Does population viability analysis software predict the behaviour of real populations? A retrospective study on the Lord Howe island woodhen Tricholimnas sylvestris (Sclater)

Population viability analysis (PVA) is used for quantitatively assessing endangerment and comparing management options. Consequently, it is essential that PVA software packages be tested to determine whether they can accurately reflect the behaviour of real populations. A retrospective PVA was performed on the Lord Howe Island woodhen Tricholimnas sylvestris (1) to compare the predictions of five different PVA packages (INMAT, GAPPS, RAMAS/age, RAMAS/metapop and VORTEX); and (2) to test the predictions of PVA computer simulations against actual field data. All packages gave similar but unrealistic results under stochastic, density-independent conditions. When a ceiling density dependence model was applied, projections based on a carrying capacity calculated from the habitat area proved too high. A PVA based on the knowledge available at the time of the woodhen recovery program would have produced overly optimistic projections. Only when the carrying capacity was estimated from the observed historical population trends did the PVA packages give realistic predictions.