Incorporating uncertainty into freshwater habitat restoration.

摘要

Recovery efforts for federally listed salmon stocks under the U.S. Endangered Species Act highlight the difficulty of making management decisions in the face of uncertainty. This study investigates the impact of prevalent sources of epistemic uncertainty (process, observation, model selection, parameter and compound) in freshwater habitat models on recovery planning decisions. Instream wood densities were generated under various management scenarios using a meta-analysis model of wood recruitment linked to a forest growth and yield model---Riparian Aquatic Interaction Simulator. Management scenarios also included active restoration in the form of wood placement. The resulting wood densities were linked to predictive models of pool formation and juvenile rearing density for coho salmon, Oncorhynchus kisutch. The short-term and long-term effects of uncertainty were incorporated using Markov Chain Monte Carlo simulation methods within a Bayesian decision framework. The utility of each management scenario was quantified and ranked according to stakeholder preferences. Contrasted with traditional point estimation methods, uncertainty changed the coho density predictions and optimal management strategies in most cases, with the greatest differences observed under model selection and compound uncertainty. The time-scale and stakeholder dependent results also demonstrate the potential influence of uncertainty using a transparent approach. In many cases, the restoration strategies selected given uncertainty were less expensive than those selected without uncertainty. Based on data from Pacific Northwest watersheds west of the Cascade Mountain range, the analytical techniques and decision-making logic used in this methodological study could be easily tailored to site specific conditions and interest groups in other areas.