Dynamic riskscapes for prey: disentangling the impact of human and cougar presence on deer behavior using GPS smartphone locations

Prey species adjust their behavior along human-use gradients by balancing risks from predators and humans. During hunting seasons, prey often exhibit strong antipredator responses to humans but may develop tolerance in suburban areas to exploit human-mediated resources. Additionally, areas with high human activity may offer reduced predation risk if apex predators avoid such locations. This study examined mule deer Odocoileus hemionus behavioral responses to risks from humans and their primary predators, cougars Puma concolor, contextualized by differences in risk levels between study sites, individual risk exposure, and human habituation. We framed our investigation using three non-mutually exclusive hypotheses: (H1) neutral impact, (H2) human shielding (human tolerance driven by cougar avoidance), and (H3) super-additive risk (human avoidance dominating behavior). We controlled for deer phenology and diel period, recognizing that deer behavior varies with these temporal dynamics. Spatiotemporal cougar encounter risk was quantified using GPS collar data, while spatiotemporal human encounter risk and use intensity were quantified using GPS smartphone data. Our results supported H2 and H3, emphasizing the significance of site- and individual-level variation in risk exposure and human use intensity. Deer managed cougar risk adaptively, but humans emerged as the dominant perceived risk, varying by study site. At the site with higher cougar density and lower human hunting pressure, deer exhibited antipredator responses to humans based on individual exposure to human activity, except during hunting season, when tolerance for cougars increased. Conversely, humans were the dominant risk at the site with lower cougar density and greater human hunting pressure. Deer behavior varied significantly across a gradient of human use, influenced by nuanced human presence and predation risks, which were discernible using human smartphone data.