Phylogenetically-Informed Priorities for Amphibian Conservation

Abstract

The amphibian decline and extinction crisis demands urgent action to prevent further large numbers of species extinctions. Lists of priority species for conservation, based on a combination of species' threat status and unique contribution to phylogenetic diversity, are one tool for the direction and catalyzation of conservation action. We describe the construction of a near-complete species-level phylogeny of 5713 amphibian species, which we use to create a list of evolutionarily distinct and globally endangered species (EDGE list) for the entire class Amphibia. We present sensitivity analyses to test the robustness of our priority list to uncertainty in species' phylogenetic position and threat status. We find that both sources of uncertainty have only minor impacts on our 'top 100' list of priority species, indicating the robustness of the approach. By contrast, our analyses suggest that a large number of Data Deficient species are likely to be high priorities for conservation action from the perspective of their contribution to the evolutionary history.

Full text

Phylogenetically-Informed Priorities for Amphibian
Conservation
Nick J. B. Isaac
1
, David W. Redding
2
, Helen M. Meredith
3
, Kamran Safi
4,5
*
1Natural Environment Research Council Centre for Ecology and Hydrology, Maclean Building, Wallingford, Oxfordshire, United Kingdom, 2University College London,
London, United Kingdom, 3Durrell Institute of Conservation and Ecology and Institute of Zoology, Zoological Society of London, London, United Kingdom, 4Max Planck
Institute for Ornithology, Department for Migration and Immuno-ecology, Radolfzell, Germany, 5Department of Biology, University of Konstanz, Konstanz, Germany
Abstract
The amphibian decline and extinction crisis demands urgent action to prevent further large numbers of species extinctions.
Lists of priority species for conservation, based on a combination of species’ threat status and unique contribution to
phylogenetic diversity, are one tool for the direction and catalyzation of conservation action. We describe the construction
of a near-complete species-level phylogeny of 5713 amphibian species, which we use to create a list of evolutionarily
distinct and globally endangered species (EDGE list) for the entire class Amphibia. We present sensitivity analyses to test the
robustness of our priority list to uncertainty in species’ phylogenetic position and threat status. We find that both sources of
uncertainty have only minor impacts on our ‘top 100‘ list of priority species, indicating the robustness of the approach. By
contrast, our analyses suggest that a large number of Data Deficient species are likely to be high priorities for conservation
action from the perspective of their contribution to the evolutionary history.
Citation: Isaac NJB, Redding DW, Meredith HM, Safi K (2012) Phylogenetically-Informed Priorities for Amphibian Conservation. PLoS ONE 7(8): e43912.
doi:10.1371/journal.pone.0043912
Editor: Matthew Charles Fisher, Imperial College Faculty of Medicine, United Kingdom
Received April 12, 2012; Accepted July 26, 2012; Published August 30, 2012
Copyright: ß2012 Isaac et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: No current external funding sources for this study.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: ksafi@orn.mpg.de
Introduction
The current biodiversity crisis demands pragmatic triage
solutions. Lists of priority species are an important tool for the
effective allocation of scarce conservation resources. Such lists are
typically dominated, at the national and global scales, by species of
high conservation concern, usually those in the Endangered and
Critically Endangered categories of the IUCN Red List. Increas-
ingly however, the notion that species’ contribution to phyloge-
netic diversity should also be considered, has been gaining traction
[1–5].
Amphibians are in the grip of an unprecedented extinction crisis
[6]. One third of species are listed as threatened and a quarter are
categorised as Data Deficient. Around 43% of species are
considered to be in decline [7]. Large scale declines have occurred
over the last few decades [8], and future decades are expected to
see the extinction of many hundreds of species [9,10]. The
amphibian extinction crisis has been attributed variously to habitat
loss and fragmentation [11], disease [12,13], environmental
contamination [14], overexploitation [15], introduced species
[16], climate change [17,18], and interactions between multiple
threats [19–24].
Faced with this crisis, a set of conservation priorities for
amphibian species is urgently needed. At present, only the three
IUCN categories of extinction risk can distinguish among the
approximately 2000 threatened species, of which over 400 are
Critically Endangered. In this paper, we generate a set of global
priorities for amphibian conservation based both on threat status
and phylogenetic position using the currently available data. We
show that a working hypothesis for the species level phylogeny of
the entire class of nearly 6000 species can be generated from a
small number of synthetic sources, namely a cladogram of higher
taxa and an authoritative taxonomy. We calculate species
‘evolutionary distinctiveness’ (ED) scores based on this phylogeny,
and combine them with categories of extinction risk to generate an
‘EDGE’ list for all amphibians. We present sensitivity analyses to
test the robustness of our priority list to uncertainty in both sources
of data used to compile them: the branching structure of the
phylogeny and the categorization of species’ extinction risk. We
also explore the impact of different choices about the way in which
EDGE scores are generated from the combination of phylogenetic
and extinction risk assessment data.
Materials and Methods
Our phylogeny is largely based on three sources: the amphibian
‘tree of life’ described by Frost et al. [25], the species-level
taxonomy of Amphibian Species of the World (ASW) [26], and the
molecular phylogeny of Roelants et al. [27]. Species’ extinction
risk categories were extracted from the Global Amphibian
Assessment (GAA) [6]. In cases where the species taxonomy of
the GAA deviated from that of the ASW, we treated the ASW as
authoritative.
Our general aim was to produce a phylogeny that was both
maximally inclusive (i.e. containing nearly all amphibian species)
and maximally resolved (given the available data). Achieving this
goal necessitated a number of ad hoc decisions about the
placement of certain species and the precise nature of the
branching patterns, and for many clades the desire for inclusivity
was in conflict with the desire for phylogenetic resolution. For this
PLOS ONE | www.plosone.org 1 August 2012 | Volume 7 | Issue 8 | e43912

reason, we designed some of our analyses to address directly the
issues around uncertainty in the phylogenetic position of large
numbers of species.
Higher-level Topology
The primary source of topological information was the
amphibian ‘tree of life’ described by Frost et al. [25] in a large
monograph. The phylogeny, depicted in their figure 50, is based
on both morphological and molecular data: it contains 526 tips,
most of which correspond to amphibian genera, and is almost fully
resolved, containing 522 internal nodes.
We pruned the Frost et al.’s [25] ‘tree of life’, to produce a
‘higher taxon tree’ to which assignment of ASW species would be
relatively uncontroversial. A total of 169 tips were pruned. This
includes the speciose genus Litoria, of which the Frost et al.
phylogeny includes just 10 out of 162 species: our higher taxon
tree contains just a single tip for the entire genus. Likewise, about
1/3 of the 169 pruned tips were in the speciose families Ranidae
and Bufonidae.
We then added a 23 additional clades that were not included in
Frost et al. [25]. From the ‘Comments’ field in ASW we placed
Chiropterotriton,Crossodactyloides,Cynops,Frostius,Kurixalus,Leptobran-
chella,Salamandrina,Spelaeophryne,Zachaenus, and the Leptodactylus
pentadactylus and Triturus vulgaris groups. From Roelants et al. [27]
we placed Caudacaecilia,Glyphoglossus,Hylophorbus,Luetkenotyphlus,
Microcaecilia,Praslinia,Proteus and Xenorhina. Finally, we placed
Onychodactylus and Protohynobius from Zhang et al. [28], Itapotihyla,
Megastomatohyla and Tepuihyla from Faivovich et al. [29] and
Barygenys from Van Bocxlaer [30].
Species-level Topology
We assigned each species in ASW [26] to each one of these
higher taxa. In most cases, this was straightforward because the
tips of the higher taxon tree were mostly at genus level. Generally,
we used a star phylogeny i.e. an unresolved multifurcating tree for
species within higher taxa. For genera containing subgenera or
‘species group’ names in ASW, we treated these taxonomic units as
monophyletic clades, thus providing extra resolution. However,
this introduced problems for some large genera in which not all
species have been assigned membership to any subgenus or species
group. We decided assignment to a genus under ASW represented
valid phylogenetic information, so we sought ways to include these
‘orphan species’ without losing the additional resolution provided
by this additional information. Our approach depended on the
size of the genus and the number of intra-genus clades. For the
large genera Philautus (145 species) and Platymantis (55 species),
both of which contain species groups that include around two
thirds of their species complement, we assigned the remaining
third to an ‘orphan’ clade within each genus. For 163 species in 18
genera where the proportion of orphans was relatively small, we
assigned the orphans to species groups at random. This included
members of Eleutherodactylus (n = 89 orphans out of 483 species),
Rhacophorus (24/70) and Xenopus (7/16).
For some taxa, material in ASW indicated that phylogenetic
data was available to add further resolution. In some cases this was
a simple observation of relatedness, e.g. ‘probable sister species’; in
other cases it referred to an external study on the phylogeny of the
group in question. We used all such information where available,
combined with species group assignments (described above). For
example, we used Emerson et al.’s [31] phylogeny of Limnonectes to
generate resolution within species groups, for a total of 17
subgeneric clades: 24 species were assigned to one of these clades
with confidence, 16 species were assigned to a random clade
within known species group, and 10 were assigned completely at
random.
Just three out of 382 higher taxa represent taxonomic units
above the genus. These were the clades defined by the following
species in Frost et al. [25]: Argenteohyla siemersi,Hamptophryne boliviana
and Phyllomedusa vaillantii. We used ASW to determine which
genera were likely close relatives, often based on their status in
previous taxonomic monographs. We then treated these genera as
monophyletic within the suprageneric tip, and assigned species to
them as described above.
A total of 5713 species were assigned to higher taxa,
representing around 97% of valid extant amphibian species and
only 153 species could not be assigned to any of the higher taxa.
Dating the Phylogeny
The ages of deep nodes come from Roelants et al. [27] who
presented a molecular phylogeny of 171 amphibian species.
Specifically, we used the version of Roelants et al.’s tree that was
constrained to be compatible with Frost et al’s [25] tree of life
(figure 3 in Roelants et al. [27]). Node ages below Roelants et al.
were derived by assuming a ‘pure-birth model’ of cladogenesis
(following [32,33]). The pure-birth model is a popular null model
of evolutionary diversification (e.g. [34–36]) and is based on a
Markov process. Specifically, it estimates the age of a node as T *
ln(a)/ln(b), where T is the age of the parent node and a and b are
the number of species descended from the focal node and the
parent node, respectively [32]. The full composite phylogeny can
be found as supporting information online (Phylogeny S1).
Evolutionary Distinctiveness and EDGE Scores
We estimated species’ contribution to phylogenetic diversity
using the ‘Evolutionary Distinctiveness’ (ED) algorithm described
by Isaac et al. [37], with a modification to the way in which scores
were corrected for polytomies (nodes with .2 descendents). Isaac
et al. used a statistical fit to simulated data in order to correct the
ED scores of branches descended from polytomies. This correction
factor decreases to zero for nodes with large numbers (.20) of
descendants, leading to an underestimate of the ED of many
species in poorly-resolved areas of the phylogeny (i.e. most species
in our amphibian phylogeny). Instead, we used a ‘pure birth
model’ of cladogenesis to derive a correction factor based on the
expected (i.e. mean) ED, given all the possible resolutions of the
polytomy [38]. This empirical correction factor yields ED scores
that are almost identical to those derived from a recently-
developed Bayesian method for resolving polytomies in dated
phylogenies [4,39].
We calculated ED scores for each amphibian species using the
caper package [40] in R [41]. Using the ‘EDGE’ algorithm
previously used for mammals [4,37], we combined these values
with the extinction risk scores taken from Global Amphibian
Assessment [6] to create our reference EDGE scores (figure 1).
Data deficient species were excluded from this analysis. We
created a further ‘candidate’ list of data deficient high ED scoring
species (in the top 5% of ED scores) as targets for future threat
assessment.
The ‘EDGE algorithm’ of Isaac et al. [37] is not the only way to
combine ED scores with extinction risk categories, and the issue of
how to convert these categories into an ordinal scale remains an
issue [42,43]. The EDGE algorithm treats each category as a
quasi-probability in which each step is associated with increasing
the extinction risk by a factor of two. The main alternative is the
‘expected loss’ (EL: [44]) algorithm, which is based on the actual
probability of extinction over 100 years, using values of 0.1%, 1%,
10%, 67%, 99.9% for categories LC, NT, VU, EN and CR
Amphibian Conservation Prioritization
PLOS ONE | www.plosone.org 2 August 2012 | Volume 7 | Issue 8 | e43912

respectively, thereby giving much higher weight to CR and EN
categories (compared with the EDGE approach). We compared
the makeup of the top 100 species produced by both methods, and
several variants thereof. One variant, named ‘IUCN500’ [42], is a
modification of the EL approach but with extinction probabilities
estimated over a much longer time period (i.e. 500 years), with
probabilities of 0.5%, 5%, 39%, 99.6%, 100% [39]. The other two
are variants on the EDGE calculation of Isaac et al. [37], in which
extinction risk increases by 1.25 fold and 5 fold respectively, for
each increase in threat categories. For each of these five methods,
we expressed the makeup of the list as the running mean ED score
of the top n ranked EDGE species, for all values of n from 1:100.
We compared these five empirical distributions two extreme
selection criteria, one based solely on ED, the other selecting first
all CR species then EN, in decreasing order of ED. Ideally, we
would like a distribution that falls midway between these two
extremes.
Analyses and Simulations
We tested how uncertainty in the underlying data could affect
the species chosen for conservation attention by the EDGE listing
process. We examined the robustness of our priority list calculated
using the standard EDGE algorithm, to four specific forms of
uncertainty: a) the placement of species on the phylogeny (‘ED
errors’), b) changes to species’ Red List status (‘GE errors’), c)
future reassignment of species currently listed as Data Deficient
(DD) and d) a sensitivity test varying the number of species for
which there were errors in the data (i.e. 2% of the species have ED
or GE errors compared to 25%).
We refer to the ED scores, extinction risk estimates and EDGE
scores described above as the ‘unmodified’ or ‘reference’ sets. For
each perturbation scenario (described in detail below), we
generated 1000 replicate datasets at each level of perturbation
and calculated EDGE scores for all species in each dataset. Given
that the EDGE listing process has been previously been used to
choose the top 100 ranked species to target for conservation
Figure 1. Species level phylogeny of 4339 amphibian species, colour-coded by species’ EDGE scores. Data Deficient and Extinct
species have been omitted.
doi:10.1371/journal.pone.0043912.g001
Amphibian Conservation Prioritization
PLOS ONE | www.plosone.org 3 August 2012 | Volume 7 | Issue 8 | e43912

attention [45], we used the similarity (i.e. the proportion of shared
species) of the top 100 species as our overall measure of effect size.
a) ED errors: perturbing the phylogeny. The phylogeny is
assembled in an ad hoc manner, so we wanted to be sure that our
conclusions were robust to incorrect assignments. Assuming that
most amphibian genera are monophyletic, the main errors in our
phylogeny derive from treating subgeneric entities as clades, and
the placement of species among these entities (see above). We
simulated the impact of these decisions by altering the topology of
the tree at random 1000 times. At each simulation, we selected
10% of species at random and moved them each to a different, but
closely-related, clade. The severity of incorrect assignments was
tested by sequential analyses moving another 1000 sets of
randomly selected 10% of all species one, two, four, eight and
finally sixteen clades away.
b) GE errors: altering the categories of extinction
risk. We simulated the impact of uncertainty in each species’
extinction risk categorization. This is important because most
changes on the Red List are due to advances in knowledge, rather
than genuine changes in status [4]. For each simulation, we
selected 10% at random and then moved them either up or down
(again at random) one threat category (e.g. Vulnerable to
Endangered or Near Threatened). This process was repeated
allowing 10% of species to move two, three and then four
categories up or down. In all cases, if a Least Concern status
species was chosen to be moved down it was kept at Least Concern
and, conversely, if a Critically Endangered species was chosen to
be more severely threatened it was kept at Critically Endangered.
c) Data deficient species. Approximately a quarter of
amphibian species are categorized as Data Deficient (DD) [6].
An unknown proportion of these species are, in reality, not at risk
of extinction whilst others are likely to be threatened. To assess
potential impact that DD species could have on EDGE scores, the
DD species were randomly assigned threat categories at the same
ratio of CR:EN:VU:NT:LC as for the set of species for which
threat categories are known. We then repeated this simulation
assuming that DD species were more threatened than expected.
Again we randomly assigned threat categories at the same ratio as
before, but then manually increased the newly-assigned threat
categories by one level. We repeated the analysis three more times,
first increasing each DD species newly-assigned threat level by two
categories, and then also decreasing each by one and two levels
respectively. Again, whenever Least Concern status was chosen to
be less threatened, it was kept at Least Concern and, conversely,
when critically endangered species were chosen to be more
severely threatened; they were kept at that level. Unlike the other
perturbations, in which species can either increase or decrease in
EDGE score if selected, the simulated top 100 sets resulting from
this process differ only in the number of currently DD species that
displace the existing top 100.
d) Multiple sources of uncertainty. Finally, we tested how
the total amount of uncertainty could affect the priority list. In the
above scenarios, we changed 10% of species at random and
examined each source of uncertainty separately: here we explore
the effect of varying this number and include both perturbations to
the phylogeny and changes to the extinction risk categories (i.e.
both ED and GE errors), in order to test whether the uncertainty is
additive or multiplicative. We simulated a scenario in which a
proportion of species had been wrongly assigned by one or two
threat categories and placed between one and two clades from
their location on our reference phylogeny, with Data Deficient
species treated as in c, above. We first chose 5% of all species
randomly and altered their ED and/or GE scores as set out above.
We repeated the analysis with the same parameter values but
increased the number of species sequentially to 10, 15, 20, 30 and
40% of all species.
Results
We calculated ED scores for 5713 amphibian species, of which
1344 were Data Deficient and 35 extinct, meaning that we could
calculate EDGE scores for 4334 species (figure 1, for details see
Table S1). The top scoring species was Archey’s Frog, Leiopelma
archeyi, a Critically Endangered (CR) frog from New Zealand,
followed by the Chinese Giant Salamander, Andrias davidianus (also
CR, see supplemental material). The only non CR species in the
top ten was the Purple Frog, Nasikabatrachus sahyadrensis, as it has
the 7th highest ED score across all amphibians and is considered
as Endangered (EN) by the IUCN. Of the top 100 species, 75 were
classified as CR, 15 EN and 10 vulnerable (VU). There were 47
‘candidate’ (DD but high ED) species, all but 10 of which are
caecilians (table 1). The frog species Hymenochirus boulengeri,
Hymenochirus feae,Mixophyes hihihorlo and the salamanders Ambystoma
flavipiperatum,Ambystoma rivulare,Ambystoma silvensis,Protohynobius
puxiongensis were the highest-ranking non-caecilian candidate
species.
Different listing procedures produced ranking lists with different
weighting of the two component values of the EDGE listing
approach (figure 2). The weighting used for the mammal EDGE
prioritisation (‘EDGE log (2)’) in amphibians struck a reasonable
balance between threat and ED for much of the top 100, but is
slightly biased towards the threat component. Expected loss (EXP
LOSS) showed a similar pattern of slight bias towards threat status
as the EDGE list based on the log(2) listing. The approach that
appears to take the most even-handed choice of species, with
respect to the two input variables, is the Expect Loss approach
used with probabilities that predicted 500 years into the future
(Exp Loss 500).
Our perturbation of species’ ED and GE scores had very little
impact on the makeup of the ‘EDGE top 100’ (figure 3). Small
perturbations (2 clades or 2 threat categories) changed only a small
proportion of the priority list (similarity = 0.9). Even under severe
perturbation of 10% of the species’ ED or GE values, the top 100
of the original EDGE list maintained a similarity of 0.85 with the
reference set of unperturbed scores (figure 3 top left and right
panels). The impact of Data Deficient (DD) species is much
greater: when assuming that DD species were as threatened as
expected (DD category = 0 on figure 3 lower left panel) then the
similarity was 0.8 on average (in other words, 20 currently DD
species would be listed in the top 100), but similarity dropped to
0.5 if DD species are on average two categories more threatened
than expected.
When all three forms of uncertainty were combined, the
similarity was lower still (figure 3 lower right panel). Low levels of
both ED and GE errors (2 clades and 2 threat categories for 10%
of species), plus assignment of DD species in the expected
proportions, yielded similarity of around 0.7, which is roughly
what would be expected from running each perturbation
separately. Under the extreme scenario where 40% of species
were perturbed, similarity dropped further, but only to around 0.6.
In other words, quadrupling the level of perturbation causes just
10 changes to the makeup of the top 100 EDGE species.
Discussion
In view of the unprecedented species decline, particularly
among amphibians, immediate conservation action is necessary.
However, the high number of threatened amphibian species will
likely overwhelm global conservation efforts and resources, even if
Amphibian Conservation Prioritization
PLOS ONE | www.plosone.org 4 August 2012 | Volume 7 | Issue 8 | e43912

these efforts were to be intensified dramatically. Conservation
action therefore must prioritise its actions and focus its attention
and resources toward alleviating the situation for the most pressing
cases. Basing prioritisation on phylogenetic uniqueness of species
(ED), in addition to extinction risk status, captures not only the
non-randomness of extinction (with respect to phylogenetic
position), but also the fact that evolutionarily distinct species could
have important ecological roles and that their loss would result in
an over-proportional loss of evolutionary history [46,47]. Here, we
provide such a prioritization for the entire Class Amphibia. Our
analyses show that the set of priority species is robust to the ad hoc
nature of our phylogenetic tree and uncertainties in the extinction
risk assessment of large numbers of species.
Our ‘EDGE list’ of amphibians is already a focus for
conservation activities (http://www.edgeofexistence.org). This is
important because threatened amphibians with high ED are no
more likely to receive conservation mitigation than by chance, and
just 15% of the top 100 high EDGE scoring amphibian species
threatened with extinction are receiving active conservation
attention [48]. The EDGE Amphibians project has supported
conservation efforts and capacity building for over 15 top priority
species (including Andrias davidianus in China, Boulengerula niedeni in
Kenya, Rhinoderma darwinii in Chile, Proteus anguinus in Croatia and
Nasikabatrachus sahyadrensis in India), funding training initiatives and
conservation actions, with even greater aims to continue
expanding the project’s scope of activities into the future. The
EDGE Amphibians project has increased global awareness of
amphibian species, providing international audiences with further
reasons to become interested in lesser known species and
amphibians in general. The project has thus far raised over £2
million for amphibian conservation initiatives around the world
and the EDGE listing has played a major role in raising the profile
of poorly known but highly distinctive species internationally. The
EDGE website provides full details of high-priority species and
ongoing conservation activities, and has proved to be a useful
platform in leveraging support for amphibian conservation,
illustrating how a science-based conservation prioritisation tool
focusing on evolutionary distinctiveness can capture the interest of
a wide range of conservation supporters and stakeholders. Whilst
our focus here, and on the EDGE website, is on the highly-
threatened species making up the top 100, the full has wider
applications for conservation, such as mapping global hotspots of
evolutionary distinctiveness and EDGE.
The production of our amphibian EDGE list was only possible
by first assembling a species-level phylogeny. Whilst our approach
is somewhat ad hoc, it is consistent with the principles of
phylogenetic ‘supertree’ construction [49,50]. Although in the
future we can expect to obtain more accurate phylogenies based
on molecular data, conservation must act in a timely manner given
the urgency of the situation and the very real risk of imminent
amphibian species extinctions globally. A complete molecular
phylogeny of amphibians is unlikely to be available for many years,
despite the enormous pace of developments in the molecular
biology and bioinformatics, by which time it is likely that many
species will have gone extinct [8,9]. The phylogeny that we have
produced will be a valuable tool for comparative studies of
extinction risk [51,52] and the randomness (or otherwise) of
extinction risk [51,53], as well as questions about the evolutionary
history of amphibians [54–56]. Eventually, the combination of
spatial, environmental and phylogenetic information could be used
to predict the potential threat status of Data Deficient species [57].
Our simulations showed that even substantial amounts uncer-
tainty about species’ phylogenetic position and threat status have
only a minor on the set of priority species identified by the EDGE
Table 1. The 47 candidate amphibian species with high ED
scores and ‘‘Data Deficient’’ IUCN assessment staus.
Rank Family Species ED score
1 Rhinatrematidae Epicrionops columbianus 81.3908
2 Rhinatrematidae Epicrionops lativittatus 81.3908
3 Rhinatrematidae Epicrionops marmoratus 81.3908
4 Rhinatrematidae Epicrionops parkeri 81.3908
5 Rhinatrematidae Epicrionops peruvianus 81.3908
6 Caeciliidae Herpele multiplicata 73.1665
7 Caeciliidae Luetkenotyphlus brasiliensis 63.6999
8 Caeciliidae Geotrypetes angeli 59.3842
9 Caeciliidae Geotrypetes pseudoangeli 59.3842
10 Caeciliidae Boulengerula changamwensis 56.8488
11 Caeciliidae Boulengerula denhardti 56.8488
12 Caeciliidae Boulengerula fischeri 56.8488
13 Pipidae Hymenochirus boulengeri 52.5783
14 Pipidae Hymenochirus feae 52.5783
15 Myobatrachidae Mixophyes.hihihorlo 50.1187
16 Caeciliidae Dermophis costaricensis 50.0494
17 Caeciliidae Dermophis glandulosus 50.0494
18 Caeciliidae Dermophis gracilior 50.0494
19 Caeciliidae Dermophis oaxacae 50.0494
20 Caeciliidae Dermophis occidentalis 50.0494
21 Caeciliidae Microcaecilia rabei 49.7193
22 Caeciliidae Microcaecilia supernumeraria 49.7193
23 Caeciliidae Gegeneophis carnosus 45.7398
24 Caeciliidae Gegeneophis danieli 45.7398
25 Caeciliidae Gegeneophis fulleri 45.7398
26 Caeciliidae Gegeneophis krishni 45.7398
27 Caeciliidae Gegeneophis seshachari 45.7398
28 Caeciliidae Gegeneophis madhavaorum 45.7398
29 Caeciliidae Gegeneophis nadkarnii 45.7398
30 Ambystomatidae Ambystoma flavipiperatum 42.3185
31 Ambystomatidae Ambystoma rivulare 42.3185
32 Ambystomatidae Ambystoma silvensis 42.3185
33 Hynobiidae Protohynobius puxiongensis 42.1579
34 Caeciliidae Siphonops insulanus 41.7074
35 Caeciliidae Siphonops leucoderus 41.7074
36 Caeciliidae Crotaphatrema bornmuelleri 37.1099
37 Caeciliidae Crotaphatrema lamottei 37.1099
38 Caeciliidae Crotaphatrema tchabalmbaboensis 37.1099
39 Caeciliidae Atretochoana eiselti 35.9600
40 Ichthyophiidae Uraeotyphlus interruptus 35.3800
41 Ichthyophiidae Uraeotyphlus malabaricus 35.3800
42 Ichthyophiidae Uraeotyphlus menoni 35.3800
43 Ichthyophiidae Uraeotyphlus narayani 35.3800
44 Ichthyophiidae Uraeotyphlus oxyurus 35.3800
45 Mantellidae Wakea madinika 34.9872
46 Microhylidae Adelastes hylonomos 30.5161
47 Limnodynastidae Notaden weigeli 29.3150
doi:10.1371/journal.pone.0043912.t001
Amphibian Conservation Prioritization
PLOS ONE | www.plosone.org 5 August 2012 | Volume 7 | Issue 8 | e43912

approach. Wholesale changes to the mammal taxonomy and
reassessment of all species’ Red List status led to a change in the
identity of around 15 species making up the top 100 EDGE
mammals [4] (i.e. similarity = 0.85). Taxonomic and Red List
instability are likely to be greater for amphibians than mammals,
due to substantial uncertainty around cryptic species complexes in
Figure 2. The mean ED scores of the top 100 species chosen using five different methods to create EDGE lists. Thick black lines indicate
upper and low limits where species are chosen purely by having the highest ED score irrespective of threat (upper line) and just the most threatened
(lower line) species are chosen. Lines represents the mean ED of the top 1:n top ranked species by each EDGE listing process. Note logarithmic y-axis.
doi:10.1371/journal.pone.0043912.g002
Figure 3. Results from simulations to explore the impact of uncertainty on the makeup of 100 highest ranked EDGE amphibian
species. In each case, ‘similarity’ is the proportion of species shared with the unperturbed reference set, based on 1000 simulated datasets.
Confidence intervals are drawn in grey but lie too close to the mean to be visible. Panel a) shows the impact of perturbing the evolutionary
distinctiveness component (ED) by moving 500 (10%) randomly-selected species to closely related clades. Panel b) shows a similar relationship when
500 species have their threat categories perturbed. Panel c) shows the effect of different assumptions about true threat categories of Data Deficient
(DD) species: with ‘DD category = 09, DD species were assigned randomly, according to the distribution of non-DD species; with DD category .0we
assume that DD species are on average more threatened than expected. Panel d) shows the impact of multiple perturbations, with increasing the
numbers of species perturbed. See text for further details.
doi:10.1371/journal.pone.0043912.g003
Amphibian Conservation Prioritization
PLOS ONE | www.plosone.org 6 August 2012 | Volume 7 | Issue 8 | e43912

the tropics [54,55]. Our perturbation of the input data has shown
the top 100 species are rather resilient to errors and increased
knowledge. The J-shaped distribution of ED scores is likely to be
the main reason for this, as although the highest ED score is
around 190 million years, only 5% of species have scores greater
than 25 million years and 75% of species have scores under
12.5my. Therefore, if assessed and threatened, the small number
of highly distinct species will remain in the top 100 unless a serious
mistake has been made in the phylogenetic (and likely morpho-
logical) analyses of these species.
By far the most substantial source of uncertainty in our analyses
surrounds the true conservation status of species currently listed as
Data Deficient. Our list of ‘candidate’ species should be targeted
for data collection in order to make full Red List assessments as a
matter of urgency. The candidate list is dominated by caecilian
species, which are typically cryptic and poorly understood. The
whole group is in need of major taxonomic reassessment before
detailed conservation targets can be established [25]. Reassuringly,
their principally fossorial nature means that they may be, in many
cases, relatively common but undetected [56]. If true, this would
be a rare piece of good news among the devastation of amphibian
biodiversity that continues all around us. In practical terms, the
EDGE approach can successfully catalyze conservation action for
little known and often overlooked amphibian species. It is proving
itself to be a very useful prioritization tool in the development of
conservation initiatives and also has considerable potential to
continue raising awareness of the plight of amphibians globally.
Supporting Information
Table S1 EDGE and ED scores of all amphibians (see
text for details).
(CSV)
Phylogeny S1 The composite phylogeny dervied as
described in the text to build the EDGE and ED scores
with. The file can be read and converted in other formats using
the open source programming environment R using the read.tree
function of the library ape.
(TRE)
Acknowledgments
We are grateful to Olaf Bininda-Emonds and Arne Mooers for advice and
constructive criticism, and to Gordon Smith for technical assistance. We
would also like to thank Stefan Lo¨tters and an annonymous referee for
constructive comments on an earlier version of our manuscript.
Author Contributions
Conceived and designed the experiments: NJBI. Performed the experi-
ments: NJBI HM DWR. Analyzed the data: NJBI DWR. Contributed
reagents/materials/analysis tools: DWR. Wrote the paper: KS NJBI DWR
HM.
References
1. Witting L, Loeschcke V (1995) The optimization of biodiversity conservation.
Biological Conservation 71: 205–207.
2. Faith DP (2012) Conservation priorities and phylogenetic pattern. Conservation
Biology 10: 1286–1289.
3. Diniz-Filho JAF (2004) Phylogenetic diversity and conservation priorities under
distinct models of phenotypic evolution. Conservation Biology 18: 698–704.
4. Collen B, Turvey ST, Waterman C, Meredith HMR, Kuhn TS, et al. (2011)
Investing in evolutionary history: implementing a phylogenetic approach for
mammal conservation. Philosophical Transactions of the Royal Society B-
Biological Sciences 366: 2611–2622.
5. Faith DP (1992) Conservation evaluation and phylogenetic diversity. Biological
Conservation 61: 1–10.
6. Stuart SN, Chanson JS, Cox NA, Young BE, Rodrigues ASL, et al. (2004) Status
and Trends of Amphibian Declines and Extinctions Worldwide. Science 10: 1–4.
7. Mendelson JR, Lips KR, Gagliardo RW, Rabb GB, Collins JP, et al. (2006)
Biodiversity. Confronting amphibian declines and extinctions. Science 313: 48.
8. Alford RA (2011) Ecology: Bleak future for amphibians. Nature 480: 461–462.
9. Wake DB (2012) Ecology. Facing extinction in real time. Science 335: 1052–
1053.
10. Stuart SN, Chanson JS, Cox NA, Young BE, Rodrigues ASL, et al. (2004) Status
and trends of amphibian declines and extinctions worldwide. Science 306: 1783–
1786.
11. Houlahan JE, Findlay CS, Schmidt BR, Meyer AH, Kuzmin SL (2000)
Quantitative evidence for global amphibian population declines. Nature 404:
752–755.
12. McCallum ML (2007) Amphibian decline or extinction? Current declines dwarf
background extinction rate. Journal of Herpetology 41: 483–491.
13. Cushman SA (2006) Effects of habitat loss and fragmentation on amphibians: A
review and prospectus. Biological Conservation 128: 231–240.
14. Berger L (1998) Chytridiomycosis causes amphibian mortality associated with
population declines in the rain forests of Australia and Central America.
Proceedings of the National Academy of Sciences 95: 9031–9036.
15. Daszak P, Cunningham AA, Hyatt AD (2003) Infectious disease and amphibian
population declines. Diversity and Distributions 9: 141–150.
16. Boone M, Cowman D, Davidson C, Hayes T, Hopkins W, et al. (2007)
Evaluating the role of environmental contamination in amphibian population
declines. In: Gascon, C; Collins, JP; Moore, RD; Church, DR; McKay J, et al.,
editor. Amphibian Conservation Action Plan. Gland, Switzerland and Cam-
bridge, UK: IUCN/SSC Amphibian Specialist Group. 32–35.
17. Carpenter SR, Turner M (2000) Opening the black boxes: Ecosystem science
and economic valuation. Ecosystems 3: 1–3.
18. Adams M (1999) Correlated factors in amphibian decline: exotic species and
habitat change in western Washington. The Journal of wildlife management 63:
1162–1171.
19. Arau´jo MB, Thuiller W, Pearson RG (2006) Climate warming and the decline of
amphibians and reptiles in Europe. Journal of Biogeography 33: 1712–1728.
20. Pounds JA, Crump ML (1994) Amphibian declines and climate disturbance: The
case of the golden toad and the harlequin frog. Conservation Biology 8: 72–85.
21. Blaustein AR, Kiesecker JM (2002) Complexity in conservation: lessons from the
global decline of amphibian populations. Ecology Letters 5: 597–608.
22. Sodhi NS, Bickford D, Diesmos AC, Lee TM, Koh LP, et al. (2008) Measuring
the meltdown: drivers of global amphibian extinction and decline. PloS one 3:
e1636.
23. Pounds JA, Bustamante MR, Coloma LA, Consuegra JA, Fogden MPL, et al.
(2006) Widespread amphibian extinctions from epidemic disease driven by
global warming. Nature 439: 161–167.
24. Hof C, Arau´jo MB, Jetz W, Rahbek C (2011) Additive threats from pathogens,
climate and land-use change for global amphibian diversity. Nature 480: 516–
519.
25. Frost DR, Grant T, Faivovich J, Bain RH, Haas A, et al. (2006) The amphibian
tree of life. Bulletin of the American Museum of Natural History: 1–370.
26. Frost DR (2007) Amphibian Species of the World: an online reference, version 5.
27. Roelants K, Gower DJ, Wilkinson M, Loader SP, Biju SD, et al. (2007) Global
patterns of diversification in the history of modern amphibians. Proceedings of
the National Academy of Sciences of the United States of America 104: 887–
892.
28. Zhang P, Chen Y-Q, Zhou H, Liu Y-F, Wang X-L, et al. (2006) Phylogeny,
evolution, and biogeography of Asiatic Salamanders (Hynobiidae). Proceedings
of the National Academy of Sciences of the United States of America 103: 7360–
7365.
29. Faivovich J, Haddad CFB, Garcia PCA, Frost DR, Campbell JA, et al. (2005)
Systematic review of the frog family hylidae, with special reference to hylinae:
phylogenetic analysis and taxonomic revision. Bulletin of the American Museum
of Natural History: 6–228.
30. Van Bocxlaer I, Roelants K, Biju SD, Nagaraju J, Bossuyt F (2006) Late
Cretaceous vicariance in Gondwanan amphibians. PloS one 1: e74.
31. Emerson SB, Inger RF, Iskandar D (2000) Molecular systematics and
biogeography of the fanged frogs of Southeast Asia. Molecular phylogenetics
and evolution 16: 131–142.
32. Purvis A (1995) A composite estimate of primate phylogeny. Philosophical
transactions of the Royal Society of London Series B, Biological sciences 348:
405–421.
33. Bininda-Emonds ORP, Cardillo M, Jones KE, MacPhee RDE, Beck RMD, et
al. (2008) The delayed rise of present-day mammals (vol 446, pg 507, 2007).
Nature 456: 274. Available:,Go to ISI.://WOS:000261039300047.
34. Pybus OG, Harvey PH (2000) Testing macro-evolutionary models using
incomplete molecular phylogenies. Proceedings of the Royal Society of London
Series B-Biological Sciences 267: 2267–2272.
35. Nee S (2007) Inferring speciation rates from phylogenies. Evolution 55: 661–668.
Amphibian Conservation Prioritization
PLOS ONE | www.plosone.org 7 August 2012 | Volume 7 | Issue 8 | e43912

36. Agapow P-M, Purvis A (2002) Power of eight tree shape statisti cs to detect
nonrandom diversification: A comparison by simulation of two models of
cladogenesis. Systematic Biology 51: 866–872.
37. Isaac NJB, Turvey ST, Collen B, Waterman C, Baillie JEM (2007) Mammals on
the EDGE: Conservation Priorities Based on Threat and Phylogeny. PLoS ONE
2: e296.
38. Mooers AØO, Atkins RA (2003) Indonesia’s threatened birds: over 500 million
years of evolutionary heritage at risk. Animal Conservation 6: 183–188.
39. Kuhn TS, Mooers AØ, Thomas GH (2011) A simple polytomy resolver for
dated phylogenies. Methods in Ecology and Evolution 2: 427–436.
40. Orme CDL, Freckleton RP, Thomas GH, Petzhold T, Fritz SA, et al. (2011)
caper: Comparative analyses of phylogenetics and evolution in R.
41. R Development Core Team, R Development Core Team (R), Team RDC
(2011) R: A language and environment for statistical computing. Availa-
ble:http://www.r-project.org.
42. Mooers AØ, Faith DP, Maddison WP (2008) Converting endangered species
categories to probabilities of extinction for phylogenetic conservation prioriti-
zation. PloS one 3: e3700.
43. Mace GM, Lande R (1991) Assessing extinction threats: toward a reevaluation of
IUCN threatened species categories. Conservation Biology 5: 148–157.
44. Redding DW, Mooers AØO (2006) Incorporating evolutionary measures into
conservation prioritization. Conservation Biology 20: 1670–1678.
45. Zoological Society of London (2008) Edge of Existence programme. http://
www.edgeofexistence.org.
46. Myers N, Knoll AH (2001) The biotic crisis and the future of evolution.
Proceedings of the National Academy of Sciences of the United States of
America 98: 5389–5392.
47. Sechrest W, Brooks TM, da Fonseca GAB, Konstant WR, Mittermeier RA, et
al. (2002) Hotspots and the conservation of evolutionary history. Proceedings of
the National Academy of Sciences of the United States of America 99: 2067–
2071.
48. Sitas N, Baillie JEM, Isaac NJB (2009) What are we saving? Developing a
standardized approach for conservation action. Animal Conservation 12: 231–
237.
49. Bininda-Emonds ORP, Gittleman JL, Steel MA (2002) The (Super)tree of life:
Procedures, problems, and prospects. Annual Review of Ecology and
Systematics 33: 265–289. doi:10.1146/annurex.ecolysis.33.010802.150511.
50. Bininda-Emonds OR, Gittleman JL, Purvis A (1999) Building large trees by
combining phylogenetic information: a complete phylogeny of the extant
Carnivora (Mammalia). Biological reviews of the Cambridge Philosophical
Society 74: 143–175.
51. Cooper N, Bielby J, Thomas GH, Purvis A (2008) Macroecology and extinction
risk correlates of frogs. Global Ecology and Biogeography 17: 211–221.
doi:10.1111/j.1466–8238.2007.00355.x.
52. Collen B, McRae L, Deinet S, De Palma A, Carranza T, et al. (2011) Predicting
how populations decline to extinction. Philosophical transactions of the Royal
Society of London Series B, Biological sciences 366: 2577–2586.
53. Purvis A, Agapow PM, Gittleman JL, Mace GM (2000) Nonrandom extinction
and the loss of evolutionary history. Science 288: 328–330.
54. Vieites DR, Wollenberg KC, Andreone F, Ko¨hler J, Glaw F, et al. (2009) Vast
underestimation of Madagascar’s biodiversity evidenced by an integrative
amphibian inventory. Proceedings of the National Academy of Sciences of the
United States of America 106: 8267–8272.
55. Stuart BL, Inger RF, Voris HK (2006) High level of cryptic species diversity
revealed by sympatric lineages of Southeast Asian forest frogs. Biology letters 2:
470–474.
56. Gower DJ, Wilkinson M (2005) Conservation Biology of Caecilian Amphibians.
Conservation Biology 19: 45–55.
57. Safi K, Pettorelli N (2010) Phylogenetic, spatial and environmental components
of extinction risk in carnivores. Global Ecology and Biogeography: 352–362.
Amphibian Conservation Prioritization
PLOS ONE | www.plosone.org 8 August 2012 | Volume 7 | Issue 8 | e43912