A Bayesian kernel density estimator for evaluating animal home range

Background/Question/Methods

Kernel density estimation is commonly used both to evaluate the home range size of animals and to evaluate the probability any location within an animal’s home range is used.  However, evaluating uncertainty surrounding home range size and the probability a location is used is challenging with current techniques.  We apply Bayesian techniques for estimating the posterior density of bandwidth matrices for a bivariate kernel density estimator.  Estimating bandwidth matrices with Bayesian techniques allows incorporation of uncertainty into matrix elements, which can be used to evaluate uncertainty in both home range size and the probability a location within the home range is used.

Results/Conclusions

We evaluate the utility of Bayesian kernel density estimation with a simulation study.  We show that home ranges can be estimated with reasonable field effort.  However, with too few samples there is considerable uncertainty at the outer boundary of the distribution, leading to considerable uncertainty in estimates of home range size.  We demonstrate the utility of this technique by estimating uncertainty in home range size and the probability a location within the home range is used by elk (Cervus elaphus) in the Black Hills, South Dakota.