Comparison of coded‐wire tagging with parentage‐based tagging and genetic stock identification in a large‐scale coho salmon fisheries application in British Columbia, Canada

Abstract Wild Pacific salmon, including Coho salmon Onchorynchus kisutch, have been supplemented with hatchery propagation for over 50 years in support of increased ocean harvest and conservation of threatened populations. In Canada, the Wild Salmon Policy for Pacific salmon was established with the goal of maintaining and restoring healthy and diverse Pacific salmon populations, making conservation of wild salmon and their habitats the highest priority for resource management decision‐making. A new approach to the assessment and management of wild coho salmon, and the associated hatchery production and fishery management is needed. Implementation of parentage‐based tagging (PBT) may overcome problems associated with coded‐wire tag‐based (CWT) assessment and management of coho salmon fisheries, providing at a minimum information equivalent to that derived from the CWT program. PBT and genetic stock identification (GSI) were used to identify coho salmon sampled in fisheries (8,006 individuals) and escapements (1,692 individuals) in British Columbia to specific conservation units (CU), populations, and broodyears. Individuals were genotyped at 304 single nucleotide polymorphisms (SNPs) via direct sequencing of amplicons. Very high accuracy of assignment to population (100%) via PBT for 543 jack (age 2) assigned to correct age and collection location and 265 coded‐wire tag (CWT, age 3) coho salmon assigned to correct age and release location was observed, with a 40,774—individual, 267—population baseline available for assignment. Coho salmon from un‐CWTed enhanced populations contributed 65% of the catch in southern recreational fisheries in 2017. Application of a PBT‐GSI system of identification to individuals in 2017 fisheries and escapements provided high‐resolution estimates of stock composition, catch, and exploitation rate by CU or population, providing an alternate and more effective method in the assessment and management of Canadian‐origin coho salmon relative to CWTs, and an opportunity for a genetic‐based system to replace the current CWT system for coho salmon assessment.


| INTRODUC TI ON
Wild Pacific salmon, including Coho salmon Onchorynchus kisutch, have been supplemented with hatchery propagation for over half a century in support of increased ocean harvest and conservation of threatened populations. The potential negative effects of hatchery production and the associated exploitation pressure that may be applied to natural populations have long been recognized and have gained prominence in recent years (Araki, Berejikian, Ford, & Blouin, 2008;Hilborn, 1992;Jones, Cornwell, Bottom, Stein, & Anlauf-Dunn, 2018;McClure et al., 2008), leading to calls for increased responsibility in management of hatchery production and of the mixed-stock fisheries it supports (Flagg, 2015;HSRG, 2014). In Canada, the Wild Salmon Policy (WSP) for Pacific salmon was established with the goal of maintaining and restoring healthy and diverse Pacific salmon populations, making conservation of wild salmon and their habitats the highest priority for resource management decision-making (Fisheries and Oceans Canada 2005). Fisheries and hatchery supplementation (termed enhancement in Canada) are to be managed in such a way as to ensure that wild populations are safeguarded and harvest benefits are sustainable. Wild salmon populations are identified and maintained in conservation units (CUs) that reflect their geographic, ecological, and genetic diversity.
Coho salmon are caught in commercial, recreational, and First Nations fisheries in British Columbia, and determination of the impact of these fisheries is of fundamental importance to status assessment for wild populations of conservation concern and management of large-scale hatchery production. Current and historical assessment of fisheries impacts has been conducted with the application of coded-wire tags (CWTs; Jefferts, Bergmann, & Fiscus, 1963). CWTs are applied to juvenile fish prior to their hatchery release and recovered from adult fish heads collected from fisheries, hatchery broodstocks, and in-river escapement sampling. Once recovered, the tags are decoded to determine the hatchery origin and age of the individual fish. Originally, only coho salmon marked with a CWT also received an adipose fin clip prior to hatchery release, with the externally visible clip mark allowing CWT-marked fish to be identified visually and sampled from fisheries or river collections.
Since the late 1990s, all coho salmon released from many hatcheries in southern British Columbia (BC), Washington, and Oregon have received an adipose fin clip (termed mass marking) in order to facilitate mark-selective fisheries intended to harvest hatchery salmon only, with most clipped individuals carrying no CWT. This approach has resulted in reduced exploitation of naturally spawned coho salmon, especially in sport fisheries, but the presence of many adipose-clipped salmon without a CWT has impaired the efficiency of CWT recovery. In spite of implementation of an electronic tag detection system to pre-screen a portion of the commercial catch to identify salmon with a CWT, the processing of many heads without a CWT from voluntary recreational recoveries and the increasing costs of CWT application and recovery have caused degradation of the information obtained from the current Canadian coho salmon assessment program.  (1986)(1987)(1988)(1989)(1990)(1991)(1992). The current expected catch of a specific stock is estimated as the product of (the expected abundance of the stock in a fishery) * (the base period exploitation rate) * (a correction factor) that relates the expected catch or effort in a particular year relative to that observed in the base period. Stock distribution and migration is assumed constant over time and is represented by the average distribution of CWT recoveries during the base period. However, differences between the abundance, distribution, and migration pattern of stocks during the base period and the year being evaluated will decrease the accuracy of the estimated stock-specific exploitation rates for the fishery under evaluation. Significant deficiencies in FRAM model predictions for Chinook salmon fishery assessment due to the use of incomplete and outdated baseline data have recently been demonstrated (Moran, Dazey, LaVoy, & Young, 2018).
For coho salmon, utility of the CWT system has been eroded by the extensive release of adipose fin-clipped individuals without CWTs and by the sharply declining number of CWTs recovered from recent fishery sampling. Through direct query of the Regional coded-wire tags, Coho salmon, fishery management, genetic stock identification, genotyping by sequencing, parentage-based tagging was 3.3%, and those of Canadian origin was 2.5%, of the respective annual numbers for the base period, largely due to a substantial reduction in catch but also coincident with reduction in the number of CWTs applied. Since estimation of fishery impacts with the FRAM is dependent on CWT recovery information, the greatly decreased number of CWT recoveries in recent years will have increased the variance of the estimated stock-specific catch and exploitation rates in fisheries (Hinrichsen et al., 2016;Reisenbichler & Hartmann, 1980).
A new, cost-effective approach to the assessment and management of wild coho salmon, and the associated hatchery production and fishery management is needed. Anderson and Garza (2006) noted that parentage-based tagging (PBT) provides equivalent information (hatchery of release, age of individual) for hatchery fish as do CWTs; implementation of PBT thus may overcome problems associated with CWT-based assessment and management of coho salmon fisheries in BC. Additionally, PBT provides a means of improved hatchery broodstock management, as well as assessment of hatchery-wild interactions in salmonids. Unlike CWT-based management, PBT-informed hatchery and fishery management would benefit from the complete adipose-clipping of hatchery-produced salmon. A significant advantage of the combination of mass marking and PBT implementation is the capability to identify visually, sample, and if desired, remove hatchery fish of local and stray origin in threatened wild populations. Moreover, PBT entails genotyping the entire hatchery broodstock and enables the identification of all hatchery progeny by parentage assignment (Anderson, 2012;Wang, 2016), thus enabling a "mark rate" of virtually 100% of hatchery fish. Steele et al. (2013) demonstrated the equivalency of CWT and PBT in an initial evaluation of population and age assignment in steelhead trout (Oncorhynchus mykiss) of the Snake River basin in the Columbia River drainage. Hess et al. (2016) expanded the approach by using both PBT and genetic stock identification (GSI) to investigate run timing of steelhead trout in the upper Columbia River drainage.
These applications confirmed the capability of a combined PBT-GSI technology to provide equivalent or better identification of fish as the CWT method, but were limited in geographic scale. In this study, we examine whether the existing PBT and GSI system of identification for Canadian-origin coho salmon (Beacham et al., 2017) can provide the information required for improved assessment and management of coho salmon mixed-stock fisheries in British Columbia, covering a much broader geographic range of coho populations from Alaska to Oregon.
Although proper management of hatchery production and associated fishery exploitation requires properly defined objectives and monitoring (Flagg, 2015), it is not currently possible to evaluate each Canadian enhancement project separately (Tompkins, Hamilton, Bateman, & Irvine, 2016). For coho salmon, CWTs are not applied to releases from some of the largest hatcheries in southern BC (Chilliwack River, Capilano River, Chehalis River, Conuma River, Nitinat River, and Tenderfoot Creek) due to funding limitations, and thus, their specific contributions to highly mixed-stock ocean fisheries are unknown. In fact, CWTs are applied only to coho salmon juveniles released from six Fisheries and Oceans Canada large production hatcheries: three on the east coast of Vancouver Island (Quinsam River, Puntledge River, and Big Qualicum River), one on the west coast of Vancouver Island (Robertson Creek), one in the lower Fraser River drainage (Inch Creek), and one in the Thompson River (a major tributary of the Fraser River) drainage (Spius Creek, where the Salmon River and Eagle River juveniles may be reared). CWTs are also applied at smaller facilities (Seymour River near Vancouver, Toboggan Creek and Zymacord River in the Skeena River drainage in northern BC) and some naturally spawned index populations in northern BC.
The 43 coho salmon CUs originally identified by Holtby and Ciruna (2007) have been modified to a current number of 44 CUs. Price, Rosenberger, Taylor, and Stanford (2014), Price, English, Rosenberger, MacDuffee, and Reynolds (2017) suggested that any suitable assessment technique must provide individual resolution for all CUs to meet the conservation requirements of Canada's WSP. There can be possible inconsistencies between CUs and existing fishery management units (MUs) (Irvine & Fraser, 2008), with differences between CUs and MUs challenging managers who are responsible for both assessing biological status of CUs and fishery objectives for MUs (Holt & Irvine, 2013). Given the limited distribution of coho salmon populations marked with CWTs in BC, it is clear that CWTs cannot provide the CU resolution recommended by Price et al. (2014), Price et al. (2017) for wild population assessment, nor the MU resolution noted by Holt and Irvine (2013) for hatchery and fishery assessment and management.
The current study is an evaluation of the application of the PBT-GSI methodology outlined by Beacham et al. (2017) to coho salmon fisheries in BC to determine whether the genetic technologies can be used to provide more information on fishery contributions by hatchery and CU than is available from CWTs. Commercial and recreational coho salmon fisheries, and river escapements for selected populations, were sampled for both CWTs and genotypes. We evaluated the population-level resolution obtained from CWTs and the genetic methodology by CU for all 2017 and some 2016 fisheries in which coho salmon were caught, catch estimation by CU for the fisheries sampled, and stock-specific exploitation rate for selected populations of coho salmon in BC. Genotyping by sequencing methodology was used to genotype coho salmon at 304 single nucleotide polymorphisms (SNPs) in 304 amplicons. Complete broodstock genotyping for PBT analysis was conducted in 2014 for 20 hatcheryenhanced populations that included genotyping 6,061 individuals (96.4% genotyping success rate), and a stock identification baseline comprising some 267 populations ranging from southeast Alaska to Oregon was employed for GSI. A comparison of the population-specific contributions to mixed-stock fisheries, catch, and exploitation rates estimated with CWTs and PBT-GSI technologies was made. We conclude that a genetic approach can emulate and improve upon the results available from the current CWT program for assessment and management of coho salmon enhancement and fisheries in BC, and provide critical information to improve wild coho salmon assessment and conservation.  Figure 1).

| Fishery sample collection
In 2017, for the northern (Area F) freezer troll fishery, selected freezer boats (28% of fleet) were required to keep heads of all coho salmon caught, with the mark type (adipose fin clipped or not) unknown for an individual head. Upon landing, the heads were counted and checked electronically for CWTs and randomly sampled to a maximum of 50 heads per delivery. If a CWT was detected, the head was sent to a central CWT head recovery laboratory in Vancouver, BC, where the DNA sample was subsequently taken. If a CWT was successfully recovered and decoded from an individual head from commercial, recreational, or First Nations fisheries, and a genotype was successfully obtained for the individual sample, then the genotypes of all of these individuals were pooled into a single mixedstock sample of known origin and known age in order to evaluate accuracy of stock compositions by CU and population. Field DNA samples were taken only from individuals with no CWTs detected.
In the northern ice boat troll fishery, the clip status of the fish in the catch was known. Samples of clipped coho salmon were obtained from this fishery as an ancillary aspect of standard Fisheries and Oceans Canada contract catch sampling for CWTs. Only clipped individuals were examined through this program, with similar sampling protocols as outlined in the freezer boat sampling. In particular, only clipped fish not containing a CWT were sampled in the field through this program, with heads containing a CWT sent to the head recovery laboratory in Vancouver. Clipped fish constituted 1.6% of the catch in the fishery, largely due to the wild origin of the catch, so it was important to obtain samples from unclipped individuals as well. genotypes were assigned to specific release groups, provide information on the relative rates of return of different release groups.
As relative rates of return for the release groups were considered ancillary to the main focus of the study, no results from these returns were included in the current study.

| Exploitation rate
We estimated exploitation rate of adult coho salmon in BC fisheries via both CWTs and genetics. Exploitation rate for a population was defined as adult catch/(adult catch-escapement). For CWTs, the observed number of CWTs was corrected by "no-pin" tag loss rates and was expanded by the population's tag-specific marking rate summed over tag codes and expanded again by the sampling rate for the fishery in order to estimate catch of hatchery-origin individuals. The observed number of CWTs in escapement sampling was expanded in a similar manner in order to estimate the hatchery contribution to the escapement. For genetics, the seasonal kept catch of adipose fin-clipped individuals in a fishery was multiplied by a seasonal stock composition estimate in order to estimate population-specific catch.
The abundance of hatchery-origin escapement was calculated as the estimated escapement multiplied by the proportion of adipose finclipped adults observed in the escapement. If part of the juvenile production was not adipose fin clipped upon release from the hatchery, then the hatchery contributions to both catch and escapement were underestimated by this method.

| Genetic stock identification baseline
The initial baseline was outlined by Beacham et al. (2017)

| Library preparation and genotyping
The detailed procedure for library preparation and genotyping was outlined by Beacham et al. (2017)

| Identification of individuals
As noted previously, PBT and GSI were used concurrently to estimate fishery stock composition. Initially, PBT was used, and the analysis was conducted where the genotypes of individuals to be identified were matched to the genotypes of prospective parents (COLONY, Jones & Wang, 2010;Wang, 2016). If all individuals in a hatchery broodstock are sampled and subsequently genotyped, then all offspring from the broodstock are genetically marked. The genotypes of individuals of unknown origin are statistically compared with the genotypes of potential parents, and if a match is made, the offspring are assigned to the parents, and thus, the origin and age of the individuals are determined. Parentage assignment software was utilized to assign offspring to parents, as COLONY can produce assignments when the genotype of one of the parents is missing, either due to a missing parental sample, or failure to produce a parental genotype from an existing sample. Given that PBT assignments for 20 potential populations were evaluated for each fishery sample, COLONY was run with all broodstock sampled during 2014 input as a single unit for analysis of fishery samples, with no differentiation among populations. Although the COLONY assumption of a single population in the parent pool was violated, analysis of known-origin samples indicated that very high levels of accuracy were achieved in assignments when pooling of potential parents in contributing populations was conducted. Two-parent assignments were accepted only when both assigned parents originated from the same population, otherwise the individual was passed to genetic stock identification (GSI) for potential assignment. Two-parent and single-parent assignments were accepted only when the probability of correct assignment was ≥0.85 for the parent pair, otherwise the individual was passed to potential assignment by GSI. Additionally, for single-parent assignments to be accepted, the PBT assignment to population had to be part of the CU assigned to the individual via GSI. Individuals for which no prospective parents were identified in the broodstock were passed to GSI for potential assignment.
Polygamous mating was assumed for the COLONY analysis. Simple pairwise comparisons between offspring and potential parents were conducted. The baseline for individuals sampled in the 2016 escapements (jacks) and 2017 fisheries included all broodstocks sampled in 2014, as these individuals are predominately three years of age (Sandercock, 1991). Jacks in the 2017 escapement were identified via body size and subsequent assignment to parents in the 2015 hatchery broodstocks. Individuals with more than 120 missing genotypes were eliminated from further analyses. An estimated genotyping error rate of 1% was used for COLONY assignments. Previously, Beacham et al. (2017)

| Estimation of stock composition for knownorigin samples
Genotypes were available from 573 jacks sampled in 2016 across three CUs and eight populations in southern BC. Estimated stock composition by CU and reporting group for this combined sample was accurate, with an error utilizing only GSI of ≤0.4% by CU for three CUs present in the sample (Table 2). With both PBT and GSI utilized, the average error declined to ≤0.2% by CU for three CUs TA B L E 2 Accuracy of regional (United States) and conservation unit ( Table 2). The average error for 45 CUs or reporting groups absent in the sample was 0.0% by both GSI and PBT-GSI. By population, 543 (94.8%) of the jacks were assigned via PBT with 100% accuracy with respect to population of origin and age (Table 3). Estimated stock composition for 20 populations where it was possible to use both PBT and GSI in estimation of stock composition was accurate. For the eight populations present in the sample of jacks, the average error utilizing GSI only was 0.5% per population, declining to 0.2% utilizing both PBT and GSI, under the assumption that the collection location was an accurate reflection of jack origin.
There was an average 0.0% error for the 12 populations with no representation in the sample with utilization of either GSI or PBT-GSI (  (Table 4). There was an average 0.0% error for the 13 populations with no representation in the sample with utilization of either GSI or PBT-GSI (Table 4). Similar to the sample of jacks, PBT when combined with GSI produced more accurate estimates of population-specific stock composition than available with only GSI.

| Evaluation of 2016 fishery sampling
As broodstock sampling at selected hatcheries in southern BC did   Table   S2).
The final fishery sampled was a demonstration commercial troll fishery in central coastal BC (A6 and A7). Contributions from local CUs were higher in samples from more inshore areas (A7-7, August 2, 12% southern CUs and regions), with contributions from southern CUs higher in the more seaward areas (A6-9, September 8, 62% southern CUs and regions). Like the west coast of Vancouver Island, migrating stocks were more likely to be found in more seaward locations.

| Comparison of CWT and PBT individual identification
As genotypes were available from 82.9% of the samples provided

| Application of PBT to Canadian fishery samples
PBT was applied to identification of 1,230 individuals in a number of fisheries in BC from which CWTs could potentially be recovered, with the intent of combining PBT and GSI to provide high-resolution estimates of stock composition in the samples from these fisheries (Table 6). There were also 269 additional individuals identified via PBT, primarily from direct sampling of the creel catch in recreational fisheries in southern BC, where both adipose fin-clipped and unclipped individuals were included in the samples (Table 7)

| Estimation of catch of hatchery-origin populations
Assessment of the impact of fisheries on specific CUs or populations within CUs via genetics requires that accurate, high-resolution estimates of stock composition of the catch are available. We obtained these estimates of stock composition with respect to CU (Supporting Information Tables S2-S5) and populations within CUs (Supporting Information Table S6) to estimate the population-specific catch of adipose fin-clipped individuals for the fisheries outlined in Table 8. For each fishery, the clipped catch that was kept for individual populations was estimated via seasonal stock composition estimates of the clipped catch (Supporting Information Table   S6). Hatchery-origin contributions to the kept catch for the genotyped populations were estimated to be the largest in the Strait of Georgia recreational fishery, with the Capilano River and Chilliwack River populations comprising 53% of the recreational catch (Table 8).
Hatchery-origin contributions to the kept catch were the largest for recreational fisheries in Johnstone Strait, the Strait of Georgia, Juan de Fuca Strait, and the west coast of Vancouver Island.

| Estimation of exploitation rate
Estimation of exploitation rates (ER) for populations was conducted with CWTs when available, and with genetics if escapement estimates were available. For the Quinsam River population, the ER estimated via CWTs (31%) and genetics (28%) were similar (Table 10). Larger contributions of hatchery-origin catch and escapement estimated via CWTs than genetics were likely attributable to a portion of the production that was marked with CWTs but were not adipose fin clipped upon release.
For the Puntledge River population, where only 12.8% of the juvenile production was clipped, the ER estimated via genetics was higher (32%) than via CWTs (10%) ( adipose fin clipped upon release, possibly accounting for part of the difference in estimated ERs. ERs for the Salmon River and Coldwater River populations estimated via CWTs were about 6% less than those estimated via genetics, which was largely attributable to higher estimated fishery catches of these populations via genetics than was obtained via CWTs (Table 10). Both of these populations originate from CUs of conservation concern, and the ERs for these populations were among the lowest observed ERs of the seven index populations examined.
Analogues to ERs were also determined for eight populations where no CWTs were applied. They ranged from 20% for the Conuma River population enhanced in a large hatchery on the west coast of Vancouver Island to 96% for the Goldstream River population, a population at the southern end of Vancouver Island where production is supplemented by a small volunteer-staffed hatchery (

| D ISCUSS I ON
Canadian commercial and recreational fisheries for coho salmon have been severely restricted since the late 1990s, but comprehensive evaluation of the benefits derived from reduced exploitation and mass marking of hatchery production has not been possible. Large

| Accuracy of estimation of stock composition
One key difference between the CWT method and PBT-GSI method as applied to salmon assessment relates to the inability of the CWT approach to provide estimates of stock composition of the catch.
CWT recoveries are used to estimate total contributions from tagged populations through "expansions" to account for the CWT marking In 2014 broodyear, 12.8% of Puntledge fry release was clipped and tagged, and the remainder was released unclipped. In 2017 escapement sampling, a CWT loss rate of 31% was observed due to sampling for otoliths and CWTs from the same individual. c

Total escapement to the Stamp
River plus Robertson Creek was 21,175 adults. 9,511 adults is the hatchery return, which is all that was sampled. d Total escapement to the Nitinat River was 4,883 adults. Clip rate is from swim-ins so total adults is just swim-ins, same as Robertson Creek. e Minor components of the escapement were not enumerated; however, the exploitation rate can be considered reasonably accurate due to the very high percentage of total return that were enumerated in the hatchery swim-in count at Nitinat River, Capilano River, and Chilliwack River. f Significant components of the escapement were not enumerated; consequently, exploitation rate was overestimated by a large but unknown percentage.
precluding the estimation of stock composition for the entire fishery sample that includes fish from tagged and untagged populations. In GSI provided the foundation for stock composition analysis in the study, and it has been demonstrated to provide reliable estimates of coho salmon stock composition for CUs or regional groups, with an average error of ≤1.0% observed in known-origin samples.
Estimates for CUs or regions that displayed higher error rates could likely be improved by increasing the number of populations in the baseline used to represent the CU or region. Reliable estimation of single-population contributions to fishery samples was enhanced by the addition of PBT, with the average error rate observed with GSI declining by over 50% relative to that observed with GSI-PBT. In particular, for the seven populations that each comprised <5% of the known-origin samples, the average error with GSI was 1.36%, while that with GSI-PBT was 0.16%. Accurate stock composition estimates of rare populations via GSI have been traditionally difficult (Reynolds & Templin, 2004;Winans et al., 2001), but if PBT can be applied in conjunction with GSI for these rare populations, then accurate estimates of stock compositions can be obtained even for these rare populations when they occur in the mixture sample. For example, in the 2017 CWT sample, three populations each comprised ≤1.0% of the sample, but the maximum error for the three populations when both a GSI and PBT approach was followed was 0.1%. This level of resolution for rare populations in samples from mixed-stock salmon fisheries will likely be of importance in management of specific fisheries.
If 100% of a hatchery broodstock is successfully genotyped, there is an expectation that 100% of the offspring from the broodstock should be identifiable via PBT. In actual practice for the two known-origin samples, assignment rates of 91.6% (CWT sample) and 94.8% (jack sample) were achieved. Failure to assign some individuals may be a result of incomplete sampling of the hatchery broodstock, failure to genotype successfully all samples that were provided, or genotyping error rates in either some of the broodstock individuals or offspring that precluded assignment to the correct parents.
The origin of individuals not assigned was subsequently estimated via GSI, and the high levels of population-specific accuracy (≥99.7%) for those populations where it was possible to implement both PBT and GSI for population-specific stock composition indicated that observed assignment rates >91% via PBT were acceptable for high-resolution stock identification analysis. Genotypes were obtained successfully from 82.9% (2,533 genotypes from 3,054 samples) of the samples provided by the central laboratory employed for potential recovery of CWTs from the samples processed from commercial, recreational, and First Nations fisheries. Genotyping success improved during the course of sample delivery, with initial tissue samples large and genotyping success of them reduced, but subsequently smaller tissue samples were affixed to the Whatman sheets, presumably drying quicker and preserving DNA quality, thus resulting in higher genotyping success. Also, the status of initial tissue quality for the samples was uncertain, as some samples deliv-

| Exploitation rate
ER for a population is catch/(catch + escapement), and estimation of fishery exploitation rates is one of the key outcomes of assessments. In theory, some clipped hatchery production may not be associated with a CWT and thus would not be included in expansions of ob- Under these circumstances, the presence of hatchery-produced individuals that returned to the river rather than the enhancement facility was unaccounted for in the estimation of enhanced (hatchery-produced) abundance, and the exploitation rate would be overestimated. To use these locations as index systems, non-broodstock escapement surveys would need to be conducted in order to provide a comprehensive assessment of hatchery-origin escapement. Use of PBT-GSI technology provides a simple means to develop additional index populations where some hatchery production occurs by the simple adoption of broodstock genotyping, juvenile mass marking, fishery sampling, and estimation of the escapement and its clip rate.
The estimated ER of 96% for the Goldstream River population is of note. Goldstream River-origin individuals were identified via PBT in four fisheries (Tables 6 and 7) and five fisheries via PBT and GSI (Table 8). Estimated stock composition of the clipped portion of the northern troll fishery was 2.7% (SD = 0.6%), yet no individuals were identified via PBT. This fishery was estimated to have contributed about 20% of the estimated catch of the population (Table 8), and we assumed that a reliable estimate of catch was obtained. The estimated hatchery abundance of the escapement was the lowest of any of the populations surveyed (Table 10) In essence, for coho salmon in southern BC in which mass marking is applied and which display low stray rates among populations, estimation of fishery exploitation rates requires only genotyping of individuals in fishery samples and hatchery broodstocks. For wild populations, where there is a fence or weir in operation, small clips could be taken from potential spawners at the fence to provide the basis for genetic identification of their offspring in fishery samples (Ford, Pearsons, & Murdoch, 2015). Some portion of the smolts could be captured and adipose fin-clipped, and subsequent sampling

| Conservation unit management
The

| Hatchery management
A genetic method of assessment enables hatchery broodstock management and assessment of hatchery production with either harvest augmentation or conservation goals. The dramatic decline in coho salmon abundance that occurred in British Columbia during the 1990s spurred the implementation of mass marking of hatchery-produced fish to enable mark-selective recreational fisheries in which only hatchery-produced fish were harvested. High levels of hatchery production were suspected to be a contributing factor to the poor survival of wild coho salmon, leading to an increased awareness of the need to manage hatchery production and assess hatchery-wild interactions (Beamish et al., 2010). In coho salmon, mass marking enables hatchery managers to ensure the inclusion of naturally produced fish in the broodstock if desired, and removal of hatchery-produced fish at fences or weirs in the natural environment to control the relative influences of the natural and hatchery environment on hatchery-supplemented populations in which gene flow between the two spawning environments takes place (Mobrand et al, 2005).
Moreover, mass marking combined with parentage analysis enables assessment of the reproductive success of hatchery-produced fish that return to spawn in the natural environment (Abadia-Cardosa, Anderson, Pearse, & Garza, 2013;Ford et al., 2015).

| Utility of PBT-GSI for Coastwide Management
The necessity of maintaining a viable CWT system for salmon as- ity and cost-effectiveness of developing a coordinated coastwide tag recovery system using PBT, stipulating that a transition from the coastwide CWT system to a PBT system would make require that: 1. The PBT system generates at least the same information currently generated from the CWT system via run reconstruction (cohort) analyses of estimated recoveries from individual CWT release groups.
2. The PBT system would have long-term annual operating costs no greater than or, ideally, substantially less than those of the existing CWT system.

3.
The cost of a coastwide PBT system was substantially less than that of the existing CWT system or that PBT delivers additional or novel information, not provided by the existing CWT system, to inform management of fisheries for coho and Chinook salmon (PSC, 2015). We undertook no comprehensive cost comparison between the PBT-GSI and CWT technologies in the current study, but a preliminary cost analysis can be made as follows. Approximately 820,000 CWTs were applied to offspring from the 2014 Canadian hatchery broodstocks and wild escapement, at an estimated cost of $169,000 (820,000 * $0.17 + $30,000; Beacham et al., 2008 with that provided by the CWT program. Beacham et al. (2018) provided evidence that for Chinook salmon in BC, a PBT-GSI assessment method was substantially cheaper than the existing CWT program.
In this case, the cost of CWT application and recovery and reading from fishery and escapement samples was compared with the cost of genotyping broodstock, fishery, and escapement samples.
Further details were outlined by Beacham et al. (2018).  noted that there is a common concern among some fishery managers and staff that "investigation of new technological approaches to provide data for salmon fishery management diverts monies that can be used to maintain the existing CWT program" (Pacific Salmon Commission Joint CWT Implementation Team, 2015). Increasingly however, the deficiencies provided by the CWT-based FRAM model for both Chinook  and coho salmon are revealed by application of PBT-GSI technologies.
Additionally, many GSI projects are already routinely conducted because CWTs do not provide adequate information for fisheries management decisions and assessment (e.g., Beacham et al., 2008;Bellinger et al., 2015;Satterthwaite et al., 2014). The current dissatisfaction of Canadian managers of northern and central coastal coho salmon fisheries and calls for better assessment tools for the management of mixed-stock and in-river fisheries in northern BC (Price et al., 2017) also highlight the need for an improved management regime. Additional cost savings may accrue from implementation of a PBT-GSI management system, as GSI projects currently conducted on an ancillary basis to the CWT program are merged into routine fishery sampling, avoiding duplication of effort and expense.
The strongest benefits of a PBT-GSI management system come from the additional information that it can provide, not only for improved fishery management but also for wild population con- Much of the increased utility of the PBT-GSI methodology results from the fact that complete broodstock sampling ensures close to a 100% mark rate for juveniles. This rate compares very favorably with the current marking rate of approximately 10% for CWTs at selected hatcheries. The reduced cost per PBT-tagged fish recovered versus CWT-tagged fish stems from the high PBT mark rate, and the ability to identify fish from all hatchery facilities enables complete catch analysis through the representative sampling of both adiposeclipped and unclipped individuals.

| Future developments
The current study has demonstrated the PBT-GSI capability to identify BC-origin coho salmon to specific Canadian hatcheries and CUs, provoking consideration of replacement of the current CWT system for coho salmon assessment in BC with a PBT-GSI based approach. The 304-SNP panel used in the current study to genotype the 2014 coho salmon broodstocks and their jack returns at selected hatcheries (Beacham et al., 2017), and for genotyping samples from coho salmon fisheries in 2016 and 2017 has since been upgraded.
A 492-SNP panel now exists, with additional loci originating from research conducted in a Genome Canada large-scale applied research project, and has been used to genotype the 2016 and 2017 coho salmon broodstocks at an expanded number of hatcheries. It is anticipated that this enhanced SNP panel and the increased number of facilities at which broodstock genotyping has occurred will provide improved stock composition results relative to those of the current study when applied to coho salmon fishery samples in 2019.
If Canada were to implement a GSI-PBT method of assessment for coho salmon in place of the CWT program, then complete assessment of exploitation rates for Canadian populations would require genetic analysis of samples from American fisheries if CWTs are retained as the assessment tool for American coho salmon populations and fisheries. Should a GSI-PBT method of analysis be deemed practical for American assessment purposes, it is conceivable that a coastwide GSI-PBT assessment method could be implemented for coho salmon fisheries.

| SUMMARY
This study has demonstrated the potential for implementation of a comprehensive PBT-GSI methodology for management and assessment of coho salmon in British Columbia that will remedy noted deficiencies of the current CWT-based management system. Most importantly, the genetic technology provides an immediate tool for identification of coho salmon to CU, a requirement for implementation of management of wild populations as mandated by the WSP for Pacific salmon, and a task that would be prohibitively expensive using CWTs. Moreover, the PBT-GSI technology benefits from the mass marking of hatchery-produced salmon, thereby facilitating improved hatchery broodstock management, monitoring of wild-enhanced fish interactions, and the evaluation of hatchery contributions to harvest. The ability to identify readily hatchery-produced salmon has been recognized as an imperative for managing the risks and assessing the benefits of hatchery production of salmonids at the domestic, bilateral, and international levels (Ruggerone & Irvine, 2018). In Canada, extensive coho salmon conservation and enhancement efforts conducted for two decades require comprehensive evaluation and possible modification that cannot be achieved under the current management system. The genetic methodology developed in this study provides an opportunity for conservation-based management of Canadian coho salmon in which the economic benefit of hatchery production can be reaped without the imposition of undue and unknown risk to wild populations.

DATA A RCH I V I N G
Multi-locus genotypes for all sampled jacks, as well as individu-

ACK N OWLED G EM ENTS
A very substantial effort was undertaken to obtain samples from