Thursday, August 7, 2008 - 9:00 AM

COS 83-4: Rank-occupancy curves: a potential new tool to compare species distributions

David G. Jenkins, University of Central Florida


The distribution of species among sites is important to the assessment of rarity and spatial distributions of biodiversity. Species distributions among sites are typically reported as a species presence/absence matrix with species in rows (R) and sites in columns (Q). Q-mode analyses compare sites for both details of composition and summary statistics, including species richness and beta diversity. R-mode analyses compare species for details with species co-occurrence and nestedness analyses, but surprisingly no summary statistics appear to exist beyond frequency distributions of species’ occupancy. Occupancy frequency distributions are often used in theoretical and applied ecology, but are affected by multiple factors (binning, study design, habitat heterogeneity, range position, etc.) and remain a puzzle. Rank-occupancy curves (analogous to rank-abundance curves) derive from occupancy frequency distributions but avoid problems of binning and retain species information to potentially enable better tests of ecological, systematic and phylogeographic hypotheses for rank-occupancy.  


As an initial test of rank-occupancy, I compared empirical rank-occupancy curves from diverse data sets to eight classes predicted by McGeoch and Gaston (2002) and that correspond to mixtures of core and satellite species.  Data sets represented 11-119 species (among plants, invertebrates, amphibians, fish, mammals) and 6-2183 sites. Only three of the eight classes were obtained, corresponding to occupancy-frequency distributions that are (a) bimodal, symmetrical, (b) bimodal, satellite-mode dominant, and (c) unimodal satellite.  The prevalence of distributions dominated by satellite species across multiple studies of diverse taxa in various places suggests that concerns for inadequate study design (sample grain, number, and extent) are generally mitigated.  Habitat heterogeneity and relatively large sample extent are probably responsible for curves weighted to satellite species, while curves that also include core species are relatively homogeneous in habitat and/or have relatively small sample extent. With further testing, rank-occupancy distributions may help to: (a) evaluate study scale relative to dispersal or range scales, (b) test ecological, systematic, and phylogeographic expectations for spatial distributions; (c) identify poorly-distributed native species for conservation; and (d) compare changing distributions through time (e.g., invasives or threatened species).