Spatial Bayesian modeling of presence-only data

When the only available information is the true presence of a species at few locations of a study area we refer to the data as presence-only data. Presence-only data problem can be seen as a missing data problem with asymmetric and partial information on a presence-absence process. This problem often characterizes ecological studies requiring the prediction of potential spatial extent of a species in suitable areas. Here we propose a Bayesian logistic spatial model adapted to presence-only data with environmental covariates available over the entire area. The spatial dependence among the observations is modelled indirectly as a latent Gaussian Markov field over the landscape, through a data augmentation MCMC algorithm we are able to estimate regression parameters jointly with the prevalence.