A Model Using Phenotypic Characteristics to Detect Introgressive Hybridization in Wild Westslope Cutthroat Trout and Rainbow Trout

Abstract

Introgressive hybridization is a substantial threat to native populations of cutthroat trout Oncorhynchus clarki ssp. To assess the status of native cutthroat trout and protect existing pure (nonhybridized) populations, fisheries managers need to identify introgressive hybridization in wild populations. Genetic techniques are the most reliable methods for detecting introgression but are typically expensive and time-consuming. Phenotypic characteristics are generally easy to measure in the field and have been investigated for their value in identifying hybrids in several genera of fish. We developed a practical quantitative tool for detecting introgressive hybridization in westslope cutthroat trout O. c. lewisi by fitting a classification tree model to the phenotypic characteristics of known pure and hybrid individuals. We then tested it as a means of making site-level assessments of the level of introgression. The genotypes of individuals were determined using noncoding sequences of nuclear DNA. Four dominant phenotypic characteristics (basibranchial teeth, throat slashes, spot shape, and relative head length) were significantly (P < 0.05) related to the genetic identity of the individual fish. The overall cross-validation and out-of-sample error rates of the classification tree (13.7% and 12.5%, respectively) were much lower than the visual-assignment error rate (38.8%) and suggested a substantial improvement in the accuracy of identifying individual fish in the field. The classification and prediction error rates were higher for hybrid trout than for westslope cutthroat trout. Site-level predictions of introgression made with our tree model had an overall error rate of 31%, which was 49% lower than that of estimates based on visual field identification by experienced observers. Model error for the site-level predictions was highest when identifying populations of westslope cutthroat trout with less than 50% introgression and lowest when identifying populations with more than 50% introgression. Our results indicate that a phenotype-based model is easy to use in the field and improves the accuracy and consistency of visual identifications of westslope cutthroat trout and hybrids of westslope cutthroat trout and rainbow trout O. mykiss. Our model or a similar one can be used to estimate the genetic status of local (site) populations.