Data reporting madness
Data reporting is extremely important when publishing exposure ages or erosion rates derived from cosmogenic-nuclide measurements, for the following reason. Basically, computing an exposure age requires two things: i) a number of field and laboratory measurements: the location, elevation, etc. of the site, the cosmogenic-nuclide concentrations; and ii) a number of assumed values for production rates, decay constants, scaling factors, etc. The sample-specific field and lab observables don’t change once measured. However, generally accepted values for production rates and scaling factors change all the time, as new production rate studies produce new information about production rate systematics. What this means is that, although the actual sample-specific observations in an exposure-dating paper can be though of as valid for all time, the exposure ages inferred from these observations will not be correct in future. They will be superseded by new information about production rates, and future readers will have to recalculate the ages to take account of this new information. For this reason, it’s critical that any cosmogenic-nuclide publication report ALL the source data needed to compute exposure ages or erosion rates — not just the exposure ages or erosion rates themselves. If this information isn’t there, the publication will be useless to future readers and the project will have been a big waste of time and money.
Anyone who talked to me at any length in the last few years, or had their paper reviewed by me, has probably heard this lecture before. The same lecture appears in the paper describing the CRONUS-Earth online exposure age/erosion rate calculators. One of the goals of these calculators, in fact, was to improve the data-reporting situation by providing a clear list of information that readers would need to have in order to recalculate exposure ages using the online calculators. In the last few months, however, a number of other people have started to harass cosmogenic-nuclide users on this subject in a variety of print outlets. Here I review these various publications and try to make sense of the resulting forest of instructions.
First, in October 2007, myself, Nat Lifton, Joerg Schaefer, and Tibor Dunai sent in a manuscript to EOS that was intended to highlight the whole issue and tell people how to do it right. The last draft of this manuscript can be downloaded here. This manuscript was a bit more confrontational than others to be discussed subsequently. First, it featured a census of the current literature that showed that zero out of ten randomly selected papers that involved Be-10 exposure ages or erosion rates actually contained enough information to recalculate them. Second, it included a discussion of why the term ‘Be-10 years’ should not be used, which appears in a similar form here. Alas, this manuscript became trapped in some sort of Kafka-esque editorial limbo at AGU, and was never published. Here I reproduce my 2009 letter to EOS editor John Geissman withdrawing it:
I regret that I and my co-authors would like to withdraw this manuscript, for the following reasons.
We originally submitted this manuscript to EOS in October, 2007, as a proposal for a ‘Feature’ article. It was sent to three reviewers, who expressed general approbation of the content but indicated that it should rather be published as a ‘Forum’ article. In May, 2008, we submitted a manuscript in the ‘Forum’ format. In September, 2008, we received an additional three reviews that all found it acceptable for publication, but suggested it would be better as a ‘Feature’ article. Thus, EOS staff instructed us to revise it into ‘Feature’ format and resubmit it. We sent in a revised manuscript in January, 2009 which contained the same content that had been approved for publication in six previous reviews, expanded into the ‘Feature’ format. At that point, we anticipated that it would be published. However, your letter appended below indicates that you have subjected this manuscript to a seventh review, which has recommended rather extensive changes prior to publication.
The extraordinary length of this editorial process means that the purpose and content of this paper have largely been overtaken by events. The examples in the paper are obsolete, and a variety of other publications and online exposure-age calculation schemes have mitigated the condition which originally motivated us.
We regret that Paul Bierman, the most recent reviewer, has been put to the inconvenience of reviewing this manuscript to no purpose. We appreciate Paul’s efforts and we apologize that they will go to waste.
Greg Balco, Berkeley Geochronology Center
Simultaneously with the tail end of this notable failure, Tibor Dunai and Fin Stuart published a similar manuscript in Quaternary Geochronology. Here is a link to this article (which unfortunately requires a library subscription to QG. It is hard to imagine, however, that any reader on this blog is not part of a major university with such a subscription). This is basically the same as the Balco and others manuscript discussed above, with the sarcastic and/or confrontational bits removed and an increased emphasis on describing not only the direct measurements and observations but also the method that was used to calculate the exposure ages or erosion rates from the measurements.
Even more recently, Kurt Frankel, Bob Finkel, and Lewis Owen published this article in EOS (again, AGU login needed). It covers more or less the same material, follows the Balco line rather than the Dunai line in emphasizing observational data over information about how ages and erosion rates were calculated, and adds its own emphasis on reporting more of the internal laboratory information, including sample weights, isotope ratios, and blank concentrations, that go into computing a cosmogenic-nuclide concentration.
Finally, I was contacted a couple of months ago by the editors of GSA Bulletin, who wanted to add some material to the `Instructions for authors’ outlining what information ought to appear in papers involving cosmogenic-nuclide measurements. These instructions were to be as compact as possible. I crushed my former three-page diatribe into four sentences as follows:
“For cosmogenic-nuclide exposure ages, authors must report: i) the location, elevation, thickness, density, and geologic context of each sample; ii) any corrections made for surface erosion and topographic or other shielding (or a statement that no such corrections were applied); iii) nuclide concentrations and measurement uncertainties; and iv) for nuclides whose production rates are composition-dependent, the major element composition of bulk rock and target mineral separates as well as the concentrations of important trace elements. For nuclides measured by accelerator mass spectrometry, authors must state the name and assumed isotope ratio of the standard to which the measurements were normalized. For example, the input data required to compute Be-10 or Al-26 exposure ages with the CRONUS-Earth online exposure age calculator (http://hess.ess.washington.edu) meets these requirements. Authors should also describe the production rate scaling scheme and calibration data set used in computing the ages, as well as the date and version number of any software used for this purpose. “
To summarize, if we include Dunai and Stuart, Frankel et al., the remarks in the Balco et al., 2008 exposure age calculator paper, and the abortive EOS article, but not the proposed GSA instructions for authors, we have three published and one unpublished instruction manuals for cosmogenic-nuclide data reporting. For the most part these say basically the same thing with varying levels of smugness, but they are inconsistent in some of the important details.
Why all this attention now? One possibility is that this sort of instructional article, although mind-numbingly boring, is likely to be highly cited. If you live and die by your citation index, it is to your advantage to publish articles of this type. From a less cynical perspective, however, this effort is actually a useful public service. It is important to get this right if the vast amount of money being spent on exposure-dating is not to be wasted from the perspective of future utility of the measurements. With adequate data reporting, the set of exposure ages that can in future be recalculated according to a common method and applied to answer large-scale paleoclimate questions continues to grow. My guess is that as more people try to do this — synthesize existing exposure-dating studies to answer larger questions — and discover how pathetic and scattershot is the data reporting in the existing literature, they become inspired to improve the situation.
A more important issue for the person who is actually going to publish exposure ages is: which of these only partially consistent prescriptions should you follow?
There are two main differences among them. The first is that Balco and Frankel emphasize reporting the actual measurements and observables, and don’t emphasize reporting what method, what production rates, etc. you used in computing ages from these measurements. Dunai on the other hand puts a lot more emphasis on reporting how you got from measurements to ages. From my perspective this issue is a no-brainer. I don’t care how you got from measurements to ages. When I read an exposure-dating paper, the first thing I am going to do is extract the measurements and feed them to the online exposure-age calculator so that I can compare them to other ages on a common basis. So in my opinion authors need only give the barest minimum of information as to how the ages were actually calculated. My emphasis is firmly on reporting all the measurements and observations that only the author knows and I can’t reproduce myself. This is the critical part.
The second difference is that Frankel, Finkel, and Owen would like you to report a lot more lab data, specifically the sample weight, measured isotope ratios, carrier amounts, and blank concentrations; whereas Balco and Dunai are happy with the final nuclide concentration and some summary information about how important the blanks are. The former position is presumably based on two considerations — determining how important the blank correction is from raw data, and determining what sort of ratio was actually being measured. Both of these are important in evaluating the reliability of the concentration measurement. The latter position trusts the author to get this right, and asks only for a summary value without the nitty-gritty of the author’s internal lab operations. In the past I’ve followed the latter line — just give me the summary result — but I am sympathetic to the desire to verify the author’s blank correction and error-propagation scheme. It is possible that sometimes people don’t get error propagation quite right.
Summary: You must include every observation or measurement needed to compute the exposure ages. It is helpful to include additional laboratory information that would help readers to evaluate the importance of the blank corrections and what sort of isotope ratios were actually being measured. Once these goals are accomplished, it is not necessary to go into very much detail about the method used to calculate the ages, or the production rates used…because, as discussed at length above, these things will most likely be obsolete soon.
Incomplete list of references:
Balco G., Stone J., Lifton N., Dunai T., 2008. A simple, internally consistent, and easily accessible means of calculating surface exposure ages and erosion rates from Be-10 and Al-26 measurements.Quaternary Geochronology 3, pp. 174-195.
Dunai T.J., Stuart F.M., 2009. Reporting of cosmogenic nuclide data for exposure age and erosion rate determinations. Quaternary Geochronology, v. 4, p. 437.
Frankel K.L., Finkel R.C., Owen L.A., 2010. Terrestrial Cosmogenic Nuclide Geochronology data reporting standards needed. (why are they not needed for extraterrestrial cosmogenic-nuclide geochronology?). EOS Transactions AGU, v. 91, p. 31.