Selection and spatial arrangement of rest sites within northern tamandua home ranges

The distribution of suitable rest sites is considered to be a key determinant of spatial patterns in animal activity. However, it is not immediately evident which landscape features satisfy rest site requirements or how these sites are configured within the home range. We used Global Positioning System (GPS)/accelerometer telemetry to investigate rest site selection at the home-range scale for northern tamanduas Tamandua mexicana on Barro Colorado Island (BCI), Panama. We developed models specifying each tamandua as the individual experimental unit and averaged coefficients to produce population-level estimates. Tamanduas had on average 17.8 (+/- 8.1) rest sites within their home range and used 1.36 (+/- 0.51) on any given day. These rest sites tended to be located in the core of tamandua home ranges, with active locations associated with the periphery of the home range. Rest sites were positively associated with (1) a high density of Attalea butyracea palm trees; (2) elevation; (3) tall vegetation. There was a slight negative relationship between the distribution of rest sites and slope, and no apparent relationship between rest site selection and relative distance to forest canopy gaps. From focal animal observations, we identified that tamandua rest sites were typically located in trees (90%), with 25% (12 of 49) occurring in palms. We contend that northern tamanduas on BCI selected vegetated arboreal rest sites because of reduced likelihood of detection from terrestrial predators in these sites. Our models identified considerable individual variation in rest site selection, which suggests that the practice of pooling individuals and fitting models at an aggregate level may be inappropriate for certain types of habitat selection research