ENDANGERED SPECIES RESEARCH
Endang Species Res
Vol. 5: 309–323, 2008
doi: 10.3354/esr00133
Printed December 2008
Published online October 29, 2008
Contribution to the Theme Section ‘Fisheries bycatch: problems and solutions’
OPEN
ACCESS
Reducing seabird bycatch in the Hawaii longline
tuna fishery
Eric Gilman1, 2,*, Donald Kobayashi3, 4, Milani Chaloupka5
1
IUCN Global Marine Programme, 2718 Napuaa Place, Honolulu, Hawaii 96822, USA
Centre for Ecology and Conservation, University of Exeter, Cornwall Campus, Penryn TR10 9EZ, UK
3
US NOAA Fisheries, Pacific Islands Fisheries Science Center, 2570 Dole Street, Honolulu, Hawaii 96822, USA
4
Department of Environmental Sciences, University of Technology, Sydney, New South Wales 2007, Australia
5
Ecological Modeling Services, PO Box 6150, University of Queensland, St. Lucia, Queensland 4067, Australia
2
ABSTRACT: Mortality in longline fisheries represents a global threat to some species of pelagic
seabirds. Regulations were adopted in 2001 to reduce seabird bycatch in the Hawaii longline tuna
fishery. We used a Poisson generalized additive regression modeling approach to evaluate the
change in seabird bycatch rates from the pre- to post-regulation period, and to evaluate the efficacy
of alternative combinations of seabird bycatch reduction methods employed during the post-regulation period. Informative covariates of temporal and geo-referenced spatial effects of fishing effort and
sampling variation commonly found with count data were included in the model to provide a better
inference of the effect of employing required changes in fishing gear and methods. We found a significant 67% (95% CI: 62 to 72) reduction in the seabird bycatch rate following the introduction of
regulations. Post-regulation, sets employing 4 different combinations of seabird avoidance methods
all resulted in significant reductions to the pre-regulation seabird catch rate. Employing side-setting
and 60 g weights was the only combination with no seabird captures during the post-regulation
period. Using heavier branch line weights and treated bait (thawed and dyed blue) both significantly
reduced seabird catch rates. Temporal and spatial distribution of fishing effort and the timing of initiating setting were also significant factors affecting seabird bycatch rates: time-area closures and
restricting the timing of setting could further reduce seabird bycatch. A substantial proportion of
seabird captures occurred south of the area where mitigation measures are required: moving the
southern boundary farther south would further reduce seabird catches.
KEY WORDS: Albatross · Bycatch · Fisheries · Longline · Mitigation · Seabird
Resale or republication not permitted without written consent of the publisher
Mortality in longline fisheries is the most critical
global threat to most species of albatrosses and large
petrels (Gales 1998, Brothers et al. 1999, Gilman et
al. 2005). Primarily while fishing gear is being set,
seabirds are hooked or entangled, dragged underwater, and drown as the gear sinks. The species of
seabirds most frequently caught on longlines are albatrosses and petrels in the Southern Ocean, Arctic fulmar Fulmarus glacialis in North Atlantic fisheries and
albatrosses, gulls, and fulmars in North Pacific fisheries (Brothers et al. 1999). Longlining occurs through-
out the world’s oceans, has been used since the 19th
century, and is practised by small-scale domestic artisanal fisheries with small, open vessels to modern
mechanized industrialized fleets from distant-water
fishing nations with large vessels. Pelagic longlines,
where gear is suspended from a mainline drifting
freely in the pelagic environment, at depths anywhere
from the sea surface to 400 m, mainly target large
tunas Thunnus spp., swordfish Xiphias gladius, other
billfishes Istiophoridae spp., dolphin fish (mahimahi)
Coryphaena spp. and sharks. Longlines can be up to
100 km long and carry up to 3500 baited hooks (Beverly et al. 2003, Gilman et al. 2008). Incidental bycatch
*Email: eric.gilman@iucn.org
© Inter-Research 2008 · www.int-res.com
INTRODUCTION
310
Endang Species Res 5: 309–323, 2008
of Laysan Phoebastria immutabilis and black-footed P.
nigripes albatrosses in North Pacific longline fisheries,
including the Hawaii pelagic longline tuna and swordfish fisheries, is a conservation and management concern (Cousins & Cooper 2000, US Fish and Wildlife
Service 2002, 2004, Gilman & Freifeld 2003, Lewison &
Crowder 2003, Niel & Lebreton 2005).
In 2007 there were 129 active domestic Honolulubased pelagic longline vessels targeting tuna, setting
38.8 million hooks (Gilman 2008, Gilman et al. 2008,
US National Marine Fisheries Service 2008). The observer coverage rate of the fishery was about 4% from
1994 to 2000, after which it was increased to about
20%. The longline tuna fishery lands primarily bigeye
Thunnus obesus, yellowfin T. albacares and albacore
T. alalunga tunas as well as several incidental pelagic
species. The fishery annually lands about 9800 t of fish
worth USD 43 million (Gilman 2008). Vessels are 10 to
31 m long. The tuna fishery operates year-round.
Vessels make trips of about 21 d duration, and about
15 trips per year, with about 10 sets per trip and
2000 hooks per set. Fishing grounds are in the western
and central Pacific Ocean, within the Exclusive Economic Zone adjacent to Hawaii and on the high seas.
From January through March, effort is concentrated
between 15 and 35° N latitude and 150 and 180° W longitude. From April through June fishing expands to the
south and spreads farther east and west to about
145° W and 170° E. Baited hooks are set at depths
between 35 and 224 m. Vessels use a wire trace, place
a lead-centered swivel of 45 to 80 g within 1 m of the
hook, use fish for bait (a mix of frozen mackerel, saury
and sardine) and do not use lightsticks. The gear soaks
during the day and is hauled at night. Mainline1 length
is about 50 km and is deployed with a hydraulic
shooter. The primary hook used is a tuna 3.6 ring hook,
with some use of 14/0, 15/0 and 16/0 circle hooks
(Gilman et al. 2008). This is a general description of the
gear and operational characteristics of the Hawaii
longline tuna fleet; however, there may be large variability in fishing gear and methods between vessels in
the fleet and even for an individual vessel.
Based on results from controlled and comparative
experiments of seabird avoidance methods in the
Hawaii longline fleet (McNamara et al. 1999, Boggs
2001, Gilman et al. 2007a), fishery management
authorities adopted and amended regulatory measures
for the Hawaii longline fleet to reduce seabird catch
rates. Since regulations were first adopted in June
2001, total seabird captures in the combined Hawaii
longline tuna and swordfish fisheries dropped 96%
from an estimated 2433 in 2000 to 88 in 2006 (Van Fossen 2007). Current regulations, which came into effect
on 18 January 2006, require Hawaii longline vessels
targeting tuna, when fishing north of 23° N, to either:
(1) Side-set, attach weights that are a minimum of
45 g to branch lines within 1 m of the hook, and deploy
a bird curtain aft of a mainline shooter. The mainline
must be deployed a minimum of 1 m forward from the
stern. The bird curtain must be a minimum of 3 m long
with 3 or more streamers; or,
(2) Attach weights that are a minimum of 45 g to
branch lines within 1 m of the hook; when seabirds are
present, discharge fish, offal (fish parts) or spent bait
while setting or hauling from the opposite side of the
vessel from where gear is being set or hauled; and use
only completely thawed, blue-dyed bait to match a
color quality control card (US National Marine Fisheries Service 2005).
Side-setting means setting longline gear from the
side of the vessel rather than the conventional position
at the stern (Fig. 1). When implemented to prescription, crew throw baited hooks forward and close to the
side of the vessel hull where seabirds, such as albatrosses, are unable or unwilling to pursue them. Ideally, by the time the stern passes, the hook has sunk
beyond the reach of seabirds (Gilman et al. 2007a).
Employment of a bird curtain in combination with
1
The mainline is monofilament line that is deployed horizontally off a reel, onto which branch lines with baited hooks,
buoys, and radio beacons are attached. A branch line is the
line that attaches to the mainline with a clip, is generally 6 to
10 m in length in this fishery, and has the following components: clip attached to the mainline, 6 to 10 m of monofilament nylon line, 45 or 60 g lead center swivel, wire leader,
and a baited hook
Fig. 1. Deck positions for side- versus stern-setting on longline fishing vessels (illustration courtesy of Nigel Brothers)
311
Gilman et al.: Reducing seabird longline bycatch
side-setting is believed to further reduce seabird
access to baited hooks being set by preventing foraging seabirds from manoeuvering close to the vessel
hull near where the setting operation is taking place
(Gilman et al. 2007a). Adding weights to branch lines
increases the baited hook sink rate, reducing the risk
of seabirds being able to access baited hooks as they
are being set (Brothers et al. 1999, Boggs 2001). The
intent of dyeing bait dark blue, by reducing the contrast between the bait and sea color, is to make it more
difficult for birds to detect the bait when foraging from
above (McNamara et al. 1999, Boggs 2001, Minami &
Kiyota 2002, Gilman et al. 2005, 2007a). To dye bait
blue to achieve regulatory-required darkness, bait is
supposed to be completely thawed and soaked in a tub
with dissolved blue food coloring (Virginia Dare FD&C
Blue No. 1) powder at a concentration of 4 g l–1 of water
for 1 to 4 h (Fig. 2).
A comparative study of the efficacy of side-setting,
blue-dyed bait and other seabird avoidance methods
found that side-setting resulted in a significantly lower
seabird catch rate than blue-dyed bait, and that sidesetting provided substantial operational benefits
(Gilman et al. 2007a). Other seabird avoidance methods were found to be relatively impractical for employment by crew (Gilman et al. 2007a). Although Gilman
et al. (2007a) tested the comparative single factor efficacy of 3 types of seabird avoidance methods, this
study used observer data to compare the efficacy of
different combinations of seabird avoidance strategies
employed by the fleet during conventional commercial
fishing operations.
We analyzed observer data from the US National
Marine Fisheries Service for the Hawaii longline tuna
fishery to calculate and compare seabird bycatch rates
for pre- and post-regulation periods requiring the
employment of seabird avoidance methods, and different combinations of methods employed to avoid catching seabirds during the post-regulation period.
METHODS
Pre- vs. post-regulation period. To compare pre- vs.
post-regulation seabird capture rates, we analyzed
data from the Hawaii longline observer program for
Hawaii-based longline tuna-targeting sets (defined by
the US National Marine Fisheries Service [2005] as sets
containing ≥15 hooks between floats) for the periods
before and after seabird avoidance regulations came
into effect. The analysis was conducted employing
only sets where one or more albatross was observed
present during setting or hauling operations and/or a
seabird was captured. For sets where no albatrosses
were present, the observation that no albatrosses were
captured is a result of an absence of albatrosses at the
fishing grounds and is not a reflection of the efficacy of
any methods employed to avoid catching birds, hence
the decision not to include these sets in the analysis.
We only considered the presence or absence of albatross species to determine whether or not to include a
set in the analysis, as captures of other seabird species
are very rare events in this fishery of 310 seabirds
observed captured in this fishery from 2 March 1994 to
4 September 2007, 6 (2%) were a species other than
black-footed or Laysan albatrosses. The pre-regulation
period used for the purposes of the present study
started on 9 May 2000, the date the observer program began recording seabird abundance during fishing operations, and ended on 9 June 2001, the day
before seabird regulations went into effect. The postregulation study period was from 10 June 2001
through 4 September 20072.
The number of birds hauled aboard is used to estimate the total number of seabirds captured during the
set, despite evidence that this method underestimates
total bird capture (Brothers 1991, Gales et al. 1998,
Gilman et al. 2003, 2007a). Observers are not required
to observe the entire setting of the gear, and therefore
all seabird captures occurring during setting are not
necessarily observed. Thus, observations of bird captures during setting operations were not used for this
analysis.
It was not possible to normalize seabird bycatch
rates for albatross abundance, as conducted in previ-
2
Fig. 2. Bait is completely thawed and dyed blue by soaking
in a large tub with dissolved blue food coloring to achieve
regulatory-required darkness
At the time of conducting the query to the US National
Marine Fisheries Service observer program database, the US
National Marine Fisheries Service had not completed validating some of the observer data from longline tuna trips
included in this analysis
312
Endang Species Res 5: 309–323, 2008
ous experiments (Gilman et al. 2003, 2005, 2007a). This
was because seabird abundance was not estimated in
a consistent manner throughout the entire study
period, and observations were not made during the
setting portion of the fishing operation of every set
included in the analysis prior to 28 September 2004.
Spatial and temporal trends in seabird catch rates for
the pre- and post-regulation sets were estimated using
a nonparametric regression modeling approach known
as generalized additive modeling (GAM). This method
allows (1) flexible specification of the error and link
functions, and (2) arbitrary specification of the functional form for each predictor included in the model
(Hastie & Tibshirani 1990). GAMs are the preferred
choice for analysis of spatial and temporal trends in
data series that may comprise nonlinear behavior and
involve multi-level sampling designs (Fahrmeir & Lang,
2001, Wood 2006). The GAM approach relaxes the normality assumptions of the standard linear regression
modeling approach and supports flexible link specification, while the functional form (linear, nonlinear) for
each predictor is estimated from the data using nonparametric smoothers while conditioning on other covariates in the model. The standard regression model
for count data, such as number of seabirds caught in a
specific time period, assumes the Poisson probability
model structure (McCullagh & Nelder 1989). Therefore,
we fitted Poisson GAMs to the seabird catch using an
offset term — offset(ln[hooks]) — to account for sampling effort measured as hooks per set. Hence, the response variable is the number of seabirds caught per
set, but is now a standardized catch rate as sampling effort was also explicitly accounted for using the offset
term. Informative covariates or factors included in the
model were (1) time of day of set, (2) geographic location where sets were initiated, (3) season of set, and (4)
whether the set was made during the pre-or post-regulation period. This nonparametric regression model was
fitted using thin plate regression splines to model any
nonlinear (time of day of initiating a set) or 2-dimensional effects (location of set). A quasi-Poisson error
structure was employed to account for over-dispersed
count data, as the data set comprised a large number of
sets with zero seabird catches (McCullagh & Nelder
1989). All smoothness parameters were determined using generalized cross-validation (Wood 2006). Models
were fitted using the mgcv package (Wood 2006) available for the open source statistical modeling program R
(Ihaka & Gentleman 1996). Sullivan et al. (2006) employed a similar statistical method in a study of methods
to reduce seabird mortality from striking warp cables in
factory trawl vessels, accounting for over-dispersion by
using a negative binomial error structure in a generalized linear model (GLM). Instead, we employed a
quasi-Poisson error structure in a GAM to address over-
dispersion because this is a far more flexible and general model that accounts for both over-dispersion and
the nonlinear function form of the covariates (see Wood
2006).
We calculated estimates of seabird catch rates for the
pre- and post-regulation periods, based on a binomial
estimator with Clopper-Pearson confidence intervals
(Agresti 2002), classified by the following informative
factors which could affect the bird bycatch rates: ≤45 g
vs. > 45 g branch line weighting, timing of the start of
the set and season. We also determined the proportion
of sets included in the sample that were made outside
of the area where seabird bycatch reduction methods
are required.
Alternative combinations of seabird avoidance
methods. For a second study component, to compare
alternative combinations of seabird avoidance methods, observer data collected from the Hawaii longline
tuna fishery were analyzed for the period 15 August
2003 through 4 September 2007. On 15 August 2003
observers began recording whether vessels were setting from the stern or side of the vessel; hence the decision not to include observer program data prior to this
date. For this study component, a set was included in
the analysis if (1) one or more albatrosses was recorded
as being present during the fishing operation; and/or
(2) one or more seabirds (any species) was observed
hauled aboard following that set.
Sets were categorized into one of 4 combinations of
seabird avoidance methods:
(1) Side-setting with 45 g weights attached to branch
lines within 1 m of hook;
(2) Side-setting with 60 g weights attached to branch
lines within 1 m of hook;
(3) Stern-setting with 45 g weights attached to
branch lines within 1 m of hook; or
(4) Stern-setting with 60 g weights attached to
branch lines within 1 m of hook.
Spatial and temporal trends in seabird catch rates for
the 4 categories of sets employed during the postregulation period and sets employed during the preregulation period were estimated by fitting Poisson
GAMs to the seabird catch. Informative covariates
included in the model were (1) time of day of set, (2)
location where sets were initiated, (3) season of set, (4)
whether or not bait was treated (thawed and dyed
blue); and (5) which of the 4 pre-regulation categories
the set fits or otherwise if the set was during the preregulation period. The GAM was fitted using thin plate
regression splines to model any nonlinear effect (time
of day of initiating a set) or 2-dimensional effect (location of set). Log link (ln[hooks]) was used in the offset
term to account for hooks per set. A quasi-Poisson error
structure was again employed to account for overdispersed count data (McCullagh & Nelder 1989). All
Gilman et al.: Reducing seabird longline bycatch
smoothness parameters were determined using generalized cross-validation (Wood 2006). Models were fitted using the mgcv package (Wood 2006) available for
the program R (Ihaka & Gentleman 1996).
We also combined these 4 categories into 2 categories of stern- vs. side-setting, and 2 categories of 45
vs. 60 g weights employed during the post-regulation
period to determine if there was a significant difference in seabird catch rates between these 2 pairs of
factors. Spatial and temporal trends in seabird catch
rates were estimated by fitting Poisson GAMs to the
seabird catch. For the side- vs. stern-setting comparison, informative covariates included in the model were
(1) time of day of set, (2) location where sets were initiated, (3) season of set, (4) whether or not bait was
treated (thawed and dyed blue); and (5) size of branch
line weight. For the 45 vs. 60 g comparison, informative covariates included in the model were (1) time of
day of set, (2) location where sets were initiated, (3)
season of set, (4) whether or not bait was treated
(thawed and dyed blue); and (5) whether sets were
made from the side or the stern of the vessel.
We calculated estimates of seabird catch rates for the
data, based on a binomial estimator with ClopperPearson confidence intervals (Agresti 2002), classified
by the following informative factors which could affect
the bird bycatch rates: side- vs. stern-setting, 45 vs.
60 g branch line weighting, timing of the start of the
set and season. We also determined the frequency of
voluntary employment of different seabird bycatch
reduction methods at grounds where seabird avoidance methods are not required.
There were 215 sets where an albatross was
observed during the fishing operation and/or a seabird
was observed captured during the haul which were
excluded from this main analysis because they did not
fit into one of the 4 categories (e.g. sets where a towed
deterrent or tori line was deployed, sets that were
made at night, sets with line weighing less than 45 g
including no weight, sets with atypical line weights,
and sets where a leader length was >1 m). As with the
first study component, the number of birds hauled
aboard was used to estimate the total number of
seabirds captured during the set, and it was not possible to normalize seabird capture rates by albatross
abundance.
RESULTS
Pre- vs. post-regulation period
There were 702 sets of 1 337 224 hooks during the
pre-regulation period, during which 107 seabirds were
observed captured. There were 3800 sets of 7 727 429
313
hooks during the post-regulation period, during which
166 seabirds were observed captured. Table 1 provides
a summary of the data used in the Poisson regression
model. The pre- and post-regulation nominal seabird
bycatch rates were 0.080 (95% CI: 0.066 to 0.097) and
0.021 (95% CI: 0.018 to 0.025) seabirds per 1000 hooks,
respectively, a significant 74 percent reduction in the
pre-regulation period seabird catch rate (Table 1).
Fig. 3 presents the Poisson GAM model fit to the
seabird catch rate data for the combined 4502 pre- and
post-regulation period sets. The 3 covariates or factors
(time of starting setting operations, season in which a
set was made, and location at the start of sets) included
in the model were all significant effects. Timing of initiating setting was a significantly nonlinear effect.
Seabird catch rates were lowest during October to
December, and seabird catch rates in all 4 quarters of
the year were significantly different from each other
(Fig. 3a). Seabird catch rates were lowest during sets
initiated between 0:00 to ca. 05:00 h, dipped slightly
around 15:00 h and then increased during early
evening (Fig. 3b). Higher seabird catch rates occurred
around the main Hawaiian Islands, with the highest
rates in the northwestern sector at ca. 25°N, 170°W
(Fig. 3c). Based on the Poisson GAM model, conditioned on the 4 covariates or factors, the seabird catch
rate decreased significantly by 67% (95% CI: 62 to 72)
from the pre- to post-regulation period (Fig. 3d). The
model was a reasonable fit to the large data set and
accounted for 38.2% of the model deviance.
Of the prescribed seabird bycatch reduction methods, branch line weighting is the one gear design that
was also conventionally employed during the preregulation period. Perhaps due to the adoption of the
seabird regulation line weighting requirement, the
mean amount of weight used during the post-regulation period was significantly different and higher,
although only by a small amount: During the pre-regulation period, vessels conventionally employed branch
line weighting with a mean of 47.4 g ± 0.4 SD, whereas
the mean for the post-regulation period was 49.9 g ±
0.1 SD. When the Poisson GAM model was modified to
explicitly account for this variability in branch line
weighting effect on seabird bycatch rate, by including
branch line weighting as a covariate, branch line
weighting was found to have a significant linear effect
on seabird catch rate (p < 0.01). We did not include
branch line weighting as a covariate in the model as
this is one of the seabird bycatch reduction methods
included in the regulations, and we needed the GAM
model to be affected by the pre- vs. post-regulation
variability in the seabird bycatch reduction methods
prescribed in the regulations. Unlike branch line
weighting, the other seabird avoidance methods in the
regulations (side-setting, dyed and thawed bait, bird
Endang Species Res 5: 309–323, 2008
314
Table 1. Four dimensional contingency table providing summary statistics of seabird capture rates based on a binomial estimator
with Clopper-Pearson confidence intervals (Agresti 2002) for Hawaii longline tuna sets for the time periods prior to and following
the introduction of regulations to reduce seabird bycatch, for sets with seabird captures observed during gear hauling
and/or where albatrosses were present during setting or hauling
Period
Weight within
1 m of hook (g)
Pre-regulations:
2 Mar 1994 – 9 Jun 2001
≤45
Season
Time of
initiating set
(h)
No. of
sets
No. of
hooks
Seabirds
captured
(hauled
aboard)
Jan–Jun
≤7:00
> 7:00
≤7:00
> 7:00
≤7:00
> 7:00
≤7:00
> 7:00
71
298
31
72
32
142
12
44
144 268
556 336
58 501
132 271
65 650
268 391
24 471
87 336
5
63
0
2
2
35
0
0
0.035
0.113
0.000
0.015
0.030
0.130
0.000
0.000
0.011–0.081
0.087–0.145
0.000–0.063
0.002–0.055
0.004–0.110
0.091–0.181
0.000–0.151
0.000–0.042
702
1 337 224
107
0.080
0.066–0.097
365
1248
173
831
229
540
131
283
727 802
2 504 716
360 516
1 716 557
470 735
1 101 414
263 751
581 938
6
89
1
10
10
50
0
0
0.008
0.034
0.003
0.006
0.021
0.045
0.000
0.000
0.003–0.018
0.029–0.044
0.0001–0.015
0.003–0.011
0.010–0.039
0.034–0.060
0.000–0.014
0.000–0.006
3800
7 727 429
166
0.021
0.018–0.025
Jul–Dec
> 45
Jan–Jun
Jul–Dec
Total
Post-regulations:
10 Jun 2001 – 4 Sep 2007
≤45
Jan–Jun
Jul–Dec
> 45
Jan–Jun
Jul–Dec
≤7:00
> 7:00
≤7:00
> 7:00
≤7:00
> 7:00
≤7:00
> 7:00
Total
curtain and management of discards) were generally
only employed during the post-regulation period.
There was uneven seasonal distribution of effort,
with 67, 10, 1 and 22% of sets made during the first
(January to March) through 4th quarters, respectively,
during the pre-regulation period, and 36, 27, 11, and
26% of sets made during the first through 4th quarters,
respectively, during the post-regulation period. The
Poisson GAM model explicitly accounted for the effect
of seasonal distribution of effort on seabird bycatch
rate. During the pre-regulation period, the seabird
catch rates by quarter were 0.09 (95% CI: 0.07 to 0.11),
0.16 (95% CI: 0.10 to 0.25), 0.00 (95% CI: 0.00 to 0.43),
and 0.007 (95% CI: 0.001 to 0.02) seabirds per 1000
hooks, respectively, using a binomial estimator. During
the post-regulation period, the seabird catch rates by
quarter were 0.03 (95% CI: 0.02 to 0.03), 0.04 (95% CI:
0.03 to 0.05), 0.005 (95% CI: 0.001 to 0.012), and 0.003
(95% CI: 0.001 to 0.007) seabirds per 1000 hooks,
respectively, using a binomial estimator. The seabird
capture rates were significantly higher during the first
2 quarters of the year during both the pre- and postregulation periods. The seabird catch rates were significantly lower during the first 2 quarters of the postregulation period relative to the first 2 quarters of the
Seabird bycatch rate
(per 1000 hooks)
Point
95% CI
estimate
pre-regulation period, but this was not the case for the
latter 2 quarters of the year.
Of the 4502 sets where an albatross was observed
present during setting or hauling operations and/or a
seabird was hauled to the vessel during gear retrieval,
2448 (54%) began at a location south of 23° N. Of the
176 sets where one or more albatrosses were captured,
74 (42%) began at a location south of 23°N. Regulations require Hawaii longline tuna vessels to employ
seabird bycatch reduction methods only when fishing
north of 23° N.
Alternative combinations of seabird avoidance
methods
A total of 2001 sets of 4 236 556 hooks were included in
this study component comparing seabird catch rates of 4
alternative combinations of seabird avoidance methods.
There were 60 sets with one or more seabirds observed
captured. Of these, 53 sets had one or more confirmed
albatross species caught. Table 2 provides a summary of
the data used in the Poisson regression model.
Post-regulations, seabird catch rates in all 4 categories
of sets were significantly lower than the pre-regulation
Gilman et al.: Reducing seabird longline bycatch
315
Fig. 3. Nonparametric Poisson regression model fitted to the seabird catch in sets made by the Hawaii longline tuna fishery during periods before (n = 702 sets) and after (n = 3800 sets) mandatory seabird avoidance measures came into effect, for sets with
seabird captures observed during gear hauling and/or with albatrosses present during setting or hauling. (a) Seabird catch rate
as a seasonal effect conditioned on the other 3 factors of time of initiating setting, location of initiating sets, and pre- and postregulation period. (b) Set time effect in the model. (c) Two-dimensional spatial (setting location) effect of catch rate; the US Exclusive Economic Zone seaward boundary is shown (solid black lines). (d) Pre- and post-regulation period effect conditioned on the
other 3 covariates. In (a) and (d) solid bars = mean, dashed bars = 95% confidence interval, y-axis = centered response scale. In (b)
solid curves = model fit, dashed curves = 95% pointwise confidence bands. Reference levels are centered at zero and are as
follows: (a) the first quarter of the year (January to March), (d) the pre-regulation period
period catch rate (Fig. 4d), based on the Poisson GAM
conditioned for the 3 factors. Side-setting with 45 g
weights located within 1 m of the hook resulted in a
seabird catch rate 40% (95% CI: 28 to 58) lower than the
pre-regulation seabird catch rate. No seabirds were
caught in sets employing the combination of side-setting
with 60 g weights located within 1 m of the hook. Stern
setting with 45 g weights located within 1 m of the hook
resulted in a seabird catch rate 60% (95% CI: 44 to 82)
lower, and stern setting with 60 g weights located within
1 m of the hook 41% (95% CI: 27 to 62) lower than the
pre-regulation seabird catch rate. There was no significant difference in seabird catch rates between the 3
categories of sets where birds were caught (Fig. 4d). The
4 covariates or factors — time of starting setting opera-
tions, season in which a set was made, location at the
start of sets, and whether or not bait was treated (dyed
blue and thawed) — were found to have significant effects on seabird bycatch. Timing of initiating setting was
a significantly nonlinear effect. Fig. 4a shows that
seabird catch rates were lowest during October to December, and that the seabird catch rate in the 4th quarter
was significantly lower than in the other 3 quarters.
Seabird catch rates were lowest during sets initiated at
0:00 h, increased as time of setting advanced through ca.
04:30 h and remained at that level for start times through
mid-afternoon (Fig. 4b). As observed in Fig. 3c, higher
seabird catch rates occurred around the main Hawaiian
Islands and the highest rates were in the northwestern
sector ca. 25° N, 170° W (Fig. 4c). Sets employing blue-
Endang Species Res 5: 309–323, 2008
316
Table 2. Four-dimensional contingency table providing summary statistics of seabird capture rates based on a binomial estimator
with Clopper-Pearson confidence intervals (Agresti 2002) for Hawaii longline tuna sets from 15 August 2003 to 4 September 2007
Bird bycatch
reduction
method
Weight within
1 m of hook (g)
Season
Time of
initiating set
(h)
No. of
sets
No. of
hooks
Seabirds
captured
(hauled
aboard)
Side-setting
45
Jan–Jun
≤7:00
> 7:00
≤7:00
> 7:00
≤7:00
> 7:00
≤7:00
> 7:00
25
114
13
102
28
63
8
15
47 108
244 063
31 346
218 476
60 659
138 200
19 074
34 996
1
10
0
1
0
0
0
0
0.021
0.041
0.000
0.005
0.000
0.000
0.000
0.000
0.001–0.118
0.020–0.075
0.000–0.118
0.0001–0.026
0.000–0.061
0.000–0.027
0.000–0.193
0.000–0.105
368
793 922
12
0.015
0.008–0.026
196
556
51
420
45
215
24
126
413 905
1 169 204
110 073
905 543
86 891
440 904
48 602
267 512
2
25
0
6
0
7
0
0
0.005
0.021
0.000
0.007
0.000
0.016
0.000
0.000
0.001–0.017
0.014–0.032
0.000–0.034
0.002–0.014
0.000–0.042
0.006–0.033
0.000–0.076
0.000–0.014
1633
3 442 634
40
0.012
0.008–0.016
Jul–Dec
60
Jan–Jun
Jul–Dec
Total
Stern-setting
45
Jan–Jun
Jul–Dec
60
Jan–Jun
Jul–Dec
≤7:00
> 7:00
≤7:00
> 7:00
≤7:00
> 7:00
≤7:00
> 7:00
Total
dyed and thawed bait had a seabird catch rate 22%
(95% CI: 15 to 31) lower than sets using untreated bait;
the difference was statistically significant (Fig. 4e). The
model was a reasonable fit to the large data set, accounting for ca. 45.4% of the model deviance.
Based on a Poisson GAM model fit to 2 categories of
sets made during the post-regulation period of those
made from the side vs. the stern of the vessel, conditioned on the factors of time of starting setting, season,
location at the start of sets, branch line weighting, and
whether or not bait was thawed and dyed blue, there
was no significant difference in seabird bycatch rates
between side- vs. stern-setting at the 95% confidence
level (p = 0.14), but there was a significant difference
at the 85% level (p < 0.15). Side-setting resulted in
seabird catch rate 21% (95% CI: –8 to 42) lower than
stern-setting.
There was a significant difference in seabird catch
rates between sets made during the post-regulation
period with 45 g weights located within 1 m of the
hook and sets with 60 g weights within 1 m of the
hook, when employing a Poisson GAM model fit to
sets employing 45 vs. 60 g weights, conditioned on
the factors of time of starting setting, season, geolocation of the start of sets, side- vs. stern-setting,
and whether or not bait was thawed and dyed blue
(p < 0.01). Sets with 60 g weights resulted in a seabird
catch rate 63% (95% CI: 45 to 88) lower than sets
with 45 g weights.
Seabird bycatch rate
(per 1000 hooks)
Point
95% CI
estimate
Of the 2001 sets in this study component, 883 sets
(44% of the sample) were initiated south of 23° N
where either an albatross was observed to be present
during setting or hauling and/or a seabird was captured. One or more of the seabird avoidance methods
were employed during these 883 sets. Side-setting was
employed in 131 sets, blue-dyed bait was used in 44
sets, and weights of 45 g or more were used in 855 of
the sets (no branch line weights were used in 28 sets).
In the 869 sets employing weights, weights were
attached to branch lines within 1 m of the hook in all
but 6 of the sets, in 55 sets offal was discarded on the
side of the vessel opposite to that where the sets were
made, a tori (bird scaring) line was deployed during 13
sets, a towed buoy was deployed during 9 sets, and 1
set was made at night.
DISCUSSION
Seabird bycatch rates
A Poisson GAM, conditioned on time of day of setting, season and location of setting predicted that the
seabird capture rate declined significantly by 67% following the introduction of seabird regulations. By explicitly accounting for these covariates and factors, this
modeling approach provided a strong inference of the
effect of regulatory measures involving changes in fish-
Gilman et al.: Reducing seabird longline bycatch
317
Fig. 4. Nonparametric Poisson regression model fitted to the seabird catch in sets made by the Hawaii longline tuna fishery during the period prior to seabird avoidance regulations coming into effect (n = 702 sets) and for 4 categories of seabird bycatch
avoidance methods (see ‘Methods; Alternative combinations of seabird avoidance methods’) employed by the Hawaii longline
tuna fishery during the post-regulations period (n = 254, 114, 1223, and 410 sets in the order displayed), for sets with seabird captures observed during gear hauling and/or albatrosses present during setting or hauling. (a) Seabird catch rate as a seasonal effect conditioned on the other 4 factors time of initiating setting, location of initiating sets, 5 categories of sets, and bait treatment
(untreated vs. dyed blue and thawed). (b) Set time effect in the model. (c) Two-dimensional spatial (setting location) effect on
catch rate; the US Exclusive Economic Zone seaward boundary is shown. (d) Effect of the 5 categories of sets conditioned on the
other covariates. Pre = pre-regulations period; Po1 = side-setting with 45 g weights located within 1 m of the hook; Po2 = sidesetting with 60 g weights located within 1 m of the hook (note, the y-value of –132.9 puts this category off the scale of the figure);
Po3 = stern-setting with 45 g weights located within 1 m of the hook; Po4 = stern-setting with 60 g weights located within 1 m of
the hook. (e) Bait treatment effect in the model: N = no (untreated), Y = yes (dyed blue and thawed). In (b) solid curves = model fit,
dashed curves = 95% pointwise confidence bands. In (a,d,e) solid bars = mean, dashed bars = 95% confidence interval, y-axis =
centered response scale. Reference levels are centered at zero and are as follows: (a) the first quarter of the year (January to
March), (d) the pre-regulation period, (e) no bait treatment
318
Endang Species Res 5: 309–323, 2008
ing gear and fishing methods. The actual, observed
change in seabird bycatch rates from pre- to postregulation periods was a significant 74% reduction
(Table 1). Results demonstrate that there are several effective methods for seabird avoidance in pelagic longline fisheries. After the regulations came into effect,
sets employing 4 different combinations of seabird bycatch reduction methods all reduced seabird catch rates
by more than 40% relative to the pre-regulation rate.
Results suggest that substantial seabird bycatch
reduction is realized by the use of 60 g weights, bluedyed and thawed bait, and perhaps side-setting, which
are components of current regulations. The timing of
setting and temporal and spatial distribution of effort
also significantly affected seabird catch rates, and represent potential additional strategies for further reductions in seabird bycatch in the Hawaii longline tuna
fishery. Additional assessment is needed to determine
to what degree bycatch would likely be reduced by
alternative, additional management measures involving restrictions on fishing gear and methods or temporal and spatial restrictions on fishing effort, and at what
cost.
To provide a rough understanding of potential
seabird conservation benefits, restricting the initiating
of sets to 7:00 h or earlier (≤7:00 h) results in a seabird
catch rate that is 69% (and significantly) lower than
when sets are made after 7:00 h (>7:00 h): sets initiated ≤7:00 h resulted in a seabird catch rate of 0.011
seabirds per 1000 hooks (95% CI: 0.007 to 0.017),
while sets initiated after 7.00 h resulted in a seabird
catch rate of 0.036 (95% CI: 0.031 to 0.041) (based on
a binomial estimator with Clopper-Pearson confidence
intervals) (Table 1, Fig. 3b). It is likely that avoiding
setting during local dawn and dusk periods, when
albatrosses most actively forage, is key (Brothers et al.
1999). Similarly, sets made during the first half of the
year resulted in a seabird catch rate an order of magnitude higher than those in the latter half of the year:
In the first half of the year, the seabird catch rate of
0.045 (95% CI: 0.039 to 0.050) was 91% higher than
during the second half of the year, when it was 0.0040
(95% CI: 0.002 to 0.007) (Table 1, Fig. 3a). Sets made
north of 24° N and west of 170° W had a seabird catch
rate of 0.84 birds per 1000 hooks (95% CI: 0.62 to
1.11, only 28 sets), an order of magnitude and significantly higher than the bird catch rate of sets made in
the rest of the fishing grounds (0.025 birds per 1000
hooks [95%CI: 0.022 to 0.028], Table 1, Fig. 3d).
These 3 examples do not explicitly account for other
factors that affect seabird bycatch rates. For example,
the 28 sets made in the alleged seabird bycatch
hotspot might have included an outlier (a single set
with a high number of bird captures), or the high
seabird catch rate could have been due to some
aspect of the fishing gear or methods employed in
these sets. These rough assessments strengthen the
finding that timing of setting and temporal and spatial
distribution of effort significantly affected seabird
catch rates, and present a potential management tool
for further seabird bycatch reductions which warrant
additional assessment.
Experimental vs. commercial conditions
Despite the observed efficacy of a bycatch avoidance
method during experimental conditions, fishers may
not employ the method as prescribed or, indeed, at all
if it is not convenient and economically viable and if
incentives, including enforcement, are insufficient
(Gilman et al. 2005, Cox et al. 2007). In the present
study we observed smaller reductions in bird catch
rates due to side-setting and blue-dyed bait (21 and
22% reductions, respectively) than found in experiments testing the single factor effects of these seabird
bycatch reduction methods. Two experiments in the
Hawaii longline fishery found the single factor effect of
employing blue-dyed fish bait reduced seabird captures by 63 and 95%, and one experiment found that
side-setting eliminated seabird captures (McNamara
et al. 1999, Gilman et al. 2005, 2007a). The smaller
reductions in seabird catch rates observed here relative to the experiments may be because these 2 seabird
bycatch reduction methods allowed for deviation from
their experimental and prescribed employment (e.g.
crew may not have thrown baited hooks as close to the
vessel hull or as far forward when side-setting as in the
experiment, crew may not have completely thawed
bait, soaked the bait in dye for sufficient time, or used
the prescribed concentration of blue dye as in the
experiments).
The observed 63% lower seabird catch rate for sets
employing 60 g weights relative to those using 45 g is
larger than expected, given that the difference in the
baited hook sink rates between these 2 gear designs
is only about 0.1 m s–1 (Brothers & Gilman 2007)3.
Because line weighting is a seabird bycatch reduction
strategy that does not result in differences in employment as a result of crew behavior, we would expect
bird catch rate reductions to be consistent with those
found in controlled experiments. Boggs (2001) found a
92% reduction in seabird contacts with fishing gear in
the Hawaii longline swordfish fishery as a result of
3
The mean baited hook sink rate of branch lines containing a
45 g swivel at 300 mm from a tuna 3.6 hook was observed to
be 1.2 m s–1. This increased to 1.3 m s–1 for a branch line
containing a 60 g swivel attached at 300 mm from a tuna 3.6
hook (Brothers & Gilman 2007)
Gilman et al.: Reducing seabird longline bycatch
adding a 60 g weight at the hook, but did not determine the effect on seabird catch rate; thus, his results
are of limited utility for comparison with the observed
effect of line weighting reported here.
Benefits and limitations of studies of observer data
The use of observer data to compare catch rates from
a marine capture fishery is inherently limited in not
allowing for the control of confounding factors. Gilman
et al. (2005) observed that, when comparing seabird
bycatch rates for avoidance methods from 2 different
experiments, even when normalized for seabird abundance, the combined effect from numerous variables
can result in significantly different seabird bycatch
rates even for the same treatment used in different
experiments. Because results in this and most other
studies were not reported as normalized for seabird
abundance, this alone substantially reduces the ability
to meaningfully compare results between studies
(Gilman et al. 2003, 2005). Albatross abundance is
likely to change by area for each season of different
years, as albatross at-sea abundance may be correlated with the proximity to breeding colonies during
the breeding season as well as the location of largescale oceanographic features and short-lived hydrographic features such as eddies and fronts (Hyrenbach
et al. 2000). In addition to seabird abundance during
fishing operations, variability may also result from several additional factors not explicitly accounted for in
our models, including fishing gear and methods (e.g.
use of deck lighting at night, offal discharge practices,
type and condition of bait, length of branch lines, size
and type of hooks, crew practices for deploying branch
lines), environmental parameters (e.g. weather, bird
behavior, bird species complex), and consistency in
observers’ methods and fishers’ behavior (Brothers
1991, Brothers et al. 1999, Gilman 2001). For example,
with seabird regulations in effect, fishers may have an
increased incentive to conceal caught seabirds from
onboard observers, for example, by dropping branch
lines containing caught seabirds before an observer
notices the bycatch. This is a documented problem in
some fisheries with seabird bycatch management measures (Gales et al. 1998, Gilman et al. 2005). Seabird
foraging behavior around fishing vessels may also be
influenced by variations in regional climate, such as
phases of the El Niño Southern Oscillation (during less
productive La Niña phases, seabirds may be more
abundant and aggressive around fishing vessels), and
may further be variable depending on the seabird species complex in the vicinity of a fishing vessel.
While controlled and comparative studies permit
drawing more definitive conclusions on causality of
319
differences in catch rates between treatments, analyses using observer program data typically enable
much larger sample sizes relative to controlled and
comparative experiments. Furthermore, analysis of
observer data enables an assessment of the status and
trends in bycatch rates in the fishery, fundamental
information needed to guide responsible fisheries
management. Such information is not as reliable
when obtained from scientific experiments, because
during these experiments, fishing methods and gear,
and fisher behavior, are not likely characteristic of
normal operations (Gilman et al. 2005, Cox et al.
2007). Observational studies often arise in the area of
medicine, where it may be unethical to employ a control group. Similar ethical, as well as ecological, issues
arise when dealing with endangered and threatened
species when it behooves researchers to consider
alternative study designs to controlled experiments,
including comparative studies and observer data
analyses.
Seabird abundance during setting
During the study period (9 May 2000 to 4 September
2007), observers did not record albatross presence during setting or hauling for 7% of sets with observed
albatross captures (13 of 181 sets), indicating that the
current protocol for recording seabird abundance
could be improved. Gilman et al. (2003, 2007a) employed a standardized method to estimate mean albatross abundance during setting: Every 15 min throughout each set a count of each seabird species within a
500 by 500 m square area astern of the vessel was
recorded (within 250 m of port and starboard of the
center of the vessel stern and within 500 m behind the
vessel). The Hawaii longline observer program should
define a similar area around the vessel for observers to
improve consistency of measurements of mean seabird
abundance during sets. More importantly, increasing
the number of seabird abundance estimates during
setting operations, and collecting data throughout the
set, and not just during the initial 30 min, would enable
a better characterization of seabird abundance for the
entire set. Currently, observers estimate seabird abundance primarily during gear hauling. Observations of
seabird abundance during hauling likely provide an
inaccurate characterization of seabird abundance during the period when birds are being captured because
(1) seabird captures occur primarily during setting in
this fishery (Gilman et al. 2003, 2005), (2) albatross
abundance is generally lower at night than during the
day, and (3) it is difficult to accurately estimate bird
abundance around the vessel in the dark (McNamara
et al. 1999).
320
Endang Species Res 5: 309–323, 2008
However, because observers need to observe each
haulback in full in order to record the number of
seabirds captured, interactions with other protected
species (sea turtles and marine mammals), handle and
release any protected species brought to the vessel
alive during the haul, and record other fundamental
information, an additional requirement for an observer
to also watch entire setting operations would leave
insufficient time to sleep and eat. However, it may be
feasible for observers to record albatross abundance
during the first and last hour of each set, which would
better characterize seabird abundance during setting
than the current method.
side of the setting operation (6%), blue-dyed bait (5%),
tori line (1%), towed buoy (1%), and night setting
(0.3%).
These results are consistent with the findings of
Gilman et al. (2005, 2007a) that side-setting presented
several operational benefits to Hawaii longline vessel
crew, while blue-dyed bait was reported by crew to be
impractical for several reasons. This suggests that compliance with a required employment of side-setting is
likely to be higher than for other seabird avoidance
methods, even when an observer is not present and in
fisheries lacking resources for a high degree of surveillance and enforcement.
Fishing grounds where seabird bycatch is
problematic
CONCLUSIONS
A large proportion of albatross interactions with the
Hawaii longline tuna fishery occurred south of 23° N,
the southern boundary ofthe area for required employment of prescribed seabird avoidance methods by
Hawaii longline tuna vessels. Management authorities
originally selected this boundary to reduce the risk of
interactions with the short-tailed albatross Phoebastri
albatrus (US Fish and Wildlife Service 2002, 2004, US
National Marine Fisheries Service 2005). However, the
stated purpose of current regulations is to reduce interactions with all seabird species, not just the listed
endangered short-tailed albatross (US Western Pacific
Fishery Management Council 2004, US National Marine Fisheries Service 2005). Based on observations of
where the fleet catches seabirds, to more effectively
minimize seabird bycatch rates in the Hawaii longline
tuna fishery, fishery management authorities should
consider moving the boundary for the prescribed use
of seabird avoidance measures farther south.
Voluntary use of seabird avoidance strategies
Of the seabird avoidance methods voluntarily
employed by the Hawaii longline tuna vessels when
fishing at grounds where seabird avoidance methods
are not required, besides the conventional practice of
attaching weighted swivels near the hook, fishers
demonstrated the most frequent voluntary use of side
setting. The analysis of the sets included in this analysis for the employment of seabird avoidance methods
at fishing grounds where bird avoidance methods are
not required showed that fishers frequently voluntarily
attach weights of 45 g or more within 1 m of the hook
(92% of sets). Of the other seabird avoidance methods,
side-setting was employed the most frequently (15%
of sets), followed by discarding offal on the opposite
Catch rates of seabirds were significantly reduced
after seabird protection regulations came into effect in
the Hawaii longline tuna fishery. This study documented the efficacy of different combinations of commercially employed seabird bycatch reduction methods, demonstrating that there are several extremely
effective methods for seabird avoidance in pelagic
longline fisheries. Findings indicate that in the Hawaii
longline tuna fishery, using heaver branch line weights
(in particular, 60 instead of 45 g weights within 1 m
from the hook), blue-dyed and thawed bait instead of
untreated bait, and possibly side-setting instead of
stern-setting are effective methods, involving differences in fishing gear and practices, to achieve large
and significant reductions in seabird catch rates. Furthermore, time-area closures and restrictions on the
timing of setting could further reduce seabird bycatch,
as these factors were observed to have significant
effects on seabird catch rates.
The efficacy of a bycatch avoidance method observed during experimental conditions may not be
achieved during commercial fishing operations because the methods in the experiments may not be
employed as prescribed or may not be used at all by
fishers if they are not convenient and economically
viable and if incentives, including enforcement, are
insufficient (Gilman et al. 2005, Cox et al. 2007). This
may be the case with blue-dyed and thawed bait and
side-setting. However, results suggest that, at least
when an observer is onboard, the Hawaii longline tuna
fleet is employing seabird avoidance measures and
this has resulted in large and significant reductions in
seabird catch rates relative to the pre-regulations
period, even when fishing at grounds where the
employment of these measures are not mandatory.
Consistent with previous results of commercial demonstrations in this fishery (Gilman et al. 2007a), sidesetting, a conventional practice, was the second most
Gilman et al.: Reducing seabird longline bycatch
common measure (after branch line weighting) voluntarily employed by vessels on fishing grounds where
seabird avoidance methods were not required. This
suggests that compliance with required employment of
side-setting is likely to be higher than other seabird
bycatch reduction methods.
Efforts by the Hawaii longline fleet alone to reduce
seabird bycatch will not reverse North Pacific albatross
population decline. The Hawaii longline fleet is a very
small component of the total longline fishing effort in
the North Pacific, representing less than 3% of total
longline hooks deployed in the Pacific Ocean each
year (Majkowski 2007). Of the 61 species of seabird
affected by longline fisheries, 26 are threatened with
extinction, including 19 species of albatrosses, among
them the Laysan and black-footed albatrosses, and
there is compelling evidence that longline mortality is
a significant component in the decline of many of these
species (Gales et al. 1998, Brothers et al. 1999, Lewison
& Crowder 2003, Niel & Lebreton 2005).
The seabird avoidance methods found to be effective
in the Hawaii fishery may likewise be effective in other
longline fisheries. However, different seabird avoidance methods may be appropriate for different longline fisheries due to differences in the diving abilities
of seabird species that interact with each fishery, vessel designs, and fishing gear and methods (Brothers et
al. 1999, Gilman et al. 2005). In particular, the very rare
occurrence of interactions with deep-diving species of
seabirds in the Hawaii fishery, and use of relatively
large weights proximate to the hook, are important differences that need to be taken into account when considering the applicability of results from this study to
other fleets. Trials in individual fisheries must precede
advocacy for the introduction of specific seabird avoidance methods.
Despite the availability of effective avoidance methods that also increase fishing efficiency, most longline
fleets do not employ effective seabird avoidance
methods (Brothers et al. 1999, Gilman et al. 2005).
Some Regional Fisheries Management Organizations
(RFMOs) have recently made progress: 5 have adopted
legally binding conservation measures related to
reducing seabird bycatch in pelagic and demersal
longline and trawl fisheries (Gilman et al. 2007b).
However, these RFMO seabird conservation measures
need to be improved. For instance, the areas where
some of these measures are required do not include
higher latitude fishing grounds, where seabird interactions have been observed to be problematic. The measure adopted by the Western and Central Pacific Fisheries Commission does not require vessels < 24 m in
length to employ seabird avoidance measures in areas
north of 23° N; however, the present and previous
studies have documented high seabird bycatch rates
321
by vessels in this size category in this area. Furthermore, compliance by many member states with these
RFMO seabird conservation measures is likely low, as
observer programs and national management frameworks are generally weak or nonexistent, preventing
definitive assessments. For example, 40 nations worldwide are engaged in longline fishing, of which only 15
have observer programs (Beverly & Chapman 2007).
There have been few voluntary longline industry
actions to address seabird bycatch (Gilman & Lundin
2008).
Recognizing this context of global longline fisheries
management, it is necessary to maximize industry’s
sense of ownership for effective seabird avoidance
measures and to provide industry with economic and
other incentives for voluntary compliance. Identifying
measures that are practical and convenient for the crew
as well as being economically viable — or better yet,
that provide operational and economic advantages —
will maximize the chance that effective bycatch reduction methods will be used by the fishing industry
(Gilman et al. 2005). Scientifically rigorous eco-labeling
and other certification programs for marine capture
fisheries, and adoption of suitable sustainable seafood
sourcing policies by retailers and seafood buyers, provide large market-based and social incentives for some
fisheries to meet sustainability criteria (e.g. Alaska
Seafood Marketing Institute 2001, Johnston et al. 2001,
FAO 2008). These emerging market-based incentives
may eventually become the most effective factor for the
tuna industry to improve the sustainability of their practices (IUCN and Western Pacific Fishery Management
Council 2008).
Acknowledgements. Funding for this study was kindly provided by the US National Marine Fisheries Service, Pacific
Islands Regional Office. Brent Miyamoto, US National Marine
Fisheries Service, Pacific Islands Fisheries Science Center,
extracted observer data for the Hawaii longline fishery, provided insightful comments and suggestions for the design of
data queries, and provided helpful comments on a draft of this
report. Stuart ‘Joe’ Arceneaux, Kevin Busscher, and Lewis
Van Fossen, US National Marine Fisheries Service, Pacific
Islands Regional Office, provided information on observer
data collection protocols and offered constructive comments
on drafts of the manuscript. Evan Howell and Dean Courtney,
US National Marine Fisheries Service, Pacific Islands Fisheries Science Center, 3 anonymous reviewers, and the Theme
Section editor provided helpful comments on a draft manuscript.
LITERATURE CITED
Agresti A (2002) Categorical data analysis. Wiley, New York
Alaska Seafood Marketing Institute (2001) Whole foods stores
carry Alaska salmon with eco-label. Marketline (June): 1
Beverly S, Chapman L (2007) Interactions between sea turtles
and pelagic longline fisheries. Western and Central Pacific
322
➤
➤
➤
➤
Endang Species Res 5: 309–323, 2008
Fisheries Commission, Scientific Committee, 3rd Regular
Session, 13–24 August 2007, Honolulu. Information Paper
WCPFC-SC3-EB SWG/IP-01. Western and Central Pacific
Fisheries Commission, Palikir
Beverly S, Chapman L, Sokimi W (2003) Horizontal longline
fishing methods and techniques: a manual for fishermen.
Secretariat of the Pacific Community, Noumea
Boggs CH (2001) Deterring albatrosses from contacting baits
during swordfish longline sets. In: Melvin E, Parrish K
(eds) Seabird bycatch: trends, roadblocks, and solutions.
University of Alaska Sea Grant AK-SG-01-01, Fairbanks,
AK, p 79–94
Brothers NP (1991) Albatross mortality and associated bait
loss in the Japanese longline fishery in the southern
ocean. Biol Conserv 55:255–268
Brothers N, Gilman E (2007) Technical assistance for Hawaii
pelagic longline vessels to change deck design and fishing
practices to side set. Hawaii Longline Association, US
NOAA Fisheries Pacific Islands Fisheries Science Center
and Pacific Islands Regional Office, and Western Pacific
Regional Fishery Management Council, Honolulu, HI.
Available at: www.wpcouncil.org/pelagic/Documents/
Exec_sum_Side_set_tech_assist_HI.pdf
Brothers NP, Cooper J, Lokkeborg S (1999) The incidental
catch of seabirds by longline fisheries: worldwide review
and technical guidelines for mitigation. FAO Fish Circ
937. Food and Agriculture Organization of the United
Nations, Rome
Cousins K, Cooper J (eds) (2000) The population biology of the
black-footed albatross in relation to mortality caused by
longline fishing. Report of a workshop held in Honolulu,
Hawaii, 8–10 October 1998 by the US Western Pacific Regional Fishery Management Council, Honolulu, HI
Cox TM, Lewison RL, Zydelis R, Crowder L, Safina C, Read J
(2007) Comparing effectiveness of experimental and
implemented bycatch reduction measures: the ideal and
the real. Conserv Biol 21:1155–1164
Fahrmeir L, Lang S (2001) Bayesian inference for generalised
additive mixed models based on Markov random field priors. Appl Stat 50:201–220
FAO (2008) Ecolabels and marine capture fisheries: current
practices and emerging issues. Globefish Research Programme, Vol 91, Food and Agriculture Organization of the
United Nations, Rome
Gales R (1998) Albatross populations: status and threats. In:
Robertson G, Gales R (eds) Albatross biology and conservation. Surrey Beatty and Sons, Chipping Norton,
p 20–45
Gales R, Brothers N, Reid T (1998) Seabird mortality in the
Japanese tuna longline fishery around Australia, 1988–
1995. Biol Conserv 86:37–56
Gilman E (2001) Integrated management approach to address
incidental mortality of seabirds in longline fisheries. Aquat
Conserv Mar Freshw Ecosyst 11:391–414
Gilman E (2008) USA Western and Central Pacific Ocean. In:
DeYoung C (ed) State of World Marine Capture Fisheries
Management – 2006: Pacific Ocean. FAO Fish Tech Pap
488/1. Food and Agriculture Organization of the United
Nations, Rome
Gilman E, Freifeld H (2003) Seabird mortality in north Pacific
longline fisheries. Endang Spec Update 20:35–46
Gilman E, Lundin C (2008) Minimizing bycatch of sensitive
species groups in marine capture fisheries: lessons from
commercial tuna fisheries. In: Grafton Q, Hillborn R,
Squires D, Tait M, Williams M (eds) Handbook of marine
fisheries conservation and management. Oxford University Press, Oxford
➤ Gilman
➤
➤
➤
➤
➤
➤
➤
E, Boggs C, Brothers N (2003) Performance assessment of an underwater setting chute to mitigate seabird
bycatch in the Hawaii pelagic longline tuna fishery. Ocean
Coast Manag 46:985–1010
Gilman E, Brothers N, Kobayashi D (2005) Principles and
approaches to abate seabird bycatch in longline fisheries.
Fish Fish 6:35–49
Gilman E, Brothers N, Kobayashi D (2007a) Comparison of
the efficacy of three seabird bycatch avoidance methods
in Hawaii pelagic longline fisheries. Fish Sci 73:208–210
Gilman E, Moth-Poulsen T, Bianchi G (2007b) Review of
measures taken by inter-governmental organizations to
address problematic sea turtle and seabird interactions in
marine capture fisheries. FAO Fish Circ 1025. Food and
Agriculture Organization of the United Nations, Rome
Gilman E, Clarke S, Brothers N, Alfaro-Shigueto J and others
(2008) Shark interactions in pelagic longline fisheries. Mar
Pol 32:1–18
Hastie T, Tibshirani R (1990) Generalized additive models.
Monogr Stat Appl Probab 43. Chapman & Hall, London
Hyrenbach KD, Forney KA, Dayton PK (2000) Marine protected areas and ocean basin management. Aquat Conserv Mar Freshw Ecosyst 10:437–458
Ihaka R, Gentleman R (1996) R: a language for data analysis
and graphics. J Comput Graph Statist 5:299–314
IUCN and Western Pacific Fishery Management Council
(2008) Sustainable Tuna Roundtable, 2008 Meeting
Report, April 21, 2008, Brussels. International Union
for the Conservation of Nature (IUCN), Gland and
Western Pacific Regional Fishery Management Council,
Honolulu, HI
Johnston RJ, Wessells C, Donath H, Asche F (2001) A contingent choice analysis of ecolabeled seafood: Comparing
consumer preferences in the United States and Norway.
J Agric Resour Econ 26:20–39
Lewison R, Crowder L (2003) Estimating fishery bycatch and
effects on a vulnerable seabird population. Ecol Appl
13:743–753
Majkowski J (2007) Global fishery resources of tuna and
tuna-like species. Food and Agriculture Organization of
the United Nations, Rome
McCullagh P, Nelder JA (1989) Generalized linear models,
2nd edn. Monogr Stat Appl Prob 37. Chapman & Hall,
London
McNamara B, Torre L, Kaaialii G (1999) Hawaii longline
seabird mortality mitigation project. Western Pacific
Regional Fishery Management Council, Honolulu, HI
Minami H, Kiyota M (2002) Effect of blue-dyed bait on reduction of incidental take of seabirds. Fisheries Research
Agency, Shizuoka
Niel C, Lebreton J (2005) Using demographic invariants to
detect overharvested bird populations from incomplete
data. Conserv Biol 19:826–835
Sullivan BJ, Brickle P, Reid TA, Bone DG, Middleton DAJ
(2006) Mitigation of seabird mortality on factory trawlers:
trials of three devices to reduce warp cable strikes. Polar
Biol 29:745–753
US Fish and Wildlife Service (2002) Revision of the US Fish
and Wildlife Service’s November 28, 2000 biological opinion for the effects of the Hawaii-based domestic longline
fleet on the short-tailed albatross (Phoebastria albatrus).
US Fish and Wildlife Service, Honolulu, HI
US Fish and Wildlife Service (2004) Biological opinion on
the effects of the reopened shallow-set sector of the
Hawaii-based longline fishery on the short-tailed albatross (Phoebastria albatrus). US Fish and Wildlife Service,
Honolulu, HI
Gilman et al.: Reducing seabird longline bycatch
323
US National Marine Fisheries Service (2005) Fisheries off
west coast states and in the western Pacific; pelagic fisheries; additional measures to reduce the incidental catch
of seabirds in the Hawaii pelagic longline fishery. Fed
Regist 70:75075–75080
US National Marine Fisheries Service (2008) The Hawaiibased longline logbook summary report, October–
December 2007. Data Report DR-08-002, Pacific Islands
Fisheries Science Center, National Marine Fisheries Service, Honolulu, HI
US Western Pacific Fishery Management Council (2004)
Additional measures to reduce the incidental catch of
seabirds in the Hawaii-based longline fishery. A Regulatory Amendment to the Western Pacific Pelagic Fisheries
Management Plan. US Western Pacific Fishery Management Council, Honolulu, HI
Van Fossen L (2007) Annual report on seabird interactions
and mitigation efforts in the Hawaii longline fishery for
2006. National Marine Fisheries Service, Pacific Islands
Regional Office, Honolulu, HI
Wood S (2006) Generalized additive models: an introduction
with R. Chapman & Hall, London
Editorial responsibility: Rebecca Lewison,
San Diego, California, USA
Submitted: October 19, 2007; Accepted: August 19, 2008
Proofs received from author(s): October 18, 2008