Mapping cetacean sounds using a passive acoustic

monitoring system towed by a Wave Glider on the

Southwestern Atlantic Ocean

Lis Bittencourta, William Soares-Filhob, Isabela Maria Seabra de Limaa,Sudhir Paic, Jose Lailson-Brito Jr.a, Leonardo Martins Barreirab, Alexandre

Freitas Azevedoa, Luiz Alexandre A. Guerrad

aLaboratorio de Mamıferos Aquaticos e Bioindicadores (MAQUA), Universidade doEstado do Rio de Janeiro (UERJ), Rio de Janeiro, RJ, Brazil

bInstituto de Pesquisas da Marinha (IPqM), Rio de Janeiro, RJ, BrazilcSchlumberger Robotics Services, Houston, TX, USA

dPETROBRAS/CENPES, Av. Horacio Macedo, 950 - Cidade Universitaria, Rio deJaneiro, RJ, 21941-915, Brazil

Abstract

Passive acoustic monitoring techniques have been shown to be useful to mon-

itor marine fauna efficiently since they do not depend on good visibility

and weather conditions. An unmanned surface vehicle, type Wave Glider,

equipped with an autonomous recording system collected acoustic data on the

Brazilian offshore waters of the Southwestern Atlantic Ocean between 6–25

February 2016. Nearly continuous data was provided during this period (light

and dark hours) for the 377 km traveled. There were 165184 high-frequency

detections in 31 events, being 1839 whistles, 389 burst-pulse sounds and

162956 echolocation clicks. Low-frequency detections summed up to 705, all

tonal signals, in 5 events. Although high-frequency detections occurred at all

hours of the day, the majority was at dark hours. Low-frequency detections

Email address: laguerra@petrobras.com.br (Luiz Alexandre A. Guerra)

Preprint submitted to Deep-Sea Research Part I May 12, 2018

did not occur at all hours of the day, but the majority also occurred during

dark hours. High-frequency and low-frequency detections represented 26%

and 3% of the recording hours, respectively. Duration, the number of emis-

sions, and ocean depth varied among acoustic events. Events composed by

high-frequency sounds were separated into seven different groups, probably

different species. One of the events presented whistles that were previously

recorded in the Rio de Janeiro Coast for Steno bredanensis (Rough-toothed

dolphins). All of the low-frequency events were composed of a type of Bal-

aenoptera brydei (Bryde’s whales) call. This type of data collection is un-

precedented for the Southwestern Atlantic Ocean and highlighted the use of

the sampled area by delphinids on different days and different times of the

day, including the dark hours.

Keywords: Passive Acoustic Monitoring, Cetaceans, Unmanned Surface

Vehicle, Wave Glider, South Atlantic Ocean, Brazil, Upwelling

1. Introduction1

The continental shelf edge is a productive ocean environment, and the oc-2

currence of large marine vertebrates is often associated with oceanographic3

features (Baumann-Pickering et al., 2016). Most studies focused on assessing4

diversity are performed with ship surveys, relying mostly on visual observa-5

tion data and thus restricted to animals that can be seen from the sur-6

face. Passive acoustic monitoring (PAM) can bring reliable information on7

marine vertebrates movements because many aquatic animals have species-8

specific sound emissions and can be acoustically identified (Klinck et al.,9

2012; Luczkovich et al., 2008; Oswald et al., 2007). Therefore, PAM tech-10

2

niques have been shown efficient to find and monitor marine fauna, allowing11

scientists to identify patterns of seasonal variation and habitat use in distinct12

environments (Baumgartner and Fratantoni, 2008; Bittencourt et al., 2016;13

Lammers et al., 2008; Locascio, 2010; Luczkovich et al., 2008; Moore et al.,14

2006; Wall et al., 2013; Wiggins et al., 2013; Wilson et al., 2013).15

Autonomous data samplers are capable of performing monitorings for16

long periods at a low cost and are non-invasive to the environment, allow-17

ing data collection even during rough weather conditions at remote locations18

worldwide. These recorders can either be fixed, moored to the marine floor19

or ocean buoys (Sousa-Lima et al., 2013; Van Parijs et al., 2009), or mo-20

bile, coupled to autonomous vehicles, like ocean gliders (Baumgartner and21

Fratantoni, 2008; Wall et al., 2012). While submarine buoyancy-driven glid-22

ers can trace vertical and horizontal paths on the water column collecting23

data, Wave Gliders harness energy from waves to endure long cruises trans-24

porting sensors close to the surface (Rudnick et al., 2004; Hine et al., 2009).25

Traditionally, gliders have been deployed to collect a variety of chemical, bi-26

ological and physical oceanographic variables (Villareal and Wilson, 2014;27

Martin et al., 2009). More recently, studies with gliders started collecting28

acoustical data (Baumgartner and Fratantoni, 2008; Wall et al., 2017, 2013;29

Bingham et al., 2012), opening a window to the soundscape of vast oceanic30

areas.31

Data regarding marine vertebrates’ occurrence in the South Atlantic Ocean32

is still scarce. Previous studies have identified several species of whales and33

dolphins in the waters covering the continental shelf and the continental slope34

of South America (Di Tullio et al., 2016a; Zerbini, 2004), but efforts are con-35

3

centrated on visual surveys or stranding data. In the present study, we used36

an unmanned surface vehicle coupled to a passive acoustic monitoring system37

to sample offshore waters in a region of the Brazilian coast well known for38

upwelling occurrence, searching for possible patterns of acoustical presence39

of cetaceans.40

2. Materials and Methods41

Acoustic data collection occurred through the deployment of an unmanned42

surface vehicle type Wave Glider model SV2 on the Brazilian offshore wa-43

ters of the southwestern Atlantic during the austral summer of 2016. The44

Wave Glider used for this study consists of a submerged glider with wings45

connected by a 7-m umbilical to a float element that remains at the surface.46

The Wave Glider converts wave motion into forwarding thrust, while solar47

panels charge onboard batteries that power the vehicle’s command and con-48

trol system, communications and sensors (Hine et al., 2009; Carragher et al.,49

2013).50

The recording system was composed of a Reson/Teledyne TC4014-5 hy-51

drophone (-180 dB re 1V/µPa, 15 Hz–480 kHz) and a SAIL Decimus R©digital52

acquisition and processing system (sampling rate of 50 kHz in 16 bits, with53

26 dB gain) housed in a tow-fish attached to the glider by a 15-m cable54

assembly with floats and weights fitted to its length (Figure 1).55

The Wave Glider traveled a distance of 377 km between 6–25 February,56

performing constant and uninterrupted recordings, generating 24 one-hour57

files per day. The only two occasions in which a whole 24-hour period was58

not sampled were on 6 February, on which nine hours were sampled because59

4

the equipment deployment occurred in the middle of the day; and on 1260

February, on which there was an equipment failure, and recordings were61

restricted to nine non-consecutive hours.62

The local depths were obtained from GEBCO-2014 Grid, available at63

http://www.gebco.net (Weatherall et al., 2015).64

Sound analyses were performed in Raven 1.5 (Program, 2011). Five de-65

tectors were employed to look for signals of biological origin in two differ-66

ent frequency scales (Table 1). Frequency and duration parameters used67

in each detector configuration were chosen based on literature available on68

marine fauna sound signals. Detectors with spectrograms of 512-point Han-69

ning windows and 50% overlap were used to find tonal and pulsed sounds70

above 2 kHz, which presented characteristic frequencies and durations of sig-71

nals produced by delphinids. Recordings were decimated from 50 kHz to72

6 kHz sampling rate with tuneR package (Krey et al., 2011) from R soft-73

ware so that low frequency sounds produced by whales and fishes could be74

found. For these sounds, we used detectors in Raven 1.5 with spectrograms75

of 512-point Hanning window and 75% overlap. Although each detector had76

different frequency and duration parameters, signal-to-noise ratio above 5 dB77

was maintained for all five detectors.78

All the detections were inspected by two experienced observers (LB and79

IMSL) and classified as positive or negative. Detection was considered neg-80

ative when it was identified as being a signal of non-biological origin (e.g.,81

glider generated noise, shipping generated noise) and positive when it could82

be identified as a signal of biological origin. The positive detections were then83

classified according to the signals characteristics: tonal signals, burst-pulsed84

5

sounds, echolocation click as shown in Figure 2. High-frequency tonal signals85

were referred to as whistles. A single detection event may sometimes include86

more than one type of sound, and in these cases, all sounds were counted.87

Consecutive detections with a time interval shorter than one hour were con-88

sidered as belonging to the same detection event, meaning an encounter with89

a group of animals or an encounter with single vocalizing animals. In case90

only one vocalization was detected and separated by more than one hour91

from others, this detection was accounted for in the number of total detec-92

tions but not considered as an individual detection event. It is unlikely that a93

group of delphinids would produce only a single sound emission of any kind,94

indicating that it belonged to another event in which the group of delphinid95

was too far from the glider.96

Detection events were quantified and mapped to obtain the spatial dis-97

tribution of encounters with animals in the sampling area. Temporal distri-98

bution of detection events was also observed by quantifying the number of99

detections per hour and per day. To investigate if the number of detections100

differed through different hours of the day, a Mann-Whitney U Test was used101

to compare the number of whistles, burst-pulses and echolocation clicks be-102

tween light hours and dark hours, after verifying data distribution was not103

normal.104

Of the detected signals, we analyzed the whistles with clear and strong105

contours (Oswald et al., 2004) that did not surpass the superior frequency106

limit of the recording system and that did not overlap with other signals.107

Seven acoustic parameters were extracted from tonal signals, as follow: initial108

frequency (IF), ending frequency (EF), minimum frequency (MinF), maxi-109

6

mum frequency (MaxF), delta frequency (DF), duration (DUR), number of110

inflection points (INF).111

We used the principal component analysis (PCA) to estimate the possible112

number of species recorded in the area. This analysis finds the combination113

of acoustic variables that is responsible for as much variation as possible114

(Quinn and Keough, 2002). The factors with the highest eigenvalues were115

considered to be the ones responsible for most of the variations. The corre-116

lations between each variable and the main factors of the PCA were used to117

define which of these variables were the most important. The main compo-118

nents were then used in place of the original acoustic variables in a posterior119

discriminant function analysis (DFA) so that the assumption of homoscedas-120

ticity of the DFA could be met (Quinn and Keough, 2002). The means of121

the canonical variables of the centroids of each event in relation to the two122

main roots of the DFA was plotted. The position of each centroid relative to123

each other and the roots were used to define the estimated number of species124

recorded.125

Also, an effort to identify the species, or at least the taxa, through visual126

inspection of detected signals was performed. Good quality signals were127

selected and had their contour visually compared to the bibliography and to128

the available delphinid signals in the MAQUA sound database.129

3. Results130

3.1. Temporal Distribution and Mapping131

In total, there were 165184 high-frequency detections, of which 165164132

were in detection events, and 20 were in isolated occurrences. Echolocation133

7

clicks were the most common high-frequency sound, with 162956 detections,134

followed by 1839 whistles and 389 burst-pulsed sounds. Low-frequency detec-135

tions summed up to 705, being all tonal signals. High-frequency detections136

occurred on all days of the sampling period, representing 26% of all record-137

ing hours (Fig. 3). In contrast, low-frequency detections occurred on five138

days of the sampling period, representing 3% of all recording hours (Fig. 3).139

High-frequency detections occurred in all hours of the day at least once, but140

the majority of detections occurred during dark hours with peak detection141

time being 04:00 am (Fig. 4). Low-frequency detections did not occur in all142

hours of the day, but the majority of detections also occurred during dark143

hours, possibly biased by a large detection peak at 02:00 am (Fig. 4).144

There were 31 high-frequency detection events (HF1 to HF31), which145

lasted from one up to eleven hours (Table 2). The composition of sound146

types varied among detection events. There were events composed solely of147

echolocation clicks and events composed solely of whistles, but both were of148

rare occurrence. Events composed of multiple sound types were the most149

common. The number of echolocation clicks did not differ between light150

hours and dark hours events (Mann-Whitney U, p > 0.05), but there was a151

difference for whistles (Mann-Whitney U, z = −2.09, p = 0.03) and burst-152

pulses (Mann-Whitney U, z = −2.49, p = 0.01), which were more numerous153

during dark hours.154

Mapping of detection events showed that delphinid encounters occurred155

across the entire sampling area (Fig. 5). Because the Wave Glider moved156

constantly, longer encounters covered larger distances.157

There were 5 low-frequency detection events (LF1 to LF5), which lasted158

8

from one up to three hours (Table 3), with only tonal signals of the same type159

occurring. Less frequent than high-frequency events, low-frequency events160

covered smaller areas as shown on the map in Figure 6.161

3.2. Species Identification162

The first three components of the PCA presented the highest eigenvalues163

and were responsible for 80.4% of the variation found in the data. The164

first component presented 41.7% of the total variance and the second, 19.9%165

of the variance. The frequency variables, especially maximum frequency,166

ending frequency and minimum frequency presented high correlations (> 0.7)167

with the first component. The number of inflection points presented high168

correlation, while delta frequency, minimum frequency, and duration had a169

moderate correlation (between 0.4 and 0.6) with the second component.170

The PCA plot of the events shows that there is overlap among several171

events while some do not mix. That is the case of HF5 and HF11 that are172

apart between themselves and from most of the other events (Figure 7). The173

plot of DFA centroids shows that events were separated into seven different174

groups, probably seven different species (Figure 8). Although the centroids175

of HF5, HF11 and HF12 were separated by both roots, the remaining events176

were grouped, which means that they share similar whistle characteristics177

and could be recordings from the same delphinid species. Examples of this178

are the group formed by HF7 and HF26, and the one formed by HF31, HF4,179

HF1, and HF21.180

The whistles from one of the detection events could be identified as part181

of the acoustic repertoire from rough-toothed dolphins, Steno bredanensis182

(Fig. 9). The whistle contour types registered in this study had positive183

9

matches with whistles found both in the literature and in the MAQUA sound184

database.185

All low-frequency sounds were identified as being a type of Brydes whale186

(Balaenoptera brydei) call (Fig. 10). The calls were all descendent in fre-187

quency, with: mean minimum frequency of 71.2 ± 8.7 Hz; mean maximum188

frequency of 182.0 ± 11.8 Hz; mean delta frequency of 110.8 ± 14.2 Hz; and189

mean duration of 442.8 ± 115.5 ms. The calls occurred in bouts of two or190

three, with bouts usually separated by two seconds on average, repeated191

during approximately one hour.192

4. Discussion193

Our results demonstrated that delphinid species used the studied area194

daily during the sampling period. Since recordings were continuous, except195

for two system failure occasions, the observed temporal distribution is free of196

recording cycle effect (Riera et al., 2013). Despite the fact that previous stud-197

ies had already identified the presence of delphinid species in the study area198

using visual observer methods (Di Tullio et al., 2016a), the higher number of199

delphinid acoustic detections during dark hours in this study highlights the200

importance of using PAM to assess cetacean occurrence. This methodology201

can be used independent of good light and weather conditions, which are202

necessary for visual surveys.203

The number of hours and sound emissions found in each detection event204

varied and this could be either due to delphinid group size, to the distance205

in which the animals were to the Wave Glider, or even to different acoustic206

behavior from different species. Events that lasted longer and presented a207

10

higher number of sound emissions could be composed by larger groups of208

delphinids and could also be involved in higher activity rates (dos Santos209

et al., 2005). Some of the detection events grouped together in statistical210

analysis not only have similar whistle characteristics but also occurred during211

similar times of the day, such as events HF8 and HF19. Temporal patterns212

of vocalization behavior can differ among odontocetes, with certain species213

producing more sound emissions during the night while others produce more214

during the day (Baumann-Pickering et al., 2016). Although the number215

of detected echolocation clicks did not differ between light and dark hour216

events, the number of detected whistles and burst-pulses did, corroborating217

that events from different hours of the day could be of different species.218

Habitat partitioning among different species is known to occur in these219

waters (Di Tullio et al., 2016b; Moreno et al., 2005), and now nocturnal data220

of cetacean presence can be added to this knowledge. It is noteworthy that221

the Wave Glider covered a broad range of depths while navigating through222

areas possibly occupied by distinct species which can show a preference for223

different depths and slope variations (Ballance et al., 2006; Dalla Rosa et al.,224

2012; De Stephanis et al., 2008). The delphinid encounters occurred across225

all depths, but no detection event covered an area in which local depth varied226

more than 400 m, suggesting that each registered group could have preferred227

depth association.228

Previous studies have shown that oceanographic features can influence229

the cetacean distribution (Ballance et al., 2006; Tynan et al., 2005), and230

that feeding habits and prey distribution are major components of habitat231

partitioning (Canadas et al., 2002; Friedlaender et al., 2006; Hastie et al.,232

11

2004).233

It is known that oceanographic features can influence cetacean distri-234

bution (Ballance et al., 2006; Tynan et al., 2005). Feeding habits and prey235

distribution are major components of habitat partitioning of cetacean species236

in a given area (Canadas et al., 2002; Friedlaender et al., 2006; Hastie et al.,237

2004). The occurrence of upwelling characterizes the region in the vicinities238

of the Cape Frio (23◦S, 42◦W) where relatively cold waters, rich in nutrients,239

surface creating a high biological productivity region (Valentin, 1984; Costa240

and Fernandes, 1993; Matsuura, 1996). The change of coastline orientation241

from north-south to east-west and the proximity of the 100-m isobath are242

favorable topographic conditions for the coastal upwelling next to the Cape243

Frio (Rodrigues and Lorenzzetti, 2001), but the primary process is the wind-244

driven Ekman transport due to persistent E-NE winds that prevail in spring245

and summer in the region (Emılsson, 1961; Ikeda et al., 1974; Castro and Mi-246

randa, 1998). Other mechanisms work to rise cold waters offshore the Cape247

Frio, like the passage of cyclonic eddies and meanders of the Brazil Current248

(Campos et al., 2000; Castelao et al., 2004) and vertical transport driven by249

the wind stress curl (Castelao and Barth, 2006).250

Upwelling regions are highly productive and aggregate diverse marine251

fauna, attracting top marine predators (Croll et al., 2005; Tynan et al., 2005).252

Since the Wave Glider passed through an area closely related to an upwelling253

system in the Southwestern Atlantic Ocean, it is possible that the great254

abundance of delphinid detections is due to prey concentration associated255

with nutrient-rich waters.256

Confirmation of species identification through whistles contour was pos-257

12

sible only for one event. Few studies characterized whistles of the species258

found in the Southwestern Atlantic Ocean and most of these used record-259

ings from coastal areas (de Andrade et al., 2015; Azevedo et al., 2010, 2007;260

Lima et al., 2012). Another difficulty is the possibility that more than one261

species has been recorded in the same detection event. Thus the presence of262

many whistle types with varying characteristics made the exclusion of some263

species of varied whistle repertoire impossible. Mixed-species associations264

among delphinids are common in several areas and present a challenge to265

species classification and identification through acoustic methods (Oswald266

et al., 2003). More bioacoustics studies focusing on acoustic repertoire char-267

acterization are necessary to fill the data gap on oceanic waters from the268

South Atlantic.269

The HF5 was the only event in which species identification was possible.270

The whistles from Rough-toothed dolphins, Steno bredanensis, were com-271

pared to a previous characterization of the species’ whistle repertoire in the272

coastal region of the Rio de Janeiro State Coast, adjacent to the Guanabara273

Bay (Lima et al., 2012). We found contour similarities between whistles274

from the database and the present study. The whistles were in most part275

differentiated from other delphinid species both in frequency and duration276

(Lima et al., 2016). Furthermore, in the Southwestern Atlantic Ocean rough-277

toothed dolphins are known to occur both in coastal waters (Lima et al., 2016,278

2012) and offshore waters (Wedekin et al., 2014), and the record from the279

present study adds information about the area of occurrence of this species.280

Despite species identification having been restricted, the acoustic param-281

eters extracted from whistles presented significant variations and no over-282

13

lapping among some events, which means that the sampled area was used283

by at least seven delphinid species. Previous studies have shown that whis-284

tle repertoires have species-specific characteristics that result in interspecific285

differences (Lima et al., 2016; Oswald et al., 2007; Steiner, 1981). Therefore,286

highest differences among whistle variables of different events in the present287

study may be related to the recordings of different species. The lowest dif-288

ferences registered among whistles of some events may reflect intraspecific289

variation, which can be influenced by the social context and environmental290

factors at the time in which each group was recorded (Azevedo et al., 2010;291

Bonato et al., 2015; Lopez, 2011; May-Collado, 2013). However, it is likely292

that one species has been recorded on more than one event, so this could293

account for some of the overlap observed among events. Also, some degree294

of overlap among whistles of different species is to be expected, as shown in295

previous studies of interspecific comparisons (Oswald et al., 2007, 2003).296

Considering that the PAM system was towed close to the surface, it is pos-297

sible that this has restricted the system’s capacity of registering deep-diving298

animals such as beaked whales and sperm whales (Physeter microcephalus).299

Although the sample rate employed in this study is not adequate to record300

beaked whale signals since their fundamental frequency is usually higher than301

100 kHz (Baumann-Pickering et al., 2014; Zimmer et al., 2008), sperm whale302

clicks, otherwise, fall within the recording system frequency range (Antunes303

et al., 2011; Rendell and Whitehead, 2005), and the species has been sighted304

in the area in previous studies (Di Tullio et al., 2016a; Zerbini, 2004). It305

is possible that either the sound detectors were not adequate to find sperm306

whale clicks, or that ecological factors influenced their absence.307

14

All calls produced by Brydes whales in the present study were of the same308

type and identified as a call first described in the eastern tropical Pacific ocean309

(Oleson et al., 2003), also registered in the Gulf of Mexico (Sirovic et al.,310

2014), indicating that this call type might be shared across different Brydes311

whale populations. During the austral summer, Brydes whales occur close312

to the southeastern coast of Brazil (Zerbini et al., 1997), thus, considering313

the sampling season, no baleen whale except for Brydes whale was expected314

to be recorded. Although it is possible that some unidentified signals were315

produced by minke whales, known for their mechanic sounding vocalizations316

(Gedamke et al., 2001; Rankin and Barlow, 2005), it is also possible that317

these signals are undescribed vocalizations produced by Brydes whales.318

Furthermore, data from PAM coupled with unmanned vehicle technology319

is unprecedented for the Southwestern Atlantic Ocean. This study was con-320

ducted through 20 days, and it produced large volumes of data, generating321

evidence that several species of delphinids use the area. The high number of322

detected sound emissions on some events also hints at the presence of large323

groups of delphinids engaging on high activity rates, indicating an abundant324

and diverse community of dolphins on the outer continental shelf and slope325

of the Southwestern Atlantic Ocean during the summer. Further research326

covering acoustic characterization and behavior of cetacean species, season-327

ality of habitat use and relationship with oceanographic features should be328

conducted to increase knowledge regarding this habitat.329

15

5. Acknowledgements330

This work was supported by PETROBRAS and the Brazilian Oil Reg-331

ulatory Agency (ANP), within the project PT-128.01.12041 - “New Tech-332

nologies and Innovations in Sensors and Metocean Sampling”. The acous-333

tic survey was executed in a partnership of PETROBRAS/CENPES, West-334

ernGeco/LROG, and Brazilian Navy/IPqM. Liquid Robotics Oil and Gas335

(LROG), previously a joint venture between Liquid Robotics and Schlum-336

berger, became wholly owned by Schlumberger on August 29, 2016, and has337

been renamed Schlumberger Robotics Services. We would like to acknowl-338

edge Fabian Van Der Werth and Levent Alkan, former employees of LROG,339

for the system setup and field operations. The authors would like to thank340

IPqM and Faculdade de Oceanografia (UERJ) for data analysis support.341

LB received a scholarship from CAPES. IMSL received a scholarship from342

FAPERJ. AFA and JL-B have research grant from CNPq (PQ-1D), FAPERJ343

(JCNE) and UERJ (Prociencia).344

6. References345

Ricardo Antunes, Tyler Schulz, Shane Gero, Hal Whitehead, Jonathan Gor-346

don, and Luke Rendell. Individually distinctive acoustic features in sperm347

whale codas. Animal Behaviour, 81(4):723–730, 2011.348

Alexandre F Azevedo, Alvaro M Oliveira, L Dalla Rosa, and J Lailson-Brito.349

Characteristics of whistles from resident bottlenose dolphins (Tursiops350

truncatus) in southern Brazil. The Journal of the Acoustical Society of351

America, 121(5):2978–2983, 2007.352

16

Alexandre F Azevedo, L Flach, Tatiana L Bisi, Luciana G Andrade, Paulo R353

Dorneles, and J Lailson-Brito. Whistles emitted by Atlantic spotted dol-354

phins (Stenella frontalis) in southeastern Brazil. The Journal of the Acous-355

tical Society of America, 127(4):2646–2651, 2010.356

Lisa T Ballance, Robert L Pitman, and Paul C Fiedler. Oceanographic in-357

fluences on seabirds and cetaceans of the eastern tropical pacific: a review.358

Progress in Oceanography, 69(2-4):360–390, 2006.359

Simone Baumann-Pickering, Marie A Roch, Robert L Brownell Jr, Anne E360

Simonis, Mark A McDonald, Alba Solsona-Berga, Erin M Oleson, Sean M361

Wiggins, and John A Hildebrand. Spatio-temporal patterns of beaked362

whale echolocation signals in the North Pacific. PLoS One, 9(1):e86072,363

2014.364

Simone Baumann-Pickering, Jennifer S Trickey, Sean M Wiggins, and Erin M365

Oleson. Odontocete occurrence in relation to changes in oceanography at366

a remote equatorial Pacific seamount. Marine Mammal Science, 32(3):367

805–825, 2016.368

Mark F Baumgartner and David M Fratantoni. Diel periodicity in both sei369

whale vocalization rates and the vertical migration of their copepod prey370

observed from ocean gliders. Limnology and Oceanography, 53(5part2):371

2197–2209, 2008.372

Brian Bingham, Nicholas Kraus, Bruce Howe, Lee Freitag, Keenan Ball,373

Peter Koski, and Eric Gallimore. Passive and active acoustics using an374

autonomous wave glider. Journal of field robotics, 29(6):911–923, 2012.375

17

Lis Bittencourt, Mariana Barbosa, Eduardo Secchi, Jose Lailson-Brito, and376

Alexandre Azevedo. Acoustic habitat of an oceanic archipelago in the377

Southwestern Atlantic. Deep Sea Research Part I: Oceanographic Research378

Papers, 115:103–111, 2016.379

M Bonato, E Papale, G Pingitore, S Ricca, A Attoumane, A Ouledi, and380

C Giacoma. Whistle characteristics of the spinner dolphin population381

in the Comoros Archipelago. The Journal of the Acoustical Society of382

America, 138(5):3262–3271, 2015.383

Edmo JD Campos, Denise Velhote, and Ilson CA da Silveira. Shelf break384

upwelling driven by brazil current cyclonic meanders. Geophysical Research385

Letters, 27(6):751–754, 2000.386

A Canadas, R Sagarminaga, and S Garcıa-Tiscar. Cetacean distribution387

related with depth and slope in the mediterranean waters off southern388

spain. Deep Sea Research Part I: Oceanographic Research Papers, 49(11):389

2053–2073, 2002.390

Peter Carragher, Graham Hine, Patrick Legh-Smith, Jeffrey Mayville, Rod391

Nelson, Sudhir Pai, Iain Parnum, Paul Shone, Jonathan Smith, and Chris-392

tian Tichatschke. A new platform for offshore exploration and production.393

Oilfield Review, 2014(25):4, 2013.394

Renato M Castelao and John A Barth. Upwelling around cabo frio, brazil:395

The importance of wind stress curl. Geophysical Research Letters, 33(3),396

2006.397

18

Renato M Castelao, Edmo JD Campos, and Jerry L Miller. A modelling398

study of coastal upwelling driven by wind and meanders of the brazil cur-399

rent. Journal of Coastal Research, pages 662–671, 2004.400

BM de Castro and LB de Miranda. Physical oceanography of the western401

atlantic continental shelf located between 4n and 34s. In A. R. Robinson402

and K. H. Brink, editors, The sea, chapter 8, pages 209–251. Wiley New403

York, Hoboken, N. J., 1998.404

Paulo Alberto S Costa and Flavio da Costa Fernandes. Seasonal and spatial405

changes of cephalopods caught in the cabo frio (brazil) upwelling ecosys-406

tem. Bulletin of Marine Science, 52(2):751–759, 1993.407

Donald A Croll, Baldo Marinovic, Scott Benson, Francisco P Chavez, Nancy408

Black, Richard Ternullo, and Bernie R Tershy. From wind to whales:409

trophic links in a coastal upwelling system. Marine Ecology Progress Series,410

289:117–130, 2005.411

Luciano Dalla Rosa, John KB Ford, and Andrew W Trites. Distribution412

and relative abundance of humpback whales in relation to environmental413

variables in coastal british columbia and adjacent waters. Continental Shelf414

Research, 36:89–104, 2012.415

Luciana Guimaraes de Andrade, Isabela Maria Seabra Lima, Halerson416

da Silva Macedo, Rafael Ramos de Carvalho, Jose Lailson-Brito, Leonardo417

Flach, and Alexandre de Freitas Azevedo. Variation in Guiana dolphin (So-418

talia guianensis) whistles: using a broadband recording system to analyze419

19

acoustic parameters in three areas of southeastern Brazil. acta ethologica,420

18(1):47–57, 2015.421

Renaud De Stephanis, Thomas Cornulier, Philippe Verborgh,422

Juanma Salazar Sierra, Neus Perez Gimeno, and Christophe Guinet.423

Summer spatial distribution of cetaceans in the strait of gibraltar in424

relation to the oceanographic context. Marine Ecology Progress Series,425

353:275–288, 2008.426

Juliana Couto Di Tullio, Tiago BR Gandra, Alexandre N Zerbini, and Ed-427

uardo R Secchi. Diversity and distribution patterns of cetaceans in the428

subtropical Southwestern Atlantic outer continental shelf and slope. PloS429

one, 11(5):e0155841, 2016a.430

Juliana Couto Di Tullio, Tiago BR Gandra, Alexandre N Zerbini, and Ed-431

uardo R Secchi. Diversity and distribution patterns of cetaceans in the432

subtropical southwestern atlantic outer continental shelf and slope. PloS433

one, 11(5):e0155841, 2016b.434

Manuel E dos Santos, Sonia Louro, Miguel Couchinho, and Cristina Brito.435

Whistles of bottlenose dolphins (Tursiops truncatus) in the Sado Estuary,436

Portugal: characteristics, production rates, and long-term contour stabil-437

ity. Aquatic Mammals, 31(4):453, 2005.438

Ingvar Emılsson. The shelf and coastal waters off southern brazil. Boletim439

do Instituto Oceanografico, 11(2):101–112, 1961.440

Ari S Friedlaender, Pat N Halpin, Song S Qian, Gareth L Lawson, Peter H441

Wiebe, Deb Thiele, and Andrew J Read. Whale distribution in relation to442

20

prey abundance and oceanographic processes in shelf waters of the western443

antarctic peninsula. Marine Ecology Progress Series, 317:297–310, 2006.444

Jason Gedamke, Daniel P Costa, and Andy Dunstan. Localization and visual445

verification of a complex minke whale vocalization. The Journal of the446

Acoustical Society of America, 109(6):3038–3047, 2001.447

Gordon D Hastie, BEN Wilson, LJ Wilson, KM Parsons, and Paul Michael448

Thompson. Functional mechanisms underlying cetacean distribution pat-449

terns: hotspots for bottlenose dolphins are linked to foraging. Marine450

Biology, 144(2):397–403, 2004.451

Roger Hine, Scott Willcox, Graham Hine, and Tim Richardson. The Wave452

Glider: A wave-powered autonomous marine vehicle. In OCEANS 2009,453

MTS/IEEE Biloxi-Marine Technology for Our Future: Global and Local454

Challenges, pages 1–6. IEEE, 2009.455

Y Ikeda, LB De Miranda, and NJ Rock. Observations on stages of upwelling456

in the region of cabo frio (brazil) as conducted by continuous surface tem-457

perature and salinity measurements. Boletim do Instituto Oceanografico,458

23:33–46, 1974.459

Holger Klinck, Sharon L Nieukirk, David K Mellinger, Karolin Klinck,460

Haruyoshi Matsumoto, and Robert P Dziak. Seasonal presence of cetaceans461

and ambient noise levels in polar waters of the North Atlantic. The Journal462

of the Acoustical Society of America, 132(3):EL176–EL181, 2012.463

S. Krey, O. Mersmann, S. Schnackenberg, G. Guenard, A. Preusser,464

A. Thieler, J. Mielke, and C. Weihs. Package “tuner”, 2011.465

21

Marc O Lammers, Russell E Brainard, Whitlow WL Au, T Aran Mooney,466

and Kevin B Wong. An ecological acoustic recorder (ear) for long-term467

monitoring of biological and anthropogenic sounds on coral reefs and other468

marine habitats. The Journal of the Acoustical Society of America, 123469

(3):1720–1728, 2008.470

Isabela Maria Seabra Lima, Luciana Guimaraes de Andrade, Rafael471

Ramos de Carvalho, Jose Lailson-Brito, and Alexandre de Freitas Azevedo.472

Characteristics of whistles from rough-toothed dolphins (Steno bredanen-473

sis) in Rio de Janeiro coast, southeastern Brazil. The Journal of the Acous-474

tical Society of America, 131(5):4173–4181, 2012.475

Isabela MS Lima, Luciana G Andrade, Lis Bittencourt, Tatiana L Bisi,476

Leonardo Flach, Jose Lailson-Brito Jr, and Alexandre F Azevedo. Whistle477

comparison of four delphinid species in Southeastern Brazil. The Journal478

of the Acoustical Society of America, 139(5):EL124–EL127, 2016.479

James Vincent Locascio. Passive acoustic studies of estuarine fish populations480

of southwest Florida. University of South Florida, 2010.481

Bruno Dıaz Lopez. Whistle characteristics in free-ranging bottlenose dolphins482

(Tursiops truncatus) in the Mediterranean Sea: influence of behaviour.483

Mammalian Biology-Zeitschrift fur Saugetierkunde, 76(2):180–189, 2011.484

Joseph J Luczkovich, David A Mann, and Rodney A Rountree. Passive acous-485

tics as a tool in fisheries science. Transactions of the American Fisheries486

Society, 137(2):533–541, 2008.487

22

JP Martin, CM Lee, CC Eriksen, C Ladd, and NB Kachel. Glider obser-488

vations of kinematics in a gulf of alaska eddy. Journal of Geophysical489

Research: Oceans, 114(C12), 2009.490

Y Matsuura. A probable cause of recruitment failure of the brazilian sardine491

sardinella aurita population during the 1974/75 spawning season. South492

African Journal of Marine Science, 17(1):29–35, 1996.493

Laura J May-Collado. Guyana dolphins (Sotalia guianensis) from Costa494

Rica emit whistles that vary with surface behaviors. The Journal of the495

Acoustical Society of America, 134(4):EL359–EL365, 2013.496

Sue E Moore, Kathleen M Stafford, David K Mellinger, and John A Hilde-497

brand. Listening for large whales in the offshore waters of Alaska. AIBS498

Bulletin, 56(1):49–55, 2006.499

Ignacio B Moreno, Alexandre N Zerbini, Daniel Danilewicz, Marcos C500

de Oliveira Santos, Paulo C Simoes-Lopes, Jose Lailson-Brito Jr, and501

Alexandre F Azevedo. Distribution and habitat characteristics of dolphins502

of the genus Stenella (Cetacea: Delphinidae) in the southwest Atlantic503

Ocean. Marine Ecology Progress Series, 300:229–240, 2005.504

Erin M Oleson, Jay Barlow, Jonathan Gordon, Shannon Rankin, and John A505

Hildebrand. Low frequency calls of bryde’s whales. 2003.506

Julie N Oswald, Jay Barlow, and Thomas F Norris. Acoustic identification507

of nine delphinid species in the eastern tropical Pacific Ocean. Marine508

mammal science, 19(1):20–037, 2003.509

23

Julie N Oswald, Shannon Rankin, and Jay Barlow. The effect of recording510

and analysis bandwidth on acoustic identification of delphinid species. The511

Journal of the Acoustical Society of America, 116(5):3178–3185, 2004.512

Julie N Oswald, Shannon Rankin, Jay Barlow, and Marc O Lammers. A513

tool for real-time acoustic species identification of delphinid whistles. The514

Journal of the Acoustical Society of America, 122(1):587–595, 2007.515

Bioacoustics Research Program. Raven pro: interactive sound analysis soft-516

ware. version 1.4, 2011.517

Gerry P Quinn and Michael J Keough. Experimental design and data analysis518

for biologists. Cambridge University Press, 2002.519

Shannon Rankin and Jay Barlow. Source of the North Pacific “boing” sound520

attributed to minke whales. The Journal of the Acoustical Society of Amer-521

ica, 118(5):3346–3351, 2005.522

Luke Rendell and Hal Whitehead. Spatial and temporal variation in sperm523

whale coda vocalizations: stable usage and local dialects. Animal Be-524

haviour, 70(1):191–198, 2005.525

Amalis Riera, John K Ford, and N Ross Chapman. Effects of different anal-526

ysis techniques and recording duty cycles on passive acoustic monitoring527

of killer whales. The Journal of the Acoustical Society of America, 134(3):528

2393–2404, 2013.529

Regina R Rodrigues and Joao A Lorenzzetti. A numerical study of the effects530

of bottom topography and coastline geometry on the southeast brazilian531

coastal upwelling. Continental Shelf Research, 21(4):371–394, 2001.532

24

Daniel L Rudnick, Russ E Davis, Charles C Eriksen, David M Fratantoni,533

and Mary Jane Perry. Underwater gliders for ocean research. Marine534

Technology Society Journal, 38(2):73–84, 2004.535

Ana Sirovic, Hannah R Bassett, Sarah C Johnson, Sean M Wiggins, and536

John A Hildebrand. Bryde’s whale calls recorded in the Gulf of Mexico.537

Marine Mammal Science, 30(1):399–409, 2014.538

Renata S Sousa-Lima, Thomas F Norris, Julie N Oswald, and Deborah P539

Fernandes. A review and inventory of fixed autonomous recorders for540

passive acoustic monitoring of marine mammals. Aquatic Mammals, 39541

(1):23, 2013.542

William W Steiner. Species-specific differences in pure tonal whistle vocal-543

izations of five western North Atlantic dolphin species. Behavioral Ecology544

and Sociobiology, 9(4):241–246, 1981.545

Cynthia T Tynan, David G Ainley, John A Barth, Timothy J Cowles,546

Stephen D Pierce, and Larry B Spear. Cetacean distributions relative547

to ocean processes in the northern california current system. Deep Sea548

Research Part Ii: Topical studies in Oceanography, 52(1-2):145–167, 2005.549

Jean L Valentin. Spatial structure of the zooplankton community in the cabo550

frio region (brazil) influenced by coastal upwelling. In Tropical Zooplank-551

ton, pages 183–199. Springer, 1984.552

Sofie M Van Parijs, Chris W Clark, Renata S Sousa-Lima, Susan E Parks,553

Shannon Rankin, Denise Risch, and Ilse C Van Opzeeland. Management554

and research applications of real-time and archival passive acoustic sensors555

25

over varying temporal and spatial scales. Marine Ecology Progress Series,556

395:21–36, 2009.557

Tracy A Villareal and Cara Wilson. A comparison of the pac-x trans-pacific558

wave glider data and satellite data (modis, aquarius, trmm and viirs). PloS559

one, 9(3):e92280, 2014.560

Carrie C Wall, Chad Lembke, and David A Mann. Shelf-scale mapping561

of sound production by fishes in the eastern Gulf of Mexico, using au-562

tonomous glider technology. Marine Ecology Progress Series, 449:55–64,563

2012.564

Carrie C Wall, Peter Simard, Chad Lembke, and David A Mann. Large-scale565

passive acoustic monitoring of fish sound production on the West Florida566

shelf. Marine Ecology Progress Series, 484:173–188, 2013.567

Carrie C Wall, David A Mann, Chad Lembke, Chris Taylor, Ruoying He,568

and Todd Kellison. Mapping the soundscape off the Southeastern USA by569

using passive acoustic glider technology. Marine and Coastal Fisheries, 9570

(1):23–37, 2017.571

Pauline Weatherall, KM Marks, Martin Jakobsson, Thierry Schmitt, Shin572

Tani, Jan Erik Arndt, Marzia Rovere, Dale Chayes, Vicki Ferrini, and573

Rochelle Wigley. A new digital bathymetric model of the world’s oceans.574

Earth and Space Science, 2(8):331–345, 2015.575

LL Wedekin, MR Rossi-Santos, C Baracho, AL Cypriano-Souza, and576

PC Simoes-Lopes. Cetacean records along a coastal-offshore gradient in the577

26

Vitoria-Trindade Chain, western South Atlantic Ocean. Brazilian Journal578

of Biology, 74(1):137–144, 2014.579

Sean M Wiggins, Kaitlin E Frasier, E Elizabeth Henderson, and John A580

Hildebrand. Tracking dolphin whistles using an autonomous acoustic581

recorder array. The Journal of the Acoustical Society of America, 133582

(6):3813–3818, 2013.583

Ben Wilson, Steven Benjamins, and Jim Elliott. Using drifting passive584

echolocation loggers to study harbour porpoises in tidal-stream habitats.585

Endangered Species Research, 22(2):125–143, 2013.586

Alexandre N Zerbini. Distribuicao e abundancia relativa de cetaceos na Zona587

Economica Exclusiva da regiao Sudeste-Sul do Brasil. Number 599.5 (81)588

DIS. 2004.589

Alexandre N Zerbini, Eduardo R Secchi, Salvatore Siciliano, and Paulo Cesar590

Simoes-Lopes. A review of the occurrence and distribution of whales of the591

genus Balaenoptera along the Brazilian coast. Report of the International592

Whaling Commission, 47:407–417, 1997.593

Walter MX Zimmer, John Harwood, Peter L Tyack, Mark P Johnson, and594

Peter T Madsen. Passive acoustic detection of deep-diving beaked whales.595

The Journal of the Acoustical Society of America, 124(5):2823–2832, 2008.596

27

Figure 1: The Wave Glider model SV2 connected to the PAM system through a 15-m tow

cable on deck after recovery.

28

Table 1: Detectors parameters used in Raven 1.5 to look for biological signals from the

offshore waters of the Southwestern Atlantic Ocean.

Detector Parameters Spectogram Frequency Limit Focal Sound Types

Det1 Minimum frequency: 8 kHz 25 kHz Whistles and burst-pulsed sounds

Maximum frequency: 22 kHz

Minimum duration: 60 ms

Maximum duration: 900 ms

Minimum separation: 50 ms

Det2 Minimum frequency: 2 kHz 25 kHz Whistles and burst-pulsed sounds

Maximum frequency: 7 kHz

Minimum duration: 60 ms

Maximum duration: 900 ms

Minimum separation: 50 ms

Det3 Minimum frequency: 15 kHz 25 kHz Echolocation clicks

Maximum frequency: 24 kHz

Minimum duration: 5 ms

Maximum duration: 10 ms

Minimum separation: 10 ms

Det4 Minimum frequency: 50 Hz 3 kHz Tonal and burst-pulsed sounds

Maximum frequency: 300 Hz

Minimum duration: 200 ms

Maximum duration: 600 ms

Minimum separation: 300 ms

Det5 Minimum frequency: 300 Hz 3 kHz Tonal and burst-pulsed sounds

Maximum frequency: 800 Hz

Minimum duration: 100 ms

Maximum duration: 600 ms

Minimum separation: 100 ms

29

Table 2: High-frequency acoustic detection events obtained through passive acoustic mon-

itoring in Brazilian offshore waters in the Southwestern Atlantic Ocean during the austral

summer.

Detection event Duration (h) Time of day interval Detection composition Local depth (m)

HF1 8 15 pm to 23 pm 8815 EC, 62 W, 32 BP 339 ± 78 (243–427)

HF2 7 00 am to 07 pm 37452 EC, 30 W, 54 BP 711 ± 74 (626–841)

HF3 1 14 pm to 15 pm 55 EC 1612

HF4 1 10 am to 11 am 16 EC, 8 W 1643

HF5 2 06 am to 08 am 9349 EC, 18 W, 12 BP 1629 ± 1 (1628–1629)

HF6 1 09 am to 10 am 31 clicks EC 1627

HF7 5 20 pm to 01 am of the following day 4564 EC, 159 W, 23 BP 1629 ± 5 (1621–1633)

HF8 3 04 am to 07 am 2722 EC, 18 W, 6 BP 1504 ± 6 (1497–1508)

HF9 4 21 pm to 01 am of the following day 18714 EC, 448 W, 47 BP 1569 ± 11 (1555–1579)

HF10 2 02 am to 04 am 10651 EC, 20 W, 11 BP 1218 ± 11 (1213–1255)

HF11 6 21 pm to 03 am of the following day 1757 EC, 122 W, 6 BP 697 ± 35 (667–756)

HF12 5 11 am to 15 pm 2995 EC, 59 W, 3 BP (352–468)

HF13 2 03 am to 05 am 1594 EC, 3 W, 2 BP 144 ± 1 (144–145)

HF14 2 19 pm to 21 pm 3 EC, 39 W, 6 BP 174 ± 5 (170–177)

HF15 3 04 am to 07 am 7 W, 4 BP 417 ± 91 (334–515)

HF16 1 10 am 28 EC 805

HF17 2 03 am to 05 am 763 EC 1364 ± 33 (1341–1387)

HF18 2 18 pm to 20 pm 4180 EC, 4 W, 48 BP 1566 ± 6 (1562–1570)

HF19 2 04 am to 06 am 11118 EC, 91 W, 53 BP 1791 ± 21 (1776–1806)

HF20 1 11 am 389 EC 1966

HF21 7 23 pm to 06 am 2575 EC, 26 W, 9 BP 2058 ± 23 (2031–2084)

HF22 2 03 am to 05 am 3 BP 1638 ± 19 (1624–1651)

HF23 2 03 am to 05 am 1188 EC, 2 BP 1177 ± 21 (1162–1191)

HF24 1 06 am 9017 EC, 9 W, 4 BP 1120

HF25 1 09 am 18 W 1026

HF26 11 22 pm to 09 am 18682 EC, 249 W, 52 BP 378 ± 160 (177–573)

HF27 1 11 am 36 EC 145

HF28 3 13 pm to 16 pm 85 EC 131 ± 1 (130–132)

HF29 5 19 pm to 00 am 654 EC, 74 W 379 ± 1 (224–513)

HF30 8 00 am to 08 am 15519 EC, 177 W, 12 BP 515 ± 27 (488–555)

HF31 2 09 am to 11 am 148 W 470

Sum 111 - 162952 EC, 1823 W, 389 BP 935 ± 619 (130–2084)

30

Table 3: Low-frequency acoustic detection events obtained through passive acoustic mon-

itoring in Brazilian offshore waters in the Southwestern Atlantic Ocean during the austral

summer.

Detection event Duration (h) Time of the day interval Detection composition Local depth (m)

LF1 2 08 am to 10 am 139 tonal signals 1041 ± 59 (999–1083)

LF2 2 20 pm to 22 pm 48 tonal signals 1576 ± 2 (1576–1579)

LF3 3 19 pm to 22 pm 122 tonal signals 841 ± 82 (756–920)

LF4 3 00 am to 03 am 393 tonal signals 677 ± 13 (667–692)

LF5 1 14 pm 3 tonal signals 392

Sum 11 - 705 tonal signals 926 ± 372 (392–1579)

Figure 2: Examples of biological signals categories as classified by two observers during

detection validation. (A) echolocations clicks (overlapped with a whistle), (B) burst-pulsed

sound, and (C) tonal signal, referred to as whistles in the case of delphinid.

31

Figure 3: Number of hours with high-frequency and low-frequency detections in all days

of the sampling period.

32

Figure 4: Sum of high-frequency (A) and low-frequency (B) detections per hour of the day

of the total sampling period.

33

Figure 5: Mapping of high-frequency detection events through unmanned vehicle pas-

sive acoustic monitoring on Brazilian offshore waters in the Southwestern Atlantic Ocean

between 6-25 February 2016.

34

Figure 6: Mapping of low-frequency detection events through unmanned vehicle passive

acoustic monitoring on Brazilian offshore waters in the Southwestern Atlantic Ocean be-

tween 6-25 February 2016.

35

Figure 7: Principal components analysis of 13 high-frequency events of delphinid detec-

tions registered during passive acoustic monitoring in Brazilian offshore waters in the

Southwestern Atlantic Ocean.

36

Figure 8: DFA centroids separating high-frequency events into seven groups, indicating

probable seven delphinid species registered during passive acoustic monitoring in Brazilian

offshore waters in the Southwestern Atlantic Ocean.

37

Figure 9: Examples of whistles of the same contour type in two situations: (A) whistle

recorded in the present study; and (B) whistle recorded during visual surveys of rough-

toothed dolphins and found in the MAQUA sound database.

38

Figure 10: Brydes whale calls recorded through unmanned vehicle passive acoustic moni-

toring on Brazilian offshore waters in the Southwestern Atlantic Ocean.

39