REMOVING SAMPLING BIASES FROM DIVERSITY CURVES: THE UNEXPECTED ROLES OF ALPHA AND BETA DIVERSITY

The reconstruction of diversity curves is confounded by the incomplete preservation and sampling of the paleontologic record. However, the analytical standardization of sampling intensity among time intervals can ameliorate some biases. Large-scale, literature-based diversity studies often use only presence-absence data, since relatively few sources report local abundance data. Several standardization methods that use subsampling have been proposed for such studies. For example, the OW method randomly subsamples lists of taxa from different fossil localities, summing the number of taxa on each until a certain number of total taxon occurrences is achieved. (Classical rarefaction works slightly differently, but gives similar results.) The O2W method is similar, but it squares the number of taxa on lists before summing. Here, we discuss the accuracy of these methods in the context of the way biodiversity is structured into alpha (within-habitat) and beta (between-habitat) components. We show that no method is universally applicable; the method of choice depends on such factors as the amount of beta diversity (and thus the geographic scale of the study). For datasets containing local abundance data from the Upper Jurassic of Western Europe and the Eocene of many continents, none of the methods accurately correct for local sampling biases. However, OW and O2W are members of a methodological continuum (OxW, where x can take any value), and an intermediate method (O1.4W) seems to perform well for these regional and global datasets. Recently, Alroy, Marshall et al. (2001) applied analytical standardization methods to a dataset consisting of marine localities from the Middle Ordovician to Carboniferous and Late Jurassic to Paleogene. Their O2W diversity curves challenged the reality of the Cenozoic rise in biodiversity seen in Sepkoski's curves, suggesting that this rise is an artifact of biased sampling. However, we show that O2W overcorrects for sampling biases in this instance. The remaining diversity curves are either consistent with the traditional curve or ambiguous due to the limited temporal and taxonomic coverage of Alroy et al.'s analyses: there is insufficient evidence to claim that heterogeneous sampling intensity drives the observed Cenozoic taxonomic radiation.