Available via license: CC BY-NC 4.0
Content may be subject to copyright.
ORIGINAL ARTICLE
Counting sperm whales and visualising their dive profiles using
two-hydrophone recordings and an automated click detector
algorithm in a longline depredation context
Samuel Pinson* and Ga¨etan Richard**
ENSTA Bretagne, Team Siso, 2 rue Fran¸cois Verny, 29806 Brest Cedex 9, France
ARTICLE HISTORY
Compiled May 4, 2022
ABSTRACT
Odontocetes depredating fish caught on longlines is a serious socio-economic
and conservation issue. A good understanding of the depredation behaviour by
odontocetes is therefore required. Within this purpose, a method is described to
follow diving behaviour of sperm whales, considered as proxy of their foraging
activity. The study case took place around Kerguelen Islands from the Patagonian
toothfish fishery. The method uses the coherence between direct-path sperm
whale clicks, recorded by two synchronized hydrophones, to distinguish them from
decoherent clicks that are reflected by the water surface or seefloor (due to surface
roughness). Its low computational cost permits to process large dataset and bring
new insights on sperm whales behaviour. Detection of sperm whale clicks permits
to estimate the number of sperm whales and to assess their diving behaviour.
Three diving behaviour were identified as ”Water Column” (individual goes down
and up), ”Water Wander” (individual seems to go up and down multiple times
in the water column), and ”Seafloor” (individual spend time on the seabed).
Results suggest that sperm whales have different diving behaviours with spe-
cific dives as they are either ”interacting” or ”not-interacting” with a hauling vessel.
KEYWORDS
Sperm whale; Echolocation clicks; Coherence; Automatic detection; Depredation.
1. Introduction
The intensification of fishing activity and rarefaction of resources has led to an increase
in competition with marine predators worldwide over the last decades (Northridge
and Hofman 1999; Read 2008). Such competition has led some marine predators to
learn how to remove fish from lines or nets. This behaviour, named depredation,
impacts substantially fishing activity, since it increases fishing costs and fishing efforts
to finish the quotas (Gillman et al. 2006; Peterson et al. 2014; Richard et al. 2017)
depredating species, with risk of lethal retaliation or entanglement (Donoghue et al.
2002; Gillman et al. 2006; Read 2008), and for the fish stocks, since depredation is
not often accounted in fishery stock assessments (Peterson and Hanselman 2017).
*Now at IRENav, ´
Ecole Navale, Rue du Poulmic, 29160 Lanv´eoc
**Now at Soci´et´e d’Observation Multi-Modale de l’Environnement, 38 rue Jim Sevellec, 29200 Brest
Contact S. Pinson Email: samuel.pinson@ecole-navale.fr
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Although most fisheries are impacted by a broad range of large marine vertebrates
depredating, most cases involve odontocetes interacting with longlines (Donoghue
et al. 2002; Gillman et al. 2006; Northridge and Hofman 1999; Read 2008; Werner
et al. 2015; Reeves et al. 2013). Longlines are composed of snoods (thin cords)
connecting unprotected hooks along a mainline, increasing fishers’ selectivity by
catching only targeted fish species, but at the same time making a resource easily
accessible for marine predators (Gillman et al. 2006; Read 2008). Longlines could
be deployed either within the water column, named pelagic longlines, or set on the
seafloor, named demersal longlines. The fishing process is composed of three phases
: (i) the setting, when hooks are baited and longlines are deployed at sea; (ii) the
soaking, when the fish are caught as the longline is left at sea with no boat activity
nearby; and (iii) the hauling phase, when the longline is recovered on board.
Depredation behaviour on longlines is highly variable depending on the type of
gear (pelagic or demersal) and on the odontocete species involved. Pelagic longlines
deployed close to the surface are always easily accessible to odontocetes and depreda-
tion appears to occur throughout the whole fishing process (Dalla Rosa and Secchi
2007; Forney et al. 2011; Passadore et al. 2015; Rabearisoa et al. 2012; Thode et al.
2016). On the other hand, demersal longlines are thought to be mostly depredated
during hauling (Guinet et al. 2014; Hucke-Gaete et al. 2004; Mathias et al. 2012;
Straley et al. 2014; Towers et al. 2019; Roche et al. 2007). However, recent studies
suggested that sperm whales and killer whales may also depredate from demersal
longlines during soaking (Richard et al. 2020, 2022; Janc et al. 2018; Cieslak et al.
2021; Towers et al. 2019). Although proofs that killer whales actually depredate on
soaking demersal longlines are still not clearly confirmed, pieces of evidence for sperm
whales doing so are clearer. Indeed, (Janc et al. 2018) revealed that an increasing
soaking time was associated to higher numbers of sperm whales at hauling, suggesting
that individuals arrived at the longlines during the soaking phase. Alternative types
of data such as videos, passive acoustic monitoring and bio-logging are required
to better understand the underwater dimension of depredation, notably during
soaking. Additionally, (Towers et al. 2019) revealed positive correlations between the
maximum dive depths of sperm whales, obtained from time-depth recorders, and the
depths of nearest longline, suggesting potential depredation on soaking longlines too.
Finally, (Richard et al. 2020) observed plausible evidence of sperm whale depredation
on demersal soaking longlines using accelerometers fixed on longlines’ hooks, with
an event confirmed by the entanglement of a sperm whale. Indeed, settings have
specific acoustic signature which could attract attention of odontocetes before soaking
(Richard et al. 2022). Nevertheless, this study was based on a limited dataset and the
occurrence of depredation behaviour on soaking longlines could not be quantified.
In order to better describe and quantify depredation on soaking longlines, passive
acoustic monitoring (PAM) is of great interest as it may cover a longer temporal scale
around soaking longlines. Such approach was recently used to describe a common
foraging activity by killer whales around soaking longlines (Richard et al. 2022).
However, this study only relied on presence/absence of acoustic signals produced
by killer whales, and there was no confirmed evidence of killer whales foraging on
the seafloor or depredation on soaking longlines. Different setup in PAMs, using for
instance array of several hydrophones, could nevertheless target this issue by method
of spatial localisation of sound . Studies about the depredation issue in Alaska (Thode
et al. 2015; Mathias et al. 2013) resulted in the description of sperm whales dive
2
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
profiles around hauling longlines. Although sperm whale are likely to depredate on
seafloor, how often remains unknown (Richard et al. 2020).
In this paper, the focus is on the sperm whale activity during the whole fishing
process, i.e. from setting to soaking as in previous studies on killer whales (Cieslak
et al. 2021; Richard et al. 2022). A passive acoustic monitoring method is described,
allowing the dive behaviour of sperm whales to be visualized and classified, and thus
to be used as a proxy of foraging activity. The objective is to correlate such behaviour
to soaking demersal longlines, when no concurrent visual observations are available.
The study case took place in the Kerguelen Economic Exclusive Zone (Southern
Ocean) from the Patagonian toothfish fishery. The method uses the coherence
between direct-path sperm whale clicks, recorded by two synchronized hydrophones,
to distinguish them from decoherent clicks that are reflected by the water surface
or the seafloor (due to surface roughness). The automatic detection of sperm whale
clicks is thus based on recorded signals spatial coherence and has a low computational
cost. However, the distance to and dive depth of the whales could not be determined
due to uncertainties in the geometry of the two hydrophones.
The paper is organized as follows. The experiment and recordings are described
in the section Material (2). The section Methods (3) describes the signal-processing
method for automatic click detection (3.1) and how it can be used to analyze the
data (3.2) which allows to interpret some diving behaviours (3.3). The Results section
(4) focuses on the whole experiment campaign data analysis which is argued in the
Discussion (5).
2. Material
The experiment took place in January 2017 within the economic exclusive zones
of Kerguelen (49◦0’S, 70◦20’E) from a fishing vessel. Sound recordings and visuals
observations were performed during fishing operations. Fishing regulations prohibit
setting during daylight to avoid seabird bycatch (Weimerskirch et al. 2000). Several
longlines were set at night and primarily hauled during the day. Longlines can stay
from 6 hours to several days at sea. During longline hauling, trained fishery observers
recorded interactions with sperm whales, defined as presence repeatedly diving whales
within an approximate 500 m to 1 km range from the vessel (Roche et al. 2007;
Tixier et al. 2010) (data available through the Pecheker database, accessible from the
Natural History Museum of Paris (Martin and Pruvost 2007)). Longline positions
(latitude and longitude), seafloor depths (500 – 2,500 m), setting times and hauling
times were recorded.
An autonomous recorder (EA-SDA14, RTSys), paired with two hydrophones, was
deployed on a longline. The acoustic recorder was clamped at 100 m deep on the
downline connecting the buoy to the anchor (Fig. 1). The hydrophones were set 4.5 m
apart vertically. However true geometry is unknown because of longline motions due to
sea-surface waves and water currents. The recorder was programmed to record contin-
uously during the whole longline deployments (i.e. from 6 hours to several days) with
a sampling rate of 39 kHz. During the recording period, the fishing boat was moving
independently of the longline provided with the acoustic recorder, and continued the
fishing (i.e. setting and hauling other longlines).
3
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Figure 1. Sketch of a longline deployment on which 2 hydrophones and a recorder are attached.
3. Methods
The idea of the method is to take advantage of spatially incoherent reflections from
the sea surface and the sea bottom (due to surface roughnesses) to identify the sperm
whale click direct path in recorded signals. Indeed it is possible for reflected paths to
be higher in amplitude than the direct one and amplitude-only click detectors may
not be able to discriminate those different kinds of recorded clicks. Thus using two
hydrophones which distance between them is sufficiently large, it is possible to identify
the direct path from the reflected ones through their loss of spatial coherence. So it is
hypothesized that the distance between the hydrophones is big enough compared with
the dominant signal wavelength, to decorrelate sea-surface and sea-bottom reflections
recorded by the two hydrophones.
Once the sperm whale click direct paths are detected, the goal is to image the
successions of signal portions showing the time delay evolution of wave-guide reflections
relative to the direct path.
3.1. Signal processing
Sperm whale clicks are described in details by Zimmer et al. (2005). They are
generated by the phonic lips in the front of the forehead, that initially transmits the
click backwards through the spermaceti organ, towards the frontal sac, which acts
like a mirror and reflects the click forward through the ”junk”. Thus sperm whale
clicks are composed of pulses which result from multiple internal reflections within
the sound generator in the forehead of the whale. So they are separated by a fixed
interval that depends on the size of the forehead. Successive pulses (called P0, P1, P2,
etc) have a frequency band between 3 kHz and 15 kHz. Additionally a low-frequency
component below 3 kHz is present at P0. An example of a recorded click is presented
in Fig. 2a with its time-frequency analysis in Fig. 2b where the two first pulses P0
and P1 are visible at t= 0 and t= 8 ms. The low frequency component associated
with the P0 pulse is not visible in Fig. 2b because of the acoustic signature of the
4
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Figure 2. (a) Example of a recorded sperm whale click and (b) its time frequency analysis (color scale in dB
with arbitrary reference). (c) Real part of the filtered signal (thin line) and its envelop (thick line).
fishing vessel (i.e. the broadband noise below 3 kHz).
For detection purposes, the recorded signals are modified into analytical signals.
To do so the signal frequency spectrum that has to be narrow band, is filtered with
a Gaussian function centered at f= 7.5 kHz and with 2.5 kHz frequency band (at
-6 dB). This filtering also remove most of the boat noise energy below 3 kHz. Thus
the recorded signals p1,2from hydrophones 1 and 2 are preprocessed in the following
manner:
pf
1,2(t)=(p1,2(t)+iH{p1,2(t)})∗w1(t),(1)
where ∗is the convolution operator, H{p1,2(t)}is the Hilbert transform of p1,2(t)
and where the Fourier transform of w1(t) is the filter described above. Real part and
absolute value of the filtered click is presented in Fig. 2c.
Coherence between two signals is usually an integral of the product of these two
signals with a varying time delay between them and divided by the signal autocorre-
lations. But here the pattern to be detected is composed of two pulses P0 and P1, so
the coherence between pf
1(t) and pf
2(t) needs to be integrated only over the duration
of a pulse by using a sliding window w2(t). This sliding time window is chosen to be a
4 ms long (at mid-amplitude) Gaussian function to be able to frame individual pulses.
Thus the sliding coherence Coh(t, τ ) used for click detection is:
Coh(t, τ ) =
Zpf
1(t0)pf∗
2(t0−τ)w2(t0−t)dt0
rZ
pf
1(t0)
2w2(t0−t)dt0Z
pf
2(t0−τ)
2w2(t0−t)dt0+n2
(2)
5
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
where pf∗
2is the complex conjugate of pf
2,τis the time delay between pf
1(t) and pf
2(t)
at which the coherence is evaluated, tis the time where the sliding window is centered
on and, nis the signal noise level estimated from pf
1,2by fitting a Gaussian function
with the signal amplitude histogram.
Fig. 3b shows the sliding coherence of the filtered signal in Fig. 3a where four
clicks are visible and where the first one is the same as in Fig. 2a. Note that
Equation 2 is evaluated only for |τ|< d12/c ≈3.2 ms where d12 ≤4.6 m is the
distance between the hydrophones and c≈1450 m/s is the water sound speed.
One can distinguish echoes (likely from the sea-surface) about 180 ms after the
direct paths in Fig. 3a. The sliding coherence shows in Fig.3b values close to one
at direct path clicks. The sea-surface reflections are not detected because they are
incoherent. It appears that there is some coherence for τ≈ −1 ms. It is believed
that this coherence comes from the fishing-boat cavitation bubble implosions for
which the sounds are also coherent between the two hydrophones. As sperm whale
clicks are also detected when emitted near the sea surface (see next section), it is
hypothesized that Loyd mirror effect does not affect significantly the coherence.
Indeed, the Fresnel zone size is getting smaller when the source (or receiver) is getting
closer to the surface and thus, coherence may be less affected by the surface roughness.
Finally, the coherence to be used for click detection is:
coh(t) = max
τCoh(t, τ ),(3)
where the time delay for which the coherence is maximum is:
τcoh(t) = argmax
τ
Coh(t, τ ).(4)
Fig. 3c shows the results for coh(t). Direct path clicks from the example show
a coherence close to one (the other spikes are believed to originate from fishing
boat cavitation noise). Fig. 3d shows τcoh(t) for which the coherence is maximum.
It appears somehow random but its value only has meaning when coh(t) is close to one.
The click arrival time is defined at the maximum of amplitude of the P0 pulse. The
coherence function coh(t) creates a window in which looking for P0 and P1 pulses.
Nevertheless, looking for two local maxima corresponding to P0 and P1 in |pf
1(t)|(thin
line in Fig. 4) is difficult because it presents rapid fluctuations within each pulse. So
|pf
1(t)|is convolved with a smoothing Gaussian window:
ps
1(t) = |pf
1(t)| ∗ w2(t),(5)
where w2(t) is the same window as for the sliding coherence. Thus ps
1(t) (thick line in
Fig. 4) has its two maximum amplitude peaks corresponding respectively to P0 and P1.
One can see that both pulses P0 and P1 are contained by the coherence (dashed line in
Fig. 4 which corresponds to the coherence in Fig. 3c). The first step of the algorithm
to detect sperm whale direct path clicks is to fix a coherence threshold (fixed at 0.6
empirically) to frame signal portions that will be identified as coherent. Fig. 3c shows
that some boat-noise spikes are higher than 0.6. It appeared that those spikes are
shorter than sperm whale clicks. So the second step is to remove all coherent portions
6
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Figure 3. (a) Real part of the filtered signal pf
1(t) in which four sperm whale clicks associated with their
sea-surface reflections (≈180 ms later) are visible and (b) its corresponding sliding coherence Coh(t, τ ) in
which clicks appear with a coherence close to one while sea-surface reflections are invisible. (c) The coherence
signal coh(t) used for click detection and (d) the associated time delays between the two hydrophones τcoh(t).
Figure 4. (thin line) The filtered signal envelop |pf
1(t)|of the sperm whale click presented in Fig. 2a and
(thick line) its smoothed version ps
1(t). The dashed line is the coherence signal coh(t).
shorter than 10 ms (fixed empirically). This second step is sufficient to remove most
of the coherent portions due to the fishing boat. The third step is to locate the two
highest local maxima within each coherent portion of ps
1(t). The first local maximum
is then identified as the P0’s arrival time. Indeed the relative amplitude between P0
and P1 depends on the whale orientation (Zimmer et al. 2005) and thus the maximum
may be either P0’s or P1’s.
3.2. Data analysis
Once each direct path click arrival time is identified by the detection algorithm, it
is possible to store short portions of ps
1(t) for which t= 0 correspond to the P0
local maxima and display the relative time of arrival of the boundary reflections.
The result obtained from the 01/20/2017 18-hour record is presented in Fig. 5b and
the corresponding τcoh are plotted in Fig. 5a. In principle, it should be possible to
7
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Figure 5. (a) Click time delays τcoh between the two hydrophones of the detected clicks. Numerous whales
can be identified by the time delays following individual trajectories, e.g. between 11:00 and 13:30, where 7
different trajectories indicate that the individuals seem to split up before convening again. (b) Short portions
of
ps
1(t)
(colour scale in dB with arbitrary reference) for which the vertical axis is the time relative to the
detected direct-path clicks. Signal portions are displayed horizontally as a function of the acquisition time. (c)
Water column dive, where the surface echo delay times monotonically increase while the whale dives, and then
smoothly decrease when the whale ascends. (d) Sea floor dive, where the surface echo delay times increase
during descent to 200ms at 1:08. At this point weaker sea bottom echoes, starting with a 200ms delay appear,
then smoothly decrease to zero at 1:18 when the whale reaches the bottom. During this time interval, the sea
surface reflections smoothly decrease to 125ms probably due to outgoing whale displacement. (e) Wander dive,
where the surface echo time delays fluctuate while the whale changes dive depth.
calculate dive depth of and distance to the whales from Fig. 5b, by analyzing relative
travel times of sea surface and seafloor echoes (Thode 2004, 2005) but the uncertainties
on the hydrophone positions here are too important to make it possible. Nevertheless
it has been possible to find three different sperm whale dive patterns in Fig. 5b.
The first one (Fig. 5c) is a simple dive into the water column and corresponds to a
monotonic increase of sea-surface echo travel time followed by its monotonic decrease,
named ”water-column dive”. The second pattern (Fig. 5d) named ”seafloor dive” has
a similar sea-surface echo evolution, plus a sea-bottom echo crossing it (at 1:08 in
Fig. 5d) and then merging with the direct path (at 1:18 in Fig. 5d) meaning that
the whale dove down to the sea-bottom. The third pattern (Fig. 5e) suggests that
the whale is wandering in the water column as the sea-surface echo evolves back and
forth, and thus named ”wander dive”. Dive behaviour is sometimes hard to identify
in Fig. 5b as some echoes from different whales are entangled. Indeed, it is possible
to identify numerous whales in Fig. 5a as direct-path time delays follow individual
trajectories. In particular, one can identify 7 different trajectories between 11:00 and
13:30 where the individuals seem to split up before gathering again.
3.3. Diving behaviour analysis
Using a descriptive approach, diving behaviours described in previous section 3.2 are
associated with fishing activities (Fig. 6a). The fishing vessel can be either setting,
8
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Figure 6. (a) Distance between recorder and fishing activities during the study period (January 15-20). Thick
gray lines correspond to setting of long lines and thin black lines to hauling. Boxes with thick lines mark sound
recordings, boxes with grey shade indicate visual observations of sperm whale interactions during hauling. (b)
Diving behaviours (wander, column and seafloor dives) as identified from sound recordings. (c) Number of
visually (open circle) and acoustically (x) identified whales during the study period.
hauling or travelling (i.e., not engaged in any fishing activities). Visually observed and
acoustically detected sperm whales are considered to be the same individuals if the
fishing vessel (where observers are) is within the acoustic detection range, of about
30 km (Mathias et al. 2013), from the longline equipped with the hydrophones. These
identified dives were then considered as ”interacting with hauling”. Conversely, dives
were considered as ”not interacting with hauling”, if at least one of the three follow-
ing conditions is met: (i) the fishing vessel was either setting or travelling; or (ii) no
whale was visually detected from the fishing vessels; or (iii) the visual observation oc-
curred outside the acoustic detection range of sperm whale clicks from the hydrophone.
Although acoustically detected whales could not be positively identified as the one vi-
sually observed from the fishing vessel, individuals observed from vessels within the
acoustic detection range of clicks were likely to be recorded by the hydrophones. Nev-
ertheless, dives not associated with a hauling vessel (i.e. not interacting with hauling)
were more confident since no whale was observed from the fishing vessels or within
the acoustic detection range of clicks.
4. Results
4.1. Sperm whale counting and behaviour analysis
A total of 81.6 hours of acoustic data was recorded in connection with five deployments
(mean duration of 16.3±4.1 hours per deployment). In these recordings a total of 51
dives were identified, with a total duration of 25 hours (mean duration of 0.5±0.3
hours per dive). In 12 events, with a total duration of 17 hours (i.e. ≈40 % of the
total dive; mean duration per event 1.4±0.8 hours) the dive types could not be
identified due to clicks from different whales being superposed, making it impossible
9
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
to associate the correct surface or seafloor reflections. Among the 51 identified dives,
57 % were associated as water-column diving behaviour (Fig. 5c and Fig. 6b), 25 %
as seafloor diving behaviour (Fig. 5d and Fig. 6b) and 16 % as wander-dive behaviour
(Fig. 5e and Fig. 6b).
The number of sperm whales for every observation varied from 1 to 7 individuals
(Fig. 6c). Counting from acoustic detection was similar to the number of individuals
estimated by observers from the fishing vessel at hauling (Fig. 6c). However, the last
recording revealed that during three haulings the acoustic method estimated more
individuals than the visual observation from the fishing vessel (Fig. 6c). Additionally,
the acoustic method also allows for estimating the number of individuals in the area,
within detection range of ca. 30km, when no visual observations are carried out, e.g.
during setting and travel.
4.2. Sperm whale interaction with boat
More diving sperm whales were detected as not interacting with hauling, n=30, than
as interacting with hauling, n=21 (Fig. 7a). When not interacting with hauling, the
whales seemed mainly engaged in water column dives, which constituted 63 % of the
totally 30 dives. The other 37 % dives were to the seafloor (Fig. 7a), since no wander
behaviour was detected at these time (i.e. not interacting with hauling). Besides, the
majority of seafloor dives, 11 dives representing 85 % of all such dives, were associated
with not interacting with hauling. Seafloor dives were then almost exclusive to not
interacting with hauling behaviour (Fig. 7b). Conversely, wander dives occurred exclu-
sively when interacting with hauling (Fig. 7b), and represented 43 % of the interacting
dives (Fig. 7a). When interacting with hauling, the whales were mainly engaged in wa-
ter column dives, representing 48 % of all 21 dives. Only two seafloor dives (i.e. 15 %
of these dives) were identified in connection with interacting with hauling (Fig. 7a).
Altogether, these results suggest that sperm whales have different diving behaviours
with specific dives as they are either interacting with hauling or not-interacting with
hauling.
5. Discussion
Taking advantage of the bad weather conditions by using a coherence function to
distinguish the acoustic direct path from the reflected ones permits to automatically
detect sperm whale clicks. As the technique is based on simple signal processing,
the computation time is of about a quarter of the recording time on a laptop
computer. In principle, it should be possible to estimate the range of the whale
from the two hydrophones by using the angle of arrival of the direct click and the
time delay of the sea-surface reflection (Thode 2004, 2005; Mathias et al. 2013).
To do so, it is necessary that the hydrophones remains on vertical position and
the distance between them stays steady. But such trivial theoretical requirements
remain challenging in the real world, especially when deployments are logistically con-
straint by fishing operation in rough conditions as encountered in the Southern Ocean.
The passive acoustic method shows great potential to estimate the number of
sperm whales around longlines, at least within a 30 km range from the two element
hydrophone array. This acoustic approach could then be complementary to traditional
10
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Figure 7. (a) Proportion of dive types in sperm whales interacting and not interacting with hauling. The
majority of dives were water column dives (N=29), followed by seafloor dives (N=13) and wander dives (N=9).
(b) Proportion of interacting vs not interacting with hauling in connection with the three dive types. Although
most of the water column dives were not associated with hauling, an equal number of such dives and wander
dives were associated with hauling.
visual counting from the vessel at hauling (Roche et al. 2007; Tixier et al. 2010).
However, the acoustic method has the added advantage that whales can be counted
also when the fishing vessels are not hauling, and thus have no visual observers
in action (Richard et al. 2022). Also, in some of the deployments reported here,
the acoustic method counted more whales than reported by the visual observers,
indicating that the latter may underestimate the real hauling interaction rates.
Indeed, missing the presence of some individuals is likely to understate interaction
rates (Roche et al. 2007; Tixier et al. 2010). However, assessing whether depredation
events are missed by visual observations requires a good monitoring of sperm whales’
behaviours around soaking longlines.
The second advantage with this passive acoustic method is that it allows for the
identification of different dive behaviours. Although it turned out to be difficult to
identify the different dive types when several whales are clicking simultaneously, still
a large proportion of the whales in this study could be individually distinguished
to allow for the identification of their dive types. The association of wander dives
with hauling suggests that this may be a depredation-specific behaviour, where the
repetitive descents and ascents could be the whale moving along the longline and
raking it, as described in Alaskan fisheries (Mathias et al. 2009). Also, the higher
proportion of water column dives than seafloor dives in connection with hauling is
consistent with the shallow dive behaviour described for sperm whales depredating
on Alaskan longline fisheries (Mathias et al. 2013). However, this dive type may
also be a natural foraging behaviour for pelagic cephalopods (Whitehead 2009),
since it was the main dive type in whales not interacting with hauling. In this case,
foraging on pelagic prey would represent here 63 % of their diving behaviours when
not interacting with a vessel, against 37 % of their dives considered as feeding on a
demersal fish as Patagonian toothfish (Collins et al. 2010). Conversely, the higher
proportion of seafloor dives when not interacting with a hauling vessel may either
reveal a natural foraging behaviour on demersal prey as Patagonian toothfish (Abe
and Iwami 1989) or a seafloor depredation behaviour (Richard et al. 2020).
11
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Although the method does not enable to distinguish between natural foraging at
seafloor or depredation behaviour on soaking longlines, the optimal foraging theory
(Charnov 1976; Pyke 1984) would suggest that individuals are more likely to use the
easiest resource, and thus that they may depredate on soaking longlines. Within this
assumption, 22 % of the detected dives (i.e. 11 seafloor dive while not interacting
with hauling among the 51 dives) in this study would then be associated as missed
soaking depredation events (Richard et al. 2020). Conversely, 41 % of all dives were
clearly associated to interaction behaviour at hauling. This observation may have
major implications for estimation of interaction rates and perhaps even depredation
rates. The real depredation rate is the difference in catch per unit effort, and may be
the sum of both seafloor and hauling depredation (Tixier et al. 2010; Gasco et al.
2015). Hence, only counting possible hauling depredation may underestimate the
depredation rate.
6. Conclusion
This passive acoustics method, based on two hydrophones and the application of
coherence-based analysis, allows for the estimation of the number of sperm whales
within a area of 30 km and the classification of their dive behaviour. Three dive types
were identified: (i)Water column dive, where the whale monotonically descended in
mid water, and then monotonically ascended, (ii)seafloor dive where the whales dove
all the way down to the seafloor, and (iii )wander dive where the whale did multiple
descent and ascents in mid water. These behaviours could be interesting proxies for
a better understanding of interactions with fisheries. Future research should focus on
two aspects. First a greater robustness of the acoustic recorder geometry would make
it possible to estimate the depth and distance of the sperm whales. This would make it
possible to better discriminate possible depredation from natural foraging by localising
more precisely clicking individuals among the longlines and within the water column.
Second, to increase the acoustic dataset as well as the visual observations, in order to
refine the estimation of the interaction and/or depredation rates.
Acknowledgments
We warmly thank the captains, their crew and the fishery observers for their help
in the data collection. We also thank the Natural History Museum (Mus´ee National
d’Histoire Naturelle) of Paris for providing access to the PECHEKER Database. We
are also very grateful for the support of the Fondation d’Entreprise des Mers Aus-
trales, the Syndicat des Armements R´eunionais des Palangriers Cong´elateurs, fishing
companies, the Direction des Pˆeches Maritimes et de l’Aquaculture, Terres Australes
et Antarctiques Fran¸caises (the Natural Reserve and Fishery units). We thank Dr.
Julien Bonnel, Dr. Christophe Guinet and Dr. Flore Samaran for their help and con-
tribution within the Orcadepred project. We warmly thank the anonymous reviewer
from a previous submission who greatly helped improving the manuscript with his
insightful comments.
12
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Ethical statement
All instrument deployments followed the ethics policies of the French Southern and
Antarctic Lands (TAAF) and were authorized by the R´eserve Naturelle Nationale
(RNN des TAAF) through approval A-2017-154.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Funding
This work was supported by the Direction G´en´erale de l’Armement (DGA); and by
the Agence Nationale de la Recherche (ANR) under the OrcaDepred program.
References
Abe T, Iwami T. 1989. Notes on fishes from the stomachs of whales taken in the Antarctic. II.
On Dissostichus and Ceratias, with an appendix (Japanese names of important Antarctic
fishes). Proceedings of the NIPR Symposium on Polar Biology. 2:78–82.
Cieslak M, Tixier P, Richard G, Hindell M, Arnould JP, Lea MA. 2021. Acoustics and photo-
identification provide new insights on killer whale presence and movements when interacting
with longline fisheries in south east australia. Fisheries Research. 233:105748.
Collins MA, Brickle P, Brown J, Belchier M. 2010. Chapter Four - The Patagonian Toothfish:
Biology, Ecology and Fishery. In: Lesser M, editor. Advances in Marine Biology. vol. 58.
Academic Press; p. 227–300.
Dalla Rosa L, Secchi ER. 2007. Killer whale (Orcinus orca) interactions with the tuna and
swordfish longline fishery off southern and south-eastern Brazil: a comparison with shark in-
teractions. Journal of the Marine Biological Association of the United Kingdom. 87(01):135–
140.
Donoghue M, Reeves R, Stone GS. 2002. Report of the workshop on interactions between
cetaceans and longline fisheries, Apia, Samoa. New England Aquarium Press.
Forney KA, Kobayashi DR, Johnston DW, Marchetti JA, Marsik MG. 2011. What’s the catch?
Patterns of cetacean bycatch and depredation in Hawaii-based pelagic longline fisheries:
Patterns of cetacean bycatch and depredation in Hawaii. Marine Ecology. 32(3):380–391.
Gasco N, Tixier P, Duhamel G, Guinet C. 2015. Comparison of two methods to assess fish
losses due to depredation by killer whales and sperm whales on demersal longlines. CCAMLR
Science. 22:1–14.
Gillman E, Brothers N, McPherson G, Dalzel P. 2006. A review of cetacean interactions with
longline gear. Journal of Cetacean Research and Management. 8:215–223.
Guinet C, Tixier P, Gasco N, Duhamel G. 2014. Long-term studies of Crozet Island killer
whales are fundamental to understanding the economic and demographic consequences of
their depredation behaviour on the Patagonian toothfish fishery. ICES Journal of Marine
Science. 72(5): 1587–1597
Hucke-Gaete R, Moreno CA, Arata J, Ctr BW. 2004. Operational interactions of sperm whales
and killer whales with the Patagonian toothfish industrial fishery off southern Chile. Ccamlr
Science. 11:127–140.
Janc A, Richard G, Guinet C, Arnould JP, Villanueva MC, Duhamel G, Gasco N, Tixier P.
2018. How do fishing practices influence sperm whale ( Physeter macrocephalus ) depreda-
tion on demersal longline fisheries? Fisheries Research. 206:14–26.
13
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Martin A, Pruvost P. 2007. Pecheker, relational database for analysis and management of
fisheries and related biological data from the French Southern Ocean fisheries monitoring
scientific programs. Mus´eum National d’Histoire Naturelle.
Mathias D, Thode A, Straley J, Folkert K. 2009. Relationship between sperm whale (Physeter
macrocephalus) click structure and size derived from videocamera images of a depredating
whale (sperm whale prey acquisition). The Journal of the Acoustical Society of America.
125(5):3444–3453.
Mathias D, Thode AM, Straley J, Andrews RD. 2013. Acoustic tracking of sperm whales in the
Gulf of Alaska using a two-element vertical array and tags. The Journal of the Acoustical
Society of America. 134(3):2446–2461.
Mathias D, Thode AM, Straley J, Calambokidis J, Schorr GS, Folkert K. 2012. Acoustic and
diving behavior of sperm whales (Physeter macrocephalus) during natural and depredation
foraging in the Gulf of Alaska. The Journal of the Acoustical Society of America. 132(1):518–
532.
Northridge SP, Hofman RJ. 1999. Marine mammal interactions with fisheries. Conservation
and management of marine mammals:99–119.
Passadore C, Domingo A, Secchi ER. 2015. Depredation by killer whale (Orcinus orca) and
false killer whale (Pseudorca crassidens) on the catch of the Uruguayan pelagic longline
fishery in Southwestern Atlantic Ocean. ICES Journal of Marine Science. 72(5):1653–1666.
Peterson MJ, Hanselman D. 2017. Sablefish mortality associated with whale depredation in
Alaska. ICES Journal of Marine Science. 74(5):1382–1394.
Peterson MJ, Mueter F, Criddle K, Haynie AC. 2014. Killer Whale Depredation and Associated
Costs to Alaskan Sablefish, Pacific Halibut and Greenland Turbot Longliners. PLoS ONE.
9(2):e88906.
Charnov E. 1976. Optimal foraging, the marginal value theorem. Theoretical population biol-
ogy, 9(2): 129–136.
Pyke GH. 1984. Optimal Foraging Theory: A Critical Review. Annual Review of Ecology and
Systematics. 15:523–575.
Rabearisoa N, Bach P, Tixier P, Guinet C. 2012. Pelagic longline fishing trials to shape a
mitigation device of the depredation by toothed whales. Journal of Experimental Marine
Biology and Ecology. 432:55–63.
Read AJ. 2008. The Looming Crisis: Interactions between Marine Mammals and Fisheries.
Journal of Mammalogy. 89(3):541–548.
Reeves R, McClellan K, Werner T. 2013. Marine mammal bycatch in gillnet and other entan-
gling net fisheries, 1990 to 2011. Endangered Species Research. 20(1):71–97.
Richard G, Bonnel J, Beesau J, Calvo E, Cassiano F, Dramet M, Glaziou A, Korycka K,
Guinet C, Samaran F. 2022. Passive acoustic monitoring reveals feeding attempts at close
range from soaking demersal longlines by two killer whale ecotypes. Marine Mammal Science.
38(1):304–325.
Richard G, Samaran F, Guinet C, Bonnel J. 2021. Settings of demersal longlines reveal acoustic
cues that can inform toothed whales where and when to depredate.JASA Express Letters.
1(1):016004.
Richard G, Bonnel J, Tixier P, Arnould JP, Janc A, Guinet C. 2020. Evidence of deep-sea
interactions between toothed whales and longlines. Ambio. 49(1):173–186.
Richard G, Guinet C, Bonnel J, Gasco N, Tixier P. 2017. Do commercial fisheries display
optimal foraging? The case of longline fishers in competition with odontocetes. Canadian
Journal of Fisheries and Aquatic Sciences. 75(6): 964–976
Roche C, Guinet C, Gasco N, Duhamel G. 2007. Marine mammals and demersal longline
fishery interactions in Crozet and Kerguelen Exclusive Economic Zones: an assessment of
depredation levels. CCAMLR Science. 14:67–82.
Straley J, Schorr G, Thode A, Calambokidis J, Lunsford C, Chenoweth E, O’Connell V,
Andrews R. 2014. Depredating sperm whales in the Gulf of Alaska: local habitat use and long
distance movements across putative population boundaries. Endangered Species Research.
24(2):125–135.
14
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
Thode A. 2004. Tracking sperm whale (Physeter macrocephalus) dive profiles using a towed
passive acoustic array. The Journal of the Acoustical Society of America. 116(1): 245–253.
Thode A. 2005. Three-dimensional passive acoustic tracking of sperm whales(Physeter macro-
cephalus) in ray-refracting environments. The Journal of the Acoustical Society of America.
118(6):3575–3584.
Thode A, Mathias D, Straley J, O’Connell V, Behnken L, Falvey D, Wild L, Calambokidis J,
Schorr G, Andrews R, et al. 2015. Cues, creaks, and decoys: using passive acoustic mon-
itoring as a tool for studying sperm whale depredation. ICES Journal of Marine Science.
72(5):1621–1636.
Thode A, Wild L, Straley J, Barnes D, Bayless A, O’Connell V, Oleson E, Sarkar J, Falvey D,
Behnken L, et al. 2016. Using line acceleration to measure false killer whale (Pseudorca
crassidens) click and whistle source levels during pelagic longline depredation. The Journal
of the Acoustical Society of America. 140(5):3941–3951.
Tixier P, Gasco N, Duhamel G, Viviant M, Authier M, Guinet C. 2010. Interactions of Patag-
onian toothfish fisheries with killer and sperm whales in the Crozet islands Exclusive Eco-
nomic Zone: an assessment of depredation levels and insights on possible mitigation strate-
gies. CCAMLR Science. 17:179–195.
Towers JR, Tixier P, Ross KA, Bennett J, Arnould JPY, Pitman RL, Durban JW. 2019.
Movements and dive behaviour of a toothfish-depredating killer and sperm whale. ICES
Journal of Marine Science. (76):298–311.
Weimerskirch H, Capdeville D, Duhamel G. 2000. Factors affecting the number and mortality of
seabirds attending trawlers and long-liners in the Kerguelen area. Polar Biology. 23(4):236–
249.
Werner TB, Northridge S, Press KM, Young N. 2015. Mitigating bycatch and depredation of
marine mammals in longline fisheries. ICES Journal of Marine Science. 72(5):1576–1586.
Whitehead H. 2009. Sperm Whale. In: Encyclopedia of Marine Mammals. Elsevier; p. 1091–
1097.
Zimmer WMX, Tyack PL, Johnson MP, Madsen PT. 2005. Three-dimensional beam pattern
of regular sperm whale clicks confirms bent-horn hypothesis. The Journal of the Acoustical
Society of America. 117(3):1473–1485.
15
.CC-BY-NC 4.0 International licenseperpetuity. It is made available under a
preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in
The copyright holder for thisthis version posted September 7, 2022. ; https://doi.org/10.1101/2022.09.06.506716doi: bioRxiv preprint
