Analyzing occupancy data collected utilizing multiple detectors over a single time period
Traditional occupancy models using presence/absence data require multiple observation periods for detection and presence probabilities to be estimable. When camera trap data is used, this requirement typically results in the creation of artificial time periods that may or may not make biological sense. We propose a sampling scheme that uses only a single sampling period but utilizes multiple cameras in a single plot. When one or more cameras in a plot capture an animal, the detection probability can simply be estimated from the number of cameras that capture the individual and the traditional confounding of the probability of presence and the probability of detection given the animal is present is avoided. We implement this model using a hierarchical Bayesian model and tested the method by performing a computer simulation study with animals following a bias random walk through a plot with multiple cameras.
In a landscape with uniform resistance to movement, the proposed method works as well as the multiple observational periods method using the same sampling effort (ie number of camera days). We extend these results to a landscape with non-uniform resistance and consider optimal camera placement in such an environment. Finally, this method is applied to wildlife trail-usage data from the San Francisco Bay Ecoregion to estimate occupancy probabilities across a range of habitats and species.