Note: This article is from Conservation Magazine, the precursor to Anthropocene Magazine. The full 14-year Conservation Magazine archive is now available here.

Does Population Viability Analysis Underestimate Extinction Risk?

July 24, 2001

Many conservation biologists have been skeptical of population viability analysis (PVA), which wildlife managers use to predict species’ extinction risks and rank management options. But in a pair of recent papers, researchers show that while some PVA models can underestimate extinction risk, they are reasonably accurate when used correctly.

The study showing that PVA can underestimate extinction risk was by Barry Brook, then at Macquarie University in Sydney, Australia, and his co-authors, and is in the June 2000 issue of Conservation Ecology. This is the first standardized comparison of widely used PVA packages.

The study showing that PVA predicts population declines accurately was by Brook et al. and is in the 23 March issue of Nature. This is the first comprehensive evaluation of the reliability of PVA predictions.

While previous PVA comparisons showed that the predicted extinction risk sometimes varied widely from package to package, the inputs in these comparisons were not standardized. In their Conservation Ecology paper, Brook and his colleagues used six life-history types representative of those modeled in recent PVAs and analyzed each with five widely used PVA packages (two individual-based packages, GAPPS and VORTEX, and three matrix-based packages, INMAT, RAMAS Metapop, and RAMAS Stage). Five of the life-histories were based on real populations (including a reptile, birds, and mammals) with detailed monitoring and ecological data. The sixth life-history was hypothetical and represented a bird- or mammal-like species with low growth rate and low levels of environmental variation.

Brook and co-authors found a striking discrepancy between the individual-based and the matrix-based PVA packages: the extinction risk predicted by the former was an average of 16 percent higher than that predicted by the latter. One of the biggest differences between these two types of PVA packages is that individual-based models consider the sexes separately and so account for variation in sex ratio, whereas standard matrix-based models consider the sexes together and so do not account for variation in sex ratio.

To test whether this difference explains the 16 percent discrepancy in predicted extinction risk, the researchers used RAMAS Stage, which is matrix-based but can consider the sexes separately and so account for sex ratio variation. When RAMAS Stage was set to consider the sexes together, the predicted extinction risk was 1.6 percent — similar to that of the standard matrix-based models. In contrast, when RAMAS Stage was set to consider only females, the predicted extinction risk was about 13 percent — similar to that of the individual-based models. Because the discrepancy in predicted extinction risk essentially disappeared when RAMAS Stage was set to account for sex ratio variation, the researchers concluded that the standard matrix-based packages were underestimating the true extinction risk.

“To keep matrix-based PVA packages from underestimating the extinction risk, managers using matrix-based PVA packages should consider only the sex that is limiting for breeding,” say Brook and his co-authors. While this generally means considering only females, there are two exceptions. First, only males should be considered for those rare species in which males are the limiting sex, such as the emu Dromaius novaehollandiae, where the male incubates eggs from many females and rears the chicks himself. Second, both sexes should be considered for monogamous species because in this case either can be limiting.

As long as the PVA packages tested were used appropriately, the researchers found that they predicted very similar extinction risks, suggesting that all five are valid for comparing management options.

Brook et al. support this conclusion in their Nature paper. They tested accuracy of the same five PVA packages with existing data from long-term studies of 21 wildlife populations ranging from 11 to 57 years. They used the first half of the data from each population to set the model parameters and the second half to test the accuracy of the model predictions. The results showed that the population declines predicted by the PVA packages closely matched the declines that had actually occurred.

“PVA predictions are surprisingly accurate, given adequate data,” says Brook. “Our research validates PVA as a useful tool for managing threatened populations.”

Brook is now at Northern Territory University in Darwin, Australia; his co-authors on the Conservation Ecology paper are Mark A. Burgman of the University of Melbourne, Victoria, Australia, and Richard Frankham of Macquarie University in Sydney, Australia.

Brook’s co-authors on the Nature paper are Julian O’Grady, Andrew P. Chapman, and Richard Frankham of Macquarie University in Sydney, Australia; Mark A. Burgman of the University of Melbourne, Victoria, Australia; and H. Resit Akcakaya of Applied Biomathematics in Setauket, New York.

Further Information:
Brook, B.W., M.A. Burgman, and R. Frankham. 2000. Difference and congruencies between PVA packages: the importance of sex ratio for predictions of extinction risk. Conservation Ecology 4(1):6.

Brook, B.W., et al. 2000. Predictive accuracy of population viability analysis in conservation biology. Nature 404:385-387.

—Robin Meadows

What to Read Next