Content uploaded by Diana O. Fisher
Author content
All content in this area was uploaded by Diana O. Fisher on Apr 30, 2020
Content may be subject to copyright.
RESEARCH
PAPER
Trajectories from extinction: where are
missing mammals rediscovered?geb_624415..425
Diana O. Fisher*
The University of Queensland, School of
Biological Sciences, St Lucia 4072, Queensland,
Australia
ABSTRACT
Aim To determine where mammals that are presumed to be extinct are most likely
to be rediscovered,and to test predictions of two hypotheses to explain trajectories
of decline in mammals. Range collapse is based on the premise that extinction rates
at the edge of species ranges are highest because habitat is suboptimal, so declining
species are predicted to survive longer near the centre of their ranges. We predicted
that under range collapse, remnant populations are most likely be rediscovered
within their former core range. Conversely, if threats usually spread across ranges,
declining species will be pushed to the periphery (range eclipse), so rediscoveries
are predicted at the edge of the pre-decline range. If so, species would be more likely
to be rediscovered in marginal habitat, and at higher elevations than the sites from
which they disappeared.
Location World-wide.
Methods Using data on 67 species of mammals which have been rediscovered, I
tested whether species were disproportionately rediscovered in the outer 50% of
their former range area or at higher elevations than their last recorded locations,
and which species characteristics were associated with rediscovery location and
habitat change, using both the phylogenetic generalized least squares method to
account for phylogenetic non-independence and linear models of raw species data.
Results Species affected by habitat loss were more likely to be rediscovered at the
periphery than the centre of their former range, consistent with range eclipse
caused by the spread of habitat destruction. High human population pressure
predicted which species changed habitat between their previous records and redis-
covery. Coastal species experienced higher human population densities, and were
more likely to be rediscovered at the periphery of their former ranges, and there was
some evidence of an up-slope shift associated with higher human populations at
lower elevations.
Main conclusion The locations of rediscoveries of species affected by habitat loss
were consistent with range eclipse through a mechanism of spreading habitat loss
and human population pressure, rather than with range collapse. Searches for
mammals that have declined from habitat loss should include range edges and
marginal habitat, especially in areas of high human population density.
Keywords
Biological invasion, elevation, extinction trajectory, habitat loss, mammals,
overkill, rediscovery.
*Correspondence: Diana O. Fisher, School of
Biological Sciences, Goddard Building (8), The
University of Queensland, St Lucia 4072,
Queensland, Australia.
E-mail d.fisher@uq.edu.au
INTRODUCTION
Species presumed to be extinct are often rediscovered (MacPhee
& Flemming, 1999), and large amounts of money and effort
have been spent attempting to rediscover some high-profile
extinct vertebrates such as thylacines in Australia and ivory-
billed woodpeckers in the USA (Paddle, 2002; Wilcove, 2005).
Despite their frequently high scientific profiles and conservation
Global Ecology and Biogeography, (Global Ecol. Biogeogr.) (2011) 20, 415–425
© 2010 Blackwell Publishing Ltd DOI: 10.1111/j.1466-8238.2010.00624.x
http://wileyonlinelibrary.com/journal/geb 415
importance (Diamond, 1987), until the recent study of Fisher &
Blomberg (2010), which found that mammals affected by
habitat loss with relatively large ranges are most likely to be
rediscovered, rediscoveries have been considered as idiosyn-
cratic events, so there has been no previous quantitative evalu-
ation of where species are rediscovered (Diamond, 1987;
MacPhee & Flemming, 1999).
The abundant centre hypothesis suggests that rediscoveries of
rare or declining species are most likely in the best habitat at the
core of their ranges rather than in marginal habitat at the edges,
because animal density and condition decrease at the range
periphery (Caughley et al., 1988; Sagarin & Gaines, 2002). The
abundant centre hypothesis predicts that ranges should collapse
inwards as the species declines to extinction, assuming that
peripheral populations go extinct most rapidly, leaving the last
remnants where the population was densest: a decline trajectory
of ‘range collapse’ – also termed ‘the demographic hypothesis’
(Channell & Lomolino, 2000a,b), ‘melting range’ (Rodriguez,
2002) or ‘demographic decline’ (if population densities in the
core range are relatively stable) (Hemerik et al., 2006). Support
for this idea includes studies by Nathan et al. (1996), who
showed that Israeli birds in a peripheral part of their geographic
range were more likely to go extinct, and Moritz et al. (2008),
who found evidence of inwards range collapse associated with a
decrease in habitat and climate suitability in two alpine
mammals. Hemerik et al. (2006) also argued that figures in
Channell & Lomolino (2000b) are consistent with range collapse
in about half of declining species.
There are obvious reasons to begin looking for a missing
species in the vicinity of its last recorded location; suitable
habitat might persist near the most recent site, and species at
very low densities can be virtually undetectable, although their
spatial distribution has not changed (Rout et al., 2009). On the
other hand, there are plausible reasons to expect rediscoveries to
be remote from the location of extinct populations. Recent
large-scale studies have failed to support the premise of the
abundant centre pattern of density distribution (Sagarin et al.,
2006). The range collapse hypothesis has been challenged by the
finding that severe declines at the continental scale in species
generally, and specifically in mammals and birds (Channell &
Lomolino, 2000a,b), New Zealand reptiles and frogs (Towns &
Daugherty, 1994) and a European butterfly (Thomas et al.,
2008), more often follow a trajectory outwards from the range
centre to the periphery. This decline trajectory has been termed
‘the contagion hypothesis’ (Channell & Lomolino, 2000a,b) or
‘range eclipse’ (Hemerik et al., 2006), the term used here. The
hypothesized reason for range eclipse is that anthropogenic
threats such as habitat loss and invasive predators usually spread
across species ranges in a contagion-like manner, pushing the
declining species to the edge of its former range (Towns &
Daugherty, 1994; Channell & Lomolino, 2000a; Fisher et al.,
2003). A recent global study has supported the premise of con-
tagion in the spatial distribution of habitat clearing (Boakes
et al., 2010). Rediscoveries might therefore be expected to occur
closer to the range periphery than most historical records of
species occurrence. Because search effort is expected to focus on
the historical core range, missing species might be more likely to
persist undetected for longer at the range periphery, especially if
this is in areas of low human population density. Species that
can retreat to marginal habitat may also be more likely to survive
threats affecting their historical core ranges (Caughley & Gunn,
1995).
Agents of extinction spread in three dimensions, but until
recently elevation has not received as much research attention as
two-dimensional decline trajectories in mammals. Recent
up-slope range shifts in species distributions have been exten-
sively studied in the context of climate change (Moritz et al.,
2008), and even considered as a ‘fingerprint of global warming’
(Raxworthy et al., 2008), but high elevations might also serve as
refuges from threats that spread from the coast (Channell &
Lomolino, 2000a). Historical and current urbanization, defores-
tation and other anthropogenic habitat changes are concen-
trated at low elevations (Ciesin, 2000). For example, Peh (2007)
found that montane areas of Southeast Asian islands contained
twice the proportion of undisturbed forest as lowlands, and Hall
et al. (2009) found that ten times the proportion of tropical
forest had been lost from low elevations as from montane areas
in a high-biodiversity region of Tanzania. Therefore we might
expect to often find missing species at higher elevations than the
sites from which they disappeared, even if they pre-date the era
of recent climate warming.
The aim of this study is to determine where rediscovered
species are most often found with respect to their former ranges,
habitats and elevational ranges, in order to help target search
efforts to places where missing species are most likely to persist.
I also aim to test if the location of mammal rediscoveries is
consistent with the range eclipse or range collapse hypotheses of
species decline. I ask: (1) how far away from last sightings do
rediscoveries occur; (2) are rediscoveries likely to occur in dif-
ferent habitat from the known historical habitat type and at
higher elevations than past records of missing species; (3) how
does human population density affect the location of rediscov-
eries; and (4) are missing mammals more likely to be found in
the range centre or at the range periphery?
MATERIALS AND METHODS
Data and definitions
I assembled a global database of mammal species that had been
reported globally extinct or possibly extinct but have been redis-
covered (Appendix S1 in Supporting Information), and estab-
lished the status history of species included in the 2009 IUCN
Red List (IUCN, 2009), using past and present IUCN Red Lists
and related publications, primary literature, books and the
Committee on Recently Extinct Organisms mammal database
(MacPhee & Flemming, 1999: http://creo.amnh.org/).I omitted
species that have been ‘rediscovered’ through taxonomic revi-
sion, and only included mammals that are named as full species
(not subspecies) and are currently considered taxonomically
valid by the IUCN (species authorities are reported in the Red
List; IUCN, 2009).
D. O. Fisher
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd416
I recorded the date and location of last sighting and rediscov-
ery for each species. To determine if the location of rediscoveries
depended on the type of threat, I recorded available data on the
cause of extinction. In 36 of the 67 species, a single threat was
reported in the published literature as the cause of decline to
apparent extinction. In the other 31 species,one of the three main
classes of threats was reported to be the most important; either
habitat loss (deforestation, agricultural clearing, fragmentation,
degradation, overgrazing), overkill (harvesting, hunting, exploi-
tation, persecution or bycatch) or invasion (invasive predators or
diseases). I therefore recorded the main or single threat and
additional threats interacting with the major extinction driver,
such as loss of vegetation cover exacerbating predation.
To test if rediscoveries were at higher or lower elevations
than last sighting locations, I assigned an elevation category to
each species (coastal =up to 50 m a.s.l., lowland =50 to 1000 m
a.s.l., highland =occurs over 1000 m a.s.l.) and an elevation
rank at last recorded locations and rediscovery sites, in 100-m
intervals, for species in which collecting site elevations or
precise locations were published. Elevation data were collated
from the published literature on species rediscoveries (Appen-
dix S1) and from species accounts in the 2009 Red List. These
were usually the same sources as the location data. Species were
classified as having the potential opportunity to change habi-
tats if they were not restricted to sea-level environments
(beach, saltmarsh, mangrove forest), mountain tops or areas
without topographical variation (inland plains or uniformly
low-elevation islands).
To find if rediscoveries are likely in habitat which differs from
the known historical habitat type, I recorded detailed descrip-
tions of habitat and vegetation type at last recorded locations
and rediscovery sites. To examine the effect of human popula-
tion density on rediscovery location, I estimated human popu-
lation density at the last sighting and rediscovery sites in 1990
(data from the closest available time to the mean rediscovery
date) using the Gridded Population of the World database
(Ciesin, 2000), imported into ArcGIS. These data have a reso-
lution of approximately 4.5 km (Ciesin, 2000).
Because the distance between last sighting and rediscovery is
likely to scale with geographic range size, which should therefore
be included as a covariate, I recorded approximate historical
range (geographical range area before recent declines, not
current area of occupancy) in km2for species with published
estimates or maps. Because the amount of information on
former ranges varied, I assigned categories of range rank,
defined as the historical range with a precision of one order of
magnitude (1 =up to 1 km2,8=1,000,000 to 10,000,000 km2). I
also recorded the potentially important covariates of island
status, i.e. restricted to islands or not (only one species, the
dibbler, is known to occur on a continental coast as well as on
islands), and body mass (g).
I also recorded search effort (number of reported search expe-
ditions targeted to the species after it was reported extinct, and
before it was rediscovered).These reported searches are likely to
be underestimates, but I assumed that reports are proportional
to the true number of searches, because expert authors of the
IUCN Red List are required to document search effort in
their accounts of extinct and possibly extinct species, and
publications that report rediscoveries invariably discussed
the frequency of previous sightings and unsuccessful search
expeditions.
Spatial data
In order to test how far away from last sightings rediscoveries do
occur,and whether missing mammals more likely to be found in
the range centre or at the range periphery, I calculated the dis-
tance between last recorded locations and rediscovery sites using
the SoDA package in R (Chambers,2008), which uses the method
of Vincenty (1975) to calculate geodetic distances in metres from
map coordinates, then converted these to kilometres (R Devel-
opment Core Team, 2009). Spatial data files accompany species
range maps published by the IUCN (IUCN, 2009). However,
many were not suitable for plotting historical ranges, because
most species accounts lacked data in the field describing the range
area from which the species has been extirpated. Some maps (e.g.
the black-footed ferret, Mustela nigripes) did show detailed pre-
decline distributions, but some (e.g. most Australian species)
showed current area of occupancy. Therefore I used additional
published sources of pre-decline (historical) range maps for
rediscovered species (Oliver & Roy, 1993; Jiménez, 1996; Carr &
Robinson, 1997; Nadler et al., 2002; Monda et al., 2007 Van Dyck
& Strahan, 2008). For the 53 species with adequate data, I digi-
tized the Lambert’s azimuthal equal area projection range maps
published by the IUCN (IUCN, 2009), and overlaid historical
range boundaries onto each map, plotted with respect to coast-
lines, and coordinates of towns and landmarks such as national
parks, located using Google Earth®. I then mapped the locations
of last sighting and rediscovery sites with respect to these histori-
cal range boundaries. In order to test if species were more likely to
be rediscovered nearer the historical range periphery or range
centre, I plotted a polygon inside the former range area by
connecting points at the same distance from the range periphery
as the rediscovery site. I used this to calculate the proportional
area of each range (in pixels) that was closer to the range periph-
ery than the rediscovery site, so that species at the periphery were
scored zero (no part of the range closer to the periphery) and
species at the range centre were scored 1 (all of the rest of the
range closer to the periphery).
Statistical analysis
To test if rediscoveries were at higher or lower elevations than last
sighting locations, I used linear mixed-effects models with
species as a random factor, elevation as the dependent variable
and location (last sighting or rediscovery location) as the inde-
pendent variable. I used the same approach to model differencein
human population density between the two locations (with
species as a random factor and (log) human population density as
the dependent variable). To incorporate the two observations of
each species as a random factor, I treated the species phylo-
geny as a pedigree, using the lmekin function in the kinship
Spatial bias in mammal rediscovery
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd 417
package in R to fit the variance–covariance matrix and estimate
fixed factor effects and the proportion of variance explained by
phylogenetic effects among species (Atkinson & Therneau,
2008). I used chi-squared tests of the raw data (with Yates’ cor-
rection) to test if rediscovery sites occurred more often than
expected at the range periphery or range centre (defined above).
To test which variables were correlated with the distance
between last sightings and rediscoveries, and rediscovery in mar-
ginal habitat, I used both the phylogenetic generalized least
squares (GLS) method to account for phylogenetic non-
independence, and linear models of raw species data (in R, R
Development Core Team, 2009). Potential correlates were geo-
graphic range, elevation, island-dwelling, threat, body size,
human population density, number of searches and years
missing. For the least squares method, I used a composite phy-
logeny based on the recent mammal supertree (Bininda-
Emonds et al., 2007; Fisher & Blomberg, 2010), with the ape
package for R (Paradis et al., 2002). For all multiple regression,
mixed-effects models, and GLS models,I selected the best model
using the minimum Akaike information criterion (AIC). For
tests in which there was no difference between phylogenetic GLS
and linear models of raw species data (GLM), I report only the
phylogenetic GLS results.
RESULTS
How far away from last sightings do
rediscoveries occur?
I identified 67 species of mammals that have been rediscovered
after they were presumed or feared to be extinct (Appendix S1).
A third of rediscovered species were found within 50 km of their
last known location, the median distance was 128 km and the
arithmetic mean distance was 294 ⫾53 km (SE) (Fig. 1). The
geometric mean was 86 km. Three species were found again at
their original locations (Bulmer’s fruit bat, Aproteles bulmerae,
in Papua New Guinea; Juan Fernández fur seal, Arctocephalus
philippii), in Chile; the Vanikoro flying fox, Pteropus tubercula-
tus, in the Solomon Islands). The Australian heath mouse,
Pseudomys shortridgei, was found furthest from its last recorded
location (2200 km away).
Rediscovery distance was significantly positively correlated
with species range size and elevation category [(log) range, t=
4.0, d.f. =59, P<0.001; elevation, t=2.5, d.f. =59, P=0.017], but
not independently with time missing or any of the other vari-
ables modelled (number of years missing, number of searches,
threat, body mass, island-dwelling, or human population
density). Coastal species were found closer to their original site
than lowland and highland species (coastal, 79 ⫾46 km (SE);
lowland, 362 ⫾72; highland, 256 ⫾117).
This was consistent with the restricted distributions of most
rediscovered coastal species. The range sizes of lowland and
highland species did not differ, but on average, coastal species
had ranges that were half the size of the ranges of lowland and
highland species (F2, 60 =7.8, P<0.001; coastal, rank 3.63 ⫾0.54
(SE); lowland, 5.07 ⫾0.17; highland, 4.82 ⫾0.42). There was
no difference in the reported number of searches conducted
before species were rediscovered at different elevation categories
(F2, 64 =1.8, P=0.17; coastal, 2.6 ⫾0.59 (SE); lowland, 1.7 ⫾
0.26; highland, 1.4 ⫾0.29).
Are rediscoveries likely to occur in different habitat
from the known historical habitat type?
Fifteen per cent of rediscovered mammals (10 species) were
found in a different type of habitat from that of their last
Figure 1 The frequency distribution of distances between the last sighting and rediscovery locations for rediscovered species.
D. O. Fisher
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd418
recorded locations (Table 1). All of these were tropical forest
species threatened mainly by habitat loss (eight species) or
mainly by overkill and secondarily habitat loss (two species).
The only significant covariate of this habitat change was human
population density. Species that were rediscovered in a different
habitat experienced human population densities around 11-fold
higher than species found in their original habitat (species redis-
covered in a different type of habitat from that of their historical
locations experienced a mean human population density of 455
⫾207 people km-2, and those rediscovered in the same habitat
as expected from historical records experienced a mean human
population density of 77 ⫾16 people km-2,z=2.8, P=0.006).
Some of the species that changed habitats were rediscovered
in higher-elevation forest types than their original lowland
records, usually associated with steep limestone outcrops, and
some were found in anthropogenic vegetation (plantations of
shrub and tree crops or regrowth of native vegetation) (Table 1).
These were all areas of woody vegetation, usually near primary
forest remnants. I found no cases where rediscovered mammals
occurred in cropland, anthropogenic pasture or other
dramatically different vegetation types from the forest that they
previously occupied.
Are rediscoveries likely to occur at higher elevations
than past records?
There was an average up-slope shift of 35% between the last
recorded location and rediscovery site in species that were not
restricted to a small elevational range by specialization on
coastal or mountain-top habitats or lack of topographical varia-
tion within their ranges. In these species, the mean elevation
rank of last recorded sites was 5.2 ⫾1.2 (c. 520 m) and the
elevation of rediscovery sites was 7.0 ⫾1.2 (c. 700 m). However,
the mixed effects model testing for a difference in elevation
between locations was the only one to differ between the GLM
(raw data) and GLS (phylogenetic effects) statistical methods.
According to the GLM, this difference was strongly significant
(t=2.8, d.f. =32, P=0.008), but according to the GLS it was not
(t=1.1, d.f. =32, P=0.26).
Table 1 Mammals rediscovered in a habitat which is different from their former (pre-decline) habitat.
Species Former habitat Habitat where rediscovered
Philippine bare-backed fruit bat (Dobsonia
chapmani)
Lowland primary forest, including at sea level Secondary forest on karst limestone outcrops,
on steep slopes
Tonkin snub-nosed monkey (Rhinopithecus
avunculus)
Low-elevation and highland primary tropical
forest
Tropical evergreen forests associated with
karst limestone hills and mountains
Eastern black-crested gibbon (Nomascus
nasutus)
Lower montane and limestone forests, in a
wet tropical monsoon climate, at an
altitude range of 50–900 m
Restricted to limestone forests on inaccessible
karst outcrops ranging 640–800 m in
elevation
Bougainville monkey-faced bat (Pteralopex
anceps)
Mature tropical forest, including lowland
forest close to the coast of Bougainville
Now found almost entirely in primary forest
in upland areas, high-elevation mossy
forest
Crested genet (Genetta cristata) Tropical lowland rain forest Edge of a farm where cassava and low scrub
abutted secondary forest
Malabar civet (Viverra c ivettina) Lowland forests, swamp and riparian forests
on the coastal plain. Most records were in
valleys around riparian areas
Confined to thickets (understorey of shrubs
and grasses) in cashew plantations and
highly degraded lowland forests
Rio de Janeiro arboreal rat (Phaenomys
ferrugineus)
Primary rain forest Herbaceous vegetation next to forested hills
and farmland
Brazilian arboreal mouse (Rhagomys
rufescens)
Lowland tropical (Atlantic) forest Heavily degraded secondary growth of
semi-deciduous tropical forest, dominated
by exotic molasses grass, Melinis
minutiflora
Salim Ali’s fruit bat (Latidens salimalii) Tropical montane evergreen forest Coffee and cardamom plantation adjacent to
a forest reserve
Thin-spined porcupine (Chaetomys
subspinosus)
Lowland tropical (Atlantic) forest Extensive areas of forest cover, including
restinga (low-canopy coastal forest on
sand, a mix of species including palms and
mangroves) and abruca forest (cocoa
plantations where a number of larger
native trees are left standing)
Sources: Santos et al. (1987), Kurup (1989), Heard & Van Rompaey (1990), Flannery (1991), Wirth (1992), Bates et al. (1994), Bonvicino et al. (2001),
Parnaby (2002), Nadler (2003), Alcala et al. (2004), Paguntalan et al. (2004), Percequillo et al. (2004), IUCN (2009).
Spatial bias in mammal rediscovery
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd 419
How does human population density affect the
location of rediscoveries?
Human population density did not differ between the last
recorded location and rediscovery site for individual species, but
sites affected by invasion had sparser human populations than
sites affected by habitat loss or overkill (difference between
habitat loss and invasion, t=3.1, d.f. =56, P=0.002; difference
between invasion and overkill, t=0.38,d.f. =56, P=0.70; densit y
at sites where species were affected by invasion 28.5 ⫾15.5 per
km2, habitat loss 170.7 ⫾69.3, and overkill 104.1 ⫾35.4).
Human population density also differed significantly between
high and coastal elevation sites (t=-2.3, d.f. =56, P=0.02;
coastal, 280 ⫾227 per km2(SE); lowland, 114 ⫾29; highland,
27 ⫾7).
Are missing mammals more likely to be found in the
range centre or at the range periphery?
Sixty per cent of rediscoveries were closer to the range periphery
than the centre; this proportion was not significant. Thirty-two
species were in the outer 50% of the range area (e.g. Fig. 2a
and b) and 21 species were in the inner 50% (e.g. Fig. 2a) (chi-
squared =2.3, d.f. =1, P=0.13). However, the proportional area
of the range that was closer to the periphery than the rediscovery
site was significantly correlated with the type of threat, human
Figure 2 Estimated historical ranges (shaded) of some rediscovered mammals, showing the location of rediscovery sites (filled circles) with
respect to range boundaries. Maps are based on azimuthal equal area projection range maps (IUCN, 2009). (a) Six Australian examples. The
long-tailed dunnart (Sminthopsis longicaudata) and heath rat (Pseudomys shortridgei) were rediscovered in the most peripheral 0–25% of
the former range (ⱕ25% of the former range is closer to a range edge), and the Julia Creek dunnart (Sminthopsis douglasi), desert rat
kangaroo (Caloprymnus campestris), Hastings River mouse (Pseudomys oralis) and Leadbeater’s possum (Gymnobelideus leadbeateri)were
rediscovered in the most central 0–25% of the former range (>75% of the former range is closer to a range edge). (b) Two Caribbean
species. The Cuvier’s hutia (Plagiodontia aedium) and almiqui (Solenodon cubanus) were rediscovered in the most peripheral 0–25% of the
former range (Hispaniola and Cuba, respectively).
D. O. Fisher
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd420
population density and elevation (difference between invasion
and habitat loss, t=2.9, d.f. =49, P=0.005; difference between
overkill and habitat loss, t=2.4, d.f. =49, P=0.02; Fig. 3).
Seventy-eight per cent of coastal species were found in the
peripheral 50% of their former range, compared with 63% of
lowland species and 20% of highland species. Species rediscov-
ered at the range centre experienced twice the human popula-
tion density (33.1 ⫾2.0 per km2) of those rediscovered at the
periphery (13.5 ⫾1.4) (effect of human population density
rank, t=2.6, d.f. =53, P=0.01; effect of elevation rank, t=2.3,
P=0.03).
DISCUSSION
Mammal rediscovery data were only consistent with the range
eclipse hypothesis in species affected by habitat loss. It would
therefore be most effective to include range edges in searches for
missing mammals that were affected by habitat loss. There was a
significant bias towards peripheral rediscovery locations in
species affected by habitat loss (71% rediscovered at a range edge
compared with 40% of species affected by overkill). Channell &
Lomolino (2000a) argued that range eclipse is a general response
to anthropogenic threats, and made no explicit predictions
about variation in decline trajectories with specific causes. They
proposed that range contraction towards the periphery occurs
when a threatening processes such as urbanization or defores-
tation moves across the species’ range, until the species persists
only at the edge of its former distribution, where the threat does
not reach. This suggests that range eclipse should predominate
in species affected by habitat loss, which is likely to reduce or
eliminate species in progressively larger parts of the range
without affecting density in the remaining habitat (Rodriguez,
2002). In contrast, based on evidence from declining birds,
Rodriguez (2002) suggested that overkill tends to reduce overall
population density without progressively removing species from
parts of the range. If so, species declining from overkill would be
expected to show range collapse. The predicted decline trajec-
tory from invasive predators and diseases is unclear. On the one
hand, biological invasions usually spread from a single entry
point (Cassey et al., 2004). On the other hand,the mechanism of
impact is similar in invasive predators and human predators
(Owens & Bennett, 2000). Both threats reduce prey survival and
density rather than range. The idea that overkill results in range
collapse was not supported by the rediscovered mammal data,
because species affected by invasion and overkill showed no
clear pattern in decline trajectories (Fig. 3).
Rediscovery attempts invariably include surveys of the most
recent known location of the species (e.g. Kabay & Start, 1976;
Kenyon, 1977; Woods et al., 1985; Pritchard, 1989; Páez &
García, 2003; Turvey et al., 2007), but fewer than 5% of redis-
covered mammals were located at the place from which they had
disappeared. Therefore it is important not to restrict searches for
any missing mammals to the region of their last location. The
distribution of rediscovery distances had a long tail, with most
cases distant from previous records. Half of all rediscovered
species were found more than 100 km away from their last loca-
tion, including four species found more than 1000 km away
(Fig. 1). The distance at which a species can be rediscovered is
constrained by range size, so it is not surprising that range and
rediscovery distance were strongly correlated.
The idea of range eclipse has previously been tested by com-
paring the location of former and current species ranges, but the
hypothesis also predicts that declining species are eliminated
from core habitat and persist in marginal habitat that is less
favourable to their threats. A substantial minority of rediscov-
ered mammals (15%) persisted in refuges from habitat destruc-
tion, and less commonly from hunting. These species had
experienced much greater human population densities than
species that were rediscovered in a similar habitat to that in their
original locations, suggesting that human population pressure
pushed them into marginal environments (or excluded them
from non-marginal environments). There have been no previ-
ous tests of the generality of this mechanism, but there are some
cases in other taxa. For example, the takahe, Porphyrio mantelli,
seems to have been pushed into a refuge from threats, in sub-
optimal habitat. The takahe is a New Zealand bird which disap-
peared from its former stronghold of riparian lowland forest,
and was rediscovered in alpine tussock grassland, where it avoids
high densities of introduced predators but does not thrive
(Bunin & Jamieson, 1995). Further evidence of this mechanism
comes from the recent rediscovery of the Australian armoured
mist frog, Litoria lorica. This frog was ‘known to be a rain forest
specialist endemic to the Wet Tropics Bioregion’, presumed
extinct due to the invasive chytrid fungus disease (IUCN, 2009),
but it was rediscovered in 2008 in dry eucalypt forest, which
appears to be suboptimal habitat but is a less favourable envi-
ronment for the disease (C. Hoskin, pers. comm.).
Figure 3 Location of rediscovery sites (peripheral, outer 50% of
range area, or central, inner 50% of range area) with respect to
the main cause of decline in rediscovered species.
Spatial bias in mammal rediscovery
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd 421
Mammal species with rediscovered remnant populations sur-
viving only in marginal habitat are all tropical forest mammals
threatened mainly by habitat loss, disproportionately in areas of
very high human population density. These species survived
only in rockier and steeper habitat than most of the historical
range, which would be less favourable for agricultural clearing
and forestry (e.g. the Bougainville monkey-faced bat, formerly
found in tropical lowland rain forest in Papua New Guinea and
the Solomon Islands, and now restricted to high-elevation moss
forest), or in disturbed forest and plantations (e.g. the malabar
civet, formerly found in the now non-existent habitat of coastal
riparian forest valleys in the Western Ghats region of India, and
briefly rediscovered in the dense understorey of a cashew plan-
tation; Table 1).
I propose that the range eclipse idea can also be extended to
three dimensions. Tropical deforestation and urbanization are
strongly biased towards low elevations (Ciesin, 2000; Peh, 2007;
Hall et al., 2009). Declining species should therefore be elimi-
nated from low-elevation habitat first, and rediscovery sites are
predicted to be at higher mean elevations than historical distri-
butions. As predicted,there was an average up-slope shift of 35%
between the last recorded location and rediscovery site in species
with the potential to change their elevational distribution, but
the evidence that rediscovery sites were more likely to be at
higher elevations than historical distributions was equivocal,
because the phylogenetic comparison was not significant. The
reason for disagreement between the two statistical methods is
most likely to be that phylogenetic comparative analysis is gen-
erally more conservative, because the method compares lineages
rather than tree tips (species) (Fisher et al., 2003). Additionally,
the six murid rodents and three soricid shrews in the data set
were all at low elevations, although taxonomic clustering within
elevations did not occur in the other groups with several species,
such as pteropodid bats.
Other analyses of elevational range shift in response to habitat
change have focused on North American mammals. My finding
of an up-slope shift is consistent with the descriptive analysis of
Laliberte & Ripple (2004), who found that many species such as
cougar, elk and Dall’s sheep have disproportionately lost the
portions of their historic ranges that were at low elevations.
Beever et al. (2003) found that over a large area pikas were more
likely to have been extirpated from lower-elevation sites, which
were more disturbed by roads and contained less habitat.At the
local scale, Moritz et al. (2008) stated that vegetation change at
low sites, but not at high elevations, was a likely factor in many
up-slope shifts of small mammals in Yosemite National Park,
but Larrucea & Brussard (2008) concluded that habitat change
at higher elevations moved the elevational distribution of
pygmy rabbits downwards.
Consistent with the idea that the threat from human popula-
tion pressure pushes declining species into marginal parts of
their range at low elevations, human population density at last
sighting and rediscovery sites was 10 times greater in coastal
rediscovered mammals than in species at high elevations
(>1000 m a.s.l.) and four times higher than species at low eleva-
tions (51 to 1000 m a.s.l.). Species that persisted only at the
periphery of their former ranges were more likely to be coastal.
Coastal species were missing for longer before they were redis-
covered, despite being found after more searches within smaller
areas than species at higher elevations, implying that their
remnant peripheral populations had often declined to near-
undetectable densities. This suggests that missing mammals are
not more likely to be eventually found at range edges simply
because search effort has been most intense at the centre. Pub-
lished accounts of several rediscovered coastal species also
suggest that they occurred at extremely low densities during the
time that they were missing. For example, 21 individuals of the
San Jose Kangaroo rat, Dipodomys insularis, were captured in
2005 after 15 years of unsuccessful targeted trapping on the
same small Mexican island (Espinosa-Gayosso & Álvarez-
Castaneda, 2006). The coastal Brazilian rodent Phyllomys uni-
color had not been recorded since 1824, when a single individual
was trapped in 2004. It had been the target of intensive unsuc-
cessful searches around the area where it was rediscovered in
2001, 2002, 2003, 2004 and 2005, and has not been captured
again since 2004 (Leite et al., 2007; IUCN, 2009).
CONCLUSIONS
The location of rediscoveries of mammals affected by habitat
loss was consistent with Channell and Lomolino’s (2000a)
hypothesis of contagion, or range eclipse, through a mechanism
of spreading habitat loss and human population pressure. The
pattern of decline suggests that these threats spread, and declin-
ing mammal distributions contracted, both outward towards
range peripheries and upward from coastal elevations, between
the times of previous location records and rediscovery. These
results imply that especially intense, repeated search effort is
needed to locate coastal mammals that have declined to near
extinction, and that these efforts are more likely to be successful
near the range periphery in coastal species. Searches for species
apparently exterminated by habitat loss and urbanization
should include peripheral areas of the probable former range,
upland rocky sites and moderately disturbed habitat. It is likely
that more tropical forest mammals currently feared extinct in
areas of high human population density survive in secondary
forest remnants; we need to find and protect these sites.
ACKNOWLEDGEMENTS
I thank Kate Jones, Jaime Jiminez, David and Meredith Happold,
Andrew Cockburn, Hideki Endo, Rainer Hutterer, Carla
Kishinami, Stefan Klose, Friederike Spitzenberger, Craig Hilton-
Taylor and Richard Fuller for providing or helping us to locate
species data. I thank Richard Fuller for advice and help with
population density mapping in ArcGIS, and Simon Blomberg
for comments, and statistical advice and help, particularly with
methods of constructing mixed effects models to incorporate
phylogenetic effects in R. This work was supported by funding
from the Australian Research Council (ARF DP0773920).
D. O. Fisher
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd422
REFERENCES
Alcala, E.L., Paalan, R.B., Averia, L.T. & Alcala, A.C. (2004)
Rediscovery of the Philippine bare-backed fruit bat (Dobsonia
chapmani, Rabor) on southwestern Negros Island, the
Philippines. Silliman Journal,45, 123–136.
Atkinson, B. & Therneau, T. (2008) Kinship: mixed-effects Cox
models, sparse matrices, and modeling data from large pedi-
grees. R Package, Available at: http://cran.r-project.org/web/
packages/kinship/index.html (accessed 5 November 2009).
Bates, P.J.J., Harrison, D.L., Thomas, N.M. & Muni, M. (1994)
The Indian fruit bat Latidens salimalii Thonglongya, 1972
(Chiroptera: Pteropodidae) rediscovered in southern India.
Bonner Zoologische Beitraege,45, 89–98.
Beever, E.A., Brussard, P.E. & Berger, J. (2003) Patterns of
apparent extirpation among isolated populations of pikas
(Ochotona princeps) in the Great Basin. Journal of Mammal-
ogy,84, 37–54.
Bininda-Emonds, O.R.P., Cardillo, M., Jones, K.E., MacPhee,
R.D.E., Beck, R.M.D., Grenyer, R., Price, S.A., Vos, R.A., Gittle-
man, J.L. & Purvis, A. (2007) The delayed rise of present-day
mammals. Nature,446, 507–512.
Boakes, E.H., Mace, G.M., McGowan, P.J.K. & Fuller, R.A.
(2010) Extreme contagion in global habitat clearance. Pro-
ceedings of the Royal Society B: Biological Sciences,277, 1081–
1085.
Bonvicino, C.R., Oliveira, J.A., D’Andrea, P.S. & Carvalho, R.W.
(2001) Rediscovery of the endemic Atlantic forest rodent
Phaenomys ferrugineus (Sigmodontinae): new data on its
morphology and karyology. Boletim do Museo Nacional (Rio
de Janeiro), Nova Serie, Zoologia,467, 1–12.
Bunin, J.S. & Jamieson, I.G. (1995) New approaches toward a
better understanding of the decline of takahe (Porphyrio man-
telli) in New Zealand. Conservation Biology,9, 100–106.
Carr, S.G. & Robinson, A.C. (1997) The present status and dis-
tribution of the desert rat-kangaroo Caloprymnus campestris
(Marsupialia: Potoroidae). South Australian Naturalist,72,
4–27.
Cassey, P., Blackburn, T.M., Russell, G.J., Jones, K.E. &
Lockwood, J.L. (2004) Influences on the transport and
establishment of exotic bird species: an analysis of the parrots
(Psittaciformes) of the world. Global Change Biology,10, 417–
426.
Caughley, G. & Gunn, A. (1995) Conservation biology in theory
and practice. Blackwell Science, Oxford.
Caughley, G., Grice, D., Barker, R. & Brown, B. (1988) The edge
of the range. Journal of Animal Ecology,57, 771–785.
Chambers, M. (2008) Software for data analysis: programming
with R. Springer, New York.
Channell, R. & Lomolino, M.V. (2000a) Dynamic biogeography
and conservation of endangered species. Nature,403, 84–86.
Channell, R. & Lomolino, M.V. (2000b) Trajectories to extinc-
tion: spatial dynamics of the contraction of geographical
ranges. Journal of Biogeography,27, 169–179.
CIESIN (2000) Gridded population of the world. Center for Inter-
national Earth Science Information Network (CIESIN) of the
Earth Institute, Columbia University, Available at: http://
sedac.ciesin.columbia.edu/plue/gpw (accessed 1 September
2009), New York, USA.
Diamond, J.M. (1987) Extant unless proven extinct? Or, extinct
unless proven extant? Conservation Biology,1, 77–79.
Espinosa-Gayosso, C.V. & Álvarez-Castaneda, S.T. (2006) Status
of Dipodomys insularis, an endemic species of San Jose Island,
Gulf of California, Mexico. Journal of Mammalogy,87, 677–
682.
Fisher, D.O. & Blomberg, S.P. (2010) Correlates of rediscovery
and the detectability of extinction in mammals. Proceedings
of the Royal Society B: Biological Sciences, doi:10.1098/
rspb.2010.1579.
Fisher, D.O., Blomberg, S.P. & Owens, I.P.F. (2003) Extrinsic
versus intrinsic factors in the decline and extinction of Aus-
tralian marsupials. Proceedings of the Royal Society B: Biologi-
cal Sciences,270, 1801–1808.
Flannery, T.F. (1991) Emperor, king, and little pig: the three
rats of Guadalcanal. Australian Natural History,23, 635–
641.
Hall, J., Burgess, N.D., Lovett, J., Mbilinyi, B. & Gereau, R.E.
(2009) Conservation implications of deforestation across an
elevational gradient in the Eastern Arc Mountains, Tanzania.
Biological Conservation,142, 2510–2521.
Heard, S. & Van Rompaey, H. (1990) Rediscovery of the crested
genet. Mustelid and Viverrid Conservation,3, 1–4.
Hemerik, L., Hengeveld, R. & Lippe, E. (2006) The eclipse of
species ranges. Acta Biotheoretica,54, 255–266.
IUCN (2009) IUCN Red List of threatened species. Available at:
http://www.iucnredlist.org/ (accessed 1 September 2009),
Species Survival Commision, Gland, Switzerland.
Jiménez, J.E. (1996) The extirpation and current status of wild
chinchillas Chinchilla lanigera and C. brevicaudata.Biological
Conservation,77, 1–6.
Kabay, E.D. & Start, A.N. (1976) Results of the search for the
potoroo in the south west and south coast of Western Australia
1975/1976. Western Australian Department of Fisheries and
Wildlife, Perth.
Kenyon, K.W. (1977) Caribbean monk seal extinct. Journal of
Mammalogy,58, 97–98.
Kurup, G.U. (1989) The rediscovery of Malabar civet (Viver ra
civettina Blyth) in India. Tigerpaper (Bangkok),16, 13–14.
Laliberte, A.S. & Ripple, W.J. (2004) Range contractions of
North American carnivores and ungulates. Bioscience,54,
123–138.
Larrucea, E.S. & Brussard, P.F. (2008) Shift in location of
pygmyrabbit(Brachylagus idahoensis) habitat in response to
changing environments. Journal of Arid Environments,72,
1636–1643.
Leite, Y.L.R., Lóss, S., Rego, R.P., Costa, L.P. & Bonvicino, C.R.
(2007) The rediscovery and conservation status of the Bahian
giant tree rat Phyllomys unicolor (Mammalia : Rodentia :
Echimyidae) in the Atlantic forest of Brazil. Zootaxa,1683,
51–57.
MacPhee, R.D.E. & Flemming, C. (1999) Requiem aeternam: the
last five hundred years of mammalian species extinctions.
Spatial bias in mammal rediscovery
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd 423
Extinctions in near-time (ed. by R.D.E. MacPhee), pp. 333–
371. Kluwer Academic/ Plenum, New York.
Monda, K., Simmons, R.E., Kressirer, P., Su, B. & Woodruff, D.S.
(2007) Mitochondrial DNA hypervariable region-1 sequence
variation and phylogeny of the concolor gibbons, Nomascus.
American Journal of Primatology,69, 1285–1306.
Moritz, C., Patton, J.L., Conroy, C.J., Parra, J.L., White, G.C. &
Beissinger, S.R. (2008) Impact of a century of climate change
on small-mammal communities in Yosemite National Park,
USA. Science,322, 261–264.
Nadler, T. (2003) Rediscovery of the eastern black crested
gibbon Nomascus nasutus in Vietnam. The Gibbon’s Voice,6,
1–3.
Nadler, T., Momberg, F., Dang, N.X. & Lormee, N. (2002) Leaf
monkeys. Vietnam primate conservation status review,pp.
145–164. Fauna and Flora International and Frankfurt Zoo-
logical Society, Hanoi.
Nathan, R., Safriel, U.N. & Shirihai, H. (1996) Extinction and
vulnerability to extinction at distribution peripheries: an
analysis of the Israeli breeding avifauna. Israel Journal of
Zoology,42, 361–383.
Oliver, W.L.R. & Roy, S.D. (1993) The Eurasian suids (Sus and
Babyrousa): the pygmy hog (Sus salvanius). Pigs, peccaries, and
hippos: status survey and conservation action plan (ed. by
W.L.R. Oliver), pp. 121–129. Species Survival Commission,
Gland, Switzerland.
Owens, I.P.F. & Bennett, P.M. (2000) Ecological basis of extinc-
tion risk in birds: habitat loss versus human persecution and
introduced predators. Proceedings of the National Academy of
Sciences USA,97, 12144–12148.
Paddle, R. (2002) The last Tasmanian tiger: the history and
extinction of the thylacine. Cambridge University Press,
Cambridge.
Páez, R.B. & García, I.R. (2003) Current status of the carabali
hutia from south of Isla de la Juventud,Mysateles meridionalis.
Orsis,18, 7–11.
Paguntalan, L.M.J., Pedregosa, M.D. & Gadiana, M.J.C. (2004)
The Philippine bare-backed fruit bat Dobsonia chapmani
Rabor, 1952: rediscovery and conservation status on Cebu
Island. Silliman Journal,45, 113–122.
Paradis, E., Strimmer, K., Claude, J., Noel, Y. & Bolker, B. (2002)
Ape: analyses of phylogenetics and evolution. Available at:
http://cran.r-project.org/ (accessed 5 November 2009).
Parnaby, H.E. (2002) A taxonomic review of the genus Pteralopex
(Chiroptera: Pteropodidae), the monkey-faced bats of the
south-western Pacific. Australian Mammalogy,23, 145–162.
Peh, K.S.H. (2007) Potential effects of climate change on eleva-
tional distributions of tropical birds in Southeast Asia.
Condor,109, 437–441.
Percequillo, A.R., Goncalves, P.R. & De Oliveira, J.A. (2004) The
rediscovery of Rhagomys rufescens (Thomas, 1886), with a
morphological redescription and comments on its systematic
relationships based on morphological and molecular (cyto-
chrome b) characters. Mammalian Biology,69, 238–257.
Pritchard, J.S. (1989) Ilin Island cloud rat extinct? Oryx,23, 126.
R Development Core Team (2009) R: a language and environ-
ment for statistical computing., R Foundation for Statistical
Computing. Available at: http://www.R-project.org (accessed
5 November 2009), Vienna, Austria.
Raxworthy, C.J., Pearson, R.G., Rabibisoa, N., Rakotondrazafy,
A.M., Ramanamanjato, J.B., Raselimanana, A.P., Wu, S.,
Nussbaum, R.A. & Stone, D.A. (2008) Extinction vulnerabil-
ity of tropical montane endemism from warming and
upslope displacement: a preliminary appraisal for the
highest massif in Madagascar. Global Change Biology,14,
1703–1720.
Rodriguez, J.P. (2002) Range contraction in declining North
American bird populations. Ecological Applications,12, 238–
248.
Rout, T.M., Salomon, Y. & McCarthy, M.A. (2009) Using sight-
ing records to declare eradication of an invasive species.
Journal of Applied Ecology,46, 110–117.
Sagarin, R.D. & Gaines, S.D. (2002) The ‘abundant centre’ dis-
tribution: to what extent is it a biogeographical rule? Ecology
Letters,5, 137–147.
Sagarin, R.D., Gaines, S.D. & Gaylord, B. (2006) Moving beyond
assumptions to understand abundance distributions across
the ranges of species. Trends in Ecology and Evolution,21,
524–530.
Santos, I.B., Oliver, W.L.R. & Rylands, A.B. (1987) Distribution
and status of two species of tree porcupines Chaetomys subs-
pinosus and Sphiggurus insidiosus in south-east Brazil. Dodo,
24, 43–60.
Thomas, C.D., Bulman, C.R. & Wilson, R.J. (2008) Where within
a geographical range do species survive best? A matter of scale.
Insect Conservation and Diversity,1, 2–8.
Towns, D.R. & Daugherty, C.H. (1994) Patterns of range con-
tractions and extinctions in the New Zealand herpetofauna
following human colonization. New Zealand Journal of
Zoology,21, 325–339.
Turvey, S.T., Pitman, R.L., Taylor, B.L., Barlow, J., Akamatsu, T.,
Barrett, L.A., Zhao, X.J., Reeves, R.R., Stewart, B.S., Wang,
K.X., Wei, Z., Zhang, X.F., Pusser, L.T., Richlen, M., Brandon,
J.R. & Wang, D. (2007) First human-caused extinction of a
cetacean species? Biology Letters,3, 537–540.
Van Dyck, S. & Strahan, R. (2008) The complete book of Austra-
lian mammals. Reed Books, Sydney.
Vincenty, T. (1975) Direct and inverse solutions of geodesics on
the ellipsoid with application of nested equations. Survey
Review,23, 88–94.
Wilcove, D.S. (2005) Rediscovery of the ivory-billed wood-
pecker. Science,308, 1422–1423.
Wirth, R. (1992) Tonkin snub-nosed monkey (Rhinopithecus
avunculus) rediscovered. Asian Primates,2, 1–2.
Woods, C.A., Ottenwalder, J.A. & Oliver, W.L.R. (1985) Lost
mammals of the Greater Antilles: the summarised findings of
a ten weeks field survey in the Dominican Republic, Haiti and
Puerto Rico. Dodo,22, 23–42.
D. O. Fisher
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd424
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online
version of this article:
Appendix S1 Mammals that have been rediscovered after
having been considered extinct, or possibly extinct.
As a service to our authors and readers, this journal provides
supporting information supplied by the authors. Such materials
are peer-reviewed and may be reorganized for online delivery,
but are not copy-edited or typeset. Technical support issues
arising from supporting information (other than missing files)
should be addressed to the authors.
BIOSKETCH
Diana Fisher is an ecologist at the University of
Queensland. Her current research focuses on causes of
modern extinction and predisposition to extinctions
and declines in mammals and other vertebrates,
detectability of extinction, and the conservation ecology
of Australian mammals. She also works on the
evolutionary ecology of marsupial mating systems,
maternal care and life-history strategies.
Editor: Brian McGill
Spatial bias in mammal rediscovery
Global Ecology and Biogeography,20, 415–425, © 2010 Blackwell Publishing Ltd 425
