In eg a ing imme sion wi h GPS da a imp o es
beha iou al classi ica ion o wande ing alba osses and
shows sca enging behind ishing essels mi o s na u al
o aging
A P B Ca nei o
1
,MPDias
1,2
, S Oppel
3
, E J Pea main
1
, B L Cla k
1
, A G Wood
4
, T Cla elle
5
&
R A Phillips
4
1 Bi dLi e In e na ional, Camb idge, UK
2 Cen e o Ecology, E olu ion and En i onmen al Changes, cE3c & Depa men o Animal Biology, Faculdade de Ci^
encias, Uni e sidade
de Lisboa, Lisbon, Po ugal
3 Royal Socie y o he P o ec ion o Bi ds, The Da id A enbo ough Building, Camb idge, UK
4 B i ish An a c ic Su ey, Na u al En i onmen Resea ch Council, Camb idge, UK
5 Global Fishing Wa ch, Washing on, Dis ic o Columbia, USA
Keywo ds
ine-scale; ishe ies; hidden Ma ko models;
imme sion; ada ; andom o es models;
seabi ds; essel.
Co espondence
Ana P. B. Ca nei o, Bi dLi e In e na ional,
The Da id A enbo ough Building, Pemb oke
S ee , Camb idge CB2 3QZ, UK.
Email: [email p o ec ed]
Richa d A. Phillips, B i ish An a c ic Su ey,
Na u al En i onmen Resea ch Council, High
C oss, Madingley Road, Camb idge CB3
0ET, UK.
Email: [email p o ec ed]
Edi o : Ka l E ans
Associa e Edi o : Jaime Ramos
Recei ed 13 June 2021; accep ed 04
Janua y 2022
doi:10.1111/ac .12768
Abs ac
Ad ances in biologging echniques and he a ailabili y o high- esolu ion fishe ies
da a ha e imp o ed ou abili y o unde s and he in e ac ions be ween seabi ds and
fishe ies and o e alua e mo ali y isk due o byca ch. Howe e , i emains unclea
whe he mo emen pa e ns and beha iou di e be ween bi ds o aging na u ally
o sca enging behind essels and whe he his could be diagnos ic o fishe ies
in e ac ions. We deployed no el logge s ha eco d he GPS posi ion o bi ds a
sea and scan he su oundings o de ec ada ansmissions om essels and
imme sion (ac i i y) logge s on wande ing alba osses Diomedea exulans om
Sou h Geo gia. We ma ched hese da a o emo ely sensed fishing essel posi ions
and used a combina ion o hidden Ma ko and andom o es models o in es iga e
whe he i was possible o de ec a cha ac e is ic signa u e om he seabi d ack-
ing and ac i i y da a ha would indica e fine-scale essel o e lap and in e ac ions.
Including imme sion da a in ou hidden Ma ko models allowed wo dis inc o ag-
ing beha iou s o be iden ified, bo h indica i e o A ea Res ic ed Sea ch (ARS)
bu wi h o wi hou landing beha iou (likely p ey cap u e a emp s) ha would no
be de ec able wi h loca ion da a alone. Bi ds app oached essels du ing all beha-
iou al s a es, and he e was no clea pa e n associa ed wi h his ype o sca eng-
ing beha iou . The andom o es models had e y low sensi i i y, pa ly because
o aging e en s a essels occu ed e y a ely, and did no con ain any diagnos ic
mo emen o ac i i y pa e n ha was dis inc om na u al beha iou s away om
essels. Thus, we we e unable o p edic accu a ely whe he o aging bou s
occu ed in he icini y o a fishing essel, o na u ally, based on beha iou alone.
Ou me hod p o ides a cohe en and gene alizable amewo k o segmen ips
using auxilia y biologging (imme sion) da a and o efine he classifica ion o o -
aging s a egies o seabi ds. These esul s ne e heless unde line he alue o using
ada de ec o s ha de ec essel p oximi y o emo ely sensed essel loca ions o
a be e unde s anding o seabi d–fishe y in e ac ions.
In oduc ion
Inciden al mo ali y (byca ch) in fishe ies is one o he majo
h ea s o seabi d popula ions wo ldwide and pa icula ly o
alba osses and pe els (Phillips e al., 2016; Dias e al.,
2019). Ad ances in biologging echniques and he
a ailabili y o high- esolu ion fishe ies da a h ough essel
moni o ing sys ems (VMS) and he au oma ic iden ifica ion
sys em (AIS) ha e ad anced ou unde s anding o in e ac-
ions be ween seabi ds and essels and associa ed mo ali y
isks (Vo ie e al., 2010; G anadei o e al., 2011; To es
e al., 2013). Howe e , he e is s ill an u gen need o be e
Animal Conse a ion  (2022) – ª2022 The Au ho s. Animal Conse a ion published by John Wiley & Sons L d on behal o Zoological Socie y o London. 1
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion-NonComme cial-NoDe i s License, which pe mi s use and dis ibu ion in any
medium, p o ided he o iginal wo k is p ope ly ci ed, he use is non-comme cial and no modi ica ions o adap a ions a e made.
Animal Conse a ion. P in ISSN 1367-9430
quan i a i e p edic ions o he isk o seabi ds om fishe y
in e ac ions a fine scales, and no s udies o da e ha e de el-
oped models based on biologging da a alone ha p edic
eeding on disca ds, bai ed hooks o o he an h opogenic
esou ces om essels. Such models could ep esen a majo
s ep-change in his esea ch field, conside ing he apid
inc ease in he a ailabili y o biologging da a o seabi ds
(Be na d e al., 2021).
Ma ine op p eda o s, including seabi ds ha usually eed
on na u ally agg ega ed p ey, will o en display A ea
Res ic ed Sea ch (ARS) beha iou when hey encoun e
p ofi able pa ches. This beha iou al model is cha ac e ised
by inc eased sinuosi y and slowe speed (Fauchald & T e aa,
2003; Weime ski ch e al., 2007; Pi o a e al., 2018) and
inc eased esidence ime in p oduc i e a eas (Weime ski ch,
Gaul , & Che el, 2005; Ga he e al., 2016). Seabi ds a e
also known o change hei fine-scale mo emen s when hey
in e ac wi h fishing essels, which also esul s in ARS beha-
iou (To es e al., 2011; Bodey e al., 2014; Co beau e al.,
2019; G 
emille e al., 2019). To disc imina e be ween na u-
al o aging and a ge ing o essels equi es independen ,
fine-scale da a on seabi d mo emen s and he loca ions o
indi idual essels. The la e , howe e , a e gene ally a ail-
able only wi hin Exclusi e Economic Zones (EEZs). Hence,
mos analyses o seabi d–fishe ies o e lap o da e ha e used
agg ega ed da a on fishing e o , such as o al longline
hooks deployed pe 5°g id cell, by mon h, which is o en
he highes esolu ion a ailable om he Regional Fishe ies
Managemen O ganiza ions (RFMOs) which manage fishing
in he High Seas (Tuck e al., 2011; Small, Waugh, & Phil-
lips, 2013; Clay e al., 2019; Ca nei o e al., 2020).
When da a a e a ailable on bi d and essel mo emen s, a
common app oach o y o dis inguish na u al o aging om
essel in e ac ions is o iden i y pe iods o ac i e o aging
wi hin acks (based on fi s -passage ime o speed- o uosi y
h esholds), o e lap hose in space and ime wi h essels,
and di e en ia e na u al o aging om essel in e ac ion
based on he p esence o absence o nea by fishing essels
(Vo ie e al., 2010; To es e al., 2011; Co beau e al., 2019;
Co beau, Colle , Pajo , e al., 2021b). A second app oach
ocuses on iden i ying beha iou al changes in he icini y o
afishing essel by examining he p obabili y o swi ching
om one beha iou al s a e o ano he (e.g. edi ec fligh a-
jec o ies when in he icini y o a fishing essel; Bodey
e al., 2014; Colle , Pa ick, & Weime ski ch, 2015;
Cianche i-Benede i e al., 2018; Le Bo , Lesc o€
el, &
G 
emille , 2018; Cla k e al., 2020b). Al hough hese s udies
demons a ed ha essels could a ec he mo emen s and
beha iou o seabi ds, i emains unclea whe he he e is a
dis inguishable mo emen pa e n, su ficien o p edic i e
disc imina ion, which is diagnos ic o whe he he bi d is
ei he o aging na u ally o sca enging behind a essel. An
app oach ha o e s conside able p omise is he use o hid-
den Ma ko models (HMMs), which a e used inc easingly
o cha ac e ising animal beha iou (McClin ock & Michelo ,
2018). In seabi d ecology, HMMs ha e been used almos
exclusi ely on loca ion da a o dis inguish be ween ARS and
ansi mo emen s, howe e , hey can also inco po a e
auxilia y biologging and en i onmen al da a (Leos-Ba ajas
e al., 2017; Pa e son e al., 2019; Clay e al., 2020; Con-
ne s e al., 2021). In pa icula , addi ional senso s could help
dis inguish sea ching beha iou om landings o cap u e
p ey, he la e po en ially signalling a byca ch isk i hey
ake place nea essels and in ol e a ge ing o longline
bai s du ing se ing o hauling. One po en ial applica ion o
hese models is o iden i y cha ac e is ic essel- ollowing
beha iou om he complex pa e ns o ac i i y o seabi ds
a sea ha could imp o e p edic ions o seabi d–fishe y in e -
ac ions in he absence o independen in o ma ion on essel
loca ions.
Wande ing alba osses Diomedea exulans a Sou h Geo gia
ha e declined ca as ophically since he 1970s (Ponce e al.,
2017), leading o he de elopmen o a conse a ion Ac ion
Plan by he Go e nmen o Sou h Geo gia and he Sou h Sand-
wich Islands (GSGSSI), and hei lis ing as a P io i y Popula-
ion by he Ag eemen on he Conse a ion o Alba osses and
Pe els (ACAP). Byca ch in fishe ies is conside ed o be hei
main h ea (Pa do e al., 2017; Ponce e al., 2017). Analyses
o he spa ial o e lap a la ge scales be ween hei a -sea dis i-
bu ion and pelagic and deme sal longline fishe ies indica ed
ha he popula ion was a highes isk in he B azil-Falklands
confluence zone, pa icula ly om he Japanese and Taiwanese
dis an -wa e una flee s (Jim
enez e al., 2016; Clay e al.,
2019). Howe e , o e lap me ics a e scale dependen and he
assump ion ha co-occu ence o seabi ds and fishe ies in a
egion leads o in e ac ion and mo ali y isk has been egula ly
highligh ed as a po en ial pi all (To es e al., 2013; Weime -
ski ch e al., 2020; Co beau, Colle , O ge e , e al., 2021a)
because he e is a lack o fine-scale da a simul aneously a ail-
able o bo h bi ds and essels. As such, a ho ough unde -
s anding o seabi d–fishe y in e ac ions a much fine scales is
highly ele an o conse a ion and o be e a ge manage-
men ac ions.
He e we used ecen ly de eloped logge s ha eco d he
GPS posi ion o bi ds a sea and egula ly scan he su ound-
ings o de ec he p esence o ada ansmissions om essels
(Weime ski ch e al., 2020). These da a may allow ecognis-
able beha iou s o be delinea ed om he mo emen s o he
bi d ha can be ma ched o he p oximi y o fishing essels a
ha ime. Combining hese da a wi h imme sion (ac i i y) da a
om geoloca o -imme sion de ices (Phalan e al., 2007;
Mackley e al., 2010; G anadei o e al., 2011; Dean e al.,
2013), we used hidden Ma ko models o cha ac e ise o ag-
ing beha iou (including landings) in mo e de ail han a
loca ion-only model (Cla k, Handby, e al., 2020a). We
ma ched hese da a wi h he posi ion o indi idual essels
ob ained om he au oma ic iden ifica ion sys em (AIS) and
used andom o es models o in es iga e whe he wande ing
alba osses b eeding a Sou h Geo gia exhibi clea ly iden ifi-
able pa e ns o mo emen and beha iou ha can be associ-
a ed wi h ei he na u al o aging o in e ac ions wi h essels.
I diagnos ic, such pa e ns could hen be applied o exis ing,
ex ensi e bi d- acking da ase s o quan i y he equency,
du a ion and p opensi y o indi idual bi ds, sexes and li e-
his o y s ages o in e ac wi h fishing essels, and he e o e
mo e accu a ely assess hei isk o byca ch.
2Animal Conse a ion  (2022) – ª2022 The Au ho s. Animal Conse a ion published by John Wiley & Sons L d on behal o Zoological Socie y o London.
Sca enging behind essels mi o s na u al o aging in alba osses A. P. B. Ca nei o e al.
Ma e ials and me hods
Da a collec ion
We used Tesa ape o a ach 85 GPS- ada logge s (XSpu -
nik; Sex an Technology) o he man le ea he s o b eeding
adul wande ing alba osses a Bi d Island, Sou h Geo gia
(54°00’S, 38°03’W) and le hem a ached o one o mo e
o aging ips du ing incuba ion (Janua y–Feb ua y), b ood-
gua d (Ma ch–Ap il) and pos -gua d chick- ea ing (July–
Augus ) pe iods in 2020. Logge s eco ded a GPS posi ion
e e y 10 min and es ed o ada ansmissions (wi hin
5 km; Weime ski ch e al., 2018) o 5 min. The same bi ds
we e also equipped wi h a geoloca o -imme sion logge (In i-
geo C-330; Mig a e Technology), which eco ded sal -wa e
imme sion e en s ha las ed ≥3 s, p o iding in o ma ion on
ac i i y pa e ns ( iming and du a ion o fligh s and wa e
landings) h oughou he o aging ip. Imme sion logge s
we e a ached o a plas ic ing on he a sus. Bi ds we e cap-
u ed on he nes du ing changeo e wi h he pa ne (incu-
ba ion o b ood-gua d) o a e eeding he chick (pos -
gua d). A achmen o he de ices always ook less han
10 min, and bi ds ypically le he colony sho ly a e being
eleased o o age a sea. To acili a e logge e ie al du ing
pos -gua d chick- ea ing, a ence was buil a ound he nes
on he 3 d day a e he adul depa ed, and nes s we e is-
i ed wice a day he ea e un il he adul e u ned and
de ices we e e ie ed (Xa ie e al., 2003). I he pa ne
a i ed be o e he ins umen ed bi d e u ned, he ence was
opened empo a ily o allow i o deli e he meal o he
chick. To al ins umen load was c. 55 g (0.6 and 0.7 o
male and emale mean adul body mass, espec i ely), which
was well below he h eshold o 3% a which de ice e ec s
end o become appa en (Phillips, Xa ie , & C oxall, 2003).
T acked bi ds comp ised an e en sp ead o sexes and ages
(9–44 yea s). B eeding success o bi ds fi ed wi h de ices
was ex emely high (97%) because we a ge ed expe ienced
b eede s (which ha e highe success han new ec ui s; F oy
e al. 2013) and did no deploy un il a leas he end o he
second incuba ion s in , a oiding he ini ial pe iod o highe
ailu e pos -laying. As such, he e was no e idence ha he
deploymen s had dele e ious e ec s.
Da a p ocessing
GPS da a we e fi s fil e ed o emo e loca ions a he nes
and o aging ips we e defined as he loca ion be ween he
las GPS fix in he colony p io o depa u e and he fi s
GPS fix a e a e u n. Un ealis ic loca ions in ol ing a el
speeds abo e 120 km h
1
we e emo ed using he SDLfil e
R package (Shimada e al., 2012). Posi ions we e linea ly
in e pola ed o a 10-min sampling equency o egula ise he
da a and fill in occasional gaps. The imme sion da a we e
summa ised as he numbe o landings ( he o al numbe o
d y–we ansi ions) and he p opo ion o ime spen on he
wa e su ace (we ) in he 10-min in e al p eceding each
GPS loca ion. The numbe o landings and he p opo ion o
ime spen on he wa e su ace we e hen ma ched
empo ally wi h GPS da a. Alba osses ha e low cos s o
fligh and landings, bu ake-o s om he wa e su ace
(we –d y ansi ions) in ol e high ene ge ic cos ; bi ds will
only land o eed o o es , and he e o e, he landing a e
p o ides a good indica ion o o aging e o (Weime ski ch
e al., 2000; Phalan e al., 2007; G anadei o e al., 2011).
Loca ions o indi idual fishing essels wi hin he s udy a ea
we e ob ained om Global Fishing Wa ch (GFW), which com-
bine public essel egis ies and machine-lea ning models o (i)
iden i y fishing essels in he AIS da a and (ii) de ec when
hey a e ac i ely fishing, wi h a fishing de ec ion accu acy o
>90% (K oodsma e al., 2018). Fo each essel (AIS) loca ion,
he ollowing in o ma ion was a ailable: unique essel iden i-
fie , da e, ime, la i ude, longi ude and fishing sco e (i.e. he
likelihood o fishing om he GFW fishing de ec ion model).
All essel loca ions wi hin 5 km and 5 min o each in e po-
la ed loca ion o he acked alba osses we e ex ac ed, bu
only essels classified as ‘ac i ely fishing’we e included in he
analysis. The empo al esolu ion o AIS da a associa ed wi h
bi d loca ions wi hin he s udy a ea was 7.8 101.9 min
(mean SD). P elimina y analyses indica ed ha including
he ype o fishing ac i i y ( awling, long-lining e c.) had no
e ec on he models, and hence he esul s p esen ed he e a e
om all ypes o fishing ac i i y pooled.
Beha iou al s a e classi ica ion
HMMs a e ypically used o iden i y beha iou al s a es om
animal mo emen based on s ep leng h and u ning angle
be ween subsequen loca ions, wi h sho -medium s ep
leng hs and high u ning angles conside ed o ep esen o -
aging (McClin ock & Michelo , 2018; Conne s e al., 2021).
The addi ion o o he da a p o ided by he concu en
deploymen o addi ional senso s may p o ide inc eased
powe o esol e beha iou al s a es (Dean e al., 2013;
McClin ock & Michelo , 2018; Cla k, Handby, e al., 2020).
We he e o e combined s ep leng hs and u ning angles wi h
he numbe o landings and p opo ion o ime spen on he
wa e su ace (ex ac ed om he imme sion da a) o cha ac-
e ise he p incipal ypes o a -sea beha iou o wande ing
alba osses using he momen uHMM package in R (McClin-
ock & Michelo , 2018). We modelled he beha iou o all
indi iduals combined, as he inclusion o indi idual e ec s
in he hidden s a e p ocess makes li le di e ence in e ms o
in e ence (McClin ock, 2021).
HMMs equi e he use o define s a e-dependen dis ibu-
ion classes and o p o ide s a ing alues o acili a e pa am-
e e es ima ion. The la e we e selec ed using k-means
clus e ing (wi h k=numbe o s a es [2–5]; Dean e al.,
2013; Cla k, Handby, e al., 2020a), wi h he excep ion o
he p opo ion o ime on he wa e su ace, which was con-
e ed in o a ca ego ical a iable ep esen ing low, in e medi-
a e and high p obabili y using s a ing pa ame e s based on
ob ious b eaks in he equency his og am (Michelo & Lan-
g ock, 2019). A gamma dis ibu ion was chosen o s ep
leng hs, w apped Cauchy o u ning angles, Poisson o
numbe o landings and ca ego ical o he p opo ion o
ime on he wa e su ace. We compa ed he fi be ween
Animal Conse a ion  (2022) – ª2022 The Au ho s. Animal Conse a ion published by John Wiley & Sons L d on behal o Zoological Socie y o London. 3
A. P. B. Ca nei o e al. Sca enging behind essels mi o s na u al o aging in alba osses
simple ( ew s a es) and complex (many s a es) models based
on he Akaike In o ma ion C i e ion (AIC) and used he
elbow c i e ion (i.e. he poin a which adding addi ional
s a es no longe esul s in subs an ial educ ions in AIC) o
selec he bes ade-o be ween model accu acy and com-
plexi y (Dean e al., 2013; Cla k, Handby, e al., 2020a).
The Vi e bi algo i hm was hen used o es ima e he mos
likely sequence o unde lying s a es om he selec ed model
o each ack (McClin ock & Michelo , 2018). An expe -
d i en app oach and anima ions o he ips we e used o
de e mine i he models we e assigning app op ia e
sequences o beha iou s, ha is o a andom selec ion o
ips, we compa ed beha iou s a es assigned by he models
and hose om isualisa ion (Clay e al., 2020).
Beha iou al bou iden i ica ion
We used he esul s o he HMM o iden i y beha iou al bou s
as sequences o posi ions in he same beha iou al s a e (see
Resul s). In o de o smoo h he da a, bou s ha consis ed o a
single loca ion we e eassigned o he same HMM s a e as he
p e ious and subsequen loca ion i he HMM s a es o he p e-
ious and subsequen loca ions we e iden ical. Wande ing alba-
osses a end fishing essels o se e al hou s (Weime ski ch
e al., 2020); he e o e, we excluded beha iou al bou s o sho
du a ion (<5 loca ions) om he analyses ( his only accoun ed
o 8% o he da a), as hese we e likely o indica e ansi
h ough he same a ea as essels, and no an in e ac ion. We
hen cha ac e ised each beha iou al bou using se e al me ics,
including he ype o bou (based on HMM classifica ion), he
mean, maximum and minimum o e all speeds, du a ion, cumu-
la i e and maximum dis ances a elled and s aigh ness o
mo emen (a measu e o maximum dis ance di ided by cumu-
la i e dis ance). Based on he majo i y o ime spen , he bou
was assigned o dayligh (including wiligh ) o da kness,
acco ding o he iming o ci il wiligh (when he sun is 6°
below he ho izon) ha we e p e iously assigned o each GPS
posi ion. We calcula ed he o al numbe o landings and he
p opo ion o loca ions wi h landing a emp s, and he p opo -
ion o ime spen on he wa e , o he du a ion o he bou .
We also ex ac ed om GFW whe he any o he essels wi hin
5 km we e ac i ely fishing a any ime du ing he bou .
Disc imina ing be ween na u al o aging
s. o aging behind essels using machine
lea ning
We used a machine-lea ning algo i hm o es whe he we
could accu a ely p edic i a bi d was o aging in he icini y
o a essel using he esul s o he HMM analysis and he
beha iou al bou classifica ion. Ou esponse a iable was
bina y and indica ed whe he each bou coincided wi h de ec-
ion o ada om a essel and i ha essel was ac i ely fish-
ing wi hin he bou (de ec ed by AIS). Fo his, we calcula ed
he p opo ion o GPS loca ions in each bou ha coincided
wi h ada de ec ion and he p opo ion o loca ions in each
bou ha we e classified as ac i ely fishing. We hen used a -
ious h esholds o de e mine whe he a beha iou al bou was
associa ed wi h an ac i ely fishing essel o o e come he
unce ain y inhe en in bo h he ada de ec ion and AIS clas-
sifica ions. Nine y-one pe cen o bou s wi h ada de ec ions
we e also associa ed wi h AIS da a, whe eas only 49% o
bou s wi h ac i ely fishing essels in e ed om AIS we e
associa ed wi h ada de ec ion. This is simila o esul s in a
p e ious s udy using he same echnology, which indica ed
ha 46.6% o AIS loca ions wi hin 5 km o bi ds esul ed in
ada de ec ion (Weime ski ch e al., 2020). We conside ed a
bou o be indica i e o a ending a fishing essel a di e en
h esholds; hese we e when 0%, 5%, 10%, 20% o 50% o
loca ions in ha bou we e associa ed wi h ada de ec ion,
and when 0%, 5%, 10%, 20% o 50% o loca ions in ha
bou we e associa ed wi h a nea by essel classified as
ac i ely fishing. We epo he p e alence (p opo ion o bou s
classified as ‘a ending a fishing essel’) o each combina ion
o h esholds in TABLE 2. We hen fi ed each andom o es
model based on a condi ional in e ence amewo k accoun ing
o co ela ed p edic o s (Ho ho n, Ho nik, & Zeileis, 2006a)
o es whe he he beha iou al s a es de i ed om he HMM
had p edic i e powe o dis inguish be ween bou s wi h and
wi hou nea by essels ha we e ac i ely fishing. We fi ed
his model in a bina y classifica ion amewo k wi h he R
package pa y (Ho ho n, Ho nik, & Zeileis, 2006b). We speci-
fied an in e nal c oss- alida ion s uc u e o ensu e ha da a
om he same indi iduals we e used ei he o fi ing o e al-
ua ing ees in he o es , which emula es a andom e ec in
linea models and accoun s o he se ial au oco ela ion o
bou s pe o med by he same indi idual (Bus on & Eli h,
2011). We used a andom subse o 65% o da a wi hou
eplacemen o build single ees and alida ed ou model by
applying he ou pu o he emaining da a o es ima e he
accu acy o p edic ions. We p esen he p edic i e accu acy o
each andom o es model as he p opo ion o c oss- alida ed
bou s ha we e co ec ly p edic ed and he sensi i i y o he
andom o es models as he p opo ion o bou s wi h ac i ely
fishing nea by essels ha we e co ec ly p edic ed.
The impo ance o a iables was calcula ed using a pe -
mu a ion p ocedu e ha assesses he loss in model p edic i e
accu acy (S obl e al., 2008; Jani za, S obl, & Boules eix,
2013; Hap elmeie e al., 2014). Fo easie in e p e a ion, he
a iable impo ance was s anda dised, wi h he mos impo -
an a iable assigned a ela i e impo ance o 100% (Oppel,
Powell, & Dickson, 2009; Oppel e al., 2017). I beha iou al
me ics de i ed om acking da a ha e su ficien p edic i e
in o ma ion o dis inguish o aging behind essels in sea-
bi ds, we would expec he sensi i i y o he model o be
high (>75% co ec classifica ions). All analyses we e pe -
o med in R 3.6.3 (R Co e Team, 2019). Unless indica ed
o he wise, all da a a e p esen ed as means SD.
Resul s
Fo aging ip cha ac e is ics
We acked 30, 28 and 27 indi idual wande ing alba osses
du ing incuba ion, b ood-gua d and pos -gua d chick- ea ing,
espec i ely (Fig. 1). GPS and ada da a we e incomple e
4Animal Conse a ion  (2022) – ª2022 The Au ho s. Animal Conse a ion published by John Wiley & Sons L d on behal o Zoological Socie y o London.
Sca enging behind essels mi o s na u al o aging in alba osses A. P. B. Ca nei o e al.
o a ew ips because o ba e y deple ion (one in incuba-
ion, h ee in b ood-gua d and six in pos -gua d chick-
ea ing). Some indi iduals we e acked o mul iple ips in
b ood-gua d and pos -gua d chick- ea ing. O e all, we
ob ained a o al o 29, 29 and 23 comple e ips ha las ed
o 12.6 6.4 days du ing incuba ion ( ange 3–39 days),
3.2 1.8 du ing b ood-gua d ( ange 2–9 days) and
8.0 6.2 days du ing pos -gua d chick- ea ing ( ange 1–
24 days), espec i ely. O he 85 indi iduals acked, de ices
on 45 bi ds (53%) de ec ed essel ada du ing a leas one
ip, and 26 (58%) o hese bi ds encoun e ed ac i ely fishing
essels (Fig. 1). Vessel ada was no de ec ed by de ices on
wel e bi ds ha encoun e ed ac i ely fishing essels acco d-
ing o he AIS da a.
Beha iou al models
The inspec ion o AIC o candida e HMMs wi h 2–5 s a es
sugges ed ha a ou -s a e HMM adequa ely desc ibed he
dominan beha iou al pa e ns o wande ing alba osses a
sea (Addi ional File 1). The ou beha iou al s a es consis ed
o a a elling s a e cha ac e ised by di ec ed fligh wi h high
speeds and low a iance in u ning angles, ew landings and
no imme sion; a es ing s a e wi h e y slow a el speeds,
in e media e a iance in u ning angles, no landings and
always we ; and wo ARS s a es cha ac e ised by slow
speeds and high a iance in u ning angles, ei he wi hou
imme sion (ARS wi hou landings) o a leas one landing
(ARS wi h landings) (Fig. 2; TABLE 1). We classified a
o al o 4349 bou s o consis en beha iou om he 85
acked indi iduals. The du a ion o bou s o ARS wi h land-
ings was c. h ee imes sho e han hose o ARS wi hou
landings. On a e age, he p opo ion o o aging ips ha
was spen a elling, es ing, in ARS wi hou landings and
ARS wi h landings was 40%, 23%, 23% and 14%, espec-
i ely (TABLE 1).
Na u al o aging s. o aging behind
essels
Vessel encoun e s ( ada de ec ions and AIS loca ions om
ac i ely fishing essels) occu ed du ing all beha iou al
s a es (Fig. 3), wi h a highe numbe o ARS wi h landings
bou s associa ed wi h essel encoun e s han expec ed
(TABLE 1;
2
=98.404; p<0.001). Howe e , despi e he
unequal equency o essel encoun e s among di e en
beha iou al bou s, we ound no p edic able pa e n associa ed
wi h o aging behind essels because he dis ibu ions o a-
el speeds, u ning angles, landings and imme sion alues
o e lapped when compa ing bou s o beha iou al s a es wi h
and wi hou essel encoun e s (Fig. 3). The use o me ics
ex ac ed om combined GPS and imme sion da a in he
HMMs did no e eal a s a e indica i e o o aging behind
essels (as indica ed by ada and AIS loca ions om
ac i ely fishing essels). Ou esul s showed a highe pe -
cen age o bou s associa ed wi h essel encoun e s o he
ARS wi h landings s a e (30.8%) when compa ed o he
ARS wi hou landings s a e (9.6%); none heless, he majo i y
o bou s in he ARS wi h landings s a e we e no associa ed
wi h essels, sugges ing ha his s a e mos ly eflec ed na u-
al o aging beha iou .
Figu e 1 Fo aging ips o wande ing alba osses acked om Bi d
Island, Sou h Geo gia, du ing incuba ion, b ood-gua d and pos -
gua d chick- ea ing in 2020, o e laid on ba hyme y. Red do s ep-
esen loca ions whe e alba osses encoun e ed essels, ei he
de ec ed by bi d-bo ne ada o om sa elli e AIS da a indica ing
ac i ely ishing essels.
Animal Conse a ion  (2022) – ª2022 The Au ho s. Animal Conse a ion published by John Wiley & Sons L d on behal o Zoological Socie y o London. 5
A. P. B. Ca nei o e al. Sca enging behind essels mi o s na u al o aging in alba osses

The andom o es models had collec i ely e y low sensi-
i i y and we e unable o iden i y bou s associa ed wi h nea by,
ac i ely fishing essels ega dless o he h esholds used o he
p opo ion o ada de ec ions and he p opo ion o loca ions
o which a nea by essel was classified as ac i ely fishing
(TABLE 2). Howe e , e en unde he lowes h esholds, an
ac i ely fishing essel was nea by only o abou 7.4% o he
4349 bou s included in he analysis (TABLE 2); his low p e a-
lence esul ed in an o e all high p edic i e accu acy because a
model p edic ing he e was no fishing essel nea by would co -
ec ly classi y >95% o bou s. Thus, we we e unable o accu-
a ely p edic whe he o aging bou s occu in he icini y o a
fishing essel, o na u ally, based on mo emen me ics.
Discussion
This s udy in es iga ed whe he i is possible o de ec a
cha ac e is ic signa u e om mo emen and ac i i y da a ha
would indica e he in e ac ion o wande ing alba osses wi h
fishing essels, a he han na u al o aging. Al hough in e-
g a ing imme sion and GPS da a allowed us o iden i y wo
di e en ARS- ype beha iou s, nei he could unambiguously
indica e essel ollowing and associa ed sca enging in his
species. Ou esul s sugges ha wande ing alba osses
app oach ac i ely fishing essels in equen ly, and hei
mo emen and ac i i y me ics du ing hose encoun e s a e
no undamen ally di e en om hei b oad ange o na u al
o aging beha iou s du ing di e en s ages o he b eeding
season (Phalan e al., 2007; F oy e al., 2015; Jim
enez e al.,
2016). We he e o e conclude ha i is cu en ly no possible
o in e essel a endance and hus byca ch isk om seabi d
mo emen and ac i i y da a alone using he me hods p e-
sen ed he e. Rega dless, bi d-bo ne ada logge s clea ly
imp o e ou abili y o in e in e ac ions (Weime ski ch e al.,
2018, 2020; Co beau e al., 2019; G 
emille e al., 2019; his
s udy).
Figu e 2 Example o one o aging ip o a wande ing alba oss acked om Bi d Island, Sou h Geo gia, du ing incuba ion in 2020, wi h eg-
ula ised GPS loca ions colou ed by beha iou s de i ed om he 4-s a e hidden Ma ko model (HMM).
Table 1 Pa ame e es ima es o he s a e-dependen p obabili y dis ibu ions om he ou -s a e hidden Ma ko model (HMM) o acked
wande ing alba osses om Sou h Geo gia based on GPS and imme sion (ac i i y) da a. The able also includes he p opo ion o ime spen
in each s a e (% loca ions) and mean SD o bou du a ions and he p opo ion o bou s wi h essel encoun e s ( ada de ec ions and AIS
loca ions om ac i ely ishing essels).
Beha iou al s a e
Bou du a ion
(h s)
% Bou s wi h
essel encoun e Speed (km/h)
Angle
conc.
N
landings
P obabili y o we %
Loca ions[0–0.2] [0.2–0.8] [0.8–1]
T a el 3.31 3.05 3.87 44.68 15.5 0.92 0.02 1 0 0 40
Res 3.41 3.01 12.93 1.59 1.08 0.8 0 0 0 1 23
ARS w/o landings 4.48 5.91 9.60 5.63 6.10 0.77 0 1 0 0 23
ARS wi h landings 1.55 1.22 30.80 13.80 16.19 0.72 2.62 0.12 0.36 0.52 14
T a el speeds a e de i ed om s ep leng hs and do no ep esen ligh (g ound) speeds o bi ds in he espec i e beha iou al s a es. Tu n-
ing angle is he angle concen a ion pa ame e (highe alues indica e less a iance and less con olu ed acks). The numbe o landings ep-
esen s he o al numbe o d y–we ansi ions. The p obabili y o we indica es he p obabili ies om low (0) o high (1) ha ime was
spen on he sea su ace.
6Animal Conse a ion  (2022) – ª2022 The Au ho s. Animal Conse a ion published by John Wiley & Sons L d on behal o Zoological Socie y o London.
Sca enging behind essels mi o s na u al o aging in alba osses A. P. B. Ca nei o e al.
Al hough seabi d–fishe ies in e ac ions ha e been he
ocus o many s udies wo ldwide (Vo ie e al., 2010; G ana-
dei o e al., 2011; Pa ick e al., 2015; Weime ski ch e al.,
2020), as a as we a e awa e, p e ious s udies ha e no
iden ified a diagnos ic pa e n o mo emen associa ed wi h
sca enging behind essels ha would allow o p edic ions.
Indeed, he ew s udies explo ing his opic ha e ound con-
as ing esul s. G anadei o e al. (2011) ound no di e ence
in mo emen o a -sea ac i i y pa e ns o black-b owed alba-
osses Thalassa che melanoph is when o aging na u ally o
in close p oximi y o awl essels a ound he Falkland
Islands. In con as , To es e al. (2011) ound ha whi e-
capped alba osses T. s eadi om he Auckland Islands
mo ed in s aigh e pa hs and a slowe speeds when ollow-
ing a squid awle han when o aging na u ally. Simila ly,
wande ing alba osses om he C oze Islands mo ed wi h
g ea e sinuosi y in he p esence o longline fishing essels
(Weime ski ch e al., 2018; Co beau e al., 2019). Howe e ,
e en hough mean alues o mo emen me ics may di e
significan ly be ween bi ds o aging na u ally o behind es-
sels, p edic ing essel in e ac ions is p oblema ic when hese
occu e y in equen ly and bi ds display mo emen pa e ns
ha also occu du ing na u al o aging.
Wande ing alba osses a e known o adop di e se s a e-
gies, including in-fligh sea ching o p ey, and si -and-wai
on he sea su ace (Weime ski ch, Wilson, & Lys, 1997; Pha-
lan e al., 2007; Weime ski ch e al., 2007). When o aging
in fligh , alba osses o en adop ARS (To es e al., 2011).
Howe e , p e ious s udies showed ha ARS beha iou in
wande ing alba osses a e p ey cap u e did no las long
and only occu ed a e inges ion o a la ge i em (Weime -
ski ch e al., 2007). Indeed, o e a la ge scale, he mos
e ec i e sea ch s a egy o wande ing alba osses is appa -
en ly o ollow a nea ly s aigh pa h, using ARS only when
Figu e 3 S a e-dependen densi y his og ams o a) obse ed speed, b) angle concen a ion, c) he numbe o landings, and d) p opo ion o
ime spen on he wa e su ace o wande ing alba osses acked om Sou h Geo gia o each beha iou bou . Di e en colou s indica e
he p opo ion o loca ions wi hin each s a e ha we e associa ed (g een) o no (pink) wi h essels (ei he de ec ed by bi d-bo ne ada o
om sa elli e AIS da a indica ing ac i ely ishing essels).
Animal Conse a ion  (2022) – ª2022 The Au ho s. Animal Conse a ion published by John Wiley & Sons L d on behal o Zoological Socie y o London. 7
A. P. B. Ca nei o e al. Sca enging behind essels mi o s na u al o aging in alba osses
hey encoun e a pa icula ly a ou able en i onmen
(Weime ski ch e al., 2007; Co beau e al., 2019). The inclu-
sion o imme sion da a in ou HMM allowed wo ARS
s a es o be iden ified, bo h in ol ing slow speeds and high
u ning angles, bu wi h one s a e including landings indica-
i e o p ey cap u e a emp s; his would no o he wise be
de ec able wi h loca ion da a only. The e o e, ou models
ep esen an imp o emen on p e ious app oaches using aux-
ilia y biologging da a o alida ion pu poses only. Landings
(de e mined om imme sion da a) p o ide a eliable indica-
o o a emp s a p ey cap u e, as alba osses should o he -
wise a oid incu ing he high ene ge ic cos o he
subsequen ake-o (Weime ski ch e al., 2000). Res ing on
he wa e – he only o he eason o land on he wa e su -
ace –appea s o be confined o da kness when low ligh
le els limi he oppo uni ies o in-fligh sea ching o p ey
(Phalan e al., 2007; Mackley e al., 2010). This will be he
case whe he sca enging behind a fishing essel o a ge ing
na u al p ey. Tha he speed o he s a e ARS wi h landings
was conside ably highe on a e age, and mo e a iable, han
he speed du ing ARS wi hou landings, may be explained
by a ew apid mo emen s o ac i ely o aging bi ds, and
also by he esolu ion o he GPS loca ions (10 min, possibly
ailing o cap u e he o al dis ance a elled [and hus speed]
du ing ci cula mo emen s in he s a e ARS wi hou land-
ings).
Ou inabili y o de e mine om mo emen s o ac i i y
whe he pa icula o aging bou s occu ed only in he icin-
i y o an ac i ely fishing essel sugges s ha wande ing
alba osses may pe cei e and eac o fishing essels and
a ou able o aging pa ches essen ially in he same way,
using isual o ol ac o y cues (Ne i , 2000; Weime ski ch
e al., 2007; Colle , Pa ick, & Weime ski ch, 2015). Fo ag-
ing behind a essel may be simply an ex ension o na u al
o aging o a gene alis species such as he wande ing alba-
oss. Indeed, pa allels ha e been d awn be ween sca enging
beha iou and in e ac ions wi h o he ypes o p edic able
an h opogenic ood sou ces (Colle , Pa ick, & Weime ski ch,
2017). Species ha ha e mo e specialised o aging s a egies
may espond di e en ly o essels. Fo ins ance, Colle ,
Pa ick, & Weime ski ch, (2017) showed ha black-b owed
alba osses we e mo e s ongly a ac ed o fishing essels
and showed a highe le el o ac i e in e ac ion han wande -
ing alba osses which o e lapped wi h he same flee . Mo e-
o e , many s udies in es iga ing seabi d–fishe ies in e ac ions
show di e ences in he deg ee o in e ac ion wi h essels
acco ding o species (Colle , Pa ick, & Weime ski ch, 2017),
popula ion (G anadei o e al., 2011), sex (Jim
enez e al.,
2016) and indi idual (Vo ie e al., 2010; G anadei o,
B ickle, & Ca y, 2014; Pa ick e al., 2015). The la ge a i-
abili y among indi iduals may obscu e pa e ns ha may
indica e o aging in associa ion wi h a essel based on
mo emen and ac i i y da a alone.
Wande ing alba osses om Sou h Geo gia ange widely
in he sou hwes A lan ic du ing he b eeding season, mos ly
in pelagic wa e s excep du ing b ood-gua d when ips end
o be es ic ed o he Sou h Geo gia shel and shel -slope
(Handley e al., 2020). Thei ex ensi e o aging ange is
such ha bi ds encoun e many fishing flee s wi h a a ie y
o ope a ional and gea cha ac e is ics (Jim
enez e al., 2014,
2020; Phillips e al., 2016). High o e laps be ween o aging
wande ing alba oss and pelagic longline fishe ies a e
epo ed in he B azil-Falklands confluence (Bugoni e al.,
2008; Jim
enez e al., 2014, 2016), pa icula ly wi h he Tai-
wanese and Japanese flee s (Jim
enez e al., 2016; Clay e al.,
2019). They also o e lap wi h deme sal longline fishe ies,
bu his is gene ally es ic ed o a con inen al shel o shel -
b eak habi a s (Clay e al., 2019). The di e si y o fishe ies,
flee s and ope a ional cha ac e is ics, including ime o se -
ing in ela ion o day/nigh ime and he use o mi iga ion
measu es, combined wi h en i onmen al condi ions, a e
known o influence he likelihood o in e ac ions (Jim
enez
e al., 2014). The e o e, a ia ion in a endance, mo emen
and landing pa e ns behind di e en ypes o fishing essel,
and when o aging on di e se na u al p ey, may gene a e
beha iou al signa u es which a e specific o each o hese
ci cums ances. The di e si y o fishing ope a ions ha may
be a ended by alba osses is likely o u he educe he p e-
dic i e powe o any gene ic model ying o dis inguish
be ween essel ollowing and o aging on na u al p ey.
Conclusions
An inabili y o p edic essel a endance based on mo emen
and ac i i y da a alone unde lines he alue o using mul iple
sou ces o in o ma ion, including bi d-bo ne ada and fine-
scale essel mo emen s o a be e unde s anding o
seabi d–fishe y in e ac ions. Fu he mo e, ou me hod p o-
ides a cohe en and gene alizable amewo k o segmen
ips using auxilia y biologging (imme sion) da a, which
Table 2 P e alence (P) and p edic i e accu acy (A) o andom
o es models aiming o dis inguish be ween beha iou al bou s o
acked wande ing alba osses om Sou h Geo gia wi h and
wi hou ac i ely ishing nea by essels as iden i ied om ada
ansmissions and essel AIS loca ions om Global Fishing Wa ch
(GFW)
Rada
h eshold
AIS h eshold
0 0.05 0.1 0.2 0.5
0–P: 0.074 P: 0.068 P: 0.057 P: 0.036
A: 0.926 A: 0.932 A: 0.943 A: 0.964
0.05 P: 0.064 P: 0.033 P: 0.031 P: 0.027 P: 0.019
A: 0.936 A: 0.967 A: 0.969 A: 0.973 A: 0.981
0.1 P: 0.056 P: 0.028 P: 0.027 P: 0.024 P: 0.018
A: 0.944 A: 0.972 A: 0.973 A: 0.976 A: 0.982
0.2 P: 0.043 P: 0.023 P: 0.023 P: 0.021 P: 0.016
A: 0.957 A: 0.977 A: 0.976 A: 0.979 A: 0.984
0.5 P: 0.023 P: 0.014 P: 0.014 P: 0.014 P: 0.011
A: 0.977 A: 0.986 A: 0.986 A: 0.986 A: 0.989
The sensi i i y o all models was ze o. Di e en h esholds o %
ada ansmissions and % o loca ions wi h a nea by essel classi-
ied as ac i ely ishing we e used o classi y a bou as a ending an
ac i ely ishing essel; he p e alence (P) indica es he p opo ion
o hese bou s ou o he o al sample o 4349 bou s om 85 indi-
iduals.
8Animal Conse a ion  (2022) – ª2022 The Au ho s. Animal Conse a ion published by John Wiley & Sons L d on behal o Zoological Socie y o London.
Sca enging behind essels mi o s na u al o aging in alba osses A. P. B. Ca nei o e al.
allow p ey cap u e a emp s o be quan ified. O he senso s
such as accele ome e s and magne ome e s may also p o ide
use ul da a o imp o ing p edic ions. Inco po a ion o da a
on wind condi ions migh also be aluable, as his modula es
fligh and o aging decisions o alba osses and o he sea-
bi ds (Clay e al., 2019). We cau ion ha he bi d-bo ne
ada did no always eco d essels ha we e nea by acco d-
ing o he AIS da a, indica ing ha di ec de ec ion o in e -
ac ions is no s aigh o wa d. None heless, ada de ec o s
can p o ide in aluable ools when fine-scale da a on essels
a e una ailable and hey can imp o e ou unde s anding o
he scale o undecla ed fishing. They can also be used o
suppo emo e su eillance o he oceans. The abili y o p e-
dic seabi d–fishe ies in e ac ions would be in aluable o
isk assessmen s and o imp o ing he a ge ing o esou ces
o byca ch mi iga ion and compliance-moni o ing, con ibu -
ing o he conse a ion o many highly h ea ened alba osses
and pe els (Phillips e al., 2016).
Acknowledgemen s
We a e g a e ul o all hose in ol ed in da a collec ion a
Bi d Island, in pa icula , Rosie Hall, Alex Dodds and James
C ymble. This pape is a con ibu ion o he Ecosys ems
componen o he B i ish An a c ic Su ey Pola Science o
Plane Ea h P og amme, unded by he Na u al En i onmen
Resea ch Council and he Bi dLi e In e na ional Ma ine p o-
g amme. Funding was p o ided by Da win Plus 092
(DPLUS092).
Re e ences
Be na d, A., Rod igues, A.S.L., Cazalis, V. & G 
emille , D.
(2021). Towa d a global s a egy o seabi d acking.
Conse . Le .,14, e12804.
Bodey, T.W., Jessopp, M.J., Vo ie , S.C., Ge i sen, H.D.,
Cleasby, I.R., Hame , K.C., Pa ick, S.C., Wakefield, E.D. &
Bea hop, S. (2014). Seabi d mo emen e eals he
ecological oo p in o fishing essels. Cu . Biol.,24,
R514–R515.
Bugoni, L., Mancini, P.L., Mon ei o, D.S., Nascimen o, L. &
Ne es, T.S. (2008). Seabi d byca ch in he B azilian pelagic
longline fishe y and a e iew o cap u e a es in he
sou hwes e n A lan ic Ocean. Endange . Species Res.,5,
137–147. h p://dx.doi.o g/10.3354/es 00115
Bus on, P.M. & Eli h, J. (2011). De e minan s o ep oduc i e
success in dominan pai s o clownfish: A boos ed
eg ession ee analysis. J. Anim. Ecol.,80, 528–538.
Ca nei o, A.P.B., Pea main, E.J., Oppel, S., Clay, T.A.,
Phillips, R.A., Bonne -Leb un, A.-.S., Wanless, R.M. e al.
(2020). A amewo k o mapping he dis ibu ion o
seabi ds by in eg a ing acking, demog aphy and
phenology. J. Appl. Ecol.,57, 514–525.
Cianche i-Benede i, M., Dell’Omo, G., Russo, T., Ca oni, C.
& Quill eld , P. (2018). In e ac ions be ween comme cial
fishing essels and a pelagic seabi d in he sou he n
Medi e anean Sea. BMC Ecol.,18, 54.
Cla k, B.L., Handby, T., Lea , E. & Webe , S.B. (2020a). Fi s
h ee-dimensional acks o he Ascension F iga ebi d
F ega a aquila highligh he impo ance o al i ude o
beha iou al s udies. Seabi d,32,1–19.
Cla k, B.L., Vig 
usd
o i , F., Jessopp, M.J., Bu gos, J.M.,
Bodey, T.W. & Vo ie , S.C. (2020b). Ganne s a e no
a ac ed o fishing essels in Iceland—Po en ial influence o
a disca d ban and ood a ailabili y. ICES J. Ma . Sci.,77,
692–700. h p://dx.doi.o g/10.1093/icesjms/ sz233
Clay, T.A., Joo, R., Weime ski ch, H., Phillips, R.A., den
Ouden, O., Basille, M., Clusella-T ullas, S., Assink, J.D. &
Pa ick, S.C. (2020). Sex-specific e ec s o wind on he
fligh decisions o a sexually dimo phic soa ing bi d. J.
Anim. Ecol.,89, 1811–1823.
Clay, T.A., Small, C., Tuck, G.N., Pa do, D., Ca nei o, A.P.B.,
Wood, A.G., C oxall, J.P., C ossin, G.T. & Phillips, R.A.
(2019). A comp ehensi e la ge-scale assessmen o fishe ies
byca ch isk o h ea ened seabi d popula ions. J. Appl.
Ecol.,56, 1882–1893.
Colle , J., Pa ick, S.C. & Weime ski ch, H. (2015).
Alba osses edi ec fligh owa ds essels a he limi o
hei isual ange. Ma . Ecol. P og. Se .,526, 199–205.
Colle , J., Pa ick, S.C. & Weime ski ch, H. (2017). A
compa a i e analysis o he beha io al esponse o fishing
boa s in wo alba oss species. Beha . Ecol.,28, 1337–1347.
Conne s, M.G., Michelo , T., Heywood, E.I., O ben, R.A.,
Phillips, R.A., Vysso ski, A.L., Sha e , S.A. & Tho ne,
L.H. (2021). Hidden Ma ko models iden i y majo
mo emen modes in accele ome e and magne ome e da a
om ou alba oss species. Mo . Ecol.,9,7.
Co beau, A., Colle , J., Fon enille, M. & Weime ski ch, H.
(2019). How do seabi ds modi y hei sea ch beha iou when
encoun e ing fishing boa s? PLOS ONE,14, e0222615.
Co beau, A., Colle , J., O ge e , F., Pis o ius, P. &
Weime ski ch, H. (2021a). Fine-scale in e ac ions be ween
boa s and la ge alba osses indica e a iable suscep ibili y o
byca ch isk acco ding o species and popula ions. Anim.
Conse .,24, 689–697. h p://dx.doi.o g/10.1111/ac .12676
Co beau, A., Colle , J., Pajo , A., Joo, R., Thellie , T. &
Weime ski ch, H. (2021b). Di e ences in o aging habi a
esul in con as ing fishe ies in e ac ions in wo alba oss
popula ions. Ma . Ecol. P og. Se .,663, 197–208.
Dean, B., F eeman, R., Ki k, H., Leona d, K., Phillips, R.A.,
Pe ins, C.M. & Guil o d, T. (2013). Beha iou al mapping
o a pelagic seabi d: Combining mul iple senso s and a
hidden Ma ko model e eals he dis ibu ion o a -sea
beha iou . J. R. Soc. In e ace,10, 20120570. h p://dx.doi.
o g/10.1098/ si .2012.0570
Dias, M.P., Ma in, R., Pea main, E.J., Bu field, I.J., Small, C.,
Phillips, R.A., Ya es, O., Lascelles, B., Bo bo oglu, P.G. &
C oxall, J.P. (2019). Th ea s o seabi ds: A global
assessmen . Biol. Conse .,237, 525–537.
Fauchald, P. & T e aa, T. (2003). Using fi s -passage ime in
he analysis o a ea- es ic ed sea ch and habi a selec ion.
Ecology,84, 282–288.
Animal Conse a ion  (2022) – ª2022 The Au ho s. Animal Conse a ion published by John Wiley & Sons L d on behal o Zoological Socie y o London. 9
A. P. B. Ca nei o e al. Sca enging behind essels mi o s na u al o aging in alba osses