454 



Fishery Bulletin 98(3) 



nificant variable represented the contribu- 

 tion of the respective morphometric variable 

 to the discrimination of silver hake samples 

 between regions. Jack-knifed cross-valida- 

 tion procedures were used to give unbiased 

 estimates of classification success (SYSTAT, 

 1997). 



Results 



Bottom ocean temperatures differed signif- 

 icantly between region and time of survey 

 (ANOVA interaction, F=11.7, df=15, 2931, 

 P<0.0001). Generally, temperatures were sig- 

 nificantly lower in the northern region than 

 in the southern region for all surveys (?-tests, 

 P<0.0001), except during the spring of 1994 

 and 1996 (although consistent differences 

 were still maintained between regions during 

 these times, albeit not significant) (Fig. 2). 

 Within each region, bottom temperatures have 

 a consistent seasonal pattern, tending to be 

 significantly lower in spring than autumn 

 for both regions (ANOVA, P<0.0001; HSD, 

 P<0.05 ), with no apparent differences between 

 years within seasons and regions (Fig. 2). 



Silver hake had significantly different growth rates 

 between sexes, year classes, and the regions from where 

 they were sampled (Fig. 3). Female silver hake grew at 

 a faster rate than males in both the northern and south- 

 ern regions; and age-length differences between the sexes 

 increased as the fish got older (ANCOVA, P<0.05). Growth 

 rates of each sex, within each region, also differed between 

 sampling years, indicative of the presence of a year class 

 effect (ANCOVA, P<0.005). Likewise, significant differ- 

 ences in growth rates were found between silver hake from 

 the northern and southern regions in each year for both 

 males and females (ANCOVA, P<0.005). 



The six otolith morphometric variables (length, width, 

 area, perimeter, circularity, and rectangularity) that were 

 measured for silver hake were log^.-transformed, except 

 for otolith width, to correct for non-normality and het- 

 erogeneity of variances. ANCOVAs detected significant 

 (P<0.0001) "region-fish length" interactions for four (oto- 

 lith area, length, width, and perimeter) of the six vari- 

 ables measured when all samples were examined together, 

 irrespective of age group, year class, or sex. However, the 

 use of individual ANCOVAs for fish of each age group 

 reduced the number of interactions to only one variable 

 for both 1-year-old (rectangularity) and 2-year-old silver 

 hake (width), and two variables (area, length) for samples 

 from 3-year-olds (Table 2). All the remaining variables for 

 the different aged samples were significantly correlated 

 with fish length (P<0.01), and therefore were corrected for 

 variable fish length with their respective common within- 

 group slope (Table 2). Consequently, length-corrected data 

 determined from the individual ANCOVAs for each age 

 group were used for the remaining analyses and indicated 

 the strong effect different ages can have on these types of 

 measurements. Typically, for any given fish length, silver 



hake from the northern region appeared to have larger 

 otolith dimensions (area, length, width, and perimeter) 

 than those fish from the southern region (Fig. 4). 



Multi- and univariate analyses were used to examine 

 confounding sources of variation that may have influ- 

 enced regional patterns of silver hake (Table 3). In gen- 

 eral, otolith morphometries were not significantly different 

 (P>0.05) between the sexes, indicating that their otoliths 

 develop at a similar rate over the age range examined. In 

 contrast, as for the growth rate analyses, significant dif- 

 ferences were found among otolith morphometries of the 

 same-age silver hake sampled from different year classes 

 for both males and females (Table 3). Likewise, similar 

 year-class differences in otolith morphometries were found 

 when the sexes were pooled together, confirming congru- 

 ency in rates of otolith development, and justifying assimi- 

 lation of sexes for statistical purposes. 



Silver hake sampled from the northern region tended 

 to have larger otolith dimensions than those from the 

 southern region (Fig. 4). Overall, when the appropriate 

 length-corrected morphometric variables were examined 

 together for silver hake of the same age group and year 

 class, irrespective of sex, significant differences were found 

 between samples from the northern and southern regions 

 (MANOVA, P<0.001) (Table 4). Fewer multivariate signif- 

 icant differences were detected for samples from 3-year- 

 olds, although these differences were probably the result of 

 samples for this age group having fewer variables to exam- 

 ine compared with samples from the other age groups. 

 This was supported by univariate tests where significant 

 differences (P<0.05) between silver hake from the north- 

 ern and southern regions were found for several variables, 

 for all age groups, and for almost all of the age group, 

 sex, and year class combinations (Table 4). Also, the oto- 



