RESEARCH REVIEW

© 2003 Blackwell Publishing Ltd. http://www.blackwellpublishing.com/journals/geb

Global Ecology & Biogeography

(2003)

12

, 361–371

Blackwell Publishing Ltd.
Predicting the impacts of climate change on the
distribution of species: are bioclimate envelope models
useful?

RICHARD G. PEARSON* and TERENCE P. DAWSON

Environmental Change Institute, School of Geography and the Environment, University of Oxford, 1 A Mansfield Road, Oxford OX1 3SZ, U.K.

E-mail: richard.pearson@eci.ox.ac.uk

ABSTRACT

Modelling strategies for predicting the potential impacts of
climate change on the natural distribution of species have
often focused on the characterization of a species’ bioclimate
envelope. A number of recent critiques have questioned
the validity of this approach by pointing to the many factors
other than climate that play an important part in determin-
ing species distributions and the dynamics of distribution
changes. Such factors include biotic interactions, evolu-
tionary change and dispersal ability. This paper reviews and
evaluates criticisms of bioclimate envelope models and dis-
cusses the implications of these criticisms for the different
modelling strategies employed. It is proposed that, although
the complexity of the natural system presents fundamental
limits to predictive modelling, the bioclimate envelope
approach can provide a useful first approximation as to the
potentially dramatic impact of climate change on biodiversity.
However, it is stressed that the spatial scale at which these
models are applied is of fundamental importance, and that
model results should not be interpreted without due consider-
ation of the limitations involved. A hierarchical modelling
framework is proposed through which some of these limita-
tions can be addressed within a broader, scale-dependent
context.

Key words

bioclimate envelope, climate change, climate
space, ecological modelling, ecological niche, hierarchy, scale.

INTRODUCTION

It is a central premise of biogeography that climate exerts a
dominant control over the natural distribution of species.
Evidence from the fossil record (Woodward, 1987; Huntley,
1999; Davis & Shaw, 2001) and from recently observed
trends (for reviews see Hughes, 2000; McCarty, 2001;
Walther

et al

., 2002) shows that changing climate has a pro-
found influence on species’ range expansion and contraction.
It is therefore expected that predicted future climate change
(IPCC, 2001) will have a significant impact on the distribu-
tion of species.
A number of modelling strategies for predicting the poten-
tial impacts of climate change on biodiversity have been
developed. These have often focused on the identification of a
species’ ‘bioclimate envelope’ (alternatively termed ‘climate
space’, Box, 1981; Huntley

et al

., 1995; Carey, 1996; Bakkenes

et al

., 2002; Berry

et al

., 2002; Pearson

et al

., 2002) either
through techniques that correlate current species distribu-
tions with climate variables or through an understanding of
species’ physiological responses to climate change (Franklin,
1995; Mack, 1996; Guisan & Zimmermann, 2000). Having
identified a species’ climate envelope, the application of
scenarios of future climate change enables the potential redis-
tribution of the species’ climate space to be estimated.
Recent studies have questioned the validity of the biocli-
mate envelope approach by pointing to the many factors
other than climate that play an important part in determining
species distributions and their dynamics over time. Notably,
Davis

et al

. (1998a,b) in their paper entitled ‘Making mis-
takes when predicting shifts in species range in response to
global warming’ identify the importance of biotic interactions
between species, arguing that bioclimate envelope-based
models are flawed. Similarly, Lawton (2000) notes a number
of factors, including the importance of species dispersal,
which may lead to erroneous results from bioclimatic models,
whilst Woodward & Beerling (1997) suggest that such models
should be disregarded and replaced by dynamic vegetation
models.

* Corresponding author.

362

R. G. Pearson and T. P. Dawson

© 2003 Blackwell Publishing Ltd,

Global Ecology & Biogeography

,

12

, 361–371

It is the aim of this paper to review, clarify and evaluate
criticisms of bioclimate envelope models and to discuss their
implications for the different predictive modelling approaches
used. The paper presents concepts that may be familiar to
many ecologists, but which we think it may be useful to col-
late and re-evaluate in the light of recent critiques. The biocli-
mate envelopes approach is first explained and placed within
the context of ecological niche theory. The main criticisms of
the approach are then discussed in turn, using selected exam-
ples from the literature. It is demonstrated that the complex-
ity of the natural system presents fundamental limits to
modelling strategies, making predictive errors inevitable.
However, it is also demonstrated that for certain species and
at certain scales, the bioclimate envelopes approach can pro-
vide useful results, giving a first approximation as to the
potentially dramatic impact of climate change on distribu-
tions. The implications of this critique for different predictive
modelling techniques employed (correlative vs. physiologi-
cally based methodologies) are discussed and the advantages
and disadvantages of each method clarified. The importance
of spatial scale is stressed and a modelling framework is
proposed whereby a hierarchy of factors is considered to
influence the distribution of species across a range of spatial
scales. This framework places bioclimatic models within a
broader, scale-dependent context. It is concluded that, in light
of the great complexity of natural systems, the bioclimate
envelopes approach can provide a valuable means of explor-
ing key characteristics of complex species-environment rela-
tionships. However, such models should only be applied, and
their results interpreted, with a thorough understanding of
the limitations involved.

BIOCLIMATIC MODELLING

The bioclimate envelope modelling approach has its foundations
in ecological niche theory. Hutchinson (1957) defined the funda-
mental ecological niche as comprising those environmental
conditions within which a species can survive and grow. Hut-
chinson proposed that the fundamental niche would completely
define the ecological properties of a species: a conceptual space
whose axes include all of the environmental variables affecting
that species (Austin

et al

., 1990; Leibold, 1995). Bioclimate
envelopes can be defined as constituting the climatic com-
ponent of the fundamental ecological niche, or the ‘climatic
niche’. Thus, bioclimatic models in their purest form consider
only climatic variables and do not include in their processing
other environmental factors that influence the distribution of
species, such as soil type and land-cover type. The definition
of a bioclimate envelope, as with Hutchinson’s definition of
the fundamental ecological niche, also does not include the
influence of biotic effects such as competition for resources.
The distinction between biotic and abiotic limitations on a
species’ distribution can be formalized in the distinction
between

fundamental

and

realized

niches (Hutchinson,
1957). The term

realized

niche describes the case whereby a
species is excluded from parts of its

fundamental

niche
because of competition and other biotic interactions (Austin

et al

., 1990; Guisan & Zimmermann, 2000). This distinction
between fundamental and realized niches is important in the
context of bioclimatic modelling, particularly with regard to
the methodologies used to characterize bioclimate envelopes.
Some bioclimatic models are based on empirical relationships
between observed species distributions and environmental
variables (e.g. Huntley

et al

., 1995; Peterson

et al

., 2001;
Bakkenes

et al

., 2002; Pearson

et al

., 2002). These models
correlate climate variables with observed distributions,
adopting the general thesis that the best indicator of a species’
climatic requirements is its current distribution. Such correla-
tive models thus characterize biocimatic envelopes based on
the

realized

niche, since observed species’ distributions are, in
reality, constrained by nonclimatic factors, including biotic
interactions. Other bioclimatic models look for a more physi-
ologically based mechanistic relationship between climate para-
meters and species response (e.g. Woodward, 1987; Prentice

et al

., 1992; Haxeltine & Prentice, 1996; Sykes

et al

., 1996).
These models aim to identify the

fundamental

niche by
modelling physiological limiting mechanisms in a species’
climatic requirements.
Early examples of modelling strategies using correlations
between climate and observed species distributions include
the work of Johnston (1924, cited in Mack, 1996), who pre-
dicted the invasive spread of prickly pear cactus in Australia
based on the climatic characteristics of the species’ home
range in North America, Hintikka (1963; cited in Hengeveld,
1990), who discriminated between the climates of locations
inside and outside the ranges of some European species based
on the variables minimum and maximum temperature. More
recently, Huntley

et al

. (1995) fitted climate response surfaces
for eight species of European higher plants by locally
weighted regression based on three climatic variables. The
fitted response surfaces were used to simulate potential future
distributions based on scenarios of climate warming.
A further example of the correlative approach is the SPE-
CIES model (Pearson

et al

., 2002). SPECIES (Spatial Evalua-
tor of Climate Impacts on the Envelope of Species) employs
an artificial neural network (ANN) to characterize bioclimate
envelopes based on observed species distributions and five
environmental inputs (derived primarily from climatic data,
but including a measure of soil-type). Application of the
model to a number of European higher plant species has
enabled predictions of the future redistribution of suitable
climate space under scenarios of climate change to be made
(Fig. 1). The environmental inputs used are thought to have
direct physiological roles in limiting the ability of plants to sur-
vive and are derived from primary climate and soils data in a
climate-hydrological process model. Thus, though fundamentally

Evaluating bioclimate envelope models

363
© 2003 Blackwell Publishing Ltd,

Global Ecology & Biogeography

,

12

, 361–371

correlative, the SPECIES model incorporates physiologically
based understanding of species’ ecology. An important
element of the SPECIES methodology is that training of the
ANN is carried out at the European scale so as to include
the full bioclimate envelope of the species being studied. The
trained network is then downscaled in its application to Great
Britain, ensuring that when applied to future climate scenarios
the model is not used to extrapolate outside its training
data range.
In contrast to models of an essentially correlative nature,
models derived from physiological considerations provide a
mechanistic basis for determining climatic limits on species
distributions (Woodward, 1987). It is the thesis of this
approach that models based on mechanistic considerations
will be more robust under changed climatic conditions than
those based on correlations between observed distributions
and current climate variables. The weakness identified in the
latter approach is that such correlations may not apply in the
future as conditions, especially for interspecies relationships,
change (Prentice & Solomon, 1991; Woodward & Rochefort,
1991; Prentice

et al

., 1992). Physiologically based models
have often focused on patterns of global biomes (Prentice

et al

., 1992; Haxeltine & Prentice, 1996). For example, Prentice

et al

. (1992) present a predictive global biome model where
the environmental limits of each plant type are assigned
based on independent physiological data and physiological
reasoning. The resulting predictions of global vegetation
patterns were found to be in good agreement with the observed
distributions, except where intensive agriculture has masked
the natural patterns.
There are fundamental limitations to the predictive capa-
city of bioclimatic models, regardless of the methodology used
to characterize the bioclimate envelope. Three of the main
criticisms of the bioclimatic approach (biotic interactions,
evolutionary change and species dispersal) are briefly
reviewed in the next section, prior to a more detailed assess-
ment of the strengths and weaknesses of the correlative and
physiologically based modelling strategies.

CRITICISMS OF BIOCLIMATIC MODELLING
Biotic interactions

An important distinction exists between how a species would
function on its own and how it actually does in the presence
of other plants and natural ‘enemies’ (Leibold, 1995; Crawley,
1997). Davis

et al

. (1998a) identified inter–species interactions
as the ‘flaw’ in bioclimate envelope modelling approaches.
They used simple microcosm experiments on assemblages of
three fruitfly species (

Drosophilia melanogaster

,

D. simulans

,

D. subobscura

) and a parasitoid wasp (

Leptopiliana boulardi

)
to demonstrate the impact of competitive interactions on
species distributions. It was shown that inter–species
interactions in experimental clines markedly altered the
Fig. 1 Simulated redistribution of suitable climate space for stiff sedge (Carex bigelowii) under future climate scenarios in Great Britain and
Ireland as predicted by the SPECIES model (Pearson et al., 2002). Climate change scenarios are those of Hulme & Jenkins (1998). Suitable climate
space is expected to be lost, with a general migration northwards as the climate changes.

364

R. G. Pearson and T. P. Dawson

© 2003 Blackwell Publishing Ltd,

Global Ecology & Biogeography

,

12

, 361–371

distributions of all three fruitfly species from those found
in single-species clines. The importance of such interactions
for predictive modelling is that if species respond individu-
ally to climate change, as is suggested by the fossil record
(Woodward, 1987; Huntley, 1999; Davis & Shaw, 2001),
the current system of species interactions will change in
future, making predictions from bioclimate envelope models
erroneous.
The fact that competition, predation and symbiosis with
other species influence a species’ distribution was recognized
long before the experiments of Davis

et al

. (1998a, 1998b).
For example, Connell (1961) studied the factors that limit the
range of a species of barnacle (

Chthamalus stellatus

) in the
intertidal zone and showed that the lower edge of the range
was set by interactions with other intertidal species, notably
competition with another barnacle (

Balanus balanoides

) and
predation by a snail (

Thais lapillus

). Similarly, Silander &
Antonovics (1982) found complex responses when experi-
mentally removing one species at a time from a salt-marsh
community and observing the reactions of the others. Results
showed that removal of one grass species (

Muhlenbergia
capillaris

) led to equal range expansions by five other plants,
whereas removing a sedge (

Fimbristylis spadicea

) resulted
in the expansion of only one other plant (the grass

Spartina
patens

).
Biotic interactions are thus shown to have important
impacts on species distributions. We are left with a view of
the natural system as a complex web of interactions and feed-
backs between species, whereby changes to the distribution of
a single species could have significant knock-on impacts on
the distributions of many other species. It is thus apparent
that modelling strategies based on bioclimate envelopes alone
may in some cases lead to predicted distributions that are, in
fact, wildly incorrect. However, it is argued that applying bio-
climatic models at macro-scales, where climatic influences on
species distributions are shown to be dominant, can minimize
the impact of biotic interactions. Indeed, the fact that a
number of bioclimatic models have been highly successful at
simulating current species distributions at certain scales is in
fundamental disagreement with the proposition that species
distributions cannot be adequately defined by climatic factors
alone.
For example, Pearson

et al

. (2002) found good agreement
between observed and simulated European-scale distributions
for 32 plant species based on correlations between observed
distributions and 5 climatic inputs (Fig. 2). Similarly, Beerling

et al

. (1995) tested the predictive capacity of climate response
surfaces for the distribution of Japanese knotweed (

Fallopia
japonica

) and concluded that the close fit between observed
and simulated distributions suggests that the species’
European distribution is climatically determined. It is thus
suggested that bioclimatic models applied at the macro-scale
are suitable for making broad predictions as to the likely
impacts of climate change on the distribution of species. The
experimental scale studied by Davis

et al

. (1998a, 1998b),
which led to their conclusion that bioclimate envelope studies
are of limited use, is far from that proposed for bioclimatic
studies.
Fig. 2 Observed European distribution of hard-fern (Blechnum spicant) alongside the distribution as simulated by the SPECIES model (Pearson
et al., 2002). Presented as an example of the good agreement achievable between observed and simulated European-scale distributions using a
bioclimatic model.

Evaluating bioclimate envelope models

365
© 2003 Blackwell Publishing Ltd,

Global Ecology & Biogeography

,

12

, 361–371

Evolutionary change

Genetic adaptation of species is rarely considered in the liter-
ature on the biotic effects of past and potential future climate
change, with range shifts frequently seen as the expected
response. It is usually expected that evolutionary change
occurs only on long time scales and that the tolerance range
of a species remains the same as it shifts its geographical
range. However, studies have shown that climate-induced
range shifts can involve not only migration into newly suit-
able areas, but also selection against phenotypes that are poor
dispersers or are poorly adapted to local conditions (Davis &
Shaw, 2001).
The potential importance of rapid evolutionary change has
been demonstrated by Thomas

et al

. (2001) who examined
insect species that have expanded their geographical ranges in
Britain over the past 20 years. Two species of bush cricket
(

Conocephalus discolor

and

Metrioptera roeselii

) were shown
to have increased fractions of longer-winged (more disper-
sive) individuals in recently founded populations, whilst two
butterfly species (

Hesperia comma

and

Aricia agestis

) have
increased the variety of habitat types that they can colonize.
Furthermore, it has been shown that rapid evolutionary
change is not confined to the range margins of highly disper-
sive species. Woodward (1990) has shown the potential for
rapid

in-situ

adaptation in plants during a long-term experi-
ment whereby populations of navelwort (

Umbilicus rupestris

)
were transplanted beyond the natural geographical limit of
the species. The species was found to evolve rapidly new low-
temperature responses of seed germination and winter sur-
vival, with temperatures that were observed to kill the species
in 1979 endured by about 50% of the transplanted popula-
tion in 1987.
The implications of rapid evolutionary change for biocli-
mate envelope modelling are important since the assumption
of niche conservatism, whereby rates of adaptation are slower
than extinction rates, will be wrong for species experiencing
sufficiently rapid adaptation. However, it should not be inter-
preted from this evidence that all species will show adaptive
responses to climatic change. Indeed, in an experimental
study on a native legume of the American Great Plains
(

Chamaecrista fasciculata

), Etterson & Shaw (2001) con-
cluded that predicted rates of evolutionary response for
plants of this kind are much slower than the predicted rate of
climate change (due to antagonistic genetic correlations
among traits within populations).
Bioclimatic studies of past climate-species distribution rela-
tionships also provide evidence that adaptation to future cli-
mates will not occur for some species. Huntley

et al

. (1989)
fitted climate response surfaces to beech (

Fagus

spp.) distribu-
tions in Europe and eastern North America. They were able
to simulate distribution patterns in Europe during the Holo-
cene using the response surface derived for North America,
and vice versa. This suggests that the North American and
European beech populations have retained similar climatic
tolerances since their separation between 25 and 10 My ago,
supporting the hypothesis that, for this species, fundamental
physiological limitations have been unaffected by evolutionary
processes over this long timescale.
Predicting adaptive changes to species in response to cli-
mate change presents a huge challenge to vegetation model-
lers and has not, to date, been accounted for within the
current bioclimatic modelling framework. It is thus apparent
that applications of bioclimate envelope models for predict-
ing distribution changes over the next century are most
appropriate for species not expected to be able to undergo
rapid evolutionary change over this timescale. This is most
likely to be the case for long-lived species and poor dispersers,
since intergenerational selection and/or selection at expand-
ing range margins is required for evolutionary processes to
take effect.

Species dispersal

Sufficiently mobile species can be expected to track the geo-
graphical position of their bioclimate envelope through dis-
persal (Graham & Grimm, 1990; Collingham

et al

., 1996).
However, the ability of a species to migrate at a sufficient rate
to keep up with the changing climate will be dependent on the
dispersal characteristics of individual species, with future
migration rates required to be at least equal to those of the
early postglacial period (Collingham & Huntley, 2000). Bio-
climate envelope models do not account for species dispersal,
but instead aim to predict the

potential

range of organisms
under changed climate. Though there is great potential to
couple bioclimate envelope models and dispersal simulations
(Carey, 1996; Peterson

et al

., 2001), it is apparent that cur-
rent predictions of potential distributions may differ greatly
from actual future distributions due to migration limitations.
The ability to migrate is a function not only of individual
species’ characteristics, but also the structure of the landscape
over which dispersal is occurring, including the presence of
natural barriers (such as mountain ranges) or the artificial
fragmentation of habitats (through, for example, the growth
of urban areas or deforestation). It can thus be expected that
in many areas of the world, where artificial landscape frag-
mentation prevails and land-uses are changing rapidly, species
will be unable to migrate at a sufficient rate to keep pace with
the changing climate. In such cases predictions of future dis-
tributions derived from bioclimatic models will be erroneous.
It is apparent then that accurate predictions of the future dis-
tribution of species will require detailed knowledge of the
ability of species to migrate through dynamic heterogeneous
landscapes within the constraint of changing bioclimate enve-
lopes. This will be the case for all but the most disperse or
sedentary species: it may be assumed that highly dispersive

366

R. G. Pearson and T. P. Dawson

© 2003 Blackwell Publishing Ltd,

Global Ecology & Biogeography

,

12

, 361–371

species will be able to fill all potential future climate space,
whilst extremely poor dispersers will occupy only those cur-
rent distributional areas that remain suitable under future cli-
mates (Peterson

et al

., 2001).
Restrictions to species’ dispersal thus present an important
limitation to the bioclimatic modelling approach. However,
there is evidence from the palaeoecological record that cli-
mate change is, in fact, sufficient to explain continental-scale
patterns of plant migrations. Indeed, the success of global
bioclimate reconstruction models in simulating Holocene
changes in distribution for some species (Huntley

et al

., 1989;
Prentice & Solomon, 1991) refutes the hypothesis that organ-
isms were unable to migrate fast enough to allow their range
limits to track climatic changes. It is expected that the ability
of species to migrate rapidly across large distances is driven
primarily by rare long-distance dispersal events (Clark

et al

.,
1998). Indeed, recent studies have highlighted the extreme
importance of long-distance dispersal events, with illustrative
examples drawn not only from palaeoecology but also from
contemporary observations of island colonization and alien
plant spread (Higgins & Richardson, 1999; Cain

et al

.,
2000). In such cases, the assumption that species are able to
migrate to occupy their suitable climate space may not be so
unrealistic.
Evidence in disagreement with the migrational lag hypo-
thesis is also provided by Johnson & Webb (1989) who exam-
ined factors determining the rates of migration of Fagaceous
trees in eastern North America. It was concluded that since
each species has the same potential dispersal rate, observed
differences in their rates of migration during the Holocene are
best accounted for by individual responses to climatic forc-
ings during this period. It is thus again apparent that when
applied at an appropriate scale bioclimate envelopes have the
potential to describe changes in species’ distributions. How-
ever, we should note that although simulations of past
climate-biota relationships provide an important test of the
performance of bioclimatic models, such studies have
coarse resolutions in both space (300–400 km) and time
(3000 years) (Davis & Shaw, 2001). These resolutions do not
match the detail required to study subcontinental changes
over the next 50–100 years.

CORRELATIVE VERSUS PHYSIOLOGICALLY-
BASED MODELS

An important criticism of the correlative approach to biocli-
matic modelling is that species distributions as we observe
them today may not be in equilibrium with the current cli-
mate, nor indeed are they necessarily determined primarily by
climate. Notably, the effects on species distributions of biotic
interactions, physical barriers to dispersal and human manage-
ment demonstrate that the realized niches used in correla-
tive bioclimate envelope methodologies may not represent
absolute limits to species ranges and that therefore future dis-
tributions may show very different realized niches.
Woodward (1990) presents an example of where the
present-day distribution of a species is not in equilibrium with
present-day climate. Careful study of small leaved lime (

Tilia
cordata

), including extensive palaeo-reconstruction by pollen
analysis (Pigott & Huntley, 1981; Huntley & Birks, 1983),
has shown that the species reached its northern limit in the
British Isles in the period between 7000 and 5000

BP

. The
present-day distribution extends to this same northern limit.
However, estimations by Woodward (1990) suggest that the
present-day reproductive limit of the species is about 200 km
south of the northern limit. It is concluded then that the
northern limit of this species is a relic from past climates,
made possible largely by the longevity of the species.
A further example of nonequilibrium between species’ dis-
tributions and current climate is provided by Peterson

et al

.
(1999) who demonstrated the impact of dispersal barriers on
the distribution of species in Central America. In their study
of the conservation of ecological niches over evolutionary
time scales, a correlative ecological niche model (incorporat-
ing inputs of mean annual temperature and precipitation) was
successfully used to predict species’ distributions based on the
ecological characteristics of sister taxa (pairs of birds, mam-
mals and butterflies). It was thus demonstrated that the eco-
logical niches of related species, which are geographically
differentiated (by the Isthmus of Tehuantepec in southern
Mexico), have been conserved in evolutionary time and are
therefore similar. Although this supports the basic assump-
tion of bioclimatic modelling that niches are conservative
over time, it is evident that the presence of a physical barrier
to dispersal means that species are unable to occupy their full
climatic niche. Correlating current climate with the observed
species distribution will therefore not identify the full

poten-
tial

climatic range of the species. Results from correlative
niche models that project future distributions under climate
change scenarios in such situations (e.g. Peterson

et al

., 2002)
should therefore be interpreted with caution.
It has, however, already been shown that a number of cor-
relative bioclimate envelope models have been successfully
used to simulate the distributions of higher plants in Europe
(Beerling

et al

., 1995; Huntley

et al

., 1995; Pearson

et al

.,
2002). These modelling results support the hypothesis that
continental-scale distributions are principally determined by
climate. It is thus suggested that many species distributions
can in fact be considered to be in equilibrium with the current
climate at the macro-scale. However, this is not the case for
all species. For example, attempts to model the European-
scale distribution of yew (

Taxus baccata

) using the SPECIES
model (Pearson

et al

., 2002) yielded relatively low agreement
between actual and simulated distributions (Fig. 3). It is
apparent that although the model identified the broad distri-
bution trends, the finer details of the distribution were not

Evaluating bioclimate envelope models

367
© 2003 Blackwell Publishing Ltd,

Global Ecology & Biogeography

,

12

, 361–371

identified. Such inaccuracies can be expected for species’ dis-
tributions that show extensive fragmentation (for nonclimatic
reasons) within a species’ climate space. In the case of yew, it
is expected that human exploitation has been sufficient to
disrupt the large-scale equilibrium between distribution and
climate (Plaisance, 1979).
Physiologically based bioclimatic models do not make the
assumption of equilibrium and are not dependent upon iden-
tifying a relationship between the current distribution and cli-
mate for characterizing the bioclimate envelope of a species.
It may thus be argued that in basing the model on physiolog-
ical limits to a species’ climatic tolerance, the bioclimate
envelope identified will better represent a species’ absolute
climatic limits than that identified through the correlative
approach. However, such models are equally limited in their
inability to account for nonclimatic influences and have a
number of important limitations that may lead to model
inaccuracies.
For example, in simulating the European distribution of
yew using a bioclimatic model based on physiological con-
straints to growth and regeneration, Sykes

et al

. (1996) were
unable to simulate the fine structure of the distribution. As
with the correlative approach (Fig. 3), they were only able to
simulate the broad envelope of the species’ range (the

poten-
tial

species range, or

fundamental

niche). Furthermore, the
results of Sykes

et al

. (1996) suggest that there are areas, such
as central France, which fall within the climatic range of yew
but where the species is not found for what are presumed to
be nonclimatic reasons. In contrast, the results obtained
through the correlative approach suggest that there are in fact
climatic reasons why the species does not occur in some of
these regions (notably, central France).
Limitations applying to the physiologically based approach
can be summarized as follows. Firstly, it is an obvious point of
definition that fundamental niches are not realized, and nor
will they be realized in the future. Thus, predicted future spe-
cies distributions based on the physiologically determined
fundamental niche are unlikely to be as accurate as those
based on correlations between the observed distribution and
the current realized niche. Secondly, there is increasing evi-
dence that the concept of undifferentiated species comprising
individuals with broad tolerances is not correct (Davis &
Shaw, 2001). Intra-species variation makes it impossible to
define precise limits to a species’ climatic tolerance since there
is no guarantee that the limits for one subpopulation at one
range margin will be exactly the same as those for another
subpopulation at another margin many miles away. Further-
more, the potential importance of rapid evolutionary change
as climatic conditions change (Woodward, 1990; Thomas

et al

., 2001) means that some species’ climatic tolerances may
alter in the future, making the fundamental niche unstable
over time.
It is apparent that there are limitations to both correlative
and physiologically based bioclimatic modelling methodolo-
gies. Though it has been proposed that physiologically based
approaches are superior (Woodward, 1987; Prentice

et al

.,
1992; Sykes

et al

., 1996), it is argued here that such models
also have important limitations and that when applied at
Fig. 3 Observed European distribution of yew (Taxus baccata) alongside the distribution as simulated by the SPECIES model (Pearson et al.,
2002). Although the broad distribution trends are identified in the simulated distribution, the finer details of the distribution are not captured.

368

R. G. Pearson and T. P. Dawson

© 2003 Blackwell Publishing Ltd,

Global Ecology & Biogeography

,

12

, 361–371
appropriate scales and to appropriate species correlative tech-
niques can give equally informative results. Since correlative
techniques do not require detailed physiological data about
individual species, they also have the advantage that they can
easily be applied to a large number of species. This enables
conclusions regarding the potential impacts of climate change
on a wide range of species, and thus habitat assemblages, to
be made (e.g. Berry et al., 2002).
A HIERARCHICAL MODELLING
FRAMEWORK
It is proposed that identifying a species’ suitable climate space
through the use of bioclimate envelope models should form
an important first step in a broader modelling framework. A
useful framework for addressing the environment-biota rela-
tionship is that of a hierarchy of factors operating at different
scales. Thus, at the continental scale, climate can be consid-
ered the dominant factor, whilst at more local scales factors
including topography and land-cover type become increas-
ingly important. Further down the hierarchy, if conditions at
higher levels are satisfied, factors including biotic interactions
and microclimate may become significant. Thus, the distribu-
tion of a species in Europe may be primarily defined by cli-
matic tolerances if the data resolution is 50 km2, whereas as
the resolution is downscaled, to perhaps 5 km2, land-cover
type may become the dominant control over species presence.
Similarly, as the resolution is downscaled to less than 1 km2,
biotic interactions may become important (Fig. 4).
Theories relating to hierarchical structure in ecological sys-
tems have been discussed by, amongst others, Kotliar &
Wiens (1990), Wu & Loucks (1995), Collingham et al.
(2000), Whittaker et al. (2001) and Willis & Whittaker
(2002). Turner et al. (2001) define a hierarchy as being a sys-
tem of interconnections wherein the higher levels constrain
the lower levels to various degrees. These levels operate
across different spatial and temporal scales, with different
processes being more important at different scales (Fig. 5).
Theoretically therefore analyses should be focused at scales at
which the phenomena of interest are dominant (Turner et al.,
2001). Bioclimate envelope modelling fits well into this hier-
archical framework, in identifying large-scale distribution
limitations at the highest, most dominant level.
The application of hierarchy theory in the present context
is supported by the tendency for biotic factors to be more
limiting when physical limiting factors are less severe. This is
characterized at the global scale by physical factors being
more restrictive at higher latitude range margins (where con-
ditions are harsher) and biotic interactions more limiting at
lower latitude range margins (MacArthur, 1972; Brown et al.,
1996; Brown & Lomolino, 1998). Thus, physical (abiotic)
factors such as climate may be considered to act at a higher
level in the hierarchy than biotic factors.
Fig. 5 Schematic example of how different factors may affect the distribution of species across varying spatial scales. Characteristic ‘scale
domains’ are proposed within which certain variables can be identified as having a dominant control over species distributions. Approximate
spatial extents have been assigned to categories of scale based in part on Willis & Whittaker (2002). It is assumed that large spatial extents are
associated with coarse data resolutions, and small extents with fine data resolutions.
Fig. 4 Diagram illustrating a hierarchical modelling framework.
Different factors affecting the distribution of species are considered
to act at different scales. For example, the left-hand section shows
species occurring in cells where both large-scale climatic and smaller-
scale land-cover requirements are met. The right-hand section shows
a downscaled portion from the larger diagram, demonstrating that at
a still finer resolution biotic competition becomes significant.

Evaluating bioclimate envelope models 369
© 2003 Blackwell Publishing Ltd, Global Ecology & Biogeography, 12, 361–371
The hierarchical theory also goes some way to explaining
the differing performance of bioclimate envelope models at
different scales. For example, Pearson et al. (2002) were able
to simulate principal trends in observed plant distributions at
the European scale, but accuracy was reduced when the
model was tested for Great Britain. This reduced accuracy is
likely to be the result of the increasing influence of factors
other than climate, including local land-cover type and inter–
species interactions, which lead to a more fragmented species
distribution at the finer scale.
The proposed hierarchical framework may be imperfect
and over-simplified, yet provides a useful starting point for
approaching the extreme complexity of the natural system.
Identifying appropriate scales of analysis for different environ-
mental drivers, thus validating the scale dependencies outlined
in Fig. 5, should be the focus of further research. However, it
is already apparent that the scale at which current bioclimatic
studies are addressed is of fundamental importance, with
climatic impacts on the distribution of species being most
influential at regional to global scales.
CONCLUSION
A number of criticisms of the bioclimate envelopes approach
have been discussed and inherent limitations of bioclimatic
models based on both correlative and physiologically based
mechanistic methodologies have been demonstrated. The dis-
cussion has also shown the great complexity of natural systems,
suggesting that there are fundamental limits to the accurate
prediction of future species distributions. Combining the com-
plexities arising from issues of biotic interactions, evolution-
ary change and species’ dispersal with the uncertainties all too
evident in predictions of future climate and land-use change,
it is apparent that accurate predictions of biogeographical
responses to future climate change are not currently possible.
The development of dynamic global vegetation models
(DGVMs), which include mechanistic representations of
physiological, biophysical and biogeochemical processes,
has demonstrated significant progress in the modelling of
vegetation–climate interactions at the global scale (Woodward
& Beerling, 1997; Cramer et al., 2001). Recent development of
these techniques for application at regional scales, including
the breaking down of ecosystem processes into key compon-
ents with characteristic spatial and temporal scales in a hier-
archical system analogous to that advocated in this paper,
shows much promise (Sykes et al., 2001). However, the com-
plexity of DGVMs makes their parameterization and valida-
tion problematic, and does not currently allow their
widespread application to specific species and regions. It has
been argued in this paper that relatively simple bioclimate
envelope models can provide a useful starting point when
applied to suitable species and at appropriate spatial scales. In
many cases, bioclimate envelope models provide perhaps the
best available guide for policy making at the current time
(Hannah et al., 2002). They have been usefully employed to
identify possible magnitudes of future changes to distribu-
tions, and to suggest which species, habitats and regions are
most at risk from climate change (Prentice et al., 1992;
Beerling et al., 1995; Huntley et al., 1995; Sykes et al., 1996;
Berry et al., 2001, 2002; Hannah et al., 2002; Midgley et al.,
2002). The importance of bioclimatic model predictions
should thus not be underestimated, though model predictions
should be interpreted with due caution and should be viewed
as first approximations indicating the potential magnitude
and broad pattern of future impacts, rather than as accurate
simulations of future species distributions.
More realistic simulations of the impact of climate change
on species distributions will require a better understanding of
the complex interactions between the many factors affecting
distributions. For example, it will be necessary to use
dynamic models to simulate the relationship between chang-
ing climate space and the potential for species to disperse
through fragmented landscapes, and to further our under-
standing of the complex dynamics of model systems consist-
ing of multiple interacting species. A hierarchical modelling
framework has been proposed through which it will be pos-
sible to integrate such factors, acting at different spatial scales.
ACKNOWLEDGMENTS
The research was funded by the European Community’s Fifth
Framework Programme (ACCELERATES project, contract
EVK2 °CT_ 2000–000610) and by a consortium of nature con-
servation organizations led by English Nature (the MONARCH
project). We thank members of the research teams and steering
groups of these projects for many helpful discussions, and R.
Lampinen of the Finnish Museum of Natural History for provid-
ing species distribution data for these projects. We are especially
grateful to Pam Berry for comments on successive versions of
the manuscript. Our thanks to Arne Anderberg, Anna-Lena
Anderberg and the Swedish Museum of Natural History for
allowing us to use their species photos. We also thank two
anonymous reviewers and M.T. Sykes for helpful comments.
REFERENCES
Austin, M.P., Nicholls, A.O. & Margules, C.R. (1990) Measurement
of the realized qualitative niche: environmental niches of five Euca-
lyptus species. Ecological Monographs, 60, 161–177.
Bakkenes, M., Alkemade, J.R.M., Ihle, F., Leemans, R. & Latour, J.B.
(2002) Assessing effects of forecasted climate change on the diver-
sity and distribution of European higher plants for 2050. Global
Change Biology, 8, 390–407.
Beerling, D.J., Huntley, B. & Bailey, J.P. (1995) Climate and the dis-
tribution of Fallopia japonica: use of an introduced species to test
the predictive capacity of response surfaces. Journal of Vegetation
Science, 6, 269–282.

370 R. G. Pearson and T. P. Dawson
© 2003 Blackwell Publishing Ltd, Global Ecology & Biogeography, 12, 361–371
Berry, P.M., Dawson, T.P., Harrison, P.A. & Pearson, R.G. (2002)
Modelling potential impacts of climate change on the bioclimatic
envelope of species in Britain and Ireland. Global Ecology and Bio-
geography, 11, 453–462.
Berry, P.M., Vanhinsberg, D., Viles, H.A., Harrison, P.A., Pearson, R.G.,
Fuller, R., Butt, N. & Miller, F. (2001) Impacts on terrestrial
environments. Climate change and nature conservation in Britain
and Ireland: modelling natural resourse responses to climate change
(the MONARCH Project) (ed. by P.A. Harrison, P.M. Berry and
T.P. Dawson), pp. 43–149. UKCIP technical report, Oxford. http://
www.ukcip.org.uk/model_nat_res/model_nat_res.html
Box, E.O. (1981) Macroclimate and plant forms: an introduction to
predictive modelling in phytogeography. Junk, The Hague.
Brown, J.H. & Lomolino, M.V. (1998) Biogeography, 2nd edn. Sin-
auer Associates, Sunderland, MA.
Brown, J.H., Stevens, G.C. & Kaufman, D.M. (1996) The geo-
graphic range: size, shape, boundaries, and internal structure.
Annual Review Ecological Systematics, 27, 597–623.
Cain, M.L., Milligan, B.G. & Strand, A.E. (2000) Long-distance dis-
persal in plant populations. American Journal of Botany, 87,
1217–1227.
Carey, P.D. (1996) DISPERSE: a cellular automaton for predicting
the distribution of species in a changed climate. Global Ecology
and Biogeography Letter, 5, 217–226.
Clark, J.S., Fastie, C., Hurtt, G., Jackson, S.T., Johnson, C., King, G.A.,
Lewis, M., Lynch, J., Pacala, S., Prentice, I.C., Schupp, E.W.,
Webb, T. III & Wyckoff, P. (1998) Reid’s paradox of rapid plant
migration: dispersal theory and interpretation of paleoecological
records. Bioscience, 48, 13–24.
Collingham, Y.C., Hill, M.O. & Huntley, B. (1996) The migration of
sessile organisms: a simulation model with measurable parameters.
Journal of Vegetation Science, 7, 831–846.
Collingham, Y.C. & Huntley, B. (2000) Impacts of habitat fragmen-
tation and patch size upon migration rates. Ecological Applica-
tions, 10, 131–144.
Collingham, Y.C., Wadsworth, R.A., Huntley, B. & Hulme, P.E.
(2000) Predicting the spatial distribution of non-indigenous ripar-
ian weeds: issues of spatial scale and extent. Journal of Applied
Ecology, 37, 13–27.
Connell, J.H. (1961) The influence of interspecific competition and
other factors on the distribution of the barnacle Chthamalus stella-
tus. Ecology, 42, 710–723.
Cramer, W., Bondeau, A., Woodward, F.I., Prentice, I.C., Betts, R.A.,
Brovkin, V., Cox, P.M., Fisher, V., Foley, J.A., Friend, A.D.,
Kucharik, C., Lomas, M.R., Ramankutty, N., Sitch, S., Smith, B.,
White, A. & Young-Molling, C. (2001) Global response of terres-
trial ecosystem structure and function to CO2 and climate change:
results from six dynamic global vegetation models. Global Change
Biology, 7, 357–373.
Crawley, M.J. (1997) The structure of plant communities. Plant ecology
(ed. by M.J. Crawley), pp. 475–531. Blackwell Science, Oxford.
Davis, A.J., Jenkinson, L.S., Lawton, J.L., Shorrocks, B. & Wood, S.
(1998a) Making mistakes when predicting shifts in species range in
response to global warming. Nature, 391, 783–786.
Davis, A.J., Lawton, J.L., Shorrocks, B. & Jenkinson, L.S. (1998b)
Individualistic species responses invalidate simple physiological
models of community dynamics under global environmental
change. Journal of Animal Ecology, 67, 600–612.
Davis, M.B. & Shaw, R.G. (2001) Range shifts and adaptive
responses to Quaternary climate change. Science, 292, 673–679.
Etterson, J.R. & Shaw, R.G. (2001) Constraint to adaptive evolution
in response to global warming. Science, 294, 151–154.
Franklin, J. (1995) Predictive vegetation mapping: geographic model-
ling of biospatial patterns in relation to environmental gradients.
Progress in Physical Geography, 19, 474–499.
Graham, R.W. & Grimm, E.C. (1990) Effects of global climate
change on the patterns of terrestrial biological communities.
Trends in Ecology Evolution, 5, 289–292.
Guisan, A. & Zimmermann, N.E. (2000) Predictive habitat distribu-
tion models in ecology. Ecological Modelling, 135, 147–186.
Hannah, L., Midgley, G.F. & Millar, D. (2002) Climate change-
integrated conservation strategies. Global Ecology and Bioge-
ograpgy, 11, 485–495.
Haxeltine, A. & Prentice, I.C. (1996) BIOME3: An equilibrium ter-
restrial biosphere model based on ecophysical constraints, resourse
availability, and competition among plant functional types. Global
Biogeochemical Cycles, 10, 693–709.
Hengeveld, R. (1990) Dynamic biogeography. Cambridge University
Press, Cambridge.
Higgins, K. & Richardson, D.M. (1999) Predicting plant migration
rates in a changing world: the role of long-distance dispersal.
American Naturalist, 153, 464–475.
Hughes, L. (2000) Biological consequences of global warming: is the
signal already apparent? Trends in Ecology and Evolution, 15, 56–
61.
Hulme, M. & Jenkins, G.J. (1998) Climate change scenarios for the
United Kingdom: scientific report. UK Climate Impacts Programme
Technical Report no. 1, Climatic Research Unit, Norwich.
Huntley, B. (1999) Species distribution and environmental change:
considerations from the site to the landscape scale. Ecosystem
management: questions for science and society (ed. by E. Maltby,
M. Holdgate, M. Acreman and A. Weir), pp. 115–130. Royal
Holloway Institute for Environmental Research, Virginia Water,
UK.
Huntley, B., Bartlein, P.J. & Prentice, I.C. (1989) Climatic control of
the distribution and abundance of beech (Fagus L.) in Europe and
North America. Journal of Biogeography, 16, 551–560.
Huntley, B., Berry, P.M., Cramer, W. & Mcdonald, A.P. (1995)
Modelling present and potential future ranges of some European
higher plants using climate response surfaces. Journal of Biogeo-
graphy, 22, 967–1001.
Huntley, B. & Birks, H.J.B. (1983) An atlas of past and present pol-
len maps for Europe: 0–13 000 B.P. Cambridge University Press,
Cambridge.
Hutchinson, G.E. (1957) Concluding remarks. Cold Spring Harbor
Symposium on Quantitative Biology, 22, 415–457.
IPCC (2001) Climate change 2001: the scientific basis. Contribution
of Working Group I to the Third Assessment Report of the Inter-
governmental Panel on Climate Change (ed. by J.T. Houghton, Y.
Ding, D.J. Griggs, M. Noguer, P.J. van der Linden, X. Dai, K.
Maskell and C.A. Johnson). Cambridge University Press, Cam-
bridge, UK.
Jalas, J. & Suominen, J. (1972) Atlas Flora Europaeae: distribution
of vascular plants in Europe, 1. The Committee for Mapping the
Flora of Europe & Societas Biologica Fennica Vanamo, Helsinki.
Jalas, J. & Suominen, J. (1973) Atlas Flora Europaeae: distribution

Evaluating bioclimate envelope models 371
© 2003 Blackwell Publishing Ltd, Global Ecology & Biogeography, 12, 361–371
of vascular plants in Europe, 2. The Committee for Mapping the
Flora of Europe & Societas Biologica Fennica Vanamo, Helsinki.
Johnson, W.C. & Webb, T. III (1989) The role of blue jays (Cyano-
citta cristata) in the postglacial dispersal of fagaceous trees in east-
ern North America. Journal of Biogeography, 16, 561–571.
Kotliar, N.B. & Wiens, J.A. (1990) Multiple scales of patchiness and
patch structure: a hierarchical framework for the study of hetero-
geneity. Oikos, 59, 253–260.
Lawton, J.L. (2000) Concluding remarks: a review of some open
questions. Ecological consequences of heterogeneity (ed. by M.J.
Hutchings, E. John and A.J.A. Stewart), pp. 401–424. Cambridge
University Press, Cambridge.
Leibold, M.A. (1995) The niche concept revisited: mechanistic models
and community context. Ecology, 76, 1371–1382.
MacArthur, R.H. (1972) Geographical ecology: patterns in the dis-
tribution of species. Harper & Row, New York.
Mack, R.N. (1996) Predicting the identity and fate of plant invaders:
emergent and emerging approaches. Biological Conservation, 78,
107–121.
McCarty, J.P. (2001) Ecological consequences of recent climate
change. Conservation Biology, 15, 320–331.
Midgley, G.F., Hannah, L., Millar, D., Rutherford, M.C. & Powerie, L.W.
(2002) Assessing the vulnerability of species richness to anthropo-
genic climate change in a biodiversity hotspot. Global Ecology and
Biogeography, 11, 445–451.
Pearson, R.G., Dawson, T.P., Berry, P.M. & Harrison, P.A. (2002)
SPECIES: a spatial evaluation of climate impact on the envelope of
species. Ecological Modelling, 154, 289–300.
Peterson, A.T., Ortega-Huerta, M.A., Bartley, J., Sanchez-Cordero, V.,
Soberon, J. & Buddemeier, R.W. (2002) Future projections for
Mexican faunas under global climate change scenarios. Nature,
416, 626–629.
Peterson, A.T., Sanchez-Cordero, V., Soberon, J., Bartley, J.,
Buddemeier, R.W. & Navarro-Singuenza, A.G. (2001) Effects of
global climate change on geographic distributions of Mexican
Cracidae. Ecological Modelling, 144, 21–30.
Peterson, A.T., Soberon, J. & Sanchez-Cordero, V. (1999) Conservatism
of ecological niches in evolutionary time. Science, 285, 1265–1267.
Pigott, C.D. & Huntley, J.P. (1981) Factors controlling the distribu-
tion of Tilia cordata at the northern limits of its geographical
range. III. Nature and causes of seed sterility. New Phytologist, 87,
817–839.
Plaisance, G. (1979) L’if. La foret privée — revue forèire
Européenne, 126, 34–47.
Prentice, I.C., Cramer, W., Harrison, S.P., Leemans, R., Monserud, R.A.
& Solomon, A.M. (1992) A global biome model based on plant
physiology and dominance, soil properties and climate. Journal
of Biogeography, 19, 117–134.
Prentice, I.C. & Solomon, A.M. (1991) Vegetation models and global
change. Global changes of the past (ed. by R.S. Bradley), pp. 365–
383. UCAR/Office for Interdisciplinary Earth Studies, Boulder,
CO.
Silander, J.A. & Antonovics, J. (1982) Analysis of interspecific inter-
actions in a coastal plant community — a perturbation approach.
Nature, 298, 557–560.
Sykes, M.T., Prentice, I.C. & Cramer, W. (1996) A bioclimatic model
for the potential distributions of north European tree species under
present and future climates. Journal of Biogeography, 23, 203–
233.
Sykes, M.T., Prentice, I.C., Smith, B., Cramer, W. & Venevsky, S.
(2001) An introduction to the European Terrestrial Ecosystem
Modelling Activity. Global Ecology and Biogeography, 10, 581–
593.
Thomas, C.D., Bodsworth, E.J., Wilson, R.J., Simmons, A.D.,
Davies, Z.G., Musche, M. & Conradt, L. (2001) Ecological and
evolutionary processes at expanding range margins. Nature, 411,
577–581.
Turner, M.G., Gardner, R.H. & O’Neill, R.V. (2001) Landscape
Ecology: in Theory and Practice. Springer-Verlag, New York.
Walther, G.R., Post, E., Convey, P., Menze, 1, A., Parmesan, C.,
Beebee, T.J.C., Fromentin, J.M., Hoegh-Guldberg, O. & Bairlein, F.
(2002) Ecological responses to recent climate change. Nature, 416,
389–395.
Whittaker, R.J., Willis, K.J. & Field, R. (2001) Scale and species rich-
ness: towards a general, hierarchical theory of species diversity.
Journal of Biogeography, 28, 453–470.
Willis, K.J. & Whittaker, R.J. (2002) Species diversity — scale mat-
ters. Science, 295, 1245–1248.
Woodward, F.I. (1987) Climate and plant distribution. Cambridge
University Press, Cambridge.
Woodward, F.I. (1990) The impact of low temperatures in control-
ling the geographical distribution of plants. Philosophical Transac-
tions of the Royal Society of London B, 326, 585–593.
Woodward, F.I. & Beerling, D.J. (1997) The dynamics of vegetation
change: health warnings for equilibrium ‘dodo’ models. Global
Ecology Biogeography Letters, 6, 413–418.
Woodward, F.I. & Rochefort, I. (1991) Sensitivity analysis of vegeta-
tion diversity to environmental change. Global Ecology Biogeo-
graphy Letters, 1, 7–23.
Wu, J. & Loucks, O.L. (1995) From balance of nature to hierarchical
patch dynamics: a paradigm shift in ecology. The Quarterly Review
Biology, 70, 439–466.
BIOSKETCHES
Richard Pearson is a doctoral student with research interests
in biogeography and spatial ecology. Particular interests
include modelling species–climate interactions, the simulation
of dispersal processes in heterogeneous landscapes, and the
role of scale in ecology.
Terence Dawson is a University Research Lecturer and
programme leader of the Terrestrial Ecology and Biodiversity
research group of the Environmental Change Institute,
University of Oxford. His research interests include the
understanding of ecosystem form and functioning at varying
spatial and temporal scales, and the study of complex human–
climate–ecosystem interactions.