Guest Post by Willis Eschenbach
The quote above is from Lord of the Rings, an exchange between Gollum and Smeagol, and it encapsulates my latest results from looking into the Shakun 2012 paper, “Global warming preceded by increasing carbon dioxide concentrations during the last deglaciation” (paywalled, at Nature hereinafter Shakun2012). I discussed the paper in my post “Dr. Munchausen Explains Science By Proxy“. Please see that post for the underlying concepts and citations.
When I left off in that post of mine, I had investigated each of the 80 proxies used in Shakun2012. I plotted them all, and I compared them to the CO2 record used in their paper. I showed there was no way that the proxies could support the title of the paper. Figure 1 recaps that result, showing the difficulty of establishing whether CO2 leads or lags the warming.
Figure 1. All proxies (green dots) from Shakun2012 (Excel spreadsheet). CO2 values digitized from Shakun 2012 Figure 1a. There is pretty good agreement between the warming and the changes in CO2.
Note that the proxies say the earth generally warmed from the last ice age, starting somewhere about 15,000 BC, and the warming lasted until about 9,000 BC. Since then, the proxies have the greatest agreement (darkest green). They say that the globe generally cooled over the length of the Holocene, the current interglacial period since the last ice age.
Today I was thinking about that single record that they used for the CO2 changes. I got to wondering what other ice core CO2 records might show about the change in CO2. So I went and downloaded every ice core CO2 record that I could find that covered the time period 26,000 BC to modern times. I found a number of ice core records that cover the period.
Then I collated all of them in Excel, saved them as a CSV file, opened the file in R, and plotted every ice core CO2 record that covered the record from 26,000 BC up to the present. I standardized them over the same period covered by the Shakun2012 CO2 data. There was excellent agreement between the Shakun2012 data and the ice core records I had downloaded … but there was also a surprise.
Figure 2 shows the surprise …
Figure 2. As in Figure 1. Black circles show Shakun2012 CO2. Additional colored dots show the ice core CO2 records which have data from 26,000 BC to the present.
Dang, I didn’t expect that rise in CO2 that started about 6,000 BC. I do love climate science, it always surprises me … but the big surprise was not what the ice core records showed. It was what the Shakun2012 authors didn’t show.
I’m sure you can see just what those bad-boy scientists have done. Look how they have cut the modern end of the ice core CO2 record short, right at the time when CO2 started to rise again …
I leave the readers to consider the fact that for most of the Holocene, eight centuries millennia or so, half a dozen different ice core records say that CO2 levels were rising pretty fast by geological standards … and despite that, the temperatures have been dropping over the last eight millennia …
And I leave everyone to ponder how far climate “science” has fallen, that a tricksy study of this nature can be published in Nature, and can get touted around the world as being strong support for the AGW hypothesis. The only thing this study supports is the need for better peer review, and at a more basic level, better science education.
My best to all, stay skeptical,
w.
Source data:
ICE CORE CO2 DATA: All ice core CO2 data are from the NOAA Paleoclimatology site, the “Ice Core Gateway” page, in the section “Gases”.
[UPDATE] A hat tip to Jostein, in the comments he points out that the original Shakun Nature paper is here (pdf).
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Keith says:
April 8, 2012 at 1:19 am
Good question, Keith. Their graphs are in years before present (1950). My graphs are in BC / AD, so the offset is about two thousand years.
Thanks,
w.
Shakun’s PhD thesis is here (3.7 MB pdf file). None of the figures in his thesis extend to times more recent than -6500 years ago.
Shakun’s Figure 4.2 is the money figure. This is from Figure 4.2 Legend: “(c) Atmospheric CO2 from the Dome C ice core (Monnin et al., 2001) on the Lemieux-Dudon et al. (2010) age model, which is the most recent and likely most accurate chronology. (d) Global mean temperature anomalies with 1σ errors due to chronological and proxy calibration uncertainties estimated from 1000 Monte Carlo simulations.” Later Figures showing CO2 trends refer back to Figure 4.2.
The Dome C ice core data shown in the Monnin, et al, (2001) Science 291, 112-114 (abstract page here, Figure only goes back to -9000 years before the present. Here’s the Monnin 2001 data archived at NCDC. The latest date is -9067 years BP. So, one wonders from where Shakun got the extra 2500 years of CO2 data shown in his Figures.
It turns out he must have gotten it here, which is another NCDC EPICA Dome C ice core data set that provides the Holocene (0-11KYrBP) Atmospheric CO2 all the way up to 435 years BP. It’s from Flückiger et al. 2002 “High-resolution Holocene N2O ice core record and its relationship with CH4 and CO2” Global Biogeochem. Cycles, 16(1), 1010, doi:10.1029/2001GB001417.
Between 10977 and 9899 years BP, the Monnin and Flückiger data sets overlap point-by-point, so they record the same data. And the Flückiger et al., 2002 data set reproduces Willis’ CO2 plot exactly. CO2 begins a steady rise at 6500 years BP, just where Shakun’s T:CO2 plot ends.
So, at the very least, Shakun used ~2500 years of the Flückiger data set without acknowledging it. And having that data set, he must have known about the rise in CO2 after -6500 years BP. And then truncated the CO2 record there, along with the temperature record.
In the text prior to his thesis Figure 4.2, Shakun wrote about the temperatures, “Temperature uncertainties (1 sigma) were taken to be 1.7C for TEX86 (Kim et al., 2008), 1.5 C for pollen and bioassemblages, 10% for ice cores (Jouzel et al., 2003), and are based on quoted calibration errors for Mg/Ca (Anand et al., 2003) and alkenones (Müller et al., 1998). We assumed temperature errors are random through time and among records, which is a maximally conservative approach as errors are likely autocorrelated..”
For those who don’t remember, random errors decrease by 1/sqrtN, where “N” is the number of records. This explains why the Shakun proxy temperature errors in his thesis Figure 4.2 are only (+/-)0.25 C, while the per-proxy errors are about (+/-)1.5 C. They’ve been averaged away as random uncorrelated error.
I assessed Jouzel’s Greenland ice core record. It displays less point scatter than any of his other published data sets, and displays at least two simultaneous error modes. This implies the error is systematic. The standard deviation of that error is 1-sigma=(+/-)0.5 C. As a systematic error, it does not decrease as 1/sqrtN.
Anand, et al., 2003, report a Ca/Mg temperature accuracy limit of (+/-)1.3 C. That’s accuracy, not precision. Accuracy limits do not decrease as 1/sqrtN. They propagate as sqrt[(sum of errors)^2].
Jouzel + Anand together, given equal weight, produce a joint accuracy limit of 1-sigma=(+/-)1.4 C.
Shakun’s Nature Figures 2&3 show proxy temperature uncertainty limits of (+/-)0.1 C over the entire 16,500 years of the record. This (+/-)0.1 C error is smaller than even in Shakun’s thesis, and about 1/10th of the limit of experimental accuracy in the the dO-18 and Ca/Mg temperatures.
The Figures live or die on that low level of T uncertainty. Increase it to a much more realistic 1-sigma=(+/-)1C and T overlaps CO2 across the full record. The entire lead-lag, T:CO2 relationship is lost.
So, Willis has again caught a good one. We have million dollar claims resting on truncated data sets and neglected error bars. Once again a triumph of climate-science and a bell-weather of Nature’s competence.
j ferguson says:
April 8, 2012 at 2:01 am
They claim to show something that refutes an earlier understanding, just as the title says.
Here’s what I observe.
1. The proxy data for all 80 proxies is given in their spreadsheet. It runs up as late in some cases as the 20th century.
2. The ice core data for the CO2 shown in their graphs is not given in their spreadsheet.
3. Despite the stated focus of the study on the glacial/interglacial transition, their CO2 data runs for 5,000 years beyond the end of that transition.
4. However, as soon as the CO2 data starts to diverge from the temperature data, it is cut off.
You are welcome to believe that this is just a cosmic coincidence, that it just shows that the gods have a sense of humor.
Me, I’ve been around agenda-driven climate scientists far too long to believe that. But don’t let me stop you.
Here’s the problem, j. If they show the full CO2 record, then they have to explain why the CO2 did all the things they claim it did during the transition, but then changed its ways after they cut off the record.
Look at Figure 2. For eight thousand years, CO2 has been rising fairly rapidly, and the global temperature has been falling. If they show that, their whole argument is threatened.
Now it is possible they just happened to cut the CO2 record off at that time, and it wasn’t related to the fact that showing it would torpedo their argument, and it just happened to be cut at the very time when the CO2 headed upwards … but my money’s betting the other way.
w.
Another excellent post Willis.
I think the key to examining the validity of their result is to look at the Younger-Dryas event. This instantly stuck me on looking at what they presented where it was occurring at a very different point in time to the Vostok record. Since they were not calling into doubt the timescale of Vostok this implies that they thought Y-D happened over 1000y later in Antarctica. So improbably that they do not suggest it overtly.
Now, seeing your digitisation of their reconstruction on top of the proxy data the deception becomes even clearer. They seem to be using processing with such a heavy damping (or there is so little resolution in the data themselves) that they have completely removed the dip in the Y-D event. This was a huge global event and the obvious key to resolving the key issue they purport to be examining.
The CO2 seems to have risen before the exit from Y-D because the data if failing to resolve it.
wmconnolley says:
April 8, 2012 at 12:48 am
—
I’ve had the joy of being able to work with mustelids for seven years- hence the name I use here- and I have to say: they are Smarter than you.
Another excellent analysis Willis. Keep slicing.
”consider the fact that for most of the Holocene, eight centuries millennia or so, half a dozen different ice core records say that CO2 levels were rising pretty fast by geological standards … and despite that, the temperatures have been dropping over the last eight millennia …”
The opposite effect in the real atmosphere that CO2 is alleged to have in theory. Is that another aspect of the CAGW hypothesis crumbling that we hear? Yup.
@Willis
If this doesn’t gain traction I won’t beat it to death, but over at http://wattsupwiththat.com/2012/04/05/more-glacial-junk-science-journalism/#comment-948766 I have been calculating the absorption of CO2 by meltwater from ice sheets for KR and Daveo. If the atmospheric CO2 level has been approximately stable for the past 8000 years and the temperature has been dropping, then there may have been a continuous increase in the amount of ice volume (a difficult number to find).
When the sea evaporates and the precipitation freezes, CO2 is expelled into the air. As the oceans cool on a large scale they pick up more CO2 so there might an approximate balance between the CO2 expelled by new ice and the drawdown of the extra CO2 by the (cooling) oceans. This would give a constant CO2 level in a cooling world if ice volume continued to increase. It seems your plot shows that possibility.
On a shorter time scale, the expulsion of CO2 by snow and ice in winter would be countered by spring-summer melting and rapid uptake. This should show up as a cyclical variation in the CO2 concentration in the NH. Lo and behold it does. As the snow and ice cover in the SH is much less variable, the CO2 variation should be lower. Lo and behold. Random source: http://www.ferdinand-engelbeen.be/klimaat/co2_measurements.html#Variations_of_CO2_due_to_the_seasons:
A counter argument could be made that ice and snow ‘cover water’ and could prevent some uptake, but my HP calculator says that ice-sourced CO2 dominates. Yes I am aware that the annual variation has been attributed to land plants but the growth is inadequate and the (larger) CO2-ice argument is not addressed.
Your plot of temp and CO2 supports the idea that accumulating ice expells CO2 into the atmosphere, trumping absorption by cooling oceans. If the Antarctic ice cap is growing, the global rise of CO2 may well match the expulsion. The calculation is simple in round numbers because the air and oceans are both about 300-400 ppm.
If instead the ice sheets are melting, that meltwater will absorb huge quantities of CO2 (relative to human emissions) and the atmospheric level would decrease, just as it does in the NH each summer.
Who were the referees? Didn’t they look at the raw data? Does Nature have a shred of integrity left?
…
Joy, beautiful spark of the gods,
Daughter of Elysium,
Joy, beautiful spark of the gods
Spark of the gods!
Words from the poem ‘Ode to joy’ (Friedrich Schiller, 1785) and used in the finale of Ludwig van Beethoven’s choral symphony No. 9 in D minor, Op.125
Willis – is Nature obliged in any way to publish this? Not sure if that’s the best choice of wording, are there precedents under which they’re obliged to publish?
Willis, thanks for the clarification on timescales. This still begs the question how did Shakun et al calculate their global temperature curve? It doesnt look like an average of the temperature data points. Did they weight the ice core isotopic data? Something pulls their line to the right particularly between 17500 bp and 13500 bp. The data cloud says their temperature curve should start to rise around 19000 bp rather than 17500. It looks like it should also go to a higher temperature around 13000 – 14000 bp. THese 2 effects would put their temperature curve to the left of the CO2 curve at least in the time period 19000 – 13000 bp
Willis,
Thanks for the expansion. It bothers me that there actually might be an intention to mislead.
The second graph, described by the text as CO2 levels, has a Y-axis labeled as temperature. Did someone forget to relabel the text in their script before generating the graph?
Willis Eschenbach says:
April 8, 2012 at 2:46 am
“Look at Figure 2. For eight thousand years, CO2 has been rising fairly rapidly, and the global temperature has been falling.”
Average it so we can see that..
It is going to take more science on your part then what you offer. The simple fact that over the last ten thousand years CO2 has risen a little over 15ppm speaks for itself. Also the fact the Northern and Southern Hemispheres comes into to temperature alignment at the time of the mild CO2 rise and temperature decrease also makes for an interesting study. (http://i39.tinypic.com/f0qkcw.jpg)
Willis:
Excellent deconstruction. Do we know what CO2 source data they used and how that matches up against the NOAA source data?
It has been my understanding that there is a difference in timing, to responses of climate change drivers, between the southern oceans and the northern oceans. Would this timing difference show and what would it be, if Willis replotted the proxies for both southern and northern oceans together with the proxy temperature record/s?
Willis:
Ignore my previous question since you answered it in the preceding reply. Is there anything we can guess on the source of their CO2 data based on their other research or from the article. Shakun et al. make strong assertions while basing it on fuzzy proxies for temperature and without providing the source of their CO2 data. They make an end of discussion assertion with pretty sketchy data – how fuzzy is that?
Yeah, some of the AGW crowd also claim that farming since about 10,000 BC has increased c02 in the atmosphere, delaying the onset of the next ice age.
They claim that the Holocene has been warmer than it should have been (even though the temperature has been dropping), and also longer than it should have been, due to farming and clearing. But to my mind these claims are not supported by the facts.
1. The Holocene has been cooler than previous interglacials. With the human c02 effect, it should have been warmer.
2. The length of the current interglacial is not unusual (we should be slipping slowly into an ice age over the next few thousand years or so).
R
If that CO2 record is really that damped (as a physical phenomenon) to the point where it fails to capture Y-D it can hardly be regarded as driving temperature change as the paper claims.
The paper clearly shows (Because they average the proxy records) that CO2 leads temperature change then drops below it http://i39.tinypic.com/f0qkcw.jpg.. Showing that CO2 should rise again to match temperatures, which it does as shown in your graph. But because you don’t average the proxy temperature records you cant see that..
Great catch Willis!! I had an inclination there was something in there…And for some reason, I can’t get a famous phrase out of my head when I come to Fig 2: “Can you tell what it is yet?” (Rolf Harris) 🙂
wmconnolley says:
April 8, 2012 at 12:48 am
Wandering off in conspiracy land again I see”
xxxxxxxxxxxxxxxxxxxxx
Ha ha ha ha.
WARNING: Now that we’ve read your unwitty comment – intelligent minds will need to brain shower.
Reblogged this on Climate Ponderings and commented:
It was Twicksy
Willis
Fascinating stuff.
You show the temperature data. Did you also plot the co2 ppm figure? If it was 280ppm at the pre industrial period what is it now, and back when the rise started?
tonyb
There is also the uncertainty associated with the age determination – which would be (in my view) a more likely place for confirmation bias to creep in. It is difficult on paired analyses to be sure – but correlation of proxy to proxy is probably a bit of a judgement call.
For example – one could anchor all of the proxies at the point of greatest change with the assumption that the change was global and instaneous – and then see how / if the CO2 in anchored samples preceeded the temperature change.
Just a thought.