The night shift: temporal activity and local use respond unevenly to human infrastructure in an Amazonian mammal community

Human infrastructure can affect wildlife by altering habitat access, human encounter risk, hunting pressure, and the spatial distribution of resources, but species may respond through different behavioral pathways. Some species may avoid infrastructure locally, others may shift the timing of activity, and many may show weak or context-dependent responses. We used camera-trap data from 483 stations across the Las Piedras region of Madre de Dios, Peru, to evaluate how distance from human infrastructure was associated with temporal activity and local use across 30 Neotropical mammal taxa. We modeled two complementary binomial responses: nocturnality, defined as the probability that independent detections occurred at night rather than during the day, and detection-night probability, defined as the probability that a species was detected on a given active camera night. Hierarchical generalized additive models estimated nonlinear species-specific and trophic-group responses to infrastructure distance while accounting for species identity, trail placement, spatial structure, survey grid, camera station, and relevant environmental covariates. The best-supported final models combined species-specific and trophic-group distance smooths, while trophic-group-only models remained strongly unsupported. Across the primary near–far contrast from 0.85 to 7.20 km from infrastructure, responses were uneven across both species and response dimensions. Lowland tapir showed the clearest temporal response, with predicted nocturnality higher near infrastructure than farther away. Detection-night responses were generally smaller in absolute magnitude, but small rodents showed higher predicted detection-night probability nearer infrastructure and collared peccary showed higher predicted detection-night probability farther from infrastructure. Brocket deer showed a similar positive but less certain detection-night trend farther from infrastructure, and tayra showed a small, supported increase in detection-night probability nearer infrastructure. Many carnivores showed weak, nonlinear, or uncertain near–far detection-night responses. Overall, Amazonian mammals did not respond to human infrastructure through a single behavioral pathway. Instead, temporal and local-use responses were concentrated in particular taxa and ecological contexts, highlighting the value of evaluating when species are active alongside where they are detected.