Multi-model inference for incorporating trophic and climate uncertainty into stock assessments

详细信息    查看全文

• 作者：James Ianelli^a ; ^{Jim.Ianelli@noaa.gov} ; Kirstin K. Holsman^b ; ^{kirstin.holsman@noaa.gov} ; André ; E. Punt^c ; ^{aepunt@uw.edu} ; Kerim Aydin^a ; ^{Kerim.Aydin@noaa.gov}
• 关键词：Model averaging ; Model ensemble ; Multi-species model ; Stock assessment ; Fisheries ; USA ; Alaska Bering Sea ; Groundfish ; Predator prey ; Climate change ; Biological reference points ; Walleye Pollock ; Pacific cod ; Arrowtooth flounder
• 刊名：Deep Sea Research Part II: Topical Studies in Oceanography
• 出版年：2016
• 出版时间：December 2016
• 年：2016
• 卷：134
• 期：Complete
• 页码：379-389
• 全文大小：1680 K
• 卷排序：134

文摘

Ecosystem-based fisheries management (EBFM) approaches allow a broader and more extensive consideration of objectives than is typically possible with conventional single-species approaches. Ecosystem linkages may include trophic interactions and climate change effects on productivity for the relevant species within the system. Presently, models are evolving to include a comprehensive set of fishery and ecosystem information to address these broader management considerations. The increased scope of EBFM approaches is accompanied with a greater number of plausible models to describe the systems. This can lead to harvest recommendations and biological reference points that differ considerably among models. Model selection for projections (and specific catch recommendations) often occurs through a process that tends to adopt familiar, often simpler, models without considering those that incorporate more complex ecosystem information. Multi-model inference provides a framework that resolves this dilemma by providing a means of including information from alternative, often divergent models to inform biological reference points and possible catch consequences. We apply an example of this approach to data for three species of groundfish in the Bering Sea: walleye pollock, Pacific cod, and arrowtooth flounder using three models: 1) an age-structured “conventional” single-species model, 2) an age-structured single-species model with temperature-specific weight at age, and 3) a temperature-specific multi-species stock assessment model. The latter two approaches also include consideration of alternative future climate scenarios, adding another dimension to evaluate model projection uncertainty. We show how Bayesian model-averaging methods can be used to incorporate such trophic and climate information to broaden single-species stock assessments by using an EBFM approach that may better characterize uncertainty.