BAYESIAN MODELS FOR SPECIES DISTRIBUTION MODELLING WITH ONLY-PRESENCE RECORDS

Abstract

One of the central issues in ecology is the study of geographical distribution of species of flora and fauna through Species Distribution Models (SDM). Recently, scientific interest has focused on presence-only records. Two recent approaches have been proposed for this problem: a model based on maximum likelihood method (Maxlike) and an inhomogeneous poisson process model (IPP). In this paper we discussed two bayesian approaches called MaxBayes and IPPBayes based on Maxlike and IPP model, respectively. To illustrate these proposals, we implemented two study examples: (1) both models were implemented on a simulated dataset, and (2) we modeled the potencial distribution of genus Dalea in the Tehuacan-Cuicatlán biosphere reserve with both models, the results was compared with that of Maxent. The results show that both models, MaxBayes and IPPBayes, are viable alternatives when species distribution are modeled with only-presence records. For simulated dataset, MaxBayes achieved prevalence estimation, even when the number of records was small. In the real dataset example, both models predict similar potential distributions like Maxent does.