154 
Fishery Bulletin 106(2) 
off south Florida from December 2001 to March 2002 
and December 2002 to March 2003 (Fig. 2). Zone 1 
represented southwest Florida and primarily consisted 
of samples from the commercial gillnet fishery near 
and to the east of the Dry Tortugas. Zone 2 represented 
south central Florida and consisted of samples from 
the recreational charter boat fishery operating south 
of Islamorada in the Florida Keys. Zone 3 represented 
southeast Florida and primarily consisted of samples 
from the commercial troll fishery from Sebastian Inlet 
to south of West Palm Beach, Florida. Collection and 
aging procedures for winter fish otoliths followed the 
same protocol as summer sampling. 
Left sagittal otoliths were digitized sulcus side down 
with an image analysis system running Image-Pro 
image analysis software (vers. 4.5, Media Cybernetics 
Inc., Bethesda, MD). Otolith samples were magnified 
by 13 x with a dissecting microscope before their im- 
ages were captured with the image analysis system. 
When left otoliths were damaged or unavailable, right 
otoliths were digitized and their mirror images were 
used for shape analysis (DeVries et al., 2002). The auto- 
trace feature in Image Pro then was used to trace the 
posterior surface of the otolith. Otolith tracing began 
at the tip of the antirostrum, was directed manually 
across the base of the rostrum, and then the software 
traced the posterior portion of the otolith. Thus, rostra 
were excluded from otolith-shape analysis because the 
anterior rostrum is fragile and often was broken during 
otolith collection (DeVries et ah, 2002). 
Fourier coefficients were computed with an algo- 
rithm within Image-Pro, and we used the mathemati- 
cally determined centroid as the center of an otolith. 
The Image-Pro algorithm used 128 vectors at equally 
spaced polar angles to create an accurate picture of the 
otolith outline. The amplitudes of the first 20 Fourier 
harmonics were calculated for analysis because each 
additional harmonic provides increasingly finer detail 
of the otolith outline. For example, 97-99% of otolith- 
shape variability in haddock ( Melanogrammus aeglefi- 
nus) is contained in the first ten harmonics (Begg and 
Brown, 2000). Fourier amplitudes were standardized 
to remove the effect of otolith size by dividing each 
amplitude by the mean radial length of the otolith. In 
addition to the first 20 standardized Fourier harmon- 
ics, the Image-Pro software calculated otolith area, 
perimeter, rectangularity, circularity, and roundness 
for a total of 25 shape variables. All variables were 
tested for univariate normality with the Shapiro-Wilks 
statistic and for homogeneity of variance with an F max 
test. Transformations were necessary for perimeter 
(natural log) and Fourier harmonics 13-16 (square- 
root) in order to meet parametric statisti- 
cal analysis assumptions of normality and 
homogeneity of variances. 
Ontogenetic effects on otolith shape were 
tested by computing the correlations of 
shape variables with fish length. Ontoge- 
netic effects were removed from each shape 
variable that was significantly correlated 
with fish length by subtracting the slope 
of the least squares linear relationship be- 
tween length and a given variable. Slope- 
corrected data were used in all subsequent 
analyses. 
Multivariate analysis of variance (MAN- 
OVA) was performed to test for potential 
shape differences between sides in a subset 
of 50 left and right sagittal otolith pairs 
(SAS, vers. 6.11, SAS Inst., Inc., Cary, 
NC). A second MANOVA also was per- 
formed to test for stock-specific differences 
in summer samples. The effect of other 
factors, including sex, age, and sampling 
year, on otolith shape parameters also was 
tested within this second MANOVA. 
Stepwise linear discriminant function 
(LDF) analysis was performed separately 
for sexes and years on otolith-shape vari- 
ables from summer sampled fish with the 
PROC STEPDISC procedure in SAS. The 
LDF procedure selected variables that 
were effective predictors of stock iden- 
tity. Jackknife cross-validation was used 
to evaluate the performance of resultant 
discriminant functions. Classification suc- 
90°W 80°W 
Figure 2 
Map of sampling locations for king mackerel ( Scomberomorous 
cavalla) in summers 2001 and 2002 in U.S. Atlantic Ocean 
waters (squares) and the Gulf of Mexico (circles). The map 
also shows the three winter sampling zones around southern 
Florida from which fish were sampled in winter 2001-02 and 
2002-03 for estimates of the Atlantic stock contribution to 
winter landings. 
