A comparative study of new methods of quantifying scale shape for stock discrimination
Watkinson, D. A.
Conventional methods of scale characterization commonly used to discriminate between stocks of different fish species have proven to be ineffective as a tool for fisheries management. This study investigated new analytical techniques for stock discrimination such as averaging of scale outline signals, wavelet signal processing methods, and computer intensive non-parametric statistics. These techniques were used to test the significance of discriminant results in a combined effort to improve the researcher's ability to discriminate between fish stocks. This study found that combining signals from several scales significantly improved the ability to discriminate between stocks. Non-parametric statistics effectively tested for significance and significant differences in analyses when assumptions of discriminant analysis are violated. Variables produced from wavelet decompositions formed significantly better discriminant functions than Fourier analysis variables for most comparisons. This research has increasedour ability to discriminate between fish stocks based upon scale morphology.