180
Planning for persistence in a changing world
2008 ). For example, recent range expansion of the
map butterfl y ( Araschnia levana ), a species whose dispersal ability depends closely on late summer temperature, was poorly predicted using generalized additive
models (Mitikka et al. , 2008 ). Similarly, Ara ú jo et al .
(2005b) used data on the distribution of 116 British
bird species in the 1970s to build climate envelope
model - based predictions of ‘ future ’ distributions in the
1990s, which they were then able to test against the
observed distribution at that time using the 1990s
climate data and the 1970s - based models. In contrast
to the result obtained by Green et al . (2008) , they found
that around 90 per cent of the observed distributions
differed markedly from those predicted. Even in terms
of predicting the directionality of range size change
(expansion versus contraction), the models performed
poorly. This apparent discrepancy of outcomes in relation to two studies of British birds might be explained
by the differing forms of data used (population trends
versus distributional range data).
Predictions of geographical range shifts owing to
climate change depend on a number of assumptions.
Some of the most important are about dispersal, i.e.
how quickly species ’ distributions can track suitable
climate space as it moves geographically. Generally, the
slower dispersal ability is assumed to be, the more dramatic the reduction in occupied area will be, because
geographical ranges will lag further and further behind
climatically suitable areas as they shift across space
(Peterson et al. , 2002 ). For some taxa, such as butterfl ies, there is good evidence that species ’ distributions
are closely tracking recent changes in climate (Wilson
et al. , 2005 ; Hickling et al. , 2006 ). However, this is
unlikely to be the case in the future for all taxa, and
especially for species needing to move across heavily
cultivated or urbanized landscapes.
Successful efforts have recently been made to couple
population models directly with predictions about
changes in distribution under land use and climate
change (Ak ç akaya et al. , 2004 ; Keith et al. , 2008 ). Such
exercises generally reveal that the extent and severity
of climate change impacts on species ’ ranges will vary
with life history characteristics. When spatial and
demographical processes are coupled directly in this
way, oversimplifying assumptions about the relationship between habitat change and extinction (e.g. inferring species loss from species – area relationships) can
be avoided (Buckley & Roughgarden, 2004 ; Thuiller
et al. , 2004 ). This appears to be a promising and
potentially important area for further research efforts.
as the tropics (Section 4.2.2 ; Collen et al. , 2008 ).
Models can help bridge this gap in the data and thereby
improve the quality of planning for conservation
(Loiselle et al. , 2003 ; also see Chapter 6 this volume).
However, as with all complex ecological models, there
are pitfalls to avoid and numerous assumptions that
need to be taken into account when interpreting their
results.
7.3.2 Modelling r ange s hifts
Assuming one can model the current distributions of
species effectively, an important next question is
whether the relationships between environmental conditions and species ’ distributions will continue to hold
into the future. If they do, then extinctions can be predicted to occur in places rendered unsuitable by environmental change, and colonizations where previously
unsuitable conditions become suitable (Figure 7.3 ).
There is no doubt that species ’ geographical ranges,
and the resulting composition of communities, shift
continuously through time. These changes occur
naturally over evolutionary time (Gaston, 2003 ; Vrba
& DeGusta, 2004 ), and climate is arguably the dominant driver of species ’ natural distributions. There is
ample evidence for this both from the fossil record
(Huntley, 1999 ; Davis & Shaw, 2001 ) and from empirical studies correlating species ’ distributions and
changes in those distributions with climate variables
(e.g. Root, 1988 ). This has led to a great deal of interest
in predicting how human - forced climate change will
alter species ’ distributions.
Once a good model of a species ’ current distribution
has been built, change in environmental parameters
(such as those refl ecting a future climate or land - use
scenario) can be simulated, and the model used to generate a prediction of the future distribution of species
in response to any given environmental change.
This general approach assumes that the relationship
between species occurrence and environmental conditions does not markedly change over time. There is
some empirical support for this assumption. For
example, population trends in 42 species of rare birds
breeding in the UK were found to be positively correlated with climate suitability trend, suggesting that
climate envelope models can successfully predict
responses to climate change (Green et al. , 2008 ).
However, such approaches do not always work well
(Davis et al. , 1998 ; Ara ú jo et al. , 2005b ; Beale et al. ,
Planning for persistence in a changing world
2008 ). For example, recent range expansion of the
map butterfl y ( Araschnia levana ), a species whose dispersal ability depends closely on late summer temperature, was poorly predicted using generalized additive
models (Mitikka et al. , 2008 ). Similarly, Ara ú jo et al .
(2005b) used data on the distribution of 116 British
bird species in the 1970s to build climate envelope
model - based predictions of ‘ future ’ distributions in the
1990s, which they were then able to test against the
observed distribution at that time using the 1990s
climate data and the 1970s - based models. In contrast
to the result obtained by Green et al . (2008) , they found
that around 90 per cent of the observed distributions
differed markedly from those predicted. Even in terms
of predicting the directionality of range size change
(expansion versus contraction), the models performed
poorly. This apparent discrepancy of outcomes in relation to two studies of British birds might be explained
by the differing forms of data used (population trends
versus distributional range data).
Predictions of geographical range shifts owing to
climate change depend on a number of assumptions.
Some of the most important are about dispersal, i.e.
how quickly species ’ distributions can track suitable
climate space as it moves geographically. Generally, the
slower dispersal ability is assumed to be, the more dramatic the reduction in occupied area will be, because
geographical ranges will lag further and further behind
climatically suitable areas as they shift across space
(Peterson et al. , 2002 ). For some taxa, such as butterfl ies, there is good evidence that species ’ distributions
are closely tracking recent changes in climate (Wilson
et al. , 2005 ; Hickling et al. , 2006 ). However, this is
unlikely to be the case in the future for all taxa, and
especially for species needing to move across heavily
cultivated or urbanized landscapes.
Successful efforts have recently been made to couple
population models directly with predictions about
changes in distribution under land use and climate
change (Ak ç akaya et al. , 2004 ; Keith et al. , 2008 ). Such
exercises generally reveal that the extent and severity
of climate change impacts on species ’ ranges will vary
with life history characteristics. When spatial and
demographical processes are coupled directly in this
way, oversimplifying assumptions about the relationship between habitat change and extinction (e.g. inferring species loss from species – area relationships) can
be avoided (Buckley & Roughgarden, 2004 ; Thuiller
et al. , 2004 ). This appears to be a promising and
potentially important area for further research efforts.
as the tropics (Section 4.2.2 ; Collen et al. , 2008 ).
Models can help bridge this gap in the data and thereby
improve the quality of planning for conservation
(Loiselle et al. , 2003 ; also see Chapter 6 this volume).
However, as with all complex ecological models, there
are pitfalls to avoid and numerous assumptions that
need to be taken into account when interpreting their
results.
7.3.2 Modelling r ange s hifts
Assuming one can model the current distributions of
species effectively, an important next question is
whether the relationships between environmental conditions and species ’ distributions will continue to hold
into the future. If they do, then extinctions can be predicted to occur in places rendered unsuitable by environmental change, and colonizations where previously
unsuitable conditions become suitable (Figure 7.3 ).
There is no doubt that species ’ geographical ranges,
and the resulting composition of communities, shift
continuously through time. These changes occur
naturally over evolutionary time (Gaston, 2003 ; Vrba
& DeGusta, 2004 ), and climate is arguably the dominant driver of species ’ natural distributions. There is
ample evidence for this both from the fossil record
(Huntley, 1999 ; Davis & Shaw, 2001 ) and from empirical studies correlating species ’ distributions and
changes in those distributions with climate variables
(e.g. Root, 1988 ). This has led to a great deal of interest
in predicting how human - forced climate change will
alter species ’ distributions.
Once a good model of a species ’ current distribution
has been built, change in environmental parameters
(such as those refl ecting a future climate or land - use
scenario) can be simulated, and the model used to generate a prediction of the future distribution of species
in response to any given environmental change.
This general approach assumes that the relationship
between species occurrence and environmental conditions does not markedly change over time. There is
some empirical support for this assumption. For
example, population trends in 42 species of rare birds
breeding in the UK were found to be positively correlated with climate suitability trend, suggesting that
climate envelope models can successfully predict
responses to climate change (Green et al. , 2008 ).
However, such approaches do not always work well
(Davis et al. , 1998 ; Ara ú jo et al. , 2005b ; Beale et al. ,