Using Monte Carlo Simulation to Evaluate Kernel-Based Home Range Estimators

Abstract

Kernel density estimators have been used to estimate home range size but little is known of their statistical properties. I applied kernel-based estimators of home range size, calculated from 95% probability contours of nonparametric density estimators, to computer-simulated radiolocation data. Four hypothetical models of home range suggested by Boulanger and White (1990) were used to evaluate bias and precision of these estimators in estimating known home range sizes. Kernel methods compared well with the best methods that are available for home range size estimation provided the appropriate level of smoothing was selected. I used brush rabbit (Sylvilagus bachmani) telemetry data to illustrate how Monte Carlo methods may also be used to assess estimator performance from field radiolocation data. A kernel estimator is preferred to a harmonic mean estimator in this example because it is less biased (i.e., the harmonic mean method has an inherent problem).