Submitted 24 April 2018Accepted 4 September 2018Published 2 October 2018

Corresponding authorDerek E Leederekwildnatureinstituteorg

Academic editorDany Garant

Additional Information andDeclarations can be found onpage 17

DOI 107717peerj5690

Copyright2018 Lee et al

Distributed underCreative Commons CC-BY 40

OPEN ACCESS

Seeing spots quantifying mother-offspring similarity and assessing fitnessconsequences of coat pattern traits ina wild population of giraffes (Giraffacamelopardalis)Derek E Lee12 Douglas R Cavener2 and Monica L Bond13

1Wild Nature Institute Concord NH United States of America2Department of Biology Pennsylvania State University University Park United States of America3Department of Evolutionary Biology and Environmental Studies University of Zuumlrich Zuumlrich Switzerland

ABSTRACTPolymorphic phenotypes of mammalian coat coloration have been important to thestudy of genetics and evolution but less is known about the inheritance and fitnessconsequences of individual variation in complex coat pattern traits such as spotsand stripes Giraffe coat markings are highly complex and variable and it has beenhypothesized that variation in coat patterns most likely affects fitness by camouflagingneonates against visually hunting predators We quantified complex coat pattern traitsof wild Masai giraffes using image analysis software determined the similarity of spotpattern traits between mother and offspring and assessed whether variation in spotpattern traits was related to fitness as measured by juvenile survival The methods wedescribed could comprise a framework for objective quantification of complexmammalcoat pattern traits based on photographic coat pattern data We demonstrated thatsome characteristics of giraffe coat spot shape were likely to be heritable as measuredby mother-offspring regression We found significant variation in juvenile survivalamong phenotypic groups of neonates defined by multivariate clustering based on spottrait measurement variables We also found significant variation in neonatal survivalassociated with spot size and shape covariates Larger spots (smaller number of spots)and irregularly shaped or rounder spots (smaller aspect ratio) were correlated withincreased survival These findings will inform investigations into developmental andgenetic architecture of complex mammal coat patterns and their adaptive value

Subjects Bioinformatics Evolutionary Studies ZoologyKeywords Adaptation Biometrics Phenomics Heritability Natural selection Remotemeasurement Phenotypic selection Coat pattern Quantitative genetics Visual animal biometry

INTRODUCTIONComplex color patterns such as spots and stripes are found on many animal speciesand these phenotypic traits are hypothesized to play adaptive roles in predator andparasite evasion thermoregulation and communication (Cott 1940 Caro 2005) Manyfoundational studies of coloration using starkly different color morphs from diverse taxa

How to cite this article Lee et al (2018) Seeing spots quantifying mother-offspring similarity and assessing fitness consequences of coatpattern traits in a wild population of giraffes (Giraffa camelopardalis) PeerJ 6e5690 DOI 107717peerj5690

such as insects (Kettlewell 1955 Wittkopp et al 2003) mice (Morse 1978 Russell 1985Bennett amp Lamoreux 2003) reptiles (Rosenblum Hoekstra amp Nachman 2004 CalsbeekBonneaud amp Smith 2008) fish (Endler 1983 Irion Singh amp Nuesslein-Volhard 2016) andbirds (Roulin 2004) demonstrated Mendelian inheritance and natural selection anddiscovered genes that cause color morph mutations (Hoekstra 2006 Protas amp Patel 2008San-Jose amp Roulin 2017) Individual variation in a complex color pattern trait of spot sizewas also part of the earliest work on genetics and inheritance (Wright 1917) Measuringindividual variation in complex color patterns especially detailed measurements such asanimal biometrics (Kuumlhl amp Burghardt 2013) can provide novel insight into developmentaland genetic architecture (Bowen amp Dawson 1977 Klingenberg 2010 San-Jose amp Roulin2017) and the adaptive value of the patterns (Hoekstra 2006 Allen et al 2011) as wellas benefitting studies of behavior (Lorenz 1937 Whitehead 1990) population biology(Holmberg Norman amp Arzoumanian 2009 Lee amp Bolger 2017) and the growing fieldof phenomics (Houle Govindaraju amp Omholt 2010) A few methods to robustly quantifycontinuous variation among individuals in complex color patterns have been developed forgeneral use (Schneider Rasband amp Eliceiri 2012 Van Belleghem et al 2018) and specifictaxa such as fishes (Endler 1980 Holmberg Norman amp Arzoumanian 2009) butterflies(Le Poul et al 2014) penguins (Sherley et al 2010) and primates (Allen Higham amp Allen2015) We see a need for more tools and techniques to reliably quantify individual variationin complex coat pattern traits in wild populations (Eizirik et al 2010Willisch Marreros ampNeuhaus 2013) and studies that use quantitative genetics and demographic methods toinvestigate heritability and adaptive significance of those traits in wildmammal populations(Kruuk Slate amp Wilson 2008 Kaelin et al 2012)

The coat patterns of Masai giraffes (Giraffa camelopardalis tippelskirchii) are complexand show a high degree of individual variation (Dagg 1968 Fig 1) Masai giraffesrsquo spotsvary in color and shape from those that are nearly round with very smooth edges (lowtortuousness) to extremely elliptical with incised or lobate edges (high tortuousness)Giraffe skin pigmentation is uniformly dark grey (Dimond amp Montagna 1976) but thespots that make up their coat markings are highly variable in traits such as color roundnessand perimeter tortuousness This variation has been used to classify subspecies (Lydekker1904) and to reliably identify individuals because patterns do not change with age (Foster1966 Bolger et al 2012Dagg 2014)Dagg (1968) first presented evidence from a small zoopopulation that the shape number area and color of spots in giraffe coat patterns may beheritable but analysis of spot traits in wild giraffes and objective measurements of spotcharacteristics in general have been lacking

It has been hypothesized that giraffe coat patterns evolved to camouflage neonateswhose primary defense against predation is concealment (Langman 1977 Mitchell ampSkinner 2003) thus the most likely fitness effects from variation in coat patterns shouldbe variation in juvenile survival Giraffe calves spend much of their time day and nighthiding in the dappled light of trees and bushes and their ability to match this backgroundshould influence detection by visually hunting predators such as lions and hyenas (Endler1978 Merilaita Scott-Samuel amp Cuthill 2017) Background matching the adaptation ofan animalrsquos coloration to mimic its average background and reduce detection by visually

Lee et al (2018) PeerJ DOI 107717peerj5690 223

Figure 1 Representative images of spot patterns of mother-calf pairs of Masai giraffes (Giraffacamelopardalis tippelskirchii) from the Tarangire ecosystem Tanzania used in this study The bluerectangle shows the area analysed using ImageJ to characterize spot pattern traits All photos by DE Lee(A) Mother-calf pair number 1 (B) mother-calf pair number 2 (C) mother-calf pair number 3 (D)mother-calf pair number 4

Full-size DOI 107717peerj5690fig-1

Lee et al (2018) PeerJ DOI 107717peerj5690 323

hunting predators is a common form of camouflage (Endler 1978Merilaita Scott-Samuelamp Cuthill 2017) Alternative hypotheses about the adaptive value of giraffe coat markingsinclude thermoregulation (Skinner amp Smithers 1990) and in this social species with goodvisual sensory perception (Dagg 2014 VanderWaal et al 2014) markings could alsofacilitate individual recognition (Tibbetts amp Dale 2007) and kin recognition (Beecher1982 Tang-Martinez 2001) To date no evidence has been presented for any of thesehypotheses

Our purpose in this study was to (1) demonstrate the use of public domain imageanalysis software ImageJ (Schneider Rasband amp Eliceiri 2012) to extract patterns fromimage data and quantify multiple aspects of the complex coat patterns of wild Masaigiraffes (2) use quantitative genetics methods (parentndashoffspring regression) to quantifythe proportion of observed phenotypic variation of a trait that is shared betweenmother andoffspring and (3) determine whether variation in complex coat pattern traits was relatedto a measure of fitness (survival) and thereby infer the effect of natural selection (viabilityselection) on giraffe coat patterns (Lande amp Arnold 1983 Falconer amp Mackay 1996)

MATERIALS amp METHODSAs a general overview our methods were to (1) collect field data in one area of Tanzaniaas digital images of giraffes to be used for spot pattern and survival analyses (2) extractpatterns from images (3) quantify giraffe patterns by measuring 11 spot traits (4) useprincipal components analysis (PCA) to reduce the dimensionality of the spot traits (5)use mother-offspring regressions to estimate the phenotypic similarity between motherand offspring of the 11 spot traits and the 1st two dimensions of the PCA (6) use k-meansclustering to assign giraffe calves into phenotypic groups according to their spot patterntraits (7) use capture-mark-recapture analysis to estimate survival and determine whetherthere are fitness differences among the phenotypic groups (8) use capture-mark-recaptureanalysis to determine whether there are fitness effects from any particular spot traits

This research was carried out with permission from the Tanzania Commission forScience and Technology (COSTECH) Tanzania National Parks (TANAPA) the TanzaniaWildlife Research Institute (TAWIRI) African Wildlife Foundation and Manyara RanchConservancy

Field Data CollectionThis study used data from individually identified wild free-ranging Masai giraffes in a1700 km2 sampled area within a 4400 km2 region of the Tarangire Ecosystem northernTanzania East Africa Data were collected as previously described in Lee et al (2016a) Wecollected data during systematic road transect sampling for photographic capture-mark-recapture (PCMR) We conducted 26 daytime surveys for giraffe PCMR data betweenJanuary 2012 and February 2016 We sampled giraffes three times per year around 1February 1 June and 1 October near the end of every precipitation season (short rainslong rains and dry respectively) by driving a network of fixed-route transects on single-lanedirt tracks in the study area We surveyed according to Pollockrsquos robust design samplingframework (Pollock 1982 Kendall Pollock amp Brownie 1995) with three occasions per year

Lee et al (2018) PeerJ DOI 107717peerj5690 423

Each sampling occasion was composed of two sampling events during which we surveyedall transects in the study area with only a few days interval between events Each samplingoccasion was separated by a 4-month interval (43 years times 3 occasions yearminus1 times 2 eventsoccasionminus1 = 26 survey events)

During PCMR sampling events a sample of individuals were encountered and eitherlsquosightedrsquo or lsquoresightedrsquo by slowly approaching and photographing the animalrsquos right sidefrom approximately 150 m at a perpendicular angle (Canon 40D and Rebel T2i cameraswith Canon Ultrasonic IS 100ndash400 mm lens Canon USA Inc One Canon Park MelvilleNew York USA) We identified individual giraffes using their unique and unchanging coatpatterns (Foster 1966 Dagg 2014) with the aid of pattern-recognition software Wild-ID(Bolger et al 2012) We attempted to photograph every giraffe encountered and recordedsex and age class based on physical characteristics We assigned giraffes to one of fourage classes for each observation based on the speciesrsquo life history characteristics and oursampling design neonate calf (0ndash3 months old) older calf (4ndash11 months old) subadult(1ndash3 years old for females 1 ndash6 years old for males) or adult (gt3 years for females gt6 yearsfor males) using a suite of physical characteristics (Strauss et al 2015) and size measuredwith photogrammetry (Lee et al 2016a) In this analysis we used only adult females andanimals first sighted as neonate calves

All animal work was conducted according to relevant national and internationalguidelines This research was carried out with permission from the Tanzania Commissionfor Science and Technology (COSTECH) Research Permit numbers 2017-163-ER-90-1722016-146-ER-2001-31 2015-22-ER-90-172 2014-53-ER-90-172 2013-103-ER-90-1722012-175-ER-90-172 2011-106-NA-90-172 Tanzania National Parks (TANAPA) theTanzania Wildlife Research Institute (TAWIRI) No Institutional Animal Care and UseCommittee (IACUC) approval was necessary because animal subjects were observedwithout disturbance or physical contact of any kind

Quantification of spot patternsWe extracted patterns and analysed spot traits of each animal within the shoulder andrib area by cropping all images to an analysis rectangle that fit horizontally between theanterior edge of the rear leg and the chest and vertically between the back and wherethe skin folded beneath the posterior edge of the foreleg (Fig 1) For color trait analysiswe used the Color Histogram procedure of ImageJ (Schneider Rasband amp Eliceiri 2012)full-color images of the analysis rectangle We extracted coat patterns using ImageJ toconvert full-color images of the analysis rectangle to 8-bit greyscale images then convertedto bicolor (black and white) using the Enhance Contrast and Threshold commands(Schneider Rasband amp Eliceiri 2012) We quantified 10 spot trait measurements of eachanimalrsquos extracted coat pattern using the Analyze Particles command in ImageJ (SchneiderRasband amp Eliceiri 2012) To account for differences in image resolution and animal size(including age-related growth) and to obtain approximately scale-invariant standardimages of each animal we set the measurement unit of each image equal to the numberof pixels in the height of the analysis rectangle Therefore all measurements are in giraffeunits (GU) where 1 GU = height of the analysis rectangle (Fig 1) We excluded spots cut

Lee et al (2018) PeerJ DOI 107717peerj5690 523

off by the edge of the analysis rectangle to avoid the influence of incomplete spots and wealso excluded spots whose area was lt000001 GU2 to eliminate the influence of speckles

We characterized each animalrsquos coat spot pattern traits within the analysis rectangleusing the following 11 metrics available in ImageJ (10 measurements plus color) numberof spots mean spot size (area) mean spot perimeter mean angle between the primaryaxis of an ellipse fit over the spot and the x-axis of the image mean circularity (4πtimes [Area][Perimeter] 2 with a value of 10 indicating a perfect circle and smaller valuesindicating an increasingly elongated shape) mean maximum caliper (the longest distancebetween any two points along the spot boundary also known as Feret diameter) meanFeret angle (the angle [0 to 180 degrees] of the maximum caliper) mean aspect ratio (of thespotrsquos fitted ellipse) mean roundness (4times[Area]πtimes [Major axis]2 or the inverse of aspectratio) mean solidity ([Area][Convex area] also called tortuousness) and mode shade([65536timesr] + [256timesg] + [b] using RGB (red green blue) values from color histogramfrom full color photos) Circularity describes how close the spot is to a perfect circle and ispositively correlated with the trait of roundness Solidity describes how smooth and entirethe spot edges are versus tortuous ruffled lobed or incised and is negatively correlatedwith the trait of perimeter Number is negatively correlated with size and perimeter withall three metrics indicating spot size See Table S2 for all correlations among traits

We quantified total phenotypic variation in spot trait values by reporting the meanSD and coefficient of variation (CV) of each trait We also quantified the repeatability(R) as the within-individual correlation among measurements (Nakagawa amp Schielzeth2010) of spot pattern trait measurement technique for the same animal made on differentphotos from different dates using a set of 30 animals with gt2 images per animal usingpackage rptR (Stoffel Nakagawa amp Schielzeth 2017)Weperformed aprincipal componentsanalysis (PCA Hotelling 1933) on the covariance matrix of the 10 spot trait measurements(standardized to z-scores) to examine the patterns of variation and covariation amongthe spot measurement data and to compute two summary dimensions explaining the10 measurements (color was not included) We performed k-means clustering to divideanimals into lsquocoat pattern phenotypesrsquo phenotypic groups based upon their spot traitcharacteristics (MacQueen 1967 Hartigan 1975) The optimal number of phenotypicgroups was determined by the gap statistic (Tibshirani Walther amp Hastie 2001) Weperformed statistical operations using R (R Core Development Team 2017) packages lmer(Bates et al 2015) FactoMineR (Le Josse amp Husson 2008) and rptR (Stoffel Nakagawa ampSchielzeth 2017)

Mother-offspring similarity of spot traitsThe (narrow sense) heritability of a trait (symbolized h2) is the proportion of its totalphenotypic variance due to additive genetic effects or available for selection to act uponParent-offspring (PO) regression is one of the traditional quantitative genetics toolsused to test for heritable additive genetic variation (Falconer amp Mackay 1996) We usedmother-offspring regression to compute similarity where heritability is 2times the slope of theregression PO regression studies cannot distinguish among phenotypic similarity due togenetic heritability maternal effects or shared environmental effects (Falconer amp Mackay

Lee et al (2018) PeerJ DOI 107717peerj5690 623

1996) it is however one of the few methods available when information on other kinrelations is lacking Pigmentation traits in mammals are known to have a strong geneticbasis (Bennett amp Lamoreux 2003 Hoekstra 2006) supporting the interpretation of POregression as indicating a genetic component We expect minimal non-random variationdue to environmental effects because the calves were all born in the same area with thesame vegetation communities during a relatively short time period of average climate andweather with no spatial segregation by coat pattern phenotype (Fig S1) The animal modelwas not an improvement because we do not know fathers and we had no known siblingsin our dataset therefore PO regression is the most appropriate tool for our estimates ofheritability with the caveat that there are potentially environmental and maternal effectsalso present

We identified 31 mother-calf pairs by observing extended suckling behavior (gt5 s)Wild female giraffes very rarely suckle a calf that is not their own (Pratt amp Anderson1979) We examined all identification photographs for individuals in known mother-calfpairs and selected the best-quality photograph for each animal based on focus clarityperpendicularity to the camera and unobstructed view of the torso

We predicted spot pattern traits of a calf would be correlated with those of its motherWe estimated the mother-offspring similarity for each of the 11 spot trait measurementsand the first two dimensions generated by the PCA When we examined the 11 individualspot traits we used the Bonferroni adjustment (αnumber of tests) to account for multipletests and set our adjusted α= 00045 We performed statistical operations in R (R CoreDevelopment Team 2017)We tested that the PO regressions for each trait met assumptionsof normality of residuals and homoscedasticity using qqPlot and ncvTest functions inpackage car in R (Fox amp Weisberg 2011)

Associations between spot patterns and juvenile survivalWe assembled encounter histories for 258 calves first observed as neonates for survivalanalysis For each calf we selected the best-quality calf-age (age lt 6 mo) photograph basedon focus clarity perpendicularity to the camera and unobstructed view of the torsoand ran the photographs through the ImageJ analysis to quantify each individualrsquos coatspot traits We analysed survival using capture-mark-recapture apparent survival modelsthat account for imperfect detectability during surveys (White amp Burnham 1999) Nocapture-mark-recapture analyses except lsquoknown fatersquo models can discriminate betweenmortality and permanent emigration therefore when we speak of survival it is technicallylsquoapparent survivalrsquo but during the first seasons of life we expected very few calves toemigrate from the study area and if any did emigrate permanently this effect on apparentsurvival should be random relative to their spot pattern characteristics

We ran two analyses of calf survival In the first we estimated age-specific seasonal(4-month seasons) survival (up to 3 years old) according to coat pattern phenotype groupswith calves assigned to groups by k-means clustering of their overall spot traits Wecompared five models a null model of one group age + three groups age times 3 groupsage + four groups and age times four groups to examine whether coat pattern phenotypesaffected survival differently at different ages In the second survival analysis we estimated

Lee et al (2018) PeerJ DOI 107717peerj5690 723

survival as a function of individual covariates of specific spot traits including linear andquadratic relationships of all 11 spot traits and the first two PCA dimensions on juvenilesurvival to examine whether directional disruptive or stabilizing selection was occurring(Lande amp Arnold 1983 Falconer amp Mackay 1996) To determine at what age specific spottraits had the greatest effect of survival we examined survival as a function of spot traitsduring 3 age periods the first season of life first year of life and first three years of life

We used Program MARK to analyse complete capture-mark-recapture encounterhistories of giraffes first sighted as neonates (White amp Burnham 1999) We analysed ourencounter histories using Pollockrsquos Robust Design models to estimate age-specific survival(Pollock 1982 Kendall Pollock amp Brownie 1995) and ranked models using AICc followingBurnham amp Anderson (2002) We used weights (W) and likelihood ratio tests as the metricsfor the strength of evidence supporting a given model as the best description of thedata (Burnham amp Anderson 2002) Due to model selection uncertainty in the analysis ofphenotypic groups we present model-averaged parameter values and based all inferenceson these model-averaged values (Burnham amp Anderson 2002) We considered factors tobe statistically significant if the 95 confidence interval of the beta coefficient did notinclude zero

Based on previous analyses for this population (Lee et al 2016a Lee et al 2016b) weconstrained parameters for survival (S) and temporary emigration (γ prime and γ primeprime) to be linearfunctions of age (symbolized lsquoArsquo) and capture and recapture (c and p) were time dependent(symbolized lsquotrsquo) so the full model was (S(A) γ prime (A) γ primeprime (A) c(t) p(t) Giraffe calf survivaldoes not vary by sex (Lee et al 2016b) so we analysed all calves together as an additionalconstraint on the number of parameters estimated We tested goodness-of-fit in encounterhistory data using U-CARE (Choquet et al 2009) and we found some evidence for lackof fit (χ2

62= 97 P = 001) but because the computed c adjustment was lt3 (c = 15) wefelt our models fit the data adequately and we did not apply a variance inflation factor(Burnham amp Anderson 2002 Choquet et al 2009)

We have deposited the primary data underlying these analyses as follows samplinglocations original data photos and spot trait data Dryad DOI httpsdoiorg105061dryad6514r

RESULTSWe were able to extract patterns and quantify 11 spot traits using ImageJ and foundmeasurements were highly repeatable with low variation in measurements from differentphotos of the same individual (Table 1) From our 31 mother-calf pairs all PO regressionsmet assumptions of normality of residuals and homoscedasticity (Fig S2) We found twospot shape traits circularity and solidity (tortuousness) (Fig S3) had significant PO slopecoefficients between calves and their mothers indicating similarity (Table 1 and Fig 2)

The first dimension from the PCA (from 258 calves including the 31 calves usedto estimate heritability) was composed primarily of spot size-related traits (perimetermaximum caliper area and number) such that increasing dimension 1 meant increasingspot size Dimension 1 explained 405 of the variance in the data (Fig 3) The second

Lee et al (2018) PeerJ DOI 107717peerj5690 823

Table 1 Summary statistics for mother-offspring regressions of spot traits of Masai giraffes in northern TanzaniaMean trait values SD (standard deviation) CV(among-individuals coefficient of variation) Repeatability (within-individual correlation among measurements from different pictures of the same individual) Parent-offspring (PO) slope coefficients F-statistics and P values are provided Statistically significant heritable traits are in bold

Number Area Perimeter Angle Circularity Maximumcaliper

Feretangle

Aspectratio

Roundness Solidity Modeshade

PCA 1stdimension

PCA 2nddimension

Mean 189 004 099 8796 051 029 882 169 063 084 6924050SD 75 001 025 1539 008 006 145 015 004 004 3930565CV 040 039 025 017 015 019 016 009 006 005 057Repeatability (R) 078 078 074 092 082 084 086 09 094 096 074SE of R 030 023 019 019 031 032 016 022 021 027 024P value (R) 0003 0002 0002 0001 0008 0009 0002 0001 0001 0002 0002PO Slope Coefficient 020 020 027 004 052 021 minus015 019 008 053 044 039 021PO Coefficient SE 023 021 018 020 016 021 015 018 017 017 022 021 019Heritability 040 040 054 008 104 042 030 038 016 106 088 078 042F129 076 087 227 004 997 101 091 111 019 973 416 345 111P value (PO) 039 036 014 084 00037 032 035 030 066 00041 005 007 030

Leeetal(2018)PeerJD

OI107717peerj5690

923

Figure 2 Mother-offspring regressions for (A) circularity and (B) solidity values of Masai giraffes innorthern Tanzania These shape traits were significantly correlated between mother and calf

Full-size DOI 107717peerj5690fig-2

Lee et al (2018) PeerJ DOI 107717peerj5690 1023

Figure 3 Contributions of 10 trait measurement variables to the first 2 dimensions of the principalcomponents analysis of giraffe spots The first dimension (Dim1) was composed primarily of spot size-related traits (perimeter maximum caliper area and number of spots) the second dimension (Dim2) wascomposed primarily of spot shape traits (aspect ratio roundness solidity and circularity) C circularityS solidity R roundness N number of spots AR aspect ratio MC maximum caliper P perimeter

Full-size DOI 107717peerj5690fig-3

dimension was composed primarily of spot shape traits (aspect ratio roundness solidityand circularity) such that increasing dimension 2 meant increasing roundness andcircularity while decreasing dimension 2 meant more tortuous edges and irregular shapesDimension 2 explained 240 of the variation in the data (Fig 3) The variance explainedby additional dimensions and the contributions of variables to the first two dimensions aregiven in Table S1 and (Fig S4) None of the dimensions from the PCA had significant POregression slopes (Table 1) Correlations among variables are given in Table S2

Gap statistics indicated either one three or four phenotypic groups was the optimalnumber of clusters for k-means clustering (Fig 4)We examined survival differences amongthree and four phenotypic groups relative to a one-group (null) model In the four-groupdefinition group 1 had medium-sized circular spots group 2 had small-sized circularand irregular spots group 3 had medium-sized irregular spots and group 4 had largecircular and irregular spots (Figs 3 and 4) Groups 1 and 2 had a large amount of overlapin PCA variable space (Fig 4) so we created three phenotypic groups by lumping thetwo overlapping groups Our survival analysis of 258 calves divided into four phenotypic

Lee et al (2018) PeerJ DOI 107717peerj5690 1123

Figure 4 Results from k-means cluster analysis of giraffe spot patterns to define phenotypic groups(A) Gap statistic for different numbers of groups (B) Four clusters mapped in PCA space

Full-size DOI 107717peerj5690fig-4

Table 2 Model selection results for giraffe calf survival according to phenotypic groups defined byspot traitsModel weights indicated some evidence for phenotypic group effects on survival NotationlsquoArsquo indicates a linear trend with age Additive models indicate groups shared a common slope coefficientbut had different intercepts multiplicative models indicated groups had different intercepts and differentslopes Minimum AICc = 323638W = AICc weight k= number of parameters

Model 1AICc W k

A+ 3 groups 0 043 36A+ 1 group 094 027 34A+ 4 groups 206 015 37Atimes 4 groups 301 009 40Atimes 3 groups 391 006 38

groups based on their spot traits indicated that the one-group model was top-rankedbut AICc weights showed there was some evidence for survival variation among the 4phenotypic groups (Table 2) The 3 phenotypic group model found significant differencesin survival according to group (Table 2 the 95 confidence interval of the beta coefficientdid not include zero for lumped groups 1 and 2=minus0717 95 CI = minus1408 to minus0002)Model-averaged seasonal apparent survival estimates indicated differences in survival of004 to 007 existed among phenotypic groups during the first season of life but thosedifferences were greatly reduced in ages 1 and 2 years old (Fig 5)

We found two specific spot traits significantly affected survival during the first seasonof life (number of spots and aspect ratio beta number of spots=minus0031 95 CI = minus0060to minus0007 beta aspect ratio=minus0466 95 CI = minus0957 to minus0002) Both number of spotsand aspect ratio were negatively correlated with survival during the first season of life(Fig 6) No other trait during any age period significantly affected juvenile survival

Lee et al (2018) PeerJ DOI 107717peerj5690 1223

Figure 5 Model-averaged seasonal (4 months) apparent survival estimates for coat pattern phenotypicgroups of giraffes defined by k-means clustering of their spot pattern traits There was evidence for sig-nificant differences in survival among phenotypic groups during the younger ages but those differenceswere greatly reduced as the animals approached adulthood (age 9ndash11 seasons) Error bars areplusmn1 SE

Full-size DOI 107717peerj5690fig-5

(all beta coefficient 95 CIs included zero) but model selection uncertainty was high(Table 3) Number of spots and aspect ratio were not correlated with each other (TableS2)

DISCUSSIONWe were able to objectively and reliably quantify coat pattern traits of wild giraffes usingimage analysis softwareWe demonstrated that some giraffe coat pattern traits of spot shapeappeared to be heritable from mother to calf and that coat pattern phenotypes definedby spot size and shape differed in fitness as measured by neonatal survival Individualcovariates of spot size and shape significantly affected survival during the first 4 monthsof life These results support the hypothesis that giraffe spot patterns are heritable (Dagg1968) and affect neonatal calf survival (Langman 1977 Mitchell amp Skinner 2003) Thefact that spot patterns affected survival could be related to camouflage but could alsoreflect pleiotropy of spot traits with other traits affecting fitness (Wilson amp Nussey 2010Lailvaux amp Kasumovic 2011) or some other effect such as shared environment (Falconer ampMackay 1996) Our methods and results add to the toolbox for objective quantification of

Lee et al (2018) PeerJ DOI 107717peerj5690 1323

Figure 6 Survival of neonatal giraffes during their first 4 months of life was negatively correlated with(A) number of spots and (B) aspect ratioNumber of spots and aspect ratio are inversely related to spotsize and roundness (the variables used when describing coat pattern phenotypic groups) Black lines aremodel estimates grey lines are 95 confidence intervals

Full-size DOI 107717peerj5690fig-6

Lee et al (2018) PeerJ DOI 107717peerj5690 1423

Table 3 Model selection results for giraffe calf survival as a linear or quadratic function of spot traitcovariates during the first season (4 months) first year and first 3 years of life Confidence intervals ofbeta coefficients for two traits excluded zero (number of spots and aspect ratio) indicating evidence forsignificant spot trait effects on calf survival during the first season of life Model structure in all cases wasS(A+Covariate)g primeprime(A)g prime(A)p(t )c(t ) with covariate structure in survival Notation lsquoArsquo indicates a lineartrend with age lsquot rsquo indicates time dependence Minimum AICc = 323987W = AICc weight k = numberof parameters Models comprising the top 50 cumulativeW are shown

Model 1AICc W k

Number of spots 1st season 0 0048 33Aspect ratio 1st season 044 0039 33Roundness2 1st 3 years 082 0032 34Angle2 1st season 087 0031 34Roundness 1st season 095 0030 33Solidity 1st season 106 0029 33Area2 1st season 111 0028 34Circularity 1st season 115 0027 33Angle2 1st 3 years 121 0026 34Null model no covariate 122 0026 32Maximum caliper 1st season 130 0025 33PCA dimension 1 1st year 163 0021 33Angle 1st 3 years 175 0020 33Solidity2 1st season 176 0020 34Perimeter 1st season 188 0019 33Feret angle2 1st season 188 0019 34PCA dimension 22 1st year 190 0019 34Feret angle 1st season 193 0018 33Number of spots2 1st season 206 0017 34

complex mammalian coat pattern traits and should be useful for taxonomic or phenotypicclassifications based on photographic coat pattern data

Our analyses highlighted a few aspects of giraffe spots that weremost likely to be heritableand which seem to have the greatest adaptive significance Circularity and solidity bothdescriptors of spot shape showed the highest mother-offspring similarity Circularitydescribes how close the spot is to a perfect circle and is positively correlated with the traitof roundness and negatively correlated with aspect ratio Solidity describes how smoothand entire the spot edges are versus tortuous ruffled lobed or incised and is negativelycorrelated with the trait of perimeter We did not document significant mother-offspringsimilarity of any size-related spot traits (number of spots area perimeter and maximumcaliper) but the first dimension of the PCAwas largely composed of size-related traits Thesecharacteristics could form the basis for quantifying spot patterns of giraffes across Africaand gives field workers studying any animal with complex color patterns a new quantitativelexicon for describing spots However our mode shade measurement was a crude metricand color is greatly affected by lighting conditions so we suggest standardization ofphotographic methods to control for lighting if color is to be analyzed in future studies

Lee et al (2018) PeerJ DOI 107717peerj5690 1523

We found that both size and shape of spots was relevant to fitness measured as juvenilesurvival We observed the highest calf survival in the phenotypic group generally describedas large spots that were either circular or irregular Lowest survival was in the groups withsmall and medium-sized circular spots and small irregular spots Both the survival byphenotype analysis and the individual covariate survival analysis found that larger spots(smaller number of spots) and irregularly shaped or less-elliptical spots (smaller aspectratio) were correlated with increased survival It seems possible that these traits enhance thebackground-matching of giraffe calves in the vegetation of our study area (Ruxton Sherrattamp Speed 2004 Merilaita Scott-Samuel amp Cuthill 2017) and that neonatal camouflagecould be an adaptive feature of complex coat patterns in other taxa (Allen et al 2011)However covariation in spot patterns and survival could also reflect a maternal effector some environmental effect The relationships among spot traits and their effects onfitness are not well studied and we are aware of no other study that measured coat patterntraits and related variation in those traits to fitness Additional investigations into adaptivefunction and genetic architecture across many taxa are needed to fill this knowledge gap

Whether or not spot traits affect juvenile survival via anti-predation camouflage spottraits may serve other adaptive functions such as thermoregulation (Skinner amp Smithers1990) or social communication (VanderWaal et al 2014) and thus may demonstrateassociations with other components of fitness such as survivorship in older age classes orfecundity Individual recognition kin recognition and inbreeding avoidance also couldplay a role in the evolution of spot patterns in giraffes and other species with complex coatpatterns (Beecher 1982 Tibbetts amp Dale 2007 Sherman Reeve amp Pfennig 1997) Differentaspects of spot traits may also be nonadaptive and serve no function or spot patterns couldbe affected by pleiotropic selection on a gene that influences multiple traits (Lamoreuxet al 2010)

Photogrammetry to remotely measure animal traits has utilized geometric approachesthat estimate trait sizes using laser range finders and known focal lengths (Lyon 1994 Leeet al 2016a) photographs of the traits together with a predetermined measurement unit(Ireland et al 2006 Willisch Marreros amp Neuhaus 2013) or lasers to project equidistantpoints on animals while they are photographed (Bergeron 2007) We hope the frameworkwe have described using ImageJ software to quantify spot characteristics with traitmeasurements from photographs will prove useful to future efforts at quantifying animalmarkings as in animal biometry (Kuumlhl amp Burghardt 2013) Trait measurements and clusteranalysis such as we performed here could also be useful to classify subspecies phenotypesor other groups based on variation inmarkings which could advance the field of phenomicsfor organisms with complex skin or coat patterns (Houle Govindaraju amp Omholt 2010)

Patterned coats of mammals are hypothesized to be formed by two distinct processes aspatially oriented developmental mechanism that creates a species-specific pattern of skincell differentiation and a pigmentation-oriented mechanism that uses information fromthe pre-established spatial pattern to regulate the synthesis of melanin (Eizirik et al 2010)The giraffe skin has more extensive pigmentation and wider distribution of melanocytesthan most other animals (Dimond amp Montagna 1976) Coat pattern variation may reflectdiscrete polymorphisms potentially related to life-history strategies a continuous signal

Lee et al (2018) PeerJ DOI 107717peerj5690 1623

related to maternal effects or a combination of both Future work on the genetics ofcoat patterns will hopefully shed light upon the mechanisms and consequences of coatpattern variation

CONCLUSIONSOur evidence that coat pattern traits were related to juvenile survival is an importantfinding that adds an incremental step to our understanding of the evolution of animalcoat patterns We expect the application of image analysis to giraffe coat patterns willalso provide a new robust dataset to address taxonomic and evolutionary hypotheses Forexample two recent genetic analyses of giraffe taxonomy both placedMasai giraffes as theirown species (Brown et al 2007 Fennessy et al 2016) but the lack of quantitative tools toobjectively analyze coat patterns for taxonomic classification may underlie some of theconfusion that currently exists in giraffe systematics (Bercovitch et al 2017)

ACKNOWLEDGEMENTSThis paper was improved by comments from two anonymous reviewers and AK Lindholm

ADDITIONAL INFORMATION AND DECLARATIONS

FundingFinancial support for this work was provided by Sacramento Zoological Society ColumbusZoo and Aquarium Tulsa Zoo Cincinnati Zoo and Botanical Gardens Tierpark Berlinand Save the Giraffes The funders had no role in study design data collection and analysisdecision to publish or preparation of the manuscript

Grant DisclosuresThe following grant information was disclosed by the authorsSacramento Zoological SocietyColumbus Zoo and AquariumTulsa ZooCincinnati Zoo and Botanical GardensTierpark BerlinSave the Giraffes

Competing InterestsThe authors declare there are no competing interests

Author Contributionsbull DerekE Lee andMonica L Bond conceived anddesigned the experiments performed theexperiments analyzed the data contributed reagentsmaterialsanalysis tools preparedfigures andor tables authored or reviewed drafts of the paper approved the final draftbull Douglas R Cavener conceived and designed the experiments contributedreagentsmaterialsanalysis tools authored or reviewed drafts of the paper approved thefinal draft

Lee et al (2018) PeerJ DOI 107717peerj5690 1723

Animal EthicsThe following information was supplied relating to ethical approvals (ie approving bodyand any reference numbers)

All animal work was conducted according to relevant national and internationalguidelines No Institutional Animal Care and Use Committee (IACUC) approval wasnecessary because animal subjects were observed without disturbance or physical contactof any kind

Field Study PermissionsThe following information was supplied relating to field study approvals (ie approvingbody and any reference numbers)

This researchwas carried outwith permission from theTanzaniaCommission for Scienceand Technology (COSTECH) Tanzania National Parks (TANAPA) the Tanzania WildlifeResearch Institute (TAWIRI) COSTECH research permit numbers 2017-163-ER-90-1722016-146-ER-2001-31 2015-22-ER-90-172 2014-53-ER-90-172 2013-103-ER-90-1722012-175-ER-90-172 2011-106-NA-90-172

Data AvailabilityThe following information was supplied regarding data availability

Lee D Cavener DR Bond M Data from Seeing spots Measuring quantifyingheritability and assessing fitness consequences of coat pattern traits in a wild population ofgiraffes (Giraffa camelopardalis) Dryad Digital Repository httpsdoiorg105061dryad6514r

Supplemental InformationSupplemental information for this article can be found online at httpdxdoiorg107717peerj5690supplemental-information

REFERENCESAllenWL Cuthill IC Scott-Samuel NE Baddeley R 2011Why the leopard got its spots

relating pattern development to ecology in felids Proceedings of the Royal Society ofLondon B Biological Sciences 2781373ndash1380 DOI 101098rspb20101734

AllenWL Higham JP AllenWL 2015 Assessing the potential information contentof multicomponent visual signals a machine learning approach Proceedings of theRoyal Society of London B Biological Sciences 28220142284DOI 101098rspb20142284

Bates D Maechler M Bolker BWalker S 2015 Fitting linear mixed-effects modelsusing lme4 Journal of Statistical Software 671ndash48 DOI 1018637jssv067i01

Beecher MD 1982 Signature systems and kin recognition American Zoologist22477ndash490 DOI 101093icb223477

Bennett DC LamoreuxML 2003 The color loci of micemdasha genetic century PigmentCell Research 16333ndash344 DOI 101034j1600-0749200300067x

Lee et al (2018) PeerJ DOI 107717peerj5690 1823

Bercovitch FB Berry PS Dagg A Deacon F Doherty JB Lee DE Mineur F Muller ZOgden R Seymour R Shorrocks B 2017How many species of giraffe are thereCurrent Biology 27R136ndashR137 DOI 101016jcub201612039

Bergeron P 2007 Parallel lasers for remote measurements of morphological traitsJournal of Wildlife Management 71289ndash292 DOI 1021932006-290

Bolger DT Morrison TA Vance B Lee D Farid H 2012 A computer-assisted systemfor photographic markmdashrecapture analysisMethods in Ecology and Evolution3813ndash822 DOI 101111j2041-210X201200212x

BowenWW DawsonWD 1977 Genetic analysis of coat color pattern variation inoldfield mice (Peromyscus polionotus) of Western Florida Journal of Mammalogy58521ndash530 DOI 1023071380000

Brown DM Brenneman RA Koepfli KP Pollinger JP Milaacute B Georgiadis NJ Louis EEGrether GF Jacobs DKWayne RK 2007 Extensive population genetic structure inthe giraffe BMC Biology 557 DOI 1011861741-7007-5-57

BurnhamKP Anderson DR 2002Model selection and multimodel inference a practicalinformation-theoretical approach New York Springer-Verlag

Calsbeek R Bonneaud C Smith TB 2008 Differential fitness effects of immunocom-petence and neighbourhood density in alternative female lizard morphs Journal ofAnimal Ecology 77103ndash109 DOI 101111j1365-2656200701320x

Caro T 2005 The adaptive significance of coloration in mammals BioScience55125ndash136 DOI 1016410006-3568(2005)055[0125TASOCI]20CO2

Choquet R Lebreton J-D Gimenez O Reboulet A-M Pradel R 2009 U-CARE utilitiesfor performing goodness of fit tests and manipulating CApture-REcapture dataEcography 321071ndash1074 DOI 101111j1600-0587200905968x

Cott HB 1940 Adaptive coloration in animals London Methuen PublishingDagg AI 1968 External features of giraffeMammalia 32657ndash669Dagg AI 2014Giraffe biology behavior and conservation New York Cambridge

University PressDimond RL MontagnaW 1976 The skin of the giraffe Anatomical Record 18563ndash75

DOI 101002ar1091850106Eizirik E David VA Buckley-Beason V Roelke ME Schaumlffer AA Hannah SS

Narfstroumlm K OrsquoBrien SJ Menotti-RaymondM 2010 Defining and mappingmammalian coat pattern genes multiple genomic regions implicated in domesticcat stripes and spots Genetics 184267ndash275 DOI 101534genetics109109629

Endler JA 1978 A predatorrsquos view of animal color patterns Evolutionary Biology11319ndash364 DOI 101007978-1-4615-6956-5_5

Endler JA 1980 Natural selection on color patterns in Poecilia reticulate Evolution3476ndash91 DOI 101111j1558-56461980tb04790x

Endler JA 1983 Natural and sexual selection on color patterns in poeciliid fishesEnvironmental Biology of Fishes 9173ndash190 DOI 101007BF00690861

Falconer DS Mackay TFC 1996 Introduction to quantitative genetics 4th edition NewYork PearsonPrentice Hall

Lee et al (2018) PeerJ DOI 107717peerj5690 1923

Fennessy J Bidon T Reuss F Kumar V Elkan P NilssonMA Vamberger M Fritz UJanke A 2016Multi-locus analyses reveal four giraffe species instead of one CurrentBiology 262543ndash2549 DOI 101016jcub201607036

Foster JB 1966 The giraffe of Nairobi National Park home range sex ratios the herdand food African Journal of Ecology 4139ndash148DOI 101111j1365-20281966tb00889x

Fox J Weisberg S 2011 An R companion to applied regression Second EditionThousand Oaks Sage

Hartigan JA 1975 Clustering algorithms New York WileyHoekstra HE 2006 Genetics development and evolution of adaptive pigmentation in

vertebrates Heredity 97222ndash234 DOI 101038sjhdy6800861Holmberg J Norman B Arzoumanian Z 2009 Estimating population size structure

and residency time for whale sharks Rhincodon typus through collaborative photo-identification Endangered Species Research 739ndash53 DOI 103354esr00186

Hotelling H 1933 Analysis of a complex of statistical variables into principal compo-nents Journal of Educational Psychology 25417ndash441

Houle D Govindaraju DR Omholt S 2010 Phenomics the next challenge NatureReviews Genetics 11855ndash866 DOI 101038nrg2897

Ireland D Garrott RA Rotella J Banfield J 2006 Development and application of amass-estimation method for Weddell sealsMarine Mammal Science 22361ndash378DOI 101111j1748-7692200600039x

Irion U Singh AP Nuesslein-Volhard C 2016 The developmental genetics ofvertebrate color pattern formation lessons from zebrafish In Current topics indevelopmental biology Vol 117 Cambridge Academic Press 141ndash169

Kaelin CB Xu X Hong LZ David VA McGowan KA Schmidt-Kuumlntzel A RoelkeME Pino J Pontius J Cooper GMManuel H 2012 Specifying and sustain-ing pigmentation patterns in domestic and wild cats Science 3371536ndash1541DOI 101126science1220893

Kendall WL Pollock KH Brownie C 1995 A likelihood based approach to capture-recapture estimation of demographic parameters under the robust design Biometrics51293ndash308 DOI 1023072533335

Kettlewell HBD 1955 Selection experiments on industrial malanism in the LepidopteraHeredity 9323ndash342 DOI 101038hdy195536

Klingenberg CP 2010 Evolution and development of shape integrating quantitativeapproaches Nature 11623ndash635 DOI 101038nrg2829

Kruuk LE Slate J Wilson AJ 2008 New answers for old questions the evolutionaryquantitative genetics of wild animal populations Annual Review of Ecology Evolu-tion and Systematics 39525ndash548 DOI 101146annurevecolsys39110707173542

Kuumlhl HS Burghardt T 2013 Animal biometrics quantifying and detecting phenotypicappearance Trends in Ecology and Evolution 28432ndash441DOI 101016jtree201302013

Lee et al (2018) PeerJ DOI 107717peerj5690 2023

Lailvaux SP Kasumovic MM 2011 Defining individual quality over lifetimes andselective contexts Proceedings of the Royal Society of London B Biological Sciences278321ndash328 DOI 101098rspb20101591

LamoreuxML Delmas V Larue L Bennett D 2010 The colors of mice a model geneticnetwork Hoboken John Wiley amp Sons

Lande R Arnold SJ 1983 The measurement of selection on correlated charactersEvolution 371210ndash1226 DOI 101111j1558-56461983tb00236x

Langman VA 1977 Cow-calf relationships in Giraffe (Giraffa camelopardalis giraffa)Ethology 43264ndash286 DOI 101111j1439-03101977tb00074x

Le S Josse J Husson F 2008 FactoMineR an R package for multivariate analysisJournal of Statistical Software 251ndash18 DOI 1018637jssv025i01

Le Poul YWhibley A ChouteauM Prunier F Llaurens V JoronM 2014 Evolution ofdominance mechanisms at a butterfly mimicry supergene Nature Communications51ndash8 DOI 101038ncomms6644

Lee DE Bolger DT 2017Movements and source-sink dynamics of a Masai giraffemetapopulation Population Ecology 59157ndash168 DOI 101007s10144-017-0580-7

Lee DE BondML Kissui BM Kiwango YA Bolger DT 2016a Spatial variation ingiraffe demography a test of 2 paradigms Journal of Mammalogy 971015ndash1025DOI 101093jmammalgyw086

Lee DE Kissui BM Kiwango YA BondML 2016bMigratory herds of wildebeests andzebras indirectly affect calf survival of giraffes Ecology and Evolution 68402ndash8411DOI 101002ece32561

Lorenz K 1937 Imprinting The Auk 54245ndash273 DOI 1023074078077Lydekker R 1904 On the Subspecies of Giraffa Camelopardalis Proceedings of the

Zoological Society of London 74202ndash229Lyon BE 1994 A technique for measuring precocial chicks from photographs The

Condor 86805ndash809MacQueen J 1967 Some methods for classification and analysis of multivariate ob-

servations In Proceedings of 5th Berkeley symposium on mathematical statistics andprobability Berkeley University of California Press 281ndash297

Merilaita S Scott-Samuel NE Cuthill IC 2017How camouflage works PhilosophicalTransactions of the Royal Society B 37220160341 DOI 101098rstb20160341

Mitchell G Skinner JD 2003 On the origin evolution and phylogeny of giraffesGiraffa camelopardalis Transactions of the Royal Society of South Africa 5851ndash73DOI 10108000359190309519935

Morse H 1978Origins of inbred mice New York AcademicNakagawa S Schielzeth H 2010 Repeatability for Gaussian and non-Gaussian data a

practical guide for biologists Biological Reviews 85935ndash956DOI 101111j1469-185X201000141x

Pollock KH 1982 A capture-recapture design robust to unequal probability of captureJournal of Wildlife Management 46752ndash757 DOI 1023073808568

Pratt DM Anderson VH 1979 Giraffe Cow-Calf relationships and social developmentof the calf in the Serengeti Ethology 51233ndash251

Lee et al (2018) PeerJ DOI 107717peerj5690 2123

Protas ME Patel NH 2008 Evolution of coloration patterns Annual Review of Cell andDevelopmental Biology 24425ndash446 DOI 101146annurevcellbio24110707175302

R Core Development Team 2017 R a language and environment for statistical comput-ing Vienna R Foundation for Statistical Computing

Rosenblum EB Hoekstra HE NachmanMW 2004 Adaptive reptile color variation andthe evolution of theMc1r gene Evolution 581794ndash1808

Roulin A 2004 The evolution maintenance and adaptive function of genetic colourpolymorphism in birds Biological Reviews 79815ndash848DOI 101017s1464793104006487

Russell ES 1985 A history of mouse genetics Annual Review of Genetics 191ndash28DOI 101146annurevge19120185000245

Ruxton GD Sherratt TN SpeedMP 2004 Avoiding attack Oxford Oxford UniversityPress

San-Jose LM Roulin A 2017 Genomics of coloration in natural animal populationsPhilosophical Transactions of the Royal Society B Biological Sciences 37220160337DOI 101098rstb20160337

Schneider CA RasbandWS Eliceiri KW 2012 NIH Image to ImageJ 25 years of imageanalysis Nature Methods 9671ndash675 DOI 101038nmeth2089[C]

Sherley RB Burghardt T Barham PJ Campbell N Cuthill IC 2010 Spotting the differ-ence towards fully-automated population monitoring of African penguins Sphenis-cus demersus Endangered Species Research 11101ndash111 DOI 103354esr00267

Sherman PW Reeve HK Pfennig DW 1997 Recognition systems In Krebs JR DaviesNB eds Behavioural ecology an evolutionary approach Oxford Blackwell Scientific69ndash96

Skinner JD Smithers RHN 1990 The mammals of the Southern African Sub-region 2ndedition Pretoria University of Pretoria

Stoffel MA Nakagawa S Schielzeth H 2017 rptR repeatability estimation and variancedecomposition by generalized linear mixed-effects modelsMethods in Ecology andEvolution 81639ndash1644 DOI 1011112041-210X12797

Strauss MK KilewoM Rentsch D Packer C 2015 Food supply and poachinglimit giraffe abundance in the Serengeti Population Ecology 57505ndash516DOI 101007s10144-015-0499-9

Tang-Martinez Z 2001 The mechanisms of kin discrimination and the evolution of kinrecognition in vertebrates a critical re-evaluation Behavioural Processes 5321ndash40DOI 101016S0376-6357(00)00148-0

Tibbetts EA Dale J 2007 Individual recognition it is good to be different Trends inEcology and Evolution 22529ndash537 DOI 101016jtree200709001

Tibshirani RWalther G Hastie T 2001 Estimating the number of clusters in a dataset via the gap statistic Journal of the Royal Statistical Society Series B (StatisticalMethodology) 63411ndash423 DOI 1011111467-986800293

Van Belleghem SM Papa R Ortiz-Zuazaga H Hendrickx F Jiggins CD McMillanWOCounterman BA 2018 patternize an R package for quantifying colour pattern vari-ationMethods in Ecology and Evolution 9390ndash398 DOI 1011112041-210X12853

Lee et al (2018) PeerJ DOI 107717peerj5690 2223

VanderWaal KLWang H McCowan B Fushing H Isbell LA 2014Multilevel socialorganization and space use in reticulated giraffe (Giraffa camelopardalis) BehavioralEcology 2517ndash26 DOI 101093behecoart061

White GC BurnhamKP 1999 Program MARK survival estimation from populationsof marked animals Bird Study 46(Supplement)120ndash138DOI 10108000063659909477239

Whitehead H 1990 Computer assisted individual identification of sperm whale flukesReports of the International Whaling Commission Special Issue 1271ndash77

Willisch CS Marreros N Neuhaus P 2013 Long-distance photogrammetric trait esti-mation in free-ranging animals a new approachMammalian Biology 78351ndash355DOI 101016jmambio201302004

Wilson AJ Nussey DH 2010What is individual quality An evolutionary perspectiveTrends in Ecology and Evolution 25207ndash214 DOI 101016jtree200910002

Wittkopp PJ Williams BL Selegue JE Carroll SB 2003 Drosophila pigmentationevolution divergent genotypes underlying convergent phenotypes Proceedings ofthe National Academy of Sciences of the United States of America 1001808ndash1813DOI 101073pnas0336368100

Wright S 1917 Color inheritance in mammalsmdashIII The rat Journal of Heredity8426ndash430 DOI 101093oxfordjournalsjhereda111864

Lee et al (2018) PeerJ DOI 107717peerj5690 2323