Viability model choice affects projection accuracy and reintroduction decisions


  • Associate Editor: Clinton Moore



Matrix model population viability analyses (PVAs) are commonly used to assess extinction risk, but their accuracy is rarely tested, especially using long-term data sets. Because the few studies that have assessed matrix models indicate that PVAs can have limited predictive power and are sensitive to model choice and assumptions, model choice can affect management decisions. We used 12 years of demographic data (1997–2009) from an introduction of a federally listed, threatened dune thistle, Cirsium pitcheri, to parameterize and test the accuracy of 5 matrix models. Models differed in their method of incorporating environmental stochasticity (matrix selection vs. element selection) as well as their correlation structure of vital rates. We compared 5-year model projections to observed population growth rate, size, and persistence in 2009. We found that all models tended to over-project population growth, size, and stage distributions, but median projections were rarely significantly different from observed values. Projected population growth simulated from the matrix selection model differed significantly from observed, likely because of greater precision compared to large variances in projections produced by element selection models. Incorporating within-year correlations or eliminating correlation structure among vital rates were modeling strategies that correctly predicted the probability of persistence of the restored population. Element selection models incorporating between-year correlations consistently underestimated persistence. Since precision was low for all models, projections should be interpreted qualitatively rather than quantitatively. Differences in PVA model projections can affect management and restoration decisions when estimating reintroduction success and determining the required number of individuals to transplant to obtain a viable population. © 2013 The Wildlife Society.