The effects of evenness on subsampling analyses of taxic richness

Wagner Peter J., 2000: The effects of evenness on subsampling analyses of taxic richness. Abstracts with Programs - Geological Society of America 32(7): 132

It has long been known that evenness (i.e., the distribution of relative abundances among taxa within a sample) affects the shape of rarefaction curves. This means that the relative richnesses (i.e., numbers of taxa within a sample) of two samples with different evenness will change at different levels of subsampling. Note that most paleobiological studies label richness "diversity". Thus, interval X can have greater richness than interval Y after subsampling 250 specimens, yet interval Y can have greater richness than interval X after subsampling 500 specimens. Published demonstrations of this relationship focus on rarefaction of individual specimens. However, many paleobiological analyses rarefy occurrences (presences at localities). This introduces several issues, including: 1) the effects of evenness on occurrence distributions and subsampling of occurrences; 2) whether subsampling of particular units (specimens, occurrences, occurrences-squared) provide truer estimates of relative richness; and, 3) methods for testing hypotheses about relative richness when evenness is demonstrably different among samples. Rarefying simulated samples of known richness and evenness can explore these issues. Occurrence distributions always are much more even than are specimen distributions from the same samples. This is true under several different distributions of specimens (e.g., log-linear, geometric, broken-stick). When simulated samples have the same richness but different evenness, the rarefaction curves from samples with low evenness are persistently lower than the rarefaction curves from samples with high evenness. However, the difference is much more pronounced when subsampling the individuals generating those occurrences. Subsampling localities, in which one samples localities until particular numbers of individuals/occurrences are found, yield similar results. Locality subsampling based on occurrences-squared shows even less effect of evenness than subsampling of occurrences, but evenness still affects curve shape. Thus, evenness always affects inferences about richness from rarefaction curves, regardless of what one subsamples. When testing hypotheses about relative richness in different samples, an alternative approach is to use Monte-Carlo simulations to determine the probability of observing X taxa in Y samples, given an empirical estimate of the most likely evenness for each sample. When evenness varies markedly among samples, this can result in very different diversity curves than derived from rarefaction.