Holliday et al.'s Gish gallop: timing of the Younger Dryas impact on four continents
Holliday et al.'s "comprehensive refutation" of the Younger Dryas impact hypothesis (YDIH) is a highly misleading Gish gallop. Earlier blog posts have shown how Holliday et al. use false, misleading or inconsistent statements, wordplay, and an inadequate scientific methodology to make their case. Essentially, they have no refutation arguments at all.
An earlier blog post addressed the presence of impact melts and microspherules in the YDB which, by themselves, confirm a cosmic impact at Abu Hureyra near the YD onset. Due to the similarity of debris at the YDB across a wide range of sites, it's likely there were cosmic impacts across a wide area within a short timespan, indicating a singular event or highly correlated series of events.
Another blog post examined the presence of nanodiamonds in the YDB. And another examined the presence of platinum in the YDB. In each case, microspherules, nanodiamonds and platinum are found precisely where they are expected based either on direct dating or proxy dating using, for example, terminal Clovis cultural artefacts, often near the bottom of the Younger Dryas black mat. The agreement is superb, essentially confirming a widespread cosmic impact near the YD onset.
In this blog post we will examine section 5 of Holliday et al. which questions the dating of YDB sites. Their view is this debris, supposing it even exists, is not synchronous across several continents and therefore this refutes the YDIH.
Note that establishing the synchroneity of events in different places nearly 13,000 years ago is difficult. It is much easier to show that two events were not synchronous, which is Holliday et al.'s aim. The reason is due to the inherent uncertainty in dating measurements. It follows that a good understanding of uncertainty in scientific measurements is required for this debate. Holliday et al. fail badly here.
Essentially, any scientific measurement is meaningless unless accompanied by an estimate of uncertainty. This includes estimates of the uncertainty in any fitted lines of regression to, for example, a set of radiocarbon data. Failure to record the uncertainty in such lines, as in Meltzer et al. (2014) and Gill et al. (2009), might be considered an oversight. However, Holliday et al. defend their failure. Their position is therefore not scientific. In fact, their position could be called pseudoscientific because it claims to be scientific when it is not.
The problem of interpreting radiocarbon data is particularly important in this debate. This is because it is very often the case that the intrinsic radiocarbon measurement error (resulting just from the radiocarbon measurement and nothing else) does not capture the true error in the measurement. The true error will have contributions from many other processes in play. To illustrate this point consider the data set provided by Gill et al. (2009) below. They use this data set to construct a (flawed) age-depth model for the specific lake sediments they investigate. However, it is clear from this data set that the error bars in the individual measurements are typically much less than the scatter in the data. Indeed, the scatter in the data covers nearly 8,000 years, while some of the intrinsic measurement errors have error bars of only a few hundred years (some are even smaller than the symbol size). This is obviously inconsistent and demonstrates that the error bars provided are almost certainly too small.
This type of situation is actually quite common. So what has gone wrong? Most likely, this problem is caused by other non-radiogenic processes in play that affect the apparent age of a sample, such as the 'old wood' effect or leaching by 'young water' ('young water' is water with 'young' dissolved carbonates which can sometimes react with a sample or simply saturate a sample leading to a radiocarbon age that is too young). Additional uncertainty is created by the relative movement of objects within sediments, e.g. via bioturbation (creepy-crawlies or disturbance by roots) and other ground disturbances like flooding.
However, these are 'known unknowns'. There could be many more 'unknown unknowns', i.e. processes that we have no knowledge of, yet they significantly affect the radiocarbon measurement of a sediment sample or its position within a sediment column. It follows that single radiocarbon measurements of sediments are unreliable for estimating the age of sediments, and instead a suite of measurements from different sediment depths is preferred. Age-depth models can then be constructed, provided a proper account of the uncertainty in the set of measurements is made.
The trend in archaeology is to prefer Bayesian modelling to generate age-depth models. In principle, this is the best available method. However, as is always the case, any model is only as good as its underlying assumptions. Unfortunately, the assumptions employed in Bayesian models of age-depth data are often either ignored or inadequately justified. The result is that age-depth models produced by Bayesian methods are often too tightly constrained and underestimate the true uncertainty in the data.
Let's consider this issue a little more carefully. As we have seen, the intrinsic error in a radiocarbon measurement usually underestimates the true uncertainty in the data. And yet, when using Bayesian modelling, all sources of uncertainty must be known accurately or else the age-depth model will be over-constrained. Therefore, when using Bayesian modelling it is quite common to either eliminate outliers (these are data points that are a long way from the trend of the others and therefore assumed to be erroneous for some unknown reason) or to create models to explain the apparent deviation of specific data points from the main trend. For example, if a cluster of data points generated from wood samples appears to be too old relative to the apparent trend in the data, then an 'old wood' model can be created to explain this deviation. In effect, such models shift data points that are apparently off the main trend so that they are closer to it.
Of course, there is a clear problem with this approach. How can we know if the old wood model, or any other kind of model used to manipulate the data, is correct? The clear answer is that we usually cannot know. In fact, we have the same problem with every data point, even those that lie closer to the apparent trend in the data. This is because, as already stated, the true uncertainty in every data point is actually unknown and likely to be underestimated by the intrinsic uncertainty in the radiocarbon measurement alone. Thus, while the use of Bayesian modelling in producing age-depth models is the best approach in principle, since not all sources of uncertainty or deviation can ever be known perfectly, Bayesian methods (or any other weighted regression method that uses the intrinsic measurement error bars, for that matter) can often produce misleading results in practice.
A similar problem also affects simple regression modelling of age-depth radiocarbon data. That is, the true uncertainty in any data point is usually under-represented by the intrinsic radiocarbon measurement, and therefore a simple linear fit to the data, or a polynomial fit, that is weighted by the intrinsic measurement uncertainty will likely under-represent the true uncertainty in the data. Again, just as for Bayesian modelling, regression modelling can be adjusted by using models, like an old wood model, for specific data points. But the same problem persists; how do we know if these models are correct? The answer is we don't, usually, so we have to remember the age-depth model is just a model.
In some cases, such as for the plot above by Gill et al. (2009), as the scatter in the data is far greater than the intrinsic measurement error and no models are used to adjust the data, it is better to use an unweighted regression method. In this case, the uncertainty in the data is determined self-consistently by the scatter in the data and not by any intrinsic measurement error.
Since section 5.1 has already been examined in an earlier blog post (it deals with the date of the onset of the YD period) and section 5.2 is irrelevant to this issue (although we will come back to it later in another context), we begin with section 5.3. My comments are in italics as usual.
5. Inadequate Dating and Stratigraphic Context
As suggested above, a variety of unsupported or misleading claims about dating of sites
critical to the YDIH permeate the early publications supporting the hypothesis (Firestone and
Topping, 2001; Firestone et al., 2006, 2007). This sort of dating misinformation carries through
many subsequent papers supporting the YDIH (Table 4). The two most critical issues of
chronology are the age of the start of the YDC and the claim that purported indicators are well
dated to the lower boundary of the YDC (i.e., the YDB) (see also Section 2).
Holliday et al. are suggesting here that there is too much uncertainty in some of the claimed YDB sites for them to be useful. Therefore, evidence from those sites should be rejected. This is wrong for reasons that will be explained later. Of course, we must deal with the evidence as it is found, not the evidence we would like. Holliday et al. are being unreasonable.
In fact, opponents of the YDIH need to show that dating of purported YDB sites is obviously mutually inconsistent. To achieve this, they need to show with high confidence that the difference in the age of the YDB at a given site and the suggested age of the YD impact is far larger than the uncertainty in any of these age estimates. For YDIH proponents the converse is true, i.e. the difference in the age of the YDB at a given site and the suggested age of the YD impact should not be far larger than the uncertainty in any of these age estimates. Since there are of order ~50 YDB sites to consider, a difference of order 4 or 5 times the standard error would be worrying. Much more than this would be fatal. However, remember that age estimates based on single radiocarbon measurements are unreliable and should be considered with caution.
The most detailed study of YDB site dating is by Kennett et al. (2015), which established using Bayesian modelling a date for the YD impact of 10,835 BCE +- 50 years (2 sigma or 95 % confidence). However, the radiocarbon calibration curve has been updated since 2015. According to Cheng et al. (2020), using the latest radiocarbon calibration curve this date is shifted to 10,875 BCE +- 50 years. Mostly, the research reviewed below uses the older calibration curve. Regardless, given that Kennett et al.'s (2015) work is based on Bayesian modelling (which uses a regression model weighted by the intrinsic radiocarbon measurement errors as well as various ad-hoc models to adjust specific data points), it likely underestimates the true uncertainty in the YDB date.
5.3. Deficient Dating of YDIH Sites
Dating fundamentals in the context of the YDIH debate are summarized by Holliday et
al. (2014, p 519) but bear repeating here from Holliday et al. (2020, p. 70) given how crucial the
issue is on both sides of the debate: “Reliable and precise numerical age control for stratigraphic
sections and associated samples is a key component of the YDIH debate. Proponents
recommend ‘very high chronological resolution to test the hypothesis’ [Kennett et al., 2008a, p.
2531].
In fact, while desirable this is not necessary since the precision in the date of any YDB site cannot be imposed in advance. As already stated, we must deal with the evidence as it is found, not as we would like it to be. All we can reasonably expect is that the date of any YDB site is not obviously inconsistent with the expected date of the YD impact. The date of the YD impact will, therefore, largely be determined by the ages of the most precisely dated YDB sites. Holliday et al.'s demand that ALL YDB sites have precise dates is unreasonable. It seems to be a rhetorical tactic. And, as we shall see later, even those sites with large dating uncertainties are very useful.
Furthermore, they argue that ‘only’ radiocarbon dates with precisions of ‘<100 years, and
preferably <60 years’ [apparently meaning 14C years] should be used for dating the YDB layer
and complain that many dates employed by others have ‘precisions from 200 years to >2,000
years’ [Kennett et al., 2008b, p. E107].
Clearly, more precise dates are more desirable. But less precise dates for the YDB are still useful, as we shall see.
They also propose that the only valid dates are those
processed with ‘modern techniques [e.g., XAD . . . or ultrafiltration].’ Given that the debate is
about whether some sort of extraterrestrial event created an environmental catastrophe at a
precise moment in geologic time, we agree that accurate and high-precision dating is essential
for testing the hypothesis.”
There is no initiating " here, so it is not clear if this is a quote or not. In fact, this claim is incorrect. Even much less precise dates are still useful for testing the hypothesis, as we shall see.
Kennett et al. (2015a, p 4344) wrote, “In a test of synchroneity, it is
ideal to have numerous, highly accurate, and precise dates to develop robust chronological
models.” Unfortunately, no dates used to support the YDIH meet these requirements, and very
few sections or samples are so accurately or precisely dated (Table 4).
Of course, precise dates are ideal. But we must deal with the evidence as it is found, not as we wish it to be. Ultimately, less precise dates are still useful, as we shall see.
LeCompte et al. (2018, p
156) complain that YDIH critics “do not use rigorous dating methods…” A more critical issue is
that the YDIH proponents do not meet this standard. Indeed, “precisions from 200 years to
>2,000 years” (Kennett et al., 2008b, p. E107) characterize the results presented by Kennett et al.
(2015a).
This is true, but not actually a problem. Holliday et. al. are being unreasonable.
Besides misunderstanding and mischaracterizing Clovis archaeology and extinctions
(Sections 1.0, 3.1, 3.2, and 5.7), the basic dating for the hypothesis proposed by Firestone et al.
(2007) was fundamentally flawed at the outset.
The YDIH predicts that the YDB should coincide with terminal Clovis-style artefacts. As this is a prediction that is part of the YDIH, we do not need a precise date for the Clovis termination in advance. All that YDIH proponents need to do is verify this prediction, i.e. show that the data from each site is not inconsistent with this prediction. Remember, radiocarbon dating has many intrinsic uncertainties - see the plot from Gill et al. (2009) above. Therefore, YDIH opponents that claim the demise of Clovis culture was gradual could be placing too much confidence in the radiocarbon dating of Clovis sites, especially if they are based on single radiocarbon measurements at a site, as is often the case.
Regarding megafaunal extinctions, there is good evidence that the YD onset coincided with significant megafaunal extinctions - see a previous blog post on this issue.
More broadly, radiocarbon dating has been a
long-standing conundrum for the YDIH (see Firestone and Topping, 2001; Southon and Taylor,
2002; Firestone 2009a,b, 2014; Gillespie 2009; Melott et al., 2015). Firestone et al. (2007, p
16017) state that “Ten Clovis and equivalent-age sites were selected because of their long established
archeological and paleontological significance, and, hence, most are well
documented and dated by previous researchers.” This is not the case, as thoroughly discussed
and documented by Meltzer et al. (2014) and summarized in Table 4. At best, only three of the
sites (Blackwater Draw, Murray Springs, and possibly Daisey Cave) have reasonable age control
and four have very poor to no age control (Chobot, Gainey, Morley, and Wally’s Beach).
By "age control" Holliday et al. mean precise radiocarbon dates for the YDB. But as already stated, we must deal with the evidence that is found, not the evidence we would like. In fact, even much less precisely dated sites are still useful, contrary to Holliday et al. Moreover, since the end of Clovis culture is a prediction of the YDIH, precise dates for terminal Clovis artefacts are not needed. If the YDB coincides with terminal Clovis artefacts then that aspect of the YDIH is verified.
Firestone et al. (2007) also allude to stratigraphic correlation with and sampling of their
purported YDB in 15 Carolina Bays but provide no stratigraphic nor geochronologic data (Table
4).
The main dating evidence is in Kennet et al. (2015) which supersedes Firestone et al. (2007).
Subsequent investigations of the YDIH produced additional attempts at age control based
on field samples or models (e.g., Firestone et al., 2007, 2010a; Bunch et al., 2012; Israde-
Alcántara et al., 2012, Kennett et al., 2009a,b; LeCompte et al., 2012; Wittke et al., 2013a; Wu et
al., 2013). From among these publications, the dating of 29 sites was evaluated by Meltzer et al.
(2014).
Again, the most comprehensive investigation of the dating of the YDB is by Kennett et al. (2015), so Meltzer et al. (2014) is redundant. But let's look at their claims anyway.
As summarized in their abstract (p E2162) “Several of the sites lack any age control,
others have radiometric ages that are chronologically irrelevant, nearly a dozen have ages
inferred by statistically and chronologically flawed age–depth interpolations, and in several the
ages directly on the supposed impact layer are older or younger than ∼12,800 calendar years
ago. Only 3 of the 29 sites fall within the temporal window of the YD onset as defined by YDIH
proponents.” Further, Meltzer et al. (2014, p E2169) note “We even relaxed one of their criteria,
namely that ‘only 14C dates with measurement precisions <100 years, and preferably <60 years,
should be used’ in assessing the supposed impact chronology and its potential effects [Kennett et
al., 2008b]. Had we applied it, we would have had to discard all luminescence ages and almost
60% of all radiocarbon ages used by YDIH proponents. Doing so would have instantly removed
all radiometric age control from 11 sites and left 8 more with only a single age that in no case
dates to the YD onset, meaning that 19 of their 26 sites with radiometric ages (group 1b) would
become essentially free floating chronologically.”
ENDNOTE 6
The approach taken by Meltzer et al. (2014) was criticized by Sweatman (2021, p 15-16,
20). He wrote “no standard errors were provided for their calculations. It is therefore not
possible to determine if any of these age differences are significant. In a technical sense,
therefore, their data is meaningless and their conclusions cannot be supported” (emphasis added).
This comment misses several key points and is factually untrue.
Wrong. It's crucial and factually true. As we shall see, it appears that neither Meltzer et al. (2014) nor Holliday et al. understand how to create proper regression models.
The text (Meltzer et al., 2014, pp
E2167-E2168) includes discussion of error and uncertainty
This is misleading. Yes, Meltzer et al. (2014) include error estimates in individual measurements, but crucially they omit error estimates in the regression lines applied to this data. Therefore, no conclusions can be drawn from those regression lines. Essentially, Meltzer et al. (2014) is not a work of science and it should be withdrawn, along with Gill et al. (2009). I don't know how these works were published - the referees must have been asleep!
and the Supplemental Data clearly
includes the standard errors in their calculations for all 29 YDIH sites reviewed.
This is misleading. As stated above, standard errors are provided for individual measurements but not for any lines of regression. Thus their conclusions are unsupported.
To be clear, the full list of tables and figures in the supplemental information of Meltzer et al. (2014) is as follows. Tables and Figures 1-16 list individual measurements for YDB sites. Table 17 lists model data for their alternative age-depth modelling. Table 18 lists the regression coefficients for their regression lines for each site, but no uncertainty estimates in these regression coefficients are provided. Table 19 lists r^2 and p values for their regression lines, but this has little to do with uncertainty estimates for the regression coefficients.
In fact, the r^2 value is an indication of how well a linear fit represents the underlying data. In other words, it's a measure of the scatter in the data relative to the fitted line. Although it is related to the uncertainty in the linear coefficients of the fitted line, it is not the uncertainty itself. In any case, Meltzer et al. ignore the r^2 value in making their conclusions. Moreover, the p-value is a measure of the probability that the linear model actually reflects a trend in the data, i.e. is there actually a linear trend or is the trend line spurious? Once again, this is only tangentially related to the regression line coefficients. Ultimately, Meltzer et al. (2014) make no use of these statistical measures in their conclusions, which in any case are not the uncertainty measures needed.
Indeed, simple inspection of the scatter in the underlying data in many of Meltzer et al.'s (2014) plots shows that there is considerable uncertainty in these regression lines that likely far outweighs the uncertainty in the date of the YDB provided by Kennett et al. (2015). It is therefore crucial that it considered.
The ages just
had to be fully outside the range of the YD/GS-1 onset age (~12.9 cal ka BP) to show that the
markers of a YDB impact did not occur in the profile when/where they were supposed to occur.
This is false and is a clear deficiency in Meltzer et al.'s (2014) and Holliday et al.'s understanding of the statistical analysis of experimental data, which is a basic requirement for any scientist. This is surprising considering so many senior scientists appear as co-authors on these papers. Did they not all check this section?
In fact, Meltzer's et al. (2014) age-depth modelling and analysis is flawed in two fundamental ways. First, they take no account of the uncertainty in their fitted regression lines, as already mentioned. Second, they make age-depth models and perform regression modelling using calibrated age estimates rather than raw C14 radiocarbon measurements. This second issue is very important for the following reason. Consider an example where 1 million radiocarbon measurements were made and each was converted to a calendar age using a calibration curve. In this case the line of best fit to the calendar age data would have only a tiny uncertainty - due to the very large number of data points. Indeed, the regression uncertainty would be far less than the radiocarbon calibration uncertainty. But this is erroneous because radiocarbon calibration uncertainty is immutable - it is simply not possible to eliminate or reduce radiocarbon calibration uncertainty this way. Thus, regression analysis, including uncertainty analysis, must be performed with the raw C14 radiocarbon data. Then, all this data (line of best fit plus their uncertainty estimates) should be converted to a calendar age in a final step through proper convolution of the uncertainty with the calibration curve uncertainty.
This point is illustrated above using Meltzer et al.'s (2014) age-depth data for Abu Hureyra (see their supplemental data). I have taken the raw C14 radiocarbon measurements provided in Table S1 of Meltzer et al. (2014) (blue symbols plotted with 1 sigma error bars). Then I determine both weighted (red line) and unweighted (blue line) linear regression models using the method of least squares. The weighted regression line takes account of the intrinsic radiocarbon error measurements. We find that most of the radiocarbon measurements fall outside of 1 sigma standard deviation from the red regression line. This strongly indicates that the radiocarbon measurement error bars do not fully account for all the uncertainty in this data (as mentioned above, this is a common occurrence and not unexpected). Therefore, it is more appropriate to use an unweighted fit and to use the standard deviation of the radiocarbon measurements relative this unweighted regression line to estimate the true measurement error. This is found to be ~ 300 radiocarbon years for each data point. Using this new measurement uncertainty the orange lines in the plot above depict linear fits that are within 2 sigma (95% probability using a normal distribution) of the unweighted best fit line. That is, they define a 2 sigma (95%) envelope of uncertainty in the best fit line.
Meltzer et al. (2014), supported by Holliday et al., claim that YDIH proponents need to show that the position of the YDB in the sediment is between the blue dashed vertical lines above, since these represent the (2 sigma) uncertainty in the date of the YDB (represented by the horizontal blue bar) against the (blue) unweighted regression line. But this neglects the uncertainty in the regression line itself, illustrated by the envelope of the orange lines. When this uncertainty is taken into account, YDIH proponents need only to show the position of the YDB in the sediment is between the red dashed vertical lines, a much less demanding requirement (at 2 sigma). The depth of the YDB at Abu Hureyra is stated to correspond to 284.7 masl (Wittke et al. and Moore et al. but not Bunch et al.). We see from the above plot that this depth is only marginally consistent at 2 sigma with the age of the YDB determined by Kennett et al. when the regression uncertainty is omitted as in Meltzer et al. (2014) (i.e. the blue lines only just straddle 284.7 masl). However, it is easily consistent at 2 sigma with the age of the YDB determined by Kennett et al. (2015) when the regression uncertainty is included (i.e. the red lines easily straddle 284.7 masl). Thus, the regression uncertainty must be considered because it much larger than the YDB age uncertainty.
The above analysis is appropriate for the simple kind of regression modelling (linear fitting) employed by Meltzer et al. (2014) and Gill et al. (2009). It is intended to show their understanding of the statistical analysis of experimental data is lacking. Kennett et al. (2015) used a more sophisticated Bayesian modelling technique for dating YDB sites, i.e. their models include additional information such as assumptions about the time-order of the samples. In principle this is a much better approach. But in practice, for the reasons discussed above, it will likely over-constrain the age-depth model leading to confidence intervals that are too small.
The issue is not whether their results and the cases they re-analyzed were significantly different.
Further, in the main text and the Supporting Data, Meltzer et al. (2014) provide ample discussion
of their methods and their statistical significance.
The only values relating to statistical significance provided by Meltzer et al. (2014) are the r^2 and p values provided. These have little to do with the uncertainty in the regression parameters as stated above. In any case, these statistical measures are not used by Meltzer et al. (2014) in framing their conclusions.
Like the alleged issue of incorrect sampling at
Arlington Canyon (Section 4.1), Sweatman (2021) and other YDIH proponents never address the
problematic nature of many of the samples, sample contexts, and resulting dates. In a technical
sense, therefore, Sweatman (2021) simply dismisses a carefully laid out analysis using an
irrelevant technicality.
It is amazing to see that Holliday et al. label Melter et al.'s (2014) analysis as 'careful' and the required regression coefficients as an 'irrelevant technicality'. In fact, the opposite is true, as shown above. I repeat, no conclusions can be drawn from Meltzer et al.'s (2014) analysis for the two fundamental reasons described above. Kennet et al. (2015) made similar criticisms of Meltzer et al. (2014), but they have been ignored by YDIH opponents, like Holliday et al., who continue to promote their junk. The fact is, Meltzer et al. (2014) is not scientific and should never have been published. Holliday et al.'s defence of Meltzer et al (2014) suggests they do not understand even basic concepts in data analysis, which is the basis of all science, thus undermining their credibility as competent scientists.
Sweatman (2021) similarly criticizes a box plot (from Holliday and Meltzer 2010, figure
3; Holliday et al., 2014, figure 2) of radiocarbon dates on the “black mat” (Section 6). Sweatman
(2021, p 16) comments that “much of the data in this plot is considered unreliable or is
unpublished.” ENDNOTE 7. All data in the figure are directly from the published citations in
Holliday and Meltzer (2010, figure 3 caption). The integrity of those dates can be evaluated from
the information within those cited sources. However, Sweatman (2021) does not elaborate on his
perceived unreliability of “much of the data.” Only two examples of problematic dates are
offered, Naco and Willcox. The Naco date was included inadvertently in Holliday et al. (2014)
and should be discarded.
Thus confirming Sweatman's (2021) point.
Sweatman’s discussion of the dating of the Willcox section provides
further circular reasoning (Table 2).
But this circular reasoning is not explained by Holliday et al.
Besides, Kinzie et al. (2014, p 478) cite unpublished data
from three sites with nanodiamond horizons purported to support the YDIH and employ circular
reasoning (Table 2).
Holliday et al. once again fail to appreciate that since the Clovis termination is a prediction of the YDIH, dating for terminal Clovis artefacts is not needed. This is a feature of the YDIH, not circular reasoning.
Shortly after the publication of the dating critique by Meltzer et al. (2014), Kennett et al
(2015a) published a paper on a Bayesian chronological approach for estimating the ages of
claimed YDB zones from many of the sites examined by Meltzer et al. (2014).
Yes, this is the key paper.
Thirty-two sites
are listed and discussed. The dating at nine is of such poor quality that Kennett et al. (2015a)
could not include their results, but they were still claimed to be YDB via circular reasoning
(Table 2). In a brief response to a critique of their Bayesian modeling, Kennett et al. (2015b,
E6723) dismiss the criticisms by stating that these “same claims previously were presented in
Meltzer et al., [2014] and were discussed and refuted in Kennett et al., [2015a] …” But in fact,
few of the criticisms enumerated by Meltzer et al. were even addressed by Kennett et al. (Table
1).
This is probably because Meltzer et al. (2014) is another Gish gallop. To refute a Gish gallop it is not necessary to deal with every spurious claim. It is sufficient only to show there are basic flaws in their understanding or reasoning. As shown above, and as already argued by Kennett et al. (2015), Meltzer et al.'s (2014) work includes fundamental flaws that render its claims unsupported. Holliday et al.'s defence of it can be regarded as pseudoscience.
YDIH proponents (e.g., Wolbach et al., 2018a,b; Sweatman 2021; Mahaney et al., 2022;
Powell 2022) largely accept the work of Kennett et al. (2015a), suggesting that the YDB is
synchronous across four continents, and thereby assert that the impact indicators were deposited
synchronously over four continents.
This is misleading. The claim in Kennet et al. (2015) is that the dating of the YDB on four continents is not inconsistent with the hypothesis. Synchroneity is the most likely explanation for this, as discussed by Sweatman (2021), but we will come to this argument in detail later.
Mahaney et al. (2022, p 17) states that the conclusions of
Meltzer et al. (2014) were “refuted using Bayesian statistics by Kennett et al., [2015a].” But like
Kennett et al. (2015a), those proponents fail to recognize or refute the identification of many
problems with the original site contexts of the dating discussed in detail by Meltzer et al. (2014;
with over 60 pages of text and tables) and by Holliday et al. (2014, 2020).
With over 60 pages of text and tables, but an obvious lack of scientific rigour, Meltzer el al. (2014) is clearly another Gish gallop. As described above, their work is junk and should be withdrawn.
Sweatman (2021, p
16) focuses on eight “high-quality sites” (using the classification of Kennett et al., 2015a for
ranking the chronologies at the 23 sites included in the age estimate). But several of these high quality
sites are problematic.
Let's see.
Six were claimed to produce “radiocarbon dates from directly
within the YDB layer” (Kennett et al., 2015a, table 1). However, one date is from the Bull Creek
site, where the radiocarbon date is stratigraphically above a claimed “impact indicator” spike
(Section 5.5, Table 5) and, ironically, YDIH proponents often use such spikes to identify the
YDB (Table 2).
As we have already seen, neither Meltzer et al. (2014) nor Holliday et al. properly understand how to treat uncertainty in regression analysis. A regression analysis like that above for Abu Hureyra will likely show that for Bull Creek the YDB indicated by the impact proxies is consistent with the claimed date of the YD impact. This is all that matters. Issues surrounding the Bull Creek data are discussed in detail in Holliday et al.’s Section 5.5 below.
Two dates (from Barber Creek and Blackville) have standard deviations >700
years (i.e., very poor precision) and thus no evidence whatsoever that the claimed YDB zone is
of YDB age.
This is false. As we shall see later, even sites with significant dating uncertainty are useful.
Two other dates (from Aalsterhut, Lingen) are for the Usselo soil in northern
Europe and of YDB age but selected from among scores of dates for dozens of sites falling far
outside of the YDB (e.g., Hoek 1997; Kaiser et al., 2009) (further discussed in Section 5.6).
Picking out dates that are conveniently YDB age has no relevance to the YDIH debate and,
moreover, is scientifically unsound.
We will look at these sites in detail later in section 5.6.
These examples of just the so-called “high-quality sites”
well demonstrate that statistical analyses, Bayesian or otherwise, cannot overcome poor sample
context, selection or precision, previously published flawed age–depth interpolations, or
unexplained and inappropriate rejection of published dates. This is a classic example of the use
of poor data resulting in the production of poor statistical results. ENDNOTE 8
Holliday et al. have not shown that the dating of any YDB site is inconsistent with the YDIH. This is all that matters. Considering that neither Meltzer et al. nor Holliday et al. even understand the concept of uncertainty in regression analysis, as shown above for Abu Hureyra, we can consider their comments as spurious.
The Table above refers to nanodiamond measurements and radiocarbon samples collected from the Bull Creek site by Kennett et al. (2009a) and Bemment et al. (2014).
It is incorrect. The data for Kennett et al. (2009) has been shifted upwards by one cell by Holliday et al.. In fact, the measurement of 25 ppb should be in BC21 and the measurement for 100 ppb should be BC 20. This is confirmed in the text of Bement et al. (2014)
“Kennett and colleagues (4) found a concentration of nds (both n- and cubic forms) centered on the boundary between two soil A horizons interpreted to be the YDB and equivalent to our samples BC20 and BC21.”
It is not clear why Holliday et al. have shifted these results upwards, but we expect this is the reason for their confusion. Clearly, when critiquing work it is imperative that the original data is not misrepresented. Any study can be “refuted” if its data is mistreated. This data is discussed again below in Holliday et al.’s Section 5.5.
(Note also that Holliday et al.’s note (4) is also incorrect: this sample is listed as “289-307” by Bement at al. (2014), not “289-298” as Holliday et al. claim. Nevertheless, it is correctly attributed to “298-307” and BC 21. Clearly, Bement et al. (2014) made a typo.)
Sweatman (2021, p 16) goes on to discuss standard deviations of the modelling results of
Kennett et al. (2015a). He seems satisfied with modeled results at 2 or 3 standard deviations (sd)
confirming a YDB age for a sample zone. But such statistical confirmation has nothing to do
with stratigraphic or chronologic reality. An age model uncertainty of 100 years (1 sd) means
that the age of a sample at 2 sd would be within a range of 400 years (or 800 years at 3 sd). Such
broad age ranges cannot confirm the identification of a moment in time in the stratigraphic
record.
Actually, it would be a range of 600 years, not 800 years, at 3 s.d.. While this might be a typo, given the previous evidence it seems more likely this is another basic deficiency in Holliday et al.’s understanding of uncertainty analysis.
All that proponents need to show is that the age of any given YDB site is not inconsistent with the age of the YD impact at some reasonable level of confidence. Demanding that the age range should "confirm the identification of a moment in time in the stratigraphic record" is a meaningless unscientific statement that fails to understand very basic concepts in data analysis and uncertainty.
The ages of the authors of this commentary could be modelled to statistically date to the
signing of the Declaration of Independence (1776 CE). The modelling could be statistically
correct, but obviously meaningless.
I'd like to know what Holliday et al. have been smoking. As the authors' age is known with certainty the comparison is meaningless.
Further problems with the results of the Bayesian age estimation are enumerated by
Holliday et al. (2020, p 70-71, 75). “Modeled age ranges with standard deviations of >300 years
up to 2405 years are presented for layers of claimed impact indicators at nine sites of ‘low
quality’ in terms of dating (their description in Kennett et al., 2015a, table2). These layers are
argued to represent the YDB based solely on the premise that if they could be YDB, they must
be the YDB” (Holliday et al., 2020, 70-71, table 5) (Table 2).
This is false. No YDIH proponent has made this argument. It is an invention of Holliday et al.
In fact, the proposal that the YDB is synchronous on four continents is based on a statistical argument, presented later.
Their conclusion also suggests that
the modelled dating of 12,255 +/- 2405 cal yr BP for the sample zone at Melrose is somehow
proof of a YDB age, which is of course preposterous.
This is nonsense and highlights Holliday et al.'s basic misunderstanding of the YDIH and its dating. As already stated, the proposal that the YDB is synchronous is based on a statistical argument and is given later.
There is also a continuing and troublesome
problem of omitting radiocarbon ages without explanation or analyses, which is a problem in
much of the YDIH literature (enumerated by Meltzer et al., 2014; Boslough et al., 2015;
Holliday et al., 2020) (Table 4).
This is simply repetition, and this complaint has already been dealt with above.
Sweatman (2021) offers several contradictory and inconsistent concluding statements of
sorts regarding the dating of the YDB. “No YDB site has yet been found to be obviously
inconsistent with a synchronous event circa 10,785 ± 50 cal BP (2 sd).” (p. 19; see also p 17).
To my knowledge, Sweatman (2021) is correct when one takes into account a proper treatment of the uncertainty in the YDB site regression modelling. See the example for Abu Hureyra above. Holliday et al. continue to show they do not understand basic concepts in data analysis, which undermines their credibility as scientists.
Besides the problematic date notation and apparent typographical errors (Section 5.1), this
statement is a non sequitur.
Nonsense. Holliday et al. have already shown they do not understand even basic concepts in uncertainty analysis.
The YDB of the YDIH, especially when it is claimed to contain
“impact markers,” must (by definition) represent a synchronous moment in time, dated to
~10,785 14C yr BP.
This statement is scientifically meaningless as no uncertainty range is provided by Holliday et al., once again demonstrating their incompetence.
To be a “YDB site” the site must clearly contain a zone accurately dated to
the YDB with high precision.
This is false. To be a potential YDB site it should i) contain a zone with impact markers consistent with those already found at other YDB sites, and ii) have a date that is not obviously inconsistent with the YD impact or be found with other proxies that are predicted to indicate the YDB, such as terminal Clovis artefacts.
As pointed out above and in Sections 5.4 to 5.7, and Table 4, many
claimed YDB zones and evidence for a synchronous “event” are not so or are not clearly shown
to be so.
The argument about whether any site is likely to be synchronous with the YD impact is statistical and given later. YDIH opponents have not shown that any of these YDB sites are obviously inconsistent with a YD impact age.
He also comments that dating at eight “high-quality sites” among the 23 dated “is
consistent with a synchronous event, which suggests all YDB sites are likely synchronous” and
“it would be surprising if the others were not all eventually found to also be consistent” (p 16).
These are baseless assertions that defy fundamental principles of objective science.
This is nonsense and highlights Holliday et al.'s basic misunderstanding. A statistical argument, given later, is used to determine whether such sites indicate a synchronous event. Sweatman's (2021) statements are fine.
Taken to its
logical conclusion, this assertion argues that there is no longer a need to date archaeological sites
or geologic sections. We can just assume high-precision dating by correlation.
Again, this is nonsense for all the same reasons as above.
Given that the
burden of proof is on the YDIH proponents, dating results from one site or a group of sites does
not confirm the dating at others. It only provides testable hypotheses, of the kind evaluated by
Jorgeson et al. (2020) (Section 5.8).
Ok, so let's now deal with the argument about the synchroneity of the YDB across four continents.
The argument is actually very simple. Earlier blog posts summarize evidence for impact proxies distributed at around 50 YDB sites on four continents. In each case the evidence indicates significant disturbance (melting, vaporization etc) of the ground, although no craters of YDB age are yet confirmed. Therefore, this evidence suggests a multitude of low-altitude airbursts distributed across several continents with a wide range of impact energies. Some were likely much larger than Tunguska, i.e. super-Tunguskas, because of the presence of significant melt glass as these sites. Note that melt glass is not apparent at Tunguska.
It's important to realise that the usual background rain of cosmic dust or the burn up of small meteorites at high altitude or even high-altitude airbursts cannot explain the YDB debris, since the YDB debris mainly has the same composition as Earth's crust and is generally quite different to meteoric material. Therefore, only major, low-altitude airburst or ground impacts can explain the YDB debris.
Summarizing the data for all YDB sites so far, we find the following;
8 sites consistent with the YD impact date with uncertainty (95%) < 200 years
10 more sites consistent with the YD impact date with uncertainty (95%) < 400 years
9 more sites consistent with the YD impact date with uncertainty (95%) < 1600 years
These sites are as follows;
Holliday et al. suggest themselves the Tunguska event was a 1 in 1000 to 1 in 10,000 year event. Clearly, the probability of so many Tunguska-like impacts within such a relatively short timespan suggests it is extremely likely that that these impact proxies were generated synchronously by a single impact event or a series of related and highly correlated events within a short timespan, i.e. the probability of these being independent Tunguska-like events (the null hypothesis) is extremely low.
This reasoning can be made statistically robust using a Poisson distribution. Let us suppose that 0.9 Tunguska events are normally expected within 900 years, conservatively taking the higher impact rate stated by Holliday et al. for Tunguska-like events. Then, the probability to find at least 18 events within a 900 year interval is 1 . 10^-17. This is essentially 0. However, 8 of these sites have even smaller age ranges (< 400 years), so this figure is an overestimate.
Even if we allow for a much higher impact rate of Tunguska-like impactors, say 1 every 500 years, we still find a vanishingly-low probability of 1 . 10^-12. Thus, a multitude of unrelated Tunguska-like impactors cannot reasonably explain the YDB.
But let's consider another model. Let's instead suppose all the YDB impact debris was created by several independent super-Tunguska airbursts that interact strongly with the ground, each with a supposed impact rate of 1 in 10,000 years, that can distribute debris over several thousand km. We would need a super-Tunguska near Abu Hureyra to account for the melt-glass found there. Three more in Western Europe, Central America and South America to account the impact debris there. Another on the East Coast of the USA to account for the melt-glass there, another on the West Coast of the USA to account for impact debris there and another near Chicago to account for the northern debris. The date resolution for each of these sites must correspond to the lowest range of all the associated YDB sites covered by each super-Tunguska impact. Therefore, we have;
Again, using the Poisson distribution the probability to find at least 7 such events within a period of 720 years is 2 . 10^-12. Therefore, we can discount this scenario too. It is extremely unlikely. Even if we use double the impact rate for these large, low-altitude airbursts we still have a probability of 2. 10^-10.
Furthermore, YDB sites are found almost everywhere we have looked on these four continents. This indicates they are very common - they are not hard to find. Therefore, we can expect the final number of YDB sites to number in the hundreds or thousands, not just ~50. This is a robust statistical argument because YDB sites are easy to find. So the probability estimates above will almost certainly be too large by very many orders of magnitude, i.e. the probability an entire hemisphere of Earth was bombarded with hundreds or thousands of unrelated Tunguska-like impacts within a thousand years is so small it can be neglected. Instead, the event was obviously a single large event or series of highly correlated closely related events.
Essentially, there is no way that a common layer of impact proxies dated to within a thousand years across four continents can signal anything other than a singular event, or highly correlated series of events.
As a check, let's consider a YDIH impact model which posits a single event, or series of highly correlated events in a short timespan, with a total magnitude of around 10,000 Mt, which corresponds to ~ 1000 Tunguskas or 10-100 super-Tunguskas, According to Figure 2 in Boslough et al. (2012), shown below, such an event is expected to occur about every 200,000 years. Therefore, the probability of one such event in the last 13,000 years is about 0.06. This is quite reasonable, and should be vastly preferred to the other models discussed above. However, the figure below shows the expected average impact rate over the last few hundreds of million to billion years. Impact rate fluctuations, i.e. coherent catastrophism, can be expected on shorter timescales when giant comets intrude into the inner solar system. Indeed, a large impact rate fluctuation is expected in recent millennia given the existence of the very large and old Taurid meteor stream and the relatively dense zodiacal dust cloud. So, the correct local impact rate could be perhaps 10 times higher, making the Younger Dryas impact a reasonably probable event.
Is such a model, i.e. a distributed event consisting of 10-1000 near simultaneous large impacting chunks of comet, reasonable? We will examine this question in the next blog post.
It follows that, provided the YDB comes into contact (above) with terminal Clovis artefacts, and those artefacts never appear significantly above the YDB, we can reasonably conclude the YD impact terminated the Clovis culture (but not, perhaps, the entire regional human population). If this is true, then separate dating of the end of Clovis culture is not needed, as it will obviously be coincident with the age of the YD impact.
5.4. Poorly Dated Platinum Anomalies
Platinum (Pt) anomalies are used by YDIH proponents to unquestioningly support the
YDIH (see Section 11). The Pt anomaly in the GISP2 ice core (Petaev et al., 2013a), which is
proclaimed by Sweatman (2021, p 20) to be “probably the most significant [YDIH] evidence so
far” post-dates the onset of the YD/GS-1 (see Section 5.1).
In fact, it post-dates an apparent signal in dD in the NGRIP ice core by around 30 years. But the definition of the YD onset by this criterion alone is contentious. See an earlier blog post about this issue.
Moore et al. (2017) report a
widespread Pt anomaly at the YDB in 11 sites across North America. Powell (2022, p 24) and
Sweatman (2021, p 17) uncritically accept those interpretations, the latter going so far as to
conclude that the Pt zone can be used to “unambiguously identify the Younger Dryas boundary
at many locations around the globe.” Unfortunately, the dating reported by Moore et al. (2017)
suffers from many of the same problems of chronology that plague the original YDB sites in
Firestone et al. (2007) (Section 5.3) (Table 4).
Let's see.
As is obvious from Moore et al. (2017,
Supplementary Information), only three sites seem to have direct dating for the Pt anomaly, but
only near the YDB (Arlington Canyon; Murray Springs, Blackwater Draw). However, Arlington
Canyon shows considerable vertical and lateral facies variation (Section 4.1). As for the other
sites, five are indirectly inferred dates based on archaeology (two of which are based on optically
stimulated luminescence (OSL) with large standard deviations), two have no numerical age
control whatsoever, and one (Sheriden Cave) has a very confusing geochronological context
(Table 4).
For the reasons given above, these arguments can be disregarded. See also an earlier blog post about the platinum data. The fact is that Moore et al. (2017) found a platinum anomaly exactly where it is predicted to occur at 11 YDB sites. The prediction is that it should occur in the same debris layer as an abundance of nanodiamonds and/or impact spherules or be coincident with terminal Clovis artefacts, and this is confirmed in every case. Holliday et al.'s belief this is merely a coincidence is preposterous.
Repeating Moore et al.’s (2017) claim, Powell (2022, p 24) states that the Pt anomaly at
three sites (Murray Springs, Blackwater Draw, Sheriden Cave) is “associated with Clovis
artifacts, representing the level at which the Clovis culture disappeared” and other sites in the
southeastern U.S. “are poorly or not directly dated and lack the black mat, but do provide a
coherent Clovis archaeological record.” Sweatman (2021 p 5) highlights the Pt anomaly at “the
Flamingo Bay site in South Carolina” claiming “a platinum abundance nearly 100 times the
average crustal value was found in association with the youngest Clovis artefacts.” But the
context of the archaeology and the Pt zones is both mixed and confused, Table 4 and Moore et al.
(2017, p 6). Setting aside the YDIH requirement that the YDB represents the termination of
Clovis, a basic tenant of many versions of the YDIH (Section 3.1), and the likelihood that the
“youngest Clovis” would be post-YDB (Waters and Stafford 2007, Waters et al., 2020),
both authors neglect to note this comment from the main text in Moore et al. (2017, p 6) “Early
Archaic artifacts in the same levels as Clovis at Flamingo Bay” indicate that “these surfaces were
stable to slowly accreting for several millennia before being buried incrementally through a
combination of slopewash and aeolian accretion…” The authors seem to believe the self contradicting
notion that a mixed archaeological assemblage spanning thousands of years of
Clovis and Early Archaic time somehow provides a precise stratigraphic age indicator for the
YDB.
Holliday et al. are mistaken, as there is no contradiction here at all. To refute this aspect of the YDIH opponents need to show that Clovis artefacts clearly occur above the YDB. To my knowledge this has never been shown because, typically, investigators of Clovis culture reject the YDIH a priori and never test this proposition. The YDIH says nothing about the effect of the YD impact on non-Clovis cultures or the timing of other such cultures.
Further, Moore et al. (2017, Supplementary Information, p 5) observe “Many sandy sites
in the eastern US contain Paleoindian and Early Archaic components within the same
stratigraphic zone or with very little separation (e.g., Topper, Kolb, and Flamingo Bay). As a
result, Pt anomalies may be expected to occur in some sites within stratigraphic sequences that
contain both Paleoindian and Early Archaic artifacts or with Early Archaic artifacts sitting
immediately above YD-age sediments. Archaeological occupations at Squires Ridge, beginning
with Early Archaic side-notched stone tool industries, are found only within and above the
deepest Pt anomaly and only pre-cultural, archaeologically sterile zones lie underneath the
deepest Pt anomaly. This is consistent with post-depositional processes and reworking of Pt enriched
sediments during periodic landform aggradation events during and after the YD event”
(emphasis added). This astonishing admission demonstrates that the Pt zone is mixed among
Early Archaic (post-YD/GS-1) artifacts and thus does not represent a discrete stratigraphic
context, nor (contrary to Powell, 2022, p 24) a “coherent Clovis archaeological record” (Table
4).
For the same reasons as above Holliday et al. are mistaken. There is no contradiction with the YDIH here at all.
This conclusion is yet another among publications where dating is based on assuming that
because a zone could be YDB in age, it must be the YDB (Table 2).
This is misleading for the same reasons as given above. The statistical argument given earlier shows the only reasonable interpretation is that the YDB signals a singular event or highly correlated series of closely-timed Tunguska-like events. Since the platinum signal is found in the same sediment layer as YD impact proxies across four continents, and it is expected to be global given the GISP2 signal, and platinum is also a good indicator of an ET impact, and the GISP2 signal is thought to indicate a cosmic impact, it is clearly associated with the YD impact.
In other words, there is no
clear age control. Given the dating problems noted above, the reference to “widespread platinum
abundance in bulk sediments near the base of YD-age black mats on at least four continents,
confirmed by several independent research groups” (Sweatman 2021, p 17) is not supported by
the evidence.
This is false. Holliday et al.'s view that these YDB sites are unrelated is preposterous and can be disregarded via the simple statistical argument above.
5.5. Inconsistent Dating of Nanodiamond Zones
Firestone et al. (2007) claim recovery of nanodiamonds from the YDB but present no
data. Kennett et al. (2009a) first presented data claiming recovery of nanodiamonds in purported
YDB zones. They discuss six sites across North America but provide plots (their figure 1) from
only three sites (with no context on stratigraphy or depths) (Table 4). The Bull Creek site in
Oklahoma is one of the three sites. Subsequent searches for nanodiamonds at Bull Creek were
reported by Bement et al. (2014), Kinzie et al. (2014), and Sexton (2016). The research and
discussions by the three groups is confusing and contradictory, however (see ENDNOTE 9 and
Section 12.6).
See also an earlier blog post on the issue of YDB nanodiamonds.
Originally, Kennett et al. (2009a) claimed a nanodiamond peak of 100 ppb at 13.0 ± 1
cal. yr BP with ≈ 25 ppb of nanodiamonds at ≈ 10 cm above that level (plotted in their figure 1).
Kennett et al. (2009a, figure 1 caption) write, “Stratigraphic profiles [showing no stratigraphy]
on left show NDs only in the YDB” and hence identifies the YDB as spanning those two levels.
Subsequently, Bement et al. (2014, table 1) reported orders of magnitude greater abundance of
nanodiamonds at 307-312 cm below surface (cmbs) but that layer was undated. Bement et al.
(2014, table 1) identified the layer 298-307 cmbs as corresponding to the 100 ppb nanodiamond
peak layer of Kennett et al. (2009a, figure 1) (Table 5) and also dated it to 11,070±60 14C yr BP
(~12,990 cal yr BP). However, Bement et al. (2014, table 1) attributes to Kennett et al. (2009a),
without explanation, a different shaped nanodiamond peak than what is plotted in Kennett et al.
(2009a, figure 1). The attributed peak has 100 ppb of nanodiamonds at 298-307 cmbs and 90
ppb at 307-312 cmbs. This attributed peak now overlaps with Bement et al.’s main peak position
at 307-312 cmbs, whereas the peak plotted by Kennett et al. (2009a, figure 1) plots 0 ppb of
nanodiamonds below the main 100 ppb peak and does not overlap with Bement et al.’s main
peak. Thus, Kennett et al. (2009a) illustrate a YDB spanning at least 10 cm at and above a date
of ~13 cal ka BP whereas Bement et al. (2014, table 1) identify a YDB spanning 14 cm at and
below the ~13 cal ka BP date, with the orders of magnitude predominant peak in nanodiamonds
clearly below that date (see further discussion in ENDNOTE 9).
Kennett et al. (2009a, 2015a), LeCompte et al. (2012), Bement et al. (2014), Wolbach et
al. (2018b, 2020), West et al. (2020a), Powell (2020, 2022), and Sweatman (2021) all accept Bull
Creek as evidence in support of the YDIH. However, as discussed in Section 12.6, the purported
YDB nanodiamond concentration measurements are not credible. But even if the concentration
measurements are accepted as accurate, as believed by YDIH proponents, the depth of the
nanodiamond peak layers are clearly below the presumed YDB at Bull Creek and are
inconsistent with the YDIH. ENDNOTE 9
First, note that Bement et al. (2014) only sampled the YDB in the same “locality” as Kennett et al. (2009); they did not sample (to our knowledge) precisely the same sediment column. Therefore, deviations in measurements can occur due to small fluctuations in stratigraphy and concentrations between the sediment columns sampled. Taking this into account, let’s consider the data in question.
Kennett et al. (2009) identify the following nanodiamond concentrations near the YDB: 1) ~ 100 ppb corresponding to a radiocarbon date 13 +- 0.1 cal yr BP, and 2) ~ 25 ppb in a section ~ 10 cm above that. Each section is ~ 10 cm thick. No depth information is provided.
Bement et al. (2014) identify the following nanodiamond concentrations near the YDB: 1) 1.9 ppm in a section dated to 11070 +- 60 RCYBP at 298-307 cm, and 2) 190 ppm in an undated section at 307-312 cms.
If we back-convert the calibrated radiocarbon date in Kennett et al. (2009) using the 2013 calibration curve we find an uncalibrated radiocarbon age of approximately 11100 +- 25 RCYBP. Bement et al. (2014) assume their undated nanodiamond peak at 307-312 cms corresponds to Kennett et al.’s (2009) nanodiamond peak at 11100 +- 25 RCYBP.
Leaving aside the possible typo in Bement et al.’s (2014) units, there is no inconsistency here. While it is true that Bement et al. (2014) have incorrectly plotted the data in Kennett et al. (2009) and potentially misidentified the units, the two data sets are consistent within error bounds (which are presumably quoted at 1 sd).
Let’s be clear; Bement et al. (2014) located a nanodiamond peak in an undated section directly below another section dated to 11070 +- 60 RCYBP. Kennett et al. (2009), on the other hand, located a nanodiamond peak at 11100 +- 25 RCYBP at an unknown depth. Any inconsistency in this data is in the minds of Holliday et al. only. Especially, when we take into account expected differences due to sampling different sediment columns, this data is perfectly in accord (except for the potential unit typo by Bement et al. (2014)).
Other claims regarding dating are equally curious, if not spurious. Kinzie et al. (2014)
present data from 24 sites purported to show YDB nanodiamond spikes (but see also Table 3 and
Section 12.6 regarding unreliability of those measurements). Eighteen of the claimed YDB zones
are poorly dated, not dated, associated with a disconformity (Table 4), or from the Usselo soil,
which formed through the Allerød and YD/GS-1 (Section 5.6). ENDNOTE 10
While there might be typo in the reported ppb/ppm units in Bement et al. (2014), as argued above there is nothing here that contradicts the YDIH. Holliday et al. seem to have misinterpreted the Bull Creek data, as indicated by their unexplained shifting of the Kennet et al. (2009) data in their Table 5 (shown earlier). The Usselo Horizon is discussed next in section 5.6.
5.6. Logical Lapses in Dating and Interpreting Usselo and Finow Soils
Following on comments elsewhere (Holliday et al., 2020, p 87), YDIH proponents
perpetuate logical lapses in the interpretation of radiocarbon dates and the dating of soils along
with no understanding of soil forming processes in discussions of dates from the Usselo
(Hijszeler, 1957) and Finow (Schlaak, 1993) soils (here simply referred to as the Usselo soil).
These soils are charcoal-rich sands that act as distinct stratigraphic markers within widespread
and genetically related layers of laterally continuous post-glacial eolian “coversand” sheets
distributed across much of northwest and northcentral Europe (e.g., van Geel et al., 1989; Hoek
1997; Vandenberghe et al., 2013; Kaiser et al., 2009; van Hoesel et al., 2012, 2014; Andronikov
et al., 2016a). The soils are not a “charcoal boundary layer” (contra Sweatman 2021, p 15).
An image of the Usselo charcoal layer (horizon) is provided above. There may be some misunderstanding of the labelling of the stratigraphy by some YDIH proponents. Nevertheless, dating of the Usselo/Finow horizon is consistent with the YDIH. Opponents of the YDIH have already shown that they do not understand how to assess the uncertainty in radiocarbon measurements or in related age-depth models.
Proponents of the YDIH (Firestone et al., 2007; Wittke et al., 2013b; Wolbach et al.,
2018b, p. 190) argue that the charcoal in the soil is evidence of catastrophic biomass burning at
the YDB or during the YD/GS-1 (depending on the author). Kennett et al. (2015a, figure 1)
claim Usselo soils at Lingen (Germany) and Aalesterhut (The Netherlands) are of YD/GS-1
onset age based only on two dates from among scores of dates for Bayesian analyses (e.g., Hoek
1997; Kaiser et al., 2009). However, the geomorphologists and soil stratigraphers who were the
principal investigators of the Usselo soil and know it best based on both field and laboratory
research
This appeal to authority has no basis in science and can immediately be rejected.
clearly demonstrate that the soil is just that: a zone of pedogenic weathering including
accumulation of organic matter such as charcoal over time (van Geel et al., 1989; Hoek 1997;
Kaiser et al., 2009; van Hoesel et al., 2012, 2013, 2014).
No arguments, other than spurious ones based on authority, are provided here to justify this assumption. Let's be clear. There is no disagreement about how soils form or about the nature of the soils above or below the charcoal layer. Instead, the debate concerns the charcoal layer itself. The dating of this layer is not inconsistent with the YDB.
Kaiser et al. (2009, figure 8) illustrate
the radiocarbon dating of 63 samples from the Usselo soil. The full range of dates spans almost
2000 years, but the bulk of the dates are from the Allerød interval (pre-YDB); far fewer date to
the YDB or YD/GS-1. The dating is consistent with the field interpretation of prolonged
pedogenesis before and during the YD/GS-1.
The focus of YDIH proponents is on the charcoal layer, not the soil. As argued by Sweatman (2021), the radiocarbon dating of the charcoal layer is not inconsistent with a moment in time and therefore the YDIH. We have already seen that Holliday et al. do not understand how to analyse uncertainty in radiocarbon (or any scientific) measurements. The timespan of 2000 years is not problematic because most of these dates are based on single radiocarbon measurements. For example, Gill et al. (2009) find dates scattered over 8000 years, suggesting individual radiocarbon dates can have an error of ~ 4000 years (2 sigma), which is far more than the intrinsic radiocarbon measurement error.
These data also directly contradict claims that half
of the charcoal dates are at or near the YDB (Sweatman 2021, p 15) or that most of the charcoal
is in or on top of the upper soil zone and marks the YDB (Kinzie et al., 2014, p 477; Kennett et
al., 2015a, p 4347, 4350; Wolbach et al., 2018b, figure A6) (Table 6). Based on the dating,
including OSL ages for the eolian deposits above and below the Usselo soil, and the evidence for
pedogenesis, van Hoesel et al. (2012, p 7651), van der Hammen and van Geel (2008, p. 360), and
Kaiser et al. (2009) all reject the claim that the Usselo soil is a rapidly deposited YDB “event”
layer.
Nevertheless, the full set of dates on the charcoal layer is not inconsistent with a moment in time and the YDIH. And we have seen that Holliday et al. do not understand how to properly assess uncertainty in experimental measurements.
YDIH proponents nevertheless persist in using the Usselo soil as a YDB marker
(Sweatman, 2021, p 12, 14, 15; Powell 2022, p 5). Sweatman (2021, p 12), for example, suggests
that the soil represents “YDB sediment” based on no data and misunderstanding what a soil
represents. He argues that the “conclusion that these sites are not synchronous should be
considered inconclusive” even though the published field data and geochronology establish the
Usselo as one of the best dated and stratigraphically consistent post-LGM terminal Pleistocene
marker soils in Europe.
Single radiocarbon measurements at most Usselo Horizon sites is not consistent with the statement "the Usselo as one of the best dated...". In any case, YDIH proponents are focussed on the charcoal layer within the soil, not the soils themselves. And this charcoal layer is consistent with a synchronous event. Holliday et al. have not presented any evidence to contradict this position.
In his review of the YDIH, Sweatman (2021) includes a number of other critiques of the
radiocarbon dating of the Usselo soil. The large body of consistent data generated by multiple
investigators, which inconveniently contradict the YDIH, is offhandedly dismissed by an
unsubstantiated remark that critics misunderstand “the nature of variance in the radiocarbon
dating of sediments” (Sweatman, 2021 p 14).
This remark is demonstrated to be true above in relation to Abu Hureyra, at least.
The dating of the Usselo/Finow soil is based on
scores of dates on individual fragments of charcoal, not sediments, from multiple sites.
Sweatman (2021, p. 12, 15, 16, 17, 20) also criticizes reliance on single radiocarbon dates.
Multiple samples for numerical age control are ideal, but in the early decades of numerical
dating, not common.
This clearly indicates that early conclusions formed about the Usselo Horizon were premature, as they were based on a misunderstanding of the nature of variance in radiocarbon dating.
A single date is not by definition in error.
As show above in relation to Gill et al (2009) and Meltzer et al. (2014), there is ample evidence that the intrinsic radiocarbon measurement error is not always (often?) a reliable estimate for the true uncertainty in the age of a sample.
There are many examples of reliable single dates.
No evidence is provided for this. The evidence above shows many are not.
Indeed, Sweatman (2021, p2-3, 16) embraces the dating of the “black mat”
by Haynes (2008) (Section 6) even though most of that dating is based on one or a few dates, and
he expresses no concern over the issue of variance in that dating.
The issue is whether the Younger Dryas boundary at any site is clearly inconsistent with the date of the YD impact. So far, no YDB site has been found to be obviously inconsistent with the date of the YD impact.
The dating of the Usselo soil also raises an important point. Sweatman (2021, p 12)
argues “the uncertainty in the age of YDB sediments is rarely captured by a single radiocarbon
measurement at a specific site. Indeed, it is standard practice to take in the region of 10
measurements at any site to create proper age-depth models so that the true age uncertainty in a
boundary layer can be reliably reported. Reliance on single measurements from any site is
unwise, as we can expect such an approach to give the false impression of asynchronous local
events for a synchronous widespread event across all sites.” This passage is rife with misleading
inferences and a fundamental lack of understanding of the Usselo soil.
Again, the above is misleading because the debate concerns the charcoal layer, not the soil. Moreover, we have already seen that Holliday et al. do not understand basic concepts in data analysis, especially relating to uncertainty, which are basic requirements for competent scientists.
More broadly, this
sweeping statement is a rather remarkable claim in support of the YDIH given that almost no site
presented in the YDIH literature meets the requirement for 10 14C measurements for each YDB
or black mat section.
This is also misleading because it supposes 10 measurements is a "requirement". It is not. It is simply a preference.
This argument could be used to consider the notion of the YDIH equally
inconclusive. He further argues (p 15) that in dating the Usselo soil “only single measurements
were made at each site. Proper age-depth models that intersect the boundary were not generated
for any of them, leaving open the possibility that the sites are synchronous and the dispersion in
dates they found was due to natural processes.” This is another unsupported assertion with no
factual basis.
This is false. It is clear that single radiocarbon measurements are unreliable and instead proper age-depth models with proper uncertainty analysis are preferred. Holliday et al. have already demonstrated that they do not understand how to assess uncertainty in radiocarbon measurements or age-depth models.
Again, similar to Kennett et al. (2015a), an assumption is made that because it
could represent the YDB, it therefore must be the YDB.
This is false. A statistical argument, see above, is used to infer that the YDB almost certainly represents an instant in time.
More to the point, an impact,
representing a moment in time (similar to a volcanic eruption such as the Laacher See; Baales et
al., 2002, and Section 5.8) produces radiometric dates that vary around a mean. The Usselo soil,
in contrast, produced many non-overlapping radiocarbon dates spanning 1400 14C years because
it is a soil. There is no possibility that the sites are synchronous.
This is false. The debate concerns the charcoal layer, not the soil layers above or below the charcoal layer. We have already seen that Holliday et al. do not understand how to properly analyse uncertainty in radiocarbon measurements or age-depth models.
The issue is whether any of these non-overlapping dates spanning 1400 14C years is inconsistent with the date of the YD impact. Given most of these measurements correspond to single radiocarbon measurements their true uncertainty is unknown. See Gill et al. (2009) above which demonstrates that the true uncertainty (i.e. scatter) in radiocarbon measurements often far exceeds that of the radiocarbon measurement itself, which is why multiple measurements leading to proper age-depth models are preferred. Thus, it cannot be concluded with any confidence that any of the single dates for the charcoal layer at these sites is clearly inconsistent with the YD impact as Holliday et al. claim.
The Usselo soil does not represent a moment in time.
As we have just argued, it is not possible to claim this with any confidence with the present radiocarbon data set for the charcoal layer in the soil.
Sweatman (2021, p 15) also asserts “the precise boundary layer at each site
corresponding to the depth of geochemical markers, rather than charcoal which is not diagnostic
for the impact event, was not determined for any site studied, and therefore it is not possible to
know if any charcoal samples were taken directly from the Younger Dryas boundary.” That is
another example of circular reasoning, however. Sweatman claims the geochemical signatures
define the YDB, while at the same time they are interpreted as impact markers largely because
they occur at the YDB and are synchronous with the YD/GS-1 onset, which is claimed
synchronous with the megafauna extinctions.
This paragraph is very confused. It is true that the Younger Dryas boundary is defined by impact markers with dates consistent with the Younger Dryas impact. There is no circular reasoning here. If impact markers similar to others in the YDB are found with a date in the region of the YD impact, then according to the statistical argument above they almost certainly represent the YD boundary.
While we agree that charcoal is not an impact
marker (see Section 9.3), YDIH proponents repeatedly claim that it is produced in great quantity
by the YDB impact. For example, “Wildfire … at the Younger Dryas boundary” is the title of
Kennett et al. (2008a) and the titles of both Wolbach et al. (2018a, b) begin “Extraordinary
biomass-burning … triggered by Younger Dryas cosmic impact.”
There is no inconsistency here.
Sweatman (2021, p 12) earlier
cites these and other papers purporting peaks in charcoal at the YDB. Therefore, charcoal is
certainly an appropriate material to use for dating the soil and the dates clearly show that it is
both older than and younger than the YDB.
We agree that charcoal is an appropriate material for dating the charcoal layer. There is an assumption that this charcoal layer corresponds to the YDB because the date of the charcoal is not inconsistent with the YDB and the charcoal layer contains impact proxies at a few sites. Moreover, charcoal is an expected product of the YD impact. Nevertheless, the YDB is technically defined by impact proxies, not charcoal. This is really quite straightforward.
Confusingly, some YDIH proponents explicitly
claim the black mat lies directly above the YDB, while others claim it is the YDB (see Section
6).
In some places, like the Usselo charcoal horizon, the YD black mat is defined by a charcoal layer. However, in many cases it is not. The YD boundary is defined by the impact proxies, and this is often, but not exclusively, found at or near the base of the YD black mat. This is very simple. The only confusion here lies in the minds of Holliday et al.
Sweatman (2021, p. 16) on the other hand apparently alludes to both, claiming the base of
the black mat (and the Usselo Soil) is the YDB and the remaining majority formed over the
YD/GS-1. The Usselo Soil cannot (and does not) represent both the YDB impact and the
YD/GS-1.
But the Usselo charcoal layer could represent the YDB, i.e. dating of this charcoal layer is not clearly inconsistent with the date of the YD impact.
The obvious conclusion based on all geologic and pedologic data, including the dating of
deposits above and below the charcoal-rich soil horizon is that the Usselo soil reflects fires
(along with pedogenesis) spanning at least ~1400 14C years, largely in the Allerød but continuing
into the YD/GS-1 (i.e., across the YDB).
This statement cannot be made with confidence. We have already seen that Holliday et al. do not understand how to properly analyse the uncertainty in radiocarbon measurements or related age-depth models. As already argued, the radiocarbon data for the Usselo charcoal layer in the soil is not obviously inconsistent with the YD impact date.
A bigger issue among YDIH proponents is a
fundamental misunderstanding of the nature of pedogenesis, which is a time transgressive
process on stable or quasi-stable landscapes. ENDNOTE 8 “[T]here is no need to invoke an
extraterrestrial cause to explain the charcoal in the fossilized soils” van der Hammen and van
Geel (2008, p 359).
A major problem with Holliday et al. is they do not properly understand the uncertainty in radiocarbon measurements or associated age-depth models.
5.7. Improved Dating of Clovis Sites and Clovis Archaeology
This section concerns the dating of Clovis artefacts and is therefore of no concern here. We are focused on the impact itself, not its secondary effects.
5.8. Radiocarbon simulations of the YDB
Inter-site variability in radiocarbon dates on purported impact proxies has remained
problematic for the YDIH, suggesting that those layers were deposited asynchronously (Holliday
et al., 2014; Meltzer et al., 2014).
On the other hand, the statistical argument present above demonstrates the YDB sites are almost certainly synchronous, or highly correlated over a short time-span.
However, YDIH proponents continue to argue that the layers
were deposited synchronously and have generally supported this argument by citing Kennett et
al. (2015a; but see Boslough et al., 2015; Holliday, 2015). Using a OxCal, with a Bayesian agemodel
implementation, Kennett et al. (2015a) estimate upper and lower chronological boundaries
for a hypothetical temporal phase containing supposed impact-indicators from 23 sites in
addition to seven paleoclimatic proxy markers of the YD/GS-1 onset. Kennett et al. (2015a, p
E4352) estimate a temporal difference between the start and end of the proxy phase somewhere
within 0–130 years (95% probability interval), concluding that synchronous deposition of all 23
layers is plausible since the range of possible years includes zero. Unfortunately, Kennett et al.
(2015a) neither plot nor describe the mean, median, or mode of this interval, so it is difficult to
assess which temporal distances are most probable–while zero years may be plausible, this
interval is also consistent with distances exceeding a century.
Agreed, 0 years is plausible, and the statistical argument above demonstrates the YDB sites are almost certainly synchronous, or at least highly correlated over a short time-span.
Further, the assumptions and
decisions involved in the creation of this phase model render its inferences problematic.
Given the many parameters and assumptions required to model the 23 site chronologies,
it is unclear to what degree this 0–130-year estimate is contingent on modeling decisions. These
decisions include the placement of stratigraphic breaks, the inclusion/exclusion of horizontally
disparate samples, chronometric hygiene protocols, as well as the distributions, types, and prior
parameter values of site-specific age models and of sample-specific outlier models. Additionally,
the probability distributions of the start and end of the proxy phase result from the choice to
include the 23 modeled YDB ages in a single temporal phase, which itself has multiple possible
distributions and prior parameter values that must be specified by the user. In combination, these
decisions compound potential problems stemming from assumptions at different modeling
levels. Despite these issues, subsequent works that favor the YDIH cite Kennett et al.’s (2015a)
model as confirmation that the proxy layers represent one event (e.g., Israde-Alcántara et al.,
2012; Moore et al., 2017), while it remains, at best, only plausibly consistent with synchroneity.
Agreed, Kennett et al. (2015) only shows plausibility for synchroneity. The argument that these impact layers are almost certainly synchronous is statistical and provided above. Essentially, the probability of the null hypothesis model (they are unrelated impacts and therefore not synchronous) is shown to be vanishingly-small.
In addition to the previously discussed evidence contradicting the assertion that the
purported proxy layers were deposited by a single event (Section 5.0), simulations published by
Jorgeson et al. (2020) illustrate that the impact proxy radiocarbon dates used by Kennett et al.
(2015a) are far more dispersed than expected for a synchronous event.
This is false because it is simply not possible to show this. Instead, Jorgeson et al. (2020) show that their models are inadequate. This was clearly explained by Sweatman (2022) in response to Jorgeson et al. (2021).
Jorgeson et al.’s (2020)
simulations iteratively sampled radiocarbon ages from a hypothetical synchronous event,
accounting for uncertainty in the radiocarbon calibration curve, laboratory measurement error,
and old wood effects. The authors compare age dispersion in the simulated samples to the age
dispersion of the observed YDB radiocarbon sample dataset and to the age dispersion in
observed radiocarbon samples of a known synchronous event, the Laacher See volcanic eruption
in Germany. The YDB radiocarbon dataset shows far more age dispersion than the simulations,
while the Laacher See volcanic eruption radiocarbon dataset displays age dispersion similar to
the simulations.
Sweatman (2021) showed that Jorgeson et al.'s (2020) arguments are flawed. Jorgeson et al. (2021) responded, but Sweatman (2022) showed their response was also flawed. Like Holliday et al. and Meltzer et al., Jorgeson et al. (2020,2022) do not understand the limitations of radiocarbon age modelling. Indeed, the key problem for Jorgeson et al. (2020,2022) is stated by Sweatman (2022) to be "an unsupportable confidence in modelling the radiocarbon data from the Younger Dryas boundary". In other words, a model is only as good as its underlying assumptions. Thus, the discrepancy between their model and YDB radiocarbon data cannot be attributed solely to the YDB radiocarbon data with any confidence. In fact, given the statistical argument above about the YDB synchroneity, it is clear the problem instead lies with the assumptions that define their model.
In fact, it is clear that their modeling approach is fundamentally flawed from the outset. As shown in the previous section, since the intrinsic radiocarbon measurement error is an unreliable estimate for the true sample age uncertainty, individual radiocarbon measurements should be viewed with caution. Instead, multiple radiocarbon measurements from the same section should be used to obtain a more accurate estimate of the true sample age uncertainty. However, Jorgeson et al. (2020) take the opposite approach, favoring individual measurements from the YDB at different sites. This is inherently unreliable.
YDIH proponents - mainly Sweatman (2021, 2022; see also Powell, 2020) - raise four
objections to Jorgeson et al.’s simulations, but each lacks merit.
We shall see.
The objections involve old wood
effects (see also Section 12.4) for the Arlington Canyon radiocarbon dates, the effects of
catastrophic geomorphic processes on the integrity of radiocarbon samples, inadequate
chronological modeling, and a failure to address the supposed geochemical evidence for the
hypothesis.
A fifth issue is added above, which is that Jorgeson et al.’s approach, which is based on single radiocarbon measurements from several sites, is inherently unreliable because single measurements cannot in principle provide a good estimate of the true variance in sample ages at each site. Instead, proper age-depth modeling at each site should be preferred.
Concerning Arlington Canyon, Sweatman (2021, 2022) argues that the old-wood offsets
used in Jorgeson et al.’s (2020) simulations are not sufficient to account for pine species at the
site, which can live up to 1000 years. As such, Sweatman argues that a synchronous event should
produce radiocarbon samples more dispersed than those simulated by Jorgeson et al., as larger
old-wood offsets would generate more temporal variability. As reported in Kennett et al.
(2008a), 7 of 11 radiocarbon samples from Arlington Canyon are wood charcoal. These wood
samples consistently predate Kennett et al.’s (2015a) modeled YDB age by only ~0–450 14C yrs,
an offset that is well accommodated by the simulated old-wood offsets (Jorgeson et al., 2022). As such,
the high dispersion in YDB radiocarbon ages is not explainable in terms of the Arlington Canyon
samples alone.
This is false. The largest decay constant used by Jorgeson et al. in their exponential old wood model is 100 years, which is clearly inadequate for modelling a small sample with such high ages of 450 years.
The remaining four (of 11) Arlington Canyon radiocarbon samples comprise an
unspecified charcoal, a “carbon spher[ule]”, a “carbon elongate”, and a “glassy carbon” sample.
Supporters of the YDIH claim that carbon spherules, carbon elongates, and glassy carbon are
remnants of burned tree sap (Israde-Alcántara et al., 2012, p E745; LeCompte et al., 2018, p 169;
Wolbach et al., 2018b, p S27, 2020, p 99; but see Scott et al., 2010, 2017 and Sections 9.3 and
12.4), and that these Arlington Canyon samples were produced by biomass burning in the wake
of the impact event (Kennett et al., 2008a).
This is false. YDIH proponents claim that charcoal samples in the YDB that contain impact proxies were produced by the YD impact. The YDIH makes no claims about samples that do not, or are not known to, contain impact proxies. Such samples could have been produced by non-impact wildfires. Since the samples used for dating Arlington Canyon are not known to contain impact proxies, Holliday et al.'s argument can be disregarded.
The spherule dates to 11,440±90 14C yr BP,
corresponding to a calibrated 95% interval of 13,458–13,163 yr BP (UCIAMS-36961; Kennett et
al., 2008a), well prior to the proposed impact.
Since the spherule is not known to contain impact proxies, no claim is made about its provenance.
The carbon elongates and glassy carbon samples
also have similar dates of 11,110±35 and 11,185±30 14C yr BP (UCIAMS-36962 and UCIAMS-
36960; Kennett et al., 2008a), corresponding to calibrated 95% intervals of 13,100–12,924 and
13,162–13,085 yr BP, respectively. Like the wood samples, these samples are consistently older
than the proposed YDB age. Unlike the wood samples, burned tree sap is not subject to old wood
effects.
Agreed, but this does not refute the YDIH or the dating of Arlington Canyon because these samples were not examined for impact proxies. They are probably from earlier wildfires.
Kennett et al. (2015a) rely on treating these three specimen types as wood charcoal with
potentially large age offsets, incorrectly allowing for a younger are for claimed impact indicators
for Arlington Canyon consistent with the YDIH.
It is true that glassy carbon, carbon elongates and carbon spherules with large age offsets could have been excluded from the dataset. However, OxCal was able to include them anyway using an old wood model. Since it is not known whether they include any impact proxies, this is fine. This is because if they were discarded it would likely make no significant difference to the age-depth model. Indeed, Holliday et al. have provided no evidence that excluding them would make any significant difference.
If the Arlington Canyon layer dates to Kennett
et al.’s (2015a) proposed YDB age, and if the carbon spherules/elongates as well as glassy
carbon are wildfire products (as YDIH proponents claim), then those from earlier wildfires were
mixed into that layer.
Agreed. This is very likely considering that Arlington Canyon is a ravine with frequent mixing of sediments due to turbulent high water flow events.
In that case the carbon spherule/elongate and glassy carbon concentration
profiles that have been published in support of the YDIH cannot be used the test the YDIH. This
is because those specimens are not impact markers, thus in order to correlate any particular
carbon spherule/elongate or glassy carbon to a possible YDB impact event (and to the same
specific wildfire), it is then necessary to date that particle to the YDB. To test the YDIH, it is
then necessary to radiocarbon date each and every carbon spherule/elongate and glassy carbon
counted through the sediment profile to construct a meaningful concentration profile based on
age-correlated abundances. This clearly has not been performed.
This is false. Holliday et al. are clearly confused. None of the samples used for dating were also examined for impact proxies. These are different measurements. Some of the samples used for dating may well have been produced by earlier wildfires and could have been excluded from the dataset, but the dating model used by OxCal was able to incorporate them using an old wood model. This is fine, and Holliday et al. have provided no evidence that their exclusion would make any significant difference. The impact proxy abundances in the YDB layer are also fine. There is no issue here.
Given the current evidence,
the parsimonious interpretation of the Arlington Canyon “proxy layer” is that it predates the
hypothesized event, consistent with Jorgeson et al.’s (2020) simulations.
Actually, given the impact proxies present in the Arlington Canyon (AC) YDB layer, based on the above statistical argument the most parsimonious explanation is that it is of YD impact age, and Jorgeson et al.'s (2020) model is inadequate.
If the samples from Arlington Canyon indeed attest to an interval of increased wildfire,
there is an alternative hypothesis. The calibrated ages of the tree samples are consistent with
sharp increase in Greenland δ18O values between the cooler GI-1b (the IACP, see Sections 3.3
and 9.2) and the warmer GI-1a interval (Rasmussen et al., 2014). At that time, Santa Barbara
Basin ocean-surface conditions, which apparently vary synchronously with Greenland climate
(Hendy et al., 2002), would have abruptly changed from cool to warm (and from cooler and drier
to warmer and wetter conditions on adjacent land areas). The samples could well be the product
of wildfire favored by that climate change and would not require an exotic explanation.
Agreed, but since these samples used for dating were not examined for impact proxies, this is irrelevant.
Sweatman (2022, p 3) also contends that inconsistent radiocarbon samples from Murray
Springs and Big Eddy should have been discarded from the simulations. Jorgeson et al. (2020)
consider the exclusion of questionable dates from Murray Springs and Big Eddy in their
supplemental simulations - exclusion of these dates does not affect the main conclusions drawn
from the simulations.
Holliday et al. fail to note Sweatman's (2022) response that "... if any of their modelled scenarios do not simultaneously account accurately for the old wood effect at Arlington Canyon, or for the SOM effect at Murray Springs, or for the anomalous data point at Big Eddy, they will likely obtain a negative result. In fact, they do not consider any scenario that allows simultaneously for all three effects." Thus, it is likely their model is inadequate.
In the second objection, YDIH proponents blame catastrophic geomorphic processes for
high variability in radiocarbon ages between layers containing purported impact proxies.
Sweatman (2021) initially argues that the radiocarbon record is consistent with synchronous
deposition. Yet, one year later, Sweatman (2022, p 1), in response to Jorgeson et al.’s
simulations, argues that high dispersion in radiocarbon dates should be expected, given the
dramatic effects of the proposed impact: “The asteroid impact... ...would alter the environment
catastrophically through a hierarchy of interlinked events and processes, many of which could
lead to an increase in the distribution of radiocarbon dates relating to the event. Ancient forests
might be felled, tsunamis, earthquakes and landslides might mix and redeposit soils, and old
sources of carbon might be redistributed. Even if some of these catastrophic processes might be
modelled, there will always remain some doubt about the suitability and completeness of such
models.”
Jorgeson et al. (2020) considered such a catastrophic event, the Laacher See volcanic
eruption. The eruption felled trees, created a temporary lake through damming of the Rhine
Valley, produced a 50-m thick tephra near the eruption center, and generated 1-m thick pumice
deposits up to 120 km from the volcano (Baales et al., 2002; Bogaard and Schmincke, 1985).
These processes left unambiguous features visible on Central Europe’s modern landscape. Even
with these dramatic eruption effects, the Laacher See tephra contains radiocarbon samples
consistent with simulations of a synchronous event (Jorgeson et al., 2020). By contrast, evidence
for the catastrophic geomorphic processes suggested by Sweatman are lacking for the proposed
YDB impact (Sections 3.3, 13.7). Impact proponents, in essence, argue that the impact produced
global catastrophic effects far exceeding those of the Laacher See eruption, while paradoxically
leaving no evidence for changes to the landscape. To our knowledge, YDIH proponents have not
offered evidence for impact related tsunamis or earthquakes.
This complaint is equivalent to suggesting that Darwin's evolution through natural selection is not valid because we have yet to find fossils for all the missing link species. It is the argument of the "gaps". Remember, absence of evidence is not evidence of absence.
In fact, the evidence is overwhelming that a large impact event near the onset of the YD period occurred. The above statistical argument shows it almost certainly occurred as a single event or a multitude of highly correlated impact events. This would have had catastrophic effects on the environment. The resulting geomorphic effects will be found if they are looked for. However, the YDIH is in its infancy as a research paradigm, so there is plenty of time for more research into this. Also, don't forget that over 100 m of sea level rise has occurred since then, thus obscuring evidence for at least some of these catastrophic effects.
In the third objection, Sweatman (2022) questions the very idea of modeling the YDB
radiocarbon dataset. Since every physical process relating to chronology cannot be known with
certainty, he argues that any unexplained variation in radiocarbon dates is unproblematic. For
example, regarding the modeling of old wood effects with an exponential distribution, Sweatman
(2022, p 3) states that “the exact ‘old wood’ model for AC [Arlington Canyon] is unknown, nor
is it known whether any exponential distribution with any value of λ is adequate” and Jorgeson et
al. (2020) “did not explore all possible forms of ‘old wood’ model. They only discuss simple
exponential forms.”
An “exact” model cannot be known for most physical processes as models are, by
definition, reductionist representations of the physical world. There are infinite possible old
wood models that could be defined; although there are theoretical reasons to expect the
distribution of old wood effects to be approximately exponential (Nicholls and Jones, 2001).
Any "theoretical reason" to expect an exponential distribution of old wood effects will be based on several assumptions. None are provided here, so the validity of the exponential old wood model used by Jorgeson et al. (2020) has not been justified. Moreover, as explained above, an old wood model with a decay rate of 100 years (the largest considered by Jorgeson et al. (2020)) is unlikely to be adequate for modelling old wood effects up to 450 years in a small sample. This is because the cumulative probability distribution for ages older than 450 years is exp(-4.5) = 0.011, yet the sample size is ~ 10 data points. A 200 year decay constant would be more appropriate in this case. It is unclear why Jorgeson et al. limited their exponential decay model to a decay constant of only 100 years when it is clearly inadequate.
Consequently, the exponential distribution is a standard model for old wood effects as
implemented in OxCal (Bronk Ramsey, 2009), the application used to estimate many
chronologies in archaeology, paleontology, and paleoclimatology. Impact proponents themselves
used OxCal to model the age of the hypothesized event, and the vast majority of their
radiocarbon samples were modeled with exponential old wood offsets (Kennett et al., 2015a).
While all models are imperfect and incomplete, the strength of the evidence produced by
a model with well-justified assumptions can be probative.
Any disagreement between a model and real data that is assumed to be synchronous either reveals the assumption of synchroneity is invalid or the model is invalid. Without further evidence, neither can be known with any confidence and therefore Jorgeson et al.'s conclusions are unsupported. However, the statistical argument provided above shows that YDB sites are almost certainly synchronous.
The key phrase above is "well-justified assumptions". In fact, the assumptions underlying Jorgeson et al.'s model are not well-justified. For example, the old wood decay constant is clearly too short. There are other problems too that are discussed below.
Jorgeson et al.’s (2020) simulations
are not just broadly inconsistent with a synchronous YDB, they demonstrate that the likelihood
of a synchronous event producing the dispersion seen in the YDB dataset is astronomically low.
This is false. They do not demonstrate asynchroneity since an alternative explanation for their results is that their model is inadequate, as discussed above. Indeed, given the presence of impact proxies in the boundary layer, it is almost certainly synchronous - see the statistical argument given earlier.
The simulations account for many sources of variability in radiocarbon dating; while there may
be other sources that are not included in the simulation, they would likely have only marginal
effects on the results.
This statement is not and cannot be justified. For example, it is notable that Holliday et al. do not address the following critique of Sweatman (2022);
"However, they do not consider a combined scenario where the Big Eddy and Murray Springs SOM data points and the AC [Arlington Canyon] data are excluded. Given that they acknowledge problems with this data through modelling the various scenarios described above, it is curious as to why they did not consider a scenario that simultaneously eliminates all the problematic data points identified. If they had, they might have obtained reasonable agreement with the YDB data. Therefore, their conclusion that the YDB likely does not represent a synchronous event is not supported."
Sweatman (2022, p 2) raises a final objection against Jorgeson et al. (2020) on the
grounds that their simulation “does not explain the physical evidence for the YD impact event at
numerous YDB sites found, and confirmed, on multiple continents as reported in dozens of
papers.” Regardless of the numerous problems with the purported physical evidence and dating
enumerated throughout this and other papers, the objective of Jorgeson et al. was not to
“explain” the claimed evidence for a YDB impact, only to illustrate that the YDB radiocarbon
record is statistically inconsistent with a synchronous event. Neither Sweatman nor other YDIH
proponents have demonstrated otherwise.
The statistical argument provided above demonstrates this.
Summary
Holiday et al. have amply demonstrated that they do not understand basic concepts in uncertainty analysis. They also demonstrate they do not understand the difference between a model prediction and circular reasoning. These concepts are required for competence in science. They should revise their understanding of probability and statistics and hypothesis testing.
In particular, the works of Gill et al. (2009) and Meltzer et al. (2014) fail to record uncertainty estimates for the coefficients of their linearly regressed age-depth models. These papers are therefore not works of science and no conclusions can be drawn from them. They should be withdrawn. Holliday et al. consider these coefficients an unnecessary technical detail. On the contrary, they are crucial and this is demonstrated here specifically for Abu Hureyra above. Thus Holliday et al.'s support for Meltzer et al. (2014) can be considered pseudoscience as it claims to be scientific, but is not.
The modelling of Jorgeson et al. (2020) is likewise incompetent as it fails to recognise the limitations of their own models which were clearly pointed out by Sweatman (2021,2022). Holliday et al. also fail to recognise these limitations.
Comments
Post a Comment