Predicting stream attractiveness to stray hatchery-origin chum salmon to aid in understanding salmon dispersal and informing hatchery management

Abstract

Understanding the processes underlying dispersal propensity in animal populations is a fundamental goal of ecologists. In metapopulations of wild Pacific (Oncorhynchus spp) and Atlantic salmon (Salmo salar, hereafter collectively referred to as "salmon"), it is recognized that dispersal, or straying, exists in tandem with philopatry and provides benefits such as gene flow and colonization of new habitat. However, straying by hatchery-produced salmon into streams can negatively affect the genetic integrity and reproductive success of wild salmon populations. Straying by hatchery-origin salmon may also confound fishery management procedures around assessing wild spawner escapement, given the difficulty in identifying hatchery salmon in the field. A first step in mitigating and managing the consequences of straying by hatchery salmon is to understand where and why hatchery salmon stray. In this study, I described the relationship between the number of hatchery-origin strays received by streams and the characteristics of those streams based on the hypothesis that certain characteristics are attractive to hatchery strays. An extensive dataset documenting the number of stray hatchery-origin chum salmon (Oncorhynchus keta) that spawned in 57 streams in Southeast Alaska was produced from hundreds of field surveys conducted over a 10-year period 2008-2019. I used these data in a generalized linear mixed effects modeling framework to predict how "attractive" a given stream was to hatchery strays based on hypothesized influential stream characteristics, such as streamflow, distance and numbers of hatchery releases, and conspecific density. I found that some streams were more attractive than others to hatchery strays: 10 of 57 streams surveyed had mean observed attractiveness indices of 39 recipient strays over time (range: 12-115) in a given survey, while the remaining 47 sites only attracted two recipient strays on average (range: 0-8). Furthermore, stream attractiveness to hatchery strays was predicted to increase by 44% with a 1-SD (27.6 million) increase in the number of hatchery-origin chum salmon released near the stream and increase non-linearly with elevated levels of intra-annual variability (CV > 0.55) of stream discharge. These results corroborate results from other studies that distance to a source population (e.g., a hatchery release site) influences the number of dispersing immigrants, or strays, received by the stream. However, additional ecological factors such as streamflow also affect the distribution of hatchery strays, indicating that inclusion of distance is necessary but not sufficient for accurate prediction. In the second part of this study, I expanded predictions to 558 additional streams throughout Southeast Alaska in 2008-2019 and in a hypothetical future year given increased hatchery releases. Only a small subset of streams (~10%) was predicted to be attractive, with mean predicted attractiveness indices of 57 recipient hatchery strays (range: 9-600). Bootstrapped coefficients of variation described uncertainty around predictions. Uncertainty was modest for predictions for streams in 2008-2019 (CV range: 0.21-0.62) but high for predictions of hypothetical future stream attractiveness (CV range: 0.70-1.15). These results suggest that the predictive modeling framework may be useful in describing patterns of stream attractiveness beyond the spatial range of the observed data, but not beyond its temporal range. Taken together, the results of this study elucidate the role of stream ecology and spatial location in attracting dispersing hatchery-origin salmon and provide insight into how predictions of stream attractiveness may be incorporated into hatchery management.