Testing a Simulation Model for Population Viability Analysis

Abstract

We conducted a field-based test of the widely available generic computer simulation model VORTEX for population viability analysis (PVA). The model was used to predict the abundance of three species of arboreal marsupials in a system of 39 remnant patches of Eucalyptus forest embedded within a 5050-ha area of exotic radiata pine (Pinus radiata) forest in southeastern Australia. The marsupial species were: greater glider (Petauroides volans), mountain brushtail possum (Trichosurus caninus), and common ringtail possum (Pseudocheirus peregrinus). Predictions were generated for scenarios in which: (1) the rate of exchange of animals between patches was varied, (2) different models for the migration of animals between habitat patches were invoked, (3) different levels of immigration (or dispersal) from a large, neighboring source area were simulated, (4) variations in habitat quality between remnant patches were incorporated in the model, and (5) the influence of the pine matrix surrounding the remnant patches was modeled. These predictions were then compared with the results of extensive field-based spotlighting surveys to estimate abundance and patch occupancy of arboreal marsupials in the remnant system. Models in which the carrying capacity of remnants was assumed to be a simple function of patch area and home range size substantially overpredicted the number of occupied patches and total abundance of animals. Only when model complexity was increased, by incorporating effects of within-patch habitat quality based on dominant forest type together with negative effects of the surrounding pine matrix on dispersal mortality, was better congruence obtained between predictions from VORTEX and observed values for patch occupancy and overall animal abundance. We knew the size, spatial location, and interpatch distances for all of the eucalypt remnants, as well as the precise time they were fragmented, information that is unavailable for the vast majority of human-modified landscapes. The biology and ecology of the three target species also were well known. Even though our studies were underpinned by extensive background information, accurate matching of model predictions to field data was possible only with a posteriori selection from an array of plausible scenarios of dispersal patterns and rates. Hence, an important over-arching conclusion from our results is that conservation biologists should be cautious in predicting the actual dynamics and response of populations in fragmented systems, even relatively simple well-known ones.