Context. Anthropogenic landscape changes have substantial effects on biodiversity and animal populations worldwide. However, anthropogenic landscape change can take a variety of forms, and its effects on wildlife species can vary by landscape context and scale. It is therefore critical that studies of the effects of anthropogenic landscape change on wildlife consider landscape context and model effects of landscape change at multiple scales. Mesopredators serve as an excellent case study of scale-dependent and even contradictory effects of anthropogenic landscape change, because mesopredator populations can respond both positively and negatively to landscape change depending on its form (e.g. agricultural production, urbanisation) and scale.

Aims. The aim of this study was to model relationships between multiple aspects of anthropogenic landscape change, including agricultural production and the proliferation of non-native shrubs, and relative abundance of three mesopredator species: Virginia opossum (Didelphis virginiana), striped skunk (Mephitis mephitis), and raccoon (Procyon lotor).

Methods. We summarised landscape variables at two scales and built both single-scale and multi-scale models linking relative abundance of mesopredators to landscape variables.

Key results. We found that relative abundance of all three species was most related to either the transition from natural areas to agricultural areas (brown-to-green gradient) or the likelihood of presence of non-native shrubs. We also found that responses to anthropogenic landscape change varied by spatial scale; for instance, skunks and raccoons had higher relative abundance in parts of Illinois with more forest cover and agriculture, respectively, but avoided more highly forested and agricultural areas, respectively, at smaller scales.

Conclusions. Mesopredator responses to landscape change were highly variable and scale-dependent, but were generally related to transitions from forest to agriculture or the presence of non-native shrubs.

Implications. Our study demonstrates the need to model effects of anthropogenic landscape change at multiple scales, given the differing results that can be achieved when landscape variables are measured at multiple scales.