Submi ed 8 Sep embe 2016
Accep ed 2 No embe 2016
Published 21 Decembe 2016
Co esponding au ho
Jamie R. S a e ,
[email p o ec ed]
Academic edi o
Ch is Cu le
Addi ional In o ma ion and
Decla a ions can be ound on
page 13
DOI 10.7717/pee j.2779
Copy igh
2016 S a e e al.
Dis ibu ed unde
C ea i e Commons CC-BY 4.0
OPEN ACCESS
Hai iness: he missing link be ween
pollina o s and pollina ion
Jamie R. S a e 1, Gus a o Liñán-Cemb ano2, Jacqueline R. Beggs1,
B ad G. Howle 3, Da id E. Pa emo e4and Ignasi Ba omeus5
1Cen e o Biodi e si y and Biosecu i y, School o Biological Sciences, The Uni e si y o Auckland, Auckland,
New Zealand
2Ins i u o de Mic oelec ónica de Se illa CSIC/Uni e sidad de Se illa, Se illa, Spain
3The New Zealand Ins i u e o Plan & Food Resea ch Limi ed, Ch is chu ch, New Zealand
4The New Zealand Ins i u e o Plan & Food Resea ch Limi ed, Hamil on, New Zealand
5In eg a i e Ecology Depa men , Es ación Biológica de Doñana (EBD-CSIC), Se illa, Spain
ABSTRACT
Backg ound. Func ional ai s a e he p ima y bio ic componen d i ing o ganism
in luence on ecosys em unc ions; in consequence, ai s a e widely used in ecological
esea ch. Howe e , mos animal ai -based s udies use easy- o-measu e cha ac e is ics
o species ha a e a bes only weakly associa ed wi h unc ions. Animal-media ed
pollina ion is a key ecosys em unc ion and is likely o be in luenced by pollina o
ai s, bu o da e no one has iden i ied unc ional ai s ha a e simple o measu e and
ha e good p edic i e powe .
Me hods. He e, we show ha a simple, easy o measu e ai (hai iness) can p edic
pollina o e ec i eness wi h high accu acy. We used a no el image analysis me hod o
calcula e en opy alues o insec body su aces as a measu e o hai iness. We e alua ed
he powe o ou me hod o p edic ing pollina o e ec i eness by eg essing pollina o
hai iness (en opy) agains single isi pollen deposi ion (SVD) and pollen loads on
insec s. We used linea models and AICCmodel selec ion o de e mine which body
egions we e he bes p edic o s o SVD and pollen load.
Resul s. We ound ha hai iness can be used as a obus p oxy o SVD. The bes models
o p edic ing SVD o he lowe species B assica apa and Ac inidia deliciosa we e
hai iness on he ace and ho ax as p edic o s (R2=0.98 and 0.91 espec i ely). The
bes model o p edic ing pollen load o B. apa was hai iness on he ace (R2=0.81).
Discussion. We sugges ha he ma ch be ween pollina o body egion hai iness and
plan ep oduc i e s uc u e mo phology is a powe ul p edic o o pollina o e ec i e-
ness. We show ha pollina o hai iness is s ongly linked o pollina ion—an impo an
ecosys em unc ion, and p o ide a igo ous and ime-e icien me hod o measu ing
hai iness. Iden i ying and accu a ely measu ing key ai s ha d i e ecosys em p ocesses
is c i ical as global change inc easingly al e s ecological communi ies, and subsequen ly,
ecosys em unc ions wo ldwide.
Subjec s Biodi e si y, Ecology, Ecosys em Science, En omology, Zoology
Keywo ds Pollina ion, Pilosi y, En opy, Func ional ai , Pollen deposi ion, Ecosys em unc ion,
Image analysis, Pollen load, SVD
How o ci e his a icle S a e e al. (2016), Hai iness: he missing link be ween pollina o s and pollina ion. Pee J 4:e2779; DOI
10.7717/pee j.2779
INTRODUCTION
T ai -based app oaches a e now widely used in unc ional ecology, om he le el o
indi idual o ganisms o ecosys ems (Cado e, Ca scadden & Mi o chnick, 2011). Func ional
ai s a e de ined as he cha ac e is ics o an o ganism’s pheno ype ha de e mine i s
e ec on ecosys em le el p ocesses (Naeem & W igh , 2003;Pe chey & Gas on, 2006).
Acco dingly, unc ional ai s a e ecognised as he p ima y bio ic componen by which
o ganisms in luence ecosys em unc ions (Gagic e al., 2015;Hilleb and & Ma hiessen,
2009). T ai -based esea ch is domina ed by s udies on plan s and p ima y p oduc i i y,
and li le is known abou key ai s o animal-media ed and mul i- ophic unc ions,
pa icula ly o e es ial in e eb a es (Didham, Lea he & Basse , 2016;Gagic e al., 2015;
La o el e al., 2013).
Mos animal ai -based s udies simply quan i y easy- o-measu e mo phological
cha ac e is ics, wi hou a mechanis ic unde pinning o demons a e hese ‘‘ ai s’’ ha e
any in luence on he ecosys em unc ion o in e es (Didham, Lea he & Basse , 2016). This
esul s in low p edic i e powe , pa icula ly whe e ai selec ion lacks s ong jus i ica ion
h ough explici ecological ques ions (Gagic e al., 2015;Pe chey & Gas on, 2006). I he
ul ima e goal o ai -based ecology is o iden i y he mechanisms ha d i e biodi e si y
impac s on ecosys em unc ion, hen ai s mus be quan i iable a he le el o he indi idual
o ganism, and be inhe en ly linked o an ecosys em unc ion (Bolnick e al., 2011;Pasa i e
al., 2013;Violle e al., 2007).
Me hodology ha allows collec ion o ai da a in a igo ous ye ime-e icien manne
and wi h di ec unc ional in e p e a ion will g ea ly enhance he powe o ai -based
s udies. Ins ead o subjec i ely selec ing a la ge numbe o ai s wi h unspeci ied links o
ecosys em unc ions, i would be be e o iden i y ewe , unco ela ed ai s, ha ha e
a s ong bea ing on he unc ion o in e es (Ca mona e al., 2016). Selec ing ai s ha
a e measu able on a con inuous scale, would also imp o e p edic i e powe o s udies
(McGill e al., 2006;Violle e al., 2012). Howe e , a g ea e ime and e o is equi ed o
measu e such ai s, exace ba ing he al eady demanding na u e o ai -based communi y
ecology (Pe chey & Gas on, 2006).
Animal-media ed pollina ion is a mul i- ophic unc ion, d i en by he in e ac ion
be ween animal pollina o s and plan s (K emen e al., 2007). A majo i y o he wo ld’s
wild plan species a e pollina ed by animals (Olle on, Win ee & Ta an , 2011), and
o e a hi d o global c ops a e dependen on animal pollina ion (Klein e al., 2007).
Unde s anding which pollina o ai s de e mine he e ec i eness o di e en pollina o s
is c i ical o unde s anding he mechanisms o pollina ion p ocesses. Howe e , cu en
ai s used in pollina ion s udies o en ha e weak associa ions wi h pollina ion unc ion
and/o ha e low p edic i e powe . Fo example La sen, Williams & K emen (2005) used
body mass o explain pollen deposi ion by soli a y bees e en when he ela ionship was
weak and non-signi ican . Many ai -based pollina ion s udies ha e subsequen ly used
body mass o simila size measu es, despi e hei low p edic i e powe . Simila ly, Hoehn
e al. (2008) used spa ial and empo al isi a ion p e e ences o bees o explain di e ences
in plan s ep oduc i e ou pu . They ound signi ican ela ionships (i.e., low P alues)
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 2/18
be ween spa ial and empo al isi a ion p e e ences and seed se , bu wi h small R2 alues,
sugges ing hese ai s ha e weak p edic i e powe . To ad ance ai -based pollina ion
esea ch we equi e ai s ha a e good p edic o s o pollina ion success.
Obse a ional s udies sugges ha insec body hai s a e impo an o collec ing pollen
ha is used by insec s o ood and la al p o isioning (Holloway, 1976;Tho p, 2000). Hai s
acili a e ac i e pollen collec ion, e.g., many bees ha e specialised hai s uc u es called
scopae ha a e used o anspo pollen o he nes o la al p o isioning (Tho p, 2000).
Addi ionally, bo h bees and lies ha e hai s dis ibu ed ac oss hei body su aces which
ac o passi ely collec pollen o adul eeding (Holloway, 1976). Di e ences in he densi y
and dis ibu ion o hai s on pollen eeding insec s likely e lec s hei eeding beha iou ,
he ypes o lowe s hey isi , and whe he hey use pollen o adul eeding and/o la al
p o isioning (Tho p, 2000). Howe e , despi e anecdo al e idence ha insec body hai s a e
impo an o pollen collec ion and pollina ion, he e is no p o en me hod o measu ing
hai iness, no is he e e idence ha hai ie insec s a e mo e e ec i e pollina o s.
He e, we p esen a no el me hod based on image en opy analysis o quan i ying
pollina o hai iness. We de ine pollina ion e ec i eness as single isi pollen deposi ion
(SVD): he numbe o conspeci ic pollen g ains deposi ed on a i gin s igma in a single isi
(King, Ballan yne & Willme , 2013;Ne’eman e al., 2010). SVD is a measu e o an insec s’
abili y o acqui e ee pollen g ains on he body su ace and accu a ely deposi hem on a
conspeci ic s igma. We p edic ha hai iness, speci ically on he body pa s ha con ac he
s igma, will ha e a s ong associa ion wi h SVD. We show ha he bes model o p edic ing
pollina o SVD o pak choi B assica apa is highly p edic i e and includes hai iness o he
ace and ho ax do sal egions as p edic o s, and he ace egion alone explains mo e han
90% o he a ia ion. Simila ly, he bes model o p edic ing SVD o kiwi ui Ac inidia
deliciosa includes he ace and ho ax en al egions and has good p edic i e powe . Ou
no el me hod o measu ing hai iness is igo ous, ime e icien and inhe en ly linked
o pollina ion unc ion. Acco dingly, his me hod could be applied in di e se ai -based
pollina ion s udies o p og ess unde s anding o he mechanisms ha d i e pollina ion
p ocesses.
MATERIALS AND METHODS
Imaging o hai iness analysis
We pho og aphed pinned insec specimens using he Visiona y Digi al Passpo po able
imaging sys em (Fig. 1). Images we e aken wi h a Canon EOS 5D Ma k II digi al came a
(5,616 ×3,744 pix). The came a colou p o ile was sRGB IEC61966-2.1, ocal leng h
was 65 mm and F-numbe was 4.5. We used en al, do sal and on al sho s wi h clea
illumina ion o minimise e lec ion om shinny insec body su aces. All pho og aphs
we e aken on a plain whi e backg ound. Raw images we e expo ed o Helicon Focus 6
whe e hey we e s acked and s o ed in .jpg ile o ma .
Image p ocessing and analysis
We p oduced code o quan i y insec pollina o hai iness using MATLAB (Ma hWo ks,
Na ick, MA, USA), and unc ions om he MATLAB Image P ocessing ToolBox. We
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 3/18
50
100
150
200
250
ab
Figu e 1 En opy image o he ace o a na i e New Zealand soli a y bee Leiop oc us paahaumaa (A)
and he co esponding en opy image (B). Wa me colou s on he en opy image ep esen highe en-
opy alues (shown by he scale ba on he igh ). Black do s on he en opy image a e nea - ound and
small objec s ha ha e been emo ed om he analysis by he p e-p ocessing unc ion.
quan i ied ela i e hai iness by c ea ing an en opy image o each insec pho og aph, and
compu ed he a e age en opy wi hin use -de ined egions (Gonzales, Woods & Eddins,
2004). To calcula e en opy alues o each image we designed h ee main unc ions. The
i s unc ion allows he use o de ine up o ou egions o in e es (RoIs) wi hin each
image. The use can de ine egions by d awing con ou s as closed polygonal lines o any
a bi a y numbe o e exes. All in o ma ion abou egions (loca ion, a ea and inpu
image ile name) is s o ed as a s uc u e in a .ma ile.
The second unc ion execu es image p e-p ocessing. We ound ha some insec s had
pollen g ains o o he a e ac s a ached o hei bodies, which would al e he en opy
esul s. Ou p e-p ocessing unc ion elimina es hese objec s om he image by unning
wo il e ing p ocesses. Fi s , he unc ion elimina es small objec s wi h an a ea less han
he use de inable h eshold (8 pixels by de aul ). Fo he i s ask, each ma ked egion
is segmen ed using an op imized h eshold ob ained by applying a spa ially dependan
h esholding echnique. Once each egion has been segmen ed, a labelling p ocess is
execu ed o all esul ing objec s and hose wi h an a ea smalle han he minimum alue
de ined by he use a e emo ed. Secondly, as pollen g ains a e o en ound in shape, he
unc ion elimina es nea -ci cula objec s. The pe ime e o each objec is calcula ed and i s
simila i y o a ci cle (S) id de ined as:
S=4π·A ea
Pe ime e 2.
Objec s wi h a simila i y coe icien no wi hin he bounds de ined by he use (5% by
de aul ) a e also emo ed om he image. Pe ime e calcula ion is ca ied ou by inding
he objec ’s bounda y, and compu ing he accumula ed dis ance om pixel cen e o pixel
cen e ac oss he bo de , a he han simply coun ing he numbe o pixels in he bo de .
The en opy il e will no p ocess objec s ha ha e been ma ked as ‘‘dele ed’’ by he
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 4/18
p e-p ocessing unc ion. This ini ial p e-p ocessing p o ides lexibili y by allowing use s o
de ine he minimum a ea h eshold and he deg ee o simila i y o objec s o a ci cle. Use s
can also disable he image p e-p ocessing by oggling a lag when unning he en opy
il e .
Once p e-p ocessing is comple e, each image is passed o he hi d unc ion, which
is he en opy il e calcula ion s age. The en opy il e p oduces an o e all measu e o
andomness wi hin each o he use de ined egions on he image. In in o ma ion heo y,
en opy (also exp essed as Shannon En opy) is an indica o o he a e age amoun o
in o ma ion con ained in a message (Shannon, 1948). The e o e, Shannon En opy, H, o a
disc e e andom a iable X ha can ake npossible alues {x1,x2,...,xn}, wi h a p obabili y
mass unc ion P(X) is gi en by:
H(X)= −
n
X
i=1
P(xi)·log2(P(xi)).
When his de ini ion is used in image p ocessing, local en opy de ines he deg ee o
complexi y ( a iabili y) wi hin a gi en neighbou hood a ound a pixel. In ou case, his
neighbou hood (o en e e ed o as he s uc u ing elemen ) is a disk wi h adius
(we call he adius o in luence) ha can be de ined by he use (7 pixels by de aul ).
Thus o a gi en pixel in posi ion (i,j) in he inpu image, he en opy il e compu es he
his og am Gij (using 256 bins) o all pixels wi hin i s adius o in luence, and e u ns i s
en opy alue Hij as:
Hij = −Gij ·log2(Gij),
whe e Gij is a ec o con aining he his og am esul s o pixel (i,j) and (·) is he do
p oduc ope a o . Using de aul pa ame e s, ou en opy il e employs a 7 pixel (13 ×13
neighbou hood) adius o in luence, and a disk-shaped s uc u ing elemen , which we
de e mined based on he size o hai s. The e o e, in he en opy image, each pixel akes a
alue o en opy when conside ing 160 pixels a ound i (by de aul ). We de e mined he
op imal adius o in luence o he en opy il e by unning ou en opy unc ion wi h he
adius o in luence se as a a iable pa ame e . We hen isually compa ed he con as in
a eas o low s. high hai iness in he esul ing en opy images (i.e., Fig. 1). We ound ha a
7 pixel adius o in luence ga e he bes con as be ween low and high hai iness a eas o
ou species se . Hai hickness alues ac oss species ypically anged be ween 3.5–4.5 pixels
and he e o e, he 7 pixel adius o in luence is app oxima ely wo imes he wid h o a
hai .
The de ini ion o he op imum adius o in luence depends on he size o he
mo phological esponsible o he complexi y in he RoI. This is de ined no only
by he physical size o hese ea u es bu also by he pixel- o-millime e scaling ac o
(i.e., numbe o pixels in he senso plane pe mm in he scene plane). Thus, al hough
7 pixels is he op imum in ou case o de ec hai s, he en opy il e unc ion akes his
adius as an ex e nal pa ame e which can be adjus ed by he use o mee hei needs.
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 5/18

The en opy il e unc ion is a p ocess ha uns o e h ee di e en en opy laye s
(ER,EG,EB), one o each o he came a’s colou channels (Red, G een, and Blue), o each
inpu image. These h ee images a e combined in o a inal combined en opy image ES,
whe e each pixel in posi ion (i,j) akes he alue ES(i,j):
ES(i,j)=ER(i,j)·EG(i,j)·EG(i,j)
Once en opy calcula ions a e comple e, ou unc ion compu es a e ages and s anda d
de ia ions o ESwi hin each o he egions p e iously de ined by he use , and w i es
he esul s in o a .cs ile (one ow pe image). En opy alues p oduced by his
unc ion a e consis en o di e en pho os o he same egion on he same specimen
(Supplemen al In o ma ion 5). The sc ip s o he image p e-p ocessing, egion
ma king and en opy analysis unc ions a e p o ided, along wi h a MATLAB u o ial
(Supplemen al In o ma ion 1–4).
Hai iness as a p edic o o SVD and pollen load
Model lowe lo al biology and pollina o collec ion
We used pak choi B assica apa a . chinensis (B assicaceae) and kiwi ui Ac inidia deliciosa
(Ac inidiaceae) as model lowe s o de e mine i ou measu emen o insec hai iness is a
good p edic o o pollina o e ec i eness.
Bo h B. apa and A.deliciosa a e impo an mass lowe ing global ood c ops (Klein e
al., 2007;Rade e al., 2009). B. apa has an ac inomo phic open pollina ed yellow lowe
wi h ou sepals, ou pe als, and six s amens ( ou long and wo sho ) (Walke , Kinzig &
Lang idge, 1999). The nec a ies a e loca ed in he cen e o he lowe , be ween he s amens
and he pe als, o cing pollina o s o in oduce hei head be ween he pe als. B. apa shows
inc eased seed se in he p esence o insec pollina o s and he lowe s a e isi ed by a
di e se assemblage o insec s ha di e in hei abili y o ans e pollen (Rade e al., 2013).
A.deliciosa is dioecious wi h indi idual plan s p oducing ei he male o emale lowe s.
Flowe s a e la ge (4–6 cm in diame e ) and ypically ha e 5–9 whi e/c eam colou ed pe als
(De i, Thaku & Ga g, 2015). Flowe s ha e mul iple s amens and s aminodes wi h yellow
an he s. Female lowe s ha e a la ge s igma wi h mul iple b anches ha o m a b ush-like
s uc u e. Bo h male and emale lowe s do no p oduce nec a bu bo h p oduce pollen,
which ac s as a ewa d o isi o s. Like B. apa,A.deliciosa lowe s a e isi ed by a di e se
ange o insec s ha di e in hei abili y o ans e pollen, and seed se is inc eased in he
p esence o insec pollina o s (C aig e al., 1988).
We collec ed pollina ing insec s o image analysis du ing he summe o Decembe
2014–Janua y 2015. Insec s we e chilled immedia ely and hen killed by eezing wi hin 1
day and s o ed a −18 ◦C in indi idual ials. All insec s we e iden i ied o species le el wi h
assis ance om expe axonomis s.
Image p ocessing
We measu ed he hai iness o 10 insec pollina o species (n=8–10 indi iduals pe species),
ac oss i e amilies and wo o de s. This included social, semi-social and soli a y bees and
pollina ing lies. Regions ma ked included: (1) ace; (2) head do sal; (3) head en al;
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 6/18
(4) on leg; (5) ho ax do sal; (6) ho ax en al; (7) abdomen do sal and (8) abdomen
en al. All en opy analysis was ca ied ou using ou image p ocessing me hod ou lined
abo e. Fo es ima es o body size, we ook mul iple linea measu emen s (body leng h,
body wid h, head leng h, head wid h, o eleg leng h and hind leg leng h) o each specimen
using digi al callipe s and a dissec ing mic oscope.
Single isi pollen deposi ion (SVD) and pollen load
Fo B. apa we used SVD da a o insec pollina o s p esen ed in Rade e al. (2009) and
Howle e al. (2011); a b ie desc ip ion o hei me hods ollows.
Pollen deposi ion on s igma ic su aces (SVD) was es ima ed using manipula ion
expe imen s. Vi gin B. apa in lo escences we e bagged o exclude all pollina o s. Once
lowe s had opened, he bag was emo ed, and lowe s we e obse ed un il an insec isi ed
and con ac ed he s igma in a single isi . The s igma was hen emo ed and s o ed in
gela ine- uchsin and he insec was cap u ed o la e iden i ica ion. SVD was quan i ied
by coun ing all B. apa pollen g ains on he s igma. Mean alues o SVD o each species
a e used in ou eg ession models.
To quan i y he numbe o pollen g ains ca ied (pollen load), sensu Howle e al. (2011),
collec ed insec s while o aging on B. apa lowe s. Insec s we e cap u ed using plas ic ials
con aining a apid killing agen (e hyl ace a e). Once dead, a cube o gela ine- uchsin was
used o emo e all pollen om he insec ’s body su ace. Pollen collec ing s uc u es (e.g.,
co biculae, scopae) we e no included in analyses because pollen om hese s uc u es is no
a ailable o pollina ion. Slides we e p epa ed in he ield by mel ing he gela ine- uchsin
cubes con aining pollen samples on o mic oscope slides. B. apa pollen g ains om each
sample we e hen quan i ied by coun ing pollen g ains in an equal-a ea subse om he
sample and mul iplying his by he numbe o equi alen sized subse a eas wi hin he o al
sample.
We measu ed SVD o A.deliciosa (n=8–12 pe pollina o species). SVD measu emen s
we e aken o insec mo emen s om s amina e o pis illa e lowe s, using a me hod
ha di e ed om B. apa. Indi idual pis illa e buds we e enclosed wi hin pape bags
2–3 days p io o opening, and we e la e used as es lowe s o e alua e pollen deposi ion
by lowe ing isi ing species. Each bag was secu ed using a wi e ie (coa ed in plas ic) ha
was gen ly wis ed o exclude pollina o s om isi ing he opening lowe s. Following
lowe opening, he bag was emo ed and he lowe pedicel abscised whe e i joined
he ine. The es lowe was hen ca e ully posi ioned using o ceps o hold he pedicel
1–2 cm om a s amina e lowe con aining a o aging insec , a oiding any con ac ing
be ween lowe s. I he es lowe was isi ed by an insec , we allowed i o o age wi h
minimal dis u bance un il i mo ed om he lowe on i s own acco d. The i s s igma
ouched by he o aging insec was hen ligh ly ma ked nea i s base using a ine black
el pen. We hen placed he ma ked s igma on o a slide and applied a d op o Alexande
s ain (Da ni, 2007). Alexande s ain was used due o i s e ec i eness o s ain s amina e
and pis illa e pollen di e en ly (pis illa e pollen—g een-blue, s amina e pollen—da k ed)
(Goodwin & Pe y, 1992).
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 7/18
S a is ical analyses
We used linea eg ession models and AICC(small sample co ec ed Akaike in o ma ion
c i e ia) model selec ion o de e mine i ou measu e o pollina o hai iness is a good
p edic o o SVD and pollen load. We cons uc ed global models wi h SVD o pollen load
as he esponse a iable, body egion as p edic o s and body leng h as an in e ac ion i.e.,
SVD o pollen load ∼body leng h * en opy ace +en opy head do sal +en opy head
en al + on leg +en opy ho ax do sal +en opy ho ax en al +en opy abdomen
do sal +en opy abdomen en al. We included body leng h in ou global model as a
p oxy o body size as i had high co ela ion coe icien s (Pea son’s > 0.7) wi h all
o he body size measu emen s. Global linea models we e cons uc ed using he lm(s a s)
unc ion. AICCmodel selec ion was ca ied ou on he global models using he unc ion
glmul i() wi h i unc ion =‘‘lm’’ in he package glmul i. We examined he e oscedas ici y
and no mali y o e o s o models by isually inspec ing diagnos ic plo s using he glmul i
package (C awley, 2002). Va iance in la ion ac o s (VIF) o p edic o a iables we e
checked o he bes models using he i () unc ion in he ca package. All analyses we e
done in R e sion 3.2.4 (R Co e Team, 2014).
RESULTS
Body hai iness as a p edic o o SVD
Fo SVD on B. apa, he ace and ho ax do sal egions we e e ained in he bes model
selec ed by AICC, which had an adjus ed R2 alue o 0.98. The subsequen op models
wi hin 10 AICCpoin s all e ained he ace and ho ax do sal egions and addi ionally
included he abdomen en al (adjus ed R2=0.98), head do sal (adjus ed R2=0.98), and
ho ax en al (adjus ed R2=0.97) and on leg (adjus ed R2=0.97) egions espec i ely
(Table 1;Fig. 2). The model wi h he ace egion included as a single p edic o had an
adjus ed R2 alue o 0.88, indica ing ha his egion alone explained a majo i y o he
a ia ion in he op SVD models.
The bes model o p edic ing SVD on A.deliciosa included he ace and ho ax en al
egions as p edic o s (adjus ed R2=0.91) (Table 1;Fig. 3). Howe e , he subsequen op
ou models we e wi hin wo AICCpoin s o he bes model and he e o e canno be
discoun ed as he po en ial op model. The ace, ho ax en al, head en al and abdomen
en al egions we e e ained in ou o he i e op models, which indica es ha hai iness
o he ace and en al egions is impo an o pollen deposi ion on A. deliciosa. Fo bo h
B. apa and A.deliciosa, body leng h and he body leng h in e ac ion we e no included in
he op models.
Body hai iness as a p edic o o pollen load
The bes model o pollen load e ained he ace egion only and had an adjus ed R2 alue o
0.81 (Fig. 4;Table 1). The subsequen bes models e ained he abdomen do sal (adjus ed
R2 alue o 0.73), he ace and head do sal (adjus ed R2=0.83), he ace and abdomen
do sal (adjus ed R2=0.82) and he abdomen do sal and on leg (adjus ed R2=0.8)
egions espec i ely. Fo pollen load, body leng h and he body leng h in e ac ion we e no
included in he op models.
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 8/18
Table 1 Reg ession models examining he e ec o en opy on SVD and pollen load. Top eg ession models examining he e ec o insec body
egion en opy on single isi pollen deposi ion (SVD) o B assica apa and Ac inidia deliciosa and pollen load o B. apa. Models a e p esen ed in
ascending o de based on AICC alues. Top models o each esponse a iable a e highligh ed in bold.
Response
a iable
Model Adj R2AICc1i wiacc wi
Face +Tho ax do sal 0.98 88.29 0.00 0.82 0.82
Face +Tho ax do sal +Abdomen en al 0.98 93.09 4.80 0.07 0.89
Face +Head do sal +Tho ax do sal 0.98 93.81 5.52 0.05 0.94
Face +Tho ax en al +Tho ax do sal 0.97 96.59 8.29 0.01 0.96
SVD (B. apa)
Face +Tho ax do sal +F on leg 0.97 97.02 8.72 0.01 0.97
Face 0.81 168.47 0.00 0.64 0.64
Abdomen do sal 0.73 171.59 3.12 0.13 0.78
Face +Head do sal 0.83 173.59 5.12 0.05 0.83
Face +Abdomen do sal 0.82 173.76 5.29 0.05 0.87
Pollen load
(B. apa)
Abdomen do sal +F on leg 0.80 174.86 6.39 0.03 0.90
Face +Tho ax en al 0.91 74.18 0.00 0.15 0.15
Abdomen do sal 0.81 74.21 0.03 0.15 0.30
Face 0.80 74.35 0.17 0.14 0.45
Head en al 0.79 74.84 0.66 0.11 0.56
SVD (A.deliciosa)
Abdomen en al 0.78 75.08 0.90 0.10 0.65
No es.
1i is he di e ence in he AICC alue o each model compa ed wi h he AICC alue o he op model. wiis he Akaike weigh o each model and acc wiis he cumula i e Akaike
weigh .
DISCUSSION
He e we p esen a igo ous and ime-e icien me hod o quan i ying hai iness, and
demons a e ha his measu e is an impo an pollina o unc ional ai . We show ha
insec pollina o hai iness is a s ong p edic o o SVD o he open-pollina ed lowe
B. apa. Linea models ha included mul iple body egions as p edic o s had he highes
p edic i e powe ; he op model o SVD e ained he ace and ho ax do sal egions.
Howe e , he ace egion was e ained in all o he op models, and when included as a
single p edic o , had a e y s ong posi i e associa ion wi h SVD. In addi ion, we show
ha hai iness, pa icula ly on he ace and en al egions, is a good p edic o o SVD o
A.deliciosa, which has a di e en lo al mo phology, sugges ing ou me hod could be
sui able o a ange o lowe ypes. Hai iness was also a good p edic o o pollen load, and
he ace egion was again e ained in he op model o B. apa. The abdomen do sal, head
do sal and on leg egions we e also good p edic o s o pollen load and we e e ained
in he subsequen op models. Ou esul s alida e he impo ance o insec body hai s
o anspo ing and deposi ing pollen. Su p isingly, we did no ind s ong associa ions
be ween SVD and body size, and op models did no con ain he body leng h in e ac ion.
Simila ly, body leng h was no e ained in he op models o pollen load. This indica es
ha ou measu e o hai iness has a g ea e p edic i e powe han body size o bo h SVD
and pollen load.
When deciding on which body egions o measu e hai iness, esea che s may i s need
o assess addi ional pollina o ai s, such as lowe isi ing beha iou . This is because
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 9/18
Goodwin RM, Pe y JH. 1992. Use o pollen aps o in es iga e he o aging beha iou
o honey bee colonies in kiwi ui o cha ds. New Zealand Jou nal o C op and
Ho icul u al Science 20:23–26 DOI 10.1080/01140671.1992.10422322.
He e a CM. 1987. Componen s o pollina o ‘‘quali y’’: compa a i e analysis o a
di e se insec assemblage. Oikos 50:79–90.
Hilleb and H, Ma hiessen B. 2009. Biodi e si y in a complex wo ld: consolida ion
and p og ess in unc ional biodi e si y esea ch. Ecology Le e s 12:1405–1419
DOI 10.1111/j.1461-0248.2009.01388.x.
Hoehn P, Tscha n ke T, Tylianakis JM, S e an-Dewen e I. 2008. Func ional g oup
di e si y o bee pollina o s inc eases c op yield. P oceedings o he Royal Socie y B:
Biological Sciences 275:2283–2291 DOI 10.1098/ spb.2008.0405.
Holloway BA. 1976. Pollen- eeding in ho e - lies (Dip e a: Sy phidae). New Zealand
Jou nal o Zoology 3:339–350 DOI 10.1080/03014223.1976.9517924.
Howle BG, Walke MK, Rade R, Bu le RC, News om-Lloyd LE, Teulon DAJ. 2011.
Can insec body pollen coun s be used o es ima e pollen deposi ion on pak choi
s igmas? New Zealand Plan P o ec ion 64:25–31.
Ja o ek S, Mackenzie K, Vande Kloe S. 2002. Compa a i e pollina ion e ec i eness
among bees (Hymenop e a: Apoidea) on lowbush bluebe y (E icaceae: Vac-
cinium angus i olium). Annals o he En omological Socie y o Ame ica 95:345–351
DOI 10.1603/0013-8746(2002)095[0345:CPEABH]2.0.CO;2.
King C, Ballan yne G, Willme PG. 2013. Why lowe isi a ion is a poo p oxy o
pollina ion: measu ing single- isi pollen deposi ion, wi h implica ions o polli-
na ion ne wo ks and conse a ion. Me hods in Ecology and E olu ion 4:811–818
DOI 10.1111/2041-210X.12074.
Klein A-M, Vaissie e BE, Cane JH, S e an-Dewen e I, Cunningham SA, K emen C,
Tscha n ke T. 2007. Impo ance o pollina o s in changing landscapes o wo ld
c ops. P oceedings o he Royal Socie y o London B: Biological Sciences 274:303–313
DOI 10.1098/ spb.2006.3721.
K emen C, Williams NM, Aizen MA, Gemmill-He en B, LeBuhn G, Minckley R,
Packe L, Po s SG, Rouls on T, S e an-Dewen e I, Vázquez DP, Win ee R,
Adams L, C one EE, G eenlea SS, Kei TH, Klein AM, Rege z J, Ricke s TH.
2007. Pollina ion and o he ecosys em se ices p oduced by mobile o ganisms: a
concep ual amewo k o he e ec s o land-use change. Ecology Le e s 10:299–314
DOI 10.1111/j.1461-0248.2007.01018.x.
K emen C, Williams NM, Tho p RW. 2002. C op pollina ion om na i e bees a isk
om ag icul u al in ensi ica ion. P oceedings o he Na ional Academy o Sciences o
he Uni ed S a es o Ame ica 99:16812–16816 DOI 10.1073/pnas.262413599.
La sen TH, Williams NM, K emen C. 2005. Ex inc ion o de and al e ed commu-
ni y s uc u e apidly dis up ecosys em unc ioning. Ecology Le e s 8:538–547
DOI 10.1111/j.1461-0248.2005.00749.x.
La o el S, S o key J, Ba dge RD, De Bello F, Be g MP. 2013. A no el ame-
wo k o linking unc ional di e si y o plan s wi h o he ophic le els o he
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 16/18

quan i ica ion o ecosys em se ices. Jou nal o Vege a ion Science 24:942–948
DOI 10.1111/j s.12083.
May ield MM, Wase NM, P ice MV. 2001. Explo ing he ‘mos e ec i e pollina o
p inciple’ wi h complex lowe s: bumblebees and Ipomopsis agg ega a.Annals o
Bo any 88:591–596 DOI 10.1006/anbo.2001.1500.
McGill BJ, Do nelas M, Go elli NJ, Magu an AE. 2015. Fi een o ms o biodi-
e si y end in he An h opocene. T ends in Ecology & E olu ion 30:104–113
DOI 10.1016/j. ee.2014.11.006.
McGill BJ, Enquis BJ, Weihe E, Wes oby M. 2006. Rebuilding communi y
ecology om unc ional ai s. T ends in Ecology & E olu ion 21:178–185
DOI 10.1016/j. ee.2006.02.002.
Naeem S, W igh JP. 2003. Disen angling biodi e si y e ec s on ecosys em unc ion-
ing: de i ing solu ions o a seemingly insu moun able p oblem. Ecology Le e s
6:567–579 DOI 10.1046/j.1461-0248.2003.00471.x.
Ne’eman G, Jü gens A, News om-Lloyd L, Po s SG, Da ni A. 2010. A amewo k o
compa ing pollina o pe o mance: e ec i eness and e iciency. Biological Re iews
85:435–451 DOI 10.1111/j.1469-185X.2009.00108.x.
Olle on J, Win ee R, Ta an S. 2011. How many lowe ing plan s a e pollina ed by
animals? Oikos 120:321–326 DOI 10.1111/j.1600-0706.2010.18644.x.
Pasa i JR, Le i T, Za ale a ES, Tilman D. 2013. Se e al scales o biodi e si y a ec
ecosys em mul i unc ionali y. P oceedings o he Na ional Academy o Sciences o he
Uni ed S a es o Ame ica 110:10219–10222 DOI 10.1073/pnas.1220333110.
Pe chey OL, Gas on KJ. 2006. Func ional di e si y: back o basics and looking o wa d.
Ecology Le e s 9:741–758 DOI 10.1111/j.1461-0248.2006.00924.x.
Po s SG, Da ni A, Ne’eman G. 2001. Pollina ion o a co e lowe ing sh ub species in
Medi e anean ph ygana: a ia ion in pollina o di e si y, abundance and e ec i e-
ness in esponse o i e. Oikos 92:71–80 DOI 10.1034/j.1600-0706.2001.920109.x.
R Co e Team. 2014. R: a language and en i onmen o s a is ical compu ing. Vienna: R
Founda ion o S a is ical Compu ing. A ailable a h p://www.R-p ojec .o g/.
Rade R, Edwa ds W, Wes co DA, Cunningham SA, Howle BG. 2013. Diu -
nal e ec i eness o pollina ion by bees and lies in ag icul u al B assica apa:
implica ions o ecosys em esilience. Basic and Applied Ecology 14:20–27
DOI 10.1016/j.baae.2012.10.011.
Rade R, Howle BG, Cunningham SA, Wes co DA, News om-Lloyd LE, Walke
MK, Teulon DAJ, Edwa ds W. 2009. Al e na i e pollina o axa a e equally e icien
bu no as e ec i e as he honeybee in a mass lowe ing c op. Jou nal o Applied
Ecology 46:1080–1087 DOI 10.1111/j.1365-2664.2009.01700.x.
Ra hcke B. 1983. Compe i ion and acili a ion among plan s o pollina ion. In:
Pollina ion biology. New Yo k: Academic P ess, 305–329.
Roubik DW. 2000. Decep i e o chids wi h Meliponini as pollina o s. Plan Sys ema ics
and E olu ion 222:271–279 DOI 10.1007/BF00984106.
Shannon C. 1948. A ma hema ical heo y o communica ion. Bell Sys em Technical
Jou nal 3:379–423 DOI 10.1002/j.1538-7305.1948. b01338.x.
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 17/18
So ensen AE. 1986. Seed dispe sal by adhesion. Annual Re iew o Ecology and Sys ema ics
17:443–463.
Tho p RW. 2000. The collec ion o pollen by bees. Plan Sys ema ics and E olu ion
222(1):211–223 DOI 10.1007/BF00984103.
Violle C, Enquis BJ, McGill BJ, Jiang L, Albe CH, Hulsho C, Jung V, Messie J. 2012.
The e u n o he a iance: in aspeci ic a iabili y in communi y ecology. T ends in
Ecology & E olu ion 27:244–252 DOI 10.1016/j. ee.2011.11.014.
Violle C, Na as M-L, Vile D, Kazakou E, Fo unel C, Hummel I, Ga nie E. 2007. Le
he concep o ai be unc ional! Oikos 116:882–892
DOI 10.1111/j.0030-1299.2007.15559.x.
Walke B, Kinzig A, Lang idge J. 1999. Plan a ibu e di e si y, esilience, and ecosys-
em unc ion: he na u e and signi icance o dominan and mino species. Ecosys ems
2:95–113 DOI 10.1007/s100219900062.
S a e e al. (2016), Pee J, DOI 10.7717/pee j.2779 18/18