Distinctive, fine‐scale distribution of Eastern Caribbean sperm whale vocal clans reflects island fidelity rather than environmental variables

Abstract

Environmental variables are often the primary drivers of species' distributions as they define their niche. However, individuals, or groups of individuals, may sometimes adopt a limited range within this larger suitable habitat as a result of social and cultural processes. This is the case for Eastern Caribbean sperm whales. While environmental variables are reasonably successful in describing the general distribution of sperm whales in the region, individuals from different cultural groups have distinct distributions around the Lesser Antilles islands. Using data collected over 2 years of dedicated surveys in the Eastern Caribbean, we conducted habitat modeling and habitat suitability analyses to investigate the mechanisms responsible for such fine-scale distribution patterns. Vocal clan-specific models were dramatically more successful at predicting distribution than general species models, showing how a failure to incorporate social factors can impede accurate predictions. Habitat variation between islands did not explain vocal clan distributions, suggesting that cultural group segregation in the Eastern Caribbean sperm whale is driven by traditions of site/island fidelity (most likely maintained through conformism and homophily) rather than habitat type specialization. Our results provide evidence for the key role of cultural knowledge in shaping habitat use of sperm whales within suitable environmental conditions and highlight the importance of cultural factors in shaping sperm whale ecology. We recommend that social and cultural information be incorporated into conservation and management as culture can segregate populations on fine spatial scales in the absence of environmental variability.

Full text

Ecology and Evolution. 2022;12:e9449.  
|
1 of 13
https://doi.org/10.1002/ece3.9449
www.ecolevol.org
Received:9August2022
|
Revised:3O ctobe r2022
|
Accepted :6Octob er2022
DOI:10.1002 /ece3.9449
RESEARCH ARTICLE
Distinctive, fine- scale distribution of Eastern Caribbean
sperm whale vocal clans reflects island fidelity rather than
environmental variables
Felicia Vachon1 | Ana Eguiguren1 | Luke Rendell2 | Shane Gero3 | Hal Whitehead1
This is an op en access arti cle under the ter ms of the CreativeCommonsAttributionL icense,whichpe rmitsuse,dis tribu tionandreprod uctioninanymed ium,
provide dtheoriginalwor kisproperlycited.
© 2022 The Author s. Ecolog y and EvolutionpublishedbyJohnWiley&S onsLtd.
1Depar tmentofBiolog y,Dalhousie
University,Halifax,NovaScotia ,Canada
2SchoolofBiolog y,UniversityofSt.
Andrews,St.An drews,UK
3Depar tmentofBiolog y,Carleton
University,Ottawa,Ontario,Canada
Correspondence
FeliciaVachon,Depart mentofBiology,
DalhousieUniversity,Halifax,NS,Canada.
Email: fvachon@dal.ca
Funding information
AGOASanc tuar y;Anim alBehav ior
Societ y;NationalGeographicSociety,
Grant /AwardNumber:NGS -62320R-
19-2;NaturalSciencesandEngineer ing
Research Council of Canada
Abstract
Environmental variables are often the primar y drivers of species' distributions as
they define t heir niche. However, individuals , or groups of individuals , may some-
timesadoptalimitedrangewithinthislargersuitablehabitatasaresultofsocialand
culturalprocesses.ThisisthecaseforEasternCaribbeanspermwhales.Whileenvi-
ronmentalvariablesarereasonably successfulin describingthegeneral distribution
of sperm whales in the region, individuals from different cultural groups have distinct
distributionsaround theLesserAntillesislands. Using datacollected over 2 yearsof
dedicatedsurveysintheEasternCaribbean,weconductedhabitatmodelingandhab-
itatsuitabilityanalysestoinvestigatethemechanismsresponsibleforsuchfine-scale
distributionpatterns.Vocalclan-specificmodelsweredramaticallymoresuccessfulat
predictingdistributionthangeneralspeciesmodels,showinghowafailuretoincorpo-
ratesocialfactorscanimpedeaccuratepredictions.Habitatvariationbetweenislands
did not explain vocal clan distributions, suggesting that cultural group segregation
in the Easte rn Caribbea n sperm whale is dri ven by traditions of site/ island fidelit y
(mostlikelymaintainedthroughconformismandhomophily)ratherthanhabitattype
specialization.Ourresultsprovideevidenceforthekeyroleofculturalknowledgein
shaping habi tat use of sperm whal es within suitab le environmental co nditions and
highlighttheimportanceofculturalfactorsinshapingspermwhaleecology.Werec-
ommend that socialandculturalinformationbeincorporatedintoconservation and
managementasculturecansegregatepopulationsonfinespatialscalesintheabsence
ofenvironmentalvariability.
KEYWORDS
Caribbean,cet acean,conservation,culture,habitatmodeling,sitefidelity,spermwhale
TAXONOMY CLASSIFICATION
Behaviouralecology,Conser vationecology

2 of 13
|
VACHON e t al.
1 | INTRODUC TION
Itisnotuncommonforspeciestoonlyoccupyalimitedrangewithin
availablesuitable habitat. Whileenvironmentalvariables areoften
theprimar ydriverofspeciesdistribution(asafailuretomeetcertain
conditionswillreducefitness),socialfac torsmightalsolimitindivid-
uals'rangewithinawidersuitablehabitat.Thisistrueforterritorial
species (e.g., wolves, Canis lupus—O'Neiletal.,2020, chimpanzees,
Pan troglody tes verus—Herbingeretal.,2001),speciesthatshowsite
fidelit y (e.g., furseals,Arctocephalus gazella—Hoffman et al.,2006;
reef fishes, Thalassoma bifasciatum—Warner,1988), as wellas prey
typespecialists(e.g.,killerwhales,Orcinus orca— Filatova et al., 2019)
andhabitatspecialists(e.g.,bottlenosedolphins,Tursiops truncatus—
Kopps et al., 2 014 , elephants, Loxodonta africana— Fishlock
et al., 2016). In case s of prey or habit at specializat ion, individ uals
learntouse,andcanspecializeon,preyorhabitatfeaturesthatare
distrib uted differently from the prey or habitat features used by
other members ofthe same species, thereby resulting ina heter-
ogenous di stribution . Territori ality, site fidelit y,p rey type spe cial-
ization , and habitat spe cialization ar e often group-level processes
that can relate to kinship and/or social learning/culture (with culture
definedasbehaviororinformationsharedwithin acommunitythat
is acquired from conspecifics through some form of social learning;
Whitehead& Rendell, 2015). For i nstance, in dividuals mi ght learn
preyorhabitatpreferencesviasociallearningwithinculturalgroups
asisthecaseinkillerwhaleecotypes(reviewedinRieschetal.,2012)
and/orviavertic alt ransmissionfromp are ntsasisthecasewithbot-
tlenosedolphin“spongers”(Krützenetal.,2005).
However,althoughtheireffectondistributioncanbequitedra-
matic, so cial factors s uch as the ones des cribed above are r arely
include d in analyse s relating to ani mal conser vation. For in stance,
habitat models, which are a widespread tool in conservation as
they allow forthe identification ofcritical habit ats forspecies're-
covery and survivalandcan offerinvaluableinformationregarding
apopulation'shealth (Redfernet al.,2006),considerenvironmental
variablesindetail but rarelyinclude cultural and socialinformation
(exceptions:Eguigurenetal.,2019; Filatova et al., 2019).
As more and more evidence suggests that culture is widespread
in the anim al kingdom (e.g., Whit en, 2017), there is increasing in-
terest in the role of cultural transmission in determining species
distribution (Brakesetal.,2021).Thismightbeparticularlyimport-
antforspecies for which many group-levelbehaviors are culturally
transmitted, such as the sperm whale (Physeter macrocephalus)(e.g.,
Cantor et al., 2015).
Sperm wh ales are deep -diving ce taceans that l ive in all of the
world's oceans (Whitehead, 2003). They have a complex social
struc ture in whic h females an d calves live at l ower latitud es year-
round in s table matril ineally-bas ed social unit s of about 10 mem-
bers (Christal et al., 199 8). Interactions between individuals and
social unit s are thenrestricted to members of the samevocal clan,
a higher-order social s tructure defined by vocal dialec t, that can
occur in sympatry(Gero et al., 2016;Rendell&Whitehead,2003).
Vocal clans can include hundreds to tens of thousands of whales
(Rendell & Whitehead, 2003), are identified by distinctive usage
of stereot yped patter ns of clicks called co das (Gero et al., 2016;
Rendell&Whitehead,2003),andhavebeendocumentedworldwide
(Amano et al., 2014; Amorim et al., 2020; Gero et al., 2016;Huijser
et al., 2020; Rendell & Whitehead, 2003). B eyond acoust ic differ-
ences,spermwhalesfromdifferentvocalclansalsodisplaydifferent
social be haviors (Cant or & Whitehead , 2015),movement patterns
(Vachon et al., 2022;Whitehead&Rendell,2004),anddistributions
(Eguiguren et al., 2 019; Vachon et al., 2022).Becausevocalclanscan
liveinsympatry andgenetic variation is insufficienttoexplainthis
behavior al variatio n (Rendell et al. , 2012), it is believed t hat vocal
clans are cultural entities, with distinctivebehaviors being socially
learnedlargelywithinsocialunits(Cantoretal.,2015).Theexistence
oftheseculturallydrivenvocalclanshas impor tant implicationsfor
thebehavior,ecology,anddistributionofspermwhales,inasimilar
waytotheecotypes of killer whales (Riesch et al., 2012).Therefore,
consider ing conservat ion metrics such a s habitat use with out ac-
counting for culture might lead to misinterpretation as culture can
alter behavior and distribution and subdivide populations in unex-
pectedways(Brakesetal.,2021;Whiten,2017 ).
The popu lation of sperm w hales in the E astern Car ibbean has
been extensively studied but, until recently, at a relatively small
spatial sc ale (i.e., largel y around a single isla nd). Since 2005, T he
DominicaSpermWhale Projec t(DSWP)hasstudied over19sperm
whale social units around Dominica (Gero et al., 2014 ),gaining im-
portantinsightonspermwhalesocialstructureandbehavior(Gero
et al., 2014, 2016).In2019and2020,weextendedthisresearcharea
and condu cted surve ys to include a wid er range along t he Lesser
Antilleanchain(from St.Kitt sandNevistoGrenada).Fromthis,we
gained insightintothewayvocal clansinfluencedthespatial orga-
nization oftheEasternCaribbean spermwhalepopulation (Vachon
et al., 2022).EasternCaribbeanvocalclans(EC1andEC2)appearto
havever ydistinc tivesmall-scaledistributions, with EC1foundpre-
dominantlyaround Dominica,Guadeloupe andSt.Vincentandthe
GrenadinesandEC2foundaroundthetwocentralislands,St.Lucia
and Martinique. This is not unheard of as sperm whale vocal clans
inthe EasternTropicalPacifichave alsobeenshowntohavesome-
what different distributions over a somewhatsimilar scale of 100s
of kilometers (Eguiguren et al., 2019).However,thecauses ofsuch
segregationhavenotbeeninvestigateduntilnow.
Weproposetwocompetinghypothesestoexplainvocalclanis-
land segregation in the Eastern Caribbean.Thefirstis habitat spe-
cializati on, where islan ds vary in the amo unt of each vocal cla n's
preferred habit at type. In this case, foraging strategies specialized
tospecifichabitattypescouldbedrivingthedistributionofEastern
Caribbean spermwhale vocalclans. Assperm whales spendabout
75% of their time fo raging (Whiteh ead & Weilgart, 1991), differ-
ences in for aging strategies relating to environmental variation could
leadtolargedifferencesinoveralldistribution.Thesecondhypoth-
esis is voca l clan-spec ific tradit ions of island pref erences that ar e
arbitrarywithrespecttothehabitateachislandoffers.Thisisakinto
aclassicstu dyofmatingsitec hoiceinblueheadwr asse(Thalassoma
bifasciatum) by Warner (1988) which first showed that preferred

|
3 of 13
VACH ON e t al.
coralheadswereinphysicaltermsnodifferentfromunusedones,a
patternrobusttotranslocationwithpersistentpreferencessocially
maintainedbytraditions.InthecaseofCaribbeanspermwhales,the
differ ent Lesser Anti lles islands mig ht be analogous to th e differ-
ent wrasse mating sites, with individuals from different vocal clans
preferentiallystayinginthevicinityofcertainislandsforreasonsof
tradition(site/islandfidelity) rather than specific physical features.
Whiletranslocationexperimentsarenotpossibleforspermwhales,
wecanaskwhet herclan-specifichabi tatprefe rence sma pontovari-
ationintheamountofpreferredhabitatacrossislandstounderstand
whetherthesepreferencesarelikelytobetraditionalornot.
Therefore,inthispaper,weattempted todif ferentiatebetween
habitat specialization and site/island fidelity by modeling sperm
whale hab itat use in th e Eastern C aribbea n, assessin g the relati ve
importance ofisland geography and habitat characteristics in pre-
dicting sperm whale presence by identifying impor tant environ-
mentalvariablesforEC1andEC2whalesindependently,andtesting
whether the distribution of these variables varies significantly
acrosstheEC1andEC2“islands.”IfEasternCaribbeanspermwhales
are habit at specialis ts, we exp ect spec ific environ mental var iables
tobecloselylinked with EC1andEC2distributionsandthereto be
stark variationinatleastsomeofthesevariablesbet weenEC1and
EC2 “islands .”On t he contrar y,if E astern Ca ribbean spe rm whale
distribution is the result of culturally mediated island/site fidelity,
weexpectislandvicinitytobeabetterpredictorofEC1/EC2sperm
whale pre sence and enviro nmental vari ables to not be signi ficant
factorsinourmodels. Suchanapproach notonlyaimsforadeeper
understandingofagroup-livingandculturalspecies'distributionand
behavior,butalso yieldsa novel approach to integrateintoconser-
vationpolicy.
2 | METHODS
2.1 | Field methods
Data were collected between the months of February and April
2019 and Januar y and March 2 020 in the East ern Carib bean. We
surveye d sperm whale pres ence between the i slands of St. Kitt s
andNevisandGrenadaalongthreetransectlines(LeewardInshore:
5–7nauticalmilesfromcoast,LeewardOffshore: 15nauticalmiles
fromcoastandWindward Offshore:5–7nauticalmilesfromshore)
(Figure 1) from a 12 m aux iliary sailb oat using a two-element hy-
drophonearray (two high-frequency Magrec HPO3 elements with
low-cut f ilter set at 2 kHz) towed behind the vessel on a 100 m
cable. O nce encountered a coustically, fema le sperm whales we re
followed, u sing the towed hydrop hone with the d irection sen sing
software Click DetectoronPAMGUARD,forhourstodays.Codasto
identif yvocalclanswererecordedviaaFirefaceUCorUMC202HD
USBaudiointerfaceconnectedtoaPCcomputerrunningsoftware
PAMGuard(Gillespieetal.,2009),samplingat96 kHzandrecording
continuouslyduringsurveys.TheGPSlocationofourresearch ves-
selwas recorded on a GPS marine chartplotter(StandardHorizon
in 2019 and Raymarinein 2020) every5 min. Given thatwe could
identifysocialunitsinrealtimeusingphotoidentification(seeGero
et al., 2014),weintentionallyspentmoretimewithgroupsofwhales
for which we had little or no prior data and, if conditions allowed,
stayed with unknown groupsuntilwehadrepeatsof multiple indi-
vidual's flukesandhad obtained atleast 80codas(this allowedfor
highconfidenceinidentifyingthevocalclanthatthegroupbelonged
to)(Vachonetal.,2022).
2.2 | Assigning GPS coordinates to vocal clans
All individuals identified on the same daywere considered part of
the same g roup if they had coor dinated behavior an d movement
(Gero et al., 2014). Their co das were used to id entify the gr oup's
voc alcl anme mb ershipfollow ingm et hodsbyHe rs he tal.(2021)(see
Vachon et al., 2022).T heGP Spositionofou rresearchve sselwa sas-
signed to a vocal clan for the length of the encounter: From the time,
we first heard the characteristic echolocation clicks of sperm whales
until we could not hear them or chose to leave the whales due to
weather orlogisticalconstraints(Whitehead, 2003).Wedidnotin-
clude enc ounters with U nit 12 (potential EC 3 vocal clan) ( Vachon
et al., 2022)inthisanalysisaswehaverelativelylittledataregarding
theirdistributioncomparedwithEC1andEC2.WeconsideredGPS
locations for which we had EC3 presence as the presence of sperm
whalesbutdidnotincludethemaseitherEC1orEC2presence.
2.3 | Habitat model variables
We included s even topographic al variables (water depth—Depth,
slope— Slope, dist ance to neares t submarine c anyon—Canyon, dis-
tance to the escarpment— Escarp, distance to the abyss—Abyss,
distance to shelf— Shelf, and distance to the center of the nearest
channel between islands—Channel); six oceanographic variables
(eastward current speed— Ecurr, northward current speed— Ncurr,
zonalvelocityvariance—Zvelv,meridionalvelocityvariance—Mvelv,
inflow through the nearest channel— Inflow, and chlorophyll-a
concentration— Chla); and four general variables (latitude—Lat,
longitude— Long, nearest island— Island, and whether the posi-
tion is leeward or windward of the Lesser Antilles island chain—
Windward)—for a total of 17 potentia l variables ( Table S1), in our
ha bi t at mod els .Th ese pre dic torva ria ble swe re cho senas the yw ere
usefulindescribingspermwhalehabitatintheMediterraneanand
South Pa cific and/or are thoug ht to relate to the aggr egation of
spermwhale'sprey,mesopelagicsquid(Claroetal.,2020; Eguiguren
et al., 2019;Pirottaetal.,2011).
Bathymetric data were obtained from the 2020 General
Bathymetric Chart of the Oceans (https://www.gebco.net/data_
and_products/gridded_bathymetry_data/) and extracted using
ArcGIS.Slopewascalculated fromtheGEBCObathymetriclayer
usingArcGISSlopetoo l. Weu se dd is t an ce togeom or phicfeatu re s
canyon, escarpment,abyss, andshelf aspredictorvariablesasin

4 of 13
|
VACHON e t al.
the habit at models of Cl aro et al. (2020). Geo morphic fea tures'
definitionsandlocationswereobtainedfromHarriset al. (2014)
v ia B l u e Ha b i t a t (w w w . b l u e h a b i t a t s . o r g )(Figure 1).Oceanographic
variables—eastward current speed, northward current speed,
zonal velocity variance, and meridional velocity variance—
were obtained from the NOAA drifter-derived climatolog y of
global near-surface current s database (Laurindo et al., 2017).
Chlorop hyll-a concent ration was ex tracte d from the NOA A vis-
ible infr ared imaging ra diometer suite ( VIIRS) sate llite data and
averaged over the last 3 months prior to each dat a point to ac-
count for thelag between primary production and sperm whale
preyavailability(Jaquet, 1996). Measures of inflow through the
nearestchannelwereobtainedfromJohnsetal.(2002).Thefour
generalpredictors wereincluded toaccountforunexplained, or
unaccounted,environmentalvariation inourdata.Nearestisland
isacategoricalvariablethatcorrespondstothenearestislandto
a GPS point ( in geodesic di stance) and wa s extrac ted using the
NeartoolinArcGIS.Windward/leewardisabinaryvariablethat
describeswhetheraGPSpoint isleeward,east,(N)orwindward
(Y)oftheLesserAntillesislandchain.
The variables depth and slope wererecordedat 0.004° spatial
resolution; variables eastward current speed, northward current
speed, zonal velocity variance, and meridional velocity variance
wererecordedat0.25°resolution,andChlorophyll-aconcentration
was recorded at 0.036° resolution. As these resolutions are lower
than that of our GPS coordinates, we used ArcGIS tools Near and
Spatial jointoextracttheclosestvaluefore achvar iabletoeachGPS
coordinate.Webelievethattheresolutionatwhichthosevariables
are availab le will not negative ly affect our mo deling approac h as
theyhavelittlesmall-scalevariability.
FIGURE 1 Mapdisplayingthe
geomorphic features used to model
spermwhaledistributionintheEastern
Caribbean.Vesseltracksdisplayedindark
gray.
E, G
E, Ga
F
A
F
F
O
, NOAA, USGS
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
S
Shelf
Abyss
Canyons
Escarpment
Geomorphic features
Slope
±
050 10025 Kilometers

|
5 of 13
VACH ON e t al.
2.4 | Habitat modeling
WeusedGPSfixesfromtheresearchvessel'schartplottertakenat
5minintervalsasourunitsofanalysis.Eachdatapointcorresponds
to specific coordinates at a certain time, along with whether sperm
whaleswereacousticallyencounteredatthatpointandtime,aswell
astheclantowhichencounteredwhalesbelongedto(datasetavail-
able as supplementar y material, Data S1). We fittedfour different
habitat modelt ypes (Presence/Absence, EC1, EC2, and Vocal clan)to
ourdata using twoindependent setsofvariables(Environment and
Island)(Figure 2,definedbelow).Here,wedescribeeachmodeltype
andtherationalefortestingthemacrossthetwovariablesets.
1. Presence/Absence:Thismodeldescribedthegeneraldistribution
of sperm whales in the Lesser Antilles, regardless of vocal clan
member ship. The re sponse vari able was 0 for ac oustic abs ence
of sperm whale and 1 for acoustic presence of sperm whales.
This allowe d us to identify key v ariables for s perm whale hab -
itat in the Lesser Antilles and assess whether modeling sperm
whale distribution independently for each vocal clan resulted
in a significant improvement in predictive accuracy.
2. EC1/EC2: These models described the distribution of sperm
whales that were assigned to the EC1 and EC2 vocal clans, respec-
tively.FortheEC1model,theresponsewas0fortheacousticab-
sence of sperm whales or the presence of EC2 and/or EC3 whales
and1fortheacousticpresenceofEC1whales.Conversely,forthe
EC2 model, the response was 1 for the acoustic presence of EC2
whales and 0 otherwise. These models allowed us to compare
theperformance of vocalclan-specifichabitat models tothat of
generalhabitatmodels(i.e.,Presence/Absence)aswellas identify
importantenvironmentalvariablesforpredictingthepresenceof
EC1andEC2whales,respectively.Theseenvironmentalvariables
werethenusedinourhabitatsuitabilityanalysis(seebelow).
3. Vocal cl an: This model w as fitted to ide ntify the var iables that
best distinguishbetweenthepresenceofEC1andEC2.The re-
sponse was 0 for EC1 acoustic presence and 1 for EC2 acoustic
presen ce. Here, a high predi ctive accurac y would suggest t hat
individuals from different vocalclansprefer contrasting habitat
modelvariablesand,therefore,suggestanimportantcontribution
ofsocialfactors(i.e.,vocalclanmembership)tospermwhaledis-
tribution.The datasetusedfortheVocal clan model was smaller
than that for the Presence/Absence, EC1, and EC2 models since
weonly usedsperm whale presencedat apoints(1sinPresence/
Absencemodel).
We tested these four habitat model types independently on
two set s of variables: eit her a full set of enviro nmental varia bles
(Environmentset), ornearestislandvariables(Islandset),andcom-
pared their predictive performance. The Island set includes vari-
ables Island and Windward, while the environment set includes all
remaining 15 environmental predictors and Windward. Weex pect
models using the Environmentvariableset to per form muchbetter
than the ones using the Islandvariablesetifspermwhalesarehab-
itatspecialistandtheoppositeifpatternsofdistribution aredriven
bysite/islandfidelity.Toavoidconfusion,modelnamesontheirown
(e.g., Presence/Absence)willrefertothemodelsperformedusingthe
Environmentvariable set and models followedby“Island”will refer
to the models performed using the Islandvariables et(e.g.,Presence/
Absence Island).
Modeling approach
Habitat models were fitted using generalized estimating equations
(GEEs;Liang&Zeger,1986),inwhichvariableswereusedaspredictors
of sperm whale presence (Presence/Absence, EC1, and EC2 models) or
vocalclanmembership(Vocal clanmodel),followingPirottaetal.(2011)
and using package geepack inR(Højsgaard et al., 2005).Thisapproach
has been us ed in other cetace an distributi on studies (e.g., E guiguren
et al., 2019; Pirottaetal.,2014; Tepsich et al., 2014)andisappropriate
whendat aarerecordedcontinuouslyalongsurveytransects.Wechose
GEEsoverothermethodssincetheyexplicitlyaccountforautocorrela-
tion (Liang&Zeger,1986).Data point s were clumped into blocks that
corresponded to sperm whale encounters. Under this framework, resid-
ua l sar e allo w edt obec o rrel a tedw i thin b lock s , butw eas s u mei n d epen d -
encebetweenblocks.Weusedencountersasourblockingvariable as
FIGURE 2 Summaryofhabitatmodelingapproach.
Sperm whale data
Presence/Absence
0: Absence
1: Presence (EC1, EC2,
EC3)
EC1
0: Absence, EC2, EC3
1: EC1
EC2
0: Absence, EC1, EC3
1: EC2
Vocal clan
0: EC1
1: EC2
Environment Island Environment Island Environment Island Environment IslandVariable Type
Model Type

6 of 13
|
VACHON e t al.
itwassuccessfullyusedinsimilarstudies(Eguigurenetal.,2 019;Pirotta
et al., 2011),andwefoun dthistob eanappropriategrou pingvar iableas 
theautocorrelationamongdatapointseventuallyconvergedat0within
each encounter (Figure S1 ).Wemodeledtherelationshipbetweenvari-
ablesandspermwhale presenceaslineartermsonly,asincludingnon-
linear relationships as in previous studies (Eguiguren et al., 2019;Pirotta
et al., 2011)onlyslightlyincreasedoverallfitandpredictiveaccuracy,at
thecostofinterpretability.
Westructuredourmodelingapproachintofivesteps(FigureS2,
described below), which were repeated independently for the
Presence/Absence, EC1, EC2, and Vocal clanmodels.Rcodeavailable
assupplementar ymaterial(CodeS1).
Preparing variables
We looked at the v ariables' distr ibutions and log ged ones, which
were highl y skewed. All var iables were th en standa rdized by sub-
tract ing the mean an d dividing by sta ndard deviatio n to facilitate
model convergence.
Removing collinearity
First,wecalculatedcorrelationcoefficientsbetweenallpairsofpre-
dictorvariables.Variableswhichhadcorrelationcoefficientsabove
0.4wereconsideredto be correlated and not includedinthesame
model.Fromthis,webuiltallpossiblecombinationsofuncorrelated
predictors into potential models, which were then tested for mul-
ticollinearityby measuring the generalized varianceinflationfactor
(GVIF)(carpackageinR).ModelswhichhadapredictorwithaGVIF
value above3were discarded,and all other potential models with
GVIFvaluesbelow3 wereusedasthefirst step in back ward step-
wise selection.
Model selection
Weused QIC (Pan, 2001), an extension of the Akaike Information
Criterion (AIC) that applies to GEE models, to compare models
using manu al backward stepwise sele ction (package MuMIn in R ,
Bartoń,2013).Westarte dfr omallthepotentialco mbi nat ionsofu n-
correlatedpredictors(step2)andcomparedtheirQIC(ΔQIC)aswe
removed a sin gle variabl e in turn. Th e model with t he lowest QI C
isthenused as the startingmodel for the nex tstep, repeating this
procedu re until the rem oval of any variab le in the model l eads to
anincreaseinQIC.ThehighertheabsolutevalueofΔQICbetween
models, the larger the gap in their predictive performance. As such,
wechosemodels withfewervariablesiftheir ΔQICwas10or less
fromtheoriginalmodeltoencouragevariableremoval.Thevariables
within the final model are then ordered according to how much their
removalincreasesQIC(fromhighesttolowest).
Model validation
The best models from step 3 were then further evaluated using
leave-one-out cross-validation where encounters were iteratively
removedfromthedata.Wecomparedthepercentageofdatapoints
thatwerecorrectlyassigned(predictiveaccuracy,Hastieetal.,2009)
betwee n the step 3 mo dels to that of th e same mode l minus one
variable. If the predictive accuracy of models with fewer variables
washigherthanthatoftheoriginalmodel,weremovedthatvariable
and sta rted this pr ocess again u ntil predic tive accurac y was high-
estforthemodelfromwhichwedidnotremovevariables.Thiswas
doneasstepwiseselectionusingQICcansometimesretainspurious
variables(Pirottaetal.,2011).
Model performance was then assessed in terms of how well mod-
elsfitthedata(goodness offit)by measuringthepropor tionofdat a
points co rrect lyassigne da sp re sencesorabs en ce s(orEC1/EC 2i nt he
vocal cla n models) using co nfusion matr ices (Fieldin g & Bell, 1997).
Totransformmodelpredictionsfromarangeof probabilitiestoabi-
nary (presence orabsence), weused the point of maximumdistance
betweenthe receivingoperating characteristic (ROC)curve and the
45-degree diagonal as the cut-off probability, usingthe R package
ROCR(Singet al., 2005). Additionally,wemeasuredmodelgoodness
offit bycalculatingtheareaundertheROCcurve(AUC), which also
reflectsoverallmodelperformance(Fielding&Bell,1997).
We finally compared the performance metrics described be-
tween models with Environment variables andIsland variables for
eachmodelt ype(Presence/Absence, EC1, EC2, and Vocal clan)tode-
terminewhetherdifferencesin distribution aredrivenprimarily by
habitatspecializationorsite/islandfidelit y.
Prediction maps
Todisplaytheresultsofourhabitatmodels,webuiltpredictionmaps
fromthebestpost-cross-validationPresence/Absence, EC1, EC2, and
Vocal clanmodels. Mapswerebuiltbyimportingour modelpredic-
tionsfromRintoArcGISPro.
2.5 | Habitat suitability analysis
Tofurtherestablish whethervocal clans have distinct distributions
asaresultofhabitatspecializationorsite/islandtraditions,wecon-
ducted a habitat suitability analysis for each Lesser Antilles island.
This was done by creating a 0.1 degree grid of GPS points that
extend ed 30 nautic al miles of fshore (repr esentative of o ur effor t,
Figure 1)leewardofeachislandandassigningthesepoints,andtheir
corresponding environmental variable values,tothe closestisland.
Fromthis,weobtainedarangeofvaluesforeachenvironmentalvar-
iable for eachisland whichwe couldthen compare between “EC1”
and “EC2” islands. Only environmental variables that were part of
the final EC1 and/or EC2modelswereincludedintheseanalysesas
theyweretheonesthatweresuggestedtoimpactvocalclandistri-
bution.Wecomparedtheenvironmentalconditionsbetweenislands
using t-te sts to test whet her each environme ntal variable sig nifi-
cantly differed betweenislands predominantlyusedby EC1andis-
landspredominantlyusedbyEC2.
We expecte d environme ntal variab les to be corre lated to pre-
ferred islands ifthe environmental variables themselves are driv-
ing vocal clan distribution (e.g., EC1 whales prefer canyons and
Dominica,GuadeloupeandSt.Vincent havemorecanyonsthanSt.
LuciaandMar tinique)anduncorrelatedifvocalclansaredistributed

|
7 of 13
VACH ON e t al.
arounddifferentislandduetositefidelitytraditions(e.g.,allislands
have similar a mounts of canyons b ut EC1 whale are only se en in
Dominica,Guadeloupe,andSt.Vincent).
3 | RESULTS
Overourtwofield seasons(Februar ytoApril2019and January to
March 2020 ), we spent 107 days at sea (Figure 1). Sper m whales
were located throughout the leeward transects, with higher con-
centrations found around Martinique, St. Lucia, and Dominica, but
were not hea rd to windward of t he islands. We had a tot al of 50
sperm whale encounters, 24 with EC1 groups, 22 with EC2 groups,
five with a n EC3 group, and o ne with both EC2 a nd EC3 (Vachon
et al., 2022),fromwhichwerecorded778 hofspermwhalevocaliza-
tions.Altogether,weobtained26,776GPScoordinatedatapoints.
3.1 | Habitat modeling
Refer to Figure 3forafullbreakdownofthePresence/Absence, EC1,
EC2, and Vocal clan habitat models at every selection step. Best
pre-cross-validation and post-cross-validation habitat models, as
well as corresponding results using the Islandv ariable set , can be
found in Table 1andTableS2withassociatedQIC,AUC,goodness
offit,andpredictiveaccuracy.Below,weexpandongeneralresults
fromeachmodeltype.
3.1.1 | Presence/Absencemodel
Thismodelhad50.62%predictiveaccuracyand69.8%goodnessof
fit in determining sperm whale presence, regardless of vocal clan,
using environmentalvariables. Sperm whaleswere more of ten en-
countere d in areas with low chl orophyll-a conc entration, clo se to
the continental shelf, relatively close to between-island channels
andfurtherawayfromcanyons(FigureS3).Thenegativecorrelation
betweenpresenceandchlorophyll-aconcentrationcouldbecaused
bytherelativelylowchlorophyll-aconcentrationsacrosstheLesser
Antilleschain, spatiallagbetweenWindwardproductivityandlee-
wardbiomassorthetemporallagbetweenprimaryproductivityand
cephalopod biomass (Jaquet, 1996; Pirot ta et al., 2011), alth ough
wetriedtoaccountforthisbyconsideringchlorophyll-aconcentra-
tion over the last 3 months as in Eguiguren et al. (2019). The final
FIGURE 3 Summaryofhabitatmodelingresultsforeachhabitatmodelateachstep(Environmentvariableset).
Presence/Absence
log transform Slope
GVIF removes 3 models
23 potenal models
Best backward model:
Pres ~ Windward +
Chla + Shelf + Zvelv +
Inflow + Channel +
Canyon
Removes: Windward,
Zvelv, Inflow
Pres ~ Chla + Shelf +
Channel + Canyon
Figure S4
EC1
log transform Slope
GVIF removes 3 models
23 potenal models
Best backward model:
Pres ~ Ecurr +
Windward + Escarp +
Abyss + Zvelv
Removes nothing
Pres ~ Ecurr +
Windward + Escarp +
Abyss + Zvelv
Figure S7
EC2
logtransform Slope
GVIF removes 6 models
20 potenal models
Best backward model:
Pres ~ Mvelv +
Windward + Inflow +
Chla + Channel + Depth
+ Zvelv
Removes: Inflow
Pres ~ Mvelv +
Windward + Chla +
Channel + Depth + Zvelv
Figure S8
Vocal clan
logtransform Slope
log transform Chla
GVIF removes 7 models
17 potenal models
Best backward model:
Pres ~ Ecurr + Channel
+ Zvelv
Removes: Channel
Pres ~ Ecurr + Zvelv
Figure S10
STEP 1:
Prepare variables
STEP 2:
Remove collinearity
Potenal models
STEP 3:
Model selecon
STEP 4:
Model validaon
Best model
STEP 5:
Predicon map

8 of 13
|
VACHON e t al.
Presence/Absence Islandmodel(Pres ~ Windward + Island)performed
better than thePresence/Absencemodel(Pres ~ Chla + Shelf + Chan
nel + Canyon) with ΔQIC of 2281.4. The Presence/Absence Island
modelhad59.61%predictiveaccuracyand65.8%goodnessoffitin
determining sperm whale presence and suggests that more sperm
whales oc cupy the water s off the cen tral islan ds of Dominic a and
Martinique (Figure S4),for reasonsnotfully explained bythe envi-
ronmentalvariablesthatweconsidered.
3.1.2 | EC1andEC2models
ModelingspermwhaledistributionindependentlyforEC1andEC2
increas ed model pre dictive accu racy, goodness of f it and lowered
QIC for both the models using environment and island variables
(Table 1).
EC1 whales prefer areas of low eastward current speed, low
zonal veloc ity variance, wit hin the escarpm ent designation, aw ay
from the abyss, leeward of the Lesser Antilles chain (Figure S5).
Bycontrast,EC2whales preferareaswith high meridional velocity
variance,lowchlorophyll-aconcentration,deeperintheocean,and
lowzonalvelocityvariance,closertochannelsleewardoftheLesser
Antilles chain (Figure S6).Unsurprisingly,variableWindward was im-
portant for both the EC1 and the EC2 models since sperm whales
werenotheardwindwardof the islandchain.Thisresult shouldbe
viewed ca utiously since th e leeward side of the is land chain was
muchmoreextensivelysurveyedthanthewindwardside(Figure 1).
Zonalvelocityvariance(Zve lv) was also impor tant forboth models
withEC1sperm whalesencounteredin areas ofhighzonalvelocity
variance and EC2 sperm whales encountered in areas with low zonal
velocityvariance(FiguresS5 and S6).
Thebest EC1model(Pres ~ Ecurr + Windward + Escarp + Abyss
+ Zvelv) and the best EC2 model(Pres ~ Mvelv + Windward + Chla
+ Channel + Depth + Zvelv) p erformed w orse than the EC1 Island
(Pres ~ Windward + Island)andEC2 Island(Pres ~ Windward + Island)
models with respective ΔQICof3115.5and501.4.Accordingtoour
predic tion maps, we exp ect EC1 sperm whal es to aggregate ne ar
Dominica,Guadeloupe,St.VincentandtheGrenadinesandSt.Kitts
and Nevis; a nd EC2 sperm wha les to aggregat e near St.Luc ia and
Martinique (Figures S7 and S8). Suc h predicti ons not only refl ect,
asexpected,thefieldobservationsthatwereusedtoconstructthis
model (Vachon et al., 2022),but also resultsfromthelong-termre-
searchoffDominicabytheDSWP,withEC2groupsseldomencoun-
tered off D ominica (onl y 2.5% of photo ide ntificati on encounter s;
Gero et al., 2016; Vachon et al., 2022).
3.1.3 | Vocalclanmodel
This model had great accuracy in distinguishing between EC1and
EC2 vocal cla n distribut ion using both th e Environment and Island
variablesets(92%and96.5%goodnessoffit,and49.7%and76.8%
predictive accuracy). EC1whales were more often encountered in
areas of low ea stward cu rrent spee d and high zonal ve locity var i-
ance, while EC2 whales were more often encountered in areas
of high east ward current speed and low zonal velocity variance
(Figure S9).
The Vocal clan Islandmod el (Pre s ~ Windward + Island)performed
better thantheVocal clanmodel(Pres ~ Ecurr + Zvelv)withΔQICof
5033.8,andEC1whalespredominantlyneartheislandsofDominica,
GuadeloupeandSt .VincentandtheGrenadinesandEC2 predomi-
nantlynearSt.LuciaandMartinique(FigureS10).
3.2 | Habitat suitability
The lower QIC and higher predictiveaccuracy of theEC1 Island,
EC2 Island, and Vocal clan Island models (Table 1) su g ges tth a tv ocal
TAB LE 1 Bestvariablecombinationsforeachmodelt ypewithassociatedQIC,ΔQIC,AUC,goodnessoffit,andpredictiveaccuracy
(post-stepwisecross-validation)
Model type Variable set QIC ΔQIC AUC
Goodness of
fit (%)
Predictive
accuracy (±SE)
Presence/
Absence
Env Chla + Shelf + Channel + Canyon 32,966.3 2281.4 0.71 69.8 50.62%(±0.02)
Island Windward + Island 30,684.9 - 0.69 65. 8 59.61%(±0.04)
EC1 Env Ecurr + Windward + Escarp +
Abyss + Zvelv
19,006. 3 311 5.5 0.79 7 7.1 56.65%(±0.03)
Island Windward + Island 15,890.8 - 0.86 72.9 72.05% (±0.04)
EC2 Env Mvelv + Windward + Chla + Channel +
Depth + Zve lv
16,52 2. 2 501.4 0.86 75.35 57.73% (±0.02)
Island Windward + Island 16,020.8 - 0.83 73.2 62.27%(±0.04)
Vocal clan Env Ecurr + Zvelv 6152.1 5033.8 0 .92 92.0 4 9.7 % (±0.05)
Island Island 1118. 3 - 0.99 96 .5 76.8%(±0.14)
Note:Usinghabitatmodelsandhabit atsuitabilit yanalyses,wepresentanddiscussaremarkableandunexpectedpatterninthedistributionof
EasternCaribbeansp ermwhales.UnliketheirPacificconspecifics ,EasternCaribbeanspermwhaleshaveshort-rangemovementsanddisplayisland
fidelityacrossmultipleyears.Suchfine-scaledistributionappear stobeculturallydrivenwithdifferentculturalgroups(calledvoc alclans)occupying
distinctiveislandsalongtheLesserAntillesasaresultoftraditionsofsitefidelit yratherthanenvironmentalvariation.

|
9 of 13
VACH ON e t al.
clandistributionmightbebetterexplainedby site/islandfidelity
than theuse of specific habitat variables. Ourhabitatsuitability
results also corroborated this conclusion as the environmental
variables that wereconsidered significant predictors of EC1 and
EC2 presence in the EC1 and EC2modelsdidnotsignificantlydif-
fer bet ween EC1 and EC2 islands, apa rt from Abyss and Depth
(t = −4.01, p-value = .007 and t = 3.6 8, p-value= .010, respec-
tively;Figure 4).Altogetherthissuggeststhatspermwhalesfrom
differ ent vocal clans do not us e different islan ds because they
haveaunique,orsignificantlydifferent,selectionofphysicalhabi-
tat properties.
Similar results were obtained if we only used surveyed grid
poi nt sr atherth antheex trapolated30n au tica lmileoffshore0 .1d e-
greegridtoc arryoutthisanalysis(FigureS11).
4 | DISCUSSION
In this stu dy,we a ttempted to tes t the competing hy potheses of
habitatspecialization and traditionalsite/islandfidelity in explain-
ingthe starkdif ferentiationinEC1and EC2vocalclandistributions
intheEasternCaribbean. Our resultssuggest that site/islandfidel-
ity,ratherthanenvironmentalvariation, isthemaindriver ofsperm
whale dis tribution in t he Lesser Antil les, with dif ferent process es
operating at the species and vocal clan levels.
At the species level, sperm whales use areas that are close to
the continental shelf and channels (Presence/Absencemodel).Such
correlations between sperm whale distribution and topography
have been documented for sperm whales elsewhere (e.g., Claro
et al., 2020;Pirottaetal.,2011;Wong&Whitehead,20 14)andmost
FIGURE 4 Habitatsuitabilit yofEC1(aquamarine)andEC2(red)islandsaccordingtosignificantenvironmentalvariablerangewithina0.1
degreegridextending30nauticalmilesleewardofeachisland.NosignificantdifferencesinvariablevaluesbetweenEC1andEC2islands.
0%
25%
50%
75%
100%
Habitat type
Shelf Escar pment SlopeCanyonAbyss
0.02
0.04
0.06
Meriodional velocity variance
−0.4
−0.2
0.0
Eastward current speed
0.025
0.050
0.075
Zonal velocity variance
−3000
−2000
−1000
0
Depth (m)
0.1
0.2
0.3
0.4
St.Kitts & Nevis Montserrat Antigua Guadeloupe Dominica Martinique St.Lucia St.Vincent & Grenadines Grenada
Chlorophyll−a concentration

10 of 13
|
VACHON e t al.
probablyreflectfoodavailabilityasverticalwatermovementassoci-
atedwithslopedareas likely promotesprimaryandsecondarypro-
ductivity (Tynanetal., 2005). However,such coarsemodelsfail to
capturethevariabilitycreatedbydifferencesinunitmovement,clan
membership,andforagingsuccessatfinerspatialscales(asreported
byJaquet & Whitehead,1996inthe SouthPacific)and seemedto
beimpacted, even at this scale, by thewhales' bias towardcertain
islands with the Presence/Absence Island model per formingbetter
than the Presence/Absence (Table 1).
Th e d r a mati c i n creas e i nthe p e r f o r m ance o fvo c a l c lan- s p e cific 
models over a general species presence model is one of the most
strikingresult sofourstudy.ThepreferenceoftheEC2vocalclan
forSt.LuciaandMartiniqueandtheEC1vocalclanforDominica,
Guadeloupe, and St. Vincent and the Grenadinesdoesnotrelate
toenvironmentalvariables,astheydonotsignificantlyorsubstan-
tiallydif feracrossislands(Figure 4),butratherseemtobecaused
bysite/islandfidelitywiththeEC1 Island, EC2 Island, and Vocal clan
Islands models performing much better than theircounterpart s
(Table 1).In this case, culture, viaconformism and homophilyto
island preference traditions,wouldact as a barriertopopulation
mixture(e.g.,Centolaetal.,2007; Riesch et al., 2012).Wesuggest
thatindividualspermwhalesstayinthevicinit yofspecificislands
becausethosearetheislandswheretheywereraised,wherethey
learnedto forage,where their closeassociates and family mem-
bers can beencountered, and wherethey can avoidinteractions
with mem bers of other voc al clans. Confo rmism and hom ophily
have alread y been repor ted in East ern Caribb ean sperm w hales
with highly stereotypical vocal repertoires (conformity, Konrad
et al., 2018)andindividualsexclusivelyassociatingwithmembers
oftheirownvocalclan(homophily,Geroetal.,2016).Itisalsonot
surprising that individual sperm whales could learn island prefer-
encesfromothermembersoftheirsocialunitsasotherbehaviors
areculturallymaintainedwithin vocal clans (e.g.,socialvocaliza-
tions [Gero et al., 2016; Rendell &W hitehead,2003], dive syn-
chrony[Cantor&Whitehead,2015], movement patterns [Vachon
et al., 2022; Whitehead & Rendell, 2004], and social structures
[Cantor & Whitehead, 2015]) andsince cultural transmission has
been suggested asthemost likely mechanism for the emergence
of vocal clans themselves (Cantor et al., 2015).
4.1 | Limitations
Thisstudyislimitedinitstemporalscope.WhileEC1and EC2distri-
bution patternswerestableoverthe2 yearsofthisstudy,andwhile
theyappeartohavebeenstablesince2005(Geroet al.,2014 , 2016;
Vachon et al., 2022), shift s could stil l occur over longe r timesca les,
as it did in the Galapagos (Cantor et al., 2016). However,while the
location of Eastern Caribbean voc al clans might change in future,
the mechanisms responsible for their spatial segregation are likely
toremainthesame.Thisstudy might also belimited bytheenviron-
mental variablesthatwereincludedinhabitatmodels. However,this
isunlikely as we cover a wide arrayofenvironmental variable types
(geomorphicfeatures, oceanographicprocesses,and biological pro-
cesses),includingvariablesthatwerepreviouslyconsideredimportant
forspermwhalehabitat(e.g.,Claroetal.,2020; Eguiguren et al., 2019 ;
Pirottaetal.,2011)andenvironmentalvariablesarerarelytotallyun-
correlated.Futureresearch couldinvestigatesperm whaleprey den-
sity (e.g., fromsquid species survey and scat samples) and examine
how prey den sity varies w ith the prese nce of differen t vocal clans
and/orthe proximity ofdifferent islands. Measures ofsperm whale
preydensityremainundocumentedintheLesserAntilles.
4.2 | Implications for conservation
Theperformanceofourhabitatmodelswasgreatlyimprovedbythe
inclusionofaculturalindicator.Wesuggestthatthelowpredictive
accuracy of our Presence/Absencemodeliscausedbyconfound-
ing varia bles across vo cal clans , something t hat could als o explain
whyotherspermwhalehabitatmodelssometimesfailtoreachhigh
predictiveaccuracywhencomparedtoothercetaceanspecies(e.g.,
Claro et al., 2020; Tepsich et al., 2 014).
Ourresultshighlight how cultural factors canleadtoimport-
ant, management-relevantvariations in the way populationseg-
ments use any given habitat, evenat relatively small geographic
scales foralarge, highlymobile,and pelagicanimal.Inthis case,
tradit ions of site/island fi delity appea r to be a more import ant
determinant of sperm whale distribution within suitable habitat
than are env ironmental var iables. Adding t his cultural len s, not
only allowedfor a betterunderstanding of population structure,
but also habitat use—two crucial variables in conservation and
management.
Likemanyotherpopulations,EasternCaribbeansperm whalesare
now facing unprecedented anthropogenic threats related to global
warming, increased ocean noise, and other human activities (e.g.,
Weilgar t, 2007; Whitehead et al., 2008). Sperm whales studied off
Dominica (predominantly EC1 units) were declining at a 4.5%/year
ratebetween2010and2015(Gero&Whitehead,2016),andthesame
might be tr ue for sperm whale s inhabiting the ot her Lesser Antill es
islands.Under these circumstances, it iscritical to builddetailed hab-
itat models which capture both impor tant culturaland environmental
variables. These habitat models cannot only be used to help protect
the population asawhole,but also identif y areasofhigh importance
foreachculturalgroup.Thisalignswithrecentconservationshiftaway
fromsolelygeneticdiversitytotheincorporationofculturaldiversit yas 
animport antcomponentofpopulations'health(Brakesetal.,2021)and
supports the recognition of sperm whale vocal clans as independent
evolutionarilysignificantunits(ESU)forconservationandmanagement.
4.3 | Implications for sperm whale ecology/
psychology
This studyaimed at incorporating both environmental and cultural
variability into the commonly used ecological and conservation

|
11 of 13
VACH ON e t al.
approachofhabitatmodeling.Byindependentlymodelingvocalclan
distribution,wewereabletogainamoredetailedinsightintosperm
whale population structure, the mechanisms responsible for their
distribution, and greatly increase habitat model accuracy. Our re-
sultssug gestthatspermwhalehabit atuseintheEasternCaribbean
ispredominantlyshapedbyculturalinformationratherthanenviron-
mental cues. Given the matrilineal social structure of these groups,
this not only highlights the importance of older females, moth-
ers, aunts, and grandmothers as repositories of knowledge within
social uni ts and voca l clans (as is the ca se in elephant s—McCo mb
et al., 2001),butalsoimpliesthatspermwhalesareabletorecognize
and commu nicate fine -scal e cultural b oundarie s in the absen ce of
physical barriers or environmentalgradient s. Over long timescales,
these bo undaries ar e unlikely to be im permeabl e (as few EC2 en-
countershavebeendocumentedinDominica;Geroetal.,2016)and
might change (e.g., Eastern Tropical Pacific vocal clan turnover—
Cantor et al., 2016), b ut nonetheles s remain cultural ly driven. As
such,ourfindings haveimplicationsbeyondtheEasternCaribbean,
and beyond s perm whales, to ou r understand ing of cultural spe-
cies.Itiscrucialtoassessthedistribution,andbehavior,ofcomplex
species inalltheircomplexity(genetic,environmental,cultural, and
theirintersections)toproperlyconserveandunderstandthem.
AUTHOR CONTRIBUTIONS
Shane Gero:Methodology(supporting);writing–reviewandedit-
ing (equal). Luke Rendell: Fundin g acquisition (le ad); investigatio n
(suppor ting); wr iting – revi ew and editing (e qual). Hal Whitehead:
Fundingacquisition(supporting); investigation(supporting);super-
vision (le ad); writing – review and e diting (equal). Felicia Vachon:
Conceptualization(lead);datacuration(lead);formalanalysis(lead);
funding acquisition (supporting); investigation (lead); methodol-
ogy(lead);project administration(lead);visualization(lead); writing
–originaldraft (lead).Ana Eguiguren:Formalanalysis (supporting);
investigation (supporting); methodology (supporting); resources
(equal);writing–reviewandediting(equal).
ACKNOWLEDGMENTS
This research wouldnot have been possible without support from
ourpar tners:CARIMAMandtheUniversityoftheWestIndies,and
funders: the National Geographic Society (NGS-62320R-19-2),the
AGOA Sanct uary, the Natura l Sciences and Eng ineering Rese arch
Council of Canada (NSERC), and the Animal Behavior Society.We
would also thank the crew that came on Balaena in 2019 and 2020 to
helpwithdatacollectionaswellasDr.EnricoPirottaforgivingadvice
onhabitat modeling. This research was conducted with permission
fromthe Departmentof Fisheries ofSt.Lucia,the Departmentof
Marine ResourcesofSt.KittsandNevis,theFisheriesDepar tment
ofGrenada,theMinistr yofAgricultureofMontserrat ,theFisheries
Division of Dominica, the Ministry of Agriculture, Forestry and
Fisheries of St. Vincent and the Grenadines and the Canadian
CouncilonAnimalCare(CACC).Wearealsogratefultothepastand
current fundersofThe DominicaSperm WhaleProjectwhosecon-
tributionsenabledthedelineationoftheclans.
CONFLICT OF INTEREST
Wehavenoconflictofinteresttodisclose.
DATA AVAIL AB ILI T Y STATE MEN T
Data are available as supplementary material and on Dr yad at:
https://doi.org/10.5061/dryad.mcvdnck4c.
ORCID
Felicia Vachon https://orcid.org/0000-0002-5883-6009
REFERENCES
Amano, M., Kourogi, A., Aoki, K., Yoshioka, M., & Mori, K. (2014).
Differences in spermwhalecodasbetwe entwowaters off Japan:
Possiblegeographicseparationofvocalclans.Journal of Mammalogy,
95(1),169–175.h t t p s : / / d o i . o r g / 1 0 . 1 6 4 4 / 1 3 - M A M M - A - 1 7 2
Amorim , T.O. S ., Rendell, L., Di Tullio, J., Secchi, E. R., Castro, F.R., &
Andriolo, A. (2020). Coda repertoire and vocal clans of sperm
whales in the western Atlantic Ocean. Deep Sea Research Part
I: Oceanographic Research Pap ers, 160, 103254. h t t p s: //d o i .
org /10.1016/j.dsr.2020.10325 4
Barto ń, K. (2013). MuMIn: Multi- model inference. In R pack age version
1.10.0. (Vol. 1). R p ackage version 1.42 .1. https://CRAN.R-proje
ct.org/packa ge=MuMIn
Brakes, P., Carroll, E. L., Dall, S. R. X., Keith, S. A., McGregor, P. K.,
Mesnic k, S. L., N oad, M. J., Re ndell, L ., Robbins , M. M., Rut z, C.,
Thornton,A., Whiten,A., Whiting, M. J., Aplin,L .M .,Bearhop,S .,
Ciucci, P.,Fishlock, V.,Ford, J. K. B.,NotarbartolodiSciara ,G., …
Garlan d, E. C. (2021). A de epening und erstand ing of animal cul-
ture suggests lessons for conservation. Proceedings of the Royal
Society B: Biological Sciences, 288(1949), 20202718. h t t p s: //d o i .
org/10.1098/rspb.2020.2718
Cantor, M., S hoemaker, L. G ., Cab ral, R. B ., Flores , C. O., Varga , M., &
Whitehead,H.(2015).Multilevelanimalsocietiescanemergefrom
cultural transmission. Nature Communications, 6(1), 8091. h t t p s: //
doi.org/10.1038/ncomm s9091
Cantor, M., & W hitehead, H . (2015). How does so cial behavior di ffer
among sperm whale clans? Marine Mammal Science, 31(4), 1275–
1290. https: //doi.org/10.1111/mms.12218
Cantor,M.,Whitehead, H., Gero,S.,& Rendell,L.(2016).Culturalturn-
over among G alápagos sperm whales. Royal Society Open Science,
3(10),160615.https://doi.org/10.1098/rsos.160615
Centola, D., González-Avella, J. C., Eguíluz, V. M., & San Miguel, M.
(2007).Homophily,culturaldrift, and theco-evolutionofcultural
groups. The Journal of Conflict Resolution, 51(6),905–929.
Christ al, J., Whiteh ead, H., & Lett evall, E. (1998). Sperm w hale social
units: Variation and change. Canadian Journal of Zoology, 76(8),
1431– 144 0.
Claro,B.,Pérez-Jorge,S.,&Frey,S.(2020).Seafloorgeomorphicfeatures
asan alternative approach into modellingthedistributionof ceta-
ceans. Ecological Informatics, 58, 101092. ht tp s://doi.org /10.1016/ j.
ecoinf.2020.101092
Eguiguren,A.,Pirotta,E.,Cantor,M.,Rendell,L.,&Whitehead,H.(2019).
Habitatuseofculturally distinct Galápagossperm whalePhyseter
macrocephalus clans. Marine Ecology Progress Series, 609, 257– 270.
https://doi.org/10.3354/meps1 2822
Fielding,A.H.,&Bell,J.F.(1997).Areviewofmethodsfortheassessment
of prediction errors in conservation presence/absence models.
Environmental Conservation, 24(1),38–49.htt ps://doi.org /10.1017/
S0376892997000088
Filatova,O.A .,Shpak,O.V.,Ivkovich,T.V.,Volkova, E.V.,Fedutin, I.D.,
Ovsyan ikova, E. N. , Burdin, A . M., & Hoy t, E. (2019). Lar ge-sc ale
habitatsegregationoffish-eatingandmammal-eatingkillerwhales

12 of 13
|
VACHON e t al.
(Orcinus orca)inthewesternNorthPacific.Polar Biology, 42(5),931–
941. h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / s 0 0 3 0 0 - 0 1 9 - 0 2 4 8 4 - 6
Fishlock,V.,Caldwell,C., &Lee,P.C.(2016).Elephantresource-usetra-
ditions. Animal Cognition, 19(2),429–433 .ht tps ://doi.org/10.1007/
s 1 0 0 7 1 - 0 1 5 - 0 9 2 1 - x
Gero, S. , Bøttche r,A ., Whiteh ead, H., & Ma dsen, P. T.(2016). So cially
segreg ated, sympatri c sperm whale cla ns in the Atlantic O cean.
Royal Societ y Open Science, 3(6),160 061.
Gero, S., Milligan, M., Rinaldi, C ., Francis, P., Gordon, J., Carlson, C.,
Steffen ,A., Tyack, P.,Evans,P.,& Whitehead, H. (2014). Behavior
andsocialstruc tureofthespermwhalesofDominica,WestIndies.
Marine Mammal Science, 30(3), 905–922. https: //doi.org/10 .1111/
mms.120 86
Gero, S., & Whitehead, H. (2016). Critical decline of the Eastern
Carib bean sperm w hale populat ion. PLoS One, 11(10), e0162019.
https://doi.org/10.1371/journal.pone.0162019
Gillespie, D., Mellinger, D., Gordon, J., McLaren , D., Redmond,
P., McHugh, R., Trinder, P., Deng, X., & Thode, A . (2009).
PAMGUARD: Semiautomated, op en source software for real-
time acoustic detection and localisation of cetaceans. The
Journal of the Acoustical Society of Am erica, 125, 2547. ht t ps : //d o i .
org /10.11 21/1.4808713
Harris, P. T., Macmillan-Lawler, M., Rupp, J., & Baker, E. K. (2014).
Geomorphologyoftheoceans.Marine Geology, 352, 4– 24. h t t ps : //
doi.org/10.1016/j.margeo.2014.01.011
Hastie , T.,T ibshirani , R., & Friedma n, J. (Eds.). (20 09). The elements of
statistical learning. Springer. https://doi.org/10.1007/978-0-
3 8 7 - 8 4 8 5 8 - 7 _ 1
Herbinger, I., Boesch, C., & Rothe, H. (2001). Territory characteris-
ticsamong threeneighboringchimpanzeecommunitiesintheTaï
National Park, Côte d'Ivoire. International Journal of Primatology,
22(2),143–167.https://doi.org/10.1023/A:10 0566 3212997
Hersh, T. A., Gero, S ., Rendell, L., & Whitehead, H. (2021).
Using identity calls to detect structure in acoustic dat aset.
Methods in Ecology and Evolution, 12, 1668–1678. ht t p s : //d oi .
org /10.1111/2041-210X .1364 4
Hoffm an, J. I., Trathan , P. N ., & Amos, W. (20 06). Geneti c tagging r e-
veals extreme site fidelity in territorial male Antarctic fur seals
Arctocephalus gazella. Molecular Ecology, 15(12), 38 41–3 847.
https ://doi.or g/10.1111/ j.1365-294X .2 006.03 05 3. x
Højsgaard,S .,Halekoh, U., &Yan, J.(2005). The RPackagegeepackfor
generalized estimating equations. Journal of Statistical S oftware,
15(1),1–11.https://doi.org/10.18637/jss.v015.i02
Huijser,L.A.E.,Estrade,V.,Webster,I.,Mouysset,L.,C adinouche,A .,&
Dulau-Drouot,V.(2020). Vocal reper toiresand insightsintosocial
struc ture of sperm whales (Physeter macrocephalus) in Mauriti us,
southwe stern Ind ian Ocean. Marine Mammal Science, 36( 2), 638–
657.http s://doi.o rg /10.1111 /mms.12673
Jaquet,N.(1996).Howspatialandtemporalscalesinfluenceunderstand-
ing of sper m whale dis tributi on: A review. Mammal Review, 26 (1), 
5 1 – 6 5 .  h tt ps ://doi.or g/10.1111/ j.1365-2907.1996 .tb0 0146 .x
Jaquet, N., & Whitehead, H. (1996). Scale-dependent correlation of
spermwhaledistributionwithenvironmentalfeaturesandproduc-
tivityin theSouth Pacific. Marine Ecology Progress Series, 135, 1– 9.
https://doi.org/10.3354/meps1 35001
Johns,W.E.,Townsend,T.L.,Fratantoni,D.M.,&Wilson,W.D.(2002).
On the Atl antic inflow to th e Caribbea n Sea. Deep Sea Research
Part I: Oceanographic Resea rch Papers, 49(2), 211–243.h t tp s : //d o i.
o r g / 1 0 . 1 0 1 6 / S 0 9 6 7 - 0 6 3 7 ( 0 1 ) 0 0 0 4 1 - 3
Konrad,C.M.,Frasier,T.R.,Rendell,L.,Whitehead,H.,&Gero,S.(2018).
Kinship and as sociation d o not explain vo cal reper toire vari ation
among individual sperm whales or social units. Animal Behaviour,
145, 131– 140. https://doi.org/10.1016/j.anbehav.2018.09.011
Kopps, A . M., Ackermann, C. Y., Sherwin, W.B ., Allen, S. J.,Bejder,L.,
&K rützen, M. (2014). Cultural transmission of tool usecombined
withhabitat specializationsleadsto fine-scalegeneticstructure in
bottlenose dolphins.Proceedings of the Royal Society B: Biological
Sciences, 281(1782),1–8.
Krützen,M.,Mann,J.,Heithaus,M.R.,Connor,R.C.,Bejder,L.,Sherwin,
W.B.,& Marler,P.(2005).Culturaltransmissionoftool useinbot-
tlenose dolphins. Proceedings of the National Academy of Science s of
the United States of America, 102(25),8939–8943.
Laurindo,L .C .,Mariano, A. J.,& Lumpkin,R.(2017).Animprovednear-
surfacevelocity climatologyfortheglobal oceanfromdrifterob-
servations. Deep Sea Re search Part I: Oce anographic Rese arch Papers,
124, 73– 92. https://doi.org/10.1016/j.dsr.2017.04.009
Liang, K .-Y.,& Zeger, S. L.(1986).Longitudinal data analysis using gen-
eralized linear models. Biometrika, 73(1), 13–22. h t t p s :// d oi .
org/10.2307/2336267
McComb , K., Moss, C ., Durant, S. M ., Baker, L., & Sayia lel, S. (2001).
Matriarchs as repositories of social knowledge in African elephants.
Science, 292(5516),491–494.
O'Neil, S. T., Vucetich, J. A., Beyer, D. E., Hoy, S. R., & Bump, J. K.
(2020). Territoriality drives preemptive habitat selection in re-
covering wolves: Implications for carnivore conservation.
The Journal of Animal Ecolog y, 89(6), 14 33 –14 47. h t tp s : //d o i.
org /10.1111/1365-265 6.13199
Pan, W. (2001). Akaike's information criterion in generalized es-
timating equations. Biometrics, 57(1), 120–125. h t tp s : //d o i .
org/10.1111/j.0006-341x.2001.00120.x
Pirotta, E., Matthiopoulos, J., MacKenzie, M., Scott-Hayward, L., &
Rendell, L. (2011). Modelling sperm whale habitat preference:
A novel approach combining transect and follow data . Marine
Ecology Progress Series, 436, 257– 272. https://doi.org/10.3354/
meps09236
Pirott a, E., Thompson , P. M., Miller, P. I., Brookes, K . L., Cheney, B.,
Barton,T.R.,Graham,I.M.,&Lusseau,D.(2014).Scale-dependent
foraging ecolog y of a marine top predator modelled using pas-
sive acous tic data. Functional Ecology, 28(1 ), 20 6–217. h t t ps : //d o i .
org /10.1111/1365-2435.12146
Redfern ,J.V.,Ferguson,M.C.,Becker,E.A.,Hyrenbach,K.D.,Good,C.,
Barlow,J.,Kaschner,K.,Baumgartner,M.F.,Forney,K.A.,Ballance,
L.T.,Fauchald,P.,Halpin, P.,Hamazaki, T.,Pershing,A .J.,Qian,S .
S.,Read,A.,Reilly,S.B.,Torres,L.,&Werner,F.(2006).Techniques
forcetacean–habitatmodeling.Marine Ecology Progress Series, 310,
271– 29 5. https://doi.org/10.3354/meps3 10271
Rendell,L.,Mesnick,S.L.,Dalebout,M.L.,Burtenshaw,J.,&Whitehead,
H. (2012). C an genetic d ifferen ces expla in vocal dial ect vari ation
in sperm whales, Physeter macrocephalus? Behavior Genetics, 42(2),
332– 3 43 . h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / s 1 0 5 1 9 - 0 1 1 - 9 5 1 3 - y
Rendell, L. E., & Whitehe ad, H. (2003). Vocal clans in sperm whales
(Physeter macrocephalus).Proceedings of th e Royal Society of London.
Series B: Biological Sciences, 270 (1512), 225–231. h t t ps : //d o i .
org /10.1098/rsp b.2002. 2239
Riesch, R., Barrett-Lennard, L. G ., Ellis, G. M., Ford, J. K. B ., &
Deecke, V. B. (2012). Cultur al traditions and the evolution of
reproductive isolation: Ecological speciation in killer whales?
Biological Journal of the Linnean Society, 106 (1), 1–17. ht t p s :// do i .
org /10.1111/j .1095-8 312 .2 012.0 1872 .x
Sing, T., Sander, O., Beerenwinkel, N., & Lengauer, T. (2005).
ROCR:VisualizingclassifierperformanceinR.Bioinformatics,
21(20),3940–3941.https://doi.org/10.1093/bioinformatics/
bti623
Tepsich,P.,Rosso,M.,Halpin,P.N.,&Moulins,A .(2014).Habitatprefer-
encesoftwodeep-divingcetaceanspeciesinthenorthernLigurian
Sea. Marine Ecology Progress Series, 508, 247–260. h t t p s :// do i .
org/10.3354/meps1 0851
Tynan,C.T.,Ainley,D.G.,Barth,J.A.,Cowles,T.J.,Pierce,S.D.,&Spear,
L.B.(2005).Cet acean distributions relativetooceanprocessesin
the nor thern California Current Syste m. Deep Sea Research Part

|
13 of 13
VACH ON e t al.
II: Topical Studies in Oceanography, 52(1), 145–167. h t t p s: //d o i .
org/10.1016/j.dsr2.2004.09.024
Vachon, F., Hersh, T. A., Rendell, L., Gero, S., & Whitehead, H. (2022).
Oceannomadsor Islandspecialists?Culturally drivenhabitatpar-
titioningcontrast sin scalebetween geographicallyisolatedsperm
whale populations. Royal Societ y Open Science, 9(5),211737.ht t p s : //
doi.org/10.1098/rsos.211737
Warner, R. R. (1988). Traditionality of mating-site preferences in
a coral reef fish. Nature, 335(6192), 719–721. ht t p s : //d o i.
org /10.103 8/335719a0
Weilgar t, L. S. (2007 ). The impact s of anthropogeni c ocean noise on
cetaceans and implications for management. Review, 85(11),1091–
1116.https://doi.org/10.1139/Z07-101
Whitehead, H. (2003). Sperm whales: Social evolution in the ocean.
UniversityofChicagoPress.
Whitehead, H., McGill, B., & Worm, B . (200 8). Diver sity of deep-
water cetaceans in relation to temper ature: Implications for
ocean war ming. Ecology Letters, 11(11), 1198–1207. ht t p s : //d oi .
org /10.1111/j .1461- 0248. 20 08 .0 123 4. x
Whitehead,H.,&Rendell,L.(20 04).Movements,habitatuseandfeeding
success of culturalclans ofSouth Pacific spermwhales. Journal of
Animal Ecology, 73(1),19 0–196.
Whitehead, H., & Rendell, L. (2015). The cultural lives of whales and dol-
phins.UniversityofChicagoPress.
Whitehe ad, H., & Weilgar t, L. (1991). Patterns of v isually obser vable
behaviour and vocalizations in groups of female sperm whales.
Behaviour, 118(3/4),275–296.
Whiten , A. (2017). A seco nd inherit ance syste m: The exten sion of bi-
ologythroughculture.Inter face Focus, 7(5), 20160142.h t t p s :// d oi .
org /10.1098/rsfs. 2016 .0142
Wong, S. N. P.,& Whitehead,H. (2014).Seasonal occurrence of sperm
whales (Physeter macrocephalus) around Kelvin Seamount in the
Sargasso Sea in relation to oceanographic processes. Deep Sea
Research Part I: Oceanographic Research Papers, 91, 10–16.h t t p s ://
doi.org/10.1016/j.dsr.2014.05.001
SUPPORTING INFORMATION
Additional supporting information can be found online in the
Suppor tingInformationsectionattheendofthisarticle.
How to cite this article: Vachon, F., Eguiguren, A., Rendell, L.,
Gero,S.,&Whitehead,H.(2022).Distinctive,fine-scale
distributionofEasternCaribbeanspermwhalevocalclans
reflectsislandfidelityratherthanenvironmentalvariables.
Ecology and Evolution, 12, e94 49. https://doi.org/10.1002/
ece3 .9449