A simple landscape-scale test of a spatially explicit population model: patch occupancy in fragmented south-eastern Australian forests

The results of a landscape-scale test of ALEX, a widely used metapopulation model for Population Viability Analysis (PVA), are described. ALEX was used to predict patch occupancy by the laughing kookkaburra and the sacred kingfisher in patches of eucalypt forest in south-eastern Australia. These predictions were compared to field surveys to determine the accuracy of the model. Predictions also were compared to a ‘‘naı̈ve’’ null model assuming no fragmentation effects. The naı̈ve null model significantly over-predicted the number of eucalypt patches occupied by the sacred kingfisher, but the observed patch occupancy was not significantly different from that predicted using ALEX. ALEX produced a better fit to the field data than the naı̈ve null model for the number of patches occupied by the laughing kookaburra. Nevertheless, ALEX still significantly over-predicted the number of occupied patches, particularly remnants dominated by certain forest types – ribbon gum and narrow-leaved peppermint. The predictions remained significantly different from observations, even when the habitat quality of these patches was reduced to zero. Changing the rate of dispersal improved overall predicted patch occupancy, but occupancy rates for the different forest types remained significantly different from the field observations. The lack of congruence between field data and model predictions could have arisen because the laughing kookaburra may move between an array of patches to access spatially separated food and nesting resources in response to fragmentation. Alternatively, inter-specific competition may be heightened in a fragmented habitat. These types of responses to fragmentation are not incorporated as part of traditionally applied metapopulation models. Assessments of predictions from PVA models are rare but important because they can reveal the types of species for which forecasts are accurate and those for which they are not. This can assist the collection of additional empirical data to identify important factors affecting population dynamics.