A total of 36 deer have been tracked in the Pennypack Ecological Restoration Trust (PERT) and adjacent properties from 2007-2013. PERT is a private, non-profit conservancy located about 25 km northeast of central Philadelphia that manages 3.3 km2 of mature forests, regenerating woodlands, riparian forests, and fields of cool- and warm-season grasses in the Pennypack Creek valley. Bryn Athyn College has studied deer movements with the goal of assessing habitat preferences and usage of the deer for management purposes. Studies of animal movements have traditionally focused on time allocation of individual animals. In this study, the movement of 32 deer instrumented with GPS/GSM radio-collars that transmit spatial and temporal data at high-frequency intervals (5 min) were monitored within PERT and surrounding lands. Twenty of the deer overlapped within an accessible area of PERT that was surveyed for vegetation presence, type, height, and density (geobotanical polygon). Brownian bridge trajectories were constructed between the deer location fixes and overlain on the geobotanical polygon within the study area. The resulting map of the “deer perceptual landscape” was then modeled based on habitat characteristics. The model was extrapolated to the entire study area and compared to the density of fixes of all tracked deer.
Results/Conclusions
The complete “perceptual landscape” map of the deer indicates locations of high deer frequency and partial probabilities of deer visits to various landscape features and habitats. The perceptual landscape appears to be predetermined by the frequency and distribution of human movements and human-made structures as well as locations of overgrown patches and neglected areas. The probability density function contains ecological covariates such as: distance to nearest building, distance to nearest road, parking lots, mature forest, young forest, lawn, annually mowed field, biannually mowed field, neglected lot, paved local road, unpaved industrial lot, field, regularly mowed field, and shrubs. Following linear regression analysis, the variables showing P>0.05 were omitted. The resulting model had five variables and was statistically significant. The perceptual landscape model was cross-validated by being applied to individuals that lived away from the geobotanical polygon. The model appears to predict the deer density with high precision.