The amount of time that juvenile salmon remain in an estuary varies among and within populations, with some individuals passing through their estuary in hours while others remain in the estuary for several months. Underlying differences in individual physiological condition, such as body size, stored energy, and osmoregulatory function, could drive individual variation in the selection of estuary habitat. Here we investigated the role of variation in physiological condition on the selection of estuarine and ocean habitat by sockeye salmon (Oncorhynchus nerka) smolts intercepted at the initiation of their 650 km downstream migration from Chilko Lake, Fraser River, British Columbia (B.C.). Behavioural salinity preference experiments were conducted on unfed smolts held in fresh water at three time intervals during their downstream migration period, representing the stage of migration at lake exit, and the expected timing for estuary entry, and ocean entry (0, 1, and 3 weeks after lake exit, respectively).

The focal population is sockeye salmon from Chilko Lake located 650 river km north of the Fraser River estuary at an elevation of 1174 m, in the interior of British Columbia, Canada (Figure 1). Out-migrating yearling sockeye salmon smolts were collected from Chilko Lake between 22:00 and 4:00 April 30 – May 1, 2019, during the peak outmigration period (Macdonald et al., 2020). Smolts were collected by dip net at a smolt fence and transferred to a riverside transport tank with continual flow-through of aerated river water. In total, 263 smolts were collected for salinity preference experiments. All fish collection, holding, and experimentation protocols were reviewed and approved by the Animal Care Committee at Simon Fraser University (Protocol number 1238B-17).

Behavioural experiments were conducted on smolts at three time intervals during their downstream migration period, representing their lake-exit, estuary-entry, and ocean-entry outmigration stages (0, 1, and 3 weeks after lake exit, respectively; see Clark et al., 2016; Stevenson et al. 2019). Ocean-entry here is defined as near full-strength sea water that is beyond the freshwater influence of the Fraser River. To visualize the salinity gradient that migrating smolts are exposed to in Figure 1, monthly average sea surface (0.5m depth) salinity values were obtained from the SalishSeaCast NEMO model (Soontiens et al, 2016; Soontiens and Allen, 2017) for May 2019.

 A random subset of smolts (n = 45) was selected to undergo lakeside salinity preference experiments to measure salinity preference, energetic reserves, and physiology immediately following outmigration (hereafter, lake-exit). The remaining smolts (n = 218) were transported to the ALCAN research facility at Simon Fraser University, Burnaby, B.C., and housed in freshwater and natural photoperiod for the duration of the experimental outmigration period (4 weeks following lake-exit, 2 – 29 May, 2019). Smolts were randomly placed in one of four 80-gallon holding tanks, each with a maximum density of approximately 15 g m^-3. Water quality in holding tanks was maintained at 8.4˚C (CI₉₅[8.25, 8.48]), 98.6 % dissolved oxygen (CI₉₅[98.04, 99.10]), and <0.25 ppm ammonia. Water temperatures were within the range of downstream migration conditions (Macdonald et al., 2020) and were low enough to prevent desmoltification (<10 ˚C, Stefansson et al., 1998).

To reflect observations of wild migrating smolts, as well as to maximize the contrast of energetic condition throughout migration, wild-caught smolts were not fed during the 4-week holding period in fresh water. Smolts were also above the critical energy threshold that limits swim performance (Hvas et al., 2021; Wilson et al. 2021), suggesting the ability to actively search for different salinities was not impaired. The decision to not feed is supported by the observation that the majority of migrating sockeye smolts caught near the mouth of the Fraser River in Mission, BC have empty stomachs (Patterson, unpublished data, personal communication), indicating that feeding is limited during the freshwater migration stage. A variable proportion (20-40% from 2004 – 2009, 29% in 2014) of juvenile sockeye collected in marine trawls in the Strait of Georgia have empty stomachs (Beamish et al., 2012; Neville et al., 2016) and show gene expression patterns that suggest recent fasting (Houde et al., 2019).

Experimental Design

We performed salinity preference experiments to infer habitat choice among migrating smolts. The design of the experimental preference tank was modified from previous studies to test the preference for three salinities simultaneously within each migratory stage (Baggerman, 1960; McInerney, 1963; Price & Schreck, 2003b). Three 20 L glass aquaria were placed side-by-side within a 120 L aquarium (Figure 2), dividing the larger aquarium into three chambers of equal size, while allowing for a 2-inch gap above the dividing walls for fish to cross between chambers. The outside and bottom of the large aquarium were darkened with black plastic to minimize disturbance to fish during the experiment. The walls between chambers were left transparent to allow smolts to view conspecifics through the dividing walls. Smolts were tested in groups of six to allow for more natural schooling behavior of sockeye salmon along the horizontal plane (Katzman et al., 2010, Tierney et al., 2009; Webster & Dill, 2007), regardless of individual salinity choice. To distinguish individual smolts during the preference test, smolts were anesthetized and marked with coloured elastomer (Northwest Marine Technology ©). Tagging occurred the day before a trial to allow for a minimum of 18 hours for recovery. The tank was illuminated uniformly from above using an LED light and surrounded by shade cloth to prevent visual disturbance. A video camera (GoPro Hero 3 ©) was positioned facing the long side of the aquarium to record fish position within all three chambers throughout the experiment. In addition, an experimenter recorded fish position (chamber and depth) and behaviour every ten minutes through viewing slots in the shade cloth.

To start the salinity preference trial, all three chambers were filled with fresh, aerated water (0.0 – 0.2 ppm, 8.0 – 9.7 ˚C, 97.2 – 101.6 % DO, n_trials = 42, Figure 2A). Six smolts were randomly selected from holding tanks and transferred to the experimental aquarium, ensuring equally dispersed, randomized placement throughout the three chambers. The smolts were allowed to acclimate to the aquarium and explore the three chambers in fresh water for 1 hour. After the 1-hour acclimation period, a horizontal salinity gradient was imposed in the experimental tank by displacing the water of each chamber with either fresh (<2.0 ppt), brackish (14–16 ppt), or salt water (30–32 ppt) (Figure 2B shows one of six possible orientations of the salinity gradient). The order of salinity of each chamber was randomized between each trial to control for any effect of chamber location or order on fish behaviour or movement. Each chamber was filled by separate water reservoirs that fed into the bottom of each chamber over approximately 5 minutes. Instant Ocean (Instant Ocean ©, Spectrum Brands 3001 Commerce St. Blacksburg, VA 24060-6671) was added to filtered, aerated water to make brackish and salt water, and each was added to the respective reservoirs. The freshwater reservoir was filled with filtered, aerated water. The higher-density salt and brackish water displaced the lower-density freshwater upwards, forming a halocline; the upper freshwater layer served as a freshwater bridge in which the fish could move freely between chambers. Displaced fresh water was simultaneously drained from the aquarium using a gravity filtration system. Preliminary trials showed that the halocline was stable for over 2 hours. Salinity probes (Marine Salinity Waterproof Tester ©, HI98319, HANNA Instruments) were fixed within each chamber to monitor salinity and temperature throughout the duration of the experiments. Once the salinity gradient was imposed, fish position (chamber and depth) was recorded every ten minutes for one hour. If a fish was in the freshwater layer above the halocline of a chamber, it was recorded as being in fresh water. In total, 44 experimental preference trials were undertaken. Eleven out of the 263 smolts tested were removed from behavioural analysis due to disturbances during experimental trials (e.g. technical failure of salinity gradient).

Physiological Sampling

Following preference trials, smolts were euthanized in tricaine methanesulfonate at a concentration of 400 mg l⁻¹ (buffered with 800 mg l⁻¹ NaHCO₃) until fish operculum beating ceased and reflex response was absent (Neiffer & Stamper, 2009). Immediately after euthanization, gill tissue samples were clipped from the right gill arch and frozen on liquid nitrogen in SEI buffer (250 mM sucrose, 20 mM Na₂EDTA, 50 mM Imidazole, pH 7.3) to preserve enzyme structure and function. Smolt carcasses were weighed and fork lengths measured before they were stored in whirlpacks and frozen on dry ice. Gill samples and carcasses were transported frozen on dry ice to the laboratory and stored at -80˚C until analysis.

Gill NKA activity is a common metric to estimate saltwater preparedness in juvenile salmonids, where higher gill NKA activity indicates higher saltwater preparedness (McCormick, 2013). Gill NKA activity was measured following the enzymatic assay protocol from McCormick (1993; see Suppl. Mat.). Eight gill samples out of 263 produced high variation among replicates (CV > 20%) and were removed from the analysis. To estimate the energetic status of individual smolts, lipids, water and ash were isolated and measured as proximate constituents following the Bligh and Dyer chloroform extraction method (Bligh & Dyer, 1959; Trudel et al., 2005; Wilson et al., 2021; see Suppl. Mat.). The energy density (ED) of each smolt was calculated from lipid (f) and protein (p) measurements using the following equation (Breck, 2008):

where ED is expressed in units kJ/g wet weight, f is the fraction of lipid measured per smolt (g/g wet weight), D_f is the energy density of lipids reported for coho salmon (O. kisutch, 39.54 kJ g^-1) (Brett & Groves, 1979; Crossin et al., 2004; Higgs, 1979), p is the fraction of protein estimated per smolt (g g^-1 wet weight), and D_p is the energy density of protein reported for coho salmon (23.64 kJ g^-1) (Crossin et al., 2004; Higgs, 1979).

Triglycerides (TAG) are the major energy storage form of lipids in fish and have high ecological and physiological relevance as indicators of growth potential (Hakanson et al., 1994; Lochmann et al., 1995). A colorimetric method was used to determine percent triglycerides for each smolt (McGowan et al., 1983; Weber et al., 2003; see Suppl. Mat.). The percent ratio of TAG to lipid represents the density of TAG within lipid stores in individual fish. All replicate estimates of TAG density were within 20% CV. Measurements of TAG density above 100% were removed from analysis (n = 2).