Fast Bayesian inference for large occupancy data sets, using the Polya-Gamma scheme

• Published in: arXiv.org
• Main Authors: Diana, Alex, Dennis, Emily, Matechou, Eleni, Morgan, Byron
• Format: Paper
• Language: English
• Published: Ithaca Cornell University Library, arXiv.org 30.07.2021
• Subjects: Autocorrelation; Bayesian analysis; Biodiversity; Datasets; Gaussian process; Occupancy; Spatial data; Statistical inference
• ISSN: 2331-8422

In recent years, the study of species' occurrence has benefited from the increased availability of large-scale citizen-science data. Whilst abundance data from standardized monitoring schemes are biased towards well-studied taxa and locations, opportunistic data are available for many taxonomic groups, from a large number of locations and across long timescales. Hence, these data provide opportunities to measure species' changes in occurrence, particularly through the use of occupancy models, which account for imperfect detection. However, existing Bayesian occupancy models are extremely slow when applied to large citizen-science data sets. In this paper, we propose a novel framework for fast Bayesian inference in occupancy models that account for both spatial and temporal autocorrelation. We express the occupancy and detection processes within a logistic regression framework, which enables us to use the Polya-Gamma scheme to perform inference quickly and efficiently, even for very large data sets. Spatial and temporal random effects are modelled using Gaussian processes, allowing us to infer the strength of spatio-temporal autocorrelation from the data. We apply our model to data on two UK butterfly species, one common and widespread and one rare, using records from the Butterflies for the New Millennium database, producing occupancy indices spanning 45 years. Our framework can be applied to a wide range of taxa, providing measures of variation in species' occurrence, which are used to assess biodiversity change.