Genetic assessment of a summer chum salmon metapopulation in recovery

Authors


Abstract

Programs to rebuild imperiled wild fish populations often include hatchery-born fish derived from wild populations to supplement natural spawner abundance. These programs require monitoring to determine their demographic, biological, and genetic effects. In 1990s in Washington State, the Summer Chum Salmon Conservation Initiative developed a recovery program for the threatened Hood Canal summer chum salmon Evolutionarily Significant Unit (ESU) (the metapopulation) that used in-river spawners (wild fish) for each respective supplementation broodstock in six tributaries. Returning spawners (wild-born and hatchery-born) composed subsequent broodstocks, and tributary-specific supplementation was limited to three generations. We assessed impacts of the programs on neutral genetic diversity in this metapopulation using 16 microsatellite loci and a thirty-year dataset spanning before and after supplementation, roughly eight generations. Following supplementation, differentiation among subpopulations decreased (but not significantly) and isolation by distance patterns remained unchanged. There was no decline in genetic diversity in wild-born fish, but hatchery-born fish sampled in the same spawning areas had significantly lower genetic diversity and unequal family representation. Despite potential for negative effects from supplementation programs, few were detected in wild-born fish. We hypothesize that chum salmon natural history makes them less vulnerable to negative impacts from hatchery supplementation.

Introduction

In the Pacific Northwest of the United States, fisheries managers are developing strategies to conserve and restore native biodiversity in Pacific salmonids. After decades of harvest beyond sustainable rates and loss of spawning and rearing habitats, over 20% of native salmon populations are imperiled or extinct (Augerot and Foley 2005). Salmon hatcheries have often been employed to mitigate losses of native fish, both as means to support native populations and to support harvest opportunities. Traditional hatcheries developed broodstocks whose progeny were released into nonancestral streams (i.e., nonlocal broodstocks); rearing and release practices promoted adaptation to hatchery environments such that nonlocal, hatchery-adapted hatchery fish fared poorly under natural conditions and potentially brought maladaptive traits into wild populations (Tessier et al. 1997; Hansen et al. 2000; McClure et al. 2008; Naish et al. 2008). Currently, impacts from hatchery fish are monitored by assessing relative fitness, genetic diversity, and effective population size (Ne) in supported populations (Wang and Ryman 2001; Hare et al. 2011). Recent studies have demonstrated lower fitness in populations receiving hatchery fish from traditional hatcheries, compared with local, wild populations (reviewed in Berejikian and Ford 2004; Araki et al. 2008; Christie et al. 2012). Studies suggested that hatchery fish with lower fitness or maladapted traits depressed the productivity of wild stocks with which they interbreed (Lynch and O'Hely 2001; Araki et al. 2007a, 2008). Further, introgression by common farmed broodstock into wild populations may decrease genetic distinction among wild populations (Hansen et al. 2009; Glover et al. 2012).

With concerns over the role of hatchery programs for restoring wild populations, supplementation hatchery programs were designed to decrease negative hatchery impacts. Supplementation hatcheries aim to temporarily increase spawner census size to conserve genetic resources and boost natural population abundance of imperiled wild populations while minimizing genetic and ecological risks commonly associated with traditional hatchery practices (Ford 2002; Goodman 2004). Supplementation programs developed to meet conservation objectives bring local, natural-origin adults, juveniles, or eggs into a hatchery to initiate the program for a specific tributary (Berejikian et al. 2008; Small et al. 2009) and release the hatchery-born juveniles into their specific tributary. When adult fish return to the supplementation program tributary, hatchery- and wild-born spawners intermix in the natural spawning area, and subsequent hatchery broodstocks for each target tributary are a mix of these hatchery- and wild-born spawners returning to their tributary. With supplementation hatcheries, it is hoped that increasing spawner abundance with local-origin supplementation fish will have positive effects of maintaining genetic diversity, increasing Ne, and supporting persistence and adaptive potential of targeted populations (Wang and Ryman 2001; Hedrick 2005; Berejikian et al. 2008). Supplementation programs have met with varying success because hatchery practices may counter intended benefits through high variance in hatchery family sizes and unequal sex ratios, which can decrease Ne and reduce genetic diversity of supplemented populations (Ryman–Laikre effects, Ryman and Laikre 1991; see Christie et al. 2012). Some evidence suggests lower reproductive success for natural-origin steelhead (Oncorhynchus mykiss Walbaum) whose parents arose from a supplementation-style hatchery in comparison with wild-origin steelhead with no hatchery ancestry (Araki et al. 2007b, 2008; Christie et al. 2012). Further, Chilcote et al. (2011) suggested that these hatchery impacts on wild salmon productivity were also found in coho salmon (Oncorhynchus kisutch Walbaum) and Chinook salmon (Oncorhynchus tshawytscha Walbaum) and occurred regardless of number of generations in the hatchery or the origin (local or exotic) of the hatchery broodstock. However, Sharma et al. (2006) observed increased productivity in supplemented wild coho salmon and Hess et al. (2012) found higher abundance and no fitness loss in supplemented wild Chinook salmon.

Although chum salmon (Onchorhynchus keta Walbaum) have a typical anadromous Pacific salmon life history, negative impacts from a supplementation hatchery program may be lower for chum salmon than for other salmonids. Chum salmon juveniles migrate to estuaries within days of emergence (Johnson et al. 1997), and hatchery chum salmon are typically released as fry after 2–3 months of rearing (chum salmon juveniles in this supplementation program were released into their target streams within 75 days of hatching). Thus, in contrast to other species, hatchery-born chum salmon spend minimal time under the selective forces within an artificial rearing environment and impacts might be primarily through broodstock selection, lack of mate choice, and unequal success of hatchery families. In this study, we explore impacts of supplementation programs in chum salmon by examining changes in genetic diversity and effective population sizes in a thirty-year time series starting in the 1970s that includes collections spanning the time from before, during, and after supplementation. Hatchery supplementation and reintroductions began in the 1990s for summer chum salmon in Hood Canal (HC) and Strait of Juan de Fuca (SJF) in Washington State in response to declines and extinctions (see Methods for details).

Materials and methods

Chum salmon life history

Chum salmon spawn mainly in lower tributary reaches up to barrier falls in coastal tributaries along the northern periphery of the Pacific Ocean, as far south as Japan and Oregon. Similar to other Pacific salmon, chum salmon are semelparous and return to spawn generally after 3 or 4 years in salt water, but returns after 2 or 5 years are not uncommon. In contrast to other Pacific salmon (except pink salmon), chum salmon juveniles spend little time in freshwater, migrating downstream to estuaries and salt water a few weeks after emergence. Chum salmon stray at similar rates to other Pacific salmon (Johnson et al. 1997), but high spawner densities may increase stray rates (reviewed in Quinn 1993).

Hood Canal summer chum salmon

Within the Puget Sound region in Washington State, summer-run chum salmon in HC and SJF (Fig. 1) are genetically and ecologically distinct from fall-run chum salmon in the region (Phelps et al. 1994; Tynan 1997) and are designated a separate ESU (Johnson et al. 1997). Due to population declines and extinctions, the HC summer chum salmon ESU was listed as threatened under the Endangered Species Act in 1999 (http://www.nwr.noaa.gov/ESA-Salmon-Listings/Salmon-Populations/Chum/Chum-Status-Reviews.cfm). A recovery program started in the early 1990s, prior to the ESA listing, enacted crucial harvest reductions and supplementation programs. In 2000, state and tribal co-managers completed the Summer Chum Salmon Conservation Initiative (SCSCI; WDFW and PNPTT 2000), a recovery plan that formalized and expanded recovery efforts. Two independent summer chum salmon populations were designated as ESU recovery units: HC and SJF; both consisted of multiple subpopulations in rivers draining to those two marine basins and both were monitored for status relative to recovery goals. There had been small releases of hatchery summer chum salmon juveniles in various HC tributaries before 1938 (Johnson et al. 1997), but the recovery program was the first directed hatchery intervention in most tributaries.

Figure 1.

Map of Hood Canal, Strait of Juan de Fuca and portions of Puget Sound. Rivers and streams are numbered and asterisks mark tributaries in the study. At right are listed the tributary names associated with numbers and the status of subpopulations in the tributaries. Map was modified from Sands et al. (2009).

As part of efforts to restore wild subpopulations of summer chum salmon, supplementation hatchery programs were initiated for three subpopulations (Lilliwaup, Quilcene, and Salmon) in 1992 and for three more subpopulations (Hamma Hamma, Jimmycomelately, and Union) in 1997, 1999, and 2000, respectively (each supplementation program used wild-origin broodstock collected from its respective tributary). Summer chum salmon from three streams with supplementation programs (Quilcene, Salmon, and Union) were also the donor stocks for programs used to reintroduce summer chum salmon in one tributary each (Big Beef, Chimacum, and Tahuya, respectively) where native populations had gone extinct. Two subpopulations, Dosewallips and Duckabush, were not deliberately supplemented. Supplementation programs were designed to reduce extinction risk and to speed recovery and recolonization while minimizing risks of deleterious genetic, ecological, and demographic effects to supplemented and unsupplemented subpopulations (WDFW and PNPTT 2000; Tynan et al. 2003). Supplementation hatcheries were scheduled to run for a maximum of three generations (12 years), and number of fry released into each tributary was based on available habitat. To best represent the demographics of each local donor population and to maintain existing genetic diversity, programs followed protocols (Schroder and Ames 2004) in which supplementation broodstocks for each tributary were collected from local spawners randomly in proportion to the timing, weekly abundance, and duration of the total return from the same tributary where hatchery-propagated fish were to be released (each tributary had its own supplementation program all with the goal of equal sex ratios and using no more than 50% of in-river spawners; see Table SI in Supplementary Information for broodstock details for each program). Matings employed partial factorial designs (Campton 2004; Busack and Knudsen 2007) to maximize genotypic diversity and effective subpopulation size (Waples and Do 1994; Withler and Beacham 1994). All hatchery fish were mass-marked either externally (fin-clipped, Quilcene program) or internally [all other programs received unique hatchery-specific otolith banding pattern marks (see Volk et al. 1987)]. Marks identified hatchery fish back to specific hatchery of origin from observed fin-clips or from otoliths collected from carcasses during annual spawner surveys (WDFW and PNPTT 2000; WDFW and PNPTC 2007). Because most hatchery fish were marked internally, after supplementation fish began to return, broodstock selected from returning spawners included unknown numbers of hatchery-born fish that were identified post-mortem. All supplementation program releases were mass-marked beginning with brood year 1997 (and chum salmon generally return after 3 and 4 years), and from 2001 onward managers identified hatchery-born spawners from internal and external marks (Table 2). Although we have no direct measure of the reproductive success for hatchery-born spawners, the proportion of hatchery-born spawners was an indication of maximum expected hatchery influence.

Collections

Collections consisted of summer chum salmon from HC and SJF subpopulations (Fig. 1; Table 1). Tissue samples for this study were collected from spawners in seven rivers and creeks in HC and two creeks in SJF from 1978 through 2009 (Table 1). Samples collected prior to initiating supplementation programs were termed ‘original’ and samples collected after supplementation fish returned were divided by birth location, ‘hatchery’ or ‘wild’. A few collections included samples from more than 1 year because there were too few samples per individual year (Table 1). The following summer chum salmon subpopulations had supplementation programs: Union, Lilliwaup, Hamma Hamma, Big Quilcene, Salmon, and Jimmycomelately (see Table 2 for numbers of hatchery-born fry planted per year and see Table S2 in Supplementary Information for years when hatchery-born spawners returned). Although Big Beef Creek summer chum salmon were extirpated by 1990, we analyzed tissues (archived scales) collected in the creek prior to extirpation (1978, 1979) and after reintroduction with Quilcene stock (2004). Summer chum salmon in Dosewallips and Duckabush rivers (unsupplemented) were monitored for comparisons with supplemented subpopulations and to document and identify strays from supplementation programs. Spawners from all HC and SJF summer chum salmon subpopulations were censused annually for supplementation hatchery- and wild-born spawners in each river (post-mortem otolith analyses identified hatchery-born spawners and their specific supplementation programs). Marked otoliths or fin-clip ratios from spawner carcasses estimated the number and proportion of hatchery-born spawners (Table 2), and scales were collected to age spawners as part of biological data for recovery monitoring (WDFW and PNPTC 2007).

Table 1. Genetic statistics for summer chum salmon collections
RegionTributaryYearName N F IS P valueLinkage (120 pairs)
5%1%
  1. Regions are abbreviated: Hood Canal = ‘HC’, Strait of Juan de Fuca = ‘SJF’. Name abbreviations used throughout document are in ‘Name’ column. Name includes collection category (O = original prior to supplementation program, H = Hatchery-origin, and W = Wild-origin during or after time of hatchery supplementation), and * indicates collections from tributaries that were not deliberately supplemented. Statistics include the Hardy–Weinberg equilibrium value expressed by FIS, and its associated P value (underlined FIS values significant before Bonferroni corrections, and bold value was significant after Bonferroni corrections), the number of locus pairs (out of 120 pairs) in linkage disequilibria at the 5% and 1% level (values in gray boxes were >5% of total pairs). Other genetic statistics, gene diversity and allelic richness, are presented in Table 4.

HCBig Beef197878BigB_O44−0.0070.652683
HCBig Beef197979BigB_O430.0450.006693
HCBig Beef200404BigB_H45−0.0030.5768174
HCUnion200000Union_O540.0160.1570112
HCUnion200303Union_W32−0.0150.763850
HCUnion200303Union_H300.0130.2706219
HCUnion200404Union_H430.0140.2255289
HCUnion200808Union_W47−0.0220.9165103
HCLilliwaup198585Lilli_O23−0.0370.9450116
HCLilliwaup199292Lilli_O460.0070.3291153
HCLilliwaup1997–9997_99Lilli_W35−0.0170.8032187
HCLilliwaup2000–0100_01Lilli_W350.0190.1504173
HCLilliwaup200101Lilli_H26−0.0060.591571
HCLilliwaup200202Lilli_H47−0.0200.87259669
HCLilliwaup2003–0403_04Lilli_H53−0.0060.65588438
HCLilliwaup2005–0605_06Lilli_W29−0.0330.9512269
HCLilliwaup200505Lilli_H600.0000.5112162
HCLilliwaup200606Lilli_H550.0220.0698112
HCLilliwaup200808Lilli_H420.0070.3315123
HCLilliwaup200909Lilli_H290.0250.1052204
HCHamma Hamma199999Hamma_O340.0330.0399224
HCHamma Hamma200101Hamma_W530.0010.466682
HCHamma Hamma200303Hamma_W40−0.0130.7611112
HCHamma Hamma2001–0301_03Hamma_H270.0180.1830104
HCHamma Hamma200808Hamma_W410.0000.4835101
HCDosewallips199292Dose_O48−0.0120.783762
HCDosewallips2000*00Dose_W530.0290.0369112
HCDosewallips2003*03Dose_W460.0480.001563
HCDosewallips2009*09Dose_W33−0.0110.713770
HCDuckabush198686Duck_O570.0140.1713124
HCDuckabush199292Duck_O46−0.0070.683792
HCDuckabush2000*00Duck_W460.0370.0141132
HCDuckabush2003*03Duck_W470.0180.135293
HCDuckabush2009*09Duck_W340.0180.1637102
HCQuilcene199292Quil_O50−0.0120.782431
HCQuilcene2008–0908_09Quil_W450.0090.2346124
SJFSalmon198686Salmon_O420.0240.6322144
SJFSalmon200000Salmon_W250.0270.0938124
SJFSalmon200000Salmon_H320.0030.1133193
SJFSalmon2003–0503_05Salmon_W340.0050.43241710
SJFSalmon200303Salmon_H270.0180.4063219
SJFSalmon2004–0504_05Salmon_H240.1310.201040
SJFSalmon200808Salmon_W250.0180.0000162
SJFSalmon200909Salmon_W230.0140.7968103
SJFJimmycomelately198686JCL_O62−0.0350.27544112
SJFJimmycomelately1998–9998_99JCL_O16−0.0190.9910165
SJFJimmycomelately200000JCL_O380.0190.72566028
SJFJimmycomelately200101JCL_W580.0090.16884522
SJFJimmycomelately2001–0403_04JCL_H75−0.0170.27928049
SJFJimmycomelately2008–0908_09JCL_W330.0010.8896113
SJFJimmycomelately200808JCL_H20−0.0280.4732125
SJFJimmycomelately200909JCL_H34−0.0420.8736196
Table 2. Number of hatchery-origin fry (for brood year) released into supplemented tributaries (fry were offspring of spawners collected in their tributary the previous fall and spawned in hatchery) and summer chum salmon escapements (wild-origin ‘wild’ and supplementation hatchery-origin ‘HOS’ spawners arriving at spawning grounds) into tributaries in Hood Canal (HC) and Strait of Juan de Fuca (SJF) from 1990 through 2011. Data are from WDFW and WWTIT (2002), WDFW and PNPTT (2007), WDFW Hatchery planting database, and FishBooks (Are Strom and Kelly Henderson, WDFW)
HCUnion Lilliwaup Dosewallips Duckabush Hamma Hamma Quilcene 
Escapement Escapement Escapement Escapement Escapement Escapement
YearWildHOSfryWildHOSfryWildHOSfryWildHOSfryWildHOSfryWildHOSfry
1990275  2  8  42  90  6  
1991208  30  250  102  71  50  
1992140  99 20 000655  617  123  743 216 441
1993251  77 12 000105  105  69  148 24 784
1994738  111 15 000225  263  370  722 343 550
1995721  79 02787  825  476  30571517441 167
1996494  76 15 0006976  2650  774  78051710612 598
1997410  27 14 20047  475  111 12 00052312672340 744
1998223  24 17 200336  226  127 280015951458343 530
1999159  13 17 400351  92  255 51 60015971640181 711
2000744 75 87622 14 8001260  464  229 55 40031152783414 353
20011491073 472415138 000757233 662280 11557249 50030483325351 709
2002872082 6363682296 0001313314 355175 1050127861 0003 2111276272 017
20037906401035 34327326103 9136510556 1600269 53631875 35610 740199392 559
200435982378 13688199 50010 2841265 7850787 240928257 00035 8382315 
20057041283 259790106 4662496162 75269 1185226117 8375920838 
200616671170 426118988 8002457120 2964171 2707358151 55010 881995 
2007188978 153372014626 127024 14167348 530247947 
2008104387 14748968 8103828102 2517151 1371256208 45038610 
200959714 60186140 210109334 2499160 59172 14920 
20108970 18850139 81625210 41100 1370101 2073NA 
201128016 7536NA11300 15060 68587 25800 
SJFSalmon JCL
Escapement Escapement 
YearWildHOSfryWildHOSfry
1990245  63  
1991172  125  
1992433 19 200616  
1993452 44 000110  
1994161 2 00015  
1995591 38 808223  
1996894 62 00030  
19977686671 82161  
199860552967 83298  
199913236734 6807 3 880
200043940790 43555 25 900
20011168147092 415251954 515
200237451772117 79724020 887
20033785186688 6106837850 307
200441031918 613104976 982
200538572285 49381757 300
20064326568 34538079 428
2007124331 46518573 840
2008154424 57348188 766
2009121819 526210292 200
201027400 739328885 630
2011226811 8141597 

Genotyping

We genotyped fish at 16 microsatellite loci [Oke-3 (Buchholz et al. 2001), Oki-1(Smith et al. 1998), Omy-1011 (Spies et al. 2005), One-101, One-102, One-106, One-108, One-111, One-114 (Olsen et al. 2000), One-18 (Scribner et al. 1996), Ots-1, Ots-2M, Ots-3M (Banks et al. 1999), Ots-103 (Small et al. 1998), Ots-G311 (Williamson et al. 2002), and Ssa-419 (Cairney et al. 2000)] for 2057 individuals from 43 collections (Table 1). DNA was extracted using silica membranes following the manufacturer's instructions (Macherey-Nagel). Microsatellite loci were amplified in six multiplexes, and alleles were scored by two researchers prior to export for analyses (see Small et al. 2009 for details of PCR, scoring, and binning).

Statistical tests

Genetic statistics were calculated to examine whether collections met expectations of random sampling and to identify any problems with loci such as disequilibrium from large-allele dropout, null alleles, or scoring problems. Statistics assessed differences in genetic diversity between original samples, supplementation program samples (hatchery-born spawners and wild-born spawners collected in the same spawning areas), and samples from unsupplemented tributaries and evaluated changes following supplementation using tests described below. The software COLONY2.0.0.1 (Wang 2008) calculated full-sibling family structure that might contribute to Ryman–Laikre effects through unequal representation of hatchery families among spawners. Samples were tested for departures from Hardy–Weinberg equilibrium (HWE) at each locus and across all loci using FSTAT 2.9.3 (Nei 1987; Goudet 2001) with 1000 permutations. Departure from HWE can be an indication that samples contain family groups, or a strong year class, or more than one subpopulation, or related parents in the previous generation, or locus problems such as null alleles or selection. We tested whether genotypes at each locus were independent with the linkage disequilibrium permutation test in GENETIX 4.03 (Belkhir et al. 2001) using 500 permutations. For each sample, GENETIX calculated the number of loci pairs in which 5% and 1% of the permutations had a smaller value than the actual value for the sample. Linkage disequilibrium can be an indication that the sample contains family groups or the offspring of matings of genetically distinct populations, or that the population is under selection, or that alleles have drifted due to a small subpopulation size. Basic genetic diversity measures including gene diversity (Nei's (1987) estimate of heterozygosity corrected to minimum eight individuals) and allelic richness (average number of alleles per locus corrected to minimum eight individuals) were calculated using FSTAT. We used Wilcoxon signed-rank tests to test for significant differences among sample categories in gene diversity, allelic richness, and Ne that might signal effects associated with supplementation. Results for all tests were adjusted for multiple comparisons (sequential Bonferroni correction, Rice 1989) to an alpha level of 0.05.

Two methods were used to estimate Ne; Ne is a fundamental parameter determining the evolutionary potential of a population, yet different techniques employ different assumptions and yield somewhat different results. Tracking changes in Ne may indicate changes associated with supplementation (Antao et al. 2010; Hare et al. 2011). We estimated Ne using linkage disequilibrium (LD-Ne, Waples 2006) in the program LDNe (Waples and Do 2008; setting the lowest frequency allele at 2% to avoid bias introduced by small collections) and estimated Ne using a method based on maximum likelihood pairwise sibship analysis (SA-Ne, Wang 2009) implemented in the program COLONY 2.0.0.1. Because chum salmon have overlapping year classes, the values from these measures estimated the number of breeders (Nb) giving rise to the collection, rather than the actual Ne. To calculate actual Ne, the Nb is multiplied by the generation time, which is 3–5 years in chum salmon. Because generation time is variable in chum salmon (especially variable in these recovering subpopulations), we simply referred to the values calculated for a single-year mixed-age collection by LD and SA as Ne and did not multiply by generation time. Because assumptions and results varied somewhat between methods, we calculated a harmonic mean (and variance) of Ne values (HmNe) from the two methods, following Waples and Do (2010). We also calculated the ratio of HmNe to the census size for spawner types as a metric to investigate differences among sample types and changes over time (Hedrick 2005).

Effective population size analyses were conducted from a metapopulation perspective (Gomez-Uchida et al. 2013): we calculated the metapopulation Ne before supplementation for the earliest original program samples (00Union_O, 85Lilli_O, 99Hamma_O, 92Quil_O, 86Salmon_O and 86JCL_O) and after supplementation for the latest collected wild-born samples (08Union_W, 05_06Lilli_W, 08Hamma_W, 08_09Quil_W, 09Salmon_W, and 08_09JCL_W). The metapopulation Ne was first calculated according to the Stepping Stone model (Maruyama 1970) and the Island model (Wright 1943) using Ne values from the LD-Ne and SA-Ne and then we calculated the harmonic mean of the metapopulation LD-Ne and SA-Ne for final metapopulation HmNe values (meta-HmNe) based on the Stepping Stone and Island models. For the Stepping Stone model the average immigration rate was estimated from the global FST values for the two sets of six collections (original and wild-born), and the same global FST values were used in denominators for the Island model.

Subpopulation comparisons

A variety of measures assessed whether supplementation had induced changes in genetic attributes and genetic structure and whether strays from supplementation programs into unsupplemented tributaries had altered population structure or substructure. To assess possible changes in genetic attributes (heterozygosity, allelic richness, and HmNe), we used Wilcoxon sign rank tests to test for significant differences in these genetic attributes among collection categories; original, supplementation hatchery-born, and wild-born (see Table 1). We tested for significant differences in genotypic distributions among all samples using GENEPOP3.3 (Raymond and Rousset 1995), and examined temporal and spatial partitioning of pairwise genetic variance with pairwise FST tests in GENETIX 4.03 (Belkhir et al. 2001). In pairwise FST tests we evaluated whether variance was significantly different from zero with 1000 permutations. A permutation test implemented in FSTAT (comparison among groups of samples) was used to test for significant differences in global FST values among categories of samples (original, supplementation hatchery-born and wild-born, see Table 1), with 10 000 permutations. To assess changes associated with supplementation, we compared pairwise FST values among original samples and among wild-born samples collected after supplementation programs had been initiated. These tests were conducted with and without samples from Dosewallips and Duckabush, the tributaries that were not deliberately supplemented. To assess changes in variance associated with hatchery supplementation programs, we compared pairwise FST values among samples of hatchery-born spawners and wild-born spawners collected in the same tributaries.

Genetic distances

We assessed subpopulation structure using a principle coordinates analysis (PCoA) implemented in GenAlEx 6.5 (Peakall and Smouse 2006; Peakall and Smouse 2012). GenAlEx calculated pairwise FST values among populations over all loci and conducted a PCoA of the pairwise values. The PCoA is a multivariate ordination technique that describes underlying patterns in a dataset. The PCoA generates axes describing genetic variance in a dataset, here pairwise FST values, with the first two axes usually describing the maximum variance. As another means to view genetic relationships, we plotted Cavalli-Sforza and Edwards (1967) chord distances among collections in a dendrogram with 10 000 bootstrap replications using programs within PHYLIP (Felsenstein 1993).

Assignment tests

To investigate whether fish sampled from the same tributary were more or less likely to be genetically similar to each other following supplementation, we used self-assignment tests in GeneClass2 (Piry et al. 2004). GeneClass employs the Rannala and Mountain (1997) algorithm in a ‘leave one out’ protocol and calculates the likelihood that an individual fish originated in the subpopulation in the tributary where it was sampled (home collection) based on the genotype of the fish and allele frequencies in the baseline collections, with the fish in question removed from its home collection in the baseline. There was no threshold likelihood value for assignment, the highest likelihood was accepted as the assignment and self-assignments were conducted on collections before and after supplementation. We tested for differences in percentage of assignments to home collection before and after supplementation using paired Student's t-tests.

Isolation by distance

Data were compared before and after supplementation for evidence of changes in isolation by distance (IBD) patterns (Slatkin 1993), using transformed pairwise FST values [FST/(1-FST)] as a genetic similarity measure (Rousset 1997). Geographical distances (kilometers) between mouths of streams were calculated using the most direct passage over open water. Mantel tests for association between pairwise FST values and distance, and reduced major axis regressions were performed using IBD 3.23 (Bohonak 2002). To assess changes that might have occurred as a result of supplementation, we used ancova to compare IBD patterns among original collections to IBD patterns among collections of wild-born fish after supplementation began. We also compared IBD patterns of wild-born samples and hatchery-born samples to evaluate differences between the sample groups. We conducted analyses with and without Dosewallips and Duckabush collections because strays from hatchery programs were documented in spawner surveys in those tributaries and may have supplemented those populations.

Results

Supplementation and escapement

Supplementation programs boosted abundance of spawners and contributed from 5% to 96% of spawners to escapements in supplemented HC and SJF subpopulations (Table 2). The proportion of hatchery-born spawners decreased as successful programs terminated according to protocols (WDFW and PNPTT 2007). From 2001–2011, strays from supplementation programs contributed from 0% to 33% (average = 10%) of spawners to Dosewallips and Duckabush (Table 2), which were not deliberately supplemented. Most strays to Duckabush and Dosewallips were from nearby supplementation programs in Hamma Hamma and Quilcene (WDFW and PNPTT 2007). Similarly, most strays in other tributaries were from nearby supplementation programs and straying may be influenced by the natural exchange rate among these subpopulations.

Subpopulation statistics

Genotypic coverage averaged 95% and ranged from 84% (One-106) to 97% (Oki-1) per locus over all individuals (N = 2086). Because samples (tissues and archived scales) were from spawner carcasses of varying freshness, genotyping success varied among samples, regardless of contemporary or historical status. In tests for Hardy–Weinberg equilibrium (HWE) at individual loci (Supplementary Information, Table S3), two tests out of 832 were significant after corrections for multiple tests and all samples except 08Salmon_W were in HWE in tests over all loci (Table 1). The Wilcoxon sign rank tests indicated no significant differences between original and wild-born samples in genetic diversity measures (gene diversity, allelic richness) or HmNe (Fig. 2, Table 3, see Tables 1 and 4 for individual collection values). However, hatchery-born spawners had lower genetic diversity and significantly lower HmNe than wild-born spawners (differences in genetic diversity were not significant when Dosewallips and Duckabush were excluded from wild samples but differences in HmNe between wild- and hatchery-born samples remained significant).

Table 3. Table of averages (avg) and harmonic means (hmean) of genetic statistics for categories of samples (original = O, hatchery-born = H, wild-born = W) detailed in Tables 1 and 4, and P values for Wilcoxon signed-rank tests for comparisons
  OWP valueNo DoseDuck 
OWP value
Gene Diversityavg0.79960.80530.43320.79460.80270.2259
Allelic Richnessavg6.987.150.23986.847.050.1819
HmNehmean52.4176.990.133245.0569.570.0629
Link 5%hmean10.0010.720.500010.5911.810.4896
  HWP valueHWP value
  1. Analyses were conducted with and without Duckabush and Dosewallips samples (DoseDuck). Gene diversity is Nei's estimate of heterozygosity corrected to a sample size of 6, and allelic richness is average number of alleles per locus corrected to a samples size of 6.

Gene Diversityavg0.79690.80530.04570.79690.80270.1774
Allelic Richnessavg6.827.150.00606.827.050.0585
HmNehmean36.1176.990.000236.1169.570.0039
Link 5%hmean14.2510.720.019214.2511.810.0794
Table 4. Population statistics for summer chum salmon collections including gene diversity (‘GeneDiv’, Nei's estimate of heterozygosity corrected to a sample size of 8) and allelic richness (‘AR’, average number of alleles per locus corrected to a sample size of 8)
RegionNameGeneDivARHmNe±2SDEscapeHmNe/N
  1. Harmonic mean Ne (HmNe) was calculated for Ne's from linkage disequilibrium (LD-Ne) and pairwise sibship analysis (SA-Ne). Escapement (Escape) was calculated using area under the curve and fish counts at traps and sums wild- and supplementation hatchery-origin escapements to natural spawning grounds. The HmNe/N is the ratio of HmNe to escapement.

HC78BigB_O0.79996.986796800.10
HC79BigB_O0.81587.253661910.19
HC04BigB_H0.80397.0658919160.03
HC00Union_O0.79136.8687107440.12
HC03Union_W0.78016.55882279060.01
HC03Union_H0.78586.68521040100.01
HC04Union_H0.80436.7248923780.02
HC08Union_W0.80137.05931910430.09
HC85Lilli_O0.81747.098818920.96
HC92Lilli_O0.80567.209513990.96
HC97_99Lilli_W0.78916.82307640.47
HC00_01Lilli_W0.81267.335110630.81
HC01_Lilli_H0.79967.217118511.39
HC02Lilli_H0.76626.03838220.01
HC03_04Lilli_H0.79376.8115512070.01
HC05_06Lilli_W0.80687.2259116850.09
HC05Lilli_H0.80627.165397900.07
HC06Lilli_H0.81367.34901011890.08
HC08Lilli_H0.80967.0975144890.15
HC09Lilli_H0.80787.3181151860.44
HC99Hamma_O0.82397.5880112550.31
HC01Hamma_W0.80787.371091411550.09
HC03Hamma_W0.80457.1783165360.15
HC01_03Hamma_H0.81517.4271143900.18
HC08Hamma_W0.81277.561633313710.12
HC92Dose_O0.81777.54151226550.23
HC00Dose_W0.81627.50911112600.07
HC03Dose_W0.81787.531202465100.02
HC09Dose_W0.81677.591061910930.10
HC86Duck_O0.81997.46101102340.43
HC92Duck_O0.81697.46155226170.25
HC00Duck_W0.81217.32100134640.22
HC03Duck_W0.80267.27921516000.06
HC09Duck_W0.80627.341332424990.05
HC92Quil_O0.80637.35125167430.17
HC08_09Quil_W0.81097.2471953530.01
SJF86Salmon_O0.78426.674365820.07
SJF00Salmon_W0.81167.06133644390.30
SJF00Salmon_H0.80696.7464134070.16
SJF03_05Salmon_W0.81737.31901376420.01
SJF03Salmon_H0.79656.7036918660.02
SJF04_05Salmon_H0.81346.77671242030.02
SJF08Salmon_W0.82397.331112515440.07
SJF09Salmon_W0.79916.92882812180.07
SJF86JCL_O0.77766.233242920.11
SJF98_99JCL_O0.75935.992171050.20
SJF00JCL_O0.75896.00226550.40
SJF01JCL_W0.77956.273642510.14
SJF03_04JCL_H0.76656.0225514270.02
SJF08_09JCL_W0.78316.46651110990.06
SJF08JCL_H0.79366.6756154810.12
SJF09JCL_H0.76386.21461021020.02
Figure 2.

Plots of genetic statistics for samples grouped into original, hatchery, and wild categories. Genetic statistics are detailed in Tables 1 and 2.

Linkage disequilibrium and pairwise sibship analyses suggested most collections included some family groups, ranging from a pair of siblings to large families (data not shown). There were 120 locus pairs examined per collection, and most collections had six or more locus pairs in linkage disequilibrium at the 5% level and 12 collections had six or more locus pairs in linkage disequilibrium at the 1% level (Table 1). The Wilcoxon signed-rank test indicated that hatchery-born spawners had significantly more locus pairs in linkage disequilibrium than wild-born spawners (P = 0.019). The collections from 02Lilli_H, 03_04Lilli_H, and 03_04JCL_H were notable for high linkage disequilibrium: COLONY calculated a single family with 14 full-siblings in 02Lilli_H (12% of pairwise relationships were full-siblings), seven families with between three and six full-siblings in 03_04Lilli_H, and six families with between three and eight full-siblings in 03_04JCL_H (4% of pairwise relationships were full-siblings). Siblings in collections were noted but not removed.

Effective population size calculations

We calculated Ne and its 95% parametric confidence interval for each collection with two methods: linkage disequilibrium (LD-Ne) and pairwise sibship analysis (SA-Ne). In most calculations, the two values were similar and the 95% confidence intervals overlapped (see data for both methods for each collection in Supplementary Information Table S4 ‘All_Ne’). Differences arose for collections with high linkage and small samples sizes, which biased LD-Ne downward (Wang and Whitlock 2003; Waples and Gaggiotti 2006), and where data were mostly lacking for a single locus, which depressed SA-Ne in 79Big Beef sample. The harmonic means of Ne values (HmNe) varied over time and space, both within and between tributaries (Table 4, Fig. 3). Where there were samples from hatchery-born and wild-born spawners collected in the same tributary in the same year, the hatchery-born samples generally had a smaller HmNe (e.g. 03Union_H and 03Union_W), except for samples from Lilliwaup Creek: in two comparisons of hatchery- and wild-born samples for that creek, the HmNe for the hatchery-born samples was equal to or greater than the HmNe for the wild-born sample. In the original Lilliwaup samples, the HmNe values were similar to those of Union River, but they declined to less than half by 1997. We lack wild-born samples from Lilliwaup Creek after 2006, but in the most recent hatchery-born samples, the HmNe values were similar to original values, suggesting that diversity may be recovering in the Lilliwaup Creek subpopulation. The HmNe in the original Jimmycomelately Creek sample was roughly half the value of the most recent wild sample, suggesting that diversity is also recovering in the Jimmycomelately Creek subpopulation. The calculated ratios of HmNe to census size (HmNe/N in Table 4) were significantly lower in wild-born samples than in original samples (Wilcoxon sign rank test, P = 0.0011 and P = 0.0137 with Dosewallips and Duckabush collections included and excluded, respectively), likely reflecting increases in census size (N) throughout the restoration program. Further, in an anova the ratio of HmNe to census size for the wild-born fish was negatively correlated with the number of years of supplementation (F1,68 = 11.2, P = 0.001).

Figure 3.

Principle coordinates plot of pairwise FST values among summer chum salmon collections from Hood Canal and Strait of Juan de Fuca. Name abbreviations follow Table 1 and colors for categories follow Fig. 2, with the exception of Big Beef (in black): ‘78 and 79 Big Beef’ collections were sampled prior to extinction, and the ‘04 Big Beef’ collection was derived from Quilcene (Quil) broodstock.

The global FST values for the six original and six wild-born samples were 0.029 and 0.021, respectively (P = 0.2414), resulting in calculated immigration rates of 8.37 and 11.65, respectively. The meta-HmNe for the original and wild-born samples was 488 and 591, respectively, with the Stepping Stone model, and 502 and 603, respectively, with the Island model (see Supplemental Information Table S4 for LD-Ne and SA-Ne values of six original and six wild-born samples and calculations for meta-HmNe values). Paired t-tests for LD-Ne and SA-Ne values indicated no significant differences between values for original and wild-born fish (P = 0.1873 for LD-Ne values for original versus wild-born fish, P = 0.3697 for SA-Ne values for original versus wild-born fish). Although meta-HmNe increased following supplementation, the difference was not significant.

In one further consideration of Ne, we computed the harmonic mean of the calculated Ne values over all collections with hatchery-born and wild-born spawners combined in collections and separated (see Supplementary Information, Table S5 ‘All_HmNe’). The calculated value for the wild-born spawners (78.95 ± 33.86) was larger than the value calculated for the hatchery- and wild-born spawners combined (55.19 ± 37.44), but the confidence intervals overlapped. Also, within some single-year examinations, the ratios of HmNe/N were higher in uncombined than in combined collections, suggesting Ryman–Laikre effects from combining hatchery- and wild-born fish into single collections.

Genetic variance patterns within and between subpopulations

Pairwise genotypic and FST tests were mostly congruent (see Supplementary Information Table S6 for pairwise test values and their associated P values). Pairwise tests indicated temporal stability within most subpopulations, with the exception of Lilliwaup Creek (see Fig. 3 for a plot of pairwise FST values along principle coordinate axes). Among Lilliwaup Creek samples, 37 of 66 pairwise genotypic comparisons were significant and 34 of 66 pairwise FST comparisons were significant. Union River samples also differed from all other HC samples and are the most isolated geographically (Fig. 1). There was little differentiation among samples from Dosewallips, Duckabush, Hamma Hamma, and Quilcene rivers. The SJF samples differed from each other and were distinct from HC samples. The original 1970s Big Beef Creek samples were most similar to those of Union River and Lilliwaup Creek, and the reintroduced sample was most similar to the sample from Quilcene River, its broodstock source, and consequently similar to Dosewallips, Duckabush, and Hamma Hamma rivers samples (Fig. 3).

Global FST values for different categories of samples did not change significantly throughout the program (Table 5). Although the global FST value for wild-born samples was lower than the value for original samples, the decrease following supplementation was not significant. The global FST for original samples matched the global FST for the hatchery samples, suggesting that the hatchery programs captured similar genetic variance among subpopulations. Including original and wild-born collections from Dosewallips and Duckabush in computations lowered global FST values, but changes were still not significant.

Table 5. Table of global FST values for categories of samples and P values for permutation test comparison among category values
 N samplesGlobal FSTP value
  1. The ‘N samples’ are the number of samples in the category. Big Beef Creek samples were excluded because the program was a reintroduction.

Original, all samples120.0240.294
Wild, all samples210.020
Original, no Dose no Duck90.0320.122
Wild, no Dose no Duck150.021
Original, only Dose and Duck30.0010.522
Wild, only Dose and Duck60.003
Hatchery160.0320.098
Wild, no Dose no Duck150.021

Genetic clusters identified in PCoA and dendrogram

Population centers formed two major clusters in the principle coordinates analysis plot (Fig. 3), and there was no distinction detected between hatchery and wild samples from the same tributaries. The first axis explained 45% of the genetic variance: SJF samples clustered on the left, and HC samples clustered on the right. The second axis explained 12% of the variance: The HC populations divided loosely into two to three clusters along the second axis. The lowest cluster included samples from Dosewallips, Duckabush, Hamma Hamma, and Quilcene rivers as well as the 2004 collection from Big Beef Creek. The topmost cluster included the original samples from Big Beef Creek (these separated along the third axis; the 1979 collection was missing most of one locus and plotted distantly from the 1978 collection), Union River samples, and the older samples from Lilliwaup Creek. The more recent Lilliwaup samples clustered mainly in the center, supporting high genetic drift in this population. The placement of the original Big Beef samples indicated that before reintroduction, summer chum salmon in Big Beef Creek were genetically more similar to other subpopulations from low-elevation tributaries on the east side of HC rather than the geographically closer subpopulations in tributaries originating in the Olympic Mountains (Dosewallips and Duckabush) on the west side of HC. A dendrogram of genetic distances among samples (Fig. 4) displayed similar overall structure. The SJF and HC clusters separated with 100% bootstrap support, with JCL forming a supported sub-branch. The Dosewallips, Duckabush, Hamma Hamma, and Quilcene samples formed an unsupported cluster, and the original samples from Lilliwaup Creek clustered with 79% bootstrap support on a branch with Union River and the original Big Beef Creek samples. The remaining Lilliwaup Creek samples clustered on three unsupported branches. There were no distinctions evident between hatchery-born and wild-born samples collected in the same tributaries.

Figure 4.

Neighbor-joining dendrogram of Cavalli-Sforza and Edwards chord distances among samples. Bootstrap values over 70% are plotted on the tree nodes.

Assignment test

Assignment tests supported a metapopulation structure with straying among close subpopulations, especially along the west side of HC (Dosewallips, Duckabush, Hamma Hamma, and Quilcene, Table 6). Where high gene flow occurs, an individual may be equally likely to have been sampled in two or more baseline populations or may assign to a closely related population. In presupplementation collections, self-assignments were low to indistinguishable from random in Dosewallips, Duckabush, Hamma Hamma, and Quilcene rivers. If we combined results from these rivers, assignments back to the combined ‘HoodMet’ were high (Table 6). Self-assignments were relatively high for Salmon and Jimmycomelately creeks and moderate in Union, Lilliwaup, and Big Beef creeks. Following supplementation, self-assignments decreased in all but Union River and Hamma Hamma River, but the decrease was not significant (Student's paired t-test, P = 0.054).

Table 6. Table of assignments using GeneClass2
 Presupplementation assignments
78_79Big Beef_W00Union_W92Lilli_W92Dose_W92Duck_W99Hamma_W92Quil_W86Salmon_W00JCL_WTotal 
78_79Big Beef 55 41137421087 
00Union11 35 011231054 
92Lilli93 20 24440046 
92Dose613 15 13550048 
92Duck10315 16 551046 
99Hamma52388 4 31034 
92Quil30310410 20 0050 
86Salmon2003201 31 342 
00JCL00000113 49 54 
           HoodMet
Correct55353015164203149 146
Total875446484634504254 178
% correct63.2264.8165.2231.2534.7811.7640.0073.8190.74 82.02
 Postsupplementation assignments
04BigBeef_Ha08Union_W09Lilli_H09Dose_W09Duck_W08Hamma_W08_09Quil_W09Salmon_W08_09JCL_WTotal 
  1. Abbreviations follow Table 1. Assignments for each collection are along rows with total fish analyzed in collection at side and bottom. Assignments back to collection of origin (‘correct’ assignments) are in bold type along diagonal. The ‘% correct’ was ‘correct’ assignments over total assigned. The category ‘HoodMet’ sums the number of assignments back to the Hood Canal ‘metapopulation’ group encompassing results for Dosewallips, Duckabush, Hamma Hamma, and Quilcene.

  2. a

    Reintroduced using Quilcene broodstock.

04BigBeef_Ha 22 1383440045 
08Union_W3 35 312300047 
09Lilli_H54 11 13140029 
09Dose_W603 4 41060033 
09Duck_W4125 9 760034 
08Hamma_W42463 10 120041 
08_09Quil_W6315511 14 0045 
09Salmon_W0000010 16 623 
08_09JCL_W02000004 27 33 
           HoodMet
Correct2235114910141627 117
Total454729333441452333 153
% correct48.8974.4737.9312.1226.4724.3931.1169.5781.82 76.47

Isolation by distance

We conducted IBD analyses for all categories of samples (Fig. 5A–D), both including and excluding Dosewallips and Duckabush samples. In all analyses, physical distance explained a significant amount (all r2 > 70%) of the genetic variance among samples (all P < 0.001). In comparisons among sample categories, because the tests for homogeneity among regressions in the ancova indicated significant differences between regressions, we were unable to compare regressions statistically. However, the IBD slope was steeper for comparisons among original samples (Fig. 5A,B) than the slope for comparisons among wild samples, suggesting that genetic variance among populations decreased with supplementation. The IBD slopes were similar for comparisons among hatchery-born and wild-born samples (Fig. 5C), but genetic distances were greater among hatchery-born fish. When analyses were limited to comparisons involving Dosewallips and Duckabush samples (Fig. 5D), the slope for the original samples was slightly steeper than the slope for the wild samples.

Figure 5.

Plots of isolation by distance patterns (transformed pairwise FST value (FST/(1 − FST) versus geographical distance) for different categories of samples: plot (A) all original samples and all wild samples, plot (B) original and wild samples (minus Dosewallips and Duckabush), plot (C) hatchery and wild samples (minus Dosewallips and Duckabush), and plot (D) only original and wild samples in comparison with Dosewallips and Duckabush. Plot (A) includes data from plots (B and D).

Discussion

This study joins a growing body of research on the genetic effects of supplementation hatchery programs on wild fish populations. We examined a time series of a summer chum salmon metapopulation (a designated ESU) from before, during, and after supplementation, providing a temporal perspective of supplementation and responses by threatened subpopulations. We found that after several years of hatchery supplementation, there was little change in genetic diversity and harmonic means of effective population sizes (HmNe) in wild-born salmon returning to spawning areas throughout the ESU. However, genetic distances within the metapopulation decreased and assignments back to collection of origin decreased following supplementation, suggesting higher gene flow within the metapopulation and lower genetic drift in subpopulations. Hatchery-born spawners collected in the same locations as wild-born spawners usually had lower genetic diversity and smaller HmNe values, suggesting that supplementation hatchery programs sampled a subset of genetic diversity in the target subpopulations and that factorial matings of hatchery broodstocks may have decreased HmNe. Higher linkage disequilibrium in hatchery-born fish indicated potential for Ryman–Laikre effect, and combining hatchery- and wild-born fish into single collections lowered ratios of effective population sizes to census sizes. However, with the exception of the Lilliwaup Creek subpopulation, suspected Ryman–Laikre effects did not increase genetic drift in wild-born spawners. Thus, possible negative impacts of supplementation appeared minimal and likely diminished as wild-born spawner abundance increased, suggesting that population recovery in this summer chum salmon ESU was unimpaired by supplementation.

Supplementation and conservation

In the Pacific Northwest, supplementation hatcheries using local broodstocks have received increased interest as a tool in fish conservation to preserve native genetic diversity and maintain a foundation of response to environmental variation. Traditional fish hatcheries were a mainstay of fish management and harvest augmentation, but they introduced problems when nonlocal, hatchery-adapted fish interacted with wild fish (Araki et al. 2008; Christie et al. 2012). Supplementation hatcheries were designed to reduce problems associated with traditional hatcheries in that domestication selection was minimized by using in-river broodstocks originally composed of wild-origin fish (subsequent broodstocks would be mixtures of hatchery- and wild-origin fish) and by reforming hatchery practices. Hatchery fish from the same gene pool as wild fish were expected to be similarly adapted to the local environment. Here we discuss studies documenting impacts from supplementation programs and hypothesizing mechanisms leading to differences between hatchery and wild fish in relation to chum salmon.

One selective force in supplementation hatchery programs is juvenile residence in a hatchery environment (McClure et al. 2008). Hatchery effects may be relatively lower for chum salmon because they out-migrate shortly after emergence, spending minimal time under hatchery rearing conditions. In contrast, most supplementation studies have been conducted in species with extended juvenile freshwater residence times [steelhead, Atlantic salmon (Salmo salar), coho and Chinook salmon (reviewed by Araki et al. 2008 and Chilcote et al. 2011)]. In these species, juvenile residence time in freshwater is a critical component of their life history, and unless they are released as unfed fry, hatchery-origin juveniles often spend a minimum of 1 year in a hatchery environment before out-migrating. Kostow (2004) observed differences in steelhead juvenile phenotypes and lower survival in hatchery-origin juveniles from the same parent pool. In a common garden experiment, Chittenden et al. (2011) found that for coho salmon juvenile rearing environment was a key influence on smolt size and multiple behaviors, regardless of ancestry. Thériault et al. (2011) implicated sexual selection at the juvenile stage, especially on males, as a factor decreasing relative reproductive success (RRS) in hatchery-origin coho salmon spawning in the wild.

Another selective force in supplementation hatchery programs is human-controlled spawning (McClure et al. 2008), and chum salmon would be impacted similar to other species. Artificial spawning bypasses mate selection, and Consuegra and Garcia de Leaniz (2008) found a positive association between disassortative mating (natural spawners choose mates with dissimilar MHC profiles) and parasite resistance: In Atlantic salmon offspring parasite load decreased their RRS. Further, the summer chum salmon supplementation programs used factorial matings, which likely decreased effective population size relative to single-pair matings (Abadia-Cardoso et al. 2013). Multiple factors likely contribute to the phenomenon that hatchery-origin fish and wild-origin fish with hatchery ancestry are less productive under natural spawning conditions (Araki et al. 2008; Chilcote et al. 2011).

In other studies, supplementation impacts were undetected or equivocal. In Berejikian et al.'s (2009) study, hatchery- and wild-born summer chum salmon spawners (Quilcene fish that were included in this study) mated randomly and there was no significant difference in offspring produced per spawner type. Sharpe et al. (2010) found no significant difference in reproductive success of wild-origin and first-generation hatchery-origin steelhead derived from the same gene pool. Further, genetic diversity in steelhead populations supplemented with native broodstock remained unchanged after 20 years (Heggenes et al. 2006) and 58 years (Gow et al. 2011). There was also no loss in genetic diversity (Eldridge and Killebrew 2008) or fitness (Hess et al. 2012) in Chinook salmon supplemented with native broodstock. In this study, we lack parentage data to address RRS (see Christie et al. 2012) and address only changes in genetic parameters. Yet, coupled with the study by Berejikian et al. (2009), the increases in wild-born spawner abundance and increases in recruits per spawner suggested that hatchery-born HC summer chum salmon spawned successfully, which may have contributed to increased census sizes and lower extinction risks and contributed to recovery in the metapopulation (Wang and Ryman 2001). The lower HmNe/N ratio in wild-born spawners could also be an indication that some of their parents were hatchery-born fish that, although abundant, had lower genetic diversity than wild-born fish because of unequal hatchery family sizes.

Metapopulation structure and supplementation

Contemporary genetic structure in HC summer chum salmon follows an IBD pattern, similar to original genetic patterns (Phelps et al. 1994; Small et al. 2009), in which the amount of genetic exchange depended on distance between spawner groups (Schtickzelle and Quinn 2007). Interestingly, the original collections from Big Beef Creek were most similar genetically to the collection from Union River, which is roughly 80 km distant, rather than to the closest tributary, Dosewallips River, roughly 8 km west across HC (see Fig. 1). This supports an ecoregional association between summer chum salmon inhabiting lowland streams on the Kitsap Peninsula on the east side of HC as suggested by Sands et al. (2009).

The relationship between genetic distance and geographical distance changed following supplementation in this summer chum salmon metapopulation – gene flow increased and genetic distances decreased – but the basic IBD pattern remained consistent over time. In another supplementation program for coho salmon, genetic distances increased following supplementation, which was attributed to genetic drift, bottlenecks, and varying success of between-river transfers (Eldridge et al. 2009). In HC summer chum salmon, each supplementation program used river-specific broodstocks rather than a common broodstock for all rivers. Thus, between-river transfers would occur via strays (nonhoming), which are documented in WDFW and PNPTC (2007). As a byproduct of increasing abundance, supplementation may have increased straying and decreased differentiation or perhaps restored straying to levels that existed prior to supplementation when subpopulations were more abundant. The pairwise FST values and assignment tests support that subpopulations are well connected through gene flow, which may have contributed to the increases in HmNe values. Alternatively, prior to supplementation census sizes were low and genetic drift may have increased differentiation; following supplementation and the fishing moratorium, increased census sizes may have countered genetic drift and decreased differentiation.

In summary, our results show that, although changes were not significant, in the HC summer chum salmon ESU contemporary genetic diversity and effective population sizes increased from original levels, suggesting that as natural production increased, possible negative impacts from supplementation (such as depressed Ne from Ryman–Laikre effects) diminished. We suspect that negative genetic impacts may be less or shorter-lived for chum salmon because their juveniles spend little time in the hatchery environment. We will continue to monitor this metapopulation to assess the long-term impacts of supplementation and success of recovery efforts.

Acknowledgements

Norm Switzler, Todd Kassler, Kyle Adicks, William Evans, Cheri Scalf and other WDFW staff, U.S. Fish and Wildlife Service and tribal staffs, and staff or volunteers from HC Salmon Enhancement Group, Long Live the Kings, North Olympic Salmon Coalition, and Wild Olympic Salmon were invaluable for sample collection and other contributions to this project. Jeff Grimm and WDFW otolith laboratory staff analyzed otolith samples. We thank the WDFW genetics laboratory technicians, past and present, who helped process samples (Alice Frye, Cheryl Dean, Sonia Peterson, Jennifer Von Bargen and Sarah Bell). The manuscript was enhanced by discussions and editorial comments from Anne Marshall, Todd Seamons, Scott Blankenship, Todd Kassler and Ken Warheit (WDFW), Adrian Spidle and Ken Currens (NWIFC), Robin Waples and Barry Berejikian (NOAA), Louis Bernatchez (editor), Craig Primmer (associate editor), reviewer Kevin Glover (IMR), and three anonymous reviewers. Special thanks to Jim Ames (WDFW), Gary Graves (NWIFC), Chris Weller (PNPTC), and Tim Tynan (NMFS), the key architects of the Summer Chum Salmon Conservation Initiative. Funding was provided by NMFS and Washington State General Funds.

Data archiving statement

Raw data for this manuscript are available in the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.k2q95.

Ancillary