Population viability and species interactions: Life outside the single-species vacuum

Population viability analysis (PVA) has become a widely used set of tools for evaluating relative extinction risk and prioritizing management options among imperiled populations. While PVA is a widely sanctioned tool in conservation biology, the field of population viability is in its infancy with respect to species interactions. In this paper, I review available methods for evaluating extinction risk when species interactions contribute significantly to population viability. This review includes an evaluation of six broad categories of species interactions (predation, disease, competition, mutualism, parasitism and host-parasitoid interactions) in population viability analysis, with a particular focus on predation as a case study. I first evaluated how often species interactions are considered when PVA is applied to population data from imperiled species. I identified 378 articles in commonly cited conservation journals, of which 24 attempted a viability analysis for populations threatened by interactions with other species. Most of these PVA's treat a putative species interaction as a constant source of mortality rather than a coupled, dynamic population process. Second, I reviewed the literature to identify the availability of time-series of abundance data for two interacting species in which at least one species was threatened or endangered. Adequate time-series data were available for both species comprising an interacting pair in only 9 out of 407 papers reviewed. Third, I used a stochastic, fully stage-structured predator prey model to create time-series data (vital rates and projection matrices) in order to quantify the efficacy of two matrix-based, single-species PVA approaches. Simple single-species PVAs confound stochastic variation with population cycles induced by species interactions (in this case predation). As a result these models provide conservatively biased forecasts of viability. Unfortunately, the data needed to construct more complex PVA's with feedback and multi-species stochasticity are rarely collected. I close with a discussion of key advances needed to "escape the population vacuum" in a move toward more realistic estimates of extinction risk.