# Identification of infectious agents in early chinook and marine coho salmon associated with cohort survival

Cohort survival was estimated using UC-matched CWT data (as determined by GSI; Fig. 2; Supplementary Material 1, Table S2). CWTs are small, serially etched bands of metal that are implanted in the snout of juvenile salmon (Johnson, 1990). Many studies of marine survival of Pacific salmon populations, particularly those in the Columbia River in northern British Columbia, rely on survival estimates determined by CWT recoveries from hatchery fish. Some of these studies have found considerable heterogeneity in survival estimates over large spatial scales, with nearby populations showing more similar survival patterns (Zimmerman et al. 2015; Ruff et al. 2017). For Chinook Salmon, we matched CUs (sometimes multiple) to CWT survival data based on previously used assignments (Supplementary Material 1, Table S2; DFO 2018; Brown et al. 2020). Since CWTs are used as indicator stocks for management units (as opposed to CUs) for coho salmon, we determined these matches for this species based on the proximity between the locations of origin of the CWTs (usually hatcheries) and CUs (Supplementary Document 1, Table S2).
For chinook and coho salmon, we used previously calculated estimates of smolt-to-adult returns (SAR; Fig. 2) based on CWT data. Since hatchery fish are released into freshwater prior to seaward migration, SAR estimates include survival during freshwater migration, in addition to seaward residence and return spawning migration. However, survival cannot be broken down between these periods. For coho salmon, an updated version of the dataset used by Zimmerman et al. (2015) was acquired as part of the Salish Sea Marine Survival Project (marinesurvivalproject.com/). We acquired SAR estimates for Interior Fraser Coho (IFR, Supplementary Material 1, Table S2) from Arbeider et al. (2020). Estimates for these datasets were based on CWTs recovered from age 3 coho salmon caught in fisheries and returned to hatcheries and spawning grounds. To generate a SAR estimate, CWT recoveries from fisheries and escapements were expanded by the fraction of total catch and escapement sampled, summed and divided by the number of CWT smolts released in the corresponding brood year. . The SAR estimation approaches for both species of salmon assume perfect detection of CWTs in sampled fish and no loss of CWT in surviving fish. For chinook salmon, we used SAR estimates generated by the Pacific Salmon Commission (TSC) Chinook Technical Committee (SAR) Exploitation Rate Analysis (ERA); Chinook Technical Committee 2019, pers. comm. Gayle Brown, CTC), as described in Welch et al. (2021). The SAR estimates for chinook we used were calculated in the same way as the SAR estimates for coho, but include a term for incidental fishing mortality (Welch et al. 2021). We note that Chinook SAR estimates are not part of an official CTC product and are different from the commonly applied measure of early sea survival generated by ERA (also known as survival at age 2( 3)). The CTC’s early estimate of marine survival specifically calculates survival before chinook are caught in fisheries, and we chose to use the SAR estimate instead for consistency with our coho SAR estimates. Additionally, since pathogen-mediated mortality for some agents in our study is not well understood, we sought to avoid limiting the potential period of impact to the early marine phase. Although an important feature of the early marine survival estimate is that it accounts for natural mortality, the SAR estimate we used and the early marine survival estimate are highly correlated (for our dataset, R2= 0.81).
Associations between infectious agent prevalence and cohort survival were estimated as follows:

$No alternative text available$

(1)

where SI is the cohort survival (SAR) for the observation I (an observation is a combination of CWT stock, season and year of ocean entry given a minimum sample size of ≥ 10 fish, to estimate pathogen prevalence; Supplementary Material 1, Tables S3, S4) , PDOI is the Pacific Decadal Oscillation, IAPI is the prevalence of infectious agents, γthere(I) is a random effect representing the variation in survival between years of ocean entry (there), and the index s indicates among CWT stocks the variation in the influence of PDO and IAP on survival (i.e., random intercept and slope). We treated IAPIour main predictor of interest, as a beta random variable to account for sampling error in the observations:

$No alternative text available$

and

$No alternative text available$

are the number of positive and negative screening results, respectively, for salmon from the stock I. Adding 1 to both values ​​implies that this forward beta for IAPI is actually an update of a flat conjugate beta prior B(1,1), with

$No alternative text available$

as the number of binomial successes and

$No alternative text available$

as the number of trials that provide information about the IAPI (Bolker 2008). PDO is a representative variable of large-scale ocean temperature regimes that is often correlated with salmon marine survival (Rupp et al. 2012; Dale et al. 2017; Gosselin et al. 2018). The PDO has been standardized by centering and dividing by two standard deviations (Gelman 2008). We assumed a random slope for the PDO, varying by stock, because the PDO could have very different implications for a stock with marine culture in a relatively cold location (e.g., the west coast of Vancouver Island ) compared to another farm in a warm location (e.g., Strait of Georgia). We assumed a random slope for the prevalence of infectious agents, varying by stock, because evolution of the immune system and variation in resistance to certain pathogens is a known feature of local adaptation in salmonids (Ching 1984 Dionne et al. 2007; Wellband and Heath 2013). Note that for Chinook Salmon, we did not distinguish between oceanic and riverine life history types, but most CWT-CU pairings with adequate sample size (≥10 fish) were dominated by populations oceanic type (i.e. migrated seaward in their first year of life), except Atnarko, Nicola, Kitsumkalum and Stillaguamish (Fig. 1; Supplementary Material 1, Table S2 (all acronyms and names complete can be found here)). For coho, we assumed that all juveniles spent one year in fresh water before their first year at sea. A random interception for the stock was included to account for stock variation in survival (Zimmerman et al. 2015 ; Ruff et al. 2017) and a random intercept for the year of ocean entry was included to account for interannual variation in ocean conditions not represented by other model terms.