Submitted 8 September 2016Accepted 2 November 2016Published 21 December 2016

Corresponding authorJamie R Stavertjamiestavertgmailcom

Academic editorChris Cutler

Additional Information andDeclarations can be found onpage 13

DOI 107717peerj2779

Copyright2016 Stavert et al

Distributed underCreative Commons CC-BY 40

OPEN ACCESS

Hairiness the missing link betweenpollinators and pollinationJamie R Stavert1 Gustavo Lintildeaacuten-Cembrano2 Jacqueline R Beggs1Brad G Howlett3 David E Pattemore4 and Ignasi Bartomeus5

1Centre for Biodiversity and Biosecurity School of Biological Sciences The University of Auckland AucklandNew Zealand

2 Instituto de Microelectroacutenica de Sevilla CSICUniversidad de Sevilla Sevilla Spain3The New Zealand Institute for Plant amp Food Research Limited Christchurch New Zealand4The New Zealand Institute for Plant amp Food Research Limited Hamilton New Zealand5 Integrative Ecology Department Estacioacuten Bioloacutegica de Dontildeana (EBD-CSIC) Sevilla Spain

ABSTRACTBackground Functional traits are the primary biotic component driving organisminfluence on ecosystem functions in consequence traits are widely used in ecologicalresearch However most animal trait-based studies use easy-to-measure characteristicsof species that are at best only weakly associated with functions Animal-mediatedpollination is a key ecosystem function and is likely to be influenced by pollinatortraits but to date no one has identified functional traits that are simple to measure andhave good predictive powerMethods Here we show that a simple easy to measure trait (hairiness) can predictpollinator effectiveness with high accuracy We used a novel image analysis method tocalculate entropy values for insect body surfaces as ameasure of hairinessWe evaluatedthe power of our method for predicting pollinator effectiveness by regressing pollinatorhairiness (entropy) against single visit pollen deposition (SVD) and pollen loads oninsects We used linear models and AICC model selection to determine which bodyregions were the best predictors of SVD and pollen loadResults We found that hairiness can be used as a robust proxy of SVD The best modelsfor predicting SVD for the flower species Brassica rapa and Actinidia deliciosa werehairiness on the face and thorax as predictors (R2

= 098 and 091 respectively) Thebest model for predicting pollen load for B rapa was hairiness on the face (R2

= 081)Discussion We suggest that the match between pollinator body region hairiness andplant reproductive structuremorphology is a powerful predictor of pollinator effective-ness We show that pollinator hairiness is strongly linked to pollinationmdashan importantecosystem function and provide a rigorous and time-efficient method for measuringhairiness Identifying and accuratelymeasuring key traits that drive ecosystemprocessesis critical as global change increasingly alters ecological communities and subsequentlyecosystem functions worldwide

Subjects Biodiversity Ecology Ecosystem Science Entomology ZoologyKeywords Pollination Pilosity Entropy Functional trait Pollen deposition Ecosystem functionImage analysis Pollen load SVD

How to cite this article Stavert et al (2016) Hairiness the missing link between pollinators and pollination PeerJ 4e2779 DOI107717peerj2779

INTRODUCTIONTrait-based approaches are now widely used in functional ecology from the level ofindividual organisms to ecosystems (Cadotte Carscadden amp Mirotchnick 2011) Functionaltraits are defined as the characteristics of an organismrsquos phenotype that determine itseffect on ecosystem level processes (Naeem ampWright 2003 Petchey amp Gaston 2006)Accordingly functional traits are recognised as the primary biotic component by whichorganisms influence ecosystem functions (Gagic et al 2015 Hillebrand amp Matthiessen2009) Trait-based research is dominated by studies on plants and primary productivityand little is known about key traits for animal-mediated and multi-trophic functionsparticularly for terrestrial invertebrates (Didham Leather amp Basset 2016 Gagic et al 2015Lavorel et al 2013)

Most animal trait-based studies simply quantify easy-to-measure morphologicalcharacteristics without a mechanistic underpinning to demonstrate these lsquolsquotraitsrsquorsquo haveany influence on the ecosystem function of interest (Didham Leather amp Basset 2016) Thisresults in low predictive power particularly where trait selection lacks strong justificationthrough explicit ecological questions (Gagic et al 2015 Petchey amp Gaston 2006) If theultimate goal of trait-based ecology is to identify the mechanisms that drive biodiversityimpacts on ecosystem function then traits must be quantifiable at the level of the individualorganism and be inherently linked to an ecosystem function (Bolnick et al 2011 Pasari etal 2013 Violle et al 2007)

Methodology that allows collection of trait data in a rigorous yet time-efficient mannerand with direct functional interpretation will greatly enhance the power of trait-basedstudies Instead of subjectively selecting a large number of traits with unspecified links toecosystem functions it would be better to identify fewer uncorrelated traits that havea strong bearing on the function of interest (Carmona et al 2016) Selecting traits thatare measurable on a continuous scale would also improve predictive power of studies(McGill et al 2006 Violle et al 2012) However far greater time and effort is required tomeasure such traits exacerbating the already demanding nature of trait-based communityecology (Petchey amp Gaston 2006)

Animal-mediated pollination is a multi-trophic function driven by the interactionbetween animal pollinators and plants (Kremen et al 2007) A majority of the worldrsquoswild plant species are pollinated by animals (Ollerton Winfree amp Tarrant 2011) andover a third of global crops are dependent on animal pollination (Klein et al 2007)Understanding which pollinator traits determine the effectiveness of different pollinatorsis critical to understanding the mechanisms of pollination processes However currenttraits used in pollination studies often have weak associations with pollination functionandor have low predictive power For example Larsen Williams amp Kremen (2005) usedbody mass to explain pollen deposition by solitary bees even when the relationship wasweak and non-significant Many trait-based pollination studies have subsequently usedbody mass or similar size measures despite their low predictive power Similarly Hoehnet al (2008) used spatial and temporal visitation preferences of bees to explain differencesin plants reproductive output They found significant relationships (ie low P values)

Stavert et al (2016) PeerJ DOI 107717peerj2779 218

between spatial and temporal visitation preferences and seed set but with small R2 valuessuggesting these traits have weak predictive power To advance trait-based pollinationresearch we require traits that are good predictors of pollination success

Observational studies suggest that insect body hairs are important for collecting pollenthat is used by insects for food and larval provisioning (Holloway 1976 Thorp 2000) Hairsfacilitate active pollen collection eg many bees have specialised hair structures calledscopae that are used to transport pollen to the nest for larval provisioning (Thorp 2000)Additionally both bees and flies have hairs distributed across their body surfaces whichact to passively collect pollen for adult feeding (Holloway 1976) Differences in the densityand distribution of hairs on pollen feeding insects likely reflects their feeding behaviourthe types of flowers they visit and whether they use pollen for adult feeding andor larvalprovisioning (Thorp 2000) However despite anecdotal evidence that insect body hairs areimportant for pollen collection and pollination there is no proven method for measuringhairiness nor is there evidence that hairier insects are more effective pollinators

Here we present a novel method based on image entropy analysis for quantifyingpollinator hairiness We define pollination effectiveness as single visit pollen deposition(SVD) the number of conspecific pollen grains deposited on a virgin stigma in a single visit(King Ballantyne amp Willmer 2013 Nersquoeman et al 2010) SVD is a measure of an insectsrsquoability to acquire free pollen grains on the body surface and accurately deposit them on aconspecific stigma We predict that hairiness specifically on the body parts that contact thestigma will have a strong association with SVDWe show that the best model for predictingpollinator SVD for pak choi Brassica rapa is highly predictive and includes hairiness of theface and thorax dorsal regions as predictors and the face region alone explains more than90 of the variation Similarly the best model for predicting SVD for kiwifruit Actinidiadeliciosa includes the face and thorax ventral regions and has good predictive power Ournovel method for measuring hairiness is rigorous time efficient and inherently linkedto pollination function Accordingly this method could be applied in diverse trait-basedpollination studies to progress understanding of the mechanisms that drive pollinationprocesses

MATERIALS AND METHODSImaging for hairiness analysisWe photographed pinned insect specimens using the Visionary Digital Passport portableimaging system (Fig 1) Images were taken with a Canon EOS 5D Mark II digital camera(5616 times 3744 pix) The camera colour profile was sRGB IEC61966-21 focal lengthwas 65 mm and F-number was 45 We used ventral dorsal and frontal shots with clearillumination to minimise reflection from shinny insect body surfaces All photographswere taken on a plain white background Raw images were exported to Helicon Focus 6where they were stacked and stored in jpg file format

Image processing and analysisWe produced code to quantify insect pollinator hairiness using MATLAB (MathWorksNatick MA USA) and functions from the MATLAB Image Processing ToolBox We

Stavert et al (2016) PeerJ DOI 107717peerj2779 318

50

100

150

200

250

a b

Figure 1 Entropy image of the face of a native New Zealand solitary bee Leioproctus paahaumaa (A)and the corresponding entropy image (B) Warmer colours on the entropy image represent higher en-tropy values (shown by the scale bar on the right) Black dots on the entropy image are near-round andsmall objects that have been removed from the analysis by the pre-processing function

quantified relative hairiness by creating an entropy image for each insect photograph andcomputed the average entropy within user-defined regions (Gonzales Woods amp Eddins2004) To calculate entropy values for each image we designed three main functions Thefirst function allows the user to define up to four regions of interest (RoIs) within eachimage The user can define regions by drawing contours as closed polygonal lines of anyarbitrary number of vertexes All information about regions (location area and inputimage file name) is stored as a structure in a mat file

The second function executes image pre-processing We found that some insects hadpollen grains or other artefacts attached to their bodies which would alter the entropyresults Our pre-processing function eliminates these objects from the image by runningtwo filtering processes First the function eliminates small objects with an area less thanthe user definable threshold (8 pixels by default) For the first task each marked regionis segmented using an optimized threshold obtained by applying a spatially dependantthresholding technique Once each region has been segmented a labelling process isexecuted for all resulting objects and those with an area smaller than the minimum valuedefined by the user are removed Secondly as pollen grains are often round in shape thefunction eliminates near-circular objects The perimeter of each object is calculated and itssimilarity to a circle (S) id defined as

S=4π middotAreaPerimeter2

Objects with a similarity coefficient not within the bounds defined by the user (5 bydefault) are also removed from the image Perimeter calculation is carried out by findingthe objectrsquos boundary and computing the accumulated distance from pixel centre to pixelcentre across the border rather than simply counting the number of pixels in the borderThe entropy filter will not process objects that have been marked as lsquolsquodeletedrsquorsquo by the

Stavert et al (2016) PeerJ DOI 107717peerj2779 418

pre-processing function This initial pre-processing provides flexibility by allowing users todefine the minimum area threshold and the degree of similarity of objects to a circle Userscan also disable the image pre-processing by toggling a flag when running the entropyfilter

Once pre-processing is complete each image is passed to the third function whichis the entropy filter calculation stage The entropy filter produces an overall measure ofrandomness within each of the user defined regions on the image In information theoryentropy (also expressed as Shannon Entropy) is an indicator of the average amount ofinformation contained in a message (Shannon 1948) Therefore Shannon EntropyH of adiscrete random variable X that can take n possible values x1x2xn with a probabilitymass function P(X) is given by

H (X)=minusnsum

i=1

P (xi) middot log2(P (xi))

When this definition is used in image processing local entropy defines the degree ofcomplexity (variability) within a given neighbourhood around a pixel In our case thisneighbourhood (often referred to as the structuring element) is a disk with radius r(we call the radius of influence) that can be defined by the user (7 pixels by default)Thus for a given pixel in position (ij) in the input image the entropy filter computes thehistogram Gij (using 256 bins) of all pixels within its radius of influence and returns itsentropy value Hij as

Hij =minusGij middot log2(Gij)

where Gij is a vector containing the histogram results for pixel (ij) and (middot) is the dotproduct operator Using default parameters our entropy filter employs a 7 pixel (13 times 13neighbourhood) radius of influence and a disk-shaped structuring element which wedetermined based on the size of hairs Therefore in the entropy image each pixel takes avalue of entropy when considering 160 pixels around it (by default) We determined theoptimal radius of influence for the entropy filter by running our entropy function with theradius of influence set as a variable parameter We then visually compared the contrast inareas of low vs high hairiness in the resulting entropy images (ie Fig 1) We found that a7 pixel radius of influence gave the best contrast between low and high hairiness areas forour species set Hair thickness values across species typically ranged between 35ndash45 pixelsand therefore the 7 pixel radius of influence is approximately two times the width of ahair

The definition of the optimum radius of influence depends on the size of themorphological responsible for the complexity in the RoI This is defined not onlyby the physical size of these features but also by the pixel-to-millimetre scaling factor(ie number of pixels in the sensor plane per mm in the scene plane) Thus although7 pixels is the optimum in our case to detect hairs the entropy filter function takes thisradius as an external parameter which can be adjusted by the user to meet their needs

Stavert et al (2016) PeerJ DOI 107717peerj2779 518

The entropy filter function is a process that runs over three different entropy layers(ER EG EB) one for each of the camerarsquos colour channels (Red Green and Blue) for eachinput image These three images are combined into a final combined entropy image ESwhere each pixel in position (ij) takes the value ES(ij)

ES(ij)= ER(ij) middotEG(ij) middotEG(ij)

Once entropy calculations are complete our function computes averages and standarddeviations of ES within each of the regions previously defined by the user and writesthe results into a csv file (one row per image) Entropy values produced by thisfunction are consistent for different photos of the same region on the same specimen(Supplemental Information 5) The scripts for the image pre-processing regionmarking and entropy analysis functions are provided along with a MATLAB tutorial(Supplemental Information 1ndash4)

Hairiness as a predictor of SVD and pollen loadModel flower floral biology and pollinator collectionWeused pak choi Brassica rapa var chinensis (Brassicaceae) and kiwifruitActinidia deliciosa(Actinidiaceae) as model flowers to determine if our measurement of insect hairiness is agood predictor of pollinator effectiveness

Both B rapa and A deliciosa are important mass flowering global food crops (Klein etal 2007 Rader et al 2009) B rapa has an actinomorphic open pollinated yellow flowerwith four sepals four petals and six stamens (four long and two short) (Walker Kinzig ampLangridge 1999) The nectaries are located in the centre of the flower between the stamensand the petals forcing pollinators to introduce their head between the petals B rapa showsincreased seed set in the presence of insect pollinators and the flowers are visited by adiverse assemblage of insects that differ in their ability to transfer pollen (Rader et al 2013)A deliciosa is dioecious with individual plants producing either male or female flowersFlowers are large (4ndash6 cm in diameter) and typically have 5ndash9 whitecream coloured petals(Devi Thakur amp Garg 2015) Flowers have multiple stamens and staminodes with yellowanthers Female flowers have a large stigma with multiple branches that form a brush-likestructure Both male and female flowers do not produce nectar but both produce pollenwhich acts as a reward to visitors Like B rapa A deliciosa flowers are visited by a diverserange of insects that differ in their ability to transfer pollen and seed set is increased in thepresence of insect pollinators (Craig et al 1988)

We collected pollinating insects for image analysis during the summer of December2014ndashJanuary 2015 Insects were chilled immediately and then killed by freezing within 1day and stored atminus18 C in individual vials All insects were identified to species level withassistance from expert taxonomists

Image processingWemeasured the hairiness of 10 insect pollinator species (n= 8ndash10 individuals per species)across five families and two orders This included social semi-social and solitary bees andpollinating flies Regions marked included (1) face (2) head dorsal (3) head ventral

Stavert et al (2016) PeerJ DOI 107717peerj2779 618

(4) front leg (5) thorax dorsal (6) thorax ventral (7) abdomen dorsal and (8) abdomenventral All entropy analysis was carried out using our image processing method outlinedabove For estimates of body size we took multiple linear measurements (body lengthbody width head length head width foreleg length and hind leg length) of each specimenusing digital callipers and a dissecting microscope

Single visit pollen deposition (SVD) and pollen loadFor B rapa we used SVD data for insect pollinators presented in Rader et al (2009) andHowlett et al (2011) a brief description of their methods follows

Pollen deposition on stigmatic surfaces (SVD) was estimated using manipulationexperiments Virgin B rapa inflorescences were bagged to exclude all pollinators Onceflowers had opened the bag was removed and flowers were observed until an insect visitedand contacted the stigma in a single visit The stigma was then removed and stored ingelatine-fuchsin and the insect was captured for later identification SVD was quantifiedby counting all B rapa pollen grains on the stigma Mean values of SVD for each speciesare used in our regression models

To quantify the number of pollen grains carried (pollen load) sensuHowlett et al (2011)collected insects while foraging on B rapa flowers Insects were captured using plastic vialscontaining a rapid killing agent (ethyl acetate) Once dead a cube of gelatine-fuchsin wasused to remove all pollen from the insectrsquos body surface Pollen collecting structures (egcorbiculae scopae) were not included in analyses because pollen from these structures is notavailable for pollination Slides were prepared in the field by melting the gelatine-fuchsincubes containing pollen samples onto microscope slides B rapa pollen grains from eachsample were then quantified by counting pollen grains in an equal-area subset from thesample and multiplying this by the number of equivalent sized subset areas within the totalsample

Wemeasured SVD for A deliciosa (n= 8ndash12 per pollinator species) SVDmeasurementswere taken for insect movements from staminate to pistillate flowers using a methodthat differed from B rapa Individual pistillate buds were enclosed within paper bags2ndash3 days prior to opening and were later used as test flowers to evaluate pollen depositionby flowering visiting species Each bag was secured using a wire tie (coated in plastic) thatwas gently twisted to exclude pollinators from visiting the opening flowers Followingflower opening the bag was removed and the flower pedicel abscised where it joinedthe vine The test flower was then carefully positioned using forceps to hold the pedicel1ndash2 cm from a staminate flower containing a foraging insect avoiding any contactingbetween flowers If the test flower was visited by an insect we allowed it to forage withminimal disturbance until it moved from the flower on its own accord The first stigmatouched by the foraging insect was then lightly marked near its base using a fine blackfelt pen We then placed the marked stigma onto a slide and applied a drop of Alexanderstain (Dafni 2007) Alexander stain was used due to its effectiveness to stain staminateand pistillate pollen differently (pistillate pollenmdashgreen-blue staminate pollenmdashdark red)(Goodwin amp Perry 1992)

Stavert et al (2016) PeerJ DOI 107717peerj2779 718

Statistical analysesWe used linear regression models and AICC (small sample corrected Akaike informationcriteria) model selection to determine if our measure of pollinator hairiness is a goodpredictor of SVD and pollen load We constructed global models with SVD or pollen loadas the response variable body region as predictors and body length as an interaction ieSVD or pollen load simbody length entropy face + entropy head dorsal + entropy headventral + front leg + entropy thorax dorsal + entropy thorax ventral + entropy abdomendorsal + entropy abdomen ventral We included body length in our global model as aproxy for body size as it had high correlation coefficients (Pearsonrsquos r gt 07) with allother body size measurements Global linear models were constructed using the lm(stats)function AICC model selection was carried out on the global models using the functionglmulti() with fitfunction = lsquolsquolmrsquorsquo in the package glmulti We examined heteroscedasticityand normality of errors of models by visually inspecting diagnostic plots using the glmultipackage (Crawley 2002) Variance inflation factors (VIF) of predictor variables werechecked for the best models using the vif() function in the car package All analyses weredone in R version 324 (R Core Team 2014)

RESULTSBody hairiness as a predictor of SVDFor SVD on B rapa the face and thorax dorsal regions were retained in the best modelselected by AICC which had an adjusted R2 value of 098 The subsequent top modelswithin 10 AICC points all retained the face and thorax dorsal regions and additionallyincluded the abdomen ventral (adjusted R2

= 098) head dorsal (adjusted R2= 098) and

thorax ventral (adjusted R2= 097) and front leg (adjusted R2

= 097) regions respectively(Table 1 Fig 2) The model with the face region included as a single predictor had anadjusted R2 value of 088 indicating that this region alone explained a majority of thevariation in the top SVD models

The best model for predicting SVD on A deliciosa included the face and thorax ventralregions as predictors (adjusted R2

= 091) (Table 1 Fig 3) However the subsequent topfour models were within two AICC points of the best model and therefore cannot bediscounted as the potential top model The face thorax ventral head ventral and abdomenventral regions were retained in four of the five top models which indicates that hairinessof the face and ventral regions is important for pollen deposition on A deliciosa For bothB rapa and A deliciosa body length and the body length interaction were not included inthe top models

Body hairiness as a predictor of pollen loadThe best model for pollen load retained the face region only and had an adjusted R2 value of081 (Fig 4 Table 1) The subsequent best models retained the abdomen dorsal (adjustedR2 value of 073) the face and head dorsal (adjusted R2

= 083) the face and abdomendorsal (adjusted R2

= 082) and the abdomen dorsal and front leg (adjusted R2= 08)

regions respectively For pollen load body length and the body length interaction were notincluded in the top models

Stavert et al (2016) PeerJ DOI 107717peerj2779 818

Table 1 Regressionmodels examining the effect of entropy on SVD and pollen load Top regression models examining the effect of insect bodyregion entropy on single visit pollen deposition (SVD) for Brassica rapa and Actinidia deliciosa and pollen load for B rapa Models are presented inascending order based on AICC values Top models for each response variable are highlighted in bold

Responsevariable

Model Adj R2 AICc 1i w i accw i

Face+ Thorax dorsal 098 8829 000 082 082Face+ Thorax dorsal+ Abdomen ventral 098 9309 480 007 089Face+Head dorsal+ Thorax dorsal 098 9381 552 005 094Face+ Thorax ventral+ Thorax dorsal 097 9659 829 001 096

SVD (B rapa)

Face+ Thorax dorsal+ Front leg 097 9702 872 001 097Face 081 16847 000 064 064Abdomen dorsal 073 17159 312 013 078Face+Head dorsal 083 17359 512 005 083Face+ Abdomen dorsal 082 17376 529 005 087

Pollen load(B rapa)

Abdomen dorsal+ Front leg 080 17486 639 003 090Face+ Thorax ventral 091 7418 000 015 015Abdomen dorsal 081 7421 003 015 030Face 080 7435 017 014 045Head ventral 079 7484 066 011 056

SVD (A deliciosa)

Abdomen ventral 078 7508 090 010 065

Notes1i is the difference in the AICC value of each model compared with the AICC value for the top model wi is the Akaike weight for each model and acc wi is the cumulative Akaikeweight

DISCUSSIONHere we present a rigorous and time-efficient method for quantifying hairiness anddemonstrate that this measure is an important pollinator functional trait We show thatinsect pollinator hairiness is a strong predictor of SVD for the open-pollinated flowerB rapa Linear models that included multiple body regions as predictors had the highestpredictive power the top model for SVD retained the face and thorax dorsal regionsHowever the face region was retained in all of the top models and when included as asingle predictor had a very strong positive association with SVD In addition we showthat hairiness particularly on the face and ventral regions is a good predictor of SVD forA deliciosa which has a different floral morphology suggesting our method could besuitable for a range of flower types Hairiness was also a good predictor for pollen load andthe face region was again retained in the top model for B rapa The abdomen dorsal headdorsal and front leg regions were also good predictors of pollen load and were retainedin the subsequent top models Our results validate the importance of insect body hairsfor transporting and depositing pollen Surprisingly we did not find strong associationsbetween SVD and body size and top models did not contain the body length interactionSimilarly body length was not retained in the top models for pollen load This indicatesthat our measure of hairiness has far greater predictive power than body size for both SVDand pollen load

When deciding on which body regions to measure hairiness researchers may first needto assess additional pollinator traits such as flower visiting behaviour This is because

Stavert et al (2016) PeerJ DOI 107717peerj2779 918

Abdomen dorsal Abdomen ventral Face

Front leg Head dorsal Head ventral

Thorax dorsal Thorax ventral

0

50

100

150

200

0

50

100

150

200

0

50

100

150

200

120 140 160 180 120 140 160 180

Mean entropy value

Sin

gle

visi

t pol

len

depo

stio

n on

stig

ma

SpeciesApis mellifera

Bombus terrestris

Calliphora stygia

Eristalis tenax

Helophilus hochstetteri

Lasioglossum sordidum

Leioproctus sp

Melangyna novaezelandiae

Melanostoma fasciatum

Odontomyia sp

120 140 160 180

Figure 2 Relationships betweenmean entropy for each body region andmean single visit pollen de-position (SVD) on Brassica rapa for 10 different insect pollinator species Black lines are regressions forsimple linear models

the way in which insects interact with flowers influences what body parts most frequentlycontact the floral reproductive structures (Roubik 2000) For some open pollinated flowerssuch as B rapa facial hairs are probably the most important for pollen deposition becausethe face is the most likely region to contact the anthers and stigma However for flowerswith different floral morphologies facial hairs may not be as important because the floralreproductive structures have different positions relative to the insectrsquos body structuresFor example disc-shaped flowers tend to deposit their pollen on the ventral regions ofpollinators while labiate flowers deposit their pollen on the dorsal regions (BartomeusBosch amp Vilagrave 2008) We found that hairiness on the face and ventral regions of pollinatorswas most important for pollen deposition on A deliciosa flowers The reproductive partsof A deliciosa form a brush shaped structure and therefore are most likely to contact theface and ventral surfaces of pollinators Accordingly where studies focus on a single plantspecies ie crop based studies it is important to consider trait matching when selectingpollinator body region(s) to analyse (Butterfield amp Suding 2013 Garibaldi et al 2015)

It is important to consider that pollinator performance is a function of both SVDand visitation frequency and these two components operate independently (KremenWilliams amp Thorp 2002 Mayfield Waser amp Price 2001) Here we focus on a single traitthat is important for pollinator efficiency (SVD) but to calculate pollinator performanceresearchers need to measure both efficiency and visitation rate Additional pollinator

Stavert et al (2016) PeerJ DOI 107717peerj2779 1018

20

40

60

80

100

20

40

60

80

100

20

40

60

80

100

140 160 180 140 160 180

Mean entropy value

Sin

gle

visi

t pol

len

depo

stio

n on

stig

ma

Apis mellifera

Bombus terrestris

Eristalis tenax

Lasioglossum sordidum

Leioproctus sp

Melangyna novaezelandiae

Abdomen dorsal Abdomen ventral Face

Front leg Head dorsal Head ventral

Thorax dorsal Thorax ventral

Species

Helophilus hochstetteri

140 160 180

Figure 3 Relationships betweenmean entropy for each body region andmean single visit pollen depo-sition (SVD) on Actinidia deliciosa for 7 different insect pollinator species Black lines are regressionsfor simple linear models

traits related to visitation rate as well as other behavioural traits such as activity patternsrelative to the timing of stigma receptivity (Potts Dafni amp Nersquoeman 2001) and foragingbehaviour eg nectar vs pollen foraging (Herrera 1987 Javorek Mackenzie amp Kloet2002 Rathcke 1983) may be important for predicting pollination performance Insome circumstances it might also be important to consider trait differences betweenmale and female pollinators particularly for some bee species Male and female beesmay have different pollen deposition efficiency due to differences in their foragingbehaviour and resource requirements For example female bees are likely to visitflowers to collect pollen for nest provisioning while males simply consume nectar andpollen during visits (Cane Sampson amp Miller 2011) For some flowers male bees have asimilar pollination efficiency compared to females (eg summer squash Cucurbita pepoCane Sampson amp Miller 2011) while for others female bees are more effective than males(eg lowbush blueberry Vaccinium angustifolium Javorek Mackenzie amp Kloet 2002)

For community-level studies that use functional diversity approaches our methodcould be used to quantify hairiness for several body regions and weighted to give betterrepresentation of trait diversity within the pollinator community This is necessarywhere plant communities contain diverse floral traits ie open-pollinated vs closed-tubular flowers (Fontaine et al 2006) Hairs on different areas of the insect body are

Stavert et al (2016) PeerJ DOI 107717peerj2779 1118

0

2500

5000

7500

10000

12500

0

2500

5000

7500

10000

12500

0

2500

5000

7500

10000

12500

120 140 160 180 120 140 160 180

Mean entropy value

Free

pol

len

load

SpeciesApis mellifera

Bombus terrestris

Calliphora stygia

Eristalis tenax

Lasioglossum sordidum

Leioproctus sp

Melangyna novaezelandiae

Melanostoma fasciatum

Odontomyia sp

Abdomen dorsal Abdomen ventral Face

Front leg Head dorsal Head ventral

Thorax dorsal Thorax ventral 120 140 160 180

Figure 4 Relationship between entropy and Brassica rapa pollen load on insects Relationships be-tween mean entropy for each body region and the mean number of Brassica rapa pollen grains carried by 9different insect pollinator species Black lines are regressions for simple linear models

likely to vary in relative importance for pollen deposition depending on trait matching(Bartomeus et al 2016) Our method requires hairiness to be measured at the individual-level (Fig S1) which makes it an ideal trait to use in new functional diversity frameworksthat use trait probabilistic densities rather than trait averages (Carmona et al 2016 Fontanaet al 2016) Combining predictive traits such as pollinator hairiness with new methodsthat amalgamate intraspecific trait variation with multidimensional functional diversitywill greatly improve the explanatory power of trait-based pollination studies

One of the greatest constraints to advancing trait-based ecology is the time-demandingnature of collecting trait data This is because ecological communities typically containmany species which have multiple traits that need to be measured and replicated (Petcheyamp Gaston 2006) To improve the predictive power of trait-based ecology and streamline thedata collection process we must firstly identify traits that are strongly linked to ecosystemfunctions and secondly develop rigorous and time-efficient methodologies to measuretraits at the individual level We achieve this by providing a method for quantifying ahighly predictive trait at the individual-level in a time-efficient manner Our methodalso complements other recently developed predictive methods for estimating difficult-to-measure traits that are important for pollination processes ie bee tongue lengthCariveau et al (2016)

Stavert et al (2016) PeerJ DOI 107717peerj2779 1218

Predicating the functional importance of organisms is critical in a rapidly changingenvironment where accelerating biodiversity loss threatens ecosystem functions (McGillet al 2015) Our novel method for measuring pollinator hairiness could be used in anystudies that require quantification of hairiness such as understanding adhesion in insects(Bullock Drechsler amp Federle 2008 Clemente et al 2010) or epizoochory (Albert et al2015 Sorensen 1986) It is also a much needed addition to the pollination biologistrsquostoolbox and will progress the endeavour to standardise trait-based approaches inpollination research This is a crucial step towards developing a strong mechanisticunderpinning for trait-based pollination research

ACKNOWLEDGEMENTSWe thank David Seldon Adrian Turner and Iain McDonald for assistance photographinginsect specimens Anna Kokeny for help collecting specimens and Stephen Thorpe forassistance identifying specimens Patrick Garvey and Greg Holwell provided insightfulcomments on the earlier manuscript We also thank two anonymous reviewers for helpfuland constructive comments on the manuscript We thank Sam Read Brian CuttingHeather McBrydie Alex Benoist Rachel Lrsquohelgoualcrsquoh and Simon Cornut for assistance infield work Lastly we thank Estacioacuten Bioloacutegica de Dontildeana for hosting JS while developingthe methodology for this paper

ADDITIONAL INFORMATION AND DECLARATIONS

FundingThis research was supported by the University of Auckland BeeFun project PCIG14-GA-2013-631653 andMBIE C11X1309 Bee Minus to Bee Plus and Beyond Higher Yields FromSmarter Growth-focused Pollination Systems The funders had no role in study designdata collection and analysis decision to publish or preparation of the manuscript

Grant DisclosuresThe following grant information was disclosed by the authorsBeeFun project PCIG14-GA-2013-631653MBIE C11X1309

Competing InterestsBrad G Howlett and David E Pattemore are employees of The New Zealand Institute forPlant amp Food Research Limited All other authors have no competing interests

Author Contributionsbull Jamie R Stavert conceived and designed the experiments performed the experimentsanalyzed the data contributed reagentsmaterialsanalysis tools wrote the paperprepared figures andor tables reviewed drafts of the paperbull Gustavo Lintildeaacuten-Cembrano and Ignasi Bartomeus conceived and designed theexperiments analyzed the data contributed reagentsmaterialsanalysis tools wrotethe paper reviewed drafts of the paper

Stavert et al (2016) PeerJ DOI 107717peerj2779 1318

bull Jacqueline R Beggs conceived and designed the experiments wrote the paper revieweddrafts of the paperbull Brad G Howlett conceived and designed the experiments performed the experimentsreviewed drafts of the paperbull David E Pattemore conceived and designed the experiments performed the experiments

Data AvailabilityThe following information was supplied regarding data availability

The raw data has been supplied as a Supplemental File

Supplemental InformationSupplemental information for this article can be found online at httpdxdoiorg107717peerj2779supplemental-information

REFERENCESAlbert A Auffret AG Cosyns E Cousins SAO DrsquoHondt B Eichberg C Eycott AE

Heinken T HoffmannM Jaroszewicz B Malo JE Maringrell A Mouissie M PakemanRJ PicardM Plue J Poschlod P Provoost S Schulze KA Baltzinger C 2015Seed dispersal by ungulates as an ecological filter a trait-based meta-analysis Oikos1241109ndash1120 DOI 101111oik02512

Bartomeus I Bosch J Vilagrave M 2008High invasive pollen transfer yet low deposition onnative stigmas in a Carpobrotus-invaded community Annals of Botany 102417ndash424DOI 101093aobmcn109

Bartomeus I Gravel D Tylianakis JM AizenMA Dickie IA Bernard-Verdier M2016 A common framework for identifying linkage rules across different types ofinteractions Functional Ecology DOI 1011111365-243512666

Bolnick DI Amarasekare P AraujoMS Burger R Levine JM NovakM RudolfVH Schreiber SJ UrbanMC Vasseur DA 2011Why intraspecific trait varia-tion matters in community ecology Trends in Ecology amp Evolution 26183ndash192DOI 101016jtree201101009

Bullock JM Drechsler P FederleW 2008 Comparison of smooth and hairy attachmentpads in insects friction adhesion and mechanisms for direction-dependence TheJournal of Experimental Biology 2113333ndash3343 DOI 101242jeb020941

Butterfield BJ Suding KN 2013 Single-trait functional indices outperform multi-traitindices in linking environmental gradients and ecosystem services in a complexlandscape Journal of Ecology 1019ndash17 DOI 1011111365-274512013

Cadotte MW Carscadden K Mirotchnick N 2011 Beyond species functional diversityand the maintenance of ecological processes and services Journal of Applied Ecology481079ndash1087 DOI 101111j1365-2664201102048x

Cane JH Sampson BJ Miller SA 2011 Pollination value of male bees the specialist beePeponapis pruinosa (Apidae) at summer squash (Cucurbita pepo) EnvironmentalEntomology 40614ndash620 DOI 101603EN10084

Stavert et al (2016) PeerJ DOI 107717peerj2779 1418

Cariveau DP Nayak GK Bartomeus I Zientek J Ascher JS Gibbs J Winfree R2016 The allometry of bee proboscis length and its uses in ecology PLoS ONE11e0151482 DOI 101371journalpone0151482

Carmona CP De Bello F Mason NWH Lepš J 2016 Traits without bordersintegrating functional diversity across scales Trends in Ecology amp EvolutionDOI 101016jtree201602003

Clemente CJ Bullock JM Beale A FederleW 2010 Evidence for self-cleaning in fluid-based smooth and hairy adhesive systems of insects The Journal of ExperimentalBiology 213635ndash642 DOI 101242jeb038232

Craig JL Stewart AM Pomeroy N Heath A Goodwin R 1988 A review of kiwifruitpollination where to next New Zealand Journal of Experimental Agriculture16385ndash399 DOI 10108003015521198810425667

Crawley MJ 2002 Statistical computing an introduction to data analysis using S-PlusChichester John Wiley amp Sons Ltd

Dafni A 2007 Pollination ecology A practical approach New York Oxford UniversityPress

Devi I Thakur BS Garg S 2015 Floral morphology pollen viability and pollinizerefficacy of kiwifruit International Journal of Current Research and Academic Review3188ndash195

Didham RK Leather SR Basset Y 2016 Circle the bandwagonsndashchallenges mountagainst the theoretical foundations of applied functional trait and ecosystem serviceresearch Insect Conservation and Diversity 91ndash3 DOI 101111icad12150

Fontaine C Dajoz I Meriguet J LoreauM 2006 Functional diversity of plantndashpollinator interaction webs enhances the persistence of plant communities PLoSBiology 40129ndash0135 DOI 101371journalpbio0040129

Fontana S Petchey OL Pomati F Sayer E 2016 Individual-level trait diversity conceptsand indices to comprehensively describe community change in multidimensionaltrait space Functional Ecology 30808ndash818 DOI 1011111365-243512551

Gagic V Bartomeus I Jonsson T Taylor AWinqvist C Fischer C Slade EM Steffan-Dewenter I EmmersonM Potts SG Tscharntke TWeisserW BommarcoR 2015 Functional identity and diversity predict ecosystem functioning betterthan species-based indices Proceedings of the Royal Society B Biological Sciences28220142620 DOI 101098rspb20142620

Garibaldi LA Bartomeus I Bommarco R Klein AM Cunningham SA AizenMABoreux V Garratt MPD Carvalheiro LG Kremen C Morales CL Schuumlepp CChacoff NP Freitas BM Gagic V Holzschuh A Klatt BK Krewenka KM KrishnanS Mayfield MMMotzke I OtienoM Petersen J Potts SG Ricketts TH RundloumlfM Sciligo A Sinu PA Steffan-Dewenter I Taki H Tscharntke T Vergara CHViana BFWoyciechowski M 2015 Editorrsquos choice review trait matching of flowervisitors and crops predicts fruit set better than trait diversity Journal of AppliedEcology 521436ndash1444 DOI 1011111365-266412530

Gonzales RCWoods RE Eddins SL 2004Digital image processing using MATLABLondon Parson Education Ltd

Stavert et al (2016) PeerJ DOI 107717peerj2779 1518

Goodwin RM Perry JH 1992 Use of pollen traps to investigate the foraging behaviourof honey bee colonies in kiwifruit orchards New Zealand Journal of Crop andHorticultural Science 2023ndash26 DOI 10108001140671199210422322

Herrera CM 1987 Components of pollinator lsquolsquoqualityrsquorsquo comparative analysis of adiverse insect assemblage Oikos 5079ndash90

Hillebrand H Matthiessen B 2009 Biodiversity in a complex world consolidationand progress in functional biodiversity research Ecology Letters 121405ndash1419DOI 101111j1461-0248200901388x

Hoehn P Tscharntke T Tylianakis JM Steffan-Dewenter I 2008 Functional groupdiversity of bee pollinators increases crop yield Proceedings of the Royal Society BBiological Sciences 2752283ndash2291 DOI 101098rspb20080405

Holloway BA 1976 Pollen-feeding in hover-flies (Diptera Syrphidae) New ZealandJournal of Zoology 3339ndash350 DOI 1010800301422319769517924

Howlett BGWalker MK Rader R Butler RC Newstrom-Lloyd LE Teulon DAJ 2011Can insect body pollen counts be used to estimate pollen deposition on pak choistigmas New Zealand Plant Protection 6425ndash31

Javorek S Mackenzie K Vander Kloet S 2002 Comparative pollination effectivenessamong bees (Hymenoptera Apoidea) on lowbush blueberry (Ericaceae Vac-cinium angustifolium) Annals of the Entomological Society of America 95345ndash351DOI 1016030013-8746(2002)095[0345CPEABH]20CO2

King C Ballantyne GWillmer PG 2013Why flower visitation is a poor proxy forpollination measuring single-visit pollen deposition with implications for polli-nation networks and conservationMethods in Ecology and Evolution 4811ndash818DOI 1011112041-210X12074

Klein A-M Vaissiere BE Cane JH Steffan-Dewenter I Cunningham SA Kremen CTscharntke T 2007 Importance of pollinators in changing landscapes for worldcrops Proceedings of the Royal Society of London B Biological Sciences 274303ndash313DOI 101098rspb20063721

Kremen CWilliams NM AizenMA Gemmill-Herren B LeBuhn G Minckley RPacker L Potts SG Roulston T Steffan-Dewenter I Vaacutezquez DPWinfree RAdams L Crone EE Greenleaf SS Keitt TH Klein AM Regetz J Ricketts TH2007 Pollination and other ecosystem services produced by mobile organisms aconceptual framework for the effects of land-use change Ecology Letters 10299ndash314DOI 101111j1461-0248200701018x

Kremen CWilliams NM Thorp RW 2002 Crop pollination from native bees at riskfrom agricultural intensification Proceedings of the National Academy of Sciences ofthe United States of America 9916812ndash16816 DOI 101073pnas262413599

Larsen THWilliams NM Kremen C 2005 Extinction order and altered commu-nity structure rapidly disrupt ecosystem functioning Ecology Letters 8538ndash547DOI 101111j1461-0248200500749x

Lavorel S Storkey J Bardgett RD De Bello F Berg MP 2013 A novel frame-work for linking functional diversity of plants with other trophic levels for the

Stavert et al (2016) PeerJ DOI 107717peerj2779 1618

quantification of ecosystem services Journal of Vegetation Science 24942ndash948DOI 101111jvs12083

Mayfield MMWaser NM Price MV 2001 Exploring the lsquomost effective pollinatorprinciplersquo with complex flowers bumblebees and Ipomopsis aggregata Annals ofBotany 88591ndash596 DOI 101006anbo20011500

McGill BJ Dornelas M Gotelli NJ Magurran AE 2015 Fifteen forms of biodi-versity trend in the Anthropocene Trends in Ecology amp Evolution 30104ndash113DOI 101016jtree201411006

McGill BJ Enquist BJ Weiher E WestobyM 2006 Rebuilding communityecology from functional traits Trends in Ecology amp Evolution 21178ndash185DOI 101016jtree200602002

Naeem SWright JP 2003 Disentangling biodiversity effects on ecosystem function-ing deriving solutions to a seemingly insurmountable problem Ecology Letters6567ndash579 DOI 101046j1461-0248200300471x

Nersquoeman G Juumlrgens A Newstrom-Lloyd L Potts SG Dafni A 2010 A framework forcomparing pollinator performance effectiveness and efficiency Biological Reviews85435ndash451 DOI 101111j1469-185X200900108x

Ollerton J Winfree R Tarrant S 2011How many flowering plants are pollinated byanimals Oikos 120321ndash326 DOI 101111j1600-0706201018644x

Pasari JR Levi T Zavaleta ES Tilman D 2013 Several scales of biodiversity affectecosystem multifunctionality Proceedings of the National Academy of Sciences of theUnited States of America 11010219ndash10222 DOI 101073pnas1220333110

Petchey OL Gaston KJ 2006 Functional diversity back to basics and looking forwardEcology Letters 9741ndash758 DOI 101111j1461-0248200600924x

Potts SG Dafni A Nersquoeman G 2001 Pollination of a core flowering shrub species inMediterranean phrygana variation in pollinator diversity abundance and effective-ness in response to fire Oikos 9271ndash80 DOI 101034j1600-07062001920109x

R Core Team 2014 R a language and environment for statistical computing Vienna RFoundation for Statistical Computing Available at httpwwwR-projectorg

Rader R EdwardsWWestcott DA Cunningham SA Howlett BG 2013 Diur-nal effectiveness of pollination by bees and flies in agricultural Brassica rapaimplications for ecosystem resilience Basic and Applied Ecology 1420ndash27DOI 101016jbaae201210011

Rader R Howlett BG Cunningham SAWestcott DA Newstrom-Lloyd LEWalkerMK Teulon DAJ EdwardsW 2009 Alternative pollinator taxa are equally efficientbut not as effective as the honeybee in a mass flowering crop Journal of AppliedEcology 461080ndash1087 DOI 101111j1365-2664200901700x

Rathcke B 1983 Competition and facilitation among plants for pollination InPollination biology New York Academic Press 305ndash329

Roubik DW 2000 Deceptive orchids with Meliponini as pollinators Plant Systematicsand Evolution 222271ndash279 DOI 101007BF00984106

Shannon C 1948 A mathematical theory of communication Bell System TechnicalJournal 3379ndash423 DOI 101002j1538-73051948tb01338x

Stavert et al (2016) PeerJ DOI 107717peerj2779 1718

Sorensen AE 1986 Seed dispersal by adhesion Annual Review of Ecology and Systematics17443ndash463

Thorp RW 2000 The collection of pollen by bees Plant Systematics and Evolution222(1)211ndash223 DOI 101007BF00984103

Violle C Enquist BJ McGill BJ Jiang L Albert CH Hulshof C Jung V Messier J 2012The return of the variance intraspecific variability in community ecology Trends inEcology amp Evolution 27244ndash252 DOI 101016jtree201111014

Violle C Navas M-L Vile D Kazakou E Fortunel C Hummel I Garnier E 2007 Letthe concept of trait be functional Oikos 116882ndash892DOI 101111j0030-1299200715559x

Walker B Kinzig A Langridge J 1999 Plant attribute diversity resilience and ecosys-tem function the nature and significance of dominant and minor species Ecosystems295ndash113 DOI 101007s100219900062

Stavert et al (2016) PeerJ DOI 107717peerj2779 1818