Clardy et al.: Relative contribution of Scomberomorus cavalla stocks to winter fisheries off South Florida 
155 
cess was estimated as the per- 
centage of individuals correctly 
classified to stock. 
The contribution of the At- 
lantic stock to winter fishery 
landings in each winter sam- 
pling zone was estimated by 
using the maximum likelihood 
(ML) modeling approach de- 
scribed in DeVries et al. (2002). 
Mixing estimates were calcu- 
lated for males and females 
separately by sample year. 
Otolith-shape variables were 
used in a two-step expectation- 
maximization (EM) algorithm 
written for the S-Plus statisti- 
cal package (Insightful Corp., 
Seattle, WA) (Millar, 1987; 
DeVries et al., 2002). Sex- and 
year-specific ML models first 
were parameterized with oto- 
lith-shape data from summer- 
sampled fish. Then, the EM 
algorithm computed estimates 
of the percentage of landings 
within a given winter sam- 
pling zone that were members 
of the Atlantic stock based on 
their otolith shape parameters. 
A bootstrap procedure (n = 500 bootstraps) was used 
to compute bias-corrected ninety percent confidence 
intervals around the maximum likelihood estimate 
(MLE) of Atlantic stock contribution. 
Results 
Summer sample sizes differed somewhat between stocks, 
sexes, and sampling years. One hundred twenty-six 
king mackerel (60 females, 66 males) were sampled 
in summer 2001, and 110 fish (56 females, 54 males) 
were sampled in summer 2002 from Atlantic waters. 
Seventy-three fish (37 females, 36 males) were sampled 
in summer 2001, and 120 fish (71 females, 49 males) 
were sampled in summer 2002 from the GOM. The age 
distributions of summer-sampled king mackerel gener- 
ally were similar between sexes, migratory groups, and 
years (Fig. 3). 
Sex-specific sample sizes were more variable from 
south Florida sampling zones during winter than dur- 
ing summer. In winter 2001-02, 153 fish (85 females, 
68 males) were sampled in zone 1, 50 fish (44 females, 
6 males) were collected in zone 2, and 142 fish (67 
females, 75 males) were sampled in zone 3. In win- 
ter 2002-03, 158 fish (85 females, 73 males) were 
collected in zone 1, 72 fish (50 females, 22 males) 
were collected in zone 2, and 153 fish (86 females, 
67 males) were collected in zone 3. The age distribu- 
tions of winter-sampled king mackerel were skewed 
toward younger fish in comparison to summer samples 
(Fig. 4). 
Correlation analysis indicated some ontogentic ef- 
fects on otolith shape may have been present. Several 
shape variables were significantly correlated with 
fish length (area, perimeter, roundness, rectangular- 
ity, circularity, Fourier harmonics 1-9, 11-14, 17, 
19, and 20); the method described above was applied 
to remove the correlation of those variables with re- 
spect to fish length. MANOVA results indicated there 
were no significant differences in otolith shape be- 
tween left otoliths and right otoliths (MANOVA, 
P<0.601). 
Morphological features of otoliths proved to be 
different between stocks, but several other factors 
also significantly affected otolith shape. Sex and 
age, as well as stock, significantly affected otolith 
shape (MANOVA, PcO.001), but sampling year did 
not (MANOVA, P=0.964). Six of 25 shape variables 
were significantly different between sexes (ANOVA, 
P<0.05). Most of the shape differences were in vari- 
ables that described gross otolith morphological fea- 
tures (area, perimeter, roundness, circularity, and 
rectangularity), and only one of the significantly dif- 
ferent variables was a Fourier harmonic. Twelve of 
25 shape variables were significantly different among 
ages (ANOVA, P<0.05), and most of the differences 
were in Fourier harmonics. Nine of 25 shape variables 
were significantly different between stocks (ANOVA, 
P<0.05). Most of the stock-specific shape differences 
