Kernel Methods for Estimating the Utilization Distribution in Home‐Range Studies

TLDR

Kernel methods offer flexible, nonparametric alternatives to parametric and Fourier‑transform approaches for estimating animal utilization distributions. The paper describes kernel methods for nonparametric estimation of utilization distributions from random locational samples in animal home ranges. The authors illustrate fixed and adaptive kernel approaches and discuss smoothing parameter choices through two data‑analysis examples.

Abstract

In this paper kernel methods for the nonparametric estimation of the utilization distribution from a random sample of locational observations made on an animal in its home range are described. They are of flexible form, thus can be used where simple parametric models are found to be inappropriate or difficult to specify. Two examples are given to illustrate the fixed and adaptive kernel approaches in data analysis and to compare the methods. Various choices for the smoothing parameter used in kernel methods are discussed. Since kernel methods give alternative approaches to the Anderson (1982) Fourier transform methods, some comparisons are made.