ORIGINAL Global patterns of specialization and … 2011...reduced (Fig. S3 in Appendix S1). We...

ORIGINALARTICLE

Global patterns of specialization andcoexistence in bird assemblages

Jonathan Belmaker1*�, Cagan H. Sekercioglu2� and Walter Jetz1�

1Department of Ecology and Evolutionary

Biology, Yale University, New Haven, CT

06520, USA, 2Center for Conservation Biology,

Department of Biology, Stanford University,

Stanford, CA 94305, USA

*Correspondence: Jonathan Belmaker,

Department of Ecology and Evolutionary

Biology, Yale University, New Haven, CT 06520,

USA.

E-mail: [email protected]�These authors contributed equally.�Current address: Department of Biology,

University of Utah, Salt Lake City, UT 84112,

USA.

ABSTRACT

Aim Increased specialization has been hypothesized to facilitate local coexistence

and thus high species richness, but empirical evaluations of the richness–

specialization relationships have been relatively scant. Here, we provide a first

assessment of this relationship for terrestrial bird assemblages at global extent and

from fine to coarse grains.

Location World-wide.

Methods We use two indices of specialization that describe species-level

resource use: diet and habitat specialization. The relationship between richness

and mean assemblage-level specialization was independently assessed at realm,

biome-realm, 12,100 km2 equal-area grid cells and fine-grained scales. To identify

assemblages that are diverse relative to environmental conditions we: (1) applied

quantile regressions, (2) statistically accounted for other environmental variables

which may constrain richness, and (3) parsed the data according to the residuals

of a model relating species richness to the environmental variables.

Results Assemblage species richness increases with both measures of

specialization at all scales. Statistically, richness appears constrained by levels of

specialization, with the highest richness values only found in specialized

assemblages. Richness is positively associated with specialization even after

accounting for gradients in resource availability. Net primary productivity and

assemblage specialization have complementary statistical effects on assemblage

species richness. Contrary to expectations based on niche partitioning of local

resources, the relationship between specialization and richness is steep even at

coarse scales.

Main conclusions The results demonstrate that for an entire clade, totalling

> 9000 species, specialization and species richness are related, at least for diverse

assemblages. The strong patterns observed across scales suggest that this

relationship does not solely originate from (1) limits on coexistence in present-

day assemblages, or (2) increased specialization in richer assemblages imposed by

species’ abilities to partition ecological space. Instead, regional-scale influences on

the species pool may determine much of the observed relationship between

richness and specialization. Although causal attribution is not straightforward,

these findings support the idea that, for the scale of our analysis, specialization

may be related to the past origination of high-diversity assemblages, rather than

their contemporary assembly.

Keywords

Biodiversity, biological inventories, birds, diet breadth, habitat breadth,

macroecology, regional scale, richness–specialization hypothesis, scale, species

richness.

Journal of Biogeography (J. Biogeogr.) (2011)

ª 2011 Blackwell Publishing Ltd http://wileyonlinelibrary.com/journal/jbi 1doi:10.1111/j.1365-2699.2011.02591.x

INTRODUCTION

Establishing a mechanistic and predictive understanding of

global richness gradients is a central theme in community and

macroecology. Generally, putative mechanisms have been

approached from evolutionary or ecological perspectives.

While evolutionary approaches focus on drivers of diversifi-

cation, ecologists have emphasized the way assemblages

subdivide niche-space locally and thus facilitate coexistence

and high levels of species richness (Chase & Leibold, 2003). A

prominent niche-based hypothesis for geographic variation in

species richness asserts that certain environments, such as the

tropics, support species that are more narrowly specialized and

thus allow tighter species packing (Hutchinson, 1959; Connell

& Orias, 1964; MacArthur, 1972).

A positive relationship between specialization and species

richness has been put forward as a key explanation for global

diversity gradients (MacArthur, 1972). Specifically, it has been

argued that the latitudinal diversity gradient originates from

the stable tropical climates, which allow narrower species

tolerances and hence specialization (Klopfer & MacArthur,

1960; Pianka, 1966; MacArthur, 1972; Stevens, 1989; Jocque

et al., 2010). However, while some researchers have found

evidence for a positive relationship between specialization and

species richness (sometimes using latitude as a proxy for

richness; Pagel et al., 1991; Eeley & Foley, 1999; Cardillo, 2002;

Dyer et al., 2007; Krasnov et al., 2008b; Mason et al., 2008),

others have not (Beaver, 1979; Lappalainen & Soininen, 2006;

Novotny et al., 2006; Belmaker, 2009; Filippi-Codaccioni et al.,

2010). Moreover, a meta-analysis failed to support the

suggested richness–specialization association (Vazquez & Ste-

vens, 2004).

Despite the prominence and potential relevance of the

richness–specialization hypothesis, to date its empirical under-

standing has been limited. In this study we address this

shortcoming by providing the first global assessment of

specialization in relation to species richness for an entire large

clade. We argue that a synthetic understanding of the

purported effect of specialization on richness requires the

additional consideration of two rarely examined issues: scale

and context-dependence.

First, the strength and directionality of the richness–

specialization relationship may change across scales. While

some attention has been given to the scale-dependence of

specialization within species (Krasnov et al., 2008a; Devictor

et al., 2010), few studies have explored the scaling of commu-

nity-level specialization. If richness is limited by the ability to

partition local resources, we may expect richness–specializa-

tion associations to be strong at fine scales, but weaker at

regional scales where constraints on coexistence and effects of

resource availability are unlikely to be major forces. If positive

richness–specialization relationships were to arise under a

scenario of richness driving specialization to facilitate the

partitioning of resources (MacArthur, 1965; Rosenzweig & Ziv,

1999), we may similarly expect the patterns to be most

pronounced at fine scales where resources are most limiting.

Alternatively, richness–specialization associations at the assem-

blage level may primarily arise from regional-scale evolution-

ary diversification of lineages. For example, specialization may

be associated with reduced dispersal and hence increased

diversification rates (Jocque et al., 2010). As these evolutionary

processes are manifested at a regional scale (Ricklefs, 2006), a

positive richness–specialization relationship would accordingly

be present at both regional and fine scales.

Second, the effect of specialization on species coexistence

may depend on environmental conditions. When richness is

depressed by external conditions (e.g. disturbance, dispersal

limitation, harsh abiotic conditions), abiotic constraints are

more likely than competition to structure species assemblages.

We may therefore be more likely to observe a strong

relationship between specialization and richness in assemblages

that exist close to carrying capacity. Here we use estimates of

net primary productivity and habitat heterogeneity to repre-

sent environmental constraints. Assemblages that contain

more species than expected given environmental conditions

are more likely to be limited by their ability to divide niche

space. Thus, we expect to find context-dependence in the

richness–specialization association: specialization may limit

richness most strongly in assemblages that are species-rich

relative to environmental conditions.

In this study we use birds to provide the first global

exploration of the geography of specialization and its associ-

ation with species richness, addressing both context- and scale-

dependence. We expect richness–specialization relationships to

be strongly positive, especially in assemblages that are species-

rich relative to their environment, and steeper at fine than

coarse scales. By assessing these relationships we expect to

identify the role of specialization in global richness gradients.

MATERIALS AND METHODS

Assemblage data

Breeding season distributions for non-marine birds were

compiled from the best available sources for a given region

or taxonomic group (Jetz et al., 2007). Based on these

distributions, species occurrence was determined within six

realms (Afrotropics, Australasia, Indo-Malaya, Nearctic, Neo-

tropics and Palaearctic), 58 biome-realms (ecosystem type

nested within realms; following Olson et al., 2001) and 9731

equal-area grid cells of 12,100 km2 (110 · 110 km grids at the

equator; cells with terrestrial surface < 20% or richness < 10

were excluded). Additionally, we compiled 398 fine-grained

assemblages from an extensive survey of the literature for well-

sampled and thoroughly documented species inventories

(Meese, 2005; Belmaker & Jetz, 2011). Here we consider

assemblages from areas of 10 to 7875 km2 (median: 488 km2,

equivalent to c. 22 · 22 km grid cells) as fine-grained, with

four larger assemblages in noticeably underrepresented regions

(for an among-realm comparison see Fig. S1 in Appendix S1 in

Supporting Information). Sub-setting the assemblages by area

(30% largest vs. 30% smallest) gave qualitatively identical

J. Belmaker et al.

2 Journal of Biogeographyª 2011 Blackwell Publishing Ltd

results (Fig. S2 in Appendix S1). While these assemblages are

fine-grained relative to typical macroecological studies, data

limitations prevent us from extending these assemblages to yet

finer grains, where biological interactions may increase in

importance, without incurring geographic and environmental

biases. Variation in fine-grained assemblage area was addressed

by including area (log10-transformed area in km2) as a

covariate in multi-predictor models. Following model diag-

nostics we log10-transformed richness.

Specialization

We used two indices of specialization that describe species-

level resource use: diet and habitat specialization. Diet

specialization was estimated from a comprehensive literature

survey, but based primarily on del Hoyo et al. (2008) (see

Wilman, 2011). Diet information included the proportional

use (frequency of use relative to all other diet categories) of

each of eight dietary categories (seeds, fleshy fruits, nectar,

invertebrates, carrion, fish, other vertebrates, and other plant

material). These were combined into a single value using the

Levins’ index (Krebs, 1999):

B ¼ 1=Rp2i ð1Þ

where pi is the proportional use of dietary category i.

Habitat specialization was estimated from the number of

distinct habitats used by each species, as reported in a fine-

scale IUCN assessment (Appendix S2). Critical to the analyses

conducted here, the habitat specialization data are globally

standardized. Although simplistic, counts of commonly used

resources frequently perform as well as more detailed data

(Krebs, 1999). Note that the number of distinct habitats

returns values that are identical to those obtained from

equation 1 when all used habitats are assigned equal propor-

tions.

The maximum values of diet and habitat specialization

indices will depend on the numbers of categories. We therefore

standardized both dietary and habitat indices to vary between 0

and 1, and then subtracted this value from one so that higher

index values indicate higher specialization:

BA ¼ 1� ðB� 1Þ=ðn� 1Þ ð2Þ

where B is the unstandardized diet or habitat specialization

index for each species and n is constant across all species and

reflects the total number of dietary states (8) or the maximum

number of habitats occupied by a species (12). Specialization

indices displayed a left skew across species and assemblages,

indicating that most species are specialized, and relatively few

are generalists (Fig. 1). However, after performing a logit

transformation (Warton & Hui, 2011) skew was substantially

reduced (Fig. S3 in Appendix S1). We therefore used the mean

across species of the logit-transformed specialization to

represent assemblage-typical specialization values. Specializa-

tion values of one were given a value of 0.975 before

transformation. Habitat and dietary specialization are consid-

erably correlated (12,100 km2 grain: r = 0.68).

At least three unavoidable methodological limitations must

be acknowledged. First, treating specialization as a species-level

attribute is a simplification. However, we expect intraspecific

variation to be small relative to the strong interspecific

differences among all birds. Second, specialization would

ideally be measured at the finest ecological resolution possible.

However, there is no obvious ‘correct’ level of ecological detail

at which specialization should be measured, and using higher

resolution specialization data will inevitably entail a loss of

geographical and taxonomical coverage. Finally, specialization

may be confounded by geographical biases in human knowl-

edge; for example tropical species are typically less known than

their temperate counterparts. However, birds are by far the

best studied taxon and therefore present the least geographical

bias. These data limitations are currently unavoidable caveats

for a broad-scale view and birds provide a best-case study

system.

Auxiliary environmental variables

To estimate richness relative to environmental constraints and

to account for the possibility that assemblage-level specializa-

tion is driven by patterns of resource availability (Devictor

Num

ber

of s

peci

es

0.0 0.2 0.4 0.6 0.8 1.0

020

0040

00

(a) Speciesn = 9,061

0.0 0.2 0.4 0.6 0.8 1.0

010

0025

00

Ass

embl

ages

(b) Biome−realm

n = 58

0.82 0.88 0.92

04

812

0.38 0.62 0.82

05

1015

Ass

embl

ages

(c) 12,100km2

n = 10,011

0.82 0.88 0.92

010

0020

00

0.38 0.62 0.82

050

015

00

Diet specialization

Ass

embl

ages

(d) Fine−grained

n = 398

0.82 0.88 0.92

040

80

Habitat specialization0.38 0.62 0.82

020

60

Figure 1 (a) The frequency distribution of diet and habitat spe-

cialization among extant terrestrial bird species globally. (b–d)

Mean assemblage diet and habitat specialization for three spatial

scales: (b) biome-realm, (c) 12,100 km2 grid and (d) fine-grained.

Means were calculated following logit transformation.

Specialization and bird richness

Journal of Biogeography 3ª 2011 Blackwell Publishing Ltd

et al., 2010; Kissling et al., 2011), we extracted key environ-

mental predictors that (1) have been shown to be strong

predictors of avian richness gradients, and (2) are directly

related to resource availability. Average annual above-ground

net primary productivity was estimated using the Lund

Potsdam Jena dynamic global vegetation model, which

accounts for land use (Bondeau et al., 2007). For the finest

scale, we estimated productivity at 1 km resolution using

MODIS imagery and the MOD17 algorithm, averaged over the

years 2000–2006. We used the number of distinct land-cover

types, based on 96-class USGS 1 km resolution GLCC product

(http://edc2.usgs.gov/glcc/), as a measure of habitat heteroge-

neity. All environmental predictors were log10-transformed

prior to analyses.

Analyses

We predict that specialization will only limit richness when

richness is high relative to the local environmental constraints.

We tested this by: (1) applying quantile regressions to pinpoint

the most diverse assemblages (we used the 0.9 quantile but

other high quantiles gave similar results; Fig. S4 in Appen-

dix S1), (2) statistically accounting for environmental variables

which may constrain richness (productivity, habitat heteroge-

neity and, for fine-grained assemblages, area) by including

them in a multiple regression framework, and (3) splitting the

data into quartiles according to the residuals of a linear model

relating species richness to environmental variables. For the

latter, we postulate that the richness–specialization relation-

ship will be stronger for higher quartiles, i.e. in assemblages

which are richest relative to the environmental constraints.

As both diet and habitat specialization show strong variation

and the directionality of the relationship is not straightfor-

ward, we used reduced major axis regressions, which are

symmetrical, in addition to ordinary least squares. Sample size

differences might hinder direct comparison between scales, so

we repeated some analyses for the 12,100 km2 grid using only

cells containing fine-grained assemblages. Pseudo-R2 for

quantile regressions was determined as 1 – (sum of weighted

distances/sum of weighted distances for an intercept-only

model). All analyses were performed in R v. 2.13.0 (R

Development Core Team, 2011). Quantile regressions were

performed using the package ‘quantreg’ (Koenker, 2011) and

reduced major axis regressions using the package ‘lmodel2’

(Legendre, 2008).

To account for spatial autocorrelation we also applied

simultaneous autoregressive (SAR) models. We used row-

standardized SAR spatial error models, examined a range of

possible neighbourhood sizes (200–3800 km), and chose the

model with the lowest Akaike information criterion (AIC)

score. We calculated pseudo-R2 values for the SAR models as

the squared Pearson correlation between observed values and

those predicted by the spatial models. Here, R2fit represents the

total amount of variance explained by the model, including the

spatial component, while R2trend represents the amount of

variance explained by the non-spatial smooth component

alone. Moran’s I global test was used to determine whether

residual autocorrelation persisted after fitting the SAR model.

The significance of Moran’s I standard deviate was assessed by

randomization. SAR analyses were performed using the

package ‘spdep’ (Bivand, 2011).

When estimated as the number of habitats within the entire

species range, habitat specialization may correlate with range

size (e.g. Davies et al., 2009). In this study we reduce this

inherent correlation by using independent specialization

estimates for each species based on observed habitat utiliza-

tion. Yet, as a correlation remained (12,100 km2 grid:

r = )0.66, fine-grained assemblages: r = )0.72) we repeated

the analysis using log10 mean range size as an additional

predictor (Fig. S5 in Appendix S1). Because range size itself is a

measure of specialization (specialized species usually have

narrow ranges), further attempts to control for range-size

variation run the risk of inadvertently controlling for special-

ization patterns.

To account for sampling effects potentially causing a

spurious richness–specialization association, we constructed a

null model in which species were randomly drawn from the

global species list, weighted by their corresponding range size.

This approach preserves local richness and the global species–

specialization distribution, but assumes no dispersal limita-

tion. We compared the shape and slope of the null

relationship to empirical patterns to determine whether a

bounded species–specialization relationship is expected by

chance alone.

RESULTS

Avian diet and habitat specialization are distributed highly

unevenly between species, with many more specialized than

generalist species (Fig. 1a). Average values of assemblage

specialization show strong variation at each of the three scales

of analysis (biome-realm, 12,100 km2 grid cells and fine-

grained assemblages), with a small trend towards higher mean

specialization at coarse scales (Fig. 1b–d). Mean assemblage-

level avian diet and habitat specialization show distinct

geographical and latitudinal variation. Both habitat and diet

specialization are highest at low latitudes, decrease towards

middle latitudes, and increase again at extremely high latitudes

(Fig. 2). Comparing 12,100 km2 grid cells, mean diet special-

ization exhibits noticeably high values across South America,

much of sub-Saharan Africa, Southeast Asia and parts of

Siberia. Habitat specialization also presents clear geographic

gradients, with high values most prominently in the Neotrop-

ics, Southeast Asia and the Tibetan Plateau and low values

across most of the Palaearctic, Australia and southern Africa

(Fig. 2).

We find strong geographic similarities between assemblage-

level specialization and known richness gradients (Fig. 2).

Clear relationship between richness and mean diet and habitat

specializations exist at the realm and biome-realm scales

(Fig. 3; upper panels). Under a global null model, the

relationship between richness and specialization is funnel-

J. Belmaker et al.


−60

−20

020

4060

(a) Diet specialization

0.790.880.890.90.93

−60

−20

020

4060

(b) Habitat specialization

0.290.480.540.630.8

−160 −120 −80 −40 0 40 80 120 160

−60

−20

020

4060

(c) Richness

10110170260560

Figure 2 The global distribution of mean (a) dietary specialization, (b) habitat specialization and (c) species richness of terrestrial birds

within 12,100 km2 grid cells. Points represent fine-grain assemblages. Columns to the left represent means across latitudinal bands using the

same colour scale as the maps.



shaped, with large variability at low richness and decreasing

variance – but otherwise no trend – with increasing richness

(Fig. 3; middle panels, dashed lines). The slopes of the null

relationship between richness and specialization are low for

both ordinary least squares (OLS) [Diet: 0.0 ()0.10–0.11 95%

confidence interval (CI)); habitat, )0.02 ()0.06–0.01)] and 0.9

quantile regressions [diet: 0.0 ()0.13–0.14); habitat,

0.01()0.10–0.10)]. This contrasts strongly with empirical

patterns for 12,100 km2 grid cells, which above c. 100 species

richness becomes increasingly restricted from above at low

values of specialization (Fig. 3; middle panels) creating a

positive richness–specialization relationship [OLS: diet

0.80 ± 0.01 (SE), habitat 0.22 ± 0.01; 0.9 quantile regression:

diet 0.81 ± 0.01, habitat 0.30 ± 0.01]. The 0.9 quantile

regression slopes are steeper for the 12,100 km2 scale than

for the fine-grained assemblages [significant interaction

between the two-level categorical predictor scale (12,100 km2

grid and fine-grained assemblages) and specialization as a

continuous predictor; for habitat: t value (P) = 3.75 (< 0.001);

for diet: 5.46 (< 0.001)]. Strong richness–specialization asso-

ciations are also found within individual realms (Fig. 3; lower

panels).

We predicted that richness–specialization associations

would remain strong, even after accounting for gradients in

environmental resource availability. Our analyses confirm

previous evidence that proxies for broad-scale gradients in

resource availability, such as net primary productivity and

habitat heterogeneity, are strong predictors of species richness

(Table 1). However, additionally fitting both diet and habitat

specialization in a multi-predictor setting reveals strong,

complementary and positive effects on richness, especially at

the 12,100 km2 grid scale (Table 1). The richness–specializa-

tion relationship also remains strong when using major axis

regressions (Table 2), which are symmetrical and thus do not

assume any cause and effect relationship.

To assess these associations in more detail, we separate

assemblages according to the residuals of the richness–

environment (productivity, habitat heterogeneity and area)

relationship. Large positive residuals represent particularly

species-rich assemblages given environmental conditions, and

Ric

hnes

s(a)

1010

010

00

(b)RealmBiome−realm

Ric

hnes

s

(c)

1032

100

316

1000

(d)

12,100km2

Fine grained

Null

Diet specialization

Ric

hnes

s

(e)

0.82 0.88 0.92

100

316

1000

Habitat specialization

(f)

0.38 0.62 0.82

PalaearcticNearctic

Indo−MalayanNeotropicalAfrotropicalAustralasia

Figure 3 The relationship between global

terrestrial bird species richness and diet and

habitat specialization. (a, b) The relationships

across realms and biome-realms. Lines rep-

resent ordinary least square regressions. (c, d)

Black points represent fine-grained assem-

blages; coloured dots 12,100 km2 grid cells

(darkness increases with dot density).

Regression lines represent 0.9 quantile

regression estimates (12,100 km2 grid: pseu-

do-R2 = 0.32, 0.20; coefficient = 0.94, 0.31;

fine-grained assemblages: pseudo-R2 = 0.24,

0.12; coefficient = 0.54, 0.18; for diet and

habitat specialization, respectively);

P < 0.001 in all cases. Regions defined by the

dashed lines represent the 95% confidence

interval of the null expectation (see Materials

and Methods). Regression lines at the

12,100 km2 scale are plotted using only grid

cells in which fine-grained assemblages are

located to maintain sample size for direct

comparison. (e, f) Richness–specialization

associations for 12,100 km2 grid cells within

each of the major realms. Lines represent 0.9

quartile regressions (solid lines correspond to

significant slopes). Only assemblages with

> 100 species were included.

J. Belmaker et al.


we expected to observe an especially strong signal of ecological

niche partitioning, facilitating coexistence, for these groups.

This prediction was only partially supported as the relationship

between richness and specialization is indeed steep in assem-

blages at high residual quartiles (quartiles 3 and 4; Table 2),

but also in assemblages belonging to low residual quartiles (e.g.

quartile 1).

Even after accounting for geographic variation in range size,

habitat and particularly diet specialization remain key predic-

tors of richness (Fig. S5 in Appendix S1). We also find that the

richness–specialization relationship is steeper for the 50% most

wide-ranged species than the 50% narrow-ranged ones (Fig. S6

in Appendix S1).

DISCUSSION

We demonstrate that, for an entire large clade across spatial

scales, specialization and species richness are strongly related.

Importantly, specialization remains positively associated with

richness even after accounting for known environmental

correlates of richness, such as productivity. Our results suggest

that avian dietary and habitat specialization set upper bounds

on species richness, indicating that high richness values are not

attained in generalist assemblages (Fig. 2). Conversely, species-

rich assemblages are more likely to contain specialized species

than assemblages that are species-poor. These findings con-

tradict several recent studies (Vazquez & Stevens, 2004;

Novotny et al., 2006) and suggest that the contribution of

specialization to global richness gradients may need to be

carefully reassessed, taking into consideration the grain, extent

and resolution of specialization categorization.

We would like to emphasize that if species richness and

specialization were not linked, but rather both associated with

areas containing more habitats or dietary types, we would

expect a negative, not a positive, relationship between special-

ization and richness. Simple random sampling of the environ-

ment would cause species to have wider diets and use more

habitats in heterogeneous and therefore species-rich areas.

Similarly, species would be habitat and diet specialists in

localities with limited diet or habitat options (e.g. few diet

categories used in cold or high-elevation regions might reflect

limitations of available feeding strategies). However, we find

the opposite pattern: an increase in specialization with

richness, which is more consistent with our expectation.

We find that the relationship between specialization and

species richness is steep even at coarse scales (realm and

biome-realm) and steeper for the 12,100 km2 scale than for

fine-grained assemblages. If the observed associations between

richness and specialization were simply the result of contem-

porary ecological filters (e.g. those precluding non-specialists

from entering local assemblages), we would expect to observe

shallower relationships at coarse scales. The same argument

holds also when viewing a reverse direction of causality. If

competition for limited local resources causes species in

diverse regions to specialize, we would predict shallower

richness–specialization associations at coarse scales where

resource partitioning is not a major limitation. Instead, it

appears that the positive richness–specialization relationship is

Table 1 Multi-predictor models of species richness in terrestrial birds (9061 species) globally. Ordinary least squares, quantile (0.9 quantile)

and simultaneous autoregressive (SAR) regressions are presented. For all regressions, predictors included diet specialization (Diet spec.),

habitat specialization (Habitat spec.), net primary productivity, habitat heterogeneity (Habitat heter.) and, for fine-grained assemblages only,

area. Neighbourhood sizes used for the SAR models were 3000, 200 and 1400 km for the biome-realm, 12,100 km2 grid and fine-grained

assemblages, respectively. Moran’s I values (P) were 0.70 (0.24), 21 (***) and )0.43 (0.7) for the biome-realm, 12,100 km2 grid and fine-

grained assemblages, respectively.

Scale Variable

Ordinary least squares Quantile regression Simultaneous autoregressive

Coef. (SE) t value Coef. (SE) t value Coef. (SE) z value

Biome- Realm Habitat spec. 0.27 (0.16) 1.6 0.24 (0.15) 1.6 0.63 (0.15) 4.1***

Diet spec. 0.50 (0.29) 1.7 0.93 (0.40) 2.4* 0.27 (0.36) 0.7

Productivity 0.09 (0.14) 0.6 )0.18 (0.16) )1.14 )0.06 (0.12) )0.5

Habitat heter. 0.32 (0.27) 1.2 0.26 (0.28) 0.92 0.28 (0.26) 1.0

Summary: R2adj = 0.37, P < 0.001 R2

Pseudo = 0.47, P < 0.001 R2trend = 0.36, R2

fit = 0.58

12,100 km2 grid Habitat spec. 0.05 (0.03) 7.8*** 0.08 (0.01) 11.1*** 0.18 (0.01) 29.8***

Diet spec. 0.54 (0.01) 41*** 0.52 (0.01) 37.1*** 0.04 (0.01) 3.9***

Productivity 0.22 (0.00) 58*** 0.17 (0.01) 27.0*** 0.07 (0.00) 25.6***

Habitat heter. 0.19 (0.01) 20*** 0.16 (0.01) 13.7*** 0.05 (0.00) 17.4***

Summary: R2adj = 0.60, P < 0.001 R2

Pseudo = 0.43, P < 0.001 R2trend = 0.41, R2

fit = 0.99

Fine-grained Habitat spec. )0.00 (0.03) )0.15 )0.03 (0.02) )1.6 )0.01 (0.04) )0.3

Diet spec. 0.21 (0.06) 3.5*** 0.55 (0.06) 8.8*** 0.31 (0.07) 4.1***

Productivity 0.18 (0.03) 6.9*** 0.17 (0.02) 9.0*** 0.16 (0.03) 5.6***

Habitat heter. 0.03 (0.05) 0.7 0.07 (0.03) 2.1* 0.06 (0.05) 1.2

Area 0.03 (0.02) 1.7 0.01 (0.01) 0.6 0.02 (0.02) 1.3

Summary: R2adj = 0.21, P < 0.001 R2

Pseudo = 0.33, P < 0.001 R2trend = 0.21, R2

fit = 0.27

Coef., coefficient; *, P < 0.05; ***, P < 0.001. Significant (P < 0.05) relationships are italicized and marked in bold.



shaped by regional evolutionary processes, which are expressed

from regional to fine grains (Ricklefs, 2008). For example, the

large radiation of the Furnariidae and Tyrannidae families

(Irestedt et al., 2006; Ohlson et al., 2008), which have relatively

narrow dietary niches, exerts a signal on the high dietary

specialization of the South American assemblages (Fig. 2).

Studies conducted at a global extent are limited by the

coarseness of the available distribution data, which applies to

specialist species even more than generalist species (Jetz et al.,

2008). By including relatively fine-grained assemblages we

show that, down to a grain equivalent to c. 22 · 22 km grid

cells (median fine-grain assemblage size was 488 km2), the

richness–specialization slope does not steepen. This suggests

that any signature of local resource partitioning on assem-

blage-level specialization may be weak, or may be most

apparent at yet finer scales.

Although the richness–specialization relationship is strong

and consistent in shape across realms, considerable scatter

remains, and a wide range of richness values can be found at

each specialization level (Fig. 3), even after accounting for

environmental conditions (Table 1). In this study, specializa-

tion estimates were obtained using relatively coarse resolution

data and it is probable that most species can differentiate finer

habitat and diet delimitations. However, even a warbler that

distinguishes between different microhabitats in a single tree

(MacArthur, 1958), may use both a montane and a lowland

forest (see categories in Appendix S2). Therefore, it is unclear

what habitat and diet resolutions are the most appropriate.

Nevertheless, using coarse habitat and diet categorizations,

which are likely to be strongly conserved phylogenetically,

could have contributed to the large variability in the richness–

specialization associations. Stronger associations may be

uncovered by using more detailed specialization data acquired

at finer scales.

Specialization should ideally be measured by comparing

observed resource utilization with resource availability at the

same geographic location. However, dietary resources and

fine-grained habitats are not available globally and therefore, in

practice, such standardization is rarely performed (Vazquez &

Stevens, 2004). Here, we use habitat variety and net primary

productivity as proxies for habitat and diet availability. The

strong and positive influence of specialization on richness was

evident even after accounting for these gradients in resource

availability (Table 1). Therefore, the richness–specialization

association does not simply mirror well-established environ-

mental gradients. Productivity, in particular, has been

Table 2 The relationship between specialization and species richness in 9061 terrestrial bird species globally. Single-predictor analyses are

presented in which the response is species richness and the predictor is either habitat or diet specialization. For the 12,100 km2 grid

and fine-grained assemblages, data were partitioned and analysed separately for different quartiles (small sample size prevented this for

the coarser scales). Quartile delineation was derived from first relating species richness to environmental variables (productivity, habitat

heterogeneity and, for fine-grained assemblages, area) and then separating the assemblages according to the residuals of this model.

Quartiles range from first (lowest 25% residuals) to fourth (highest 25% residuals).

Scale Specialization Quartiles

RMA OLS

Coef. (95% CI) Coef. (SE) t value

Realm Habitat spec. all 0.6 ()0.2–3.2) 0.5 (0.3) 1.9

Diet spec. all 1.4 (0.7–3.6) 1.2 (0.3) 3.9*

Biome-Realm Habitat spec. all 0.8 (0.6–1.2) 0.5 (0.1) 5.5***

Diet spec. all 2.2 (1.6–3.6) 0.9 (0.2) 5.1***

12,100 km2 grid Habitat spec. 1 0.77 (0.74 )0.80) 0.19 (0.01) 12.8***

2 0.60 (0.58–0.62) 0.04 (0.01) 3.0**

3 0.80 (0.77–0.83) 0.16 (0.02) 9.9***

4 0.68 (0.65–0.70) 0.33 (0.01) 27.8***

Diet spec. 1 1.70 (1.64–1.76) 0.76 (0.03) 25.1***

2 1.40 (1.34–1.45) 0.21 (0.03) 7.9***

3 1.89 (1.79–1.92) 0.81 (0.03) 24.4***

4 1.54 (1.50–1.59) 1.02 (0.02) 43.8***

Fine-grained Habitat spec. 1 0.38 (0.31–0.47) 0.02 (0.04) 0.6

2 0.20 (0.17–0.25) 0.01 (0.02) 0.7

3 0.20 (0.16–0.24) 0.05 (0.02) 2.7***

4 0.30 (0.25–0.36) 0.13 (0.03) 4.9***

Diet spec. 1 0.63 (0.52–0.76) 0.24 (0.06) 4.1***

2 )0.42 ()0.52–)0.35) )0.02 (0.04) )0.4

3 0.53 (0.44–0.64) 0.18 (0.05) 3.6***

4 0.75 (0.65–0.87) 0.51 (0.05) 9.1***

Coef., coefficient; Diet spec., diet specialization; Habitat heter., habitat heterogeneity; Habitat spec., habitat specialization; *, P < 0.05; **, P < 0.01; ***,

P < 0.001. Significant (P < 0.05) relationships (for ordinary least squares regression; OLS) or slopes with 95% confidence intervals (CI) excluding

zero (for reduced major axis regression; RMA) are italicized and marked in bold.

J. Belmaker et al.


suggested to promote specialization (MacArthur, 1965; Evans

et al., 2006), and hence may be a confounding factor when

analysing richness–specialization associations. This study dem-

onstrates that specialization has an independent contribution

to richness above and beyond that attributed to productivity.

Hence, and assuming richness is casually linked to specializa-

tion, gradients in resource availability (manifested as more

habitats or energetic niches, Hurlbert, 2004; Anderson & Jetz,

2005; Clarke & Gaston, 2006; Novotny et al., 2006; Belmaker,

2009; Kissling et al., 2011) may operate in tandem with

gradients in specialization to establish richness patterns.

Range size is in itself a measure of specialization (Stevens,

1989; Jocque et al., 2010). It is thus difficult to control for

range-size variation without running the risk of inadvertently

controlling for specialization patterns. Nevertheless, diet and

habitat specialization remain independent contributors to

richness, even after accounting for geographic variation in

range size (Fig. S5 in Appendix S1). Therefore, the richness–

specialization relationship does not solely reflect known

geographic variation in range sizes, most notably the decrease

in range size towards low latitudes (Stevens, 1989). Wide-

ranged species contribute disproportionally to richness gra-

dients (Jetz & Rahbek, 2002). In accordance, we find that the

richness–specialization relationship is strongest for the 50%

most wide-ranged species (Fig. S6 in Appendix S1). This

suggests that much of the global relationship may be

attributed to the replacement of wide-ranged generalists in

species-poor regions by wide-ranged specialists in species-rich

regions.

The positive but largely coarse-scale association between

richness and specialization supports several, not mutually

exclusive, hypotheses. First, specialization might facilitate the

origination of high-diversity assemblages, in addition (or

instead) of an effect on contemporary assembly. For example,

specialization may promote high diversification rates by

increasing reproductive isolation between populations (Haw-

thorne & Via, 2001; Jocque et al., 2010). In addition, a positive

richness–specialization relationship may form over evolution-

ary time and at coarse geographical scales if specialization

increases the effectiveness of resource use, consequently

reducing the within-clade ecological limits on diversification

(Phillimore & Price, 2008; Rabosky, 2009). In both of these

examples a positive assemblage-level richness–specialization

relationship emerges due to intrinsic differences among clades

in diversification rates and may thus be independent of local

environmental constraints (although the environment may

strongly influence specialization levels; Jocque et al., 2010).

This is in line with our results, which reveal only limited

support for context-dependence in the richness specialization

association, as we do not find an increase in the steepness of

the richness–specialization relationship at higher quartiles of

richness–environment residuals.

A second explanation for the positive richness–specializa-

tion relationship is that the association between the two is in

fact not causal, but that instead both are linked indepen-

dently to a third variable. We find the richness–specialization

association to be independent of prominent environmental

gradients (notably productivity), but it is impossible to

exclude the possibility that unmeasured environmental vari-

ables or life-history traits affect both specialization and

richness.

Finally, the direction of causality may in fact be reversed and

the positive richness–specialization relationship primarily

brought about by richness influencing patterns of specializa-

tion. Diverse assemblages may be more likely than species-

poor assemblages to coevolve towards increased specialization

(MacArthur, 1965; Rosenzweig & Ziv, 1999). However, we

expect such a pattern to be largely confined to fine scales where

competition for limited resources, postulated to drive such a

process, should be strongest. It has also been suggested that

asymmetrical interaction networks, where specialists tend to

interact with generalists, can cause a positive correlation

between specialization and richness (Vazquez & Stevens, 2004).

However, here too we may expect the pattern to be most

pronounced at fine scales where species interactions take place.

We cannot currently envision a mechanism by which higher

species richness will cause gradients in specialization primarily

at coarse scales. Nevertheless, lacking fully resolved phyloge-

netic trees and data on the past occurrences and interactions of

species, we are unable to directly quantify the effect of regional

diversity on the evolution of specialization across clades. Thus,

the goal of this study was a first geographic assessment and a

general test of ideas regarding niche-partitioning and com-

munity assembly. Separating the directionality of effects will

require further research that may need to be limited in

geographic and taxonomic scope.

In conclusion, we demonstrate that globally high bird

species richness is positively associated with specialization. We

interpret our findings across scales as indicating a relatively

minor role for niche-partitioning mechanisms in causing this

pattern. We suggest that our results support a regional and

evolutionary perspective for understanding how specialization

is associated with global geographic gradients in species

richness.

ACKNOWLEDGEMENTS

We would like to thank R. Meese for help with the fine-grained

data and D.M. Post, K. Mertes-Schwartz and N. Cooper for

comments. H. Wilman, C. de la Rosa and J. Simpson compiled

the bird diet data. The study was partially supported by

National Science Foundation (NSF) grants DBI 0960550 and

DEB 1026764 to W.J. C.H.S acknowledges support from the

Christensen Fund.

REFERENCES

Anderson, K.J. & Jetz, W. (2005) The broad-scale ecology of

energy expenditure of endotherms. Ecology Letters, 8, 310–

318.

Beaver, R.A. (1979) Host specificity of temperate and tropical

animals. Nature, 281, 139–141.



Belmaker, J. (2009) Species richness of resident and transient

coral-dwelling fish responds differentially to regional

diversity. Global Ecology and Biogeography, 18, 426–436.

Belmaker, J. & Jetz, W. (2011) Cross-scale variation in species

richness–environment associations. Global Ecology and Bio-

geography, 20, 464–474.

Bivand, R. (2011) spdep: spatial dependence: weighting schemes,

statistics and models. R package version 0.5-31. Available at:

http://cran.r-project.org/web/packages/spdep/index.html.

Bondeau, A., Smith, P.C., Zaehle, S., Schaphoff, S., Lucht, W.,

Cramer, W. & Gerten, D. (2007) Modelling the role of

agriculture for the 20th century global terrestrial carbon

balance. Global Change Biology, 13, 679–706.

Cardillo, M. (2002) The life-history basis of latitudinal diver-

sity gradients: how do species traits vary from the poles to

the equator? Journal of Animal Ecology, 71, 79–87.

Chase, J.M. & Leibold, M.A. (2003) Ecological niches: linking

classical and contemporary approaches. The University of

Chicago Press, Chicago, IL.

Clarke, A. & Gaston, K.J. (2006) Climate, energy and diversity.

Proceedings of the Royal Society B: Biological Sciences, 273,

2257–2266.

Connell, J.H. & Orias, E. (1964) The ecological regulation of

species diversity. The American Naturalist, 98, 399–414.

Davies, T.J., Purvis, A. & Gittleman, J.L. (2009) Quaternary

climate change and the geographic ranges of mammals. The

American Naturalist, 174, 297–307.

Devictor, V., Clavel, J., Julliard, R., Lavergne, S., Mouillot, D.,

Thuiller, W., Venail, P., Villeger, S. & Mouquet, N. (2010)

Defining and measuring ecological specialization. Journal of

Applied Ecology, 47, 15–25.

Dyer, L.A., Singer, M.S., Lill, J.T., Stireman, J.O., Gentry, G.L.,

Marquis, R.J., Ricklefs, R.E., Greeney, H.F., Wagner, D.L.,

Morais, H.C., Diniz, I.R., Kursar, T.A. & Coley, P.D. (2007)

Host specificity of Lepidoptera in tropical and temperate

forests. Nature, 448, 696–699.

Eeley, H.A.C. & Foley, R.A. (1999) Species richness, species

range size and ecological specialisation among African pri-

mates: geographical patterns and conservation implications.

Biodiversity Conservation, 8, 1033–1056.

Evans, K.L., Jackson, S.F., Greenwood, J.J.D. & Gaston, K.J.

(2006) Species traits and the form of individual species–

energy relationships. Proceedings of the Royal Society B:

Biological Sciences, 273, 1779–1787.

Filippi-Codaccioni, O., Devictor, V., Bas, Y. & Julliard, R.

(2010) Toward more concern for specialisation and less for

species diversity in conserving farmland biodiversity. Bio-

logical Conservation, 143, 1493–1500.

Hawthorne, D.J. & Via, S. (2001) Genetic linkage of ecological

specialization and reproductive isolation in pea aphids.

Nature, 412, 904–907.

del Hoyo, J., Elliott, A. & Christie, D.A. (2008) Handbook of the

birds of the world. Lynx Edicions, Barcelona.

Hurlbert, A.H. (2004) Species–energy relationships and

habitat complexity in bird communities. Ecology Letters, 7,

714–720.

Hutchinson, G.E. (1959) Homage to Santa Rosalia or why are

there so many kinds of animals? The American Naturalist,

93, 145–159.

Irestedt, M., Fjeldsa, J. & Ericson, P.G.P. (2006) Evolution of

the ovenbird-woodcreeper assemblage (Aves: Furnariidae) –

major shifts in nest architecture and adaptive radiation.

Journal of Avian Biology, 37, 260–272.

Jetz, W. & Rahbek, C. (2002) Geographic range size and

determinants of avian species richness. Science, 297, 1548–

1551.

Jetz, W., Wilcove, D.S. & Dobson, A.P. (2007) Projected

impacts of climate and land-use change on the global

diversity of birds. PLoS Biology, 5, 1211–1219.

Jetz, W., Sekercioglu, C.H. & Watson, J.E.M. (2008) Ecological

correlates and conservation implications of overestimating

species geographic ranges. Conservation Biology, 22, 110–

119.

Jocque, M., Field, R., Brendonck, L. & De Meester, L. (2010)

Climatic control of dispersal–ecological specialization trade-

offs: a metacommunity process at the heart of the latitudinal

diversity gradient? Global Ecology and Biogeography, 19, 244–

252.

Kissling, W.D., Sekercioglu, C.H. & Jetz, W. (2011) Bird die-

tary guild richness across latitudes, environments and

biogeographic regions. Global Ecology and Biogeography. doi:

10.1111/j.1466-8238.2011.00679.x

Klopfer, P.H. & MacArthur, R.H. (1960) Niche size and faunal

diversity. The American Naturalist, 94, 293–300.

Koenker, R. (2011) quantreg: quantile regression. R package

version 4.62. Available at: http://cran.r-project.org/web/

packages/quantreg/index.html.

Krasnov, B.R., Mouillot, D., Khokhlova, I.S., Shenbrot, G.I. &

Poulin, R. (2008a) Scale-invariance of niche breadth in fleas

parasitic on small mammals. Ecography, 31, 630–635.

Krasnov, B.R., Shenbrot, G.I., Khokhlova, I.S., Mouillot, D. &

Poulin, R. (2008b) Latitudinal gradients in niche breadth:

empirical evidence from haematophagous ectoparasites.

Journal of Biogeography, 35, 592–601.

Krebs, C.F. (1999) Ecological methodology. Addison-Wesley

Educational Publishers, Menlo Park, CA.

Lappalainen, J. & Soininen, J. (2006) Latitudinal gradients in

niche breadth and position—regional patterns in freshwater

fish. Naturwissenschaften, 93, 246–250.

Legendre, P. (2008) lmodel2: model II regression. R package

version 1.6-3. Available at: http://cran.r-project.org/web/

packages/lmodel2/index.html.

MacArthur, R.H. (1958) Population ecology of some warblers

of northeastern coniferous forests. Ecology, 39, 599–619.

MacArthur, R.H. (1965) Patterns of species diversity. Biology

Reviews, 40, 510–533.

MacArthur, R.H. (1972) Geographical ecology: patterns in the

distribution of species. Harper & Row, New York.

Mason, N.W.H., Irz, P., Lanoiselee, C., Mouillot, D. & Argil-

lier, C. (2008) Evidence that niche specialization explains

species–energy relationships in lake fish communities.

Journal of Animal Ecology, 77, 285–296.

J. Belmaker et al.


Meese, R.J. (2005) Biological inventories of the world’s protected

areas. Information Center for the Environment, University

of California, Davis, CA.

Novotny, V., Drozd, P., Miller, S.E., Kulfan, M., Janda, M.,

Basset, Y. & Weiblen, G.D. (2006) Why are there so many

species of herbivorous insects in tropical rainforests? Science,

313, 1115–1118.

Ohlson, J., Fjeldsa, J. & Ericson, P.G.P. (2008) Tyrant fly-

catchers coming out in the open: phylogeny and ecological

radiation of Tyrannidae (Aves, Passeriformes). Zoologica

Scripta, 37, 315–335.

Olson, D.M., Dinerstein, E., Wikramanayake, E.D., Burgess,

N.D., Powell, G.V.N., Underwood, E.C., D’Amico, J.A.,

Itoua, I., Strand, H.E., Morrison, J.C., Loucks, C.J., Allnutt,

T.F., Ricketts, T.H., Kura, Y., Lamoreux, J.F., Wettengel,

W.W., Hedao, P. & Kassem, K.R. (2001) Terrestrial

ecoregions of the worlds: a new map of life on Earth. Bio-

Science, 51, 933–938.

Pagel, M.D., May, R.M. & Collie, A.R. (1991) Ecological as-

pects of the geographical distribution and diversity of

mammalian species. The American Naturalist, 137, 791–815.

Phillimore, A.B. & Price, T.D. (2008) Density-dependent

cladogenesis in birds. PLoS Biology, 6, 483–489.

Pianka, E.R. (1966) Latitudinal gradients in species diversity: a

review of concepts. The American Naturalist, 100, 33–46.

R Development Core Team (2011) R: a language and envi-

ronment for statistical computing. R Foundation for Statis-

tical Computing, Vienna, Austria.

Rabosky, D.L. (2009) Ecological limits and diversification rate:

alternative paradigms to explain the variation in species

richness among clades and regions. Ecology Letters, 12,

735–743.

Ricklefs, R.E. (2006) Evolutionary diversification and the ori-

gin of the diversity–environment relationship. Ecology, 87,

S3–S13.

Ricklefs, R.E. (2008) Disintegration of the ecological commu-

nity. The American Naturalist, 172, 741–750.

Rosenzweig, M.L. & Ziv, Y. (1999) The echo pattern of species

diversity: pattern and processes. Ecography, 22, 614–628.

Stevens, G.C. (1989) The latitudinal gradient in geographical

range: how so many species coexist in the tropics. The

American Naturalist, 133, 240–256.

Vazquez, D.P. & Stevens, R.D. (2004) The latitudinal gradient

in niche breadth: concepts and evidence. The American

Naturalist, 164, E1–E19.

Warton, D.I. & Hui, F.K.C. (2011) The arcsine is asinine:

the analysis of proportions in ecology. Ecology, 92,

3–10.

Wilman, H. (2011) The energetic niche of species: integrating

single-species and ecosystem perspectives on species’ role in

communities. PhD Thesis, University of California, San

Diego, CA.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the

online version of this article:

Appendix S1 Additional figures detailing the relationship

between specialization and species richness of birds of the

world (Figs S1–S6).

Appendix S2 Summary of the IUCN assessment habitat

categories.

As a service to our authors and readers, this journal provides

supporting information supplied by the authors. Such mate-

rials are peer-reviewed and may be re-organized for online

delivery, but are not copy-edited or typeset. Technical support

issues arising from supporting information (other than

missing files) should be addressed to the authors.

BIOSKETCHES

Jonathan Belmaker is a post-doctoral researcher inter-

ested in understanding how patterns of biodiversity and the

underlying processes vary across scales and environmental

gradients. He is particularly interested in the interplay

between large scale processes and local community assem-

bly.

Cagan Sekercioglu is an assistant professor at the University

of Utah. He is a conservation ecologist, ornithologist and

tropical biologist especially interested in the effects of global

change on biodiversity patterns and ecosystem services.

Walter Jetz is interested in the ecology, biogeography and

conservation of terrestrial vertebrates and plants.

Editor: Melodie McGeoch



ORIGINAL Global patterns of specialization and … 2011...reduced (Fig. S3 in Appendix S1). We...

Documents

Transcript of ORIGINAL Global patterns of specialization and … 2011...reduced (Fig. S3 in Appendix S1). We...