Patterns of ant species diversity and turnover across 2000 km of Amazonian floodplain forest

Aim To determine the effect and relative importance of geographic and local environmental factors on species richness and turnover of ant assemblages in floodplain forests across the Amazon basin. Location Twenty-six mature forest sites scattered along the entire extension of the Amazon River in Brazil. The study area encompassed nearly 18° of longitude and 3.5° of latitude. Methods Systematic collections of ants were performed at each site during the low-water season (i.e. when forests are not inundated) using three complementary sampling methods. We used variance partitioning techniques to assess the relative effects of the spatial (latitude and longitude) and environmental (rainfall, length of the dry season and flood height) variables on ant species richness and composition. Results There was a twofold variation in the number of species per site, which was largely explained by inter-site variations in rainfall seasonality and flooding intensity. In general, there were more species at sites located in the western part of the basin, where the dry season is less severe, or near the river estuary, where precipitation is also high and flooding is less intense. Ant community composition was also affected by environmental heterogeneity. For instance, some species only occurred at those sites less affected by the river's seasonal flooding, whereas others were mostly associated with the drier or wetter regions of the basin. In addition, the turnover of species increased significantly as geographic distances increased. Nevertheless, the rate of change was small given that many species had a broad distribution across the study area. Main conclusions Ant distribution patterns along the floodplain forests of the Amazon appear to be controlled to a relatively large extent by the current gradient in flooding intensity and - most importantly - in precipitation. Altered rainfall regimes resulting from global warming and land-use change thus have the potential to influence these patterns. © 2009 Blackwell Publishing Ltd