Guest Post by Willis Eschenbach
The new Nature Magazine article on Lake Tanganyika, “Late-twentieth-century warming in Lake Tanganyika unprecedented since AD 500”, discussed a couple days ago by Anthony Watts here, was quite interesting to me. In 2003 I had contributed a “Communications Arising” to Nature Magazine regarding earlier claims that AGW was causing productivity loss in the Lake. As a result, I am very familiar with the available records for the lake.
Figure 1. Rainbow over Lake Tanganyika
I was puzzled by the claims in the new article regarding the changes in Lake Tanganyika surface temperatures, because I knew that there was almost no historical data on lake surface temperature. I wondered how they determined the surface temperature of the lake over the past 1,500 years. So I sprung the $18 to purchase the Nature paper and find out …
It turns out that they used a proxy called TEX86, which has been used in other studies. But how did they calibrate the proxy to the lake surface temperature (which they call “LST”)?
Well … they didn’t calibrate it. In their theory, no calibration is needed. However, that seems like a very problematic assumption, as there are always confounding factors for proxies that mean that they need to be calibrated to the instrumental record. Some of these factors are listed in their Supplementary Information.
How well does their reconstruction correspond with air temperatures? Well … rather than compare the reconstruction to local temperatures over the last 50 years, and despite the fact that Lake Tanganyika is in the Southern Hemisphere, they compare the reconstruction to a famous Northern Hemisphere reconstruction …
Figure 2. A most ingenious way to hide the differences between two graphs, by redacting the front information so you can’t see the back information. Note that part (a) uses the discredited Hockeystick and various Hockeystick clones (the so-called “independent reconstructions”) as its basis for comparison.
Commenting on this figure, they say (emphasis mine):
Our LST reconstruction is qualitatively similar to Northern Hemisphere temperature reconstructions (Fig. 3a), implying that Tanganyika LST largely followed global trends in temperature during the past 1,500 years, much as it has in the past half-century. As LST closely tracks air temperatures over the instrumental period, we can also infer that air temperatures in this region of East Africa varied in concert with the global average and thus were controlled primarily by the major forcings influencing temperatures over this timescale, both natural (solar radiation, volcanism) and anthropogenic (greenhouse-gas emissions; refs 19, 20). The temporal resolution of our dataset precludes comparison between Tanganyika LST and volcanic events of the past, but we can compare our record with changes in solar irradiance (total solar irradiance (TSI) anomaly, estimated from 10Be in ice cores21; Fig. 3b). TSI and Tanganyika LST share some similar centennialscale features, including maxima near 1350 and minima at 1450, 1250 and 1000. However, TSI variability clearly does not explain the dramatic twentieth-century increase in LST, which, as with global temperatures, is probably a response to greenhouse-gas forcing.
Unfortunately, in their paper they neglected to show how the Lake Tanganyika LST “closely tracks air temperatures over the instrumental period” of the “past half-century”. To remedy this lacuna, I have plotted the only two longer-term temperature stations on the lake along with the MSU data and the proxy-derived LST:
Figure 3. Ground station temperatures, UAH MSU, and proxy lake surface temperature (LST), 1950-1996
As you can see, while their proxy LST generally agrees with the air temperature over the last half of the record, it does very poorly during the first half. So no, the LST proxy reconstruction does not “closely track air temperatures over the instrumental period.”
Finally, Tierney with some other co-authors have published previously in Science Magazine (subscription required) on the Tanganyika LST. In the current (2010) paper, they say (emphasis mine):
Before the twentieth century, LST varied between 22.5 C and 24.3 C (Fig. 2a). LSTs were relatively warm between ad 500 and 700, followed by an interval of cool LSTs that lasted until ad 1100. Lake Tanganyika then experienced a period of extended warmth between 1100 and 1400, followed by a return to cooler LSTs between 1400 and 1500 and more variable temperatures until 1900. Beginning around 1900, LSTs trend upwards, rising about 2 C in 100 years (see Fig. 2 inset). Our uppermost sample from core MC1 (identified using 210Pb dating as about ad 1996), calibrates to 25.7 C.
OK, so the current paper says that in the last 1,500 years the LST has varied between a low of 22.5 C to a high of 25.7 C. During the last 50 years of the record, their proxy LST value rises by 1.6 C.
And in the current paper, they also say:
Our records indicate that changes in the temperature of Lake Tanganyika in the past few decades exceed previous natural variability.
But in their previous (2008) paper, which used the same TEX86 proxy, they had said:
Holocene lake [Tanganyika] surface temperature (LST) fluctuated between 27° and 29°C …
And during the Holocene, their 2008 paper shows a change of 1.65 C in 50 years, which is larger than the recent change shown in the 2010 paper.
Despite citing the earlier paper in their current paper, they don’t mention these discrepancies … which does make me wonder just how good their proxy is. It also make me curious about what they mean by “previous natural variability”. During the Holocene, by their own figures, the Lake Tanganyika LST was 3 C warmer, and changed temperature faster, than in the last fifty years of their more recent proxy record.
[UPDATE] You know how sometimes you have this nagging feeling that you’ve left something out, and you can’t think of what it was? When I woke up this morning, I realized what I had wanted to say.
This is truly a watershed paper in that it purports to be a study of the changes in lake surface temperature (LST) over time, but they present no measurements of the changes in the LST over time. The only actual surface temperatures mentioned in the paper are the following, all from 2003:
Our uppermost sample from core MC1 (identified using 210Pb dating as about ad 1996), calibrates to 25.7 C. This is within the range of 2003 measurements of seasonal LST for the Kalya Slope area (25.5-26.3 C; see Fig. 2 inset) and is also similar to the annual average LST measured near Mpulungu, at the southern end of the lake (26.1 C; ref. 16).
Unfortunately, reference 16 is very vague. It is:
Descy, J-P. et al. Scientific Support Plan for a Sustainable Development Policy (SPSD II), Part II: Global Change, Ecosystems and Biodiversity Atmosphere and Climate (Belgian Science Policy, 2003).
Research showed this is the Belgian CLIMLAKE project, which I had studied before, and which had some interesting results. Here’s one of them:
Figure 4. Satellite derived lake temperatures. SOURCE – CLIMLAKE FINAL REPORT.
As you can see, on a single day the surface temperature of the lake varies by 4° C from coldest to warmest. I couldn’t find their “2003 measurements of seasonal LST” or their “annual average LST”, although Figure 29 of that CLIMLAKE report does show a three year temperature record for two places on the lake, so I suppose they might have used those.
(As an aside, my high school science teacher would never have allowed such a vague citation as reference 16 above, I’d have gotten a “D” on the paper if not an “F”. “Make it easy to find”, he’d say, “point me right at it. Cite me chapter and verse.” But I digress …)
My point is, the Tierney 2010 report is a study of the change in Lake Tanganyika surface temperature over time, which contains no measurements of the change in LST over time, and which has exactly three actual surface temperature measurements, which are poorly cited, are from different parts of the lake, and are all from 2003 …
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
—
Sera takes note of the appropriateness of sending Tierney et al to the ‘Group W bench’.
Where they will each doubtless be asked:
—
Rich, I can not tell a lie…
I put Tierney et al underneath that pile of garbage….
Willis-
Your welcome. Langenberg’s work seems to reflect the benefits of 8 years of life on the Lake. Lake Tanganyika is one of those great waterbodies that can make you continually scratch your head–its complex, confusing, and at times counter-intuitve– all the things that create interest and challenge. It would be a shame to see this elegant complexity destroyed for the sake of a few wild assumptions and a temperature proxy.
BillD says:
May 20, 2010 at 12:24 pm
Gail;
Your comments about the stable water column, low oxygen and small change in temperature with depth are all correct according to my understanding. However, there is no evidence at all for geothermal (volcanic warming) of the deeper waters of the lake.
________________________________________________________________________
It is a quote from elsewhere. The volcanic activity seems to be “common knowledge” see: http://www.aquaticcommunity.com/cichlid/tang2.php
Here is another quote:
“….Volcanic activity at the lake’s floor accounts for this tiny difference in temperature and causes a high level of dissolved minerals. The lack of temperature difference means there is a lack of vertical currents. Without the mixing of the deep water and the water at the surface, oxygen cannot reach a depth greater than 300 feet, leaving the waters of the deep either too high in Hydrogen Sulfide, or too low in oxygen to sustain life forms (zambiatourism). ….” http://www.tassiecichlids.com/viewtopic.php?f=18&t=25&start=0
The reference seems to be zambiatourism.
“…The lake is 1,460m above sea level and empties into the River Ruzizi, which flows southwards into Lake Tanganyika, covering 2,700sqkm in a volcanically active area….” http://www.globalenvision.org/library/1/1732
Another article (behind a paywall ) states the “..lake and land structures are grossly similar. Low heat flow and similarity with surrounding values in Africa is consistent with lack of active volcanicity and seafloor spreading in Lake tanganyika. A new concept on the evolution of a rift is proposed….” http://www.springerlink.com/content/m02m140738h1x317/
This all seems to indicate an area with heat coming up through the ground similar to what is seen in Yellowstone and other ” volcanically active areas”
Did these researchers measure the PH of the lake? If not, why not?
If they did, what did they find? (Nothing alarming, probably.)
[REPLY – The Rift Lakes are strongly Alkaline. ~ Evan]
Doesn’t the air temp instrumental record go back to 1913? Wouldn’t you want to plot the Tierney proxy temp against the entire record, plot the trend lines and see how they correlate to see if “LST closely tracks air temperatures over the instrumental period” ?
On reflection, it is easy to see that the “Oldest Profession” is still going strong. I guess the temperature or water level or CO2 content have nothing to do with making a buck the easy way. I just never realized how much money could be made at it if you only had a PhD.
You can draw any line you want, if you have only one data point.
[old math joke]
Let me cherry pick the data, and I can prove ANYTHING!
From comments in the first Tanganyika thread, it appears that Andy Cohen is the guy to ask about the research. The summary identifies him as chief researcher. He took (and is featured in) the pictures, drilled the cores, and has been studying these rift valley lakes for some time:
http://www.geo.arizona.edu/web/Cohen/AC_page.html
http://www.geo.arizona.edu/web/Cohen/publications.html
When these articles are all on the other side of a paywall, it’s hard to sink your teeth into the specifics – say, the exact method of coring bottom samples, where they were taken and what they measured – but issues raised above about these methods, prior warming, (and if it’s there) its geothermal causes, sure raise some questions.
One problem seems to be that a lot of these “breaking news” research papers (I’m thinking also of the geology papers reviewed here a month ago, but any of these climate papers which use the word “unprecedented”) are the works of undergraduates or newly-minted grads from various schools. A picture of how these papers get written is emerging:
A professor lines up (or is given) funding from NAS or NSF to “investigate” (prove) global warming in some unique environment; he/she arranges a research expedition (“field trip”) to that exotic locale and drums up the adventurous, social aspects of the trip, priming students with intended result (“we’re going to find evidences of man-made global warming as exhibited by Lake Tanganyika”); he passes out a number of his own “peer-reviewed” papers to show how it’s done; he ascribes grades based on whose paper most closely approximates the intended result; papers which show AGW in the most damning light “win” publication.
I don’t think the challenge to respond to questions will be met. We won’t get answers from students because, well, they are just students after all. This was chiefly an academic learning experience for them. We won’t get the mentoring professor to answer because it would be impolite, if not improper, for him to intercede on behalf of his students, who, after all, wrote the paper! The publishing magazine did no peer review, claims to the contrary notwithstanding, and the long record of their responsiveness to complaints, querries is documented on this and Steve McIntyre’s web site. So I suppose we shouldn’t expect any reasoned clarification or discussion from any of these sources. Nobody is responsible.
Meanwhile, it appears that the perks for these projects are simply irresistible. If I were an undergraduate science student with an option to try this, I’d leap at it: the trips to exotic locales, the expense accounts, the notoriety, the status, the grooming for a government job, not to mention the opportunity to accompany the tallented Ms. Tierney (pictured above)… and, assuming that you enjoy all the above, the prospect that someday “all this will be yours” as an educator who can perpetuate the cycle. It’s a win-win!
JB says:
May 21, 2010 at 5:45 am
JB, which air temp record are you referring to? I don’t find any local air temps that go back that far.
—
Writes Bill Parsons:
I regret to report that I’m not a frequenter of Mr. McIntyre’s Web site. If Mr. Parsons could provide a URL to those threads on that site in which discussion of these periodicals’ “ long record of their responsiveness to complaints [and] queries,” this would be appreciated.
That aside, Mr. Parsons‘ explanation of the mechanism whereby graduate students and post-doctoral fellows are used by their AGW-invested preceptors as “cut-outs” to disseminate dezinformatsia favorable to the warmist fraud is both perceptive and reasonable.
—
Willis Eschenbach says:
May 21, 2010 at 11:49 am
JB, which air temp record are you referring to? I don’t find any local air temps that go back that far.
Tierney’s paper claims a temperature record from 1913 to 2000 citing Verburg. However Verburg (2009) shows only a 3 month data set for 1913, data for 1938, 1947, 1973 and 1975. Hardly a reliable or continuous temperature record for either paper.
I’ve written to Langenberg – if he responds any interest in posting?
My apologies, I realized everything I was looking at was lake temps.
According to the SI, they are computing their calibration equation,
LST = TEX86 * 38.874 – 3.4992
with a cross section of 13 lakes around the world, and not with a time series.
In itself, this is reasonable. They do run the regression the wrong way (by “ICE”, as if TEX86 caused temperatures, rather than by “CCE”, regressing the proxy on temperature and then inverting), but since the R2 is .92 and the t-stat on the slope is 11.0, this doesn’t make a big difference. (I’ve checked that these are indeed the ICE estimates from their 13 data points.)
A potential problem, however, is their rejection of the regression standard error (2.1dC) as the basis of computing reconstruction standard errors. The reconstruction standard errors have to be at least this high (and in fact a little higher, because of slope uncertainty and their incorrect use of ICE, but not by much in this instance). But they don’t like this number, and so instead use a “leave-one-out” “jackknife” procedure, and somehow end up using a 95% CI of 0.40 dC, corresponding to a SE of 0.20dC, too small by a factor of 10! I think they may be misunderstanding what the jackknife procedure tells you, though I’m not certain how they used it.
Rich Matarese says:
May 21, 2010 at 12:13 pm
Climate Audit is Steve McIntyre’s web site. Over the several years that I’ve frequented Climate Autdi, I’ve seen a number of stories in which Mr. McIntyre reasonably requested supplemental information, data, or clarifications from various magazines. These included Nature and Science, as I recall, but there have been mentions of others journals. My older browser and platform don’t like his web site since he reformatted it, and even Firefox shows a lot of overwritten text and columns, so I just don’t go there any more. He has a search window, in which you might search “Nature Magazine”. Or, if you wish to approach it from the other perspective, here are the East Anglia e-mails, where the cozy relationship with the journals is obliquely referenced:
http://www.eastangliaemails.com/index.php
I’m heading out to eat, but I’ll try looking for some of examples of this later. What has become pretty clear over the years is that there is an openly symbiotic relationship between the climatologists and the periodicals.
Thanks, Willis, for the thread. Hope this isn’t too far off the topic.
Panel a compares Lake T to an AR4 HS composite, but notice that the scales are different by a factor of 4! “Our LST reconstruction is qualitatively simlar to NH temperature reconstructions (Fig 3a), implying that Tanganyika LST largely followed global trends in temperature during the past 1500 years, much as it has in the past half-century.” (emphasis added) Note also that the comparison runs no more than 1300 years.
Note that the new article is in Nature Geoscience, not Nature itself, and is in the June issue which is not yet the “current issue”, but is available in “Advance online publication” form.
Hu McCulloch says:
May 21, 2010 at 3:12 pm (Edit)
Willis Eschenbach says:
Thanks, Hu. For those not familiar with the name, Hu is a frequent poster at ClimateAudit and a person whose mathematical judgement I trust.
I had not looked at their so-called 95% CI, because I didn’t see any way to even estimate it. If we have virtually no data about the lake surface temperature, how can we even estimate what the error of what is only an educated guess might be?
I note that in their SOI they list the constants used with the magical formula that relates the temperature to the degree of cyclization of aquatic archaeal glycerol dialkyl glycerol tetraethers to temperature … as they say …
These constants vary from 0.35 to 0.88 in the various lakes, which I assume reflect the actual conditions of the lakes in which they are used. Since we don’t know the actual conditions of Lake Tanganyika, it could even be outside of those parameters.
So I didn’t even try to guess how far wrong their guesses might be. And given all of that, your numbers, though reasonable, seem like they might be small.
(In that regard, I loved their statement that, after they calculated the 95%CI and decided that it was too big, that
Conservative? Conservative to me means a big 95% CI, not a small 95% CI. But I digress …)
Finally, there is no attempt to see if their analysis (of the degree of cyclization of aquatic archaeal glycerol dialkyl glycerol) might be affected by modern changes in the lake. These would include sewage pollution, increased silt load from erosion in the surrounding areas, runoffs from farms, introduction of fish species, and the like. What do those do to the degree of cyclization of aquatic archaeal glycerol dialkyl glycerol? About all of that, the authors are silent.
So Hu, I’d say that your excellent estimates would represent the absolute minimum 95% CIs, and that the true answer is likely larger.
Hu McCulloch says:
May 21, 2010 at 4:02 pm
Whoa, nice catch. I had missed that entirely … hide the incline.
Tierney draws heavily on the BSI as evidence for the link with TEX86 LSTs for a temperature related productivity. Langenburg seems to contradict Verbums finding that diatoms are an indicator for Lake Tanganyika productivity stsating picocyanobacteria may at times be the dominant form of phytoplankton. A 2009 paper in Journal of Plankton Research by Stenuite et al supports this position .
A paper by Hecky and Kling 1987. Phytoplankton ecology of the great lakes in the rift valleys of Central Africa. Arch. Hydrobiol., Beih. Ergebn. Limnol., 25, 197–228 found that in Africa rift lakes upwelling is associated with diatom production and stratification stability with the production of cyanobacteria.
A presentation by Hecky and Verburg http://www.espp.msu.edu/climatechange/…/Physical%20and%20Ecological%20Responses%20of%20the%20Great%20 showed the switch from cyanobacteria in the wet season to diatoms in the dry season on an annual basis.
My question is what has Tierney shown- her BSI seems to merely reflect one of two modes of primary production. I’m not sure how BSI says anything about total productivity in this lake.
Willis:
They are using just one recent point in time for each of 13 lakes, ranging in Annual Lake Surface Temp from 11.2 or 11.6 dC (Issyk Kul and Yellowstone) up to over 28 dC (Matano, Towuti, and Turkana). The first two entries in the table are for L. Tanganyika (Kigoma Basin and Kalya Slope), so these presumably are the tops of the 2 cores they are using for their study, though I’m not sure which is which. These had TEX86 values of .761 and .752, and ALST of 26.5 and 26.1, resp. As you point out, there is a lot of variation in temperature from one end of the lake to the other, and I’m not sure how they’re taking this into account. They may have just taken these temperature readings themselves as part of the project.
They then regress ALST on TEX86 as shown in Figure S4, for which the regression residuals have a standard error of 2.1dC, and then use this line to “forecast” temperature from past values of TEX86 at various depths down the core (not tabulated anywhere, unfortunately). Total Forecast Standard Errors from this calculation (including both the coefficient uncertainty and the observation errors) are 2.1*sqrt(1 + 1/13) = 2.2dC at the average of the calibration TEX86 values. They increase somewhat as you move away from the average, but not by much given that the slope had a t-stat of 11.0. This means the 95% CIs should extend at least 4.4dC above and below the point estimates, rendering the entire reconstruction uninformative.
Simply stating, as they do, that the sample is small and that there are additional uncertainties is no excuse for replacing the correctly computed forecast standard error with a value that is 10 times smaller. If done correctly, the “leave-one-out” procedure will give the coefficient forecast standard error (2.1*sqrt(1/13) = 0.58dC at the mean of the TEX86 values), rather than the relevant total forecast standard error, but they have somehow come up with something even smaller than that.
Re: Hu McCulloch says:
May 22, 2010 at 8:11 am
Thanks, Hu. I have changed the formatting of your posting to what I think you intended.
I agree with you on all points. I don’t think that they have the data on Lake Tanganyika to plug into their equation, because the temperature changes so much from end to end of the lake, and the data are so scanty.
I am more concerned, as are you, with their … ummm … “interesting” take on statistics. I simply don’t see how they can get from a standard error that is over 2°C to a 95% confidence interval of 0.4°C. You say that this would make their reconstruction “uninformative” … I had a rather stronger word in mind, but this is a family blog, so yours encapsulates the situation quite well. Here’s what I get using a 95% CI of 4.4 …
Figure 5. My digitized data from their Figure 3, showing the correct 95% Confidence Intervals
w.
Willis —
Thanks for fixing the blockquote in my post.
Your diagram about says it all. Except that it would be nice to put the HS on the same scale as Lake T LST, so that it appears as a flat line!
I wonder what they mean by % agreement with the NH recons?
— Hu
Willis —
How did you obtain Tierney’s point estimates to form your CI? Did she archive them somewhere I didn’t see, or do you have a program to digitize graphs? Reply by e-mail if you like: mcculloch dot 2 at osu dot edu.
Hu McCulloch says:
May 24, 2010 at 8:38 am
Some years ago I got tired of asking “scientists” for their data and being turned down. So I started digitizing off of graphs, just using my graphics program.
Today, of course, there are applications for that. I’m on the Mac, and I use a program called “GraphClick”, but I’m sure there are corresponding programs for the PC.
In this case, I used the program to automatically digitize the center line of the black zone. To that, I added their ±0.4 C error interval. Those are the three blue lines on the graph. I often copy them back onto the original graph, to check that my digitizing is correct.
To copy them back, I first make the graph in Excel, then I copy the graph and paste it into my graphics program. I use “Vectorworks” as my graphic program, and when you paste an Excel graph into VW, it comes in as a series of actual lines rather than a bitmap. This lets me size it properly to the bitmapped graphic (which I have also imported into VW).
Finally, VW allows me to do transparent overlays, like the one you see above.