MARINE MAMMAL SCIENCE, 16(2):437-447 (Apri l 2000) © 2000 by che Sociery for Manne Mammalogy

A TWO-DIMENSIONAL ACOUSTIC LOCALIZATION SYSTEM FOR MARINE MAMMALS

For the study of animal communication systems, ic is crucial co b~ able co identify che sender of a given call. T his nocion might seem crivial co inves­tigators of many cerrescrial animals. However, there are situations in which simple visual observation cannot be used even in cerrescrial environments (e.g., observations at night or in dense vegetation). In aquatic environments sender detection is almost always a problem. As a result, many srudjes on marine mammals have only been able to link calls co che behavior of a whole g roup t:lCher chan particular individuals (Evans and Dreher 1962, Taruski 1979, Sjare and Smith 1986, Ford 1989. Weilgarr and Whitehead 1990, Smolker er al. L 993).

A few scudies on wild marine mammals have been able co overcome the problem of sound source detection by using passive acoustic localization (Wat­kins and Scbevill 1974, Clark et al. 1986, M0hl et al. 1990, Frankel et al. 1995 ). This method uses differences 10 rimes of arrival of a signal ac different transducers of a microphone array to determine the source posicion. Ics main advantages are char it is completely non-invasive and that ic can be used ro monitor vocal behavior of several individuals at once. Alternatives, such as temporarily capturing individuals (e.g. , Sayigh et al. 1990) or arcaching telem­etry devices (e.g .. Tyack 1986), are difficult ro apply in the wild and can affect rhe behavior of the animals. However, the accuracy of passive localization relies heavily on rhe acoustic topography of the study sire (Spiesberger and Frisrrup 1990) and data analysis can be very rime consuming. This is one reason why ir bas nor been used more widely in behavioral studies on marine animals. The acceptable error in such studies, however, varies with the quesrion under investigation. The study of close-coocact interactions berweeo individuals, for example, requires more accurate sound source localization chan thar of long­distance communication.

In this study we present the localization accuracy of a three-element hy­drophone array that was used to study different aspects of vocal behavior in bottlenose dolphins (Tursiops truncatus) and harbor seals (Phoca vitulina). The whole localization system, including the analysis software, was based upon readily available components and can be easily installed in any shallow-water

437

438

•• a)

i ><

f r .::

0 0 • 3

•• i " f ~ r .::

0 0 ..3

•• . i I; c

f .::

0 0 .3

C) 600

500 11-JI

400

I 300 II) u c "' ~ 200 0

100

0

-100 -500

MARINE MAMMt\L SCIE NCE. VOL

.. .. t.2 n.. : ..c

.. .. t.:Z n.. :-

- -~- ~-

.. .. 1.2 n.. :-

-400 -300 -200

~

b)

~ ~ 0 u

0

, os _, _ _,

. t5

"' "' 0 u 0

•• to ·1.S

~

<r

"' 8

0

, os - t .. .S

-100

Distance (m)

I h. NO. ~. 2000

~-JVL- t · 2

-t .O . s 0 t.O t.S Tlm• : ....c

-~t- t · l

-1.0 •. S 0 1.0 1. 5 flmoe : ...c

lL_ 2·3

- 1.0 • S • .s 1.0 t .S TlnM : oec

A 3

0 100 200 300

Figurt 1. Illusrrarion of steps in localizarion procedure. (a) Three spectrograms of same signals received ac three differenc hydrophones (Number l , 2 and 3). Noce chat same signal scans ac slightly different times in each spectrogram. (b) Cross-correlation functions for first whistle of spectrograms shown in (a). Numbers on right of each function indicate which spectrograms have been cross-correlated . Peak of each corre­lation function corresponds co time difference in time of arrival of first whistle in (a) ac compared hydrophones. In this figure several minor peaks can also be found because chere are three signals in spectrograms. (c) Plor of array geometry with resulting hyper-

NOTES

area. We will discuss issues related co c:he choice of srudy sire, sec-up of che equipmenc. analysis pcoceJures, anJ what questions can be addressed with a given localization accuracy.

The hydrophone array was deployed i.n a 500-m-wide natural channel char forms the encrance ro che Beauly Firth in northeast Scotland, U.K. (57°30'N, 4°l-1' W ). Bottlenose dolphins and harbor seals frequently use this area, es­pecially during rhe summer months when they can be observed in the channel for several hours each day. Water depth increases from 6 tO 38 m from ease co west in the middle of the channel. The maximum tidal range was 5 m. The sea floor consists of mud and pebbles. Three hydrophones (High Tech Inc. SSQ94) were installed in a triangle co form a two-dimensional array (Fig. lc). Two hydrophones (numbers l and 3) were p laced on che north shore of che channel and one (number 2) on the south shore. !mer-hydrophone disr.ances were 208. 5l3, and 560 m. The hydrophone locations were determined by using rwo Magellan GPS ProMark X units wich Magellan Mscar software (Version 1.05) to p rovide differential global positioning. The accuracy of chis sys rem was ±3 m. Each hydrophone was mounted on a 30-cm flexible pole which was arrached co a 33-cm diameter trailer wheel fined wir.h lead weights (coral weight 15 kg). Each hydrophone was about 50 m from the shore ac a depth of 1-5 m depending on che scare of the ride. A 150-m cable and leaded rope connected each hydrophone co a shore sracion, which contained a Micron TX-l Ol VHF radio transmitter (transmitting power 25 mW) and power sup­plies for the rransmirrer and hydrophone. Each star.ion could be run off cwo PP3 9V batteries, although here we bad external power available. Each trans­miner was connected co a simple whip aerial. The receiving and recording station consisted of a Yagi directional ae.rial, 3 Yaesu FRG-9600 receivers and a Fosrex 3805 or a Tascam Porta Two multitrack cape recorder. The frequency response of the whole sysrem was 50 Hz co 18 kHz ± 3 dB wir.h rhe Fostex recorder (used for dolphin recordings) and 40 Hz co 12.5 kHz ± 3 dB with the Tascam recorder (used for seal recordings). Micron makes receivers char can be used co replace the Yaesu receivers char we used here . We recommend char normal VHF/UHF receivers should nor be used unless cheir frequency response has been measured as we did wirh che Yaesu receivers. The equipment was installed in an observation booth 30 m above sea level on the northern shore of che channel. The ma.ximurn disrance between transmitter and receiv­ing station was approximately 700 m in line of sight. The radio cransmitters we used could transmit over such distances , bur there were no obstacles in the cransmission path and we used a directional aerial for reception. All re­cordings were made in sea scare of Beaufort 1 or 2 and wirhour rain.

~

bolas from correlation peaks in (b) Triangles with numbers indicate posicion of hydro­phones. Numbers correspond co chose in (a). Boxed pairs of numbers on each hyperbola indicate which peak it corresponds co in (b). Poinc of intersection of hyperbolas indi­cares posicion of sound source for first whisde in (a).

440 MARINE Mt\MMAL SCIENCE. VOL. II>. NO. ~ - .!000

All clara were analyzed wich che SIGNAL sofcware loc.:alizacion moJule (En­gineering Design, Belmont, USA) (Beeman L 996). The localization of a sounJ source is based on che rime difference wich which a signal arrives ac each pair of hydrophones. These delays are preserved on che simultaneous multitrack recording of input from all hydrophones (Fig . 1a). If che speed of sound in che medium and the transducer positions are known, rhe difference in rime of arrival of rhe same sound ac a given pair of hydrophones holds informacion on che possible sound source locacions. The mean speed of sound ( :::!::SD) was determined as 1,488 ± 3 m/sec (n = 24) by measurement wich an artificial source over known discances between c:he hydrophones. Each rime delay cor­responds co a specific hyperbola of possible source locacions. The SIGNAL software calculac:es che form of chis hyperbola for each rime delay using che foUowing formula (Beeman)1:

x = t . cJ-41 + d2 y-- t~ •(2.

where

(x. y) = che locus coord inates in cwo dimensions ac each paine of che hyperbola in m

d = che separation of c:he hydrophone pair in m

c = the speed of sound in m/sec

t = che c:ime delay becween che hydrophones in sec.

In the SIGNAL software the x-axis is defined as exc:ending berween che hy­drophones. and che y-a.xis bisects them. Thus, c:he cwo hydrophones are locaced ac ( -d/2, 0) and ( + d/2, 0 ). The equation gives che x coordinate for a given y. T he software calculates x values fo r all possible y values.

If planed, the hyperbolas of all chree pairs of hydrophones (l -2, 1-3 , 2-3) have a common paine of inrersecrion chac: represents che acc:ual posicion of the caUing individual (Fig . lc). Occasionally cbese hyperbolas form a triangle rath­er chan one point of intersection. In such cases che sound source is somewhere in che area within this triangle. It is im porram co note che coordinates of all three vertices of che criangle as the localization error. To determine the time delay berween cwo hydrophones a cross-correlation of eicher frequency spec­c:rograms (Fig. lb) or original waveforms can be used. Sounds were digic:ized ar a sampling race of 50 kHz. Spectrograms were calculac:ed with a 75% overlap becween different FFT's (Fast Fourier Transforms, de: 10 msec, df: 98 Hz, FFT size: 512) resulting in an effective rime resolution of 2.5 msec. We primarily used spectrogram cross-correlation in this srudy. Even c:hough the cross-correlation of frequency spectra instead of waveforms results in a slighdy larger error due to the averaging over small time slices of che original signal

1 Personal communic3cion (e-mail) from Kim Beeman. Engineering Design. 43 Newton Street, Belmont, MassachusettS 02178, U.S.A., 1 October 1996.

NOTE.S

700

600

500

ln Zl ~ ~ '>Cl

:fl ~ )<0 XI l!l

)0

l!l :Cl )Cl lD

g 400 Cl

Xl I!J l!l

u c !! " 300 i5

9J l!l

es :fl

200 IJXI l!!l Xo

!:!

Q ~ .. 1!1

100

0

-600 -400 -200 0 200 Dlscance (ml

400 600 800

F ig11re 2. Overhead view of srudy area wi rh plor of localizarion error. Triangles indicare posicion of hydrophones. Squares indicare posicion of sound source as deter­mined bv cbeodolice. Crosses show resulrs of acouscic localizacion. Median error of sound lo~alizarion was 6.6 m. Grey areas are norrh and south shores. Poinr (0 ,0) was locarion of observation placform.

in che calcularion of rhe specu ogram. we found rhar rhe cross-correlarion of frequency spectra performed well and was sufficienr fo r our srudy. To ensure rhar che cross-correlarion used our rarger signal and nor some fearure in che background noise, we cross-checked all cross-correlar ion resulrs by also mea­suring rime delays manually with an on-screen cursor.

To measure localizarion accuracy we used an arrificial sound source (rwo pieces of meral banged mgerher) from an inflarable boar in one-meter deprh ar various locarions in rhe channel. Only one signal was used ar each locarion. . "-' . since rhe currenrs were roo srrong ro keep rhe boar stat ionary. The rest signal bad a sudden onset and was 50 msec long with a peak freq uency ar 6 kHz and a bandwidrh of approximarely 12 kHz. No marine mammals were presenr during rhese experirnems. To rest che accuracy of rhe acousric localization sysrem, we compared irs results with locations obrained by using a Sokkisha DTG digital theodolite (accuracy 10 sec).

This resr showed char the absolute localization error ranged from 2.5 ro 20.4 m (Fig. 2). The P' ro Yd quan:iles of che error distribution were 4 m, 6.6 m, and 9.2 m, respectively. The largest absolute error inside rhe rriangle formed by the array was 13 m. Figure 3 shows rhe errors in bearing and range at rhe intersection of the lines resulring from a rime delay of 0 sec between hydrophones 1 and 2 and hydrophones 2 and 3 (see Fig. l c). This reference point lies near rhe center of rhe triangle formed by rhe array. lr shows rhar the range error increases slightly with distance from the array. The bearing error appears to be greatest cowards the shore line (north and south directions), bur this is mainly because the locations rested in chose directions were closer

442 MARI1 E MAMMAL SCIENCE. VOL. 1(1. NO 2. 2oon

a)

6 -1/J Q) 4 Q) ~

C'l Q) 2 "C

~

0 0 ~ ~

w C'l -2 c:

• • • •• • • • • • • -------------~-.-... ~ .. ----------------------~.-~------------.~ • • • • •

·;:: -4 "' Q) •

IIl -6 •

0 90 180 270 360

Bearing (degrees)

b)

20 ... • -E 10 • • -~

0 ~ .... 0 w Q)

C'l c:

"' -10 a:: • • •

• • • .. .. . . --------------------------~-------------------------------------------. ~. . ·.. .

•

• •• •

• • -20

0 200 400 600 Range (m)

Figure 3. Bearing (a) and range (b) errors in relation co bearing and range from iocersection of lines resulting from a rime delay of 0 sec berween hydrophones l and 2 and hydrophones 2 and 3. On Figure 2 this poioc lies ar (102, 394). In (a) 0° and 360° refer co norrh and 90° co ease. Noce char rbese plocs would look differeoc if another reference point was chosen.

to the reference poinr than the ones ro rhe east and west. Borh the bearing and the range plot would look differently if other reference poincs were chosen.

One advantage of long distances between hydrophones is char errors of a few meters in the exact position of the hydrophones do nor affect the locali­zation accuracy as much as in arrays with short inrer-hydrophone distances. However, a disadvantage is that some sounds may nor be received on all

:-.JOTES 4-13

hydrophones. so chat Jocaliz:uion is noc always possible. We rc::corde::d fo r 4 h, LS min. J.nd -U sec over 7 J while bocrlenost: dolphins (average number of antmals presenc = 10) were presenc in che channel. In char rime we recorded l.-1 9 dolphin whisrles, of which 728 (42%) could only be recorded on one or cwo hydrophones. Thus, 58% of all wh tscles have been localized. In 6 h of recordings wich harbor seals presenc (average number of animals = 2) we recorded 875 calls. Of rhese 513 (59%) were received on all hydrophones and have been localized.

The array design presented here is relatively simple in comparison co more elaborare approaches char have been used when higher accuracy has been need­c::d (t:.g . . Freitag and Tyack 1993). Nevertheless, chis sim pler sysrem can pro­vide valuab le informacion. We have used ic co study communication in bor­rlenose dolphins <Janik 1998) and harbor seals (van Parijs 1998). In these studies we invescigaced vocal interactions of individuals char were generally separated by 100 m or more, as determined by visual observation and sound source locanons. An erro r of up co L 3 m, such as we have found inside our array. JS nor a problem in such a srudy. Clark er a/. ( 1986) used che same merhod wich a linear chree-elemenc array co monitor bowhead whale migra­non.

Ont: advantage of a simple array such as che one described here is rhe relative ease wtch whKh ir can be used in rhe field. All componems can be run off barrery power wichour che need co change bacceries for about 12 h. Trans­m i rcers and receivers are small and che whole syscem cosr less chan £5,000 ( -liSS8.00Q). However. several poincs have co be considered when placing che hydrophones. One muse be certain char there are no obstacles in che sound pach between che source and che hydrophones. Such obstacles include large rocks or shallow banks, if the sound source is in deeper water. Scrong currents can also alcer rhe sound path or reflecr some of che sound energy (for a review of how ropography can influence localizarion performance see Spiesberger and Frisrrup 1990). Another important issue is rhe spacing of the hydrophones. A larger array allows one ro localize disranc sound sources more accurately (Wac­ktns and Schevill 1972). A larger incer-hydrophone distance also increases che cime delay wich which che same sound arrives at cwo different hydrophones . Errors in the measurement of the exact posicion of each hydrophone have a smaller effect on the accuracy of localization if delays are long. Furthermore, a large array chat can be installed on boch shores of a channel allows rhe monitoring of calls inside the array where rhe locaLization error is smallest. However, for a localization co be achieved eacb call must be received on all three hydrophones. Our results showed char 58% of all dolphin whistles and 59% of all seal calls could be received on all rhree hydrophones. This was sufficient fo r our studies, bur could be coo few for ochers. Thus, th ere is a rrade-off between che size of the array and call detection on all hydrophones char muse be considered.

It is also very important co calibrate the system when it is used in different study sires, as there are many variables that can have an impact on the accuracy of passive acoust ic localizacion. Usually che ropography of the field sire is not

444 MARINE MAMMAL SCIENCE. VOL. 16 . NO ~. 2000

known in detail, so char calibration becomes even more important. Ocher sources of error are the time resolution of rhe digitized signal or of a frequency spectrogram char is used in cross-correlation , uncertainties in rhe exact posicion of hydrophones or in the speed of sound ar rhe field si re, and rhe use of a two-dimensional array in a three-dimensional environment. If rhe difference in rhe depth of rhe receiver and rhe source increases, rhe localization error will also increase. The SIGNAL software and rhe multitrack recorders char we used here can also make use of daca from a fourth hydrophone. Four hydrophones provide si.x hyperbolas of possible source locations. This adds some redundancy co the measurement which is useful co detect faulry hyperbolas. However, it also requires more processing rime, since six rather chan three pairs of hydro­phones have co be compared. Finally, rhe signal srrucrure can also have an influence on local izarion accuracy. We chose a calibration signal char was sim­ilar co a dolphin bang as found in our srudy a.rea. According co what sounds will be recorded, ir is best ro choose a signal char is as similar as possible co rhe animal sounds of interest or, if possible, use a playback of chose sounds.

For the analysis ir is important co choose the most appropriate method co determine che rime delays between different hydrophones. One way of mea­suring this delay is by cross-correlation of either rhe waveforms or che spec­trograms of rhe rwo signals. The peak of the cross-correlation function is chen used as che time delay between rhe channels. The cross-correlation of rhe original waveform of the signal gives the best rime resolution and, therefore, almost always ytelds rhe mosc accurace localization. Especially wirh shorr ( < 100 msec) rarget signals like clicks, waveform cross-correlation performs much beccer chan spectrogram cross-correlacion. However, several problems can preclude rime delay measurements by cross-correlarion. In shallow water each signal arrives ac rhe same hydrophone several rimes in very quick suc­cession because of reflections off rhe surface and the sea floor. These so-called mulriparh problems can prevenr rhe cross-correlation from providing an ac­curate measurement. In such conditions time delays can srill be measured on specrrograms for some signals. On a spectrogram of a narrow-band, frequency­modulated signal like a dolphin whistle, fo r example, rhe leading edge of rhe signal can be seen even if che Line chat represents rhe frequency contour of the whistle is relatively wide due co mulriparh problems. This leading edge can be used co measure rime delays, since ir represencs rhe first arrival of the signal , i.e .. rhe one part of rhe signal char cook che direct path from source co receiver wichour any reflections.

If cross-correlation is used one should use the whole signal, bur when mak­ing leading-edge measuremencs it is important not co use che start or end of a whistle. Rather one should use a poinr at which a certain frequency is reached, for example 8 kHz in an upsweep whistle char ranges from 4 co 12 kHz. Since signal production in animals includes fading in and our, hydro­phones char are farther away from the sender will pick up less of che start and end of the signal because the quieter pares will be attenuated quickly over distance. Therefore, rhe same signal will appear slightly shorter ar the more disranc hydrophone. Since attenuation is not linear and depends on rhe specific

NOTES

environment, rhe e."<acr rime delay between cwo channels cannot be determined by looking ac the start: and end points alone if the actual signals that have been received have different durations. This also means char exact: localizarioo of non-modulated, pure cone signals is nor possible wirh chis mechod or by using cross-correlarion unless the signal shows a significant amplirude mod­ulation or the exacc scrucrure of rhe source signal is known. Directional hy­drophones or arrays in which the inter-hydrophone distance is smaller than che wavelength of the signal (Clark 1980, Miller and Tyack 1998) can be used co determine the direction co such sounds. However, it is nor possible to determine the distance between source and receiver unless one can use cwo such syscems for triangulation.

Another problem is a low signal-co-noise (SIN ) racio in which rhe carger signal is nor dominant enough co give the peak value in che cross-correlation function. Common background noise char can obscure rime delays are boac engines, ocher marine animals, or ocher signals by rhe same animal. Dolphins, for example, can produce clicks and whistles simultaneously. To ensure char rhe computer program is using the right carger call, one should cross-correlate only in the frequency band of interest. Just as ir is crucial co rest che accuracy of each array that is used in rbe field, one has to monitor rhe performance of automatic analysis procedures carefully.

The use of small arrays such as tbis cao provide a wealth of informacion on marine mammal communication. The specific quesrions we have concentrated on were related co functional aspects and the behavioral ecology of commu­nication systems, bur arrays can also be used for marine mammal monicoring srudies (Clark et al. 1986). With the relatively inexpensive componencs now available, these arrays can be used more widely chan in che past and have grear pocencial co help us co understand marine mammal communication in more derail.

A CK).;QWLEDGM.E::\TS

We would like co thank Ross and Cromarry Disrrict Council, Ross and Cromarry Enterprise, Scottish Natural Heritage, and the European Life Prog ram for providing equipment and access co the Dolphin and Seal Interpretacion Cenrre. Highland Coun­cil, Merkinch Communiry Council, and the Royal Narional Lifeboat Institution also allowed us co place equipment in their facilities. We also thank Perer]. B. Slater for his continuous support and Bruce Greig, Gordon Hastie, and Tim Barron for help in the field. Christopher Clark. Kurr Frisrrup, Perer ]. B. Slarer, and Rebecca Thomas gave very helpful commencs on a draft of this paper. VMJ was funded by a DAAD­Dokroraodensripendium aus Mirrelo des zweireo Hochschulsooderprogramms. SVP was funded by a University of Aberdeen Scholarship.

L ITERATURE CITED

BEEMAN, K. 1996. SIGNAlloc.alization user's guide. Engineering Design, Belmont, MA CLARK, C. W . 1980. A real-rime direction finding device for determining the bearing

co cbe underwater sounds of Southern cighr whales, Eubalaena australis. Journal of the Acoustical Society of America 68:508-511.

446 MARINE MAMMAL SCIENCE. V OL. I h . N O. 2. ~OOU

CLARK. C. W., W. T. Et.usoN AND K. BEE~IA:-1 . 1986. A prelimwary accounr of che acoustic scudy conducted during the 1985 spring bowhead whale, Bc~laenu mys­tJcet!IJ. migration off Poinc Barrow, Alaska. Report of rhe Inceroacional Wha.ling Commission 36:311-316.

EvANS, W. E., AND].]. DREHER. 1962. Observations on scouring behavior anJ ::LSso­ciated sound production by the Pacific bonlenosed porpoise (Tumops gil!i Dall). Bulletin of che Southern California Academy of Sciences 6 1:21 / -226.

Fo RD, J. K. B. 1989. Acoustic behaviour of residenr killer whales (Omma orca ) off Vancouver Island. British Columbia. Canad ian Journal of Zoology 67 :727-745.

FR.\.'iKEL. A. S., C. W. CLARK, l. M. HER.\tAN AND C. M. G.-\.BRJELE. 1995 . Spacial distribution, habitat utilization, and social inreractions of humpback whales, Mt­gaptera notlaeangliae, off Hawai'i, determined using acoustic and v1sual techniques. Canadian Journal of Zoology 73:1134-1146.

FREITAG, L. E., AND P. L. TYACK. 1993. P::LSsive localization of che AdantiC bocclenose dolphm using whistles and echolocation clicks. Journal of rhe Acoustical Society oi Amenca 93:11 97-2105.

] AI'o: IK, V. M. 1998. Functional and organizational aspecrs of vocal repertoires in boc­clenose dolphins (T11niops truncatu;). Pb. D. thesis. University of Sr Andrews, St Andrews. UK. 11 7 pp.

MtuER, P. J., A:-lD P. L. TYACK. 1998. A small cowed beamforming array co Idenrify vocalizing resident killer whales (Orcinu; orca ) concurrent wirh ~ocal behavioral observarions. Deep-Sea Research II 4 S: 1389-1405 .

M0Ht.. B., A. SL'RLYKKE AND L. A. MILLER. 1990. High mtenstry narwhal dicks. Pages 295-303 in]. A. Thomas and R. A. Kastelein, eds. Sensory abilities of cetaceans: Laboracory and field evidence. Plenum Press , New York. NY.

SAYlGH, L S., P. L. T YAC", R. S. WELLS A~D M . D . Sco TT. 1990. Sig narurt whiscles of free-ranging bottlenose dolphins, Tur;iops tmncatus: mother-offspring compar­ISOns. Behavioral Ecology and Sociobiology 16:247- :260.

SJARE, B. L, ,\NO T. G. SMITH. 1986. The relationship berween behavioral activity and underwater vocalizations of the white whale. Oelphtnaptems fem'Js. Canadian Jour­nal of Zoology 6-k1824-283 l.

Sl-lOL'ZER, R. A., J NL-.:-~N AND B. B. SMUTS. 1993. Use oi SI~narure whisdes during separations and reunions by wild bottlenose dolphin mothers and infants. Behav­ioral Ecology and Sociobiology 33:393-40 2.

SPIESBERGER, J. L., AND K. M. FRtsTRUP. 1990. Passive loc::dizacion or· calling Jnimals and sensing of their acoustic environment using acousric tomography. American Naruralist 135:107-153.

T ARt:SKI, A. G. 1979. The whisde repertoire of the Norch Aclanric p ilot whale (G iob­icephala melaena) and ics relationship co behavior and environment. Pages 34 5-368 in H. E. Winn and B. L. Olla, eds. Behavior of marine animals. Volume 3. Cetaceans. Plenum Press, New York, NY.

T v ACK, P . 1986. Whistle repertoires of cwo borde nosed dol phi as, Tur;iop; tmncarus: mimicry of sigoarure whistles? Behavioral Ecology and Sociobiology 18:251-257.

VAN PARIJS, S. 1998. Reproduccive strategies of male harbour seals, Phoca vitulina. Ph.D. thesis , Universiry of Aberdeen, Aberdeen, UK 187 pp.

W ATKJNS, W. A. , AND W. E. SCHEVlLI.. 1972. Sound source location by arrival-times on a non-rigid three-dim ensional hydrophone array. Deep-Sea Research 19:691-706.

WATKINS, W. A., AND W. E. SaiEvlll. 1974. Listening co Hawaiian spinner porpoises, Stene/la cf fongiro.rtris, with a three-dimensional hydrophone array. Journal of Mammalogy 55:319-328.

WmGART, L. S .. AND H. WHITEHEAD. 1990. Vocalizations of the Norrh Adancic pilot whale (Giobicephala me/a;) as related co behavioral come.:<cs. Behavioral Ecology and Sociobiology 26:399-402.

NOTES 447

V. M. J ANIK/ School of Environmenral and Evolutionary Biology, University of Sc Andrews, Fife KY1 6 9TS, United Kingdom ; S. M. V AN P ARIJS and P. M. THOMPSON, Uruversicy of Aberdeen, Department of Zoology, Lighthouse Field Station, Cromarcy, Ross-shire IVll 8YJ, United Kingdom. Received 8 December 1998. Accepted 25 Augusc 1999.

~ Currenc address: Woods Hole Oc=ogrnphic Inscirucion. Biology Departmenr. MS 34 Red­field, Woods Hole. M:tSsachuserrs 02543. U.S.A.: e-mail: vjanik @whoi .edu.