Conservation planning in a changing world
181
bioclimatic envelope analysis (e.g. McLachlan et al. ,
2005 ). As well as occurring at a fi ne spatial scale,
below the resolution of most bioclimatic modelling
exercises (Pearson, 2006 ), such effects appear to be
highly idiosyncratic and would be hard to predict.
While there are clearly many assumptions underlying predictive distribution models, meta - analyses of
large numbers of studies have revealed changes in
species ’ distributions that are broadly concordant with
recent climate change (Fitter & Fitter, 2002 ; Parmesan
& Yohe, 2003 ; Root et al. , 2003 ). Such ecological
changes are yet to translate into widespread extinctions, although numerous studies predict that this will
be an inevitable consequence of continued warming.
For example, a global study by Sekercioglu et al . (2008)
projected that 400 – 550 species of land birds will be
extinct by the year 2100 due to climate change. In
another global study, Jetz et al . (2007) estimated that
900 bird species will show range contractions greater
than 50 per cent by 2100.
Such quantitative projections are effective at garnering newspaper headlines but, given the large number
of assumptions involved, they should be treated with
extreme caution (Ladle, 2009 ; Box 7.3 ). Although
it seems likely that impending climatic changes will
lead to an accelerating wave of future extinctions
(Pimm, 2008 ), at the present time there are too many
A second important group of assumptions are those
surrounding important biotic interactions, such as
predation and site - specifi c competition. These further
limit species ’ distributions in ways that are not simply
predictable from measures of environmental suitability
(Davis et al. , 1998 ). Ara ú jo & Luoto (2007) incorporated the interaction between the European butterfl y
the clouded Apollo ( Parnassius mnemosyne ) and the
presence/absence of four of its larval food plants,
Corydalis spp. The results supported the proposition
that both the explanatory and predictive power of
climate - based species distribution models could be signifi cantly improved when the current and modelled
future distribution of food plants was incorporated.
Many predictions also assume that evolutionary
responses to changing climates will be slow relative to
the fl ux in environmental conditions. While a fair
amount of evidence suggests that local evolutionary
change is unlikely to mitigate negative threats to
species (Parmesan, 2006 ), a recent review of data on
evolutionary response in the face of directional environmental change has challenged the validity of this
assumption (Skelly et al. , 2007 ). Once again, this
appears an important area for future research.
Finally, the persistence of species at low densities in
small areas within regions of more generally unsuitable climate is diffi cult to model using traditional
Box 7.3 Predicting g lobal e xtinctions with s pecies d istribution m odels
The most commonly used method of forecasting climate induced range changes and extinction is
a family of models known as species distribution models. A basic species distribution model has
three components:
First, the climate and habitat within the observed geographical distribution of a species are analysed
statistically. This produces a unique bioclimatic envelope (also known as ‘ climate space ’ ) which
represents the physical conditions that allow that species to fl ourish. Second, the ability of the species
to reach new habitats (dispersal) is quantifi ed. Third, one or more climate change scenarios are chosen
as the basis for forecasting the geographical distribution of the species ’ future ‘ climate space ’ .
Typically, a set of high, medium and low impact (change) scenarios are chosen and applied to
one or two signifi cant points in the future. These points are typically ‘ round number ’ years such as
2050 or 2100.
This type of model was used to forecast global extinctions under climate change in a paper by
Thomas et al . (2004) that garnered global media coverage and, at the time of writing, had been cited
over 800 times in ISI Web of Science. Thomas and his colleagues used projections of species ’
distributions (a selection of vertebrates, invertebrates and plants) for future climate scenarios to
assess extinction risks for sample regions that cover some 20 per cent of the Earth ’ s terrestrial
surface. They then used three methods to estimate extinction, based on the species – area relationship to assess probability of extinction:
181
bioclimatic envelope analysis (e.g. McLachlan et al. ,
2005 ). As well as occurring at a fi ne spatial scale,
below the resolution of most bioclimatic modelling
exercises (Pearson, 2006 ), such effects appear to be
highly idiosyncratic and would be hard to predict.
While there are clearly many assumptions underlying predictive distribution models, meta - analyses of
large numbers of studies have revealed changes in
species ’ distributions that are broadly concordant with
recent climate change (Fitter & Fitter, 2002 ; Parmesan
& Yohe, 2003 ; Root et al. , 2003 ). Such ecological
changes are yet to translate into widespread extinctions, although numerous studies predict that this will
be an inevitable consequence of continued warming.
For example, a global study by Sekercioglu et al . (2008)
projected that 400 – 550 species of land birds will be
extinct by the year 2100 due to climate change. In
another global study, Jetz et al . (2007) estimated that
900 bird species will show range contractions greater
than 50 per cent by 2100.
Such quantitative projections are effective at garnering newspaper headlines but, given the large number
of assumptions involved, they should be treated with
extreme caution (Ladle, 2009 ; Box 7.3 ). Although
it seems likely that impending climatic changes will
lead to an accelerating wave of future extinctions
(Pimm, 2008 ), at the present time there are too many
A second important group of assumptions are those
surrounding important biotic interactions, such as
predation and site - specifi c competition. These further
limit species ’ distributions in ways that are not simply
predictable from measures of environmental suitability
(Davis et al. , 1998 ). Ara ú jo & Luoto (2007) incorporated the interaction between the European butterfl y
the clouded Apollo ( Parnassius mnemosyne ) and the
presence/absence of four of its larval food plants,
Corydalis spp. The results supported the proposition
that both the explanatory and predictive power of
climate - based species distribution models could be signifi cantly improved when the current and modelled
future distribution of food plants was incorporated.
Many predictions also assume that evolutionary
responses to changing climates will be slow relative to
the fl ux in environmental conditions. While a fair
amount of evidence suggests that local evolutionary
change is unlikely to mitigate negative threats to
species (Parmesan, 2006 ), a recent review of data on
evolutionary response in the face of directional environmental change has challenged the validity of this
assumption (Skelly et al. , 2007 ). Once again, this
appears an important area for future research.
Finally, the persistence of species at low densities in
small areas within regions of more generally unsuitable climate is diffi cult to model using traditional
Box 7.3 Predicting g lobal e xtinctions with s pecies d istribution m odels
The most commonly used method of forecasting climate induced range changes and extinction is
a family of models known as species distribution models. A basic species distribution model has
three components:
First, the climate and habitat within the observed geographical distribution of a species are analysed
statistically. This produces a unique bioclimatic envelope (also known as ‘ climate space ’ ) which
represents the physical conditions that allow that species to fl ourish. Second, the ability of the species
to reach new habitats (dispersal) is quantifi ed. Third, one or more climate change scenarios are chosen
as the basis for forecasting the geographical distribution of the species ’ future ‘ climate space ’ .
Typically, a set of high, medium and low impact (change) scenarios are chosen and applied to
one or two signifi cant points in the future. These points are typically ‘ round number ’ years such as
2050 or 2100.
This type of model was used to forecast global extinctions under climate change in a paper by
Thomas et al . (2004) that garnered global media coverage and, at the time of writing, had been cited
over 800 times in ISI Web of Science. Thomas and his colleagues used projections of species ’
distributions (a selection of vertebrates, invertebrates and plants) for future climate scenarios to
assess extinction risks for sample regions that cover some 20 per cent of the Earth ’ s terrestrial
surface. They then used three methods to estimate extinction, based on the species – area relationship to assess probability of extinction: