Cooperative Fish and Wildlife Research Units Program: Tennessee
Education, Research and Technical Assistance for Managing Our Natural Resources


Fitzgerald, D., A. Henderson, K. Maloney, M. Freeman, J. Young, A. Rosenberger, D. Kazyak, and D. Smith. 2021. A Bayesian framework for assessing extinction risk based on ordinal categories of population condition and projected landscape change. Biological Conservation 253: https://doi.org/10.1016/j.biocon.2020.108866

Abstract

Many at-risk species lack standardized surveys across their range or quantitative data capable of detecting demographic trends. As a result, extinction risk assessments often rely on ordinal categories of risk based on explicit criteria or expert elicitation. This study demonstrates a Bayesian approach to assessing extinction risk based on this common data structure, using three freshwater mussel species being considered for listing under the US Endangered Species Act. The probability that a population is classified under each risk category was modeled as a function of projected landscape change using ordered probit regression, assuming observed categories reflect a latent, continuous probability of persistence. All three species were more likely than not (mean probability > 0.5) to be classified as extirpated or low condition throughout their range based on effects of urban development and hydrologic alteration. Spatial variation in estimates revealed strongholds and high-risk areas relevant to conservation decision making. Projected change in probabilities of each risk category based on multiple land-use and climate models was generally small relative to high baseline risk resulting from past landscape changes. Assessing extinction risk based on probabilities of ordinal condition as a function of landscape patterns may provide a flexible and robust approach for many at-risk taxa by adjusting species’ demographic criteria to match relative risk categories, following standardized criteria, or using expert elicitation for data-deficient species. This approach provides decision makers with a useful measure of uncertainty around ordinal classifications and provides a framework for estimating future risk based on projections of anthropogenic stressors.