Benchmarking astrochronologic methodsIt is now over thirty years since it was first demonstrated that orbital cycles are preserved in the climatic records. This forum is the place to discuss them.
7 posts • Page 1 of 1
Dear astrochronologists and cyclostratigraphers:
EARTHTIME geochronologists are dedicated to standardization of their analytical methods in order to minimize inter-laboratory bias. They are achieving this by sharing standards and tracers by which they can tune their laboratory techniques in order to obtain consistent results.
At present, no such analogous effort being undertaken by the astrochronology community. While the challenges are much different, the end result would be the same: assurance that the techniques used by researchers can detect, process, and assess astronomical signal accurately in stratigraphic data.
Here I invite discussion of what might be used to “benchmark” astrochronologic techniques. This might involve, for example, synthetic models with perfectly known signal and noise combinations against which researchers can test their techniques. A related benchmark could address practices of astronomical tuning.
In future postings, I will propose some specific benchmarks; meanwhile, I hope that others will contribute their views and ideas.
Thank you for initiating this forum Linda. There are many challenges that face the astrochronology community as we endeavor to refine orbital time scales and extend coverage further into deep-time. I am happy to find a place to discuss these issues in a more informal manner.
I would like to take a slight diversion from the “benchmarks” topic to address what I believe is a more fundamental underlying issue (but I will touch on the issue of benchmarks at the end of the posting). One of the most pressing concerns in the field of astrochronology is the development of a standard methodology for orbital time scale construction. Even a cursory review of the literature will show a lack of such standards. One extreme consequence of this problem is the existence of conflicting orbital interpretations for some stratigraphic units. If the potential of astrochronology is to be fully realized, we must focus our attention on the development of a robust methodology that removes subjectivity from the science.
What ought such a “standard methodology” look like? I will make a few suggestions to start, and I invite others to contribute their thoughts.
(1) The methodology should involve the acquisition of high-resolution quantitative data series characterized by high signal/noise (I am referring to analytical noise here). Much important cyclostratigraphic work has been conducted using semi-quantitative interpretative schemes (e.g., facies rank), but such schemes can also impart a level of subjectivity.
(2) In generating an orbital time scale, it is necessary to demonstrate that a clock is indeed preserved in the strata, while a secondary (but related) issue is the calibration of this chronometer. The fact that multiple incompatible orbital interpretations exist for some stratigraphic units exemplifies the fact that these are distinct but related concerns.
If sufficient independent time-control is available (e.g., radio-isotopic data), both of these issues can (potentially) be addressed in a straightforward manner. But this is not the case in most deep-time cyclostratigraphic studies. A wide range of methods have been developed to prospect for orbital signals in deep-time strata that lack sufficient independent time control (e.g., spectral frequency ratio methods, amplitude modulation assessment, visual comparison of the stratigraphic data series to the theoretical orbital-insolation times series, etc.). While these methods can be successful, they are not consistently applied across studies, and each method involves a level of subjectivity in its application. A standard approach needs to be developed.
Relative to this issue, I have recently proposed a new technique (termed average spectral misfit; Meyers and Sageman, 2007) to provide an objective quantitative method for orbital time scale construction. The method is applied to untuned data, and explicitly evaluates the null hypothesis of “no orbital signal”. It also provides an optimal temporal calibration of the observed stratigraphic data to an orbital target.
(3) The methodology should apply time-frequency analysis to the untuned data, to assess stratigraphic changes in the bedding periods. In general, the cyclostratigraphic record is quite messy, riddled with gaps and sedimentation rate changes, and other sources of “noise”. These issues can be addressed using time-frequency methods such as wavelet analysis and evolutive harmonic analysis.
(4) Ultimately, orbital timescales developed at one site should be confirmed by complimentary analyses at other locations.
By no means is this intended to be an exhaustive list, just a few thoughts to start a discussion. In terms of (2) and (3), the development of benchmarks to evaluate different techniques will be important. In fact, some of this work has already been completed (so called “spectral shootouts”; Kay and Marple, 1981; Thomson, 1982). In terms of creating benchmark models specifically for astrochronology, a good place to start would be a series of models containing orbital-insolation signals, distorted by sedimentation rate changes, embedded in autoregressive (“red”) noise. Let’s see how far we can push our methods before they fail.
The ultimate goal of all of this work should be a robust cyclostratigraphic methodology that allows us to quantify how well we know that our astrochronology is reliable, rather than a qualitative “it looks like Milankovitch”.
References noted above:
Kay and Marple, 1981, Proceedings of the IEEE, v. 69, p. 1380-1419.
Meyers and Sageman, 2007, American Journal of Science, v. 307, p. 773-792.
Thomson, 1982, Proceedings of the IEEE, v. 70, p. 1055-1096.
I am (of course) very interested in your new Average Spectral Misfit method. I also enjoyed your paper analyzing our old Latemar platform study under this new light.
ASM finally provides a test for how much a Milankovitch-like spectrum is actually comparable to a Milankovian template - something we all were missing severely.
However, to date, you applied the method under the assumption of constant sed. rates (i.e., on untuned series). I'd say this is a huge problem when we are dealing with long or non-pelagic series.
Could you estimate how much the ASM method is sensitive to changes in sedimentation rates?
All the best
Ref: Meyers, 2008, Geology, 36:319-322
Thank you for your comment Nereo, you raise an excellent point.
The ASM method identifies the most plausible AVERAGE sedimentation rate for a stratigraphic data series (for example, see the model in Figure 4 of Meyers and Sageman, 2007). If sedimentation rate instability is large enough, the method will ultimately fail to reject the null hypothesis with a high degree of confidence.
As you note, the results published in Meyers (2008) are based on an analysis of the entire CDL lithofacies rank series. The results indicate an optimal sedimentation rate of 49.5 cm/kyr (null hypothesis significance level = 0.293%). In other words, of the 100,000 Monte Carlo spectra simulations, only 293 yielded an ASM smaller than the CDL data. Figure DR3A (GSA Data Repository Item 2008074) illustrates the remarkable fit between the orbital target and the CDL spectrum when calibrated with a sedimentation rate of 49.5 cm/kyr.
Based on these results, we can conclude that the null hypothesis (no orbital signal in the CDL series) is highly unlikely. And when considering the CDL series in its entirety, a calibration of 49.5 cm/kyr is optimal. But one may ask, are there subsections of the CDL series that yield a different sedimentation rate with a lower null hypothesis significance level? And more generally, can the ASM method be adapted to allow for changes in sedimentation rate through a stratigraphic interval?
I am currently developing a new “time-frequency” implementation of the ASM method to specifically address this issue. I plan to debut the technique at the GSA Annual meeting next fall. So please stay tuned…
time-frequency approach is exactly what is needed. Looking forward for your results.
By the way, it is not a matter of "the Latemar is Milankovitch" or not. I suppose that the Latemar is not Milankovitch is quite an accepted "fact" at this point. Real problem is, if not Milankovitch, what else? No consistent answers yet on this point, at least in my view. And, if it is not Milankovitch, is it something after all? Your answer is yes, there is some periodical signal there. That is why I like your results.
All the best
Dunno Nereo, is it that the Latemar isn't Milankovitch or is it rather that there is still disagreement on which bundles represent which orbital signal? The way I understand it, nobody really has fully rejected the notion that there is a Milankovitch signal within the Latemar stratigraphy (though some have come close), but instead how the whole stack is tuned.
You bring up a good point though and one that I certainly don't have an answer to- if fundamental cycles and even megacycles are sub-milankovitch, then what is the driver? I'd like to know.
Ciao to all, and sorry if I didn't post quickly after Rob's comment, but here I am now.
Rob, you are right. Of course, while the Milankovitch interpretation of the Latemar elementary cycle is probably abandoned by most, still almost all students (including me) assume that lower frequency cyclicities defined by sedimentary cycle bundles should be Milankovitch at least in part.
If the elementary cycle is sub-Milankovitch (and periodic), it is quite important to understand what is it after all.
There is only one published hypothesis on their origin, namely, that they represent climate cycles triggered by a long-period tidal cycle (Kent, Muttoni and Brack in EPSL). I must admit I don't believe this one: literature dismissed the 1.8 kyrs tidal cycle as a potential climate driver (not to speak of the sea-level effect, which is really negligible), and nothing similar was ever observed in present or near-present times.
But, we must face this: there are no other explanations around.
And, it is important to ask ourselves: why the Latemar is so special that it registered such high-freq cycles? Because, if it turns out it is not special at all, then we must espect those guys (the sub-Milankovithc cycles I mean) popping out everywhere in the geological record.
All the best.
7 posts • Page 1 of 1
Who is online
Users browsing this forum: No registered users and 1 guest