Multivariate analysis of scale-dependent associations between bats and landscape structure

The assessment of biotic responses to habitat disturbance and fragmentation generally has been limited to analyses at a single spatial scale. Furthermore, methods to compare responses between scales have lacked the ability to discriminate among patterns related to the identity, strength, or direction of associations of biotic variables with landscape attributes. We present. an examination of the relationship of population- and community-level characteristics of phyllostomid bats with habitat features that were measured at multiple spatial scales in Atlantic rain forest of eastern Paraguay. We used a matrix of partial correlations between each biotic response variable (i.e., species abundance, species richness, and evenness) and a suite of landscape characteristics to represent the multifaceted associations of bats with spatial structure. Correlation matrices can correspond based on either the strength (i.e., magnitude) or direction (i.e., sign) of association. Therefore, a simulation model independently evaluated correspondence in the magnitude and sign of correlations among scales, and results were combined via a meta-analysis to provide an overall test of significance. Our approach detected both species-specific differences in response. to landscape structure and scale dependence in those responses. This matrix-simulation approach has broad applicability to ecological situations in which multiple intercorrelated factors contribute to patterns in space or time.