Geographical, environmental and intrinsic biotic controls on Phanerozoic marine diversification

The Paleobiology Database now includes enough data on fossil collections to produce useful time series of geographical and environmental variables in addition to a robust global Phanerozoic marine diversity curve. The curve is produced by a new ‘shareholder quorum’ method of sampling standardization that removes biases but avoids overcompensating for them by imposing entirely uniform data quotas. It involves drawing fossil collections until the taxa that have been sampled at least once (the ‘shareholders’) have a summed total of frequencies (i.e. coverage) that meets a target (the ‘quorum’). Coverage of each interval’s entire data set is estimated prior to subsampling using a variant of a standard index, Good’s u. This variant employs counts of occurrences of taxa described in only one publication instead of taxa found in only one collection. Each taxon’s frequency within an interval is multiplied by the interval’s index value, which limits the maximum possible sampling level and thereby creates the need for subsampling. Analyses focus on a global diversity curve and curves for northern, southern and ‘tropical’ (30°N to 30°S) palaeolatitudinal belts. Tropical genus richness is remarkably static, so most large shifts in the curve reflect trends at higher latitudes. Changes in diversity are analysed as a function of standing diversity; the number, spacing and palaeolatitudinal position of sampled geographical cells; the mean onshore–offshore position of cells; and proportions of cells from carbonate, onshore and reefal environments. Redundancy among the variables is eliminated by performing a principal components analysis of each data set and using the axis scores in multiple regressions. The key factors are standing diversity and the dominance of onshore environments such as reefs. These factors combine to produce logistic growth patterns with slowly changing equilibrium values. There is no evidence of unregulated exponential growth across any long stretch of the Phanerozoic, and in particular there was no large Cenozoic radiation beyond the Eocene. The end-Ordovician, Permo–Triassic and Cretaceous–Palaeogene mass extinctions had relatively short-term albeit severe effects. However, reef collapse was involved in these events and also may have caused large, longer term global diversity decreases in the mid-Devonian and across the Triassic/Jurassic boundary. Conversely, the expansion of reef ecosystems may explain newly recognized major radiations in the mid-Permian and mid-Jurassic. Reef ecosystems are particularly vulnerable to current environmental disturbances such as ocean acidification, and their decimation might prolong the recovery from today’s mass extinction by millions or even tens of millions of years.

M uch of the palaeobiological literature adheres to the notion that environmental factors determine levels of local, regional and global diversity. For example, claims that diversity in the deep fossil record tracks a latitudinal gradient or reflects biogeographical provincialism go back at least four decades (Fischer 1960; Stehli et al. 1969; Valentine 1970). However, even after so many years of discussion we are still at an early stage of testing such ideas with rigorous statistical analyses, thanks in no small part to the extreme difficulty of removing sampling overprints.

Solving this problem requires assembling large amounts of data resolved to the level of individual fossil localities (e.g. Knoll et al. 1979) instead of working with simple lists of geological first and last appearances that lack any contextual information (e.g. Sepkoski 1984, 1997). It also requires separating the concepts of accurate, comprehensive and uniformly sampled. I will show that these properties are not only different but impossible to attribute to a single diversity curve. Finally, it requires rethinking what it would mean to demonstrate causal relationships between diversity and either biotic or abiotic factors (McKinney and Oyen 1989).

Here, I employ these new strategies to investigate controls on Phanerozoic marine invertebrate diversity at the global scale and within three major palaeolatitudinal belts. The analysis focuses primarily on comparisons between global, southern and tropical trends because there is little northern hemisphere shelf area throughout much of the early Palaeozoic. It employs new statistical methods (Alroy 2010) that explicitly seek to reconstruct each taxonomic sampling pool’s relative size, i.e. its richness in the strict ecological sense (cf. Hurlbert 1971; Gotelli and Colwell 2001), as opposed to drawing a fixed amount of data for no other reason than that it seems fair (Alroy 1996; Miller and Foote 1996; Alroy et al. 2008). It works within a framework of rigorous time series analysis and instead of taking a piecemeal approach, it considers biotic, environmental and geographical factors simultaneously. Gaps in coverage of the literature are still evident and additional variables may need to be considered, but strong patterns in the data suggest that diversification is neither random nor governed by a single set of factors.