An analysis of feral pig (Sus scrofa) home ranges in a Hawaiian forest using GPS satellite collars

Early Polynesians first transported a subspecies of the common pig Sus scrofa, to the Hawaiian Islands in 400 A.D. Subsequent reintroductions of the Sus genus have created a varying mix of feral pig populations that are threatening forested and farmland areas. Recently resource managers have turned to Global Positioning System (GPS) telemetry devices to gather movement data to resolve problems surrounding feral pig management. In order to use these new tools effectively, the goal of this study was to evaluate methods of home range estimation and recommend a method for use with data acquired through GPS telemetry. To address this goal, I conducted two studies: (1) a comparison of home range estimation methods and (2) a GPS collar error evaluation. For the first study, I used GPS telemetry relocations from four feral pigs, collared within a heterogeneous island forest habitat to evaluate home range estimation methods. Five estimation methods were evaluated: (1) minimum convex polygon (MCP), (2) kernel density estimator (KDE) with the bandwidth or smoothing parameter (h) based on the average marginal covariance between the x- and y-coordinates (KDE-href), (3) KDE with the bandwidth h calculated using least squares cross-validation (LSCVh ) method, (4) KDE likelihood cross-validation (CVh) method, and (5) the Brownian bridge movement model (BBMM). The five methods resulted in home range estimates ranging from 11.3 ha to 85.9 ha. Estimates using the BBMM clearly reflect landscape characteristics, whereas the KDE- href over-smoothed data and the LSCVh method failed to provide estimates when animal subjects exhibited multiple centers of activity. The second study evaluated Quantum 5000 GPS collars (Telemetry Solutions Inc.) across a topographically challenging landscape in terms of location error and fix rate. Results showed an average of 154.3 m 95% Circular Error Probable (CEP) for all stationary sites; meanwhile removing stationary sites with extreme topography (i.e., at the base of near vertical cliffs) resulted in a 95% CEP of 31.7 m for the collars. Understanding telemetry error is necessary to produce more robust movement models necessary for wildlife management. In addition, the overall findings suggest that movement and home range in feral pigs may be a function of gender. This work provides important verification for the BBMM when used in wildlife studies, improving our ability to determine wildlife-habitat relationships and strengthen management decisions.