Arendt et al.: Catch rates and demographics of Caretta carettci captured from the Charleston, South Carolina, shipping channel 
101 
sapidus), horseshoe crabs (. Limulus polyphemus), mis- 
cellaneous crabs, cannonball jellyfish ( Stomolophus me- 
leagris), and miscellaneous jellyfish. Loggerheads will 
consume finfish; however, such occurrences of finfish are 
thought to be dead fishery discards (Seney and Musick, 
2007), and were excluded from multivariate analyses. 
Owing to the large-mesh webbing and streamlined body 
designs of finfish, we also suspected less efficient fin- 
fish capture relative to similar-size invertebrates that 
became entangled in or otherwise clung to the trawl 
webbing. 
Data analyses 
Loggerhead catch during 2004-07 was analyzed using a 
generalized linear model (GLM) with log-link function, 
with the log of the standardized sampling effort for each 
trawling event treated as an offset variable. Sampling 
effort was standardized to a net head rope length of 
30.5 m calculated as follows: [2 netsx(18.3 m head rope)/ 
30.5 m]x[(tom time, min)/60]. Loggerhead catch per 
trawling event best fitted the negative binomial distri- 
bution despite a significant P-value (^ 2 =17.346, df=7, 
P= 0.015) which resulted from infrequent capture of 
three or more loggerheads per trawling event. 
Final model selection was accomplished in R software 
(vers. 2.10.1; R Core Team, Vienna, Austria) through 
backward elimination stepwise regression (a=0.05) 
that generated the lowest Akaike’s information crite- 
rion (AIC) score. With chi-square analysis of deviance, 
we assessed the statistical significance of variables 
retained in the final model. Quantile residuals (Dunn 
and Smyth, 1996) were plotted against each variable 
to assess trends and model-assigned statistical signifi- 
cance of variables. Cumulative deviance attributed to 
final model variables was expressed as a percentage of 
the null deviance to characterize the extent to which 
the final model accounted for variation in catch in the 
data set. The adjusted loggerhead counts (mean ±95% 
confidence interval [Cl] ) per trawling event were used 
to examine catch rate trends among years and among 
blocks and size classes by year. 
Twenty-one terms included in the null model con- 
sisted of hydrographic and meteorological variables (9), 
vessel towing speed, prey item groupings (5), sampling 
period (factor, 1 to 8), sampling block (factor, 1 to 3), 
hour of day, and three interaction (Pearson correlation 
coefficient r>0.4) terms between 1) barometric pres- 
sure and sampling period, 2) blue crabs and water 
temperature, and 3) miscellaneous jellyfish and water 
temperature. Twelve trawling events that were con- 
ducted at stations sampled only in May 2004 and 11 
trawling events that were terminated early because of 
net hang ups or interference were not analyzed. Five 
stations missing vessel towing speed data were also 
excluded from the GLM. The wind direction for 38 
trawling events with calm winds was assigned as the 
prevalent wind direction during trawling events im- 
mediately before or after (whichever was more robust) 
winds became calm. Cloud cover for five events and 
wave height for one event were populated by using the 
same approach. 
Standardized effort enabled comparison of catch rates 
between this study and two historical data sets, one 
employing the same trawl gear as the current study 
(Dickerson et al. 1 ) and another using 18-m mongoose- 
style nets with 10-cm stretch mesh webbing (Van Dolah 
and Maier, 1993). Effort and catch for daytime only 
trawling in 1991 were obtained from Van Dolah et al. 2 
A negative binomial GLM with log-link function was 
used to compare loggerhead catch between study periods 
(1991-92 vs. 2004-07) with year and month as factors 
and the log of the sampling effort as an offset variable. 
Data for May were available in all years; however, data 
for August were absent in 1992 and 2006 and data for 
June were only available in 1991 and 2004. 
Straight-line carapace length (nuchal notch to post- 
marginal scutes, SCLnt) was compared between 2004- 
07 and 1991-92 (Dickerson et al. 1 ; Van Dolah et al. 2 ). 
Size values were not normally distributed; therefore, 
data grouped by 10-cm size classes were analyzed 
with Kruskal-Wallis analysis of variance by ranks and 
Dunn-Bonferroni pairwise comparisons (Minitab 15®; 
Minitab, Inc., State College, PA). Sex and mtDNA data 
were evaluated by using chi-square analysis (Minitab 
15®) to test for annual differences in the ratio of fe- 
males to males and variations in haplotype frequencies 
between groups of interest. Owing to a high probability 
of error for determing the sex of pubescent loggerheads 
based on hormone levels alone, sex was not assigned for 
loggerheads from 75.1 to 85.0 cm SCLnt. 
Results 
Catch and recapture data 
From the 432 trawling events conducted in the Charles- 
ton shipping channel between May 2004 and August 
2007, 220 loggerhead sea turtles were captured (Table 1). 
Eight of 220 loggerheads (3.6%) were recaptured during 
the survey of which four were recaptured during the 
same cruise, one was recaptured during the same 
season, and three were recaptured in subsequent years 
257, 453, and 705 days later. Two loggerheads captured 
by trawling <5 km from the Charleston shipping channel 
in 2001 by the South Carolina Department of Natural 
Resources (SCDNR) were recaptured in this channel 
1066 and 1396 days after initial tag and release. Only 
two loggerheads tagged during this survey were reported 
as recaptured away from the channel: a 95.4-cm SCLnt 
female captured in May 2006 nested on Cumberland 
2 Van Dolah, R. F., P. P. Maier, S. R. Hopkins-Murphy, G. F. 
Ulrich, and D. M. Cupka. 1992. A survey of turtle popula- 
tions in the Charleston Harbor entrance channel. SC Dept 
Natural Resources, Charleston, SC Final Report #14-16-0004- 
90-944 to USFWS. [Available from http://dnr.sc.gov/marine/ 
turtles/Literature/Van%20Dolah%20CNHB%20Channel.pdf, 
accessed June 2011.] 
