Arendt et a!.: Temporal trends and influences on fishery-independent catch rates for Coretta caretta in an important coastal foraging region 
475 
Table 2 
Assessment of generalized linear model fits for loggerhead sea turtle ( Caretta caretta) catch overall and with respect to preva- 
lent 5-cm size classes of minimum straight-line carapace length. Best model fits (12.2-14.9% of model deviance explained) were 
associated with the smallest and largest size classes examined, with half as much model deviance generally explained for inter- 
mediate size classes. 
Model metric 
Overall 
55.1-60.0 
60.1-65.0 
65.1-70.0 
70.1-75.0 
75.1-80.0 
AIC score, null model 
5573.2 
1080.7 
2033.2 
2285.8 
1831.1 
966.0 
AIC score, final model 
5550.1 
1052.1 
2004.0 
2258.7 
1805.5 
951.0 
Null model deviance 
3080.8 
793.6 
1407.1 
1514.7 
1325.4 
757.4 
Final model deviance 
2822.3 
697.2 
1248.8 
1403.3 
1238.2 
644.4 
Percentage of deviance explained 
8.4 
12.2 
11.3 
7.4 
6.6 
14.9 
Winyah Bay, SC to Charleston, SC to Savannah, GA to Brunswick, GA to St. 
Charleston, SC Savannah, GA Brunswick, GA Augustine, FL 
Figure 2 
Spatial variability in modeled catch (mean ±95% Cl) of loggerhead sea 
turtles ( Caretta caretta) per linear kilometer for 4097 coastal trawling 
events conducted between St. Augustine, Florida, (29.9°N) and Winyah 
Bay, South Carolina, (33.1°N) in 2000-03, 2008-09, and 2011. The 
black line denotes overall catch rates. Bars denote the following 5-cm 
size classes of minimum straight-line carapace length: 55.1-60.0 (light 
gray); 60.1-65.0 (dark gray); 65.1-70.0 (white); 70.1-75.0 (medium gray); 
75.1-80.0 (charcoal). 
Catch rate influences 
Geographic subregion, distance from 
shore, and sampling year were the 
only parameters retained as signifi- 
cant terms in all final models (Table 
3). Geographic subregion was the most 
important parameter overall (3.9% of 
deviance) but was the most important 
observed influence on catch rates for 
loggerhead sea turtles that measured 
<70. 0cm SCLmin, where it accounted for 
3. 2-7. 6% of data set deviance (Table 3). 
Catch increased significantly (F l 2 =27.3, 
r 2 =0.90, P=0.035; Fig. 2) between the 
subregion of Winyah Bay to Charleston, 
South Carolina, (mean ±95% CI = 0.174 
±0.003 turtles per km; CV=0.24) and the 
subregion of Brunswick, Georgia, to St. 
Augustine, Florida, (0.468 ±0.013 turtles 
per km; CV=0.44). 
Distance from shore accounted for 
1.5% of data set deviance overall, but 
between 0.1% and 2.5% of data set de- 
viance among 5-cm size classes (Table 
3). Catch rates decreased systematically 
with distance from shore (Fig. 3), with 
overall trends driven largely by logger- 
head sea turtles that measured 60.1- 
70.0 cm SCLmin and captured within 5 
km from shore. 
Sampling year explained 0.8% of 
data set deviance overall and between 
1.0-6. 1% of data set deviance among 5-cm size classes 
(Table 3). Annual catch rates (sea turtles per linear 
kilometer) ranged from 0.256 ±0.014 (mean ±95%CI) 
in 2009 to 0.356 ±0.019 in 2003, but rates were not 
significantly different among years (F 15 = 0.0, r 2 =0.00, 
P=0.944). Interannual differences in mean catch rates 
for loggerhead sea turtles that measured 55.1-75.0 cm 
SCLmin (Fig. 4, A and B) were not significantly differ- 
ent (F 15 =0. 5-4.1, r 2 =0.00-0.34, P=0.098-0.500). Catch 
rates for loggerhead sea turtles 75.1-80.0 cm SCLmin 
(Fig. 4B) increased significantly (F 15 =24.1, r 2 = 0.79, 
P=0.004) between 2000 (4, 2% of captures; CV=0.86) 
and 2011 (25, 19% of captures; CV=0.92). 
Four terms were retained only as significant terms 
for a subset of all final models. The interaction between 
mean water depth and distance from the closest inlet 
was retained as a significant model term in all final 
models, except for the smallest (55.1-60.0 cm SCL- 
min) and largest (75.1-80.0 cm SCLmin) size classes 
evaluated, but accounted for <0.7% of data set deviance 
(Table 3). Time of day and the interaction between sam- 
pling year and the NAO index was retained as signifi- 
