Flexible spatio-temporal modelling of occupancy using GAMs: A case study with large carnivores.

This is my speed talk for the ISEC2020 conference (http://www.isec2020.org/). The occupancy state variable may be used as an ecological indicator to inform policy decisions for biodiversity conservation. In this context, occupancy models are increasingly used to investigate the occurrence of animal species while accounting for imperfect detection. Temporal and spatial variation in occupancy may be captured by incorporating relevant covariates. Here, we propose a GAM-like approach for estimating and assessing trends in occupancy. Using the Nimble R package, we employ a fully Bayesian approach using Markov chain Monte Carlo simulations to estimate the model parameters. We illustrate our work by a study of the Eurasian lynx (Lynx lynx) in France, in which occupancy probabilities are expressed as nonlinear functions of time and geographical coordinates, possibly in interaction, using data from a semi-structured citizen science program.