THE PULL OF LITHIFICATION, GEOGRAPHY, AND COLLECTION SIZE ON PHANEROZOIC MARINE DIVERSITY PATTERNS

The 35-author revision of the Phanerozoic marine invertebrate diversity curve employing the Paleobiology Database has evolved considerably through 18 manuscript drafts. Here we discuss the more novel and significant methodological issues that have been resolved lately. The key issue is not the volume or quality of the data, which was shown early in the process to have no effect on the sampling standardized curves. Another early discovery was a large Pull of the Recent effect in range-through or boundary-crosser curves. The analyses have been enhanced recently to factor in the large Phanerozoic increase in the evenness of abundances. Subsampling quotas are now based on real or estimated specimen counts. Estimates are calibrated separately in each bin by collection-level rarefaction curves. This improved method suggests a smooth but unimpressive Cretaceous and Cenozoic radiation. Two newly analyzed factors are entirely responsible for what little remains of the increase. A third bias actually works against finding it, and had to be corrected to generate the small upswing. (1) Removing collections known to come from unlithified or poorly lithified sediments, but holding the subsampling quota constant, causes curves to drop substantially in the Cenozoic. Others have shown that this factor has a great impact on within-sample collection curves. (2) The geographic range of the sampled fossil record expanded considerably in the Cenozoic. The late Jurassic - Neogene increase disappears if the data are restricted to a 120 degree diameter circle placed independently in each bin to include the largest possible number of fossil occurrences. The restriction has no great effect on pre- or post-Pangea intervals other than the Cenozoic, so increased geographic dispersion in the Cenozoic is a bias instead of a signal. (3) Curves would crash going into the Cenozoic without accounting for the great imbalance between very large and small samples in that interval. A simple correction is to weight the probability of sampling a collection by the inverse of its total specimen count, which results in each collection contributing exactly the same number of specimens in the average subsampling trial. Although still more biases may lurk in the data, so many have been resolved that the latest estimates of Phanerozoic diversity are likely to capture some real biological signals.