Ants are dominant in tropical forests and many species nest in hollow cavities. The manner in which species are vertically stratified in these complex habitats is not known, with lack of nest sites being proposed to limit ant populations. Here, we assess ant community stratification and nest site limitation in a lowland rainforest in New Guinea using experimental addition of artificial bamboo nests of two cavity sizes (small: ~12 mm large: ~32 mm diameter) placed at ground level, in the understorey, and in the canopy. We also conducted a pilot experiment to test the utility of nest translocation. Nests were checked for occupancy after 10 weeks and half of the occupied nests were then translocated between forest plots, while keeping same vertical position. Occupancy of small nests was much higher in the understorey and canopy than at ground level (~75% vs. ~25%). Translocation was successful, as a majority of nests was inhabited by the same species before and after translocation and there was no impact of translocation to a different plot compared to the control, except for a reduction in colony size at ground level. Our experiment demonstrates a vertical stratification in community composition of ants nesting in hollow dead cavities and shows that these ants are more nest site limited in the higher strata than at ground level. Use of small artificial cavities has great potential for future experimental studies, especially for those focused on arboreal ants, as occupancy is high and translocation does not negatively affect their colony size. Abstract in Tok Pisin is available with online material.

The ant mosaic is a concept of the non‐random spatial distribution of individual ant species in trees built upon the assumption of interspecific behavioural associations. However, colony identity and environmental variance may also play a role in species distribution. Here we assess the presence of ant mosaics in a primary forest ecosystem and whether they are structured by species' aggressive behaviours or by habitat filtering. We sampled arboreal ants from vertically stratified baits exposed in 225 canopy trees in a 9‐ha plot of primary lowland forest in Papua New Guinea, the largest forest area surveyed to detect ant mosaics. We performed behavioural tests on conspecific ants from adjacent trees to determine the territories of individual colonies. We explored the environmental effects on the ant communities using information on the plot vegetation structure and topography. Furthermore, we created a novel statistical method to test for the community non‐random spatial structure across the plot via spatial randomisation of individual colony territories. Finally, we linked spatial segregation among the four most common species to experimentally assessed rates of interspecies aggression. The ant communities comprised 57 species of highly variable abundance and vertical stratification. Ant community composition was spatially dependent, but it was not affected by tree species composition or canopy connectivity. Only local elevation had a significant but rather small effect. Individual colony territories ranged from one tree to 0.7 ha. Species were significantly over‐dispersed, with their territory overlap significantly reduced. The level of aggression between pairs of the four most common species was positively correlated with their spatial segregation. Our study demonstrates the presence of ant mosaics in tropical pristine forest, which are maintained by interspecific aggression rather than habitat filtering, with vegetation structure having a rather small and indirect effect, probably linked to microclimate variability.

Previous studies have provided important scientific information on ant species richness and composition relating to the effects of elevation, sampling approaches, stratification, and forest succession. Yet, they have primarily focused on single sites or regions. Knowledge of ant ecology should also include the impact of disturbance in various forest types. Tuna baiting and hand collection methods were used to investigate diversity and community composition of ants in 16 sites sampled across Papua New Guinea, in both disturbed and pristine forest, at heights ranging from 28 to 2,728 m above sea level. We found 176 species as a result of exposing 320 tuna baits and traversing 72 hand-searched plots. Baiting samples were strongly dominated by a few common species, while the hand-collecting captured more species per plot. The Chao 2 richness estimator for both methods predicted undersampling of the local community. As expected, ant species diversity and richness significantly decrease with increasing elevation. We observed, on average, greater species diversity of ground-dwelling ant communities in disturbed compared to undisturbed forests. The effect was not significant using multivariate randomisations, since the same species dominated both forest classes. The unexpected pattern of ant species richness being locally higher in the disturbed sites is driven by our sampling of undisturbed communities at all elevations, but sampling of the disturbed communities only up to 1,600 m above sea level. Hence, future studies should consider more locations, aiming ideally for an equal sampling effort to capture disturbance stage and elevation.