Scale dependence of diversity measures in a leaf-litter ant assemblage

A reliable characterization of community diversity and composition, necessary to allow inter-site comparisons and to monitor changes, is especially difficult to reach in speciose invertebrate communities. Spatial components of the sampling design (sampling interval, extent and grain) as well as temporal variations of species density affect the measures of diversity (species richness S, Buzas and Gibson's evenness E and Shannon's heterogeneity H). Our aim was to document the small-scale spatial distribution of leaf litter ants in a subtropical dry forest of the Argentinian Chaco and analyze how the community characterization was best achieved with a minimal sampling effort. The work was based on the recent standardized protocol for collecting ants of the leaf litter ("A.L.L.": 20 samples at intervals of 10 m). To evaluate the consistency of the sampling method in time and space, the selected site was first subject to a preliminary transect, then submitted after a 9-month interval to an 8-fold oversampling campaign (160 samples at interval of 1.25 m). Leaf litter ants were extracted from elementary 1 m(2) quadrats with Winkler apparatus. An increase in the number of samples collected increased S and decreased E but did not affect much H. The sampling interval and extent did not affect S and H beyond a distance of 10 m between samples. An increase of the sampling grain had a similar effect on S than a corresponding increase of the number of samples collected, but caused a proportionaly greater increase of H. The density of species m(-2) varied twofold after a 9-month interval; the effect on S could only be partially corrected by rarefaction. The measure of species numerical dominance was little affected by the season. A single standardized A.L.L. transect with Winkler samples collected <45% of the species present in the assemblage. All frequent species were included but their relative frequency was not always representative. A log series distribution of species occurrences was oberved. Fisher's alpha and Shannon's H were the most appropriate diversity indexes. The former was useful to rarefy or abundify S and the latter was robust against sample size effects. Both parametric and Soberon and Llorente extrapolation methods outperformed non-parametric methods and yielded a fair estimate of total species richness along the transect, a minimum value of S for the habitat sampled.