Using a coherent hydrophone array for observing sperm whale range, classification, and shallow ...

13 Pages·2014·2.9 MB·English

Using a coherent hydrophone array for observing sperm
whale range, classification, and shallow-water dive
profiles
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation Tran, Duong D., Wei Huang, Alexander C. Bohn, Delin Wang,
Zheng Gong, Nicholas C. Makris, and Purnima Ratilal. “Using a
Coherent Hydrophone Array for Observing Sperm Whale Range,
Classification, and Shallow-Water Dive Profiles.” The Journal of
the Acoustical Society of America 135, no. 6 (June 2014):
3352–3363. © 2014 Acoustical Society of America
As Published http://dx.doi.org/10.1121/1.4874601
Publisher Acoustical Society of America (ASA)
Version Final published version
Accessed Tue Apr 09 20:31:37 EDT 2019
Citable Link http://hdl.handle.net/1721.1/97630
Terms of Use Article is made available in accordance with the publisher's policy
and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.
Detailed Terms
Using a coherent hydrophone array for observing sperm whale
range, classification, and shallow-water dive profiles
DuongD.Tran,WeiHuang,AlexanderC.Bohn,andDelinWang
DepartmentofElectricalandComputerEngineering,NortheasternUniversity,360HuntingtonAvenue,
Boston,Massachusetts02115
ZhengGongandNicholasC.Makris
DepartmentofMechanicalEngineering,MassachusettsInstituteofTechnology,77MassachusettsAvenue,
Cambridge,Massachusetts02139
PurnimaRatilala)
DepartmentofElectricalandComputerEngineering,NortheasternUniversity,360HuntingtonAvenue,
Boston,Massachusetts02115
(Received17September2013;revised20February2014;accepted11April2014)
Sperm whales in the New England continental shelf and slope were passively localized, in both
range and bearing, and classified using a single low-frequency (<2500 Hz), densely sampled,
towedhorizontalcoherenthydrophonearraysystem.Whalebearingswereestimatedusingtime-do-
main beamforming that provided highcoherent array gain in sperm whale click signal-to-noise ra-
tio. Whale ranges from the receiver array center were estimated using the moving array
triangulation technique from a sequence of whale bearing measurements. Multiple concurrently
vocalizing sperm whales, in the far-field of the horizontal receiver array, were distinguished and
classifiedbased on their horizontal spatial locations and the inter-pulse intervals of their vocalized
clicksignals.ThediveprofilewasestimatedforaspermwhaleintheshallowwatersoftheGulfof
Maine with 160m water-column depth located close to the array’s near-field where depth estima-
tion was feasible by employing time difference of arrival of the direct and multiply reflected click
signals received on the horizontal array. By accounting for transmission loss modeled using an
ocean waveguide-acoustic propagation model, the sperm whale detection range was found to
exceed60kminlowtomoderateseastateconditionsaftercoherentarrayprocessing.
VC 2014AcousticalSocietyofAmerica.[http://dx.doi.org/10.1121/1.4874601]
PACSnumber(s):43.30.Pc,43.30.Sf[AMT] Pages:3352–3363
I. INTRODUCTION 2004;Telonietal.,2007).Hereweshowthatitispossibleto
distinguish and classify multiple vocalizing sperm whale
Sperm whales (Physeter macrocephalus) in the
individuals located in the far-field of a single, densely
Northwest Atlantic during spring and summer are concen-
sampled,towedhorizontalcoherenthydrophonearraysystem
trated along continental slope regions from the Mid-Atlantic
using the instantaneous sperm whale position estimates in
BighttosouthofGeorgesBank(Whiteheadetal.,1992)and
both range and bearing, and the IPIs of the vocalized click
the Scotian shelf edge. While foraging, sperm whales per-
signals. Most studies of the vocalization behavior and dive
formdeepdiveslastingfromseveralminutestomorethanan
profileofspermwhaleshavebeenconfinedtodeepcontinen-
hour (Watwood et al., 2006), emitting short-duration broad-
tal slope environments bounding the Pacific and Atlantic
band clicks with frequencies ranging from several hundred
ocean (Jacquet et al., 2001; Madsen et al., 2002b; Mathias
hertz to more than 30kHz (Madsen et al., 2002b; Weilgart
et al., 2012; Watwood et al., 2006). Here we provide esti-
and Whitehead, 1988). Each click exhibits a multi-pulse
mates of the three-dimensional (3D) dive profile of a sperm
structure (Møhl, 2001; Norris and Harvey, 1972; Zimmer
whaleindividualthevocalizationsofwhichwereopportunis-
et al., 2004) arising from reflections of the acoustic signal
ticallyrecordedintheshallowwaterenvironmentoftheGulf
generatedbythephoniclipsoffthefrontalanddistalairsacs
of Maine with roughly 160m water-column depth during a
bounding the spermaceti organ of a sperm whale. The inter-
sea test of a newly developed, densely sampled, towed hori-
pulse interval (IPI) provides a measure of the length of the
zontalcoherentreceiverarraysysteminMay2013.
spermaceti organ that has been shown to be strongly corre-
Localization of an acoustic source, such as a vocalizing
lated with the size of a sperm whale individual (Antunes
sperm whale, in the far-field of a single, densely sampled,
et al., 2010; Gordon, 1990; Growcott et al., 2011; Mathias
towed horizontal coherent hydrophone array system is often
et al., 2009; Miller et al., 2013; Rhinelander and Dawson,
atwo-stageprocess.First,thebearingorhorizontaldirection
of arrival of the acoustic signal is determined by time-delay
a)Author to whom correspondence should be addressed. Electronic mail: analysisor beamforming of thesignalsreceived onthe indi-
[email protected] vidual hydrophone elements of the array. Second, the range
3352 J.Acoust.Soc.Am.135(6),June2014 0001-4966/2014/135(6)/3352/12/$30.00 VC 2014AcousticalSocietyofAmerica
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 18.51.1.88 On: Wed, 01 Jul 2015 18:28:21
or horizontal distance of the acoustic source from the re- the environment, suchasbathymetry orsound speed profile,
ceiver array center is determined by tracking changes in the is needed to estimate source range in the far-field of the
bearing of a series of acoustic emissions over time (Gong array.
et al., 2013; Nardone et al., 1984; Oshman, 1999; Ristic Other approaches for localizing sperm whales include
etal.,2004;Yardimetal.,2011).Short-aperture,towedhor- (a)hyperbolicrangingwithasmallnetworkofsinglehydro-
izontal coherent hydrophone array systems have been previ- phones (Baggenstoss, 2011; Tiemann and Porter, 2003;
ously used to record vocalizations from sperm whales. Watkins and Schevill, 1972) and (b) time-delay measure-
However, the coherent receiver array data have only been ment of click reflections from ocean boundaries acquired
applied to determine the bearing of a sperm whale. No sub- with a single hydrophone or a sparse array of hydrophones
sequentrangeestimateshavebeenmadebasedsolelyonthe (Mathiasetal.,2013;Mathiasetal.,2012;NosalandFrazer,
coherentreceiver arraymeasurements.Consequently,coher- 2006; Skarsoulis and Kalogerakis, 2006; Thode, 2004;
entarraygainofdenselysampledhydrophonearraysystems Tiemannetal.,2006;Wahlberg,2002).Hereweutilizetime
has not been previously exploited for range localization of difference ofarrivalsofthespermwhaledirectandmultiple
sperm whales. In Teloni (2005), a 128-element horizontal bottom and surface reflected click [multiple-reflection based
coherent hydrophone array system of the NATO Undersea time difference of arrival (MR-TDA)] signals after beam-
Research Center (NURC) with an array aperture length of formingtoestimatethedepthandhencethediveprofileofa
11.6mwasemployedtodeterminespermwhalevocalization sperm whale in shallow waters of the Gulf of Maine with
bearings and to separate whale clicks from different azi- 160m water-column depth. This sperm whale individual’s
muthal directions, but no range estimates or range analysis horizontalranger¼1kmwasveryclosetothearray’snear-
were provided. In Zimmer et al. (2004), click data from a field distance r (r (cid:2)L2=k(cid:3)750m, where L is the array
N N
singlespermwhale acquiredusingthesameNURC receiver aperture length and k is the wavelength) making it possible
arraywereusedtodeterminethebearingsofthewhaleindi- to estimate its depth. Depth estimation for acoustic sources
vidual.Thespermwhalewasprimarilytrackedusingadigi- at long ranges, in the far-field ðr (cid:4)L2=kÞ of a single, hori-
tal tag (DTAG) attached to its body consisting of a zontalcoherentreceiverarraysystemischallengingbecause
hydrophone used to record sounds directly from the whale the acoustic wavefield received by the array is multi-modal
(Zimmeretal.,2004).Thespermwhalerangetothereceiver in nature having undergone many surface and bottom boun-
array center was determined from click travel time differ- ces in a random ocean waveguide making the received field
ence between the DTAG hydrophone and the towed array less sensitive to the source’s depth location, except in the
hydrophones(Zimmeretal.,2004). endfiredirectionofthehorizontalarray.
Here we localize multiple sperm whales in the far-field
of a single low-frequency (<2.5 kHz), densely sampled,
II. METHOD:EXPERIMENTALDATACOLLECTION
towedhorizontalcoherenthydrophonearraysystem,provid-
ANDANALYSIS
ing estimates of both range and bearing for each sperm
whale. Because no other acoustic sensors were available to A newly developed, densely sampled, towed horizontal
usapartfromthetowedhorizontalreceiverarraysystem,the linear hydrophone array system funded by the National
whale ranges were estimated from their bearing-time trajec- Science Foundation and the Office of Naval Research was
tories. A review of methods that can be applied to passively deployed and tested in the continental slope region south of
estimate the range of an acoustic source from a single, Cape Cod between 500 and 2000m water depth on May 13
denselysampled,towedhorizontalcoherentreceiverarrayis (site B in Fig. 1) and in the Gulf of Maine in 150–180m
providedinSec.IofGongetal.(2013).Hereweemploythe water depth on May 14 and 15 (site A in Fig. 1). Passive
moving array triangulation technique developed in Gong acoustic data were collected on a sub-aperture of the array
etal.(2013)toestimate spermwhale rangesfromthemeas- with N¼32 elements having an inter-element spacing of
ured click bearings. This technique combines bearing meas- 0.75m. The hydrophone elements, each having (cid:5)188dB re
urements from spatially separated apertures of the towed lPa/V sensitivity were sampled at 5kHz with 24-bit digital
horizontal coherent receiver array and employs the conven- resolution. The array was towed by the research vessel
tionaltriangulationrangingalgorithmforlocalizingasource Endeavor at various speeds between 1 and 4kn. The water-
that may be in the near- or far-field of the array. Because column sound speed at the experiment sites were monitored
data from only a single towed horizontal coherent receiver using a conductivity-temperature-depth (CTD) sensor and
array are used here to remotely and passively localize both expendable bathythermographs(XBT). The measured sound
the range andbearing of sperm whales andto classify them, speed profiles are shown in Fig. 2 for the two test sites. The
the methods and results developed here are highly relevant array depth was maintained between 60 and 80m near the
andcanbedirectlyappliedtoaddressthefeasibilityofmoni- thermocline in the Gulf of Maine, and varied between 10
toringspermwhaleswithothertowedhorizontalcoherentre- and50matthecontinentalsloperegion.
ceiver array systems, such as those employed in naval and Toinvestigatethepresenceofspermwhaleclicks,time-
geophysical applications, where it may be important and frequency spectrograms of the received signal on each
necessary to remotely sense marine mammal activity from hydrophone were first calculated, and the spectrogram inco-
long ranges. An advantage of bearings-only range localiza- herentlyaveragedacrossseveralhydrophoneswereobtained.
tion with a densely sampled, towed horizontal coherent re- Sperm whale clicks were consistently present in all 75min
ceiver array system is that no additional information about of passive acoustic recordings on May 13 at the continental
J.Acoust.Soc.Am.,Vol.135,No.6,June2014 Tranetal.:Spermwhalerangelocalizationandclassification 3353
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 18.51.1.88 On: Wed, 01 Jul 2015 18:28:21
coverageofthebandwidthofthespermwhaleclickthatcan
extend to 30kHz (Madsen et al., 2002a; Mathias et al.,
2012; Teloni, 2005; Thode, 2004; Watwood et al., 2006;
Zimmeretal.,2004).
A. Spermwhalelocalizationinhorizontalrange
andazimuthalbearing
1. Determiningclickarrivaltimeandazimuthal
bearing
The relative horizontal azimuthal direction or relative
bearing b^0 of each sperm whale click, measured from array
broadside,wasnextestimatedusingtime-domaindelay-and-
sum beamforming (Johnson and Dudgeon, 1993). The
pressure-time series data from each hydrophone of the array
were first high-pass filtered with 300Hz cut-off frequency.
Each two-dimensional (2D) matrix of high-pass filtered
FIG. 1.Locations of the two test sites where the densely sampled, towed pressure-time series data from the 32 elements of the array
horizontal coherent receiver array was deployed to collect ambient noise within roughly 13s duration was next converted to 2D
datainMay2013.SiteAisintheGulfofMaineshallowwaterenvironment,
beam-timedatabysteeringthearrayin400azimuthaldirec-
andsiteBisinthedeepercontinentalslopeenvironment.
tions equally spaced from (cid:5)1 to 1 in sin b0, where b0 is the
azimuthalanglemeasuredfromarraybroadside.Anangleof
slope region and 1h of passive acoustic recordings in the sinb0¼(cid:5)1correspondstothebackendfiredirectionandsin
Gulf of Maine shallow waters. No active acoustic sound b0¼1correspondstotheforwardendfiredirection.Therela-
sourceswereusedinthetowedreceiverarrayseatest. tive azimuthal direction and time of arrival of each sperm
The maximum frequency of the acoustic data recorded whaleclickweredeterminedfromthelocalpeakenergylev-
bythereceiverarraysystemhereis2.5kHz.Ouranalysisof els of the 2D beam-time data. In general, the sperm whale
spermwhale clicksisthereforelimited tothelow-frequency clicks after high-pass filtering and beamforming stood
component of the click signals in the several hundred hertz between 10 and 35dB above the background. A detection
to a couple of kilohertz range that is more omnidirectional threshold of 10dB above the background was applied in the
(Mathias et al., 2013; Tiemann et al., 2006; Zimmer et al., localpeakdetectiontoreducethefalsealarmrate.
2004)andsufferlesstransmissionloss.Incontrast,thesam- Spectrogram and time-series examples of the beam-
pling frequencies of acoustic systems in previous sperm formed received click trains are shown in Figs. 3 and 4.
whale studies were significantly higher, by at least a factor Because each sperm whale click contains multiple sharp
of three (Mathias et al., 2013; Tiemann et al., 2006) to over pulseshighlylocalizedintimewithawidthoflessthan1ms
ten times that used in this study to provide a more complete per pulse (see Fig. 5), the click signal resembles the output
of a matched filter. This enables high resolution beamform-
ing in the time domain because coherent addition of the
pulses across all hydrophones decorrelates within a small
timelagofroughly1/8ms,correspondingtobearingestima-
tionaccuracies of approximately 1.7(cid:6) atarraybroadsideand
8(cid:6)neararrayendfire.Incontrast,thearrayangularresolution
is much broader, roughly k=ðLcosb0Þ¼3.7(cid:6) at broadside
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
and 2 0:886k=L¼22.7(cid:6) at endfire from planewave beam-
forming of a time-harmonic signal, where k, L, and b0 are,
respectively, the wavelength, array aperture and azimuthal
directionfromarraybroadside(JohnsonandDudgeon,1993;
Makrisetal.,1995)forthegivenarrayapertureL¼23.25m
at a frequency of 2 kHz. Examples of the beam pattern
obtainedfrombeamformingtwodistinctspermwhaleclicks,
one located near array broadside and the other near array
endfireareshowninFig.6.Thedirectionofarrivalisclearly
distinguishable since the main lobe stands more than 8dB
abovethegratinglobeinbothcases,mitigatinganypotential
effectofspatialaliasing.
Theestimatedrelativebearingsb^0 measuredwithrespect
to array broadside were then converted to absolute bearings
b^, measured from the array center with respect to true North
FIG.2.MeasuredsoundspeedprofilesatthetwotestsitesshowninFig.1. by correcting for the corresponding array heading
3354 J.Acoust.Soc.Am.,Vol.135,No.6,June2014 Tranetal.:Spermwhalerangelocalizationandclassification
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 18.51.1.88 On: Wed, 01 Jul 2015 18:28:21
FIG.3.(Coloronline)(A)Spectrogram
ofaseriesofspermwhaleecholocation
clicks recorded at frequencies up to
2.5 kHz inthe Gulfof Maine on May
14 starting at 17:19:07 EDT. The
spectrogram was calculated using a
short-time Fourier transform with win-
dow size 256 and 75% overlap. (B)
Beamformedpressuretimeseriesofthe
clicks, bandpass filtered between 1500
and2500Hz.(C)Beamformedpressure
timeseriesplottedindecibel(dB)scale.
The solid curve with error bars shows
the mean and standard deviation of
beamformedbackgroundambientnoise
level in the 1500–2500Hz band, esti-
matedfromregionsoutsideofclicks.
measurement a. To resolve the left-right ambiguity inherent of Gong et al. (2013). When the array is steered in the azi-
in linear receiver array measurement of a source bearing, the muthofspermwhaleclicks,thearraygain(Kay,1998;Urick,
left and right absolute bearing sequences were statistically 1983) obtained from coherent addition of the click signals
correlated to the measured array headings. The true bearing measuredacrossallN¼32hydrophonescanenhancethesig-
sequencewasselectedtobetheonewiththesmallercorrela- nal-to-noise ratio by approximately 10 log N (cid:3) 15dB over
10
tion coefficient because the ambiguous bearing sequence that of a single hydrophone [compare beamformed click sig-
closely followsthe array heading changes asshown inFig.2 nals in Fig. 4(C) with single hydrophone measured click
FIG.4.(Coloronline)(A)Spectrogram
of three consecutive slow clicks fol-
lowedbyatrainofecholocationclicks
recordedatfrequenciesupto2.5kHzin
theGulfofMaineonMay14startingat
17:16:15 EDT. (B) Beamformed pres-
suretimeseriesoftheclicks,bandpass
filtered between 500 and 2500Hz.
(C) Beamformed pressure time series
plottedindBscale.Thesolidcurvewith
errorbarsshowsthemeanandstandard
deviation of beamformed background
ambientnoiselevelinthe500–2500Hz
band,estimatedfromregionsoutsideof
clicks. (D) Corresponding signal
receivedonasinglehydrophone,band-
passed filtered between 500 and
2500Hz.
J.Acoust.Soc.Am.,Vol.135,No.6,June2014 Tranetal.:Spermwhalerangelocalizationandclassification 3355
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 18.51.1.88 On: Wed, 01 Jul 2015 18:28:21
movement between pairs of whale bearing measurements in
the MAT technique has to satisfy the near field condition,
A2=k(cid:7)r , where r is the whale range from the receiver
s w w
centerandkisthewavelengthoftheclicksignal.Tolocalize
and track sperm whales at ranges less than 5km from the
array, with k set to be 0.75m, the synthetic aperture length
shouldbeatleast60m.Thearraycanbetowedoverthisdis-
tance in half a minute, so that near real-time tracking of
sperm whales is feasible with this method. To track sperm
whalesatlongerrangeswiththeMATtechnique,longerob-
servationtimeswouldbenecessary.
The sperm whale inter-click interval is approximately
FIG.5.(Coloronline)Multiplereflectionarrivalpatternofthespermwhale
clicks detected on May 14 in the Gulf of Maine. The order of arrival is: 1sorlessinaclicktrainandthereceiverarrayheadingwas
Direct path; pairs of bottom and surface reflected, bottom-surface-bottom updated at roughly 12s intervals. The MAT technique was
and surface-bottom-surface reflected, etc. Between 17:26:00 and 17:32:30
applied to pairs of click bearing measurements that were at
EDT,reflectionsfromuptoseveninterfacebouncesaredetected.
least 12s apart to estimate each whale range. The whale
signalsinFig.4(D)].Thissignificantlyimprovesspermwhale range estimates obtained here are expected to have smaller
click detectability and ranging capability. Note that an inco- fractional errors than the MAT localization fractional errors
herentarrayofhydrophonesalsohasnoarraygainovernoise, reportedinGongetal.(2013).Thisisbecause,bythelawof
regardless of the number of hydrophones, because coherence large numbers (Bertsekas and Tsitsiklis, 2002; Goodman,
between sensors is necessary to accumulate array gain. 1985),thereareroughlysixtimesmorenumerousrangeesti-
Subsequentanalysesonthetemporalandspectralcharacteris- mates derived from bearing measurements embedded in
tics of the clicks were performed on the noise-suppressed noise which can be regarded as statistically uncorrelated
beamformedclicks. acrosstime forthe spermwhaleproblemcompared to Gong
etal.(2013)whereonlyonerangeestimate wasavailable at
2. Rangeestimationforspermwhalesinthenear-or every 75s interval for the source localization problem dis-
far-fieldofthetowedhorizontalcoherentreceiver cussedthere.
array
3. Simultaneousdepthandrangeestimationfora
Each sperm whale individual was localized and tracked
spermwhaleinshallowwater
from its corresponding sequence of click bearing measure-
mentsusingthemovingarraytriangulation(MAT)technique The range and depth of a sperm whale located approxi-
(Gong, 2012; Gong et al., 2013), which combines bearing mately 1km from the receiver array in the Gulf of Maine
measurements from spatially separated apertures of a towed were simultaneously estimated using MR-TDA of beam-
horizontalreceiverarrayandemploystheconventionaltrian- formed direct and singly or multiply reflected click signals
gulation ranging algorithm to localize a source in either the from sea bottom and surface. The concept for whale range
near- or far-field of the array. The whale range from the re- and depth inference is similar to that presented in Thode
ceiverarraycenterwascalculatedusingEqs.(1)through(3) et al. (2002). However, because the array depth was accu-
ofGongetal.(2013)givenapairofwhalebearingmeasure- ratelyknownfromdepthsensormeasurementsampledevery
ments. The synthetic aperture length A created by the array 10ms,therewasonefewerunknown.Asaresult,onlythree
s
arrivals: Direct path, bottom bounce, and surface bounce
were required to solve for whale range and depth as derived
and discussed in the appendix. When more than three arriv-
als were present, the localization result could be obtained
withhigheraccuracybyemployingallavailableinformation
(see the appendix). The whale range estimates obtained
using MR-TDA will be compared to those obtained with
bearings-onlyMATmethodinSec.IV.
B. InferringspermwhalesizefromIPIs
The first 10-15ms of a sperm whale click usually con-
sistsofmultiplepulsesafewmillisecondsapart(Backusand
Schevill, 1966; Møhl, 2001; Møhl et al., 2003; Norris and
Harvey, 1972), resulting from multiple reflection within the
FIG.6.Arraybeamformeroutputasafunctionofsteeringanglefromarray whale head according to the bent-horn hypothesis (Norris
broadside0(cid:6)shownfortwodistincttimeinstances.Spermwhaleclickswith and Harvey, 1972; Zimmer et al., 2003, 2004). The IPI has
relativebearing3.7(cid:6)neararraybroadsideand73.7(cid:6)neararrayendfire.The beenshowntocorrelatewiththespermacetilength(Gordon,
corresponding1-dBbeamwidthsareapproximately1.7(cid:6)nearbroadsideand
1990; Rhinelander and Dawson, 2004) and with the overall
8.0(cid:6) near endfire. Rectangular window was applied across the array
aperture. body size (Antunes et al., 2010; Growcott et al., 2011;
3356 J.Acoust.Soc.Am.,Vol.135,No.6,June2014 Tranetal.:Spermwhalerangelocalizationandclassification
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 18.51.1.88 On: Wed, 01 Jul 2015 18:28:21
Mathiasetal.,2009;Milleretal.,2013;Telonietal.,2007). following the approach of Andrews et al. (2009) and Gong
An automated moving local peak energy detector with a etal.(2010).
1msaveraging-timewindowwasappliedtothebeamformed
pressure-time series data to determine the arrival time of
III. RESULTSI:ASPERMWHALEINSHALLOW
each pulse within a spermwhaleclick signal to estimate the
WATERSOFTHEGULFOFMAINE
IPI. The result for each click was plotted and visually
inspected for accuracy. This analysis was applied to high- A. Analysisofvocalizations
pass filtered beamformed pressure-time series data to sup-
VocalizationsfromaspermwhaleindividualatsiteAin
press ambient noise and improve estimates of the IPI. Many
theshallow-watersoftheGulfofMainewererecordedusing
click signals from each whale were examined, and only
the towed receiver array for over an hour from 16:35:00 to
thosewithaclearmulti-pulsestructure wereincludedinour
17:40:00 EDT on May 14. An example of the beamformed
analysis. IPI estimates were averaged over multiple clicks
time series andspectrogram of anecholocationclick train is
(roughly20–60perwhale)toreducetheerrorintheIPIesti-
shown in Fig. 3. These measurements were made in water-
mates.AscanbenotedfromTableII,thestandarddeviation
column depths of roughly 160m where the receiver array
in the IPI estimates for each sperm whale is comparatively
was located at roughly 65m depth. The average inter-click
small, less than 10%. The whale body length, L , was then
w interval was approximately 0.7s; this implied that these
estimated here using the methods proposed by Gordon
clickscouldbecategorizedas“usualclicks”(Whiteheadand
(1990)andGrowcottetal.(2011)
Weilgart, 1990). Spectrogram analysis indicates that the
clicks contain significant energy at low frequencies in the
L ¼4:833þ1453(cid:8)IPI(cid:5)0:001(cid:8)IPI2; (1)
w;Gordon 1.5–2.5kHz range. This enabled the beamformed, high-pass
L ¼1:257(cid:8)IPIþ5:736: (2) filtered, time-series data of the clicks to stand between 15
w;Growcott
and 30dB above the mean background ambient noise level
The sampling frequency of the towed array hydrophones [Fig.3(C)].Clickrateswerefoundtovarywithinanecholo-
limited our analysis of sperm whale click signal to cation click train, as shown in Fig. 3, where the inter-click
frequencies(cid:2)2.5 kHz. In this low-frequency regime, the intervals and click amplitudes decrease slowly over a time
sperm whale multipulsed-click signals are approximately interval of about 20s. This implied that the sperm whale
omnidirectional. vocalizations could be transitioning from clicks to creaks,
which area sequence of lowenergy clicks closelyspaced in
time emitted when homing in on a prey (Madsen et al.,
C. Estimatingspermwhalemaximumdetectionrange
2002b;Milleretal.,2004).
withthelow-frequencytowedcoherentreceiverarray
system Besides these usual echolocation clicks, we also
recorded intense broadband clicks with dominant energy
The maximum detection range of a sperm whale with
contained in the 0.5-2kHz frequency range. These loud
our towed array system was determined as the range at
clicks were separated by intervals of 5s or longer (see Fig.
which the transmission loss correction led to a received
4)andcanbecategorizedasslowclicks(BarlowandTaylor,
spermwhaleclicksignallevelthatstoodtwostandarddevia-
2005; Jacquet et al., 2001; Oliveira et al., 2013; Weilgart
tions above the mean beamformed background noise level
and Whitehead, 1993). The spectra of both the echolocation
band-pass filtered between 300Hz and 2.5kHz. The two
andslowclicksrecordedherecloselymatchthoseofDTAG-
standard deviation signal excess enabled positive detection
recordedspermwhaleclickvocalizationsshowninFig.1of
ofspermwhaleswithover95%confidence.
Oliveiraetal.(2013).
The broadband transmission loss from the sperm whale
Theoccurrencesoftherecordedvocalizationsareshown
locationtothereceiverarraywascalculatedat10Hzinterval
againsttimeinFig.7withclicktypesindicated.Eachecholo-
in the bandwidth of the received clicks using the parabolic
cationclicktrainlastedroughly2minwithperiodsofsilence
equation-based range-dependent ocean waveguide-acoustic
varying between 20s and several minutes. Longer periods of
propagation model RAM (Collins, 1994). To simulate the
silence lasting between 5 and 10min observed here may be
effectofinternalwavesthatrandomizedtheacousticpropaga-
associatedwiththespermwhale’sascentinthewater-column
tion path, water-column sound speed profiles obtained from
(Zimmeretal.,2004)andrestingnearthesurface.
CTD and XBT measurements during the experiment were
By analyzing each received click in the time interval
randomly updated every 500m range, the approximate hori-
from 17:18:00 to 17:33:00 EDT, we estimated the mean IPI
zontalcorrelationlengthforlinearinternalwaves.Thebottom
for this sperm whale individual to be 3.0ms with a standard
wasassumedtobesandywithsoundspeed1800m/s,density
deviation of 0.3ms. This corresponds to a sperm whale
1800kg/m3 and attenuation 0.8dB/k. The water-column
lengthofapproximately9.3m,accordingtoEq.(2).
attenuation was set to 6(cid:8)10(cid:5)5dB/k. These waveguide pa-
rameters were previously found to provide the best-fit match
B. Trackingrangeanddepthofaspermwhaleclose
betweenmeasuredandmodeledtransmissionlossintheGulf
toarraynear-field
of Maine (Andrews et al., 2009; Gong et al., 2010). The
broadbandtransmissionlossisobtainedbyincoherentlyaver- Theestimatedbearingsofthevocalizationsobtainedvia
agingthereceivedintensityfrom50Monte-Carlorealizations time-domain beamforming are plotted in Fig. 7. The instan-
overthebandwidthofaspermwhaleclickfrom1to2.5kHz, taneous sperm whale ranges were estimated from the
J.Acoust.Soc.Am.,Vol.135,No.6,June2014 Tranetal.:Spermwhalerangelocalizationandclassification 3357
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 18.51.1.88 On: Wed, 01 Jul 2015 18:28:21
EDT to 17:33:00 EDT. All estimated delay times were used
for each recorded click, giving a maximum of 7C ¼21
2
range-depth estimates when seven reflections were detected,
and a minimum of 4C ¼6 range-depth estimates when four
2
reflections were detected when employing the MR-TDA
method. When ranges were available from the MAT tech-
nique, between four and seven depth-estimates could be
obtainedforeachclick,dependingonthenumberofavailable
reflections. The results are plotted in Fig. 8 comparing the
MATbasedresultswiththeMR-TDAbasedresults.
Theanalysisindicatesthatasthereceiverarraywasbeing
towedintheNorthbounddirection,thespermwhaleinitially
locatedroughly0.5kmawayfromthearrayat17:08:00EDT
FIG. 7.Measured bearings of sperm whale echolocation and slow clicks moved away to a range of about 1.5km in roughly 25min.
detectedintheGulfofMaineonMay14overa1-hperiodfrom16:35to Thewhaleappearedtohoverwithinthewatercolumnwithout
17:35EDT. surfacingtobreathewithestimateddepthvaryingbetween70
and 100m over this time period. The increasing whale-
measuredwhalebearingsusingthebearings-onlyMATtech-
receiverseparationestimatedusingtheMR-TDAexplainsthe
nique for the time interval from 17:08:00 to 17:33:00 EDT
more compact arrival structure in Fig. 5 because when range
andplottedinFig.8(A).
increases, the reflection arrivals get closer to the direct path,
In the shallow water Gulf of Maine environment, clicks
asdemonstratedinFig.12.Thespermwhalerangeanddepth
arriving at the receiver array from the sperm whale in close
estimatesobtainedviatheMATandMR-TDAmethodsagree
rangetothearraynear-fielddistancedisplayclearpatternsof
well, especially near array broadside (between 17:23:00 and
multiple reflection between the sea bottom and surface. The
17:28:00). At other time instances, the two methods yield
evolutionofmultipatharrivaltimesisshownforalltheclicks
results that are within 1 SD of each other. The whale range
recordedbetween17:18:00and17:33:00EDTinFig.5.The
estimatesobtainedviaMATareexpectedtobemorereliable
direct arrival time of each click was first determined and
than those obtained using MR-TDA. This is because in the
alignedintimebeforethestackedsequenceofclickswascre-
MR-TDA technique, the whale range estimates depend on
ated. Between 17:18:00 and 17:23:00 EDT, the first-order
otherparameterssuchastheunknownwhaledepthandwater
bottom and surface reflected arrivals are distinguishable in
columndepth.TherangeestimatesobtainedviaMATarenot
time, crossing each other due to vertical displacement of the
dependentontheseparameters,insteaditdependsonlyonthe
whale.Atothertimeinstances,thebottomandsurfacereflec-
measuredbearingchangeofthewhalewithtime.
tions are not clearly distinguishable. The higher-order
reflectedarrivalsaremoreprominentwithshortertime sepa-
C. Spermwhaledetectionrangeinshallowwater
ration for the later clicks in the time period analyzed.
Multiple reflections in the shallow water waveguide extends The click signals from this sperm whale located at
the received sperm whale click time duration from 10-30ms roughly 1km in range from the receiver array stood by as
ofthedirectarrival(Møhl,2001)tomorethan250ms. much as 35dB over the mean background ambient noise
The sperm whale range and depth were also simultane- level after beamforming [see Figs. 3(C) and 4(C)]. Because
ously estimated by applying the MR-TDA technique thebackgroundambientnoiselevelinthebeamofthesperm
describedintheappendixforthetimeintervalfrom17:18:00 whale after high pass filtering has roughly 5.5dB standard
FIG. 8.(Color online) (A) Single sperm whale in Gulf of Maine localization result using the two methods, MAT and MR-TDA for the period between
17:15:20and17:32:40EDT.TheellipsesrepresentcontoursoflocalizationuncertaintyateachtimeinstancewithMAT(solidcurve)andwithMR-TDA
(dashedcurve).Theoriginofthecoordinatesystemislocatedat(41(cid:6)48.780N,69(cid:6)0.060W).(B)RangeestimatesusingMATandMR-TDAbetween17:18:00
and17:32:40EDT.Theerrorbarsshowthestandarddeviationoftherangeestimatesina4-mintimewindow.(C)Depthestimatesforthesametimeperiod.
3358 J.Acoust.Soc.Am.,Vol.135,No.6,June2014 Tranetal.:Spermwhalerangelocalizationandclassification
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 18.51.1.88 On: Wed, 01 Jul 2015 18:28:21
TABLE I.Correlation coefficients between receiver array heading change
andclickbearingchangeforthecandidatepairsofleft-rightabsoluteclick
bearingclustersshowninFig.10.Theselectedtrueclickbearingclusters
areeachmarkedwithanasterisk.
Correlationcoefficients,q
Cluster Right Left N
1 0.64 0.32* 17
2 0.66 0.36* 25
3 0.63 0.09* 10
4 0.09* 0.60 16
5 0.44 0.08* 41
6 0.64 0.34* 9
7 0.63 0.26* 17
FIG. 9.Average broadband transmission loss in the frequency range of
8 0.37* 0.76 229
1.5–2.5kHzobtainedusingtheRAMmodelatdistancesupto80kmforthe
9 0.06* 0.27 42
GulfofMaineenvironment.
deviation,thespermwhaleclicksignalshaveasignalexcess
of (35-2(cid:8)5.5)¼24dB for our detector. The broadband
continental slope south of Cape Cod. Both left-right bearing
transmission loss from a whale located approximately 1km
estimates of the detected clicks are shown for roughly an
in range from the receiver array center and at a depth of
hour of recording from 23:00:00 EDT on May 13 to
80m is roughly 55dB re 1m [see Fig. 9(A)]. The signal
00:02:30 EDT on May 14 in Fig. 10. The left-right ambigu-
excess of 24dB implies that the sperm whale vocalizations
ity of the linear receiver array is resolved for each group of
can be detectable out to much longer ranges>60km, where
clicks using the technique described in Sec. IIA1. The cor-
thetransmissionlossis(55þ24)¼79dBre1m,afterbeam-
relation coefficients between the change in bearings and the
forming with the towed receiver array. In contrast, sperm
changeinarrayheadingsfornineidentifiedclustersofclicks
whale detection range with a single omnidirectional hydro-
inFig.10arelistedinTableIforboththeleftandrightbear-
phone is expected to be significantly limited [compare Fig.
ingcandidates.Aseriesofclickbearingsisdeterminedtobe
4(C) showing the result after array beamforming with Fig.
trueif(a)the correlation coefficient qisbelow 0.4,whereas
4(D) which is the result for a single hydrophone]. Because
theambiguousgrouphasq>0.6,or(b)thecorrelationcoef-
sperm whale detection range is dependent on ambient noise
ficientqisatleastfivetimessmallerthanthatoftheambigu-
level, the detection ranges given here are valid for low to
ous group. When none of these criteria are met, the true
moderate sea state of around two to three and wind speed
click bearings could not be determined and no localization
between 8 and 11kn, according to recorded measurements.
resultisobtained.
The detection ranges will be larger at lower sea states and
smallerathigherseastates.
A. Distinguishingspermwhaleindividualsusing
temporal,spectral,andspatialcharacteristicsofclicks
IV. RESULTSII:MULTIPLESPERMWHALESATTHE Wefurtherassociatethedifferentclickclusterstodistinct
CONTINENTALSLOPESOUTHOFCAPECOD sperm whale individuals based on the mean and standard
We identified over 1000 sperm whale clicks in about deviationofIPIofeachclickcluster.Themeanandstandard
75min of recording on May 13 at site B (Fig. 1) on the deviation of the IPI as wellas estimateofwhale bodylength
based on Eqs. (1) and (2) are provided in Table II for all
TABLEII.IPIforeachclusterandtheestimatedspermwhalebodylength.
Cluster hIPIi r L L Whale
IPI w,Gordon w,Growcott
number (ms) (ms) (m) (m) group N
1 3.4 0.4 9.8 10.0 A 10
2 7.7 0.7 16.0 15.4 B 8
3 3.3 0.3 9.6 9.9 A 8
4 4.2 0.3 11.0 11.0 C 11
5 4.3 0.2 11.0 11.1 C 9
6 7.6 0.7 15.8 15.3 B 5
7 4.9 0.3 11.9 12.0 D 12
8-1 4.7 0.3 11.6 11.6 E 43
8-2 4.6 0.3 11.5 11.5 E 12
8-3 4.5 0.4 11.4 11.4 E 18
FIG.10.(Coloronline)Trueandambiguousbearingsofspermwhaleclicks
receivedonthearrayonMay13atsiteBonthecontinentalslope.Whale 8-4 4.5 0.4 11.4 11.4 E 33
bearingsaregroupedintoclustersfrom1to9accordingtotheirtemporal, 9 6.9 0.5 14.8 14.4 F 17
spectral,andspatialcharacteristics.
J.Acoust.Soc.Am.,Vol.135,No.6,June2014 Tranetal.:Spermwhalerangelocalizationandclassification 3359
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 18.51.1.88 On: Wed, 01 Jul 2015 18:28:21
FIG. 11.(Color online) Localization
andtrackingresultformultiplesperm
whaleswithbearingclustersshownin
Fig.10onMay13atsiteBonthecon-
tinentalslope.Thedashedcurveisthe
trackofthereceiverarray.Theorigin
of the coordinate system is located at
(39(cid:6)50.940N,70(cid:6)19.550W).
identifiedclickclusters.BasedonthemeasuredIPI,weasso- either CorD.From TableII,the estimatednumberofsperm
ciate click clusters 1 and 3 (mean IPI (cid:3) 3.4ms) to the same whale individuals simultaneously and passively recorded by
whaleA.Similarly,clickclusters2and6areassociatedwith thetowedreceiverarrayisatleast4orasmanyas6.
whale B with much longer IPI of approximately 7.6ms.
WhaleCisassociatedwithclickclusters4and5.Clickclus-
ter 9 has very distinct IPI and corresponding whale size esti- B. Localizationandtrackingofmultiplespermwhales
infar-fieldofthetowedhorizontalcoherentreceiver
mate and is associated with whale F. Whale D is associated
array
withclick cluster 7,while whale E isassumed topossess the
IPI measurement of click cluster 8, which is composed of The sperm whale click clusters from May 13 are local-
sub-clusters 8-1, 8-2, 8-3 and 8-4. The IPI values measured ized using the bearings-only MAT technique and the results
for clusters associated with whale E lie between those of are shown in Fig. 11 grouped according to the classification
whale C and whale D, so whale E could also possibly be and association results obtained in Sec. IVA. Whale A
FIG. 12.Calculated time delays from
direct arrival for the bottom bounce
Dt (upper left), surface bounce Dt
b s
(upper right), bottom-surface bounces
Dt (lower left), and surface-bottom
bs
bounces Dt (lower right) as a func-
sb
tionofwhalerangeanddepth.There-
ceiver depth is set at 65m and the
waterdepthis160m.
3360 J.Acoust.Soc.Am.,Vol.135,No.6,June2014 Tranetal.:Spermwhalerangelocalizationandclassification
Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 18.51.1.88 On: Wed, 01 Jul 2015 18:28:21