the MSc theses of Kaufman’s students – the attention to detail in these theses makes them infinitely better resources than journal publications by their professors.

UPDATE: Here’s another Alaskan BSi series from a Kaufman student – a site also not used in Kaufman et al 2009 and not archived. In this case, the thesis has a photoimage of the data (which, as a result, cannot be readily extracted as it could from the other two theses):


Reference: McKay thesis 


  1. Steve McIntyre
    Posted Sep 19, 2009 at 8:02 PM | Permalink

    If the two cores are intact, BSi looks like a pretty useless proxy if one core goes hugely up while the other goes hugely down.

    One possible explanation for this: reports of sediment coring often report the loss of part of the core at the top. Maybe the last bit of the bottom core is missing.

    I’ve explored this possibility by examining the dating of the two cores. This leads down another complicated rabbit hole that I’ll discuss some time.

  2. bender
    Posted Sep 19, 2009 at 8:17 PM | Permalink

    Depending on the depth/date calibration, the top one could actually favour their hypothesis stronger than the bottom one. Even though the bottom one has that blade uptick at the end. Sure, the blade uptick boosts the regression slope during the instrumental calibration period. But the long downslope in the fisrt half of top figure favours the premise that the Artic was cooling for 2000 years (due to orbital forcing) and offers a better match to the GCM output. The mild uptick on the top figure may nevertheless be sufficiently sharp to get a strong calibration during the instrumental period.
    But my guess is they chose the bottom one, based on HS shape.
    This example actually helps clarify some of the points I was making earlier today.

  3. jae
    Posted Sep 19, 2009 at 8:44 PM | Permalink

    snip – please don’t use such language

  4. jae
    Posted Sep 19, 2009 at 8:53 PM | Permalink

    bender:- snip – Maybe I’m wrong, here, but I would just quote you:

    But my guess is they chose the bottom one, based on HS shape.

    LOL, amigo!

  5. jae
    Posted Sep 19, 2009 at 8:57 PM | Permalink

    Just what the hell was the basis of the snip, Steve?

    prohibited language, attribution of motives, venting, angriness, piling on. ..

    And please, no food fights with bender. Both of you.

  6. bender
    Posted Sep 19, 2009 at 9:25 PM | Permalink

    No food fights. Just the important point that the depth/date calibration matters greatly.

    Steve: Actually I chose series where there isn’t that much relative dilation – see below.

  7. bender
    Posted Sep 19, 2009 at 9:36 PM | Permalink

    “Calibration” is perhaps not as accurate a term as “mapping”. Depending on how depth maps to date in the two samples there could be quite a bit of distortion in the x direction. So much so that the two could end up correlating hugely.
    Piling on is not warranted in this case because the mapping is not known to us (only to Steve, who says it’s hairy), and – more importantly – the conclusion of the paper does not hinge all that critically on this one issue – as interesting and as worthy of audit as it is.

    • Steve McIntyre
      Posted Sep 19, 2009 at 10:02 PM | Permalink

      Re: bender (#7),

      It looks possible to me that the last few decades of the bottom series have been lost – not an uncommon event with sediment cores – and that the upspike in the bottom series may have occurred 20-30 years ago. (I’m not proposing that just yet – I’m just experimenting right now.)

      If the bottom series went right back down (as the top series does), then there is a big divergence problem for the interpretation of BSi as a temperature proxy.

      If there is no truncation and the two series are as is, then there is a big divergence problem between the two series. Until this sort of thing is resolved, I don’t see how you can use this sort of stuff in a scientific compilation. It becomes phrenology.

  8. Kenneth Fritsch
    Posted Sep 19, 2009 at 9:38 PM | Permalink

    One of the students failed to “flip” the series?

  9. Steve McIntyre
    Posted Sep 19, 2009 at 9:53 PM | Permalink

    bender, I chose series where there wasn’t a huge amount of relative dilation in the ascribed dates one to another. My query is with the lack of a downtick in the bottom panel. If this thing is a climate proxy as represented, then there sure is a divergence problem between the two lakes in the past few years.

  10. ianl
    Posted Sep 19, 2009 at 10:09 PM | Permalink


    Do the core logs include a downhole geophysical log (density, caliper, gamma) ?

    This is the surest method of highlighting core losses

    • Steve McIntyre
      Posted Sep 19, 2009 at 10:17 PM | Permalink

      Re: ianl (#11),

      The top core is Cascade Lake, reported only in a MSc thesis by Kasey KAthan, a Kaufman student http://jan.ucc.nau.edu/~dsk5/S_AK/Cascade/KK_ThesisFinal.pdf ; the bottom core is Hallet Lake, one of the Kaufman et al series, also reported in a MSc thesis by a Kaufman student(McKay) http://www.cefns.nau.edu/Academic/Geology/programs/MS-geology/Theses/McKay_2007.pdf. DAta at ftp://ftp.ncdc.noaa.gov/pub/data/paleo/paleolimnology/northamerica/usa/alaska/chugach2008.txt

      • ianl
        Posted Sep 20, 2009 at 1:24 AM | Permalink

        Re: Steve McIntyre (#13),

        At this point, I’ve read Kathan’s thesis. Cores are only up to 16m depth, so deviation from vertical is a practical non-issue.

        Kathan aquired magnetic susceptibility (MS) logging “down the hole” for each percussion-driven hole. The MS logs were then used to correlate horizons common across the 3 core holes. This is a very acceptable correlation method for horizons but does not date them of course.

        An acoustic downhole log was also aquired for each hole to profile the hole (common exploration practice). The only comment on core recovery is on p.32, where Kathan notes that the acoustic logs were about 3m deeper than the recovered cores – this, over a depth of 16m, is a core loss of about 19%. Kathan notes that this loss was at the bottom of the holes, and comparing the MS logs across the holes (where common horizons may be seen), I tend to agree.


        1) percussion tends to brutalize recovered core, but Kathan notes that the recovered cores showed no evident signs of this. Photos of core sub-sections on pp. 89 & 92 support this

        2) it is likely that core losses were at the bottom of the percussion holes, but this is not proven. The core lithological log should be matched to the acoustic log to determine this. Does it make a difference ? Since Kathan limits discussion to the recovered core sections, probably not – to the MSc thesis, at least

        3) aging of the core sub-sections is poorly constrained (several comments in the thesis note this)

        4) the BSi (biogenic silica) rating is based on the assumption that biological activity in the glacial lake is higher in warm periods than that cooler periods. On p.49, Kathan comments that increased BSi ratings are more likely to be caused by precipitation (rain) than glacial melt … this seems to negate the OMGIWTWT viewpoint

  11. Steve McIntyre
    Posted Sep 19, 2009 at 10:10 PM | Permalink

    Here’s another version going back to 500 AD which shows data in the bottom panel as per the MSc thesis – this is different than in the NCDC version. Y’see the big spike in the early portion of the bottom panel – that’s deleted from the official archived version – no notation in the archive. (If one crosschecks, the article mentions that three high values are deleted.) I really dislike the idea of deleting actual data from the official record. If they want to exclude the high values on the basis that they are turbidites (which seems to be something that varvos do), then do so in the analysis, not in the archived data.

    • bender
      Posted Sep 19, 2009 at 11:00 PM | Permalink

      Re: Steve McIntyre (#12),
      What’s wrong with failing to mention a little inhomogeneity like that 😉

    • tty
      Posted Sep 20, 2009 at 11:18 AM | Permalink

      Re: Steve McIntyre (#12),

      That spike wouldn’t by any chance be around 535 AD would it? There was apparently a huge volcanic eruption with severe global climatic effects somewhere at that time, but the volcano has never been satisfactorily identified.

  12. Hemst 101
    Posted Sep 19, 2009 at 10:28 PM | Permalink

    What is the Y axis?

    Where are these lakes located in Alaska? Are they close to each other or far apart? Looking at the charts I would say the the climate of the two lakes is not at all the same if they indicate the climate at all. And Bender, the top chart is not better as the “temperature” goes up from ~1100 to the present and does not trend down. The bottom one has a sine wave appearance (except for the RH end) being “low” where the top one is “high” a VV. The only use I can see for this proxy to the study is at the RH end and that depends on the core having been loped off. These charts seem to bear very little relation to the Kaufman “result”. Of course I know absolutely nothing about this “science” but the mind boggles!

    • James Erlandson
      Posted Sep 20, 2009 at 4:41 AM | Permalink

      Re: Hemst 101 (#14),

      Where are these lakes located in Alaska? Are they close to each other or far apart?

      From the thesis links that Steve provided …
      Hallet: 61.5°N, 146.2°W; 1128 m asl
      Cascade: 60.1°N; 159.4°W; 330 m asl

  13. deadwood
    Posted Sep 19, 2009 at 10:38 PM | Permalink

    Smart of you Steve – looking for MS theses by PI students. After the last time they should have realized you’d do this. Rather bold of them to not archive the offending data, though.

    But, as you say, their reasons must be impeccable.

  14. michel
    Posted Sep 20, 2009 at 2:55 AM | Permalink

    Puzzling about the logic of this for the better part of a day now.

    Why exactly do we think that the BSi series are proxies for temperature? I can’t figure it out. Every time you try to work through this, you come on a logical difficulty. I tried imagining a case where we go to the Moon, and find a bunch of scattered instruments which have made millenial scale plots, but we do not know of what. They do not correlate. Why do we come to the conclusion that they must be tracking some quantity which, while there are random individual site variations, can meaningfully be expressed as a planetary index?

    Yes, if you construct a series, errors, if that is what they are, will cancel out, and maybe they do with ground station readings on earth over time. But when we have yet to prove that the ground stations contain thermometers, and they could just contain random number generators? Or rainfall gauges? Or windspeed?

    Does this make sense, and am I just being dense?

    • ianl
      Posted Sep 20, 2009 at 4:02 AM | Permalink

      Re: michel (#18),

      Perhaps you’re missing Kaufman’s point – which is to demonstrate that late 20th century is “worse than” the MWP, ie. more BSi activity from melting glaciers now than then

      As my review of the Kathan thesis says, no such thing is demonstrated; rather somewhat increased precipitation (maybe)

  15. Posted Sep 20, 2009 at 6:04 AM | Permalink

    RE #13, I see that McKay is listed #3 of 10 authors, out of alphabetical order, while the last 7, from Ammann on, are just listed alphabetically. This suggests that his contribution is even more important than that of Ammann, Bradley, Briffa, etc.

    A quick Google search turns up Kasey Kathan as currently a PhD student of Scott Lamoureux at Queen’s University. It would nice to alert her and McKay of this discussion just in case they want to read it or even contribute, plus a “to” or CC to the other 9 authors. (I’ll leave this to Steve’s discretion, since this is his thread — I’ve already alerted the 10 authors about my calibration thread, but they might not think to search subsequent — or previous — topics.)

    RE #12, I see from Ken Fritsch’s graph of the Hallet Lake data (#2) at http://www.climateaudit.org/?p=6932#comment-355963, that #2 goes back another 500 years to year 0, with a pronounced “warm” spell that must contribute to the pre-blade downtrend. Even since 500, the series looks a little different than your bottom graph. Are both decadal?

  16. Craig Loehle
    Posted Sep 20, 2009 at 6:37 AM | Permalink

    If we only have two samples of a proxy, BSi (about which I know nothing), and the trend in one is up whenever the other is down, as we see here, then it takes a heroic leap of faith to use either one as a thermometer. Leap leap.

  17. Posted Sep 20, 2009 at 7:26 AM | Permalink

    Re Craig Loehle #22,

    If we only have two samples of a proxy, BSi (about which I know nothing), and the trend in one is up whenever the other is down, as we see here, then it takes a heroic leap of faith to use either one as a thermometer.

    The SI indicates that Haukadalsvatn (#17) in SW Iceland also used BSi (plus OM = Organic Material), making 3 in all.

    I initially thought that BSi must be a Boron-Silicon ratio, sort of like an MgCa series. It would clearly be nice if McKay and/or Kathan could help us out here!

    • TAG
      Posted Sep 20, 2009 at 7:50 AM | Permalink

      Re: Hu McCulloch (#24),

      initially thought that BSi must be a Boron-Silicon ratio, sort of like an MgCa series. It would clearly be nice if McKay and/or Kathan could help us out here!

      from http://channelislands.noaa.gov/pcw2/pcwglossary.html

      Biogenic Silica(Bsi) – Silica that has its origins in the shell structures of living organisms

      • TAG
        Posted Sep 20, 2009 at 7:54 AM | Permalink

        Re: TAG (#25),

        and as well

        The amount of sediment suspended in a particular body of water, sometimes called the sediment load, is a property that can be measured in a number of different ways. One particularly sensitive measure for the sediment load is Silica concentration in the water column. Scientists measure two types of silica in Plumes and Blooms project, Lithogenic Silica and Biogenic Silica. Lithogenic Silica comes from the land, also called the lithosphere, and Biogenic Silica comes from the shells of marine animals. In the Plumes and Blooms data sets these two attributes are abbreviated as Lsi and Bsi

        • TAG
          Posted Sep 20, 2009 at 8:01 AM | Permalink

          Re: TAG (#26),

          So there aare two types of silica in water, one organic Bsi and one inorganic Lsi. Lsi can compose up to 90% of the sample. Its concentration depends runoff and indirectly precipitation. Bsi comes from the diatoms in the water and their concentration can depend as well on temperature but presumably other factors.

          I wonder how the multiple confounding factors are dealt with?

    • Steve McIntyre
      Posted Sep 20, 2009 at 8:39 AM | Permalink

      Re: Hu McCulloch (#24),

      there are some other BSi data sets in circulation. I’ve spent a little time collating them and will report back some time. You’d think that this sort of comparison is the sort of thing that Kaufman could have usefully done – as opposed to doing a Briffa composite.

  18. Posted Sep 20, 2009 at 8:09 AM | Permalink

    RE TAG #25, 26,
    Would it be reasonable to hypothesize that BSi responds positively to fertilization by atmospheric CO2 in addition to temperature? Are these organisms photosynthetic, or do they eat photosynthetic stuff? Should the calibration of these series therefore hold CO2 constant, or at least ask if it is a contributing factor, as would obviously be appropriate for Tree Rings (but was not done by these authors)?

    Steve: Hu, I’m not sold on CO2 fertilization as being that big an issue with tree rings, as compared to strip bark growth pulses as being like the pulse inhomogeneity that we’re seeing with some of the varve series – wild inhomogeneities. It’s interesting for me to see some new bulk distributions of a different sort of data.

  19. Steve McIntyre
    Posted Sep 20, 2009 at 9:00 AM | Permalink

    Here’s Fig 22 from the MSc thesis of another Kaufman student (Daigle url 40 MB) showing BSi from Goat Lake, Alaska, which has a shape that the Idsos would like. This data was not used in Kaufman et al 2009 and was not archived at NCDC – and, again, I’m sure that the reasons are impeccable. This thesis also color codes the geological formations, which I found very helpful.

  20. Posted Sep 20, 2009 at 9:27 AM | Permalink

    RE Steve #30,
    What’s the gray-coded period from 100BP to 290BP, when BSi was zero or virtually zero? Had the lake dried up almost entirely? Taking logs would cause chaos on this series.

    BTW, from the -55BP date, it looks like “BP” means the traditional pre-1950, rather than pre-2000 as is sometimes used. Recall that Craig had to revise a couple of his chronologies because of this ambiguous notation.

    Again, it wouldn’t hurt to alert Thomas A. Daigle of this discussion, just in case he has some info to offer. I can’t find his current whereabouts, however.

    Here’s a nice picture of Goat Lake, plus an NSF Press Release with video promoting Kaufman 09’s contribution to the NSF mission.

    • Howard
      Posted Sep 20, 2009 at 11:13 AM | Permalink

      Re: Hu McCulloch (#31),

      What’s the gray-coded period from 100BP to 290BP, when BSi was zero or virtually zero? Had the lake dried up almost entirely? Taking logs would cause chaos on this series.

      That’s the period of the maximum LIA glacial extent over-topping the spillway where the sediment consists of rock flour. What this plot indicates is that the recent LIA was an unprecedented cooling event over the last 3,500-years.

      Re: Joel McDade (#32),

      Same experience here (GeoProbe?). Recovery rates for direct push coring are dependent on material type and density contrasts. If a dense soil (or rock) fills the end of the core with a soft low density material below, the soft material can be pushed aside. This is evident by the significant core barrel frictional boundary-layer mushrooming of previously flat layers seen in the core photos. This type of deflection is common, but I have not seen this degree of mushrooming very often. I suspect that these young low density sediments are the culprit.

      The mushrooming indicates in-barrel soil densification may also contribute to softer materials being pushed aside resulting in reduced core recovery rates. The lack of technical details of the coring operation, barrel sleeve type, sleeve length and a proper boring log makes it difficult to completely evaluate the logs. I suspect that sonic coring would result in better recovery

  21. Joel McDade
    Posted Sep 20, 2009 at 10:39 AM | Permalink

    I do a lot of soil sampling using air hammer coring. Losses do typically occur at the bottom of the core run. But individual runs are only 5 feet in length, so overall core losses occur at these intervals, not just at the bottom of the hole.

  22. Posted Sep 20, 2009 at 10:57 AM | Permalink

    This data was not used in Kaufman et al 2009 and was not archived at NCDC – and, again, I’m sure that the reasons are -snip-

    No piling on!


  23. Harry Eagar
    Posted Sep 20, 2009 at 1:18 PM | Permalink

    ‘reports of sediment coring often report the loss of part of the core at the top’

    Is this due to the activity of taking the core?

    • ianl
      Posted Sep 20, 2009 at 5:45 PM | Permalink

      Re: Harry Eagar (#36),

      The top of a sediment pile is generally looser and more unconsolidated material than the more compacted material beneath. Mostly, the top needs casing to allow the drilling to continue.

      However, in Kathan’s thesis, a comparison of core log with the acoustic scan of the hole showed the core losses to be at the bottom of the hole – perhaps because drilling cuttings (percussion) were settling in the bottom and making core recovery difficult. My post #17 notes this.

      Kathan maintains that the BSi index is a measurement of the skeletal silicate remains of diatomaceous activity. Various laboratory examinations of the core samples and sediment traps (including some electron microsocopy) were used to decide the origin of the silica shards.

      Again, Kathan’s contention is that increased levels of BSi in the area of study were a result of summer precipitation, not glacial meltwater (p.49)

      I haven’t completed a full reading of McKay’s thesis as yet, but can find no direct comment on core recovery as yet.

  24. njc
    Posted Sep 20, 2009 at 1:35 PM | Permalink

    Re #27

    A method for distinguishing between BSi and LSi has been suggested:
    Olivier Ragueneau et al Continental Shelf Research 25, (2005) 697-710

    The MSc thesis by Kathan makes no mention of LSi. In all likelihood because they use mildly basic conditions (Mortlock and Froelich (1989) ) which would not dissolve LSi.

    However, things are not so simple. When looking at trapped sediments the sample is pretreated with 1M NaOH, here one might expect LSi to be dissolved as well.

    As ever I suspect these questions arise only because we are not familiar with the laboratory background. Almost certainly there exists a background knowledge which is taken for granted in the MSc thesis and not thought to need referencing.

  25. Posted Sep 20, 2009 at 2:19 PM | Permalink

    RE njc #37,
    I gather that essentially the same problem of separating diatom skeleta from other sediments was present in at least two of the series Craig Loehle used in his reconstruction, namely deMenocal and Farmer. See http://www.econ.ohio-state.edu/jhm/AGW/Loehle/ and links there, although they were looking at Mg/Ca ratios. Apparently there is an adequate way of doing this in bulk, but I don’t know any details.

    RE Tag #27 and Steve’s reply to #28, If the organisms are diatoms, then they are photosynthetic, and at least candidates for controlling for CO2. Wikipedia has a nice article on diatoms.

  26. Robinedwards
    Posted Sep 20, 2009 at 2:27 PM | Permalink

    I just wrote something that vanished without trace! Will try again.

    I am finding this thread a bit confusing. Is it known for certain what the period of the Kaufman proxies actually was, in terms of the widely used “BP” scale? I can understand that scale readily enough. Steve’s plot in #8 uses “Ascribed year” as the x axis. When I read that Kaufman states the period of his work covers from 5600 to 3000 how do I reconcile Steve’s axis notation with this scale? The Excel file that I’ve downloaded and begun to examine uses Year as its first column, on a scale going from 5 to 2000 – if I’ve remembered correctly – in steps of 5. This confuses me a lot :-((

    Is there a full list of Kaufman’s proxies somewhere? I fear I must have missed it.

    Thanks for any help/advice.


  27. Robinedwards
    Posted Sep 20, 2009 at 2:33 PM | Permalink

    Modification to #39. Time period runs from 1995 to 5 by decades in the Excel file.

  28. Julian Flood
    Posted Sep 20, 2009 at 10:19 PM | Permalink

    Re 41

    “Because of this bloom-and-bust lifestyle, diatoms are believed to play a disproportionately important role in the export of carbon from oceanic surface waters[9][8] (see also the biological pump).” Wikipedia

    “The first stable product of photosynthetic carbon fixation by land plants is either the three-carbon molecule phosphoglycerate (in C3 plants) or the four-carbon compounds malate or aspartate (in C4 and CAM (crassulacean-acid metabolism) plants). Reinfelder et al. infer that a C4 biochemical pathway of carbon fixation also operates in marine diatoms1, 2 on the basis of their discovery of the enzyme phosphoenolpyruvate (PEP) carboxylase and of their 14C-tracer results in the marine diatom Thalassiosira weissflogii. However, we consider that further analysis is called for to demonstrate that this marine diatom meets all the criteria for C4 photosynthesis.” Nature

    e.g. A. Leynaert P. Tréguer, Christiane Lancelot and Martine Rodier 2000

    It would appear that a study of long-term oceanic diatom population changes might be interesting.


  29. Don Keiller
    Posted Sep 21, 2009 at 9:32 AM | Permalink

    snip – venting. Please don’t waste my time making me snip this sort of stuff/

  30. Layman Lurker
    Posted Oct 4, 2009 at 9:34 PM | Permalink

    A reference at CO2 Science to a Nunavut study analysing fossil chironomid distributions.


  31. MikeN
    Posted Oct 20, 2009 at 7:45 PM | Permalink

    Why can’t you extract the graph from the third paper? ust take a snapshot, and cut out the rest. I could do that.
    Steve: I wanted the digital data.

%d bloggers like this: