Content uploaded by Juliana López Marulanda
Author content
All content in this area was uploaded by Juliana López Marulanda on Feb 08, 2019
Content may be subject to copyright.
Aquatic Mammals 2017, 43(2), 162-176, DOI 10.1578/AM.43.2.2017.162
First Results of an Underwater 360° HD
Audio-Video Device for Etho-Acoustical Studies on
Bottlenose Dolphins (Tursiops truncatus)
Juliana Lopez-Marulanda,1 Olivier Adam,1, 2 Torea Blanchard,1, 3 Marie Vallée,4
Dorian Cazau,5 and Fabienne Delfour3, 6
1Institute of Neurosciences Paris Saclay, CNRS UMR 9197, University Paris Sud, Orsay, France
E-mail: juliana.lopez-marulanda@u-psud.fr
2Sorbonne Universités, UPMC Univ Paris 06, CNRS UMR 7190, Institut Jean Le Rond d’Alembert, F-75005 Paris, France
3Parc Astérix, Plailly, France
4Association Abyss, La Réunion, France
5ENSTA Bretagne Lab-STICC, CNRS UMR 6285, Université Européenne de Bretagne, Brest, France
6Laboratoire d’Ethologie Expérimentale et Comparée EA 4443 (LEEC), Université Paris 13,
Sorbonne Paris Cité, F-93430 Villetaneuse, France
Abstract presented. The results show that when the observers
stayed ahead and avoided the direct path of groups
Bottlenose dolphins (Tursiops truncatus) are highly of five to nine dolphins, only one animal emitted
social odontocetes that live in a fission-fusion soci-click trains while swimming towards the observers
ety and demonstrate production of a varied sound or after turning its rostrum in the humans’ direc-
repertoire, including clicks, whistles, and burst- tion, and this dolphin was never the one leading
pulsed sounds, as well as a diverse behavioral the group. The benefits of using this audio-video
repertoire. To better understand the species’ behav-device for underwater observations of dolphins in
ior, it is necessary to compare visual and acoustic clear water with good visibility are discussed.
observations and link vocalizations to individuals
and their specific actions. However, the task of Key Words: behavior, acoustics, hydrophone
linking sounds to individual dolphins is challeng-array, acoustic localization, bottlenose dolphin,
ing for human observers because dolphins do not Tursiops truncatus
always display specific visual cues when producing
a sound, and also because human hearing is not nat-Introduction
urally adapted to locate underwater sound sources.
To respond to these challenges, a new underwa-Bottlenose dolphins (Tursiops truncatus) are
ter 360° HD audio-video device, the BaBeL, was highly social odontocetes with a fission-fusion
designed and built. This device consists of a five- social structure (Connor et al., 2000; Mann et al.,
hydrophone array attached to two wide-angle video 2000; Gibson & Mann, 2008; Tsai & Mann, 2013).
cameras that together cover a 360° field of vision. Group members may travel very short or very
Acoustic recordings were analyzed with a custom-long distances within a habitat of limited visibil-
ized program to detect and localize sound sources ity (Connor et al., 1998). As such, communication
and to identify individual vocalizing dolphins. Data via acoustic signals is the most effective strategy
from a population of bottlenose dolphins were col-for sharing information under water (Janik, 1999).
lected during 14 boat surveys along the northwest Bottlenose dolphins display a complex and exten-
coast of Reunion Island (France) by following a sive repertoire of sounds such as clicks or pulsed
strict pre-established protocol to standardize data sounds (Au, 1993; Au & Fay, 2012), burst-pulsed
collection. A total of 21 min of audio-video were sounds (Lopez & Bernal-Shirai, 2009), and whis-
recorded when dolphins were present, and 42 click tles or tonal sounds (reviewed in Janik, 2009).
trains and 42 whistles were detected from these Dolphin sound production is enriched by non-
data. Dolphins identified as vocalizers were also acoustic communication signals during social
present for 17% (n = 7) of emitted click trains and interactions when individuals are within visual
33% (n = 14) of emitted whistles on the videos. range of one another. These animals display vari-
Therefore, an analysis of three video sequences ous body postures (Pryor, 1990), contacts (Sakai
as examples of the scope of this methodology is et al., 2006; Dudzinski et al., 2009), and bubble
163360° HD Audio-Video Device to Observe Dolphins
emissions (Marten et al., 1996). Several dolphin dolphin. Four hydrophones are needed to local-
vocalizations are associated with behavioral con-ize a moving dolphin in 3D (Watkins & Schevill,
texts; for example, burst-pulsed sounds (squawks 1972; Wahlberg et al., 2001)
and whines) have been associated with agonistic Mobile arrays of two (Dudzinski et al., 1995),
behaviors (Herzing, 1996), and low-frequency four (Au & Herzing, 2003; Schotten et al., 2004),
bray calls are related to feeding (Janik, 2000; and 16 (Ball & Buck, 2005) hydrophones have
King & Janik, 2015). been used to study dolphin vocalizations and
Studies of dolphin social behavior and commu-their associated underwater behaviors, but they,
nication rely on simultaneous descriptions of both too, presented several disadvantages. Dudzinski
visual and acoustic signals (Thomas et al., 2002). et al.’s (1995) system did not allow localization in
However, the main obstacle associated with com-the vertical axis, and the systems with four hydro-
pleting these descriptions is the difficulty in iden-phones were used only for localization of click
tifying which dolphin in a group is the vocalizer. emitters (Au & Herzing, 2003; Schotten et al.,
This challenge is caused by two factors: (1) human 2004). The 16-hydrophone array had elements
hearing is not adapted to localize to sound sources separated by 3.2 cm (Ball & Buck, 2005) but did
underwater; and (2) dolphins do not show visible, not allow the confirmation of the emitter’s iden-
regular signs when emitting sounds, like opening tity if animals were located outside of the narrow
their mouths or displaying external clues (Janik, angle of the video camera.
2009). To overcome these obstacles, several meth-As part of this study, an audio-video system
odologies have been developed. Animals have been that was non-intrusive and compact enough to be
isolated (Caldwell et al., 1990; Sayigh et al., 1990) deployed from a small boat was developed. This
or tagged (Tyack, 1991; Nowacek et al., 1998); underwater device includes five hydrophones and
however, these approaches can be considered inva-a 360° HD video recording system with a lim-
sive and might lead to modification of the subjects’ ited blind spot that allows localization of sounds
behaviors and vocalizations. Emission of bubble to free-swimming, vocalizing dolphins coming
streams concurrent with vocalizations has been from almost every direction. In this article, details
used to identify a vocal animal because sometimes about this system’s design and the software devel-
dolphins emit bubbles while whistling (McCowan oped to localize to sounds and to link them to indi-
& Reiss, 1995; Herzing, 1996); however, whistles vidually identified dolphins are provided. Three
with bubble streams are not representative of the audio-video sequences of free-ranging bottlenose
entire whistle repertoire of bottlenose dolphins dolphins have been analyzed to illustrate the ben-
(Fripp, 2005, 2006). efits of this system in dolphin ethological and
As a non-intrusive method to identify the acoustical research.
vocalizing animal, different hydrophone arrays
have been designed. These arrays allow for pro-Methods
cessing of the differences in time of arrival of the
sound to each hydrophone to determine where Recording Device
the call originated. The position of the sound Simultaneous audio and video recordings were
source is linked to the video recordings to con-collected using a waterproof audio-video system
firm which animal is in the same position as the named BaBeL (BioAcoustique, Bien Être et
sound source, thereby identifying the vocalizer. Langage) (Figure 1). The acoustic set-up was
Fixed arrays using two (López-Rivas & Bazuá- comprised of five calibrated Aquarian H2a-XLR
Durán, 2010), three (Watkins & Schevill, 1974), hydrophones connected and synchronized to a
four (Brensing et al., 2001; Quick et al., 2008), ZOOM H6 digital audio recorder. Audio record-
and eight (Thomas et al., 2002) hydrophones have ings were made at a 96-kHz sampling frequency
been used to link audio recordings to behavioral and coded on 24 bits. The recorder was placed in
observations or video recordings; however, fixed a waterproof housing rated to 60 m depth. The
arrays are not well adapted to study highly mobile, architectural design of the hydrophone array was
free-ranging dolphins, and the video recordings a compromise between a large aperture between
were often obtained from a fixed point at the sur-hydrophones and maneuverability since the
face. The main problem with acquiring behavioral system needed to be deployed from small boats
information from the surface is that the docu-with limited space and to be controllable by one
mented behaviors could possibly represent only a observer when submerged. The synchronized
very small percentage of an animal’s behavioral hydrophones were positioned to obtain the time
activity (Janik, 2009). Moreover, an array with delay of arrival to provide the 3D estimations of
two hydrophones (Lopez-Rivas & Bazuá-Durán, dolphin positions.
2010) allows data to be obtained only on the angle The video portion of the BaBel system was
of arrival and not the real position of the emitting comprised of two Kodak SP360 video cameras
164 Lopez-Marulanda et al.
both with a wide-angle field of view (214°); the method used the spatial distribution of hydro-
cameras were placed opposite of each other to the phones, the acoustic properties of the source
left and the right to allow for 360° field of vision (e.g., propagation speed and spherical propaga-
for the system. These cameras were positioned tion model), and the measure of the time differ-
below the waterproof housing of the acoustic ences of arrival (TDOA) of the acoustic wave
recorder (Figure 2). Video and audio files were from the source to the different hydrophones
stored for a posteriori analysis. (Alameda-Pineda & Horaud, 2014). The aim
was to estimate the differences in time of arrival
Custom-Made Program for Data Analysis of emitted sounds; the cross-correlation function
A geometrical localization method was used to method for whistle detection (Van Lancker, 2001)
estimate the position of an acoustic source. This and the threshold time energy for click detection
Figure 1. BaBeL (BioAcoustique, Bien Être et Langage) device: (a) Diagram and orientation of hydrophones (H); and
(b) picture of the device with 360º cameras unattached—five hydrophones installed on five deployable arms and plugged to
a ZOOM H6 (inside the adapted waterproof case).
165360° HD Audio-Video Device to Observe Dolphins
(Blanchard, 2015) were used. To display the esti-
mated position of the acoustic source in the video
image, a conversion position-pixel that took into
account the deformations of the image because
of the spherical curved lens of the Kodak SP360
video cameras was used. With these consider-
ations, a customized program to analyze data
obtained with the BaBeL system was created in
MATLAB®, Version 2013a (Mathworks, Natick,
MA, USA) to synchronize the video and audio
recordings and then to estimate the localization(s)
of each vocalizing dolphin (Blanchard, 2015)
(Figure 3). After identifying the location of the
vocalizing dolphin, video analysis allowed for the
identification of the dolphin based on recogniz-
able scars and marks.
Tests with Artificial Sounds
Two simulated whistles to test this approach with
different signal-to-noise ratios (SNRs) (Figure 4)
were created. The objective of this test was to
confirm performance of the time correlation
for acoustic signals deteriorated by underwater
acoustic propagation or when ambient noise is
present in the marine area. To verify our local-
ization method, the system was tested in a 3.1 ×
8.2 m² rectangular freshwater swimming pool.
The BaBeL was immersed in the center of the
Figure 2. Disposition of two Kodak SP360 video cameras.
Each camera has a 360° (N-S-E-W) plus 214° angle of
view. As both cameras are placed opposite to each other,
there is a 34° overlap in the images and a ≈ 50-cm blind
spot between the cameras.
Figure 4. Simulated whistles with two signal-to-noise ratios (SNRs): (a) +20 dB and (b) -10 dB.
166 Lopez-Marulanda et al.
pool at 2.5 m from the edge. Percussive sounds Bottlenose Dolphin Data Collection
were generated by knocking together two steel
Acoustic and video data were collected on free-
bars from nine different known places in the hori-
ranging bottlenose dolphins along the northwest
zontal plane of the deviceʼs gravity center. Using
coast off Reunion Island, a French territory in the
the position-pixel conversion, the position of each
Mascarene Islands in the Southwest Indian Ocean.
percussive sound source in the video image was
The species is observed in this location through-
estimated and compared to the location estimated
out the year in groups of 10 to 100 individuals (48
by the custom program.
individuals on average) and occurs in deeper water
Figure 4. Screen display to track dolphins by videos and passive acoustics. On the bottom left, estimations of the angles from
the successive clicks (in blue) and the whistle (in red). On the right, the red cross points to the emitter dolphin.
Figure 5. Disposition of BaBeL device in the water; the whole system is controlled by Observer 1. The vision range of the
camera depends on water clarity. For the Reunion Island, it is ≈10 m. The real scales are not represented in the figure.
167360° HD Audio-Video Device to Observe Dolphins
(425.6 m on average) and further offshore (1.2 to
Audio-Video Analysis
6 km from the coast) than other cetacean species in
The claps at the beginning of each recording ses-
this area (Dulau-Drouot et al., 2008).
sion were used to manually synchronize acoustic
Fourteen boat surveys were conducted from and video data with video editing software (Final
21-29 May 2015 and from 6-18 June 2015 to Cut Pro X, Version 10.1.3©, Apple Inc.). A single
search for bottlenose dolphins and collect etho- video file was created displaying the videos of
acoustical data. When a group of dolphins was the two Kodak SP360 video cameras in the same
sighted, a strict pre-established protocol was fol-window, as well as one of the five audio tracks and
lowed (see Agreement on the Conservation of its corresponding turning spectrogram (FFT size:
Cetaceans in the Black Sea Mediterranean Sea 1,024, overlap 50%, Hanning window) obtained
and Contiguous Atlantic Area [ACCOBAMS], with Audacity, Version 2.0.6 (GNU General
Resolution 4.18) to decide if observers would Public License). We chose only one track in the
enter the water to start a recording session. First, video as a referent since the five audio tracks were
the boat was positioned parallel to the animals’ used for our custom-made acoustical analysis in
travel direction at a distance of more than 50 m. MATLAB® (Blanchard, 2015).
The behavioral response of the dolphin group was The location of the vocalizing dolphin was
recorded into one of three categories: (1) “avoid-noted as “visible” when our program was able to
ance,” (2) “indifference,” and (3) “oncoming” point out one of the dolphins in the video, “ambig-
(see ACCOBAMS for definitions). If the behav-uous” when the program pointed out two dolphins
ioral response was cataloged as “indifference” that were close to each other or in the same direc-
or “oncoming,” the boat was slowly positioned tion, and “not visible” when the program pointed
100 m ahead of the first animal of the group, never to another direction indicating that the emitter
interfering with the travel direction of the animals. dolphin was far outside the range of vision of the
Once in this position, two observers slipped into video cameras, estimated at further than 10 m
the water. away in all directions but also dependent on the
Procedure in Water—One observer swam with wide-angle lens (reduces the size of objects) and
the BaBeL device submerged below the sea sur- water clarity.
face (≈1 m under the surface) (Figure 5), while the To conduct our etho-analysis, the sequences in
other observer recorded the animals on a backup which we could locate with no doubt at least one
SONY HDR-GW66 video camera. Date and time vocalizing dolphin were selected (see Appendix 2,
on all video cameras were synchronized for a pos-Figure 1); a focal-animal sampling technique was
teriori analysis. Since BaBeL was being operated used to note occurrence and duration of body
for the first time, the intent was to document all postures, tactile contacts, and other behaviors
the events. The backup video sequences might be displayed during intraspecific interactions and
used later to confirm what was observed on the during interactions towards humans in video
BaBeL wide-angle cameras, and the recorded sequences (Altmann, 1974; Mann, 1999). Since
sequences might be replayed to document the all sightings were mainly “swim by” wherein
BaBeL operator’s position and behavior in the the dolphins did not remain near the observers
water. for long, individual dolphins in each sequence
At the beginning of the recording session, two were listed in order of appearance in the video.
successive claps were made—one in front of each The “all occurrences” recording sampling method
camera in order to synchronize both videos with focused on frequencies, and durations of occur-
audio recordings during the a posteriori analysis. ring behaviors was used (Martin et al., 1993).
Both observers remained floating at the surface The analyzed sequences allowed the researchers
with their bodies oriented perpendicular to the to create a behavioral catalog (Tinbergen, 1963),
group’s travel direction, avoiding the direct path which included nonsocial and social (intraspecific
of the dolphins and letting the animals choose at and human-dolphin interactions) behaviors and
what speed and distance they approached. When sounds produced (Table 1).
dolphins slowly moved along the observers,
they swam calmly in parallel with the animals. Results
Depending on whether the dolphins stayed around
the observers or departed, recording sessions Tests with Artificial Sounds
were repeated several times on the same group by As this study is dedicated to the analysis of behav-
carefully re-orienting the boat and by informing iors, only situations when at least one dolphin was
the observers each time the dolphins swam by. A visible in the videos, at a distance of less than
recording session finished when dolphins were 10 m, were taken into account. If this dolphin
not visible for 5 min or when weather conditions emitted clicks and/or whistles during a period
prevented continued observations. with no underwater noise, then the SNR ratio
168 Lopez-Marulanda et al.
was higher than 20 dB. If the dolphin vocalizes one bottlenose dolphin body length. For distances
further away, SNR decreased and could be nega-from the device, the error of the custom program
tive. We performed our approach for positive and was 1.1 to 3.9 m (Table 3).
negative SNR (Table 2). The time differences of
arrival (TDOAs) were still correctly estimated for Data Description
SNR larger than -10 dB, which is acceptable for During 14 boat surveys, dolphins were sighted
our study because underwater noise was low com-four times, allowing collection of 21.03 min of
pared to dolphin sounds. (SNR was always posi-360° HD audio-video data with dolphins pres-
tive in our acoustic recordings.) ent. Recordings allowed the detection of 42 click
Results of the first test comparing estimations of trains and 42 whistles. The vocalizing dolphin
positions in video and audio show that differences was localized and visible on the video for seven
in estimation for azimuthal localizations are less click trains (17%). The vocalizer was not visible
than 12° except for in positions 3 and 4 (Table 3). for 25 click trains (59%); and for 10 click trains
For elevation localizations, the difference is less (24%), localization of the vocalizer was ambigu-
than 10°, except for in position 8. Positions 3, 4, ous. For whistles, localization analysis was not
and 8 can only be seen right on the edge of the possible for five whistles (12%) because of a low
image, making estimations more difficult due to SNR ratio. The vocalizing dolphin was visible on
image compression. Taking into account that the the video for 14 whistles (33%), the vocalizer was
maximal vision range of the BaBeL is estimated to not visible on the video for 18 whistles (43%), and
be 10 m depending on water clarity, a 10° differ-the localization of the vocalizer was ambiguous
ence in estimations from video and audio means for five whistles (12%).
that localization at 10 m from the BaBeL can have Three recording sessions (24 May at 0937
a maximum difference of 1.7 m from the posi-and 0949 h, and 27 May at 1316 h) were chosen
tion of the source in the video, which is less than during which it was possible to localize the
Table 1. Behavioral catalog of the dolphins observed while swimming by observers and documented with the BaBeL device
Behaviors Code Definition
Pectoral rubbing PR The dolphin touches another dolphin.
Synchronized swimming SyS Dolphins swim in synchronous manner within one body length of another dolphin,
showing parallel movements and body axes.
Swim upside down SUD The dolphin swims with its belly turned up.
Swim upside down
underneath
SUDU The dolphin swims with its belly turned up underneath a conspecific.
Side swimming SS The dolphin swims with its belly turned to the right or the left next to a conspecific.
Approach APP The dolphin approaches the observers by leaving the direction axis of its group.
Swim towards observers STO The dolphin swims towards the BaBeL device and the observers.
Turn rostrum TR The dolphin turns its rostrum in the direction of an observer.
Leave L The dolphin stops swimming towards the observers and starts to move away.
Whistle W The dolphin whistles.
Click train C The dolphin emits a click train.
Table 2. Accuracy of the time correlation on simulated signals
Simulated signal Features SNR = 20 dB SNR = -10 dB
Whistle #1 Duration: 0.1 s
Fundamental: 11 kHz
ΔTDOA = 0 ΔTDOA = 1.8 ms
Whistle #2 Duration: 0.5 s
Fundamental frequency 11 kHz
switch to 16 kHz at 0.35 s
ΔTDOA = 0 ΔTDOA = 0
169360° HD Audio-Video Device to Observe Dolphins
dolphin vocalizing to facilitate completion of neutral buoyancy, maneuverability, and simplic-
detailed analyses of dolphin behavior according to ity of deployment simultaneously. The BaBeL
the behavioral catalog (Table 1; see Appendix 1). is relatively easy to deploy from small boats to
Results show that the first animal of the group record behavior and acoustic data on free-ranging
to approach the observers did not produce click dolphins and can also be used with delphinids
trains. A click train was made after an approach under human care. Contrary to other hydrophone
and/or movement of the rostrum towards the arrays, the BaBeL system can be used to detect,
device: in the first observation, the click train was locate, and track dolphins emitting sounds in a 3D
emitted by the second individual after it turned space. The hydrophone arrays of Au & Herzing
its rostrum towards the device (see Appendix 1, (2003), Schotten et al. (2004), and Ball & Buck
Figure 1). In the second observation, the click (2005) all present hydrophones in the same plane,
train was emitted by the last individual after it making it impossible to discriminate from the
approached the device (see Appendix 1, Figure 2; audio recording if the emitter dolphin was in front
video available on the Aquatic Mammals web-of or behind the device. The design of our system
site:
www.aquaticmammalsjournal.org/index.php?
places hydrophones in different planes, allow-
option=com_content&view=article&id=10&
ing us to determine the position of the vocalizing
Itemid=147
). In the third observation, the click dolphin regardless of its direction of approach to
train was emitted by the second individual after it the observers, and the wide-angle HD 360° video
approached the device (see Appendix 1, Figure 3). cameras provide information to localize to an
The same individual produced the four whistles identified vocalizing dolphin visually. When ani-
presented in our second sequence; the first whistle mals are in the visual range of the camera, this
was emitted before the approach, and the three 360° audio-video system could greatly increase
others were emitted after leaving. This animal the number of vocalizations that can be attributed
produced no whistles while swimming towards to an individual dolphin.
the observers. Simultaneous visual and acoustic recordings
are necessary for localizing to a vocalizing dol-
Discussion phin. This system is mainly limited by visual
detection, which depends not only on water clar-
The results with the BaBeL system are promis-ity but also on the wide-angle video cameras.
ing for the study of dolphin behavior. Its accu-Wide-angle lenses affect the perspective by exag-
racy using simulated underwater sounds in a gerating the distance between objects. They make
pool was validated. The BaBeL design during subjects at moderate and far distances seem fur-
field testing was verified: the device achieved ther away than they really are. Consequently, only
Table 3. Localization performance of our custom-made program using the sound produced by two bars of steel during tests
in a pool
Position estimated from the videos
Position estimated from
the acoustic recordings Difference in estimation
Position Azimuth
(°)
Elevation
(°)
Distance
(m)
Azimuth
(°)
Elevation
(°)
Distance
(m)
Azimuth
(°)
Elevation
(°)
Distance
(m)
133.0 5.3 4.7 21.4 3.4 6.3 11.6 1.9 -1.6
25.7 1.2 4.5 6.2 -2.6 0.8 -0.5 3.8 3.7
3328.8 3.5 4.7 349.6 1.1 0.8 -20.8 2.4 3.9
4315 3.1 3.3 349.3 -3.1 1.5 -34.3 6.2 1.8
58.0 -6.1 3.0 2.9 -7.9 0.8 5.1 1.8 2.2
643.6 2.3 3.3 36.3 4.2 1.3 7.3 -1.9 2.0
761.6 5.6 2.1 55.4 -2.8 0.7 6.2 8.4 1.4
86.8 0.7 1.5 0.6 12.5 0.4 6.2 -11.8 1.1
9304.4 -0.7 2.1 303.1 -7.7 0.5 1.3 7.0 1.6
Mean -2.0 2.0 1.8
SD 15.2 6.0 1.6
170 Lopez-Marulanda et al.
dolphins vocalizing near BaBeL (within 10 m) Acknowledgments
were visually and acoustically detected. As previ-
ously demonstrated by Watkins & Shevill (1974), FD’s contribution was partially funded by
the accuracy of this acoustic localization system Focused on Nature. Parc Asterix funded previ-
should decrease as the distance of dolphins from ous preliminary fieldwork of FD and JLM and
the device increases. To improve accuracy, the dis-funded TB’s work. Materials and fieldwork were
tance between the hydrophones can be increased, financed by the CERECAR project from Abyss
but this would reduce system maneuverability. Foundation (La Réunion, France). We are very
Therefore, it is recommended that this device only grateful to the Parc Asterix Dolphinarium curator,
be used in clear water and preferably with dolphin Birgitta Mercera, and her trainers for their help
populations habituated to the presence of human in testing previous versions of our device within
swimmers, or with dolphins under human care. their group of dolphins. We would like to thank
The five possible situations that observers Abyss Foundation volunteers and Emmanuel
might encounter while using BaBeL are sum-Antongiorgi for their support in collecting
marized in Appendix 2. Contrary to using a regu-data, and Isabelle Charrier and the Bioacoustic
lar camera, an observer using such a 360° video Communication Team of NeuroPSI for her con-
system increases the possibility of capturing structive comments during the development of
ongoing behaviors regardless of his position with this work. We thank Isabella Clegg for revising
respect to the animals and his concentration level, the English of this manuscript. Finally, we would
thus reducing human error. In 59% of the detected like to give special thanks to editor Dr. Kathleen
click trains and in 43% of detected whistles, this Dudzinski for her advice and encouragement to
device could acoustically and visually detect dol-improve this manuscript.
phins, but the customized program did not point to
one of the dolphins present on the video, meaning Literature Cited
that the dolphin emitting the sound was out of our
range or vision. Therefore, in these cases, assign-Agreement on the Conservation of Cetaceans in the Black
ing the recorded sound to any of the dolphins Sea Mediterranean Sea and Contiguous Atlantic Area
present on the video would have been a mistake, (ACCOBAMS), Resolution 4.18. (n.d.). Guidelines on
demonstrating the necessity of the use of hydro-the granting of exceptions to Article II, paragraph 1, for
phone arrays to aid in interpreting dolphins’ vocal the purpose of non-lethal in situ research in the agree-
behavior location to avoid erroneous assump-ment area. Retrieved from www.accobams.org
tions about the identity of the vocalizer. When Alameda-Pineda, X., & Horaud, R. (2014). A geometric
dolphins were detected visually and acoustically, approach to sound source localization from time-delay
and the customized program pointed to one indi-estimates. IEEE/ACM Transactions on Audio, Speech,
vidual dolphin present in the video, it was found and Language Processing, 22(6), 1082-1095. https://
that the echolocating dolphin was never the first doi.org/10.1109/TASLP.2014.2317989
one of the group. Moreover, in one case, it was Altmann, J. (1974). Observational study of behavior:
found that one dolphin emitted a click train while Sampling methods. Behaviour, 49, 227-265. https://doi.
its conspecific swimming behind turned its head org/10.1163/156853974X00534
to observers. This particular dolphin might have Au, W. W. L. (1993). The sonar of dolphins. New York:
been eavesdropping on the returning echoes of Springer-Verlag. https://doi.org/10.1007/978-1-4612-
its echolocating conspecific (Gregg et al., 2007). 4356-4
Finally, echolocating dolphins swam dorsal side Au, W. W. L., & Fay, R. R. (Eds.). (2012). Hearing by
up or ventral side up, suggesting that the ani-whales and dolphins (Vol. 12). New York: Springer
mals use different body postures while produc-Science & Business Media.
ing acoustic signals. Further investigations of the Au, W. W. L., & Herzing, D. L. (2003). Echolocation sig-
body postures dolphins display while vocalizing nals of wild Atlantic spotted dolphin (Stenella fronta-
would enable better understanding of their sound lis). The Journal of the Acoustical Society of America,
production and their communication system. 113(1), 598-604. https://doi.org/10.1121/1.1518980
The BaBeL system offers a method of data col-Ball, K. R., & Buck, J. R. (2005). A beamforming video
lection to conduct an etho-acoustical analysis of recorder for integrated observations of dolphin behav-
bottlenose dolphin sound emissions potentially to ior and vocalizations. The Journal of the Acoustical
be associated with individual dolphins and their Society of America, 117 (3, Pt 1), 1005-1008. https://doi.
underwater behaviors. The BaBel and the associ-org/10.1121/1.1831284
ated software algorithms for data analysis repre-Blanchard, T. (2015). Innovation audio-visuel pour le
sent an improved tool for ethologists to record and suivi des grands dauphins [Audio-visual innovation for
collect data on all dolphins present in a 360° space monitoring bottlenose dolphins] (Master’s thesis). Ecole
via focal and group follows.
171360° HD Audio-Video Device to Observe Dolphins
Nationale Supérieure d’Ingenieurs du Mans, Le Mans, Gregg, J. D., Dudzinski, K. M., & Smith, H. V. (2007).
France. Do dolphins eavesdrop on the echolocation signals of
Brensing, K., Linke, K., & Todt, D. (2001). Sound source conspecifics? International Journal of Comparative
location by difference of phase, on a hydrophone array Psychology, 20(1), 65-88.
with small dimensions. The Journal of the Acoustical Herzing, D. L. (1996). Vocalizations and associated
Society of America, 109(1), 430-433. https://doi. underwater behavior of free-ranging Atlantic spotted
org/10.1121/1.1333421 dolphins, Stenella frontalis and bottlenose dolphins,
Caldwell, M. C., Caldwell, D. K., & Tyack, P. L. (1990). Tursiops truncatus. Aquatic Mammals, 22(2), 61-79.
Review of the signature whistle hypothesis for the Janik, V. M. (1999). Origins and implications of vocal learn-
Atlantic bottlenose dolphin. In S. Leatherwood & R. R. ing in bottlenose dolphins. In H. O. Box & K. R. Gibson
Reeves (Eds.), The bottlenose dolphin: Recent progress (Eds.), Mammalian social learning: Comparative and
in research (pp. 199-234). San Diego: Academic Press. ecological perspectives (pp. 308-326). Cambridge, UK:
Connor, R. C., Mann, J., & Watson-Capps, J. (2006). A Cambridge University Press.
sex-specific affiliative contact behavior in Indian Ocean Janik, V. M. (2000). Food-related bray calls in wild bottle-
bottlenose dolphins, Tursiops sp. Ethology, 112(7), 631- nose dolphins (Tursiops truncatus). Proceedings of the
638. https://doi.org/10.1111/j.1439-0310.2006.01203.x Royal Society B: Biological Sciences, 267(1446), 923-
Connor, R. C., Mann, J., Tyack, P. L., & Whitehead, H. 927. https://doi.org/10.1098/rspb.2000.1091
(1998). Social evolution in toothed whales. Trends in Janik, V. M. (2009). Acoustic communication in delphi-
Ecology and Evolution, 13(6), 228-232. https://doi. nids. In M. Naguib & V. M. Janik (Eds.), Advances in
org/10.1016/S0169-5347(98)01326-3 the study of behavior (1st ed., Vol. 40, pp. 123-157).
Connor, R. C., Wells, R. S., Mann, J., & Read, A. J. (2000). New York: Academic Press. https://doi.org/10.1016/
The bottlenose dolphin: Social relationships in a fission- S0065-3454(09)40004-4
fusion society. In J. Mann, R. C. Connor, P. L. Tyack, King, S. L., & Janik, V. M. (2015). Come dine with me:
& H. Whitehead (Eds.), Cetacean societies: Field Food-associated social signaling in wild bottlenose dol-
studies of dolphins and whales (pp. 91-126). Chicago: phins (Tursiops truncatus). Animal Cognition, 18(4),
University of Chicago Press. 969-974. https://doi.org/10.1007/s10071-015-0851-7
Dulau-Drouot, V., Boucaud, V., & Rota, B. (2008). Cetacean López, B. D., & Bernal-Shirai, J. A. B. (2009).
diversity off La Réunion Island (France). Journal Mediterranean common bottlenose dolphin’s repertoire
of the Marine Biological Association of the United and communication use. In Dolphins: Anatomy, behav-
Kingdom, 88(6), 1263-1272. https://doi.org/10.1017/ ior, and threats (pp. 129-148). New York: Nova Science
S0025315408001069 Publishers, Inc.
Dudzinski, K. A., Clark, C. W., & Würsig, B. (1995). López-Rivas, R. M., & Bazúa-Durán, C. (2010). Who is
A mobile video/acoustic system for simultaneous whistling? Localizing and identifying phonating dol-
underwater recording of dolphin interactions. Aquatic phins in captivity. Applied Acoustics, 71(11), 1057-
Mammals, 21(3), 187-194. 1062. https://doi.org/10.1016/j.apacoust.2010.05.006
Dudzinski, K. M., Gregg, J. D., Ribic, C. A., & Kuczaj II, Mann, J. (1999). Behavioral sampling meth-
S. A. (2009). A comparison of pectoral fin contact ods for cetaceans: A review and critique.
between two different wild dolphin populations. Marine Mammal Science, 15(1), 102-122.
Behavioural Processes, 80(2), 182-190. https://doi. https://doi.org/10.1111/j.1748-7692.1999.tb00
org/10.1016/j.beproc.2008.11.011 784.x
Fellner, W., Bauer, G. B., Stamper, S. A., Losch, B. A., & Mann, J., Connor, R. C., Barre, L. M., & Heithaus, M. R.
Dahood, A. (2013). The development of synchronous (2000). Female reproductive success in bottlenose dol-
movement by bottlenose dolphins (Tursiops truncatus). phins (Tursiops sp.): Life history, habitat, provisioning,
Marine Mammal Science, 29(3), E203-E225. https://doi. and group-size effects. Behavioral Ecology, 11, 210-
org/10.1111/j.1748-7692.2012.00609.x 219. https://doi.org/10.1093/beheco/11.2.210
Fripp, D. (2005). Bubblestream whistles are not repre-Marten, K., Shariff, K., Psarakos, S., & White, D. J.
sentative of a bottlenose dolphin’s vocal repertoire. (1996). Ring bubbles of dolphins. Scientific American,
Marine Mammal Science, 21, 29-44. https://doi. 275(2), 82-87. https://doi.org/10.1038/scientificameri-
org/10.1111/j.1748-7692.2005.tb01206.x can0896-82
Fripp, D. (2006). Bubblestream whistles are not represen-Martin, P., Bateson, P. P. G., & Bateson, P. (1993). Measuring
tative of bottlenose dolphin whistle repertoires reply behaviour: An introductory guide. Cambridge, UK:
to McCowan. Marine Mammal Science, 22, 496-501. Cambridge University Press. https://doi.org/10.1017/
https://doi.org/10.1111/j.1748-7692.2006.00028.x cbo9781139168342
Gibson, Q. A., & Mann, J. (2008). The size, composition McCowan, B., & Reiss, D. (1995). Maternal aggressive con-
and function of wild bottlenose dolphin (Tursiops sp.) tact vocalizations in captive bottlenose dolphins (Tursiops
mother-calf groups in Shark Bay, Australia. Animal truncatus): Wide-band, low-frequency signals during
Behaviour, 76, 389-405. https://doi.org/10.1016/j.anbe-mother/aunt-infant interactions. Zoo Biology, 14(4), 293-
hav.2008.01.022 309. https://doi.org/10.1002/zoo.14301404022
172 Lopez-Marulanda et al.
Nowacek, D. P., Tyack, P. L., Wells, R. S., & Johnson, M. P. Acoustical Society of America, 109, 397-406. https://doi.
(1998). An onboard acoustic data logger to record bioso- org/10.1121/1.1329619
nar of free-ranging bottlenose dolphins. The Journal of Watkins, W. A., & Schevill, W. (1972). Sound source
the Acoustical Society of America, 103(5), 2908. https:// location with a three-dimensional hydrophone
doi.org/10.1121/1.422063 array. Deep-Sea Research, 19, 691-706. https://doi.
Pryor, K. W. (1990). Non-acoustic communication in small org/10.1016/0011-7471(72)90061-7
cetaceans: Glance, touch, position, gesture, and bubbles. Watkins, W. A., & Schevill, W. E. (1974). Listening to
In J. A. Thomas & R. A. Kastelein (Eds.), Sensory abili-Hawaiian spinner porpoises, Stenella cf. longirostris,
ties of cetaceans (pp. 537-544). New York: Springer. with a three-dimensional hydrophone array.
https://doi.org/10.1007/978-1-4899-0858-2_37 Journal of Mammalogy, 55(2), 319-328. https://doi.
Quick, N. J., Rendell, L. E., & Janik, V. M. (2008). A org/10.2307/1379001
mobile acoustic localization system for the study of free-
ranging dolphins during focal follows. Marine Mammal Appendix 1
Science, 24(4), 979-989. https://doi.org/10.1111/j.1748-
7692.2008.00231.x Etho-Acoustical Description
Sakai, M., Hishii, T., Takeda, S., & Konshima, S. (2006).
First Observation—In this sequence (34 s),
Flipper rubbing behaviors in wild bottlenose dolphins
we analyzed the behaviors of nine adult dolphins
(Tursiops aduncus). Marine Mammal Science, 22(4), 966-
(Ind1 to Ind9) and one calf (Ind10) (Figure 1).
978. https://doi.org/10.1111/j.1748-7692.2006.00082.x
The first dolphin (Ind1) approached and swam
Sakai, M., Morisaka, T., Kogi, K., Hishii, T., & Kohshima,
towards the BaBeL for 15 s; and at 7 and 9 s,
S. (2010). Fine-scale analysis of synchronous breathing
Ind1 turned its rostrum towards the BaBel and
in wild Indo-Pacific bottlenose dolphins (Tursiops adun-
continued swimming towards the observers for
cus). Behavioural Processes, 83(1), 48-53. https://doi.
6 s more before leaving. Ind2 approached and
org/10.1016/j.beproc.2009.10.001
swam towards the device for 11 s before leav-
Sayigh, L., Tyack, P., Wells, R., & Scott, M. (1990).
ing, emitting a click train at 4 s. Ind3 swam by
Signature whistles of free-ranging bottlenose dolphins
the device. Ind4 swam upside down underneath
Tursiops truncatus: Stability and mother-offspring com-
Ind5 for 15 s then moved to a synchronized swim-
parisons. Behavioral Ecology and Sociobiology, 26(4),
ming position above Ind 5. Ind5 swam synchro-
247-260. https://doi.org/10.1007/BF00178318
nously for 14 s above Ind4 and then moved to a
Schotten, M., Au, W. W. L., Lammers, M. O., & Aubauer,
side swimming posture for 1 s before returning to
R. (2004). Echolocation recordings and localization of
synchronized swimming below Ind4. Ind6 swam
wild spinner dolphins (Stenella longirostris) and pan-
synchronously below Ind7 for 14 s; at 11 s, Ind6
tropical spotted dolphins (S. attenuata) using a four-
turned its head to observers and continued swim-
hydrophone array. In J. A. Thomas, C. F. Moss, & M.
ming synchronously below Ind7. At 19 s, Ind6
Vater (Eds.), Echolocation in bats and dolphins (pp.
swam upside down underneath Ind7; at 20 s, Ind6
393-399). Chicago: The University of Chicago Press.
conducted pectoral fin-to-pectoral fin rubbing to
Thomas, R. E., Fristrup, K. M., & Tyack, P. L. (2002).
Ind7 for 7 s. At 27 s, Ind6 continued swimming
Linking the sounds of dolphins to their locations and
synchronously below Ind7, and Ind7 swam syn-
behavior using video and multichannel acoustic record-
chronously with Ind6 and touched Ind6 with its
ings. The Journal of the Acoustical Society of America,
pectoral fin on two occasions for 2 s at 14 s on the
112 (4), 1692-1701. https://doi.org/10.1121/1.1494805
body and then for 1 s at 26 s on the belly. At 27
Tinbergen, N. (1963). On aims and methods of ethology.
s, Ind6 and Ind7 stopped body contact and started
Zeitschrift für Tierpsychologie, 20, 410-433. https://doi.
synchronized swimming next to each other. Ind8
org/10.1111/j.1439-0310.1963.tb01161.x
swam by the observers. Finally, Ind9 and Ind10
Tsai, Y-J. J., & Mann, J. (2013). Dispersal, philopatry, and
swam synchronously by the BaBel next to each
the role of fission-fusion dynamics in bottlenose dol-
other.
phins. Marine Mammal Science, 29(2), 261-279. https://
Second Observation—In the second sequence
doi.org/10.1111/j.1748-7692.2011.00559.x
(40 s), five adult dolphins passed in front of the
Tyack, P. L. (1991). Use of a telemetry device to iden-
BaBeL system. The first and second dolphin (Ind1
tify which dolphin produces a sound. In K. Pryor &
and Ind2) approached synchronously. Ind1 swam
K. S. Norris (Eds.), Dolphin societies (pp. 319-344).
towards the observers for 4 s, pointed its rostrum
Berkeley: University of California Press.
towards the recording system with no sound
Van Lancker, E. (2002). Acoustic goniometry: A spatio-
being detected, and left. Ind2 also swam towards
temporal approach (Doctoral dissertation). Faculté
the observers for 5 s, pointed its rostrum towards
Polytechnique de Mons.
the observers with no sound being detected, and
Wahlberg, M., Mohl, B., & Madsen, P. (2001). Estimating
left. Following behind the first two dolphins,
source position accuracy of a large aperture hydro-
Ind3 whistled, swam towards the observers for 2
phone array for bioacoustics. The Journal of the
s, and left. After leaving, Ind3 emitted three more
173360° HD Audio-Video Device to Observe Dolphins
whistles before disappearing out of the range of
Third Observation—In this 22 s sequence, six
vision of the video system. Ind4 swam by the
adult dolphins passed from left to right in front
BaBel, while Ind5 approached and swam towards
of the BaBeL (Figure 8). Ind1 swam less than 50
the observers for 4 s, emitted a click train, and
cm distance above Ind2. Ind2 swam upside down
continued swimming towards the observers for 5 s underneath Ind1 for 16 s; at 3 s, Ind2 turned its
before leaving (Figure 2). rostrum towards the observers and, at 8 s, emitted
Figure 1. First observation: (a) Screenshot of the location of the clicking dolphin (Ind2) in 360° video and backup video with
the red cross pointing to the source of the sound emission; and (b) timelines for ten individuals.
174 Lopez-Marulanda et al.
Figure 2. Second observation: (a) Screenshot of the location of the whistling dolphin (Ind3) in 360° video and backup video
with the red cross pointing to the source of the sound emission; (b) screenshot of the location of the clicking dolphin (Ind5)
in 360° video and backup video; and (c) timelines for five individuals (Ind1 to Ind5).
175360° HD Audio-Video Device to Observe Dolphins
a click train. At 11 s, Ind2 rubbed Ind1’s belly upside down underneath Ind4 for 4 s and then con-
with its pectoral fin for 8 s. At 19 s, Ind2 stopped tinued swimming synchronously with Ind4. Ind4
its contact with Ind1 and swam synchronously pectoral fin rubbed Ind5’s belly for 1 s at 3 s and
above it until the end of the sequence. then continued swimming synchronously with Ind5
Ind3 swam by the BaBel, and Ind4 synchro-and Ind6 (Figure 3).
nously swam next to Ind5 and Ind6. Ind5 swam
Figure 3. Third observation: (a) Screenshot of the location of the clicking dolphin (Ind3) in 360° video and backup video
with the red cross pointing to the source of the sound emission; and (b) timelines for six individuals.
176 Lopez-Marulanda et al.
Figure 1. Use of focal-animal sampling technique. Top left: No sound detected and dolphins out of the cameraʼs visual
range; Top middle: Sound detected but dolphins out of the cameraʼs visual range; Top right: No sound detected but dolphins
in the visual range of the camera; Bottom left: Dolphin in the visual range of the camera and sound detected, but the dolphin
vocalizing is not present in the video; and Bottom right: Dolphin in the visual range of the camera, with the sound detected
and the vocalizing dolphin present in the video.
Appendix 2
The five possible situations observers could encounter while operating the BaBeL.
