Coho salmon (Oncorhynchus kisutch) dispersal and life history variations among Humboldt Bay watersheds

The decline of Coho Salmon (Oncorhynchus kisutch) in California is the result of various anthropogenic effects across the landscape, affecting all stages of their anadromous life history. Monitoring a subset of the remaining populations is essential to evaluate the success of management actions and develop new restoration projects. Defining the appropriate spatial scale for this monitoring and restoration depends on the frequency and extent of dispersal of individuals across watershed boundaries. Coho Salmon life-cycle monitoring projects in California estimate the abundance of juveniles and adults over time in selected focal watersheds. If individuals frequently enter or leave the monitored watersheds for rearing or spawning, these abundance estimates might not accurately reflect the production and survival of individuals in the focal watershed. To address this issue, I assessed movement of Coho Salmon among watersheds along Humboldt Bay, including the life-cycle monitoring population in Freshwater Creek. Using individual tags and mark-recapture multi-state modeling, I quantified the frequency of juvenile and adult movement between Freshwater, Wood, Ryan, and Jacoby Creek over two years of life-cycle monitoring (2017-2019). Wood Creek and Ryan Creek are two connected sub-watersheds that share an estuary with Freshwater Creek, while Jacoby Creek is separated from these other watersheds by Humboldt Bay. Straying of adults among watersheds was rare (only 2 individuals out of 51 tagged adult returns strayed into a stream with potential spawning habitat). Movement of juveniles through the full marine habitat in Humboldt Bay (between Jacoby Creek and the three other streams) occurred, but at low rates (3 fish out of 2492 individuals tagged in 2017 and 5 fish out of 2614 individuals tagged in 2018). Movement of juveniles among Freshwater, Wood, and Ryan Creeks was relatively common (ranged from 250 fish out of 2492 individuals tagged in fall 2017 to 354 fish out of 2614 individuals tagged in fall 2018). I developed a multi-state model structure to estimate the probability of individuals moving among watersheds while accounting for survival and imperfect detection, but parameter estimates from the global model were unreliable due to small sample size and violations of mark-recapture assumptions. A reduced model with fewer parameters provided more reliable estimates. Apparent survival in the second interval of the most parsimonious reduced model was 47.5% in 2017-18 and 29.5% in 2018-19. The reduced model estimated that \u3c0.2% of juvenile fish crossed the bay in both years. However, 17% and 23% of juvenile fish moved between Freshwater, Ryan, and Wood Creeks in 2017-18 and 2018-19 respectively. I also performed a power analysis with simulated data to demonstrate that a greater sample size of fall-tagged individuals would likely not provide more accurate model estimates for transition probability, as many of the transition probabilities are very close to zero. These results suggest the importance of scaling up monitoring efforts to include all connected areas upstream of marine habitats and suggest that it is less essential to monitor adjacent watersheds separated by full marine habitat