Large-scale experimental assessment of coyote behavior across urban and rural landscapes

Abstract

Carnivores must navigate the complexities of human modifications to their environment. Natural resources and biodiversity decline in urban areas, while people in rural areas often pose greater direct risk through actions such as hunting. To evaluate if carnivore populations adapt their behavior to local risks in rural and urban environments, we compared behavioral responses to novel objects in coyotes (Canis latrans). We placed an attractant at arrays of 30 camera-trap stations at 16 pairs of urban and rural field sites across the USA, with a novel object placed at half of the stations. Coyotes exhibited more cautious behavior and remained farther from the attractant at all sites with the novel object; however, urban coyotes got closer to the attractant than rural coyotes. There were few behavioral differences between urban and rural coyotes and none between eastern and western coyotes. Coyotes across the USA exhibit neophobic behavior but urban coyotes, especially western coyotes, are willing to take more risk (i.e., be closer to the attractant). The consistency in most metrics of coyote behavior suggest that solutions developed in one area could be universally useful. This study also demonstrates the effectiveness of a large, collaborative approach to studying broad-scale patterns in behavioral traits.

Keywords: Bayesian statistics; Behavior; Canis latrans; Detection; Novel object; Urbanization.

Fig. 1

Map of field sites across the United States of America (USA), with color coding representing each of the adjacent urban and rural areas. (a) Adjacent sites are exemplified with close-up maps for Chicago (b), Cleveland (c), and New York City (d), USA, with red circles for urban areas and green squares for rural areas.

Fig. 2

Model-predicted probabilities of coyote detection at 16 urban and 16 adjacent rural sites surveyed across the USA. Boxes show the median, upper 75%ile, and lower 25%ile; violin plots behind boxes show the full posterior distribution. Treatment sites were those with a novel object. Asterisks indicate instances where 95% credible intervals failed to overlap, indicating a difference between groups.

Fig. 3

Model-predicted visit durations in seconds of coyotes detected at sites surveyed across the United States. Boxes show the median, upper 75%ile, and lower 25%ile; violin plots behind boxes show the full posterior distribution. Treatment sites were those with a novel object.

Fig. 4

The average proportion of time coyotes spent at different distances from an attractant placed in front of a camera trap at 16 field sites surveyed across the USA. Panel a depicts rural versus urban sites; panel b depicts control versus treatment sites, which had a novel object.

Fig. 5

The average proportion of time coyotes spent in different behavioral states at 16 field sites surveyed across the USA. Treatment sites were those with a novel object.