Introduction

The occurrence of ontogenetic, or size-related, shifts in diet or habitat are prevalent in nature and important in shaping species interactions and community structure (Werner and Gilliam 1984). Pacific salmon (Oncorhynchus spp.) populations undergo ontogenetic shifts in both diet and habitat as they migrate from freshwater to saltwater, and research suggests that year class strength is determined during this critical period (Pearcy 1992; Beamish et al. 2004; Pearcy and McKinnell 2007). A suite of environmental factors during early marine residence (e.g., physical conditions, plankton and predator abundance) have recently been shown to correlate with salmon survival (Burke et al. 2013; Miller et al. 2013, 2014), but identifying specific mechanisms that account for juvenile salmon mortality in the ocean remains a challenge. Because predation pressure and selective mortality may be higher in smaller or slower growing fish, rapid growth may provide a survival advantage during ontogeny (Zabel and Williams 2002; Claiborne et al. 2011; Duffy and Beauchamp 2011). Rapid growth may reduce the potential pool of competitors, predators, and pathogens, as well as provide sufficient energetic stores needed to survive the first winter at sea (Beamish and Mahnken 2001).

Across several taxa of fish that become piscivorous during ontogeny, many experience dramatic increases in growth following the onset of piscivory (i.e. when fish become mostly piscivorous; see reviews by Juanes 1994; Mittelbach and Persson 1998). Variations in the timing of the onset of piscivory are related to predator size, prey availability, and environmental variability (Juanes et al. 2002; Hansen et al. 2013). Bigger juveniles have larger gape widths, more developed jaw structures, larger reaction distances, better visual acuity, and swim faster than smaller conspecifics. Collectively, these attributes allow fish of a certain size expand their trophic niche. Predator and prey phenologies may also determine the onset of piscivory (Juanes 1994). Predators and prey respond differently to environmental cues across species and life stages. For example, anomalously high temperatures that shift the timing of predator migration (Anderson et al. 2013) may lead to temporal or spatial mismatches between predator and prey (Cushing 1972) and advance or delay the onset of piscivory.

For juvenile Chinook Salmon (Oncorhynchus tshawytscha), whose diets during early marine residence have been extensively studied (Peterson et al. 1982; Emmett et al. 1986; Brodeur et al. 2007), there is clear support for an ontogenetic shift from invertebrate and terrestrial insect prey to piscivory of larval and young-of-the-year (YOY) marine fishes (Brodeur 1991; Daly et al. 2009; Duffy et al. 2010). Despite this, quantitative estimates of potential prey fields, relationships between prey community and environmental variables, information on the timing and size at the onset of piscivory, and the effects of prey quality on salmon growth remain understudied (but see Schabetsberger et al. 2003; Brodeur et al. 2011; Wells et al. 2012). Juvenile salmon size and survival are related to climate across broad scales (Beamish and Bouillon 1993; Mantua et al. 1997) and expected to increase when conditions during the first few months of ocean residence are favorable (i.e. during cool, productive, upwelling conditions) and when lipid-rich prey are plentiful (Peterson et al. 2014). Therefore the onset of piscivory might occur earlier when ocean conditions are favorable.

Tracking ontogenetic shifts in diet from salmon stomach contents alone has its inherent limitations. While providing considerable taxonomic resolution, observations are temporally limited to the most recent meal, and interpretations based on quantitative measures can be biased by differences in digestion rates (Rindorf and Lewy 2011). A more informative approach combines stomach content analysis with chemical analyses of trophic biomarkers from lipids (fatty acids) and stable isotopes. Fatty acids and bulk stable isotopes of carbon and nitrogen (δ13C and δ15N) measured in the muscle tissue of consumers reflect diet integrated over weeks to months (Fry 2006; Copeman et al. 2016; Vander Zanden et al. 2015). While carbon isotopes and some fatty acid biomarkers generally reflect sources of primary production (Peterson and Fry 1987; Budge and Parrish 1998; Parrish 2013), nitrogen isotopes and other fatty acid biomarkers vary with nutrient source and consumer trophic position (Post 2002; El-Sabaawi et al. 2009; Daly et al. 2010). When combined, stomach content and trophic biomarker analyses provide a more robust method for recent (weeks to months) diet reconstruction than either method alone. The goal of our study was to identify and evaluate the timing and size at the onset of piscivory in a population of juvenile Chinook Salmon using a combination of field sampling of potential prey with measurements of salmon size, growth, stomach contents, stable isotopes, and fatty acids. We defined the onset of piscivory as the timing or size when salmon consumed more fish prey by wet weight than any other prey category.

Our study was divided into three parts. First, we characterized seasonal and annual variations in the salmon prey field by sampling potential prey over two years and identifying environmental variables associated with prey community composition. Next, we developed metrics to account for variation in juvenile salmon size, growth, and body condition. For this analysis, we selected an abundant stock group of upper Columbia summer-fall Chinook Salmon (UCSF) that was repeatedly sampled through time and identified using genetics. Subyearlings from the UCSF stock group have been detected exiting the Columbia River from May through November, with abundance peaking in July (Weitkamp et al. 2015). Because UCSF subyearlings remain concentrated nearshore along the Oregon and Washington coasts during their first few months at sea (Fisher et al. 2014; Teel et al. 2015) this stock group is ideal for a longitudinal foraging study. Lastly, we evaluated ontogenetic changes in salmon diet using direct observations (stomach content analysis), stable isotopes, and fatty acids. In response to the ontogenetic shift in habitat from freshwater to saltwater, we expected changes in salmon carbon stable isotopes and phytoplankton fatty acid biomarkers, indicating changes in the carbon pool at the base of the food web. We hypothesized that the onset of piscivory would be related to the availability of marine fish prey and that piscivorous salmon would be larger, in better condition, and grow faster than salmon feeding on invertebrates. Compared to invertebrate feeders, we also expected that piscivores would have higher stable nitrogen isotope and fatty acid marker values related to trophic position but that there would be a lag of at least a month between consumption of fish prey and expression of diet in consumer tissues (Copeman et al. 2013; Heady and Moore 2013; Vander Zanden et al. 2015). This is the first study to quantify seasonality in the salmon prey field using a trawling method designed to sample micronekton upon which juvenile salmon feed. It is also the first study to present information on ontogenetic shifts in juvenile Chinook Salmon diets using an integrative approach that combines stomach content, stable isotopes and fatty acid analyses.

Methods

Field collections

We sampled two stations located off Willapa Bay, Washington, USA (46° 40′ N) a total of 52 times during monthly cruises from late-May through late-September 2011 and 2012 (Fig. 1). Each station was sampled 2–4 times per cruise. The first station was located 9 nautical miles (nm) offshore (16.7 km) and the second station was located 14 nm (26.0 km) offshore. Sampling occurred during daylight hours when juvenile salmon feed (Brodeur et al. 2011). There was no cruise in June 2012 due to difficulties chartering a vessel. We collected animals used in this study under Scientific Research Permit 1410 issued to the Northwest Fisheries Science Center under the authority of Section 10(a)(1)(A) of the Endangered Species Act, Scientific Taking Permit 17,203 issued by Oregon Department of Fish and Wildlife, and Washington State Scientific Collection Permit 12–128.

Fig. 1
figure 1

Map of study area showing the two stations repeatedly sampled on the continental shelf off of Willapa Bay, Washington from May through September in 2011 and 2012

Full size image

We fished for salmon and their potential prey by deploying a 264 Nordic rope trawl (NET Systems Bainbridge Island, Washington) for 15 min at an average speed of 5.7 km hr.−1 from the chartered commercial fishing vessel F/V Miss Sue. The trawl had variable mesh sizes (162.6 cm at the mouth to 8.9 cm at the cod end), with a 6.1 m long, 0.3 cm knotless liner sewn into the cod end. This gear has been used to successfully sample micronekton at mid-water (30 m) depths (Phillips et al. 2009) and was the type of gear recommended by Brodeur et al. (2011) to best sample all potential salmon prey types. We recorded GPS locations at the start and end of each haul and estimated volume swept (m3) by multiplying distance trawled (m) by the mouth area of the net (336 m2). Hydrographic information was collected to within 5 m from the bottom by deploying a Seabird SBE 25 conductivity, temperature, and depth (CTD) profiler at each station. The CTD recorded water temperature (°C), salinity (psu), density (kg m−3), turbidity (mg m−3), fluorescence (mg m−3), dissolved oxygen concentration (ml l−1), and dissolved oxygen saturation (%).

Analysis of the prey field

All potential prey items collected in the trawl were sorted by size (<80 mm total length; TL and <3.0 g), frozen, and transported back to the laboratory. Up to 30 individuals per taxon and station were measured (nearest mm) and weighed (nearest 0.01 g). Prey field biomass estimates were calculated for each prey type and station by dividing the total mass of the prey by the volume of water sampled by the trawl net, standardized to μg m−3 by multiplying by 1000, and averaged by cruise (Appendix).

Size at and timing of ocean entry varies considerably for populations of salmon originating from the Columbia River basin (Weitkamp et al. 2015). Development of the prey field is also highly variable and dependent on environmental conditions, thus contributing to potential variability in the prey field juvenile salmon first encounter. We tested for seasonal variation in the salmon prey field by conducting ecological analyses on the community structure of catch data. Data were evaluated using nonparametric multi-response permutation procedures (Mielke and Berry 2001) and nonmetric multidimensional scaling (NMS; Kruskal 1964) with the Sørensen distance measure in the statistical package PC-ORD v. 6.07 (MjM Software Design). Because we were interested in seasonality of the prey field, we used nonparametric multi-response permutation tests to test the hypothesis of no difference in prey field community by month across years. We used NMS to ordinate sample units (biomass by haul) in species space to identify sample unit clusters with similar prey field community compositions. Biomass estimates were normalized using a generalized logarithmic transformation [log(x + x min ) – log(x min )] to reduce bias from overly abundant species and to account for zero truncation in the data (McCune and Grace 2002).

In NMS, the most dissimilar samples are farthest apart and the most similar samples closest together. To measure the success of the prey community ordination, we calculated a stress value from 250 runs of real data starting from a random configuration (Mather 1976). Monte Carlo simulations were conducted with an additional 250 runs of randomized data, which were then compared to the real data. Statistical significance (ɑ = 0.05) was calculated as the proportion of randomized runs with stress less than or equal to the observed stress.

Indicators of the physical ocean environment such as temperature and salinity are related to salmon recruitment (Peterson et al. 2014). We hypothesized that coastal ocean environmental variables would also be related to the salmon prey community so we evaluated associations between prey community NMS ordinations and the physical environment with correlation analyses using in situ measurements of 1 m sea surface temperature (SST), salinity, turbidity, fluorescence (a proxy for chlorophyll-a concentration), and dissolved oxygen concentration collected by the CTD. We also selected two of the regional physical variables that positively correlate with Chinook Salmon adult returns (Burke et al. 2013; Miller et al. 2013), Columbia River plume volume (www.stccmop.org) and coastal upwelling (www.pfeg.noaa.gov), for inclusion in the analysis. We obtained estimates of the three-dimensional volume (km3) of the Columbia River plume (salinity cutoff 28 PSU) using simulation database DB31 of the “Virtual Columbia River” modeling system generated by the Center for Coastal Margin Observation and Prediction (Zhang and Baptista 2008). For upwelling, we used indices from daily averages of offshore Ekman transport driven by geostrophic wind stress between 45°N and 48°N and 125°W.

Analysis of juvenile salmon

We transported juvenile salmon frozen at sea back to the laboratory where they were identified to species, measured for fork length (FL; nearest mm), and weighed (nearest 0.1 g). To estimate genetic stock of origin, we extracted fin clips from 288 juvenile Chinook Salmon and genotyped them at 13 microsatellite DNA loci following Teel et al. (2015). Salmon were assigned to stock groups using a standardized genetic database (Seeb et al. 2007), the likelihood model of Rannala and Mountain (1997), and the program ONCOR (Kalinowski et al. 2007). Based on genetic information, we assigned 61% (n = 175) of juvenile Chinook Salmon to the UCSF genetic stock group with mean ± standard error (SE) probability of assignment of 89.1 ± 1.0%. Our analysis focused on subyearlings (0-age; 96% of UCSF fish), which we classified by length (Weitkamp et al. 2015): ≤120 mm FL in May (0% of catch), ≤140 mm FL in June (1% of catch), ≤180 mm FL in July (49% of catch), ≤210 mm FL in August (21% of catch), and ≤250 mm FL in September (25% of catch). Of these fish, 40% had an adipose fin clip, passive integrated transponder (PIT), or coded wire tag (CWT), indicating likely hatchery origin. The UCSF stock group is composed of fish originating in main stem and tributary sources east of the Cascade Mountains, although both hatchery and natural production of the stock also occurs in the mid-Columbia River (Miller et al. 2013; Teel et al. 2014).

To evaluate salmon size, we compared FL and mass by month between 2011 and 2012 using Student’s t-tests. We developed a condition index for UCSF subyearlings using residuals from the linear regression of ln-transformed FL and mass (r2 = 0.99, p < 0.001; Jakob et al. 1996; Brodeur et al. 2004), and compared condition by month within years using analysis of variance (ANOVA) followed by Tukey’s Honestly Significant Difference (HSD). We increased the sample size of UCSF subyearlings used for size, growth, and condition calculations from 168 to 518 by including salmon from two other studies that used the same genetic and life history identification methods. The first study sampled UCSF subyearlings (n = 201) as they exited the Columbia River (Weitkamp et al. 2015). The second study sampled salmon (n = 149) in coastal waters from central Oregon to northern Washington (Teel et al. 2015). We included only individuals that were collected May through September in 2011 and 2012 from the mouth of the Columbia River north along the shelf to Willapa Bay. Inclusion of these additional samples did not change our results but increased statistical power to detect differences in size, growth, and condition by sampling period.

Monthly estimates of early ocean growth (G L in mm d−1) in UCSF subyearlings were calculated from differences in FL between ocean and estuary-caught individuals, assuming a uniform size and time of entry for each month:

$$ {G}_L=\frac{\left({L}_o-{L}_e\right)}{\left({t}_o-{t}_e\right)}, $$

where L o is individual FL at capture in the ocean at time t o and L e is the mean monthly FL at capture in the estuary the prior month at mean time t e . We also estimated specific growth rates (SGR in% body weight [BW] d−1) from changes in salmon weight as:

$$ SGR=\frac{\left( \ln \left[{W}_o\right]- \ln \left[{W}_e\right]\right)}{\left({t}_o-{t}_e\right)}\times 100, $$

where W 0 is individual mass at capture in the ocean at time t o and W e is the mean mass at capture in the estuary the prior month at mean time t e . Even though we did not know which month the fish entered the ocean, we considered growth estimates to be valid based on evidence that abundances of UCSF subyearlings exiting the estuary are normally distributed around June and August (Weitkamp et al. 2015) and observations that average estuary residence times for this stock are low (<1 week, Claiborne et al. 2014). We compared monthly estimates of G L and SGR between years using t-tests.

Dietary analysis and comparison with the prey field

An integrative approach for assessing salmon foraging ecology includes stomach content analysis conducted alongside salmon stable isotope and fatty acid analyses. We identified stomach contents from a subsample of 229 UCSF subyearlings collected in the lower estuary and ocean from June through September in both years. Prey items were identified to the lowest possible taxa under a dissecting scope using methods described by Brodeur et al. (2007). To quantify stomach contents, we weighed the entire stomach contents and individual prey items (nearest 0.001 g), enumerated all of the prey, and measured the total length of up to 10 prey per taxon per stomach (nearest mm).

To standardize stomach fullness, we calculated mean stomach fullness as a percent of total body weight:

$$ Stomach\ Fullness\ \left(\%\right)=\frac{stomach\ content\ weight\ (g)}{total\ fish\ weight(g)-\mathrm{stomach}\ \mathrm{content}\ \mathrm{weight}(g)}\times 100. $$

We defined stomachs as empty when fullness <0.05% according to Weitkamp and Sturdevant (2008). Stomach fullness was compared by month within years using ANOVA followed by Tukey HSD tests. To visually represent diet composition, prey taxa were grouped into 14 categories that contributed ≥2% of salmon diets by weight: nonfood (e.g., Plantae), insects (Insecta), pteropods (Pteropoda), cladocerans (Cladocera), ostracods (Ostracoda), copepods (Copepoda), isopods (Isopoda), amphipods (Amphipoda), mysids (Mysidacea), krill (Euphausiidae), shrimp larvae (Pandalidae), crab larvae (Metacarcinus magister and Cancer productus zoea and megalopae), YOY Northern Anchovy (Engraulis mordax; hereafter referred to as anchovy), and unidentified fish (Osteichthyes). Average size at the onset of piscivory was compared between years using t-tests.

Stable isotope and lipid analysis

To address ontogenetic changes in diet, we subsampled (n = 29, 3–6 fish per month) UCSF subyearlings for bulk stable isotopes of nitrogen and carbon, total lipids, and fatty acids. For isotopes, we sampled dorsal muscle tissue with skin removed from the right side just under the dorsal fin (average = 0.32 g wet weight). For lipids, we sampled dorsal muscle from the left side (average = 0.54 g wet weight). In addition, we processed a representative sample of Dungeness crab megalopae collected from May through July in each year (n = 5–6 per year) for nitrogen stable isotopes. This was done to establish a trophic baseline to estimate salmon trophic position as megalopae are primary consumers and the salmon prey item with the lowest trophic position as determined by Miller et al. (2010). We also measured fatty acids from a subsample of invertebrate prey (n = 32) and fish prey (n = 144) collected from May through September in each year to develop a piscivory biomarker based on fatty acid differences between these two main prey types.

We processed bulk stable isotopes of carbon and nitrogen at Oregon State University (College of Earth, Ocean, and Atmospheric Sciences Stable Isotope Laboratory). Dried tissue was ground into a fine powder and 1.0 ± 0.1 mg of powder packed into a tin capsule. Prepared samples and international standards (USGS40, ANU Sucrose, and IAEA-N2) were combusted at >1000 °C using a Carlo Erba NA1500 elemental analyzer. The resulting CO2 and N2 were measured by continuous-flow mass spectrometry using a DeltaPlus isotope ratio mass spectrometer. Stable isotopes of carbon and nitrogen (δ13C and δ15N) were expressed in the delta notation:

$$ \mathrm{Stable}\kern0.30em \mathrm{Isotopes}=\left(\frac{R_{\mathrm{sample}}}{R_{\mathrm{standard}}}\kern0.30em -1\right)\ast 1000, $$

where R is 15N:14N or 13C:12C. Instrument error was ±0.1‰ for carbon and ±0.2‰ for nitrogen. Because no samples had atomic C:N ratios >3.5, they were not lipid-corrected as suggested by Post et al. (2007).

We converted all δ15N values to trophic position (TP) using the notation of Post (2002):

$$ TP=\uplambda +\frac{\delta^{15}{N}_{salmon}-{\delta}^{15}{N}_{base}}{\varDelta_n}, $$

where λ is the trophic position of Dungeness crab megalopae used to establish the δ15Nbase relative to δ15Nsalmon, and Δn (Δδ15N) is the enrichment in δ15N per trophic level. For Δδ15N we used the value of 3.4‰ per trophic level from Post (2002). The δ15Nbase in this case was estimated each year from the average δ15N value determined for Dungeness crab megalopae sampled in 2011 (δ15N = 10.3 ± 0.3 SE) and 2012 (δ15N = 10.2 ± 0.6 SE) and setting λ = 2.1 according to Miller et al. (2010). We assumed a one-month lag between consumption of prey and expression of trophic position in tissues of the consumer based on the turnover model of Vander Zanden et al. (2015).

We extracted lipids in chloroform and methanol according to Parrish (1999) using a modified Folch procedure (Folch et al. 1957). To calculate fatty acid concentration (μg mg−1), a constant amount of internal standard (23:0) was added to each sample. We prepared fatty acid methyl esters (FAME) by transesterification using sulfuric acid according to Budge et al. (2006) and analyzed FAME on an HP 7890 GC FID equipped with an autosampler and a DB wax + GC column (Agilent Technologies, Inc., USA) according to Copeman et al. (2016). The column temperature began at 65 °C for 0.5 min and was increased to 195 °C (40 °C min−1), held for 15 min then increased again (2 °C min−1) to a final temperature of 220 °C. Final temperature was held for 1 min. The carrier gas was hydrogen, flowing at a rate of 2 ml min−1. Injector temperature was set at 250 °C and the detector temperature was constant at 250 °C. We identified peaks using retention times based upon Supelco standards (37 component FAME, BAME, PUFA 1, and PUFA 3). We used Nu-Check Prep GLC 487 quantitative fatty acid mixed standard to develop correction factors for individual fatty acids and integrated chromatograms using Chem Station (version A.01.02, Agilent).

Analysis of trophic markers

Fatty acids and stable isotopes integrate information about dietary history over a period of weeks to months (Fry 2006; Copeman et al. 2016; Vander Zanden et al. 2015). We used correlation analyses to compare salmon stable isotopes and fatty acids to assess their coherence as trophic biomarkers. Trophic biomarkers were selected a priori from existing literature (Table 1). The piscivory marker (the ratio of docosahexaenoic acid to eicosapentaenoic acid, DHA:EPA) was selected based on initial observations by Daly et al. (2010) that fish prey contains higher proportions of DHA relative to EPA than invertebrate prey and later confirmed by fatty analysis of prey items collected in this study (fish prey DHA:EPA = 1.6 ± 0.1 SE and invertebrate prey = 0.9 ± 0.1 SE). To evaluate the relationship between environmental variability and salmon foraging ecology, we used regression analysis to estimate the relative importance of biological and physical factors in explaining variation in trophic position. Explanatory variables included measures of salmon size (FL and condition), measures of prey availability (proportion of fish in the diet and anchovy biomass), and environmental conditions measured in the coastal environment (SST, fluorescence, upwelling, and Columbia River plume volume). We lagged all explanatory variables except FL by one month to account for tissue turnover and conducted all statistical analyses using R (R Core Team 2015).

Table 1 Summary of trophic biomarkers analyzed in this study, including stable isotopes of carbon and nitrogen (δ13C and δ15N) and fatty acid biomarkers (based on percent of total fatty acids)
Full size table

Results

Salmon prey field

During the two years of our study, we captured 22,661 potential juvenile salmon prey items representing 47 taxa (Appendix). Species richness was highest in May, with 22 taxa collected in each of 2011 and 2012, but on average, biomass estimates were lower in May (53.1 ± 26.3 μg m−3) compared to other months. Lowest diversity (3 taxa) occurred in August 2012, but average biomass estimates were highest (2585 μg m−3) during this sampling period because of large catches of anchovy, an important prey for subyearling Chinook Salmon in the California Current (Brodeur 1991; Daly et al. 2009; MacFarlane 2010). In fact, during both years of the study, highest prey biomass estimates coincided with large catches of anchovy, although the timing of peak anchovy biomass differed between years. In 2011, the highest average anchovy biomass estimates occurred in September (342 μg m−3), whereas the highest anchovy biomass estimates in 2012 occurred in August (2577 μg m−3). Over the whole study, average prey biomass estimates were lowest in September 2012 (1.8 μg m−3). Interestingly, this was just one month after biomass estimates peaked for the study. Average anchovy biomass estimates in September 2012 were >450 times lower than they were in September 2011, representing a clear shift in prey phenology between the two years.

As expected, the prey field community transitioned throughout the sampling season as the coastal environment varied. Results of the multi-response permutation procedure comparing community composition revealed significant differences in the prey field by month (p < 0.001). Pairwise comparisons showed there was no difference in the prey field community between June and July or between August and September over the entire sampling period (both with p > 0.100), which justified combining months into three seasons over both years of the study (spring = May, early summer = June and July, and late summer = August and September). A diverse, but low biomass community of invertebrate prey and YOY fishes (osmerids, flatfishes, and other groundfish in their pelagic phase) occurred in spring and early summer, which transitioned into a community dominated by anchovy by late summer.

The NMS ordination that best described the prey field dataset had three dimensions, Monte Carlo p < 0.001 and medium stress (12.5), suggesting little risk of drawing false inferences (Fig. 2, Table 2). The ordination represented 86.6% of the variation in the prey community. Axis 1 accounted for 55.5% of the variation and was positively associated with Dungeness Crab megalopae, smelt (Osmeridae), Pacific Sanddab (Citharichthys sordidus), Slender Sole (Lyopsetta exilis), and Sand Sole (Psettichthys melanosticus) during spring (May), when Columbia River plume volume was highest. Axis 1 was negatively associated with anchovy during late summer (August and September), when SST and fluorescence values were highest, indicating high productivity. Axis 2 accounted for 15.3% of the variation and separated coastal downwelling from upwelling conditions. Axis 3 accounted for 15.8% of the variation and was positively associated with rockfish (Sebastes spp.), negatively associated with California Market Squid (Doryteuthis opalescens), but was not associated with any environmental variables measured.

Fig. 2
figure 2

Results of nonmetric multidimensional scaling analysis where each point represents the log-transformed biomass (μg m−3) of invertebrate and marine fish prey sampled by haul and plotted in species space from May through September 2011 and 2012; top plots show joint plots of environmental variables (cut-off r2 = 0.30 for inclusion) for a axes 1–2, and b axes 1–3; bottom plots show joint plots of species (cut-off r2 = 0.30 for inclusion) for c axes 1–2, and d axes 1–3 (symbols are based on categorical groupings for season: spring = May, denoted by a filled black triangle, early summer = June–July, denoted by a grey circle, and late summer = August–September, denoted by an open square); abbreviations are 1 m sea surface temperature (SST), Columbia River plume volume (CR plume; km3), and unidentified (unid)

Full size image
Table 2 Pearson’s correlation coefficients for associations between nonmetric multidimensional scaling ordination axes and environmental variables or species, denoted by “*” where r2 > 0.30
Full size table

UCSF subyearlings

Subyearling UCSF Chinook Salmon was the most abundant genetic stock group sampled alongside the prey community from July through September in both years. We also collected juvenile Chinook Salmon from 14 other genetic stock groups, as well as juvenile Coho, Chum (O. keta), and Sockeye (O. nerka) Salmon, but they were not included in this analysis. Average size of UCSF subyearlings, including salmon sampled in the two other complementary studies, ranged from 89 mm FL and 6.9 g in May to 151 mm FL and 43.2 g in September (Fig. 3a, b). Salmon lengths (t-test, p = 0.014) and weights (p = 0.002) were significantly lower in June 2011 compared to June 2012, but were significantly larger by the end of September in 2011 compared to 2012 (length and mass both p < 0.001). Average early ocean growth rates were significantly (p = 0.005) slower from June to July in 2011 (0.4 mm d−1) than 2012 (0.6 mm d−1), but significantly faster (p < 0.001) from August to September in 2011 (1.2 mm d−1) compared to 2012 (0.9 mm d−1; Fig. 3c). Average specific growth rates were also significantly (p < 0.001) faster from August to September 2011 (3.1% BW d−1) compared to August to September 2012 (2.0% BW d−1; Fig. 3d). As expected, UCSF subyearlings grew fastest when anchovy were most abundant in the field and consumed. Salmon condition was also significantly higher (ANOVA and Tukey HSD, p < 0.05) when anchovy biomass in the field was greatest (Fig. 4). Across years, mean salmon condition and anchovy biomass were significantly and positively correlated (r = 0.900, p = 0.015).

Fig. 3
figure 3

Plots of average ± SE a fork length b mass, c growth rates, and d specific growth rates of subyearling Chinook Salmon (Oncorhynchus tshawytscha) assigned to the upper Columbia summer-fall genetic stock group in 2011 and 2012 (*indicates significance [Student’s t-test, p < 0.05])

Full size image
Fig. 4
figure 4

Scatterplots with separate scales on the y-axis for monthly mean ± SE condition of subyearling Chinook Salmon (Oncorhynchus tshawytscha) assigned to the upper Columbia summer-fall genetic stock group (filled triangles) and mean ± SE biomass of Northern Anchovy (Engraulis mordax) measured in net tows (open circles) sampled July through September in a 2011 and b 2012 (different subscripts indicate months when salmon condition varied significantly [ANOVA and Tukey HSD; p < 0.05])

Full size image

Stomach contents

Juvenile salmon stomach contents varied by month and year over the sampling period and reflected variations in the prey field (Fig. 5). Of the stomachs examined, 18 were considered empty. In general, proportions of fish increased in stomachs through time, except from August to September 2012, when proportions of fish in diet (and salmon growth rate) decreased. The onset of piscivory, which we defined as when diets contained more fish than any other prey category, occurred later in 2011 (September =62% of diet) than 2012 (August =50% of diet). Average salmon size at the onset of piscivory (Fig. 6) was significantly higher (t-test, p < 0.01) in 2011 (140–150 mm FL; 49% fish in diet) than in 2012 (120–130 mm FL; 43% fish in diet). All salmon >150 mm were completely piscivorous in 2011 whereas proportions of fish in diet decreased rather than increased in fish >130 mm FL in 2012, indicating that the onset of piscivory was related more to prey availability than predator size. Prey-sized (<80 mm TL) anchovy were caught in the trawl in August and September 2011 and from July through September 2012. Stomach fullness (% of total body weight) was significantly higher (ANOVA and Tukey HSD, p < 0.05) when salmon ate more anchovy (mean ± SE = 2.2 ± 0.4% in September 2011 and 2.0 ± 1.2% in August 2012), compared to when anchovy were less abundant and overall prey biomass estimates were low (0.6 ± 0.3% in September 2012).

Fig. 5
figure 5

Average percent biomass of potential prey measured from June through September in a 2011 and b 2012, and diet composition presented as percent wet mass of prey eaten by subyearling Chinook Salmon (Oncorhynchus tshawytscha) in c 2011 (n = 125) and d 2012 (n = 104; fish prey indicated by colored bars, invertebrate prey indicated by black and white bars)

Full size image
Fig. 6
figure 6

Diet composition presented as percent of wet mass of prey by length category in a 2011 (n = 125) and b 2012 (n = 104; fish prey indicated by colored bars, invertebrate prey indicated by black and white bars)

Full size image

Salmon stable isotopes and lipids

Salmon stable isotopes, trophic position, and fatty acids varied by month and year (Table 3). Correlation analyses identified significant relationships among bulk stable isotopes of carbon and nitrogen, fatty acids, and salmon size (Table 4, Figs. 7 and 8). Values of δ13C, indicating carbon source at the base of the food web, were positively correlated with salmon FL (r = 0.620, p < 0.001) and a fatty acid biomarker for marine diatoms:flagellates (the ratio of all polyunsaturated fatty acids [PUFA] containing 16 carbon atoms to all PUFA containing 18 carbon atoms, r = 0.730, p < 0.001; Fig. 7a, c). Carbon isotopes and salmon FL were both negatively correlated (δ13C r = −0.700 and FL r = −0.560, both p < 0.002) with a fatty acid biomarker for freshwater (the sum of linolenic and linoleic fatty acids [18:3n-3 + 18:2n-6]; Fig. 7b, d), indicating that the dietary carbon pools changed as salmon migrated from freshwater to saltwater. Values of δ15N were positively correlated with salmon FL (r = 0.460, p = 0.012), δ13C (r = 0.690, p < 0.001) and a fatty acid marker for piscivory (DHA:EPA, r = 0.720, p < 0.001; Fig. 8), which supports the observation that salmon diet shifted from invertebrate prey to piscine prey during early ocean residence as salmon grew (Fig. 9). Across years, salmon in September 2012 had the highest δ15N, trophic position, and DHA:EPA, even though they were not the largest fish sampled and were not piscivorous at the time of collection. Due to the lag between when prey is consumed and expressed in salmon isotopes and fatty acids, we assumed that salmon biochemistry in September 2012 reflected fish prey consumed in August 2012, the sampling period when anchovy were most abundant in the field and when salmon first became piscivorous.

Table 3 Average ± SE subyearling Chinook Salmon (Oncorhynchus tshawytscha) fork length (FL, mm), mass (g), stable isotopes of carbon (δ13C) and nitrogen (δ15N, ‰), trophic position, total fatty acids expressed as the amount per dry weight of each sample (μg g−1),% fatty acids comprising >1% of total fatty acids, and biomarker * values for fatty acids from 29 samples collected July through September 2011 and 2012
Full size table
Table 4 Significant (p < 0.05) Pearson’s correlation coefficients for comparisons between carbon and nitrogen stable isotopes (δ13C and δ15N) and fatty acid biomarkersa from 29 subyearling Chinook Salmon (Oncorhynchus tshawytscha) muscle tissue samples collected July through September 2011 and 2012
Full size table
Fig. 7
figure 7

Correlations between subyearling Chinook Salmon (Oncorhynchus tshawytscha) carbon stable isotope values (δ13C), fatty acids biomarkers (based on percent total fatty acids), and size: a the relationship between diatom to flagellate markers indicated by the ratio of all polyunsaturated fatty acids (PUFA) containing 16 carbon atoms to all PUFA containing 18 carbon atoms and δ13C, b the relationship between freshwater markers indicated by the sum of linolenic acid (18:3n-3) and linoleic acid (18:2n-6) and δ13C, c the relationship between salmon fork length (FL) and 16:18 PUFA, and d the relationship between salmon FL and 18:3n-3 + 18:2n-6 (symbols are based on categorical groupings for year and month: 2011 = filled symbols, 2012 = open symbols; July values are denoted by a triangle, August by a square, and September by a circle)

Full size image
Fig. 8
figure 8

Correlations between subyearling Chinook Salmon (Oncorhynchus tshawytscha) isotope values, fatty acid biomarkers (based on percent of total fatty acids), and size: a the relationship between carbon and nitrogen stable isotope values (δ13C and δ15N), b the relationship between the ratio of docosahexaenoic acid to eicosapentaenoic acid (DHA:EPA) and δ15N, c the relationship between salmon fork length (FL) and δ15N, and d the relationship between salmon FL and DHA:EPA (symbols are based on categorical groupings for year and month: 2011 = filled symbols, 2012 = open symbols; July values are denoted by a triangle, August by a square, and September by a circle)

Full size image
Fig. 9
figure 9

Associations between mean ± SE subyearling Chinook Salmon (Oncorhynchus tshawytscha) trophic position and a fluorescence (1-mo lag), b Columbia River plume volume (1-mo lag), c salmon fork length, d proportion of fish measured in salmon diet (1-mo lag), e biomass of juvenile Northern Anchovy (Engraulis mordax) measured in the field (1-mo lag), and f salmon condition (1-mo lag)

Full size image

Models estimating trophic position

Regression analysis of average salmon trophic position and physical and biological factors showed salmon trophic position increased from July to September from 2.5 to 2.9 in 2011 and from 2.6 to 3.0 in 2012, reflecting prey consumed from approximately June through August in both years (Table 3, Fig. 9). Monthly and inter-annual variations in trophic position were best explained by differences in fluorescence (1-mo lag), Columbia River plume volume (1-mo lag), salmon FL, and proportion of fish in salmon diet (1-mo lag; Fig. 9). These results demonstrate that all fish were transitioning from being zooplantivorous to piscivorous through time as they increased in size. Associations between trophic position and the physical environment also highlights the potential role of bottom-up productivity in regulating the timing of the onset of piscivory.

Discussion

Our analyses showed that inter-annual differences in timing and size at the onset of piscivory in a population of Chinook Salmon were related to prey availability. Seasonal variation in the prey field during ontogeny has broad implications for early ocean foraging success, growth, and survival in UCSF subyearlings. The composition of the prey field is also important for other populations of Columbia River Pacific Salmon and Steelhead (O. mykiss) that have stock-specific differences in their size and timing of ocean entry (Weitkamp et al. 2015). Our results that UCSF subyearlings became piscivorous at a smaller size and earlier in 2012 than 2011, but that the earlier onset of piscivory did not result in higher growth rates by the end of the sampling period in 2012, lends support to the idea that the total duration of piscivory, not just the onset, should be considered in terms of long-term fitness and survival. A longer period of piscivory in early marine life may increase survival by increasing salmon size at the end of the first summer at sea, as suggested by the critical size, critical period hypothesis (Beamish and Mahnken 2001). It may also be that Chinook Salmon size-at-maturity, fecundity, and fitness are related to the total duration of piscivory during marine life.

The spatial coverage of our study was limited through time therefore we do not know how representative our samples were of the prey community throughout salmon’s range. We suspect that our samples represented available prey in the Northern California Current as previous surveys of zooplankton (Lamb 2011) and ichthyoplankton (Auth 2011) resources concluded that community structure was homogenous north and south of the region sampled. In June 2016, NOAA Fisheries conducted a coastwide (northern Washington to central California) survey of the salmon prey field in shelf waters using the same gear type as we did to further investigate this issue. That survey also used zooplankton net tows (bongo and Methot nets) to capture small marine invertebrate prey such as pteropods, ostracods, copepods, amphipods, and decapod larvae that our trawl gear was less effective at sampling. Future studies could also include image analysis (e.g. in situ ichthyoplankton imaging system [ISIIS], Cowen and Guigand 2008) or acoustics to provide a more complete quantitative estimate of the prey field, especially at depths greater than we sampled where salmon might feed (Brodeur et al. 2011). This type of sampling could help better understand prey patchiness as it relates to fronts and other oceanographic features (Peterson and Peterson 2008; Ainley et al. 2009; Brodeur and Morgan 2016).

Estimates of marine growth in UCSF subyearlings were based on the assumption that salmon caught in the estuary and ocean were representative of the entire population and that changes in size were due to ocean growth and not factors such as emigration, immigration, or size-selective mortality. We contend that this is a reasonable assumption as our estimates of growth (0.4 to 1.2 mm d−1 and 0.8 to 3.1% BW d−1) were similar to estimates calculated for this stock based on otoliths (0.8 to 1.2 mm d−1 and 0.9 to 2.6% BW d−1; Miller et al. 2013; Claiborne et al. 2014). We suggest that foraging and growth patterns observed in UCSF subyearlings would be similar in salmon populations with similar life histories. One example is the Snake River fall population of Chinook Salmon, which was the second most abundant genetic stock group represented in our samples (14.2% of catch). Unlike the UCSF population, Snake falls are listed under the U.S. Endangered Species Act. These fish migrate to the ocean slightly earlier than UCSF subyearlings (Weitkamp et al. 2015), but have the same genetic lineage (Waples et al. 2004) and similar early marine distributions and growth rates (Teel et al. 2015; Weitkamp et al. 2015).

Combining stomach content analysis with stable isotope and fatty acid analyses is a powerful integrated approach for evaluating diet and for understanding biological processes. For example, carbon isotope values in fish, which represent primary producers at the base of the food web, can be difficult to interpret as fractionation of δ13C in phytoplankton tends to vary positively with SST, cell size, and growth rate, and negatively with dissolved inorganic carbon pools (CO2 and HCO3 −) (Laws et al. 1995; Burkhardt et al. 1999). Combining stable isotopes with fatty acid biomarkers can help aid in δ13C interpretation. The fatty acid biomarker approach is based on observations that phytoplankton produce essential fatty acids not biosynthesized by consumers that are then deposited in consumer tissue with minimal modification (Dalsgaard et al. 2003; Budge et al. 2006; Copeman et al. 2016). Freshwater biomarkers (18:2n-3 + 18:2n-6), likely originating from the Columbia River, varied inversely with δ13C and salmon FL, whereas marine phytoplankton biomarkers (16:18 PUFA) varied positively with δ13C and salmon FL. These results confirm a general pattern of enrichment of δ13C in salmon tissue by up to 7.5‰ associated with the ontogenetic shift in diet and habitat from freshwater to saltwater. However, it varies from other interpretations of juvenile salmon δ13C that consider there to be a gradient between enriched δ13C waters onshore and depleted δ13C values offshore within the marine realm (Miller et al. 2008). Salmon δ13C may also vary along a SST gradient with lower δ13C values at lower temperatures due to higher concentrations of dissolved CO2 in cooler waters (Hertz et al. 2015a), although we found no significant relationship between salmon δ13C and SST (r = 0.00, p = 0.99).

Completely piscivorous fish should have a trophic position of 3.5–4.0, but by September our estimates for trophic position were 2.9 in 2011 and 3.0 in 2012. There are several explanations for this. The first is that piscivorous diets were not yet reflected in salmon tissue. We assumed a one-month lag between a diet shift and when tissue actually equilibrates with that value, but muscle may take longer to equilibrate than other tissues (Heady and Moore 2013). Anchovy consumed in August may not have been reflected in salmon muscle by September. A two-month lag would suggest that September δ15N values reflected zooplanktivorous diets consumed in July. We also may have overestimated the trophic discrimination factor of 3.4‰ given that our δ15N baseline values for crab megalopae were near 10‰. Recent analyses (Hussey et al. 2014; Hertz et al. 2016) show that turnover can complicate interpretation of ontogeny in salmon. We also may have inadequately captured the trophic baseline by integrating megalopae values across season and space. A final possibility is that UCSF juveniles never became completely piscivorous, an observation supported by stomach content data and consistent with the notion that Chinook Salmon are generalist foragers that feed upon a variety of prey types (Gregory and Northcote 1993).

The use of DHA:EPA as a biomarker of piscivory is based on observations that DHA is conserved in higher trophic levels in marine ecosystems (Dalsgaard et al. 2003; Parrish 2013) and found in higher concentrations in marine fish prey relative to invertebrate prey (Daly et al. 2010; this study). Increases in the DHA:EPA ratio may also indicate a decrease in juvenile salmon condition as lipid classes get utilized differently for energy under poor feeding conditions. Storage lipids (triacylglycerides) contain small amounts of DHA, but are mobilized more readily than polar lipids, which contain proportionally higher amounts of DHA. Similarly, fasting has been shown to increase δ15N signatures by up to 0.5‰ in juvenile Chinook Salmon (Hertz et al. 2015b). Future applications of DHA:EPA and δ15N as trophic markers must consider salmon condition in addition to stomach content data, as poor feeding conditions may cause these biomarkers to increase even though no fish prey has been consumed.

Availability of anchovy prey appears to have important consequences for the onset of piscivory and growth in UCSF subyearlings. Anchovy are an abundant forage fish in the northern California Current (Litz et al. 2008) and the timing and duration over which they spawn is related to physical factors such as SST and river plume dynamics (Richardson 1973; Parnel et al. 2008). Coastal ocean conditions were similar in the two years of our study and ecosystem indicators of ocean conditions related to salmon recruitment such as Pacific Decadal Oscillation (Mantua et al. 1997), winter SST, and copepod community structure (Peterson et al. 2014) were all considered favorable. Notably, mean volume of the Columbia River plume was 4 times larger in June 2011 (200 km3) compared to June 2012 (49 km3). June and July have been identified as important months for spawning anchovy that utilize plume fronts during reproduction (Richardson 1973; Auth 2011). There was also a difference in the intensity of upwelling favorable winds during July of 2011 relative to 2012; wind speed cubed ([m s−1]3) was 2.4 times lower in July 2011 (50.5) compared to 2012 (122.5; www.pfeg.noaa.gov). Differences in plume size and wind strength may have contributed to inter-annual differences in the timing and magnitude of anchovy recruitment, which occurred later and persisted longer in 2011 than 2012. A less windy surface layer, as was observed in June and July 2011, is hypothesized to benefit first feeding anchovy during early life (Lasker 1978).

Yearling Chinook and Coho Salmon (juveniles that enter the ocean as 1.0 age fish) and juvenile Steelhead originating from the Columbia River are typically larger, migrate to the ocean earlier (April–May), and become piscivorous sooner than subyearlings (Daly et al. 2014; Weitkamp et al. 2015). Marine YOY fishes may contribute as much as 50 to 90% of the mass of larger juvenile diets during spring and early summer (Brodeur et al. 2007; Daly et al. 2009), but have been previously sampled in the field with limited success (Schabetsberger et al. 2003; Brodeur et al. 2011; Brodeur and Morgan 2016). Yearling Chinook, Coho, and Steelhead consume smelt, rockfishes, greenlings (Hexagramidae), sculpin (Cottidae), Pacific Sand Lance (Ammodytes hexapterus), and flatfishes (Pleuronectiformes, Brodeur et al. 2007; Daly et al. 2010, 2014). These prey items were well represented in our samples, but mainly collected in May when average biomass estimates were low (57.0 ± 26.3 μg m−3 SE). It is currently unknown whether the observed low biomass estimates of fish prey measured in May was related to predator density, but hypotheses addressing prey limitation and predator density dependence could be further explored by comparing yearling densities, diets (including stomach fullness), condition, and composition of the prey field. Intra- and interspecific competition among outmigrating juvenile salmon is a topic that has received considerable attention in the North Pacific Ocean and Bering Sea (see Ruggerone and Nielsen 2004 for a review), but has not been extensively investigated for fish exiting the Columbia River.

Understanding variability in feeding ecology of juvenile salmon first entering the marine environment, and in particular identifying factors relating to their ontogenetic diet shift, is important for understanding future salmon survival. Chinook Salmon from the UCSF population typically spend 2–4 years in the ocean before returning to spawn, with the majority returning after three ocean winters. Adult passage of summer and fall Chinook Salmon at Priest Rapids Dam on the Upper Columbia River (www.cbr.washington.edu) after three years in the ocean provides a measure of relative survival (Miller et al. 2013; Losee et al. 2014). Juveniles from the same year class as fish sampled in this study in 2011 and 2012 predominantly returned during summer and fall of 2014 and 2015. Adult returns of UCSF salmon was high in 2014 (198,341) – almost twice the ten-year average from 2004 to 2013 (111,008). Adult returns in 2015 were also unexpectedly high (167,440) despite a large developing El Niño (www.elnino.noaa.gov) and unprecedented warming in the Northeast Pacific known as the “The Blob” (Bond et al. 2015). Hatchery production of UCSF subyearlings has been similar over the last two decades (www.fpc.org) therefore the high returns in 2014 and 2015 were not due to increased hatchery production. Because El Niño and warm ocean temperatures are typically associated with poor salmon survival (Mantua et al. 1997; Meuter et al. 2002), high returns of 2011 and 2012 outmigrants suggests that UCSF salmon may have benefitted from favorable ocean conditions and abundant anchovy prey during early ocean residence. Comparison of salmon prey, diet, trophic biomarkers, and growth during years of poor survival could be used to test this hypothesis. During a warm and unproductive outmigration year, we might expect changes in the overall composition and abundance of salmon prey, earlier or later shifts in prey phenology and the onset of piscivory, lower growth rates, condition, and adult returns. Our integrated approach that combines sampling of salmon prey with estimates of salmon growth, stomach contents, stable isotopes, and fatty acids provides a useful framework for assessing bottom-up regulatory mechanisms impacting foraging ecology during a critical period in salmon life history.