Skip to content

Antarctic subglacial sediment has almost no beryllium-10. Repeat: Antarctic subglacial sediment has almost no beryllium-10*.

June 26, 2018

Pinned on the bulletin board in my office, between the Chart of the Nuclides and a New Yorker cartoon showing a cutaway image of the Earth labeled “Crust. Mantle. Chewy Nougat Center,” is a 2013 article in the Economist entitled “How Science Goes Wrong.” It summarizes reviews in biomedical sciences, mostly in the areas of psychology and pharmaceutical research, that showed that a disproportionate number of results in these fields were not reproducible: when other researchers tried to repeat the studies, they got different results. The article highlights funding and publication incentives that lead to overvaluation of results that are interesting and surprising, but irreproducible and likely erroneous. Since then, these studies have prompted a lot of reexamination of funding and publication practices in biomedical fields, and development of some new programs, mainly led by NIH, to do a better job of replicating important results. It is very likely that in Earth science we do an even worse job of ensuring that important results are reproducible, because we share the same incentive structures that discourage funding and publication of replicate studies, and then we add additional obstacles of fieldwork in remote regions, unique samples or data sets from inaccessible places that would be very hard to obtain again, and the fundamental problem that you can’t go back and re-run events in Earth history. But even given these constraints, we don’t do a great job of this and we should do better.

So in this post I am highlighting a study that we really know is reproducible, because it’s been done twice. It’s a recent paper in Nature by Jeremy Shakun and a variety of co-authors that shows, basically, that there is almost (although not quite) zero beryllium-10 in quartz from the AND-1B core in McMurdo Sound, Antarctica. Taken in context, this isn’t particularly surprising, because this core is in 900 meters of water and the material in the core is derived from subglacial erosion beneath the Antarctic ice sheet. We’d only expect to see Be-10 if any of the quartz had been exposed at the surface in some part of Antarctica that was ice-free and above sea level during the past few million years. There’s already a lot of very good evidence, including marine oxygen isotope evidence for the continued existence of the Antarctic ice sheets for more than 30 million years, the stratigraphic record on land in Antarctica, and the fact that parts of Antarctica that probably did deglaciate in past warm-climate periods are below sea level so would still be covered by water anyway, that there haven’t been any such ice-free areas. So we don’t expect a priori to find significant amounts of Be-10 in Antarctic subglacial sediments. The paper, however, very strongly highlights the significance of not finding anything as additional evidence that the East Antarctic Ice Sheet has been around for a while.

What you won’t find in this paper is the fact that it’s the second time this project has been done. In 2009, Dan Morgan was a graduate student at the University of Washington working with his Ph.D. advisor Jaakko Putkonen on applications of cosmogenic-nuclide geochemistry to measuring erosion and sediment transport rates in the Antarctic Dry Valleys, and Dan and Jaakko came up with the idea of looking for for cosmic-ray-produced nuclides in the then-recently-drilled ANDRILL AND-1B core. The Ferrar Glacier, which drains part of the upper Dry Valleys, supplies sediment to the general vicinity of the core site, and their idea was that the marine sediment section in McMurdo Sound might contain some record of past export of sediment from the Dry Valleys. Surface sediment in the DV is mostly derived from surfaces with very low erosion rates and has extremely high cosmogenic-nuclide concentrations, and the idea was that a small amount of this material should be detectable even if it was diluted by a lot of Be-10-free debris from subglacial erosion of fresh rock.  So Dan and Jaakko asked for and received some samples from the core as well as seed funding from PRIME Lab for a few trial AMS measurements. Dan then did essentially exactly the same thing that was done in the Shakun paper, the main exception being that he used smaller samples that yielded less quartz. Here are Dan’s results from 2009 (kindly provided by Dan at my request) on the bottom, with the results from the Nature paper (specifically, Figure 3a in the paper) on top.

What’s being shown here is a comparison between the amount of Be-10 present in process blanks (processed in exactly the same way as the samples, only with no sample) and samples. In both figures, the amounts of Be-10 in the blanks are shown in gray, and the amounts in the samples are shown in colors: all are represented by Gaussian uncertainties. The core depths in the Morgan samples are equivalent to samples A-D in the Shakun paper; E-H are deeper in the core. The main difference here is that there are a lot more blanks represented in the Shakun data set, and they are mostly lower, centered around 8000-10000 atoms Be-10 rather than the 15000-20000 in Dan’s results. The University of Vermont cosmogenic-nuclide lab, where the Shakun measurements were done, has done an enormous amount of work in the last ten years to make their Be-10 blanks very low and very stable, and it’s worked. A secondary difference is that AMS performance has gotten a bit better in the last ten years as well, so the later measurements are a bit more individually precise. So the amounts of Be-10 observed in the samples are similar in the aggregate between the two studies, but the lower blanks shown in the upper figure have the consequence that one can, in fact, conclude at good confidence that some of the samples in the Shakun paper have Be-10 present at levels above blank, whereas for the Morgan data the samples are all indistinguishable from blank. But the overall implication of each data set — that forms the core argument of the paper — is very similar: there is very little Be-10 in quartz in the AND-1B core.

Dan also in 2009-10 sent me splits of two of his samples — from 61 and 81 m depth in the core — for Ne-21 measurements at BGC. This is potentially interesting because Ne-21 is stable, so material that might have been exposed at the surface much longer ago than could be recorded by Be-10 would still retain cosmogenic Ne-21. It’s a lot more difficult to detect small amounts of cosmogenic Ne-21, because there is always a nonzero amount of non-cosmogenic Ne-21 produced by decay of naturally occurring U and Th, and it is difficult to separate cosmogenic from U/Th-derived inventories with the precision needed to detect small amounts of surface exposure. Essentially the null hypothesis is that if no cosmogenic Ne-21 is present, and given a particular neon closure age for the sediment source rocks (which should be effectively constant on million-year time scales in the core if the source area of the sediments is not highly variable) then the Ne-21 concentration should just be a constant multiple of the U/Th concentration. The following figure shows that this is, in fact, the case for two samples (here the U and Th concentrations are expressed as eU, which approximates total alpha particle production):

You can’t exclude cosmogenic Ne-21 with much confidence from only two not-particularly-precise measurements, so this is a bit of a sideshow to the Be-10 data (although an interesting one if you are a noble gas geochemist). But the point is that Ne-21 is completely accounted for by production from U and Th, which again tends to indicate that the quartz is not derived from surfaces that were ever ice-free for very long.

Dan’s measurements were never published. I am speculating a bit as to why because I was not closely involved, but I believe the thinking was that there was no expectation that any Be-10 would be found: if there was some, that would be a story, but no Be-10 equals no story. In the Shakun paper, it is the opposite. No Be-10 is a big story. As an aside, I am concerned that the paper oversells the significance of this result, because possible Antarctic ice sheet change scenarios that are hypothesized to have occurred in Plio-Pleistocene warm periods and are relevant in light of near-term climate change involve deglaciation of marine embayments where the bed of the ice sheet is below sea level. If this happened during past warm periods it would not be recorded by in-situ-produced Be-10 in quartz. So this proxy record has zero sensitivity to the type of deglaciation event that is most concerning and most plausible in the near future. But the main point that I want to make here is that, in contrast to probably too much Earth science research, we know this result is replicable, because it’s been replicated. Although the Economist would probably argue that the funding incentive structure is such that the Shakun work would never have been funded if the Morgan results had been published, it is unfortunate that there is no way to know from the present paper that this work has been done twice, and I think Dan should get some credit for having tried this nine years ago.

* Trying to make this title short and punchy also makes it incorrect. It should refer to in-situ-produced Be-10 in quartz, rather than just Be-10. In fact, some Antarctic subglacial sediment has lots of Be-10, it’s just not in-situ-produced in quartz.

No comments yet

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: