Abstract

Abstract.-Variation at eight microsatellite loci and two linked exons of a major histocompatibility complex (MHC) locus was surveyed in approximately 21,000 coho salmon Oncorhynchus kisutch sampled from 138 localities ranging from southeast Alaska to the Columbia River, the majority of the sites being in British Columbia. The observed regional population structure enabled evaluation of the utility of using microsatellite and MHC variation for estimating the stock composition of coho salmon in mixed-stock fisheries. Both MHC exons were more effective for stock identification than any of the eight microsatellite loci examined. The two MHC exons combined were nearly as effective, on average, as the eight microsatellite loci combined. Some loci were particularly effective at discriminating stocks from specific regions. Mixed-stock analysis provided accurate estimates of contributions from the threatened Thompson River and upper Skeena River stocks, even when they composed less than 5% of the sampled fish. From about 17,000 coho salmon sampled from mixed-stock fisheries in British Columbia and Washington during 1997-1999, we found that the highest estimated proportions of coho salmon originating in southeast Alaska were in Canadian fishing areas adjacent to the international border in northern British Columbia; the highest proportions of Washington-origin coho salmon were observed closest to the international border in southern British Columbia. Within major river drainages, MHC variation within appropriately sampled fisheries can be used to determine the timing of spawning returns of specific stocks and the relative or absolute stock escapements. The application of molecular genetic markers to stock structure analysis and mixed-stock analysis of anadromous salmonids has been extensive because of the economic importance of these fish and the relative ease of sampling temporally or spatially segregated spawning aggregations In 1995, we began to develop a comprehensive genetic database for coho salmon in British Columbia that would assist in identifying and selecting conservation and management units of British Columbia coho salmon. We believed the database would also provide sufficiently accurate and precise estimates of stock composition in mixed-stock samples and thereby enhance conservation-based fisheries management. We chose to survey variation at eight microsatellite loci and 1117 STOCK IDENTIFICATION OF COHO SALMON two exons (coding portion of a gene) of a major histocompatibilty complex (MHC) locus. We used a PCR-based (polymerase chain reaction) approach to ensure cost effectiveness and speed in establishing the database and to enable nonlethal sampling for mixed-stock analysis. Microsatellite loci are abundant, highly polymorphic, and noncoding (considered selectively neutral), and provide genetic information on nonselective forces, including mutation and drift. As such, they can be used to generate estimates of gene flow, effective population size, and phylogenetic relationships. Vertebrate MHC genes encode cell-surface glycoproteins that are functional in the adaptive immune system. They evolve rapidly, are highly polymorphic, and because they encode adaptive variation, are subject to natural selection. The adaptive nature of MHC genes compromises use of MHC allele frequencies to estimate parameters for which an assumption of selective neutrality is required. However, MHC allele frequencies have the potential to enhance stock specificity and thus their utility in mixed-stock analyses. Moreover, variation in MHC allele and genotype frequencies attributable to selective forces provides quantitative information on the adaptive variation among salmonid stocks that conservation efforts are directed at preserving (Miller et al. in press). The two linked class-I MHC exons surveyed in this study exhibit high levels of polymorphism, heterozygosity, and temporally stable differentiation among coho salmon populations After having received scientific advice in 1998 that the abundance of Thompson River and upper Skeena River coho salmon was at critically low levels (Stocker and Peacock 1998), the Minister of the Department of Fisheries and Oceans directed that the management of Canadian fisheries in 1998 was to be conducted with the objective of achieving a zero mortality of those salmon. Fisheries were curtailed in areas where Thompson River and upper Skeena River coho salmon were believed to be prevalent. Salmon fisheries in other areas could proceed if they were unlikely to intercept significant numbers of coho salmon, and generally, all coho salmon caught in any British Columbia fishery were to be released. Coded wire tag (CWT) analysis depends upon recovery of CWTs from dead fish, so under the 1998 management objectives, the traditional stock identification information from CWTs would not be available. However, by 1998, extensive surveys of microsatellite and MHC variation had been conducted, the general units of population structure of coho salmon had been defined, and the feasibility of DNA-based MSA had been assessed In this study, we evaluate the utility of using microsatellite and MHC data for coho salmon stock identification through simulation analyses, apply the technologies to estimate stock composition of known-origin samples of coded-wiretagged coho salmon, and outline the applications to estimating stock composition for coho salmon fisheries sampled in British Columbia and Washington during 1997-1999. Methods Collection of DNA samples and laboratory analysis.-Genomic DNA was extracted from either liver, scales, operculum punches, or fin clips from coho salmon sampled between 1987 and 1999 using the phenol-chloroform protocol of class-I MHC exons was surveyed by denaturing gradient gel electrophoresis (DGGE) Collection of the CWT sample.-In 1997, coho salmon could still be landed and retained in British Columbia fisheries. The program to recover codedwire-tagged fish was in operation, and we were able to obtain operculum punches from coho salmon that had previously been marked with CWTs and for which the CWT had been recovered and decoded for marking location (source population). We subsequently used this sample of 264 fish to evaluate the accuracy of estimated stock compositions using a sample of known origin. Collection of fishery samples.-In 1997, samples were collected from the recreational fishery off southwestern Vancouver Island and in test fisheries in the lower Fraser River in southern British Columbia. In 1998, when coho salmon were not to be retained in most fisheries in the province, sampling coho salmon from the fisheries was challenging. Sampling effort was expanded considerably; observers aboard troll, purse seine, and gillnet vessels sampled the bycatch of coho salmon before their release. Obtaining samples from the recreational fishery was difficult; there were no landings to sample, and it was not practical to place observers aboard individual vessels. Samples from these fisheries were generally obtained either from individual guides or charter boat operators, or from members of the British Columbia Wildlife Federation. The DNA samples from the 1998 and 1999 fisheries were obtained from either operculum punches or fin clips preserved in 70% ethanol. To facilitate rapid analysis of fishery samples, we generally screened them for variability at both MHC exons and at four microsatellite loci. The microsatellite loci screened for the 1997-1998 samples were Ots2, Ots3, Ots101, and Ots103, whereas the loci screened for the 1999 samples were Oki1, Oki10, Oki100, Oki101. Baseline populations.-Applying DNA variation to estimates of stock composition in mixedstock fisheries requires surveying variation in contributing populations at a sufficient number of genetic markers to provide reliable determination of population structure and, thus, estimates of stock composition. The baseline survey consisted of analysis of approximately 21,000 coho salmon in 138 populations from geographic areas where coho salmon are likely to occur in British Columbia fisheries. These populations included 1 from Oregon, 17 from Washington, 111 from British Columbia, and 9 from southeast Alaska ( Conversion of allele sizes between manual and automated sizing systems.-The ABI 377 automated sequencer was obtained in our laboratory during the 1998 fishery to shorten the processing time for the approximately 9,000 samples collected from fisheries throughout British Columbia. At that time the baseline microsatellite data consisted of manual gel data for only four (Ots2, Ots3, Ots101, and Ots103) of the eight microsatellite loci used in this analysis. For the 1998 fishery samples, we surveyed variation at Ots3, Ots101, and Ots103 on the automated sequencer and retained Ots2 on manual gels. Given the wide distribution of allele sizes of Ots101 and Ots103 and the limitation of three fluorescent dyes for microsatellites on the sequencer, we were not able at that time to analyze Ots2 on the sequencer. Estimated allele sizes at Ots3, Ots101, and Ots103 differed between the manual nondenaturing gels stained with ethidium bromide and the automated sequencer denaturing gels with fluorescently labeled alleles. To convert allele sizes between the two systems, we analyzed approximately 600 fish on both systems and determined the distributions of allele frequencies. By inspection of the allele frequencies, we were able to match specific allele sizes obtained from the sequencer to specific allele sizes from the manual gels and then convert the sizing in the automated sequencer data set to match that obtained from the manual gels. Estimated allele sizes from both systems were very highly correlated (r 2 ϭ 0.987 for Ots3, 0.998 for Ots101, and 0.999 for Ots103). In general, sizes for the same allele from the sequencer were larger than those estimated from manual gels, and the differential increased directly with allele size. Estimating stock composition.-Genotypic frequencies were determined at each locus in each population. The statistical package for the analysis of mixtures software program (SPAM; Reported stock compositions for the CWT and actual fishery samples are the point estimates of each mixture analyzed; variance estimates were derived from 100 bootstrap simulations. Each baseline population and fishery sample was sampled with replacement in order to simulate random variation involved in the collection of the baseline and fishery samples. Reported stock composition for simulated mixtures was the bootstrap mean and standard deviation. Coastal British Columbia is divided into statistical areas for salmon catch reporting and management ( Results Population Structure If a regional genetic structure among populations contributing to a fishery exists, then it is unnecessary to survey all individual populations that contribute to the fishery. The portion of the mixed-stock sample derived from unsampled populations is allocated to sampled populations from the same region, reducing the cost and complexity of establishing a baseline sufficient for mixture analysis. The sampled populations constitute the baseline used to estimate stock compositions in mixed-fishery samples. Regional structure was observed in the baseline populations, the Thompson River populations being the most distinct of 15 geographically based groups or stocks (Table 2; Comparison of Individual Loci Determining the relative power of individual loci for regional discrimination is of prime importance for practical stock identification applications. Of the 10 markers surveyed in our study, the MHC exons were individually more effective for stock identification than any of the eight microsatellite loci Coho salmon from some regions were more easily differentiated than those from other regions. The distinctive Thompson River coho were clearly well differentiated from coho salmon in other regions, regardless of the loci examined. When all 10 loci surveyed were used, coho salmon from the west coast of Vancouver Island (WCVI) were the most difficult to discriminate when mixed with populations from other regions, whereas those from the east coast of Vancouver Island (ECVI) populations were accurately discriminated Some loci were particularly effective at discriminating populations from specific regions. For example, the two MHC exons were more powerful for identifying coastal Washington and Columbia River populations than were microsatellite loci. However, the combined microsatellite loci were more effective at identifying Vancouver Island coho salmon than were the MHC exons. Although the overall discriminatory ability of Ots101 was only moderate, it was particularly effective for discriminating Thompson River coho salmon (e.g., the average estimated composition of pure samples of Thompson River coho salmon was 98% using only this single locus in the 138-population baseline; Thompson and Upper Skeena River Identifications Since 1998, Canadian salmon fisheries have been conducted to minimize mortality of Thompson River and upper Skeena River coho salmon. Accurate estimates of these two stock components in mixed-fishery samples were thus essential for proper management. We were also interested in separating Thompson River from upper Fraser River populations, a stock of uncertain status that has genetic characteristics most similar to Thompson River populations Estimates of Regional Stock Composition We evaluated whether the genetic differentiation observed among the 138 coho salmon populations included in the baseline was sufficient for mixedstock analysis aimed at estimating regional contributions to fishery samples. Three fishery-mixture samples were simulated, and stock compositions were estimated for 16 regions. For stock contributions ranging from 0% to 20% of the mixture, the estimated bootstrap mean of a region was usually within 0.0-1.5% of the actual composition in the mixture For eight regional groups of coho salmon we evaluated the accuracy of estimated stock compositions in simulated mixtures, based on compositions of the target region ranging from 0-100% and only 6 of the 10 loci surveyed being used. Very little bias was observed when the region composed less than 40% of the mixture ( Identification of Specific Populations Accurate differentiation of mixture samples to specific populations was generally not possible because not all populations contributing to a fishery sample were included in the baseline. However, situations could occur in which all populations contributing to a fishery sample could be sampled. Such a case arose for the proposed ''mark-only'' fishery for coho salmon in southern British Columbia and Washington State in which hatchery fish, marked by a clipped adipose fin, may be retained but naturally spawned fish, identified by the presence of an adipose fin, must be released. We evaluated the accuracy of the estimated stock composition for each Canadian population by simulating mixtures for six southern British Columbia hatcheries for which population-specific estimates of stock composition are required. The baseline was substantially reduced to include only the six Canadian populations, but all populations from Washington were retained. Analysis of three simulated mixtures indicated that accurate hatcheryspecific estimates of stock composition could be obtained if applied to samples from mark-only fisheries Analysis of a Sample of Known Origin The superiority of using expected over observed genotypic frequencies for baseline samples was confirmed for the mixture sample containing fish identified by their CWTs. The sum of errors in estimated stock composition was always less when expected genotypic frequencies for all loci were used than when observed genotypic frequencies for some loci were used Analysis of Fishery Samples: Southern Baseline The estimated proportion of Thompson River coho salmon in mixed-stock samples was of key importance to Canadian fishery managers in 1998 and 1999. In 1998, we were unable to distinguish reliably between Thompson River and upper Fraser River using the loci surveyed in the mixedstock sample. Indeed, it was only after the introduction of DNA analysis to the mixed-stock samples that separation of the two closely related stock groups was considered of management importance. Therefore, upper Fraser and Thompson stock estimates were combined in the 1998 mixedfishery samples, but reported separately for the 1999 samples because of the change in the loci surveyed. Estimated stock compositions of Thompson River coho salmon were never above 2% in the Pacific Salmon Commission (PSC) seine test fishery conducted from late July to late August in Area 20 (Strait of Juan de Fuca) and rarely above 2% for the PSC gill-net test fishery conducted from early July through mid-August in a similar area Recreational fishery sampling in the Strait of Georgia (Areas 14-19) indicated that coho from Vancouver Island, the lower British Columbia mainland, the lower Fraser River, and Puget Sound predominated the catch in the summer, but October samples in Area 14 indicated that ECVI stock was predominant, composing 85% of the sample (Appendix 1). By October, coho salmon from other areas have probably moved from the Strait of Georgia and closer to their respective spawning grounds. The major contributor to fisheries in Canada's Area 20 in the Strait of Juan de Fuca was the Puget Sound stock, composing nearly 40% of the coho sampled in the seine and gill-net test fisheries (Appendix 1). However, the relative proportion of the Puget Sound stock in Canadian recreational fish-1130 BEACHAM ET AL. TABLE 7.-Percentage composition (SD) of a sample of coded-wire-tagged coho salmon obtained from fisheries in British Columbia in 1997 and estimated with three sets of loci for three groups of baseline populations. Because all fish in the sample were marked with coded wire tags, the actual composition of the sample is known. Set-1 loci include ␣1, ␣2, Ots2, Ots3, Ots101, and Ots103; set-2 loci include ␣1, ␣2, Oki1, Oki10, Oki100, and Oki101; set-3 loci include ␣1, ␣2, and all eight microsatellite loci. In state 1, the expected Hardy-Weinberg genotypic frequencies were used for all loci for the appropriate baseline populations. In state 2, observed genotypic frequencies for Oki100 and Ots103 were used. Analysis of Fishery Samples: Central Baseline A major interception fishery occurs in the Queen Charlotte Strait and Johnstone Strait (Areas 11-13; The troll fishery is the predominant fishery occurring off the west coast of Vancouver Island (Areas 124-127). The area and time of highest Thompson River proportion in the fishery samples was the first two weeks in August in the northern (Area 125-127) troll fishery, the Thompson stock estimated at 3% in the samples. Generally, the upper Skeena stock was estimated at negligible levels in the samples. Most of the fish sampled originated from Vancouver Island, the southern mainland, the lower Fraser River, and Puget Sound. Higher proportions of Canadian-origin coho salmon were sampled in this fishery compared with the more southerly fishery in the Strait of Juan de Fuca (Area 20) (Appendix 2). Off the west coast of Vancouver Island, about 70-80% of the sample was estimated to be of Canadian origin, compared with about 40-50% for samples from the Strait of Juan de Fuca. Analysis of Fishery Samples: Northern Baseline In northern fisheries, the upper Skeena stock was of greatest management concern. For fisheries adjacent to the Queen Charlotte Islands (Areas 1, 2W, and 2E), this stock was only detected in a late July troll fishery on the west coast of the Queen Charlottes (2W), and then was estimated to have composed 3% of the 99-fish sample (Appendix 3). However, in Area 3, this stock composed 15% of a 153-fish sample from a seine fishery in the last half of July 1998 and 8-25% of much smaller samples from gill-net fisheries in Areas 3 and 4 taken at the same time. The Thompson River stock was estimated to have contributed only negligible amounts to these fishery samples. There were clear differences in stock composition between fisheries on the east coast and west coast of the Queen Charlotte Islands. On the east coast (2E), samples from both the seine and gillnet fisheries from mid-September to mid-October 1998 indicated that coho salmon from the Queen Charlotte Islands predominated the fishery, composing about 70% of the samples from both fisheries (Appendix 3). However, on the west coast (2W), the Queen Charlotte Islands stock composed less than 20% of the fishery samples from late July and August 1998. The estimated contributions of Alaskan-origin coho salmon were highest in Canadian fishing areas closest to the northern border. Alaskan-origin coho salmon composed up to 20% of the sample from Area 3, and although only 21 fish were sampled in Area 1, nearly 20% of that sample was estimated to have been derived from Alaskan populations. In northern British Columbia, the northcentral coast stock was the predominant contributor to fisheries; coho salmon from Alaska, the lower Skeena River, WCVI, and NVI composed, at times, significant proportions of samples. Central coast fishery samples (Areas 6 and 7) were predominated by the northcentral coast stock, with Vancouver Island and southern mainland populations at times making significant contributions (Appendix 3). Analysis within Major Watersheds: Fraser River Baseline The key question in sampling fisheries within the Fraser River drainage related to the relative abundance of Thompson River coho salmon, particularly the migration timing of the stock through the lower Fraser River. Three years of sampling by a test fishery in the lower portion of the river indicated a consistent pattern. The Thompson and upper Fraser stock composed 35% of the coho salmon in the lower Fraser River before September 22 and declined rapidly thereafter to the first week of October ( Analysis within Major Watersheds: Skeena River Baseline In 1998, all coho salmon caught in a test fishery in the lower Skeena River were analyzed to provide migration timing information on specific components of the run. In 1999 we analyzed a subsample of the returns that was taken over time and in proportion to run abundance. Several identifiable substocks exist within both the upper and lower portions of the watershed. The upper Skeena stock (incorporating populations upstream from the confluence of the Skeena and Babine rivers) encompasses the following substocks: upper drainage tributaries, the Babine River, and the Bulkley and Morice rivers. The lower Skeena stock encompasses the following substocks: mid-Skeena, Lakelse Lake and River tributaries, and the lower drainage tributaries. Conservation concerns over the marine fisheries addressed the entire upper Skeena stock. To satisfy local management concerns, we estimated contributions of each of the three upper Skeena substocks. Not surprisingly, coho salmon from the upper Skeena drainage were most prevalent in samples from the early portion of the returns sampled in the lower river test fishery. Composing less than 20% of the sample in the last half of July, upper drainage coho salmon declined to negligible proportions after the first week of September (Appendix 5). Both Babine River and Bulkley and Morice coho salmon were largely absent after mid-September. Conversely, the later-spawning lower river substock reached its relative peak of abundance in September, composing over 50% of the coho salmon in the river at the time. The substock was also the predominant stock in the total return to the river, composing nearly 40% of the total return in both 1998 and 1999. The least abundant substocks were those from the upper drainage and Babine River, each composing less than 10% of the total returns in both years. Discussion Ideal technologies for mixed-stock analysis are those based on biological variation in characters that differ substantially among stocks, show little temporal or annual variation within stocks, and can be screened in a rapid, nonlethal, and cost-effective manner for both baseline and mixed-stock samples. The PCR-based survey of single-locus allele frequencies at microsatellite DNA and MHC loci meet these criteria and can be used for mixedstock analysis in a species for which no alternative methodology is available. The stock composition estimates reported in this study have been derived from fisheries throughout coastal British Columbia and in the Strait of Juan de Fuca in Washington. They agree with general expectations for coho salmon distribution and migration timing. These include higher proportions of fish of U.S. origin in fishing areas closest to the northern and southern international boundaries, substantial mixing of stocks in fisheries off the west coasts of both the Queen Charlotte Islands and Vancouver Island, and higher proportions of upper drainage stocks in the early returns to both the Fraser and Skeena rivers. Stock compositions of the two stocks of greatest current management concern, the Thomp-1133 STOCK IDENTIFICATION OF COHO SALMON son River and upper Skeena River, were generally within expectations, which should enable development of effective fishery management options to conserve both stocks. This study indicated that microsatellite and MHC variation can provide reliable estimates of stock composition to any fishery in British Columbia. The accuracy of stock composition estimates was enhanced by assuming a Hardy-Weinberg distribution of genotypic frequencies within baseline population samples for loci at which observed genotypic frequencies did not conform to expected values. This is probably because genotypes may have occurred in a mixture that was not observed in baseline samples, given the limited number of fish sampled per population and the high degree of polymorphism at the loci surveyed. Under the assumption of Hardy-Weinberg equilibrium, all possible genotypes resulting from the observed alleles at a locus are assigned some probability of occurring in the baseline population. This enables assigning, at some positive probability level, a genotype in a mixture sample to populations lacking the observed genotype in the baseline samples. The development, evaluation, and implementation of the coho salmon genetic database provided some practical information of value to future applied studies. The inclusion of loci that may be affected by natural selection on appropriate geographic and temporal scales can increase the stock identification capabilities for a database to be used in mixed-stock analysis. In this study, an increase from four to eight microsatellite loci did not result, on average, in improved accuracy of estimates when the microsatellite data were used with the MHC data. There was, however, an increase in precision of the estimates when all 10 genetic markers were used. Similar results were observed by Conservation concerns engender strict requirements for limiting fishing mortality of stock components that are almost certain to be present at very low abundance in mixed-stock samples. The management objective of zero mortality for Thompson River and upper Skeena River coho salmon created an urgent requirement for information on their presence in different fisheries, even those targeting other salmonid species. Mortality estimates of a nontarget stock component such as these are the product of three factors: the encounter rate of the component by a particular fishery, the assumed mortality of nontarget fish after release, and the estimated proportion of the component stock among the fish released. When projected mortalities of Thompson or upper Skeena coho salmon were considered to be contrary to the policy of zero mortality, the respective Canadian fisheries were curtailed, regardless of the species they targeted. When conservation concerns for a particular stock drive the management of a fishery, it is clear that accurate estimation of stock composition of the nontarget stock in fishery samples is vital. However, when there are conservation concerns for a particular stock, it is typically in very low abundance in fishery samples, and the accurate estimation of a stock composition of very low percentage is particularly challenging. Fortunately, in southern British Columbia, Thompson River coho salmon are very distinct genetically, and our simulation analyses and estimation of this component in the CWT sample indicated that its accurate estimation in fishery samples is possible using the DNA technology outlined in our study. Similarly, in northern British Columbia, accurate estimation of the upper Skeena River component was possible, even when the stock composed a very small (Ͻ5%) portion of the sample. In general, very little bias was observed when coho salmon from a particular region (not just Thompson River or upper Skeena River) composed 40% or less of the mixture. The provision of stock composition estimates 1134 BEACHAM ET AL. for coho salmon specific to area, time, and gear allows fishery managers to evaluate the impact of options such as changing fishing boundaries or times to apply conservation measures at a local level. One such situation existed in the 1999 recreational fishery samples from Bamfield and Ucluelet in Area 23. The recreational fishery at Bamfield (situated on the southern shore of the entrance to Barkley Sound) was centered more inshore in Barkley Sound, whereas the recreational fishery at Ucluelet was at the northern shore of the entrance and more seaward. Thompson River coho salmon were not detected in the 247-fish sample from Bamfield, but 4.8% of the 115-fish sample from Ucluelet was estimated to have been of Thompson River origin. Use of DNA analysis of freshwater test fishery samples can provide stock-and substock-specific escapement estimates and information of run timing. For example, the relative proportions of different stocks were estimated for coho salmon in the Skeena River, which combined with absolute abundance information for one of the stock components, can provide a method to estimate escapements for the other stock components. Similar applications of microsatellite variation for estimating sockeye salmon escapement have been described by The two technologies available for stock identification of coho salmon in mixed-stock samples (CWTs and nuclear DNA analysis) differ in their strengths and weaknesses. Coded-wire tagging normally consists of inserting a 1-mm binary-coded wire into the nasal cartilage of juvenile salmon and removing the adipose fin as a means of identifying tagged fish as adults. Only populations accessible to juvenile handling can be marked with CWTs, and the population must be tagged each year. Catches of adult salmon are usually sampled to find adults without an adipose fin, CWTs are recovered from these tagged fish, and then the data from CWT recoveries are expanded to account for the juvenile tagging rate, the catch sampling rate, and any lost heads or lost tags. The tool is very useful for providing catch estimates of tagged populations (Bernard and Clark 1996), but the application to estimating stock composition is limited by the generally low percentage of fish tagged in tagged populations and the mixed presence of untagged populations in the adult return samples. Tagged populations are assumed to be representative of the distribution of untagged populations within the same geographic area, which may not be true. In genetic analysis, baseline information is typically not available from all populations encountered in a fishery. However, for species in which genetic differentiation is generally based on isolation by distance, there tends to be a regional structuring of populations. Thus, for management decisions requiring regional estimates of stock composition, all contributing populations do not to have to be sampled. That portion of a mixedstock sample derived from unsampled populations is usually allocated to sampled populations from the same region, reducing the cost and complexity of establishing a sufficient baseline for mixture analysis. Thus, DNA-based mixed-stock analysis and CWTs are complementary technologies for identifying coho salmon stocks and estimating their exploitation. For regional stock composition estimates, DNA technology is more likely to provide accurate estimates of stock compositions than CWT recoveries. For estimates of catch, CWTs are more likely to provide reliable results than genetic analyses, which typically lack baseline information for all populations contributing to a fishery. Nearby unsampled populations are probably similar genetically to the sampled and CWT-marked population in question, so genetic analyses are likely to overestimate the contribution of the intended population. When all populations contributing to a fishery sample have been included in the baseline, it may be possible to use genetic data to provide catch estimates of specific populations. Our evaluation indicated that the current genetic database for coho salmon could be used to monitor hatchery-specific catch in ''mark-only'' fisheries introduced to conserve wild populations. All of the limited number of hatchery populations have been included in the baseline, and 10 genetic markers have revealed sufficient differentiation among populations to enable accurate estimates of catch by hatchery. Similarly, if management needs exploitation rate estimates for specific populations, once the catch is estimated and estimates of escapement are available, then exploitation rate can be estimated without any reference to CWTs. Widespread application of microsatellite and MHC variation to estimating stock composition of coho salmon in mixed-stock fisheries in British Columbia was conducted because of acute conservation concerns for specific stocks. The mixedstock analysis enabled accurate estimates of stock composition in mixed-stock fishery samples, even 1135 STOCK IDENTIFICATION OF COHO SALMON for stocks that were released alive after sampling, which alleviates concerns over sampling-induced mortality in the endangered stocks. The genetic database developed for coho salmon is a model that will probably be applied to an increasing number of exploited populations that give rise to the twin management concerns of identifying conservation units and detecting their presence in mixedstock fisheries.