480 
Fishery Bulletin 1 10(4) 
tion abundance or simply from aggregation of turtles in 
an area at the time of sampling. Nevertheless, because 
in-water studies can provide more temporally refined 
indications of pending nesting recruitment than nest- 
ing beach surveys alone, we encourage further in-water 
studies. As evidenced by this study and those published 
by Epperly et al. (2007), Pons et al. (2010), and Mur- 
ray (2011), even zero-rich data sets (for which, cur- 
rently, there are no alternatives for in-water sea turtle 
data), when analyzed with the appropriate statistical 
technique (Maunder and Punt, 2004), are valuable for 
assessments of temporal trends. This approach implies, 
of course, that the trend attributes can be adequately 
explained and that high zero catches do not reflect poor 
survey design. 
Despite consistency in model terms affiliated with 
model deviance, very little model deviance was actu- 
ally explained. The modest amount of model deviance 
explained likely stemmed from random sampling con- 
ducted in an open marine system where juvenile log- 
gerhead sea turtles may not be randomly distributed. 
Specifically, nonrandom distribution is suggested by 
site fidelity documented by telemetry studies in estua- 
rine (Byles, 1988; Morreale, 1999; Avens et al., 2003; 
Mansfield et al., 2009) and coastal (Renaud and Car- 
penter, 1994; Arendt et al., 2012b) habitats. Habitat 
preferences would seem a likely explanation for non- 
random distributions, particularly given the suggestion 
by Hopkins-Murphy et al. (2003) that loggerhead sea 
turtles associate with dense, live-bottom habitats. Un- 
fortunately, dense, live-bottom habitats are not condu- 
cive to trawling operations and, as such, these habitats 
were avoided where possible in this study. Furthermore, 
use of large-mesh trawl webbing and mud rollers on the 
trawl foot rope was included to minimize the collection 
of sponges and gorgonians whose collection is needed to 
distinguish probably hard and hard habitats from not 
hard habitats. Nevertheless, nearly half of sampling 
events occurred where habitats were characterized as 
probably hard or hard, and, in contrast to the sugges- 
tion by Hopkins-Murphy et al. (2003), catch rates from 
these sampling events were significantly more reduced 
than rates from not hard habitats. Seafloor type was 
either excluded or retained as a nonsignificant term in 
half of the final models. However, in the absence of data 
on gear efficiency and performance in these different 
habitats, it is not possible to rule out the importance 
of habitat features on spatial distribution patterns at 
this time. 
Four model parameters (geographic subregion, dis- 
tance from shore, seafloor type, and year) consistently 
accounted for at least two-thirds (and upwards to 97%) 
of explained model deviance, and different contribu- 
tions were associated with each parameter among the 
various size classes examined. In pelagic habitats, en- 
vironmental parameters, such as temperature, Chl-a, 
and mesoscale eddies, influence the spatial distribu- 
tion of loggerhead sea turtles (Mansfield et al., 2009; 
Kobayashi et al., 2011). However, SST and Chl-a each 
were retained only as a nonsignificant term in just one 
final model in this study. This observation is in line with 
localized distribution patterns reported by Arendt et al. 
(2012b), an outcome that would have been expected to 
be more variable if spatial distributions fluctuated in re- 
sponse to fine-scale hydrographic changes. Hydrographic 
conditions can create density gradients that are known 
to greatly influence loggerhead sea turtle distributions 
in the winter (Epperly et al., 1995b) but are less likely 
to occur during the 2 months surrounding the summer 
solstice, as well as where this study was conducted. As 
noted by Atkinson et al. (1983), “The large heat capacity 
of water insures a highly damped response to daily air 
temperature cycles, but cycles at seasonal and inter-an- 
nual time scales have a large effect. Similar arguments 
apply to the inner shelf salinity field, which is controlled 
by seasonal and inter-annual cycles of river discharge.” 
Accordingly, when coupled with prevalent southwesterly 
winds, excessive freshwater runoff in spring 2003 set up 
a coastwide cold-water upwelling event (see discussion in 
Maier et al. 4 ). Concurrent with altered circulation pat- 
terns, the greatest number of captures (and recaptures) 
of loggerhead sea turtles occurred in 2003. 
Conclusions 
This study is the first to report on coastal loggerhead 
sea turtle catch rates in a large and central portion of 
one of the most important foraging grounds for this 
species in the NW Atlantic basin (Bowen et al., 2004). 
Our inability to detect a significant trend among annual 
catch likely was the result of the short duration of our 
study relative to the life history of this species, and 
simultaneous increases in variance concurrent with 
increased catch rates. Stable to increasing catch for 
loggerhead sea turtles that corresponds with matur- 
ing or mature individuals is encouraging for continued 
recovery of this threatened species in the NW Atlantic, 
a population that fares better than most populations of 
this globally distributed species (Wallace et al., 2010b). 
Regionally, the data presented herein begin to address 
one of 3 demographic recovery criteria that stipulate that 
increases in the in-water abundance of juvenile sea tur- 
tles must occur throughout a network of monitoring sites 
for at least one generation (NMFS and USFWS, 2008). 
Analysis of trawl data previously has received mixed 
reviews (Bjorndal and Bolten, 2000), primarily because 
of nuances specific to data sets (e.g., a single fishery-de- 
pendent data set collected at a single location after per- 
ceived historic stock decimation). The randomized sam- 
pling design employed in this study minimized temporal 
and spatial bias and maximized temporal and spatial 
coverage. Randomized sampling design increased the 
probability that observed catch was proportional to ac- 
tual abundance, rather than hyperstable, which could 
have resulted from intensely sampling areas of high 
abundance (Hilborn and Walters, 1992). Given that a 
wide range of sea turtle sizes were captured, we also 
do not suspect that data reported herein represent a 
hyperdepleted scenario, where only a portion of the pop- 
