Warmest since….when?

There’s a new study out in GRL (press release), the third (at least) in the last couple of years using plant remains that have been newly uncovered by receding ice caps in Greenland and Baffin Island to estimate the most recent date at which temperatures must have been about the same as current. I’ve now seen four numerous notices on Twitter about it, and a popular media story on it has just popped up. That story contains the basic storyline statement: “…summer temperatures in the Canadian Arctic are higher today than they’ve been in at least 44,000 years, and are likely the warmest they’ve been in 120,000 years.

Interestingly, the author of that piece says “I’m a journalism professor and science journalist. Miller told me about his results last spring, and I’ve been waiting for his study to be published in a peer-reviewed journal before writing about them.” Well, given that he’s a journalist and had six months, it would have been nice if he’d asked some basic level investigative questions during that time instead of taking stated conclusions on faith.

When I first started reading this I was actually excited about novel evidence that might be relatively decisive regarding how unique current temperatures are in the arctic, relative to the Holocene/Pleistocene. Alas, as I continued reading that hope faded quickly and I now instead marvel at how the authors could state the conclusions they do, given the methods and data presented.

The study dates plant and lichen remains (using C-14 methods) found right at the edge of melting ice caps, on Baffin Island, Canada. Because ice caps expand and contract only by snow accretion and ice melting, and not much by movement, they don’t disturb the ground like glaciers, and it is therefore possible to find intact plant remains that have not been moved and mixed by the ice. Dating them gives the approximate last date that no ice covered that particular piece of ground, and therefore presumably, the date that air temperatures were last about the same as current.

The authors made over 300 specimen collections, with 145 datings, along a ca 1000 km transect on the east side of Baffin Island (see their Figure 1 below). Most specimens were mosses in the genus Polytrichum, with some fruticose lichens collected from rocks at a few higher elevation sites. Of the 145 dates, 135 (93%) dated to < 5000 kya, all of which were Polytrichum, (no lichens) and ranging from 236 to 4900 years, with a mean of 1243.

Miller etal Fig S1Figure 1. Eastern Canadian Arctic DEM showing collection sites, with an expanded map of the southern field area. Circles (color-coded by their 14C age) identify the 135 sites where rooted plants were collected the year of their exposure along receding ice cap margins that have been 14C dated 47 ka. The margins of the LIS at the LGM and 9 ka [A. S. Dyke, 2004] solid lines, and the 1 km LGM surface elevation contour (dotted line; [A.S. Dyke et al., 2002]) show that sites dated >47 ka were elevationally above the LIS and supported only local ice caps. P: Penny Ice Cap; B: Barnes Ice Cap. A tight cluster of 5 coastal sites (circled) exhibit anomalously low changes in their ELAs and are excluded from the regressions in Fig. 2A).

The other ten samples are much older, all > 47 ky. These samples include six lichen and four Polytrichum, collected from four sites. Those sites are all clustered nearby each other (diamond symbols in the figure), and the four moss samples are from one sampling location. The authors state in another publication describing their methods, that they prefer moss over other plant materials, such as those used in Greenland in a study published by others last year, because it has a relatively short life cycle, which presumably this keeps the dating tighter. Therefore, we can assume that at the three sites where lichens were used, there was no Polytrichum available, or they would have used it. Why is there no old moss there? Does using lichens instead, which differ in longevity, make a difference?

Note the open circle near the bottom of the figure, the one that contains the three orange circles. About this the authors state: “The five largest low-elevation outliers with respect to this trend are a tight cluster of coastal sites (circled in Fig. 1), where sea ice persists longest in summer [Moore, 2006]; hence, these sites are unlikely to be representative of the regional snowline.“, and that they “exhibit anomalously low changes in their ELAs [equilibrium snowline altitudes]“. In simple terms, these sites are unusual with regard to the snow and ice dynamics in the local area. Well, as the figure clearly shows, that area is very close to the cluster of sites (diamonds) where all the older samples were taken. Further, those samples are also 270 meters higher on average than the 135 samples aged at < 5 kya, (which have a mean elevation of 900m). So, it is entirely reasonable to ask whether the combination of the higher elevation and whatever factor(s) creates the unusual ice and snow persistence in the area, aren't partly or wholly responsible for the existence of these older samples, i.e. that they are local exceptions not reflective of the larger climatic phenomenon of interest.

But even if they are not, we have four sites clustered together at one end of the 1000km sampling transect that give very anomalous results relative to the 135 samples collected all along that transect. So why in the world are they focusing on those four sites, to the exclusion of the much more geographically extensive 135? How can the authors just blow past this fact without discussing why in any way? Reviewers, HELLO?? The authors then note the oldest radiocarbon ages are near the limit of the method and then hypothesize that these are therefore probably under-estimates. But by how much? This they address by referencing Greenland ice cores that (they state) show that one has to go back at least 110,000 years to get temperatures warm enough to have open ground. And the Discovery article and press release of course, mention this larger number.

The authors also include rough calculations of time required to melt a cap of 70m maximal thickness (assumed, based on hill diameter and stated ice weight constraints), assuming current, observed melt rates of 0.5 m/yr (getting the time required wrong in the process, stating it as 100 years, when it’s obviously 140). This means that it would have taken an extended period of warmth (140 yrs) to fully remove some ice caps of a given initial thickness, in the past. But we don’t know the original thicknesses of these caps before the dated plants grew there, so what’s the point? This is a confused discussion. The rest of the paper goes on to use estimated rates of snowline change since the mid 20th century to argue that “many climate models underestimate the magnitude of Arctic amplification, even on relatively long timescales.“, which of course implies that said amplification is going to be worse than predicted.

The authors conclude with this statement, which really pretty much gives away their bias:

“These findings add additional evidence to the growing consensus that anthropogenic emissions of greenhouse gases have now resulted in unprecedented recent summer warmth that is well outside the range of that attributable to natural climate variability.”

No it does not thank you very much. The study doesn’t even address natural variability. And I thought the consensus was supposedly already pretty much full grown…that’s what I’ve been hearing anyway. And lastly, an area of a few square miles on Baffin Island upon which the thesis rests, does not deserve the general phrase “Arctic Canada” used in the title.

Addendum: Reported ages for the 135 Polytrichum samples < 5000 years old:
236 358 436 441 441 454 460 531 538 544 544 555 557 558 559 559 559 562 563 564 569 569 577 579 587 597 599 605 609 638 642 642 644 646 648 651 651 651 651 653 653 655 655 656 656 657 657 658 658 658 658 693 693 705 707 708 710 712 713 714 721 724 735 738 751 834 901 904 904 912 1003 1058 1061 1062 1066 1067 1070 1071 1073 1073 1073 1073 1073 1077 1077 1081 1102 1104 1104 1104 1113 1154 1166 1169 1228 1331 1378 1515 1534 1549 1581 1601 1603 1607 1612 1630 1633 1635 1635 1640 1664 1671 1680 1694 1698 1766 1846 1984 2000 2296 2549 2631 2636 2993 3027 3425 3712 3741 3824 4082 4166 4193 4376 4398 4900

25 thoughts on “Warmest since….when?

  1. an area of a few square miles on Baffin Island upon which the thesis rests, does not deserve the general phrase “Arctic Canada” used in the title.

    Dr. Judith Curry has noted essentially the same on her “Climate Etc.” blog:

    how representative of the Arctic is their findings from Baffin Island? Well, it doesn’t even seem to be too representative even of Ellesmere Island and West Greenland.


    • That’s a relatively minor issue though. The really big problem with the paper is that they have all these 135 moss samples that are dated to a mean of ~ 1200 years. They never explain what those imply for the climates of the time. I don’t know know why people are not picking up on that.

  2. he rest of the paper goes on to use estimated rates of snowline change since the mid 20th century to argue that “many climate models underestimate the magnitude of Arctic amplification, even on relatively long timescales.“, which of course implies that said amplification is going to be worse than predicted.

    HUH? there statement is backwards

    More models Overestimate the amplification than underestimate it


    • I think there’s a couple of potential issues with those analyses: 1) use of linear trends; 2) cutting off at 1999/2000.

      Just focussing on Greenland the model map comparison for bcc-csm1-1 shows a huge 1900-1999 trend difference (>2ºC). Using data from climate explorer I found that this is mostly because the BEST linear trend is actually zero over this period. However, plotting the bcc-csm1-1 realisation anomalies against BEST Greenland up to 2012 and pinning 11-year averages together at 1900 reveals that the BEST end point in the 2000s actually gets relatively close to the model: ~0.3ºC less warming than the lowest model trend and ~0.6ºC less warming than the average of three runs.

      The present warmth of the 2000s is a long way above the linear trend through 1900-1999, and even the 1900-2012 trend. This means using a linear trend to represent temperature evolution effectively assumes Greenland warmth in the current decade is “noise” rather than signal, which I think is yet to be established.

      Looking at the whole Arctic and the scatter plot for 1950-2000 trends, I’ve found using 1950-2012 trends tells a different story with BEST amplification being about the same as the CMIP5 mean.

  3. Pingback: Climate Audit

  4. What’s your take on Judith’s blog?

    Recall that Kara/Laptev Seas is in the heart of the lynch pin region for the Stadium Wave. Note, Marcia Wyatt did not have any data sets from the Baffin/Ellesmere region of the Canadian Arctic to include in the stadium wave analysis.

    Clearly, there is substantial spatial variability of climate variability in the Arctic, with Opel et al. noting a see-saw between the Eurasian vs North American Arctic and seasonal variations (annual vs summer). Especially interesting is the absence of MWP and LIA in some of these high latitude data sets.

    In any event, extrapolating from one location in the Arctic to inferring Arctic-wide change is clearly not supported. It further seems that single locations don’t have a very large radius of influence, viz the differences between Baffin and Ellesmere.

    The natural internal variability in the Arctic seems to be an exceedingly complex dance between atmospheric circulations, sea ice, ocean circulations and ice sheet dynamics, on a range of timescales. We have some hints about how all this interacts, but much is unknown. In light of this, simplistic inferences about global warming in the Arctic seem unjustified.

  5. Having read Curry, McIntyre, and now Bouldin on this paper (but not the paper, alas) it looks like this one would be a good example for how to tell dodgy science from good science (as discussed with kch here on an earlier thread). The critical comments all focus on the seeming lack of curiosity of the authors as to how well the methodology supports the conclusions, with big gaping holes between what they found and what they claim. I submit that this continuity of argument is the best measure of good science: are alternative explanations for the data exhaustively described and robustly refuted in the paper? Can the reader follow a logical argument from the beginning to the end, or are there “just so” connections between data, correlation and causation?

    • Matt Skaggs -

      Richard Telford also has a number of interesting points to make wrt this paper – although I admittedly almost stopped reading at ‘fake climate skeptic’. At least he wears his bias openly, I guess.

      What struck me with Telford’s posts (as well as his comments at CA) are the number of assumptions he – and presumably the authors – are making about the timeline and physical processes of the moss growth and ice accumulation/retreat. I’m not sure they’re wrong, but the assumptions don’t seem to be explained so much as insisted upon.

      So is it dodgy science? I don’t really think it falls that far (much of the argument makes sense to me), but I’d certainly put it into the category of ‘unproven to me’ science. It definitely falls into the category of ‘overpromoted because we like the stated results’ science, but that’s a different thing.

    • I agree with you kch. Much of the study is in fact quite interesting. We just want to understand the chain of logic used so we can evaluate how reasonable their stated conclusions are.

  6. Pingback: More on Mosses (Miller et al 2013) | Musings on Quantitative Palaeoecology

  7. Based on Richard Telford’s comments at his blog, I have to admit that I may not have fully understood all the assumptions underlying this paper. I was sick and might have rushed both the reading and the interpretation thereof, and on top of that I somehow lost my copy of Miller etal (2013; Quaternary Science Reviews, 77, 239-245) in which they describe their methods in more detail and provide support for the findings of Lowell etal (2013)* on Greenland. When I’m feeling better I will have to look at it all again and write a second piece with any corrections if need be.

    *Lowell et al., 2013, Late Holocene expansion of Istorvet ice cap, Liverpool Land, east Greenland. Quaternary Science Reviews 63: 128-140.

  8. I’ve read the Miller et al paper and the comments by Jim, Steve McIntyre and Richard Telford. I have no issues with the data and the first order interpretation for these small plateau ice caps. The post 5000 year ice expansion and recent retreat seems to me to be very well recorded by the Holocene radiocarbon dates. I also think the observation that the ten samples (6 mosses, 4 lichens) with dates on the order of 40 to 50,000 years probably do represent Eemian (120,000 year) growth. In a timeline we then have Eemian moss growth implying no ice cover, followed by a long period of ice cover through the last glaciation through to the Holocene, when 5000 years bp again the plateau largely had no ice cover with a few small outliers of ice remaining. There was then ice advance through to the little ice age and then retreat in the modern period.
    So far, so straightforward. However, where I take issue with the interpretation are the assumptions that link ice cover directly to temperature. Miller et al are right to point out that summer insolation largely as a result of precession is lower than 5 or 6 thousand years ago. These are very interesting observations.
    However to then argue that the recent melting is a result of CO2 induced greenhouse warming that offsets this decline in summer insolation and not some other cause is a leap of faith. For example small changes in ice albedo as a result of dust etc. may have significant impacts on ice melting and retreat of these small plateau ice caps. Then there are local effects associated with local geography and micro climates. The authors themselves mention that local geography may result in areal variability in behaviour of the ice sheets. Imprinted on these may well be natural cycles. I think the assumptions behind the estimates of melting rates, thickness of the plateau ice sheets, relationship to temperature, rates of change of temperature, albedo changes (whether cloud or surface based) need scrutiny. Finally to then stretch the observation to the whole of the Canadian arctic is ridiculous.

  9. Any chance that it may have been warmer in the past, but for a shorter duration, whilst in the present, the warming may be less, but much longer lived?

    Could this scenario give the same results?

    • I think McIntyre suggested this at climateaudit, in the comments. The answer seems to be that the shape of the mountain prevents a cap thicker than 70 meters, which vanishes in an instant (a century or so) if the temperatures get above freezing.
      His other suggestion that I think is interesting is that this is an awful lot to base on a continental-scale map from Dyke 2002, that implies that this part of Baffin Island was not included in the Laurentide Ice Sheet. From the map shown there, it was just off the edge. Move that edge a drop seaward and the whole story-line vanishes: that could have been a whole lot thicker than 70 meters.

      As the author here says, there are a lot of assumptions building into this result. If it works out, it’s a remarkable feat of deduction. Any of the assumptions turn out not to be right, the result is gone.

    • The ice cap depth limit is not a finding of McIntyre–it’s explained right in the paper. The idea that they vanish quickly is up for debate also, because the 0.5 m/yr rate is simply the current rate, which may or may not have held in times past, including recent times.

      The idea that the ice was actually a lot deeper than ~70m would imply that the sampled areas were under a moving glacier, and we know from the condition of the mosses that that is pretty much ruled out.

    • Thanks. I hadn’t meant to imply that McIntyre supplied the answer; I had picked it up from the posts about the paper.
      About the moving glacier, I’m not familiar enough: What if there was an icecap on top of the mosses and then it all got covered by a glacier. Could the cap protect the mosses?

      I saw, by the way, Marcia Wyatt at Judith Curry’s site, who seems to be bringing a modification of McIntyre’s suggestion (as far as I can see). That is that the presence of the massive glacier nearby affects the weather in a big way, and even higher solar insolation of the time might not be enough to melt the cap. This is a suggestion of Koerner (or follows from one?)

    • Or alternatively, that they are basing their calculations of ice cap shrinkage rates in the past, on the present (not past century) rate of melt, which is higher than the rate computed over the last 100-150 years, and which may also be impacted by black carbon on the surface.

  10. A question for anyone interested in pointing me at the answer:

    In the comments at Climate Audit, Steve McIntyre shows a map from Dyke et al 2002 of the Laurentide Ice Sheet which indicates that the sheet maxed out at over 2000m depth in the very near area discussed in Miller’s paper. What process took place that allowed that depth of ice to melt completely away, while a ~70m thick ice cap less than 300 km away survived to only melt now? What am I missing here?

    • The glacier would be flowing from high to low elevations, with enormous ice losses at sea level (where it’s much warmer) not compensated by ice gain at the higher elevations. The ice caps conversely, are stationary (and much above sea level, and hence colder) so their ice volume changes only from the combined effects of winter snowfall and summer temps.

    • Okay, thanks – that’s pretty much how I understood the overall retreat of the LIS. But…

      The specific process I’m not getting is the long-term survival of such a (relatively) small ice cap from which the moss emerged. The preservation of the moss seems to imply that the ice cap did not at any time grow deep enough to start glacial movement, or it would have presumably scraped the ground clean of moss. At the same time, the current exposure of the moss implies that the ice cap has remained relatively intact for the last 44ky or more.

      I’m finding this, in combination with the continental-scale climactic changes that must have occurred with the formation and retreat of the LIS, to be hard to accept as anything other than an anomaly – a very long term maintenance of a fairly narrow range of temperature and precipitation, only to end now.

      Not that there aren’t lessons to be learned from such an anomaly, but is it really possible extrapolate out to ‘Unprecedented recent summer warmth in Arctic Canada’ on the basis of it?

  11. This seems to be a lot of brouhaha about a paper that hasn’t even been copy-edited yet, but since I am snowbound in the Canadian Subarctic, I admit to a certain level of interest.

    Miller et al. GRL in press: “Key Points: This study provides the first direct evidence that recent summer temperatures in the Eastern Canadian Arctic, now exceed those of any century in the Holocene, including the Holocene Thermal Maximum, when regional summer insolation was 9% greater than at present. We also show that the inferred magnitude of regional late Holocene cooling substantially exceeds that simulated by the CMIP5 climate models.”

    This is less inflammatory than the press release and the title of the paper (which seems to be competing with the recent and much more modest 600 year claim in Tingley & Huybers Nature letter doi:10.1038/nature11969). Still, Miller et al. do seem to be protesting too much. A few ancient moss samples from one small area do not make convincing ‘first direct evidence’ of anything. Given the complex topography and climate of the area (Telford’s simple diagrams appreciated – but misleading), they are no more than possibly Eemian relics (assuming the C14 work is valid) exposed by ice retreating for an unknown reason. If they were so important, then I would expect Nature to be where this bloody shirt was being waved, not a journal with an IF of less than 4 (although I find IF of no use in predicting what papers I think are interesting – it is a good guide to what the chatterati will be going on about).

    Well, it is interesting seeing what a house of cards can be built on such a flimsy premise if it fits the meme and the degrees to which the Watts and Telfords will talk past each other, but the only good thing I can see in this paper is that they were looking at the mosses and lichens. I don’t suppose they had much choice, but still nice.

    • Just want to note here, for those who’ve seen their comments not immediately appear: that’s due to the setting whereby any comment coming from an IP address that has previously had a comment approved, is automatically approved, but otherwise is held in the queue (and thus not approved until I get to it, which can sometimes be a while–as in this case).

    • although I find IF of no use in predicting what papers I think are interesting – it is a good guide to what the chatterati will be going on about

      Indeed, nor with respect to quality. It’s illuminating when one finds papers in the Journal of Great Lakes Research that are far better written, and more importantly, which directly and clearly discuss all sources of error and variation that could potentially have affected their primary conclusions, than do papers in GRL, Nature etc. IF’s are essentially worthless metrics IMO, very similar in root concept to emphases on the importance of “consensus” as a criterion as to who is right.

      The “talking past each other” is now apparently mandatory at most climate change discussions on the web as far as I can tell. And distorting the other person’s argument of course.

Comments are closed.