Core areas, a frequently reported characteristic of animal space use, are used to answer many ecological questions. Numerous methods to estimate core areas have been applied, but current methods are either based on arbitrary decision rules (e.g., 50% isopleth), or are unable to differentiate between clustered and random point patterns. Using arbitrary rules for defining core areas ignores the biological basis for the existence (or lack of existence) of core areas. Organisms scale differently across species, seasons, and social classes, thus the actual spatial data from an animal should drive core area estimation rather than pre-defined rules. We present a novel statistical approach, based on kernel density methods, to estimate the isopleth that best delineates core areas when available data are limited to temporally independent animal locations with continuous spatial support on a bounded region. A key concept to this method is the optimal partitioning of a clustered spatial point process into a set of ≥ 2 complete spatially random point processes such that the union of their support is equal to the home range. We account for the uncertainty in the core area estimate by allowing it to be a random set that is parameterized and characterized with Bayesian methods. We present data from coyotes (Canis latrans), bobcats (Lynx rufus), and red-shouldered hawks (Buteo lineatus) collected from the same study site in Texas to determine if the optimal isopleth for core areas differs between species and to determine if there is any support for the common use of the 50% isopleth for identifying core areas.
Results/Conclusions
Optimal isopleths for characterizing individual core areas ranged between 18.7 and 71.5%. There was significant overlap between species in the optimal isopleths, indicating no species-specific range of isopleths. We also found no evidence to support the use of the 50% isopleth to subjectively define core areas. Our method is the first to accommodate the uncertainty in core area boundaries and can be extended to define areas of concentrated space use at multiple spatial scales (i.e., sub-core areas). Additionally, it can easily be generalized to other home range estimation methods and applications, such as epidemiology.
