Abstract
Linear discriminant analysis (LDA) based on scale patterns was used to develop a methodology of estimating regional origins of chum salmon. Age-4 fish were sampled in 2004–2006 from 12 river stocks of the Okhotsk and Sea of Japan (SJ) regions from Hokkaido to Honshu. The scale radius at the first annulus of each fish was separated into i intervals and the radius of each interval was divided by the number of scale circuli within the interval to quantify scale patterns. The i variables and five other morphometric measurements were used in a stepwise LDA to classify the following regional groups: Hokkaido and Honshu (I), Okhotsk and SJ (II), Okhotsk, Hokkaido SJ and Honshu SJ (III). Percentages of correctly classified fish (hit rates) improved with increased i but tended to be close to asymptotic values in all cases. Hit rates for each river stock in case (I) ranged from 74.3% to 100% (mean 97.2%), estimated by direct maximum likelihood methods using predictor variable sets from the best models for LDAs. Hit rates were lower in cases (II) and (III). This study demonstrated that scale patterns are useful for classifying the origins of chum salmon, at least between Hokkaido and Honshu.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have