Begg et at: Stock discrimination of Scomberomorus queenslandicus and 5. munroi 
659 
Canonical variate I 
Bowen-2 
Moreton Bay-2 
Rockhampton-1 
Rockhampton-2 
Figure 5 
Discrimination between school mackerel samples based on the concentrations 
of 7 trace elements (± 95% confidence ellipses around the sample mean for 
each area). 
from Rockhampton and Moreton Bay. 
The samples of 1-year-old fish from 
Rockhampton had PC I scores that 
were significantly different from all 
the 2-year-old samples, and PC II 
scores that also were significantly dif- 
ferent from the other samples, but 
not from Moreton Bay (HSD, P<0.05). 
Analyses indicated that school 
mackerel samples are best separated 
into three groups: 1) Bowen 2-year- 
old fish; 2) Moreton Bay and Rock- 
hampton 2-year-old fish; and 3) 
Rockhampton 1-year-old fish (Wilks’s 
lambda=0.195) (Fig. 5). Similar dis- 
criminant patterns were observed 
when the “significant” elements (K, 
Mg, Mn, Na, P; Wilks’s lambda=0.214) 
were used in isolation. Significant dif- 
ferences in the discriminant scores 
were found between these groups for 
both the first canonical variate (CV I) 
( ANOVA, F= 74. 10, df=3, 80, PcO.OOOl) 
and the second (CV II) (ANOVA, F=6.61, df=3, 80, 
P<0.0001). Like the results of the PCA analyses, Bowen 
samples had significantly different CV I scores than 
those from Rockhampton and Moreton Bay; no differ- 
ences were found between Rockhampton and Moreton 
Bay 2-year-old fish; and Rockhampton 1-year-old 
samples had CV I scores significantly different from 
all the other groups. The Rockhampton 1-year-old 
samples also had CV II scores that were significantly 
different from the other groups, with the exception of 
fish from Bowen (HSD, P< 0.05). 
Variation in Na, P, and Mg concentrations were 
primarily responsible for the separations apparent 
among the samples (89%) according to the first ca- 
nonical variate (Table 2). Bowen fish tended to have 
lower Na and Mg and higher P concentrations than 
Moreton Bay and Rockhampton samples (Fig. 2). 
Approximately 62% of the school mackerel samples 
were classified into their correct groupings, indicating 
an overlap in otolith composition of Rockhampton and 
Moreton Bay samples (Table 3). Overall classification 
success increased to 77% when these areas were pooled, 
and individual classification rates improved for Bowen 
2-yr-old fish (86%), Moreton Bay and Rockhampton 
2-year-old fish (70%), and Rockhampton 1-year-old fish 
(80%), providing further support for the notion of three 
distinct groups in elemental composition of otoliths. 
Spotted mackerel 
The same suite of 11 elements that were used for 
school mackerel were measured and analyzed for 
Table 2 
Discrimination between samples of school mackerel deter- 
mined by the pooled within-group correlations of elemen- 
tal concentrations with the significant (P< 0.05) canonical 
variates I and II, and the cumulative proportion of the ex- 
plained variance accounted for by each function for seven 
elements (Wilks’s lambda=0.195). 
Canonical variate 
Element 
I 
II 
Na 
0.53 
0.07 
P 
-0.48 
0.02 
Mn 
0.17 
-0.72 
K 
-0.20 
-0.40 
Mg 
0.42 
-0.38 
S 
-0.13 
0.24 
Sr 
-0.13 
-0.11 
Cumulative proportion 
0.89 
0.97 
spotted mackerel (Fig. 6). Iron, Li, and Ca were ex- 
cluded from statistical analyses for the same reasons 
as they were for school mackerel. Concentrations of 
Na (ANCOVA, F=14.02, df=l, 141, P<0.0003; r= 
-0.0004597) and Sr (ANCOVA, F= 6.03, df=l, 141, 
P< 0.02; r=-0.9227) in spotted mackerel otoliths were 
highly correlated with fish length; therefore the data 
were corrected for length with the respective regres- 
sion coefficient for the length covariate. Manganese 
was not used in statistical analyses because a sig- 
nificant interaction in concentrations existed be- 
