An empirical model of species coexistence in a spatially structured environment.

Ecological theory has long supported the idea that species coexistence in a homogeneous habitat is promoted by spatial structure, but empirical evidence for this hypothesis has lagged behind theory. Here we describe a Neotropical ant-plant symbiosis that is ideally suited for testing spatial models of coexistence. Two genera of ants, Allomerus cf. demerarae and three species of Azteca are specialized to live on a single species of ant-plant, Cordia nodosa, in a Western Amazonian tropical rain forest. Empirically, using census data from widely separated localities, we show that the relative colonization abilities of the two ant genera are a function of plant density. A parameterized model shows that this pattern alone is sufficiently robust to explain coexistence in the system. Census and experimental data suggest that Azteca queens are better long-distance flyers, but that Allomerus colonies are more fecund. Thus, Azteca can dominate in areas where host-plant densities are low (and parent colony-sapling distances are long), and Allomerus can dominate in areas where host-plant densities are high. Existing spatial heterogeneity in host-plant densities therefore can allow regional coexistence, and intersite dispersal can produce local mixing. In conclusion, a dispersal-fecundity trade-off appears to allow the two genera to treat spatial heterogeneity in patch density as a niche axis. This study further suggests that a spatially structured approach is essential in understanding the persistence of some mutualisms in the presence of parasites