Dippold et al.: Growth, mortality, and movement of Rachycentron canadum 
465 
1500 
1000 
500 
ATL 
□ FLGC 
FLK 
FLPH 
LA 
NcGOM 
TX 
100 
>. 
o 
c 
0) 
3 
CT 
d) 
50 
25 
ooa50T-c\jcO'tfLntDi"coGRO-<-c\jco'3-Lncoi'-.oocr>Oi-cMco''3- 
coooroo)(JiO)a)(J)fflffiO)(j)ooooooooooi-t-t-T-r 
0 ) 5 ) 0 ) 0 ) 0 ) 0 ) 0 ) 0 ) 0 ) 0 ) 0 ) 0)00000000000000 0 
■^•r-T--r-T-T-T-T-'r-T-T-T-C\JC\JC^OOC\JC\JC\JCVCMC\JC\JCMC\JCNJC\J 
ATL 
FLGC 
FLK 
FLPH 
LA 
NcGOM 
TX 
i-wcoTintoscoo) 
Ot-NCITiPONCDOIO 
oooooooooo 
0500 ) 0503050 ) 0)000000000000000 
t—t— v— t t— t— t— CMC\| C \| C MC\i<M C MOj C 'J C \l C M C MC\jCM C \J 
Year 
Figure 3 
Total number of cobia (Rachycentron canadum) (A) tagged and 
(B) recaptured by year in each of 7 geographic zones in the Gulf 
of Mexico and South Atlantic Ocean during 1988-2014. The 7 
defined zones are the South Atlantic Ocean (ATL),Florida Gulf 
Coast (FLGC), Florida Keys (FLK), Florida panhandle (FLPH), 
Louisiana (LA), northcentral Gulf of Mexico (NcGOM), and Tex¬ 
as (TX). 
vorship and tag-recovery (4 candidate models). Time- 
dependent parameters were those that varied by year 
and time-independent parameters were those that 
were constant through the duration of the tagging pro¬ 
gram. The global model (fully parameterized) was de¬ 
fined as having time-dependent survivorship, Sit), and 
time-dependent tag-recovery, fit). Global model fit was 
evaluated by using the constant noted as c, which is 
an estimate of dispersion used in Program MARK. To 
estimate dispersion, a simulation procedure is used in 
which data are generated at varying levels of c and a 
logistic model is then fitted to estimate c for the global 
model. A c value of less than 3.0 indicates adequate 
model fit. After the suite of candidate models were 
fitted to the tag-recapture data and the global mod¬ 
el was determined to adequately fit the data, model 
support was evaluated by using AIC (Burn¬ 
ham and Anderson, 2002). The model with the 
greatest support (lowest AIC value) was used 
to estimate mean annual Z. The estimates of 
Z derived in this study were then compared 
with the mortality values reported in both the 
Gulf of Mexico and South Atlantic Ocean cobia 
stock assessments (SEDAR 1 ) by converting the 
value of S estimated in this study to an esti¬ 
mate of Z with the equation 
S - e- z . (7) 
Results 
Tagging program 
A total of 17,875 cobia were tagged from 1988 
to 2014. The number of individuals tagged an¬ 
nually ranged from 113 to 1423 individuals. 
A majority (57%) of tagging occurred between 
1990 and 1998 (Fig. 3). The reported length 
of tagged individuals ranged from 178 mm to 
1549 mm FL (Fig. 4A). A total of 1137 indi¬ 
viduals were recaptured, and the number of 
recaptured individuals annually varied from 3 
to 94 individuals. The annual number of in¬ 
dividuals recaptured was greatest from 1990 
to 1998 (Fig. 3). The reported length of recap¬ 
tured individuals ranged from 305 to 1448 mm 
FL (Fig. 4B), and the time between tagging 
and recapture ranged from 1 to 2973 days at 
large (Fig. 4C). Of the 7 zones defined in this 
study, the northcentral Gulf of Mexico zone 
had the greatest number of tagged and recap¬ 
tured individuals and the Texas zone had the 
fewest number of tagged and recaptured indi¬ 
viduals (Table 1). 
Growth 
Only individuals for which lengths at tagging 
and recapture were recorded were used in the 
growth analysis (n=926). All reported lengths 
were converted to FL using the linear relationship be¬ 
tween TLs and FLs (FL=0.91TL+0.23; coefficient of de¬ 
termination [r 2 ]=0.98) developed from lengths reported 
in this study. Sex of tagged and recaptured individuals 
was not reported in this cooperative tagging program 
and therefore we modeled sex-combined length-at-age. 
The 3 nonlinear length-at-models were fitted, and the 
relative model support was evaluated with AIC (Table 
2). Of the 3 candidate models, the GROTAG VBGF was 
best supported on the basis of calculated values of w ; 
(~ 1.0). However, on the basis of the mean and 95% 
CIs of the L„ parameter, there was no difference in 
the mean estimates of L„ among the 3 candidate mod¬ 
els. The mean value of k did vary (on the basis of 95% 
CIs) for each of the 3 candidate models. Specifically, 
the mean value of k estimated in the GROTAG VBGF 
