Impacts of climate change on Swiss biodiversity: An indicator taxa approach
Peter B. Pearmana,⇑, Antoine Guisanb, Niklaus E. Zimmermanna
aLand Use Dynamics, Swiss Federal Research Institute WSL, Zurcherstrasse 111, 8903 Birmensdorf, Switzerland
bDepartment of Ecology and Evolution, Biophore, University of Lausanne, 1015 Lausanne, Switzerland
a r t i c l e i n f o
Received 5 January 2010
Received in revised form 18 November 2010
Accepted 27 November 2010
Available online 22 December 2010
Boosted regression trees
Species distribution models
a b s t r a c t
We present a new indicator taxa approach to the prediction of climate change effects on biodiversity at
the national level in Switzerland. As indicators, we select a set of the most widely distributed species that
account for 95% of geographical variation in sampled species richness of birds, butterflies, and vascular
plants. Species data come from a national program designed to monitor spatial and temporal trends in
species richness. We examine some opportunities and limitations in using these data. We develop eco-
logical niche models for the species as functions of both climate and land cover variables. We project
these models to the future using climate predictions that correspond to two IPCC 3rd assessment scenar-
ios for the development of ‘greenhouse’ gas emissions. We find that models that are calibrated with Swiss
national monitoring data perform well in 10-fold cross-validation, but can fail to capture the hot-dry end
of environmental gradients that constrain some species distributions. Models for indicator species in all
three higher taxa predict that climate change will result in turnover in species composition even where
there is little net change in predicted species richness. Indicator species from high elevations lose most
areas of suitable climate even under the relatively mild B2 scenario. We project some areas to increase in
the number of species for which climate conditions are suitable early in the current century, but these
areas become less suitable for a majority of species by the end of the century. Selection of indicator spe-
cies based on rank prevalence results in a set of models that predict observed species richness better than
a similar set of species selected based on high rank of model AUC values. An indicator species approach
based on selected species that are relatively common may facilitate the use of national monitoring data
for predicting climate change effects on the distribution of biodiversity.
? 2010 Elsevier Ltd. All rights reserved.
Anthropogenic climate warming could result in an average glo-
bal temperature increase of 4.0 ?C by the end of the 21st century
(Meehl et al., 2007). A recent global assessment of the impacts of
climate change suggests that mountain ecosystems will experience
unprecedented rates of warming during the 21st century, two to
three times greater than observed during the 20th century
(Nogués-Bravo et al., 2006). Mountain ecosystems are likely sensi-
tive to global warming owing to the reduction of terrestrial area
with increasing elevation (Guisan et al., 1995; Theurillat et al.,
1998; Diaz et al., 2003; Beniston, 2006). Over the last two decades,
continuing climate warming has been associated with increases in
the northern latitudinal limits of birds in the northern hemisphere
(Thomas and Lennon, 1999), changes in animal movement and
habitat use (Walther et al., 2002) and advancing flowering phenol-
ogy (Fitter and Fitter, 2002). Ongoing changes in temperature and
other climate parameters at high elevations are expected to have
strong effects on plant communities and associated animal assem-
blages (Beniston et al., 1996; Walther, 2003). Impacts of global
warming have appeared in the Alps and include the slow upward
shift of tree line (Gehrig-Fasel et al., 2007) and the ranges of alpine
and nival plants (Grabherr et al., 1994; Pauli et al., 1996, 2007;
Walther et al., 2005; Vittoz et al., 2006). In this study, we address
the continuing effects of climate change on patterns of species
richness in Switzerland using ecological niche models and three
groups of indicator species that represent geographic patterns of
species richness in three divergent higher taxa.
The use of indicator species has seen considerable discussion.
Some studies suggest that species at risk (protected by the Endan-
gered Species Act in the United States or highly rated on the IUCN
Red List) could be associated with patterns of total species richness
(Mikusinski et al., 2001; Lawler et al., 2003; Warman et al., 2004)
and have useful indicator properties (Pearman et al., 2006).
However, rare species can have special habitat requirements that
limit their coincidence with areas of high total species richness
(Prendergast et al., 1999; Chase et al., 2000; Aubry et al., 2005;
Orme et al., 2005). The geographical distribution of species
richness is determined at both continental and regional extents
by the spatial occurrence patterns of species that are widely
0006-3207/$ - see front matter ? 2010 Elsevier Ltd. All rights reserved.
⇑Corresponding author. Tel.: +41 44 739 2524.
E-mail addresses: email@example.com (P.B. Pearman), firstname.lastname@example.org
(A. Guisan), email@example.com (N.E. Zimmermann).
Biological Conservation 144 (2011) 866–875
Contents lists available at ScienceDirect
journal homepage: www.elsevier.com/locate/biocon
distributed (Jetz and Rahbek, 2002; Vazquez and Aizen, 2003;
Lennon et al., 2004; Pearman and Weber, 2007). This suggests that
a subset of widely distributed (i.e., common) species can best
indicate general patterns of species richness over geographic
extents that are pertinent for habitat management, biodiversity
monitoring and prediction of the impacts of climate change. More-
over, while national-level assessment of biodiversity trends is
mandated by multinational agreements that promote biodiversity
conservation (Weber et al., 2004), the use of multiple indicator
taxa in an approach to national-level assessment and prediction
of the impacts of climate change on biodiversity has not been
developed previously. The prediction and assessment of climate
change impacts using common indicator species could be advanta-
geous. This is because the greater area of occurrence of these
species likely provides better coverage within national boundaries,
in comparison with the same number of narrowly distributed or
uncommon species (Pearman and Weber, 2007). Further, focus
on the potential response of species to climate change can provide
estimates of changes in species composition that are not available
when composite measures, such as species richness or diversity
indices, are used without reference to species identity.
In this paper we report research using an indicator species ap-
proach and data from the Swiss Biodiversity Monitoring Program
(BDM) Z7 field program (Plattner et al., 2004) to predict the im-
pacts of climate change on the species richness of birds, butterflies,
and vascular plants in Switzerland. The indicator species that we
examine are widely distributed species that represent most (90%)
of the geographic variation in the richness of all species in their
respective higher taxa. We use ecological niche models (ENMs)
and a variety of climate and land cover data to achieve a compre-
hensive understanding of the potential impact of climate warming
on patterns of species richness in Switzerland. In this study we in-
clude land cover variables to investigate the relative importance of
land cover and climate variability to current species distributions.
We use predictions of future climate that are derived from general
circulation models and the ENMs of the indicator species to predict
future potential distributions of the species and of indicator species
richness. We examine the efficacy of two groups of species as indi-
cators of current patterns of species richness, species for which
models exhibit superior performance (as measured by AUC) and
the models of a group of the most-prevalent species. While change
in species richness is part of the story of the predicted impacts of
climate change, we also investigate turnover in assemblage com-
position and how it develops with time. Our findings suggest that
turnover in assemblages of the indicator species we investigate
will be a notable characteristic of ecological responses to climate
2.1. Species occurrence data
Data on species occurrences for this study come from an exist-
ing database of samples taken on Switzerland’s landscape diversity
(‘Z7’) sampling sites. This biodiversity monitoring program has
been described in detail elsewhere (Plattner et al., 2004; Kéry
and Schmid, 2006; Pearman and Weber, 2007). There are 380 reg-
ularly spaced, 1-km2square cells (quadrats) of the Z7 sample that
are aligned within the approximately 41,295 1 km2units of the
Swiss national coordinate system and are a sample thereof. An
additional 140 quadrats are distributed in southern and western
Switzerland and result in these areas having approximately twice
the prevailing density of samples. These quadrats were excluded
from the present analysis so as to maintain a single density of sam-
ples across the entire country. The establishment and sampling of
this set of 1 km2cells constitutes one of several steps taken to meet
Switzerland’s commitments resulting from the Rio de Janeiro
Convention on Biological Diversity (www.biodiv.org). Unlike many
atlas datasets, the data were collected with a documented
sampling protocol that corresponds to the 1 km2cell size of the
cells. This is particularly advantageous when modeling species
distributions because predictions based on models that are
calibrated with data from regularly spaced samples can far
out-perform models based on data from ad hoc collections where
a design is absent (Edwards et al., 2006). We used occurrence data
to indicate species presence. Species were assumed to be absent at
sites at which they were not observed. While this underestimates
the number of species actually at a site and biases observed species
prevalence downward (Kéry and Schmid, 2006), the structure of
the data (i.e. no repeat visits conducted within a robust design)
do not permit an estimate of probability of detection for species
in all three groups and at all sites.
One fifth of the Z7 sites were sampled each year. These sites
were a regularly spaced subsample of all sites and spanned the
entire extent of the sampling area. A second selection of regularly
spaced sites, not including the first sites and also constituting
one-fifth of the total number of sites, was sampled in the second
year, and so on until by the end of the 5th year all sites had been
sampled. We used data from the first 5 years of sampling for each
of the three taxonomic groups. At each site, the three taxa were
sampled following standard protocols that were specific to each
taxonomic group. The sampling efficiency for each taxon has been
reported elsewhere (Plattner et al., 2004; Kéry and Schmid, 2006).
2.2. Selection of indicator species
Widely distributed species of birds, butterflies, and plants
determine the geographic pattern of species richness in the Z7
BDM data and the richness of most prevalent 30% of species is
highly correlated with overall species richness (Pearman and
Weber, 2007). Our study focuses on these widely distributed
species. Thus we avoid rare species that may be difficult to detect
and contribute little to overall geographic patterns of species
richness (Pearman and Weber, 2007).The relatively large number
of occurrences of the chosen species in the Z7 dataset suggests that
the models constructed for these species are unlikely to be influ-
enced by the vagaries of field sampling to as great a degree as
models of species that are uncommon or, equivalently, have very
few observed occurrences. For the current analysis, we selected
for modeling species that were present in at least 20 sites. We
focus on the most prevalent ?30% of species that, in each taxon,
present a pattern of species richness that is highly correlated
(rp> 0.95) with total species richness of the taxon. We report
results for the modeling of each group of these indicator species,
as well as each species (see online supplementary materials).
2.3. Environmental data
2.3.1. Land cover data
We created a habitat classification to capture variation in land
cover that is likely relevant to the composition of natural commu-
nities and the resulting species richness. At the same time, we at-
tempted to avoid a proliferation of categories, which would result
in a large number of variables and challenge our ability to calibrate
models meaningfully with a data set of only 380 or so samples. We
reviewed land cover categories for the Swiss ‘Arealstatistik 1992/
1997’ land cover summary, which are available within the GEO-
STAT database (Bundesamt für Statistik, www.bfs.admin.ch), to
form seven land cover types (Appendix I). We then used a 100 m
resolution rasterized version of the original lattice map in a geo-
graphic information system (GIS) to evaluate land cover in each
P.B. Pearman et al./Biological Conservation 144 (2011) 866–875
square kilometer grid cell of the Swiss national coordinate system.
We converted these values into a proportion of each land cover
type within each of the 41,295 square kilometers in the Swiss na-
tional grid system.
2.3.2. Climate data and variables
Construction of gridded environmental variables at our working
resolution of 1 km required a number of steps. We assembled a
vector of five climate variables (Table 1), at a resolution of
100 m, that were chosen to be closely related to physiological lim-
its and requirements for metabolism in mountain plants (Körner,
2003), and are similar to variables that have functioned well in
previous attempts to model distributions of mountain plants
(Zimmermann and Kienast, 1999), the geographic distribution of
birds (Tingley et al., 2009) and of butterflies (Lutolf et al., 2009).
These variables were degree-days above 0 ?C, yearly and winter
mean potential evapotranspiration (PET), yearly average precipita-
tion, and moisture index (yearly precipitation divided by yearly
PET). Climate data represented five distinct periods. The current
period was represented by average of monthly values for the per-
iod 1961–1990. We interpolated these climatological normals from
weather station data collected from across Switzerland. For
complete details on the derivation of bioclimatic variables see
Zimmermann and Kienast (1999) and Guisan et al. (2007). For this
study, these variables were averaged within 1 km2grid cells.
We used general circulation model output (below) to represent
climate during the intervals 1991–2020, 2021–2050, 2051–2075
and 2076–2100. We used data that were derived from the Hadley
Center Coupled Model (HadCM3), which was calibrated to project
future climates under the A1FI and B2 scenarios of the Third
Assessment Report of the Intergovermental Panel on Climate
Change (IPCC, 2001) as these were the most recent available at
the time. The climate data were obtained directly from the CRU
website representing the TYN SC 1.0 data set (Mitchell et al.,
2004) and available in a 100(ca. 18 km) resolution. These data were
downscaled in three steps: First, we derived anomalies of future
climate compared to the baseline (1961–1990) on a monthly basis
using the same 100spatial resolution originating from the CRU TS
1.2 data set (Mitchell et al., 2004). Second, the anomalies were
re-projected from the Geographic to the Swiss National coordinate
system and spatially interpolated to a resolution of 1 km using in-
verse distance weighted interpolation. In a third step, the spatial
resolution was further refined to 100 m using bilinear interpola-
tions and the anomalies were added to the 100 m maps of climato-
logical normals, above. For this study, these climate values were
also averaged within 1 km2grid cells.
2.3.3. Topo-environmental variables
We also evaluated five topo-environmental variables (Table 1)
that were chosen to represent relevant local variation at a resolu-
tion of 1 km2, but which do not likely respond directly to climate
change. These variables represent environmental variation that is
determined primarily by topography and soils. They included solar
radiation, average slope, topographic position, soil water holding
capacity and soil coarse fragment content. Potential global solar
radiation was calculated in ArcInfo using a method modified after
Kumar et al. (1997). Average slope was determined by evaluating
the mean inclination in degrees across the one kilometer grid cells.
Topographic position expresses the degree of convexity or concav-
ity around a center cell (Zimmermann et al., 2007), by evaluating a
range of window sizes in a moving window algorithm. Soil water
holding capacity (in mm) expresses the amount of water that soils
can hold. Coarse fragment content expresses the fraction (in %) of
coarse debris that reduces the capacity of soils to hold back water.
Both these variables were taken from the Soil Suitability Map of
the GEOSTAT database of Switzerland (www.bfs.admin.ch, BFS,
1992). These variables were assumed to remain at their original
values during all time periods.
2.4. Ecological niche modeling
2.4.1. Modeling algorithm
We modeled species distributions using an iterative com-
puter learning algorithm called the gradient boosting machine
(Friedman, 2001). We used the algorithm as implemented in the
R statistical package ‘gbm: Generalized Boosted Regression
Models’, available on the R website (http://cran.r-project.org).
Boosted regression trees are becoming increasingly popular in pre-
dictive modelling because of their often superior performance in
prediction (Elith et al., 2006; Guisan et al., 2007). A full description
of gbm and a users guide was recently published (Elith et al., 2008).
We employed the ‘area under the receiver operator characteristic
curve’ (AUC) as a criterion for evaluating the fit of gbm models to
the calibration dataset and in evaluating predictive performance
(Fielding and Bell, 1997). This measure of model fit is suited for
absences because it requires no arbitrarily defined threshold prob-
ability with which to establish prediction of species presence. In
calibrating the model for each species, we generated an estimate
of AUC by conducting 10-fold cross-validation on a calibration
dataset that was selected randomly from the complete dataset
and that conserved the overall proportion of presences/absences
found in the full dataset (Efron and Tibshirani, 1998; Randin
et al., 2006).
We retrieved directly from the gbm model (in fact, an R-object)
the proportional contribution of each climate variable to the mod-
el. We also calculated the marginal influence of the three most
important variables that were associated with the current species
distribution as determined by the gbm algorithm (Ridgeway,
2007). Models built with gbm can ‘overfit’ the data when a large
number (several thousands) of regression trees is combined into
the final model. For this reason, we used the ‘out of bag’ criterion
(Ridgeway, 2007) for determining a conservative estimate of the
number of regression trees upon which the projections of species
distributions should be based. Measures of model performance
were also based on models using the number of regression trees
suggested by application of this criterion.
One problem when predicting species distributions is that niche
models provide projections of species distributions in the form of
Variables, abbreviations, and four alternative model parameterizations. An ‘x’
indicates that the variable is included in the model under the corresponding
numbered parameterization. The variables included in the model under a particular
parameterization exhibited a correlation of rp< 0.7.
Variable (abbrev.) Parameterization
Forest, mature (pc.forest)
Other forest (pc.othfor)
Intensive agriculture (pc.intagr)
Extensive agriculture (pc.extagr)
Impervious surface (pc.imperv)
Other land cover (pc.others)
Degree-days 0 (dd00_av)
PET, yearly average (etyy_av)
PET, winter (etwi_av)
Precipitation, yearly average (pryy_av)
Moisture index (miyy_av)
Solar rad., year, mean (sryy_av)
Slope, average (slop_av)
Soil water holding cap. (swhc_av)
Soil quality (scfc)
P.B. Pearman et al./Biological Conservation 144 (2011) 866–875
probabilities that the species is found within map cells. The making
of a distribution map entails deciding on a threshold probability
value above which the species is predicted to be present. We used
the criterion of the maximum coefficient of agreement (maximum
Kappa, Cohen, 1960) to determine this threshold, based on an anal-
ysis of the performance of the models in correctly predicting the
pattern of species occurrences and absences in the data with which
the models were calibrated.
2.4.2. Model parameterization
The variables one selects to use as explanatory or predictor
variables and whether they enter into the model as linear or
higher-order terms determines the parameterization of the model.
In many modeling exercises, the inclusion of many correlated
variables can result in models that lack fit, predictive ability, and
stability, meaning that the results are sensitive to the inclusion or
omission of a small number of observations. To avoid this, we
grouped variables into three categories: habitat variables, climate
variables, and topo-environmental variables that are determined
only variables derived from climate data are projected to change
variables for correlation using 460 Z7 samples points. In composing
Pearson correlations of less that 0.70. While we eventually consid-
ered seven different model parameterizations, we settled on four
of these to explore in detail (Table 1) in this paper.
The modeling results were analyzed graphically by mapping the
sum total of species that were predicted to be present in each 1-km
map cell. We also calculated two components of species turnover
in order to identify both the time frame and geographic pattern
of change in predicted species assemblage structure at the scale
of 1-km2. We calculated predicted percent gain of species as 100
times the proportion of the species predicted present at a future
time point that were not predicted present during the original time
period. We calculated the predicted percent loss of species as 100
times the proportion of species that were predicted present during
the original time period that were no longer predicted present by
the future time period in question. As in all models of this
type, the validity of projection of the climate-distribution relation-
ship to future climates assumes that climate-distribution relation-
ships remain unchanging over the time period that is considered.
2.4.3. Indicator species evaluation
We ranked species by decreasing magnitude of model AUC and
also by decreasing species prevalence. We then examined the cor-
relation between observed species richness in the field data and
the predicted species richness of groups of increasing numbers of
species by including successive species in order of their rank. We
determined modeled species richness that resulted from projecting
the models of the species in each group to the current climate. We
then determined the correlation between the predicted species
richness values and the observed total species richness for each
taxon across the sampled sites.
3.1. Model performance
Models for species in each group are, on average, good (Tables
A2–A4 in online supplementary materials; Swets, 1988). The mod-
els for bird species fit the original Z7 data somewhat better than
the models for plants or butterflies, as shown by the average values
of 10-fold cross-validated AUC. For all three groups, the average
model AUC (±std. dev.) indicates that the models are useful (Araújo
et al., 2005; birds: 0.84 ± 0.10; butterflies: 0.81 ± 0.10; vascular
plants: 0.82 ± 0.09). Nonetheless, there is a wide range of AUC val-
ues, despite the tight distribution around the median value (Fig. 1).
Modeled species richness demonstrates a higher correlation
with the species richness in the observed data when species are
collected into groups in order of rank prevalence (Fig. 2) than when
collected into groups in order of model performance (AUC). In the
case of birds, the modeled richness of only very few of the most-
prevalent species (10–20) closely approximates species richness
of all avian species in the observed data (Fig. 2b). We observe
similar but less striking patterns for modeled species of the
most-prevalent butterflies and plants (Fig. 2).
3.2. Variable influence
For birds, the best models for most species are most-strongly
influenced by habitat variables, which we did not modify as part
of climate change scenarios. (Table 2, and Table A2 in supplemen-
tary online materials). Percent mature forest is the most influential
variable in models for 15 bird species while percent intensive
agriculture is the most influential variable in models of another
five species. Among the variables that describe climate and express
climate change, degree-days is the most influential variable in best
models for seven bird species and yearly potential evapotranspira-
tion is the most influential variable in best models for only two
Best models of butterfly distributions are primarily influenced
by degree-days (31 models, Table 2, Table A3 in online supplemen-
tary materials). Yearly potential evapotranspiration acts as the pri-
mary influence on best models of an additional three species. The
variables ‘percent mature forest’ and ‘percent intensive agriculture’
were most influential in models for fewer species, 10 and two of
species best models, respectively. Similar to the case for butterflies,
degree-days influences more best models for plant species (214 of
Fig. 1. Boxplot of 10-fold cross-validated AUC from projection of models onto the
data used in their training. The median, first quartiles, and range are plotted for
each of the three study taxa.
Most-influential variables, as percent of best models for most-prevalent species.a
Influential variableSpecies group
Percent mature forest
Percent intensive agriculture
Yearly evapotranspiration (PET)
Mean soil water holding capacity
aThese species were selected by ranking all species in order of prevalence, then
selecting in order of prevalence until the species richness of the selected species at
the sampled sites was correlated with overall species richness at a level of rp= 0.95.
P.B. Pearman et al./Biological Conservation 144 (2011) 866–875
530 models) than other variables. Percent mature forest acts as the
strongest influence in models for an additional 62 species (Table 2,
Table A4 in online supplementary materials). Thus, the variables
mature forest and degree-days are the most influential variables
for models in all three species groups, but the relative frequency
with which these two variables are most influential is reversed
for bird species in comparison with the other two groups (Table 2).
3.3. Current modeled species richness
The patterns of modeled species richness of the groups of
indicator species vary among the higher taxa (Fig. A1 in online sup-
plementary materials). The greatest modeled richness of bird and
richness of butterfly indicator species at middle elevations.
3.4. Future species richness and turnover
The models for each of the three indicator groups show the ten-
dency for future species richness at relatively high elevations to be
greater than current levels (see Figs. A2–A4 in online supplemen-
tary materials). There is a consistent gain in the modeled number
of bird indicator species with suitable climatic conditions in the
Alps (Figs. 3a and b; A2b and d, online supplementary materials).
The models indicate loss of suitable climatic conditions for bird
indicator species primarily in central southern Switzerland
(Fig. 3c and d). Species that are modeled to gain suitable climatic
conditions at high elevation greatly outnumber those species for
which models predict loss of suitable conditions under both the
A1 and B2 scenarios (Fig. A2b and d).
In contrast to the results on bird indicators, the results for but-
terfly indicators show a trend towards increasing number of spe-
cies with newly suitable climate conditions at relatively high
elevations, and decreases in future predicted species richness in
the middle elevations to the north and south of the Alps (Fig. A3,
online supplementary materials). These middle elevations are
modeled originally to have the highest richness of butterfly indica-
tor species (Fig. A1b, online supplementary materials). The models
predict that many areas that currently present suitable conditions
for butterflies at middle and high elevations become unsuitable
over the next 100 years (Fig. 4c and d). Further, the number of but-
terfly indicator species that we model to lose suitable climate con-
ditions at middle elevations just north and south of the Alps
exceeds the number modeled to gain area with suitable climate
(Fig. A3b and d). The modeling results predict net decrease in rich-
ness of butterfly indicator species in currently occupied areas over
the next 100 years under both the A1 and (relatively mild) B2 sce-
narios (Fig. A3b and d, online supplementary materials). Areas that
we predict to increase in species richness (Fig. A3b, d) actually
show almost complete turnover in the species that find suitable
climatic conditions by the end of the century (Figs. 4b, d).
Predicted trends for the group of plant indicator species under
the A1 scenario are intermediate to those predicted for birds and
butterflies. We predict species richness of plant indicators to in-
crease at relatively high elevations throughout the Swiss Alps
(Fig. A4 in online supplementary materials). The Engadine valley
in the extreme western part of Switzerland appears to acquire a cli-
mate that may support the influx of a substantial number of newly
arriving plant species by the year 2100, under both the A1FI
(Fig. 5b), and milder B2 (not shown) scenarios. The results predict
substantial loss of suitable conditions for plant indicators in the
middle and high elevations in the Swiss Alps (Fig. 5d). Similar to
the case of butterflies, some high elevation areas that are predicted
to experience increased suitability for additional species early in
the century (Fig. A4a, online supplementary materials) end up
being suitable for fewer species of plants as the century closes
(Fig. A4b). Even where there is a net increase in the number of spe-
cies that may find suitable climate (Fig. A4b), there is a large turn-
over in the species that are predicted to find suitable conditions
4.1. General patterns
By studying indicator species that are closely related to geo-
graphic patterns of species richness at a resolution of 1 km
(Pearman and Weber, 2007), the predictions provided here reflect
on species richness, turnover in community composition, and the
amount of area with both suitable climate and habitat available
for each of the species under consideration. However, compared
to an earlier study where plant species richness was modeled di-
rectly as a dependent variable (Wohlgemuth et al., 2008), we also
find that the numbers of plants and butterflies for which suitable
conditions currently exist are highest at middle elevations. Condi-
tions may be suitable for comparatively many species in middle
elevation. However, this pattern may alternatively signify a mid-
domain effect (Colwell and Lees, 2000) in which shared constraints
on species range limits leads to higher species richness in the mid-
dle of a study area. In contrast, the diversity of birds shows no such
pattern (Fig. A1a, online supplementary materials). This difference
Fig. 2. Plots of the correlation between the species richness of modeled species at
sampled Z7 sites (under current climatic conditions) and species richness of all
recorded species. The modeled species are ranked in descending order of the
magnitude of the variable on the abscissa, which has subsequently been scaled
between 0 and 1 to account for small discrepancies among taxa in terms of the total
number of sampled sites. (a) Species sorted in order of decreasing value of 10-fold
cross-validation of AUC on the training data. (b) Species sorted in order of
decreasing prevalence in the Z7 dataset.
P.B. Pearman et al./Biological Conservation 144 (2011) 866–875
between the distribution of birds compared to plants and butter-
flies is not due to differences between the taxa in model perfor-
mance (Fig. 1). In each taxon, the distribution of observed species
richness was well-represented by the modeled distribution of suit-
able environmental conditions.
The results of our analysis suggest that climate change will have
broad impacts on species richness, across Switzerland. Climate
forecasts project particularly strong impacts in mountain areas,
where temperature change will exceed mean change globally (Diaz
et al., 2003; Beniston, 2006; Nogués-Bravo et al., 2006) An increase
in species richness at high elevations as these areas become war-
mer and suitable for additional species has already been observed
(Braun-Blanquet, 1957; Hofer, 1992; Walther et al., 2005; Vittoz
et al., 2006). As climate warms, it comes as no surprise that species
are predicted to extend their distribution to higher elevations, or
be faced with decreasing area with suitable climatic conditions if
they are unable to migrate to follow the geographical displacement
of areas with suitable climate. However, coarse-grained data likely
do not contain information on microclimatic variation that may be
important to plants.
4.2. Extinction and microclimate variation
Predictions of extinction due to climate change in mountainous
regions may overlook the existence of microhabitats and therefore
over-estimate the loss of area with suitable climatic conditions for
species (Randin et al., 2009). Nevertheless, the prediction of disap-
pearance of areas with suitable climate, as measured at a scale of
1 km2, suggests that availability of suitable conditions for many
high-elevation species is greatly diminished. The climate and land
cover data that were used in this study to characterize 1 km2grid
cells were taken from rasters with a resolution of 100 m. Thus, the
presence of small areas of suitable conditions may be captured to
some degree in our 1 km climate and habitat layers. The results
we obtained largely corroborate results from studies with larger
grid cell size that suggest the loss of suitable conditions for many
species at high elevations (e.g. Thuiller et al., 2005). Our study
further predicts that many areas of middle elevation in the alpine
region will show substantial change in community composition, as
indicated by turnover of plant and butterfly indicator species that
are predicted originally to have suitable habitat at intermediate
elevation (Figs. 4 and 5).
4.3. Turnover in species composition
One potentially important phenomenon is that changes in
species richness in these groups only tell part of the story of the
impacts of climate change. Both butterflies and vascular plants will
likely show increases in species richness in eastern Switzerland by
the year 2050 under the A1FI scenario (Figs. A3 and A4, online sup-
plementary materials). By the year 2100, communities at both
middle and low elevations have predicted losses of approximately
50–80% of the original number of species that found suitable cli-
mate. The magnitude of species losses in the eastern Swiss Alps
Fig. 3. Modeled turnover in indicator species of birds in Switzerland, under the A1FI scenario of the IPCC. The change in species richness corresponds to the average predicted
climate during the years 2021–2050 (a and c) and 2076–2100 (b and d). The upper panels, ‘Gain’ (a and b), present the number of species predicted to be present under future
conditions that were not predicted to be present under the initial average climate conditions for the period 1961–1990, as a precent of all species predicted present under
future climate conditions. The lower panels, ‘Loss’ (c and d), present the number of species that are predicted to be present during initial climatic conditions, but are no longer
modeled present in the future, as a precent of the number of species modeled present under the initial conditions.
P.B. Pearman et al./Biological Conservation 144 (2011) 866–875
is markedly greater later in the period (Figs. A3d and A4d, online
supplementary materials). Thus, the loss of species later in the per-
iod reduces the early trend toward increases in species richness of
butterflies and plants in the eastern Swiss Alps.
4.4. Effects of climate change and habitat conversion
Groups of species (i.e., in higher taxa) may differ in their rela-
tionships to variables that change both spatially and temporally
(Debinski et al., 2006). In our study, the distribution of suitable
conditions for bird species appears to be more closely related to
habitat variables than to climate variables. Variables reflecting for-
est cover and agriculture are assumed in this study not to respond
to climate change. However, both climate change and cultural fac-
tors will likely bring changes in the distribution of mountain for-
ests and meadows (Gehrig-Fasel et al., 2007). Other studies have
found that land cover change alone will likely drive substantial
changes in butterfly assemblages in the Alps (Lutolf et al., 2009),
and that both predicted climate change and land use change (spe-
cifically, urban sprawl) combine to affect predictions of future neo-
phyte species richness (Nobis et al., 2009). Future work should
include estimates of the impact of interactions between climate
change and social factors on the distribution of forest, intensive
and extensive agriculture, and urban and other developed areas.
Incorporation of temporal changes on the scale of the next
100 years may improve the accuracy of predictions for species
for which patterns of suitable environmental conditions are pri-
marily related to land cover variables. Nonetheless, we predict sub-
stantial impacts of climate change on species in these three higher
taxa, apart from changing habitat and land cover patterns. Finally,
climate change affects patterns of extreme climate in addition to
mean values. Inclusion of measures of climate variability and their
projected development over the next century might lead to sub-
stantial (10–20%) improvement in the models and forecasted
changes in species distribution (Zimmermann et al., 2009).
4.5. Indicator properties
We have tested the indicator properties of collections of the
niche models of species in three higher taxa. Our analysis demon-
strates that the models of approximately 10–40% of species with
the highest prevalence in the BDM Z7 dataset can be used to pre-
dict site species richness within their higher taxon (Fig. 2). Models
of these prevalent species perform better as indicators than do the
models of the species with the highest rank values of model AUC.
This suggests that the often-observed relationship between the
distribution of common species and patterns of overall species
richness (Lennon et al., 2004; Pearman and Weber, 2007) is suffi-
ciently robust to render variation in AUC among species unimpor-
tant to establishing the correlation with observed species richness.
Whether these indicator relationships hold up under continued cli-
mate change is an open question. Our data suggest that predicted
patterns in the future distribution of common species may also be
indicative of species richness. Nonetheless, the most effective indi-
cator species could be the ones that are currently common, or they
might be species that have predicted high prevalence in the future.
4.6. Modeling with monitoring data and limitations
Modeling the effects of climate change may be facilitated and
improved through use of data from designed monitoring programs.
The BDM Z7 samples do not present exhaustive censuses of the
Fig. 4. Modeled turnover in indicator species of butterflies in Switzerland, under the A1FI scenario of the IPCC. Otherwise as in Fig. 2.
P.B. Pearman et al./Biological Conservation 144 (2011) 866–875
species that are present in the sampled 1 km2cells and it is clear
that species go unobserved during sampling (Plattner et al.,
2004; Kéry and Schmid, 2006). Nonetheless, clear advantages in
the predictive performance of ENMs result from model calibration
with data from a designed sampling program (Edwards et al.,
2006). Further, the availability of species absences that are based
on a designed sampling program removes the influence on model-
ing results that arbitrary decisions during the selection of pseudo-
absence data can have (Chefaoui and Lobo, 2008; VanDerWal et al.,
2009). The use of the BDM Z7 data in calibrating niche models
provides advantages that are consistent with these considerations.
Nonetheless, these models are subject to the assumption of an
equilibrium between species distribution and environmental con-
ditions, which may not be justified due to human impacts on land
cover (e.g. Gehrig-Fasel et al., 2007). Here, we have presented the
modeled potential distribution of future suitable conditions. The
results on species richness and turnover with climate change
assume that species are able to disperse to track the geographic
distribution of suitable conditions. This is more likely the case for
birds and butterflies than it is for plants.
A further consideration when using monitoring data, such as
those we used here, is that many species have ranges that are sub-
stantially larger than the extent of the monitoring data. The BDM
Z7 data do not include information on the environmental condi-
tions that are coincident with the lower elevation limit of the dis-
tribution of some species. This deficit can limit the ability of niche
models to describe the falloff in probability of species occurrence
toward the ‘hot-dry’ end of environmental gradients. In the present
study, this likely leads to underestimation of the loss of areas with
suitable climate for species at low elevations, specifically in
northern Switzerland and in the canton Ticino in the south of the
country. Loss of suitable conditions for mountain species and esti-
mates of turnover in mountainous areas are likely less affected in
analyses of the current data because climatic suitability is lost
primarily for species that only find suitable conditions at higher
elevations currently. The extent of this effect is impossible to
assess more precisely without conducting expanded analysis that
uses species occurrence data and environmental data that capture
the entire environmental gradients over which the species are
distributed. Such an exercise would help assess the accuracy of
predictions of loss of area with suitable climate for species that
are typical of low elevations in Switzerland. However, such data
are not available at 1 km2over most of Europe and would likely
come from lower resolution atlas data, illustrating a real trade-
off between data resolution and geographic extent. New methods
should be developed to address this issue and concomitantly to
understand the implications that accompany the simultaneous
use of data of various resolutions in ENMs.
Species occurrence data that come from designed sampling
programs for monitoring species distributions and biodiversity
provide a rich source of information for addressing questions in
biogeography, effects of climate change, and conservation. This
study demonstrates that ecological niche models of the most-
prevalent species can perform better in predicting current levels
of species richness than do an equal number of the best performing
models, as measured by AUC. This implies that predictions of
Fig. 5. Modeled turnover in indicator species of plants in Switzerland, under the A1FI scenario of the IPCC. Otherwise as in Fig. 2.
P.B. Pearman et al./Biological Conservation 144 (2011) 866–875
future species richness may be more reliable when made with spe-
cies distribution models of the currently most-prevalent species.
This constitutes an empirically-based hypothesis that should be
tested (1) as species distributions in the study region respond to
ongoing climate change and (2) in additional mountainous regions
where the same pattern might be detected. Similarly, our study
suggests that species turnover in communities in the Swiss Alps
could be substantial, although species richness may not be greatly
affected in some areas. This suggests a second empirical hypothesis
that will be testable over time locally and also constitutes a test-
able pattern that might hold for other mountain ranges in Europe
The comments of T.C. Edwards, C.F. Randin and two anonymous
reviewers led to improvements in earlier versions of the manu-
script. We thank the consulting firm Hintermann and Weber AG,
and Matthias Plattner for making the Z7 data from Biodiversity
Monitoring Switzerland available for analysis. P.B.P. was supported
during analysis and writing by the EU FP6 ECOCHANGE Project and
by Hintermann and Weber AG. P.B.P. and N.E.Z. acknowledge the
support of the Sinergia Project CRSII3-125240 ‘SPEED’ and P.B.P.
acknowledges the support of Project 3100A0-122433 ‘ENNIS’, both
funded by the Swiss National Science Foundation.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.biocon.2010.11.020.
Araújo, M.B., Pearson, R.G., Thuiller, W., Erhard, M., 2005. Validation of species-
climate impact models under climate change. Global Change Biology 11, 1504–
Aubry, S., Magnin, F., Bonnet, V., Preece, R.C., 2005. Multi-scale altitudinal patterns
in species richness of land snail communities in south-eastern France. Journal of
Biogeography 32, 985–998.
Beniston, M., 2006. Mountain weather and climate: a general overview and a focus
on climatic change in the Alps. Hydrobiologia 562, 3–16.
Beniston, M., Fox, D.G., Adhikary, S., Andressen, R., Guisan, A., Holten, J.I., Innes, J.,
Maitima, J., Price, M.F., Tessier, L., 1996. Impacts of climate change on mountain
regions. In: Intergovernmental Panel on Climate Change. Second Assessment
Report, IPCC. Cambridge University Press, Cambridge (Chapter 5).
BFS, 1992. Bodeneignungskarte. Bundesamt für Statistik. Neuchatel, Switzerland.
Braun-Blanquet, J., 1957. Ein Jahrhunder Florenwandel am Piz Linard (3414 m). Bull.
Jard. Bot. Etat Brux. Volume Jubilaire Walter Robyns, pp. 221–232.
Chase, M.K., Kristan, W.B., Lynam, A.J., Price, M.V., Rotenberry, J.T., 2000. Single
species as indicators of species richness and composition in California coastal
sage scrub birds and small mammals. Conservation Biology 14, 474–487.
Chefaoui, R.M., Lobo, J.M., 2008. Assessing the effects of pseudo-absences on
predictive distribution model performance. Ecological Modelling 210, 478–486.
Cohen, J., 1960. A coefficient of agreement for nominal scales. Educational and
Psychological Measurement 20, 37–46.
Colwell, R.K., Lees, D.C., 2000. The mid-domain effect: geometric constraints on the
geography of species richness. Trends in Ecology & Evolution 15, 70–76.
relationships betweenbird and
environmental change. Ecological Applications 16, 380–393.
Diaz, H.F., Grosjean, M., Graumlich, L., 2003. Climate variability and change in high
elevation regions: past, present and future. Climatic Change 59, 1–4.
Edwards, T.C., Cutler, D.R., Zimmermann, N.E., Geiser, L., Moisen, G.G., 2006. Effects
of sample survey design on the accuracy of classification tree models in species
distribution models. Ecological Modelling 199, 132–141.
Efron, B., Tibshirani, R.J., 1998. An Introduction to the Bootstrap. Chapman and Hall/
CRC Press, Boca Raton, Florida.
Elith, J., Graham, C.H., Anderson, R.P., Dudik, M., Ferrier, S., Guisan, A., Hijmans, R.J.,
Huettmann, F., Leathwick, J.R., Lehmann, A., Li, J., Lohmann, L.G., Loiselle, B.A.,
Manion, G., Moritz, C., Nakamura, M., Nakazawa, Y., Overton, J.M., Peterson, A.T.,
Phillips, S.J., Richardson, K., Scachetti-Pereira, R., Schapire, R.E., Soberon, J.,
Williams, S., Wisz, M.S., Zimmermann, N.E., 2006. Novel methods improve
prediction of species’ distributions from occurrence data. Ecography 29, 129–
Elith, J., Leathwick, J.R., Hastie, T., 2008. A working guide to boosted regression trees.
Journal of Animal Ecology 77, 802–813.
Fielding, A.H., Bell, J.F., 1997. A review of methods for the assessment of prediction
errors in conservation presence/absence models. Environmental Conservation
Fitter, A.H., Fitter, R.S.R., 2002. Rapid changes in flowering time in British plants.
Science 296, 1689–1691.
Friedman, J.H., 2001. Greedy function approximation: a gradient boosting machine.
Annals of Statistics 29, 1189–1232.
Gehrig-Fasel, J., Guisan, A., Zimmermann, N.E., 2007. Tree line shifts in the Swiss
Alps: climate change or land abandonment? Journal of Vegetation Science 18,
Grabherr, G., Gottfried, M., Pauli, H., 1994. Climate effects on mountain plants.
Nature 369, 448.
Guisan, A., Holten, J.I., Spichiger, R., Tessier, L. (Eds.), 1995. Potential Ecological
Impacts of Climate Change in the Alps and Fennoscandian Mountains.
Conservatoire et Jardin Botaniques de Genève, Geneva, Switzerland.
Guisan, A., Zimmermann, N.E., Elith, J., Graham, C.H., Phillips, S., Peterson, A.T., 2007.
What matters for predicting the occurrences of trees: techniques, data, or
species’ characteristics? Ecological Monographs 77, 615–630.
Hofer, H.R., 1992. Veränderungen in der Vegetation von 14 Gipfeln der
Berninagebietes zwischen 1905 und 1985. Ber. Geobot. Inst. ETH Stiftung
Rübel Zürich 58, 39–54
IPCC, 2001. In: Houghton, J.T., Ding, Y., Griggs, D.J., Noguer, M., van der Linden, P.J.,
Dai, X., Maskell, K., Johnson, C.A. (Eds.), Climate Change 2001: The Scientific
Basis. Contribution of Working Group I to the Third Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge University Press,
Cambridge, United Kingdom.
Jetz, W., Rahbek, C., 2002. Geographic range size and determinants of avian species
richness. Science 297, 1548–1551.
Kéry, M., Schmid, H., 2006. Estimating species richness: calibrating a large avian
monitoring programme. Journal of Applied Ecology 43, 101–110.
Körner, C., 2003. Alpine Plant Life: Functional Plant Ecology of High Mountain
Ecosystems. Second Edition. Springer, Berlin.
Kumar, L., Skidmore, A.K., Knowles, E., 1997. Modelling topographic variation in
solar radiation in a GIS environment. International Journal of Geographical
Information Science 11, 475–497.
Lawler, J.J., White, D., Sifneos, J.C., Master, L.L., 2003. Rare species and the use of
indicator groups for conservation planning. Conservation Biology 17, 875–882.
Lennon, J.J., Koleff, P., Greenwood, J.J.D., Gaston, K.J., 2004. Contribution of rarity and
commonness to patterns of species richness. Ecology Letters 7, 81–87.
Lutolf, M., Bolliger, J., Kienast, F., Guisan, A., 2009. Scenario-based assessment of
future land use change on butterfly species distributions. Biodiversity and
Conservation 18, 1329–1347.
Meehl, G.A., Stocker, T.F., Collins, W.D., Friedlingstein, P., Gaye, A.T., Gregory, J.M.,
Kitoh, A., Knutti, R., Murphy, J.M., Noda, A., Raper, S.C.B., Watterson, I.G.,
Weaver, A.J., Zhao, Z.-C., 2007. In: Solomon, S., Qin, D., Manning, M., Chen, Z.,
Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L., (Eds.), Global Climate
Projections. In Climate Change 2007: The Physical Science Basis. Contribution
of Working Group I to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change. Cambridge University Press, Cambridge, United
Kingdom and New York, NY, USA.
Mikusinski, G., Gromadzki, M., Chylarecki, P., 2001. Woodpeckers as indicators of
forest bird diversity. Conservation Biology 15, 208–217.
Mitchell, T.D., Carter, T.R., Jones, P.D., Hulme, M., New, M., 2004. A comprehensive
set of high-resolution grids of monthly climate for Europe and the globe: the
observed record (1901–2000) and 16 scenarios (2001–2100). Tyndall Center
Working Paper No. 55. Tyndall Center for Climate Change Research. University
of East Anglia, East Anglia, UK.
Nobis, M.P., Jaeger, J.A.G., Zimmermann, N.E., 2009. Neophyte species richness at the
landscape scale under urban sprawl and climate warming. Diversity and
Distributions 15, 928–939.
Nogués-Bravo, D., Araújo, M., Errea, M.P., Martinez-Rica, J.P., 2006. Exposure of
global mountain systems to climate warming during the 21st century. Global
Environmental Change – Human and Policy Dimensions 17, 420–428.
Orme, C.D.L., Davies, R.G., Burgess, M., Eigenbrod, F., Pickup, N., Olson, V.A., Webster,
A.J., Ding, T.S., Rasmussen, P.C., Ridgely, R.S., Stattersfield, A.J., Bennett, P.M.,
Blackburn, T.M., Gaston, K.J., Owens, I.P.F., 2005. Global hotspots of species
richness are not congruent with endemism or threat. Nature 436, 1016–1019.
Pauli, H., Gottfried, M., Grabbherr, G., 1996. Effects of climate change on mountain
ecosystems. Upward shifting of alpine plants. World Resource Review 8, 382–
Pauli, H., Gottfried, M., Reiter, K., Klettner, C., Grabherr, G., 2007. Signals of range
expansions and contractions of vascular plants in the high Alps: observations
(1994–2004) at the GLROIA master site Schrankogel, Tyrol, Austria. Global
Change Biology 13, 147–156.
Pearman, P.B., Penskar, M.R., Schools, E.H., Enander, H.D., 2006. Identifying potential
indicators of conservation value using natural heritage occurrence data.
Ecological Applications 16, 186–201.
Pearman, P.B., Weber, D., 2007. Common species determine richness patterns in
biodiversity indicator taxa. Biological Conservation 138, 109–119.
Plattner, M., Birrer, S., Weber, D., 2004. Data quality in monitoring plant species
richness in Switzerland. Community Ecology 5, 135–143.
Prendergast, J.R., Quinn, R.M., Lawton, J.H., 1999. The gaps between theory and
practice in selecting nature reserves. Conservation Biology 13, 484–492.
Randin, C.F., Dirnbock, T., Dullinger, S., Zimmermann, N.E., Zappa, M., Guisan, A.,
2006. Are niche-based species distribution models transferable in space?
Journal of Biogeography 33, 1689–1703.
P.B. Pearman et al./Biological Conservation 144 (2011) 866–875
Randin, C.F., Engler, R., Normand, S., Zappa, M., Zimmermann, N.E., Pearman, P.B.,
Vittoz, P., Thuiller, W., Guisan, A., 2009. Climate change and plant distribution:
local models predict high-elevation persistence. Global Change Biology 15,
Ridgeway, G., 2007. Generalized boosted models: a guide to the gbm package.
Swets, J.A., 1988. Measuring the accuracy of diagnostic systems. Science 240, 1285–
Theurillat, J.-P., Felber, F., Geissler, P., Gobat, J.-M., Fierz, M., Fischlin, A., Küpfer, P.,
Schlüssel, A., Velluti, C., Zhao, G.-F., 1998. Sensitivity of plant and soil
ecosystems of the Alps to climate change. In: Cebon, P., Dahinden, U., Davies,
H.C., Imboden, D., Jäger, C.C. (Eds.), Views from the Alps. Regional Perspectives
on Climate Change. MIT Press, London, pp. 225-308.
Thomas, C.D., Lennon, J.J., 1999. Birds extend their ranges northwards. Nature 399,
Thuiller, W., Lavorel, S., Araujo, M.B., Sykes, M.T., Prentice, I.C., 2005. Climate change
threats to plant diversity in Europe. Proceedings of the National Academy of
Sciences of the United States of America 102, 8245–8250.
Tingley, M.W., Monahan, W.B., Beissinger, S.R., Moritz, C., 2009. Birds track their
Grinnellian niche through a century of climate change. Proceedings of the
National Academy of Sciences of the United States of America 106, 19637–
VanDerWal, J., Shoo, L.P., Graham, C., William, S.E., 2009. Selecting pseudo-absence
data for presence-only distribution modeling: how far should you stray from
what you know? Ecological Modelling 220, 589–594.
Vazquez, D.P., Aizen, M.A., 2003. Null model analyses of specialization in plant–
pollinator interactions. Ecology 84, 2493–2501.
Vittoz, P., Jutzeler, S., Guisan, A., 2006. Flore alpine et réchauffement climatique:
observation de trois sommets valaisans à travers le 20ème siècle. Bulletin de la
Murithienne 123 (2005), 49–59.
Walther, G.-R., Beißner, S., Burga, C.A., 2005. Trends in the upward shift of alpine
plants. Journal of Vegetation Science 16, 541–548.
Walther, G.R., 2003. Plants in a warmer world. Perspectives in Plant Ecology,
Evolution and Systematics 6, 169–185.
Walther, G.R., Post, E., Convey, P., Menzel, 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.
Warman, L.D., Forsyth, D.M., Sinclair, A.R.E., Freemark, K., Moore, H.D., Barrett, T.W.,
Pressey, R.L., White, D., 2004. Species distributions, surrogacy, and important
conservation regions in Canada. Ecology Letters 7, 374–379.
Weber, D., Hintermann, U., Zangger, A., 2004. Scale and trends in species richness:
considerations for monitoring biological diversity for political purposes. Global
Ecology and Biogeography 13, 97–104.
Wohlgemuth, T., Nobis, M.P., Kienast, F., Plattner, M., 2008. Modelling vascular plant
diversity at the landscape scale using systematic samples. Journal of
Biogeography 35, 1226–1240.
Zimmermann, N.E., Edwards, T.C., Moisen, G.G., Frescino, T.S., Blackard, J.A., 2007.
Remote sensing-based predictors improve distribution models of rare, early
successional and broadleaf tree species in Utah. Journal of Applied Ecology 44,
Zimmermann, N.E., Kienast, F., 1999. Predictive mapping of alpine grasslands in
Switzerland: species versus community approach. Journal of Vegetation Science
Zimmermann, N.E., Yoccoz, N.G., Edwards, T.C., Meier, E.S., Thuiller, W., Guisan, A.,
Schmatz, D.R., Pearman, P.B., 2009. Climatic extremes improve predictions of
spatial patterns of tree species. Proceedings of the National Academy of
Sciences of the United States of America 106, 19723–19728.
P.B. Pearman et al./Biological Conservation 144 (2011) 866–875