Life‐history traits predict species responses to habitat area and isolation: a cross‐continental synthesis

L E T T E RLife-history traits predict species responses to habitat

area and isolation: a cross-continental synthesis

Erik Ockinger,1* Oliver

Schweiger,2 Thomas O. Crist,3

Diane M. Debinski,4 Jochen

Krauss,5 Mikko Kuussaari,6

Jessica D. Petersen,7 Juha Poyry,6

Josef Settele,2 Keith S.

Summerville8 and Riccardo

Bommarco1

Abstract

There is a lack of quantitative syntheses of fragmentation effects across species and

biogeographic regions, especially with respect to species life-history traits. We used data

from 24 independent studies of butterflies and moths from a wide range of habitats and

landscapes in Europe and North America to test whether traits associated with dispersal

capacity, niche breadth and reproductive rate modify the effect of habitat fragmentation

on species richness. Overall, species richness increased with habitat patch area and

connectivity. Life-history traits improved the explanatory power of the statistical models

considerably and modified the butterfly species–area relationship. Species with low

mobility, a narrow feeding niche and low reproduction were most strongly affected by

habitat loss. This demonstrates the importance of considering life-history traits in

fragmentation studies and implies that both species richness and composition change in

a predictable manner with habitat loss and fragmentation.

Keywords

Connectivity, fragmentation, Lepidoptera, meta-analysis, quantitative synthesis, species

richness, species–area relationship.

Ecology Letters (2010) 13: 969–979

I N T R O D U C T I O N

Local species richness is predicted to decrease with

decreasing area of habitat patches and increasing isolation

from conspecific populations (MacArthur & Wilson 1967;

Hanski 1999; Fahrig 2003). Not all species, however,

respond equally to habitat loss and fragmentation, and it

has been suggested that the variability in species� responses

to habitat area and fragmentation can be explained by life-

history traits (Henle et al. 2004; Ewers & Didham 2006;

Prugh et al. 2008). When combinations of traits make

species more susceptible to habitat loss and fragmentation,

communities in small and isolated habitat patches will not

only contain fewer species than large and well-connected

ones, but will also differ in species composition. In highly

fragmented landscapes, communities are generally pre-

dicted to become increasingly similar in composition, and

dominated by few generalist species with similar traits

(McKinney & Lockwood 1999; Ekroos et al. 2010). How-

ever, predicting such responses may not be quite so simple.

For example, a recent study indicated non-univocal

responses of different species groups, indicating potential

interactions between niche breadth and dispersal capacity

(Dormann et al. 2007).

1Department of Ecology, Swedish University of Agricultural

Sciences, P.O. Box 7044, 75007 Uppsala, Sweden2UFZ, Helmholtz Centre for Environmental Research, Depart-

ment of Community Ecology, Theodor-Lieser-Strasse 4, 06120

Halle, Germany3Department of Zoology, Miami University, Oxford, OH, USA4Department of Ecology, Evolution, and Organismal Biology,

Iowa State University, Ames, IA 50011, USA5Population Ecology Group, Department of Animal Ecology I,

University of Bayreuth, Universitatsstrasse 30, D-95447

Bayreuth, Germany

6Finnish Environment Institute, Research Programme for Bio-

diversity, P.O. Box 140, FI-00251 Helsinki, Finland7Department of Entomology, Iowa State University, 424 Science

II, Ames, IA 50011, USA8Department of Environmental Science and Policy, Drake Uni-

versity, Olin Hall, Des Moines, IA 50311-4505, USA

*Correspondence: E-mail: [email protected]

Ecology Letters, (2010) 13: 969–979 doi: 10.1111/j.1461-0248.2010.01487.x

� 2010 Blackwell Publishing Ltd/CNRS

Because species persistence in fragmented habitats

depends on local extinction and colonization processes,

life-history traits associated with these processes are likely to

determine species responses to habitat loss and fragmenta-

tion. Dispersal capacity, niche breadth and reproductive

potential are three traits hypothesized to be key determi-

nants of species persistence in fragmented landscapes

(Kotiaho et al. 2005; Ewers & Didham 2006), each leading

to clearly defined expectations.

The importance of immigration from nearby populations

for population persistence (Brown & Kodric-Brown 1977)

and re-colonization following local extinction events (Han-

ski 1999) is likely to increase with decreasing patch and

population sizes. Hence, species with poor dispersal capacity

are predicted to be more strongly affected by reduction in

habitat area and connectivity compared to more mobile

species (Lomolino 1984; Ockinger et al. 2009). Conversely,

species with high dispersal capacity can move between

habitat patches and more efficiently utilize a fragmented

resource, making them less sensitive to increasing geo-

graphical isolation (Hanski & Ovaskainen 2000).

Species with narrow feeding niches are able to fulfil their

resource requirements in only the largest and most well-

connected habitat patches, where the probability of co-

occurrence of essential resources is largest. Resource

specialists are also less likely to utilize resources in the

�matrix� outside the primary habitat patches (Brotons et al.

2003). As a consequence, specialists will experience their

habitat as more fragmented compared to generalists, which

increases the sensitivity of specialists to habitat loss and

fragmentation (With & Crist 1995; Ewers & Didham 2006).

In fragmented habitats, species with high reproductive

potential may suffer less because a large number of

offspring implies a large number of potential emigrants,

increasing the chance of colonization of empty habitat

patches. It also results in more rapid population growth

which enhances the chances of colonization, successful

population establishment, and population recovery follow-

ing environmental change (Henle et al. 2004).

Despite the fairly well established predictions, there are

few empirical studies explicitly addressing whether species

with contrasting life-history traits are differentially affected

by habitat fragmentation (Henle et al. 2004). Life-history

traits tend to be highly correlated, and studying only the

effects of a single trait may impede the detection of trait-

mediated responses (Mattila et al. 2006). Therefore, there is a

need to analyse how species with certain combinations of

traits are affected by habitat loss and fragmentation.

Here, we focus on Lepidoptera because butterflies and,

to a lesser extent, moths provide several advantages.

Especially, butterfly species have been proven to be good

indicators of environmental change and overall biodiversity

(Thomas 2005). Further, they are charismatic and highly

attractive and thus well studied. Therefore, there is

comparably good knowledge about species traits and their

substantial variation available and there are many studies

dealing with questions of habitat loss and fragmentation

(Koh et al. 2004; Thomas & Hanski 2004; Dover & Settele

2009). Some, but not all, of the case studies have

supported predictions of positive relationships between

species richness and connectivity (Hanski & Poyry 2007).

Given the large number of case studies, it is surprising that

no quantitative syntheses have been conducted on the

effects of habitat loss and fragmentation on the diversity

of Lepidoptera in general and on trait-mediated responses

in particular.

In this study, we synthesize data on the responses of

Lepidoptera (butterflies and moths) to habitat loss and

fragmentation to test two more general predictions: (1)

species richness increases with patch area and connectivity,

and (2) species characterized by low dispersal ability

(indicated by small wing span), narrow larval feeding niche

and low reproductive potential have the strongest responses

to decreasing patch area and connectivity. We tested these

predictions with analyses that combined information on life-

history traits with patterns of species occurrence in habitat

patches from 24 independent data sets, covering a wide

range of landscapes and biotopes in Europe and North

America. We demonstrated that species richness responses

to decreasing habitat area and connectivity strongly

depend on species� dispersal capacity, niche breadth and

reproduction.

M E T H O D S

Data sets

We used literature searches and personal contacts to collect

data sets relating species richness of Lepidoptera to habitat

loss and fragmentation. We restricted our search to

community-level studies, where species richness was

observed at the patch level and where at least seven

well-defined habitat patches were systematically sampled

for butterflies, moths or both using standardized sampling

methods (transect counts, observation plots or standard-

ized traps). We excluded small-scale studies with experi-

mentally created arrays of patches where individuals freely

moved among patches (e.g., Summerville & Crist 2001),

and focused on broad-scale studies with larger isolated

patches where spatially structured dynamics occur due to

local extinction and dispersal limitation. We also excluded

studies where information on species incidence was

compiled rather than systematically sampled (Dennis &

Shreeve 1997). We restricted our search for data sets to

Europe and North America, due to uncertainties in

taxonomy and limited availability of data on life-history

970 E. Ockinger et al. Letter


traits for other regions. In most cases, some or all of the

data needed for analysis were unpublished and we

contacted authors for more detailed data. Of the 24 data

sets (20 published and 4 unpublished) fulfilling our criteria,

there were 19 (15 from Europe and 4 from North

America) on butterflies (in some cases these also included

diurnal moths) from transect walks or plot counts and 5

(two from Europe and three from North America) on

nocturnal moths from light traps. A summary of the data

sources is provided in Appendix S1. For further informa-

tion on these studies, we refer the reader to the original

publications. From each data set, we used the observed

identities and richness of species.

Measures of area and connectivity

We used patch areas and connectivity metrics given in the

publications or provided by the authors. Connectivity

metrics differed among data sets, but two types of metrics

dominated: a connectivity index that weights the distances

and sizes of habitat patches in the surroundings of the focal

study patch (Hanski 1999) (N = 11), and the proportion of

similar habitat within a certain radius around the focal

habitat patch, here referred to as buffer-index (N = 8). For

three of these studies (data set numbers 4, 8 and 9 in

Appendix S1), we had access to both metrics.

Even though it may be more biologically relevant to

weight neighbouring patches by their distance to the patch

of interest than to measure the proportion of habitat within

a certain radius relative to the patch of interest, these two

types of indices are often strongly correlated (Moilanen &

Nieminen 2002; Winfree et al. 2005). In the three data sets

where we had access to both indices, they were significantly

correlated (r = 0.52, N = 193, P < 0.001). Assuming that

the two indices are largely interchangeable, we included

either one or the other in analyses as approximate measures

of connectivity. For data sets that provided both indices, we

used the distance-weighted measure. Any systematic differ-

ences in measures of connectivity (e.g., type of matrix or

spatial scale) between studies were accounted for in the

generalized linear mixed-effects models, where we allowed

the effects of connectivity to vary randomly among datasets.

Studies using other isolation indices (N = 4, see Appen-

dix S1) as well as one study lacking connectivity measures

(see Appendix S1) were excluded. Hence, analyses of the

effect of connectivity included 19 of the 24 data sets. In all

analyses, we used log-transformed values of both area and

connectivity to improve linearity.

Life-history traits

Average wing span of female Lepidoptera was obtained

from field guides (see Appendix S2) and used as a proxy

for dispersal capacity. There is good evidence that

morphometrics such as body size can be used as a proxy

for dispersal capacity (Greeleaf et al. 2007). Field studies

have found strong correlations between wing span and

migration rates in Lepidoptera (Nieminen 1996). For the

butterfly species covered in the book by Bink (1992) (all

species found in northwest Europe) and included in our

analyses, there was a significant correlation between

mobility (nine classes) as assessed by expert opinions and

wing span (r = 0.30, P < 0.001, N = 123). Because we

used the number of species sharing a certain trait as the

dependent variable in our analyses, we could not insert

wing span as a continuous variable in the model. Instead,

we classified species as having large or small wing span

based on whether the value was larger or smaller than the

median (32 mm) wing span for all species included in the

analysis.

As a measure of larval feeding niche breadth, we used the

range of plants on which the Lepidoptera species feed as

larvae. Species whose larval feeding niche consists of a

single plant genus were classified as food specialists whereas

species feeding on more than one plant genus were classified

as generalists. Information on larval diet for all butterfly and

moth species included in any of the data sets were extracted

from published sources (see Appendix S2).

For a subset of the butterfly species, we also had data on

maximum egg production (Bink 1992). This included all

butterfly species from studies in northern and western

Europe (Finland, Sweden, Germany, a total of 10 data sets),

and > 90% of species recorded in the studies from

Switzerland and Czech Republic (data sets 5, 6 and 15 in

Appendix S1). Butterflies for which data on egg production

were available were classified as having high reproduction if

the maximum number of eggs produced exceeded 160

(which was the median value for all species included in our

analysis), and as having low reproduction if the maximum

number of eggs produced was 160 or less. There was a

moderately strong correlation between wing span and

reproduction (r = 0.39, P < 0.001, N = 123), but no

difference in wing span (t-test: t1338 = 0.11, P = 0.91) or

egg production (t121 = 0.61, P = 0.54) between specialist

and generalist species.

Statistical analyses

We analysed data from the two sampling methods, transect

walk (including plot counts; sampling butterflies and diurnal

moths) and light traps (sampling nocturnal moths) sepa-

rately. Species detection probabilities differ between the two

methods, and moths are much more diverse than butterflies

in temperate environments (total number of observed moth

species ranged from 214 to 493 across studies compared

with 17–83 observed butterfly species; see Appendix S1).

Letter Habitat fragmentation and life-history traits 971


Therefore, the analyses can be considered as two indepen-

dent tests of the same hypotheses.

The observed species richness was used as the dependent

variable in generalized linear mixed-effects models with

Poisson error distribution and log link function in the

lme4 package (Bates & Mechler 2008) of R software

(R Development Core Team 2008). We built random-effects

models, where slopes and intercepts were allowed to vary

among data sets (Gelman & Hill 2006; Bolker et al. 2009) as

each study differed in range of habitat areas, levels of patch

connectivity, and the species richness of life-history catego-

ries. The random-effects models also account for any

differences in spatial scale or sampling effort among studies

due to sampling frequency or numbers of sampling units

(transects or traps). First, we analysed overall continuous

effects of habitat area and connectivity on species richness

irrespective of traits, and tested if these differed between

continents and major habitat types. We then performed

separate analyses using either habitat area or connectivity as

continuous variables and life-history traits as categorical

predictor variables. All main effects of area, connectivity,

traits, and their interactions were screened both as fixed

effects and random effects of data set. Treatment contrasts

were used for testing differences among the trait states.

Hierarchical model selection was used to select the best-

fitting model by sequentially removing higher-order ran-

dom-effects interactions followed by random main effects.

Model selection was based on the minimum Akaike�sInformation Criterion (AIC). After selecting the combina-

tion of random effects that resulted in the minimum AIC,

hierarchical model selection of fixed effects was conducted

to determine the combination of fixed and random effects

with the minimum AIC (Bolker et al. 2009). We checked for

overdispersion in Poisson regression models using the

dispersion scale factor, which was < 1.5 in all the models

examined; hence, it was unnecessary to use quasipoisson

errors and quasi-AIC (QAIC) to correct for overdispersion

(Bolker et al. 2009). The statistical significance of each fixed

and random effect in the best-fitting model was determined

by likelihood ratio tests on models with and without each

effect (Pinhero & Bates 2000; Bolker et al. 2009).

To test our main hypothesis – life-history traits influence

the relationship between species richness and habitat area

and connectivity – we used likelihood ratio tests to explicitly

compare the explanatory power of our best-fitting models

with those that did not include any interactions between

area or connectivity and life-history traits (Pinhero & Bates

2000; Bolker et al. 2009). The first set of analyses assessed

the effect of patch area on species richness and included

four combinations of classes of wing span and niche

specialization, i.e. (1) generalist species with small wing span,

(2) generalist species with large wing span, (3) specialist

species with small wing span, and (4) specialist species with

large wing span, as explanatory variables. Because connec-

tivity measures were not available for all data sets, we ran a

second set of analyses with the same species trait variables,

but this time included both the effect of patch area and

connectivity on species richness. Finally, we examined the

effect of patch area on species richness with four

combinations of classes of niche specialization and repro-

ductive potential, i.e. (1) generalist species with low

reproduction, (2) generalist species with high reproduction,

(3) specialist species with low reproduction, and (4)

specialist species with high reproduction, as explanatory

variables. This analysis only included European butterfly

species for which we had data on reproduction. Appen-

dix S1 indicates which data sets were included in each

analysis.

R E S U L T S

General patterns

In total, the 24 data sets contained 1485 species and 815

habitat patches. Across all data sets, there was a positive

effect of patch area on species richness (mean slope of the

species–area relationship on a log–log scale = 0.120,

SE = ± 0.028; likelihood ratio test, v2 = 106.3, d.f. = 3,

P < 0.001) and connectivity (mean slope on a log–log

scale = 0.028, SE = ± 0.018, v2 = 30.7, d.f. = 3, P <

0.001) when variables were treated as random effects across

data sets. This pattern differed slightly when butterflies and

moths were analysed separately. For butterflies, the mean

slope (on a log–log scale) of the species–area relationship

was estimated to be 0.124 (SE = ± 0.033, v2 = 85.5,

d.f. = 3, P < 0.001), and the species–connectivity relation-

ship to be 0.045 (SE = ± 0.017, v2 = 14.3, d.f. = 3,

P = 0.026). For moths, area had a mean slope of 0.108

(SE = ± 0.035, v2 = 30.7, d.f. = 3, P < 0.001) and con-

nectivity had an overall slightly negative effect of )0.058

(SE = ± 0.038, v2 = 11.7, d.f. = 3, P < 0.01) on species

richness.

The slope of the relationships between species richness

and area (F1,18 = 2.07, P = 0.15) and connectivity (F1,18 =

0.58, P = 0.45) did not differ between continents, or

between grassland, forest and urban habitats (area: F2,16 =

0.62, P = 0.54; connectivity: F2,16 = 1.25, P = 0.29).

Effects of life-history traits

Life-history traits significantly modified the response of

species richness to habitat patch area and connectivity. For

butterflies, the models including interactions between traits

and patch area had higher explanatory power than compet-

ing models without those interactions (Table 1a–c). For

moths the model which included an interaction between



Table 1 The best-fitting (lowest AIC) generalized linear mixed-effects models for the effects of: (a) wing span (small or large) and niche

breadth (specialist or generalist) on the relationship between butterfly species richness and patch area; (b) wing span and niche breadth on the

relationships between species richness, patch area and connectivity; (c) reproductive potential (high or low) and niche breadth on the

relationship between species richness and patch area; and (d) wing span and niche breadth on the relationships between moth species

richness, patch area, and connectivity. To demonstrate differences among trait states (tests based on treatment contrasts), the signs of the

coefficients are coded in a relational way: large relative to small wingspan, generalist relative to specialist niche breadth, and high relative to

low reproductive potential

Model

(a) Butterflies:

area, wing

span & niche

breadth

(19 data sets)

(b) Butterflies:

area, wing span,

niche breadth &

connectivity

(15 data sets)

(c) Butterflies: area,

niche breadth &

reproductive

potential

(15 data sets)

(d) Moths: area,

wing span, niche

breadth &

connectivity

(4 data sets)

Coefficients of fixed effects

Area )0.110 )0.074 )0.111 )0.104

Connectivity )0.049 +0.009

Large +0.704 +0.745 +0.427

Generalist +1.148 +1.082 +1.206 +2.638

High reproduction +0.186

Area · large +0.010 )0.130 –

Area · generalist )0.775 – )0.056 –

Area · high reproduction )0.016

Large · generalist +0.036 – )0.714

Generalist · high reproduction )0.130

Area · large · generalist )0.095 – –

Area · generalist · high reproduction )0.128

Connectivity · large – +0.063

Connectivity · generalist – –

Connectivity · large · generalist – –

Variances for random effects of data sets

Area 0.010 0.010 0.008 0.004

Connectivity 0.003 0.005

Large 0.387 0.474 0.086

Generalist 0.577 0.280 0.118 0.234

High reproduction 0.312

Area · large 0.007 – –

Area · generalist – 0.017 0.015 –

Area · high reproduction 0.178

Large · generalist 0.393 0.331 –

Generalist · high reproduction –

Area · large · generalist – – –

Area · generalist · high reproduction –

Connectivity · large – –

Connectivity · generalist – –

Connectivity · large · generalist – –

Model fits and hypothesis tests

AIC 2683 2007 1980 403

DAIC vs. no connectivity 8 7

Likelihood ratio test vs. no connectivity v2 = 25.31,

d.f. = 8

P = 0.0014

v2 = 17.1,

d.f. = 5

P = 0.0043

DAIC vs. no area · trait or connectivity

· trait interactions

170 135 135 10



connectivity and wing span explained the data better than

the competing model without that interaction term

(Table 1d). Competing models with DAIC £ 10 compared

to the best model (c.f. Burnham & Anderson 2002) are

presented in Table S1.

For butterflies, we first analysed the effect of patch area

in relation to wing span and niche breadth, including 19

data sets. The full model, including all interactive effects,

had the lowest AIC, but a competing model including only

main effects and the area · niche breadth interaction had

only slightly higher AIC (DAIC = 2) and marginally lower

explanatory power (likelihood ratio test: v2 = 7.76,

d.f. = 3, P = 0.0513; Table 1a). Patch area had a greater

positive effect on the richness of small-winged than on

large-winged species and on richness of specialist than on

generalist species (Fig. 1). The difference between special-

ists and generalists in the effect of area on species richness

tended to be larger among large-winged species (Fig. 1),

but the heterogeneity among datasets was also large

(Fig. 2).

To test for the effect of connectivity, we repeated the

analysis with the 15 butterfly data sets including both

connectivity measures and patch area. The most parsimo-

nious model contained a positive effect of connectivity on

species richness, but no interaction between connectivity

and any individual life-history trait (Table 1b). We then

repeated the analysis of species–area relationships including

data on reproductive potential and niche breadth (15

datasets for which we had data on reproductive potential).

The model with the lowest AIC included interactive effects

of area and niche breadth, area and reproductive potential,

as well as a three-way interaction between area, niche

breadth and reproductive potential (Table 1c; Fig. 3). The

slope of the relationship between species richness and area

Table 1 continued

Model

(a) Butterflies:

area, wing

span & niche

breadth

(19 data sets)

(b) Butterflies:

area, wing span,

niche breadth &

connectivity

(15 data sets)

(c) Butterflies: area,

niche breadth &

reproductive

potential

(15 data sets)

(d) Moths: area,

wing span, niche

breadth &

connectivity

(4 data sets)

Likelihood ratio test vs. no area · trait

or connectivity · trait interactions

v2 = 200.1,

d.f. = 15

P < 0.0001

v2 = 50.8,

d.f. = 9

P < 0.0001

v2 = 47.0

d.f. = 9

P < 0.0001

v2 = 11.4,

d.f. = 1

P = 0.0007

AIC, Akaike�s Information Criterion.

Variables tested but not retained in the best-fitting model are denoted with �–�. Area and connectivity are log-transformed. The model

coefficients are for the fixed effects, and variances are for the random effects of dataset included in each best-fitting model. Random

effects of area, connectivity, and traits varied significantly (likelihood ratio tests, P < 0.05) among datasets (see Figs 1–3). The DAIC

values and likelihood ratio tests compare best-fitting models to those without connectivity, area · trait interactions, or connectivity · trait

interactions.

Inte

rcep

t

–10

12

3

Large Small Large Small

–0.2

0.0

0.2

0.4

0.6

Slo

pe

Generalists Specialists

Figure 1 Intercepts and slopes from the best [lowest Akaike�sInformation Criterion (AIC)] generalized linear mixed-effects

models testing the effect of wing span (large or small) and niche

breadth (specialist or generalist) on the response of species

richness to habitat patch area. The box plots refer to the

variability among 19 European and North American butterfly

datasets. A large intercept indicates high species richness within a

particular trait group; slope indicates the magnitude of change in

species richness with increasing patch area within a particular trait

group.



is steeper for specialists than for generalists, and is steeper

for species with low reproductive potential than for species

with high reproductive potential. The effect of reproductive

potential, however, is largest among specialists and partic-

ularly for specialists with low reproductive potential, which

have a steeper species–area relationship than the other

groups (Fig. 3).

Nocturnal moths were analysed in the same manner as

butterflies but constraints in data availability restricted the

analysis to comparisons of wing span and niche breadth.

The model with the best fit included patch area, connec-

tivity, wing span, niche breadth and an interactive effect of

connectivity and wing span. Surprisingly, the effect of

connectivity on species richness was slightly negative in

some data sets, indicating higher species richness in

more isolated habitat patches. This effect was especially

pronounced for the richness of species with small wing

span, indicating a low mobility. But because there was a

significant negative correlation (r = )0.51) between patch

area and connectivity in these data sets (i.e., the isolated

patches tended to be large), patch area is probably the key

factor driving this pattern.

Fixed effects model EUR Czech Republic EUR Finland

EUR Germany EUR Sweden EUR Switzerland

NA Iowa NA Missouri

–4 –2 0 2 4 6

–10

12

34

–4 –2 0 2 4 6

–10

12

34

–4 –2 0 2 4 6–1

01

23

4

–4 –2 0 2 4 6

–10

12

34

–4 –2 0 2 4 6

–10

12

34

–4 –2 0 2 4 6

–10

12

34

–4 –2 0 2 4 6

–10

12

34

–4 –2 0 2 4 6

–10

12

34

–4 –2 0 2 4 6

–10

12

34

NA Ohio

Log area (ha)

Log

spec

ies

richn

ess

Large generalists Small generalists

Large specialistsSmall specialists

Figure 2 The fixed effects model (without the random effects of data set) of richness and log area for wing span (small or large) and niche

specialization (specialist or generalist) for 19 butterfly datasets, and the random-effects models for a subset of 8 of the 19 datasets,

demonstrating the variability in the relationships among species richness, log area, and traits among datasets. Subsets were selected to illustrate

a large geographical extent and range of variability among datasets.



D I S C U S S I O N

Our quantitative synthesis demonstrated consistent and

significant effects of habitat patch area and connectivity on

butterfly and moth species richness across two continents

and a wide range of habitat types. More importantly, we

demonstrated that life-history traits associated with dispersal

capacity, niche breadth and reproduction modify the

response of species richness to habitat area and fragmen-

tation. As a consequence, both species richness and

composition change with habitat loss and fragmentation

(McKinney & Lockwood 1999; Bommarco et al. 2010).

Specifically, our results show that with decreasing patch size,

butterfly communities will be increasingly dominated by

mobile generalists, whereas species with low dispersal

power, narrow feeding niches, and low reproductive rates

will become rarer. An increasing dominance of mobile

generalist over sedentary specialist herbivores could have

profound implications for ecosystem function (Schweiger

et al. 2010) and lead to cascading effects on higher trophic

levels (Dunn et al. 2009).

Surprisingly, habitat fragmentation studies often do not

support the predicted relationships between habitat area or

connectivity and species richness (see reviews by Debinski

& Holt 2000; Ewers & Didham 2006; Hanski & Poyry

2007). Our results suggest that variability in species

responses may be due to their life-history traits. The

explanatory power of models relating species richness to

habitat area and connectivity increased strikingly when

accounting for life-history traits associated with dispersal

capacity, niche breadth, and reproduction. Communities

inhabiting highly fragmented landscapes, which tend to be

largely dominated by highly mobile generalist species with

high reproductive rates, may respond less dramatically to

further habitat loss and fragmentation compared to com-

munities with a more even distribution of species traits.

For butterflies, our synthesis confirmed the predictions

that resource specialists, short-winged species (reflecting low

mobility), and species with low reproduction are more

sensitive to habitat loss and fragmentation than resource

generalists and mobile species with high reproductive rates.

For moths, the negative effect of connectivity was more

pronounced for species with small wing span.

Typically, generalist species have both a wider range of

available resources within a focal patch and are more likely

to find alternative resources outside the habitat patch, which

makes them less susceptible to decreased patch size.

Previous studies have found a relationship between niche

breadth and extinction risk, both in Lepidoptera (Kotiaho

et al. 2005) and in other taxa (Goulson et al. 2005; Boyles &

Storm 2007). Our results suggest that the strong negative

effect of habitat loss on specialist species can be an

important driver behind this pattern.

The effect of reproductive potential on responsiveness to

habitat loss was more pronounced among specialists than

generalists. Resource specialists are more severely affected

than generalists by habitat loss, but specialists with high

reproductive potential appear to compensate for this, at

least to some extent. Populations of species with high

reproductive potential can recover more rapidly after habitat

perturbations, which can be important in small habitat

patches where environmental stochasticity is likely to be

high (Hanski 1999; Henle et al. 2004). Species with high

reproduction can also produce more colonizers, implying

higher chances of re-colonization following local population

extinctions.

The steeper slope of the species–area relationship for

butterflies with a smaller wing span compared to long-

winged species is similar to what Bommarco et al. (2010)

observed for generalist bees, and is best explained by their

Inte

rcep

t–1

01

23

High Low High Low

–0.2

0.0

0.2

0.4

0.6

Slo

pe

Generalists Specialists

Figure 3 Intercepts and slopes from the best [lowest Akaike�sInformation Criterion (AIC)] generalized linear mixed-effects

models testing the effect of reproductive potential (high or low)

and niche breadth (specialist or generalist) on the response of

species richness to habitat patch area. The boxplots refer to the

variability among 15 European butterfly datasets. Large intercepts

indicate higher richness among groups with different traits; slopes

indicate differences in rate of change in richness with patch area

among groups with different traits.



lower mobility (Nieminen 1996). Poor dispersal capacity

(Lomolino 1984), leading to stronger effective isolation, is

generally predicted to increase the slope of the species–area

relationship (Connor & McCoy 1979). Small habitat patches

can typically harbour only small populations, which are

prone to local extinction. Populations of mobile species,

however, can persist due to immigration from other

populations (Brown & Kodric-Brown 1977) and also have

a higher chance to re-colonize habitat patches following

local extinction (Hanski 1999).

The large variability in both the direction and the strength

of the effect of wing span among data sets could partly be

explained by a nonlinear effect of dispersal capacity on

population persistence (Thomas 2000) and complex trade-

offs with other traits not included in our analysis. Further,

wing span does not only reflect dispersal capacity, but is also

correlated with other life-history traits associated with body

size. Large-bodied species with high energetic needs

generally have lower population densities and thereby are

more sensitive to habitat loss compared to small-bodied

species (Prugh et al. 2008).

Surprisingly, life-history traits modified the response of

species richness to habitat area but not to connectivity.

Several individual studies have found life-history traits, such

as life span (Lindborg 2007), mobility (Ockinger et al. 2009)

and dispersal mode (Tremlova & Munzbergova 2007), to

influence species� response to connectivity. In contrast, the

meta-analysis by Prugh et al. (2008) did not show interactive

effects of life-history traits and connectivity across species.

The potential effects of species traits can be masked, if there

are high levels of heterogeneity in the effects of connectivity

among data sets due to the landscape matrix (Brotons et al.

2003) or fragmentation history (Kuussaari et al. 2009). The

negative response of nocturnal moths to increasing con-

nectivity can, to a large extent, be explained by the negative

correlation between area and connectivity in the moth data

sets. In the studies we examined, isolated habitat patches

had more species because these patches were large. Over-

sampling of species from matrix habitats by light traps in

highly fragmented landscapes (Summerville & Crist 2004),

or a negative correlation between habitat quality and

connectivity, are other potential mechanisms that could

contribute to this unexpected pattern.

The observed differences between butterflies and moths

could represent differences in the life histories between the

two groups. The most obvious difference is that moths are

nocturnal and mainly rely on olfactory cues to navigate

through the landscape, whereas butterflies are assumed to

use visual cues to a much larger extent (c.f. Hamback et al.

2007). But because butterflies were mainly sampled in

grassland habitats and moths mainly in forest habitats, and

by means of different methods (transect and plot counts vs.

traps) in the studies included in our analysis, we cannot

separate the effect of habitat type from that between diurnal

and nocturnal Lepidoptera or between sampling methods.

Species with different traits could differ in detectability. This

could be a potential problem in some studies, but only if the

species characteristics that influence detectability depend on

habitat area or connectivity. And even if such a bias exists

for some of our studies, our main conclusions would only

be affected when this bias was similar across all studies. We

find this quite unlikely and the large variance among the

intercepts and slopes of the trait-habitat relations (Fig. 2),

even having different signs, indicates that there is no such

systematic bias.

The variability in the impact of species traits on their

response to fragmentation (Fig. 2) highlights the importance

of meta-analyses and quantitative syntheses to understand

the consequences of global change. By including primary

data from multiple studies from different geographical

regions and habitat types on two continents, we can make

general predictions about species distributions in changing

landscapes. In our synthesis, we demonstrate that even

though there are some general patterns in species� responses

to habitat loss and fragmentation, these are largely contin-

gent upon life-history traits associated with dispersal, niche

breadth and reproduction. As a result, the remnant

communities in highly fragmented landscapes will be

dominated by a few mobile, generalist species. Because

mobile generalists are also more likely to be able to cope

with the ongoing climate change (Schweiger et al. 2010),

the combined pressures of habitat fragmentation and

climate change (Travis 2003) are likely to lead to increa-

sing global biotic homogenization (McKinney & Lockwood

1999) on one hand, and will have serious consequences

for the function and resilience of ecosystems (Potts et al.

2010).

A C K N O W L E D G E M E N T S

We thank K.-O. Bergman, M. Franzen, T. Kadlec &

M. Konvicka, M. Nieminen, J. Roland, B. Schmid &

A. Swengel for kindly giving us access to unpublished data.

We also thank Lyubomir Penev for organizing the work-

shop that formed the basis of this work. This study was

funded by the 6th framework EU-project �COCONUT –

Understanding effects of land use changes on ecosystems to

halt loss of biodiversity� (SSPI-CT-2006-044346).

R E F E R E N C E S

Bates, D. & Mechler, M. (2008). Linear Mixed Effects Models

Using S4 Classes, Ver 0.999375-27. University of Wisconsin,

Madison. Available at: http://lme4.r-forge.r-project.org.

Bink, F.A. (1992). Ecologische Atlas van de Dagvlinders van Noordwest-

Europa. Schuyt & Co Uitgevers en Importeurs, Haarlem.



Bolker, B.M., Brooks, M.E., Clark, C.J., Geange, S.W., Poulsen,

J.R., Stevens, M.H.H. et al. (2009). Generalized linear mixed

models: a practical guide for ecology and evolution. Trends Ecol.

Evol., 24, 127–135.

Bommarco, R., Biesmeijer, J.J., Meyer, B., Potts, S.G., Poyry, J.,

Roberts, S.P.M. et al. (2010). Dispersal capacity and diet breadth

modify the response of wild bees to habitat loss. Proc. R. Soc. B,

in press.

Boyles, J.G. & Storm, J.J. (2007). The perils of picky eating: dietary

breadth is related to extinction risk in insectivorous bats. PLoS

ONE, 2, e672.

Brotons, L., Monkkonen, M. & Martin, J.L. (2003). Are fragments

islands? Landscape context and density-area relationships in

boreal forest birds Am. Nat., 162, 343–357.

Brown, J.H. & Kodric-Brown, A. (1977). Turnover rates in insular

biogeography: effect of immigration on extinction. Ecology, 58,

445–449.

Burnham, K.P. & Anderson, D.R. 2002. Model Selection and Multi-

Model Inference. Springer Verlag, Berlin.

Connor, E.F. & McCoy, E.D. (1979). The statistics and biology of

the species-area relationship. Am. Nat., 133, 791–833.

Debinski, D.M. & Holt, R.D. (2000). A survey and overview of

habitat fragmentation experiments. Conserv. Biol., 14, 342–355.

Dennis, R.L.H. & Shreeve, T.G. (1997). Diversity of butterflies on

British islands: ecological influences underlying the roles of area,

isolation and the size of the faunal source. Biol. J. Linn. Soc., 60,

257–275.

Dormann, C.F., Schweiger, O., Augenstein, I., Bailey, D., Billeter,

R., de Blust, G. et al. (2007). Effects of landscape structure and

land-use intensity on similarity of plant and animal communities.

Global Ecol. Biogeogr., 16, 774–787.

Dover, J. & Settele, J. (2009). The influences of landscape structure

on butterfly distribution and movement: a review. J. Insect Con-

serv., 13, 3–27.

Dunn, R.R., Nyeema, C.H., Colwell, R.K., Koh, L.P. & Sodhi, N.S.

(2009). The sixth mass coextinction: are most endangered

species parasites and mutualists? Proc. R. Soc. B., 276, 3037–

3045.

Ekroos, J., Heliola, J. & Kuussaari, M. (2010). Homogenization of

lepidopteran communities in intensively cultivated agricultural

landscapes. J. Appl. Ecol., 47, 459–467.

Ewers, R.M. & Didham, R.K. (2006). Confounding factors in the

detection of species responses to habitat fragmentation. Biol.

Rev., 81, 117–142.

Fahrig, L. (2003). Effects of habitat fragmentation on biodiversity.

Annu. Rev. Ecol. Evol. Syst., 34, 487–515.

Gelman, A. & Hill, J. (2006). Data Analysis Using Regression and

Multilevel ⁄ Hierarchical Models. Cambridge University Press,

Cambridge.

Goulson, D., Hanley, M.E., Darvill, B., Ellis, J.S. & Knight,

M.E. (2005). Causes of rarity in bumblebees. Biol. Conserv., 122,

1–8.

Greeleaf, S.S., Williams, N.M., Winfree, R. & Kremen, C. (2007).

Bee foraging ranges and their relationship to body size. Oecologia,

153, 589–596.

Hamback, P.A., Summerville, K.S., Steffan-Dewenter, I., Krauss, J.,

Englund, G. & Crist, T.O. (2007). Habitat specialization, body

size, and family identity explain Lepidopteran density–area

relationships in a cross-continental comparison. Proc. Natl. Acad.

Sci. USA, 104, 8368–8373.

Hanski, I. (1999). Metapopulation Ecology. Oxford University Press,

Oxford.

Hanski, I. & Ovaskainen, O. (2000). The metapopulation capacity

of a fragmented landscape. Nature, 404, 755–758.

Hanski, I. & Poyry, J. (2007). Insect populations in fragmented

habitats. In: Insect Conservation Biology (eds Stewart, A.J.A., New,

T.R. & Lewis, O.T.). CABI, Wallingford, pp. 175–202.

Henle, K., Davies, K.F., Kleyer, M., Margules, C. & Settele, J.

(2004). Predictors of species sensitivity to fragmentation. Bio-

divers. Conserv., 13, 207–251.

Koh, L.-P., Sodhi, N.S. & Brook, B.W. (2004). Ecological corre-

lates of extinction proneness in tropical butterflies. Conserv. Biol.,

18, 1571–1578.

Kotiaho, J.S., Kaitala, V., Komonen, A. & Paivinen, J. (2005).

Predicting the risk of extinction from shared ecological charac-

teristics. Proc. Natl. Acad. Sci. USA, 102, 1963–1967.

Kuussaari, M, Bommarco, R., Heikkinen, R.K., Helm, A., Krauss,

J., Lindborg, R. et al. (2009). Extinction debt: a challenge for

biodiversity conservation. Trends Ecol. Evol., 24, 564–571.

Lindborg, R. (2007). Evaluating the distribution of plant life-

history traits in relation to current and historical landscape

configurations. J. Ecol., 95, 555–564.

Lomolino, M.V. (1984). Mammalian island biogeography: effects of

area, isolation and vagility. Oecologia, 61, 376–382.

MacArthur, R.H. & Wilson, E.O. (1967). The Theory of Island Bio-

geography. Princeton University Press, Princeton.

Mattila, N., Kaitala, V., Komonen, A., Kotiaho, J.S. & Paivinen, J.

(2006). Ecological determinants of distribution decline and risk

of extinction in moths. Conserv. Biol., 20, 1161–1168.

McKinney, M.L. & Lockwood, J.L. (1999). Biotic homogenization:

a few winners replacing many losers in the next mass extinction.

Trends Ecol. Evol., 14, 450–453.

Moilanen, A. & Nieminen, M. (2002). Simple connectivity mea-

sures in spatial ecology. Ecology, 83, 1131–1145.

Nieminen, M. (1996). Migration of moth species in a network of

small islands. Oecologia, 108, 643–651.

Ockinger, E., Franzen, M., Rundlof, M. & Smith, H.G. (2009).

Mobility-dependent effects on species richness in fragmented

landscapes. Basic Appl. Ecol., 10, 573–578.

Pinhero, J. & Bates, D.M. (2000). Mixed-Effects Models in S and

S-PLUS. Springer Verlag, New York.

Potts, S.G., Biesmeijer, J.C., Kremen, C., Neumann, P., Schweiger,

O. & Kunin, W.E. (2010). Global pollinator declines: drivers and

impacts. Trends Ecol. Evol., in press, DOI: 10.1016/j.tree.

2010.01.007.

Prugh, L.R., Hodges, K.E., Sinclair, A.R.E. & Brashares, J.S.

(2008). Effect of habitat area and isolation on fragmented

animal populations. Proc. Natl. Acad. Sci. USA, 105, 20770–

20775.

R Development Core Team (2008). The R foundation for Statistical

Computing, ver 2.8. Vienna University of Technology, Vienna.

Available at: http://www.r-project.org/.

Schweiger, O., Biesmeijer, J.C., Bommarco, R., Hickler, T., Hulme,

P.E., Klotz, S. et al. (2010). Multiple stressors on biotic inter-

actions: how climate change and alien species interact to affect

pollination. Biol. Rev., in press. DOI:10.1111/j.1469-185X.2010.

00125.x.

Summerville, K.S. & Crist, T.O. (2001). Effects of experimental

habitat fragmentation on patch use by butterflies and skippers

(Lepidoptera). Ecology, 82, 1360–1370.



Summerville, K.S. & Crist, T.O. (2004). Contrasting effects of

habitat quantity and quality on moth communities in fragmented

landscapes. Ecography, 27, 3–12.

Thomas, C.D. (2000). Dispersal and extinction in fragmented

landscapes. Proc. R. Soc. Lond. B, 267, 139–145.

Thomas, J.A. (2005). Monitoring change in the abundance and

distribution of insects using butterflies and other indicator

groups. Phil. Trans. R. Soc. B., 360, 339–357.

Thomas, C.D. & Hanski, I. (2004). Metapopulation dynamics in

changing environments: butterfly responses to habitat and cli-

mate change. In: Ecology, Genetics and Evolution of Metapopulations

(eds Hanski, I. & Gaggiotti, O.E.). Elsevier Academic Press,

Amsterdam, pp. 489–514.

Travis, J.M.J. (2003). Climate change and habitat destruction: a

deadly anthropogenic cocktail. Proc. R. Soc. Lond. B, 270, 467–

473.

Tremlova, K. & Munzbergova, Z. (2007). Importance of species

traits for species distributions in fragmented landscapes. Ecology,

88, 965–977.

Winfree, R., Dushoff, J., Crone, E.E., Schultz, C.B., Budny, R.V.,

Williams, N.M. et al. (2005). Testing simple indices of habitat

proximity. Am. Nat., 165, 707–717.

With, K.A. & Crist, T.O. (1995). Critical thresholds in species�responses to landscape structure. Ecology, 76, 2446–2459.

S U P P O R T I N G I N F O R M A T I O N

Additional Supporting Information may be found in the

online version of this article:

Table S1 Best-fitting and alternative generalized linear

mixed-effects models.

Appendix S1 Studies included in the analyses.

Appendix S2 References to sources of information on life-

history traits.

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

supporting information supplied by the authors. Such

materials 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.

Editor, Fangliang He

Manuscript received 5 January 2010

First decision made 12 February 2010

Manuscript accepted 29 March 2010



Life‐history traits predict species responses to habitat area and isolation: a cross‐continental synthesis

Documents