Parentage‐based tagging combined with genetic stock identification is a cost‐effective and viable replacement for coded‐wire tagging in large‐scale assessments of marine Chinook salmon fisheries in British Columbia, Canada

Abstract Wild Pacific salmon, including Chinook salmon Oncorhynchus tshawytscha, have been supplemented with hatchery propagation for over 50 years in support of increased ocean harvest, mitigation for hydroelectric development, 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. For policy implementation, a new approach to the assessment and management of Chinook salmon and the associated hatchery production and fisheries management are needed. Implementation of genetic stock identification (GSI) and parentage‐based tagging (PBT) for marine fisheries assessment may overcome problems associated with coded‐wire tag‐based (CWT) assessment and management of Chinook salmon fisheries, providing at a minimum information equivalent to that derived from the CWT program. GSI and PBT were used to identify Chinook salmon sampled in 2018 and 2019 marine fisheries (18,819 individuals genotyped) in British Columbia to specific conservation units (CU), populations, and broodyears. Individuals were genotyped at 391 single nucleotide polymorphisms via direct sequencing of amplicons. Very high accuracy of assignment to population and age (>99.5%) via PBT was observed for 1994 Chinook salmon of ages 2–4 years, with a 105,722–individual, 380–population baseline available for assignment. Application of a GSI‐PBT system of identification to individuals in 2019 fisheries provided high‐resolution estimates of stock composition, catch, and exploitation rate by CU or population, with fishery exploitation rates directly comparable to those provided by CWTs for 13 populations. GSI and PBT provide an alternate, cheaper, and more effective method in the assessment and management of Canadian‐origin Chinook salmon relative to CWTs, and an opportunity for a genetics‐based system to replace the current CWT system for salmon assessment.


| INTRODUC TI ON
Supplementation of Pacific salmon abundance via hatchery production has been conducted for many years in Canada with two principal objectives, primarily increasing ocean harvest for selected species or enhancing production from specific populations of conservation concern. Canada initiated a Salmonid Enhancement Program in the 1970s with the objective of doubling catch of Pacific salmon in British Columbia (BC; Hilborn & Winton, 1993). This objective was never achieved, but led to the simultaneous exploitation of hatcheryenhanced and wild populations in mixed-stock fisheries, potentially leading to over-exploitation of wild populations. Management of mixed-stock fisheries is a matter of continuing concern (Flagg, 2015;HSRG, 2014). The effects of straying of hatchery-produced individuals into wild-spawning populations are also of concern (Araki et al. 2008;Jones et al. 2018;McClure et al. 2008). Canada responded to concerns over declining wild population abundance by developing As some Chinook salmon populations are enhanced through hatchery production, easy access to the juveniles produced is obtained. To aid in assessment of mixed-stock fisheries, some portion of the juveniles is marked with coded-wire tags (CWTs; Jefferts et al. 1963) prior to release from the hatchery. The presence of these tags is implied in returning adults either by absence of the adipose fin which was clipped at the time of tagging or by means of an electronic tag detection (ETD) system applied to individual salmon sampled in fisheries, hatcheries, or on the spawning grounds. These "indicator" populations, coupled with CWT marking of selected wild populations in the United States but generally not in Canada, provide the basis for current fishery assessment and management regimes for Chinook salmon. The key assumption underlying an assessment method employing indicator populations is that the indicator population displays characteristics that are representative of the other untagged (naturally spawning) populations within the management unit or geographic region that it is intended to represent. Once recovered, the tags are decoded to determine the hatchery origin and age of the individual fish. Originally, only individuals 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.
However, since the 1990s, in order to facilitate fisheries that exploited Chinook salmon produced only in hatcheries, most Chinook salmon released from many hatcheries in Washington, Oregon, and the Columbia River drainage were mass marked by receiving an adipose fin clip, but not necessarily a corresponding CWT.
The Chinook Technical Committee of the Pacific Salmon Commission (PSC) uses a Chinook Model to generate key outputs of relevance to the PSC's annual fishery management cycle, where preseason prediction of abundance is a key driver of subsequent fisheries in southeast Alaska and BC (PSC-CTC, 2018). Fishery exploitation rates are derived from CWT recoveries and are used in model calibration. There is also a Fisheries Regulation Assessment Model (FRAM) that is used as the primary analytical and assessment tool for fisheries off the U.S. west coast, where CWT recoveries are used to estimate stock-specific abundance and exploitation rates (PFMC, 2008). The utility of the CWT system for fisheries assessment has been eroded by the extensive release of the previously mentioned adipose fin-clipped individuals without CWTs (mass marking). Without coastwide implementation of electronic sampling, inadequate application and recovery of double-index CWTs, as well as misalignment of fisheries regulations with assessment programs may occur (PSCSFEC, 2016). Additionally, 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 et al. 2018 between Canada and the United States through a Memorandum of Understanding. In 2004, given the impairment of CWT recovery through mass marking of some American hatchery production, the PSC convened an expert panel to examine limitations of the CWT program and to evaluate the capacity of alternative technologies to provide data to improve assessment of Chinook salmon. One finding of the panel was that a parentage-based tagging (PBT) approach as proposed by Anderson and Garza (2005) could provide the equivalent of CWT recovery data (hatchery of release and age of the sampled individual), and could be easily integrated with a genetic stock identification (GSI) system to provide stock of origin for all fish from PBT hatcheries (PSC, 2005). There was a recognition that a geneticsbased assessment method could provide equivalent information to that of a CWT-based method, but an empirical demonstration of the equivalency was mandatory.
A rockslide in the middle portion of the Fraser River drainage in southern BC was discovered in late 2018 which was recognized as having the potential of severely impeding upstream migration of salmon utilizing spawning habitat in the middle and upper portions of the drainage. Accordingly, substantial restrictions on Chinook salmon fisheries in BC were implemented in 2019 in order to minimize fisheries exploitation on populations that would utilize this habitat. There was prior minimal hatchery enhancement of populations in the region, and the restrictive fisheries management regime implemented heightened interest among some stakeholders in mark-selective (adipose fin clip) fisheries that would target hatchery production from other areas. In BC, mass marking of Chinook salmon hatchery production via an adipose fin clip has not been implemented, although a pilot project was initiated for mass marking of Sarita River and Conuma River hatchery production on the west coast of Vancouver Island (WCVI) in 2020. There is the potential for more substantial mass marking of BC hatchery Chinook salmon production in the future. Mass marking via an adipose fin clip can impair the recovery of CWTs, as now many adipose fin-clipped individuals do not carry a CWT, thus requiring operation of an ETD system to allow practical recovery of CWTs. In spite of operation of an ETD system to screen a portion of the commercial catch to identify salmon with a CWT, and increased tagging rates on the indicator populations, reduced marine survival rates due to a prolonged low productivity regime in the Pacific Northwest and associated low harvest rates has resulted in fewer tags obtained from the current CWT assessment program. If a program of routine mass marking of Chinook salmon hatchery production in BC is implemented, further expense and difficulty in CWT recovery for Chinook salmon in BC may occur.
A genetics-based assessment method can incorporate both GSI and PBT methods to produce high-resolution stock composition and age structure of catch in mixed-stock fisheries. As proposed by Anderson and Garza (2005), illustrated as potentially possible by Anderson and Garza (2006), and outlined by Steele et al. (2019), PBT uses molecular-based approaches to conduct large-scale parentage assignments and has resulted in the unprecedented ability to identify genetically millions of hatchery-origin salmonids. Assignments are made to parents of known origin, and with that information, it is possible to determine the origin and age of individuals sampled in fisheries. Application of a GSI-PBT system of identification of coho salmon (O. kisutch) in fisheries and escapements (number of salmon that "escape" fisheries and return to fresh water to spawn) in BC 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 (Beacham, Wallace, Jonsen, McIntosh, Candy, Willis, Lynch, Moore et al. 2019).
Recent molecular and analytical improvements have made it possible for a genetics-based assessment system to provide an alternative to the current CWT-based system. Direct DNA sequencing, coupled with automated scoring of the genotypes, results in costeffective genotyping and unprecedented ability to provide accurate estimates of stock composition or individual identification to very discrete geographic regions or CUs. As noted by , it is a new era in the application of genetic variation to resource management and forensic analysis. Integration of PBT and GSI into a single application can produce fishery stock composition estimates of very high resolution, as well as origin and age of individuals sampled when identified via PBT, the same information as provided by CWTs as first described by Anderson and Garza (2005) and subsequently demonstrated in studies by Hess et al. (2016), Beacham et al. (2017Beacham et al. ( , 2018 Chinook salmon is the most important Pacific salmon species in terms of CWT application, the most diverse in terms of age structure of returning adults, and was the species of most concern to a panel examining deficiencies in the CWT program (PSC, 2005). The challenge in evaluating a PBT application in Chinook salmon is equivalent to that posed with coho salmon; specifically, this requires that a GSI-PBT approach provide the equivalent of CWT recovery data in an empirical demonstration on a coastal scale. Age at maturity in Chinook salmon in BC (mainly ages 2-6 years) is more variable than that of coho salmon (2-4 years), and thus would potentially present the largest challenge for Pacific salmon in correct assignment of individuals via PBT, as correct assignment to wider age span is required. Beacham et al. (2018) provided an initial indication that implementation of a GSI-PBT evaluation method may be possible for Chinook salmon. However, empirical demonstration of correct age assignments across the full suite of age of return was not available, nor was empirical demonstration of the technology to mixed-stock fishery samples available, where both GSI and PBT are applied in estimation of stock composition, with a coastwide baseline of populations available for utilization in the analyses.
The current study is an evaluation of the application of the GSI-PBT methodology outlined by Beacham et al. (2018) to selected Chinook salmon fisheries in BC to determine whether GSI and PBT can be used to provide more information on fishery contributions by hatchery and CU than is available from CWTs. Several improvements have been made to the methods and results outlined by Beacham et al. (2018). First, the single nucleotide polymorphism (SNP) panel has been enhanced from 321 to 391 SNPs available for genotyping. Second, the baseline has been substantially enhanced to include populations from Russia, Alaska, the Yukon Territory, BC, the Pacific Northwest, and California, allowing greater resolution in estimation of stock composition in samples from mixed-stock fisheries. Third, we evaluated the population-level resolution obtained from CWTs and the GSI-PBT methodology by CU for some 2018 and 2019 fisheries in which Chinook salmon were caught, along with catch estimation by CU for the fisheries sampled. Complete broodstock genotyping for PBT analysis was conducted since 2013 for selected hatcheryenhanced populations, and a stock identification baseline comprising some 380 populations ranging from Russia to California was employed for GSI. After evaluation of the results of the fishery sampling program in 2019, we conclude that a genetic approach can emulate and improve upon the results available from the current CWT program for assessment and management of Canadian Chinook salmon enhancement and fisheries in BC, and provide F I G U R E 1 Map indicating geographic locations for fishery sampling (a) and 48 populations for which parentage-based tagging was applied in estimation of stock composition or origins of 2019 hatchery broodstocks (b) critical information to improve wild Chinook salmon assessment and conservation.

| Fishery sample collection
The initial sampling for both GSI and PBT application occurred in 2018 fisheries, as 2018 marked the first year in which PBT identifications could potentially be made across the suite of most likely age of return, and the GSI baseline was ready for initial application.
The intent of the fishery analysis in 2018 was to evaluate the extent to which PBT identifications could be made in commercial and recreational fisheries, as well as to evaluate the performance of the GSI baseline for mixed-stock analysis. A total of 6286 individuals were genotyped from fishery samples collected in 2018. Samples were pooled for analysis by fishery and gear as outlined in Figure 1a.
Fisheries that occurred within the defined geographic regions were separated by sector and gear (commercial troll, commercial net, recreational, First Nations, and test fisheries).
In the northern BC troll fishery, offloads were selected at random and every 5th or 10th fish was selected from the load to a maximum of 50 fish from a single load. Individuals with a CWT were not excluded from the sample. Tissue samples from sampled individuals were subsequently provided for genotyping. Samples from the northern BC recreational fishery were provided by operators of fishing lodges or through direct Fisheries and Oceans Canada (DFO) creel sampling. Samples were also obtained from a gillnet test fishery at Tyee operating at the mouth of the Skeena

River.
In southern BC, samples were available from a First Nations troll fishery (T'aaq-wiihak fishery) conducted off the WCVI. In the WCVI creel survey, if individual fishers agreed, heads of Chinook salmon were sampled and otoliths collected for subsequent possible hatchery identification, as well as a tissue sample for genetic analysis.
Samples from the 2018 WCVI recreational fishery were genotyped subject to their having been screened previously for individuals identified via CWT or by a specific hatchery otolith mark. Individuals with an adipose fin clip but no CWT were deemed to be of US origin.
If individuals were identified by these methods, then they were not subsequently provided for genotyping. In the Strait of Georgia (SoG) and the Juan de Fuca Strait (JDF) recreational fishery off Victoria, BC, samples from the recreational fishery were obtained from a DFO creel survey program supplemented by samples provided by the Avid Anglers. The Avid Anglers are a "citizen science" group of volunteers who, if given the opportunity, fish year round, and collect biological information and tissue samples. The Capilano River hatchery derby samples were obtained from a recreational fishery that occurred in July within approximately 15 km of the mouth of the Capilano River. Johnstone Strait recreational samples were obtained from the creel survey program. In each sample analyzed over all BC fisheries, the number of individuals identified via PBT relative to the number of genotypes in the sample was tabulated, and summarized over sample, fishery, and season.
Minimum size limits were in effect for recreational fisheries in the SoG, with individuals <62 cm fork length defined as sublegal and required to be released. The origins of the sublegal individuals were unknown via CWTs, as the CWT could not be recovered from an individual released alive, and sublegal individuals would not be observed in the creel sampling program. Fin clip samples were obtained from released sublegal individuals via the Avid Anglers sampling program, and GSI and PBT analyses were subsequently conducted on these individuals.

| Evaluating accuracy of PBT age determination
Samples of 1333 adipose fins of juveniles that were clipped as part of CWT marking were obtained in 2017 from 16 hatchery populations, as these individuals were the offspring of the hatchery broodstock sampled in 2016. These juveniles constituted samples of known origin and age and were subsequently used in evaluation of population and broodyear assignment, with the baseline available for potential parentage assignment including hatchery broodstocks genotyped prior to 2017. Genetic tagging rates in these 16 populations were calculated as outlined by Satterthwaite et al. (2015). Tagging rate in a year for two-parent assignments was estimated as = (the proportion of the broodstock successfully genotyped) 2 . Genetic tagging rates for single-parent assignments was estimated as = 1 − (proportion of broodstock not genotyped) 2 . The expected number of PBT identifications was estimated as sample size × genetic tagging rate for both two-parent and one-parent identifications.
In 2017, hatchery broodstocks were genotyped, and at four hatcheries it was possible to match individuals that were marked with CWTs with age and origin of individuals estimated via PBT. The objective of the analysis was to evaluate accuracy of assignments of the individuals marked with CWTs with respect to population and age. The baseline for potential assignment included hatchery broodstocks genotyped prior to 2017. Once the parents of the juveniles and those individuals with CWTs were identified, they were subsequently removed from the baseline and parentage assignment conducted again in order to evaluate the rate of false positive assignments.

| Baseline
The initial baseline was outlined by Beacham et al. (2018) and con-  Table S1, with the populations from BC arranged by CU, and with Russian and United States of America (US) populations arranged by geographic (reporting) region. The CU boundaries for southern BC are indicated in Figure S1, while those for northern BC are indicated in Figure S2. The SNPs genotyped in the expanded panel are outlined in Table S2, along with primer sequences for the amplicons and F ST and heterozygosity estimates for the SNPs. Some SNPs were found to have duplicate positions when aligned to the Chinook reference genome; these are most likely a result of genome assembly artifacts and not true duplications (K. Christensen, Univ. of Victoria, pers. comm.). For markers with "Y" in the "Multiple" column in Table S2, the position represented reflects one of the possible locations.

| Library preparation and genotyping
The detailed procedure for library preparation and genotyping was outlined by Beacham et al. (2018), and a summarized version were assembled to provide multi-locus genotypes that were the basic input for PBT analysis. Genotypes had to be available for at least 150 SNPs for an individual to be retained in the baseline. In a test where the DNA of the same 768 individuals was genotyped on two occasions, an average genotyping error rate of 1.14% (1839 discrepancies in 161,280 single-locus genotype comparisons) was observed over the 319 SNPs scored (Beacham et al. 2018). The species identification SNP OkiOts_120255-113 (Starks et al. 2016) and sex identification SNP Ots_SEXY3-1 were omitted from subsequent parentage and GSI analyses, leaving 389 SNPs for subsequent analysis.

| Heterozygosity and F ST analysis
Expected and observed heterozygosities by locus over all baseline populations were determined with adegenet (Jombart & Ahmed, 2011). Estimation of F ST by locus was conducted with ape (Paradis & Schliep, 2018). BWA mem 0.7.17-r1188 (Li, 2013) was used to align amplified sequences to the reference alignments subsequently filtered using filter_sam_file.py in snp-placer (commit 8bd5e72 -https://github.com/CNuge /snp-placer), and a modified version of SNP-placer written in R (https://github.com/erond eau/snp-placercommit: eabfc78) was used to place the markers onto the Chinook reference genome assembly GCF_002872995.1 (Christensen et al., 2018). Bedtools getfasta v2.26.0 (Quinlan & Hall, 2010) was used to extract flanking sequences before and after alignment, and manually reviewed to ensure correct position was identified. Multiple best-alignments (MapQ = 0 in sam file) were flagged in Table S2; the majority were determined most likely to be a result of genome assembly artifacts and not true duplications, and alternate mapping locations were removed from the genome using bedtools maskfasta v2.26.0 (Quinlan & Hall, 2010) for subsequent alignments and plotting. Positional information was used to plot F ST by marker along the genome using R (R core team, 2020).

| Identification of individuals
Since the inception of genotyping of hatchery broodstocks in 2013, the number of hatcheries participating in the program has increased, so that in the 2019 fisheries, it was potentially possible to assign parents from 48 hatchery ( Figure 1b). The same techniques as outlined by Beacham et al. (2018) for PBT and GSI analysis were used in the current study. Summarized briefly, initially PBT was used for individual assignment, 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). COLONY was run with all broodstock sampled during a single year input as a single unit for analysis of fishery or broodstock samples, with no differentiation among populations.
Five age classes (2-6 years) may occur in fishery and broodstock sampling, and thus five runs of COLONY may have been conducted for each year of fishery or broodstock sampling. The choice of utilizing COLONY over SNPPIT (Anderson, 2012) was primarily to generate both two-parent and single-parent assignments. Where incomplete brood sampling was obtained, accurate single-parent PBT-based assignments could still be made, especially for hatcheries with incomplete sampling or underperforming genotyping.
COLONY was run by broodyear for two reasons. First, it allowed partitioning the brood age classes across multiple computer systems, speeding up the analysis overall. Second, binning parents into year classes allowed two-parent assignments only to individuals that could have been conceivably been crossed, limiting occasional cross-year assignments that we presume resulted from close relatives and/or small effective population sizes. Two-parent assignments were accepted only when both assigned parents originated from the same population in the same year and the probability of correct assignment was ≥0.95 for the parent pair. Restrictions were placed on acceptance of single-parent assignments. First, if a two-parent assignment had already been made for an individual in a particular year, any subsequent single-parent assignments in alternate years were rejected. Second, if single-parent assignments in alternate years were observed for the same individual, the assignment with the higher probability was accepted, subject to the probability being at least 0.05 higher than the competing assignment, and having the assignment probability ≥0.95. Third, an additional constraint on the single-parent assignment before it was accepted was that both the PBT assignment and GSI assignment corresponded to populations in the same CU. Individuals for which no prospective parents were identified in the fishery samples or hatchery broodstocks available for analysis were passed to GSI for potential assignment. Polygamous mating was assumed

| Exploitation rate
Although it was potentially possible to assign parents from 48 hatchery populations for individuals sampled in the 2019 fishery, only a portion of these populations was marked with CWTs, and escape- Note: The baseline available for potential parentage assignment included hatchery broodstocks genotyped prior to 2017. N is number of juveniles genotyped for each population, as well as the number not subsequently assigned via PBT, and those assigned via 2 and 1 parents should have been observed had the entire fishery been sampled was 20 × (5000/1000) = 100 CWTs. In order to estimate the total contribution of Robertson Creek hatchery-origin Chinook salmon to the fishery, the estimated number of tags is further expanded by the proportion of juveniles that had been marked with CWTs in a year. For example, if 10% of the Robertson Creek juveniles had been tagged prior to release from the hatchery, then the estimated number of tags (100)

| Accuracy of estimation of age and population of origin via PBT
The initial test of accuracy of age and population of origin via PBT was derived from genotyping 1333 juveniles from the 2016 hatchery broodstock production from 16 populations. No assignments were obtained for 8.1% of the juveniles genotyped, with failure of assignments dominated by the Chuckwalla River population, where only 38.6% (32/83) of the juveniles were assigned (Table 1). However, only 29.5% (5/17) of potential parents were successfully genotyped from the 2016 broodstock (Table S3) (Table 2). Once the parents had been identified, they were removed from the baseline and the parentage analysis conducted again in order to evaluate the level of false positive assignment. Like the juveniles analyzed previously, no parental assignments were made, providing a 0% false positive rate.
In the current study, no errors were made in assigning individuals to population of origin via two-parent PBT assignments (1707 assignments), but 4.3% (287 assignments) of single-parent assignments were incorrect to population. The assignments errors were confined to some

| GSI applied to 2018 fishery sampling
Fishery samples were obtained from 12 fisheries in 2018 and stock compositions were estimated for the samples ( genotyped from 12 fisheries, and estimated stock compositions corresponded to expectations based upon the locations of the fisheries.

| GSI applied to 2019 fishery sampling
In 2019, sampling was conducted for 17 fisheries in BC, with 12.533 Chinook salmon genotyped from the fishery sampling. A more detailed description of the GSI results available from the fishery sampling was outlined in Supplementary Results 2.4. In general, estimated stock compositions were consistent with expectations based upon geographic locations of the fisheries. Fisheries in more terminal locations were estimated to have been comprised almost entirely of Chinook salmon originating from a single CU (Table S5).

| Strait of Georgia legal versus sublegal catch
In the recreational fishery in the SoG in 2018 and 2019, individual Chinook salmon had to exceed 62 cm in length in order to be retained (legal), while those individuals shorter than this limit were released (sublegal). In 2018, substantial differences in stock composition were observed between legal-sized and sublegal-sized catches in the same month in the northern SoG fishery (Figure 3). For example, the ECVI-

| PBT applied to fishery sampling
In 2018 Creek individuals identified in virtually all fisheries outside of the SoG.
As in 2018, fishery PBT identifications for the Chilliwack River population (12.4%) were largely restricted to fisheries in Johnstone Strait, the SoG, and JDF, similar to the fishery PBT identifications of the Capilano River broodstock (5.0%). Notably, two fishery PBT identifications were made for the Nicola River population, a population of conservation concern and one in which virtually all of the hatchery production is currently marked with CWTs. One PBT identification was observed in July samples from the recreational fishery in the northern SoG, and one identification in July samples from the southern SoG (Table S7).

| Fishery age composition derived from CWTs and PBT
With 3591   =3.86, p > 0.14) or 2019 ( 2 ( 2 ) = 1.36, p > 0.50). CWT marking for the Nitinat River population only started in 2016, so there was no opportunity for recovery of CWTs from older-aged individuals that were observed with PBT. Fishery-derived age compositions for a population in a year were similar between CWTs and PBT.
Legal-sized individuals were primarily age 3 (83%) and 4 (16%) years, whereas sublegal individuals were primarily age 2 (68%) and 3 (32%) years, with little difference in the age composition in the fishery in the northern and southern regions of the SoG (Table 3).
Age composition of the sublegal catch could only be obtained via PBT, as the sublegal individuals had to be released, and thus it was not possible to determine age via CWTs.

| GSI applied in combination with PBT to fishery sampling
When PBT and GSI are applied in combination, it is possible to evaluate fishery samples for both age and stock composition. Sampling in the recreational fishery in the SoG indicated that Chinook salmon are present year round in the SoG, and that substantial differences in stock composition were observed in the northern and southern portions of the SoG. For example, in the northern SoG, winter residents were primarily sublegal-sized individuals from the Puntledge River, Qualicum River, and Cowichan River populations, whereas those in the southern SoG were primarily from Puget Sound populations (Table 4, Figure 4). Individuals from the lower Fraser fall age 0.3 CU were relatively more abundant in the southern SoG and arrived in April, with legal-sized individuals relatively more abundant than sublegal individuals. Chinook salmon from the summer South Thompson age 0.3 CU were more prevalent in the southern SoG and arrived in August and September. Seasonal, regional, size-related variability stock compositions were observed in the recreational fishery in the SoG.

| Estimation of catch by CU
The application of GSI and PBT enabled estimation of regional and CU contributions of Chinook salmon to the 2019 marine catch in BC.
Approximately 396,000 Chinook salmon were estimated to have been caught in marine fisheries in BC, with approximately 6% originating from CUs in northern and central BC, 1% from the southern BC mainland CUs, 17% from Fraser River CUs, 8% from ECVI CUs, 39% from WCVI CUs, and 29% from the western US (Table 5). In the Fraser River drainage, the South Thompson River summer age 0.3 CU was by far the dominant contributor to the BC catch (67% of drainage contribution, 11.6% of BC total catch), followed by the Lower Fraser fall age 0.3 CU (21% of drainage contribution, 3.6% of BC total catch). For the ECVI region, the Qualicum River-Puntledge River fall age 0.3 CU was the dominant contributor (63% of regional contribution, 4.6% of BC total catch) followed by the Cowichan River-Koksilah River fall age 0.x CU (33% of regional contribution, 2.4% of BC total catch). For the WCVI region, the WCVI south fall age x.3 CU was by far the dominant contributor (79% of regional contribution, 30.8% of BC total catch). The Robertson Creek population in this CU was estimated by itself to have contributed 24.0% of the total BC catch. The Nootka Sound-Kyuquot Sound fall age x.3 CU contributed 21% to the WCVI regional total and 8.0% to the BC catch, with the Conuma River population itself contributing 6.5% of the total BC catch. In 2019, Chinook salmon from Puget Sound contributed a minimum of 6.7% to the total BC catch, as did Columbia River-origin Chinook salmon (10.4%), Oregon-origin (4.6%), and California-origin (0.8%).   (Table S10).

| Assessment of hatchery broodstocks
As in 2018, no PBT assignments were observed in the Harrison River broodstock as the population is largely naturally spawned, but substantial assignments (73%) observed in the summer Puntledge River broodstock, a population which is largely hatchery origin. Note: CU catch was estimated as (monthly catch) * (population-specific monthly stock composition) summed over all populations in the CU in the baseline. Fisheries were as follows: 1) northern troll, 2) northern sport, 3) central sport, 4) central net and First Nations food, social, and ceremonial (FSC) fishery, 5) Johnstone Strait sport, 6) Strait of Georgia-north sport, 7) Strait of Georgia-south sport, 8) Juan de Fuca Strait sport, 9) west coast Vancouver Island sport, 10) WCVI troll, 11) WCVI gillnet and seine, 12) First Nation WCVI troll, and 13) all fisheries. a Includes an estimated 9% (7680 Chinook salmon) of WCVI sport catch which were adipose fin clipped but not otolith marked and which were assumed to be of US hatchery origin. Stock composition was not available for these estimated 7680 Chinook salmon and has not been included in individual US geographic regions.

TA B L E 5 (Continued)
of 150 individuals). Conversely, offspring from previously defined summer-returning parents were incorporated into the fall-return broodstock at a rate of 0.9% in 2018 (7 of 815 individuals) and 1.3% in 2019 (13 of 968 individuals).

| Estimation of exploitation rate (ER)
One of the main assessment requirements for fishery management is estimation of ER. Importantly, either CWTs or genetics can be applied to estimate population-specific ER. There was a wide range in observed population ERs in Canadian marine fisheries, ranging from the low exploitation rate population Nicola River (2.4% CWT, 1.2% genetics) to the higher rate population Robertson Creek (72.2% CWT, 71.7% genetics; Table 6). For the 13 populations evaluated, there was generally a close agreement between 2019 Canadian marine fishery ERs estimated via CWTs and genetics (r (11) = 0.960, p < 0.01; Figure 5). The greatest discrepancy in ER between the two methods (16.7%) was observed in the Cowichan River population, which was due to the under estimation of CWTs in the escapement sampling (see Section 4.2). With the Cowichan River population removed from the analysis, there was very close alignment between ERs estimated through CWTs and genetics for the remaining 12 populations (r (10) = 0.979, p < 0.01).
Age-specific exploitation rates are also of interest in fisheries assessment. For the three populations in which age 6 year comparisons were possible, there was good agreement between the two assessment methods (r (1) = 0.995, p < 0.01; Table 6).
There was poor agreement in exploitation rates of age 5 year Chinook salmon between the two methods in seven populations (r (5) = 0.572, p > 0.10), but this was largely due to the Big Qualicum River population (CWT 0.0%, genetics 73.8%). With the Big Qualicum River population removed, there was good agreement between the two techniques (r (5) = 0.958, p < 0.01).

| D ISCUSS I ON
The first accomplishment of the study was to demonstrate that PBT assignments were accurate with respect to identification of population of origin and individual age. The second accomplishment was to illustrate that the number of PBT identifications conformed to expectations of tag rates by delivering appropriate levels of observed assignments. The third accomplishment of the current study was to provide GSI-derived high-resolution stock composition estimates for 2018 and 2019 Canadian fisheries. This study marked the first time that Chinook salmon fisheries impacts in Canada could be evaluated by CU, and thus has enabled an assessment that was sufficiently informative for conservationbased management as envisaged in the WSP. There is no other method of fishery assessment that can provide this level of resolution for mixed-stock analysis. The fourth accomplishment was to merge a wide-ranging PBT-based assessment of fishery impacts with the GSI-based assessment to evaluate stock composition and age structure of both legal-sized and sublegal catches of Chinook salmon, illustrating that GSI and PBT can be applied in combination to provide information unavailable from CWTs. The fifth accomplishment was to sample 2019 fisheries with enough intensity for the genetic analysis to provide a realistic comparison with an existing CWT sampling program that has been in existence for many years. Merging these five accomplishments together has provided an opportunity to evaluate the finding an expert panel who predicted that PBT could provide the equivalent of CWT recovery data and could be easily integrated with a GSI program system to provide stock of origin for all fish sampled in fisheries (PSC, 2005).

| Accuracy of estimation of stock composition
One major application of a GSI-PBT approach to fisheries assessment is to estimate the stock composition of the catch, which can include contributions from populations that are not marked with CWTs. One major difference between CWT and GSI-PBT fisheries applications is the inability of the CWT approach to provide esti- Note: Age-specific exploitation rates were determined only for those populations in which at least 20 PBT identifications were observed in fishery sampling.
a Includes 49 age 7 year Kitsumkalum, and age composition derived from scale analysis. b Hatchery proportion of escapement determined via adipose fin clip rate.

TA B L E 6 (Continued)
Both PBT and GSI can be used to identify individuals to specific populations, but only PBT can determine the age of the individual, as once the parents are identified in the hatchery broodstock, age of the individual is easily determined by calculating the difference between the year of broodstock sampling and fishery sampling.
When the accuracy of identification of population of origin and age were combined with the levels of accuracy previously reported by Beacham et al. (2018) for population and age identification, we concluded that the level of accuracy for both population and age determined via PBT was sufficiently high (>99.8%) to justify application to mixed-stock fishery sample evaluation.

| Exploitation rate
Exploitation rates of Chinook salmon in 2019 Canadian marine fisheries derived from CWTs and genetics were highly comparable for 12 of 13 populations evaluated, with the Cowichan River population the exception. For that population, the exploitation rate derived from CWTs (51.2%) was higher than from genetics (34.5%). However, the number of CWTs observed in the escapement (estimated 592 CWTs) was considerably less than the number of adipose fin-clipped individuals (2488)  Age-specific exploitation rates across populations were comparable between CWTs and genetics for ages 3-6 years, with the age 5-year discrepancy limited to the Big Qualicum River population. In that instance, 2.4 times as many PBT identifications were made (94) as CWTs recovered (39) in the fishery sampling, with sampling error with respect to CWTs likely attributable for the discrepancy. Jack exploitation rates were estimated as 0.0% for three populations via CWTs (Big Qualicum, Harrison, Atnarko), whereas exploitation rates estimated via genetics ranged from 8.9% to 57.3%. This discrepancy in estimated exploitation was a result of no jacks from these populations observed in fishery samples via CWTs but they were present in the escapement through CWT or scale observations. In contrast, jacks were observed in the fishery samples for these populations via PBT. Given the observed accuracy in identifying jacks via PBT, it was unlikely that this discrepancy was a result of errors in age determination of individuals in fishery samples via PBT. PBT identification did not require an individual to be adipose fin clipped before detection, so perhaps there was less likelihood of an adipose fin-clipped jack to be sampled in fisheries than an intact jack. It is possible that individual fishers may have declined to have jacks sampled in the creel survey, as they may not have wanted an already small fish to be beheaded for CWT detection.

| Hatchery management
As the CWT system is impaired by inadequacies in sampling and assessment associated with mass marking of hatchery fish, some of which are released without a CWT, Canada has not implemented mass marking of Chinook salmon hatchery production partially over concerns for the integrity of the current CWT system of assessment.
An unintended consequence of this decision is the inability to harvest only hatchery-origin individuals in mark-selective fisheries, as there is no way to identify visually hatchery-origin individuals. This lack of harvest can subsequently lead to straying of hatchery-origin individuals into wild populations, and the inclusion of strays into hatchery broodstocks for populations of conservation concern. The question naturally arises as to whether a more appropriate fisheries assessment system can be applied that is complementary to mass marking of hatchery production via a visible mark such as an adipose fin clip.
The obvious solution to the previous question is a genetics- Whereas CWTs are of limited use in the study of hatchery-wild interactions, the non-lethal and simple tissue sampling has made genetic analysis of interactions commonplace in ecological studies (Sekino et al. 2005;Denson et al. 2012;Ashton et al. 2016). Moreover, the hatchery pedigree that can be obtained using PBT enables direct estimation of inbreeding and outbreeding effects in hatchery production and estimation of genetic parameters such as heritability (Kozfkay et al. 2008, Berejikian et al. 2017).

| Utility of PBT-GSI for fisheries assessment
In 2004, the PSC convened an expert panel to examine limitations of the CWT program for both Chinook salmon and coho salmon, and to evaluate the capacity of alternative technologies to provide data to improve assessment of salmon. The panel noted that PBT could provide the equivalent of CWT recovery data, but that an empirical demonstration was needed to validate theoretical PBT results that suggested broad feasibility (PSC, 2005). With no large-scale PBT applications developed in the intervening years, the PSC again commissioned in 2014 an evaluation of the feasibility 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 require that: 1. The PBT system generate 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).
The result of the 2014 PSC request for the previously noted evaluation led to the non peer-reviewed report of Satterthwaite et al. (2015). Various scenarios were explored in the report, and the report has been currently interpreted as concluding that the transition from a CWT-based assessment system to a genetics-based system is not cost effective or feasible (PSC Southern Endowment Fund Committee, pers. comm.).
The results of the current study, as well as those of Beacham When fisheries sampling was conducted, no distinction was made between Chinook salmon displaying an adipose fin clip and unclipped individuals. Therefore, a GSI-PBT program could proceed on the basis of no marking or limited visual marking of hatcheryorigin salmon as is done now with the CWT program. However, if all hatchery-origin Chinook salmon were visually marked with an adipose fin clip before hatchery release, fishery sampling could be targeted toward only the hatchery component, only the wild component, or both, depending upon the objective of the sampling. In essence, a GSI-PBT approach to fishery assessment provides an opportunity for mass marking of Canadian Chinook salmon hatchery production prior to hatchery release, and at no increased cost to an assessment program. In the CWT-based assessment program, release of adipose fin-clipped individuals without a CWT leads to degradation of the assessment system and was the reason for convening the expert panel in 2004 to assess alternatives for the system at the time. One issue associated with the reluctance of Canada to mass mark Chinook salmon hatchery production has been the potential impact on the assessment capability of the CWT program and the associated increased costs in conducting a suitable CWT-based assessment program. With no way to visually distinguish between hatchery-origin and wild-origin Chinook salmon, concerns for the integrity of the CWT assessment system have stifled appropriate management of hatchery broodstocks and development of markselective fisheries. A GSI-PBT method of assessment would facilitate improvements in both of these areas.
One concern often raised in implementation of mark-selective fisheries is the estimation of incidental mortality of unmarked released individuals. As noted by Satterthwaite et al. (2015), current attempts to evaluate the effect of mark-selective fisheries rely on double-index tag groups of individuals, in which hatchery releases of adipose fin-clipped and CWT-marked individuals are paired with individuals that receive a CWT but no corresponding adipose fin clip.
However, CWTs cannot be obtained from individuals released in a mark-selective fishery, so the impact of specific fisheries on mortality of unmarked individuals cannot be evaluated directly. Relative measures of survival may be estimated by recovery of CWTs from clipped and unclipped individuals in the escapements. In theory, a similar approach could be followed with a genetics-based evaluation, with PBT being used to identify hatchery-origin individuals in fishery sampling. If the clip status of the individual is recorded, differential impacts of the fishery on retained and released individuals could be directly estimated. Genetics provides the opportunity to evaluate fishery impacts on the released unmarked portion of the catch in a mark-selective fishery, provided that a DNA sample is obtained prior to release of the unmarked individual.
In implementation of GSI-PBT system of assessment for Canadian fisheries, the question of the appropriate sampling level for fisheries arises if a population of low abundance or conservation concern is present in fisheries. For Chinook salmon in BC, the population of conservation concern centers on the Nicola River population, a spring-run population in the lower Thompson River, a major tributary of the Fraser River drainage. As Chinook salmon fishery sampling incorporating PBT did not commence until 2018, and initially only in a limited manner, direct comparisons between CWT and PBT identifications were available only for 2019. In 2019, the number of Nicola River-origin CWTs recovered from marine fisheries in Canada was one CWT, and virtually 100% of hatchery production of this population is already marked with CWTs. Thus, there is no way to amplify the marking rate for this population. For genetics, two PBT identifications were observed in 2019 marine fishery sampling. Thus, there was little evidence to indicate that genetic analysis was less likely to identify Nicola River Chinook salmon, even though <3.5% of the catch was genotyped. Thus, there is no evidence to indicate that PBT analysis based upon current genetics sampling programs are less able to detect populations of conservation concern compared with the current CWT assessment program. Should the level of fisheries genotyping be increased to 23,000 individuals as noted previously, 84% higher than actually conducted in 2019, the increased sampling effort can be directed toward those fisheries and times where Nicola Riverorigin individuals are known to have been caught, such as in lower Fraser River fisheries.

| Future developments
The current study has demonstrated the GSI-PBT capability to identify BC-origin Chinook salmon to specific Canadian hatcheries and

| Summary
Parentage-based tagging provided an effective method to identify the age and hatchery of origin of individual Chinook salmon. GSI provided identification of mixed-stock fishery samples to CU, a requirement for implementation of management of wild populations as mandated by Canada's WSP for Pacific salmon. With 3.24% of the 2019 Chinook salmon catch in BC genotyped, fishery impact assessments via genetics were generally comparable to those via CWTs.
When GSI and PBT were combined in a concurrent application, the GSI-PBT method of fisheries assessment for Canadian populations outperformed a CWT-based assessment method on many levels, including stock composition resolution, cost, and ancillary information.
Moreover, unlike CWTs, GSI-PBT-based assessment benefits from the external (allowing visual recoveries) mass marking of hatcheryproduced 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 easily hatchery-produced salmon has been recognized as important for managing the risks and assessing the benefits of hatchery production of salmonids at the domestic, bilateral, and international levels.

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

CO N FLI C T O F I NTE R E S T
There are no conflicts of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
Multi-locus genotypes for all individuals sampled in 2018 and 2019 fisheries, as well as all hatchery broodstocks, will be available at DRYAD upon manuscript acceptance.