Northeast Fisheries Science Center Reference Document 04-11
Bycatch of Sea Turtles
in the Mid-Atlantic Sea Scallop
(Placopecten magellanicus)
Dredge Fishery during 2003
(2nd Edition)
by Kimberly T. Murray
National Marine Fisheries Serv., Woods Hole Lab., 166 Water St., Woods
Hole, MA 02543
Print
publication date October 2004;
web version posted November 9, 2004
Citation: Murray, K.T. 2004. Bycatch of sea turtles in the Mid-Atlantic sea scallop (Placopecten
magellanicus) dredge fishery during 2003. 2nd ed. U.S. Dep. Commer., Northeast Fish. Sci.
Cent. Ref. Doc. 04-11; 25 p.
Download complete PDF/print version
PREFACE
The NEFSC has re-estimated takes of loggerhead sea turtles in the
Mid-Atlantic sea
scallop fishery during 1 June to 30 November 2003, first estimated
in the Northeast
Fisheries Science Center Reference Document 04-11, issued in August
2004. This
revision of the document uses improved data about the location of trips
used in the
analysis.
The resulting estimate changed by 119 animals, from 630 to
749. The August bycatch
analysis used actual location data from only those trip reports with
recorded lat./long.
locations, and prorated the remaining trips (about 32% of the trips
used) to areas under
analysis. The revised analysis uses trip reports with good trip location
information
entered either as lat./long. or as loran coordinates (only 1.5% of
trips are prorated).
Trips and takes that had been previously prorated to areas
proved to be distributed differently than was assumed in the first analysis.
The analytical method used is unchanged, as are the resulting bycatch
rates and the coefficient of variance (CV=0.28) for the estimates. The
conclusions of the analysis, that sea surface temperature and tow duration/speed
have more influence on the likelihood of loggerhead takes than gear configuration
or other observed fishery practices, are also unchanged.
Previous edition available in PDF here.
ABSTRACT
During 2003, fisheries observers
aboard commercial vessels in the U.S. sea scallop dredge fishery documented
sea turtle bycatch in the Mid-Atlantic region, from Long Island, New
York to Cape Hatteras, North Carolina. This study utilizes
Generalized Linear Model and Generalized Additive Model fitting techniques
to identify environmental factors and gear characteristics that influence
bycatch rates of sea turtles in the 2003 Mid-Atlantic scallop dredge
fishery, and predicts total bycatch in this area during June-November
2003. Significant factors affecting sea turtle bycatch were sea surface
temperature and tow speed. In estimating total bycatch, rates were stratified
based on sea surface temperature because tow speed was not available
in fishing logbooks. Highest bycatch rates occurred in surface temperatures
greater or equal to 22°C. Total estimated bycatch of sea turtles
during 1 June to 30 November 2003 in the Mid-Atlantic scallop dredge
fishery was 749 animals (C.V. = 0.28).
INTRODUCTION
During 2001 and 2002, fisheries
observers aboard commercial vessels in the sea scallop dredge fishery
documented sea turtle bycatch in two regions of the Mid-Atlantic (Murray
2004). These areas, termed the 'Hudson Canyon Controlled Access Area'
and the 'Virginia Beach Controlled Access Area' were closed in April
1998 to allow juvenile scallops to recover and reopened in May 2001
on a conditional basis. Observers sampled approximately 11% of the
commercial dredge effort in the Hudson Canyon Access Area during 2001
and 2002, and 16% of the effort in the Virginia Beach Access Area during
2001. No trips were observed in the Virginia Beach Access Area during
2002 due to low commercial fishing effort in the area. Outside of these
two areas, observer coverage was less than 1%. Over both years, observers
documented 27 sea turtle interactions in the Controlled Access Areas
(CAAs). Estimated bycatch in the Hudson Canyon Area was 69 turtles
in 2001, and 95 turtles in 2002. Estimated bycatch in the Virginia
Beach Access Area was 5 turtles in 2001, and 0 turtles in 2002.
The spatial extent of observer
coverage in the commercial scallop dredge fishery expanded in 2003.
This increase in spatial coverage was needed to properly assess bycatch
outside of the CAAs. Bycatch in these two areas was not estimated in
2001-2002 because of scientific concerns that bycatch rates differed
throughout the broader Mid-Atlantic based on environmental factors,
fishing practices, and dredge gear characteristics. Accounting for
differences in bycatch rates based on these factors reduces bias in
the total bycatch estimate.
The purpose of this paper
is to identify factors that influenced the bycatch rate of turtles
in the Mid-Atlantic sea scallop dredge fishery in 2003, and to provide
a total bycatch estimate for this fishery. This bycatch assessment
is based on new and expanded information about the fishery in the Mid-Atlantic,
taking into consideration fishing practices, environmental variables,
and gear characteristics both inside and outside of the Controlled
Access Areas.
METHODS
Spatial and Temporal Boundaries used for Bycatch Rate Extrapolation
The 2003 fishing year for
the U.S. commercial sea scallop fishery occurred from 1 March 2003
to 28 February 2004 (NEFMC 2002) in the Gulf of Maine, Georges Bank,
and Mid-Atlantic regions. This analysis deals solely with trips operating
in the Mid-Atlantic region from Long Island, NY to Cape Hatteras, NC
(approximately 41°09'N/71°00'W to 35°15'N/71°00'W),
from 1 June to 30 November 2003.
It is assumed here that
there are no turtle interactions in the scallop fishery in the Georges
Bank and Gulf of Maine regions; as such, commercial fishing effort
north of Long Island, NY was excluded from the bycatch analysis. Compared
to the Mid-Atlantic fishery, the Georges Bank and Gulf of Maine scallop
fisheries operate in different ecological conditions. These fisheries
operate north of the general range of loggerhead turtles (~ 41°N
latitude as a northern limit; Shoop and Kenney 1992). There has never
been an observed turtle interaction in the Georges Bank and Gulf of
Maine areas; however, observer coverage (% trips observed) in these
areas has been low (< 1.0% in 2001, 2002 and 2003).
A bycatch estimate is provided
for the period June to November 2003, rather than throughout the entire
fishing year, because outside of this time frame turtle interactions
with the sea scallop dredge fishery are unlikely to occur. Loggerhead
sea turtles inhabit northern temperate waters seasonally, appearing
in early summer and remaining for several months until migrating south
in the fall (Morreale 1999). By December, turtles are near or south
of Cape Hatteras, NC (Shoop and Kenney 1992). Historically, turtle
interactions have been observed in the scallop dredge fishery from
late June to late October. The potential for interactions also exists
during November due to the overlap of turtles and dredge fishing effort
in the southern most area of the fishery.
Data Sources
Observer Coverage and Observed Turtle Interactions
From June - November 2003,
observer coverage (percentage of dredge hours observed) in the entire
Mid-Atlantic was 2.7%. Coverage was higher in the Hudson Canyon Controlled
Access Area (9.7%) than outside this area (1.4%) (Figure
1).
Twenty-two turtle interactions
were observed in the Mid-Atlantic region, of which 16 were in the Hudson
Canyon Controlled Access Area, and 6 were outside this area (Table
1, Figure 2). Five turtles were released alive and uninjured, 1 was
fresh dead, 12 were alive and injured, 1 was resuscitated, and 3 were
alive yet whose condition was unknown. Four (18%) interactions occurred
during July, 6 (27%) during August, 2 (9%) during September, and 10
(46%) during October.
Seventeen (77%) of the interactions
were loggerheads (Caretta caretta), and 5 (23%) were not positively
identified. Four of the 5 unidentified turtles occurred on trips that
also interacted with a loggerhead. According to observer logs, instances
where turtles could not be identified occurred when the turtle fell
from the gear and the observer did not have enough time to positively
identify the turtle. Since 2001, only loggerhead species have been
positively identified in the Mid-Atlantic scallop dredge fishery. Because
loggerheads have been the only species of sea turtles observed in the
scallop dredge fishery since a dedicated observer program began in
2001, and 4 out of the 5 unidentified interactions took place in loggerhead
territory, it is assumed that the 5 unidentified species were loggerheads.
Therefore, the 2003 total bycatch estimate is considered to be for
loggerhead turtles only.
In addition, two turtles,
each on separate trips during October, were captured when an observer
was off-watch (Table 2). One trip was inside the Hudson Canyon Controlled
Access Area and the other was outside. These 2 interactions occurred
when the observer was not collecting information on the haul. These
turtles were not included in the count of turtles used to calculate
bycatch rates, though they are subsumed in the extrapolated estimate.
Furthermore, 6 severely decomposed turtles wrapped in gillnet gear
were captured on a single trip during September. These interactions
occurred in the Hudson Canyon Access Area. These turtles were also
not included in the count of turtle interactions because the mortalities
may have occurred from previous interactions with gillnet gear.
Commercial Fisheries Data
Any vessel issued a federal
fishery permit with Vessel Trip Report (VTR) requirements must submit
a report for each trip taken. In the report, vessel operators are required
to provide information such as, date and time sailed, date and time
catch landed, quantity and size of gear, average depth fished, statistical
area fished, latitude and longitude, average tow time, species composition
and catch weight, and total number of hauls.
Vessel Trip Reports from
scallop dredge fishermen operating in Mid-Atlantic waters from Long
Island, NY to Cape Hatteras, NC from June to November 2003 were used
in this analysis. Number of dredge hours from trips without coordinate
positions (1.5%) were prorated between the Mid-Atlantic and Hudson
Canyon Controlled Access Area regions based on the percentage of dredge
hours in these regions from trips with known coordinates.
The frequency of commercial
and observed dredge hauls were compared for variables available in
both the commercial and observer data to assess how representative
the observer data were of the commercial fleet.
Temperature Data
It was necessary to acquire
temperature values for each Observed and VTR fishing event in order
to model and extrapolate bycatch rates for the total mortality estimate.
Sea surface temperature at each position reported in the Observer and
VTR databases was extracted from NOAA Advanced Very High Resolution
Radiometer (AVHRR) Coastwatch Satellite Images. A Visual Basic routine
was used to extract temperatures from 7-day composite images (3 days
forward and backward of the haul date), using a 3x3 cell window at
1 km resolution. Therefore, a
9 km2 area of coverage around each coordinate position was used to
extract sea surface temperature. Within the 3x3 cell search radius, the pixel
representing the warmest temperature was used to avoid temperatures affected
by cloud coverage.
Sea
surface temperature values could not be obtained for 4% of the VTR
data due to missing coordinate positions on the VTR logs or to unclear
satellite images. For these fishing events, sea surface temperature
was predicted using a generalized additive model based on year, month,
day, and statistical area in which the vessel fished.
Modeling Approach
Generalized Linear Model
(GLM) and Generalized Additive Models (GAM) fitting techniques were
used to understand and predict bycatch rates of sea turtles in relation
to environmental variables, fishing practices, and gear characteristics
in the U.S. commercial sea scallop dredge fishery in the Mid-Atlantic
region.
Before a GLM was constructed,
a GAM helped group continuous variables into categories (Figure
3).
Fitting the GLM model with categorized variables was necessary to extrapolate
bycatch rates to derive a total estimate of the bycatch of turtles
in scallop dredges. All of the variables tested in the GLM model were
first fitted to a GAM, in which the parameters of the continuous prediction
variables were estimated by a smoothing spline. Variable values were
grouped according to whether they had a positive or negative influence
on the bycatch rate (i.e., the group explained more or less of the
bycatch rate).
Development of a GLM Bycatch Model
The bycatch rate of turtles
is defined as:
Number of Observed Turtles
Number of Observed Dredge Hours
where
Dredge Hour (1) = Number of Dredges*Dredge
Haul Duration (Hrs)
Because bycatch events were counts ranging from zero to one, a logistic
regression was used to model the probability of turtle bycatch per unit
dredge hour, i.e., the bycatch rate (GLM function, SPLUS 6.2). The model
can be written as:
log(turtlebyc / dredgehr) = ß0 = ß1x1 + ß2x2+......ßixi
where 
i is
a parameter coefficient, and xi a predictor
variable describing environmental, gear, or fishing characteristics.
In order to model the bycatch rate (i.e., adjusting for varying time
spans), the logarithm of dredgehr becomes an offset variable with a coefficient
set to 1.0 (Allison 1999; Palka and Rossman 2001). Thus the model can
be rewritten as:
log(turtlebyc) = log(dredgehr)
= ß0 = ß1x1 + ß2x2+......ßixi
A forward stepwise selection method was used to determine the best
fitting model. Model parameters were estimated by maximizing the log-likelihood
function. The null model was the first model in the stepwise process
and was specified with the offset variable alone:
H0 : log(turtle bycatch) = log(dredgehr)
At each step, a new variable was added to the null model (Table
3)
and tested against the previous model formulation (ANOVA function, Chi-square
test) to determine the better fitting model. Main effects of each variable
were tested in the stepwise selection process. If a P-value was less
than or equal to 0.05, then the additional variable was considered to
explain more of the variability in bycatch than a model without that
variable.
The order in which variables were tested in the forward stepwise model
was determined by first ranking variables based on their Akaike Information
Criterion (AIC) value relative to other variables in the model (StepAIC
function, SPLUS 6.2). The AIC is defined as:
where 
is
the numerical value of the log-likelihood at its maximum point and
K is the number of estimable parameters (Burnham and Anderson, 2002).
The
AIC is a measure of the goodness of fit that includes the level of
parsimony, defined as a model that fits the data well and includes as
few parameters
as necessary (Burnham and Anderson, 2002).
In the StepAIC process, a sequence of models is fitted to the null
model in an automated process, where the first variable added has resulted
in the greatest change in AIC relative to all other variables in the
scope of the model. If the AIC value decreases, the new combination of
variables in the model fit the data better. A 2nd variable
is then fitted to the previous model based on the greatest change in
AIC and so on, until the AIC value no longer reduces. Each variable in
the final sequence was then tested individually in the forward stepwise
selection process described above. A low AIC value relative to another
model does not necessarily mean the additional variable is significant
at 
=0.05. P-values, therefore, were used for model selection.
Alias patterns in the full GLM model (i.e., a model with all categorical
variables included) were examined to assess correlation among the explanatory
variables. In addition, scatter plots and pearson correlation coefficients
between variables were examined before variables were grouped into the
GAM categories to assess possible interactions.
Bycatch Rate Estimates
The spatial and temporal
stratification of bycatch rates in the Mid-Atlantic was determined
by the explanatory variables in the best-fitting GLM. Parameter estimates
from the best-fitting model were used to predict the bycatch rate for
each stratum.
The coefficient of variation
(C.V.) for each bycatch rate was estimated by bootstrap resampling
(Efron and Tibshirani, 1993). The resampling unit was a single trip.
Replicate bycatch rates were generated based on the best-fitting GLM
model, by sampling with replacement 1000 times from the original data
set. The C.V. was defined as the standard deviation of the bootstrap
replicate bycatch rate in a stratum divided by the bycatch rate for
that stratum estimated from the original data.
An annual C.V. from stratified
estimates was calculated by taking the square root of the sum of bycatch
rate variances over all strata, and dividing by the sum of bycatch
rates over all strata (Wade and Angliss, 1997):
Annual CV = 
where xi = the bootstrap replicate mean bycatch rate in each
stratum
Total Bycatch
The total estimated turtle
bycatch in each stratum was calculated as the product of predicted
bycatch per dredgehr (i.e., the predicted bycatch rate) for that stratum
and the total number of dredge hours by the commercial fishery in that
stratum:
 Predicted Bycatchi |
x (Total Dredge Hours)i |
Dredge Hour i |
|
where i = stratum
Total bycatch was the sum of the stratified bycatch
estimates.
____________________
1. In
previous years most hauls were 1 hour long with little variation
around this time. With an increased sample size in 2003, there was
more variation
in haul durations, making it necessary to standardize for the amount
of time each dredge was in the water.
RESULTS
Data Representativeness
Based on the spatial distribution
of commercial dredge fishing effort in the Mid-Atlantic, the Hudson
Canyon Controlled Access Area was over-sampled in 2003 (Table
4). Small
vessels
(from 42-77 ft) were under-sampled, as were small dredges (<=10 ft frame width).
Fifteen percent of dredge hauls accomplished by commercial vessels in the Mid-Atlantic
used dredges less than or equal to 10 feet, while only 2% of this size dredge
was sampled by observers. Commercial effort in shallow depths (2-26 fm) was under-sampled,
and over-sampled in mid-depth ranges (27-31 fm).
Some variables, such as
tow speed, could not be used to estimate total bycatch because these
data are not recorded on VTR records. Thus, for these variables, it
is assumed that unobserved hauls have the same characteristics as the
observed hauls used in the bycatch model.
Factors Influencing Bycatch
Significant factors affecting
sea turtle bycatch were sea surface temperature and tow speed (Table
5). In estimating total bycatch, rates were stratified based on temperature
because tow speed was not available from fishing logbooks. Highest
probability of bycatch occurred in
surface waters 22°C or warmer. Hence bycatch rates were stratified
based on whether
temperatures were higher or lower than 22°C.
Bycatch rates inside and
outside the Hudson Canyon Controlled Access Area were identical. In
addition, rates did not differ due to use and number of rock and tickler
chains, vessel length, or frame width of the dredge.
Total Mortality Estimate
The total estimated bycatch
of sea turtles from June to November, 2003 in the Mid-Atlantic sea
scallop dredge fishery is 749 turtles (C.V.=0.28) (Table
6). Five of
the 22 observed turtles were released alive and uninjured, implying
a 22.7% survival rate and a 77.3% mortality/injury rate. Hence, of
the 749 interactions, an estimated 579 turtles are considered to have
died or have injuries.
Out of the 749 interactions,
122 (16%) occurred in the Hudson Canyon Controlled Access Area, and
627 (84%) outside of this area (Table 6a). Interactions in the Virginia
Beach Controlled Access Area are subsumed in the total estimate. However,
because there was only 1 commercial trip in this Access Area between
June and November 2003, the likelihood of a turtle interaction in this
Access Area was very low.
DISCUSSION
Factors Influencing Bycatch
Stratifying bycatch rates
by sea surface temperature accounts for much of the temporal and spatial
variation in bycatch rates of turtles in the Mid-Atlantic. Observed
hauls in waters
22°C or warmer occurred in offshore waters from New York harbor down
to the mouth of the Chesapeake, with waters closer to shore (i.e. west of the
27fm isobath) warming sooner than offshore (Figure 4). In general, months for
these warm waters, as indicated by the observed hauls, range from early July
to end of October. The timing of surface temperature warming and cooling on shelf
waters in the Mid-Atlantic depends on a variety of factors, including air temperatures,
the position of the Gulf Stream, water transport from the north, and the timing
of vernal warming
or fall overturn (Jossi and Benway 2003).
Tow speed was also a significant
factor affecting bycatch of turtles. Had bycatch rates been stratified
by tow speed, highest rates would have occurred during hauls having
tow speeds between 4.4 and 4.9 knots.
Tow speed should be investigated
further as a factor influencing the bycatch of turtles. Tow speed may
affect how dredges come in contact with the bottom, influencing the
potential for a turtle encounter. According to James Kendall (1998), "with
dredge gear, as well as with a lot of other towed gear including trawls,
speed plays an important role.....the fact is that if fishermen tow
too fast, they will actually fly the gear right up off the bottom much
as with a kite, because the gear becomes functionally weightless." It
is possible that there is an optimal speed to catch scallops that allows
for the dredge to maintain contact with the bottom, leading to a higher
probability of a turtle encounter if turtles are foraging or crouching
on the bottom. Presence or absence of a pressure plate, which helps
keep the dredge frame on the bottom, did not significantly affect turtle
bycatch, though other related factors not considered in this analysis
may play a role.
Bycatch Patterns
Over the last three
years, sea surface temperature has been a significant predictor of
sea turtle bycatch in both the Mid-Atlantic Controlled Access Areas
[2001-2002], and over the broader Mid-Atlantic from New York to North
Carolina [2003]. In 2001-2002, a higher probability of turtle bycatch
occurred after waters warmed to 19C, and in 2003, higher
probabilities occurred after waters warmed to 22C. These
differences may reflect inter-annual variation in sea surface temperatures,
turtle distributions, and shifting patterns in the fishery. Nevertheless,
there may be a consistent minimal threshold temperature from year to
year above which turtle bycatch is more likely to occur, though the
minimal temperature threshold at which the observed interaction occurs
is likely to fluctuate from year to year.
From 2001-2003,
turtle bycatch occurred between depths of 25-35 fm. In general, the
fishery operated in depths from 5-50 fm, with 40-50% operating in depths
shallower than 25 fm over these three years. While depth was not a
significant predictor of bycatch in this analysis, the area in which
turtle bycatch occurred over the last three years is worth noting.
Hot spots for turtle bycatch may consistently occur within certain
depths that do not coincide with the entire range of the fishery. More
sampling coverage should be accomplished in shallower depth ranges
to further explore this idea.
Data Caveats
While the best-fitting
GLM model for predicting turtle bycatch in the scallop dredge fishery
included tow speed, this co-variate had to be removed from the final
model because tow
speed information is not available in VTR logbooks. Removing the effect
of tow speed from the final model forces the assumption that bycatch rates are
the same for vessels fishing at different speeds. This may cause the bycatch
estimate to be biased in certain directions depending on the variability of tow
speeds throughout the Mid-Atlantic.
The disproportionate amount
of observer coverage in the area outside of the Hudson Canyon Access
Area relative to total commercial effort may have obscured the ability
to detect differences in bycatch rates among the variables examined.
For example, there was very little coverage on boats using dredges
10 feet wide or smaller. The low sample size of small dredges may have
prevented the detection of a difference in rates due to dredge size.
This analysis assumes that bycatch rates are the same for all dredge
sizes. If rates are higher or lower in small dredges the total estimated
bycatch will likewise be biased high or low. Many vessels using small
dredges are fishing under a General Category permit for scallops, where
scallop catch is limited to 400 pounds shucked scallops outside the
Controlled Access Areas and 100 pounds inside. More coverage is needed
on boats using small dredges and/or of the General Category permit
boats to ensure that fishing behavior on these boats does not increase
the potential for turtle bycatch.
In this analysis, main effects
of each categorical variable were considered in the bycatch model because
there did not appear to be any interaction effects between variables.
In other words, the effect of vessel length and depth on bycatch rates,
for instance, are considered independently from one another in the
GLM. The clustered nature of many of the parameter values, however,
may have statistical consequences on the model results. For instance,
the majority of smaller vessels may tow in shallower waters and at
lower tow speeds. So, the range of parameter values is not represented
equally across all variables. Further work needs to address whether
this clustering influences the model's ability to detect significant
effects of a single variable.
Further work should also
investigate whether the behavior of particular vessels, other than
what has been examined in this analysis, increase the likelihood of
a turtle interaction. For example, 1 of the observed trips (Trip K,
Table 1) took 4 turtles, 2 of the observed trips (I and L, Table 1)
took 3 turtles, and 3 of the observed trips (A, C, and J, Table 1)
took 2 turtles. Hence, 16 (73%) of the 22 observed interactions involved
multiple interactions. These vessels may have been behaving in a manner
that increased the likelihood of catching a turtle in the dredge. Identifying
such possible behavioral patterns may facilitate mitigation options
for reducing interactions. Furthermore, some of these options may need
not apply to the entire commercial fleet.
Finally, given the association
between sea surface temperature and turtle bycatch, further work should
examine bycatch rates over a range of different temperature strata.
Exploring changes in rates as a function of temperature may indicate
whether predicted bycatch rates are sensitive to small changes in temperature.
Furthermore, this research would help inform the design of possible
time and area closures intended to reduce turtle bycatch, by examining
the percentage of bycatch reduction achieved at various temperature
ranges.
Other Fisheries
The bycatch of turtles in
the Mid-Atlantic scallop dredge fishery represents only a portion of
the total bycatch of sea turtles in Northwest Atlantic Ocean fisheries.
For example, trawl fisheries targeting a variety of species also operate
in the same area and months when turtles are present in the Mid-Atlantic.
Trawls that target scallops may have a different bycatch rate of turtles
than trawls that target fish because scallop trawl gear is of somewhat
different design, is fished closer to the bottom, and is fished at
different speeds than trawl gear for fish. However, there has been
very little observer coverage with which to estimate bycatch in the
scallop trawl fishery. Based on reported trips, coverage (% trips observed)
from 2001-2003 was roughly 0.2%. There were 7 trips observed during
2001-2004 [4 in 2001; 1 in 2002; 0 in 2003; and 2 in 2004
through July] and no sea turtles were observed on any of these trips.
Observer coverage is currently scheduled for late summer and fall of
2004. This coverage may start to shed light on the magnitude of turtle
bycatch in the scallop trawl fishery.
ACKNOWLEDGMENTS
Many thanks to Debra Palka for her statistical guidance in the bycatch
assessment, as well as to Andre Punt for his comments on the general
modelling approach. David Mountain was extremely helpful in assembling
2003 Coastwatch Images and writing the Visual Basic routine to extract
temperature data. Paul Rago, Dvora Hart, Debra Palka, Fred Serchuk, and
Richard Merrick provided insightful comments on earlier drafts of the
manuscript.
REFERENCES
Allison, P. 1999. Logistic Regression Using the SAS System: Theory
and Application. SAS Institute Inc., Cary, NC. 304 pp.
Burnham, K.P. and D.R. Anderson. 2002. Model Selection and Multimodal
Inference: A Practical Information-Theoretic Approach, 2nd Ed.
Springer-Verlag, New York, 488p.
Efron, B. and R. Tibshirani. 1993. An Introduction to the Bootstrap.
Chapman&Hall, New York, 436 p.
Jossi, J. and R. Benway. 2003. Variability of temperature and salinity
in the Middle Atlantic Bight and Gulf of Maine based on data collected
as part of the MARMAP Ships of Opportunity program, 1978-2001. NOAA Technical
Memorandum NMFS-NE-172. Accessible at http://www.nefsc.noaa.gov/nefsc/publications/tm/tm172/
Kendall, J. 1998. Scallop Dredge Fishing. In Effects of Fishing
Gear on the Sea Floor of New England (E. Dorsey and J. Pederson eds.),
pp 90-93. Conservation Law Foundation, Boston, MA.
Morreale, S. 1999. Oceanic migrations of sea turtles. PhD dissertation,
Cornell University. 144 p.
Murray, K. (in press). Magnitude and distribution of sea turtle bycatch
in the sea scallop (Placopecten magellanicus) dredge fishery
in two areas of the Northwestern Atlantic Ocean, 2001-2002. Fishery
Bulletin 102(4).
NEFMC [New England Fishery Management Council]. 2002. Framework Adjustment
15 to the Atlantic Sea Scallop Fishery Management Plan. Newburyport,
MA. Accessible at http://www.nefmc.org/scallops/index.html.
Palka, D. and M. Rossman. 2001. Bycatch estimates of coastal bottlenose
dolphin (Tursiops truncatus) in U.S. Mid-Atlantic gillnet fisheries
for 1996-2000. Northeast Fisheries Science Center Reference Document
01-15. 77p.
Shoop, R. and R. Kenney. 1992. Seasonal distributions and abundances
of loggerhead and leatherback sea turtles in waters of the northeastern
United States. Herpetological Monographs 6:43-67.
Wade, P. and R. Angliss. 1997. Guidelines for Assessing Marine Mammal
Stocks: Report of the GAMMS Workshop April 3-5, 1996, Seattle, Washington.
U.S. Dep. Commer., NOAA Tech. Memo. NMFS-OPR-12, 93 p.
