160 



Fishery Bulletin 91(1), 1993 



Table 1 



Sampling periods for age-0+ snapper Pagrus auratus, 1987 

 and 1988 year-classes. 



Period 



1987 



1988 



27 January 

 2 March 



28 April 

 30 June 

 24-27 August 

 19 October 

 14 December 



4 February 

 15 March 

 6 April 



30 May-7 June 

 25 July^t August 



31 October 



29 Nov-20 Dec 



on somatic growth rate. Regressions were fitted to plots 

 of size-variables vs. FL, and the residuals were plotted 

 against somatic growth rate. The latter was estimated 

 by the equation 



Somatic growth rate = (FL-8)/(post-metamorphic age), 



where the constant 8 represents approximate mean 

 length of snapper at metamorphosis (Fukuhara 1985 

 and 1991, Foscarini 1988, Battaglene & Talbot 1992). 

 Post-metamorphic age-at-capture was estimated from 

 transverse sections by counting daily increments be- 

 tween the metamorphic mark and the section margin 

 (see Francis et al. [In press] for validation of daily 

 increments). Post-metamorphic age was used rather 

 than post-hatch age because only about 10% of our 

 sections contained cores; use of post-hatch age would 

 have severely limited sample sizes. Similar analyses 

 were not performed on data from Period 1 because of 

 small sample size, nor on data from Periods 4-7 be- 

 cause daily increments deposited during winter are 

 not resolvable with a light microscope (Francis et al. 

 In press). 



Results 



Figure 1 shows plots of sagitta size-variables vs. FL 

 for all sampling periods and age-classes combined. 

 Sagitta weight increased exponentially with FL 

 (Fig. 1A). Plots of sagitta length, height, and sulcal 

 width vs. FL were convex, with slopes decreasing over 

 the range 35-300 mmFL (Fig. IB). 



Data for 0+ snapper of the 1987 and 1988 year- 

 classes collected in Periods 1-7 were extracted for fur- 

 ther analysis by ANCOVA. Because only linear rela- 

 tionships can be analyzed by ANCOVA, sagitta weight 

 and FL were log,,, transformed before the relationship 

 between them was analyzed. Relationships between 

 the other three size-variables and FL are clearly 



nonlinear (Fig. IB). However, ANCOVAs fit linear re- 

 gressions to individual samples (i.e., sagittae of snap- 

 per from one year-class caught in one period), which 

 span only short segments of the lower end of the FL 

 range shown in Fig. IB. All samples were tested for 

 nonlinearity by regressing size-variables against FL, 

 and plotting the residuals against FL. There were no 

 trends in the residuals, so the untransformed data 

 were used in the ANCOVAs. 



The first set of four ANCOVAs (one for each size- 

 variable) tested the effects of year-class and sampling 

 period on sagitta size, using FL (or log„)FL) as the 

 covariate. There were no significant interaction terms 

 involving year-class, and the year-class factor itself 

 was not significant (p>0.05) in any ANCOVA. There- 

 fore, data for the two year-classes were pooled for sub- 

 sequent analyses. 



