474 
Fishery Bulletin 1 10(4) 
Table I 
Temporal distribution of sampling effort (number of events), loggerhead sea turtle (Caretta caretta ) captures (no. of Cc) and recap- 
tures (by recapture year) in the sea turtle trawl survey conducted in 2000-03, 2008-09, and 2011 in a coastal foraging region in 
the southeastern United States. 
Year 
Number of events 
Tagged 
no. of Cc 
Recaptured 
2000 
2001 
2002 
2003 
2008 
2009 
2011 
Total 
2000 
621 
172 
0 
0 
1 
4 
0 
0 
0 
5 
2001 
603 
177 
0 
0 
2 
1 
2 
0 
5 
2002 
684 
209 
1 
1 
0 
0 
2 
4 
2003 
714 
250 
0 
0 
0 
0 
0 
2008 
589 
167 
0 
0 
0 
0 
2009 
586 
152 
0 
1 
1 
2011 
410 
135 
1 
1 
Total 
4207 
1262 
0 
0 
2 
7 
1 
2 
4 
16 
Quantile residuals (Dunn and Smyth, 1996) were then 
plotted against each variable to assess trends and mod- 
el-assigned statistical significance of variables. Cumula- 
tive deviance attributed to all final model parameters 
was expressed as a percentage of null deviance to char- 
acterize the extent to which the final model accounted 
for catch variation. Linear regression was used to as- 
sess model fits (AIC vs. counts and SD) and annual 
mean modeled catch of loggerhead sea turtles. Confi- 
dence intervals ([Cl], 95%) around mean catch were 
computed with 7-statistics from Table B3 in Zar (1996). 
Results 
Sampling effort and catch distribution 
Random sampling (Moran’s lndex=0.00, Z-score = 0.67, 
P=0.501) was attempted for 4207 trawling events during 
7 sampling years between 2000 and 2011. Of 1262 cap- 
tured loggerhead sea turtles, 16 were captured twice 
during this survey (Table 1) up to 9 years later (mean 
±SD=3.9 ±3.2 years). Two loggerhead sea turtles tagged 
during this study were recaptured by other programs 
(after 0.3 and 1.9 years), and 14 loggerhead sea turtles 
tagged by other programs (3.4 ±3.4 years earlier) also 
were captured in this study. No loggerhead sea turtles 
died during this study, and only 3 turtles required 
resuscitation. Six loggerhead sea turtles tagged and 
released in this study were subsequently (3.0 ±1.8 years 
later) reported as stranded dead, 5 of them on beaches 
adjacent to the survey area and the sixth one in the 
northern Gulf of Mexico. 
Sampling effort was significantly different among 
years and subregions (^ 2 18 =455, P<0.001). Greatest 
annual sampling effort occurred in 2002-03 (343-355 
h; 18-19% of total) followed by 2000-01 (300-309 h; 
16% of total) and then 2008-11 (194-204 h; 10-11% 
of total). The greatest amount of sampling effort was 
expended between Brunswick and Savannah, Georgia, 
(533 h; 28% of total), followed by Savannah, Georgia, 
to Charleston, South Carolina, (501 h; 26% of total), 
St. Augustine, Florida, to Brunswick, Georgia, (465 h; 
24% of total), and Charleston to Winyah Bay, South 
Carolina, (404 h; 22% of total). 
Loggerhead sea turtles were captured in 23% (951) 
of sampling events, with up to 7 individual loggerhead 
sea turtles captured in one sampling event. A single log- 
gerhead sea turtle was the most common positive catch 
observed and accounted for 20% of sampling events in 
2011 (83 of 410) to 15% of sampling events in 2009 (87 
of 586); however, the ratio of zero- to single-catch events 
was not significantly different among years (;t 2 6 =7.5, 
P=0.275). Double catch of loggerhead sea turtles oc- 
curred in 4% (185) of sampling events and was not 
significantly different among years (^ 2 6 =3.1, P=0.800). 
Three or more loggerhead sea turtles were captured 
only in 1% (46) of attempted sampling events between 
2000 and 2011. 
Model fits 
Catch rate trends were analyzed for 1227 loggerhead sea 
turtles captured in 4097 sampling events with complete 
effort and companion data. Final model AIC scores 
ranged from 951.0 (75. 1-80. 0cm SCLmin) to 5550.1 
(overall; Table 2). The count of sampling events with 
zero turtle catches, both overall and for each of 5 preva- 
lent 5-cm size classes, was significantly and positively 
associated with final model AIC (F 14 =1123.6, r 2 =1.00, 
P<0.001) and SD (F M = 1010.5, r 2 =L00, P<0.001). The 
overall final model explained 8.4% of deviance in the 
overall data set, with between 6.6% and 14.9% of devi- 
ance explained for prevalent 5-cm size classes (Table 2). 
In the final overall model, 12 of 25 parameters (48%) 
were retained, with between 6 (24%) and 12 (48%) 
parameters retained in the final model for prevalent 
5-cm size classes (Table 3). Among parameters retained 
in final models, only 50% (3 of 6) to 67% (8 of 12) were 
deemed significant (Table 3). 
