Quantifying the space requirements of a population is a fundamental problem in spatial ecology, particularly as it relates to the identification of important utilization areas and the designation of protected areas for conservation and wildlife management.

Traditionally, population space use estimation techniques scale up from the individual to the population level by aggregating individual animal tracks and then using a single pooled distribution estimator like minimum convex polygons (MCP) or kernel density estimation (KDE). These techniques fail to account for the high levels of temporal autocorrelation in modern tracking datasets, and estimates are often sensitive to the number of individuals sampled.

We introduce a new population kernel density estimator (PKDE) that accounts for temporal autocorrelation in tracking data, propagates uncertainty from the individual to the population level, accounts for inter-individual variation when scaling up to the population level, and is not highly sensitive to the number of individuals tracked. Through a combination of simulated data and empirical GPS tracking datasets from three species: (a) grizzly bear (Ursus arctos horribilis); (b) lowland tapir (Tapirus terrestris); and (c) bobcat (Lynx rufus), we demonstrate that PKDE produces minimally biased estimates of population-level space use compared to conventional methods like MCP and KDE.

The use of conventional estimators can lead to substantial underestimation of population space usage, making them unsuitable for area-based conservation planning. The statistically efficient PKDE estimator provides relatively unbiased estimates of space use with fewer individuals sampled. This method has been made available as a function in the ctmm R package.