Population viability assessment of salmonids by using probabilistic networks
Public agencies are being asked to quantitatively assess the impact of land management activities on sensitive populations of salmonids. To aid in these assessments, we developed a Bayesian viability assessment procedure (BayVAM) to help characterize land use risks to salmonids in the Pacific Northwest. This procedure incorporates a hybrid approach to viability analysis that blends qualitative, professional judgment with a quantitative model to provide a generalized assessment of risk and uncertainty. The BayVAM procedure relies on three main components: (1) an assessment survey in which users judge the relative condition of the habitat and estimate survival and reproductive rates for the population in question; (2) a stochastic simulation model that provides a mathematical representation of important demographic and environmental processes; and (3) a probabilistic network that uses the results of the survey to define likely parameter ranges, mimics the stochastic behavior of the simulation model, and produces probability histograms for average population size, minimum population size, and time to extinction. The structure of the probabilistic networks allows partitioning of uncertainty due to ignorance of population parameters from that due to unavoidable environmental variation. Although based on frequency distributions of a formal stochastic model, the probability histograms also can be interpreted as Bayesian probabilities (i.e., the degree of belief about a future event). We argue that the Bayesian interpretation provides a rational framework for approaching viability assessment from a management perspective. The BayVAM procedure offers a promising step toward tools that can be used to generate quantitative risk estimates in a consistent fashion from a mixture of information.