next message in archive
next message in thread
previous message in archive
previous message in thread
Index of Subjects
> they were being receiving a greater d
--Apple-Mail-41-435100132
Content-Type: text/plain;
charset=US-ASCII;
format=flowed;
delsp=yes
Content-Transfer-Encoding: 7bit
Hi Steve,
On 23-Apr-12, at 7:33 PM, Stephen R. Shaw wrote:
> Hi Chris, Dave: [buyer beware -- this contains some numbers but no
> locations]
> Putting off marking some student reports .... Chris' comments seem
> reasonable, but I share some of Dave's general misgivings in his
> earlier post about the quality of the some parts of the swift guano
> article.
>
> Where does the '2.2 cm per year' guano accumulation that they
> estimate, come from? The deposit is said to be 2 m deep ('2' = one
> significant figure -- a guess is that it is not uniformly flat on
> top so that this may be an estimate of only the edge of the part
> excavated). There are three guano horizons available in the main
> paper, [1] the top (0 cm, sealed 1992), [2] the cesium spike (70 cm
> down, supposedly 1963) and [3] the bottom (200 cm down, opened
> 1928). As Dave said earlier, none of these give their '2.2 cm/year'
> answer: [1] & [2] give 2.41, [2] & [3] give 3.82, while [1] & [3]
> give 3.17 cm/year. Another estimate is possible, buried in their
> supplementary Figure S1, where an estimated 100 cm were sampled from
> ~1944-1992, which gives 2.08 (different analyses start at different
> years and I couldn't find it stated when Fig S1 started, and its X
> axis is too sparsely labelled to make this out accurately, but it
> says '48 years' in the Abstract, and 100cm/48yr = 2.08).
Why not ask Joe Nocera directly? I've corresponded with him and he's
very open to providing information with respect to the paper. Whether
it's 2.2 cm/year or 3.1 cm/year (200 cm/64 years = 3.1 cm/year), what
salient difference does it make?
> In the only data Figure (Fig. 2a-d) of the main paper, the last
> three lines of the legend are rendered unintelligible by some sort
> of text transposition. Didn't any of the 10 authors notice this to
> correct it? (Did most even read it?)
There's no such problem in the version I have which I downloaded from
the site. Maybe there is a software glitch at your end
> Do these peculiarities affect the paper's conclusions? It probably
> doesn't really matter if the depth per year was at the extremes of
> 3.8 or 2.1 cm/yr, but it does shake confidence in the reader: if
> they can't even get such simple arithmetic across clearly, how
> careful/sloppy were they about the other measurements?
But Steve, without inquiring into this, you are assuming this is an
error. Maybe you should find out the basis of the calculation. Do you
honestly think that a paper published in the Proceeding so the Royal
Society by ten recognized academics which employed really
sophisticated science and mathematics (read their description of their
statistical analyses) suffers from simple mistakes in arithmetic?
Wouldn't the first assumption be that one needs clarification on this
point rather than the authors are sloppy and erroneous with simple math?
> Two other issues could use airing. Two meters of solid guano must
> press down heavily, so you might expect that the layers near the
> bottom would become compacted relative to those at the top, over the
> years.
I don't know that that's the case. Does guano really compact?
> The numbers above (2.41 for the top part versus 3.82 cm/year at the
> bottom) suggest the opposite. How could that be, mis-identification
> of the Cs spike, bad sampling, or what? An obvious need is for a
> tree-ring type analysis to really identify the years, perhaps
> looking for annual pollen spikes. Their sampling method is too
> crude to reveal this, but driving a vertical geology-type core down
> an undisturbed part of the deposit should work, and one of the
> authors lives in a geology department. This issue of possible
> compaction or reworking is not even mentioned in the paper.
Reworking is mentioned in the paper. Note, for example: "... changes
in nitrogen provenance of post depositional processing such as
ammonification or nitrate utilization ..." (paragraph 2, page 5) which
refers to bacterial or other reworking of the deposits in such a way
as to change the ^14N/^15N ratios leading to the comparatively large
fluctuations in ^15N as shown in Figure 2c.
> Second and perhaps more ominous, no metabolite DDE measurements are
> reported in the paper from before the mid-1940s. On the face of it,
> this may seem sensible conservation of effort because DDT was first
> detected in the wild in 1939 (they say somewhere). On the other
> hand, the article's reviewers should have absolutely required 2-3
> measurements from the 1929-1938 interval. If any substantial DDE had
> been discovered at these depths before 1939, it would mean that DDE
> had been eluted and washed down from the levels above. If such
> eluted amounts were substantial, this could compromise the basis for
> the entire depth analysis. Why didn't they look at this? Reviewers
> asleep at the switch?
My understanding is that all analyses were done at levels of the
1928-1992 deposit. The paper reports data from ~ 1945-1992 since the
authors are focusing on the impact of DDT on the Chimney Swifts. If
you look at Figure 2d you will see that the first data point shown
(from circa 1945) indicates that levels of DDE were at almost 0 p.p'.
DDT was not "first detected in the wild in 1939" (and the authors do
not say this). It's insecticidal properties were first noted in 1939
and it was not in widespread use until 1945 (see the legend to Figure
2).
Again, if you feel there is some uncertainty on this point, before
jumping to the conclusion that the author's did not conduct DDE
measurements prior to 1945, I would inquire first. The answer is an
email away.
> On Chris' point 2, '...indeed there was a slight (and statistically
> significant) rise until 1992 (when the chimney in this study was
> capped)...' The supplementary material discusses with some
> sophistication the curve fitting, which seems to involve fitting
> linear functions to chunks of data then combining these in the stats
> program 'R'. This produced