Tradeoffs between homing and habitat quality for spawningsite selection by hatchery-origin Chinook salmon

Jeremy M. Cram & Christian E. Torgersen &

Ryan S. Klett & George R. Pess & Darran May &

Todd N. Pearsons & Andrew H. Dittman

Received: 31 May 2011 /Accepted: 16 April 2012# Springer Science+Business Media B.V. 2012

Abstract Spawning site selection by female salmon isbased on complex and poorly understood tradeoffs be-tween the homing instinct and the availability of appro-priate habitat for successful reproduction. Previousstudies have shown that hatchery-origin Chinook salm-on (Oncorhynchus tshawytscha) released from differentacclimation sites return with varying degrees of fidelityto these areas. To investigate the possibility that homingfidelity is associated with aquatic habitat conditions, wequantified physical habitat throughout 165 km in theupper Yakima River basin (Washington, USA) andmapped redd and carcass locations from 2004 to 2008.Principal components analysis identified differences insubstrate, cover, stream width, and gradient among rea-ches surrounding acclimation sites, and canonical corre-spondence analysis revealed that these differences inhabitat characteristics were associated with spatial

patterns of spawning (p<0.01). These analyses indicatedthat female salmon may forego spawning near theiracclimation area if the surrounding habitat is unsuitable.Evaluating the spatial context of acclimation areas inrelation to surrounding habitat may provide essentialinformation for effectively managing supplementationprograms and prioritizing restoration actions.

Keywords Salmon . Spawning . Supplementation .

Habitat . Homing . Straying

Introduction

Anadromous salmonids typically home to their natalstream for spawning (Wisby and Hasler 1954; Dittmanand Quinn 1996; Quinn 2005), and recent research

Environ Biol FishDOI 10.1007/s10641-012-0026-1

J. M. Cram (*) : R. S. KlettUniversity of Washington,School of Environmental and Forest Sciences,Seattle, WA 98195, USAe-mail: jcram@u.washington.edu

C. E. TorgersenU.S. Geological Survey, Forest and Rangeland EcosystemScience Center, Cascadia Field Station,University of Washington,School of Environmental and Forest Science,Seattle, WA 98195, USA

G. R. PessFish Ecology Division, Northwest Fisheries Science Center,National Marine Fisheries Service,National Oceanic and Atmospheric Administration,2725 Montlake Boulevard East,Seattle, WA 98112, USA

D. MaySchool of Aquatic and Fishery Sciences,University of Washington,Box 355020, Seattle, WA 98195, USA

T. N. PearsonsGrant County Public Utility District,P.O. Box 878, Ephrata, WA 98823, USA

A. H. DittmanNational Oceanic and Atmospheric AdministrationFisheries, Northwest Fisheries Science Center,2725 Montlake Boulevard East,Seattle, WA 98112, USA

suggests that within a stream, salmon may also hometo specific reaches or natal incubation sites (Neville etal. 2006; Quinn et al. 2006). Furthermore, hatcherysalmon that are transported away from their rearingsite generally return as adults to the site from whichthey are released (Donaldson and Allen 1958; Ricker1972). These findings indicate that hatchery-originsalmon released from acclimation facilities may becapable of returning to those areas for spawning. How-ever, Dittman et al. (2010) found that hatchery-originspring Chinook salmon (Oncorhynchus tshawytscha)with different imprinting histories showed varyingdegrees of fidelity to release areas and most spawnedin the same reaches that were preferred by wild-originsalmon, more than 25 km from their respective releasesites. The acclimation process may be effective whenhabitat conditions are suitable, but when acclimationfacilities are located in areas of marginal spawninghabitat, factors other than homing may be importantfor spawning site selection.

During their freshwater spawning migration, salmonoften travel hundreds of kilometers and exhibit explor-atory behavior that exposes them to a range of spawninghabitat conditions (Keefer et al. 2008). The ultimatechoice of spawning location is driven, at least in part,by the availability of appropriate environmental condi-tions for redd construction and larval survival, which arelargely controlled by physical habitat conditions andgeomorphological processes (Beechie et al. 2008;Milleret al. 2008). For example, the broad-scale (102 – 103 m)distributions of spawning Chinook and coho salmon(Oncorhynchus kisutch) have been correlated with par-ticular channel morphologies, such as pool-riffle channeltypes (Montgomery and Buffington 1998; Montgomeryet al. 1999; Hanrahan 2007). Due to the hierarchicalnature of lotic systems, pool-riffle channel types containthe velocity, substrate, and depth ranges preferred byspawning Chinook salmon (Frissell et al. 1986; Bjornnand Reiser 1991). Geist (2000); Geist et al. (2002)showed that hyporheic exchange, which increases inpool-riffle reaches, is associated with spawning site se-lection by Chinook and chum salmon (Oncorhynchusketa). Intermediate-scale habitat factors (101 – 102 m)that are associated with spawning site selection havebeen less frequently studied than broad- or fine-scale(< 101 m) characteristics (Fausch et al. 2002), althoughchannel type, bifurcation, and sinuosity have beenshown to be important predictors of spawning at thisscale (Baxter and Hauer 2000; Fukushima 2001;

Coulombe-Pontbriand and LaPointe 2004). Despite nu-merous previous studies pertaining to salmonid spawn-ing habitat requirements and redd site selection, little isknown about the effects of environmental variation onhoming and straying behavior among hatchery-originsalmon (Blair and Quinn 1991; Quinn 1993).

Hatchery supplementation is an approach to preserveand rebuild populations of anadromous salmonids(Berejikian et al. 2008). Supplementation programs inthe Columbia River Basin collect gametes from wildbroodstock, fertilize eggs in a hatchery, and rear thejuveniles there for some period of time. In some casesthe juveniles are released as smolts from the hatchery, tomigrate to sea and continue their life cycle. However, inother cases the juveniles are transported to acclimationfacilities (e.g., large outdoor holding ponds) where theyare held until release to facilitate imprinting on thoselocations within a watershed, away from the hatchery(Bugert 1998). Acclimation facilities are designed toexpand the spatial distribution of spawning into underu-tilized areas and to minimize potentially deleteriousinteractions (e.g. competition and interbreeding) withwild conspecifics (Mobrand et al. 2005; Kostow2009). However, for the acclimation process to succeed,hatchery-origin salmon must return to their acclimationarea and spawn nearby. If suitable spawning habitat islacking near an acclimation area, salmon may stray intoanother stream or reach for spawning, or return to thevicinity of the hatchery itself, where they spent their firstmonths of life.

In this study, we investigated the influence of accli-mation site location versus habitat characteristics onspawning site selection by hatchery-origin female Chi-nook salmon. We employed a novel approach that inte-grated spatially continuous surveys of aquatic habitat,redds, and carcasses over multiple years. The objectivesof this study were to (1) differentiate among acclimationareas based on physical habitat characteristics and (2)relate these differences to the observed spawning distri-bution of hatchery-origin salmon.

Methods

Study area

The Yakima River drains a 15 928 km2 watershed on theeastern slopes of the Cascade Mountains in WashingtonState and flows into the Columbia River near Richland,

Environ Biol Fish

WA, USA. This study was conducted in the upperYakima River basin (Fig. 1) at elevations ranging from768 m at Keechelus Dam (Yakima R., rkm 97) to 470 mat Ellensburg, WA (Yakima R., rkm 1), including theupper 97 km of the Yakima River, 12 km of the CleElum River (drainage area 0598 km2), and 31 km of theTeanaway River (drainage area 0536 km2). Almost allspring Chinook salmon in the upper Yakima Riverspawn upstream of Ellensburg, WA (Dittman et al.2010). Forested portions of these watersheds are situatedon public and private lands that are managed for timberproduction, whereas developed and arid sections arepredominantly private lands used for agriculture andresidential purposes. All three rivers would naturallyhave snowmelt-dominated hydrographs peaking duringlate spring, but the Cle Elum and Yakima rivers are

dammed, and flow is managed for agricultural purposes,delaying peak discharge until the summer. Easton Dam(Yakima R., rkm 76) is currently the only damwithin thestudy area that allows fish passage. Cle Elum dam (CleElumR., rkm 12) is impassable and was the end point ofour Cle Elum River habitat, redd, and carcass surveys.

The upperwatershed receives substantial precipitation(350 cm year-1; http://www.wrcc.dri.edu) and supportsdense riparian forests of Douglas fir (Pseudotsuga men-ziesii), western red cedar (Thuja plicata), red alder (Alnusrubra), and salmonberry (Rubus spectabilis), while thelower Yakima River is surrounded by an arid landscapethat receives little precipitation (< 25 cm year-1; http://www.wrcc.dri.edu) and is sparsely forested by ponderosapine (Pinus ponderosa) and black cottonwood (Populustrichocarpa). The study area is located in the Cascade

Fig. 1 Study area location in the upper Yakima River basin,Washington, including the Cle Elum and Teanaway rivers. Thehighlighted portions of river indicate the survey extents. Stars

represent the central hatchery and satellite acclimation facilities(Clark Flat, Jack Creek, and Easton) from which fish werereleased; dark bars are dam locations

Environ Biol Fish

geological province, which is dominated by sedimentaryand metamorphic rocks, although much of the upperTeanaway and Cle Elum basins are overlain by sand-stone (Leland 1995). Local geography and topographycreate a range of riverine habitat conditions, from alluvialfloodplain reaches with multiple channels to confined,bedrock channel types.

Broodstock collection, rearing, and marking proto-cols for the Yakima Klickitat Fisheries Project SpringChinook Salmon Program have been described byDittman et al. (2010) and Knudsen et al. (2006). Eggswere fertilized and incubated at the Cle Elum Hatcheryand juveniles were transferred to outdoor raceways forapproximately 10 more months of rearing. In Februaryof their second year, prior to the parr-smolt transforma-tion, salmon were transferred to one of three satelliteacclimation, imprinting, and release facilities: Easton(Yakima rkm 74), Clark Flat (Yakima rkm 19), orJack Creek (Teanaway rkm 29) (Fig. 1). Surfacewater collected adjacent to the site was used to imprintthe fish to the site. After 2–8 weeks of acclimation (Aprilto early June), smolts were allowed to volitionallymigrate from the facilities. All salmonweremarkedwithrelease-site-specific coded wire tags (CWT), color-coded visible implant elastomer eye tags (NorthwestMarine Technology, Inc.), and adipose fin clips. Uponreturn as adults, all hatchery-reared salmon wereallowed to spawn naturally within the basin. The YakimaRiver spring Chinook population is not listed for Endan-gered Species Act protection.

Aquatic habitat, redd, and carcass surveys

Physical habitat data from 165 km ofmainstem (97 km),tributary (43 km), and floodplain habitats (25 km) werecollected between 5 September 2007 and 5 October2007. A spatially continuous approach was used toinvestigate habitat characteristics at multiple spatialscales throughout the study area (Torgersen et al.2006). Individual habitat units were identified visuallyas pools, riffles, glide-like pools, or glide-like rifflesbased on changes in depth, channel gradient, and bed-form morphology (Bisson et al. 1982). Data collectedfor each unit included wetted (distance between wettedmargins) and active (bankfull) channel widths, meanand maximum depths, percent of unit length of cover(wood, vegetative, or boulder), and substrate (gravel,cobble, boulder, bedrock, fines). Substrate and coverwere visually estimated by the field crew (Hankin and

Reeves 1988; Latulippe et al. 2001). Depth was mea-sured relative to the snorkeler’s body length (calibratedwith a measuring tape up to 2 m). Habitat unit widthsand lengths were measured using a laser rangefinder(ASC Scientific). Individual habitat survey units weremapped in the field using ArcMap 7.1 (ESRI) on a tabletcomputer with an integrated global positioning system(GPS).

Redd and carcass surveys were conducted within thestudy area from 2004 to 2008 (Dittman et al. 2010).Individual carcasses and redds were identified by thefield crew and georeferenced using a GPS. For eachcarcass, sex and origin (hatchery or wild) were notedin the field, and acclimation facility was determinedeither in the field according to the elastomer eye tag orin the laboratory after processing the CWT. Carcass datawere used to infer spawning site selection by femaleChinook salmon (Murdoch et al. 2009a), and redd dataprovided the overall spawning distribution. Amongsemelparous species, female salmon select a spawningsite, spawn, and then defend the nest from superimpo-sition by other females until death, whereas males do notdefend the redd and may seek other mating opportuni-ties (Quinn 2005; Murdoch et al. 2009b). Therefore,female carcasses are more likely to be recovered neartheir spawning site, making marked hatchery-originfemales ideal subjects for studying homing (Murdochet al. 2010). Spawning data were spatially joined to theappropriate habitat units using a GIS.

Individual habitat units containing empirical habitatand spawning data were supplemented with GIS-derivedNetMap data (Benda et al. 2007). NetMap is an interac-tive GIS database with data primarily derived from dig-ital elevation models (DEM). The NetMap data fieldsincluded channel gradient, active width, stream power,and tributary effect (an estimate of the downstreameffects of tributaries on mainstem habitat (sensu Bendaet al. 2004)). Additionally, sinuosity was calculated foreach habitat unit using Hawth’s Tools (Beyer 2004).

Data analysis

Spatially explicit physical habitat, carcass, and redd datawere summarized in 1-km reaches for analysis. Most1-km reaches spanned several habitat units, so data weresummarized as means (e.g. percent gravel), maximums(e.g. maximum depth), minimums (e.g. minimumdepth), or counts (e.g. redds) of the original data fields.Inter-annual variation in habitat conditions was assumed

Environ Biol Fish

to be minimal from 2004 to 2008 because extreme highflow events did not occur (based on 99th percentileexceedance flows). This assumption was validated bypair-wise Pearson’s correlations of redd counts within1-km reaches across years. Locally weighted scatterplotsmoothing (LOWESS) was used to identify peaks inspawning density in the Yakima and Teanaway Rivers,but not in the Cle Elum River because its 11 reacheswere not adequate for the analysis (Trexler and Travis1993). LOWESS was calculated using a sampling pro-portion of 0.2 and a quadratic polynomial in SigmaPlot(version 10.0). Only 2007 redd data were used for thisanalysis, because the corresponding habitat data werecollected during that year. Principal components analy-sis (PCA) of a correlation matrix was used to exploresimilarities among reaches based on their habitat char-acteristics. The number of variables in the analysis wasminimized by first running PCA with all habitat varia-bles, and then removing redundancies based on visualinterpretation of the resulting habitat vectors. For exam-ple, field-measured wetted and active widths werestrongly correlated; active width was removed becauseit had less of an effect on the ordination, although theDEM-derived active width remained in the analysis. Arandomized broken-stick model was used to test forstatistical significance of the principal components axes(Legendre and Legendre 1998).

Canonical correspondence analysis (CCA) was usedto summarize patterns in spawning assemblage structure(i.e. the composition of Easton-, Clark Flat-, Jack Creek-,and wild-origin spawners in a reach, henceforth referredto as “spawning groups”) according to dominantgradients in co-variation between spawning site selec-tion and aquatic habitat (Palmer 1993). Specifically,we used CCA to assess the tradeoffs between hom-ing instinct (for hatchery-origin females) and desir-able habitat characteristics. We hypothesized that ifhoming behavior were the dominant factor in spawningsite selection by hatchery-origin females, then (1) reachesnear each acclimation facility would form distinct clus-ters in ordination space based on their spawning assem-blage, (2) the weighted averages for each hatchery-originspawning group would be near their acclimation facilityin ordination space, and (3) inter-annual variation inordination space within each spawning group would beminimal. We interpreted deviations from these expectedpatterns as evidence for straying based on habitat quality.The position in ordination space of spawning groups inrelation to their acclimation sites provided information

on the degree of straying after accounting for differencesin habitat quality. Inter-annual dispersion within spawn-ing groups was interpreted as a measure of annual vari-ation in habitat stability and suitability at different flowsand spawner densities. The CCA predictor matrix includ-ed the same habitat variables used for PCA, and theresponse matrix was composed of female carcass countdata for fish released from Easton, Jack Creek, and ClackFlat acclimation facilities, and wild-origin females. Rea-ches with no female carcasses identified in any year wereremoved from the environmental and spawning matrices(Yakima rkm 13, and Teanaway rkm 24, 30, and 31). Inorder to reduce the impact of noisy environmental dataand redundant variables, habitat data were log trans-formed, and carcass counts were column-standardizedand arcsine-square-root transformed (McCune 1997).An analysis of variance (ANOVA) was used to test forsignificance among the terms used in the CCA. Allstatistical analyses were performed using R statisticalsoftware (www.r-project.org).

Results

Habitat surveys

The spatially continuous habitat data revealed longi-tudinal variation in substrate, cover, and depththroughout the study area. The most prevalent sub-strate type in the Yakima River was gravel, whereasthe Teanaway and Cle Elum rivers were dominated bycobble (Table 1). The Yakima River contained thedeepest maximum and mean depths, the widest chan-nel, and most abundant cover of the three rivers. TheTeanaway River had minimal cover and a higher pro-portion of riffle channel types and bedrock substratethan the other rivers. Downstream from Keechelus andCle Elum dams, channel structure appeared to beinfluenced by impoundment and flow regulation forapproximately 1 km: channels were highly incised,dominated by large cobbles, and lacking variation indepth (Cram, unpubl. data). However, Easton Dam didnot have the same geomorphic effects.

Redd distribution

Longitudinal profiles of redd density (n0573) in theYakima River in 2007 revealed two scales of spatialvariation (Fig. 2). Local maximums in density occurred

Environ Biol Fish

at 15 to 25 km intervals but were nested within a larger-scale gradient of increasing spawning density in anupstream direction towards Easton Dam (Fig. 2, peaksA – E). The Cle Elum and Teanaway rivers containedfewer redds (40 and 13, respectively) than the YakimaRiver. The LOWESS profile of the Teanaway Rivershowed spawning peaks at rkm 6 and 20. In the CleElumRiver, redd densities were highest at rkm 2 and 11.Pairwise Pearson’s correlations of redd distributionsfrom 2004 to 2008 ranged from 0.84 to 0.97, indicatingthat spawning distribution patterns were consistentamong years. These high correlations among years alsoprovided indirect evidence that aquatic habitat condi-tions were consistent during this time period.

Variation in habitat conditions and redd density

Principal components analysis confirmed that therewas significant variation in habitat conditions among

1 km reaches (Fig. 3, Table 2). The first two principalcomponent axes explained 41 % of the variation inhabitat characteristics among reaches (p<0.01). Thefirst axis was driven primarily by stream active width,depth, and gradient, whereas the second axis wasdriven by wetted width and cover availability. UpperYakima reaches were primarily located in quadrant 1of the ordination, Teanaway River reaches were inquadrant 2, Cle Elum and lower Yakima reaches werein quadrant 3, and the lower and middle Yakimareaches were in quadrant 4. The Yakima River wasthe longest and most physically diverse of the threerivers surveyed, as represented by the spread amongYakima River reaches in all quadrants of the PCA.Highly diverse habitat conditions among the 11 CleElum River reaches caused them to occur in threequadrants of the PCA. Despite being nearly threetimes longer than the Cle Elum, the Teanaway Riverwas the least diverse of the three rivers and was

Table 1 Extensive aquatichabitat survey data from fall 2007sorted by totals and averages.SD standard deviation

River

Yakima Cle Elum Teanaway

Totals

Habitat units 373 30 152

Total length (km) 108 11.9 31.6

Glide-like pool (%) 33 40 26

Glide-like riffle (%) 17 13 22

Pool (%) 18 13 4

Riffle (%) 32 34 48

Averages Mean SD Mean SD Mean SD

Unit length (m) 303.5 246.1 406.3 235.0 209 213.0

Gravel (%) 54.6 21.8 34.7 14.2 28.3 17.7

Cobble (%) 33.4 17.7 47.3 22.5 53.9 24.7

Boulder (%) 5.8 9.8 8.7 12.4 4.3 6.4

Fines (%) 2.5 9.9 3.3 7.8 0.1 1.6

Bedrock (%) 2.0 8.4 0.3 1.9 10.7 22.8

Maximum depth (m) 1.4 0.9 1.1 1.1 0.5 0.5

Mean depth (m) 0.5 1.2 0.1 0.5 0.1 0.3

Wetted width (m) 24.3 13.5 21.7 7.9 9.7 3.9

Active width (m) 28.5 14.5 39.2 18.0 18.2 7.6

Boulder cover (%) 8.9 15.5 8.4 14.8 1.6 5.0

Vegetative cover (%) 13.2 20.1 4.2 18.4 3.1 7.5

Wood cover (%) 8.5 15.6 9.8 16.0 2.6 8.8

Sinuosity 1.2 0.3 1.1 0.1 1.1 0.1

Environ Biol Fish

characterized by a narrow channel, high channelgradient, and abundant cobble and bedrock.

The PCA indicated that the Easton (upper Yakima),Clark Flat (lower Yakima), and Jack Creek (upperTeanaway) acclimation areas were located in markedlydifferent habitat types. The Easton acclimation area(Fig. 3, quadrant 1) had relatively dynamic channels thatcontained woody and vegetative cover, high pool-riffleratio, and predominantly gravel substrate. The JackCreek acclimation area (Fig. 3, quadrant 2) was associ-ated with high-gradient riffles, cobble, and bedrock sub-strate, whereas wetted width and boulder substrate wereassociated with the Clark Flat area (Fig. 3, quadrant 4).Unlike the other acclimation sites, the central hatchery(located in the center of the ordination) was not stronglyassociated with any major habitat gradients.

Nearly half of the maximums in spawning densityidentified using LOWESS occurred in the first quadrant

of the PCA, which was associated positively with cover,gravel, multiple channels, sinuosity, and channel unit type(pools, glides). The reaches in quadrant 1 were locatedprimarily in the Yakima River, upstream from the CleElum River confluence. Among the five local maximumsthat fell outside of quadrant one, the Cle Elum reach wasimmediately downstream of a dam, which blocked pas-sage of spawners. Two of the LOWESS peaks from thelower Yakima and one from the Teanaway were moreclosely associated with characteristics and reaches inquadrant 1 than they were to those of geographicallyproximate reaches. These reaches were characterized byhabitat that resembled high-density spawning areas and,in turn, supported more spawning than their surroundingreaches. The combination of LOWESS and PCA indicat-ed that spawning female salmon preferred reaches char-acterized by gravel substrate, woody or vegetative cover,and complex, low gradient channel types.

Fig 2 Locally weightedscatterplot smoothing (LOW-ESS) of redd data from 2007binned in 1-km reaches(black dots) in the Yakima(a), Teanaway (b), and CleElum (c) rivers. Circles indi-cate local peaks in spawningdensity. Letters identify spe-cific peaks that appear inmultivariate ordination(see Fig. 3b)

Environ Biol Fish

Tradeoffs between habitat and homing

A total of 3583 female carcasses were identified from2004 to 2008 (Table 3). The suite of habitat variablesthat was used to constrain the CCA explained 14 % ofthe variation among reaches in spawning assemblagespace on the first two axes (ANOVA, p<0.01, Fig. 4,Table 4). The first CCA axis was driven by substrate,vegetative cover, depth, and multiple channels, and thesecond axis was driven by stream gradient, active width,and depth (Fig. 4a). There was differentiation amongreaches from each river based on their respective spawn-ing assemblages. Yakima River reaches showed the

most variation (i.e., they occurred in all quadrants),whereas Cle Elum and Teanaway river reaches occurredalmost entirely in quadrant 2. Within quadrant 2, therewas very little overlap between Teanaway reaches andreaches from the Yakima and Cle Elum, and there was asmall degree of overlap between Yakima and Teanawayreaches in quadrants 2 and 3 (Fig. 4a).

There was distinct separation among acclimation rea-ches in spawner space; the Easton reach was located onthe right side of the ordination, which contained highlyproductive spawning habitat, as indicated by the LOW-ESS peaks (Fig. 2a) and PCA (Fig. 3b). Clark Flat andJack Creek were in quadrants 2 and 3 of the CCA,

Fig. 3 Principal componentsanalysis of aquatic habitatcharacteristics in 1-kmreaches in the Yakima, CleElum, and Teanawayrivers. Reaches in habitatspace are plotted with respectto a biplot indicating thedirection and magnitude ofloadings (a). Redd densitypeaks from LOWESS(reaches A – I, Fig. 2) and thelocations of hatchery facili-ties are shown in habitatspace (CFClark Flat, JC JackCreek, EA Easton,CH centralhatchery). Habitat nearhatchery facilities is averagedin 3-km reaches (b)

Environ Biol Fish

respectively, and were associated with habitat conditionsthat were not preferable for spawning (i.e., higher gradi-ent and larger percentage of cobble substrate). The sep-aration among acclimation facilities in the CCAwas dueto females homing to their release location for spawning,resulting in different spawning assemblage structures.The inter-annual spread within the Easton spawninggroup and its overlap with the wild-origin spawninggroup was influenced primarily by one female fromEaston spawning upstream of Easton Dam in 2005, anarea that was otherwise exclusively used by wild-originfemales (Fig. 4b). The inter-annual variation in spawn-ing distribution was less for Easton females, which

encountered productive spawning habitat while homingto their release site, than it was for females from otheracclimation facilities. Jack Creek and Clark Flat femalesshowed broader inter-annual spread and tended to befarther from their release sites in ordination space(Fig. 4b). Wild-origin females spawned throughout thesurvey area, but were generally on the right side of theordination

The results of the CCA analysis supported ourhypotheses that there would be distinct clusters ofspawning groups and reaches only if homing werethe dominant factor affecting spawning site selection.The separation among acclimation sites and reaches inordination space did indicate homing fidelity for somefemales, but the general pattern was for the spawninggroups to cluster around the origin, indicating thatfemales selected similar spawning locations and hab-itat conditions regardless of acclimation site. The loca-tions in ordination space of the spawning groupsrelative to their acclimation sites provided an indica-tion of the degree to which a given spawning groupresponded to habitat versus acclimation location(Fig. 4). For example, spawning groups from JackCreek and Clark Flat were farther from (and to theright of) their acclimation site in ordination space thanthe Easton females were from their acclimation site.Thus, female salmon released from Clark Flat andJack Creek acclimation facilities generally spawnedin reaches with more complex channel types, gravelsubstrate, and cover availability than was offered bytheir release areas.

Discussion

Acclimation areas

Environmental conditions near each acclimation facil-ity were markedly different and contributed to varyingfidelity among spawning groups to their release areas.Previous research focused on either the spawning hab-itat preferences of salmon (Geist 2000; McHugh andBudy 2004) or on homing behavior (Dittman and Quinn1996), but rarely has the potential tradeoff betweenhabitat and homing been investigated (Dittman et al.2010). Acclimation facilities characterized by marginalhabitat could be expected to have higher stray rates thanthose in more productive areas because reproductivesuccess is dependent on females identifying and

Table 2 Structure loadings (> 0.40) from the first two axes ofPCA (Fig. 3) on spatially continuous habitat data binned in 1-kmreaches

PCAVariable PC1 PC2

Gravel 0.53

Cobble −0.60Boulder −0.45Bedrock

Wetted width 0.49 −0.74Gradient −0.72Active Width 0.75 −0.53Sinuosity 0.51

Tributary effects −0.42 0.48

Vegetative cover 0.63

Wood cover 0.65

Maximum depth 0.60

Glides 0.47 0.42

Pool

Riffle −0.46Multiple channels 0.46

Table 3 Female carcasses per year by spawning group

Spawning group Year

2004 2005 2006 2007 2008

Clark Flat 218 10 61 40 110

Easton 92 13 81 37 137

Jack Creek 144 87 93 35 93

Wild 1033 747 270 143 139

Environ Biol Fish

utilizing appropriate spawning and incubation environ-ments (Williamson et al. 2010).

The Jack Creek acclimation area is 29 km upstream inthe TeanawayRiver, which poses challenges tomigratoryadults related to cover and flow, as well as substratelimitations for spawning. Discharge in the TeanawayRiver is primarily derived from snowpack, with high

runoff in spring and early summer, and low flow in thefall (< 0.3 m3/s in low water years). Therefore, salmonthat return to the Teanaway River in spring may movedownstream as the water level declines and temperatureincreases due to a lack of cover (i.e. narrow channels andfew pools) (sensu Torgersen et al. 1999). Alternative-ly, salmon that hold in the Yakima River until fall

Fig. 4 Canonical correspon-dence analysis of spawningassemblage from 2004 to2008 in 1-km reaches in theYakima, Cle Elum, andTeanaway rivers and thelocations of hatcheryfacilities (stars) are shown inhabitat space (CF Clark Flat,JC Jack Creek, EA Easton,CH central hatchery).Reaches and hatchery facili-ties are plotted with respect tothe direction and magnitudeof their associations withhabitat characteristics thatconstrain variation amongreaches in spawningassemblage space (a). Theinset of individual spawningassemblage groups (CFClarkFlat, JC Jack Creek, EAEaston, W wild; the numbersafter the abbreviation indicatethe year of observation)illustrates patterns near thecenter of the ordination (b)

Environ Biol Fish

may encounter difficult passage conditions throughthe Teanaway River and, thus, spawn elsewhere.Spawning density may also be limited by a lack ofsuitable gravel, which is flushed out during peakflows resulting in substrates dominated by cobbleand bedrock (Table 1). Dittman et al. (2010) found anexceptionally high stray rate among Jack Creek-releasedindividuals: 97 % spawned outside of the TeanawayRiver in 2005. Inter-annual variability in flow and den-sity of Chinook throughout the system may also affectstraying among Jack Creek salmon.

The Clark Flat acclimation area is situated down-stream of the major tributaries where the river is pre-dominantly a single, wide, exposed channel that hasminimal floodplain interaction due to (1) armoring onthe outside of most meanders and (2) the natural con-straint of the channel by adjacent hills. There is abun-dant boulder and bedrock substrate, but spawninggravels are limited in areas that have adequate depthand velocity for spawning. There is minimal wood andvegetation along the river margins and mid channel,making deep water the only available cover type. Thecombination of low spawning habitat quality and lack ofcover for over-summer holding may cause most salmonreleased from Clark Flat to continue upstream in searchof better habitat. Dittman et al. (2010) showed that, on

average, Clark Flat salmon spawned 29 km upstreamfrom their release site.

Of the three acclimation facilities, Easton was lo-cated in the most productive spawning area. As indi-cated by Fig. 2, most spawning occurred in theYakima River from rkm 40 to 76, reaches character-ized by desirable spawning or holding features such assinuosity, cover, channel bifurcation, gravel, pools,and water depth (Montgomery et al. 1999; Torgersenet al. 1999; Fukushima 2001; Mull and Wilzbach2007). The same habitat characteristics were associat-ed with quadrants 1 and 4 of the CCA, which wereelected by wild- and Easton-origin females and wereassociated with the most productive spawning areas ofthe PCA. The positions of Clark Flat and Jack Creekspawning groups on the left side of the CCA ordina-tion indicated that many spawned in low density neartheir acclimation areas, but most strayed into the moredesirable areas already occupied by wild- and Easton-origin salmon.

Where do straying salmon go?

If habitat was equal throughout the survey area and theacclimation process alone determined spawning siteselection, then the distribution of redds would haverevealed distinct peaks along the longitudinal profile ateach acclimation site. Also, CCAwould have resulted ingreater differentiation among spawning groups in theordination, such that they were more closely associatedwith their release site than with other spawning groups.However, due in part to habitat conditions in the Tean-away and lower Yakima rivers, most females from thoseareas spawned closer to the peak in wild spawning nearthe central hatchery than they did to their release sites.These salmon may have homed correctly and then elec-ted to spawn elsewhere due to low spawning habitatquality near their release site. Alternatively, this mayhave been attributable to sequential imprinting that oc-curred at the central hatchery before juveniles weretransferred to their acclimation facility (Dittman et al.2010) or to social behavior related to conspecifics inupstream areas (Mull and Wilzbach 2007). However,social or imprinting factors are unlikely to be the prima-ry drivers of straying behavior, given the frequency withwhich the Cle Elum River was used for spawning byhatchery-origin females. Females that spawned in theCle Elum River were almost exclusively wild- or JackCreek-origin salmon.

Table 4 Analysis of variance of aquatic habitat loadings oncanonical correspondence axes (**: α00.01,*:α00.05)

CCAVariable Df Chi square F N. Perm Pr (>F)

Gravel 1 0.05 5.30 99 0.01 **

Cobble 1 0.01 1.61 99 0.17

Boulder 1 0.01 1.28 99 0.31

Bedrock 1 0.03 3.05 99 0.01 **

Wetted width 1 0.05 5.19 99 0.01 **

Gradient 1 0.02 1.72 99 0.10

Active width 1 0.04 4.94 99 0.01 **

Sinuosity 1 0.01 1.68 99 0.15

Tributary effects 1 0.01 1.62 99 0.11

Vegetative cover 1 0.03 2.92 99 0.01**

Wood cover 1 0.03 2.97 99 0.01**

Maximum depth 1 0.02 1.79 99 0.09

Glides 1 0.03 3.30 99 0.01**

Pool 1 0.01 0.75 99 0.76

Riffle 1 0.02 2.13 99 0.05*

Multiple channels 1 0.01 0.85 99 0.76

Environ Biol Fish

It is possible that the Jack Creek salmon recognizedtributary habitat and chose to spawn there because theTeanaway River was unsuitable. The Cle Elum andTeanaway rivers are geographically proximate, drain tothe south, and have similar geologies, but the Cle ElumRiver offers better migration and spawning conditionsas it is wider and deeper (see Table 1). Within eachspawning group, dispersion among yearly averages inthe ordination may have been due to inter-annual habitatdifferences or density related factors. Females releasedfrom acclimation facilities in low-quality spawning rea-ches quickly occupied the suitable redd sites and dis-placed others from their spawning group into moredistant reaches. Easton- and wild-origin females showedthe least variation across years, likely because wildsalmon have always exploited the best available habitatand Easton salmon are acclimated in a productivespawning area, making straying unnecessary. The highinter-annual spread among Jack Creek and Clark Flatfemales was driven by frequent straying in search ofmore suitable spawning habitat than was available neartheir release sites.

The patchy distribution of redds in 2007 was likelydriven by habitat discontinuities (sensu Poole 2002),which were assumed to be relatively consistent duringour study period based on the Pearson’s correlationsamong redd densities. In most reaches, naturally occur-ring aquatic habitat variation influenced spawning siteselection, but in a few cases dams may have contributedas well. Areas immediately downstream of the Cle Elumand Easton Dams may have supported higher densityspawning than would have naturally occurred if thebarriers had not been present. For nearly 3 km down-stream of the Cle Elum dam, the river is incised andlacks small sediment and wood. Spawning in this areamay have been motivated by the barrier, rather thandesirable habitat conditions. However, reaches down-stream of Easton Dam ostensibly were not affected bythe dam’s interception of wood and sediment beyond thefirst few hundred meters. Therefore, the high spawningdensity near Easton Dam is likely attributable to bene-ficial habitat conditions despite the barrier upstream.

Management implications

Acclimation facilities attempt to expand the spatial dis-tribution of a population, creating a metapopulationstructure, wherein each acclimation facility is a source,and intermediate areas may be sinks or sources for wild

populations (Schlosser and Angermeier 1995). Accli-mation facilities were moderately successful at dis-tributing hatchery-origin spawners among areas thatwere not extensively used by wild-origin salmon, butrelatively low numbers of salmon returned to theseareas. Salmon released from the two sites that werelocated in areas of marginal spawning habitat quality(Clark Flat and Jack Creek) showed low fidelity totheir acclimation areas, but their distribution wasaffected by homing. Females that strayed from theseacclimation sites may be the most likely to colonizeother reaches in the watershed, as they are unable tospawn near their release site.

Peaks in spawning density that occurred near rela-tively unproductive areas, such as downstream of ClarkFlat, are of particular importance because they aresource areas for wild-origin salmon and spawninggrounds for supplementation salmon. There is evidencethat naturally spawning hatchery-origin Chinook salm-on select inferior spawning locations and produce feweroffspring than wild-origin conspecifics (Williamson etal. 2010). However, hatchery-origin salmon may bepredisposed to select inferior spawning sites based onthe location of acclimation facilities within a watershed.Additionally, hatchery-origin females may be excludedfrom more desirable redd sites by wild-origin salmonthat emerged from suitable redds and intend to return tosites of proven quality. By creating the habitat condi-tions found in high-density spawning areas through site-and process-based restoration near acclimation areas,the spatial organization of spawning may shift as thetradeoffs between homing and habitat are reduced.

This study has demonstrated the importance thatspawning habitat has on influencing the spatial distribu-tion of spawning. Acclimation sites are intended to influ-ence the distribution of spawners and this approach maybe effective in areas that have quality spawning habitat.However, habitat enhancements may be more effective atinfluencing spawning distribution, particularly in areasthat have unsuitable spawning habitat. Acclimation andhabitat enhancement might be used in different areasdepending upon the existing habitat conditions of thedesired spawning location. Supplementation programsseeking to rebuild salmon populations could relativelyquickly and inexpensively quantify habitat in a similarmanner to the extensive habitat survey we conducted.Field-derived data coupled with readily available GISdata offer insights into the spawning potential of currentor future acclimation areas. Existing programs could

Environ Biol Fish

utilize such information to plan restoration actions, whilefuture projects could optimize their placement of accli-mation facilities in areas that are likely to meet theirobjectives.

Acknowledgments We thank the Yakama Nation for providinglogistical support and identifying redds. Additionally, we aregrateful to NOAA Fisheries, USGS Western Region BiologicalResource Division State Partnership Program, and the WaterCenter at the University of Washington for providing funding.This publication is partially funded by the Joint Institute for theStudy of the Atmosphere and Ocean (JISAO) under NOAACooperative Agreement No. NA17RJ1232, Contribution # 1844.Tanya Cram, James Chu, and Ethan Welty were essential volun-teers assisting with the habitat surveys. We also thank Hiroo Imakiand Patricia Haggerty for GIS assistance. Thomas Quinn, JulianOlden, Phil Roni, and two anonymous reviewers provided valu-able comments. Any use of trade, product, or firm names is fordescriptive purposes only and does not imply endorsement by theU.S. Government

References

Baxter CV, Hauer FR (2000) Geomorphology, hyporheic exchange,and selection of spawning habitat by bull trout (Salvelinusconfluentus). Can J Fish Aquat Sci 57:1470–1481

Beechie TJ, Moir HJ, Pess GR (2008) Hierarchichal physicalcontrols on salmonid spawning location and timing. In: SearDA, DeVries P (eds) Salmonid spawning habitat in rivers:physical controls, biological responses, and approaches toremediation. American Fisheries Society, Bethesda, pp 83–101

Benda L, Poff NL, Miller D, Dunne T, Reeves G, Pess G,Pollock M (2004) The network dynamics hypothesis:how channel networks structure riverine habitats. BioScience54:413–427

Benda L, Miller D, Andras K, Bigelow P, Reeves G, Michael D(2007) NetMap: a new tool in support of watershed scienceand resource management. For Sci 53:206–219

Berejikian BA, Johnson T, Endicott RS, Lee-Waltermire J(2008) Increases in steelhead (Oncorhynchus mykiss) reddabundance resulting from two conservation hatchery strat-egies in the Hamma Hamma River, Washington. Can J FishAquat Sci 65:754–764

Beyer HL (2004) Hawth’s analysis tools for ArcGIS, Availableat http://www.spatialecology.com/htools

Bisson PA, Nielsen JL, Palmason RA, Grove LE (1982) Asystem of naming habitat types in small streams, withexamples of habitat utilization by salmonids during lowstreamflow. In: Armantrout NB (ed) Acquisition and utili-zation of aquatic habitat inventory information. AmericanFisheries Society, Western Division, Bethesda, pp 62–73

Bjornn TC, Reiser DW (1991) Habitat requirements of salmonidsin streams. In: Meehan WR (ed) Influences of forest andrangeland management on salmonid fishes and their habitats.American Fisheries Society Special Publication, Bethesda,pp 83–138

Blair GR, Quinn TP (1991) Homing and spawning site selectionby sockeye salmon (Oncorhynchus nerka) in Iliamna Lake,Alaska. Can J Zool 69:176–181

Bugert RM (1998) Mechanics of supplementation in theColumbia River. Fisheries 23:11–20

Coulombe-Pontbriand M, LaPointe M (2004) Geomorphic con-trols, riffle substrate quality, and spawning site selection intwo semi-alluvial salmon rivers in the Gaspe Peninsula,Canada. River Res Appl 20:577–590

Dittman AH, Quinn TP (1996) Homing in pacific salmon:mechanisms and ecological basis. J Exp Biol 199:83–91

Dittman AH, May D, Larsen DA, Moser ML, Johnston M, FastD (2010) Homing and spawning site selection by supple-mented hatchery- and natural-origin Yakima River springChinook salmon. Trans Am Fish Soc 139:1014–1028

Donaldson LR, Allen GH (1958) Return of silver salmon,Oncorhynchus kisutch (Walbaum) to point of release. TransAm Fish Soc 87:13–22

Fausch KD, Torgersen CE, Baxter CV, Li HW (2002) Landscapesto riverscapes: Bridging the gap between research and con-servation of stream fishes. BioScience 52:483–498

Frissell CA, Liss WJ, Warren CE, Hurley MD (1986) A hierarchi-cal framework for stream habitat classification: viewingstreams in a watershed context. Environ Manag 10:199–214

Fukushima M (2001) Salmonid habitat-geomorphology rela-tionships in low-gradient streams. Ecology 82:1238–1246

Geist DR (2000) Hyporheic discharge of river water into fallchinook salmon (Oncorhynchus tshawytscha) spawningareas in the Hanford Reach, Columbia River. Can J FishAquatSci 57:1647–1656

Geist DR, Hanrahan TP, Arntzen EV, McMichael GA, MurrayCJ, Chien YJ (2002) Physicochemical characteristics of thehyporheic zone affect redd site selection by chum salmonand fall Chinook salmon in the Columbia River. N Am JFish Manag 22:1077–1085

Hankin DG, Reeves GH (1988) Estimating total fish abundanceand total habitat area in small streams based on visualestimation methods. Can J Fish Aquat Sci 45:834–844

Hanrahan TP (2007) Bedform morphology of salmon spawningareas in a large gravel-bed river. Geomorphology 86:529–536

Keefer ML, Caudill CC, Peery CA, Boggs CT (2008) Non-directhoming behaviours by adult Chinook salmon in a large,multi-stock river system. J Fish Biol 72:27–44

Knudsen CM, Schroder SL, Busack CA, Johnston MV, PearsonsTN, Bosch WJ, Fast DE (2006) Comparison of life historytraits between first-generation hatchery and wild upperYakima river spring Chinook salmon. Trans Am FishSoc 135:1130–1144

Kostow K (2009) Factors that contribute to the ecological risksof salmon and steelhead hatchery programs and somemitigating strategies. Rev Fish Biol Fish 19:9–31

Latulippe C, Lapointe MF, Talbot T (2001) Visual characterizationtechnique for gravel-cobble river bed surface sediments;Validation and environmental applications contribution tothe Programme of CIRSA (Centre Interuniversitaire deRecherche sur le Saumon Atlantique). Earth Surf ProcessLandforms 26:307–318

Legendre P, Legendre L (1998) Numerical ecology. Elsevier,Amsterdam, p 853, xv

Leland HV (1995) Distribution of phytobenthos in the YakimaRiver basin, Washington, in relation to geology, land-use,

Environ Biol Fish

and other environmental-factors. Can J Fish Aquat Sci52:1108–1129

McCune B (1997) Influence of noisy environmental data oncanonical correspondence analysis. Ecology 78:2617–2623

McHugh P, Budy P (2004) Patterns of spawning habitat selectionand suitability for two populations of spring Chinook salmon,with an evaluation of generic versus site-specific suitabilitycriteria. Trans Am Fish Soc 133:89–97

Miller DJ, Burnett KM, Benda LE (2008) Factors controllingavailability of spawning habitat for salmonids at the basinscale. In: Sear DA, DeVries P (eds) Salmonid spawninghabitat in rivers: physical controls, biological responses,and approaches to remediation. American Fisheries Society,Bethesda, pp 103–120

Mobrand LE, Barr J, Blankenship L, Campton DE, Evelyn TTP,Flagg TA, Mahnken CVW, Seeb LW, Seidel PR, SmokerWW, Hatchery G (2005) Hatchery reform in WashingtonState: Principles and emerging issues. Fish Rev Sci 30:11–23

Montgomery DR, Buffington JM (1998) Channel processes, clas-sification, and response. In: Bilby RJ, Naiman RE (eds) Riverecology and management. Springer, New York, pp 13–42

Montgomery DR, Beamer EM, Pess GR, Quinn TP (1999)Channel type and salmonid spawning distribution andabundance. Can J Fish Aquat Sci 56:377–387

Mull KE, Wilzbach MA (2007) Selection of spawning sites bycoho salmon in a northern California stream. N Am J FishManag 27:1343–1354

Murdoch AR, Pearsons TN, Maitland TW (2009a) Use ofcarcass recovery data in evaluating the spawning distributionand timing of spring Chinook salmon in the Chiwawa River,Washington. N Am J Fish Manag 29:1206–1213

Murdoch AR, Pearsons TN, Maitland TW (2009b) The number ofredds constructed per female spring Chinook salmon in theWenatchee River basin. N Am J Fish Manag 29:441–446

Murdoch AR, Pearsons TN, Maitland TW (2010) Estimating thespawning escapement of hatchery- and natural-origin springChinook salmon using redd and carcass data. N Am J FishManag 30:361–375

Neville HM, Isaak DJ, Dunham JB, Thurow RF, Rieman BE(2006) Fine-scale natal homing and localized movement asshaped by sex and spawning habitat in Chinook salmon:

insights from spatial autocorrelation analysis of individualgenotypes. Mol Ecol 15:4589–4602

Palmer MW (1993) Putting things in even better order - theadvantages of canonical correspondence analysis. Ecology74:2215–2230

Poole GC (2002) Fluvial landscape ecology: addressing unique-ness within the river discontinuum. Freshw Biol 47:641–660

Quinn TP (1993) A review of homing and straying of wild andhatchery-produced salmon. Fish Res 18:29–44

Quinn TP (2005) The behavior and ecology of Pacific salmonand trout. University of Washington Press, Seattle, p 378

Quinn TP, Stewart IJ, Boatright CP (2006) Experimental evidenceof homing to site of incubation by mature sockeye salmon,Oncorhynchus nerka. Anim Behav 72:941–949

Ricker WE (1972) Hereditary and environmental factors affectingcertain salmonid populations. In: Simon RC, Larkin PA (eds)The stock concept in Pacific salmon. University of BritishColumbia, Vancouver, pp 19–160

Schlosser IJ, Angermeier PL (1995) Spatial variation in demo-graphic processes of lotic fishes: conceptual models, empir-ical evidence, and implication for conservation. In: NielsenJL (ed) Evolution and the aquatic ecosystem: defining uniqueunits in population conservation. American Fisheries SocietySymposium, Bethesda, pp 392–401

Torgersen CE, Price DM, Li HW, McIntosh BA (1999) Multiscalethermal refugia and stream habitat associations of Chinooksalmon in northeastern Oregon. Ecol Appl 9:301–319

Torgersen CE, Baxter CV, Li HW, McIntosh BA (2006) Land-scape influences on longitudinal patterns of river fishes:spatially continuous analysis of fish-habitat relationships.Am Fish Soc Symp 48:473–492

Trexler JC, Travis J (1993) Nontraditional regression analyses.Ecology 74:1629–1637

Williamson KS, Murdoch AR, Pearsons TN, Ward EJ, Ford MJ(2010) Factors influencing the relative fitness of hatchery andwild spring Chinook salmon (Oncorhynchus tshawytscha) inthe Wenatchee River, Washington, USA. Can J Fish AquatSci 67:1840–1851

Wisby W, Hasler A (1954) Effect of olfactory occlusion onmigrating silver salmon (O. kisutch). J Fish Res BoardCan 11:472

Environ Biol Fish