Content uploaded by Arial M. Brewer
Author content
All content in this area was uploaded by Arial M. Brewer on Dec 02, 2023
Content may be subject to copyright.
Communication in Cook Inlet beluga whales: Describing the
vocal repertoire and masking of calls by commercial ship noise
Arial M. Brewer,
1,a),b)
Manuel Castellote,
2,a)
Amy M. Van Cise,
1
Tom Gage,
3
and Andrew M. Berdahl
1
1
School of Aquatic and Fishery Sciences, University of Washington, Seattle, Washington 98195, USA
2
Cooperative Institute for Climate, Ocean and Ecosystem Studies, University of Washington, Seattle, Washington 98195, USA
3
Alaska Department of Fish and Game, Anchorage, Alaska 99518, USA
ABSTRACT:
Many species rely on acoustic communication to coordinate activities and communicate to conspecifics. Cataloging
vocal behavior is a first step towards understanding how individuals communicate information and how
communication may be degraded by anthropogenic noise. The Cook Inlet beluga population is endangered with an
estimated 331 individuals. Anthropogenic noise is considered a threat for this population and can negatively impact
communication. To characterize this population’s vocal behavior, vocalizations were measured and classified into
three categories: whistles (n¼1264, 77%), pulsed calls (n¼354, 22%), and combined calls (n¼15, 1%), resulting
in 41 call types. Two quantitative analyses were conducted to compare with the manual classification. A
classification and regression tree and Random Forest had a 95% and 85% agreement with the manual classification,
respectively. The most common call types per category were then used to investigate masking by commercial ship
noise. Results indicate that these call types were partially masked by distant ship noise and completely masked by
close ship noise in the frequency range of 0–12 kHz. Understanding vocal behavior and the effects of masking in
Cook Inlet belugas provides important information supporting the management of this endangered population.
V
C2023 Acoustical Society of America.https://doi.org/10.1121/10.0022516
(Received 2 May 2023; revised 28 October 2023; accepted 6 November 2023; published online 30 November 2023)
[Editor: Rebecca A. Dunlop] Pages: 3487–3505
I. INTRODUCTION
Many species rely on vocal communication as a mecha-
nism for mate selection, to share resource information, avoid
predators, and organize collective movement (Bradbury and
Vehrencamp, 2011;Kershenbaum et al., 2016). Vocal reper-
toire analysis can provide a baseline for studies on conspe-
cific communication and vocal learning in young animals,
and a mechanism to assess population structure. The vocal
repertoire of a wide variety of species has been documented,
including birds (Saunders, 1983;Trejos-Araya and
Barrantes, 2014), primates (Hammerschmidt and Fischer,
2019;Macedonia, 1993), mustelids (Lemasson et al., 2014;
Leuchtenberger et al., 2014;McShane et al., 1995), and
marine mammals (Brady et al., 2020;Ford, 1989;Martin
et al., 2021;Phillips and Stirling, 2001;Sayigh et al., 2013;
Weilgart and Whitehead, 1997). Among marine mammals,
studies have shown that vocal repertoire may be integral to
the maintenance of population structure (Sharpe et al., 2019;
Van Cise et al., 2017;Whitehead et al., 1998;Yurk et al.,
2002) and is an important tool for mate attraction (Tyack,
1981). Communication via acoustic signaling is especially
important in the marine environment, in which visibility is
often limited and animals must rely primarily on sound to
navigate their surroundings and communicate with conspe-
cifics (Dudzinski et al., 2009).
Beluga whales (Delphinapterus leucas) are toothed
whales in the family Monodontidae and have a circumpolar
distribution. There are 21 recognized populations (Kovacs
et al., 2021), including five distinct populations in Alaska
(Hill and DeMaster, 1998). The smallest of these five is the
Cook Inlet beluga (CIB) population, which is non-migratory
(Hobbs et al., 2005;Laidre et al., 2000) and is geographi-
cally and genetically isolated (O’Corry-Crowe et al., 1997).
The CIB population had an estimated abundance of 1300
individuals in the late 1970s (Shelden et al., 2015), but
declined rapidly in the late 20th century. Despite the restric-
tion of subsistence hunting in 1999, the CIB population has
continued to decline (Shelden and Wade, 2019). In 2000,
this population was designated as depleted under the Marine
Mammal Protection Act (U.S. Federal Register, 2000) and
then listed as endangered under the U.S. Endangered
Species Act in 2008 (U.S. Federal Register, 2008). This list-
ing led to the designation of critical habitat in 2011 (Fig. 1),
which is comprised of 7800 km
2
of marine habitat (U.S.
Federal Register, 2011). The most recent analysis of abun-
dance estimates 331 individuals in the population (95% con-
fidence interval: 290–386) (Goetz et al., 2023). This
population remains endangered, despite federal protection,
designation of critical habitat, and the implementation of a
recovery plan, which lists threats ranked from low to high
level of concern. Three threats ranked as high level of
a)
Also at: Marine Mammal Laboratory, Alaska Fisheries Science Center,
National Marine Fisheries Service, National Oceanic and Atmospheric
Administration, Seattle, WA 98115, USA.
b)
Email: arialb@uw.edu
J. Acoust. Soc. Am. 154 (5), November 2023 V
C2023 Acoustical Society of America 3487
ARTICLE
...................................
concern include catastrophic events (e.g., oil spills, mass
strandings), cumulative effects of multiple stressors (e.g.,
co-exposure to chemical pollutants and noise), and anthro-
pogenic noise (National Marine Fisheries Service, 2016).
One of the fundamental knowledge gaps that remain for
the CIB population is information surrounding their social-
ity and communication. Beluga whales are a highly gregari-
ous and vocal species, producing a wide array of
vocalizations, including whistles, pulsed calls, combined
calls, and echolocation clicks (Au et al., 1985;Fish and
Mowbray, 1962). Whistles are narrowband, tonal signals
that can be flat (i.e., little or no frequency modulation) or
frequency modulated and can contain harmonics that are
lower in amplitude. Pulsed calls are bursts of broadband pulses
in which the harmonic interval corresponds to the pulse repeti-
tion rate (PRR) (Watkins, 1966). Combined calls, also known
as biphonic or mixed calls, occur when two concurrent signals
are produced by the same individual, often consisting of a tonal
and pulsed component (Karlsen et al., 2002). Belugas are also
known to exhibit a graded vocalization structure, in which
vocalizations can transition into others on a continuum (Sjare
and Smith, 1986). Vocal repertoire has been documented for
several beluga populations, including Cunningham Inlet,
Canada (Sjare and Smith, 1986); St. Lawrence, Canada
(Faucher, 1988); Bristol Bay, USA (Angiel, 1997); Svalbard,
Norway (Karlsen et al., 2002); White Sea, Russia (Belikov and
Bel’kovich, 2006,2007,2008); Churchill River, Canada
(Chmelnitsky and Ferguson, 2012); and the eastern Beaufort
Sea, USA (Garland et al.,2015). A previous study investigated
spatial and temporal calling behavior in the CIB population
using broad call categories but did not describe vocal repertoire
due to limitations in duty cycle and sampling period (Blevins-
Manhard et al.,2017).
Northern Cook Inlet has the highest concentration of
belugas during ice-free months and also has the largest poten-
tial for negative impacts from anthropogenic noise (Small
et al., 2017). This area is in close proximity to the Port of
Alaska (Anchorage) as well as the Joint Base Elmendorf-
Richardson military base, resulting in persistent levels of
anthropogenic noise. In this region, commercial shipping
noise is the most prominent source of anthropogenic noise, in
both percent of overall time and mean duration (Castellote
et al., 2018). The masking of vocalizations by commercial
shipping noise, and consequently the disruption of communi-
cation, could be one of the main underlying mechanisms of
anthropogenic impact. Anthropogenic noise can negatively
affect marine mammals in a multitude of ways, including
temporary or permanent hearing threshold shifts, changes in
behavior, and auditory masking (Branstetter and Sills, 2022;
DeRuiter et al.,2013;Finneran, 2015;Holt et al., 2011;
Martin et al.,2023;Parsons, 2017;Tyack and Janik, 2013).
Auditory masking is often considered the most prevalent and
occurs when one sound interferes with an individual’s ability
to detect and discriminate another sound (Branstetter and
Sills, 2022;Erbe et al., 2016).
Due to the extreme turbidity of Cook Inlet waters and
the highly vocal nature of these animals, passive acoustic
monitoring (PAM) has proved to be a valuable tool for
understanding the spatio-temporal distribution of this popu-
lation without affecting their behavior. Several long-term
PAM studies have been conducted in Cook Inlet to examine
the year-round seasonal distribution and foraging occurrence
of belugas (Castellote et al., 2016;Castellote et al., 2018;
Castellote et al., 2020;Lammers et al., 2013). While these
studies provide important insights into the spatio-temporal
movement of these whales, little effort has been focused on
investigating the vocal repertoire and how vocalizations
may be masked by anthropogenic noise.
We produce the first description of CIB vocal repertoire
in two critical habitat locations across multiple seasons,
which can be used in future studies of acoustic communica-
tion and group coordination within the CIB population. We
also investigate the degree of masking that commercial ship
noise may have on common CIB vocalizations in the fre-
quency range of 0–12 kHz, which provides strong indication
that ship noise may have a profound impact on vocal com-
munication in this population.
II. METHODS
A. Study area and acoustic recordings
Cook Inlet is an estuary in south-central Alaska that
stretches roughly 370 km from Knik Arm to the Gulf of
Alaska. Two arms, Knik and Turnagain, extend from the
northern reaches of Cook Inlet, surrounding Anchorage, the
most populous city in the state. Cook Inlet is known for its
dramatic tidal cycles, strong currents, and extreme turbidity.
These factors, combined with ice coverage in the winter,
make Cook Inlet an extremely challenging place to conduct
field studies year-round. The Cook Inlet Beluga Acoustics
Program, located at the National Oceanic and Atmospheric
Administration’s Marine Mammal Laboratory, in partner-
ship with the Alaska Department of Fish and Game, has
been deploying passive acoustic recorders and monitoring
the occurrence of cetaceans and anthropogenic noise since
2008 (Castellote et al., 2016;Castellote et al., 2018;
Castellote et al., 2020;Lammers et al., 2013;Small et al.,
2017). For this study, we selected two locations within the
CIB critical habitat where large concentrations of belugas
have been documented, both visually and acoustically, dur-
ing various seasons (Fig. 1). The waters off the Susitna
River Delta (hereafter Susitna), which is the core of the CIB
documented range (Rugh et al., 2010), draws high concen-
trations of belugas during summer months while Trading
Bay is frequented by belugas in winter and spring
(Castellote et al., 2020;Shelden et al., 2015).
The raw data used for this study are from a previous
project investigating beluga presence and seasonality
(Polasek, 2021). We recorded beluga vocalizations and
anthropogenic noise using bottom-mounted DSG-ST acous-
tic recorders (Loggerhead Instruments, Sarasota, FL) with
an HTI-96-min hydrophone (flat frequency response from
2 Hz to 34 kHz and sensitivity of –201 dB re 1 V/lPa). The
acoustic recorders sampled at 24 kHz with a þ33 dB gain.
3488 J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al.
https://doi.org/10.1121/10.0022516
Table Iprovides details regarding location, depth, recording
period, and duty cycle of the acoustic recorders. Recordings
were on a duty cycle to preserve battery life and extend the
duration of the recording period to cover a broader temporal
range. Given the 24 kHz sampling rate, we are only able to
describe vocal repertoire and anthropogenic masking in the
frequency range up to 12 kHz. Belugas are considered mid-
frequency odontocetes (National Marine Fisheries Service,
2018), with hearing capabilities above 100 kHz (Castellote
et al., 2014;Klishin et al., 2000;Mooney et al., 2020).
While some beluga vocalizations contain acoustic energy
above 12 kHz, previous studies have found that the funda-
mental frequencies of whistles, as well as the key acoustic
properties used for classification in pulsed and combined
calls, occur within the sampled frequency range (Belikov
and Bel’kovich, 2006,2008;Chmelnitsky and Ferguson,
2012;Panova et al., 2019).
B. Acoustic analysis
Previously, we documented beluga acoustic encounters
that were used for this study (Polasek, 2021). We define an
acoustic encounter as a grouping of vocalizations during a
given time and designate a new encounter when 60 min or
more elapses with no vocalizations (Lammers et al., 2013).
To capture the vocal repertoire, we analyzed multiple
months across the sampling period and documented call
types for each location. We qualitatively analyzed 84
encounters across 60 days in Susitna and 90 encounters
across 54 days in Trading Bay. From those data, we only
included vocalizations that met the following criteria: (1)
the beginning and end point were clearly distinguishable,
(2) the signal contour could be clearly defined, and (3) the
signal-to-noise ratio (SNR) was greater than 10 dB, which
was measured using Raven Pro 1.6 (Ithaca, NY) (K. Lisa
Yang Center for Conservation Bioacoustics, 2023). We
included vocalizations that met these criteria in further
quantitative analysis and annotated using Raven Pro 1.6
as 10 s–long smoothed spectrograms over the full fre-
quency range of 0–12 kHz, with a 1024 point fast Fourier
transform (FFT), Hanning window, and 75% overlap. We
did not include echolocation clicks in this analysis due to
the frequency sampling limitations of the acoustic
recorders.
TABLE I. Acoustic data used for vocal repertoire analysis.
Location Latitude (N) Longitude (W) Depth (m) Recording period Duty cycle
Susitna 61 10.482 150 30.012 18.3 5/13/2018–9/9/2018 5 min every 15 min
Trading Bay 60 53.134 151 38.610 21.9 9/15/2018–4/26/2019 5 min every 10 min
FIG. 1. (Color online) Map of Cook Inlet, Alaska, indicating locations of two acoustic moorings used in this study, CIB critical habitat, and designated Port
of Alaska commercial shipping lanes.
J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al. 3489
https://doi.org/10.1121/10.0022516
C. Manual classification of call types
We classified vocalizations, hereafter referred to as
calls, based on aural and visual examination of spectrograms
following the protocol and classification scheme originally
developed by Sjare and Smith (1986) for beluga and updated
by Garland et al. (2015). Following these classification
schemes, we divided calls into three categories: whistles,
pulsed calls, and combined calls, which were further broken
down into call types and sub-types. Call types reflect the
contour shape of the fundamental frequency (e.g., ascend-
ing, descending, flat, modulated), while sub-types reflect the
structure of the call (e.g., segmented, sequenced).
We implemented a call classification protocol based on
previous beluga vocal repertoire studies (Chmelnitsky and
Ferguson, 2012;Garland et al., 2015;Sjare and Smith,
1986). Calls documented a minimum of three times are con-
sidered to be stereotyped call types or sub-types, following
precedent from previous analyses of vocal repertoire in
belugas and other species (Chmelnitsky and Ferguson,
2012;Garland et al., 2015;Selbmann et al., 2023;Sharpe
et al., 2019;Van Cise et al., 2017). We assigned call contour
shape by the contour of the fundamental frequency seen in
at least 50% of the call and labeled call types following the
nomenclature in Garland et al. (2015), with abbreviations
based on the category and contour shape. We considered
separate calls to be those that were separated by >0.2 s, fol-
lowing Chmelnitsky and Ferguson (2012). If a call was
within 0.1 s of another call of the same type, we assigned
these as segments following Garland et al. (2015), and the
call was labeled as segmented. For example, if a flat whistle
(flatws) had multiple tonal segments of <0.1 s in time, we
designated this as a flat segmented whistle (flatws.seg) and
we grouped all the segments within the same annotation
box. If two different calls were within the 0.2 s, we anno-
tated them as independent calls. For this study, we added
two new rules to the previously published protocol based on
call types documented in Cook Inlet. First, we defined a call
sequence as multiple units of the same call type repeated in
a series separated by 0.2 s. For example, a descending
whistle (dws) repeated in series with time gaps 0.2 s
between units is designated as a descending whistle
sequence (dws.seq) and all units in this sequence are
grouped within the same annotation box. We followed the
same protocol for pulsed calls and whistles, adding one
adjustment to the classification of pulsed calls to account for
instantaneous steps in PRR, which is the second additional
rule to the protocol. Pulsed calls that are continuous in time
and contain one or more instantaneous steps in PRR are des-
ignated “.bc” for “band change.” For example, a descending
pulsed call (pulse.d) with one or more instantaneous steps in
PRR, is classified as a descending pulsed call with a band
change (pulse.d.bc). For combined calls, we assigned a new
number as new contour shapes were discovered with a “CI”
identifier for Cook Inlet (CI.c.1, CI.c.2, etc.).
We assessed vocal repertoire richness with a rarefaction
curve implemented in the vegan package in R (Oksanen
et al., 2022) for each location separately and combined.
When the curve begins to asymptote, few to no new call
types are being added to the repertoire as new samples are
added, and it can be assumed that the acquired repertoire of
calls is nearly complete. To test for differences in call type
and call category compositions among locations, we imple-
mented a Pearson’s v
2
statistic in R (R Core Team, 2022).
We also conducted a qualitative comparison of call types
following methods from previous beluga repertoire studies
(Belikov and Bel’kovich, 2007,2008;Chmelnitsky and
Ferguson, 2012;Garland et al., 2015;Karlsen et al., 2002)
to identify which CIB call types are shared and unique. We
conducted a visual comparison of call types from our study
with published spectrograms in Cunningham Inlet, Canada
(Sjare and Smith, 1986); St. Lawrence, Canada (Faucher,
1988); Bristol Bay, USA (Angiel, 1997); Svalbard, Norway
(Karlsen et al., 2002); White Sea, Russia (Belikov and
Bel’kovich, 2006,2007,2008); Churchill River, Canada
(Chmelnitsky and Ferguson, 2012); and the eastern Beaufort
Sea, USA (Garland et al., 2015).
D. Quantitative classification of call types
We isolated each beluga call via an annotation box in
Raven Pro 1.6. For whistles, following the previously estab-
lished methodology by Garland et al. (2015), we only
included the fundamental frequency in the annotation box
and therefore, frequency measurements were made only on
the fundamental component of the whistle. Because higher
frequencies attenuate faster than lower frequencies, the pres-
ence of whistle harmonics as well as the high frequency
components of pulsed and combined calls will vary with
source sound level and distance from the recorder. For
pulsed and combined calls, we included the entire broad-
band signal in the annotation box following Garland et al.
(2015), with the caveat that the upper frequency limit is
highly dependent on signal attenuation as well as the upper
limit of our sampling rate. Therefore, our analysis is limited
to the harmonic components and broadband signals of
pulsed and combined calls below 12 kHz. For each anno-
tated call, Raven Pro generated the following acoustic mea-
surements: duration, minimum frequency, maximum
frequency, bandwidth, center frequency, and peak frequency
(Table II). We manually measured start frequency, end fre-
quency, frequency trend, number of inflections, number of
segments, number of steps, number of units, and PRR in
Raven Pro for a subset of 10% of each call type (Table II).
We chose high quality calls (i.e., high SNR, non-overlap-
ping) at random throughout all encounters from both loca-
tions to capture a representative sample. When 10% was
below n¼10, we measured ten calls and if there were fewer
than ten occurrences, we measured all calls within that call
type. For all call types, we calculated the mean and standard
deviation of each measurement.
Following the methodology of Garland et al. (2015),we
conducted both a classification and regression tree (CART)
analysis and a Random Forest analysis on our randomly
3490 J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al.
https://doi.org/10.1121/10.0022516
selected 10% subset of calls to compare with our manual
call classification. A CART analysis is robust to outliers,
non-normal and correlated data, and not only gives a classi-
fication, but also an estimate of the misclassification proba-
bility (Breiman et al., 1984). A Random Forest analysis
creates multiple trees, or a forest, which is used to evaluate
the error rate (out-of-bag error) and the importance of each
predictor (Breiman, 2001). All measurements described in
Table II were used for both analyses. For the CART analy-
sis, we used the R package rpart (Therneau et al., 2022).
We performed a tenfold cross-validation and set terminal
nodes to a minimum sample size of three. Nodes were split
using the Gini Index, which is a measure of node impurity
(Breiman et al., 1984). We then performed upward pruning
until the best predictive tree was obtained. For the Random
Forest analysis, we used the R package randomForest (Liaw
and Wiener, 2002). Since Random Forest models estimate
error internally, we did not need to implement additional
cross-validation (Breiman, 2001). Following Garland et al.
(2015), we set the number of trees to 1000.
E. Anthropogenic noise analysis
Because commercial ship noise has been identified as
the top priority focus for noise mitigation management
actions, we focused our call masking analysis on commer-
cial ship noise. Previous studies have described the acoustic
signature and occurrence of commercial ship noise in Cook
Inlet, which was verified by Port of Alaska ship logs
(Castellote et al., 2018;Polasek, 2021). To obtain a repre-
sentative example of commercial ship noise in this area, we
extracted commercial ship noise from the same acoustic
data used for repertoire analysis in Susitna, as this location
is the core of the designated critical habitat and sees persis-
tent levels of anthropogenic noise due to its proximity to the
Port of Alaska commercial shipping lanes (Fig. 1). We
extracted two 5 min–long sound clips representing the cen-
ter (i.e., highest amplitude) and the edge (i.e., marginal
amplitude above background levels) of the acoustic foot-
print of a commercial ship passing through the commercial
shipping lanes to capture the range of received ship noise
levels at this location, which is 2000 m away from the com-
mercial shipping lanes. We defined the center of the com-
mercial ship footprint as the 5 min portion with the Doppler
effect pattern seen in the spectrogram and the edge as the
last 5 min portion of the ship encounter. To compare the
center and edge frequency spectral content of the ship noise
with the spectral content of beluga calls received at the
same location, we selected a minimum of three clips repre-
senting the spectral variability of each common call type per
category. We selected clips of high quality calls based on
the mean center frequency values of each call type. High
quality calls were those that had no overlapping calls and
had a SNR greater than 10 dB, which was measured using
Raven Pro 1.6. We calculated the received sound pressure
level (SPL) in one-third octave bands for the representative
calls for each common call type and the commercial ship
center and edge footprint using the MATLAB Acoustic
Ecology Toolbox package (Bioacoustics Research Program,
Ithaca, NY) (Dugan et al., 2011). We then plotted the spec-
trum levels from each common call type against the center
and edge ship noise spectra to determine the level of over-
lap. We included the composite beluga audiogram from
Erbe et al. (2016), which is comprised of the lowest hearing
thresholds from multiple individuals across different popula-
tions, to show which portions of the call and ship spectra are
within beluga hearing thresholds. It has previously been
shown that narrower octave bands (e.g., 1/12) may be a bet-
ter estimate for the noise-masking potential in belugas
(Erbe, 2008); however, we use the 1/3 octave band to facili-
tate direct comparison between our results and the beluga
composite audiogram, which was computed in one-third
octave bands per the American National Standards
Institute’s standards for measuring odontocete audiograms
(ANSI, 2018;Erbe et al., 2016).
III. RESULTS
A. Manual classification of call types
We extracted and classified a total of 1633 calls based
on the classification system implemented in previous beluga
repertoire studies (Chmelnitsky and Ferguson, 2012;
Garland et al., 2015;Sjare and Smith, 1986). In Susitna, we
classified 944 calls across five encounters spanning May
through September 2018. In Trading Bay, we classified 689
calls across ten encounters spanning January through April
2019. Of the 1633 total calls classified, 1626 were repeated
three or more times and therefore considered to be stereo-
typed call types, resulting in 41 call types. The rarefaction
curve reflects a near complete repertoire as very few new
call types are added with additional data sampled (Fig. 2).
TABLE II. Description of measurements used in the quantitative classifica-
tion of call types.
Measurement Abbreviation Description
Duration (s) Dur. Length of call
Minimum frequency (Hz) Min. Minimum frequency
Maximum frequency (Hz) Max. Maximum frequency
Bandwidth (Hz) BW Maximum–Minimum frequency
Start frequency (Hz) Start Start frequency of fundamental
End frequency (Hz) End End frequency of fundamental
Frequency trend (ratio) Trend Ratio of start/end frequency
Center frequency (Hz) Center Frequency that divides call into
two intervals of equal energy
Peak frequency (Hz) Peak Frequency at the spectral peak
Inflections (#) Inflect. Number of slope reversals
Segments (#) Seg. Number of segments (temporal
gap between segments)
Steps (#) Steps Number of frequency steps (no
temporal gap between steps)
Units (#) Units Number of units within a
sequence
Pulse repetition rate (/s) PRR Number of pulses per second (for
pulsed and combined calls)
J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al. 3491
https://doi.org/10.1121/10.0022516
Of the 41 stereotyped call types we identified in Cook
Inlet, 25 were whistles (n¼1264, 77.7%) and 15 were
pulsed calls (n¼354, 21.8%). For combined calls, we only
identified one call type that met the repetition requirement
of the protocol (CI.c.5, n¼8, 0.5%), with an additional six
combined calls that were documented fewer than three times
each. Both call type and call category composition differed
significantly among the two locations (Pearson’s v
2
pvalue
¼0.0005 for both). Since data from Susitna captured calling
behavior in summer and Trading Bay captured calling
behavior in winter and spring, we cannot conclude whether
the compositional differences were due to location or sea-
son. Figure 3shows call types in order of prevalence and by
category for Cook Inlet total (Susitna and Trading Bay com-
bined) and call category composition broken down by loca-
tion. Whistles were the predominate call category in Cook
Inlet, followed by pulsed calls, then combined calls.
In this study, we expanded on the call type comparison
table from Garland et al. (2015) to include call types docu-
mented in Cook Inlet and their relation to other beluga popula-
tions to provide a visualization of similarities and differences
in repertoire content (Table III). Of the 41 call types we docu-
mented in the CIB population, 18 were not documented in any
other population, 7 call types were documented across all pop-
ulations, and 16 call types were documented in some, but not
all, populations based on published spectrograms of calls.
1. Whistles
Whistles are the most common call category in the CIB
vocal repertoire, consisting of 77.7% of calls. We identified
eight whistle contour categories (ascending, descending,
flat, modulated, n-shape, r-shape, u-shape, trill), which we
further broke down into 25 unique call types based on con-
tour and structure (segmented, sequenced, terminal tail)
(Fig. 4). The three most common whistle types were
descending whistle (dws, n¼359, 22.1%), flat whistle
(flatws, n¼326, 20%), and modulated whistle (modws,
n¼157, 9.7%). We documented several whistle types with
terminal tails, which we denoted with “.t,” and resulted in
the addition of new call types dws.seq.t, modws.t, and
modws.seg.t to the repertoire. We also documented side-
ways S-shaped modulated whistles, which we denoted with
FIG. 2. (Color online) Rarefaction curve depicting CIB vocal repertoire
richness at each location sampled and in total. These curves are created by
creating an ensemble of curves by randomly re-ordering the calls and then
plotting the average of the ensemble.
FIG. 3. (Color online) (A) CIB call types shown in order of prevalence by call category, (B) Susitna location call category composition, (C) Trading Bay
call category composition.
3492 J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al.
https://doi.org/10.1121/10.0022516
a “.sa” label and resulted in the addition of new call types
modws.sa and modws.sa.seq. For whistles, we only included
the fundamental frequency when measuring acoustic proper-
ties because not all of the whistles in our study displayed
harmonics. Acoustic measurement results for whistles are
presented in Table IV.
2. Pulsed calls
Pulsed calls comprise 21.8% of the CIB vocal reper-
toire. We identified five pulsed call contour categories
(ascending, descending, flat, modulated, n-shape). We
further broke down contours into 15 unique call types
based on contour and structure (segmented, change in
PRR) (Fig. 5). The three most common pulsed call types
were flat pulsed call (pulse.flat, n¼114, 7%), flat seg-
mented pulsed call (pulse.flat.seg, n¼65, 4%), and
descending pulsed call (pulse.d, n¼42, 2.6%). We
included the entire broadband signal, which for some high
SNR cases was truncated by our upper frequency limit of
12 kHz. Acoustic measurement results for pulsed call
types are presented in Table V.
TABLE III. A visualization of CIB call types that have been documented in other beluga populations, based on published spectrograms (Angiel, 1997;
Belikov and Bel’kovich, 2006,2007,2008;Chmelnitsky and Ferguson, 2012;Faucher, 1988;Garland et al., 2015;Karlsen et al., 2002;Sjare and Smith,
1986) (See the supplementary material for the full table that includes call type nomenclature from each study.
1
)
Cook Inlet, US
Current study
Eastern
Beaufort Sea, US
Garland et al. (2015)
Churchill River,
Canada
Chmelnitsky and
Ferguson (2012)
White Sea, Russia
Belikov and
Bel’kovich
(2006,2007,2008)
Svalbard, Norway
Karlsen et al.
(2002)
Bristol Bay,
US
Angiel (1997)
St. Lawrence,
Canada
Faucher (1988)
Cunningham
Inlet, Canada
Sjare and
Smith (l986)
aws
aws.seg —— ——
aws.seq — — — — — — —
dws
dws.seg —— ——
dws.seq — — ————
dws.seq.t — — — — — — —
flatws
flatws. seg --
modws
modws.seg ——
modws.m — — — — — — —
modws.m.seg — — — — — — —
modws.sa — — — — — — —
modws.sa.seq — — — — — — —
modws.t — — — — — — —
modws.seg.t — — — ———
nws
nws.seg —— ——
nws.seq — ————
rws ————
uws
uws.seq — — ————
trill —— —
dottrill — — — — — — —
pulse.a ——
pulse.d ———
pulse.d.seg — — — — — — —
pulse.d.bc — — — — — — —
pulse.flat
pulse.flat.seg — — ———
pulse.flat.seg. 1 — — — — — — —
pulse.flat.seg.2 — — — — — — —
pulse.flat.seg.3 — — — — — — —
pulse.flat.seg.4 — — — — — — —
pulse.flat.bc — — — — — — —
pulse.mod — ———
pulse.mod.seg — — — — — — —
pulse.mod.bc — — — — — — —
pulse.n — ————
CI.c.5 — — — — — — —
J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al. 3493
https://doi.org/10.1121/10.0022516
3. Combined calls
Combined calls were the least common call category
documented in Cook Inlet, comprising 0.5% of the reper-
toire. We identified one stereotypic combined call type,
CI.c.5 (n¼8, 0.5%), and documented an additional six com-
bined calls that were encountered fewer than three times
each (Fig. 6). The six combined calls documented that did
not meet our protocol requirements were CI.c.1 (n¼1),
CI.c.2 (n¼1), CI.c.3 (n¼1), CI.c.4 (n¼1), CI.c.6 (n¼1),
and CI.c.7 (n¼2). Since we documented CI.c.5 eight times,
we can be confident that the call was not two independent,
overlapping calls. The remainder of the combined calls we
documented did not meet the repetition requirement of the
protocol but are included here for descriptive purposes and
had high SNR clearly showing the characteristics of a com-
bined call, but we cannot assure the stereotypic nature of
these calls in this study. For combined calls, we included the
entire broadband signal, which for some was truncated by
our upper frequency limit of 12 kHz. Acoustic measurement
results for CI.c5 are presented in Table VI.
B. Quantitative classification of call types
Following the methodology from Garland et al. (2015),
we conducted both a CART and Random Forest analysis to
compare with our manual classification. For both analyses,
we used a subset of calls per call type, which resulted in 369
total calls. We did not include the single stereotyped com-
bined call type, CI.c.5, in our analyses, as this was the only
call type in this category. The remaining 40 call types (25
whistle types, 15 pulsed call types) were included in both
analyses using all measurements described in Table II.
The most informative acoustic variables to CART con-
struction were maximum frequency and bandwidth. These
were followed by, in descending order: center frequency,
minimum frequency, start frequency, number of segments,
end frequency, frequency trend, maximum PRR, duration,
peak frequency, minimum PRR, number of units, number of
inflections, and number of steps. These variables provided
the analysis with 90% classification of call types (root node
error) with a misclassification rate of 4.9%. Forty-three ter-
minal nodes were created, which is three more than the 40
call types defined using the manual call classification system
(see supplementary material for CART figure).
1
Those three
additional nodes were sub-divisions within three of the man-
ually classified call types (aws, dws.seq, pulse.flat). These
were based on minimum frequency for aws (1600 Hz cut-
off), center frequency for dws.seq (891 Hz cutoff), and fre-
quency trend for pulse.flat (1.3 cutoff). The tree was heavily
influenced by contour shape, with different branches repre-
senting contour categories (ascending, descending, flat,
modulated). The first branching separated flat whistles
FIG. 4. (Color online) CIB whistle types organized by contour. Spectrograms are 1024 point FFT, Hanning window, and 75% overlap, generated in Raven
Pro 1.6.
3494 J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al.
https://doi.org/10.1121/10.0022516
TABLE IV. Descriptive statistics of CIB whistle types [mean 6standard deviation (SD)].
Call type N % total Dur. (s) Min. (Hz) Max. (Hz) BW (Hz) Start (Hz) End (Hz) Center (Hz) Peak (Hz) Inflect. Seg. Units
aws 40 2.5 0.4 60.2 3010. 4 62378.7 4639.8 63588.1 1629. 4 61702.4 2828.9 62197.3 4210.7 63332.1 3637.5 62743.2 3642.2 62791.8 0 60160160
aws.seg 8 0.5 0.7 60.3 4649 61532.8 7058 61634.8 2409 61191.4 4649 61532.8 7038.3 61623.3 5768.6 61717 6023. 4 61982.2 0 60 2.6 61.1 1 60
aws.seq 9 0.6 1 60.4 5356.7 61464.3 9409.9 62535.9 4053.2 61307.1 5901 61898.1 9155.1 62494.3 6830.7 61787.8 6489.6 61684.1 0 60160 4.2 61.6
dws 359 22.1 0.5 60.3 1091.6 61008.5 1948.5 61327.9 857 6672.9 2186 61214.3 991.8 6999 1439.5 61083.8 1426 61080.4 0 60160160
dws.seg 28 1.7 0.9 60.3 1684 61710.7 3274.3 61992.4 1590.3 61015.8 3348.1 61375.2 1446.2 61376.3 2162.9 61961.9 2164.6 61955.1 0 60360.9 1 60
dws.seq 79 4.9 0.9 60.6 2356 62606.6 3376.4 62731.3 1020.5 6441.4 1556 6388.3 742.8 6290.4 2799.4 62806.9 2790 62868.7 0 60160 8.9 63.3
dws.seq.t 4 0.2 1.2 60.5 600.3 6139.3 1749.9 6475.7 1149.6 6359.7 1638.3 6565.7 642.4 6200 779.3 6105.5 808.6 6102.2 0 60160 6.2 64.3
flatws 326 20.0 0.6 60.3 1461.3 61415 1672.5 61416.6 211.2 665 2092.2 62165.3 2007.6 62187.4 1576.1 61415.1 1576.9 61416.5 0 60160160
flatws.seg 27 1.7 1.1 60.5 1875.6 61670.3 2132.5 61766.8 256.9 6125.7 1247.3 6409.9 1228.9 6418.1 2013.9 61736.5 2012.2 61732.5 0 60562160
modws 157 9.7 1 60.5 2088.1 61803.9 3292.3 62179.5 1204.2 6861.7 2403.5 61427.1 2090.5 61549.6 2589 61955.7 2569.2 62003.3 8.1 63.8 1 60160
modws.seg 76 4.7 1.2 60.5 2319.7 61733.5 4110.3 62679.9 1790.5 61192.6 2041.1 62844.4 1519.5 61745.2 2950.3 62119 2879.1 62067.7 7.3 63.9 6 63.8 1 60
modws.m 4 0.2 0.9 60.3 3386.6 61627.2 5066 62167.6 1679.4 6668.3 3604.7 61932.1 3606.8 61708.3 4218.8 62102.3 3908.2 61687.5 3 60160160
modws.m.seg 5 0.3 1.1 60.1 646.1 667.3 1413.6 679.3 767.4 675.4 824 6104.9 646.1 667.3 928.1 673.4 909.4 6137.1 3 60360160
modws.sa 3 0.2 0.2 60 4891.7 6390 8294.2 6234.3 3402.4 6160.7 8294.2 6234.3 4891.7 6390 6210.9 693.8 5914.1 6500.7 2 60160160
modws.sa.seq 7 0.4 1.5 60.6 5424.8 6303.1 8249.5 6413.4 2824.7 6665.9 7874.4 6602.4 5745.2 6327.5 6699.8 6167.1 6468.8 6325.3 8.6 63160 4.3 61.7
modws.t 5 0.3 0.9 60.2 766.5 6635.2 2734.5 6852.1 1968 61075.7 2497.5 6886.1 818.9 6632 1246.9 6891 1190.6 6913.4 9.4 64.4 1 60160
modws.seg.t 5 0.3 1.1 60.2 597.6 6188.3 1772.5 6116 1175 6199.4 905.1 6187.1 1146.3 6595.6 1021.9 6309.9 993.7 6325 6.8 64.4 5.8 62160
nws 21 1.3 0.4 60.2 1488.5 62010 2378.2 62004.6 889.7 6310.5 756.4 6291 728.4 6341.9 2021.2 62064.8 2021.2 62108.2 1 60160160
nws.seg 3 0.2 0. 5 60.1 2084.3 62952.6 3356.1 62789.6 1271.8 6226.3 2306.5 62760.4 2084.3 62952.6 2820.3 63127 3000 62996.7 1 60360160
nws.seq 5 0.3 0.8 60.6 2967.4 61904.8 3462.3 61832.9 494.9 6209 3078.6 61825.6 3032.7 61816.3 3150 61968.1 3173.4 62015.5 5.2 66.1 1 60 5.6 65.9
rws 6 0.4 0.3 60.1 1513.7 674.7 2019.2 676.6 505.5 694.8 1513.8 674.7 1941.8 656.9 1906.3 670.6 1878.9 6127 0 60160160
uws 26 1.6 0.4 60.3 2787.8 62102.1 4179.8 63146.1 1392.1 61325.4 3537 62886 3531.8 62889.8 3229.9 62404.1 3168.6 62362.1 1 60160160
uws.seq 10 0.6 1.3 60.5 3382.6 6935.1 5092.3 61505.1 1709.7 6906.7 4942.3 61538.6 4600.2 61361.8 3808.6 61054.7 3728.9 61041 7.5 63.9 1 60 7.5 63.9
trill 48 3.0 1.2 60.8 3629.4 6774.6 4129.9 6724 500.4 6200.2 3276.3 6815.7 3321 6763.9 3928.2 6762.7 3917.5 6766.2 11.8 65.1 1 60 13.1 65.7
dottrill 3 0.2 0.9 60.4 4473.4 61348.7 5034.1 61294.5 560.7 6179.2 4625.1 61504.8 4787 61062.9 4765.6 61232.7 4726.6 61115.5 0 601602169.6
J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al. 3495
https://doi.org/10.1121/10.0022516
(flatws, flatws.seg) from other calls using bandwidth, then
number of inflections was used to separate the remaining
calls with no inflections (e.g., dws, aws, pulse.a, pulse.d,
pulse.flat) from those with inflections (e.g., modws, nws,
uws, pulse.mod, pulse.n).Within these branches, whistles
and pulsed calls were separated using the minimum and
maximum PRR measurements.
The Random Forest model also yielded similar results to
the manual classification, resulting in an 85% agreement (out-
of-bag error rate ¼15.18%). The most important acoustic vari-
ables for the model in terms of accuracy were number of
inflections, number of segments, bandwidth, and frequency
trend (Fig. 7). In general, call types with small sample sizes
(dws.seq.t, modws.m, modws.seg.t, modws.t, nws.seq, pulse.-
flat.bc, pulse.flat.seg.4, pulse.mod.bc) had higher misclassifica-
tion rates (0.4–1.0) than those with larger sample sizes, such as
dws, flatws, modws, and pulse.flat, which varied from 0.0 to
0.3 (see supplementary material for the Random Forest confu-
sion matrix and call type error rates).
1
C. Anthropogenic noise analysis
We assessed the effect of close and distant commercial
ship noise on the three most common call types in the whis-
tle and pulsed call categories and the single stereotyped
combined call type. Figure 8shows a long-term spectral
average (LTSA) (10 s and 5 Hz averaging) of 24 h (8 h of
duty-cycled data), and a zoomed in spectrogram of 3 h (1 h
of duty-cycled data) of the acoustic footprint of the commer-
cial ship encounter we used for masking analysis at the
Susitna location.
We plotted the spectra of the three most common whis-
tle types (dws, flatws, modws), pulsed call types (pulse.d,
pulse.flat, pulse.flat.seg), and the single most common
stereotyped combined call type (CI.c.5) with close and dis-
tant ship noise and in relation to the composite beluga
audiogram (Fig. 9). The call type spectra (shown in blue)
are completely masked in all seven call types under close
ship noise conditions (shown in red). Under distant ship
noise conditions (shown in yellow), all call types are par-
tially masked. In the distant ship noise condition, partial
masking in whistles occurred from roughly 500–5000 Hz in
descending whistles (dws), 500–6000 Hz in flat whistles
(flatws) with a break from masking around the spectral
peaks at 800 and 1600 Hz, and 500–800 Hz in modulated
whistles (modws). For pulsed calls, partial masking occurred
from roughly 500–2500 Hz with a break from masking
around the spectral peaks at 800 and 2000 Hz in descending
pulsed calls (pulse.d), 500–800 Hz in flat pulsed calls (pul-
se.flat), and 500–2800 Hz with a slight break in masking at
the spectral peak of 1000 Hz in flat segmented pulsed calls
(pulse.flat.seg). For the combined call, CI.c.5, partial mask-
ing occurred from roughly 500–2800 Hz with a slight break
in masking around 1500 and 2000 Hz. An important caveat
to note is that we were not able to describe masking above
12 kHz due to the frequency sampling limitations of the
acoustic recorders used in this study.
IV. DISCUSSION
Our study provides the first description of the vocal rep-
ertoire of the endangered CIB population. We further
endeavor to provide a broader context for geographic vari-
ability in beluga vocal repertoire by conducting a qualitative
comparison to other published descriptions of beluga vocal
repertoires worldwide. Finally, we describe the levels of
masking of the most commonly used call types from
FIG. 5. (Color online) CIB pulsed call types organized by contour. Spectrograms are 1024 point FFT, Hanning window, and 75% overlap, generated in
Raven Pro 1.6.
3496 J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al.
https://doi.org/10.1121/10.0022516
TABLE V. Descriptive statistics of CIB pulsed call types (mean 6SD).
Call type N % total Dur.(s) Min. (Hz) Max. (Hz) BW (Hz) Start (Hz) End (Hz) Center (Hz) Peak (Hz) Inflect. Seg. Steps Min. PRR Max. PRR
pulse.a 21 1.3 0.9 60.4 1068.6 6743.6 10279.1 62749.3 9210.5 62583.1 1006.1 61009.5 1567.2 61338 2405.1 61643.4 1857.1 61401.7 0 60160060 470.9 6243.9 631.9 6261.3
pulse.d 42 2.6 0.8 60.3 917.2 6417.8 8608.6 62906.2 7691.3 62807.2 1889.6 61230 954.5 6564.2 1774.6 61084.8 1607.1 61086.4 0 60160060 931.1 6364.6 1235.5 6382.2
pulse.d.bc 12 0.7 1.4 60.6 803.7 6302.6 9241.3 61899.5 8437.6 61996.8 2258.2 61055.9 1017 6397.8 1533.2 6368.7 1525.4 6385.6 0 60160 1.3 60.5 723.1 6116.8 1672.2 6296.9
pulse.d.seg 16 1.0 1.4 60.4 740.46629.3 9307.7 62143.7 8567.3 61967.9 1138.3 6697.8 667.2 6470.4 2572.3 61558.5 1699.2 61408 0 60 3.3 60.8 0 60 295.8 6169.4 486.6 6181.2
pulse.flat 114 7.0 0.9 60.4 1387.9 61052.6 7747.5 63271.1 6359.6 63106.7 1286.7 6922.5 1216.3 6963.9 2571.1 61586.9 2380.1 61607.3 0 60160060 848.1 6437.4 897.5 6439.8
pulse.flat.
bc
3 0.2 2 61 1143.2 6341.9 9829.5 6829.6 8686.3 6700.2 2366.7 6993.1 1734.4 6338.3 1421.9 6352.6 1406.2 6353.9 0 60160 1.7 60.6 800.7 666.1 1329.1 6321.1
pulse.flat.
seg
65 4.0 1.8 61.4 1285.1 6953.2 10403.1 61943.3 9118 62127.1 1835.8 61474.5 1870.4 61497.1 3298.6 61413.8 3160.1 61864.9 0 60 2.9 61.9 0 60 237.5 6143.4 403.9 681.1
pulse.flat.
seg.1
6 0.4 0.9 60.2 793.7 6368.3 4831.1 61708.5 4037.4 61504.2 1852 6115.4 1630 662.6 1574.2 6428.2 1566.4 6447 0 60260060 884 646.7 1803.5 632.0
pulse.flat.
seg.2
11 0.7 1 60.2 663.3 673.9 3511.5 61284.8 2848.2 61347.8 922.3 6100.6 1582.2 669.4 1031.2 6198.3 1007.8 6228.4 0 60360060 982.9 659.5 1516.6 6168.3
pulse.flat.
seg.3
16 1.0 1 60.3 722.6 666.8 6181.7 62493.5 5459.1 62465.6 1519.8 6294.7 2767.1 64182.8 1706.5 6350.3 1646.5 6350.3 0 60260060 921.4 691.4 1620.1 688.7
pulse.flat.
seg.4
6 0.4 1 60.3 842.6 6240.6 5109.1 61724.1 4266.5 61698 964.8 6231.8 1547 6154.8 1406.2 6332.8 1390.6 6426.7 0 60260060 957.8 666.8 1578.6 655.9
pulse.mod 16 1.0 1.2 60.6 1482.7 6901.8 9330.3 62020.8 7847.5 61582.5 2606.7 61197.3 2469.9 61129.5 3153.8 61571.7 2893.1 61648.5 6.4 64.9 1 60060 459.6 6240.5 614.5 6344.3
pulse.mod.
bc
4 0.3 2.4 60.4 985.9 6197.1 9522.7 62058.5 8536.8 62047.7 2244.8 61297 1719 6663.8 2343.7 61135.2 2419.9 61138.8 6.2 63.1 1 60 1.2 60.5 578.9 6245.0 1310.2 6402.1
pulse.mod.
seg
10 0.6 1 60.2 1335 6300.8 9833 61337.2 8498 61394 3253.6 61556.6 3273.2 61605.4 4591.4 61128.2 4410.9 61744.5 7.2 64360.9 0 60 366.7 6245.3 737.8 6236.7
pulse.n 12 0.7 0.5 60.2 1807.8 61401.6 8960.4 62940.1 7152.6 62764.3 1489.5 61201.3 1432.2 61177.2 2933.6 62067.7 2996.1 62112.3 1 60160060 838.4 6382.2 1141.0 6409.2
J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al. 3497
https://doi.org/10.1121/10.0022516
commercial ship noise at the Susitna location, which is in
the core of the CIB critical habitat.
We documented CIB calls during multiple seasons
across two core locations of the critical habitat, Susitna
and Trading Bay, and manually classified calls into 41 call
types. We compared our manual call classification system
with two quantitative methods that have proved to be suc-
cessful when classifying beluga vocal repertoire (Garland
et al., 2015). Our CART and Random Forest analyses had
a 95% and 85% agreement with the manual call classifica-
tion, respectively. In the CART analysis, 43 terminal
nodes were created, which was 3 more than our manual
call classification (see supplementary material for CART
figure).
1
Those additional nodes were sub-divisions within
three of the manually classified call types (aws, dws.seq,
pulse.flat). Aws and dws.seq were split by two frequency
measurements, minimum frequency, and center frequency,
respectively. These differences in minimum and center
frequency indicated that the manual classification of these
two call types could have been split into additional sub-
types based on frequency; however, in our manual classifi-
cation, we chose to keep them classified into a single aws
and dws.seq call type since the contour remained the
same. Pulse.flat was sub-divided based on differences in
frequency trend. For the CART analysis, 30% of the
pulse.flat call contours showed a slight decrease in fre-
quency at the terminal portion; the remaining 70% were
flat throughout the contour. The call contour was used as
the main criterion for our manual classification, and call
classification was driven by the contour seen in at least
50% of the call. Since a slight decrease in frequency
occurred at the very end of the call contour and more than
50% of the call was flat, we designated these call types as
pulse.flat.
Our results indicate significant differences in both call
type and call category composition between Susitna and
Trading Bay. We could not test the effect of season due to
the sampling dates of the acoustic recorders used; however,
a previous study investigated spatial and temporal calling
behavior of the CIB population and also found a significant
difference in call category composition both spatially and
temporally (Blevins-Manhard et al., 2017). Spatio-temporal
differences in call use may be linked to preferred seasonal
habitats and behavior (e.g., feeding, molting, calving).
Previous studies have examined which environmental varia-
bles may contribute to CIB summer habitat preference and
found that there is a greater probability of belugas being pre-
sent closer to rivers with medium flow accumulation,
Chinook salmon runs, tidal mudflats, and areas with sandy
coastlines (Goetz et al., 2007;Goetz et al., 2012). A long-
term photo identification study documented habitat use and
distribution of the CIB population and found that encounters
had predictable seasonal patterns and distinct hot spots that
were stable across years (McGuire et al., 2020). Spatio-
temporal habitat preference, as well as behavior, may be
driving the call type and call category compositional differ-
ences found in our study.
A. Beluga communication
Belugas are known to be one of the most vocal cetacean
species, producing a wide array of calls. Possibly the most
unique of those are combined calls, where two signals are
produced simultaneously by the same individual. Along
with belugas, this phenomenon has also been described in
other odontocetes, such as killer whales (Orcinus orca),
false killer whales (Pseudorca crassidens), short-finned pilot
whales (Globicephala macrohynchus), and bottlenose dol-
phins (Tursiops truncatus)(Caldwell and Caldwell, 1967;
Filatova et al., 2009;Ford, 1989;Karlsen et al., 2002;Lilly
and Miller, 1961;Murray et al., 1998;Papale et al., 2015;
Ridgway et al., 2015;Sayigh et al., 2013;Van Cise et al.,
2017;Vergara and Barrett-Lennard, 2008). Belugas also
exhibit a graded call structure, in which call types can transi-
tion into others on a continuum (Karlsen et al., 2002;Sjare
and Smith, 1986). Graded call systems have also been docu-
mented in killer whales, false killer whales, and pilot
whales, which all exhibit complex social structures (Ford,
1989;Murray et al., 1998;Sayigh et al., 2013). Belugas are
known to be very gregarious and have highly complex sys-
tems of social organization and interaction (O’Corry-Crowe
et al., 2020). Our study has found that the CIB population,
like other beluga populations, exhibit a rich and complex
vocal repertoire. These findings align with the social com-
plexity hypothesis for communicative complexity, described
in Freeberg et al. (2012), which posits that species that
FIG. 6. (Color online) CIB combined calls (both stereotyped and non-stereotyped). Asterisks denote calls that did not meet the protocol repetition require-
ment but are shown for reference for future studies. Spectrograms were 1024 point FFT, Hanning window, and 75% overlap, generated in Raven Pro 1.6.
Note CI.c.2 to CI.c.5 show aliasing artifacts in the upper frequencies due to acoustic energy above the Nyquist frequency of 12 kHz.
3498 J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al.
https://doi.org/10.1121/10.0022516
FIG. 7. Mean decrease in out-of-bag accuracy caused by excluding individ-
ual acoustic variables from the Random Forest model. Acoustic variables
are listed in order of descending importance.
FIG. 8. (Color online) Commercial ship noise recorded at Susitna. (A)
LTSA of 24 h (8 h of duty-cycled data) with beluga acoustic encounters and
commercial ship noise. The commercial ship encounter used in masking
analysis is shown in black outline. (B) Zoomed in spectrogram of 3 h (1 h
of duty-cycled data) of the commercial ship encounter. The two 5 min sec-
tions shown in bolded outline represent the center (shown in red) and edge
(shown in yellow) of the ship’s acoustic footprint.
TABLE VI. Descriptive statistics (mean 6SD) of CIB combined call type CI.c.5 (n¼8, 0.5%).
Whole call Whistle Pulse
Dur. (s) Min. (Hz) Max. (Hz) BW (Hz) Center (Hz) Peak (Hz) Min. (Hz) Max. (Hz) BW (Hz) Start (Hz) End (Hz) Trend Inflect. Min. PRR Max. PRR
1.9 60.4 1315.5 6985.3 11989.7 619.1 10674.2 6995.1 3257.8 6968.5 3293 61128.8 2711.9 6130.4 4632.6 6247.9 1920.7 6316.5 4542.9 6283.2 2830.8 6289.8 1.6 60.2 2.1 60.4 128.4 614.5 709.5 627.5
J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al. 3499
https://doi.org/10.1121/10.0022516
exhibit larger and more complex social units show greater
communicative complexity and these complex communica-
tion systems can express a wide range of emotional and
motivational states between group members (Freeberg et al.,
2012).
It has been hypothesized that some pulsed and/or com-
bined calls may aid in group cohesion. Calls used for this
specific purpose are referred to as contact calls, some of
which may possibly encode individual or group identity
(Vergara and Mikus, 2019). The use of contact calls
between group members has been studied in belugas in both
aquaria and the wild (Mishima et al., 2015;Mishima et al.,
2018;Morisaka et al., 2013;Panova et al., 2017;Panova
and Agafonov, 2023;Van Parijs et al., 2003;Vergara et al.,
2010;Vergara and Mikus, 2019). The use of individual
identity and group cohesion calls has been extensively stud-
ied in delphinids, particularly in bottlenose dolphins, which
are known to produce highly stereotyped signature whistles
containing individual identity and broadcast information
about the caller (Caldwell and Caldwell, 1965;Janik et al.,
2006;Janik and Slater, 1998;Sayigh et al., 1999). In killer
whales, pods exhibit unique calls that are thought to func-
tion as contact signals to maintain group cohesion (Ford,
1989;Yurk et al., 2002). Narwhals (Monodon monoceros),
which are the closest living relative to the beluga and only
other species in the family Monodontidae, have also been
shown to exhibit evidence of signature contact calls in both
adults (Shapiro, 2006) and mother–calf pairs (Ames et al.,
2021).
Belugas are highly social cetaceans and the CIB popula-
tion in particular live in an extremely turbid environment in
which acoustic signaling is the most effective form of sen-
sory modality. Identification calls at the level of the individ-
ual or group could aid in group cohesion efforts during
periods of isolation or separation via predator presence, tidal
influence, or anthropogenic disturbance. In Cook Inlet, we
documented several pulsed and combined calls that may
serve as contact calls given their structure (e.g., pulse.flat,
pulse.flat.seg, CI.c.6, CI.c.7). Though this study was not
able to investigate contact call use, we hope the classifica-
tion of these calls will be useful for future studies that aim
to investigate beluga contact calls in Cook Inlet.
B. Comparisons among beluga vocal repertoires
worldwide
Geographic variation in vocal repertoires among beluga
populations can occur through drift or active social learning
and may be a reliable indicator of divergence and population
FIG. 9. (Color online) One-third octave band frequency spectral levels of common CIB call types (blue), and commercial ship noise at close (red) and distant
(yellow) conditions received at the Susitna acoustic recorder. The audiogram (dashed gray) is the smoothed minimum beluga whale audiogram, also in one-
third octave bands (from Erbe et al., 2016). Whenever any portion of the call type signal is below the audiogram, hearing is audiogram limited and the signal
is inaudible. Whenever any portion of the call type signal is above the audiogram, hearing is noise-limited and masking occurs when any portion of the call
type signal is below the commercial ship noise.
3500 J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al.
https://doi.org/10.1121/10.0022516
structure among groups. In this study, we were not able to
conduct a formal, quantitative geographic variation analysis;
however, we have expanded on the call type comparison
table from Garland et al. (2015) to include call types docu-
mented in Cook Inlet and their relation to other beluga pop-
ulations (Table III). Our study shows that the CIB
population exhibits a number of unique call types and it is
likely that the divergence of the CIB vocal repertoire is due
to the population’s long-term geographic and genetic isola-
tion from other beluga populations (O’Corry-Crowe et al.,
1997).
Among whistles, multiple call types have been docu-
mented across all eight populations studied, including
ascending, descending, flat, modulated, n-shaped, and u-
shaped whistles. Overall, flat whistles appear to be a domi-
nant call type across beluga populations, being either the
most common or second most common call type across all
eight populations studied. In Cook Inlet, flat whistles
(flatws) comprised 20% of the vocal repertoire, which was
the second most common call type. Descending whistles,
the most common call type in Cook Inlet (dws, 22.1%),
were also the most common call type in the eastern Beaufort
Sea population (Garland et al., 2015). Segmented whistles
(aws.seg, dws.seg, flatws.seg, modws.seg, modws.m.seg,
modws.seg.t, nws.seg) were present in Cook Inlet, primarily
modulated segmented whistles (modws.seg, 4.7%) and were
also documented in the eastern Beaufort Sea (Garland et al.,
2015), Svalbard (Karlsen et al., 2002), Bristol Bay (Angiel,
1997), St. Lawrence (Faucher, 1988), and Cunningham Inlet
(Sjare and Smith, 1986) populations. Whistles in sequences
(aws.seq, dws.seq, dws.seq.t, modws.sa.seq, nws.seq, uws.-
seq) were also present in Cook Inlet, primarily descending
whistle sequence (dws.seq, 4.9%), which was also docu-
mented in the White Sea population (Belikov and
Bel’kovich, 2007). Trills were documented in Cook Inlet
(3%) and also occurred in the eastern Beaufort Sea (Garland
et al., 2015), Svalbard (Karlsen et al., 2002), St. Lawrence
(Faucher, 1988), and Cunningham Inlet (Sjare and Smith,
1986) populations while dotted trills (0.2%) were only docu-
mented in Cook Inlet.
Among Cook Inlet pulsed call types, only flat pulsed
calls were documented in all other populations. Along with
flat whistles, flat pulsed calls seem to be a universal call
type among all beluga populations studied thus far. Flat
pulsed calls (pulse.flat, 7%) were the most common pulsed
call type in Cook Inlet, and second most common call types
in Svalbard (Karlsen et al., 2002) and Bristol Bay (Angiel,
1997). Segmented pulsed calls (pulse.d.seg, pulse.flat.seg,
pulse.flat.seg.1–4, pulse.mod.seg) were present in Cook
Inlet, primarily flat segmented pulsed calls (pulse.flat.seg,
4%), which were the second most common pulsed call type.
This call type was also documented in the White Sea
(Belikov and Bel’kovich, 2008) and Cunningham Inlet
(Sjare and Smith, 1986). The third most common pulsed call
type in Cook Inlet was descending pulsed call (pulse.d,
2.6%) which was also documented in the eastern Beaufort
Sea (Garland et al., 2015), White Sea (Belikov and
Bel’kovich, 2008), Bristol Bay (Angiel, 1997), and
Cunningham Inlet (Sjare and Smith, 1986) populations. No
combined calls found in Cook Inlet were documented
elsewhere.
Due to the lack of standardization among beluga vocal
repertoire studies, one important caveat to highlight is that
differences in repertoires could potentially be in part due to
non-comparable recording effort across studies (e.g., num-
ber of days recorded, single season, and single location sam-
pling, effect of human and vessel presence on beluga vocal
behavior when collecting recordings, recorder sampling
rate, and duty cycle). In this study, our analysis was based
on 1633 calls, other studies varied from 460 calls in St.
Lawrence, Canada (Faucher, 1988) to 2839 calls across
three studies in the White Sea, Russia (Belikov and
Bel’kovich, 2006,2007,2008). Because our study is the
only one to use a rarefaction curve to assess vocal repertoire
richness and completeness, it is unclear how complete the
vocal repertoire is for other beluga populations.
C. Anthropogenic noise masking
In the presence of vessel noise, belugas have been
shown to reduce their overall calling rate and increase repe-
tition of specific calls, as well as shift the frequency range of
their calls upward (Lesage et al., 1999). Vergara et al.
(2021) modeled the communication range of various age
classes of belugas in the presence of vessel noise in the St.
Lawrence Estuary and showed that communication ranges
of young belugas were significantly reduced in the presence
of noise, which can have particularly detrimental effects on
mothers and dependent calves as it may prevent maintaining
contact with one another. Another study in St. Lawrence
showed that belugas vary vocalization levels as a function
of noise, indicating a Lombard vocal response (Lombard,
1911;Scheifele et al., 2005).
In Cook Inlet, previous studies have documented the
occurrence and potential impacts of anthropogenic noise
(Castellote et al., 2018;Mooney et al., 2020;Polasek, 2021;
Small et al., 2017), which has been listed as a high level
concern for the recovery of the CIB population (National
Marine Fisheries Service, 2016). In particular, commercial
shipping noise has been found to be the most prevalent
anthropogenic noise source within Cook Inlet, showing a
wide spatial distribution and temporal duration (Castellote
et al., 2018). Because of this, communication masking by
commercial shipping noise has been highlighted as a poten-
tial concern (Castellote et al., 2018). The masking of com-
munication signals can be problematic for a species that
relies heavily on acoustic communication, rather than vision
for navigation, foraging, group cohesion, and predator
avoidance. Our study provides empirical support that all
seven of the most common call types in the CIB vocal reper-
toire were partially masked by distant commercial ship noise
and completely masked by close commercial ship noise in
the frequency range up to 12 kHz. Auditory masking occurs
when one sound interferes with an individual’s ability to not
J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al. 3501
https://doi.org/10.1121/10.0022516
only detect, but also discriminate and recognize another
sound (Branstetter and Sills, 2022;Erbe et al., 2016). While
some beluga calls can have acoustic energy above 12 kHz,
the crucial components of the calls below 12 kHz will be
masked and the animal’s ability to discriminate, recognize,
and process the encoded information will be impacted.
This region of Susitna is in the core of the CIB critical
habitat and within 2000 m of the Port of Alaska commercial
shipping lanes. We consider these results a conservative
measure of masking of beluga communication in other areas
of the critical habitat along the commercial shipping lanes
(shown in Fig. 1). Due to the shallow nature and high glacial
sediment load of Cook Inlet (Sharma and Burrell, 1970),
sound propagation in this environment is strongly attenuated
(Au and Hastings, 2008). Therefore, the sound levels for the
commercial ship footprint reported 2000 m away would
likely occur at greater distances in the deeper waters of the
CIB critical habitat.
It has been estimated that roughly 486 commercial ships
use the Port of Alaska annually, with an average of 8–10
ships per week (Eley, 2006). Our results suggest that every
time a commercial vessel transits through the Port of Alaska
commercial shipping lanes, CIB communication is heavily
impacted within their core habitat. Our study fills a critical
research gap in which we quantify masking levels as they
relate to commonly used CIB call types. We recommend
that future research use acoustic tags on CIB to further
investigate the effect of masking by anthropogenic noise on
the behavior of individuals in the population, in particular in
areas of their critical habitat encroached by commercial
shipping noise, such as the Susitna Delta or lower Knik
Arm.
V. CONCLUSION
We have provided the first description of CIB vocal rep-
ertoire, which complements previous studies of vocal reper-
toire in other beluga populations, providing a critical
baseline framework that can be built upon to quantify geo-
graphic variability in vocal repertoire among populations, or
context-specific acoustic behavior within the Cook Inlet
population. We have also provided a quantification of call
masking by commercial ship noise for the most commonly
used CIB call types in the frequency range of 0–12 kHz.
Our masking analysis results indicate that, in the Susitna
area, beluga communication is heavily impacted during
each commercial ship passage, with all seven of the most
commonly used call types being partially or fully masked
during this noise disturbance.
ACKNOWLEDGMENTS
We thank Kim Goetz, Sarah Converse, Jessica Crance,
and Lori Polasek for providing comments that improved this
manuscript. We thank Kim Shelden for creating the map in
Fig. 1; Chris Garner, Justin Jenniges, and Del Westerholt for
field support; and Paul Wade for project support. We thank
Dimitri Ponirakis, Brian Branstetter, and Catherine Berchok
for helpful discussions regarding the masking portion of the
study. Funding was provided by the NOAA Fisheries
Species Recovery Grants to States Program (Grant No.
NA17NMF4720071 to the Alaska Department of Fish and
Game), Hilcorp Alaska LLC, Georgia Aquarium, Shedd
Aquarium, SeaWorld-Busch Gardens Conservation Fund,
and the H. Mason Keeler Endowed Professorship in Sports
Fisheries Management. The authors have no conflicts to
disclose. The data that support the findings of this study are
available from the corresponding author upon reasonable
request.
1
See supplementary material at https://doi.org/10.1121/10.0022516 for a
beluga call type comparison table; CART figure; the Random Forest con-
fusion matrix and call type error rates.
Ames, A. E., Blackwell, S. B., Tervo, O. M., and Heide-Jørgensen, M. P.
(2021). “Evidence of stereotyped contact call use in narwhal (Monodon
monoceros) mother-calf communication,” PLoS One 16, e0254393.
Angiel, N. M. (1997). “The vocal repertoire of the beluga whale in bristol
Bay, Alaska,” Master’s thesis, University of Washington, Seattle, WA.
ANSI (2018). ANSI S3/SC1.6-2018, Procedure for Determining
Audiograms in Toothed Whales through Evoked Potential Methods
(American National Standards Institute, New York).
Au, W. W. L., Carder, D. A., Penner, R. H., and Scronce, B. L. (1985).
“Demonstration of adaptation in beluga whale echolocation signals,”
J. Acoust. Soc. Am. 77, 726–730.
Au, W. W L., and Hastings, M. C. (2008). Principles of Marine
Bioacoustics (Springer, New York).
Belikov, R. A., and Bel’kovich, V. M. (2006). “High-pitched tonal signals
of beluga whales (Delphinapterus leucas) in a summer assemblage off
Solovetskii Island in the White Sea,” Acoust. Phys. 52, 125–131.
Belikov, R. A., and Bel’kovich, V. M. (2007). “Whistles of beluga whales
in the reproductive gathering off Solovetskii Island in the White Sea,”
Acoust. Phys. 53, 528–534.
Belikov, R. A., and Bel’kovich, V. M. (2008). “Communicative pulsed sig-
nals of beluga whales in the reproductive gathering off Solovetskii Island
in the White Sea,” Acoust. Phys. 54, 115–123.
Blevins-Manhard, R., Atkinson, S., and Lammers, M. (2017). “Spatial and
temporal patterns in the calling behavior of beluga whales,
Delphinapterus leucas, in Cook Inlet, Alaska,” Mar. Mammal Sci. 33,
112–133.
Bradbury, J. W., and Vehrencamp, S. L. (2011). Principles of Animal
Communication, 2nd ed. (Sinauer Associates, Inc., Sunderland, MA).
Brady, B., Hedwig, D., Trygonis, V., and Gerstein, E. (2020).
“Classification of Florida manatee (Trichechus manatus latirostris) vocal-
izations,” J. Acoust. Soc. Am. 147, 1597–1606.
Branstetter, B. K., and Sills, J. M. (2022). “Mechanisms of auditory mask-
ing in marine mammals,” Anim. Cogn. 25, 1029–1047.
Breiman, L. (2001). “Random forests,” Mach. Learn. 45, 5–32.
Breiman, L., Friedman, J. H., Olshen, R. A., and Stone, C. G. (1984).
Classification and Regression Trees (Chapman and Hall, London, UK).
Caldwell, M. C., and Caldwell, D. K. (1965). “Individualized whistle con-
tours in bottle-nosed dolphins (Tursiops truncatus),” Nature 207,
434–435.
Caldwell, M. C., and Caldwell, D. K. (1967). “Intraspecific transfer of
information via the pulsed sound in captive odontocete cetaceans,” Anim.
Sonar Syst. 2, 879–936.
Castellote, M., Mooney, T. A., Quakenbush, L., Hobbs, R., Goertz, C., and
Gaglione, E. (2014). “Baseline hearing abilities and variability in wild
beluga whales (Delphinapterus leucas),” J. Exp. Biol. 217, 1682–1691.
Castellote, M., Small, R. J., Lammers, M. O., Jenniges, J. J., Mondragon, J.,
and Atkinson, S. (2016). “Dual instrument passive acoustic monitoring of
belugas in Cook Inlet, Alaska,” J. Acoust. Soc. Am. 139, 2697–2707.
Castellote, M., Small, R., Lammers, M., Jenniges, J., Mondragon, J.,
Garner, C., Atkinson, S., Delevaux, J. M. S., Graham, R., and Westerholt,
D. (2020). “Seasonal distribution and foraging occurrence of Cook Inlet
beluga whales based on passive acoustic monitoring,” Endang. Species
Res. 41, 225–243.
3502 J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al.
https://doi.org/10.1121/10.0022516
Castellote, M., Thayre, B., Mahoney, M., Mondragon, J., Lammers, M. O.,
and Small, R. J. (2019). “Anthropogenic noise and the endangered Cook
Inlet beluga whale, Delphinapterus leucas: Acoustic considerations for
management,” Mar. Fish. Rev. 80, 63–88.
Chmelnitsky, E. G., and Ferguson, S. H. (2012). “Beluga whale,
Delphinapterus leucas, vocalizations from the Churchill River, Manitoba,
Canada,” J. Acoust. Soc. Am. 131, 4821–4835.
DeRuiter, S. L., Southall, B. L., Calambokidis, J., Zimmer, W. M. X.,
Sadykova, D., Falcone, E. A., Friedlaender, A. S., Joseph, J. E., Moretti,
D., Schorr, G. S., Thomas, L., and Tyack, P. L. (2013). “First direct mea-
surements of behavioural responses by Cuvier’s beaked whales to mid-
frequency active sonar,” Biol. Lett. 9, 20130223–20130225.
Dudzinski, K. M., Thomas, J. A., and Gregg, J. D. (2009). “Communication
in marine mammals,” in Encyclopedia of Marine Mammals, 2nd ed.,
edited by W. F. Perrin, B. W€
ursig, and J. G. M. Thewissen (Academic
Press, San Diego, CA), pp. 260–269.
Dugan, P. J., Ponirakis, D. W., Zollweg, J. A., Pitzrick, M. S., Morano, J.
L., Warde, A. M., Rice, A. N., Clark, C. W., and Van Parijs, S. M. (2011).
“SEDNA - Bioacoustic Analysis Toolbox,” in Oceans-11 Mts,IEEE,
Kona (IEEE, New York).
Eley, W. D. (2006). Cook Inlet Vessel Traffic Survey: Report to Cook Inlet
Regional Citizens Advisory Council (Cape International, Inc. & Nuka
Research and Planning Group, LLC, Juneau, AK).
Erbe, C. (2008). “Critical ratios of beluga whales (Delphinapterus leucas)
and masked signal duration,” J. Acoust. Soc. Am. 124, 2216–2223.
Erbe, C., Reichmuth, C., Cunningham, K., Lucke, K., and Dooling, R.
(2016). “Communication masking in marine mammals: A review and
research strategy,” Mar. Pollut. Bull. 103, 15–38.
Faucher, A. (1988). “The vocal repertoire of the St. Lawrence estuary popu-
lation of beluga whale (Delphinapterus leucas) and its behavioral, social,
and environmental contexts,” Master’s thesis, Dalhousie University,
Halifax, Nova Scotia, Canada.
Filatova, O. A., Fedutin, I. D., Nagaylik, M. M., Burdin, A. M., and Hoyt,
E. (2009). “Usage of monophonic and biphonic calls by free-ranging resi-
dent killer whales (Orcinus orca) in Kamchatka, Russia Far East,” Acta
Ethol. 12, 37–44.
Finneran, J. J. (2015). “Noise-induced hearing loss in marine mammals: A
review of temporary threshold shift studies from 1996 to 2015,”
J. Acoust. Soc. Am. 138, 1702–1726.
Fish, M., and Mowbray, W. H. (1962). “Production of underwater sound by
the white whale or beluga, Delphinapterus leucas,” J. Mar. Res. 20,
149–162.
Ford, J. K. B. (1989). “Acoustic behaviour of resident killer whales
(Orcinus orca) off Vancouver Island, British Columbia,” Can. J. Zool. 67,
727–745.
Freeberg, T. M., Dunbar, R. I. M., and Ord, T. J. (2012). “Social complexity
as a proximate and ultimate factor in communicative complexity,” Philos.
Trans. R. Soc. B. 367, 1785–1801.
Garland, E. C., Castellote, M., and Berchok, C. L. (2015). “Beluga whale
(Delphinapterus leucas) vocalizations and call classification from the
eastern Beaufort Sea population,” J. Acoust. Soc. Am. 137, 3054–3067.
Goetz, K. T., Montgomery, R. A., Ver Hoef, J. M., Hobbs, R. C., and
Johnson, D. S. (2012). “Identifying essential summer habitat of the endan-
gered beluga whale Delphinapterus leucas in Cook Inlet, Alaska,”
Endang. Species Res. 16, 135–147.
Goetz, K. T., Rugh, D. J., Read, A. J., and Hobbs, R. C. (2007). “Habitat
use in a marine ecosystem: Beluga whales Delphinapterus leucas in Cook
Inlet, Alaska,” Mar. Ecol. Prog. Ser. 330, 247–256.
Goetz, K. T., Shelden, K. E. W., Sims, C. L., Waite, J. M., and Wade, P. R.
(2023). “Abundance of belugas (Delphinapterus leucas) in Cook Inlet,
Alaska, June 2021 and June 2022,” Alaska Fisheries Science Center
Processed Report No. 2023-03, National Oceanic and Atmospheric
Administration, National Marine Fisheries Service, Seattle, WA.
Hammerschmidt, K., and Fischer, J. (2019). “Baboon vocal repertoires and
the evolution of primate vocal diversity,” J. Hum. Evol.126, 1–13.
Hill, P. S., and DeMaster, D. P. (1998). “Alaska marine mammal stock
assessments, 1998,” U.S. Dep. Commer., NOAA Tech. Memo. NMFS-
AFSC-97, 165.
Hobbs, R. C., Laidre, K. L., Vos, D. J., Mahoney, B. A., and Eagleton, M.
(2005). “Movements and area use of belugas, Delphinapterus leucas,ina
subarctic Alaskan estuary,” Arctic 58, 331–458.
Holt, M. M., Noren, D. P., and Emmons, C. K. (2011). “Effects of noise lev-
els and call types on the source levels of killer whale calls,” J. Acoust.
Soc. Am. 130, 3100–3106.
Janik, V. M., Sayigh, L. S., and Wells, R. S. (2006). “Signature whistle
shape conveys identity information to bottlenose dolphins,” Proc. Natl.
Acad. Sci. U.S.A. 103, 8293–8297.
Janik, V. M., and Slater, P. J. B. (1998). “Context-specific use suggests that
bottlenose dolphin signature whistles are cohesion calls,” Anim. Behav.
56, 829–838.
Karlsen, J. D., Bisther, A., Lydersen, C., Haug, T., and Kovacs, K. M.
(2002). “Summer vocalisations of adult male white whales
(Delphinapterus leucas) in Svalbard, Norway,” Polar Biol. 25, 808–817.
Kershenbaum, A., Blumstein, D. T., Roch, M. A., Akc¸ay, C¸., Backus, G.,
Bee, M. A., Bohn, K., Cao, Y., Carter, G., C€
asar, C., Coen, M., DeRuiter,
S. L., Doyle, L., Edelman, S., Ferreri-Cancho, R., Freeberg, T. M.,
Garland, E. C., Gustison, M., Harley, H. E., Huetz, C., Hughes, M.,
Hyland Bruno, J., Ilany, A., Jin, D. Z., Johnson, M., Ju, C., Karnowski, J.,
Lohr, B., Manser, M. B., McCowan, B., Mercado, E., III, Narins, P. M.,
Piel, A., Rice, M., Salmi, R., Sasahara, K., Sayigh, L., Shiu, Y., Taylor,
C., Vallejo, E. E., Waller, S., and Zamora-Gutierrez, V. (2016). “Acoustic
sequences in non-human animals: A tutorial review and prospectus,”
Biol. Rev. 91, 13–52.
Klishin, V. O., Popov, V. V., and Supin, A. Y. (2000). “Hearing capabilities
of a beluga whale, Delphinapterus leucas,” Aquat. Mamm. 26, 212–228,
available at https://www.aquaticmammalsjournal.org/article/volume-26-
issues/.
K. Lisa Yang Center for Conservation Bioacoustics. (2023). “Raven Pro:
Interactive Sound Analysis Software,” Cornell Lab Ornithology, V 1.6.4.
Kovacs, K. M., Romano, T. A., Reeves, R. R., Hobbs, R. C., Desportes, G.,
Brennan, R., and Castellote, M. (2021). “Polar Research Special
Cluster—Beluga whales (Delphinapterus leucas): Knowledge from the
wild, human care and TEK,” Polar Res. 40, 8235.
Laidre, K. L., Shelden, K. E. W., Rugh, D. J., and Mahoney, B. A. (2000).
“Beluga, Delphinapterus leucas, distribution and survey effort in the Gulf
of Alaska,” Mar. Fish. Rev. 62, 27–36, available at https://spo.nmfs.noaa.
gov/content/beluga-delphinapterus-leucas-distribution-and-survey-effort-
gulf-alaska.
Lammers, M. O., Castellote, M., Small, R. J., Atkinson, S., Jenniges, J.,
Rosinski, A., Oswald, J. N., and Garner, C. (2013). “Passive acoustic
monitoring of Cook Inlet beluga whales (Delphinapterus leucas),”
J. Acoust. Soc. Am. 134, 2497–2504.
Lemasson, A., Mikus, M.-A., Blois-Heulin, C., and Lod
e, T. (2014). “Vocal
repertoire, individual acoustic distinctiveness, and social networks in a
group of captive Asian small-clawed otters (Aonyx cinerea),” J. Mammal.
95, 128–139.
Lesage, V., Barrette, C., Kingsley, M. C. S., and Sjare, B. (1999). “The
effect of vessel noise on the vocal behavior of belugas in the St. Lawrence
River Estuary, Canada,” Mar. Mammal Sci. 15, 65–84.
Leuchtenberger, C., Sousa-Lima, R., Duplaix, N., Magnusson, W. E., and
Mour~
ao, G. (2014). “Vocal repertoire of the social giant otter,” J. Acoust.
Soc. Am. 136, 2861–2875.
Liaw, A., and Wiener, M. (2002). “Classification and regression by
randomForest,” R News 2, 18–22.
Lilly, J. C., and Miller, A. M. (1961). “Vocal exchanges between dolphins,”
Science 134, 1873–1876.
Lombard, E. (1911). “Le signe de l’
el
evation de la voix” (“The sign of rais-
ing your voice”), Ann. Mal. L’creille 37, 101–109.
Macedonia, J. M. (1993). “The vocal repertoire of the ringtailed lemur
(Lemur catta),” Folia Primatol. 61, 186–217.
Martin, M., Gridley, T., Harvey Elwen, S., and Charrier, I. (2021). “Vocal
repertoire, micro-geographical variation and within-species acoustic parti-
tioning in a highly colonial pinniped, the Cape fur seal,” R. Soc. Open
Sci. 8, 202241.
Martin, M. J., Halliday, W. D., Storrie, L., Citta, J. J., Dawson, J., Hussey,
N. E., Juanes, F., Loseto, L. L., MacPhee, S. A., Moore, L., Nicoll, A.,
O’Corry-Crowe, G., and Insley, S. J. (2023). “Exposure and behavioral
responses of tagged beluga whales (Delphinapterus leucas) to ships in the
Pacific Arctic,” Mar. Mammal Sci. 39, 387–421.
McGuire, T. L., Himes Boor, G. K., McClung, J. R., Stephens, A. D.,
Garner, C., Shelden, K. E. W., and Wright, B. (2020). “Distribution and
habitat use by endangered Cook Inlet beluga whales: Patterns observed
J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al. 3503
https://doi.org/10.1121/10.0022516
during a photo-identification study, 2005–2017,” Aquat. Conserv. 30,
2402–2427.
McShane, L. J., Estes, J. A., Riedman, M. L., and Staedler, M. M. (1995).
“Repertoire, structure, and individual variation of vocalizations in the sea
otter,” J. Mammal. 76, 414–427.
Mishima, Y., Morisaka, T., Itoh, M., Matsuo, I., Sakaguchi, A., and
Miyamoto, Y. (2015). “Individuality embedded in the isolation calls of
captive beluga whales (Delphinapterus leucas),” Zool. Lett. 1, 27.
Mishima, Y., Morisaka, T., Mishima, Y., Sunada, T., and Miyamoto, Y.
(2018). “Redefinition and sexual difference of contact calls in belugas
(Delphinapterus leucas),” Aquat. Mamm. 43, 538–554.
Mooney, T. A., Castellote, M., Jones, I., Rouse, N., Rowles, T., Mahoney,
B., and Goertz, C. E. C. (2020). “Audiogram of a Cook Inlet beluga whale
(Delphinapterus leucas),” J. Acoust. Soc. Am. 148, 3141–3148.
Morisaka, T., Yoshida, Y., Akune, Y., Mishima, H., and Nishimoto, S.
(2013). “Exchange of ‘signature’ calls in captive belugas (Delphinapterus
leucas),” J. Ethol. 31, 141–149.
Murray, S. O., Mercado, E., and Roitblat, H. L. (1998). “Characterizing the
graded structure of false killer whale (Pseudorca crassidens) vocal-
izations,” J. Acoust. Soc. Am. 104, 1679–1688.
National Marine Fisheries Service (2016). Recovery Plan for the Cook Inlet
Beluga Whale (Delphinapterus Leucas) (National Marine Fisheries
Service, Alaska Region, Protected Resources Division, Juneau, AK).
National Marine Fisheries Service (2018). 2018 Revisions to: Technical
guidance for assessing the effects of anthropogenic sound on marine
mammal hearing (V 2.0): Underwater thresholds for onset of permanent
and temporary threshold shifts. U.S. Dept. of Commer., National Oceanic
and Atmospheric Administration, Technical Memorandum NMFS-OPR-
59, p. 167.
O’Corry-Crowe, G., Suydam, R., Quakenbush, L., Smith, T. G., Lydersen,
C., Kovacs, K. M., Orr, J., Harwood, L., Litovka, D., and Ferrer, T.
(2020). “Group structure and kinship in beluga whale societies,” Sci. Rep.
10, 11462.
O’Corry-Crowe, G. M., Suydam, R. S., Rosenberg, A., Frost, K. J., and
Dizon, A. E. (1997). “Phylogeography, population structure and dispersal
patterns of the beluga whale Delphinapterus leucas in the western
Nearctic revealed by mitochondrial DNA,” Mol. Ecol. 6, 955–970.
Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., and
McGlinn, D. (2022). “Vegan: Community ecology package,” https://
cran.r-project.org/web/packages/vegan/index.html.
Panova, E., and Agafonov, A. (2023). “Possible occurrence of contact calls
in all-male groups of free-ranging beluga whales,” J. Zool. 320, 29–41.
Panova, E., Agafonov, A., Belikov, R., and Melnikova, F. (2017).
“Vocalizations of captive beluga whales, Delphinapterus leucas:
Additional evidence for contact signature ‘mixed’ calls in belugas,” Mar.
Mammal Sci. 33, 889–903.
Panova, E., Agafonov, A., Belikov, R., and Melnikova, F. (2019).
“Characteristics and microgeographic variation of whistles from the vocal
repertoire of beluga whales (Delphinapterus leucas) from the White Sea,”
J. Acoust. Soc. Am. 146, 681–692.
Papale, E., Buffa, G., Filiciotto, F., Maccarrone, V., Mazzola, S., Ceraulo,
M., Giacoma, C., and Buscaino, G. (2015). “Biphonic calls as signature
whistles in a free-ranging bottlenose dolphin,” Bioacoustics 24, 223–231.
Parsons, E. C. M. (2017). “Impacts of Navy sonar on whales and dolphins:
Now beyond a smoking gun?,” Front. Mar. Sci. 4, 295.
Phillips, A. V., and Stirling, I. (2001). “Vocal repertoire of South American
fur seals, Arctocephalus australis: Structure, function, and context,” Can.
J. Zool. 79, 420–437.
Polasek, L. (2021). “Foraging ecology and habitat use of cook inlet beluga
whales (Delphinapterus leucas) final report,” in ESA Section 6 Award
#NA17NMF4720071 (Alaska Department of Fish and Game, Marine
Mammal Program, Juneau, AK).
R Core Team (2022). R: A Language and Environment for Statistical
Computing, R Foundation for Statistical Computing. https://www.R-pro-
ject.org/.
Ridgway, S., Samuelson Dibble, D., Van Alstyne, K., and Price, D. (2015).
“On doing two things at once: Dolphin brain and nose coordinate sonar
clicks, buzzes, and emotional squeals with social sounds during fish
capture,” J. Exp. Biol. 218, 3987–3995.
Rugh, D. J., Shelden, K. E. W., and Hobbs, R. C. (2010). “Range contrac-
tion in a beluga whale population,” Endang. Species Res. 12, 69–75.
Saunders, D. A. (1983). “Vocal repertoire and individual vocal recognition
in the short-billed white-tailed black cokcatoo, Calyptorhynchus funereus
latirostris CArnaby,” Wildl. Res. 10, 527–536.
Sayigh, L., Quick, N., Hastie, G., and Tyack, P. (2013). “Repeated call
types in short-finned pilot whales, Globicephala macrohynchus,” Mar.
Mammal Sci. 29, 312–324.
Sayigh, L. S., Tyack, P. L., Wells, R. S., Solow, A. R., Scott, M. D., and
Irvine, A. B. (1999). “Individual recognition in wild bottlenose dolphins:
A field test using playback experiments,” Anim. Behav. 57, 41–50.
Scheifele, P. M., Andrew, S., Cooper, R. A., Darre, M., Musiek, F. E., and
Max, L. (2005). “Indication of a Lombard vocal response in the St.
Lawrence River beluga,” J. Acoust. Soc. Am. 117, 1486–1492.
Selbmann, A., Deecke, V. B., Filatova, O. A., Fedutin, I. D., Miller, P. J.
O., Simon, M., Bowles, A. E., Lyrholm, T., Lacey, C., Magnusdottir, E.
E., Maunder, W., Wensveen, P. J., Svavarsson, J., and Samarra, F. I. P.
(2023). “Call type repertoire of killer whales (Orcinus orca) in Iceland
and its variation across regions,” Mar. Mammal Sci. 39(4), 1136–1160.
Shapiro, A. D. (2006). “Preliminary evidence for signature vocalizations
among free-ranging narwhals (Monodon monoceros),” J. Acoust. Soc.
Am. 120, 1695–1705.
Sharma, G. D., and Burrell, D. C. (1970). “Sedimentary environment and
sediments of Cook Inlet, Alaska,” Am. Assoc. Pet. Geol. Bull. 54,
647–645.
Sharpe, D. L., Castellote, M., Wade, P. R., and Cornick, L. A. (2019). “Call
types of Bigg’s killer whales (Orcinus orca) in western Alaska: Using
vocal dialects to assess population structure,” Bioacoustics 28, 74–99.
Shelden, K. E. W., Goetz, K. T., Rugh, D. J., Calkins, D. G., Mahoney, B.
A., and Hobbs, R. C. (2015). “Spatio-temporal changes in beluga whale,
Delphinapterus leucas, distribution: Results from aerial surveys (1977-
2014), opportunistic sightings (1975-2014), and satellite tagging (1999-
2003) in cook inlet, Alaska,” Mar. Fish. Rev. 77, 1–31, available at
https://spo.nmfs.noaa.gov/content/spatio-temporal-changes-beluga-whale-
delphinapterus-leucas-distribution-results-aerial.
Shelden, K. E. W., and Wade, P. R. (2019). “Aerial surveys, distribution,
abundance, and trend of belugas (Delphinapterus leucas) in Cook Inlet,
Alaska, June 2018,” Alaska Fisheries Science Center Processed Report
No. 2019-09, National Oceanic and Atmospheric Administration,
National Marine Fisheries Service, Seattle, WA.
Sjare, B. L., and Smith, T. G. (1986). “The vocal repertoire of white whales,
Delphinapterus leucas, summering in Cunningham Inlet, Northwest
Territories,” Can. J. Zool. 64, 407–415.
Small, R. J., Brost, B., Hooten, M., Castellote, M., and Mondragon, J.
(2017). “Potential for spatial displacement of Cook Inlet beluga whales
by anthropogenic noise in critical habitat,” Endang. Species Res. 32,
43–57.
Therneau, T., Atkinson, B., and Ripley, B. (2022). “Recursive partitioning
and regression trees,” https://cran.r-project.org/package=rpart.
Trejos-Araya, C., and Barrantes, G. (2014). “Natural history and acoustic
repertoire of the large-footed finch (Pezopetes capitalis), an endemic,
highland bird of Costa Rica and Western Panama,” Ornitol. Neotrop. 25,
261–271.
Tyack, P. (1981). “Interactions between singing Hawaiian humpback
whales and conspecifics nearby,” Behav. Ecol. Sociobiol. 8, 105–116.
Tyack, P. L., and Janik, V. M. (2013). “Effects of noise on acoustic signal
production in marine mammals,” in Animal Communication and Noise,
edited by H. Brumm (Springer, Berlin, Heidelberg), Vol. 2, pp. 251–271.
U.S. Federal Register (2000). Designating the Cook Inlet, Alaska, stock of
beluga whale as depleted under the Marine Mammal Protection Act
(MMPA), No. FR 65, National Marine Fisheries Service, National
Oceanic and Atmospheric Administration, U.S. Department of
Commerce, pp. 34590–34597.
U.S. Federal Register (2008). Endangered and threatened species; endan-
gered status for the Cook Inlet beluga whale, No. 205, pp. 62919–62930.
U.S. Federal Register (2011). Designation of critical habitat for Cook Inlet
beluga whale, No. 69, pp. 20180–20214.
Van Cise, A. M., Roch, M. A., Baird, R. W., Aran Mooney, T., and Barlow,
J. (2017). “Acoustic differentiation of Shiho- and Naisa-type short-finned
pilot whales in the Pacific Ocean,” J. Acoust. Soc. Am. 141, 737–748.
Van Parijs, S. M., Lydersen, C., and Kovacs, K. M. (2003). “Sounds pro-
duced by individual white whales, Delphinapterus leucas, from Svalbard
during capture (L),” J. Acoust. Soc. Am. 113, 57–60.
3504 J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al.
https://doi.org/10.1121/10.0022516
Vergara,V.,andBarrett-Lennard,L.G.(2008). “Vocal development
in a beluga calf (Delphinapterus leucas),” Aquat. Mamm. 34,
123–143.
Vergara, V., Michaud, R., and Barrett-Lennard, L. (2010). “What can cap-
tive whales tell us about their wild counterparts? Identification, usage,
and ontogeny of contact calls in belugas (Delphinapterus leucas),” Int. J.
Comp. Psychol. 23, 278–309.
Vergara, V., and Mikus, M. (2019). “Contact call diversity in natural beluga
entrapments in an Arctic estuary: Preliminary evidence of vocal signa-
tures in wild belugas,” Mar. Mammal Sci. 35, 434–465.
Vergara, V., Wood, J., Lesage, V., Ames, A., Mikus, M.-A., and
Michaud, R. (2021). “Can you hear me? Impacts of underwater noise
on communication space of adult, sub-adult and calf contact calls of
endangered St. Lawrence belugas (Delphinapterus leucas),” Polar
Res. 40, 5521.
Watkins, W. A. (1966). “The harmonic interval: Fact or artifact in spectral
analysis of pulse trains,” Mar. Bio-Acoust. 2, 15–43.
Weilgart, L., and Whitehead, H. (1997). “Group-specific dialects and geo-
graphical variation in coda repertoire in South Pacific sperm whales,”
Behav. Ecol. Sociobiol. 40, 277–2851.
Whitehead, H., Dillon, M., Dufault, S., Weilgart, L., and Wright, J. (1998).
“Non-geographically based population structure of South Pacific sperm
whales: Dialects, fluke-markings and genetics,” J. Anim. Ecol. 67, 253–262.
Yurk, H., Barrett-Lennard, L., Ford, J. K. B., and Matkin, C. O. (2002).
“Cultural transmission within maternal lineages: Vocal clans in resident
killer whales in southern Alaska,” Anim. Behav. 63, 1103–1119.
J. Acoust. Soc. Am. 154 (5), November 2023 Brewer et al. 3505
https://doi.org/10.1121/10.0022516
