Evaluating sampling sufficiency and the use of surrogates for assessing ant diversity in a Neotropical biodiversity hotspot

Ecologists often seek sampling protocols that are both effective and relatively simple, i.e. those that provide a balance between the advantages obtained through sampling completeness and the costs involved in species sorting. Here we explored ways of simplifying a protocol devised for assessing geographic patterns of ant species richness and composition in the savanna-dominated Cerrado landscape of central Brazil. This protocol, which retrieved up to 88.5% of the expected number of species, was employed in five different sites located up to 1200 km apart from each other. In each site three transects (>1 km apart) were established, and within each transect 80 arboreal and 80 ground pitfall traps were installed. We then evaluated the degree of congruence in species richness and composition between data originated from the full data set and various subsets of the data, that either contained a reduced number of samples (in different spatial configurations) or potential surrogates of the entire ant fauna. Our main findings show that by sorting specimens from just a subset of the ant genera we retrieved most of the information existing in the full data set, even though the amount of work involved in ant sorting was reduced by about 50%. Reducing the number of samples taken in each site, as expected, also reduced the amount of work involved in ant sorting. However, in most of our analyses the level of congruence with the full data set was lower and/or became more erratic when sampling intensity was reduced to one-third or less of the original number of samples. More species were retrieved when samples were spread over multiple transects than when they were located in only one or two transects, indicating that even at relatively small scales (a few kilometers) there is a turnover of ant species within the same type of habitat. Based on these results we suggest that at least half of the samples from our original protocol are necessary to adequately describe large-scale patterns of ant species richness and beta diversity in Neotropical savannas. We also suggest that, whenever possible, sampling within a given site should be done in multiple locations in order to maximize the number of species collected.