Content uploaded by John D. Filmalter
Author content
All content in this area was uploaded by John D. Filmalter on Sep 13, 2018
Content may be subject to copyright.
291
© The Ecological Society of America www.frontiersinecology.org
Human-induced impacts on marine populations are
typically thought of in terms of the harvesting of
resources, which is generally regarded as the largest
direct anthropogenic effect on marine ecosystems.
However, other more complex issues – including global
warming (Schmittner 2005; Cheung et al. 2010), ocean
acidification (Caldeira and Wickett 2003), trophic cas-
cades due to the removal of entire ecosystem tiers
(Jackson et al. 2001; Myers et al. 2007), marine pollu-
tion (Halpern et al. 2008), and technological shifts in
fishing practices – may ultimately have impacts equal-
ing or even surpassing those caused by ongoing harvest-
related exploitation; impacts associated with these
processes are harder to quantify because they are not
easily observable or go unobserved at short timescales.
The current methodology used to assess marine popula-
tion health often excludes such processes and instead
relies almost entirely on a single parameter, the amount
of biomass extracted, which has been observed and
quantified for decades. Although the impacts of these
processes are often complex and difficult to measure and
remedy, their mitigation in some cases may be relatively
simple. Here we quantify and provide solutions for a
previously unknown impact of a technological shift in
the global tuna (Scombridae) purse-seine fishery, repre-
senting an extensive new source of mortality for a
pelagic shark species already designated as Near
Threatened by the International Union for
Conservation of Nature (Bonfil et al. 2009).
During the past 20 years, the tropical tuna purse-seine
fishery, operating throughout oceans worldwide, has
changed its typical fishing practice. Traditionally, tuna
schools were caught when feeding at the surface or when
associated with marine mammals (Scott et al. 2012) or
drifting logs (Freon and Dagorn 2000). More recently,
the use of artificial fish aggregating devices (FADs) has
become widespread. FADs work by taking advantage of
the propensity of tropical tunas to aggregate around
floating objects (Parrish and Edelstein-Keshet 1999).
Once deployed, a FAD is left to drift freely in the open
ocean for several months, with its spatial location moni-
tored remotely via a satellite-tracked buoy (Dagorn et al.
2012). The FADs are then revisited by fishing vessels
and the aggregated tuna and associated bycatch species
captured. This fishery enhancement tool now accounts
for >40% of all of the world’s annual tropical tuna catch
(4 million tons; Miyake et al. 2010). These FADs usually
consist of bamboo poles bound with old netting, which
extends to varying depths below the water’s surface. The
subsurface structure of a FAD is believed to aid in the
attraction of small fish and serves to increase drag, ensur-
ing ocean currents rather than wind drive the direction
of its drift. These functions are important to fishers, who
consider them essential for the formation of tuna aggre-
gations. This netting can entangle marine animals, but
because of the difficulty involved in observing such mor-
tality events, they have largely been ignored by marine
scientists and resource managers. The popularity of
FADs in tuna purse-seine fleets has led to global con-
cerns over the increased capture of undesirable sized
tunas and bycatch, which include vulnerable pelagic
RESEARCH COMMUNICATIONS RESEARCH COMMUNICATIONS
Looking behind the curtain: quantifying
massive shark mortality in fish aggregating
devices
John David Filmalter1,2,3*, Manuela Capello4, Jean-Louis Deneubourg4, Paul Denfer Cowley2, and Laurent Dagorn1
Increasing catch rates are considered the main impact of dynamic fisheries practices on marine ecosystems,
but other effects can be equally important and are often ignored. Here we quantify a major, previously
unknown source of shark mortality: entanglement in drifting fish aggregating devices, now widely used in
the global tropical tuna purse-seine fishery. Using satellite tagging and underwater observational data, we
developed two novel, independent, and complementary approaches, which quantify and highlight the
scale of this problem. Entanglement mortality of silky sharks (Carcharhinus falciformis) in the Indian Ocean
was 5–10 times that of the known bycatch of this imperiled species from the region’s purse-seine fleet. More
importantly, these estimates from a single ocean (480 000–960 000 silky sharks) mirror those from all world
fisheries combined (400 000–2 million silky sharks), a situation that clearly requires immediate manage-
ment intervention and extensive monitoring.
Front Ecol Environ 2013; 11(6): 291–296, doi:10.1890/130045 (published online 27 Jun 2013)
1Institut de Recherche pour le Développement, Sète, France *(john-
david.filmalter@ird.fr); 2South African Institute for Aquatic Bio-
diversity, Grahamstown, South Africa; 3Department of Ichthyology and
Fisheries Science, Rhodes University, Grahamstown, South Africa;
4Service d’Ecologie Sociale, Université Libre de Bruxelles, Brussels,
Belgium
Quantifying shark mortality in FADs JD Filmalter et al.
292
www.frontiersinecology.org © The Ecological Society of America
shark species (Gilman 2011). The silky shark
(Carcharhinus falciformis) constitutes up to 90% of the
elasmobranch bycatch in this fishery (Gilman 2011).
Localized depletion is a concern in many parts of this
shark’s circumglobal distribution (Bonfil 2008; Anderson
and Jauharee 2009; Bonfil et al. 2009) because it is also
captured in greater numbers by other fishing gears
including longlines and gillnets (Bonfil 2008; Gilman
2011; Hall et al. 2012). Juveniles of this species regularly
associate with drifting objects (Anderson and Jauharee
2009; Filmalter et al. 2011), accounting for their preva-
lence in FAD fishing sets. This behavior also results in
their entanglement in the netting of the FADs them-
selves, a previously unknown source of mortality. The
now widespread use of FADs could pose a major risk to
silky shark populations and requires quantification.
Currently no methods exist for investigating “ghost fish-
ing” on the high seas. This work aims to provide the first
quantitative results for evaluating the scale of this prob-
lem in the Indian Ocean, as well as outlining new exper-
imental and analytical approaches that can be used for
future assessments in other fisheries or oceans.
nMaterials and methods
Satellite tagging: estimating the average time before
entanglement
A total of 43 silky sharks were captured and fitted with
pop-up satellite archival tags (PSATs; product name
“MiniPAT”, Wildlife Computers, Redmond, WA) during
six cruises conducted between 2010 and 2012 offshore of
the Republic of Seychelles and in the northern
Mozambique Channel (Figure 1a). Sharks were caught
either by handline from research vessels (n= 13) or dur-
ing purse-seine operations on commercial vessels (n=
30). Those caught from purse-seine vessels were tagged
to investigate their post-release survival. Tags were
attached either to a single threaded nylon rod secured
through the first dorsal fin (n= 11) or by a nylon anchor,
to which the tag was tethered, which was inserted into
the dorsal musculature at the base of the first dorsal fin
(n= 32). For the analysis presented here, only sharks
that did not display direct post-release mortality (ie
immediately sinking to a depth of >1600 m after release)
were included, which led to a total of 29 individuals
(WebTable 1).
Analysis of the detailed time-series depth profiles
received from the tags clearly indicated that some of the
individuals had succumbed as a result of FAD entangle-
ment (WebPanel 1). The data series revealed an abrupt
cessation of vertical movements followed by a constant
depth reading close to the surface for extended periods
(0.34–2.40 days; WebTable 2). After death by entangle-
ment, the tagged sharks then sank to a depth >1600 m,
causing the tag to automatically release and float to the
surface (Figure 1b).
Using the data received from PSATs, we were able to
estimate the daily probability of a silky shark becoming
entangled in a FAD. Achieving this required that we
consider the observation duration of all tagged sharks,
as this differed between individuals owing to the pre-
mature shedding of many tags. Considering t= 0 as the
tagging time, we denote the total number of sharks
that are still observed at time tas Ntot(t). Because
Ntot(t) varies in time due either to tag shedding or to
Figure
1. (a) Study area showing PSAT tagging and pop-up locations as well as locations of underwater observations at drifting
FADs in the western Indian Ocean. (b) Typical depth profile received from a PSAT on a silky shark that became entangled. (c)
Juvenile silky sharks entangled in the subsurface structure of a drifting FAD.
(a) (b) (c)
0
50
1650
1700
1750
1800
1850
0 1 16 17 18 19
Time (days)
Depth (m)
F Forget, © ISSF
40˚ 50˚ 60˚
10˚
0˚
–10˚
–20˚
JD Filmalter et al. Quantifying shark mortality in FADs
293
© The Ecological Society of America www.frontiersinecology.org
dX0=–µX0+ X1
(Equation 5)
{
dt
dXj=–jXj–µXj+ µX(j–1) + (j+ 1)X(j+1) for j≠ 0
dt
where µ is the probability of a FAD entangling a silky
shark and is the probability of the entangled shark drop-
ping out of the net. To adopt the most parsimonious
model, we considered that the parameters µ and were
constant for all FADs. For any value of entangled sharks
per FAD j, Equation 5 leads to the stationary solutions:
Xj=1 (µ )j
X0
(Equation 6)
j!
Considering that the sum of the number of FADs with j
entangled sharks must equal the total number of observed
FADs (denoted as NFAD), the estimated daily probability
of a FAD entangling a shark µ is expressed as:
µ = – ln (X0) (Equation 7)
NFAD
The factor was obtained from a survival curve analysis of
the observed time individuals spent entangled (dead) in
the net before sinking, from the PSAT dataset (WebFigure
1; WebTable 2). For this purpose, we adopted the most
parsimonious model for survival events – that is, a single
exponential model of the form f(t) = exp(–t) – and fitted
this to the observed entanglement duration data.
Several factors may influence the time an entangled
shark will remain in the netting of a FAD. Many species
of fish – including the oceanic triggerfish (Canthidermis
maculata) and rainbow runner (Elagatis bipinnulata) –
commonly aggregate around FADs in the Indian Ocean.
When a shark becomes entangled, these fish often feed
on its carcass. Some entangled sharks were also observed
with large portions of musculature removed, very likely
by other, larger sharks. As the carcass is broken up due to
this predation, it falls from the net and sinks. In addition,
the manner in which sharks become entangled can vary
greatly, from being strongly meshed behind the gills and
around the head to simply being wrapped up with the net
hooked on the jaw. These and other factors, such as the
prevailing state of the sea, which will influence the jerk-
ing motion of the net, interact to determine how long the
shark’s carcass will remain in the net.
nResults
Four of the 29 sharks tagged became entangled after
4–94 days at liberty (Table 1). The exponential model fit-
ted to the tagging observation duration data produced an
exponent of 0.024 (Figure 2). Following Equation 1, the
average time required for an individual silky shark swim-
ming in an environment with FADs to become entangled
was estimated as 300 ± 45 days. From growth models
entanglement events, we can express its temporal vari-
ation through:
dNtot (t)
dt =–␣eNtot(t)–␣sNtot(t) (Equation 1)
where ␣edenotes the probability of entanglement and ␣s
the probability of tag shedding; both probabilities are
assumed to be time-independent. From Equation 1, we
can express the number of sharks that are still observed at
time tas:
Ntot(t) = Ntot(0)exp[–(␣e+ ␣s)t] (Equation 2)
where Ntot(0) is the total number of sharks released after
tagging. The total number of entangled sharks at time t,
denoted by Ne(t), then reads:
{1–exp[–(␣e+␣s)t]}
Ne(t) = ∫t
0␣eNtot(t)dt = ␣eNtot(0) ␣e+ ␣s
(Equation 3)
that, for large observation times t∞, leads to the following
expression for ␣e:
␣e= Ne(t∞)(␣e+ ␣s) (Equation 4)
Ntot(0)
The value of the exponent (␣e+ ␣s) was obtained
through a survival curve analysis of the number of sharks
still observed at time t(Equation 2), fitting the cumula-
tive distribution of Ntot(t) with an exponential model.
From Equation 4, the average time required for a silky
shark swimming in an environment with FADs to
become entangled was estimated as 1/␣e.
Underwater observations: estimating the daily
probability of a FAD entangling a shark
In addition to the tagging data, underwater observations
were conducted between 2010 and 2012 by divers at 51
FADs with subsurface netting (Figure 1, a and c).
During these observations, the presence and number of
entangled sharks were noted (WebTable 3). Using a
bootstrap resampling method (Efron and Tibshirani
1993) run with 1000 iterations, we then estimated the
average and standard error of the number of FADs with
zero, one, and two entangled silky sharks. From this
dataset, we could estimate the daily probability of a
FAD entangling a shark, taking into account both the
possibility that a single FAD could entangle multiple
sharks and that sharks could remain entangled for sev-
eral days. Given X0as the number of FADs with zero
entangled sharks and Xjas the number of FADs with a
non-zero number of entangled sharks j, we expressed
their time dependence through the following system of
differential equations:
Quantifying shark mortality in FADs JD Filmalter et al.
(Joung et al. 2008) and catch size frequencies (Amandé et al.
2010), we know that silky sharks found around FADs are
generally in their first 3 years of life. Integrating the esti-
mated average entanglement time of 300 days into an expo-
nential survival model, we found that the entanglement
mortality is 71% ± 4% after one year. Following this rate,
the number of sharks avoiding entanglement after 3 years is
only 2.6% ± 0.4% (Figure 3a).
Results from the underwater observation dataset
revealed that 35% of FADs surveyed had at least one
entangled silky shark (Table 2). The estimated value of
was = 0.85 ± 0.1 days–1 (see WebFigure 1), which led to
an average time spent entangled in the net of approxi-
mately 1.2 ± 0.2 days. Integrating this result into
Equation 7, we estimated the daily probability of a FAD
entangling a shark as µ = 0.35 ± 0.08.
To test the validity of our model (Equation 5), we per-
formed a chi-square test of homogeneity between the the-
oretical Poisson distribution (Equation 6) and the experi-
mental distribution of FADs with zero, one, and two
entangled sharks (Table 2). The test indicates that we
can accept the null hypothesis of homogeneity between
the two distributions for values of µ between 0.25 (4 days)
and 0.6 (2 days), which validates our model and results
(see WebFigure 2).
To extrapolate our predictions to the ocean basin, we
considered different numbers of FADs active
per day in the Indian Ocean (WebPanel 2).
Assuming the presence of between 3750 and
7500 active FADs (Figure 3b), estimates of
between 480 000 ± 110 000 and 960 000 ±
220 000 silky sharks are killed per year,
respectively.
nDiscussion
The high frequency of entanglement events
that we observed provides preliminary evi-
dence that the impact of FADs is severe.
Although uncertainty is inherent and unavoid-
able in this type of study (Piatt and Ford 1996),
this first quantitative estimate serves to high-
light both the extent of this issue and the need
for immediate attention. Our study is based on
two approaches that are not typically used con-
comitantly: behavioral information (obtained
through satellite tagging) and count statistics
(generally used in population analyses, but
novel here because they incorporate underwater observa-
tions from the pelagic realm). The fact that these two inde-
pendent experimental protocols both signal high rates of
entanglement reinforces our predictions.
The mortality rates reported here are concerning, and
lead directly to questions regarding their effects on popula-
tions. Owing to the absence of catch data from other types
of fishing gear believed to substantially impact silky sharks
in the Indian Ocean (gillnets and longlines), any attempt at
assessing their population status in this region is impossible
(Bonfil 2008). However, the little information available
suggests strong declines in recent decades (Anderson and
Jauharee 2009). The fact that juvenile silky sharks are still
regularly encountered at FADs suggests that either a por-
tion of the population may not be exposed to high FAD
densities or the population effect of entanglement is
delayed. Addressing the first hypothesis requires informa-
tion on the spatial density distribution of juvenile silky
sharks as well as FADs. However, these parameters are still
unknown for the Indian Ocean. As for the second hypothe-
sis, a delay in the population effect could be the result of
several interacting factors. First, the individuals affected by
entanglement are typically within the first 3 years of age.
Second, silky sharks mature after approximately 10 years
(Joung et al. 2008), and females generally produce between
six to 12 offspring every 2 years (Bonfil 2008). Third, the
use of FADs has recently increased greatly. As such, severe
population impacts may only be observed in years to come.
To contextualize our results in relation to the observed
bycatch mortality of silky sharks from the tuna purse-seine
fishery in the Indian Ocean, we used data reported in
Dagorn et al. (2012) on the amount of silky sharks (tons)
caught per 1000 tons of tuna (WebTable 4). We then con-
verted this to the number of individuals using the weight at
110 cm fork length, the peak of the observed bycatch
294
www.frontiersinecology.org © The Ecological Society of America
exp (–0.024 t)
Figure
2. Semi-logarithmic plot of the survival curve of the observation
durations for silky sharks tagged with PSATs fitted with an exponential function
of the form f(t) = exp[–(␣e+ ␣s)t]. Solid black circles represent entangled
individuals; open circles represent individuals that avoided entanglement.
0 20 40 60 80 100
Time (days)
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.09
0.08
0.07
0.06
0.05
0.04
Fraction of sharks with tags
Table 1. Time at liberty before entanglement for four
individual silky sharks tagged with PSATs
Shark ID Time at liberty (days)
494
875
19 17
13 4
JD Filmalter et al. Quantifying shark mortality in FADs
length frequency (Roman-Verdesoto and Orozco-Zoller
2005; Amandé et al. 2008), using published conversion fac-
tors (approximately 15 kg; Joung et al. 2008). This led to an
average estimate of 82 000 silky sharks taken as bycatch in
the Indian Ocean purse-seine fishery each year, an order of
magnitude less than the mortality estimates presented here.
Furthermore, following the same method, the global purse-
seine fishery catches an average of 158 000 silky sharks
annually (Dagorn et al. 2012), which is still inferior to our
lower estimate from the Indian Ocean. Our estimates are
more comparable with the estimate of silky shark bycatch
from longline vessels in the Central and South Pacific more
than 20 years ago (900 000 individuals; Bonfil 1994) or the
range of a more recent estimate from all of the world’s fish-
eries combined, obtained from the shark fin trade in
Southeast Asia (Figure 4; Clarke et al. 2006). Comparing
our results with the situations in other oceans is currently
impossible, because no similar data exist. Furthermore, we
argue that extrapolation using regional FAD deployment
figures alone should be avoided, given that entanglement
probability is likely to vary between oceans due to both
FAD design and silky shark abundance.
We recognize that the sample sizes used in this study are
limited and that increasing sample sizes would improve the
accuracy of our extrapolations. However, while improving
data collection through widespread monitoring is impera-
tive, we argue that priority should be given to solving this
issue. Despite these small sample sizes, both independent
datasets suggest that FAD entanglement poses an immense
threat to silky shark populations. Yet simple, cost-effective
solutions exist that would promote the conservation of this
species. Redesigning FADs by excluding meshed materials
would eliminate this problem while sustaining the produc-
tion of the fishery. Although these subsurface structures are
of importance, the use of netting is not. Alternative materi-
als that are biodegradable and provide zero chance of entan-
glement, such as sisal ropes, can offer effective substitutes.
Although the findings of this study reflect the impacts of
a technological change on a single species of shark, the
problem identified here is of wider importance. Fisheries
managers require more adaptive approaches. While efforts
are currently underway to improve the monitoring of
catches and bycatch, we have shown that this information
is not always adequate for detecting all impacts of changing
fishery practices. As marine resources become scarcer, the
technologies used to maintain efficient economic extrac-
tion rates will continue to develop. It is the responsibility
of scientists and managers to identify such changes, along
with all of their possible impacts, as and when they occur.
FADs have been used with increasing frequency worldwide
for the past 20 to 30 years, but it is only now that the unex-
pected impact on silky shark populations in the Indian
Ocean has been detected. Clearly, such retrospective
approaches will not lead to long-term sustainability of fish-
eries. Anticipation is essential to mitigating negative
human-induced impacts on ecosystems and as such should
be a cornerstone of resource management in the future.
nAcknowledgements
We thank F Poisson, A Vernet, and F Forget for assistance
with fieldwork. This research was financially supported by:
the commission of the European communities, FP 7,
295
© The Ecological Society of America www.frontiersinecology.org
Table 2. Number and frequency of FADs (with a net)
found with 0, 1, and 2 entangled sharks from under-
water observations
Number of entangled Number of observations Frequency
sharks per FAD (± SE) (± SE)
0 33 (± 3.5) 65% (± 7%)
1 14 (± 3.1) 27% (± 6%)
2 4 (± 2.0) 8% (± 4%)
Notes: SE = standard error.
0 1 2 3 4 5
Age (years)
(a) (b)
Fraction avoiding
entanglement
1.0
0.5
0.0
Number of entangled
sharks
0 2000 4000 6000 8000
Number of daily active FADs
1.2 ×106
0.8 ×106
0.4 ×106
0.0
Figure
3. (a) Predicted survival curve of juvenile silky sharks due to entanglement, using an average time before entanglement of 300
days. The proportion of sharks surviving is 29% after one year, 9% after 2 years, and 3% after 3 years. (b) Estimated annual
number of entangled sharks in the Indian Ocean as a function of the number of FADs active per day, from the estimated daily
probability of a FAD entangling a shark of µ= 0.35. Error bars in (a) and (b) indicate standard errors.
Quantifying shark mortality in FADs JD Filmalter et al.
“Theme 2–Food, agriculture, fisheries and biotechnology”,
through the research project MADE, contract No 210496;
the Bycatch mitigation project of the International Seafood
Sustainability Foundation; the ORTHONGEL-IRD project
on bycatch mitigation; the Marie Curie IEF fellowship, pro-
ject QUAESITUM, FP7 PEOPLE, contract No 299519;
and the Belgium National Science Foundation (FNRS).
This work does not necessarily reflect the commission’s
views and in no way anticipates its future policy in this area.
nReferences
Amandé MJ, Ariz J, Chassot E, et al. 2010. Bycatch of the European
purse seine tuna fishery in the Atlantic Ocean for the
2003–2007 period. Aquat Living Resour 23: 353–62.
Amandé MJ, Chassot E, Chavance P, et al. 2008. Silky shark
(Carcharhinus falciformis) bycatch in the French tuna purse-
seine fishery of the Indian Ocean. Victoria, Seychelles: Indian
Ocean Tuna Commission, IOTC WPEB – 2008/016.
Anderson RC and Jauharee R. 2009. Opinions count: declines in
abundance of silky sharks in the central Indian Ocean reported
by Maldivian fishermen. Victoria, Seychelles: Indian Ocean
Tuna Commission, IOTC-2009-WPEB-08.
Bonfil R. 1994. Overview of world elasmobranch fisheries. Rome,
Italy: FAO.
Bonfil R. 2008. The biology and ecology of the silky shark,
Carcharhinus falciformis. In: Camhi MD, Pikitch EK, and Babcock
EA (Eds). Sharks of the open ocean: biology, fisheries and conser-
vation. Oxford, UK: Oxford Blackwell Publishing Ltd.
Bonfil R, Amorim A, Anderson C, et al. 2009. Carcharhinus falci-
formis. In: International Union for Conservation of Nature
(IUCN) 2012: IUCN red list of threatened species. Version
2012.2. www.iucnredlist.org/details/39370/0. Viewed 2 May 2013.
Caldeira K and Wickett ME. 2003. Anthropogenic carbon and
ocean pH. Nature 425: 365.
Cheung WWL, Lam VWY, Sarmiento JL,
et al. 2010. Large-scale redistribution of
maximum fisheries catch potential in
the global ocean under climate change.
Glob Chang Biol 16: 24–35.
Clarke SC, McAllister MK, Milner-
Gulland EJ, et al. 2006. Global esti-
mates of shark catches using trade
records from commercial markets. Ecol
Lett 9: 1115–26.
Dagorn L, Holland KN, Restrepo V, et al.
2012. Is it good or bad to fish with
FADs? What are the real impacts of the
use of drifting FADs on pelagic marine
ecosystems? Fish Fish; doi:10.1111/
j.1467-2979.2012.00478.x.
Efron B and Tibshirani R. 1993. An intro-
duction to the bootstrap. New York,
NY: Chapman and Hall.
Filmalter JD, Dagorn L, Cowley PD, et al.
2011. First descriptions of the behavior
of silky sharks, Carcharhinus falciformis,
around drifting fish aggregating devices
in the Indian Ocean. Bull Mar Sci 87:
325–37.
Freon P and Dagorn L. 2000. Review of
fish associative behaviour: toward a
generalisation of the meeting point
hypothesis. Rev Fish Biol Fish 10:
183–207.
Gilman EL. 2011. Bycatch governance and
best practice mitigation technology in
global tuna fisheries. Mar Policy 35:
590–609.
Hall NG, Bartron C, White WT, et al. 2012. Biology of the silky
shark Carcharhinus falciformis (Carcharhinidae) in the eastern
Indian Ocean, including an approach to estimating age when
timing of parturition is not well defined. J Fish Biol 80:
1320–41.
Halpern BS, Walbridge S, Selkoe KA, et al. 2008. A global map of
human impact on marine ecosystems. Science 319: 948–52.
Jackson JBC, Kirby MX, Berger WH, et al. 2001. Historical over-
fishing and the recent collapse of coastal ecosystems. Science
293: 629–38.
Joung S-J, Chen C-T, Lee H-H, et al. 2008. Age, growth, and repro-
duction of silky sharks, Carcharhinus falciformis, in northeastern
Taiwan waters. Fish Res 90: 78–85.
Miyake MP, Guillotreau P, Sun C-H, et al. 2010. Recent develop-
ments in tuna industry: stocks, fisheries, management, process-
ing, trade and markets. Rome, Italy: FAO.
Myers RA, Baum JK, Shepherd TD, et al. 2007. Cascading effects of
the loss of apex predatory sharks from a coastal ocean. Science
315: 1846–50.
Parrish JK and Edelstein-Keshet L. 1999. Complexity, pattern, and
evolutionary trade-offs in animal aggregation. Science 284:
99–101.
Piatt JF and Ford RG. 1996. How many seabirds were killed by the
Exxon Valdez oil spill? Am Fish Soc Symp 18: 712–19.
Roman-Verdesoto M and Orozco-Zoller M. 2005. Bycatch of
sharks in the tuna purse-seine fishery of the eastern Pacific
Ocean reported by observers on the Inter-American Tropical
Tuna Commission, 1993–2004. La Jolla, CA: Inter-American
Tropical Tuna Commission.
Schmittner A. 2005. Decline of the marine ecosystem caused by a
reduction in the Atlantic overturning circulation. Nature 434:
628–33.
Scott MD, Chivers SJ, Olson RJ, et al. 2012. Pelagic predator asso-
ciations: tuna and dolphins in the eastern tropical Pacific
Ocean. Mar Ecol-Prog Ser 458: 283–302.
296
www.frontiersinecology.org © The Ecological Society of America
Figure
4. Estimated range of silky shark mortality due to FAD entanglement from the
Indian Ocean (top) as compared with estimated silky shark mortality from all world
fisheries from the shark fin trade in Hong Kong (bottom). Red lines indicate annual
incidental capture in purse-seine fisheries at each scale.
FAD entanglement in
Indian Ocean
Fin trade in
Hong Kong
82 000
195 000
480 000 960 000
400 000 2 000 000
200 000 400 000 600000 800000 1 000 000 1200 000 1 400 000 1 600 000 1 800 000 2 000 000
0
© The Ecological Society of America www.frontiersinecology.org
JD Filmalter et al. – Supplemental information
WebPanel 1.
Determining entanglement events and longevity
To verify that the data from the tags represented an entanglement
event and not the behavior of the silky sharks (Carcharhinus falci-
formis), we developed a method based on the vertical behavior dis-
played by each individual throughout tag deployment. In addition to
sharks outfitted with pop-up satellite archival tags (PSATs) becoming
entangled, one C falciformis (77 cm total length) tagged with a pres-
sure-sensitive acoustic tag (Vemco, Halifax, Canada), which had been
surgically implanted inside the shark’s peritoneal cavity, also became
entangled. The time-series data from this tag were recorded and
transmitted via a satellite-linked acoustic receiver (VR4-Global,
Vemco), which was attached to the drifting fish aggregating device
(FAD) where the shark had been tagged 5.42 days before. The data
from this tag were included in the estimation of entanglement
longevity. Here we define the time teas the point at which the shark
became entangled and the time tsas the point at which the shark
sank from the net, with the tag still attached (WebFigure 3a). We
considered the time interval (ts–te) as the “entanglement longevity”.
To distinguish the entanglement event from other periods of
reduced vertical movements and to establish the temporal
boundaries teand tsof the time spent in the net, we developed the
following approach. Time bins of 30 minutes were created, and
average swimming depth Djand its associated variance 2
Dj were
calculated. We then identified periods of low vertical movement
(LVM) as time windows where the shark’s swimming behavior was
characterized by a constant depth and very small vertical dis-
placement within the temporal bin, based on two criteria:
1. The variance of the depth 2
Dj of each bin constituting the LVM
time window was smaller than 1 m.
2. For each bin jconstituting the LVM time window, the consecu-
tive bin j+1 also constituted the same LVM time window if the
relative difference among their depth was smaller than 2 m
(|Dj– Dj+1|< 2m).
These criteria were chosen according to the precision of the
depth sensor of the tag (0.5 m). In this way, we identified several
LVM time windows in the tagged sharks’ behavior. We calculated
the distribution, average value, and standard deviation of these
LVM time windows, apart from the last one recorded before sink-
ing, which was a candidate for an entanglement event. If the tagged
shark was not observed sinking from the net (n= 1), we consid-
ered the entanglement longevity to be the last recorded period at
a constant depth, which was not at the surface (>3 m). WebTable
2 shows the estimated entanglement times. The average LVM time
window, other than the entanglement times, calculated for the
entangled sharks corresponded to 0.64 ± 0.61 hours. The distrib-
ution of LVM time windows was highly skewed toward very short
times <30 min (WebFigure 4). Therefore, we could safely consider
that the entanglement times in WebTable 2 corresponded to true
entanglement events, since their durations were several standard
deviations away from the average LVM window length.
Certainty of entanglement event
There are several possible interpretations of the vertical pattern
observed in the PSAT datasets; however, none explain the pat-
tern as well as that of an entanglement event.
Before considering the PSAT data, the interpretation of the
acoustic tag data is equally as important and less complicated. This
tag was internally implanted and is negatively buoyant. It can be
detected only when within the range of the receiver, which is fixed
to the FAD. The shark in question ceased all vertical displacement
after 5.42 days of “normal” behavior, and was continuously detected
for 21.43 hours. The tag then sank out of the reception range. At
this particular FAD, there were three other silky sharks tagged with
acoustic tags, none of which displayed any change in behavior similar
to the shark in question, suggesting that the receiver was functioning
normally. Before these sharks were tagged, underwater observa-
tions were performed and the sharks and FAD-associated net were
observed. Coupled with the regular observation of non-tagged
sharks entangled in the nets of FADs, it is highly probable that this
individual became entangled. After the 21.43 hours, either the shark
fell from the net or the tag fell from its peritoneal cavity as other
FAD-associated fish consumed its carcass.
The interpretation of PSAT data is more complicated given
that these tags are buoyant and will float when deployment is
terminated. There are several possible reasons for a tag deploy-
ment to be terminated, but again, none explain the observed ver-
tical pattern as well as FAD entanglement.
First, if a tag becomes detached prematurely, either by the
anchor working out of the musculature, or the tag being pulled
off by another animal, it will float at the surface. There is no rea-
son for it to suddenly sink after several hours or days of floating.
Second, if the tagged shark dies, it and the tag will sink, but
then the observed constant depth prior to sinking is unlikely to
occur. The analysis above clearly demonstrates that this constant
depth is not part of the shark’s natural behavior. If the shark was
dying due to injuries from tagging or predation, it seems equally
unlikely for it to be able to maintain a constant depth for such an
extended period in the throes of death.
The shark possibly became entangled in something other than the
netting of a FAD, such as a drifting gillnet; however, there are several
lines of evidence to suggest that this is less likely. First, we have
shown, through underwater observations, that these sharks do
become entangled in FADs, irrespective of whether they were
tagged. Second, as these sharks associate with drifting objects for
extended periods (at least several days at a time), they are constantly
exposed to the possibility of entanglement. Finally, although poorly
documented, the use of pelagic gillnets is known to be far less com-
mon in the equatorial western Indian Ocean, where the suspected
entanglements occurred, than in waters north of 10˚N. Conversely,
the distribution of FADs encompasses the entire area where sharks
were tagged and where suspected entanglements occurred.
For all of these reasons, the possibility that something other
than entanglement led to the consistent patterns observed in
the four PSAT data series seems highly unlikely.
Species concerned
Only two species of elasmobranch regularly associate with drift-
ing objects, the silky shark and the oceanic whitetip shark
(Carcharhinus longimanus). Silky sharks are far more common, rep-
resenting more than 90% of the elasmobranch catch in FAD sets.
It is certainly the associative behavior that leads to the entangle-
ment of silky sharks. Furthermore, because those that associate
are juveniles, they match the mesh size of the netting that is regu-
larly used in FAD construction. As such, the potential for the net
to act as a gillnet is almost optimized for this small size of shark.
The juveniles of other common pelagic shark species, such as the
blue shark (Prionace glauca), typically occur in more temperate
waters beyond the bounds of the FAD fishing activity. The possi-
bility that another species of shark can become entangled in a
FAD cannot be denied but it is certainly far less likely.
Supplemental information JD Filmalter et al.
www.frontiersinecology.org © The Ecological Society of America
WebPanel 2. Estimating total FAD numbers in the Indian Ocean
Quantification of the number of Carcharhinus falciformis killed through entanglement annually in the Indian Ocean required the esti-
mation of the number of FADs active on a daily basis. The fishery consists almost entirely of Spanish, Seychelles, and French flagged
vessels. As information on FAD deployment is not readily available, we used fleet-specific data and trends. The Spanish fleet (13 ves-
sels) reported the deployment of 3692 FADs with nets in 2010 to the Indian Ocean Tuna Commission (data from IOTC). Additionally,
eight vessels, with Spanish skippers but flagged in Seychelles, did not report these data but were assumed to operate in the same man-
ner, and as such, deploy 2272 FADs annually. The French fleet (13 vessels) has a self-imposed limit of 200 tracking buoys per vessel
per year and consequently deploys a maximum of 2600. We know that 58% of these buoys are deployed on FADs (Moreno et al.
2007a), leading to 1508 FADs annually. Adding these three figures gives an approximate estimate of 7500 FADs deployed annually that
could entangle sharks. This estimate is probably only accurate in terms of the order of magnitude, due to poor data reporting since
the emergence of the FAD fishery.
As FADs undergo cycles of serial ownership, their absolute history and life span is masked. Some evidence suggests that FAD life-
time may be as long as one year (Moreno et al. 2007a), whereas a more conservative estimate suggests 6 months. As such, we
obtained estimates of 3750 and 7500 FADs active on a daily basis. This compares well with the estimated 2500 FADs active daily,
obtained from interviews with skippers in 2004 and 2005 (Moreno et al. 2007b).
WebTable 1. Metadata of juvenile silky sharks (Carcharhinus falci-
formis) tagged with PSATs at drifting FADs in the Indian Ocean
Identity Tagging Total length Observation time
code date (cm) Sex (days)
1 13/03/2010 88 F 18
2 15/03/2010 109 M 100
3 15/04/2011 91 M 26
4 20/04/2011 98.5 M 94*
5 20/04/2011 103 F 11
6 18/06/2011 98 F 6
7 20/06/2011 93 M 32
8 20/06/2011 102 M 75*
9 13/04/2012 109 F 119†
10 13/04/2012 111.8 F 30†
11 14/04/2012 111.3 – 29†
12 14/04/2012 116 – 58†
13 14/04/2012 116 M 4*†
14 26/03/2011 137 F 2
15 27/03/2011 127 F 6
16 28/03/2011 140 M 27
17 28/03/2011 86 F 44
18 31/03/2011 100 F 15
19 01/04/2011 87 F 17*
20 01/04/2011 98 F 21
21 01/04/2011 87 M 36
22 02/04/2011 155 – 13
23 25/05/2011 122 F 53
24 25/05/2011 150 M 40
25 02/04/2012 104 M 28
26 02/04/2012 113.6 F 40
27 03/04/2012 132 M 77
28 03/04/2012 155 M 42
29 06/05/2012 104 – 77†
Notes: Sharks 1–13 formed part of behavioral studies whereas sharks 14–29 were tagged to study
post-release survival. Asterisks denote individuals that became entangled. Tags were programmed to
release after 100 days except those marked with the single dagger, which were set for 150 days.
WebTable 2. Entanglement
times estimated for four
sharks outfitted with PSATs
and one acoustically*
tagged shark
Identity Entanglement
code time (hours)
64* 21.43
47.92
8 54.42
19 37.83
13 25.17
JD Filmalter et al. Supplemental information
© The Ecological Society of America www.frontiersinecology.org
WebTable 3. Data from under-
water observations at FADs with
nets
Number of Number of
sharks free sharks
FAD entangled observed
11 0
22 0
30 2
40 6
50 0
60 0
71 0
81 25
91 15
10 0 10
11 0 10
12 1 1
13 2 4
14 0 3
15 1 2
16 0 2
17 2 1
18 0 5
19 0 1
20 0 3
21 1 5
22 0 0
23 0 3
24 1 0
25 0 7
26 0 0
27 0 0
28 0 0
29 1 0
30 0 0
31 0 0
32 0 0
33 1 0
34 0 0
35 0 0
36 0 0
37 0 3
38 0 0
39 1 3
40 0 1
41 0 0
42 1 1
43 1 0
44 2 1
45 0 0
46 0 2
47 0 1
48 0 2
49 0 2
50 1 2
51 0 7
WebTable 4. Data used to estimate silky shark bycatch in number of individ-
uals from each ocean area, taken from Dagorn et al. (2012)
EPO AO IO WCPO Total
Tons of sharks per 1000 t FAD tuna 1.9 1.8 6 1.1 10.8
Landed BET (t) 70 000 15 750 21000 47 500 154 250
Landed SKJ (t) 151 680 80 900 151 590 857 920 1242 090
Landed YFT (t) 35 700 14 040 55 590 178 560 283 890
Total tuna (t) 257 380 110 690 228180 1 083 980 1680 230
Total sharks (t) 440.1198 179.3178 1232.172 1073.1402 2924.7498
Number of silky sharks 29 341 11 955 82145 71 543 194 983
Notes: EPO = eastern Pacific Ocean, AO = Atlantic Ocean, IO = Indian Ocean, WCPO = western and central
Pacific Ocean. Data are reported by species: BET = bigeye tuna (Thunnus obesus), SKJ = skipjack tuna (Katsuwonus
pelamis), and YFT = yellowfin tuna (Thunnus albacares).
nWebReferences
Dagorn L, Holland KN, Restrepo V, et al. 2012. Is it good or bad
to fish with FADs? What are the real impacts of the use of
drifting FADs on pelagic marine ecosystems? Fish Fish;
doi:10.1111/j.1467-2979.2012.00478.x.
Moreno G, Dagorn L, Sancho G, et al. 2007a. Using local eco-
logical knowledge (LEK) to provide insight on the tuna
purse seine fleets of the Indian Ocean useful for manage-
ment. Aquat Living Resour 20: 367–76.
Moreno G, Dagorn L, Sancho G, et al. 2007b. Fish behaviour
from fishers’ knowledge: the case study of tropical tuna
around drifting fish aggregating devices (DFADs). Can J
Fish Aquat Sci 64: 1517–28.
Supplemental information JD Filmalter et al.
www.frontiersinecology.org © The Ecological Society of America
WebFigure 1. Survival curve of entanglement durations. Semi-
log plot of the survival curve of entanglement durations from
juvenile silky sharks tagged with pressure-sensitive electronic tags
fitted with an exponential function of the form f(t) = exp(–t).
exp (–0.85 t)
0.0 0.5 1.0 1.5 2.0 2.5
Time (days)
1.0
0.8
0.6
0.4
0.2
In fraction of sharks in the FAD net
WebFigure 2. Validating model predictions. Pvalues of the com-
parison between underwater observation data of silky sharks entangled
in FADs and the theoretical Poisson distribution calculated with
different values of µ. The test of comparison was the chi-square test.
0.2 0.4 0.6 0.8 1.0
µ
0.8
0.6
0.4
0.2
0.0
Pvalue of the chi-squared Pearson test
WebFigure 3. Depth profiles of an entangled and a free individual. (a) Typical
depth profile of a tagged shark that became entangled during the monitoring
period, from tagging to entanglement. Arrows labeled teand tsindicate the
entanglement time and sinking time, respectively. (b) Typical depth profile of a
tagged shark that did not become entangled.
(a) (b)
0
500
1000
1500
Depth (m)
tets
0 5 10 15 20
Time (days) 0 10 20 30 40
Time (days)
WebFigure 4. Distribution of LVM windows. Distribution of
LVM times from the entangled sharks.
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Frequency
N= 360
0 1 2 3 4 5 6 7 8 9 10
LVM duration (hours)
