Using Integrated Models to Improve Management of Imperiled Salmon and Steelhead

Abstract

Exploited wild animals include some of the most charismatic and celebrated species globally. They feed civilizations, are foundational to economies, are the subject of socially and culturally important recreational and ceremonial harvests and provide ecosystem services beyond human use. Yet, despite their importance to humanity, the history of management of these species is checkered, demonstrating both successes involving stable populations and long-term sustained harvests, and failures including overexploited species where population collapses and even extinction have occurred. Sustainable exploitation depends entirely on the fundamental principle of compensatory recruitment, wherein population growth rates approach zero when populations are close to their carrying capacities, but as abundance declines due to harvesting, per capita growth rates increase in response to decreasing competition, thereby enabling populations to recover toward their carrying capacities. However, to ensure harvest is implemented sustainably, answers to a few fundamental questions are required: 1) how many animals does a population contain? How many are being harvested? and, 3) What is the relationship between adult abundance and recruitment? Although natural resource managers have attempted to answer these questions for decades, the need for improved answers is greater than ever for species like Pacific Salmon (Oncorhynchus spp.) in the continental United States, where most populations are listed under the Endangered Species Act and face elevated risk of extinction yet remain subject to exploitation. This dissertation uses modern statistical modeling to address each of the fundamental questions necessary for the sustainable management of exploited but imperiled Pacific Salmon in Washington State. Chapter 1 estimates the relationship between spawner abundance and recruitment to the smolt life stage for coastal steelhead (O. mykiss) populations in Washington State. By re-parameterizing the Beverton-Holt and hockey-stick stock-recruitment functions, the study incorporated habitat characteristics to explain variation in biological reference points among populations. The findings indicated strong effects of habitat quantity and quality on smolt carrying capacity, with estimates of capacity and the spawner abundance producing the maximum sustained yield (SMSY) aligning with previous studies but demonstrating previously underappreciated sensitivity to smolt and kelt survival. The study highlighted the importance of geomorphic features in informing estimates of smolt capacity in stream-rearing salmonids and identified a minimum smolt-to-adult return rate required for population viability at around two percent based on maximum per-capita smolt productivities around 50 in the least productive populations. Chapter 2 focuses on estimating the abundance of coho salmon (O. kisutch) for Washington State’s populations in the Lower Columbia Evolutionarily Significant Unit (ESU). The study used data from an ESU-wide monitoring program initiated in 2010, which included spawning ground surveys, mark-recapture experiments, and trap-and-haul counts of migrating coho at dams, complemented by fishery catch estimation programs. The multivariate state-space integrated population model that was developed provided estimates of coho salmon population parameters, revealing that coho salmon abundance ranged between 12-86,000 wild spawners (median = 28,310) from 2010-2022. Several populations periodically exceeded their ESA recovery goals but remained below densities seen in healthy populations elsewhere. Chapter 3 develops a Bayesian multivariate state-space model to improve estimates of catch in sport fisheries using creel survey and effort count data. This model, applied to data from the 2021 Skagit River wild winter steelhead catch-and-release sport fishery, robustly quantified uncertainty and was developed to flexibly accommodate multiple study designs. The model estimates of catch were used within a probabilistic management framework, helping fishery managers balance the risks of overfishing and unnecessarily restricting fishing opportunities. This approach demonstrated how robust quantification of uncertainty could optimize study designs for desired precision and cost efficiency.