Clardy et al.: Relative contribution of Scomberomorus cavalla stocks to winter fisheries off South Florida 
157 
Table 1 
Accuracies of jackknifed classifications from stepwise linear discriminant function models computed with otolith shape param- 
eters to estimate summer king mackerel (Scomberomorus cavalla) stock identity. The model column identifies which sex- and 
year-specific models are examined. Numbers in the parameters column represent Fourier harmonics; Ro = Roundness, Re = 
Rectangularity, and P = Perimeter. Remaining columns indicate the percentage of fish correctly classified to the Atlantic and 
Gulf of Mexico (GOM) stocks with the jackknife algorithm. 
Model 
Parameters 
Accuracy (%) 
for Atlantic region 
Accuracy (%) 
for GOM region 
Mean accuracy 
(%) 
Females 2001 
Ro, Re, 3, 7, 20 
81.7 
71.1 
76.4 
Males 2001 
3, 5, 6, 8, 9, 10 
69.7 
67.6 
67.8 
Females 2002 
P,Ro, 2,9, 13, 15, 16 
67.9 
70.8 
69.4 
Males 2002 
P, Re, 2, 8, 11, 13 
70.4 
61.2 
65.8 
Table 2 
Maximum likelihood estimates ( MLE ) of the contribution (%) of Atlantic stock king mackerel ( Scomberomorus cavalla ) to winter 
landings in each of three south Florida winter sampling zones by sex and year, with 90% bias-corrected confidence intervals (Cl) 
provided. The model column indicates which zone and year was estimated. 
Model 
MLE females 
90% Cl 
MLE males 
90% Cl 
Zone 1, 2001-02 
61.0 
32.2-82.7 
61.0 
40.2-73.9 
Zone 2, 2001-02 
48.6 
20.1-67.2 
99.9 
60.9-100.0 
Zone 3, 2001-02 
82.8 
62.8-99.8 
76.0 
57.0-99.7 
Zone 1, 2002-03 
14.5 
0.0-28.9 
45.0 
21.2-70.0 
Zone 2, 2002-03 
41.3 
20.9-68.9 
83.1 
49.4-100.0 
Zone 3, 2002-03 
40.4 
24.2-59.5 
71.9 
51.5-99.4 
from bootstrapped MLEs of Atlantic stock contribu- 
tion to south Florida winter king mackerel landings. 
Imprecision in those estimates prohibits more definitive 
conclusions about the relative contribution of GOM and 
Atlantic stocks to winter fisheries off south Florida. 
Nonetheless, it is possible to infer from our results that 
the Atlantic stock contributes substantially more than 
the zero percent of winter south Florida landings that is 
currently assumed by fishery management groups. 
Most of the otolith-shape differences between king 
mackerel stocks were observed in gross morphological 
variables and low-order Fourier harmonics. Low-order 
Fourier harmonics are related to general otolith shape, 
whereas high-order Fourier harmonics are related to 
increasingly fine-scale variation (Bird et al., 1986). 
DeVries et al. (2002) reported that gross otolith mor- 
phological parameters and low-order Fourier harmonics 
are significant contributors to otolith-shape variability 
in female king mackerel in southwest Florida, but they 
also reported many high-order Fourier harmonics to be 
significant as well. 
Sex effects on king mackerel otolith shape were sig- 
nificant for every gross morphological variable but for 
only one Fourier harmonic; this results indicates that 
sex-specific shape differences exist at a general level. 
Sex effects are not surprising given that sexually dimor- 
phic growth occurs in king mackerel; females achieve 
higher growth rates than males (Johnson et al., 1983; 
Manooch et al., 1987; Sturm and Salter, 1989; DeVries 
and Grimes, 1997). DeVries et al. (2002) examined only 
female king mackerel as a precaution against potential 
sex effects due to sexually dimorphic growth observed 
in this species. Most otolith shape studies that have 
tested for sex effects have found no significant differ- 
ences between males and females (Bird et al., 1986; 
Castonguay et al., 1991; Bolles and Begg, 2000; Begg et 
al., 2001). In studies where sex effects were significant, 
other factors were deemed more influential (Campana 
and Casselman, 1993). 
Otolith-shape variables in the models that best clas- 
sified king mackerel migratory groups were not con- 
sistent between sampling years. This result indicates 
that new shape-analysis models should be developed 
each summer and used only to estimate the migratory 
group composition of landings of the next winter. It is 
unclear why parameters in a discriminant function 
model may be important one year but of little value in 
distinguishing stocks the next year. However, interan- 
nual variability in growth rates between stocks may 
explain why LDFs do not perform well from one year to 
