Urbanization is driving many species to inhabit modified landscapes, but our understanding of how species respond to this remains limited. Bats are particularly vulnerable due to their life-history traits but have received little attention. We describe the roosting behavior and roost site selection, including maternity roosts, for the Gambian epauletted fruit bat (Epomophorus gambianus) within a modified forest-savannah transition ecological zone in Ghana, West Africa. We compared characteristics of roost and non-roost sites to test the hypotheses that roost site selection is non-random and that maternity roost site selection differs from non-maternity roosts. Male bats were more likely to switch roost (mean= 0.49 ± 0.23 bat days, N=23) than females (mean= 0.33 ± 0.18 bat days, N=7) while linear distances between roosts used by males (255 ± 254 m) were significantly longer than for females (102 ± 71m) (t = 4.50, df =86, P <0.0001). Roost trees were more likely than non-roost trees to be bigger, taller, occur closer to buildings, and be in relatively open and less mature plots; maintaining such trees in modified landscapes could benefit the species. Lactating bats selected a subset of roost trees but significantly, those that contained a greater number of bats, a strategy which may reflect predator avoidance, or other social co-operation benefits. Although there was a preference for five tree species, other trees with preferred characteristics were also used. Our findings contribute to the understanding of how species utilize modified landscapes, which is important in the management of biodiversity in the Anthropocene.

Roost and non-roost tree characteristics data, including maternal roosts for Epomophorus gambianus. 

Tree height was measured using a Nikon® Forestry Pro laser rangefinder, DBH was measured at 1.4 m above the ground and crown diameter was measured as the longest on-the-ground extent across the crown of each tree using a tape measure.

For measeurements of tree Density,  Basal Area, a 17.8 m radius circular buffer (ca 218 0.1-ha area) was delineated around each tree to create plots. For each plot, the total tree basal area "BA" (m²/ha) and the tree density "TD" (trees/ha) per plot were calculated using the "points in polygon" tool in QGIS.

Distance to Building was measeured ast the shortest didtance to a building or structure (in meters).

Please email the corresponding author if you have questions.