Ecological traits influence the phylogenetic structure of bird species co-occurrences worldwide

LETTER Ecological traits influence the phylogenetic structure of bird

species co-occurrences worldwide

Jean-YvesBarnagaud,1*†

W.DanielKissling,2*†

BrodySandel,1,3Wolf L. Eiserhardt,1

C�a�ganH.S�ekercio�glu,4,5

Brian J. Enquist,6,7

ConstantinosTsirogiannis,1,3 and

Jens-ChristianSvenning1

Abstract

The extent to which species’ ecological and phylogenetic relatedness shape their co-occurrence pat-terns at large spatial scales remains poorly understood. By quantifying phylogenetic assemblagestructure within geographic ranges of >8000 bird species, we show that global co-occurrencepatterns are linked – after accounting for regional effects – to key ecological traits reflecting diet,mobility, body size and climatic preference. We found that co-occurrences of carnivorous, migratoryand cold-climate species are phylogenetically clustered, whereas nectarivores, herbivores, frugivoresand invertebrate eaters tend to be more phylogenetically overdispersed. Preference for open or for-ested habitats appeared to be independent from the level of phylogenetic clustering. Our resultsadvocate for an extension of the tropical niche conservatism hypothesis to incorporate ecologicaland life-history traits beyond the climatic niche. They further offer a novel species-oriented perspec-tive on how biogeographic and evolutionary legacies interact with ecological traits to shape globalpatterns of species coexistence in birds.

Keywords

Diversification, functional traits, geographical range, macroecology, phylogenetic fields,phylogenetic community structure, species coexistence.

Ecology Letters (2014)

INTRODUCTION

How contemporary environmental gradients influence large-scale patterns of species diversity has been extensively studied(Gaston & Blackburn 2000; Hawkins et al. 2003), with increas-ing attention to the contribution of historical and evolutionaryprocesses (Ricklefs 2004; Wiens & Donoghue 2004). The tropi-cal niche conservatism hypothesis provides an integrative his-torical-environmental explanation to latitudinal changes inspecies richness and global-scale species co-occurrence patterns(Wiens & Donoghue 2004). It suggests that climatic stabilityand large areas have favoured higher net diversification rateswithin the tropics, and that most tropical clades have beenunable to expand outside the tropics because niche conserva-tism prevents species from adapting to extra-tropical climates.Macroecological studies that have found support for such his-torical processes underlying large-scale diversity gradients havebeen largely based on methods that focus on species assem-blages which co-occur at particular locations (e.g. Cardillo2011; Kissling et al. 2012a). This approach takes a spatialfocus, in contrast with macroevolutionary studies which oftenfocus on how diversification can be explained by species traits,without an explicit spatial perspective (Bennett & Owens 2002;Bokma 2004; Ackerly et al. 2006). A combined spatially explicit

and species-oriented perspective unifying these two frameworksis needed to provide an integrated perspective on the ecologicaland historical-evolutionary drivers of species co-occurrences(Villalobos et al. 2013; Tobias et al. 2014).Phylogenetic methods that have long been used to study the

evolutionary structure of species co-occurrences in space rarelyaccount for species’ attributes (Webb et al. 2002; Cavender-Bares et al. 2009). The recent species-oriented concept of‘phylogenetic fields’ (Villalobos et al. 2013) overcomes this limi-tation by describing the phylogenetic structure of speciesco-occurrences within a given focal species’ geographical range(Fig. 1). A phylogenetic field quantifies the phylogenetic related-ness of all the species that co-occur with the focal species (Vill-alobos et al. 2013). Hence, the field is a property of a species,and provides a framework for linking macroscale species co-occurrence patterns to other species properties, such as ecologi-cal or life-history traits. Importantly, phylogenetic fields allowinvestigating how broad-scale historical processes such as speci-ation, extinction and dispersal combine to shape present-day co-occurrence patterns of evolutionarily related species (Villaloboset al. 2013). For instance, allopatric speciation should promotelow levels of range overlap among closely related species andthus overdispersed phylogenetic fields (Pigot & Tobias 2013),while high levels of overlap (clustered phylogenetic fields) could

1Section for Ecoinformatics & Biodiversity, Department of Bioscience, Aarhus

University, DK-8000, Aarhus C, Denmark2Institute for Biodiversity and Ecosystem Dynamics (IBED), University of

Amsterdam, P.O. Box 94248, 1090 GE, Amsterdam, The Netherlands3Center for Massive Data Algorithmics (MADALGO), Aarhus University,

DK-8000, Aarhus C, Denmark4Department of Biology, University of Utah, 257 S. 1400 E., Salt Lake City, UT,

84112, USA

5KuzeyDo�ga Derne�gi, Ortakapı Mah. Șehit Yusuf Cad., No:93 Kat:1 Merkez,

Kars, 36100, Turkey6Department of Ecology and Evolutionary Biology, University of Arizona, P.O.

Box 210088,Tucson, 85721, AZ, USA7The Santa Fe Institute, 1399 Hyde Park Rd, Santa Fe, NM, 87501, USA

*Correspondence: E-mail: [email protected];

[email protected]

†These authors contributed equally to this work.

© 2014 John Wiley & Sons Ltd/CNRS

Ecology Letters, (2014) doi: 10.1111/ele.12285

be the consequence of post-speciation range changes or localisedradiations (Ericson 2012). Phylogenetic fields can also be shapedby species’ dispersal abilities (Wiens & Donoghue 2004) andniche convergence or divergence among co-occurring species(Johnson & Stinchcombe 2007; Villalobos et al. 2013). As a con-sequence, phylogenetic fields are likely to be related to specifictraits that reflect variations in contemporary resource use (e.g.dietary adaptations, Silvertown et al. 2006; Pigot & Tobias2013), large-scale habitat or climatic preferences (Streelman &Danley 2003; Ackerly et al. 2006), dispersal behaviour (Belliureet al. 2000) and body size (Bokma 2004).Birds are well suited for analyses of phylogenetic fields as

their global distributions (BirdLife International & Nature-Serve 2011), ecological traits (Del Hoyo et al. 2013) andphylogenetic relationships (Jetz et al. 2012) are extensivelydescribed. Although allopatric speciation is thought to domi-nate bird diversification (Edwards 2005), fast transition ratesto secondary sympatry (Pigot & Tobias 2013) and high levelsof niche conservatism (Peterson et al. 1999) could make theirphylogenetic fields prone to clustering (Fig. 1). Clustering ofphylogenetic fields might be particularly pronounced for birdclades characterised by high diversification rates, isolated geo-

graphic occurrence and low dispersal ability (Fig. 1), factorsthat increase the probability of localised speciation (Ricklefs2006; Kisel & Barraclough 2010; Jetz et al. 2012). In contrast,low diversification rates, high dispersal ability and strongcompetitive exclusion could trigger phylogenetic overdisper-sion (Fig. 1). Ecological traits related to bird habitat use, diet,body size and mobility should therefore be associated with thestructure of species’ phylogenetic fields (Table 1). Reflectingdeep-time evolutionary legacies (Ericson 2012), the relation-ships between phylogenetic fields and ecological traits in birdsmight also be a consequence of evolutionary conservatism.Here, we tested whether phylogenetic fields are related to

key ecological traits in more than 8000 terrestrial bird species(Fig. 1, Table 1). We first explored the imprint of biogeo-graphical history on phylogenetic fields by investigating theirregional variation. We then tested specific a priori hypotheseson the relationships between phylogenetic fields and species’ecological traits (see Table 1 for details and references). Weexpected that (1) forest and open-habitat species should differin their degree of phylogenetic clustering; (2) species-rich die-tary guilds in tropical areas (e.g. insectivores, frugivores andnectarivores) should show high levels of overdispersion; (3)high mobility (i.e. propensity to post-natal dispersal and irreg-ular population-level dispersal events) should lead to phyloge-netic overdispersion while regular migration should be relatedto clustering; (4) small-bodied species should show higherphylogenetic clustering than large-bodied species; and (5) astropical niche conservatism has allowed only few families toexpand outside the tropics, cold-climate species should haveclustered phylogenetic fields. We show that the level of phylo-genetic clustering and overdispersion in bird species co-occur-rences worldwide is related to several of these key ecologicaltraits, and that some of these relationships are affected by astrong historical-evolutionary component.

MATERIAL AND METHODS

Bird distributions

We compiled a comprehensive data set of global bird distribu-tions, representing conservative extent-of-occurrence extrapola-tions of the world-scale breeding ranges for 9886 bird species(out of 9920 recognised extant species, BirdLife International &NatureServe 2011). Distribution maps were derived from avariety of sources, including >5.8 million specimens and impor-tant bird area occurrences, other records (e.g. published litera-ture, survey reports etc.), distribution atlases derived fromsystematic surveys, distribution maps in field guides and otherhandbooks and expert opinion (BirdLife International & Natu-reServe 2011). The data set uses a globally consistent taxo-nomic standard and currently represents the best electronicallyavailable knowledge on global bird species occurrences. Weoverlaid these maps, originally in polygon format, onto a gridin cylindrical equal area projection with >39900 cells of110 9 110 km resolution (equivalent to c. 1� 9 1� near theequator; Kissling et al. 2012b), from which we subsequentlyexcluded cells with more than 50% of water and the Antarcticcontinent, retaining 10599 cells. This excludes many islandsand thus 850 island endemic species, ensuring that our results

Figure 1 Conceptual representation of phylogenetic fields based on

phylogenetic relationships, traits and species co-occurrences. Phylogenetic

fields reflect the phylogenetic structure of the species assemblage that

occurs within the geographic range of a focal species. For species A, the

co-occurrences of species B and C lead to a clustered field (shown in red

in the phylogeny), while species D has a phylogenetically overdispersed

field with species E and F. Several evolutionary, biogeographic and

ecological processes may affect species co-occurrences and can be

mediated via species traits.


2 J.-Y. Barnagaud et al. Letter

Table

1Key

ecologicaltraitsofbirdsandrelatedhypotheses

onhow

they

mightinfluence

thephylogenetic

structure

(overdispersion,clustering)ofspeciesco-occurrences(‘phylogenetic

fields’)at

biogeographic

scales

Trait

Description

Hypotheses

Potentialprocesses

References

Habitatpreference

Forest

vs.open

habitats

Whether

aspeciesprefers

open

orforested

habitats

Forest

andopen-habitatbirds

show

differentlevelsof

phylogenetic

structure,butthe

directionoftheeffect

isunclear

Thespatialgrain

ofresource

distributiondiffers

amongforests

andopen

habitats

CuetoandDeCasenave1999

Diet

Dietary

guild

Speciesare

divided

into

sixdietary

guilds

accordingto

theirdominantdiet

(frugivores,herbivores/seed

eaters,

nectarivores,insectivores,carnivores,

omnivores)

Overdispersionin

tropicalguilds

(e.g.insectivores,frugivores,

nectarivores)

Ecologicalcompetitionlimits

geographicalco-occurrence

ofclosely

relatedspecies

Kisslinget

al.2012b;Pigot&

Tobias2013

Mobility

Postnataldispersal

Whether

post-fledginglong-distance

dispersaloccurs

ornot

Highratesofdispersalfavour

overdispersion

Dispersalcreatesopportunitiesfor

extra-rangecolonisationand

allopatric

speciationover

evolutionary

timescales

Belliure

etal.2000

Irregulardispersal

Irruptiveorirregularmovem

ents

Highratesofdispersalfavour

overdispersion

Belliure

etal.2000

Long-distance

migration

Whether

periodic

(seasonal)migration

alonglatitudinalgradients

occurs

ornot

Restrictedopportunitiesfor

allopatric

speciationpromote

clustering

Inspiteoftheirhighmobility,

migratory

speciesexhibitlow

physiological,behaviouraland

ecologicalflexibility

B€ ohning-G

aeseet

al.1998

Altitudinalmigration

Whether

periodic

(seasonal)migration

alongelevationalgradients

occurs

ornot

Restrictedopportunitiesfor

allopatric

speciationpromote

clustering

B€ ohning-G

aeseet

al.1998

Bodysize

Bodymass

Bodymass

ingrams(species-level

average,

continuoustrait)

Bodymass

isnegativelycorrelated

withclustering

Smaller

speciesmightdiversify

faster

thanlarge-bodiedspecies,andthe

latter

mightbemore

affectedby

competitiveexclusion

Bokma2004;Olsonet

al.2009

Climaticpreference

Tem

perature

optimum

Averagetemperature

over

aspecies’

geographic

range(continuoustrait)

Overdispersionin

tropicalspecies,

clusteringin

cold-climate

species

Tropicalnicheconservatism

,withfew

familieshavingdispersedoutsidethe

tropics

Wiens&

Donoghue2004;

Hortalet

al.2011


Letter Traits and phylogenetic patterns of co-occurrence 3

were not affected by the high rates of recent man-driven extinc-tions that affected insular bird assemblages (Pimm et al. 2006).We further eliminated all families including at least one pelagicor marine species to exclude seabirds and terrestrial specieswith close marine relatives (n = 719 species), and lumped 34species for taxonomic consistency with the phylogenetic data.We eventually deleted species for which all ecological traits(n = 28) or phylogenetic information (n = 21) were missing,and those whose restricted ranges prevented computation ofphylogenetic fields (n = 70). Our final set therefore included8164 terrestrial and freshwater species. We assigned each spe-cies to the region (listed in Table S1: 19 among the 20 zoogeo-graphic regions defined by Holt et al. 2013) encompassing mostof its geographic range.

Phylogeny

We computed phylogenetic fields based on the recently pub-lished mega phylogeny of the world’s bird species (Jetz et al.2012), which we pruned to the above-mentioned 8164 bird spe-cies. The original tree was inferred using a two-step protocol inwhich time-calibrated phylogenetic trees were estimated forwell-supported bird clades and subsequently grafted onto abackbone tree representing deep phylogenetic relationships. Inthis procedure, about two thirds of bird species (6663) wererepresented by molecular information, whereas the remainingspecies were assigned a phylogenetic position based on genus orhigher taxa membership (cf Jetz et al. 2012 for metho-dological details). This tree currently represents the mostcomprehensive and up-to-date synthesis of phylogenetic infor-mation for birds, allowing species-level inference in spite ofremaining uncertainty. To test the sensitivity of our results tophylogenetic uncertainty, we repeated all computations and sta-tistical models using 100 trees randomly drawn from the set of10 000 trees provided by Jetz et al. (2012). These trees are reali-sations of the probability space of possible trees based onpseudo-posteriors of the original Monte Carlo Markov chains.The 10 000 trees were available for two separate backbones; weused those from the second-stage Ericson model for our analysis(available at http://birdtree.org, see details in Jetz et al. 2012).

Traits

We considered eight ecological traits corresponding to majordifferences in bird lifestyles and grouped them into five maintrait categories: habitat, diet, mobility, body size and climaticpreference (Table 1, family-level overview in Table S1). Thesetraits are partly correlated due to biogeographic and evolu-tionary constraints. Since statistical removal of collinearityamong multiple traits by multivariate statistical procedureswould have resulted in a loss of biological interpretability anda drastic reduction of sample size, we tested the effect of eachtrait separately, but explored correlations among traits using aHill & Smith multivariate analysis (Hill & Smith 1976) on the6266 species for which all traits were known (Appendix S1).

Habitat preferenceWe assigned each of the 8164 species to a binary habitat trait,separating forest from open-habitat species, derived from the

habitats classification scheme (version 3.0) of the InternationalUnion for Conservation of Nature (IUCN, www.iucnredlist.org; downloaded April 2012). We defined forest species asthose occurring in IUCN habitat classes 1.1–1.9 (n = 6396).All other species occur in open habitats (n = 1768).

Dietary guildsFor each species with dietary information (n = 8099), we dis-tinguished seven dietary categories (plants, seeds, fruits, nec-tar, invertebrates, carrions, vertebrates) from a comprehensiveliterature survey (S�ekercio�glu et al. 2004; updated with DelHoyo et al. 2013). The relative importance of dietary catego-ries for a given species’ diet was ranked, with all scores addingup to 10 (see Kissling et al. 2012b for a similar methodology).Due to low sample sizes, we summed the scores of plants andseeds, and those of carrion and vertebrates and then assignedeach species to one of six dietary guilds according to the pre-dominant diet (score > 5): frugivores (n = 1156), herbivores/seed eaters (n = 1037), nectarivores (n = 454), invertebrateeaters (n = 4745), carnivores (n = 406) and omnivores ifnone of the diet categories exceeded a score of 5 (n = 301).

MobilityWe compiled four binary trait categories related to dispersaland migration (see Table 1, data also extracted from DelHoyo et al. 2013) to describe species’ mobility over large spa-tial extents (n = 8140 species). The first two mobility traitscorresponded to seasonal migrations across latitude (‘long-dis-tance migration’, n = 1544 migrants) or altitude (‘altitudinalmigration’, n = 968 migrants). The other two refer to a spe-cies’ propensity to perform non-periodic long-distance dis-persal events, either after leaving the nest (‘post-nataldispersal’, n = 422 dispersers) or due to irruptive or irregularmovements (‘irregular dispersal’, n = 956 dispersers).

Body sizeBody size was represented by mean body mass across sexes (ifavailable), retrieved from Dunning (2008) for a total of 6314species, and log-transformed to reduce the effects of extremevalues.

Climatic preferenceWe approximated species’ climatic preferences (n = 8164) asthe average temperature it encounters over its geographicrange (i.e. the mean temperature across the grid cells in whichit occurs). Temperature data (°C 9 10) were averaged fromannual means retrieved from Worldclim version 1.4 (www.worldclim.org; Hijmans et al. 2005). As a supplement, we fur-ther calculated species’ range-level average net primary pro-ductivity and mean range latitude (data described inAppendix S2): these two variables were highly correlated withtemperature and associated results are therefore only pre-sented as Supporting Information in Appendix S2.

Calculation of phylogenetic fields

Phylogenetic fields quantify the phylogenetic structure ofspecies co-occurrence within a focal species’ geographicalrange (Villalobos et al. 2013) by comparing the observed



phylogenetic dispersion in the pool of species co-occurringwith the focal species to a null expectation (Villalobos et al.2013). We calculated phylogenetic fields with a modified ver-sion of the Net Relatedness Index (NRI, Webb et al. 2002)for all focal species occurring in more than one grid cell.Our index (‘focal’ NRI: NRIF) was defined as a standar-dised effect size varying from overdispersed phylogeneticfields (negative values) to clustered phylogenetic fields(positive values):

NRIF;i;t ¼ �1�MPDF;i;t;obs �meanðMPDF;i;t;randÞstandard deviationðMPDF;i;t;randÞ

MPDF,i,t,obs is the observed mean pairwise phylogeneticdistance (in million years) between a given focal species iand the n species it co-occurs with, for a given phylogenetictree t, calculated with a computationally efficient algorithm(Tsirogiannis et al. 2012). It differs from the mean pairwisedistance measure which usually underlies the NRI (Webbet al. 2002) in that it includes only the distances betweenthe focal species and the n co-occurring species, rather thanall pairwise distances in the assemblage. Thus, the NRIF ismore a property of the focal species than of the wholeassemblage. MPDF,i,t,rand is the null distribution of MPDF

values for focal species i, and tree t across 100 null modelreplicates. The species pools used in the null model wereconstrained by two conditions. First, a species could only bedrawn for the null expectation if it occurred in at least oneof the regions (sensu Holt et al. 2013) in which the focalspecies is present. Hence, MPDF,i,t,obs is compared to a nullexpectation drawn from a regional (rather than global)species pool, ensuring that worldwide differences in phyloge-netic fields are not simply driven by differences in speciesrichness and composition among zoogeographic regions.Second, the probability of drawing a given species was pro-portional to its range size, which implies that large-rangingspecies are more likely to be drawn than range-restrictedspecies. Hence, the NRIF accounts for differences in species’range sizes as well as in the richness of regional speciespools. We preferred this null expectation over one based ona global species pool or one where species are drawn inde-pendently of geographic range size because it relates range-wide co-occurrences to the specific region in which a focalspecies is present, and it minimises any spurious effectsrelated to differences in species range sizes.

Statistical analyses

We first used an ordinary-least-squares (OLS) linear regressionmodel with a normally distributed error term (referred to as‘non-phylogenetic regression’) to relate species’ NRIF to a suiteof species’ ecological traits. For categorical traits (habitat pref-erence, dietary guild and mobility), parameters were made iden-tifiable by setting a reference category to 0. For body mass andtemperature preference (continuous variables), we reported thelinear slope coefficient. Because not all traits were measured forall species (see above, ‘Traits’) and because our aim was toinvestigate the magnitude of the fields–trait relationships ratherthan to compare the effects among trait categories, we per-formed five separate models: (i) forest vs. open-habitat use;

(ii) dietary guild; (iii) mobility (post-natal dispersal + irregulardispersal + long-distance migration + altitudinal migration);(iv) body size, and (v) climatic preference. In all models, weadditionally included the core zoogeographic region of thefocal species’ occurrence as a covariate to account for biogeo-graphic differences in phylogenetic fields. In this region effect,each species is attributed to the region hosting the largestfraction of its range (see ‘Bird distributions’, above). Weused Akaike’s Information Criterion (AIC, Burnham & Ander-son 2002) to measure the gain of likelihoods of models with setsof traits, as compared to models with zoogeographic regiononly (discriminating models separated by more than two AICunits).We also investigated the phylogenetic component of the

trait effects with generalised least square regressions (hereafter‘phylogenetic regression’, implemented under the phylolm Rlibrary, Ho & An�e 2013) to test whether the effect of traits onphylogenetic fields contains a deep evolutionary signal. Weused the same fixed effect structures as in non-phylogeneticregressions, but additionally imposed a correlated error struc-ture based on a phylogenetic distance matrix that assumes Pa-gel’s evolutionary model (Pagel 1999). The associatedparameter (Pagel’s k) increases towards 1 as the phylogenyhas an increasing impact on model structure (Pagel 1999).

Sensitivity analyses

We assessed the sensitivity of the relationships between phylo-genetic fields and traits to phylogenetic uncertainty by per-forming the non-phylogenetic and phylogenetic regressions for100 separate phylogenetic trees and their associated phyloge-netic field distributions. We reported the regression coeffi-cients �95% confidence intervals separately for each of these100 models in the results. We further evaluated the influenceof marginally overlapping species by recomputing phyloge-netic fields and reestimating their relationships with traits,restricting species assemblages to those species overlapping bymore than 5% or 10% with the focal species’ geographicrange (Appendix S3).

RESULTS

Structure of bird phylogenetic fields

Bird phylogenetic fields varied widely across the 8164 species(Fig. 2). Overall, 49% of the species (n = 4017) had overdi-spersed phylogenetic fields, while 40% (n = 3282) had clusteredphylogenetic fields. Only 865 phylogenetic fields (11%) did notexhibit any clear phylogenetic structure (95% confidence inter-vals overlapping 0, Fig. 2). Across species-rich bird families(with >100 species), rails and allies (Rallidae), birds of prey(Accipitridae) and the toucans and barbets (Ramphastidae) hadthe most clustered phylogenetic fields, whereas ovenbirds (Fur-nariidae), sunbirds (Nectariniidae) and cisticolas (Cisticolidae)were particularly overdispersed (for a list with NRIF averagedat the family level see Table S2). As expected, phylogeneticfields showed regional variation according to the focal species’core zoogeographic region of occurrence (Fig. 3): Eurasian,Central Asian and Nearctic species had the most clustered



fields, while species with the most overdispersed fields occurredin tropical areas, especially in Africa and South America.

Relationships between phylogenetic fields and traits

Sample sizes were relatively large across all trait categories,ensuring that trait effects were reasonably well estimated even

for the less common categories. Models with sets of traits inaddition to a region effect exhibited lower AIC than modelswith a region effect only, indicating that all life-history traitsexcept habitat preference contributed to explain phylogeneticfields (Table 2). Open-habitat and forest species did not differin their field structures (Table 2, Fig. 4A). Dietary guild hada strong effect in non-phylogenetic regressions (Table 2),mainly driven by the high level of clustering in carnivore spe-cies (Fig. 4B). Omnivores exhibited the highest variance intheir phylogenetic fields among all guilds, and nectarivoresand herbivores were the most overdispersed, although theconfidence intervals of their regression parameters partly over-lapped with invertebrate eaters and frugivores. Among mobil-ity traits, the most clustered phylogenetic fields were relatedto altitudinal and latitudinal migratory behaviours, whilepost-natal and irregular dispersal showed a weaker clusteringeffect (Fig. 4C). In non-phylogenetic regression, body masswas positively correlated with the level of clustering, indicat-ing that large species had the most clustered fields (Fig. 4D).Finally, warm-climate species had more overdispersed fieldsthan cold-dwelling ones (Fig. 4E).

Imprint of phylogeny

Pagel’s k was close to 0.4 in all phylogenetic regressions(Table 2), suggesting that phylogenetic fields contain a signalrelated to the clade’s evolutionary history. Accordingly, DAICindicated that the effect of dietary guilds on phylogenetic fieldsdisappeared when including a phylogenetic covariance matrix,while body mass and mobility still improved the models rela-tive to a region only model, but to a smaller extent than innon-phylogenetic regressions (Table 2, Fig. 4). The effects ofaltitudinal and latitudinal migrations persisted in phylogeneticregressions, contrary to those of post-natal and irregular dis-persal (Fig. 4C). Furthermore, the effect of body mass wasreversed, with large-bodied species being more overdispersed

Figure 3 Variation of bird phylogenetic fields across zoogeographic regions. The map shows phylogenetic fields averaged across all species present in each

of the 19 regions, from regions dominated by overdispersed fields (lighter colours) to clustered fields (darker colours).

Figure 2 Distribution of phylogenetic fields of 8164 bird species, ranked

by increasing order of phylogenetic clustering. Phylogenetic fields vary

from species that co-occur with less closely related species across their

geographic range (overdispersion, negative values) to those that co-occur

with more closely related species (clustering, positive values) compared to

a null expectation of phylogenetically random assemblages. For each

species, phylogenetic fields are represented by their mean (solid line) and

5–95% quantiles (dark and light grey shades) across 100 replicated

phylogenetic trees.



than smaller ones after accounting for phylogeny. Finally, andcontrary to all other traits, species’ climatic preferenceimproved the AIC more in phylogenetic than in non-phyloge-netic regressions (Table 2), but its regression coefficientsremained largely similar, with warmer climate species havingmore overdispersed phylogenetic fields than cold-dwelling spe-cies (Fig. 4E).

Sensitivity to phylogenetic uncertainty, marginal species and

multitrait correlations

The Hill & Smith analysis on the set of species for whichall traits were available (n = 6266) revealed associationsamong traits (Appendix S1). For instance, carnivory wasassociated with large body mass, high rates of post-nataland irregular dispersal and preference for open habitats.Moreover, frugivore species were mainly associated withwarm climates, and to a lesser extent with forested habitatsand sedentariness. Finally, altitudinal or latitudinal migrantswere negatively associated with warm climates, highlighting

(a)

(b)

(c)

(d)

(e)

Figure 4 Effects of ecological traits on phylogenetic fields of birds, estimated as regression coefficients in non-phylogenetic regressions (red) and

phylogenetic regressions (blue). The 95% confidence intervals are reported for 100 models, each based on a different phylogenetic tree drawn from Jetz

et al. (2012)’s original Monte Carlo Markov chains. For categorical traits (a: habitat preference; b: dietary guilds; c: mobility), effects are expressed as the

difference of phylogenetic fields relative to a reference level fixed to 0 (in brackets). Slope coefficients are reported for continuous traits (d: body size and

e: climatic preference). All regressions include species’ main zoogeographic region of occurrence as a covariate.

Table 2 Performances of regression models relating bird species’ phyloge-

netic fields to their ecological traits. DAIC values express the difference

between a model with species’ traits + main zoogeographic region of

occurrence, and a model with regions only

Trait category

Non-phylogenetic regression Phylogenetic regression

D AIC D AIC Pagel’s k

Habitat use �1.23 � 0.57 �0.64 � 0.49 0.44 � 0.03

Dietary guild 169.66 � 7.94 �16.24 � 0.70 0.44 � 0.03

Mobility 121.59 � 6.45 67.07 � 5.56 0.44 � 0.03

Body size 131.12 � 10.47 3.83 � 0.89 0.44 � 0.03

Temperature 63.63 � 7.28 91.00 � 9.54 0.43 � 0.03

Positive DAIC values >2 (highlighted in bold) indicate that ecological

traits improve the likelihood of the model relative to a model with

regions only. Pagel’s k measures the strength of the phylogenetic corre-

lation structure in phylogenetic regressions. DAIC and Pagel’s k are

shown as mean � SD over 100 models in which phylogenetic fields

and phylogenetic covariance matrices (in phylogenetic regressions) are

derived from 100 phylogenetic trees drawn from Jetz et al. (2012)’s

original Monte Carlo Markov chains. AIC, Akaike’s information

criterion.



their ability to breed in strongly seasonal (e.g. northern lati-tude) habitats.Phylogenetic fields showed little variation across the 100

randomly sampled phylogenetic trees (Fig. 2), and our conclu-sions on trait effects were thus not substantially affected byuncertainty in the underlying phylogenetic trees (Fig. 4).Removing marginal species did not qualitatively change theresults (compare Fig. 4 with figures in Appendix S3).

DISCUSSION

Quantifying phylogenetic fields across broad biogeographicalscales offers a novel species-oriented perspective to testhypotheses on geographical patterns of co-occurrence(Villalobos et al. 2013). Applying this approach, we presentthe first comprehensive analysis of relationships betweenecological traits and the phylogenetic structure of birdassemblages worldwide. Accounting for regional speciespools and range size, we found that some bird species tendto co-occur with close relatives, while others tend to co-occur with distant relatives (overdispersion), with a trendtowards clustering in extra-tropical species. We found thatspecies’ ecological characteristics explained a substantial pro-portion of species co-occurrence patterns at large spatialscales, but the intensity of their influence as well as theirphylogenetic signal varied substantially from trait to trait.Effects of diet and body size on phylogenetic fields showeddeep evolutionary signals, whereas effects of mobility andclimatic preference appeared to be less evolutionarily con-strained. Species that only marginally co-occurred with afocal species had little impact on the magnitude of trait sig-nals in phylogenetic fields.

Patterns of phylogenetic clustering and overdispersion

About half of the bird species showed phylogenetic overdi-spersion in the co-occurrence structure with other species.This overdispersion could be explained by the predominanceof allopatric speciation associated with low secondary rates ofsympatry (e.g. in Furnariidae, Table S2, Pigot & Tobias 2013;Tobias et al. 2014), which has been suggested to dominateamong terrestrial birds (Edwards 2005). However, 40% ofbird species had clustered phylogenetic fields, indicating thatsym- and parapatric speciation and/or allopatric speciationwith high secondary rates of sympatry could also be commonamong birds. In addition to speciation mode, biogeographichistory of a region could also be a mediating factor (Fig. 1).Interestingly, phylogenetic fields were strongly affected byregion, with species of northern temperate and arctic regionsshowing high levels of clustering, whereas tropical speciestended to have overdispersed fields. This pattern contrastswith previous results for Neotropical bats, for which overdi-spersion in phylogenetic fields has been found at northern(rather than tropical) latitudes (Villalobos et al. 2013). Acrossthe bird clade, we suggest that only few families have beenable to colonise and diversify in colder realms (consistent withtropical niche conservatism, Wiens & Donoghue 2004; Haw-kins et al. 2006), leading to a predominance of clustering inextra-tropical regions. Distributional constraints imposed by

current and past climatic changes on Arctic-Siberian speciescould further explain the high levels of clustering comparedwith more southern species (Hortal et al. 2011). This is consis-tent with higher clustering in regions that have been climati-cally unstable in the past, relative to regions with lowpalaeoclimatic temperature oscillations (Kissling et al. 2012a).However, it could also indicate that allopatric speciation asso-ciated with high secondary rates of sympatry is most commonin harsh and unstable environments, which requires furthertesting.

Trait and niche conservatism

For some traits (especially dietary preferences and body size),parameter estimates differed between non-phylogenetic andphylogenetic regressions. This indicates an imprint of traitconservatism on phylogenetic fields and suggests that suchecological characteristics have evolved little after the earlyseparation of ancestors. However, more recent trait evolutionwithin clades (Peterson et al. 1999; Ericson 2012) might alsohave occurred at a much finer scale than our coarse trait cat-egorisations, as proposed for instance in the dietary prefer-ences or specialisation of New World warblers (Mac Arthur1958). The relatively higher proportion of carnivores atnorthern latitudes, as compared with that of frugivores, seedeaters or invertebrate eaters, could explain the relative overdi-spersion of other dietary guilds in our results (see the bimo-dal latitudinal species richness gradient for carnivores inKissling et al. (2012b)). Interestingly, the diet trait effect onphylogenetic fields disappeared in phylogenetic regressions,indicating that carnivory is a highly conserved dietary strat-egy within the avian tree of life. An even more pronouncedchange in parameter estimates between non-phylogenetic andphylogenetic regression occurred for body size (from overdi-spersion to clustering), suggesting that this trait shows largertrait conservatism than coarse dietary strategies. We note thatbody size was uncorrelated with range size in the subset ofspecies used for our analyses (Pearson’s r = 0.1 on a log-logscale), which should ensure that the null model underlyingthe computation of species’ phylogenetic fields did not favourthe inclusion of large-bodied species in random null assem-blages. In contrast to body size and diet, however, the effectof climatic preference on phylogenetic fields did not show anevolutionary signal. Hence, we hypothesise that much of thewidely recognised tropical niche conservatism, initially attrib-uted to temperature and water availability effects (Wiens &Donoghue 2004), may actually arise via constraints imposedby species ecological traits such as body size and dietary pref-erence rather than directly via their range-wide climaticniches.

Migration vs. dispersal

Migratory behaviour had a detectable imprint on phylogeneticfields, whereas dispersal behaviour had no or at most a weakeffect. In line with our hypothesis (Table 1), altitudinal andlatitudinal migrants showed clustered phylogenetic fields rela-tive to non-migratory birds, maybe because the evolution ofmigratory behaviour adds up with low flexibility in habitat



preferences, nesting phenology or physiological constraints(B€ohning-Gaese et al. 1998). However, the same clusteringeffect might also arise from habitat or geophysical constraints,as these can confine migratory birds to certain regions. Forinstance, many altitudinal migrants are found in associationwith mountain ranges (e.g. Andes, Himalaya), whereas latitu-dinal migrants usually use the seasonal excess of resources atnorthern latitudes (Hurlbert & Haskell 2003). In contrast tomigrants and non-migrants, species with high (irregular orpost-natal) dispersal ability did not differ in phylogeneticfields relative to less mobile species. We acknowledge that ourcoarse dispersal categories might disregard an influence of dis-persal behaviour on phylogenetic fields, and that additionalanalyses with actual dispersal distances would be desirable infuture analyses. However, such direct dispersal estimates areextremely difficult to quantify and hence are lacking for themajority of bird species (Dawideit et al. 2009).

Correlations among multiple traits

Some of the effects of individual traits on phylogenetic fieldscould be mediated by correlations among multiple traits(Appendix S1; Bennett & Owens 2002). For instance, the effectof body size could at least in part mediate the overdispersion ofdietary guilds relative to carnivory. Frugivory, preference forwarm temperature and sedentariness are also strongly related(Appendix S1). This could indicate that climatic niche conser-vatism will be difficult to disentangle from the effects of ecolog-ical and morphological traits for some bird species (e.g.frugivores and non-migratory birds), as covariation betweenclimate and ecological adaptations might be due to indirectclimatic effects on large-scale distributions of higher trophiclevels (Kissling et al. 2007). Finally, latitudinal migration wasnegatively related to warm climate preference and frugivory,which indicates that some ecological strategies are exclusive(i.e. obligate frugivores occur in tropical climates and are notlatitudinal migrants). In spite of these correlations, none of theinvestigated traits were fully represented by the other traits andthus each of them deserves a separate examination. Hence, aspecies’ response to selection pressures is driven by the jointinfluence of its ecological characteristics (B€ohning-Gaese et al.1998; Bennett & Owens 2002), such that trait correlations arechallenging to disentangle in an evolutionarily meaningful way.Based on our results, we suggest that climatic preference, die-tary adaptation and migratory behaviour are the key ecologicaltraits that jointly shape broad-scale bird co-occurrence patternsover evolutionary timescales.

CONCLUSION

Using a novel species-oriented phylogenetic perspective, weprovide evidence that key ecological traits – on top of regio-nal effects – shape geographical patterns of species co-occur-rence in birds worldwide. Some trait effects on phylogeneticfields (e.g. via body size and diet) are deeply conserved in thebird tree of life, whereas others appear not to be (e.g. climaticpreference). This suggests that a more differentiated theory of‘tropical niche conservatism’ needs to be developed that doesnot only focus on climatic niches (Wiens & Donoghue 2004),

but also takes various ecological and life-history traits intoaccount. Beyond the traits used in our study, other factorssuch as the fine-scale division of resources and space (oftenapproximated by morphological measurements, Grant 1968)or host-pathogen dynamics (Ricklefs 2010) could also contrib-ute to shape the phylogenetic and geographic structure of spe-cies co-occurrences. Furthermore, intraspecific variation in aspecies’ phylogenetic field could be quantified for each gridcell of a species geographic range. This would allow investi-gating how phylogenetic structure in co-occurrence variesfrom a species’ range core to its margins, where abiotic condi-tions often become harsher (Sexton et al. 2009). We thereforesuggest that, in line with our results, shifting from a purelysite-based to a species-oriented, range-wide approach may beespecially fruitful to understand how species’ evolutionary his-tories and current ecological conditions combine to shapelarge-scale gradients of biodiversity and species coexistence.

ACKNOWLEDGEMENTS

We thank Jens Mogens Olesen, Jonathan Chase and threeanonymous referees for comments. This study is a contribu-tion by Center for Informatics Research on Complexity inEcology (CIRCE), funded by the Aarhus University ResearchFoundation under the AU Ideas program. W.D.K. is sup-ported by a University of Amsterdam (UvA) starting grant.C.H.S. thanks the University of Utah and the ChristensenFund for their support and is grateful to Sherron Bullens,Debbie Fisher, David Hayes, and especially Beth Karpas,Kathleen McMullen and Jason Socci for their dedicated helpwith the world bird ecology database. B.S. and C.T. were sup-ported by MADALGO-Center for Massive Data Algorith-mics, a Center of the Danish National Research Foundation.

AUTHORSHIP

W.D.K. & J.-Y.B. conceived the study, J.-Y.B., W.D.K.,J.-C.S. formulated hypotheses, W.D.K. and C.H.S. compileddata, J.-Y.B., W.D.K., B.S., W.L.E. and C.T. performedanalyses, J.-Y.B. & W.D.K. wrote the manuscript, all authorscommented on draft manuscripts.

REFERENCES

Ackerly, D.D., Schwilk, D.W. & Webb, C.O. (2006). Niche evolution and

adaptive radiation: testing the order of trait divergence. Ecology, 87,

S50–S61.Belliure, J., Sorci, G., Møller, A.P. & Clobert, J. (2000). Dispersal

distances predict subspecies richness in birds. J. Evol. Biol., 13, 480–487.

Bennett, P.M. & Owens, I.P.F. (2002). Evolutionary Ecology of Birds: Life

Histories, Mating Systems, and Extinction. Oxford University Press,

Oxford.

BirdLife International & NatureServe. (2011). Bird Species Distribution

Maps of the World. BirdLife International and NatureServe,

Cambridge, UK, and Arlington, USA.

B€ohning-Gaese, K., Gonz�alez-Guzm�an, L.I. & Brown, J.H. (1998).

Constraints on dispersal and the evolution of the avifauna of the

Northern Hemisphere. Evol. Ecol., 12, 767–783.Bokma, F. (2004). Differential rates of morphological divergence in birds.

J. Evol. Biol., 17, 933–940.



Burnham, K.P. & Anderson, D.R. (2002). Model Selection and

Multimodel Inference: a Practical Information-Theoretic Approach, 2nd

edn.. Springer-Verlag, Heidelberg.

Cardillo, M. (2011). Phylogenetic structure of mammal assemblages at

large geographical scales: linking phylogenetic community ecology with

macroecology. Philos. Trans. R. Soc. Lond. B Biol. Sci., 366, 2545–2553.

Cavender-Bares, J., Kozak, K.H., Fine, P.V.A. & Kembel, S.W. (2009).

The merging of community ecology and phylogenetic biology. Ecol.

Lett., 12, 693–715.Cueto, V.R. & de Casenave, J.L. (1999). Determinants of bird species

richness: role of climate and vegetation structure at a regional scale. J.

Biogeogr., 26, 487–492.Dawideit, B.A., Phillimore, A.B., Laube, I., Leisler, B. & B€ohning-Gaese,

K. (2009). Ecomorphological predictors of natal dispersal distances in

birds. J. Anim. Ecol., 78, 388–395.Del Hoyo, J., Elliott, A. & Sargatal, J. (2013). Handbook of the Birds of

the World. Lynx Edicions, Barcelona, Spain.

Dunning, J.B. (2008). Handbook of Avian Body Masses, 2nd edn.. CRC

Press, Boca Raton.

Edwards, S.V. (2005). Speciation in birds: genes, geography, and sexual

selection. Proc. Natl Acad. Sci., 102, 6550–6557.Ericson, P.G.P. (2012). Evolution of terrestrial birds in three continents:

biogeography and parallel radiations. J. Biogeogr., 39, 813–824.Gaston, K.J. & Blackburn, T. (2000). Pattern and Process in

Macroecology. Blackwell Publishing, Oxford.

Grant, P.R. (1968). Bill size, body size, and the ecological adaptations

of bird species to competitive situations on islands. Syst. Zool., 17, 319–333.

Hawkins, B.A., Field, R., Cornell, H.V., Currie, D.J., Guegan, J.F.,

Kaufman, D.M. et al. (2003). Energy, water, and broad-scale

geographic patterns of species richness. Ecology, 84, 3105–3117.Hawkins, B.A., Diniz-Filho, J.A.F., Jaramillo, C.A. & Soeller, S.A. (2006).

Post-Eocene climate change, niche conservatism, and the latitudinal

diversity gradient of New World birds. J. Biogeogr., 33, 770–780.Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G. & Jarvis, A.

(2005). Very high resolution interpolated climate surfaces for global

land areas. Int. J. Climatol., 25, 1965–1978.Hill, M.O. & Smith, A.J.E. (1976). Principal component analysis of

taxonomic data with multi-state discrete characters. Taxon, 25, 249–255.Ho, L.S.T. & An�e, C. (2013). Package “Phylolm” - Phylogenetic Linear

Regression. Available at: http://cran.r-project.org/web/packages/phylolm/

index.html. Last accessed 17 February 2014.

Holt, B.G., Lessard, J.-P., Borregaard, M.K., Fritz, S.A., Ara�ujo, M.B.,

Dimitrov, D. et al. (2013). An update of Wallace’s zoogeographic

regions of the world. Science, 339, 74–78.Hortal, J., Diniz-Filho, J.A.F., Bini, L.M., Rodr�ıguez, M.�A., Baselga, A.,

Nogu�es-Bravo, D. et al. (2011). Ice age climate, evolutionary

constraints and diversity patterns of European dung beetles. Ecol. Lett.,

14, 741–748.Hurlbert, A.H. & Haskell, J.P. (2003). The effect of energy and

seasonality on avian species richness and community composition. Am.

Nat., 161, 83–97.Jetz, W., Thomas, G.H., Joy, J.B., Hartmann, K. & Mooers, A.O. (2012).

The global diversity of birds in space and time. Nature, 491, 444–448.Johnson, M.T.J. & Stinchcombe, J.R. (2007). An emerging synthesis

between community ecology and evolutionary biology. Trends Ecol.

Evol., 22, 250–257.Kisel, Y. & Barraclough, T.G. (2010). Speciation has a spatial scale that

depends on levels of gene flow. Am. Nat., 175, 316–334.Kissling, W.D., Rahbek, C. & B€ohning-Gaese, K. (2007). Food plant

diversity as broad-scale determinant of avian frugivore richness. Proc.

R. Soc. B Biol. Sci., 274, 799–808.Kissling, W.D., Eiserhardt, W.L., Baker, W.J., Borchsenius, F., Couvreur,

T.L.P., Balslev, H. et al. (2012a). Cenozoic imprints on the

phylogenetic structure of palm species assemblages worldwide. Proc.

Natl. Acad. Sci. U. S. A., 109, 7379–7384.

Kissling, W.D., Sekercioglu, C.H. & Jetz, W. (2012b). Bird dietary guild

richness across latitudes, environments and biogeographic regions.

Glob. Ecol. Biogeogr., 21, 328–340.Mac Arthur, R. (1958). Population ecology of some warblers of

Northeastern coniferous forests. Ecology, 39, 599–619.Olson, V.A., Davies, R.G., Orme, C.D.L., Thomas, G.H., Meiri, S.,

Blackburn, T.M. et al. (2009). Global biogeography and ecology of

body size in birds. Ecol. Lett., 12, 249–259.Pagel, M. (1999). Inferring the historical patterns of biological evolution.

Nature, 401, 877–884.Peterson, A.T., Soberon, J. & Sanchez-Cordero, V. (1999). Conservatism

of ecological niches in evolutionary time. Science, 285, 1265–1267.Pigot, A.L. & Tobias, J.A. (2013). Species interactions constrain

geographic range expansion over evolutionary time. Ecol. Lett., 16,

330–338.Pimm, S., Raven, P., Peterson, A., Sekercioglu, C.H. & Ehrlich, P.R.

(2006). Human impacts on the rates of recent, present, and future bird

extinctions. Proc. Natl. Acad. Sci. U. S. A., 103, 10941–10946.Ricklefs, R.E. (2004). A comprehensive framework for global patterns in

biodiversity. Ecol. Lett., 7, 1–15.Ricklefs, R.E. (2006). Global variation in the diversification rate of

passerine birds. Ecology, 87, 2468–2478.Ricklefs, R.E. (2010). Host-pathogen coevolution, secondary sympatry

and species diversification. Philos. Trans. R. Soc. B-Biol. Sci., 365,

1139–1147.S�ekercio�glu, C� .H., Daily, G.C. & Ehrlich, P.R. (2004). Ecosystem

consequences of bird declines. Proc. Natl Acad. Sci., 101, 18042–18047.Sexton, J.P., McIntyre, P.J., Angert, A.L. & Rice, K.J. (2009). Evolution

and ecology of species range limits. Annu. Rev. Ecol. Syst., 40, 415–436.Silvertown, J., Dodd, M., Gowing, D., Lawson, C. & McConway, K.

(2006). Phylogeny and the hierarchical organization of plant diversity.

Ecology, 87, S39–S49.Streelman, J.T. & Danley, P.D. (2003). The stages of vertebrate

evolutionary radiation. Trends Ecol. Evol., 18, 126–131.Tobias, J.A., Cornwallis, C.K., Derryberry, E.P., Claramunt, S.,

Brumfield, T. & Seddon, N. (2014). Species coexistence and the

dynamics of phenotypic evolution in adaptive radiation. Nature, 506,

359–363.Tsirogiannis, C., Sandel, B. & Cheliotis, D. (2012). Efficient computation

of popular phylogenetic tree measures. In: Algorithms Bioinforma.,

Lecture Notes in Computer Science (eds. Raphael, B. & Tang, J.).

Springer, Berlin, Heidelberg, pp. 30–43.Villalobos, F., Rangel, T.F. & Diniz-Filho, J.A.F. (2013). Phylogenetic

fields of species: cross-species patterns of phylogenetic structure and

geographical coexistence. Proc. R. Soc. B Biol. Sci., 280, 20122570.

Webb, C.O., Ackerly, D.D., McPeek, M.A. & Donoghue, M.J. (2002).

Phylogenies and community ecology. Annu. Rev. Ecol. Syst., 33, 475–505.Wiens, J.J. & Donoghue, M.J. (2004). Historical biogeography, ecology

and species richness. Trends Ecol. Evol., 19, 639–644.

SUPPORTING INFORMATION

Additional Supporting Information may be downloaded viathe online version of this article at Wiley Online Library(www.ecologyletters.com).

Editor, David MouillotManuscript received 3 September 2013First decision made 10 October 2013Second decision made 17 March 2014Manuscript accepted 25 March 2014



Ecological traits influence the phylogenetic structure of bird species co-occurrences worldwide

Documents

Transcript of Ecological traits influence the phylogenetic structure of bird species co-occurrences worldwide