Gillanders et al : Aging methods for Seriola lalandi 



821 



als greater than 3 in the model 1 fit were removed 

 from the data set. Initially, a simple three-param- 

 eter model was fitted (model 2 in Table 4), which 

 indicated growth rates of 255 mm and 176 mm for 

 400-mm and 600-mm fish, respectively. The first 

 additional parameter selected was that describ- 

 ing the shape of the growth curve (model 3 in 

 Table 4). Model 3 showed annual growth rates of 270 

 and 140 mm. There was a significant improvement 

 in fit when a term describing growth variability was 

 added (model 4 in Table 4). Additional parameters 

 (e.g. seasonal growth terms) did not result in sig- 

 nificant improvements of fit. The best fitting model 

 was therefore case-1 model in Table 1. Plots of 

 standardized residuals against length at tagging, 

 time at liberty, and expected growth increment 

 showed no pattern (correlations were 0.022, 0.003, 

 and 0.005, respectively) suggesting that the model 

 was appropriate. 



The best fitting model for the kingfish length-fi-e- 

 quency data set identified five cohorts aged from 1.73 

 to 5.73 years (Table 5; Fig. 8). The standard devia- 

 tion of the predicted length estimates for each age 

 class increased with age. Parameters for first length 

 bias and seasonal growth were also included in the 

 best fitting model (Table 5; Fig. 9). Estimates of the 

 mean length-at-age ( and standard deviation ) for the 

 length-frequency data set are given in Table 5. The 

 mean lengths and proportions of the modes predicted 

 by MULTIFAN generally fitted the observed data 

 I Fig. 8). There were large numbers of small fish 

 (e.g. 1 yr) in the catch during the summer months 

 (e.g. November and December), whereas 2-yi- fish 

 dominated the catch at other times. Small numbers 

 of large fish (e.g. gi-eater than 5 yr) were found 

 throughout the year (Fig. 8). The seasonal form of 

 the von Bertalanffy growth equation showed that 

 the projected value of L,, was 1252 mm FL and the 

 rate of change in growth increment (K) was 0.189 

 (Fig. 9). 



Comparison of annual growth between age- 

 based (age-length and length-frequency) and 

 length-based (mark-recapture) data, although not 

 strictly comparable, showed a decrease in growth 

 with age and size offish (Fig. 10). Estimates of 

 annual growth were similar among the three 

 methods for 2-4 year fish (=550-750 mm SL) but 

 varied by =50 mm for 1-yr-old fish. 



Discussion 



All structures showed patterns of growth that were, 

 to varying degrees, quantifiable. Delineation of each 

 zone was, however, sometimes difficult, as has been 



q 



> 

 ■o 



P 





1.0,- 



08 



0.6 



0.4 



0.2 



0.0 



1.0 



— Vertebrae 

   Scales 

 — Otoliths 



_L 



_L 



0.5 - 



-0.5 



-1.0 



12 3 4 



Estimated most probable age (no. zones) 



Vertebrae D 



Scales 



Otoliths 



r 



_L 



_L 



12 3 4 



Estimated most probable age (no. zones) 



Figure 6 



(A) Aging error versus estimated most probable age for king- 

 fisli aged with otoliths, scales, and vertebrae. Estimated most 

 probable age was calculated separately for each structure by 

 using the methods of Richards et al. ( 1992), see text for fur- 

 ther details. Errors around the first and last ages are indi- 

 cated by the vertical lines. (B) Relative bias versus estimated 

 most probable age for kingfish aged by otoliths, scales, and ver- 

 tebrae. Relative bias was calculated from a model that incorpo- 

 rated two age readings from each of the three structures. 



found in other studies (e.g. Brennan and Cailliet, 

 1989; Manooch and Potts, 1997), and the clarity of 

 zones varied among individuals for all structures. 

 Because pelagic fishes including Seriola spp. are 

 known to be difficult to age, these results were not 

 surprising. 



To accurately reflect the age of a fish, the zones must 

 be formed on a regular and determinable time scale. 



