THE SPATIOTEMPORAL DYNAMICS OF INSULAR ANT METAPOPULATIONS

The ant species inhabiting 129 small Bahamian cays were surveyed annua lly over a 5-yr period. Stepwise multiple linear regression analyses r evealed that plant species number was the best single predictor of ant species richness. Vegetated area or cay elevation (depending on the a nalysis) were second selected variables. Variation in the selected pre dictor variables explained 56.8-71.7% of the variation in ant species number. An ''all-subsets'' regression approach yielded similar results . Stepwise multiple logistic regression analyses revealed that the pro bability of occurrence of Dorymyrmex pyramicus was positively related to vegetated area and elevation, whereas that of Pheidole punctatissim a was positively related to plant species number and distance. Species turnover was documented on 42% of the cays inhabited by ants over the course of the study. Mean annual relative turnover was low and variab le, ranging from 0.83 to 22.37%/yr when calculated on a per-species ba sis, and from 2.13 to 5.70%/yr when calculated on a per-island basis. Sampling error was quantified, and pseudoturnover was found to inflate observed turnover for only one species (P. punctatissima). A comparis on of immigration and extinction rates indicated the existence of a dy namic equilibrium between these two processes. The probabilities of im migration and extinction (pooled over all species) were both positivel y related to vegetation height. Observed turnover rates for ants were lower than those reported from other, more limited studies of insular ants and from studies of insular arthropods in general, and were more similar to documented vertebrate turnover rates. Incidence functions r evealed that the three most common species, Brachymyrmex cf. obscurior , D. pyramicus, and P. punctatissima, displayed different patterns of distribution in relation to insular ant species richness.