Estimation of precise ages from the marginal increment widths of differently growing sardine (Sardinops sagax) otoliths

More than 7000 sardine (Sardinops sagax) otoliths collected from commercial landings on the west, south, and east coasts of South Africa during the period 1980-1996 were aged in 1997. Annual and marginal increment widths were measured in all of the otoliths. Estimated precise ages were assigned to the marginal increments, judging their widths in relation to the widths of the preceding and expected annual increments. Three clustering strategies and quadratic discriminant analysis were employed to rank the otoliths of 5-year-old sardines in groups characterised by different growth patterns. The grouped 5-year-olds constituted a calibration data set which served to classify the otoliths in age-groups 1-4. In each age-group, three different data sets, hereafter called methods, were tested for their capacity to accurately predict precise ages from otolith measurements and fish lengths. Method 1 estimated the size of the marginal increment relative to the expected annual increment width at the following age. Methods 2 and 3 estimated precise age directly using the initial age estimations. Multiple regressions provided the weights for the significant (p<0.05) regressors. The performances of the three methods using (a) grouped or (b) pooled otolith specimens were compared by their sums of squared residuals and by testing the frequency distributions of the residual ages for significant differences with the Kolmogorov-Smirnov test statistic. The means and standard deviations of the residuals were plotted over the precise ages in each age-group and were examined for systematic trends. None of the methods provided the ultimate argument for its unrestricted use. Method 2 was given preference if slow-growing fish become proportionally more abundant at the older ages and/or growth rates vary markedly among year-classes. The specific advantages and disadvantages of each method are discussed in detail. (C) 2000 Elsevier Science B.V. All rights reserved.