Estimating Population Density From Presence-Absence Data Using a Spatially Explicit Model

Presence-absence (detection/non-detection) data are routinely collected in wildlife studies where identification of individuals is impossible or impractical and where the detection method may be able to detect only the presence of an individual rather than a count (e.g., track or scat surveys). Estimating population density from presence-absence data usually is assumed to be difficult or impossible unless certain restrictive assumptions are made or supplementary information is collected. Recently, Chandler and Royle (2013) presented an extension of a spatially explicit capture-recapture model that estimates population density from spatially replicated counts in unmarked populations. We extended the model of Chandler and Royle (2013) to situations where only presence-absence data can be collected. The model assumes that individuals can be detected at multiple sample units, producing spatially correlated detections. A spatially explicit model of the detection process is then fit to the correlated detection data using Bayesian methods. We report on the performance of the model using simulation and illustrate its use with a practical example estimating the abundance and density of red foxes (Vulpes vulpes) from remote camera surveys in the Grampians National Park in southeastern Australia. Results from simulations suggest the model produces unbiased estimates of density if device spacing is less than the radial length of a typical home range and the number of encounter occasions is high (i.e., at least 10). Application of the model to camera detection data from foxes in the Grampians National Park resulted in an estimated density of 0.22foxes/km(2) (95% CI: 0.16-0.53). For this dataset, precision of the density and detection parameter estimates were increased by the use of an informative prior distribution for the home-range-scale parameter. The current model should apply widely to a range of sampling situations that result in spatially correlated detection/non-detection data such as bait take, scat surveys, tracking stations, and chew cards, to name a few. (c) 2015 The Wildlife Society.