MARINE ECOLOGY PROGRESS SERIESMar Ecol Prog Ser

Vol. 613: 151–169, 2019https://doi.org/10.3354/meps12896

Published March 21

1. INTRODUCTION

Marine prey are highly heterogenous across spaceand time (Barry & Dayton 1991), creating both oppor-tunities and challenges for mobile consumers. Forexample, infrequent pulses of prey due to seasonaldynamics and prey phenology can create temporaryprey hotspots for predators (Croll et al. 2005, Yang etal. 2008). Further variation of prey abundance acrossspace and time may be driven by major habitat tran-

sitions, abiotic preferences of prey, environmentaldynamics, and top-down effects of predators (Telesh& Khlebovich 2010, Lannin & Hovel 2011, David et al.2016). Some of the difficulties faced by mobile con-sumers that result from the heterogeneity of preyacross seascapes are the risks of starvation and pre-dation (Letcher & Rice 1997, Pitchford 2001). To copewith this spatio–temporal variation in prey, predatorsrespond behaviourally by modifying their distribu-tions (e.g. migration) and search patterns (e.g. Lévy

© The authors 2019. Open Access under Creative Commons byAttribution Licence. Use, distribution and reproduction are un -restricted. Authors and original publication must be credited.

Publisher: Inter-Research · www.int-res.com

*Corresponding author: marbeide@sfu.ca

Integrating prey dynamics, diet, and biophysical factors across an estuary seascape

for four fish species

Michael Arbeider1,*, Ciara Sharpe1, Charmaine Carr-Harris2,3, Jonathan W. Moore1

1Earth to Oceans Research Group, Simon Fraser University, 8888 University Drive, Burnaby, British Columbia V5A 1S6, Canada2Skeena Fisheries Commission, 3135 Barnes Crescent, Kispiox, British Columbia, V0J 1Y4, Canada

3Present address: Fisheries and Oceans Canada, 417 2nd Ave W, Prince Rupert, British Columbia, V8J 1G8, Canada

ABSTRACT: Estuary food webs support many fishes whose habitat preferences and populationdynamics may be controlled by prey abundance and distribution. Yet the identity and dynamics ofimportant estuarine prey of many species are either unknown or highly variable between regions.As anthropogenic development in estuaries increases, so does the need to understand how theseenvironments may be supporting economically, culturally, and ecologically important fishes.Here, we examine how important estuary fishes integrate their prey across the seascape and whatmay influence prey dynamics. Specifically, we surveyed juvenile coho salmon Oncorhynchuskisutch, juvenile sockeye salmon O. nerka, Pacific herring Clupea pallasii, and surf smelt Hypo -mesus pretiosus diets along with zooplankton abundance in the estuary of the Skeena River(British Columbia, Canada) at a relatively fine scale. We found diets were highly variable, evenwithin a species, but 1 or 2 prey composed most diet contents per species. Juvenile coho salmonprimarily consumed terrestrial insects and larval fish, whereas sockeye salmon primarily con-sumed harpacticoid copepods. In contrast, small pelagic fish (Pacific herring and surf smelt) pri-marily consumed calanoid copepods, which were the most abundant prey in the environment. Wefound that certain prey groups were correlated with biophysical factors. For example, calanoidcopepod abundance was positively correlated with salinity, whereas harpacticoid copepod abun-dance was highest over eelgrass sites. Identifying key prey species and how they distribute withinthe estuary seascape is an integral link in understanding the food-web foundation of fish habitatuse in areas under pressure from anthropogenic development.

KEY WORDS: Juvenile salmon · Small pelagic fish · Estuary · Diet · Prey · Oncorhynchus · Clupea ·Hypomesus

OPENPEN ACCESSCCESS

Mar Ecol Prog Ser 613: 151–169, 2019

walk) to increase encounter rates with prey patches(Croll et al. 2005, Sims et al. 2008). Alternatively,mobile predators may integrate across staggered,smaller pulses of prey that occur over the variableseascape (e.g. prey waves instead of hot spots) toachieve more extended and consistent feedingopportunities (Armstrong et al. 2016). Different spe-cies of predators have different prey preferences andforaging abilities and employ different search move-ments to survive and thrive in heterogeneous preyseascapes. Thus, understanding the prey dynamicsof seascapes is a key component of understandingthe ecology of their consumers (Boström et al. 2011).

Estuaries can be prey-rich places for planktivorousfishes (Selleslagh et al. 2012, Levings 2016), but theyare driven by multiple biophysical processes thatproduce particularly dynamic prey fields. Here, werefer to estuaries as the tidally influenced portions ofrivers that have saltwater influence and the con-stituent bays that have freshwater influence (Perillo1995). Productivity in estuaries is derived from thecombination of riverine inputs and upwelled oceannutrients, as well as local production from seagrasses, salt marshes, benthic and epiphytic algae,and microbes (Cloern et al. 2014). Each of thesesources have seasonal patterns, often creating largephytoplankton blooms followed by zooplanktonblooms (Cloern 1996, Mackas et al. 2012). In addi-tion, zooplankton within an estuary are challengedby complicated hydrodynamic effects of tides andcurrents that interface with a variety of habitats toremain in their optimal environment (Palmer 1988,David et al. 2016). For example, rising tides overintertidal areas can push pelagic zooplankton intohigh density patches (David et al. 2016). In contrast,benthic and epibenthic zooplankton are known tohave higher site fidelity than pelagic species becausethey can bury or attach themselves to their substrateand avoid this redistribution (Palmer 1988). Temper-ature and turbidity are also strong drivers of zoo-plankton habitat preference and are linked with theirgrowth and reproductive development (Morgan et al.1997). In addition to natural variability, estuaries arealso undergoing changes from anthropogenic activ-ity that may impact the dynamics of natural factors(López Abbate et al. 2015). Thus, a dynamic mosaicof zooplankton prey provide the resource base forplanktivorous fishes that may rely on estuaries forstaging or important nursery habitats (Beck et al.2001, Nagelkerken et al. 2015, Sheaves et al. 2015).

Multiple economically, culturally, and ecologicallyimportant small fishes such as juvenile coho salmonOncorhynchus kisutch, juvenile sockeye salmon

O. nerka, adult Pacific herring Clupea pallasii, andadult surf smelt Hypomesus pretiosus are supportedby zooplankton and other food sources in estuariesalong the West Coast of North America (Table S1 inthe Supplement at www.int-res. com/ articles/ suppl/m613 p151 _ supp. pdf). Food webs supporting juvenilecoho salmon are well researched, though many stud-ies are from systems in the California Current Systemlike the Columbia River estuary or estuaries in PugetSound (Table S1). These studies found large re -gional, seasonal, annual, and ontogenetic variabilityin juvenile coho salmon diets (Brodeur et al. 2007b,Daly et al. 2009, Bollens et al. 2010, Levings 2016).Juvenile coho salmon are considered generalists,eating decapod larvae, amphipods, pteropods, cope-pods, euphausiids, eggs, and various other larvalcrustaceans but are predominantly piscivorous andinsectivorous (Brodeur 1991). Studies on sockeyesalmon diets in estuaries reported that they con-sumed euphausiids, cirripeds, mysids, larval fish, andcalanoid copepods as well as other crustaceans inminor amounts (Simenstad et al. 1982, Birtwell et al.1987, Ajmani 2011). In contrast, Pacific herring andsurf smelt can have variable diets but generally con-sume copepods and other crustaceans in coastal en -vironments (Miller & Brodeur 2007, Hill et al. 2015).To the best of our knowledge, few published reportson the estuarine diets of adult Pacific herring existand none for adult surf smelt (Table S1). Overall,there is variable scientific understanding of the estu-ary diets of juvenile coho and sockeye salmon, Pacificherring, and surf smelt in Northeast Pacific estuaries.There are even fewer studies linking diets and preydistribution patterns (Bollens et al. 2010) to the un -derstanding of the role of estuaries as staging andnursery habitats (Nagelkerken et al. 2015, Sheaves etal. 2015, McDevitt-Irwin et al. 2016).

This study focusses on fish diets and prey distribu-tions in the Skeena River estuary, an important areafor juvenile salmon and one that was also under con-sideration for anthropogenic development at the timeof writing (CEAA 2016). The estuary region we focuson is within the larger Skeena River estuary, an areapreviously identified as having particularly highabundances of juvenile salmon during their migra-tion: 2- to 8-fold greater abundance of juvenile sal -mon than other regions over several years of obser-vation (Carr-Harris et al. 2015). The region supportsjuvenile salmon from throughout the Skeena water-shed, with at least 40 different populations identifiedin the estuary (Carr-Harris et al. 2015, Moore et al.2015) that enter at different times (Carr Harris et al.2018), most likely forage there, and reside for vari-

152

Arbeider et al.: Integrating estuarine prey dynamics

able amounts of time (Moore et al. 2016). The aver-age estimated residencies for sockeye and coho sal -mon were 2 and 14 d, respectively (Moore et al.2016). Further work within this region identified different fish abundances associated with abioticaspects of estuary habitat (Sharpe 2017); however,how the estuarine zooplankton community supportsalmon and other small pelagic fish in the SkeenaRiver estuary remains unknown.

Here, we quantify the spatial and temporaldynamics of estuarine prey for 4 fish species andtheir relationships with biophysical aspects of theirseascape. Specifically, we studied juvenile cohosalmon, juvenile sockeye salmon, Pacific herring,and surf smelt in the estuary of the Skeena Riverin northern British Columbia, Canada. We asked(1) how are prey distributed in the estuary acrossspace and time, (2) what are the most consumedand selected prey of these 4 fish species, (3) dobiophysical factors of the estuary co-vary orpredict diet variability, and (4) can variability inprey abundance be predicted by biophysical fac-tors? We discovered that diets varied greatlyacross the small spatial and temporal scale of ourstudy for each species, particularly salmon, andthat a few prey taxa had consistently high abun-dances across the seascape while other prey wereassociated with different biophysical factors, suchas salinity or the presence of eelgrass. These find-ings provide insight into important prey dynamicsand identify biophysical factors through whichpotential change could impact food webs support-ing key fish species — a recognized knowledgegap in on-going decision-making and planningprocesses in the Skeena River watershed (Pickardet al. 2015).

2. MATERIALS AND METHODS

We investigated the spatio–temporal dynamics ofzooplankton prey along with the diets of 4 fish spe-cies in the estuary of the Skeena River. We mappedimportant prey abundance across our sampling sitesand tested for trends in abundance between sites andsampling periods. Next, we ranked prey importanceand selectivity with 2 common metrics and scoreddiet overlap across individuals within species tomeasure small-scale diet variability. Subsequently,we assessed whether variation in important preyabundance in diet samples and in the seascape couldbe predicted by biophysical factors through general-ized linear regression.

2.1. Study area

The Skeena River is the second-largest watershedentirely in British Columbia, Canada, draining anarea of 55 000 km2. The Skeena River mixes with theocean in Chatham Sound, a semi-enclosed basin of~1500 km2 (Ocean Ecology 2014), by travellingthrough 3 major passages. Our study area is situatedat the end of the northern-most passage, Inverness,which directs ~25% of the total flow of the SkeenaRiver (Trites 1956). The study region (Fig. 1) is in thetraditional territory of the Tsimshian First Nationsand is a focus of a research program developed incollaboration with Lax Kw’alaams Fisheries andSkeena Fisheries Commission to improve under-standing of estuarine use by juvenile salmon and thebroader estuarine food web.

2.2. Field sampling and laboratory methods

For diet analysis, we lethally sampled (SimonFraser University Animal Care 1107B-11; Fisheriesand Oceans Canada licence XR 82 2016) 111 sock-eye salmon Oncorhynchus nerka, 57 coho salmonO. kisutch, 57 Pacific herring Clupea pallasii, and35 surf smelt Hypomesus pretiosus across 17 sam-pling occasions (from here on, referred to as ‘sets’;Table 1). The 17 sets occurred across 9 sites, i.e. somesites had >1 sampling occasion while others only had1. The 9 sites where samples were ob tained were asubset of 25 sites from the ongoing research program(Fig. 1). Sites were chosen to represent 4 main habi-tat types available in the estuary — eelgrass, sandybay, rocky shoreline, and open water (Sharpe2017) — or were part of the long-term monitoringproject (Carr-Harris et al. 2015). We collected fishwith 2 sizes of purse seine; the larger net measured73.2 m long by 9.1 m deep with 5.1 cm webbing at thetow end and 1.3 cm webbing at the bunt, and thesmaller net measured 45.7 m long by 5.5 m deep with1.3 cm webbing at the tow end and 0.64 cm webbingat the bunt end. The 2 sizes of net were used so wecould target the entire water column of sites withvarying depths without snagging the net on the estu-ary substrate. We enumerated each species of fishand calculated a catch per unit effort (CPUE) as ameasurement of salmon abundance. Relative abun-dances from the smaller purse seine were standard-ized to the larger purse seine by multiplying thesmall net catches by the large net area (length bywidth) and net tow duration, then dividing by thearea and tow duration of the small net. We aimed to

153

Mar Ecol Prog Ser 613: 151–169, 2019

tow the seine hauls from the different sizes of net atsimilar speeds so that they would catch a similardiversity of fish. Fish samples were retained whenthere were at least 5 individuals of a species from anygiven set available for collection. We collected fishbetween 10 May and 21 June 2016 (see Table 1 for

specific dates), to capture juvenilesalmon around the peak of their outmi-gration, immediately storing lethalsamples in seawater buffered 5% for-malin solution.

Fish and diet samples were furtherprocessed in the laboratory. We meas-ured fork length (mm) and wet weight(g) (outside pat dried with paper towel)of all fish before excising their stom-achs. Stomach contents were analyzedby identifying prey to the lowest possi-ble taxonomic level. Abundance, totalwet weight, and state of digestion wasreported for each prey taxa in eachstomach. Prey that was too digested tobe identified was removed from thesubsequent analysis. When diet con-tents could be identified to taxonomicgroup but were broken into parts, pre-venting an accurate count of indivi -duals, we estimated abundance byusing prey-specific linear regressionsof known abundance on weight fromour diet samples (Table S2 in the Sup-plement). One sockeye salmon, 4 cohosalmon, and 1 Pacific herring hadempty stomachs, leaving 110 sockeyesalmon, 53 coho salmon, 56 Pacificherring, and 35 surf smelt diets in theanalysis (Table 1).

We concurrently sampled for zoo-plankton in the environment at the18 small-purse seine sites when fishwere sampled (Fig. 1). Zooplanktonwere collected over 4 time periods:13−20 May, 24 May to 1 Jun, 6− 10Jun, and 20−24 Jun (n = 71, one sam-pling occasion was missed due tosafety concerns from ocean condi-tions). We used a 250 μm WP2 plank-ton net towed by hand verticallyfrom a boat from 5 m below the sur-face to standardize the volume ofwater that was sampled. Sampleswere stored in a seawater buffered5% formalin solution. We stained

zooplankton with Rose Bengal to make them morevisible, partitioned them with a Folsom planktonsplitter, and sorted them until at least 400 individu-als or the entire sample had been identified. Weused a taxonomic level that was comparable to zoo-plankton identified within the diet samples and

154

Day of Site Set Coho Sockeye Herring Smeltyear no.

131 Porpoise Channel 1 4 4 0 5134 Lelu–rock 2 0 10 10 5141 Flora 1 3 2 2 0 0141 Kitson 4 0 10 10 0145 Inverness Lelu 5 9 0 3 0147 Flora 1 6 5 6 0 0153 Porpoise Channel 7 5 10 0 0153 Kinahans West 8 0 10 0 0158 Inverness Lelu 9 5 5 5 5158 Inverness NP 10 5 11 3 5158 Flora 2–eelgrass 11 0 5 5 0159 Kinahans–open water 12 0 11 5 0160 Flora 1 13 3 5 4 5161 Kitson 14 4 5 5 5161 Kitson–open water 15 0 5 0 0168 Flora 1 16 5 6 0 0173 Flora 1 17 5 11 5 5

Table 1. Number of non-empty diet samples for coho salmon, sockeye salmon, Pacific herring, and surf smelt across the sampling period and sites

Fig. 1. Sampling locations according to net type used to capture fish across theSkeena River estuary. Note that vertical zooplankton tows were done con -currently at small seine net sampling events. Map inserts indicate location ofsampling region in relation to (A) the mouth of the Skeena River, and (B) the

coastline of British Columbia, Canada

Arbeider et al.: Integrating estuarine prey dynamics

enumerated each group. We used abundance, cor-rected by the size of partition, as the final variablebecause all samples were from the same depth(5 m) and, therefore, volume of water (3.9 m3).

2.3. Prey abundance across the seascape

We identified 6 prey groups to investigate theirspatial and temporal trends based on their commonoccurrence in diets of our study’s focal fish in otherstudies. Harpacticoid and calanoid copepods, Cirri-pedia cyprids, decapod zoea, pteropods, and oiko-pleurans were chosen based on their prevalence inprior diet research that we compiled (Table S1) andrepresentation in the zooplankton tows (i.e. we didnot investigate prey groups such as larval fish thatcould avoid the plankton net nor terrestrially derivedinsects which would be concentrated at surfacewaters). We tested for differences in zooplanktonabundance between sites and periods, for each taxa,using the non-parametric Kruskal-Wallis test byranks. If there was a significant difference (α = 0.05level) between groups, we used Dunn’s test to deter-mine which sites or periods were different. We deter-mined the direction of any differences in abundancegraphically.

2.4. Importance, selectivity, and variabilityof prey in diets

We used 2 indices that calculate consumption andselectivity of different prey by predators to determinewhich prey were most important and selected for bythe study fish. First, we quantified prey consumptionby the abundance, weight, and frequency at which itwas consumed for every individual fish as a metric ofprey importance to each fish species using a modifiedindex of relative importance (IRI) (Bottom & Jones1990):

(1)

We calculated an IRI score for each prey taxa ( j) foreach individual fish (i). A represents the percentabundance of prey j in fish i. B represents the percentwet weight biomass of prey j in fish i. FO is the per-cent frequency of occurrence of prey j across all individuals of a given species. The IRI metric consid-ers prey ‘importance’ as best described by both its percent abundance and percent wet weight biomasswithin a diet because abundance and weight rela-tionships are not equivalent across taxa (e.g. 1 fish

larvae may account for a high percentage of preybiomass but a low percent of abundance, while manysmall copepods may do the opposite). Multiplyingthe cumulative percent of prey abundance and bio-mass by its percent frequency of occurrence scoresrare prey lower than common prey and helps stan-dardize IRI scores across varying individuals. Thus,individuals that did not consume a certain prey wereremoved from the IRI calculation after the FO wascalculated so subsequent calculation of standarderror and mean IRI for each prey per predator speciesdo not include any zeros.

Second, we quantified prey electivity to investigatewhich food resources were appearing more often inthe diets than expected by chance. Electivity indicesare commonly used to provide inference on realizedselectivity within a given prey seascape. We usedChesson’s α-electivity index (Chesson 1978) to rankthe electivity of fish for each prey taxa ( j):

(2)

where N is the number of prey taxa considered (N =7, 9, 11, and 13 for coho salmon, sockeye salmon, her-ring, and smelt, respectively), dj/pj is the relative fre-quency ratio of the proportion of prey j in the diet (d)of an individual fish and in the plankton (p) of itsassociated site, and Σ(di/pi) is the sum of this ratio forall prey taxa included in the analysis. The neutralelectivity threshold, which suggests that a prey isbeing eaten in an equivalent proportion to what itwould be encountered at by random in the environ-ment, is defined for each predator as 1/N. Weremoved some species from analysis including preythat could readily avoid capture in the plankton net(e.g. larval fish, crab megalopa, cumaceans, isopods)or occurred in <5% of tow samples (e.g. terrestrialinsects) because they artificially inflated the electiv-ity denominator (p) due to systemic sampling error orgeneral rarity (Brodeur et al. 2011). Prey that werenot eliminated from this process but were still notpresent at some sites were assigned a p that was 1order of magnitude smaller than the smallest meas-ured p so that there were no zeros in the denomina-tor. Since zooplankton samples were only taken atthe 18 small purse seine sites during concurrent fishsampling (i.e. within 100 m of the seine at the sametime), diet samples of fish caught from large purseseine sets were matched with the nearest planktonsample in time (average ± SD: 2.7 ± 2.3 d) and space(833 ± 905 m). If multiple plankton samples weretaken within 250 m of the diet sample, we selectedthe sample closer in time.

IRIij ij ij jA B FO= + ×( ) ( )

d

pdp

i Njj

j

i

i∑α = ⎛

⎝⎞⎠ =, for 1, … ,

155

Mar Ecol Prog Ser 613: 151–169, 2019

We quantified the amount of variability in diet samples across all individuals within a fish speciesby using Schoener’s (1970) percent similarity index(PSI):

(3)

where Px,j is the percent wet weight of prey j in thestomach of individual x, Py,j is the percent wet weightof prey j in the stomach of individual y, and n is thetotal richness of prey consumed of the fish speciesconcerned.

A PSI of 100 represents complete diet overlap, andzero represents complete dissimilarity. We removedrare prey, those with a FO <5%, before analysis. Wedetermined our sample sizes were adequate to char-acterize diet composition because, once rare preywere removed, each species cumulative prey curvereached an asymptote.

We also quantified gut fullness across all individu-als within each fish species. Gut fullness is alsoreferred to as stomach fullness, an index of feedingintensity (Bottom & Jones 1990), or feeding index(Price et al. 2013), and is commonly calculated as apercentage of body weight (%BW):

(4)

where MT is the total wet weight of the fish priorstomach removal, and MP is the total wet weight ofstomach contents.

2.5. Predicting prey in diets and acrossthe seascape

We subsequently investigated patterns that couldpredict variability of important or selected preyacross individuals. We chose prey (Table S3) withhigher than average IRI and α scores per fish speciesand regressed their abundance across individualdiets (including zeros) against a suite of covariatesusing generalized linear models with mixed effects(GLMM). The covariates included fish fork length,turbidity (Secchi disk depth), water temperature, dis-tance from shore, total number of fish in set (inCPUE), and day of year. We formulated severalhypotheses about the relationships between each ofthese variables, with explanations and examples forwhy they could be negative or positive (Table S4).We included a random effect for set in each modelbecause individuals caught in the same net are con-sidered non-independent.

We used a similar approach to the diet variabilityanalysis to investigate relationships between varia-tion in prey abundance with biophysical covariates ofthe estuary seascape. We used the same 6 prey fromour prior prey analysis (Table 2): harpacticoid andcalanoid copepods, Cirripedia cyprids, decapod zoea,pteropods, and oikopleurans. We used GLMMs toregress prey abundance counts from zooplanktontows against temperature, salinity, time of tow, mainhabitat type (eelgrass, sandy bay, rocky shore, oropen water), and site distance from shore. We did notinclude turbidity because it had a Pearson correlationcoefficient of 0.7 with salinity, which is known to be astrong determinant of zooplankton distributions. Weexplored hypotheses about the relationships be -tween each of these variables, with explanations forwhy they could be negative or positive (Table S5).We included site as a random effect in all models toaccount for the expected covariation (non-indepen-dence) within sites that may be present across sam-pling time periods.

2.6. Generalized linear model specifications

GLMMs were used to provide information for ourthird and fourth questions (predicting prey in dietsand the seascape). For both the diet variability andprey variability analysis, we fit single fixed-effectGLMMs with their respective random effect due tothe limited amount of data and risk of overfitting(Babyak 2004, Hitchcock & Sober 2004). We com-pared model fits of Poisson, negative binomial 1, andnegative binomial 2, with log links using Akaike’sinformation criterion corrected for small sample size(Burnham & Anderson 2002) to determine which dis-

%BW 100 [ ( )]M M MP T P= × −

PSI 100 1 0.5, , ,P Px y x j y jj

n∑( )= − −⎡⎣ ⎤⎦

156

Prey Abundance Density Occurrencespecies

Calanoid 1630 (1563) 418 (401) 1.00copepods

Pteropoda 92 (109) 24 (28) 1.00Decapoda 14 (17) 3 (4) 0.72zoea

Oikopleura 333 (582) 85 (149) 0.92Cirripedia 282 (715) 72 (183) 0.99cyprids

Harpacticoid 45 (109) 12 (28) 0.76copepods

Table 2. Mean (SD) abundance and density (ind. m−3) persample (3.9 m3 vertical plankton tow) of select prey speciesand their frequency of occurrence across samples (n = 71)

Arbeider et al.: Integrating estuarine prey dynamics

tribution family was appropriate for each response-predictor variable combination. We fit models in R(R Core Team 2017) using the package glmmTMB(Brooks et al. 2017), which estimates parameters bymaximizing likelihood. We tested the likelihood thatcovariates had a significant effect on improvingmodel fit against an intercept-only model with a like-lihood ratio test at the α = 0.05 level. However,because we did multiple comparisons using singlecovariate models for each response variable, weincreased the probability of committing a Type I error(i.e. rejecting H0 when H0 is true) (Cabin & Mitchell2000). Thus, we applied a Bonferroni correction(Bonferroni 1936, Dunn 1961) to α of α/m, where m isthe number of likelihood ratio tests per response vari-able. However, we discuss all results even if theywere subsequently rejected by the Bonferroni cor-rection because statistical power to detect effects inbehavioural research can often be low (Jennions2003, Nakagawa 2004). All covariates were centeredand scaled (subtracted the mean from each observa-tion and divided by 1 standard deviation) so that theireffects could be comparable and to improve modelconvergence. If a covariate had a significant effect,we visually inspected their fit by examining the Pear-son’s and standardized residuals plotted against fit-ted values and checked for patterns. Subsequently,we graphically inspected the trends against real data

for biological significance and confidence around theaverage prediction.

3. RESULTS

3.1. Prey abundance across the seascape

We found differences in abundance between preygroups, across sampling locations, and between timeperiods within several prey species. Calanoid cope-pods had the highest average abundance and werepresent in every sample (Table 2), making them thenumerically dominant and most ubiquitous prey.Pteropods were the only other prey present in everysample but occurred at considerably lower averageabundance than calanoid copepods. Cirripedia cyp -rids and oikopleurans had the next highest averageabundances and frequencies of occurrence, followedby the more sporadically distributed harpacticoidcopepods and decapod zoea. Three of the 6 specieswe tested for variability in abundance (calanoidcopepods, pteropods, and decapod zoea) had statisti-cally different medians at the α = 0.05 level betweensites (Fig. 2). Differences between sites could suggestthat they are not entirely independent and that somesites are possibly hotspots within the seascape forthese species (Figs. 2 & 3). No single site appeared to

157

5

6

7

8

9

1

2

3

4

5

6

0

1

2

3

4

0

2

4

6

8

140 150 160 170

0

2

4

6

8

140 150 160 170

0

2

4

6

140 150 160 170

A** B** C**

D* E* F

ln(a

bund

ance

+ 1

)

Day of the year

Fig. 2. Spatial and temporal patterns of abundance of (A) Calanoida, (B) Pteropoda, (C) Decapoda zoea, (D) Oikopleura, (E)Cirripedia cypris, and (F) Harpacticoida. Points connected by lines are samples from the same site within the Skeena estuary.Note that not all y-axes are equivalent and contain broken axes. Letters denoted by ** and * had significant site or sampling

period differences, respectively, as detected by Kruskal Wallis tests with p < 0.05

Mar Ecol Prog Ser 613: 151–169, 2019

have consistently higher abundance of all species,but some sites had consistently below average abun-dance, which could represent an area with low preyavailability for our focal predators. Two prey taxa,Cirripedia cyprids and oikopleurans, had statisticallydifferent medians between periods (Fig. 2). OnlyPeriod 2 (13−20 May) had a higher median than theother periods for Cirripedia cyprids, whereas Peri-ods 2 and 4 (20−24 June) had higher medians thanPeriods 1 and 3 for oikopleurans (Fig. 2). Becauseonly 2 of 6 prey showed differences between sam-pling periods and not in a consistent manner, we sug-gest that Period may not have as an important effectas Site, and thus we did not include Period as a ran-dom effect in subsequent models. Harpacticoid cope-pods did not have either temporal or site-specific dif-ferences in their abundance.

3.2. Importance, selectivity, and variabilityof prey in diets

We observed large variation in importance (IRI)and electivity (α) scores across individual fish withina species. There was considerable variability withinthe highest mean IRI scoring prey items for each fishspecies, with the values often ranging from ~0 to 1order of magnitude larger than the mean. In the dis-cussion, we refer to prey that have >2-fold the aver-age IRI score as primary prey and those around theaverage as secondary prey.

For juvenile coho salmon Oncorhynchus kisutch,terrestrial-based insects and larval/juvenile fisheshad the highest mean IRI scores, 2.4- and 2.1-foldhigher than the third-highest prey, respectively(Fig. 4A). Insects were primarily Diptera (64% by

158

Fig. 3. Spatial distribution and relative mean abundance across all time periods of (A) Calanoida, (B) Pteropoda, (C) Decapodazoea, (D) Oikopleura, (E) Cirripedia cypris, and (F) Harpacticoida across sampling sites in the Skeena River estuary. Bubblesize is scaled by the relative abundance within a species, i.e. the size of bubbles is not comparable between species, onlywithin. Large bubbles of species that had site level differences (A–C) could represent hotspots for those prey. White is

water; grey is land

Arbeider et al.: Integrating estuarine prey dynamics 159

Amphipoda

Cirripedia cyprisDecapoda megalopa

CumaceaE. furcilia

Teleostei juvenile

PteropodaHarpacticoida

Insecta

Isopoda

Calanoida

Decapoda zoea

0 2000 4000 6000 0.0 0.2 0.4 0.6

Amphipoda

Cirripedia cypris

Decapoda megalopa

Cirripedia nauplius

CladoceraCumacea

Pteropoda

Harpacticoida

InsectaOikopleura

Calanoida

Decapoda zoea

0 2000 4000 6000 0.0 0.1 0.2 0.3 0.4

Amphipoda

TunicateCirripedia cypris

Decapoda megalopa

Cirripedia nauplius

Cladocera

E. calyptopsis

E. furcilia

Teleostei egg

Pteropoda

Harpacticoida

Oikopleura

Calanoida

Decapoda zoea

0 3000 6000 9000 0.0 0.1 0.2 0.3

Amphipoda

Tunicate

Cirripedia cypris

Decapoda megalopa

Cirripedia nauplius

Cladocera

E. calyptopsis

E. furcilia

Teleostei juvenile

Pteropoda

HarpacticoidaOikopleura

Polychaeta

Calanoida

Decapoda zoea

0 5000 10000 15000

Index of relative importance0.0 0.1 0.2 0.3 0.4 0.5

Chesson's alpha score

A

B

C

D

E

F

G

H

Coho

Sockeye

Herring

Sm

elt

Fig. 4. Index of relative importance (IRI) and Chesson’s electivity scores of each prey species with standard error for (A,E) cohosalmon, (B,F) sockeye salmon, (C,G) Pacific herring, and (D,H) surf smelt. Dotted lines represent the overall average IRI score foreach fish species (A−D) and the neutral alpha selectivity threshold for each fish species (E−H). E.: Euphausiacea. Alpha scoresabove the neutral selectivity threshold suggest that a species is represented in the diet more than it is represented in the envi-ronment (i.e. proportionally higher in the diet), whereas values below the line suggest that the prey is proportionally higher inthe environment. Note the reduced diversity of prey presented with Chesson’s alpha because poorly sampled prey were re-moved. Only the prey that occurred in >5% of diets are presented for both indices for coho salmon, sockeye salmon, and Pacific

herring, with a 10% occurrence cut-off for surf smelt because of the higher diversity of prey consumed in small amounts

Mar Ecol Prog Ser 613: 151–169, 2019

abundance), followed by Hemiptera (26%), Cole -optera (6%), and others (Collembola, Hymenoptera,Trichoptera, and Ephemeroptera). Only 23% of juve-nile and larval fish were identified to family or lower,which were either Pleuronectidae (86%) or Pacificherring Clupea pallasii (14%). Decapod zoea (pri-marily from infraorder Brachyura), harpacticoid cope -pods, gastropods (Limacina pteropods when identifi-able to genera), and amphipods also had higher thanaverage mean IRI scores. Amphipods were primarilyGammaridea (83%) with a notable 47% of Gam-maridea being a high-intertidal family, Talitridae.The remaining amphipods (17% by abundance)were Hyperiidae. Coho salmon had the highest electivity for decapod zoea followed by amphipods(Fig. 4E). The mean harpacticoid copepod electivityscore was slightly below the neutral selectivitythreshold. All other prey groups scored below theneutral selectivity threshold, i.e. were consumed atlower frequencies than their relative frequencies inthe environment.

In juvenile sockeye salmon O. nerka, harpacticoidcopepods had the highest mean IRI score, 2-foldhigher than their next highest ranked prey (Fig. 4B).Calanoid copepods, Cirripedia cyprids, gastropods(Limacina pteropods when identifiable to genera),and terrestrial-based insects (majority Diptera andHemiptera) also had higher than average mean IRIscores. It is interesting to note that 5 individuals alsoconsumed adult stages of the salmonid parasite fromthe family Caligidae, which has also been observedin southern British Columbia (Price et al. 2013). Juve-nile sockeye salmon had the highest electivity forharpacticoid copepods followed by decapod zoea,Cirripedia cyprids, and amphipods (Fig. 4F).

Calanoid copepods had the highest mean IRIscores for both adult Pacific herring and adult surfsmelt Hypomesus pretiosus, 4- and 5.5-fold higherthan their next highest ranked prey, respectively, foreach fish species (Fig. 4C,D). Ascidian tunicates andCirripedia cyprids were the only other Pacific herringprey whose mean IRI scores werehigher than the overall mean IRI. Her-ring had the highest electivity fordecapod zoea followed by Cirripediacyprids, calanoid copepods, and gas-tropods with nearly neutrally electivityfor euphausiid calyptopsis (Fig. 4G).Gastropods, all of which were uniden-tifiable beyond Gastropoda, were theonly other surf smelt prey whose meanIRI score was above average. Smelthad the highest electivity for Hyperi-

idea amphipods followed by decapod zoea and hadneutral electivity for oikopleurans, ephausiid calyp-topsis and furcilia, gastropods, and calanoid cope-pods (Fig. 4H). Although both herring and smelt hada high electivity value for decapod zoea, it was onlyfound in 47 and 51% of individuals, respectively, andzoea were not consumed in large quantities. Highelectivity and generally low presence in the dietcould mean that when zoea were encountered,despite their rarity, they were opportunistically tar-geted by both predators.

We found that diets and gut fullness within eachspecies were highly variable with low diet overlapbetween individuals. The mean PSI values across allindividuals within a species were all <50 on averageexcept for surf smelt (Table 3), meaning that dietswere frequently >50% different across individuals ofthe same species. When we examined PSI pair-wisecomparisons that were done between individualsfrom the same set (still of the same species), we foundthat the average PSI values increased, but 2 speciesstill had PSI values below 50%: coho salmon andPacific herring. Average percent gut fullness wascomparable to other studies on coho and sockeyesalmon in estuary and nearshore environments if notslightly higher (Bottom & Jones 1990, Healey 1991,Brodeur et al. 2007a, Price et al. 2013). In contrast,surf smelt and Pacific herring had less than half thegut fullness of both salmonids (Table 3).

3.3. Predicting prey in diets and acrossthe seascape

We found 3 relationships between prey diet abun-dance and several biophysical factors that hadparameter estimates that were statistically differentthan a null model by using likelihood ratio tests at theBonferroni corrected α = 0.05 level, and 8 addi -tional relationships when α was uncorrected (Fig. S1,Table S6 in the Supplement). The abundance of lar-

160

Fish species Mean Mean set Mean gut FL rangePSI (SE) PSI (SE) fullness (SD) (mean) (mm)

Coho salmon 20.2 (0.85) 26.9 (3.92) 0.91 (0.86) 84−136 (102)Sockeye salmon 23.7 (0.36) 50.6 (4.70) 0.83 (1.11) 59−109 (82)Pacific herring 32.6 (0.79) 41.6 (6.58) 0.38 (0.73) 68−168 (125)Surf smelt 52.3 (1.30) 57.1 (7.53) 0.34 (0.34) 106−168 (134)

Table 3. Average percent similarity index (PSI) values with standard error foreach fish species between all individuals and between only individuals fromthe same sets, mean gut fullness (% body weight) with standard deviation, and

fork length (FL) range

Arbeider et al.: Integrating estuarine prey dynamics

val fish in coho salmon diets decreased with increas-ing distances from shore (Fig. S1A). Insect abun-dance in coho salmon diets decreased with increas-ing temperature (Fig. S1B), possibly due to increasedwater volume (flow) from the river and increaseddelivery of upriver insects. Day of year was positivelycorrelated with harpacticoid copepod abundance insockeye salmon diets (Fig. S1G). The following 8relationships were only significant at the uncorrectedα = 0.05 level. Increased total set CPUE (total abun-dance) decreased the number of insects in cohosalmon diets (Fig. S1C) and the number of calanoidcopepods in Pacific herring diets (Fig. S1I), possiblybecause of exploitative competition. Insect and deca-pod zoea abundance in coho salmon diets increasedwith Secchi depth (Fig. S1D,E), possibly due toincreased ability of fish to see insects stranded in sur-face waters and decapods within the water column inclearer water. Decapod zoea abundance in cohosalmon diets also increased with coho salmon length(Fig. S1F), but this relationship appears to be drivenby outlying data points. Water temperature was pos-

itively correlated with harpacticoid abundance insockeye salmon diets (Fig. S1H). However, day ofyear and temperature were highly correlated for setswith sockeye salmon samples (Pearson’s correlationcoefficient = 0.65) and day of year fit better uponvisual inspection. Calanoid copepod abundance inPacific herring diets increased with distance awayfrom shore (Fig. S1J), but the confidence intervals onthis relationship are particularly large. Calanoidcopepod abundance in surf smelt diets decreasedwith fish length (Fig. S1K), a possible indicator thatlarger individuals were targeting a different prey forconsumption.

We found 3 relationships between prey abundancein the environment with biophysical factors that hadparameter estimates statistically different than zeroat the Bonferroni-corrected α = 0.05 level and 3 addi-tional relationships when α was uncorrected (Fig. 5,Table S7); however, each relationship’s biologicalsignificance is variable. The following 3 relationshipswere statistically significant after the Bonferroni cor-rection. Calanoid copepod and oikopleuran abun-

161

Fig. 5. Predicted relationships between the abundance of several important prey in the environment and biophysical aspectsof the seascape within the Skeena estuary. Each light grey point is a site’s measurement across sampling periods. The darkmiddle lines and the white triangles in (F) are the average model predictions. The outer gray lines and whiskers are the 95%confidence intervals calculated from the respective GLMMs. (A−C) are statistically significant at α < 0.05 after the Bonferroni

correction. (D−F) are only statistically significant before the Bonferroni correction

Mar Ecol Prog Ser 613: 151–169, 2019

dance were positively correlated with salinity, possi-bly driven by these species’ natural salinity toler-ances. Both relationships with salinity appeared tohave a significant biological effect (Fig. 5A,B).Pteropoda abundance was negatively correlatedwith temperature, but there was quite a bit of vari-ability in abundance around the middle tempera-tures (Fig. 5C). The following 3 relationships wereonly statistically significant when α was uncorrected.Pteropoda abundance was positively correlated withsalinity (Fig. 5D). Cirripedia cyprid abundance wasnegatively correlated with temperature (Fig. 5E).Harpacticoid copepods were more abundant on aver-age at sites with eelgrass substrate compared to sitesover open water or close to rocky shores and had aslight tendency to be more abundant than sites thatwere sandy bays (Fig. 5F). No biophysical factorspredicting decapoda zoea abundance had statisticalsupport.

4. DISCUSSION

Estuaries are particularly valued as key nurseryhabitats for a variety of fish species (Beck et al. 2001),where the dynamics of prey resources are an integralcomponent of the nursery function (Sheaves et al.2015). Here we found that 4 co-occurring estuary fishspecies relied on multiple prey that were dynamicacross space and time. Our results also highlight highvariability in diet contents within a small region andeven between fish of the same species from the sameseine set, whereas prior research often contrasts dietsover seasons, years, or regions (Simenstad et al. 1982,Brodeur et al. 2007b, Hill et al. 2015). Biophysical fac-tors predicted some of the variability in fish diets andprey in the environment. Thus, we provide rare em -pirical evidence for the spatio–temporal dynamics ofprey and how predators integrate across them withina major estuary. The spatio–temporal prey mosaichas been previously suggested as a critical but under-studied dimension of the role of estuaries as impor-tant refuges and nursery habitats (Nagelkerken et al.2015, Sheaves et al. 2015).

4.1. Prey distribution across space and time

Our study fills recognized knowledge gaps for theSkeena River estuary (Pickard et al. 2015) by further-ing our understanding about how prey are distrib-uted during the period of highest juvenile salmonabundance (Carr-Harris et al. 2015, Sharpe 2017).

Prey abundance and distribution can determine theiravailability to predators (Griffiths 1973, 1975), so it isimportant to understand these features of the preymosaic when considering how predators integratewith prey. Calanoid copepods showed consistent dif-ferences in abundance between sites, but becausethey were the most abundant and ubiquitous zoo-plankton prey, the sites with low abundance still hadhigher abundance than most other prey groups.Calanoid copepods’ relatively high abundance hasthe potential to make them one of the most availableprey (Griffiths 1975). Cirripedia cyprids and oiko-pleurans were present in moderate abundance andshowed different temporal patterns across sites, withCirripedia cyprids having a single temporal peak inabundance while oikopleurans had two. Peaks inabundance could be interpreted as differing bloomphenologies between these 2 groups and could affecttheir availability by matching or mismatching pre -dators’ estuary timing (Cushing 1990). Decapod zoeaand harpacticoid copepods had the overall lowestaverage abundance and sporadic distributions, whichcould increase the search intervals (or de creaseencounter rates) for this patchier prey (Sims et al.2008). We did not effectively sample for larval fish orterrestrial insects (Brodeur et al. 2011), which arealso known to be common prey (Table S1 in the Sup-plement). But overall, the prey field in the SkeenaRiver estuary appears to be saturated by calanoidcopepods with temporally variable abundances ofCirripedia cyprids and oikopleurans and low andpatchy abundance of decapod zoea and harpacticoidcopepods.

4.2. Important and selected prey

Our study also addresses a gap in knowledge forjuvenile coho Oncorhynchus kisutch and sockeyesalmon O. nerka diets in the Skeena River estuary(Pickard et al. 2015), as well as for the co-occurringand highly abundant small pelagic fishes in the area(Sharpe 2017): Pacific herring Clupea pallasii andsurf smelt Hypomesus pretiosus. Despite the ob -served diet heterogeneity within sets, each speciesoften consumed 1 or 2 prey most often and in highabundance or weight, here referred to as primaryprey, followed by a few secondary prey that wereconsumed at a magnitude more than all remainingprey. Primary prey for coho salmon were insects andlarval fish, which is consistent with prior research inBritish Columbia (Manzer 1969, Osgood 2016). Al -though we did not include insects and larval fish in

162

Arbeider et al.: Integrating estuarine prey dynamics

the electivity analysis, their ubiquity in diets acrossother regions along the west coast of North America(Table S1) and relatively high energy density to otherprey (Duffy et al. 2010) leads us to presume theywere likely selected for. Coho salmon also selectedfor decapod zoea and gammarid amphipods in theestuary environment. Harpacticoid copepods werethe most important and selected for prey by sockeyesalmon, followed by decapod zoea, amphipods, andCirripedia cyprids. We believe that this research isthe first record of juvenile sockeye salmon primarilyforaging on harpacticoid copepods in estuaries.Harpacticoid copepods were also an important sec-ondary prey for coho salmon in our estuary and are aprimary prey in other estuaries for coho as well as forchum O. keta, pink O. gorbusha, and ocean-typeChinook O. tshawytscha salmon (Healey 1979, 1980,Godin 1981, Simenstad et al. 1982, Macdonald et al.1987, Northcote et al. 2007). By contrast, the primaryprey for both Pacific herring and surf smelt werecalanoid copepods. Whereas herring selected forcalanoid copepods, Cirripedia cyprids, and decapodzoea, surf smelt only selected for amphipods anddecapod zoea, with neutral affinity for calanoid cope-pods. All 4 species consumed pteropods as a second-ary prey and selected for decapod zoea, and all butPacific herring selected for amphipods, suggestingthat these prey could be an energetically desirableor easily caught prey across predator taxa (Emlen1966).

The drivers of the difference in the primary cope-pod prey between salmonid and small pelagic fishin this study can be examined in the context of onto-genetic niche theory (Werner & Gilliam 1984) andby how prey activity can affect its availability topredators (Griffiths 1973). Small pelagic fish fedheavily on calanoid copepods, whereas juvenilesalmonids relied more on harpacticoid copepods.Harpacticoid copepods are generally more seden-tary than calanoid species because they are epiben-thic and phytal, primarily feeding on epiphytic andmacroalgae as well as detritus, bacteria, and fungi(Chandler & Fleeger 1987, Steinarsdóttir et al.2010), whereas calanoid copepods primarily feedactively in the water column (Mauchline 1998). Inaddition, harpacticoid copepods have slower burstspeeds, relative to species of calanoid copepodswho are known for their evasive behaviour (Leising& Yen 1997, Buskey et al. 2002). Juvenile sockeyesalmon are facultative planktivores, primarily tar-geting and consuming one prey item at a time (Laz-zaro 1987), and are known to select for slower,larger prey in lakes when it is available (Eggers

1982). This is consistent with our observation thatjuvenile sockeye salmon selected for the slower har -pacticoid copepods while consuming calanoid cope-pods less than their relative availability (with re -gards to abundance) in the environment. Even incoastal environments where harpacticoid copepodsare often not found in the water column as an alter-native prey, juvenile sockeye salmon eat calanoidcopepods at proportions near or less than theiravailability in the environment (Price et al. 2013),possibly due to the difficulty to capture them. Incontrast, Pacific herring consumed calanoid cope-pods selectively, and surf smelt consumed themwith neutral selectivity (i.e. neither selected for noravoided). Pacific herring and surf smelt may be bet-ter adapted to handle the quick, abundant, andpelagic calanoid copepods because they can createstrong suction using their round mouths and buccalcavities, and even filter-feed at high prey densities(Gibson & Ezzi 1985, Lazzaro 1987). The physiologi-cal adaptations of small pelagic fish are likelyhighly selected for because they spend much oftheir ontogeny at sizes relatively close to their fullsize, whereas juvenile salmon quickly grow tolarger sizes and can use different foraging tactics ondifferent types of prey (Werner & Gilliam 1984, Dalyet al. 2009, Duffy et al. 2010). This contrast in forag-ing patterns between juvenile salmon and smallpelagic species illuminates key differences in howthese fish species may integrate with the prey layerof estuarine seascapes.

Differences within the diets of small pelagic fishmay suggest subtle behavioural differences thatdrive how they integrate with their prey in estuariestoo. We found Pacific herring selected Cirripediacyprids and barely consumed hyperiid amphipods,whereas surf smelt highly selected for hyperiidamphipods and consumed Cirripedia cyprids at rela-tively low quantities. Hyperiid amphipods and Cirri-pedia cyprids could be distinguishable to herring andsmelt as they have considerably different morpholo-gies, colour and refraction, and swimming patterns(Giske et al. 1994). The differences in foraging pat-terns between these 2 pelagic fish could be the resultof differences in visual capabilities or light attenua-tion of prey that makes one more discernable thanthe other (Giske et al. 1994), or the distribution ofpredators and prey in the water column is such thatthey overlap/encounter the prey more frequently(Eggers 1977), or the differences could relate to dif-ferences in feeding morphology (Labropoulou &Eleftheriou 1997) or some type of other learned pref-erence (Brown & Laland 2003).

163

Mar Ecol Prog Ser 613: 151–169, 2019

4.3. Biophysical factors and diet variability

Variability of prey abundance in diet samples foreach predator was linked to some biophysical fac-tors. We suggest that abiotic environmental factorscould affect how predators integrate with their preyacross the seascape in addition to inherent physio-logical constraints of the predators. We found thatincreasing Secchi depth, our index for water clarity,increased the abundance of 2 important prey incoho salmon diet samples: decapod zoea and terres-trial insects. Less turbid conditions may increasecapture success of live decapod zoea (Berg & North-cote 1985, Gregory & Northcote 1993) or increasethe line of sight to surface waters where expired ornon-evasive terrestrial insects concentrate (Tschap-linski 1987). Insect prey in coho salmon diets alsoincreased with colder water temperatures. As rivertemperatures were cooler than ocean temperaturesduring our sampling period, cool surface tempera-tures may correlate with increased riverine preysubsidies that are flushed down from upstream intocertain areas (Tschaplinski 1987). Finally, calanoidcopepod abundance decreased in Pacific herringdiets as the set total CPUE of all fish increased.Since calanoid copepods were eaten by herring,smelt, and sockeye salmon in this study, this inverserelationship may suggest per capita calanoid cope-pod consumption rates decrease when the combinedabundance of these multiple predators is high (Arditi& Ginzburg 1989, Sih et al. 1998).

We found that biophysical characteristics did notexplain variation of calanoid copepod abundance insurf smelt diets nor of Cirripedia cyprids in sockeyesalmon and Pacific herring diets. The lack of statisti-cal or biologically relevant relationships betweenthese predator–prey pairs suggests that they are notaffected by biophysical processes or that we did notidentify the correct process. We also must acknowl-edge the possibilities that processes such as sto-chastic variation and insufficient sample size couldaffect our observed results. Furthermore, prey mayhave been consumed elsewhere and thus notdirectly related to the biophysical factors from thesite. However, the effects from biophysical factorswe detected, as well as others, influence how differ-ent predators integrate with their prey across theestuary seascape in time and space.

We found an additional seasonal pattern of har -pacticoid copepod abundance in juvenile salmondiets that was not linked directly to any biophysicalvariables. The abundance of harpacticoid copepodsin sockeye salmon diet samples increased with day of

year. In theory, zooplankton production and abun-dance may increase over time in the spring and sum-mer towards a seasonal maxima (Mackas et al. 2012);however, we did not find a relationship withharpacticoid abundance across time in our study. Thelack of an increasing trend of harpacticoid abun-dance in the environment could be explained by top-down control of predators in areas with high produc-tion of prey (Rudstam et al. 1994, Yang et al. 2008).Harpacticoid copepod production may have beenincreasing, but their abundance did not because ofhigh predation rates by juvenile sockeye salmon(Sibert 1979). Interestingly, the prevalence of harpac -ticoid copepods increased across time within cohosalmon diet samples also (relationship not statisti-cally significant), and harpacticoid copepods wereconsumed above average compared to other preytypes. Thus, it may be possible for juvenile salmon tohave collectively consumed harpacticoid copepods inthe environment at a rate equivalent (or evengreater) to their production as a form of top-downcontrol (Healey 1979, Sibert 1979, Godin 1981, Fuji-wara & Highsmith 1997).

4.4. Biophysical factors and plankton abundance

Although we found no biophysical predictors ofharpacticoid copepod abundance in diets, we didfind that they were more abundant over eelgrasshabitats than other habitats in the seascape. Site-spe-cific hotspots of harpacticoid abundance were notdiscernable across time. Yet at sites where eelgrasswas present, mean abundance in the water columnwas higher than at sites beside rocky shoreline orover open water. Eelgrass is known to support higherdensities of harpacticoid copepods and is likely apopulation source (Hosack et al. 2006, Kennedy et al.2018). Multiple studies have shown that juvenilesalmon are capable of consuming large proportionsof total harpacticoid production (Healey 1979, Godin1981, Fujiwara & Highsmith 1997). Therefore, degra-dation of eelgrass may affect prey productivity andcould affect salmon foraging behaviour and poten-tially survival. In our system, sockeye and cohosalmon abundances were consistently highest in theregion of a particularly large eelgrass bed, FloraBank (Carr-Harris et al. 2015, Sharpe 2017). We spec-ulate that the Flora Bank eelgrass habitat may be animportant source of harpacticoid copepods for theseyoung salmon. Eelgrass habitat has previously beenidentified as a conservation priority because of itsrole as a productive food source for multiple juvenile

164

Arbeider et al.: Integrating estuarine prey dynamics

fish species in other estuaries (McDevitt-Irwin et al.2016). The Flora Bank region was the proposed loca-tion of major industrial developments (Moore et al.2015, CEAA 2016), and previous assessment reportsidentified information on food web-habitat connec-tions as ‘high prioritization’ data gaps (Pickard et al.2015). Establishing these food-web habitat connec-tions in areas poised for development is an impor-tant step in identifying the potential habitat value ofestuarine seascapes to species of interest. The nextsteps are to identify species-specific predator–preyresponses to possible impacts on habitats involved insupporting the estuary prey mosaic, such as eelgrassbed fragmentation or reductions in shoot density(Lannin & Hovel 2011, Ljungberg et al. 2013, Chacin& Stallings 2016).

The abundances of other zooplankton prey withinthe estuarine seascape were related to biophysicalprocesses through space and time in multiple ways.Calanoid copepods, pteropods and decapod larvaeshowed site level consistencies in their abundanceover time, which could indicate that certain locationsare acting as prey hotspots. The best fit predictor ofvariability in calanoid copepod abundance was salin-ity, a common gradient in estuaries and driver of zoo-plankton distributions (Telesh & Khlebovich 2010).Increasing salinity was correlated with increases inpteropods, a secondary prey for all of our predators,and oikopleurans, a prey that was marginally con-sumed by sockeye salmon in this study but is oftenfound in the diets of other juvenile salmon (Manzer1969, Landigham et al. 1998, Brodeur et al. 2007b).Although the salinity gradient and abundance pat-terns of calanoid copepods and pteropods were asso-ciated with sites, there was still variability within thesalinity gradient at sites across time. Oikopleuransdid not show any site-level persistence in abundancepatterns despite being correlated with salinity. Wesuggest that this prey group is responding todynamic environment forcing rather than being stat-ically abundant in a specific location. Learning howprey are influenced by biophysical processes, likesalinity (Telesh & Khlebovich 2010), or habitat fea-tures like eelgrass patches (Lannin & Hovel 2011,Ljungberg et al. 2013) is an integral layer of under-standing the prey mosaic of estuaries.

5. CONCLUSION

Here we integrated understanding of the spatialand temporal dynamics of zooplankton and their consumption by 4 species of fishes, but it is important

to consider potential limitations of our study. We discovered that juvenile coho salmon Oncorhynchuskisutch primarily consumed terrestrial insects andfish; however, because these prey items are not ade-quately sampled by vertical plankton tows with themesh size we used (Brodeur et al. 2011), we couldnot assess the abundance of these important preysources across space or time. Our inference onharpacticoid copepod abundance is also based onwhat we sampled in the water column, yet they areassociated with meiobenthic and epiphytic habitat(Alheit & Scheibel 1982, Steinarsdóttir et al. 2010).Juvenile salmon are more likely feeding in the watercolumn (Clark & Levy 1988) and less so directly fromsubstrate or blades of eelgrass. Thus, we believe thatour zooplankton sampling likely represented relativedensities that these fish might be encountering butlacks the ability to properly identify epibenthic zoo-plankton population sources, like that of harpacticoidcopepods. With the example of harpacticoid copepodabundance in the environment, it is also difficult totease apart the effects of bottom-up biophysical pro-cesses or phenology of zooplankton and top-downeffects from predation. Furthermore, diet samplesonly represent a snapshot of what an individual fishwas eating and only from locations where fish werepresent at the time of sampling. Isotope and fatty acidanalyses could provide additional longer-term per-spectives on diet trends (Daly et al. 2010, Selleslaghet al. 2015), but we believe that our study capturesvariability in diets and provides a picture of primaryand secondary prey types consumed in the SkeenaRiver estuary. Last, when calculating Chesson’s al -pha, we assumed that diet snapshots were represen-tative of the site where samples were taken from, butthe duration required to travel between sites by fishis less than that of egestion (Brett & Glass 1973,Brodeur & Pearcy 1987). However, we comparedChesson’s alpha results from spatially averaged zoo-plankton abundances and found only minor differ-ences. Thus, our study has important limitations butalso contributes to the relatively understudied fieldsof the prey basis of nursery function in estuaries(Sheaves et al. 2015).

Collectively, our study highlights how 4 culturally,economically, and ecologically important fish speciesintegrate with prey differently across the dynamicseascape of a major estuary. We found that few bio-physical factors covaried with herring Clupea pallasiiand smelt Hypomesus pretiosus diets other than totalCPUE, suggesting that their diets may be influencedmost by the number of fish present and less so by abi-otic conditions. In contrast, multiple abiotic variables

165

Mar Ecol Prog Ser 613: 151–169, 2019

covaried with the abundance of certain prey in juve-nile coho salmon diets, suggesting that there mightbe stronger impacts on their foraging success withabiotic changes. In addition to integrating with preyacross the biophysical dynamics of the seascape,populations of juvenile salmon enter the SkeenaRiver estuary at a diversity of times and may interactwith different peaks of zooplankton abundance indifferent seascape conditions (Carr Harris et al.2018). Thus, different populations’ diets may beaffected differently depending on when they enterthe estuary.

The spatial and temporal asynchronies in differentprey abundances and the ubiquity and abundance ofothers within the Skeena River estuary may provideextended and buffered foraging opportunities for co-occurring mobile consumers. Juvenile salmon dietsmay benefit from a diverse prey portfolio that buffersthem from fluctuations in a single prey item or allowsthem to capitalize on easily captured prey (Arm-strong et al. 2016). Harpacticoid copepods are oneexample of a non-evasive prey and could occur inadequate abundances in patches, such as over eel-grass habitats (Kennedy et al. 2018), to supportsalmon. Pacific herring and surf smelt appear moreadapted to forage on the most abundant prey group(Hill et al. 2015), the highly evasive but ubiquitouscalanoid copepods in this system. However, calanoidcopepod abundance is correlated with salinity, andcalanoid copepod distribution may change if riverflow changes the salinity gradient with climatechange or anthropogenic development (Sherwood etal. 1990, Nohara et al. 2006). Our work adds to thegrowing appreciation that estuary seascapes havedynamic and complex prey mosaics that underpintheir function as nursery and foraging habitats(Nagelkerken et al. 2015, Sheaves et al. 2015)and may be affected through multiple biophysicalprocesses.

Acknowledgements. This research is part of a collaborationbetween Lax Kw’alaams First Nation, Skeena FisheriesCommission, and Simon Fraser University. Any opinions,findings, conclusions, or recommendations expressed in thispaper are those of the authors and do not necessarily reflectthe views of the partner organizations. We sincerely thankBill Shepert, Jen Gordon, John Latimer, Harvey James Russell, Wade Helin, Jim Henry Jr., Devin Helin, KatelynCooper, and Brandon Ryan from Lax Kw’alaams FisheriesDepartment along with David Doolan from Metlakatla Fish-eries Program for providing field and logistical support. Wealso thank Biologica Environmental Services Ltd. for theirthorough identification, enumeration, and weighing of thediet samples. We are grateful to many members of the Earthto Oceans Research Group and Salmon Watersheds Lab for

advice and inspiration. This research was supported byLiber Ero Foundation, Natural Sciences and EngineeringResearch Council, and the Coast Opportunities Fund.

LITERATURE CITED

Ajmani AM (2011) The growth and diet composition of sock-eye salmon smolts in Rivers Inlet, British Columbia. Uni-versity of British Columbia, Vancouver

Alheit J, Scheibel W (1982) Benthic harpacticoids as a foodsource for fish. Mar Biol 70: 141−147

Arditi R, Ginzburg LR (1989) Coupling in predator–preydynamics: ratio-dependence. J Theor Biol 139: 311−326

Armstrong JB, Takimoto G, Schindler DE, Hayes MM,Kauffman MJ (2016) Resource waves: phenologicaldiversity enhances foraging opportunities for mobileconsumers. Ecology 97: 1099−1112

Babyak MA (2004) What you see may not be what you get: abrief, nontechnical introduction to overfitting in regres-sion-type models. Psychosom Med 66: 411−421

Barry JP, Dayton PK (1991) Physical heterogeneity and theorganization of marine communities. In: Kolasa J, PickettSTA (eds) Ecological heterogeneity. Springer, New York,NY, p 270−320

Beck MW, Heck KL, Able KW, Childers DL and others(2001) The identification, conservation, and manage-ment of estuarine and marine nurseries for fish andinvertebrates. Bioscience 51: 633−641

Berg L, Northcote TG (1985) Changes in territorial, gill-flar-ing, and feeding behavior in juvenile coho salmon(Oncorhynchus kisutch) following short-term pulses ofsuspended sediment. Can J Fish Aquat Sci 42: 1410−1417

Birtwell IK, Nassichuk MD, Beune H (1987) Underyearlingsockeye salmon (Oncorhynchus nerka) in the estuary ofthe Fraser River. In: Smith HDL, Margolis L, Wood CC(eds) Sockeye salmon (Onchorhynchus nerka) in the es -tuary of the Fraser River. Fisheries and Oceans Canada,Ottawa, p 25–35

Bollens SM, Hooff RV, Butler M, Cordell JR, Frost BW (2010)Feeding ecology of juvenile pacific salmon (Oncorhyn-chus spp.) in a northeast Pacific fjord: diet, availability ofzooplankton, selectivity for prey, and potential competi-tion for prey resources. Fish Bull 108: 393−407

Bonferroni C (1936) Teoria statistica delle classi e calcolodelle probabilita. Pubbl R Ist Super Sci Econ Commeri-ciali Fir 8: 3−62

Boström C, Pittman SJ, Simenstad C, Kneib RT (2011) Sea-scape ecology of coastal biogenic habitats: advances,gaps, and challenges. Mar Ecol Prog Ser 427: 191−217

Bottom DL, Jones KK (1990) Species composition, distribu-tion, and invertebrate prey of fish assemblages in theColumbia River estuary. Prog Oceanogr 25: 243−270

Brett JR, Glass NR (1973) Metabolic rates and critical swim-ming speeds of sockeye salmon (Oncorhynchus nerka) inrelation to size and temperature. J Fish Res Board Can30: 379−387

Brodeur RD (1991) Ontogenetic variations in the type andsize of prey consumed by juvenile coho, Oncorhynchuskisutch, and Chinook, O. tshawytscha, salmon. EnvironBiol Fishes 30: 303−315

Brodeur RD, Pearcy WG (1987) Diel feeding chronology,gastric evacuation and estimated daily ration of juvenilecoho salmon, Oncorhynchus kisutch (Walbaum), in thecoastal marine environment. J Fish Biol 31: 465−477

166

Arbeider et al.: Integrating estuarine prey dynamics

Brodeur RD, Daly EA, Schabetsberger RA, Mier KL (2007a)Interannual and interdecadal variability in juvenile cohosalmon (Oncorhynchus kisutch) diets in relation to envi-ronmental changes in the northern California Current.Fish Oceanogr 16: 395−408

Brodeur RD, Daly EA, Sturdevant MV, Miller TW and others(2007b) Regional comparisons of juvenile salmon feedingin coastal marine waters off the west coast of NorthAmerica. Am Fish Soc Symp 57: 183−203

Brodeur RD, Daly EA, Benkwitt CE, Morgan CA, Emmett RL(2011) Catching the prey: sampling juvenile fish andinvertebrate prey fields of juvenile coho and Chinooksalmon during their early marine residence. Fish Res108: 65−73

Brooks ME, Kristensen K, van Benthem KJ, Magnusson Aand others (2017) glmmTMB balances speed and flex -ibility among packages for zero-inflated generalized linear mixed modeling. The R Journal 9:378–400

Brown C, Laland KN (2003) Social learning in fishes: areview. Fish Fish 4: 280−288

Burnham KP, Anderson DR (2002) Model selection andmulti-model inference: a practical information-theoreticapproach, 2nd edn. Springer, New York, NY

Buskey EJ, Lenz PH, Hartline DK (2002) Escape behavior ofplanktonic copepods in response to hydrodynamic dis-turbances: high speed video analysis. Mar Ecol Prog Ser235: 135−146

Cabin RJ, Mitchell RJ (2000) To Bonferroni or not to Bonfer-roni: When and how are the questions. Bull Ecol Soc Am81: 246−248

Carr-Harris C, Gottesfeld AS, Moore JW (2015) Juvenilesalmon usage of the Skeena River estuary. PLOS ONE10: e0118988

Carr Harris CN, Moore JW, Gottesfeld AS, Gordon JA andothers (2018) Phenological diversity of salmon smoltmigration timing within a large watershed. Trans AmFish Soc 147: 775−790

Canadian Environmental Assessment Agency (CEAA) (2016)Decision Statement Issued under Section 54 of the Cana-dian Environmental Assessment Act, 2012. CanadianEnvironmental Assessment Agency, Ottawa

Chacin DH, Stallings CD (2016) Disentangling fine- andbroad-scale effects of habitat on predator−prey interac-tions. J Exp Mar Biol Ecol 483: 10−19

Chandler GT, Fleeger JW (1987) Facilitative and inhibitoryinteractions among estuarine meiobenthic Harpacticoidcopepods. Ecology 68: 1906−1919

Chesson J (1978) Measuring preference in selective preda-tion. Ecology 59: 211−215

Clark CW, Levy DA (1988) Diel vertical migrations by juve-nile sockeye salmon and the antipredation window. AmNat 131: 271−290

Cloern JE (1996) Phytoplankton bloom dynamics in coastalecosystems: a review with some general lessons fromsustained investigation of San Francisco Bay, California.Rev Geophys 34: 127−168

Cloern JE, Foster SQ, Kleckner AE (2014) Phytoplanktonprimary production in the world’s estuarine–coastal eco-systems. Biogeosciences 11: 2477−2501

Croll DA, Marinovic B, Benson S, Chavez FP, Black N, Ter-nullo R, Tershy BR (2005) From wind to whales: tropiclinks in a coastal upwelling system. Mar Ecol Prog Ser289: 117−130

Cushing DH (1990) Plankton production and year-classstrength in fish populations: an update of the match/mis-

match hypothesis. In: Blaxter JHS, Southward AJ (eds)Advances in marine biology, Vol 26. Elsevier, New York,NY, p 249−293

Daly E, Brodeur RD, Weitkamp L (2009) Ontogenetic shiftsin diets of juvenile and subadult coho and chinooksalmon in coastal marine waters: important for marinesurvival? Trans Am Fish Soc 138: 1420−1438

Daly E, Benkwitt CE, Brodeur RD, Litz MN, Copeman L(2010) Fatty acid profiles of juvenile salmon indicate preyselection strategies in coastal marine waters. Mar Biol157: 1975−1987

David V, Selleslagh J, Nowaczyk A, Dubois S and others(2016) Estuarine habitats structure zooplankton commu-nities: implications for the pelagic trophic pathways.Estuar Coast Shelf Sci 179: 99−111

Duffy EJ, Beauchamp DA, Sweeting RM, Beamish RJ, Bren-nan JS (2010) Ontogenetic diet shifts of juvenile Chinooksalmon in nearshore and offshore habitats of PugetSound. Trans Am Fish Soc 139: 803−823

Dunn OJ (1961) Multiple comparisons among means. J AmStat Assoc 56: 52−64

Eggers DM (1977) The nature of prey selection by planktiv-orous fish. Ecology 58: 46−59

Eggers DM (1982) Planktivore preference by prey size. Ecol-ogy 63: 381−390

Emlen JM (1966) The role of time and energy in food prefer-ence. Am Nat 100: 611−617

Fujiwara M, Highsmith RC (1997) Harpacticoid copepods: potential link between inbound adult salmon and out-bound juvenile salmon. Mar Ecol Prog Ser 158: 205−216

Gibson RN, Ezzi IA (1985) Effect of particle concentration onfilter- and particulate-feeding in the herring Clupeaharengus. Mar Biol 88: 109−116

Giske J, Aksnes DL, Fiksen Ø (1994) Visual predators, envi-ronmental variables and zooplankton mortality risk. VieMilieu 44: 1−9

Godin JGJ (1981) Daily patterns of feeding behaviour, dailyrations, and diets of juvenile pink salmon (Oncorhynchusgorbuscha) in two marine bays of British Columbia. CanJ Fish Aquat Sci 38: 10−15

Gregory RS, Northcote TG (1993) Surface, planktonic, andbenthic foraging by juvenile chinook salmon (Oncorhyn-chus tshawytscha) in turbid laboratory conditions. Can JFish Aquat Sci 50: 233−240

Griffiths D (1973) The food of animals in an acid morlandpond. J Anim Ecol 42: 285−293

Griffiths D (1975) Prey availability and the food of predators.Ecology 56: 1209−1214

Healey MC (1979) Detritus and juvenile salmon productionin the Nanaimo estuary: I. Production and feeding ratesof juvenile chum salmon (Oncorhynchus keta). J Fish ResBoard Can 36: 488−496

Healey MC (1980) Utilization of the Nanaimo river estuaryby juvenile Chinook salmon, Oncorhynchus tshawu -tscha. Fish Bull 77: 653−668

Healey MC (1991) Diets and feeding rates of juvenile pink,chum, and sockeye salmon in Hecate Strait, BritishColumbia. Trans Am Fish Soc 120: 303−318

Hill A, Daly E, Brodeur R, Hill A, Daly E, Brodeur R (2015)Diet variability of forage fishes in the Northern CaliforniaCurrent System. J Mar Sci 146: 121−130

Hitchcock C, Sober E (2004) Prediction versus accommoda-tion and the risk of overfitting. Br J Philos Sci 55: 1−34

Hosack GR, Dumbauld BR, Ruesink JL, Armstrong DA(2006) Habitat associations of estuarine species: compar-

167

Mar Ecol Prog Ser 613: 151–169, 2019

isons of intertidal mudflat, seagrass (Zostera marina),and oyster (Crassostrea gigas) habitats. Estuaries Coasts29: 1150−1160

Jennions MD (2003) A survey of the statistical power ofresearch in behavioral ecology and animal behavior.Behav Ecol 14: 438−445

Kennedy LA, Juanes F, El-Sabaawi R (2018) Eelgrass asvaluable nearshore foraging habitat for juvenile Pacificsalmon in the early marine period. Mar Coast Fish 10: 190−203

Labropoulou M, Eleftheriou A (1997) The foraging ecologyof two pairs of congeneric demersal fish species: impor-tance of morphological characteristics in prey selection.J Fish Biol 50: 324−340

Landigham JH, Sturdevant MV, Brodeur RD (1998) Feedinghabits of juvenile Pacific salmon in marine waters ofsoutheastern Alaska and northern British Columbia. FishBull 96: 285−302

Lannin R, Hovel K (2011) Variable prey density modifies theeffects of seagrass habitat structure on predator−preyinteractions. Mar Ecol Prog Ser 442: 59−70

Lazzaro X (1987) A review of planktivorous fishes: their evo-lution, feeding behaviours, selectivities, and impacts.Hydrobiologia 146: 97−167

Leising AW, Yen J (1997) Spacing mechanisms within light-induced copepod swarms. Mar Ecol Prog Ser 155: 127−135

Letcher BH, Rice JA (1997) Prey patchiness and larval fishgrowth and survival: inferences from an individual-based model. Ecol Model 95: 29−43

Levings CD (2016) Ecology of salmonids in estuaries aroundthe world: adaptations, habitats, and conservation. UBCPress, Vancouver

Ljungberg P, Hasper TB, Nilsson PA, Persson A (2013)Effects of small-scale habitat fragmentation on preda-tor−prey interactions in a temperate sea grass system.Mar Biol 160: 667−675

López Abbate MC, Molinero JC, Guinder VA, Dutto MS andothers (2015) Microplankton dynamics under heavyanthropogenic pressure. The case of the Bahía BlancaEstuary, southwestern Atlantic Ocean. Mar Pollut Bull95: 305−314

Macdonald JS, Birtwell IK, Kruzynski GM (1987) Food andhabitat utilization of juvenile salmonids in the CampbellRiver estuary. Can J Fish Aquat Sci 44: 1233−1246

Mackas DL, Greve W, Edwards M, Chiba S and others(2012) Changing zooplankton seasonality in a changingocean: comparing time series of zooplankton phenology.Prog Oceanogr 97−100: 31−62

Manzer J (1969) Stomach contents of juvenile Pacific salmonin Chatham Sound and adjacent waters. J Fish Res BoardCan 26: 2219−2223

Mauchline J (1998) The biology of calanoid copepods. Aca-demic Press, New York, NY

McDevitt-Irwin JM, Iacarella JC, Baum JK (2016) Reassess-ing the nursery role of seagrass habitats from temperateto tropical regions: a meta-analysis. Mar Ecol Prog Ser557: 133−143

Miller TW, Brodeur RD (2007) Diets of and trophic relation-ships among dominant marine nekton within the north-ern California Current ecosystem. Fish Bull 105: 548−559

Moore JW, Gordon J, Carr-Harris C, Gottesfeld AS, WilsonSM, Russell JH (2016) Assessing estuaries as stopoverhabitats for juvenile Pacific salmon. Mar Ecol Prog Ser559: 201−215

Moore JW, Carr-Harris C, Gottesfeld AS, MacIntyre D andothers (2015) Selling First Nations down the river. Sci-ence 349: 596

Morgan CA, Cordell JR, Simenstad CA (1997) Sink or swim?Copepod population maintenance in the ColumbiaRiver estuarine turbidity-maxima region. Mar Biol 129: 309−317

Nagelkerken I, Sheaves M, Baker R, Connolly RM (2015)The seascape nursery: a novel spatial approach to iden-tify and manage nurseries for coastal marine fauna. FishFish 16: 362−371

Nakagawa S (2004) A farewell to Bonferroni: the problemsof low statistical power and publication bias. Behav Ecol15: 1044−1045

Nohara D, Kitoh A, Hosaka M, Oki T (2006) Impact of cli-mate change on river discharge projected by multimodelensemble. J Hydrometeorol 7: 1076−1089

Northcote T, Gregory R, Magnhagen C (2007) Contrastingspace and food use among three species of juvenilePacific salmon (Oncorhynchus) cohabiting tidal marshchannels of a large estuary. Can Tech Rep Fish Aquat Sci2759, Fisheries and Oceans Canada, St John’s

Ocean Ecology (2014) Skeena River estuary juvenile salmonhabitat. Report prepared by Ocean Ecology for SkeenaWild Conservation Trust and Skeena Watershed Conser-vation Coalition. Prince Rupert

Osgood G (2016) Historical diets of forage fish and juvenilePacific salmon in the Strait of Georgia, 1966−1968. MarCoast Fish 8: 580−594

Palmer MA (1988) Dispersal of marine meiofauna: a reviewand conceptual model explaining passive transport andactive emergence with implications for recruitment. MarEcol Prog Ser 48: 81−91

Perillo GME (1995) Definitions and geomorphologic classifi-cations of estuaries. Dev Sedimentol 53: 17−47

Pickard D, Porter M, Olson E, Connors B, Kellock K, Jones E,Connors K (2015) Skeena River estuary assessment: technical report. Pacific Salmon Foundation, Vancouver

Pitchford WJ (2001) Prey patchiness, predator survival andfish recruitment. Bull Math Biol 63: 527−546

Price M, Glickman B, Reynolds J (2013) Prey selectivity ofFraser River sockeye salmon during early marine mi -gration in British Columbia. Trans Am Fish Soc 142: 1126−1133

R Core Team (2017) R: a language and environment for sta-tistical computing. R Foundation for Statistical Comput-ing, Vienna

Rudstam LG, Aneer G, Hildén M (1994) Top-down control inthe pelagic Baltic ecosystem. Dana 10: 105−129

Schoener TW (1970) Non-synchronous spatial overlap oflizards in patchy habitats. Ecology 51:408-418

Selleslagh J, Lesourd S, Amara R (2012) Comparison ofmacrobenthic assemblages of three fish estuarine nurs-eries and their importance as foraging grounds. J MarBiol Assoc UK 92: 85−97

Selleslagh J, Blanchet H, Bachelet G, Lobry J (2015) Feedinghabitats, connectivity and origin of organic matter sup-porting fish populations in an estuary with a reducedintertidal area assessed by stable isotope analysis. Estu-aries Coasts 38: 1431−1447

Sharpe C (2017) Abiotic and biotic dimensions of habitat forjuvenile salmon and other fishes in the Skeena Riverestuary. MS thesis, Simon Fraser University, Burnaby

Sheaves M, Baker R, Nagelkerken I, Connolly RM (2015)True value of estuarine and coastal nurseries for fish: in -

168

Arbeider et al.: Integrating estuarine prey dynamics

corporating complexity and dynamics. Estuaries Coasts38: 401−414

Sherwood CR, Jay DA, Bradford Harvey R, Hamilton P,Simenstad CA (1990) Historical changes in the ColumbiaRiver Estuary. Prog Oceanogr 25: 299−352

Sibert J (1979) Detritus and juvenile salmon production inthe Nanaimo Estuary: II. Meiofauna available as food tojuvenile chum salmon (Oncorhynchus keta). J Fish ResBoard Can 36: 497−503

Sih A, Englund G, Wooster D (1998) Emergent impacts ofmultiple predators on prey. Trends Ecol Evol 13: 350−355

Simenstad CA, Fresh KL, Salo EO (1982) The role of PugetSound and Washington coastal estuaries in the life history of Pacific salmon: an unappreciated function.In: Kennedy VS (ed) Estuarine comparisons. Elsevier,New York, NY, p 343−364

Sims DW, Southall EJ, Humphries NE, Hays GC and others(2008) Scaling laws of marine predator search behaviour.Nature 451: 1098−1102

Steinarsdóttir MB, Ingólfsson A, Ólafsson E (2010) Fieldevidence of differential food utilization of phytal har -pacticoids collected from Fucus serratus indicated byδ13C and δ15N stable isotopes. Estuar Coast Shelf Sci88: 160−164

Telesh IV, Khlebovich VV (2010) Principal processes withinthe estuarine salinity gradient: a review. Mar Pollut Bull61: 149−155

Trites RW (1956) The oceanography of Chatham Sound,British Columbia. J Fish Res Board Can 13: 385−434

Tschaplinski PJ (1987) The use of estuaries as rearing habitats by juvenile coho salmon. Fisheries and OceansCanada, Research Branch, Pacific Biological Station,Nanaimo

Werner EE, Gilliam JF (1984) The ontogenetic niche andspecies interactions in size-structured populations. AnnuRev Ecol Syst 15: 393−425

Yang LH, Bastow JL, Spence KO, Wright AN (2008) Whatcan we learn from resource pulses? Ecology 89: 621−634

169

Editorial responsibility: Antonio Bode,A Coruña, Spain

Submitted: September 27, 2018; Accepted: February 14, 2019Proofs received from author(s): March 12, 2019