Assessment Strategies for Data-Limited Chinook Salmon Stocks of Western Alaska

Abstract

Management strategies for Alaskan Pacific salmon species are conducted in the face of substantial uncertainty, particularly in large drainage systems. In addition to the interacting sources of partial observability and environmental stochasticity, substantial uncertainty exists around the appropriate way to structure an assessment. Assessment approaches vary in their statistical and biological complexity, which inherently leads to trade-offs and questions about the optimal assessment approach. In this thesis, I develop and investigate a range of assessment approaches that address multiple sources of uncertainty by casting them in a Bayesian state-space modeling framework. Chapter 1 presents an introduction to the topics of Pacific salmon management in Alaska, assessment strategies, and uncertainty to provide the necessary background for the following chapters. Chapter 2 investigates trade-offs of estimating abundance and the characteristics of the population dynamics in an integrated analysis as opposed to a sequential analysis. Chapter 3 investigates potential management implications of declining trends in age- and size-at-maturity that have been widely observed but are typically not addressed in assessment models. Chapter 4 revisits a previously developed habitat-based predictive model for stocks that lack adequate assessment data by developing a hierarchical framework and applying a suite of variable selection techniques to decide which habitat variables have predictive credibility. Themes and topics that are addressed continuously in this thesis include spawner-recruit analysis, measurement error, and methods to incorporate multiple sources of uncertainty so that their implications can be more fully addressed.