Improving Inventory Efficiency: A Case Study of Leaf-Litter Ant Diversity in Madagascar

For most invertebrates, ecologists lack efficient inventory methods for assessing geographic patterns of species richness, complementarity (distinctness), and areas of endemism. I evaluated the efficiency of quantitative inventory methods developed for leaf-litter ants in tropical rain forests in eastern Madagascar. The aim was to maximize the number of species captured per sampling effort in a systematic design subject to standard statistical analysis. I used species and complementarity accumulation curves to evaluate the efficiency of the inventory design based on all ant species sampled and based on four species-rich genera that could potentially act as surrogates for estimating total ant diversity. I evaluated: (1) efficiencies of pitfall and Winkler sifting methods to capture leaf-litter ant assemblages, (2) effects of sample size and spacing on completeness and ranking of species richness, (3) completeness of complementarity values, and (4) four species-rich ant genera that could potentially act as surrogates for estimating total ant richness. Inventories were conducted at 15 sites along elevational gradients at four localities. I collected 117044 worker ants belonging to 381 species from pitfall and leaf-litter samples. Species-accumulation curves showed decreased rates of species accrual with increased sampling but were still increasing slowly at the end of each transect. Leaf-litter samples had a greater rate of species accumulation than pitfall samples and were more cost effective. Rates of species accumulation were not improved by taking samples 10 or 15 m apart, compared with samples taken 5 m apart. Relative ranking of between-site patterns of species richness and complementarity stabilized before 25 samples and was not predicted to change with the addition of 50 more stations to each transect. The nonparametric incidence-based coverage estimator (ICE) of species richness was not useful for assessing sampling completeness because of a strong relationship between sample size and estimated species richness. Defining completeness based on the expected percentage increase in species from a doubling of sampling effort was more useful. To project the increase in species richness from increased sampling, I fit a nonasymptotic logarithmic model to the observed species accumulation curve. Four potential surrogate genera (Pheidole, Strumigenys, Tetramorium, and Hypoponera, containing 68, 51, 51, and 50 species, respectively) were efficiently surveyed at each elevation; each was significantly correlated with the richness of species and complementarity values based on all other ant species. Strumigenys showed the highest correlation in both species richness and complementarity.