Siliceous and Delicious

Slow start to the day, again. Instead of working the pre-Cenozoic rad project, as called for by my schedule, I instead knocked out a quick spreadsheet to calculate how much radiolarian test material I’d need to pick for the silicon isotope project. Interestingly, the only mass estimates for individual radiolarians I could find gave wildly diverging results. According to the Abelmann paper, which uses geometric models similar to the ones I used in my thesis, C. davisiana should weigh about 38 µg—each! This would be fantastic, because we’d only need two shells to make the measurement. Alas, the Moore paper (based on actually weighing specimens) estimates the mass of radiolarians (average across assemblages) as closer to 0.05-0.1 µg. This means 100s to 1000s of shells, which is clearly not feasible. I hope Abelmann is right, but I can’t say I trust her numbers more than Moore’s. I’ll have to try it out for myself, and keep my fingers crossed in the mean time.

Spent quite a while picking through Dave’s material with my trusty 0000 paintbrush, but man, those rads are tiny. They are substantially smaller than the ones I’ve been picking out of the Ohio concretion residues, they’re really quite delicate, and that makes me somewhat nervous with regard to getting enough Si to make the isotope measurement. I had discussed ways of weighing the rads with Rama, and he had suggested weighing a small falcon or PCR tube before and after adding the radiolarians. But after a single picking session, I noticed that the weight of the tube had changed by milligrams—orders of magnitude more than the change should have been due to adding radiolarians. This, no doubt, was to be blamed on adding and subtracting water with the paintbrush, the mass of which will far exceed the flimsy radiolarians floating in it. How, then, to weigh them properly? Well, the first thing is to get access to a proper scale, since the one we have in the lab neither has the required precision (it doesn’t down to µg), nor does it deliver a steady weight (the last digit wanders all over the place). The second thing will be to find a way to weigh the buggers without dealing with fluctuating amounts of water. For example: weigh a glass slide (or petri dish, for that matter), pick radiolarians onto it, then heat to evaporate the water and weigh again.

The FIB session I had booked for the afternoon was cancelled (the machine is down, once again), so I fled the office and decided instead to head to the Atomic Bean Café with Beau to do a little bit of caffeine-fuelled out-of-office work. Decided to continue working on the diatom diversity project, since I had gotten decent mileage out of that on Monday. Ran the O2W algorithm 50 times and plotted up the means. Strangely, two of the time bins (looks like it’s 54.5 and 49.5 Ma) have zero values—if they are ‘inquorate’ (i.e. insufficient occurrences-squared to meet the threshold value), they should be NA-valued, not zero. Anyway, if I remove those zero values, the plot looks like this:

For comparison, the by-list, unweighted plot from a couple of moons ago:

The two plots look qualitatively similar, though the ~35 Ma peak in diversity is much less pronounced in the O2W plot than in the UW plot. In fact, that peak barely registers… and I’m not entirely sure I’d even be able to defend an increasing diversity trend from 40-0 Ma at all… it’s basically flat-lined. It might be interesting to repeat these sampling-standardization exercises with different thresholds, and comparing the shape of curves that way—obviously, those time bins with the fewest samples will disappear first, but what is the shape of the curve that remains? Is it reasonably robust as the sampling threshold rises?

The next step—which I think is the final step in redoing Rabosky’s results—is to apply the sampling probability transformation to the data. First, I also need to figure out why the getStats results summary for the O2W subsampling appears to return zero values for all stats in two time bins, and why the first few time bins have NA values for their nrecords entry (which, if I’m remembering correctly, is the same as occurrences). If they have nonzero diversity, they must also have nonzero nrecords.