COS 66-2
Multi-scale effects of exurban development on birds at protected and unprotected sites: An application of an occupancy model accounting for false positive and false negative detections
Exurban development, the construction of low-density residential homes in a rural landscape, is the fastest growing type of land use in the United States and is prominent in the southern Appalachian region. A potential consequence of exurban development is the loss and fragmentation of native habitat. We developed a Bayesian model that accounts for false positive and false negative detections to make inferences about how the occupancy of six forest-dwelling, Neotropical migrant birds is related to multi-scale attributes of exurban development. Many ecological models estimate the probability of false negative detections. However, false positive detections are also known to occur in ecological data, and if this type of imperfect detection is not accounted for in models, estimates will be biased. We built on previous attempts to account for false positive and false negative detections in occupancy models, while addressing some of the criticisms of past models. Through simulations, we evaluated our model parameterization before modeling the relationship between avian occupancy and land use at National Forest, land trust, and unprotected sites in Macon County, North Carolina. We performed Bayesian model selection and model averaging with a Bayesian Information Criterion weights approximation, and we evaluated models’ predictive ability.
Results/Conclusions
Our model parameterization generated accurate and precise posterior distributions even when the true positive detection probability was less than the false positive detection probability, there was a low rate of confirmed observations, and there were observation confirmation errors. It may also be possible to identify phantom species (species that were absent from all sites but were detected erroneously) with our occupancy models. Results from modeling the effects of exurban development indicated that landscape- and local-scale covariates influenced posterior occupancy probabilities more than site-scale covariates and that landscape composition and elevation had a greater effect on posterior occupancy probabilities than configuration. The Black-throated Blue Warbler and Wood Thrush had the lowest posterior occupancy probabilities of the six focal species. National Forest sites had high occupancy, but land trust sites exhibited patterns similar to unprotected sites. Our findings can provide guidance to land use planners and land trusts as they decide how to respond to exurban development. Also, our study demonstrates the application of an improved occupancy model that can generate more accurate inference by accounting for both types of imperfect detection while describing heterogeneity.