Page 1
Climate change effects on animal and plant phylogeneticdiversity in southern AfricaDOROTHEA V . P IO 1 , 2 , 3 , † , ROB IN ENGLER 1 , † , H . P ETER L INDER 4 , ARA MONAD JEM5 ,
F ENTON P .D . COTTER I LL 6 , P ETER J . TAYLOR 7 , 8 , M . CORR IE SCHOEMAN9 ,
BEN JAMIN W . PR ICE 1 0 , 1 1 , MART IN H . V I LLET 1 1 , GEETA E ICK 1 2 ,
N ICOLAS SALAMIN 1 , 2 , ‡ and ANTOINE GUISAN1 , 1 3 ,‡1Department of Ecology & Evolution, University of Lausanne, Lausanne, 1015, Switzerland, 2Swiss Institute of Bioinformatics,
University of Lausanne, Lausanne 1015, Switzerland, 3Fauna & Flora International, London, WC2N 6DF, UK, 4Institute for
Systematic Botany, University of Zurich, Zollikerstrasse 107, Zurich 8008, Switzerland, 5All Out Africa Research Unit,
Department of Biological Sciences, University of Swaziland, Private Bag 4, Kwaluseni, Swaziland, 6AEON - Africa Earth
Observatory Network, Geoecodynamics Research Hub, Department of Botany and Zoology, University of Stellenbosch, Private Bag
X1, Matieland 7602, South Africa, 7Durban Natural Science Museum, P. O. Box 4085, Durban, South Africa , 8Department of
Ecology and Resource Management, School of Environmental Sciences, University of Venda, P/Bag X5050, Thohoyandou 0950,
South Africa, 9School of Life Sciences, University of Kwazulu-Natal, Westville campus, Durban 4000, South Africa, 10Department
of Ecology and Evolutionary Biology, University of Connecticut, 75 North Eagleville Road, Storrs, CT 06269, USA, 11Department
of Zoology & Entomology, Rhodes University, Grahamstown 6140, South Africa, 12Institute of Ecology and Evolution, University
of Oregon, Eugene Oregon USA, 13Institute of Earth Sciences, University of Lausanne, Lausanne 1015, Switzerland
Abstract
Much attention has been paid to the effects of climate change on species’ range reductions and extinctions. There is
however surprisingly little information on how climate change driven threat may impact the tree of life and result in
loss of phylogenetic diversity (PD). Some plant families and mammalian orders reveal nonrandom extinction pat-
terns, but many other plant families do not. Do these discrepancies reflect different speciation histories and does cli-
mate induced extinction result in the same discrepancies among different groups? Answers to these questions
require representative taxon sampling. Here, we combine phylogenetic analyses, species distribution modeling, and
climate change projections on two of the largest plant families in the Cape Floristic Region (Proteaceae and Restiona-
ceae), as well as the second most diverse mammalian order in Southern Africa (Chiroptera), and an herbivorous
insect genus (Platypleura) in the family Cicadidae to answer this question. We model current and future species distri-
butions to assess species threat levels over the next 70 years, and then compare projected with random PD survival.
Results for these animal and plant clades reveal congruence. PD losses are not significantly higher under predicted
extinction than under random extinction simulations. So far the evidence suggests that focusing resources on climate
threatened species alone may not result in disproportionate benefits for the preservation of evolutionary history.
Keywords: Chiroptera, Cicadidae, climate change, extinction, niche models, phylogenetic diversity, Platypleura, Proteaceae,
Restionaceae, southern Africa
Received 30 August 2012 and accepted 1 November 2013
Introduction
The effects of climate change on the distribution, abun-
dance, and migration of plant and animal species are
already apparent both in terrestrial and marine ecosys-
tems (Parmesan & Yohe, 2003; Perry et al., 2005; Chen
et al., 2011) and these effects are predicted to become
even more striking in the coming decades (Thomas
et al., 2004; Thuiller et al., 2005, 2011; Broennimann
et al., 2006; Engler et al., 2011). Warming experiments
also suggest negative effects of climate change on
native plant diversity (Gedan & Bertness, 2009).
In Southern Africa, the effects of climate warming are
particularly daunting (West et al., 2012), as magnified
by human demographic pressure on intact landscapes
with widespread and burgeoning land-use change that
converts wild lands into agro-ecosystems (Vetter, 2009;
Willis & Bhagwat, 2009). Together, these two factors
have been leading to the gradual desertification of huge
portions of the landscape (Vetter, 2009). Subtropical
thickets are turning into pseudosavannas of scattered
woody species, where tree mortality rates exceed
recruitment. In the Karoo region of South Africa, heavy
Correspondence: Antoine Guisan, Nicolas Salamin, tel. +41
21 692 4154, fax +41 21 692 4165, e-mail: [email protected] ,
[email protected] †Co-first authors‡Co-last authors
1© 2014 John Wiley & Sons Ltd
Global Change Biology (2014), doi: 10.1111/gcb.12524
Global Change Biology
Page 2
grazing and drought have led to the loss of palatable
shrubs and increased dominance by unpalatable woody
species and annuals (Vetter, 2009; Willis & Bhagwat,
2009). A significant portion of miombo savanna wood-
lands across south-central Africa have been converted
to agro-ecosystems since the mid-20th century and this
trend continues (Campbell, 1996; Du Toit & Cumming,
1999; Willis & Bhagwat, 2009).
Over the past decade, several authors have predicted
that native species richness would be reduced and turn-
over rates would increase in the Southern African biota
as a result of global change (Midgley et al., 2003, 2006;
Bomhard et al., 2005; Broennimann et al., 2006; Thuiller
et al., 2006; Midgley & Thuiller, 2007). Declines in
groups of species with similarities in life history traits,
in association with particular environments or host spe-
cies, and susceptibility to anthropogenic pressures are
likely to be influenced by their shared evolutionary his-
tory (Purvis et al., 2000). This can lead to differential
probabilities of (nonrandom) extinction between lin-
eages and can have consequences on future levels of
species and genetic diversity. Conversely, a phyloge-
netically random risk pattern implies that species’ fates
are not largely determined by traits that show a strong
tendency to have similar values among closely related
species.
Assessing climate threat on phylogenetic diversity
(PD) allows predicting whether related taxonomic
groups will require more urgent attention, thus
enabling conservation efforts to be more targeted and
effective for specific groups. There are other advanta-
ges to quantifying the effects of climate change on
PD. Phylogenies represent the evolutionary history of
a taxonomic group and thus PD should encompass
information on multiple traits, represented within
phylogenies into a simple index of ecological,
phenotypic, and functional similarity. PD is thought
to represent an objective and cost effective proxy for
functional diversity and ecosystem function (Srivast-
ava et al., 2012). Although studies demonstrating a
clear link between PD and ecosystem function are
still largely lacking (Winter et al., 2012), there is now
some evidence that PD has an important influence on
aboveground productivity and ecosystem stability
(Cadotte et al., 2012). Ecologically, a phylogenetic
approach to measuring diversity makes it possible to
move beyond traditional taxonomic classification,
toward understanding where on the tree of life
habitat differentiation and adaptation are taking place
(Kembel et al., 2011). Moreover, several studies
suggest that larger PD should increase the evolution-
ary potential and resilience of species to adapt to
environmental change (Meynard et al., 2011; Sgro’
et al., 2011). The use of PD in making inferences
about complex evolutionary processes in a system, as
well as its role as a proxy for ecological function
makes it very powerful.
Despite the important effect that differential threat
probabilities can have on future biodiversity levels,
little attention has been paid so far to exploring the
relationship between predicted climate change effects
and PD (Thuiller et al., 2011). Using IUCN threat
status across taxonomic groups, some authors have
found evidence for nonrandom extinction processes
in angiosperm, avian, and mammalian phylogenies
worldwide (Purvis et al., 2000; Vamosi & Wilson,
2008; Bromham et al., 2012), and in angiosperm
phylogeny in Australia (Sjostrom & Gross, 2006). For
mammals, threat status has been shown to be related
both to anthropogenic pressures and life history
traits, hinting that extinction might be nonrandom
(Davies et al., 2008). Yet, recent results for the same
three taxonomic groups (plants, birds and mammals)
in Europe, but using climate-specific predicted extinc-
tions provide a contrasting view, by showing no
higher loss of evolutionary history in these groups
than expected by chance alone (Thuiller et al., 2011).
This reopens the question of random vs. nonrandom
threat in the tree of life and calls for additional stud-
ies on the effects of climate-specific threat.
Here, we aim to contribute to this debate by provid-
ing evidence from Southern African animal and plant
taxa. We used two major components of South Africa’s
Cape flora (Proteaceae and Restionaceae), the second
most diverse mammal order in Southern Africa (Chi-
roptera) as well as an insect genus (Platypleura) in the
family Cicadidae to make predictions on how threat
and PD may be affected by climate change 70 years
from now. The choice of data was dictated by the
availability and reliability of novel distribution and
genetic datasets for a diverse set of plants and animals.
Moreover, Platypleura were selected because insect
datasets are generally underrepresented in this type of
study and can be expected to provide an interesting
comparison with plants and small mammals. Cicadas
are large-bodied, sap-sucking insects with relatively
high population densities and specific in their plant
associations, (e.g., Price et al., 2007; Price, 2010). In
Southern Africa, almost all species show some degree
of endemism (Price et al., 2007; Price, 2010).
More specifically, we investigated the likely effects of
climate-specific threat on PD (by using contractions of
suitable climate as a proxy for extinction risk) and
whether surviving PD was higher or lower than that
expected under random extinction processes. This pro-
vided a timely assessment of potential climate change
effects on the probability of survival between lineages
of different Southern African taxonomic groups.
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
2 D. V. PIO et al.
Page 3
Materials and methods
Species occurrence data
Species occurrence datasets were assembled for four animal
and plant groups occurring in Southern Africa: Proteaceae,
Restionaceae, Chiroptera and Platypleura. Georeferenced
occurrence data were obtained (i) for the Proteaceae from the
Protea Atlas (Rebelo, 2012; covering 82% of all species); (ii) for
the Restionaceae based on herbarium specimens collected by
one of us (H.P. Linder et al., unpublished; last accessed 2006;
covering all species); (iii) for Chiroptera from museum
records, as well as from several unpublished surveys and
represent the most comprehensive distribution dataset for
Southern African bat species existing today (Monadjem et al.,
2010b; 88; % of all species), and (iv) for Platypleura from an
unpublished dataset (M.H. Villet, unpublished; 66% of all spe-
cies). The determination of all specimens of Restionaceae was
re-assessed, and species which were indeterminate, or which
could be named only to generic level, were removed from the
data matrix. No specimens that were not personally checked
by HPL were included. The georeferenced data were then
assessed by mapping the distribution of each species individu-
ally, and distribution points deemed to be unlikely (wrong
soils, geographical outliers, or in the ocean) were revisited and
corrected (they were not simply pruned from the matrix). The
re-identification is done over a period of 14 years. The genus
Platypleura is monophyletic in Africa, and the putative Asian
species are being transferred to other genera because they are
phylogenetically independent linages. The African species
form two sister clades, one (34 species) endemic to South
Africa and the other (12 species) originating in Southern
Africa and spreading northwards into East and West Africa,
but not overlapping geographically with the first clade. Five
species of the latter clade from Western Africa are clustered
on a branch very distant from the root of the tree where they
have limited effect on the phylogenetic distances between the
Southern African species and on PD calculations (Price, 2010).
The spatial accuracy associated with observation records was
of approximately 2 min for the Proteaceae and Restionaceae,
from accurate GPS localities to quarter degree squares
(15′ 9 15′) for the Chiroptera, and of 1 min for the Platypleura.
At the latitude of South Africa, 1 min corresponds to
approximately 1.6 km.
Climatic data
We used fine resolution 1′ 9 1′ (~ 1.6 9 1.6 km at this
latitude) climatic data for calibrating and projecting the Prote-
aceae and Restionaceae models, since the spatial accuracy of
the occurrence data allowed for finer scale predictions. We
used a set of seven little-correlated (correlation coefficient
<0.7), ecologically meaningful variables for plants, adapted
from the Worldclim database for the Cape region, namely:
annual evapotranspiration, evapotranspiration of the wettest
quarter, annual precipitation, precipitation of the wettest
quarter, precipitation of the driest quarter, annual tempera-
ture, and temperature of the coldest quarter. Climatic data for
the calibration and projection of Chiroptera and Platypleura
models were obtained from the CRU CL 2.0 dataset (New
et al., 2000) generated at a grid size of 10′ 9 10′ (~ 16 9 16 km
at this latitude). These data consisted in six little-correlated
variables (correlation coefficient <0.7) representing the major
climatic gradients in Africa, namely: mean annual potential
evapotranspiration, annual growing-degree days, minimum
temperature of the coldest month, maximum temperature of
the warmest month, mean annual temperature, and annual
sum of precipitation.
Climate projections for the year 2080 (average of 2070–2090)
were produced by perturbing the current climatic data with
anomalies derived from climatic projections produced by the
HadCM3 general circulation model using the A1FI IPCC SRES
scenario (Nakicenovic & Swart, 2000). This scenario is consid-
ered severe, with concentrations of CO2 increasing from
380 ppm in 2000 to 800 ppm in 2080, resulting in a global tem-
perature rises of 3.6 K. For South Africa, the A1FI scenario
projects an increase in yearly average temperature of
2.8 � 0.7 K (mean � 1 SD) and a decrease of 59 � 20 mm in
annual sum of precipitations. When considering only the Cape
Floristic region, the projected increase in temperature is of
2.2 � 0.3 K, and the projected decrease in rainfall of
45 � 13 mm.
Species distribution modeling
As our species datasets were lacking true species absences,
we generated random pseudo-absences (i.e. randomly
selected locations that were considered as species absences;
Elith et al., 2006). For each species, pseudo-absences were
generated in a number 10 times more abundant than the
species’ presences. Pseudo-absences were down-weighted
for model calibration to ensure equal prevalence between
presences and pseudo-absences (presences were given a
weight of 1.0, pseudo-absences a weight of 0.1; following
Elith et al., 2006).
Species distribution models were calibrated for each species
with more than 20 occurrences using six different modeling
techniques: generalized linear models (GLM; Mccullagh &
Nelder, 1989), generalized additive models (GAM; Hastie &
Tibshirani, 1986), boosted regression trees (GBM; Ridgeway,
1999), random forest models (Breiman, 2001), multivariate
adaptive regression splines (MARS; Friedman, 1991) and clas-
sification tree analysis (CTA; Breiman et al., 1984). GLMs and
GAMs were calibrated using a binomial distribution and a
logistic link function. A bidirectional stepwise procedure was
used for explanatory variable selection, based on the Akaike
information criterion (Akaike, 1974). Up to second order poly-
nomials (linear and quadratic terms) were allowed for each
explanatory variable in GLMs, and up to third order splines in
GAMs. GBMs were calibrated with a maximum number of
trees set to 5,000, fivefold cross-validation procedures to select
the optimal numbers of trees to be kept and a value of five as
maximum depth of variable interactions. Random forest mod-
els were fitted by growing 750 trees with half the numbers of
available predictors sampled for splitting at each node. MARS
models were fitted with a maximum interaction degree equal
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
CLIMATE CHANGE & PHYLOGENETIC DIVERSITY LOSS 3
Page 4
to 2. All models were calibrated using the BIOMOD package
(Thuiller et al., 2009) in the R software (R Development Core
Team, 2009).
The predictive power of each individual model was evalu-
ated through a repeated data-splitting procedure (for details
see Thuiller et al., 2009). A model was trained on 70% of the
data before being evaluated against the remaining 30% using
two measures: the area under the receiver operating character-
istic curve (AUC; Hanley & Mcneil, 1982) and the true skill sta-
tistic (TSS; Allouche et al., 2006). This data-splitting procedure
was repeated 10 times and the evaluation values averaged.
The final models used to carry out spatial projections were
calibrated using 100% of the data. To avoid using poorly cali-
brated models, only projections from models with AUC >0.75and TSS >0.5 were considered in all subsequent analyses.
Model projections were carried out over various spatial
extents, depending on the group of species: Proteaceae and
Restionaceae models were projected only over the Cape Floris-
tic region, while and Platypleura and Chiroptera models were
projected over respectively all of South Africa and the entire
southern part of the African continent. All models were pro-
jected at the same spatial resolution as the data with which
they were calibrated (1′ for Proteaceae and Restionaceae, 10′
for Platypleura and Chiroptera).
We assumed unlimited (universal) dispersal, which makes
our area loss or extinction rates conservative.
Model averaging (ensemble forecast) was performed by
weighting the individual model projections respectively by
their AUC or TSS scores and averaging the result, a method
shown to be particularly robust (Marmion et al., 2009). Each
ensemble forecast model was then reclassified into two differ-
ent binary projections (i.e. the species is either projected pres-
ent or absent), using the threshold that would respectively
maximize jointly the percentage of presences and absences
correctly predicted (Liu et al., 2005), and maximize the TSS
value.
Since generating pseudo-absence introduced a random
component into our models, the entire modeling procedure
was replicated 11 times, each time with a new set of pseu-
do-absences (pseudo-absence replicates). The final projected
distribution of a species was then obtained by combining
together the projections from our 11 replicates. The number of
replicates was not increased further due to computational and
data storage resources-constraints. Two methods were used:
majority rule and average. In the majority rule method, the
binary projections obtained from each replicate are summed
and the final value is equal to the majority across the 11 repli-
cates, which is why the number of replicates needs to be
uneven (e.g., if six replicates predicted a presence and five and
an absence, the majority method assigns a presence). In the
average method, the continuous ensemble forecasts for each
pseudo-absence replicate were averaged. The resulting aver-
aged projection was then reclassified into binary projections
using the average reclassification threshold of all 11 replicates.
In addition, regressions to test the relationship between the
size of present and future distributions as well as between
present distributions, and the absolute overlap between pres-
ent and future distributions, were run.
Phylogenetic tree building
We used dated species level phylogenies for each taxonomic
group. Reconstructions always had a Bayesian component,
but the exact options used changed between the datasets
because MCMC methods must be adjusted to ensure conver-
gence and appropriate mixing. We measured tree imbalance
and stemminess (Heard, 1992; Pybus & Harvey, 2000; Davies
et al., 2012) of each tree using rescaled Colless values and the
apTreeshape package in R. These measures of the branching
pattern of a tree are interesting attributes because they can
give us indications on macro-evolutionary processes that have
shaped species evolution and can be important when assess-
ing patterns of extinction (e.g., Cadotte & Davies, 2010). For
example, negative rescaled Colless values indicate more
balanced phylogenies (as opposed to pectinate or more comb-
like), negative stemminess values indicate ‘tippy’ trees (trees
with longer branches toward the tips) and positive stemmi-
ness values indicate ‘stemmy’ trees (trees with longer
branches toward the root; Davies et al., 2012).
Proteaceae. A calibrated phylogenetic tree for the Proteaceae
based on 23 genes was assembled from pre-existing data (see
Valente et al., 2010; Pio et al., 2011) and all other available
sequences for the South African (and some Australian) Protea-
ceae in GenBank following the method of Sanderson & Mcma-
hon (2007). The tree comprising 284 taxa was built using
MrBayes 3.1.2 (Huelsenbeck et al., 2001). Two runs of four
Markov chain Monte Carlo chains were run for 10 million gen-
erations using the GTR+Gamma model of DNA evolution and
default priors. The best-fit model was estimated through like-
lihood-ratio tests. The convergence of the two runs and the
stabilization of the model parameters were assessed using Tra-
cer 1.5 (Drummond & Rambaut, 2007) and we removed the
first 5,000 trees as burn-in. The tree with the highest posterior
probability was then dated with a penalized likelihood
method (Sanderson, 2002) as implemented in the ape package
in R (Paradis et al., 2004) using the fossils described in Sauquet
et al. (2009). Penalized likelihood was used here instead of
relaxed clock analyses to alleviate the computational cost due
to the size of the matrix used. To check the consistency of the
date estimates, we also ran penalized likelihood on 100 ran-
domly sampled trees from the posterior distribution given by
MrBayes (see Pio et al., 2011 for details).
Restionaceae. The tree for this group was published previ-
ously (Hardy et al., 2008) and here we used the tree estimated
based on plastid DNA sequences (ITS). A more recent study
(Litsios et al., accepted, Evolution) has added one nuclear
locus to this tree, but the topology remains unchanged.
Chiroptera. A species level phylogeny for 89 species of bats
was built using two mitochondrial markers: cytochrome b and
16S rRNA. DNA was extracted from fresh skin samples
collected in Malawi and Mozambique between August and
December 2007 (Monadjem et al., 2010a), and muscle samples
obtained from the Durban Science Museum, South Africa.
Both fresh and museum specimens were stored in 95%
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
4 D. V. PIO et al.
Page 5
ethanol, and DNA was extracted using a phenolchloroform-
isoamyl procedure (Sambrook et al., 1989). We did not use the
mammal supertree because the taxonomy of Southern African
bat species has undergone recent changes and our species
coverage is more extensive. A set of six custom primers were
used to amplify the two genes, which were sequenced using
the Sanger method (Sanger et al., 1977). Existing sequences
available in GenBank by 1/10/09 were also integrated in the
analysis. Sequences were first aligned with a multiple
alignment program, ClustalX (Larkin et al., 2007) and then
adjusted manually with the sequence alignment editor BioEdit
(Hall, 1999). Sequences from the two genes were concatenated
to form a supermatrix after checking that single gene
phylogenetic analyses did not show incongruent nodes
supported by more than 80% bootstrap.
Trees were built with both Maximum Likelihood (ML) and
Bayesian methods. We used PhyML 2.4.4 (Guindon &
Gascuel, 2003) and MrBayes 3.1.2 (Huelsenbeck et al., 2001).
The best model for the combined dataset was GTR+G+I as
determined by likelihood-ratio tests. Three mammals, the
horse (Equidae, Perissodactyla), the pangolin (Manidae, Pholi-
dota) and the sea lion (Phocidae, Carnivora) were designated
as outgroups. ML bootstrap support was determined from 200
replicates with NNI branch swapping (Salamin et al., 2003).
The MrBayes tree was obtained by running three independent
analyses, each with four Markov chains for 50 million genera-
tions with sampling performed every 1000 trees, thereby gen-
erating 50,000 sample points per analysis. Burn-in values were
determined in Tracer 1.5 by checking for convergence of two
independent analyses as well as the stabilization of the model
parameters. The first 5,000 samples were discarded and a
consensus tree with posterior probability values was obtained
in TreeAnnotator 1.5.3 by pooling trees from the three
independent runs (both extensions of BEAST 1.5.3, see below).
Because of congruence between trees obtained by using
maximum likelihood and Bayesian methods, we chose to use
and date the tree generated with Bayesian methods for
consistency with the other groups in this study.
The relaxed Bayesian clock method (Drummond et al.,
2006) was used to generate a calibrated tree using BEAST 1.5.3
(Drummond & Rambaut, 2007). As for the previous analyses,
the GTR + G + I model was used. Three fossil constraints
were introduced on the tree and prior distributions followed a
log-normal distribution with mean taken from the fossil
record and variance adjusted to include the uncertainty associ-
ated with the fossil. The first constraint (corresponding to the
last shared ancestor of the Chiroptera) was set to 65 � 5 MYA,
at or following the Cretaceous-Tertiary boundary (Eick et al.,
2005; Teeling et al., 2005). The second fossil constraint, corre-
sponding to the last known common ancestor shared by the
families Rhinolophidae and Hipposideridae was set to 46 � 5
MYA (Eick et al., 2005; Teeling et al., 2005) and a third was set
to 75 � 10 MYA for the carnivore–pangolin split (Mckenna &
Bell, 1997).
Cicadidae: Platypleurini: Platypleura. A tree for this group
was based on the larger phylogeny of the cicada tribe
Platypleurini (Price, 2010). From this dataset, 26 South African
species of cicadas in the genus Platypleura (Amyot & Serville
1842) were selected for a species level phylogeny based on
two mitochondrial markers (COI and 16S) and one nuclear
marker (EF-1a), with Albanycada albigera (Walker 1850) used as
an outgroup.
Bayesian Inference analyses were conducted using MrBayes
v.3.1.2 (Huelsenbeck & Ronquist, 2001) under the
GTR + I + G model, carried out on the University of Oslo Bio-
portal facility cluster (www.bioportal.uio.no). The analysis
comprised four independent runs of 10 million generations,
using random starting trees with four chains (one cold, three
hot), sampling every 1000 generations with an initial branch
length prior set, using the ‘brlenspr = Unconstrained : Expo-
nential (100.0)’ command. Stationarity in each analysis was
assessed using the potential scale reduction factor (PSRF) data
and plots of likelihood scores, tree length, and average stan-
dard deviation of split frequencies against generation. Only
trees generated at stationarity were used to calculate the
posterior probabilities.
PD calculation
Although many measures of PD exist (Schweiger et al., 2008;
King, 2009; Vamosi et al., 2009; Magnuson-Ford et al., 2010;
Morlon et al., 2011; Pio et al., 2011), we used rooted PD (Rodri-
gues & Gaston, 2002), as it was recently shown to be one of
the most appropriate ways of accounting for evolutionary his-
tory and relatedness between taxa in a conservation context
(Pio et al., 2011). First, rooted PD (hereafter referred to simply
as PD) was calculated for the entire tree.
Global comparison of predicted vs. random remaining PD
In this analysis predicted and random remaining PD were
compared by sequentially pruning phylogenetic trees and cal-
culating remaining PD after each pruning event. Using the
results from present and future (2080) distribution models,
species were pruned from each tree according to the percent-
age of overall occupancy predicted to be lost over time. The
species predicted to undergo the largest loss in occupancy
from present to future distribution was dropped from the tree
in the first pruning event, the two species predicted to
undergo the largest and second largest losses were dropped
from the tree in the second pruning event and so on. After
each pruning event, the remaining PD was calculated and
compared to the PD remaining from pruning events where
the same number of randomly selected species were dropped.
10′000 replicates of random remaining PD were computed for
each pruning event, allowing us to obtain a P-value of
predicted PD as compared to the distribution of the random
replicates.
Spatially explicit comparison of predicted PD vs. randomPD
The relationship between predicted and random PD was
further investigated by spatially quantifying the difference
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
CLIMATE CHANGE & PHYLOGENETIC DIVERSITY LOSS 5
Page 6
between predicted remaining PD and random remaining PD
in each of the study area’s grid cells. First, species richness
and PD were computed in each grid cell under current and
future (A1 2080) climatic conditions. Then, for each grid cell,
which had at least 2 species predicted to be present under
future climatic conditions, and for which a decrease in species
richness is projected under our climate change scenario, a
number of randomly selected species equal to the difference
between the number of species predicted under current and
future climatic conditions was discarded. PD was then com-
puted for the remaining species (random remaining PD). This
operation was replicated 1,000 times for each grid cell. Finally,
we computed the mean PD of the 1,000 replicates, as well as
the 5% quantile and the P -value of predicted PD as compared
to the distribution of the 1,000 random replicates.
Results
Species distribution models
Models were fitted for 167 species of Proteaceae (of
330; 51%), 147 species of Restionaceae (of 297; 50%),
50 species of Chiroptera (of 75; 67%), and 26 species
of Platypleura (44; 59%). These models allowed pre-
dicting the potential distributions of all species under
present and future (year 2080) climatic conditions
(Fig. 1). Most models demonstrated robust predictive
power with a mean AUC of 0.98 and 0.99 and a
mean TSS value of 0.98 and 0.99, for majority rule
and average methods, respectively (evaluations val-
ues are averages obtained from the split-sample pro-
cedure). Less than 2% of models were rejected based
on insufficient predictive power. Since the AUC and
TSS-based projections with both the majority and
average methods were highly correlated (r > 0.98 in
all groups), hereafter we present only results from
the ‘TSS majority’ method. In addition, size of cur-
rent species distributions was found to be a good
predictor for both size of future distributions (R2 of
0.17, 0.26, 0.07 and 0.26 with respective P -values of
<0.0001, <0.0001, 0.045 and <0.001 for Proteaceae,
Restionaceae, Chiroptera, and Platypleura, respec-
tively) and size of the overlap between present and
future distributions (R2 of 0.23, 0.47, 0.39, 0.75 with
P -values of <0.001, <0.0001, <0.0001, and <0.0001 for
Chiroptera, Platypleura, Restionaceae and Proteaceae
respectively.
Phylogenetic tree imbalance and stemminess
Platypleura and Restionaceae communities were rela-
tively balanced and stemmy, while Restionaceae and
Chiroptera were relatively tippy in comparison (Protea-
ceae: 5.62; 11.32; Restionaceae: 4.77; �3.19; Platypleura:
1.24; 4.13, Chiroptera: �0.43; �3.18).
Global comparison of predicted vs. random remaining PD
Forty-seven, 24, 145, and 167 comparisons between pre-
dicted and random remaining PD (for Chiroptera,
Platypleura, Restionaceae, and Proteaceae, respectively)
were computed from sequential pruning events
(Fig. 2). Two of the comparisons for Chiroptera, three
for Restionaceae and 22 for Proteaceae (none for Platy-
pleura) resulted in predicted remaining PD being signif-
icantly lower than random remaining PD. Although not
significant, predicted remaining PD in Proteaceae is
higher than random as long as 38% of species are still
on the tree. After this threshold and as species continue
to be pruned predicted remaining PD becomes lower
than random. In Restionaceae, predicted remaining PD
is higher than random while the first 15% of species are
pruned (as well as between 25% and 69%) and mostly
lower than random after 69% of species have been
pruned. In Chiroptera predicted remaining PD is on
the contrary, lower until 65% of the species have been
pruned and mostly higher thereafter. In Cicadas,
predicted remaining PD is always slightly higher than
random.
Spatially explicit comparison of predicted PD vs. randomPD
Summary statistics across grid cells revealed that
P -values of predicted PD as compared to the distribu-
tion of the 1,000 random replicates were not correlated
with numbers of species present under current climatic
conditions, numbers of species predicted under future
conditions, PD values under current climatic condi-
tions, or PD values under future climatic conditions.
Grid cells which underwent significantly higher PD
losses as a result of predicted local extinctions as
opposed to randomly simulated local extinctions repre-
sented 12%, 13%, 17%, and 6% for Chiroptera, Platyple-
ura, Proteaceae, and Restionaceae respectively (Fig. 3).
Discussion
This study predicts that high concentrations of PD for
Proteaceae, Restionaceae, Chiroptera, and Platypleura
will contract considerably within the next 70 years but
that climate change will not result in higher nor lower
PD losses, on average, compared to random extinction
simulations (Fig. 1).
Our results for the Proteaceae are consistent with cli-
mate change simulations previously carried out in a
study, which assessed the relative potential impacts of
future land use and climate change on the threat status
of this plant family (Bomhard et al., 2005). However,
we found that additional species are predicted to
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
6 D. V. PIO et al.
Page 7
(a)
(b)
Fig. 1 Predicted phylogenetic diversity (PD) and normalized discrepancy [(PD minus species richness (SR)] patterns are represented
for present and future (2080) climatic conditions in four diverse animal and plant groups: Proteaceae (a), Restionaceae (b), Chiroptera
(c), and Platypleura (d). Highest values (in red) denote larger amounts of present and future PD or larger normalized discrepancies.
More specifically, in the latter case, red grid cells denote areas where values for PD are higher than for SR (species present account for a
disproportionately high amount of PD) while blue grid cells denote areas where SR is higher than PD (species present account for a
comparatively low amount of PD).
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
CLIMATE CHANGE & PHYLOGENETIC DIVERSITY LOSS 7
Page 8
(c)
(d)
Fig. 1 continued
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
8 D. V. PIO et al.
Page 9
become extinct. This is because our simulations extend
for a further 60 years into the future. Our results for
Chiroptera also concur with a study evaluating
whether national parks met their mandate under cli-
mate change and land transformation scenarios by
assessing the sensitivity of 277 (nonvolant) mammals
on the African continent (Thuiller et al., 2006). This
study found that the Kalahari will likely lose the high-
est number of mammals on the African continent (Thu-
iller et al., 2006); similarly our results predicted this
region to become very barren in terms of future surviv-
ing bat PD.
Extinctions have received attention from many
researchers because they result in both a shift in com-
munity structure and interspecies interactions (Wil-
liams & Jackson, 2007; Stralberg et al., 2009), and the
deletion of portions of the history of life (Purvis et al.,
2000). An already existing substantial body of literature
tries to assess whether present species’ extinction risk is
nonrandom, exhibiting patterns of phylogenetic clump-
ing (Bennett & Owens, 1997; Purvis et al., 2000; Sakai
et al., 2002; Pilgrim et al., 2004; Sjostrom & Gross, 2006;
Faith, 2008; Vamosi & Wilson, 2008). Studying how
extinctions are distributed phylogenetically ultimately
allows us to investigate what portions of evolutionary
history may be eliminated and consequently how eco-
systems and their functioning may be affected.
Although evidence that PD constitutes a meaningful
proxy for functional diversity and evolutionary poten-
tial is still scarce, this body of literature is growing rap-
idly (Cadotte et al., 2009; Flynn et al., 2011; Cadotte,
2013) and a recent study sets out clear future research
priorities to create a solid conceptual basis for the use
and relevance of PD in conservation (Winter et al.,
2012).
Fig. 2 Predicted and mean random remaining phylogenetic
diversity (PD) for Proteaceae, Restionaceae, Platypleura, and
Chiroptera are plotted. Species predicted to undergo the largest
losses in total occupancy from present to future distributions
are dropped from their phylogenetic trees sequentially, one at a
time, until only two species are left (the minimum required to
calculate PD).
Fig. 3 All mapped grid cells represent areas that undergo local (only from that specific grid cell) extinctions as a result of climate
change for 4 groups of plants and animals: Restionaceae, Proteaceae, Platypleura, and Chiroptera. Red cells represent areas where larger
amounts of phylogenetic diversity (PD) are predicted to be lost through climate change induced local extinctions as compared to ran-
dom expectations, while black cells represent areas where there is no difference between predicted PD losses and those resulting from
random simulations.
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
CLIMATE CHANGE & PHYLOGENETIC DIVERSITY LOSS 9
Page 10
Several studies on birds and nonvolant mammals,
but more recently also on angiosperm families, have
shown evidence for nonrandom extinction processes
(Bennett & Owens, 1997; Russell et al., 1998; Purvis
et al., 2000; Sjostrom & Gross, 2006; Vamosi & Wilson,
2008). Threat in birds for instance seems negatively cor-
related with body size and relative fecundity, which
may have predisposed certain lineages to extinction
(Bennett & Owens, 1997). Studies on angiosperms
found that elevated risk was found in smaller clades
(Vamosi & Wilson, 2008). However, other studies on
regional and global floral datasets found contrasting or
mixed evidence for nonrandom threat or rarity patterns
(Schwartz & Simberloff, 2001; Webb & Pitman, 2002;
Pilgrim et al., 2004; Davies et al., 2011; Thuiller et al.,
2011).
Some of the studies that found evidence for nonran-
dom extinction, have pointed out that once phylogenet-
ically distributed traits have already mediated
considerable extinction, then many monotypic genera
or families might be the last survivors of once-larger
clades and cause disproportionate PD loss (Purvis et al.,
2000). Nonrandom patterns may then be likely if a large
percentage of each group has already been lost. In our
simulations, however, both at the landscape and at the
grid-cell level, there was no evidence of PD losses
depending on the number of species already pruned
from a tree.
Using detailed phylogenetic information, a recent
study shows that plant extinctions from the Cape
Floristic Region may result in little loss of evolutionary
history, and that this may be true because biodiversity
hotspots such as the Cape Floristic Region are the
product of recent speciation (Davies et al., 2011). The
authors further argue that the processes driving extinc-
tion in relict lineages are different from those for young
diversifying lineages and that their findings may be
particularly relevant to recently radiated plant groups.
Although focusing on the same region and some of the
same groups, there are several important differences
which prevent direct comparison of these interesting
findings with the present study. Firstly, Davies et al.
(2011) are investigating extinction as a measure of
accumulated threats, not specifically climate change.
Second, the present study includes an animal group
(Chiroptera) for which Southern Africa is not the center
of speciation.
In the majority of the published literature, extinction
is indeed to be simulated through a threat status as a
measure of accumulated threats and most often
considers drastic land-use changes. Gradually changing
ecosystems, however, (for example, those that change
in response to climate change) may show different
patterns from systems subject to abrupt anthropogenic
use (Winter & Schweiger, 2011). Climate-specific effects
on the tree of life are the focus of this study and have
only been investigated once before (Thuiller et al.,
2011). Using European plants, birds, and mammals,
Thuiller et al. (2011) showed that PD loss predicted to
happen through climate change is neither higher nor
lower than expected by chance (Thuiller et al., 2011).
The present study provides similar evidence from four
Southern African groups. Simulated extinction events
based on sequential pruning of species predicted to
undergo range contractions in this study did not result
in higher (nor lower) PD losses when compared to ran-
dom simulations, except in a small number of cases. A
possible explanation of this pattern is that a species’
range, the trait which places species at risk of extinction
from climate change in this study, is a trait with very
weak or no phylogenetic signal (Carotenuto et al.,
2010).
With the ranges of most species in this study being
reduced in response to climatic changes, the degree of
habitat connectivity will likely play an important role
in the maintenance of diversity. Because of the limita-
tions intrinsic to predicting range contractions,
increased threat, and indeed to incomplete sampling,
this study does not claim specific conservation recom-
mendations for particular areas or taxonomic groups in
Southern Africa. Moreover, we do not investigate the
entire tree of life, and some species may occur outside
the study area or may shift their ranges away from the
study area. Thus, a loss within this region must not nec-
essarily be considered a total loss. What this study
does, however, is to suggest likely trends in PD loss
within a diverse groups of organisms. Predicting cli-
mate change induced effects on the tree of life globally
certainly requires further study to tease apart how dif-
ferent taxa and geographical areas may be affected but
the evidence accumulated so far suggests that focusing
resources on climate threatened species alone may not
result in disproportionate benefits for the preservation
of evolutionary history.
Acknowledgements
We thank the hundreds of volunteers who collected the datathat make up the Protea Atlas. We are indebted to Dr RichardGrenyer for comments on an earlier draft. The feedbackreceived by three anonymous referees and Dr Oliver Schweigerallowed us to make exceptional improvements on an earlier ver-sion of this manuscript. DP is funded by the Marie-Curie EarlyStage Researcher (ESR) Fellowship as part of the EU-Hotspotsproject (http://www.kew.org/hotspots) to AG and NS. NS isfunded by the Swiss National Science Foundation grant no3100A0-116412. BWP is funded by the US National ScienceFoundation grant no DEB-0955849. MHV and BWP is funded bythe National Research Foundation of South Africa grant no2069059. We thank the Swiss Institute of Bioinformatics for
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
10 D. V. PIO et al.
Page 11
access to Vital-IT, their high performance computer center. Col-lecting and georeferencing the Restionaceae specimen dataset,and the construction of the Restionaceae phylogeny, wasfunded by SNF 31-66594-01 to H.P. Linder.
References
Akaike H (1974) A new look at statistical model identification. IEEE Transactions on
Automatic Control, 19, 716–722.
Allouche O, Tsoar A, Kadmon R (2006) Assessing the accuracy of species distribution
models: prevalence, kappa and the true skill statistic (TSS). Journal of Applied
Ecology, 43, 1223–1232.
Bennett PM, Owens IPF (1997) Variation in extinction risk among birds: chance or
evolutionary predisposition? Proceedings of the Royal Society of London Series B-Bio-
logical Sciences, 264, 401–408.
Bomhard B, Richardson DM, Donaldson JS et al. (2005) Potential impacts of future
land use and climate change on the red list status of the Proteaceae in the Cape
Floristic Region, South Africa. Global Change Biology, 11, 1452–1468.
Breiman L (2001) Random forests. Machine Learning, 45, 5–32.
Breiman L, Friedman JH, Olshen RA, Stone CJ (1984) Classification and Regression
Trees. Wadsworth, Belmont, CA.
Broennimann O, Thuiller W, Hughes G, Midgley GF, Alkemade JMR, Guisan A
(2006) Do geographic distribution, niche property and life form explain plants’
vulnerability to global change? Global Change Biology, 12, 1079–1093.
Bromham L, Lanfear R, Cassey P, Gibb G, Cardillo M (2012) Reconstructing past spe-
cies assemblages reveals the changing patterns and drivers of extinction through
time. Proceedings of the Royal Society of London Series B-Biological Sciences, 279,
4024–4032.
Cadotte MW (2013) Experimental evidence that evolutionarily diverse assemblages
result in higher productivity. Proceedings of the National Academy of Sciences of the
United States of America, 110, 8996–9000.
Cadotte MW, Davies TJ (2010) Rarest of the rare: advances in combining evolutionary
distinctiveness and scarcity to inform conservation at biogeographical scales.
Diversity and Distributions, 16, 376–385.
Cadotte MW, Cavender-Bares J, Tilman D, Oakley TH (2009) Using phylogenetic,
functional and trait diversity to understand patterns of plant community produc-
tivity. Public Library of Science ONE, 4, e5695.
Cadotte MW, Dinnage R, Tilman D (2012) Phylogenetic diversity promotes ecosystem
stability. Ecology, 93, 223–233.
Campbell B (1996) The Miombo in Transition: Woodlands and Welfare in Africa. CIFOR,
Jakarta.
Carotenuto F, Barbera C, Raia P (2010) Occupancy, range size, and phylogeny in eur-
asian pliocene to recent large mammals. Paleobiology, 36, 399–414.
Chen IC, Hill JK, Ohlemuller R, Roy DB, Thomas CD (2011) Rapid range shifts of spe-
cies associated with high levels of climate warming. Science, 333, 1024–1026.
Davies TJ, Fritz SA, Grenyer R et al. (2008) Phylogenetic trees and the future of mam-
malian biodiversity. Proceedings of the National Academy of Sciences of the United
States of America, 105, 11556–11563.
Davies TJ, Smith GF, Bellstedt DU et al. (2011) Extinction risk and diversification are
linked in a plant biodiversity hotspot. Public Library of Science Biology, 9, e1000620.
Davies TJ, Cooper N, Diniz-Filho JAF, Thomas GH, Meiri S (2012) Using phylogenetic
trees to test for character displacement: a model and an example from a desert
mammal community. Ecology, 93, S44–S51.
Drummond AJ, Rambaut A (2007) BEAST: bayesian evolutionary analysis by sam-
pling trees. BMC Evolutionary Biology, 7, 214–222.
Drummond AJ, Ho SYW, Phillips MJ, Rambaut A (2006) Relaxed phylogenetics and
dating with confidence. Public Library of Science Biology, 4, 699–710.
Du Toit JT, Cumming DHM (1999) Functional significance of ungulate diversity in
African savannas and the ecological implications of the spread of pastoralism.
Biodiversity & Conservation, 8, 1643–1661.
Eick GN, Jacobs DS, Matthee CA (2005) A nuclear DNA phylogenetic perspective on
the evolution of echolocation and historical biogeography of extant bats (Chirop-
tera). Molecular Biology and Evolution, 22, 1869–1886.
Elith J, Graham CH, Anderson RP et al. (2006) Novel methods improve prediction of
species’ distributions from occurrence data. Ecography, 29, 129–151.
Engler R, Engler R, Randin CR et al. (2011) 21st century climate change threatens
mountain flora unequally across Europe. Global Change Biology, 17, 2330–2341.
Faith DP (2008) Threatened species and the potential loss of phylogenetic diversity:
conservation scenarios based on estimated extinction probabilities and phyloge-
netic risk analysis. Conservation Biology, 22, 1461–1470.
Flynn DFB, Mirotchnick Jain M, Palmer MI, Naeem S (2011) Functional diversity and
phylogenetic diversity as predictors of biodiversity-ecosystem-functional relation-
ships. Ecology, 92, 1573–1581.
Friedman JH (1991) Multivariate adaptive regression splines. Annals of Statistics, 19,
1–67.
Gedan KB, Bertness MD (2009) Experimental warming causes rapid loss of plant
diversity in New England salt marshes. Ecology Letters, 12, 842–848.
Guindon S, Gascuel O (2003) A simple, fast, and accurate algorithm to estimate large
phylogenies by maximum likelihood. Systematic Biology, 52, 696–704.
Hall TA (1999) BioEdit: a user-friendly biological sequence alignment editor and
analysis program for windows 95/98/NT. Nucleic Acids Symposium Series, 41,
95–98.
Hanley JA, Mcneil BJ (1982) The meaning and use of the area under a receiver operat-
ing characteristic (roc) curve. Radiology, 143, 29–36.
Hardy CR, Moline P, Linder HP (2008) A phylogeny for the African Restionaceae and
new perspectives on morphology’s role in generating complete species phyloge-
nies for large clades. International Journal of Plant Sciences, 169, 377–390.
Hastie T, Tibshirani R (1986) Generalized additive models. Statistical Science, 1,
297–318.
Heard SB (1992) Patterns in tree balance among cladistic, phenetic, and randomly
generated phylogenetic trees. Evolution, 46, 1818–1826.
Huelsenbeck J, Ronquist F (2001) MrBayes: bayesian inference of phylogenetic trees.
Bioinformatics, 17, 754–755.
Huelsenbeck JP, Ronquist F, Nielsen R, Bollback JP (2001) Bayesian inference of phy-
logeny and its impact on evolutionary biology. Science, 294, 2310–2314.
Kembel SW, Eisen JA, Pollard KS, Green JL (2011) The phylogenetic diversity of
metagenomes. PLoS One, 6, e23214.
King I (2009) The need for the incorporation of phylogeny in the measurement of bio-
logical diversity, with special reference to ecosystem functioning research. BioEs-
says, 31, 107–116.
Larkin MA, Blackshields G, Brown NP et al. (2007) Clustal w and clustal x version
2.0. Bioinformatics, 23, 2947–2948.
Liu CR, Berry PM, Dawson TP, Pearson RG (2005) Selecting thresholds of occurrence
in the prediction of species distributions. Ecography, 28, 385–393.
Magnuson-Ford K, Mooers A, Paquette SR, Steel M (2010) Comparing strategies to
preserve evolutionary diversity. Journal of Theoretical Biology, 266, 107–116.
Marmion M, Parviainen M, Luoto M, Heikkinen RK, Thuiller W (2009) Evaluation of
consensus methods in predictive species distribution modelling. Diversity and
Distributions, 15, 59–69.
Mccullagh P, Nelder JA (1989) Generalized Linear Models. Chapman & Hall, London,
UK.
Mckenna MC, Bell SK (1997) Classification of Mammals above the Species Level. Columbia
University Press, New York.
Meynard CN, Devictor V, Mouillot D, Thuiller W, Jiguet F, Mouquet N (2011) Beyond
taxonomic diversity patterns: how do a, b and c components of bird functional
and phylogenetic diversity respond to environmental gradients across France?
Global Ecology and Biogeography, 20, 893–903.
Midgley GF, Thuiller W (2007) Potential vulnerability of Namaqualand plant diver-
sity to anthropogenic climate change. Journal of Arid Environments, 70, 615–628.
Midgley GF, Hannah L, Millar D, Thuiller W, Booth A (2003) Developing regional
and species-level assessments of climate change impacts on biodiversity in the
Cape Floristic region. Biological Conservation, 112, 87–97.
Midgley GF, Hughes GO, Thuiller W, Rebelo AG (2006) Migration rate limitations on
climate change-induced range shifts in Cape Proteaceae. Diversity and Distribu-
tions, 12, 555–562.
Monadjem A, Schoeman MC, Reside A et al. (2010a) A recent inventory of the bats of
Mozambique with documentation of seven new species for the country. Acta Chir-
opterologica, 12, 371–391.
Monadjem A, Taylor PJ, Cotterill FPD, Schoeman MC (2010b) Bats Of Southern Africa:
A Biogeographic and Taxonomic Synthesis. University of the Witwatersrand, Johan-
nesburg, South Africa.
Morlon H, Schwilk DW, Bryant JA et al. (2011) Spatial patterns of phylogenetic diver-
sity. Ecology Letters, 14, 141–149.
Nakicenovic N, Swart R (2000) Emissions Scenarios: A Special Report of Working Group
III of the Intergovernmental Panel on Climate Change. Cambridge University Press,
Cambridge.
New M, Hulme M, Jones P (2000) Representing twentieth-century space-time climate
variability. part II: development of 1901-96 monthly grids of terrestrial surface
climate. Journal of Climate, 13, 2217–2238.
Paradis E, Claude J, Strimmer K (2004) APE: analyses of phylogenetics and evolution
in r language. Bioinformatics, 20, 289–290.
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
CLIMATE CHANGE & PHYLOGENETIC DIVERSITY LOSS 11
Page 12
Parmesan C, Yohe G (2003) A globally coherent fingerprint of climate change impacts
across natural systems. Nature, 421, 37–42.
Perry AL, Low PJ, Ellis JR, Reynolds JD (2005) Climate change and distribution shifts
in marine fishes. Science, 308, 1912–1915.
Pilgrim ES, Crawley MJ, Dolphin K (2004) Patterns of rarity in the native British flora.
Biological Conservation, 120, 161–170.
Pio DV, Broennimann O, Barraclough TG et al. (2011) Spatial predictions of phyloge-
netic diversity in conservation decision making. Conservation Biology, 25,
1229–1239.
Price BW(2010) Historical Biogeography of the Tribe Platypleurini Schmidt, 1918 (Hemip-
tera: Cicadidae) with a Focus on Southern Africa. unpublished PhD thesis. Rhodes
University, Grahamstown.
Price BW, Villet MH, Barker NP (2007) Patterns and processes underlying evolu-
tionary significant units in the Platypleura stridula L.species complex (Hemipter-
a: Cicadidae) in the Cape Floristic region, South Africa. Molecular Ecology, 16,
2574–2588.
Purvis A, Agapow PM, Gittleman JL, Mace GM (2000) Nonrandom extinction and the
loss of evolutionary history. Science, 288, 328–330.
Pybus OG, Harvey PH (2000) Testing macroevolutionary models using incomplete
molecular phylogenies. Proceedings of the Royal Society of London, Series B: Biological
Sciences, 267, 2267–2272.
R Development Core Team (2009) R: A Language and Environment for Statistical Com-
puting. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-
07-0, Available at: http://www.R-project.org.
Rebelo AG(2012) Protea Atlas Project.
Ridgeway G (1999) The state of boosting. Computing Science and Statistics, 31,
172–181.
Rodrigues ASL, Gaston KJ (2002) Maximising phylogenetic diversity in the selection
of networks of conservation areas. Biological Conservation, 105, 103–111.
Russell GJ, Brooks TM, Mckinney MM, Anderson CG (1998) Present and future
taxonomic selectivity in bird and mammal extinctions. Conservation Biology, 12,
1365–1376.
Sakai AK, Wagner WL, Mehrhoff LA (2002) Patterns of endangerment in the Hawai-
ian flora. Systematic Biology, 51, 276–302.
Salamin N, Chase MW, Hodkinson TR, Savolainen V (2003) Assessing internal sup-
port with large phylogenetic DNA matrices. Molecular Phylogenetics and Evolution,
27, 528–539.
Sambrook J, Fritsch E, Maniatis T (1989) Molecular Cloning: A Laboratory Manual. Cold
Spring Harbor Laboratory Press, New York.
Sanderson MJ (2002) Estimating absolute rates of molecular evolution and divergence
times: a penalized likelihood approach. Molecular Biology and Evolution, 19, 101–
109.
Sanderson MJ, Mcmahon MM (2007) Inferring angiosperm phylogeny from EST data
with widespread gene duplication. BMC Evolutionary Biology, 7 (Suppl. 1): S3.
Sanger F, Nicklen S, Coulsen AR (1977) DNA sequencing with chain-terminating
inhibitors. Proceedings of the National Academy of Sciences of the United States of Amer-
ica, 74, 5463–5467.
Sauquet H, Weston PH, Anderson CL, Barker NP, Cantrill DJ, Mast AR, Savolainen V
(2009) Contrasted patterns of hyperdiversification in Mediterranean hotspots.
Proceedings of the National Academy of Sciences of the United States of America, 106,
221–225.
Schwartz MW, Simberloff D (2001) Taxon size predicts rates of rarity in vascular
plants. Ecology Letters, 4, 464–469.
Schweiger O, Klotz S, Durka W, Kuhn I (2008) A comparative test of phylogenetic
diversity indices. Oecologia, 157, 485–495.
Sgro’ CM, Lowe AJ, Hoffman AA (2011) Building evolutionary resilience for conserv-
ing biodiversity under climate change. Evolutionary Applications, 4, 326–337.
Sjostrom A, Gross CL (2006) Life-history characters and phylogeny are correlated
with extinction risk in the Australian angiosperms. Journal of Biogeography, 33,
271–290.
Srivastava DS, Cadatte MW, MacDonald AM, Marushia RG, Mirotchnick N (2012)
Phylogenetic diversity and the functioning of ecosystems. Ecology Letters, 15,
637–648.
Stralberg D, Jongsomjit D, Howell CA, Snyder MA, Alexander JD, Wiens JA, Root TL
(2009) Re-shuffling of species with climate disruption: a no-analog future for Cali-
fornia birds? PLoS One, 4, e6825.
Teeling EC, Springer MS, Madsen O, Bates P, O’brien SJ, Murphy WJ (2005) A molec-
ular phylogeny for bats illuminates biogeography and the fossil record. Science,
307, 580–584.
Thomas CD, Cameron A, Green RE et al. (2004) Extinction risk from climate change.
Nature, 427, 145–148.
Thuiller W, Lavorel S, Araujo MB, Sykes MT, Prentice IC (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.
Thuiller W, Broennimann O, Hughes G, Alkemade JRM, Midgley GF, Corsi F (2006)
Vulnerability of African mammals to anthropogenic climate change under conser-
vative land transformation assumptions. Global Change Biology, 12, 424–440.
Thuiller W, Lafourcade B, Engler R, Ara�ujo MB (2009) BIOMOD - a platform for
ensemble forecasting of species distributions. Ecography, 32, 369–373.
Thuiller W, Lavergne S, Roquet C, Boulangeat I, Lafourcade B, Araujo MB (2011)
Consequences of climate change on the tree of life in Europe. Nature, 470, 531–534.
Valente LM, Reeves G, Schnitzler J et al. (2010) Diversification of the African genus
Protea (Proteaceae) in the cape biodiversity hotspot and beyond: equal rates in dif-
ferent biomes. Evolution, 64, 745–759.
Vamosi JC, Wilson JRU (2008) Nonrandom extinction leads to elevated loss of angio-
sperm evolutionary history. Ecology Letters, 11, 1047–1053.
Vamosi SM, Heard SB, Vamosi JC, Webb CO (2009) Emerging patterns in the compar-
ative analysis of phylogenetic community structure. Molecular Ecology, 18, 572–592.
Vetter S (2009) Drought, change and resilience in South Africa’s arid and semi-arid
rangelands. South African Journal of Science, 105, 29–33.
Webb CO, Pitman NCA (2002) Phylogenetic balance and ecological evenness. System-
atic Biology, 51, 898–907.
West AG, Dawson TE, February EC, Midgley GF, Bond WJ, Aston TL (2012) Diverse
functional responses to drought in a Mediterranean-type shrubland in South
Africa. New Phytologist, 195, 396–407.
Williams JW, Jackson ST (2007) Novel climates, no-analog communities, and ecologi-
cal surprises. Frontiers in Ecology and the Environment, 5, 475–482.
Willis KJ, Bhagwat SA (2009) Biodiveristy and climate change. Science, 326, 806–807.
Winter M, Schweiger O (2011) Ecosystem rates of transformation matter. Science, 333,
936–937.
Winter M, Devictor V, Schweiger O (2012) Phylogenetic and nature conservation:
where are we? Trends in Ecology and Evolution, 28, 199–204.
© 2014 John Wiley & Sons Ltd, Global Change Biology, doi: 10.1111/gcb.12524
12 D. V. PIO et al.