The scale of effect depends on operational definition of forest cover-evidence from terrestrial mammals of the Brazilian savanna

Context Determining the appropriate scale at which to study species' interactions with their environment is a great challenge. Objective We investigated the spatial extent at which landscape structure affects the occurrence of four species of terrestrial mammalian herbivores in the Brazilian savannas and examined whether those scales could be explained by species ecological traits and how forest habitat was operationally defined. Methods Using maps of forest cover, camera trapping and occupancy modelling, we determined the relations between three landscape metrics (percentage of forest cover, patch density and edge density) and the occurrence of four species. To determine the optimal scale of effect for each species, we computed landscape metrics at different spatial extents (from 0.5 to 10 km radius) from camera trap locations and for three forest maps, each using different operational definitions of a forest pixel: minimum of tree cover of 25, 50 or 75%. Results The occupancy models revealed scales of effect of 0.5 to 2 km with the scale of effect being similar among three of the species. However, the probability of a scale of effect being detected depended upon how forest is operationally defined, being greater when forest was defined with greater tree cover, particularly for forest-specialist species. Conclusions Besides biological traits, the way habitat is operationally defined shapes our ability to detect the scale of effects. Thus, it is necessary not to adopt a multi-scale approach, but also to use multiple operational definitions of habitat, considering particularities of how each species interact with their environment.

Automated summary

The optimal scale of effect for the occurrence of terrestrial mammalian herbivores in Brazilian savannas was found to be 0.5 to 2 km, with the scale of effect being similar among three of the species. The way forest habitat is operationally defined significantly influences the probability of detecting the scale of effects, especially for forest-specialist species.