At the second meeting of Kaufman’s PIs, one of the scientists plaintively asked:
But shouldn’t we aim to do a synthesis that is only lake seds (at least as first step)?
This logical building block was pushed aside (thereby allowing Briffa’s Yamal series to be recycled for the nth time) on the following grounds:
some modeling experiments will require tests vs paleodata from other parts of the world (not just our lake sites)
As we observed to (bender’s) surprise the other day (see comments in thread …), Kaufman et al 2009 did not actually contain model results for the past 2000 years although that was supposedly part of the program – Kaufman Figure 4 uses results from a CSM model for the period 5600-3600BP. Nothing is mentioned in Kaufman et al 2009 about Ammann and Schneider’s runs (though they were discussed at PI meetings)
Although the archiving record of Kaufman’s program at NOAA is very incomplete, there have been some useful additions to the Alaskan BSi record from Kaufman’s MSc students that we’ve been discussing over the last few days. Yesterday, I mentioned Daigle’s Goat Lake BSi series, which showed a remarkable contrast between LIA sediments and MWP sediments. An excerpt is shown below.
Should any MSc students read this thread, I though that Daigle’s graphic was considerably more effective than the corresponding graphics in theses by Kathan and McKay, as it nicely integrated the geological classification of the sediments with the quantitative information from BSi % and Organic Material %. Low values of BSi% and OM% are associated with a type of gray mud, while high values of BSi% and OM% are associated with “gyttja”.
Only McKay’s Hallet Lake series was used in Kaufman et al 2009. Here is the Hallet Lake BSi series (as rendered into Kaufman decadal averages and rendered into SD units) – showing both the Kaufman archived version and my replication. The NCSC archive observes that a new version was filed in Nov 2008 and possibly Kaufman used an older version of the Hallet series. (I’ve exactly replicated about 10 Kaufman series from original data and am 100% confident that I’ve got his decadal averaging and re-scaling method. In my emulation, I’ve included 3 BSi upspikes reported in the MSc data, but expunged in the NCDC version.)
In either version, Hallet Lake BSi has a 20th century uptick, a depressed medieval warm period (always attractive to the Team) and an elevated early first millennium.
Figure 2 – Versions of Hallet Lake BSi in SD Units – Kaufman archive and emulated from MSc data.
In order to orient readers to a broader range of BSi values than the Hallet Lake site used in Kaufman, the graphic below compares BSi % over the Holocene for 5 Alaskan sites ranged from west to east. Information from three of these sites (Goat, Cascade, Hallet) are from Kaufman’s MSc thesis (I manually typed in the Goat Lake values as these were annoyingly in a photo-format.) Two sites are from other groups and archived at NCDC. Only the Hallet BSi value is used in Kaufman 2009.
The contrasts in Daigle’s Goat Lake series are obviously much sharper than the corresponding series from Kathan (Cascade Lake) and McKay (Hallet Lake) (or the other two sites.) This seems like an extremely interesting and useful result – precisely the sort of thign that one would like to have seen discussed in a “synthesis that is only lake seds (at least as first step)”. All of the following series show BSi (plotted as % rather than SD units). Take a look and I’ll comment further below.
Figure 3. Five Alaskan BSi % series.
Obviously the BSi contrast at Goat Lake is 1-2 orders of magnitude greater than at the other sites. In most forms of analysis, it’s easier to work from strong contrasts than from weak contrasts and I don’t think that there’s any exception here.
The big contrast at Goat Lake is between the very low BSi% in the grey (predominantly inorganic) mud of the Little Ice Age and the LGM and the elevated BSi% associated with gyttja in the Holocene Optimum, the MWP and with modern warm period sediments. Daigle observed that the LIA glacier advance at Goat Lake was unprecedented since the LGM ( I didn’t check whether he used the u-word, but he made the point.) BSi% at Goat Lake in the MWP is very high (as you can see) though not quite at the overall maximum; the anti-MWP animus in the “community” is so great that despite the elevated BSi values compared to the LIA and modern period, Daigle is obliged to observe the following:
Warmer temperatures during the MWP are not recognized in the Goat Lake productivity signal, in which OM remains constant and BSi decreases from 31% at 1000 AD to 26% at 1200 AD (Figure 22).
In comparison with the striking contrast at Goat Lake, there is negligible contrast in the Hallet Lake BSi record employed by Kaufman. Re-examining the McKay thesis, the Hallet Lake sediments unsurpisingly are reported to be grey mud (not gyttja) – the color of the lower mud (with higher BSi values) is a “dark grey” while the more recent portion is a “light” grey. At nearby Greyling Lake (the original NSF30 target), a gyttja zone is reported, but Kaufman did not take any BSi measurements and focused on nearby Hallet Lake.
As a third party looking at this proxy for the first time (albeit one coming with experience in geological literature and proxy literature), it’s hard for me to avoid the view that the minor Hallet Lake BSi fluctuations are extremely unlikely to be usable as any sort of useful climate record (likewise, the Cascade Lake BSi record.) It seemed to me that Goat Lake represented a much better building block, but that’s just a first impression.
I echo the frustration of the scientist who asked:
But shouldn’t we aim to do a synthesis that is only lake seds (at least as first step)?
I have less than zero interest in yet another CPS-style reconstruction blending Yamal with a bunch of other data with no common “signal” which is what we got. But it would have been very interesting to see a “synthesis” of lake sediments in which the work was carried out with consistent methods, thereby permitting an overall assessment of the value of the proxy. If this be “vicious commentary”, so be it.
Climate Audit 
23 Comments
“am 100% confident that I’ve got his decadal averaging and re-scaling method”
Careful. You will force them to add a pseudo random noise sequence to protect their IP.
Steve, if Cascade, Mica, and Hallet were plotted over a smaller ordinate percent interval, it would be possible to visually compare them with Goat. But in any case, looking at them as they are, there doesn’t seem to be a common signal among them. One wonders what would be learned by taking an average.
In any case, Goat seems to show a very pronounced RWP, a more modest MWP, and an even more modest Modern WP. Not much evidence for anything extraordinary in the current climate at Goat Lake.
Hallett Lake contributes greatly to the monotonic downtrend Kaufman found before the CSH (current HS) set in, because it was normalized to unit variance during the relataively brief period 980-1980AD when all 23 series were observed. [Steve: you mean 980-1800]
HM reply: My bad. But in fact, all 23 were observed 970-1810, so they could have used 2 more decades than they did, and still have had all present. This makes a 12% difference in the variance and therefore a 6% difference in the weight for one of the early series — I forget which — but leaves the rest about the same.
If it and the 13 series that are observed between 0-460AD and up to 1980 were normalized to the full period 0-1980 in order to reconstruct 0-460AD, Hallett would receive a much smaller weight in this portion of the reconstruction, and the appearance of a monotonic trend would be much weaker.
I’m not claiming that CPS is a sensible way to aggregate proxies, just noting that this is what CPS would seem to call for, if consistently applied.
Re: Hu McCulloch (#2),
It would be interesting to calculate the ratio of weights arising from longer standardization versus shorter standardization. I can probably do it pretty quickly.
Re: Hu McCulloch (#2),
when the original authors have already converted the proxy to temperature, the procedure of Loehle and McCulloch 🙂 (using the temperature estimates), seems more logical than converting the temperatures back to SD units (though ultimately this only affects respective weights).
It’s nice to see that at least one person wanted to stick with the original program, no matter that their concerns got brushed aside.
It strikes me as odd to convert to unit variance, but then use everything as a temperature proxy. What normalization acknowledges is that there are site-specific differences in how the sediments respond to conditions. Some respond with a large variance, and some with a much smaller variance. But from what’s been reported in the threads about the theses, and from comments in the threads, these sites seem to likely (or possibly? or primarily?) respond to other conditions such as precip., not to mention site-specific effects of post-deposition disturbances and extraction problems.
So it seems a bit shaky to just toss everything in as an overarching temperature proxy. Combined with selection bias in sediments chosen and combined with Yamal and friends, then… well I’ll just stop there.
Re: Clark (#5),
From what I can gather, the BSi variation is related to temperature because a warm period with a retreated glacier and high biologic activity results in sedimentation high in BSi. During cold periods, the primary sedimentation is from glacier scraping and grinding abiotic fresh rock.
This is most clearly seen at Goat Lake and may be a decent temperature proxy. For lakes with very low BSi variation, the temperature signal is likely buried among many other signals.
I suspect that the geography and hydrology of each watershed determines which lake gives the best BSi/temperature relationship. I have not seen a conceptual model to evaluate the appropriate sites for using this particular proxy. Maybe someone else can site such a model that I missed.
Re: Howard (#9),
It would be great to see that proven out. Even by overlay of reasonable temp values and multiple samples. If even one of these proxies worked out I would have a hundred times improved confidence in this type of “science”.
Re: Howard (#9),
The amount of variation does not at all indicate whether something is a strong or weak proxy. That is a matter of covariance, not variance. That the variance differs between graphs is not necessarily damning. It’s just another mysterious inhomogeneity of unknown cause. Just another item on Ken’s wish list of things to be explained.
Re: bender (#12),
You are right, at least theoretically. From a geologic gut perspective, the lack of variation in actual mass production from a macroscopic hydrogeologic feature that may be dependent on climate shifts is a sign of a poor proxy. With low variation, any old fart or burp will be construed as some signal of a real movement. In my experience, people get their panties in a knot over nothing all the time.
Always remember, the map is not the terrain.
RE SM #6, see #16 over in the “Invalid Calibration in Kaufman 2009” thread, since this relates to general calibration issues.
The following comment pertains to Blue Lake varve thickness – one of Kaufman’s series that I’m working on. While it’s not a BSi series, I couldn’t help but note it on the nearest thread:
That would definitely qualify as “weak”.
Re: Steve McIntyre (#8), Surely this is not representative!?
Logical building blocks get in the way of fast-tracking policy-oriented science. We have a deadline.
so what happens with the results when there is only a synthesis of lake seds?
Re: cba (#14),
“a synthesis of lake seds” refers to a formal review of the state of the science. It’s not something you generate overnight.
Wegman — “when there are major policy decisions to be made based on statistical assessments …evaluation by statisticians should be standard practice.”
We can dream.
Howard: Re: Howard (#16),
Thirty years ago, at a mine we were constructing in the middle of nowhere, a certain drawing indicated that the sewage drainage field for the mine shop complex had to be located just so, out to the west of the shops.
.
But when we went out to see what the ground actually looked like, a hill was found rather than a depression. So it was back to the drawing board for the Head Office engineering staff.
Re: Scott Brim (#19), The solution was obvious; just invert the hill!
Re: Paul Penrose (#20),
Yep. Maybe head office did a principal components analysis of the topography 🙂
Re: Paul Penrose (#20)
Re: Steve McIntyre (#21)
.
A look at the base topography map revealed that it was perfectly OK. The original engineer had simply interpreted the contour lines as a hole rather than as a hill, and had quite obviously not looked at the elevations that were actually assigned to those contour lines. Otherwise his design was a near perfect solution to the contractual requirements.
Not being a sediment scientist, but having to keep sedimentation in mind when performing my job (prescribed burning), has any thought been given to the possibility of increased sedimentation after wildfires upstream from these lakes? Even if the average temperature changed (up or down) for a period of time after a fire, one could reasonably surmise that there could still be increased sedimentation from erosion of exposed ground, for at least a few years, due to the short growing season in Alaska. Any enlightenment (pro or con) would be appreciated.
Steve, thanks for all of the incredible work that you and your colleagues put into this and for your efforts at bringing sanity back into science.