Pattern without bias: local-regional richness relationship revisited

For decades, ecologists have been testing for species saturation by using regression analysis to determine the relationship between local and regional species richness. The cumulative result of scores of studies and meta-analyses has led to a general consensus that evidence of species saturation is relatively uncommon. However, the bias induced on the regression by the arbitrary choice of local and regional area has threatened to undermine this consensus and has even led to the proposal to abandon the regression method entirely. Nonetheless, the use of regression analysis to determine local-regional richness relationships continues. We performed a meta-analysis of almost 100 data sets using a log-ratio-based regression model that is free of the scale-induced bias inherent in more commonly used models. This reanalysis contradicts the general consensus: saturated relationships are at least as common as unsaturated patterns. However, a discouragingly large proportion of studies produce no discernable patterns at all. An explanation for such a plethora of species saturation patterns poses a challenge to our understanding of the processes that govern community assembly.