Guest Post by Willis Eschenbach
I have to learn to keep my blood pressure down … this new paper, “Abrupt onset of the Little Ice Age triggered by volcanism and sustained by sea-ice/ocean feedbacks“, hereinafter M2012, has me shaking my head. It has gotten favorable reports in the scientific blogs … I don’t see it at all. Anthony provides an overview of the paper here.
First, the authors say:
Here we present precisely dated records of ice-cap growth from Arctic Canada and Iceland showing that LIA summer cold and ice growth began abruptly between 1275 and 1300 AD, followed by a substantial intensification 1430– 1455 AD. Intervals of sudden ice growth coincide with two of the most volcanically perturbed half centuries of the past millennium.
OK, precisely dated records show that big volcanoes equals Little Ice Ages. Peaks in ice growth coincide with volcanoes. Got it. Then in the paper they discuss the later interval of “sudden ice growth”. They start by saying …
The PDF peak [in ice growth] between 1430 and 1455 AD corresponds with a large eruption in 1452 AD …
What’s the problem with these claims? Figure 1, which is from their study with a couple of my annotations, shows the problem …
Figure 1. Original Caption Figure 2. … (b) Global stratospheric sulfate aerosol loadings [Gao et al., 2008]. (c) Ice cap expansion dates based on a composite of 94 Arctic Canada calibrated 14C PDFs (probability distribution functions). I added the vertical red line down from the top of the “D” panel that shows the full size of the sulfates from the eruption in 1258 (258 teragrams [megatonnes]). The vertical blue line, also added, indicates the timing of the following large eruption in 1455.
I get nervous when people cut off important data in a graph, it’s a bad sign regarding their transparency … but I digress …
I always look for alternative ways to verify what the authors are showing. In this case, the GAO et al 2008 aerosol loadings shown in figure 1(B) are calculated loadings using a record of the volcanoes and a climate model. Me, I always prefer actual data. Fortunately, we have very accurate data thanks to the ice core record from a place with the lovely name of Dronning Maud Land. You may not recognize it by its Norwegian name, but when I say “Queen Maud Land”, everyone knows where that is … well, everyone but me, I had to look it up …
Figure 2. Location of Dronning Maud Land, home of ice. And ice cores.
Ice cores record how much sulfate has fallen on the ice during past years. Sulfate comes from volcanoes, and is ejected high into the stratosphere. From there it is mixed worldwide, and eventually it settles out on the ice. The sulfate record from two different ice cores in Dronning Maud Land agree to within a couple of years, so we can have confidence that they are accurate.
Next, before I go further, what is the “probability density function (PDF)” that the paper uses? It is a function that gives the probability of an event occurring in a certain year. For example, carbon-14 dating of some dead moss might give the date it died as say 1135. Are we sure it died in exactly 1135? No way, that’s just the most probable value. It might have died in 1134, or 1136. It might also have died in 1130 or 1140, but the probability of it being either of those years is much lower than the probability that the date is actually 1135. The probability density function is the function that gives us the probability of the event actually occurring in each years. Typically it looks like the famous “bell curve” or Gaussian curve, peaked in the middle and fading to zero on either side. It may be asymmetrical, with different probabilities that the event is before or after the most probable date. It is a good way to aggregate data
With that as prologue, here is the overview of the two records. One is the ice expansion record from the M2012 paper. The other is the volcanic sulfate record from the Maud Dronning Land ice cores.
Figure 3. Volcanic sulfate records from Maud Dronning Land (blue and green) and the ice cap expansion records from Baffin Island (purple line). The PDF values are the probability percentages multiplied by 100, so for example if the scale reads “400” that means 40% (0.40).
Right away you can see some curious things. There is a large expansion of the ice cap (increasing purple line) in the century from 900 to 1000, but nary a volcano in sight. They say in the paper that “cold summers can be maintained by sea-ice/ocean feedbacks long after volcanic aerosols are removed …”, but what started and maintained the cold summers from 900 to 1000?
Then there’s the claim that the intervals of sudden ice growth in 1280 and 1435 occur during “two of the most volcanically perturbed half centuries of the past millennium” … I’ll buy that for the year 1280, but 1435? One lousy volcano in the half century around 1435, it wasn’t even as “volcanically perturbed” as the last half of the 20th century or the first half of the 19th century.
Intrigued by these problems with their claims, I looked closer. Figure 4 shows a closeup of the time in question:
Figure 4. As in Figure 3, volcanic sulfate records from Maud Dronning Land (blue and green) and the ice cap expansion records from Baffin Island (purple line).
More oddities. First, the expansion of the ice cap started in 1215, about 45 years before the eruption in 1258. Then in 1250, the rate of ice cap expansion increased, almost a decade before the eruption. And while you would expect an immediate increase in the rate of ice cap expansion, the increase doesn’t begin until about [1470].
But that’s nothing compared to the other end of the period. The peak ice cap expansion occurs in 1435, a full two decades before the eruption in [1455]. Nor does the eruption speed up the ice cap expansion. In fact, the expansion slows markedly after the 1455 eruption.
Now, you may recall that I quoted the start of a sentence above, which said:
The PDF peak between 1430 and 1455 AD corresponds with a large eruption in 1452 AD …
Um … well … they are being most expansive with their claim that the 1435 peak and the eruption “correspond”. The volcano is well after the expansion in ice area. How do they explain this?
Well, the sentence goes on to say:
… although the ages of the three largest 5-year bins appear to precede the eruption date. In contrast to the earlier 13th Century peak, the second PDF peak occurs at the end of a 150-year interval of variable but falling snowline (Figure 2c), raising the possibility that the PDF peak plausibly reflects a brief natural episode of summer cold that preceded the large 1452 AD eruption. Alternatively, the apparent lead of kill dates with respect to the 1452 eruption may be a consequence of combined measurement and calibration uncertainties.
To me, that’s special pleading. Not only that, but it destroys their entire case. Here’s why:
If the 1435 peak “plausibly reflects a brief natural episode”, then why should we believe the much smaller 1280 peak is not just another “brief natural episode”?
Alternatively, if the timing of their “precisely dated” 1435 record is really off by twenty years due to “combined measurement and calibration uncertainties”, then why on earth should we believe the timing of the “precisely dated” 1280 peak?
I’m sorry, but I just don’t see the evidence that volcanoes had anything to do with the changes in the Baffin Island ice cap. And their whole sea/ice feedback claim? I note that the claim is supported by … well … I fear all it is supported by is models all the way down.
w.
PS—An oddity. The 1258 volcanic eruption was the largest in the last 2,000 years … and as far as I can determine, nobody knows where it occurred.
DATA: All data used in this post is available here as a comma-separated (CSV) file.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
“First, the expansion of the ice cap started in 1215, about 45 years before the eruption in 1258. ”
err no. the PDF starts to change in 1215. entirely different thing than you describe.
The ice expansion does not start in 1215. The PDF, the probability, starts to change in 1215.
This refutation is pretty weak. The paper undoubtedly has problems, but Willis is being a bit lazy with this write up.
“but what started and maintained the cold summers from 900 to 1000?”
This is an interesting question, the answer to which may support or refute their assertions, but merely asking the question does nothing. They don’t seem to be saying ice cap expansion can only be caused by volcanoes, merely that the two in question were. Yes finding what caused the earlier expansion might reveal the cause for the later two, or it might not. Willis’ comment does nothing to answer this question.
“Then there’s the claim that the intervals of sudden ice growth in 1280 and 1435 occur during “two of the most volcanically perturbed half centuries of the past millennium” … I’ll buy that for the year 1280, but 1435? One lousy volcano in the half century around 1435, it wasn’t even as “volcanically perturbed” as the last half of the 20th century or the first half of the 19th century.”
Again, the writers asserted that the time periods were two among other active periods. NOT the two most but two of the most. I would think that of the 24 half centuries (chronologically) having 2 of the top 4 or 5 most active would make this a valid statement. I don’t know if both periods are in the top 4 or 5, but Willis offers no evidence to the contrary. He merely says he doesn’t like it because of two periods that are more active than one of the authors.
Willis’ other points are a little better, though with the typo’s it is not clear to me if he is clear on the original paper.
I didn’t read the original paper, but I do know that nothing in this article is particularly damning of it. Very weak tea indeed.
Steven Mosher says:
April 13, 2012 at 8:41 am
““First, the expansion of the ice cap started in 1215, about 45 years before the eruption in 1258. ”
err no. the PDF starts to change in 1215. entirely different thing than you describe.
The ice expansion does not start in 1215. The PDF, the probability, starts to change in 1215.”
Nice catch, Steve. The paper is therefore unfalsifiable, and therefore unscientific.
Doug Proctor says:
April 13, 2012 at 7:59 am
Is it possible that the internet, blogging and, for the first time, a large number of non-academic, but technically proficient readers, are revealing how much shoddy scientific work has always been published?….
_____________________
You might want to read some of Dr. Scott Armstrong’s Papers:
“Bafflegab Pays,” 1980: http://marketing.wharton.upenn.edu/ideas/pdf/armstrong2/bafflegab.pdf
This is my favorite one on experts. “Incomprehensible, You Say? Brilliant!” 1980 http://marketing.wharton.upenn.edu/ideas/pdf/Armstrong/Mass%20Media/Bulletin%201980.pdf
“Forecasting: Of Suckers and Seers” 1985: http://marketing.wharton.upenn.edu/ideas/pdf/Armstrong/Mass%20Media/The%20Press%201985.pdf
He has a whole bunch of interesting papers on research and peer review: (see bottom of page) http://marketing.wharton.upenn.edu/people/faculty/Armstrong/Mass_Media.cfm#knowledge
elmer says:
April 13, 2012 at 8:13 am
You can’t “precisely date” ice cores. Tree rings are annual rings, ice core rings are not. It snows all year round in the Arctic and Antarctic.
__________________________________
They use O18 to date ice core and other proxies This is why Pat Franks article was so fiercely contested http://wattsupwiththat.com/2012/04/03/proxy-science-and-proxy-pseudo-science/
The ice core dating looks reasonably accurate for ice cores… maybe. That is if there was snowfall and it did not sublimate. http://www.iceandclimate.nbi.ku.dk/research/strat_dating/annual_layer_count/ice_core_dating/
In the interior of Antarctica precipitation is as low as 50 mm per year, while in coastal regions it can be as high as 250 mm per year. It is also very windy. That is why I say “maybe”
Good article. As I was just noting elsewhere, Bond et al. 2001 note, regarding cold events like the Little Ice Age:
“Our correlations are evidence, therefore, that over the last 12,000 years virtually every centennial time scale increase in drift ice documented in our North Atlantic records was tied to a distinct interval of variable and, overall, reduced solar output.”
http://one.geol.umd.edu/www/preprints/Bond_et_al.pdf
The alarmist side has tried to get rid of the Medieval Warm Period and to pretend the Little Ice Age was caused by anything other than solar variation primarily. In fact, the wikipedia article on the LIA suggests human activity affecting CO2 levels as a plausible cause, which would be comedy except some people actually fall for that; CO2 levels meanwhile varied by only a handful of ppm like this illustrates:
ftp://ftp.ncdc.noaa.gov/pub/data/paleo/icecore/antarctica/epica_domec/edc-co2.txt
Their tactics break down extra much when one considers more than the past millenium but also the rest of the Holocene, where correlation with solar activity (not as the sole influence but often a predominant one along with cosmic rays affected by it) occurs time and time again.
The more I think about this paper, the more it appears to be similar to the convoluted logic and date manipulation used by Immanuel Velikovsky in his seminal work, “Worlds in Collision”. In it he freely changed dates of various catastrophes around the world to support his contention that the ‘exodus’ events were the result of a comet colliding with the earth. Their reasoning bears striking similarities to the tortured logic of Velikovsky.
Pamela Gray, I get up early to fire up my wood stove too.
The chronology of volcanic and climatic events in this paper is based on 14C dating of tundra vegetation, yet the authors are talking about accuracies of only a few years. There is no way you can get that kind of accuracy from radiocarbon dates. Radiocarbon ages are expressed in radiocarbon years before present. A radiocarbon year is not the same as a calendar year because of variation in the production rate of 14C in the atmosphere caused by variation in the neutron flux that creates 14C from nitrogen. So any radiocarbon date has to be adjusted by means of a calibration curve to convert radiocarbon years to calendar years. The problem is that the calibration curve to make this conversion is not a straight line–it’s a wiggly line with more than one possible intercept corresponding to sevedral possible dates, (i.e., a 14C date might correspond to anywhere from 1-3 or more calendar dates, rather than one date). So I pulled up the standard 14C calibration curve to see what kind of accuracy one could claim for ages in the past 1200 years. For example, if the measured 14C age was 600 radiocarbon years (i.e, around 1400 AD), the calibrated calendar age intercept could be any of three ages, 640, 590, or 560 calendar years (or between the calendar years of 1372 to 1452). . For a 14C age of 350 radiocarebon years, the possible calendar year intercepts vary from 450 to 300 calendar years (or 1562 t0 1712 AD). Yet the authors of this paper are attempting to make correlations between volcanic and climatic events within a matter of only a few years. The bottom line here is that 14C dates cannot possibly provide the accuracy claimed in this paper.
In the 21 year period 1990-2010 inclusive, while Arctic Ice was steadily declining in the nearby Kamchatka peninsula and the Kurile Islands there were on average about 2.5 volcanic eruptions annually.
Kamchatka Volcanoes
Bezymianny: 2010, 2009, 2008, 2007, 2000-05, 1998, 1997, 1996-97, 1994-95, 1993-94, 1992, 1991, 1990
Klyuchevskoy: 2010, 2009, 2007, 2005, 2002-04, 2000, 1998, 1996-97, 1994-95, 1993, 1992, 1991
Sheveluch: 2000-10, 1998, 1997, 1993-94, 1991, 1990
Avachinsky: 2001, 1991
Karymsky: 1996-2009
Koryaksky: 2009, 2008
Mutnovsky: 2000
Kurile Islands Volcanoes
2010, 2x 2009, 2008, 2007, 2×2005, 2003, 2002, 1999, 1997, 1996,1991
These eruptions caused some of the ‘Sudden Stratospheric Warming’ ( http://hal.archives-ouvertes.fr/docs/00/56/34/77/PDF/SSW.pdf ) in the winter months, but an AGW advocate would solemnly testify that it is the CO2 which has saved the Northern Hemisphere from the onset of another Little Ice Age
If there was causation – “volcanos cause ice increase”, then there should have been SOME ice growth around the double erruptions recorded about 1825. Zero, zip, nada. Theory falsified.
As well as this paper we have had Booth et al claiming that most unforced variation in the measured temperature record was due to aerosols. Is there a movement with the climate science community to use aerosols to explain what they cannot otherwise explain. Perhaps it is time for paper claiming that the Yellowstone Erruption of 2015 caused the levelling off of temperature in the early 21st centruy.
Gail Combs says:
April 13, 2012 at 5:01 am
Interesting, although it leans heavily on Michael Mann’s ridiculous work, Mann, M.E., Fuentes, J.D. and Rutherford, S., 2012, Underestimation of volcanic cooling in tree-ring-based reconstructions of hemispheric temperatures. Nature Geoscience, doi:10.1038/ngeo1394. Even my stomach is not strong enough for that one …
w.
elmer says:
April 13, 2012 at 8:13 am
SOURCE: UCAR
Google is your friend …
w.
Arno Arrak says:
April 13, 2012 at 8:28 am
It’s not that the volcanic cooling doesn’t exist. It’s just that it is immediately offset by shifts in the clouds and the albedo. The earth’s temperature is thermostatically controlled through a host of mechanisms, so volcanoes just don’t end up doing much.
w.
Steven Mosher says:
April 13, 2012 at 8:41 am
Thanks, Steven, but that’s a difference that doesn’t make a difference, because the PDF values for individual events are tall and don’t last long … otherwise we wouldn’t see the peaks, which are very narrow (five years or so). So within a couple of years of 1215 the probability is approximately 1 that the expansion has started …
By the way, does anyone have a copy of the Supplementary Information for the paper?
w.
Bill says:
April 13, 2012 at 8:51 am
Oh, go away with accusations of “laziness”. I’m the only one who has done a refutation at all.
In addition, I show that the volcano comes after the ice age it’s supposed to cause, hardly a “weak” refutation. You’re just trying to salvage your chestnuts from the fire. Go insult someone else, your science-fu is lacking. Your most hilarious statement is:
Typical. You are too lazy to read the paper, but you are quite willing to accuse me of laziness, and to give us your uninformed opinion.
w.
Paul Wanamaker says:
April 13, 2012 at 9:51 am
Not true at all. Note the right side of the graph in Figure 1 says “continuous ice cover”. The study was done on Baffin Island (off of Greenland). Once the ice cap was fully established, more cold (from volcanoes or any source) couldn’t expand it any further.
w.
Don Easterbrook says:
April 13, 2012 at 9:36 am
…….
Absolutely agree.
(Willis Eschenbach says: April 13, 2012 at 10:28 am)
……….
Arctic ice 14C dating is prone to significant errors, radioactive deposition are subject to atmospheric conditions as well as diffusion process within the ice.
Here is graphic illustration how a NASA’s scientist may have made errors in dating the solar activity during the Dalton Minimum 1800-1820.
http://www.vukcevic.talktalk.net/DM.htm
Steve Mosher: data is available, no code required.
Putting PDF (Probability Density Function) on the Y-axis is pure balderdash.
It is to imply some significance with high values with meaningful events above the noise.
A PDF is a function with many Ys for many Xs (or Ts)/ So exactly what are they plotting vs time? Max value Y at it’s X? So we see as visually important high precision of insignificant events and as unimportant where there is wide uncertainty of significant events. Backwards! We should also visualize that every event in the data results in a PDF and the the events must superimpose at each and every point in time. Finally, we are to believe that the uncertainties contributing to each PDF’s event timing are independent of each other, when in fact each event is most certainly shares common elements of uncertainty; if one event has +/- 10 years of uncertainty, I’d bet the neighboring event in time has the same, and linked, uncertainty.
Far better to plot the aerosol loading on Y and express timing PDF as a 90% confidence error bar in the X direction.
Figure 3. Volcanic sulfate records from Maud Dronning Land (blue and green) and the ice cap expansion records from Baffin Island (purple line). The PDF values are the probability percentages multiplied by 100, so for example if the scale reads “400″ that means 40% (0.40).
That would be times 1000 to change 0.40 (40%) to 400.
Y-axis label says “…PDF (percent * 10)”
Resuming reading…
Re: Willis 10:41am
I didn’t mean criticism of your article as criticism of you personally. I chose “lazy” over ham-handed or clumsy, but perhaps you prefer those descriptors. If you’re going to go to the trouble of doing something, you might as well do it properly.
I didn’t read the article in question. I did read your article. Yours was unpersuasive from a logical point of view. I make no representation about the validity of the original piece. Don Easterbrook has an illuminating comment upthread that seemed to pretty well sh%t-can the piece for the carbon dating error margins.
You make no reply to my substantive criticisms of your piece. Do you take them as granted?
There were enough apparent typo’s to suggest that the article was not proofread. If not, doing so in the future may help clarify presentation and formulation.
My favorite post of yours over the years was on the Fair Weather Gale; well written, entertaining and informative. Many other posts have been incisive and penetrating. This one is not. Sorry if that bums you out.
Ice peaks without eruptions, ice peaks with eruptions, eruptions without icepeaks. Not a strong covariation between ice peaks and eruptions.
Willis, thanks again.
Anopheles says:
April 13, 2012 at 5:30 am
I am shocked that no-one can see how ice causes volcanoes from this data.
That was my thought when I first encountered this back in January. The “data” they have assembled more strongly implies that increasing isostacy from expanding ice triggers significant volcanism, but even at that it wouldn’t be much better than a vague suggestion, but still better than the chicken before the egg analysis they present.
It seems like I remember reading letters to the editor to journals from old publications that read like this. IOW someone would write a paper, someone would evicserate it, and then folks would either also eviciserate the original paper or they would savage the critics methodology. When did the journals start sidelining the critical responses and forcing folks to find sites like this one to get both sides of a story? Anybody can be guilty of expectational bias, but there should be an expectation that you might get ridiculed if you fell into that trap.