A New Alaskan d18O Series

PAGES2K Arctic introduced a lake sediment d18O series from Kepler Lake, Alaska that hadn’t been used in previous studies. Although O18 data is a workhorse of paleoclimate, O18 data from Alaska (or, for that matter, anywhere in the Arctic hemisphere between 90E and 90W – going east) is very scarce. Thus, the appearance of a new d18O series from Alaska is of considerable interest. I’ll show why by comparing the new data to other O18 information:

The Mt Logan series, shown in the top panel below, is almost unique as a high-resolution “Pacific” hemisphere d18O series. As has been discussed in past CA posts, it goes the “wrong way” in the 19th and 20th century – attributed by the specialist authors (Fisher et al) to changes in regional circulation.

The Kepler Lake series is shown in the third panel/ Two cores were drilled. Both cores show a decline in O18 values comparable to the decline in O18 values in the nearby Mt Logan ice core. The resolution in the ice core is considerable greater, but the Kepler Lake values are sufficiently similar to provide a form of confirmation of the Mt Logan results. However, there are obviously differences in detail, both between the two sediment cores from Kepler Lake and between the sediment cores and the Mt Logan core. My own surmise is that some amount of the difference arises from dating errors in one or both sediment cores. If this data were given to an exploration geophysicist, my sense is that they would treat the Mt Logan data as the best-dated and try to use that as a dating model for the lower-quality sediment data (if they even bothered using the sediment data.)

The second panel shows an older and shorter Mt Logan ice core drilled by Holdsworth in the late 1980s (used in Mann et al 2008). It also shows a decline in d18O values in the 19th century, but the timing of the decline is somewhat later in the Holdsworth core: is this due to inconsistency in dating between the two cores? Or is it an authentic feature? Dunno. One wishes that multiproxy authors would engage in this sort of question.

Finally, the bottom panel shows an earlier lake sediment d18O series (Farewell Lake, Alaska), which was used in Ljungqvist 2010, a multiproxy study cited in AR5. Although Farewell Lake is reasonably close to Kepler Lake, the two series obviously have an entirely different appearance. The Farewell Lake data has a much lower resolution (about 52 years) and completely lacks even the gross features of the more highly resolved data. While the inclusion of “lower frequency” data has become somewhat fashionable in the more recent multiproxy reconstructions, the comparison shows the need for caution in including such muddy data: the Farewell Lake data is useless in a 1000-year reconstruction.

Figure 1. ALaska O18 series. Top – Mt Logan (Fisher et al 2006); second – earlier version of Mt Logan (HOldsworth); third – Kepler Lake (Gonyo et al, used in PAGES2K Arctic); bottom – Farewell Lake AK (Hu et al ) used in Ljungqvist 2010.

Despite the almost total absence of O18 data in the Pacific hemisphere, remarkably little interest has been shown by specialists in remedying the situation. Obviously many sediment cores have been taken in the eastern Canadian Arctic, but no d18O results have been reported for any of them. In 2002, Lonnie Thompson drilled an ice core on the Alaska-Yukon border (Bona-Churchill), the results of which should have been eagerly awaited by specialists. However, more than decade later, the core remains both unarchived and unpublished. I drew attention to the delay over six years ago in early CA posts. At the time, I speculated that Thompson would have promptly published Bona-Churchill if it had had “good” results; I therefore speculated that, like Mt Logan, Bona-Churchill results went the ‘wrong way”. Be that as it may, there is only a single d18O ice core series in the Pacific hemisphere, as compared to dozens in the north Atlantic hemisphere.

Remarkably, despite its uniqueness, the Mt Logan series has been excluded from the collection of Arctic ice cores used in two prominent recent multiproxy studies (TIngley and Huybers 2013; PAGES2K Arctic).

One of Tingley and Huybers’ “justifications” for excluding Mt Logan was that it was “out of phase with paleotemperature series”. The next figure compares Mt Logan O18 to Lomonosovfonna, Svalbard O18, one of the most easterly ice cores in the North Atlantic hemisphere. In a sense, the Mt Logan series is indeed “out of phase” with the Svalbard series, but this doesn’t justify deletion of the Pacific hemisphere series. Tingley and Huybers also observed that the original authors of the Mt Logan series (Fisher et al) had “explained” the decline in O18 values as due to a change in “source region”. If so (and it is entirely plausible), it opens up the prospect that anomalously high O18 values in other areas might also be due to a change in “source region”. If changes in source region are a source of error, there is all the more reason to closely examine and analyse all the d18O series, rather than expunging inconvenient series from the network, thereby, so to speak, hiding the decline in d18O values, as Tingley and Huybers did. Their intentions may well have been “good” but the effect is a form of ex post screening.

Figure 2. Top – Mt Logan; bottom – Lomosovonovfonna, Svalbard

Although PAGES2K Arctic also had an extensive collection of ice core isotope series, they also expunged the Mt Logan series from their network, again setting out seemingly objective selection criteria. Since Kepler Lake is comparable in its major features to Mt Logan, it’s hard to understand how one series “passed” the PAGES2K criteria, while the other one failed. However, these are the same folks who used the contaminated Igaliku series and used the Hvitarvatn data upside-down, so any search for precision is bound to be fruitless.


  1. Timothy Sorenson
    Posted Dec 16, 2013 at 9:01 PM | Permalink

    I believe you meant “In 2002, Lonnie Thompson drilled…”

    Steve – fixed.

  2. Brian H
    Posted Dec 16, 2013 at 9:06 PM | Permalink

    So, what does it all mean? What potential interpretations of the “out of phaseness” would the various authors be trying to forestall?

    Steve: The full quote is

    “Details and references for the ice core records used in the analysis are available in Table S.2. We exclude the Mount Logan series that is included in [35 – Kinnard et al 2011] because the original reference [36 -Fisher et al 2006] indicates it is a proxy for precipitation source region and is out of phase with paleotemperature series.

    My take is that “out of phase” in this context means that it goes down in the 19th century when they expected it to go up. But regardless of their intention, in my opinion, if they believe that Arctic ice core O18 series are a temperature proxy, then they are obliged to take all of them – not do ex post selection based on correlation with temperature. This is the same issue that’s been discussed here over and over.

  3. bernie1815
    Posted Dec 16, 2013 at 11:48 PM | Permalink

    What led to your surmise that Kepler cores had possible dating problems?

    Steve: the dating of Mt Logan is going to be more accurate than the dating of Kepler Lake, which is interpolated by radiocarbon and only approximate. My surmise is that the O18 histories at Mt Logan and Kepler Lake would have been similar enough that the downturns should be sort-of synchronous. Similarly, there are discrepancies between the two Kepler Lake cores. Some would be due to “noise”, but there is play within the radiocarbon errors as well.

    • Posted Dec 17, 2013 at 5:14 PM | Permalink

      My eyes want to wiggle-match the graph timelines. It might be a step forward in accurate dating for these graphs. I wonder what a 100 yr filtered Logan would look like next to Kepler.

      • bernie1815
        Posted Dec 18, 2013 at 1:17 PM | Permalink

        I think we may have noticed the same shape similarity/distortion. But it is always dangerous to put your eye to a keyhole. 😉

  4. Posted Dec 16, 2013 at 11:58 PM | Permalink

    d18O from a lake sediment is a new one on me. Usually it’s from ice cores like Mt. Logan and is expected to correlate positively with temperature (except when it doesn’t, like Bona Churchill …).

    Which way is lake varve d18O expected to go with temperature, and why? Some lake water (recent rainfall and snowfall) will be entrained in the sediments, but these sediments are porous and the water in them is essentially groundwater, which percolates gradually through the soil. If the water isn’t staying in place, what does it tell us about climate history?

    I’ve been laying low of late, working on other things. But I’ve been trying to learn R, and so may one day be able to read your codes, Steve!

    Steve: Kepler Lake isn’t “varved”. The O18 in the lake reflects the precipitation. The Pumacocha O18 values also relate to nearby Quelccaya – an interesting relationship that I mean to post on some time.

    • tty
      Posted Dec 17, 2013 at 11:00 AM | Permalink

      Another possible problem with arctic lake d18O series is, that if the lake is fed with glacial melt water, part of the water will be derived from precipitation that fell hundreds or thousands of years earlier.

      • mpainter
        Posted Dec 17, 2013 at 1:48 PM | Permalink

        So it seems, and a big problem and not resolvable, I think.

  5. Geoff Sherrington
    Posted Dec 17, 2013 at 12:29 AM | Permalink

    Steve, you surmised “If this data were given to an exploration geophysicist, my sense is that they would treat the Mt Logan data as the best-dated and try to use that as a dating model for the lower-quality sediment data (if they even bothered using the sediment data.) ”
    You did not ask for an experienced exploration geochemist’s view, so I offer one gratuitously for you to snip if you wish.
    1. Why is this type of analysis being done by you after the horse has bolted and policies of global importance decided on global data like this?
    2. There is no immediately useful signal for further proxy work. A paper should perhaps have been written saying this, to avoid more people going down dry gulches.
    3. The aim of the exercise is to correlate a proxy response with temperature at given times, not so much with another proxy. The within-location variability (two adjacent cores) is huge and the between-location error is worse, allowing that comparison is sometimes between ice core and sediment core. Lacking separate criteria, I would not select any series as a reference. I’d reject them all.
    4. As a geochemist, I would have taken the data to a sedimentologist and asked for reports on the viability of processes to reconstruct the past from the measured information and information like it from elsewhere.
    Are there any well-qualified sedimentologist reading this who would care to comment on the hurdles to be overcome in reconstruction of relevant past events in the sedimentary record?
    As with the preceding post, I’m aware of complications in past erosional history, glaciation (not my strong suit), catchment area changes, lagged responses, etc, some of which are probable complications that my limited reading has not encountered in this type of application (lake sediment proxies for temperatures).

    • ianl8888
      Posted Dec 17, 2013 at 5:43 PM | Permalink

      Do we know if any downhole high-resolution geophysical logs are available ?

  6. Adam Gallon
    Posted Dec 17, 2013 at 6:14 AM | Permalink

    Is there actually any observational evidence, that any of these proxies are solely temperature dependent?

    • Posted Dec 19, 2013 at 11:56 AM | Permalink

      The word “solely” is surely out of reach for all temperature proxies –whether it be tree rings, ice cores or dO18. Preponderantly would be a nice goal to aim for but Steve M’s relentless dissection of the touching faith in proxie untilization does not inspire confidence in reaching such a goal. And any hope of reaching such a goal is bedevilled by the absence of adequate caution.

  7. Posted Dec 17, 2013 at 8:16 AM | Permalink

    Steve, you’re downright dangerous when you get your eyes on data that others have overlooked, ignored or treated lightly.

    Good post Steve.

    Geoff; Know any sediment specialists that you can involve? As usual, your comments intrigue me. I have my doubts about some sediment analysis statements I’ve read. Truthfully, I am not sure those sediment analysis were actually performed by someone trained in sedimentology.

    • Geoff Sherrington
      Posted Dec 17, 2013 at 8:47 PM | Permalink

      In summary, I use an adage that ‘knowing when to get into a matter is 1/3 of the challenge, knowing when to get out is 2/3’.
      The climate science tendency is to spend too long down the path of diminishing returns, to the point where there can be more discussion about the adjustments than about the basic meaning of the data. Exploration geochemistry has a practical purpose that promotes the ‘look carefully then move on sharply’ tendency. One does not ‘adjust’ geochem data. Maybe the two fields should not be compared, but I think they can be, and should have been, with benefit.
      I’ve been retired too long to know current sedimentologists. However, even before the specialist stage, several complicating problems are apparent from this and the previous thread on lake sediments. Some have already been mentioned here, like the mixing of isotope signatures in lake liquids. I see evidence that causes doubt about the usual simplifying assumption, that of uniformitarianism. (Such as natural dam breaks).
      I’d pass by the data above because an early step would be likely to fail. That is the calibration step. Even if you assume zero error in the measured temperature/time data, the error involved between the 2 cores from the same locality would be so large that the resulting inferred proxy temperatures would not match adequately. Note that I don’t prefer one core over another because one might better fit preconceptions about past temperatures derived from other proxies. That would be a mild case of data dredging, which is a horrible way to go, but is common in climate science.

  8. EdeF
    Posted Dec 17, 2013 at 10:10 AM | Permalink

    Both Svalbard and Mt. Logan show a real plummet in deltaO18 in the early 1800s, then
    a rebound. From the link below, which has a great primer in the introduction on
    the causes of depletion in dO18, its concentration in precipitation can be a function of temperature, distance from the coast, elevation and the amount of precipitation,
    and storm direction. Two sites relatively close by could have much different
    levels. I am thinking of two sites on either side of the Sierra Nevada, one at about
    4k ft elevation in the heart of the maximum rainfall zone, and one site on the eastern
    flank of the mountain, at higher altitude, in the rainfall rainshadow area, which
    also may get monsoon moisture in the summer from the south. I suspect when I look more
    closely at Mt. Logan and surrounding lakes, this may be the case. My other thought was
    in the nature of the possible lack of good instrumented temperature stations in the far, far hinterlands of the cold, white polar region. Would be interesting to see
    how far they had to go to find a wx station for some of these remote sites. I am guessing Mt. Logan teamed with Phoenix, Az?

  9. Posted Dec 17, 2013 at 10:12 AM | Permalink

    Interesting “new” proxies (in the form of ancient tree stumps) were recently discovered under the retreating Mendenhall Glacier in Alaska this summer:


    I’d love to see the tree ring analysis done on these ~2000+ year old unfossilized old stumps.

    • Posted Dec 17, 2013 at 10:15 AM | Permalink

      That should read ~1200+ year old stumps.

    • Speed
      Posted Dec 17, 2013 at 11:25 AM | Permalink

      From the article …

      “Those trees are newly thawing under the glacier for the first time in 2,000 years, which is kind of neat,” she said. “It’s like going to the tomb of King Tut or something like that. The tomb of the king spruce tree.”

      I guess that 2,000 years ago it was pretty warm there.

  10. JohnGM
    Posted Dec 17, 2013 at 11:29 AM | Permalink

    The earliest trees newly revealed under the Mendenhall Glacier are dated to 1400-1200 years ago:

    “The most recent stumps she’s dated emerging from the Mendenhall are between 1,400 and 1,200 years old. The oldest she’s tested are around 2,350 years old. She’s also dated some at around 1,870 to 2,000 years old.”

    • JohnGM
      Posted Dec 17, 2013 at 11:31 AM | Permalink

      More clearly: “The latest trees newly revealed under the Mendenhall Glacier are dated to 1400-1200 years ago”

  11. Paul Dennis
    Posted Dec 17, 2013 at 6:05 PM | Permalink

    Steve, one needs to be careful in interpreting the Kepler Lake data. I don’t know the study but from a brief web search it looks as though it is by Gonyo, Yu and Bebout and was initially published as an abstract at the AGU fall meeting in 2008. It looks like they analysed the d18O of inorganically precipitated calcite.

    There are two controls on the isotope composition of calcite: (i) the water temperature at the time of precipitation, and (ii) the isotopic composition of the water. Assuming the lake water to be meteoric then one expects to see higher d18O values associated with warming temperature trends. However, a warming trend results in reduced isotope fractionation between precipitated calcite and parent water that somewhat cancels out the effect of temperature on the meteoric water composition.

    As a very rough guide there is about a 0.7 per mille increase in d18O in meteoric water for an increase of 1 degree C in temperature. For the same 1 degree rise in temperature the fractionation between calcite and water is reduced by about 0.25 per mille. There is no easy way to unravel the two effects given knowledge of the calcite isotope composition only. One is left with an arm waving exercise when trying to interpret the results.

    In the Kepler Lake record there is an increase in the d18O value of calcite during the period of the LIA. This is in the opposite direction to what one might infer from knowledge of isotopes in the water cycle and the thermodynamics of isotope partitioning during calcite precipitation so it is suggested that there has been a change in air mass trajectories and rainwater source region with a consequent change in the lake water isotope composition.

    Unfortunately this cannot be uniquely determined from the stable isotope data alone and remains at best a suggestion.

    • Posted Dec 18, 2013 at 11:47 AM | Permalink

      Thanks, Paul. A further confounding factor is that only 1/3 of the O in the carbonate in the calcite can come from meteoric H2O. The remaining 2/3 would then come from atmospheric CO2, whose 18O could be doing something completely different from that in atmospheric H2O. Or, if the carbonate in the water is from dissolved limestone upstream, all of its 18O could be irrelevant to the current atmosphere.

      • Paul Dennis
        Posted Dec 18, 2013 at 12:40 PM | Permalink

        Hu, this is not strictly correct but I can see why you have written this! In practise the oxygen isotope composition of the dissolved inorganic carbon is in equilibrium with and buffered by the isotopic composition of the water. The number of oxygen atoms in water molecules is many orders of magnitude greater than that in dissolved inorganic carbon. Thus under dynamic equilibrium with constant hydration and dehydration of dissolved CO2 one ends up with a pool of bicarbonate and carbonate ions that are in isotopic equilibrium with the water. The kinetics are pH dependent but reasonably fast being on the order of tens of minutes to hours, rather than days or weeks.
        However, if the calcite/aragonite or whatever carbonate phase is precipitating is doing so rapidly (e.g. as a result of a phytoplankton bloom), or rapid CO2 degassing associated with springs etc. then all bets are off as to whether or not the oxygen isotope composition of the calcite is in equilibrium with the water. One could relatively easily check if the calcite in this case is in equilibrium by measuring modern calcite from the lake and it’s associated lake water. If the data is available I might do the calculation tonight.

        • Pat Frank
          Posted Dec 18, 2013 at 6:29 PM | Permalink

          Paul, one might naïvely think that the calcite precipitation from a phytoplankton bloom would be a flocculent non-crystalline hydrous solid. Is this correct?

          If so, given a high surface area and intimate contact with water of a flocculent solid, would the carbonate solid-solution equilibration produce oxygen-isotopic equilibrium with lake water?

          Or would crytallization happen quickly enough for re-equilibration to be suppressed?

        • Paul Dennis
          Posted Dec 19, 2013 at 3:15 AM | Permalink

          Pat, I’m not sure what the state of crystallinity would be but there is certainly the possibility of hydrated phases being present. There is also a high probability that they are precipitating so rapidly that they are out of isotopic equilibrium with the water. We have documented some of these disequilibrium effects for carbonates associated with the freshwater green alga Chara. See Andrews et al., 2004, Equilibrium and disequilibrium stable isotope effects in modern charophyte calcites: implications for palaeoenvironmental studies, Palaeo3, 204, 101-114.

          In general we find that such carbonates are out of equilibrium with the lake water. The degree of disequilibrium does not have to be very great to cause problems. The range of likely d18O compositions for Holocene temperature changes is going to be less than 1 per mille at any one site. Disequilibrium or kinetic fractionation can cause shifts of several per mille.

        • Posted Dec 19, 2013 at 11:32 AM | Permalink

          Thanks again, Paul. I was wrongly assuming that the CO3– groups, once formed, would be stable until outgassed, but apparently they revert to CO2 and back in the water with great frequency. Do 18O and 16O participate in this process at an equal rate, and if not, does it result in noticeable fractionation in the calcite?

          A further problem, raised by TTY above, is that if the lake is glacier fed, its H2O could be centuries or even millennia old.

          Are we to gather than lake sediment d18O is pretty useless as a temperature proxy?

        • Posted Dec 19, 2013 at 11:34 AM | Permalink

          (That should be CO3(-2), but my double – became an en-dash.)

        • Paul Dennis
          Posted Dec 19, 2013 at 12:19 PM | Permalink

          Hu, the differing rates of reaction for the 16O and 18O are what ultimately lead to the equilibrium isotope fractionation. The equilibrium constant for a reaction involving 18O divided by the equilibrium constant for the same reaction but involving 16O gives the fractionation factor. Of course the equilibrium constants are just the ratios of the rate constants for the forward and backward reactions.

          TTY’s problem is a very real one. The crux of lake marl palaeotemperature estimates is being able to measure the isotope partitionin between the precipitated carbonate and it’s parent water. In virtually all palaeo studies it is not possible to estimate the water isotope composition so one can’t estimate the fractionation and thus temperature.

          Unless one has very detailed studies of the lakes modern hydrology (including source regions for precipitation, rainfall amounts, seasonality etc. and their effects on the surface and groundwater isotope composition) and subsequently understands the ‘plumbing’ system of the aquifer for groundwater fed lakes it is nigh on impossible to determine temperatures from isotopes of carbonates.

          The problem is that the phase space occupied by the carbonate isotope composition has so many degrees of freedom that virtually an infinite array of temperature, water source regions, humidity, rainfall amount etc. can be invoked to explain the data.

          What is needed is an isotopic thermometer that is independent of water isotope composition. We, and others, are working on such a thermometer. It is based on the ordering of 18O and 13C in the carbonate lattice. Because the 13C-18O bond is stronger than the 13C-16O and 12C-18O bonds there is a tendency for the isotopes to clump and not be randomly distributed throughout the lattice. The degree of ordering can be measured though it is a very small deviation from a random distribution. Currently the overall precision of the measurement in terms of temperature is on the order of 1 degree C (+/-2 sd). Whilst this is a fantastic achievement it is not yet good enough for studies of modern/Holocene material without doing very many replicates to reduce the se of the mean. I’ve had 2 masters students working on a local lake in which we combined many tens of measurements of single depths in a core to improve the standard error and it looks like it might be possible to begin to characterise temperature fluctuations associated with the LIA in the UK. We’re in the process of writing this work up.

          The beauty of the technique is it is completely independent of the water isotope composition. Thus we can tie down temperature, and then using the d18O of the carbonate estimate the water isotope composition from the experimentally determined temperature dependence of oxygen isotope fractionation between carbonate and water. With both T and water d18O we can really begin to look at the effects of possible changes in hydrology, weather patterns etc.

          Sorry for the long note!

        • Pat Frank
          Posted Dec 19, 2013 at 5:55 PM | Permalink

          Thanks, Paul, I got the paper. I also found your, Pentecost, et al., (2006) “Charophyte growth in small temperate water bodies: Extreme isotopic disequilibrium and implications for the palaeoecology of shallow marl lakes” Palaeogeography, Palaeoclimatology, Palaeoecology 240, 389-404. All very interesting stuff. It looks like a lot of careful work. I plan to read through it.

          I wonder, though, how did the curator at the Natural History Museum, London react when you said you wanted to disassemble their specimen collection? 🙂

        • Pat Frank
          Posted Dec 19, 2013 at 6:18 PM | Permalink

          Paul, I gather from your comments about your new isotope clustering method, that the degree of clustering is dependent on temperature because of the partitioning effect of bond strengths. Is that right? To be independent of water isotope ratios, is it that you have developed a normalized clustering fraction?

          Normalized cluster fraction would be independent of the isotope fraction of the source. One should always get the same ratio of isotope clusters at a given temperature, no matter the specific enrichment. Is that about right? The method would require a statistically valid population, but that’s not hard to achieve when even tiny fractions include 10^13 atoms.

          One question, though. If the experiment is done using a mass spec and liberated CO2, one will lose approximately one carbonate O18 out of three. Do you correct for that?

          On the other hand, can one use an FTIR method and look for the isotopically shifted bands in the solid? That might solve the O18 loss problem, but are IR instruments precise enough for that?

          Thinking about this a little further, I found now that people have developed laser ablation of carbonates for stable isotope counting; called “LASSIE.” With modern high-powered lasers, presumably you can sample extremely small areas of a solid carbonate surface — less than 1 um^2?

          In any case, congratulations on discovering and developing the method! It sounds *very* exciting and useful.

        • Paul Dennis
          Posted Dec 20, 2013 at 3:32 AM | Permalink

          Pat, first and importantly I can’t claim any intellectual priority over development of the clumped isotope technique. That honour belongs to John Eiler and his research group at CalTech who in about 2004 started to look at the possibility of isotope ordering in CO2 and carbonates. My contribution at UEA has been to develop a new isotope ratio mass spectrometer specifically optimised to make these measurements. With it we can now begin to develop an understanding of the underlying physical chemistry, the opportunities and importantly the limits of the technique.

          You are right on the money with your description of internal normalisation. At it’s simplest level we measure the 47/44 (13C18O16O/12C16O16O) ratio of a gas sample and normalise this with respect to the 47/44 ratio we would expect were the isotopes to be stochastically distributed. We determine the stochastic distribution from the measured 18O/16O and 13C/12C ratios (46/44 and 45/44).

          You are also right that we have to be careful to ensure that fractionation fduring acid reaction is carefully controlled. In practise we calibrate by looking at the ordering in CO2 derived from carbonate by reaction at a constant temperature. To compare across laboratories who are using different reaction schemes we need to understand the effect of reaction conditions on the derived CO2.

          FTIR is a possibility. Several groups (largely the commercial manufacturers [Los Gatos, Picarro, Aerodyne Research] are looking at IR spectroscopy of CO2, but not the solid state. AS yet the sensitivity isn’t good enough though theoretically it might be possible. One of the limits is temperature stability of the instrument over long data acquisition times. You end up building a very expensive but highly sensitive thermometer!

          LASSIE is interesting but has encountered problems. I think LASSIE is based on an IR laser system that has the unfortunate effect of heating a small spot size on the sample to high enough temperatures to break down the carbonate to CO2 and CaO. The high temperature ensures no fractionation between the CO2 and CaO. Unfortunately the thermal gradients are such that there is isotopic exchange between the ‘calcined’ spot and adjacent material. Another approach might be UV laser ablation which is a lower temperature ablation of the sample. This combined with reaction with fluorine, or another fluorine compound (ClF3, BrF5) might work. I don’t know if anyone has looked at this option.

          Following your suggestion I’ve started on a review of isotope thermometry but am undecided if (i) to submit for publication, (ii) extend it to a research manual, and/or (iii) to put it all freely available on the web. I’ve set up a web site called ‘Earth Surface Temperature Exploration Project’ at http://www.earthstep.org.
          As yet there is virtually no content other than me playing with WordPress and MathJax to see if I can find a format I like!

          If anyone is interested further in these techniques please feel free to write to me at p dot dennis at uea dot ac dot uk. It’s probably better than cluttering up Steve’s thread.

    • Posted Dec 18, 2013 at 12:04 PM | Permalink

      Yet another factor is that the carbonate in old sediments or bones can have been replaced by fresh carbonate from groundwater. Thus, radiocarbon labs do not date bone by its carbonate, but rather from its much more stable bone collagen. See http://www.radiocarbon.com/carbon-dating-bones.htm .

      • Duster
        Posted Dec 18, 2013 at 6:21 PM | Permalink

        Reliable labs will also cross-check the derived C-14 date by looking at both C-13/C-12 ratios and d-O18. Even good labs can make mistakes. Best to partition samples and submit the portions to different labs. Learned that the hard way.

    • Ben
      Posted Jan 21, 2014 at 5:06 PM | Permalink


      RE: “Because the 13C-18O bond is stronger than the 13C-16O and 12C-18O bonds there is a tendency for the isotopes to clump and not be randomly distributed throughout the lattice.”

      Interesting. How do we know that neutron capture occurred prior to C-O bonding?

      Doesn’t it make a difference if neutron capture occurred after bonding vs before? If it does, how does one filter out the neutron capture post-bonding from the 13C-18O pre-bond?

      • Paul Dennis
        Posted Jan 23, 2014 at 5:48 PM | Permalink

        Ben, as far as I am aware neutron capture for production of stable isotopes of carbon and oxygen occurs only during nucleosynthesis during the helium and carbon burning stages of stellar evolution. This means that the total number of oxygen and carbon isotopes on Earth are constant. The implication is that the concentration of 13C-18O bonds in a mineral such as calcite is not affected by neutron capture processes.

  12. Howard
    Posted Dec 18, 2013 at 12:17 PM | Permalink

    These are obviously not temperature proxies, they are likely storm activity proxies.

  13. John Reistroffer
    Posted Dec 18, 2013 at 2:30 PM | Permalink

    I wonder whether the water from glacial melt that flows into the lake during warmer climatic conditions, contaminates or dilutes the O18 signal from the precipitation from the atmosphere (such rain and snow) that is filling the lake at the time of the varve deposition?

    If you have deposition of lacustrine deposits during a relatively warm period, then water from glacial runoff contaminates the signal from the ongoing precipitation from the atmosphere, because the melting glaciers were deposited under different O18 conditions, presumably during colder periods, when the glaciers formed.

    This could account for the inverted O18 values.

    • Howard
      Posted Dec 18, 2013 at 2:50 PM | Permalink

      Great point. These plots look like they have many different signals competing for significance. Perhaps this is why Thompson is sitting on Bona-Churchill.

  14. Paul Dennis
    Posted Dec 18, 2013 at 3:05 PM | Permalink

    I’ve had a very quick look at the data for Kepler Lake (available at ftp://ftp.ncdc.noaa.gov/pub/data/paleo/paleolimnology/northamerica/usa/alaska/kepler2012.txt). I haven’t had a chance to get hold of the paper yet (Gonyo, A.W., Z. Yu, and G.E. Bebout. 2012.
    Late Holocene change in climate and atmospheric circulation inferred
    from geochemical records at Kepler Lake, south-central Alaska.
    Journal of Paleolimnology, Vol. 48, No. 1, June 2012, pp. 55-67.
    DOI: 10.1007/s10933-012-9603-8)

    Lake water Isotope compositions for Kepler and surrounding lakes lie on an evaporation trend with a range of 12 per mille d18O from precipitation values (ca. -20 per mille) through to the most evaporatively enriched at -8 per mile (Wasilla Lake). Kepler Lake looks to be amongst the least evaporated of the lakes. I don’t know when the lake was sampled (time of year etc.) or if the compositions reported are a single sample or represent averages. I note that the authors have suggested that Kepler Lake was subject to greater evaporation during the LIA. If so then it would be next to impossible to determine the lake water isotope composition and the composition of the original precipitation.

    The d18O of the inorganically precipitated calcite is ca. -16.5 per mille when reported on the vpdb scale. Taking the lake water isotope composition of -16.5 per mille (vsmow scale) the precipitation temperature determined from isotopes is about 20 degrees C. This is not inconsistent with summer maximum air temperatures of 19 degrees C at Kepler Lake (http://www.myweather2.com/Fishing/United-States-Of-America/Alaska/Kepler-Lake-MatanuskaSusitna/climate-profile.aspx?month=6). It’s likely that the carbonate precipitation is a spring/summer process.

    The range in d18O for Kepler Lake calcite is just 0.7-0.8 per mille with average maximum values of -15.7 per mille in the middle of the core. In the absence of changes in air mass circulation one would interpret such a shift as representing a trend towards carbonate precipitation during slightly warmer conditions. How much warmer precisely is difficult to say because it is not possible to deconvolve the effects of temperature on precipitation isotope composition, of evaporation on changes in lake water isotope composition and of water temperature on the fractionation between calcite and water. i.e. the system is completely under-determined and making a reasoned interpretation not possible.

    • Matt Skaggs
      Posted Dec 18, 2013 at 5:09 PM | Permalink

      Paul Dennis wrote:
      “…the system is completely under-determined and making a reasoned interpretation [is] not possible.”

      On behalf of the peanut gallery, thanks for stopping by and taking the time to write this. I notice that there have been plenty of studies on various paleo aspects of these lakes in the Matanuska valley, with many of them dating from the 1980s and then a more recent pulse (nowadays, you can core one of these lakes and finish your grueling day with a cup of Starbucks in Wasilla). All the proxy studies seem to have serious limitations, but if I were a limnologist, I would be tempted to perform a meta-analysis to see if I could tease out a common temperature signal using all the proxies and all these lakes as a group.

    • Pat Frank
      Posted Dec 18, 2013 at 6:37 PM | Permalink

      Whenever Paul comments here on temperature proxies, one immediately sees how challenging is that field. It’s physics convolved with chemistry, convolved with biology, convolved with climate. Incredible.

      One then sees what a monstrous mockery has been produced by the consensus charade-mongers.

      • Geoff Sherrington
        Posted Dec 18, 2013 at 7:54 PM | Permalink

        It is lamentable that these problems are coming to public view after important global policy changes have been made. Again, thank you, Steve.
        Re this exercise on lake sediments, if one has a clear mind, rather than one driven by a preferred outcome, it is easy to list, in advance, a number of uncertainties that can be expected to confuse the interpretation. If the uncertainties are unmanageable, the experiment should not proceed. An overall effect of unmanageable uncertainties, as Paul notes, can be that “the system is completely under-determined and making a reasoned interpretation not possible.”
        This applies beyond the oxygen isotope matter to the whole of the study of proxies from lake sediments in the presence of glaciers.

        (I mentioned colloidal processes earlier, in the context of sediment formation and lithification. To follow this theme see http://www.smedg.org.au/John%20Elliston%20Book%2007.pdf)

        • Howard
          Posted Dec 18, 2013 at 8:42 PM | Permalink

          Geoff: There is nothing to follow except a promotion for a $75 book that sounds like an infomercial from a get rich scheme. Do you have any meat on this topic to link to? I ask because colliodal processes are very interesting and I’m curious, however, my state-side geo-cynic has been triggered.

    • Posted Dec 18, 2013 at 8:05 PM | Permalink

      “system is completely under-determined and making a reasoned interpretation not possible.”

      So much potential yet no answers.

      • Howard
        Posted Dec 18, 2013 at 8:46 PM | Permalink

        Jeff: First off: Congrats on Mr Sausage.

        I’m sure an appropriately timed blade could be teased out of what you engineers call noise (it’s actually a conglomerate breccia of signals)

  15. Kenneth Fritsch
    Posted Dec 18, 2013 at 5:30 PM | Permalink

    “If so (and it is entirely plausible), it opens up the prospect that anomalously high O18 values in other areas might also be due to a change in “source region”. If changes in source region are a source of error, there is all the more reason to closely examine and analyse all the d18O series, rather than expunging inconvenient series from the network, thereby, so to speak, hiding the decline in d18O values, as Tingley and Huybers did. Their intentions may well have been “good” but the effect is a form of ex post screening.”

    SteveM, there would appear currently to be no motivation for the current lot of climate scientist to devote efforts to this more basic work you suggest. If reviewers and some climate scientists of a different bent were available and making the necessary criticisms and pointing to the very basic errors in selecting temperature proxies after the fact there might be some motivation. I do not see that happening while the modus operandi in climate science is not rocking the consensus boat or even to give that impression.

  16. Kenneth Fritsch
    Posted Dec 19, 2013 at 11:23 AM | Permalink

    True scientists are just as interested in following up “bad” or unexpected results with further study as they are confirming results. It appears to me that as a general case in climate science the same science principle applies – after substituting a “hearty hand wave” for “further study”.

    I would suppose that the thought is that, when the mission is so important and urgent for mankind and that a consensus of expectations exists, only confirmation of that consensus is required and negative results are “wrong” by committee. How else can the wrongheaded notion of selecting proxies after the fact be explained?

  17. Geoff Sherrington
    Posted Dec 19, 2013 at 9:14 PM | Permalink

    Paul Dennis,
    You are into some fascinating work.
    The following question is not based on experiment, it is just a thought.
    You note that ” … the differing rates of reaction for the 16O and 18O are what ultimately lead to the equilibrium isotope fractionation.”
    I’m thinking about the sources of reactants, which themselves are products of fractionation reactions earlier in a sequence.
    Example, if you study calcite, as Hu notes, the oxygen can theoretically come from CO2 or H2O or maybe partially from other sources. Each of these sources will have been formed during its own fractionation reaction, and so on ad infinitum as Swift wrote about fleas.
    (A analogy is a uranium decay series whereby disequilibrium can be caused by depletion/addition of an isotope or several isotopes higher in the series.)
    Does this pre-fractionation effect require adjustment or has it been found too small to need consideration?
    Next, if we consider water alone as a source of O18 in calcite, there are various possible precursors as has been noted. Some are easy to visualise in a relatively open system, but we might be dealing with a somewhat closed system where a part of the water comes from dehydration of hydrous minerals during compaction, within the sediment and perhaps inaccessible to lake waters. Such bound water could be imagined to have a rather different isotope history to an alternative, like open lake water. Its availability for reaction could depend on factors that have little to do with outside air temperatures and the time at which it reacts need not be related to seasons.
    These complications would be known to you, so your work on crystal structures could be seen as a way to investigate them better to see if. more generally, it really possible to relate oxygen isotopes to ambient temperature. The loose, practical equations that are commonly used have many such potential complications that seem to have been glossed over.

    • Paul Dennis
      Posted Dec 20, 2013 at 3:36 AM | Permalink

      Geoff, you are right here vis-a-vis precipitation steps and incorporation of different CO3 anions with varying fractionations etc. Some have sought to explain an apparent pH effect on the oxygen isotope fractionation between water and calcite as a result of incorporation of varying amounts of HCO(3-) and CO3(2-) anions into the carbonate lattice. Others have observed that there is an ‘orientation’ dependence on the isotope fractionation in the sense that rapidly growing faces have a different isotope composition to more slowly growing faces. Of course these must all be disequilibrium effects but whether these pre-cursor phenomena as you call them are ever equilibrated is a moot point. Can I refer you to the long note I left for Frank above where you’ll find my email. I’m very aware that this discussion whilst very interesting is taking us away from Alaskan lake proxies!

      • michael hart
        Posted Dec 20, 2013 at 2:25 PM | Permalink

        Paul, do you also look at biological oxygen fractionation in non-equilibrium carbonate precipitation, such as by carbonic anhydrase?
        For example, Watkins et al

        Click to access Watkins_etal_2013_EPSL.pdf

        • Paul Dennis
          Posted Dec 20, 2013 at 3:23 PM | Permalink

          Michael, I’m very grateful for you drawing my attention to the Watkins et al paper. It was completely under my radar! We’ve done work on biological systems in both the natural and laboratory environment. e.g. the papers I referred to above in the discussion with Pat Frank and also an experimental study of biogenic isotope fractionation in freshwater snails.
          It is true that more than 50 years after Urey’s seminal work we still haven’t a robust calibration of oxygen isotope fractionation between calcite and water. Whilst seemingly simple it is a difficult experiment to do with kinetic effects undoubtedly the key player. Using carbonic anhydrase to ensure equilibrium amongst the DIC pool and water as Watkins et al have done may help to isolate kinetic effects associated with crystal growth such as diffusion, step migration etc.
          We’re currently precipitating calcite at very slow growth rates from saturated solutions by controlling the rate of CO2 diffusion through a gas permeable membrane. The aim is to ensure (i) equilibrium in the DIC pool and (ii) no kinetic effects associated with crystal growth.
          I’m not entirely convinced by the suggestion that cave calcites are in isotopic equilibrium. It is true that on a macroscopic scale the overall kinetics of speleothem growth are slow. However, they grow from thin water films in which CO2 degassing kinetics are sufficiently fast that equilibrium amongst the DIC pool is not maintained. This is evidenced by the co-variation of d18O and d13C along growth layers. We also find that speleothems are out of equilibrium wrt to clumped isotopes.
          I do like the approach being taken by Watkins et al. and am sure that such theoretical models will lead to great insight.
          As I said above I’m more than happy to continue this discussion but am very wary of derailing this topic away from Alaskan Lakes. If anything, though, the discussion does highlight the great difficulties in applying some of these proxies and how aware we need to be of problems and limitations.

          Steve: don’t worry about connection to Alaska. AN interesting discussion. I’ve been trying to understand d18O data from different environments.

        • Howard
          Posted Dec 21, 2013 at 11:30 PM | Permalink

          Is carbonic anhydrase found in varves? I would expect the dominant biological mechanism to be natural anaerobic respiration in the organic layers utilizing oxygen by reducing nitrate, iron oxide, sulfate, etc. to increase carbonate dissolution. Due to the low permeability of the varves the lake water should have a relatively minor influence as compared with the biological and chemical makeup of the sediments. The dominant transport mechanism in the sediment would be molecular diffusion with a net gradient from the sediment to the lake, not the other way round.

        • Paul Dennis
          Posted Dec 22, 2013 at 3:32 PM | Permalink

          Howard, Gonyo and co-workers don’t give any details about the carbonate phase in the Kepler Lake sediments other than stating they are inorganic precipitates. The carbonate is most likely seasonal (late spring-summer) and possibly associated with biologic activity drawing down CO2 (phytoplankton blooms) and/or water warming. This carbonate rains down to the sediment surface. I’m only speculating here. It’s almost certainly not an in-situ product within the sediment column where you describe the likely anaerobic processes. Again the paper doesn’t detail the sediment chemistry.

        • michael hart
          Posted Dec 22, 2013 at 7:09 PM | Permalink

          Howard, if there is any biological activity at all (even from dead or dormant bacteria, decaying vegetation etc), then I would say the likelihood of carbonic anhydrase activity is very high. It is ubiquitously expressed by all life forms in significant amounts, and requires no additional energy inputs or co-factors. Even stalagmites/tites will possess a ‘bio-film’. It is one of the so called “perfect enzymes” in that it raises the, actually quite slow, rate of CO2-bicarbonate equilibration up to the diffusion-limited maximum rate that appears to be so often assumed in carbon-cycle models. This is a rate acceleration of between 5 and 7 orders of magnitude. (Yes, about a million times faster, give or take a factor of 10 !)

        • Howard
          Posted Dec 25, 2013 at 5:47 PM | Permalink

          Thanks Michael. Another paper on Kepler Lake
          Indicates ostracods and not vegetative changes were most sensitive to climactic changes over a 2700 year long core. Ostracods could be a source of calcium carbonate which ranges from 68% to 88% and organic matter, which ranges from ~3% to >7% with very low bulk densities 0.2- to 0.66-g/cm3:
          I can’t imagine the difficulty working with this 6-foot core of goo. In any event, it looks like there is plenty of organic matter to create a reducing environment for anaerobic bacteria to live and dissolve the abundant calcite.

  18. MrPete
    Posted Dec 20, 2013 at 1:58 PM | Permalink

    For what it’s worth, it’s quite easy either using buttons on the CA-Assistant editor, or even manually, to make some of the text in this thread look much better.

    I did the following by typing HCO3- then highlighting the 3 and – separately and pressing subscript/superscript buttons: HCO3

    Or, type it manually: surround “3” with <sub> and </sub> and “-” with <sup> and </sup>

    (We also support LaTeX but most people don’t know much about that.)

  19. Ian H
    Posted Dec 20, 2013 at 8:08 PM | Permalink

    Could somebody who knows where to get the data and what they are doing please compare the November temperature data in Russia to the October data. As you know the recorded data shows that November in Russia was apparently unusually warm. but is this real or merely another instance of the Russians getting the month wrong


    Steve: I was curious as well – the month is right. “Hot spot” for Siberian locations in November doesn’t mean that the locations are “hot” – some sites have average November average temperature of (say) -21 deg C so -16 deg C. would be a 5 degree hot spot.

    • Ian H
      Posted Dec 20, 2013 at 8:42 PM | Permalink

      Thanks Steve. I’m much more inclined to trust that this is real now that I know you have audited it.

  20. Jeff Id
    Posted Dec 25, 2013 at 1:50 PM | Permalink

    We were hit by the same storm. We were without power for two days so we went to the folks. It put a dent in the Christmas planning but everything worked out with a bit of effort.

    When I was about 15, we had a huge overnight ice storm which put a layer of ice down on top of 6 inches of fresh snow that was so hard you couldn’t stomp your foot through it. The next day was cold and cloudless so every stick and twig sticking through the smooth snow surface looked like a chunk of clear glass. We used sticks to cut footholds into the local sled hill and took metal runner sleds to the top. The speed was incredible with shattering glass-like twigs flying over our heads as the fronts of the sleds cut through. I have never seen anything like it since but seeing the trees in that crystalline condition always reminds me of that day.

%d bloggers like this: