Modeling utilization distributions in space and time

By:  and 



W. Van Winkle defined the utilization distribution (UD) as a probability density that gives an animal's relative frequency of occurrence in a two-dimensional (x, y) plane. We extend Van Winkle's work by redefining the UD as the relative frequency distribution of an animal's occurrence in all four dimensions of space and time. We then describe a product kernel model estimation method, devising a novel kernel from the wrapped Cauchy distribution to handle circularly distributed temporal covariates, such as day of year. Using Monte Carlo simulations of animal movements in space and time, we assess estimator performance. Although not unbiased, the product kernel method yields models highly correlated (Pearson's r - 0.975) with true probabilities of occurrence and successfully captures temporal variations in density of occurrence. In an empirical example, we estimate the expected UD in three dimensions (x, y, and t) for animals belonging to each of two distinct bighorn sheep {Ovis canadensis) social groups in Glacier National Park, Montana, USA. Results show the method can yield ecologically informative models that successfully depict temporal variations in density of occurrence for a seasonally migratory species. Some implications of this new approach to UD modeling are discussed. ?? 2009 by the Ecological Society of America.
Publication type Article
Publication Subtype Journal Article
Title Modeling utilization distributions in space and time
Series title Ecology
DOI 10.1890/08-1131.1
Volume 90
Issue 7
Year Published 2009
Language English
Larger Work Type Article
Larger Work Subtype Journal Article
Larger Work Title Ecology
First page 1971
Last page 1980
Google Analytic Metrics Metrics page
Additional publication details