|Document Type:||Contract Report|
|Title:||Columbia River stock identification study: validation of genetic method|
|Author/Editor:||George B. Milner, David J. Teel, Fred M. Utter, C. L. Burley|
|Publisher:||National Marine Fisheries Service|
|Contracting Agency:||Bonneville Power Administration. Portland, Oregon|
|Keywords:||Genetic stock identification|
The reliability of a method for obtaining maximum likelihood estimates of component stocks in mixed populations of salmonids through the frequency of genetic variants in a mixed population and in potentially contributing stocks was tested in 1980. A data base of 10 polymorphic loci from 14 hatchery stocks of spring chinook salmon of the Columbia River was used to estimate proportions of these stocks in four different "blind" mixtures whose true composition was only revealed subsequent to obtaining estimates.
The averaged differences of estimated and actual values of the 14 stocks in the four mixtures ranged from 0.9 to 7.4%; the precision of the estimates (measured by calculated standard deviations) ranged from 2.1 to 9.1%. Both accuracy and precision tended to improve when geographic groupings rather than individual stocks were considered, and dropped to a mean difference of 2.3% and a standard deviation of 1% when only two groups (stocks above or below Bonneville Dam) were considered. The accuracy and precision of these blind tests have validated the genetic method as a valuable means for identifying components of stock mixtures.
Properties of the genetic method were further examined by simulation studies using the pooled data of the four blind tests as a mixed fishery. Replicated tests with sample sizes between 100 and 1,000 indicated that actual standard deviations on estimated contributions were consistently lower than calculated standard deviations; this difference diminished as sample size increased. Both the accuracy and precision of estimates of a mixture involving two populations were greatest when only those two populations were included in the data base; the best estimates from a data base of all 14 populations were for that stock that was genetically most distinct from the remainder of the group.
Estimates of 87% above and 13% below Bonneville Dam were made on a sample of 123 fish collected in Astoria, Oregon, following a 24–hour winter fishery. The moderate precision of this estimate would have increased with a larger sample size and an a priori reduction in the presumed complexity of the fishery.
Costs of $10.29 per fish were estimated for sampling, analyzing, and estimating the stock composition of a mixed fishery sample of 500 fish if a turnaround time of 24 hours from landing of fish to obtaining estimates were required. These costs would be reduced if a longer time interval were permissible.
It is recommended that future applications of the method be preceded by simulation studies that will identify appropriate levels of sampling required for acceptable levels of accuracy and precision. Variables in such studies include the stocks involved, the loci used, and the genetic differentiation among stocks.