Evidence that dispersal barriers influence blaberoid cockroach assemblages in a neotropical savanna–forest matrix

The forest refuge hypothesis to explain Amazonian diversity places an importance on dry savanna as a landscape feature isolating populations of humid forest-dwelling animals. Leaf-litter cockroaches are a useful group to study this as they are terrestrial and occupy a high-humidity microclimate. We examine the role of savanna, forests and flood zones in affecting blaberoid cockroach assemblage composition in the north Rupununi of Guyana, an expansive savanna–forest matrix. We test whether stretches of dry savanna and flood zones are barriers to dispersal for leaf-litter cockroaches. Our study area is the ˜10 000 hectares of Karanambu Eco-Lodge. We collected 1061 individuals of 17 species from 28 sites. Using a landscape classification based on satellite imagery, we generated models of relative isolation of the 28 sites using a cost of travel algorithm. We also used non-metric multi-dimensional scaling and Mantel tests to determine whether the similarity of cockroach assemblages could be explained by different landscape categories (savanna, forests and flood zones) or by geographic distance. We identified a single model that best predict observed variation in cockroach abundance and species richness. Differences in assemblage similarity were largely due to differences in the savanna–forest habitat axis. Geographic distance was also found to effect assemblage similarity. Our data suggest that savannas are barriers to dispersal, whereas flood zones are not. Future investigation of savanna driving speciation in some species is warranted, and this is a key feature of the forest refuge hypothesis.