Assessing spatial PVA models of arboreal marsupials using significance tests and Bayesian statistics

The predictions of stochastic metapopulation models of four species of arboreal marsupial were compared to field data on patch occupancy. The species examined were the greater glider Petauroides volans, the mountain brushtail possum Trichosurus caninus, the ringtail possum Pseudocheirus peregrinus and the yellow-bellied glider Petaurus australis. The models were developed for a system of 39 eucalypt patches in southeastern Australia, embedded within an exotic plantation of radiate pine Pinus radiata. Additionally, two alternative (null) models were developed for each species: one in which it was assumed that there was no impact of fragmentation; and a second that only modeled local population dynamics with no dispersal between patches. Congruence between the probability of occupancy of patches predicted by the metapopulation model and the actual occupancy observed in the field was assessed using tests of significance based on logistic regression. The relative performance of the three competing models was assessed for each species using Bayesian statistics. The results demonstrated that the metapopulation model made reasonable predictions for the greater glider and the yellow-bellied glider. However, none of the models predicted the observed increase in population density of mountain brushtail possums and ringtail possums in patches relative to continuous forests. Incorporating edge effects and inter-specific interactions would improve the predictions for these two species. (C) 2001 Elsevier Science Ltd. All rights reserved.