As the human footprint upon the landscape expands, wildlife seeking to avoid human contact are losing the option of altering their spatial distribution and instead are shifting their daily activity patterns to be active at different times than humans.  In this study, we used game cameras to evaluate how human development and activity were related to the daily activity patterns of the nine-banded armadillo (Dasypus novemcinctus) along an urban to rural gradient in Arkansas, USA during the winter of 2020-2021.  We found that armadillos had substantial behavioral plasticity in regard to the timing of their activity patterns; >95% of armadillo activity was nocturnal at six of the study sites whereas between 30 and 60% of activity occurred during the day at 3 other sites.  The likelihood of diurnal armadillo activity was best explained by the distance to downtown Fayetteville (the nearest population center) and estimated ambient sound level (both indices of human activity) with armadillos being most active during the day at quiet sites far from Fayetteville.  Furthermore, armadillo activity occurred later during the night period (minutes after sunset)at sites near downtown and with higher anthropogenic sound.  Anecdotal evidence suggests that the observed activity shift may be in response to not only human activity but the presence of domestic dogs.  Our results provide further evidence that human activity has subtle nonlethal impacts on even common, widespread wildlife species.  Because armadillos have low body temperatures and basal metabolism, being active during cold winter nights likely has measurable fitness costs.  Nature reserves near human population centers may not serve as safe harbors for wildlife as we intend, and managers could benefit from considering these nonlethal responses in how they manage recreation and visitation in these natural areas.

We used motion-triggered wildlife game cameras to document the presence and activity patterns of nine-banded armadillos at each study site. We used a combination of Spypoint Force Dark (Spypoint Inc, Victoriaville, Quebec, Canada) and Browning Strike Force (Browning, Morgan, Utah, USA) model cameras. To determine camera locations, we overlaid each study site with a series of 150 m x 150 m grids. Each grid was assigned a number and we randomly chose grids in which cameras would be placed. No grid received more than one camera at a time although at some sites multiple cameras were placed within the same grid at different times of the year. We used this approach to ensure cameras were dispersed across the study sites. Within selected grids, we placed cameras in locations with clear lines of sight for at least 15 m and in locations that would not be visible from roads or trails to avoid theft or the photography of people. We placed cameras on either trees or tripods 50 cm above the ground. We programmed cameras to operate 24 hrs per day, take bursts of 3 photos at each trigger, and to reset after 5 seconds. We downloaded memory cards from cameras approximately every 2 weeks. Due to constraints on the number of cameras, cameras were not deployed at all sites simultaneously and not all sites had equal sampling effort. At each specific location where a camera was deployed, we recorded the coordinates so we could later use a GIS (ArcMap 10.2; ESRI Inc, Redlands, CA, USA) to calculate landscape variables around each camera.

We reviewed all photographs using the Timelapse 2.0 software (Greenberg et al. 2019). Timelapse 2.0 allowed us to rapidly review photographs, extract metadata (date, time), and to assign species identity to each group of photographs. Timelapse 2.0 also allowed us to group photographs into sequences. We assumed all photographs taken within 5 min of one another were one sequence of the same individual and these photographs were combined into one unique detection to reduce double-counting individuals. We chose 5 min for the sequence length because most detections of armadillos foraging in front of cameras lasted for less than 5 min and allowed even slow-moving individuals to move out of range of the camera. We scored each armadillo detection as occurring during day or night based on the sunrise and sunset time for the 15^th day of that month so that our definitions of diurnal and nocturnal activity varied across the study to account for changing daylengths. We used the time of the first armadillo detection within the sequence to assign the timing of activity.

See attached file to understand data sheet.