Access to this full-text is provided by Wiley.
Content available from Diversity and Distributions
This content is subject to copyright. Terms and conditions apply.
426
|
wileyonlinelibrary.com/journal/ddi Diversity and Distributions. 2020;26:426–440.
Received: 19 June 2019
|
Revised: 8 November 2019
|
Accepted: 12 December 2019
DOI: 10.1111/ddi.13029
BIODIVERSITY RESEARCH
Identifying conservation priorities in a defaunated tropical
biodiversity hotspot
Andrew Tilker1,2 | Jesse F. Abrams1 | An Nguyen1 | Lisa Hörig1 | Jan Axtner1 |
Julie Louvrier1 | Benjamin M. Rawson3 | Hoa Anh Quang Nguyen3 |
Francois Guegan4 | Thanh Van Nguyen1,5 | Minh Le5,6 | Rahel Sollmann7 |
Andreas Wilting1
1Department of Ecological Dynamics,
Leibniz Institute for Zoo and Wildlife
Research, Berlin, Germany
2Global Wildlife Conservation, Austin, TX,
USA
3World Wide Fund for Nature, Hanoi,
Vietnam
4World Wide Fund for Nature, Vientiane,
Lao PDR
5VNU-Central Institute for Natural
Resources and Environmental Studies,
Vietnam National University, Hanoi, Vietnam
6Department of Environmental Ecology,
Faculty of Environmental Sciences, VNU-
University of Science, Vietnam National
University, Hanoi, Vietnam
7Department of Wildlife, Fish, and
Conser vation Biology, University of
California Davis, Davis, CA, USA
Correspondence
Andrew Tilker, Department of Ecological
Dynamics, Leibniz Institute for Zoo and
Wildlife Research, Berlin, Germany.
Email: tilker@izw-berlin.de
Funding information
Leibniz Institute for Zoo and Wildlife
Research; Critical Ecosystem Partnership
Fund; Safari Club International Foundation;
Point Defiance Zoo and Aquarium,
Grant/Award Number: Dr. Holly Reed
Conser vation Fund; German Federal
Ministry for the Environment, Nature
Conser vation, and Nuclear Safety (BMBU);
Fulbright exchange program; Kreditanstalt
für Wiederaufbau (Kf W); German Federal
Ministry of Education and Research, Grant/
Award Number: BMBF FKZ: 01LN1301A
Editor: Luigi Maiorano
Abstract
Aim: Unsustainable hunting is leading to widespread defaunation across the trop-
ics. To mitigate against this threat with limited conservation resources, stakeholders
must make decisions on where to focus anti-poaching activities. Identifying priority
areas in a robust way allows decision-makers to target areas of conservation impor-
tance, therefore maximizing the impact of conservation interventions.
Location: Annamite mountains, Vietnam and Laos.
Methods: We conducted systematic landscape-scale surveys across five study sites
(four protected areas, one unprotected area) using camera-trapping and leech-de-
rived environmental DNA. We analysed detections within a Bayesian multispecies
occupancy framework to evaluate species responses to environmental and anthro-
pogenic influences. Species responses were then used to predict occurrence to
unsampled regions. We used predicted species richness maps and occurrence of en-
demic species to identify areas of conservation importance for targeted conservation
interventions.
Results: Analyses showed that habitat-based covariates were uninformative. Our final
model therefore incorporated three anthropogenic covariates as well as elevation,
which reflects both ecological and anthropogenic factors. Conservation-priority spe-
cies tended to found in areas that are more remote now or have been less accessible
in the past, and at higher elevations. Predicted species richness was low and broadly
similar across the sites, but slightly higher in the more remote site. Occupancy of the
three endemic species showed a similar trend.
Main conclusion: Identifying spatial patterns of biodiversity in heavily defaunated
landscapes may require novel methodological and analytical approaches. Our results
indicate that to build robust prediction maps it is beneficial to sample over large spa-
tial scales, use multiple detection methods to increase detections for rare species,
include anthropogenic covariates that capture different aspects of hunting pressure
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium,
provided the original work is properly cited.
© 2020 The Authors. Diversity and Distributions published by John Wiley & Sons Ltd
|
427
TILKER E T aL.
1 | INTRODUCTION
Tropical biodiversity is declining at an alarming rate as a result of
intense anthropogenic pressures (Bradshaw, Sodhi, & Brook, 2009).
Although habitat loss and degradation are major drivers of these de-
clines (Rosa, Smith, Wearn, Purves, & Ewers, 2016; Alroy, 2017), un-
sustainable hunting is increasingly emerging as the primary threat to
wildlife in tropical biodiversity hotspots (Benítez-López et al., 2017).
Large- and medium-sized mammals tend to be particularly vulner-
able to hunting because they often occur at lower average popula-
tion densities and have lower intrinsic rates of increase and longer
generation times (Bodmer, Eisenberg, & Redford, 1997; Cardillo et
al., 2005; Davidson, Hamilton, Boyer, Brown, & Ceballos, 2009).
Indeed, the “empty forest syndrome” that Redford (1992) warned
about almost three decades ago is now a commonplace phenome-
non and, given the ever-increasing demand for wildlife products in
the world's tropical regions (Rosa et al., 2016; Ripple et al., 2016),
this trend is unlikely to slow in the coming years. Without urgent
and effective measures to address overexploitation, tropical wild-
life populations will continue to decline, and species extinctions will
follow. Confronting the pantropical defaunation crisis has become
one of the most important challenges facing conservation today
(Bradshaw et al., 2009).
Defaunation has been particularly severe in South-East Asia,
where high human densities, a thriving illegal wildlife trade, weak
protected area governance and rapid infrastructure development
have synergistically contributed to unsustainable, industrial-scale
hunting (Duckworth et al., 2012; Harrison et al., 2016; Wilcove,
Giam, Edwards, Fisher, & Koh, 2013). Within South-East Asia,
the Annamites ecoregion on the border of Vietnam and Laos has
undergone severe defaunation as a result of widespread illegal
hunting (Harrison et al., 2016; Timmins et al., 2016). Poaching
in the Annamites is primarily accomplished by the setting of in-
discriminate wire snares (Gray et al., 2018). Numerous mammals
are regionally extinct (Walston et al., 2010; Brook et al., 2014),
and even once common species now survive at low densities
(Duckworth et al., 2016). High levels of unsustainable hunting
pressure are particularly worrisome from a conservation per-
spective, because the region is home to several endemic mam-
mal species. Mammals restricted to this ecoregion include the
saola Pseudoryx nghetinhensis, large-antlered muntjac Muntiacus
vuquangensis, Annamites dark muntjac species complex Muntiacus
rooseveltorum/ truongsonensis, Owston's civet Chrotogale owsto ni
and Annamite striped rabbit Nesolagus timminsi (Long, Hoang, &
Truyen, 2005; Sterling & Hurley, 2005; Tordoff, Timmins, Smith,
& Vinh, 2003). Taken together, the high poaching pressure and
unique biodiversity in the Annamites make it one of the highest
priority tropical regions in South-East Asia for the prevention of
imminent hunting-driven extinctions.
To maximize the effectiveness of conservation interventions to
prevent unsustainable hunting in tropical biodiversity hotspots, it is
imperative to make optimal use of limited conservation resources.
In the Annamites, the magnitude of the snaring crisis (Gray et al.,
2018), coupled with nascent protected area enforcement capacities
and lack of sufficient resources, has overwhelmed efforts to ade-
quately reduce this threat at the landscape level. Given these lim-
itations, targeting snare-removal efforts to specific areas within a
landscape may be critical to reduce snaring to levels that would allow
population recovery. To implement this approach, it is first neces-
sary to identify priority areas. In the Annamites, areas that harbour
threatened and endemic species are top priorities for targeted in situ
protection measures. These species often occur at low densities and
are therefore particularly susceptible to local extirpation. To identify
priority areas, it is important to apply appropriate analytical tech-
niques. Species distribution modelling provides an ideal framework
for mapping spatial patterns of biodiversity and thus identifying
conservation-priority areas (Guisan et al., 2013; Rodríguez, Brotons,
Bustamante, & Seoane, 2007).
There are, however, two fundamental challenges to the mod-
elling of species distributions in tropical rain forest environments.
First, tropical mammal species are often difficult to detect because
they are rare, elusive and occur at low densities. Second, even when
these species can be detected, it may be difficult to obtain enough
data to construct robust species distribution models (Cayuela et al.,
2009), particularly in defaunated areas, where mammal populations
are depleted.
Advances in non-invasive survey methods and statistical mod-
elling techniques provide ways to address these challenges. Two
non-invasive methods have revolutionized surveys for tropical mam-
mals: camera-traps (Tobler, Carrillo-Percastegui, Pitman, Mares, &
Powell, 2008) and high-throughput sequencing of environmental
DNA (eDNA) (Bohmann, Schnell, & Gilbert, 2013). Camera-trapping
is a well-established method and has been used to gather data on
even the rarest of tropical mammal species (Ganas & Lindsell, 2010;
and analyse data within a Bayesian multispecies framework. Our models further sug-
gest that more remote areas should be prioritized for anti-poaching efforts to prevent
the loss of rare and endemic species.
KEYWORDS
Annamites, camera-trapping, defaunation, environmental DNA, multispecies occupancy,
species richness, tropical rain forest, unsustainable hunting
428
|
TILKER ET aL.
Raloff, 1999; Whitfield, 1998). The use of eDNA is relatively new
but shows considerable promise. Invertebrate-derived DNA (iDNA)
approaches using terrestrial haematophagous leeches, in particu-
lar, have proven adept at detecting tropical mammals (Schnell et al.,
2018, 2012; Weiskopf et al., 2018). Recently, Abrams et al. (2019)
showed that combining camera-trapping and iDNA leech data have
the potential to improve detection probabilities for tropical mam-
mal species beyond what would be provided by each method inde-
pendently. The joint camera-trap and iDNA approach thus open new
possibilities for obtaining detections of elusive tropical rain forest
mammals, which in turn can be used to build robust species distri-
bution models.
Even with improved detection methods and combined datasets,
however, it may not be possible to obtain sufficient records for rare
species. This shortfall represents a major issue because the rar-
est species are often the species of highest conservation concern.
Multispecies occupancy models offer an analytical framework to
address this challenge, as species with few detections borrow infor-
mation from more abundant species, which improves precision of
the parameter estimates for rare species (Drouilly, Clark, & O'Riain,
2018; Li, Bleisc h, & Jiang, 2018; Tobl er, Hartley, Carr illo-Perc asteg ui,
& Powell, 2015). Because species-specific responses to covariates
can be projected to unsampled areas, this approach can be used to
generate maps of species potential occurrence (MacKenzie et al.,
2017; Sollmann et al., 2017).
Here, we collected a landscape-scale systematic camera-trap-
ping and iDNA dataset across a protected area complex in the
central Annamites landscape to identify priority areas for targeted
conservation interventions. We used a multispecies occupancy
framework and environmental and anthropogenic covariates to es-
timate species occurrence and predict species richness across the
surveyed landscape. Our prediction maps provide insight into where
to focus conservation efforts among individual study sites at the
landscape scale, and more specifically can inform deployment of
snare-removal teams within protected areas. We discuss our results
within the context of informing targeted conservation interventions
to prevent further defaunation, and species extinctions, within trop-
ical biodiversity hotspots.
2 | METHODS
2.1 | Study area
We conducted landscape-scale surveys in a large contiguous forest
in the central Annamites landscape of Vietnam and Laos. The study
area spans both countries and is divided into five administrative
units. In Vietnam, we surveyed three sites: Bach Ma National Park
(NP), the Hue Saola Nature Reserve (NR) and the Quang Nam Saola
NR. In Laos, we surveyed the eastern section of Xe Sap National
Protected Area (NPA) and an adjacent ungazetted forest near the
village of Ban Palé (Figure 1). Together these areas comprise ap-
proximately 900 km2 of mountainous terrain with elevations ranging
between 100 and 2,000 m asl. The dominant habitat type is wet ev-
ergreen tropical rain forest. Although the wider central Annamites
region has experienced extensive past disturbance from both de-
foliation and logging, habitat loss, degradation, and fragmentation
within the past 20 years has been minimal within our study sites
(Matusch, 2014; Meyfroidt & Lambin, 2008). At the landscape scale,
forest structure and habitat type are consistent across the study
sites, characterized by mature secondary forest with a multi-tiered
closed canopy (Figure S1). The Vietnam sites are surrounded by a
densely populated matrix consisting of human settlements, agricul-
tural fields and timber plantations. Human population density in the
Lao sites is low and, aside from small-scale shifting cultivation, the
landscape surrounding the survey areas has not been heavily modi-
fied. However, Vietnamese incursion into these areas for poaching
and illegal gold mining is widespread (Tilker, 2014) and has been fa-
cilitated by the recent construction of a road connecting Vietnam
and Laos that bisects the Palé area.
Poaching pressure is high across the landscape (Wilkinson, 2016;
WWF, 2017). Measures to mitigate illegal hunting differ in intensity
and effectiveness among the five sites. Patrolling in Bach Ma NP is
not intensive and has received less technical and financial support
than the adjacent sites. The Hue and Quang Nam Saola NRs have
benefited from WWF investment in enforcement since 2011 under
the Carbon and Biodiversity (CarBi) project, maintaining active
Forest Guard patrol teams to strengthen enforcement capacities in
the field and provision of capacity development in patrol strategy,
data collection and adaptive management for park staff. The Forest
Guard teams are comprised of local community members and their
primary role is to remove wire snares and destroy poacher camps
(Wilkinson, 2017). Between 2011 and 2017, the patrols removed
>110,000 snares from the Hue and Quang Nam Saola NRs (WWF,
2017). The eastern section of Xe Sap NPA has also benefited from
WWF-supported snare-removal operations, although these efforts
have not been as regular or intensive as in the Saola NRs. There are
no active patrols in Palé, as it is outside of the Xe Sap NPA.
2.2 | Data collection and preparation
We conducted systematic camera-trapping and leech surveys from
No vember 20 14 to De cembe r 2016. We set up a tot al of 140 came ra-
trap stations: 53 stations in Bach Ma NP, 21 in the Hue Saola NR, 25
in the Quang Nam Saola NR, 15 in eastern Xe Sap NPA and 26 in the
Palé area (Figure 1; Table S2). Stations were spaced approximately
2.5 km apart (mean = 2.47 ± 0.233), aiming at spatial independence
of sampling locations, and left in the forest for a minimum of 60 days
(mean = 71.60 ± 16.39). Cameras were set 20–40 cm off the ground,
operational 24 hr/day and programmed to take a three-photo burst
with no delay between photographic events. To maximize detec-
tion probabilities, we set two camera-traps (Hyperfire Professional
PC850, Reconyx®) at each station facing different directions. We
treated the two cameras as a single station in our analyses. Camera-
trap data were managed using the package camtrapR (Niedballa,
|
429
TILKER E T aL.
Sollmann, Courtiol, & Wilting, 2016). We excluded arboreal species
from our final species list, as these species are unlikely to be reli-
ably detected by camera-traps placed at ground-level (Abrams et al.,
2018). We also removed rodents and squirrels, given the difficulty
of identifying these mammals to species level using camera-trap im-
ages alone, and all domestic animals.
We complemented camera-trapping with the collection of ter-
restrial haematophagous leeches around the camera-trap stations.
Leeches were collected once during camera-trap setup and again
du ring re t r i eval. In Vie tnam, leeches we re co lle cted in 20 × 20 m sa m -
pling plots set up to assess microhabitat characteristics (see below).
In Laos, we collected leeches in a grid around each camera-trap sta-
tion, with one camera-trap station per grid. We altered the leech
collection strategy in Laos because sampling occurred during the dry
season; increasing spatial coverage around the stations allowed us
to collect leech numbers similar to the Vietnam sites. We separated
the two types of leeches, brown and tiger, because the leeches po-
tentially differ in their feeding behaviour (Schnell et al., 2015). All
leeches of the same type from the same station and occasion were
combined and processed as one leech bulk sample. Leeches were
immediately placed in RNAlater and stored long-term at −20°C.
Lee c he s we re pr o ces s e d us i n g th e la bo r ato r y pr o ce d u re s an d bi o -
informatics pipeline described in Axtner et al. (2019). The workflow
is designed to minimize the risk of false positives that could arise
from laboratory artefacts or misidentification during taxonomic as-
signment. To address these risks, it employs different levels of repli-
cation (i.e. extraction, PCRs), a curated reference database, and the
probabilistic taxonomic assignment method PROTAX (Somer vuo,
Koskela, Pennanen, Henrik Nilsson, & Ovaskainen, 2016) that has
been shown to be robust even when reference databases are incom-
plete (Richardson, Bengtsson-Palme, & Johnson, 2017; Rodgers et
al., 2017). Leech samples were digested, DNA was extracted, and
then mitochondrial target DNA of host species was amplified with
PCR and sequenced using Illumina high-throughput sequencing. We
trained PROTAX models and weighted them towards 127 mammal
and bird species expected to occur in the study area by assigning a
prior probability of 90% to these species and a 10% probability to all
others (Somervuo et al., 2016; see Table S2 for full-weighted species
list). Our protocol was slightly modified from A xtner et al. (2019) in
that we amplified the mitochondrial marker 16S in six PCR replicates
for all samples and used genetic markers 12S and CytB only for sam-
ples where taxonomic assignment was still uncertain due to inter-
specific invariance or missing references (e.g. porcupines, viverrids,
muntjacs). We accepted a species assignment when it was present in
at least two independent PCR replicates (Abrams et al., 2019; Axtner
et al., 2019). As with the camera-trapping data, we excluded arboreal
FIGURE 1 Map of camera-traps and leech collection stations across five areas in the central Annamites landscape
430
|
TILKER ET aL.
species, rodents, squirrels and domestic animals from the final spe-
cies list.
2.3 | Covariates
We hypothesized that mammal occurrence may be influenced by
both environmental and anthropogenic factors. We measured three
environmental features that characterize different aspects of micro-
habitat structure: canopy closure, vegetation density and leaf litter.
We used canopy closure as an indication of forest degradation, with
lower values representing more disturbed habitat (Chazdon, 2003).
Previous studies have shown that vegetation density may be an im-
portant microhabitat feature for some tropical mammals (Goulart et
al., 2009; Martin, Cavada, Ndibalema, & Rovero, 2015; Mathai et al.,
2017). Leaf litter impacts multiple aspects of vegetation community
composition (Facelli & Pickett, 1991). It is also an important micro-
habitat for invertebrates and small vertebrates (Burghouts, Ernsting,
Korthals, & De Vries, 1992; Vitt & Caldwell, 1994), which are impor-
tant food resources for insectivores and small carnivores.
To assess microhabitat features, we set up a 20 × 20 m plot
around the camera-trap stations, with the centrepoint halfway be-
tween the two cameras, and oriented along the cardinal axes. To
measure canopy closure, we took vertical photographs at the cen-
trepoint and at the corners of the grid. Canopy photographs were
manually converted to black and white images using the GNU Image
Manipulation Program (GIMP team, 2017). We calculated percent-
age canopy closure (white pixels) for each image using R 3.4.0 (R Core
Team, 2018). Values for each image were averaged to give a single
canopy closure value for each station. To measure vegetation den-
sity, we took ph otographs in each cardinal direction of a 1 × 1.5 m or-
ange sheet positioned 10 m from the centrepoint. Photographs were
processed using the canopy closure protocol, giving a single average
vegetation density value for each station. We measured leaf litter
percentage cover in nine 1 × 1 m subplots located at the centrepoint,
10 m from the centrepoint in each cardinal direction and at the plot
corners. Each subplot was visually assigned a value from 0 to 4 based
on the amount of leaf litter versus bare ground visible in each plot.
Leaf litter values were averaged to give a single value for each sta-
tion. For a detailed explanation of the microhabitat assessment, see
Abrams et al. (2018).
In addition to the environmental covariates, we measured an-
thropogenic features that approximate hunting pressure. We use
proxies for hunting pressure, rather than direct measures, for two
reasons. First, we are not aware of any existing datasets that di-
rectly measure hunting pressure within our study sites. Second,
robustly assessing poaching represents a difficult undertaking
because illegal hunting is such a cryptic phenomenon. Although
some studies have used presence or absence of people from cam-
era-trapping data to represent direct measures of hunting pressure
(Dias, Lima Massara, Campos, & Henrique Guimarães Rodrigues,
2019), such measures are not applicable in our landscape, because
some local communities are allowed to legally enter the study
sites to collect non-timber forest products. Further complicat-
ing the situation is the fact that these local people may engage in
both legal non-timber forest product collection and illegal hunt-
ing in order to maximize potential profit. Given the difficulties in
assessing hunting directly, we used measures of accessibility as
proxies for hunting pressure in our study sites. Previous studies
have shown accessibility and hunting to be correlated (Espinosa,
Branch, & Cueva, 2014; Koerner, Poulsen, Blanchard, Okouyi, &
Clark, 2017; Rao, Myint, Zaw, & Htun, 2005). We used three co-
variates that capture different aspect of accessibility: distance
from major cities, village density and least-cost path from major
roads. We used city distance as a proxy for hunting pressure
captured at the landscape scale. Although we measure distance
to the nearest major city (Hue or Da Nang, both with popula-
tion > 350,000), we also interpret this covariate as an approxi-
mation of accessibility to the densely populated coastal areas of
Vietnam. We chose to measure distance to the cities, rather than
other points along the urbanized coastal areas, because Hue and
Da Nang are known to be major hubs for the illegal wildlife trade
(Sandalj, Treydte, & Ziegler, 2016; Van Song, 2003). Given the vol-
ume of bushmea t that pas ses through these market s (Sandalj et al.,
2016), it is likely that these urban population centres create sub-
st ant ial na tural re sour ce dema n d shado ws acro ss the lan dscap e, as
has been shown in other tropical regions (Wilkie, Bennett, Peres,
& Cunningham, 2011). We derived the city distance covariate by
calculating the Euclidean distance from the camera-trap stations
to the nearest major city using the package gDistance (Van Etten,
2017), then taking the lower of the two values. The city distance
covariate is measured in metres, with increasing values indicating
more remote areas. We then took the log of the covariate to ap-
proximate the nonlinear effect that increasing distance likely has
on accessibility. Village density serves as a proxy for hunting at
the local scale. Local villagers often supplement their income by
providing bushmeat to the bushmeat markets in regional towns
and cities, and are therefore a primary driver of poaching in the
central Annamites (MacMillan & Nguyen, 2014). Studies in other
tropical regions have demonstrated mammal depletions surround-
ing local villages (Abrahams, Peres, & Costa, 2017; Koerner et al.,
2017; Rao et al., 2005). To calculate village density, we first cre-
ated a ground truthed point shapefile layer documenting local vil-
lages around our study sites. We then created a heatmap in q gis
2.18.9 (QGIS Development Team, 2016) using the village shapefile
as the input point layer. To create the heatmap, we used the de-
fault quartic kernel decay function and set the radius to 15 km.
The village density radius was chosen so that all individual sam-
pling stations in our study landscape were covered in the final
heatmap. Observations in the field indicate that all stations, even
those in the most in the most remote areas, were subject to some
level of hunting pressure. We then used the extract function in
the raster package (Hijmans, 2019) to obtain heatmap values for
each station. The village density covariate is unitless, with lower
values indicating areas that are more remote. Finally, the least-
cost path covariate also serves as a proxy for hunting pressure at
|
431
TILKER E T aL.
the local scale. However, it differs from the village density mea-
sure in two fundamental ways. First, the least-cost path covariate
explicitly incorporates accessibility based on terrain ruggedness
characteristics, therefore providing a more accurate representa-
tion of remoteness than linear measures. Second, we calculated
the least-cost path covariate over three time periods (1994, 2004,
and 2014) to better capture the amount of time that an area has
been subjected to poaching pressure. The least-cost path covari-
ate therefore captures both spatial and temporal dimensions. To
create the least-cost path covariate, we first used the timelapse
function in Google Earth Engine (Gorelick et al., 2017) to generate
a GIS layer of major roads in and around our study site for three
time periods: 1994, 2004 and 2014. We converted the roads layer
to points in QGIS. Next, we used the R package movecost (Alberti,
2019) to calculate travel time along least-cost path routes from the
statio ns to the neares t 50 points along the road layer, using a shut-
tle radar topography mission (SRTM) 30 m digital elevation model
as the cost surface raster, and then selected the lowest value as
the final least-cost path value. We averaged the three values to
give a si ngl e leas t-co st pat h valu e for ea ch station , which we us e as
an approximation of the time that an area has been accessible over
the past 20 years. The roads least-cost path covariate is measured
in hours, with higher values indicating areas that take longer to
access, and are therefore more remote.
We also included elevation as a covariate in our models. We
consider elevation as both an anthropogenic and ecological covari-
ate. Because higher elevation areas are more difficult to access, el-
evation serves as a measure of remoteness within our landscape.
Elevation is also linked to a complex range of ecological attributes in
the central Annamites, including subtle variations in forest structure
and microclimate (Long, 2005; Tordoff et al., 2003).
We standardized all covariates. We tested for correlations be-
tween all possible pairs of covariates using Pearson's correlation
plots. None of our covariates were highly correlated (|r| < .6; Figure
S2).
2.4 | Modelling framework
We adopted a hierarchical multispecies occupancy model to es-
timate species occupancy and richness (Dorazio & Royle, 2005;
Dorazio, Royle, Söderström, & Glimskär, 2006). Occupancy mod-
els estimate the probability of species occupancy, ψ, while ac-
counting for species detection, p, using temporally replicated
detection/non-detection data collected across multiple sampling
locations (MacKenzie et al., 2002). To convert camera-trapping
data to an occupancy format (i.e. a detection matrix), we divided
the active camera-trapping time for each station into 10-day sam-
pling periods, yielding a minimum of six occasions for each sta-
tion. For each station and sampling period, the detection matrix
for a given species received an entr y of “1” when the species was
detected at least once during the 10-day period by at least one of
the two cameras com prisin g the station. We chose to use a 10-day
sampling period to minimize zero inflation in the detection history
matrix. We treated each leech collection event as a separate oc-
casion for the stations. We defined zij as the true occupancy state
(0 or 1) of species i at sampling station j. Occupancy state can be
mo del led as a Be rnou lli ra ndo m vari able with th e success proba bil-
ity ψij, the occupancy probability of species i at site j. We defined
pijk as detection probability for species i at station j during the kth
sampling occasion, and yijk the observation (i.e. yijk = 1 if species i
is observed at site j, occasion k, and 0 oth erw is e). Obser ving a spe-
cies is conditional on its occurrence, so that yijk can be modelled as
a Bernoulli random variable with success probability zij · pijk.
Covariate effects on both parameters can be modelled on the
logit scale. We included habitat and anthropogenic covariates on ψij
to investigate their potential effects on species occurrence. To avoid
overparameterizing the model, we first ran single-covariate models
using each of the seven covariates that we selected a priori, and as-
sessed covariate importance by evaluating effect sizes for each spe-
cies in the community. Because the environmental covariates did not
show strong effects on occupancy, with all species having 95% BCIs
ov erlappi ng ze r o and most sp eci es sh owi n g ove rlap pin g 75% Bay e sian
confidence inter vals (BCIs) (Figure S3), these covariates were not in-
cluded in the final model. Our final community model included four
covariates on ψij: city distance, village density, roads least-cost path
and elevation. Following Abrams et al. (2019), we used survey method
(camera-trap, brown leech or tiger leech) as a covariate on p. We ac-
counted for varying survey effort by including number of days each
came ra-t rap station was oper ational dur ing each 10-day occasio n (i .e .
20 days if both camera-traps at one station were operating for the
10 days) or number of leeches per sample on p.
We implemented the models in a Bayesian framework using
JAGS (Plummer, 2003) accessed through the package rjags
(Plummer, 2018). We used vague priors (e.g. normal distributions
with mean zero and variance 100 for community-level occupancy
and detection coefficients). We ran three parallel Markov chains
with 250,00 0 iterations, of which we discarded 50,0 00 as burn-in.
We assessed chain convergence using the Gelman–Rubin statis-
tic, with values close to 1 indicating convergence (Gelman, Carlin,
Stern, & Rubin, 2004). We report results as posterior mean and
standard deviation. We consider a coefficient to have strong sup-
port if the 95% Bayesian confidence interval (95% BCI, the 2.5%
and 97.5% percentiles of the posterior distribution) does not over-
lap zero and moderate support if the posterior 75% BCI does not
overlap zero. The full model description is provided in Appendix
S1.
To test for spatial autocorrelation in response variables not ac-
counted for by predictor variables, we followed an approach put
forth by Moore and Swihart (2005). We calculated Moran's I using
the moransI function from the R package lctools for residuals from
occupancy models using a neighbourhood distance of 2.5 km (aver-
age spacing of our sampling stations). We only found evidence of low
to moderate spatial autocorrelation in occupancy model residuals in
only 2 of the 23 species analysed. We acknowledge that for these
species we may underestimate occupancy. However, our analysis
432
|
TILKER ET aL.
is concerned with comparisons of patterns across study sites, not
among species, and for a given species, any bias should be similar
across the sites. Estimates of Moran's I and associated p-values for
species are shown in Appendix S2.
To predict species richness across the landscape, we first di-
vided the study area into 200 × 200 m grid cells. We included
proposed extensions for the Hue and Quang Nam Saola NRs in
the prediction area. Next, we derived covariate values for each
cell. For the city distance, village density and roads least-cost
path covariates, we followed the same protocols described above.
Elevation values were extracted from an SRTM 30 m digital eleva-
tion mo del (see Fi gur e S4 for cova ria te ra ste rs). We used es timate s
of the coefficients from the multispecies model linking covariates
to occupancy probability to predict occupancy values for each
species and grid cell and then summed the occupancy probabili-
ties for all species per cell to produce species richness maps. To
highlight areas of high richness for conservation-priority species,
we prod uce d a sepa rat e spec ies richness map for the end emic spe-
cies and those listed as Near Threatened or higher on Th e IUCN
Red List of Threatened Species. To provide a further level of detail
for the endemic species, we also produced single-species occu-
pancy maps for Annamite striped rabbit, Annamite dark muntjac
and Owston's civet. We note that, although our sampling stations
only covered part of the study sites, covariate values at the sta-
tions were largely representative of values across the sites. When
we filtered the raster cells to remove cells that fell outside the
range of our covariates at the sampling stations, we found that a
small number raster cells were excluded. However, we present the
full prediction maps here, both because the differences between
the complete and filtered rasters were minor, and for visualization
purposes (see Figure S5 for modified prediction maps).
We also used a modified Bray–Curtis index to assess compo-
sitional dissimilarity among the five study sites. The Bray–Curtis
index calculates dissimilarity values by comparing composition in a
reference assemblage with one or more target assemblages (Bray &
Curtis, 1957). We adapted the index to compare predicted species
occupancy probabilities between all possible site combinations, fol-
lowing the general framework proposed by Giacomini and Galetti
(2013). To do this, we sampled random values from the posterior
distributions of species-specific occupancy probabilities for both
the focal and target study sites. We repeated this procedure 30,000
times using Monte Carlo sampling to generate a distribution of val-
ues and took the mean of the posterior distribution. The final value
gives an indication of how dissimilar the predicted community-level
occupancies are among the sites. Dissimilarity values can range be-
tween −1 and 1. A value of 0 indicates no differences in occupancy
between the focal and reference sites, a value of 1 indicates com-
plete dissimilarity with the reference site having higher occupancies
than the focal site, and a value of −1 indicates complete dissimilarity
with the focal site having higher occupancies than the reference site.
We calculated Bray–Curtis dissimilarities first for the entire commu-
nity and then for endemic and threatened species. Further details
on the Bray–Curtis dissimilarity index are provided in Appendix S3.
3 | RESULTS
We obtained data from 139 camera-trap stations totalling 17,393 trap-
nights (Table S1). The camera-trapping yielded 5,261 independent de-
tections (∆ = 60 min between subsequent pictures of the same species
at the same camera-trap) of 27 terrestrial mammals. We identified all
mammals to species, with the exception of the ferret badgers (Melogale
personata and M. moschata) and pangolins (Manis pentadactyl and M. ja-
vanica), which we identified to the genus level due to the difficulty of
identifying to species using camer a-trap photograp hs, an d the Annamite
dark muntjac species complex Muntiacus rooseveltorum/truongsonensis,
due to its unresolved taxonomic status. Our final species list resulted in
22 mammals. We obtained 193 leech samples totalling 2,043 leeches
(1,888 brown, 155 tiger) from 98 stations (mean leeches/station = 21,
standard deviation = 22; Table S1). We were able to amplify and se-
quence DNA from 104 samples. PROTAX identified 25 mammals to the
species level and seven to the genus level. The final species list from
the leeches included 19 terrestrial mammals. Overall, the two survey
methods provided similar species lists. The exceptions were pangolin,
pig-tailed macaque Macaca leonina, spotted linsang Prionodon pardicolor
and yellow-bellied weasel Mustela kathiah, which were detected only in
the camera-traps, and marbled cat Pardofelis marmorata, which was de-
tected only in the leeches. The final species list used for the community
occupancy analysis included 23 mammals. Four of these were threat-
ened, three were Annamite endemics, and one species fit both catego-
ries. The full species list and classifications can be found in Table S3.
Detection probabilities (p) within the mammal community var-
ied among species and with respect to survey method. Estimates of
occupancy (ψ) showed extreme heterogeneity among individual spe-
cies. Estimates of p and ψ can be found in Figure S6, and full model
results are provided in Table S4.
Species-specific responses to the covariates were highly
variable within the community (Figure 2). Village density had
the strongest negative response at the community level, with a
community-level 95% BCI that did not overlap zero; at the spe-
cies level, 16 species had a moderate negative and four species
had a strong negative relationship with this covariate. No species
responded positively to village density. There was a moderate
positive relationship with elevation at the community level, with
nine and four species showing moderate and strong associations
with higher elevations, respectively. However, we also observed
negative relationships with elevation in four species (one moder-
ate and three strong). Due to mixed species-specific responses
to distance to cities, the community response was close to zero.
Specifically, six species showed a moderate positive relationship
with city distance, while three species showed a moderate neg-
ative relationship, and four species showed a strong negative re-
lationship. Responses to the road least-cost path covariate were
generally weaker than the other three covariates, and the com-
munity response was close to zero. Five species had a moderate
positive association with the road least-cost path covariate, while
four species (three moderate and one strong) showed a negative
relationship with this covariate.
|
433
TILKER E T aL.
Average predicted species richness for the full community was
substantially lower than the total number of species detected in the
study and was similar among the study sites (Table 1, Figure 3a). The
Palé area had slightly higher predicted species richness than the other
four areas. Richness of threatened and endemic species followed a
similar pattern, with all sites showing low richness relative to the total
number of conservation-priority species detected and the highest
richn ess in the Palé area (Tabl e 1, Fi gur e 3b). Pr edi cted occupan cie s for
the three Annamite endemic mammals showed heterogeneity among
species and sites (Table 1, Figure 4). Annamite dark muntjac had the
highest predicted occupancy, followed by Annamite striped rabbit,
followed by Owston's civet. All three endemics had highest predicted
occupancies in the Palé area, followed by Xe Sap NPA, Quang Nam
Saola NR, Hue Saola NR and finally Bach Ma NP. Single-species occu-
pancy predictions for all species can be found in Figure S7.
The Bray–Curtis dissimilarity index values showed similar values
for full community occupancies among Bach Ma NP, the Hue and
Quang Nam Saola NRs, and Xe Sap NPA (Table 2). However, the Palé
area had negative and high defaunation index values when compared
to every other site, indicating that occupancies for the full suite of
species are higher for this area. Dissimilarity values for endemic
and threatened species showed a similar pattern. The fact that Palé
area showed negative and higher dissimilarity values for both the
full community and for conservation-priority species suggests that,
FIGURE 2 Standardized beta coefficients (mean, 95% BCI, 75% BCI, on the logit scale) showing covariate effects on species occupancy
Grey bars show relationships in which the 75% Bayesian credible interval (BCI) does not overlap zero, black bars indicate that the 75%
interval does not overlap zero but the 95% interval does overlap zero, and red bars indicate that the 95% interval does not overlap zero. The
community response is shown in the lower panel
TABLE 1 Predicted species richness and species occupancies (mean ± SD) for five study sites in the central Annamites landscape, from
multispecies community occupancy model fit to 23 mammal species
Study site
All sitesBach Ma NP
Quang Nam
Saola NR Hue Saola NR Xe Sap NPA Palé
Species richness
Full community 6.52 ± 1.39 6.32 ± 1.00 7.00 ± 1.24 6.89 ± 0.98 8.25 ± 1.13 7.02 ± 1.41
Threatened and endemic species 1.79 ± 0.77 1.80 ± 0.65 1.85 ± 0.66 2.10 ± 0.70 2.97 ± 0.85 2.12 ± 0.89
Species occupancy
Annamite striped rabbit 0.15 ± 0.04 0.24 ± 0.04 0.20 ± 0.05 0.28 ± 0.03 0.36 ± 0.05 0.24 ± 0.09
Annamite dark muntjac 0.24 ± 0.18 0.35 ± 0.15 0.30 ± 0.18 0.41 ± 0.17 0.66 ± 0.19 0.39 ± 0.24
Owston's civet 0.05 ± 0.02 0.06 ± 0.02 0.05 ± 0.02 0.08 ± 0.02 0.13 ± 0.05 0.07 ± 0.05
Note: Full community indicates richness for all 23 species. Threatened and endemic species indicates richness for 10 species that are endemic and/ or
listed as Near Threatened or higher on The IUCN Red List of Threatened Species. Species occupancy shows predicted occupancies for each of the three
Annamite endemic mammals. Occupancy values range from 0 to 1.
434
|
TILKER ET aL.
FIGURE 3 Predicted species richness across five study sites in the central Annamites landscape with histogram showing proportion of
cells in each study area for predicted species numbers (left panel) and prediction map (right panel) based on community occupancy model fit
to camera-trapping and iDNA data for 23 mammal species. (a) Predicted richness for all species. (b) Predicted species richness for threatened
and endemic species
FIGURE 4 Predicted occupancies for three Annamite endemic species with violin plot showing predicted occupancy values for each
of the five study sites (left panel) and prediction map (right panel) based on community occupancy model fit to camera-trapping and iDNA
data for 23 mammal species. (a) Predicted occupancy for Annamite striped rabbit. (b) Predicted occupancy for Annamite dark muntjac. (c)
Predicted occupancy for Owston's civet. Note that, for visualization purposes, occupancy values are scaled independently for each species.
Single-species prediction maps with standardized scaling for occupancy values can be found in Figure S7
|
435
TILKER E T aL.
within the context of species occurrence, the site has undergone less
severe defaunation than the other sites.
4 | DISCUSSION
Our study highlights the landscape-scale effects of unsustainable
hunting on the occurrence and distribution of terrestrial mammals
within a tropical biodiversity hotspot. For a structurally intact tropi-
cal rain forest habitat, average predicted species richness (7.02) was
low (Table 1; Figure 3a, b). For comparison, Deere et al. (2018) found
a predicted species richness of 14.12 in logged forests in Malaysian
Borneo and a richness of 4.54 in adjacent oil palm plantations. Given
the largely homogenous landscape-scale forest structure and habi-
tat in our study sites, the low predicted species richness is likely in-
dicative of a community that has undergone severe hunting-driven
defaunation. The extent of faunal impoverishment is further sup-
ported by the fact that we failed to record almost half of the mam-
mal community would be expected to occur in these sites based on
historical distribution maps (Tilker et al., 2019).
At the landscape level, predicted richness was broadly similar
among the five study sites, although the more remote Palé area
showed the highest richness values, especially for the endemic and
threatened species (Figure 3). The Bray–Curtis dissimilarity values
show a similar geographical pattern, with the Palé area having higher
dissimilarity values compared to the other four areas, indicating less
defaunation (Table 2). For the three endemic species, Bach Ma NP
showing the lowest predicted occupancy, followed by the Hue Saola
NR, Quang Nam Saola NR, Xe Sap NPA and Palé. These findings in-
dicate a strong landscape-scale defaunation gradient for the three
endemic species (Figure 4). Our covariate responses suggest that
this gradient reflects an increasing level of remoteness (Figure 2).
Bach Ma NP lies near densely populated coastal areas of Vietnam,
has lower average elevations and has been accessible for decades
by a well-established road network. In the westernmost section is
the Palé area, which is far from major cities, has few villages, higher
elevations, and has only recently been accessible by road. The Saola
NRs and Xe Sap NPA fall between these two extremes. Furthermore,
these areas have had some level of active enforcement in the last few
years, which may have slowed the decline of mammal populations.
Our results provide information that is directly applicable to
conservation planning in this landscape. From a biogeographical
perspective, protecting the Palé area is a top priority for the conser-
vation of threatened and endemic species. Indeed, it may be the only
place in our sur vey sites to harbour Owston's civet, Asian black bear
Ursus thibetanus and marbled cat (Figure S7). Our predictive maps
offer a robust scientific framework to support ongoing initiatives to
grant Palé formal protected area status as a first step to implement-
ing active protection measures. Our maps also provide information
to guide targeted snare-removal efforts within protected areas. This
information is especially useful for the Hue and Quang Nam Saola
NRs, where WWF and local partners are operating snare-removal
teams, but have not yet been able to significantly reduce snaring
pressure across the wider protected area complex (Wilkinson, 2016).
We suggest that, to maximize the impact of snare-removal efforts on
conservation-priority species, the teams should focus on the more
remote areas along the border of the two reserves, and in the bor-
der area of the Quang Nam Saola NR and Bach Ma NP. It is possible
that these areas have maintained higher occupancies of conserva-
tion-priority species because they are more difficult to access and,
as a result, have cumulatively experienced less snaring pressure.
However, remoteness will not protect these areas for long. An in-
crease in road development in recent years has created a situation
where even the most remote locations in the Saola NRs can now be
reached within a single day from the nearest access point, meaning
that no area is inaccessible for a motivated hunter. Given the likely
relationship between accessibility and increased hunting pressure, it
seems inevitable that, in the absence of scaled-up enforcement ef-
forts, snaring pressure will continue to increase in the more remote
areas, especially as other parts of the protected areas become in-
creasingly empty and poachers are forced to travel further distances
to maintain comparable levels of offtake (Kümpel, Milner-Gulland,
TABLE 2 Bray–Curtis dissimilarity values (mean ± SE) calculated using predicted occupancy values per site for all mammal species (top
value) and for threatened and endemic species (bottom value)
Bach Ma NP Quang Nam Saola NR Hue Saola NR Xe Sap NPA Palé
Bach Ma NP 0−0.0006 ± 0.0205 0.0406 ± 0.0119 0.0350 ± 0.0233 0.1338 ± 0.0315
0.0431 ± 0.0205 0.0049 ± 0.0119 0.0841 ± 0.0233 0.2547 ± 0.0315
Quang Nam Saola
NR
0.0006 ± 0.0205 00.0412 ± 0.0206 0.0357 ± 0.0 061 0.1346 ± 0.0193
−0.0431 ± 0.0205 −0.0382 ± 0.0205 0.04127 ± 0.006 0.2150 ± 0.0193
Hue Saola NR −0.0406 ± 0.0119 −0.0412 ± 0.0205 0−0.0 055 ± 0.0206 0.0938 ± 0.0297
−0.0049 ± 0.0119 0.0382 ± 0.0206 0.0793 ± 0.0206 0.2503 ± 0.0297
Xe Sap NPA −0.0350 ± 0.0233 −0.0357 ± 0.0061 0.0055 ± 0.0206 00.0995 ± 0.0174
−0.0840 ± 0.0233 −0.0413 ± 0.006 −0.0792 ± 0.0206 0.1754 ± 0.0174
Palé −0.1338 ± 0.0315 −0.1346 ± 0.0193 −0.0938 ± 0.0297 −0.0995 ± 0.0174
−0.2547 ± 0.0315 −0.2149 ± 0.0192 −0.2502 ± 0.0297 −0.1754 ± 0.0174 0
Note: Values range from 0 to 1, with 0 indicating complete similarity in species occupancies between the sites and 1 indicating complete dissimilarity.
436
|
TILKER ET aL.
Cowlishaw, & Rowcliffe, 2010). Although threatened and endemic
species appear to be absent from much of Bach Ma NP, there are
isolated high-elevation areas that should be considered for inten-
sive anti-poaching efforts. Our models indicate that the border areas
of eastern Xe Sap NPA Area have undergone moderate-to-severe
defaunation. Given that this area appears to be heavily hunted by
Vietnamese poachers (Tilker, 2014), such a finding is not unexpected.
The eastern section of the protected area is nonetheless a top prior-
it y for cont inu ed prote c tio n effo r ts, bo th bec aus e it ma y be a str ong-
hold for the endemic Annamite striped rabbit (Figure S7), and more
generally because effective enforcement can serve as a buffer from
furth er cross-bo rder in cur sions into the Pal é area. In a bes t-ca se sce-
nario, reducing snaring pressure in core areas within this landscape
could not only prevent the local extirpation of conservation-priority
species, but also allow their populations to rebound (see Steinmetz,
Chutipong, Seuaturien, Chirngsaard, & Khaengkhetkarn, 2010 for a
case study on large mammal population recovery following mitiga-
tion of unsustainable hunting pressure).
In many ways, our study landscape exemplifies classical “empty
forest syndrome” (Redford, 1992). All large- and medium-sized
predators (with the exception of Asiatic black bear), as well as all
megaherbivores, appear to be locally extirpated (Tilker et al., 2019).
Large ungulates have been hunted out from most of the landscape
(Figure S7). Yet our findings show that even in this empty forest,
conservation-prior it y species still pe rsist, alb eit at extremely low oc-
cupancies. Based on these results, we suggest that the conservation
potential of defaunated landscapes should not automatically be dis-
missed in the absence of comprehensive surveys. It is important that
such surveys use sufficient sampling effor t and be conducted over
a large spatial extent for two reasons. First, species often show ex-
treme spatial heterogeneity in defaunated landscapes because local
extinctions necessarily result in reduced, often patchy, distributions.
Surveys over wider areas are more likely to detect remnant popula-
tions. Second, working over larger spatial scales may better capture
the underlying factors influencing species distribution, which can be
especially important in landscapes characterized by complex anthro-
pogenic pressures operating at multiple spatial scales. In our study,
it was only by sampling the wider forest complex that we were able
to adequately characterize the full spectrum of anthropogenic fac-
tors that appear to impact species occurrence patterns. Large-scale
surveys require substantial resources. We acknowledge that, with
multiple competing conservation objectives and finite resources,
landscape-scale surveys may not always be possible. However, we
note that because this approach can enhance the efficiency of tar-
geted interventions, it is possible that limited conservation resources
may be saved in the long term.
To overcome the challenge of detecting rare and elusive tropi-
cal mammal species, we used two complimentary survey methods:
camera-trapping and leech collection. Although camera-trapping de-
tected more species overall, leeches provided our sole detection for
marbled cat and doubled the number of records for two rare species,
Owston's civet and Asian black bear. Moreover, while camera-trap-
ping detection probabilities were higher for most species in our
analysis, leeches had higher average detection rates for both Asian
black bear and the endemic dark muntjac (Figure S7). Our results are
consistent with the findings of Abrams et al. (2019) that demonstrate
the advantages of using both camera-trapping and eDNA to increase
detection probabilities for tropical mammals. We further suggest
that because utilizing multiple methods may increase detections
of rare species, this approach could be especially important when
surveying faunally impoverished systems. Future surveys using joint
detection methods need not rely only on leeches but could use other
sources of eDNA , such as water (Ushio et al., 2017) or ticks (Gariepy,
Lindsay, Ogden, & Gregor y, 2012), or incorporate other non-invasive
sampling techniques, such as acoustic monitoring devices (Kalan et
al., 2015). The Bayesian modelling approach that we used, adapted
from Abrams et al. (2019), is flexible with regard to the underlying
detection method used to generate spatial or temporal replicates.
We found that species occurrence in our study area appears to
be primarily driven by anthropogenic factors, with no strong influ-
ence from the habitat covariates that we assessed in our models
(Figure S3). This finding was unexpected, given the importance of
vegetation structure in explaining mammal occurrence patterns in
other tropical rain forests (Goulart et al., 2009; Mathai et al., 2017;
Sollmann et al., 2017). One possible explanation is that, as anthro-
pogenic pressures in a landscape increase, ecological relationships
weaken. Several hypothetical scenarios could give rise to this sit-
uation. Spatially non-random hunting pressure could, for example,
differentially impact areas of preferred habitat, leaving higher occu-
pancies in less suitable areas. Alternatively, intensive hunting across
a landscape could drive stochastic local extinctions, leaving remnant
populations that are distributed randomly with respect to habitat.
Regardless of the underlying process, the failure of habitat-based
indices to reflect faunal biodiversity, thus the “environmental de-
coupling” of species–habitat relationships, has broad implications.
Biodiversity assessments that rely solely on remote-sensed habi-
tat-based measures may provide information that is inaccurate be-
cause they do not accurately capture species occurrence patterns.
Recently, a growing number of scientists have called for the devel-
opment of standardized remote-sensing parameters, often referred
to as Satellite Remote-Sensing Essential Biodiversity Variables (SRS-
EBVs), to monitor biodiversity at the global scale (O'Connor et al.,
2015; Pettorelli et al., 2016; Skidmore et al., 2015). While we ac-
knowledge the value of earth observation data to provide insight
into biodiversity patterns and processes at large scales (Bush et al.,
2017), our results indicate that remote-sensed habitat-based mea-
sures may provide little information on the status or distribution of
wildlife in defaunated landscapes. In tropical rain forests subject to
hunting pressure, there is likely no substitute for large-scale in situ
surveys to collect primary biodiversity data.
Our results underscore the importance of incorporating an-
thropogenic factors in studies that seek to explain or predict
species occurrence in landscapes characterized by high human
pressure. Furthermore, our findings suggest that to build robust
distribution models it may be beneficial to incorporate a diverse
suite of anthropogenic covariates that capture different aspects
|
437
TILKER E T aL.
of this pressure. We used measures of village density and city dis-
tance as proxies for accessibility at the local and landscape scales,
respectively. Previous studies have shown the impact of similar
accessibility measures on wildlife communities at different spatial
scales (Koerner et al., 2017; Schuette, Wagner, Wagner, & Creel,
2013; Torres et al., 2018). Our least-cost path covariate adds an
additional dimension to these accessibility measures, both be-
cause it takes into account the ruggedness of the terrain in our
landscape, and because it is calculated over a 20-year window.
Finally, we use elevation as a proxy for both local accessibility
and a complex set of ecological attributes. The relative contribu-
tion of anthropogenic and environmental traits to species occur-
rence along elevational gradients in the Annamites represents an
intriguing question. Future studies in the region that measure a
wider range of microhabitat characteristics, and ideally are con-
ducted in areas under less severe hunting pressure, may provide
insight into this issue.
Given the current magnitude of hunting across the world's
tropical rain forests (Benítez-López et al., 2017; Harrison, 2011;
Ripple et al., 2016), and future projections for population growth
(Gerland et al., 2014) and road expansion in developing countries
(Laurance et al., 2014), it is likely that defaunation will become in-
creasingly prevalent in tropical regions. Confronting the pantrop-
ical defaunation crisis will require a well-resourced, multi-faceted
approach from conservation stakeholders worldwide. Because
specific threats and potential solutions necessarily depend on
local context, effective strategies to prevent unsustainable
hunting must be site-specific. One constant that is applicable to
conservation initiatives in all tropical hotspots, however, is that
resources are limited. We show that, within this context, under-
standing spatial patterns of defaunation can help stakeholders
prioritize areas for conservation activities and therefore more ef-
fectively use finite conservation resources.
ACKNOWLEDGEMENTS
We thank the staff of the WWF-CarBi project for providing ex-
tensive logistical support; Bach Ma NP, the Thua Thien Hue and
Quang Nam Saola NRs, and Xe Sap NPA for providing permissions
and personnel to conduct this research; and our field team leaders
in Vietnam and Laos. Funding for the surveys was provided by the
German Federal Ministry of Education and Research (BMBF FKZ:
01LN1301A), Leibniz-IZW, Point Defiance Zoo and Aquarium, Safari
Club International and Critical Ecosystem Partnership Fund. AT re-
ceived support through a Fulbright scholarship. The CarBi project
was provided by Internationale Klimaschutzinitiative of the Federal
Ministry for the Environment, Nature Conservation, and Nuclear
Safety (BMBU) and Kreditanstalt für Wiederaufbau (KfW).
DATA AVAIL AB I LI T Y STATE MEN T
Underlying data and R code that support this study are available on
the DRYAD digital repository https ://doi.org/10.5061/dryad.ns1rn
8pnx.
ORCID
Andrew Tilker https://orcid.org/0000-0003-3630-8691
Jesse F. Abrams https://orcid.org/0000-0003-0411-8519
Jan Axtner https://orcid.org/0000-0003-1269-5586
Julie Louvrier https://orcid.org/0000-0003-1252-1746
Benjamin M. Rawson https://orcid.org/0000-0003-4141-5985
Minh Le https://orcid.org/0000-0002-2953-2815
Rahel Sollmann https://orcid.org/0000-0002-1607-2039
Andreas Wilting https://orcid.org/0000-0001-5073-9186
REFERENCES
Abrahams, M. I., Peres, C. A., & Costa, H. C. (2017). Measuring local
depletion of terrestrial game vertebrates by central-place hunt-
ers in rural Amazonia. PLoS ONE, 12(10), e0186653. https ://doi.
org/10.1371/journ al.pone.0186653
Abrams, J. F., Axtner, J., Bhagwat, T., Mohamed, A., Nguyen, A., Niedballa,
J., … Wilting, A. (2018). Studying terrestrial mammals in tropical rainfor-
ests. A user guide for camera-trapping an d environmental DNA. Berlin,
Germany: Leibniz-IZW.
Abrams, J. F., Hoerig, L., Brozovic, R., Axtner, J., Crampton-Platt, A.,
Mohamed, A., … Wilting, A. (2019). Shifting up a gear with iDNA:
From mammal detection events to standardised surveys. Journal of
Applied Ecology, 56(7), 1637–1648.
Alberti, G. (2019). movecost: An R package for calculating accumulated
slope-dependent anisotropic cost-surfaces and least-cost paths.
Softwa re X, 10, 100331. https ://doi.org/10.1016/j.sof tx.2019.100331
Alroy, J. (2017). Effects of habitat disturbance on tropical forest bio-
diversity. Proceedings of the National Academy of Sciences, 114(23),
6056–6061.
Axtner, J., Crampton-Platt, A., Hörig, L., Mohamed, A., Xu, C. C. Y., Yu,
D. W., & Wilting, A. (2019). An efficient and improved laboratory
workflow and tetrapod database for larger scale eDNA studies.
GigaScience, 8(4). https ://doi.org/10.1093/gigas cienc e/giz029
Benítez-López, A., Alkemade, R., Schipper, A . M., Ingram, D. J., Verweij,
P. A., Eikelboom, J. A. J., & Huijbregts, M. A . J. (2017). The impact of
hunting on tropical mammal and bird populations. Science, 356(6334),
180–183.
Bodmer, R . E., Eisenberg, J. F., & Redford, K . H. (1997). Hunting and the
likelihood of extinction of Amazonian mammals. Conservation Biology,
11(2), 460–466. https ://doi.org/10.1046/j.1523-1739.1997.96022.x
Bohmann, K., Schnell, I. B., & Gilbert, M. T. P. (2013). When bugs re-
veal biodiversit y. Molecular Ecology, 22(4), 909–911. https ://doi.
org /10.1111/me c.12221
Bradshaw, C. J., Sodhi, N. S., & Brook, B. W. (2009). Tropical turmoil:
A biodiversity tragedy in progress. Frontiers in Ecology and the
Environment, 7(2), 79–87. https ://doi.org/10.1890/070193
Bray, J. R., & Curtis, J. T. (1957). An ordination of the upland forest com-
munities of southern Wisconsin. Ecological Monographs, 27(4), 325–
349. https ://doi.org/10.2307/1942268
Brook, S. M., Dudley, N., Mahood, S. P., Polet, G., Williams, A. C.,
Duckworth, J. W., … Long, B. (2014). Lessons learned from the loss of
a flagship: The extinction of the Javan rhinoceros Rhinoceros sonda-
icus annamiticus from Vietnam. Biological Conservation, 174, 21–2 9.
Burghouts, T., Ernsting, G., Korthals, G., & De Vries, T. (1992). Litter fall,
leaf litter decomposition and litter invertebrates in primary and se-
lectively logged dipterocarp forest in Sabah, Malaysia. Philosophical
Transactions of the Royal Society of Lo ndon. Series B: B iological Scien ces,
335(1275), 407–416.
Bush, A., Sollmann, R., Wilting, A., Bohmann, K ., Cole, B., Balzter, H., …
Yu, D. W. (2017). Connecting Earth observation to high-throughput
biodiversity data. Nature Ecology & Evolution, 1(7), 0176. https ://doi.
org /10.1038/s41559- 017-0176
438
|
TILKER ET aL.
Cardillo, M., Mace, G. M., Jones, K. E., Bielby, J., Bininda-Emonds, O. R.,
Sechrest, W., … Purvis, A . (20 05). Multiple causes of high extinction
risk in large mammal species. Science, 309(5738), 1239–1241.
Cayuela, L., Golicher, D. J., Newton, A. C., Kolb, M., De Alburquerque,
F. S., Arets, E. J., … Pérez, A. M. (2009). Species distribution mod-
eli ng in the tropics: Problems, potenti alities, and the role of biolog-
ical data for effective species conservation. Tropical Conservation
Science, 2(3), 319–352. ht tps ://doi.o rg /10.1177/19400 82909
00200304
Chazdon, R. L. (20 03). Tropical forest recovery: Legacies of
human impact and natural disturbances. Perspective s in Plant
Ecology, Evolution and System atics, 6(1–2), 51–71. ht tps ://doi.
org /10.1078/1433-8319- 00042
Davidson, A . D., Hamilton, M. J., Boyer, A. G., Brown, J. H., & Ceballos,
G. (2009). Multiple ecological pathways to extinction in mam-
mals. Proceedings of the National Academy of Science s of the United
States of America, 106(26), 10702–10705. https ://doi.org/10.1073/
pnas.09019 56106
Deere, N. J., Guillera-Arroita, G., Baking, E. L., Bernard, H., Pfeifer, M.,
Reynolds , G. , … Struebig, M. J. (2018). High carbon sto ck forests pro-
vide co-benefits for tropical biodiversity. Journal of Applied Ecology,
55(2), 997–1008. https ://doi.org/10.1111/1365-2664.13023
Dias, D. D. M., Lima Massara, R., de Campos, C. B., & Henrique Guimarães
Rodrigues, F. (2019). Human activities influence the occupancy prob-
ability of mammalian carnivores in the Brazilian Caatinga. Biotropica,
51(2), 253–265. https ://doi.org/10.1111/btp.12628
Dorazio, R. M., & Royle, J. A. (2005). Estimating size and com-position
of biological communities by modeling the occurrence of species.
Journal of the American Statis tical Association, 100, 389–398. https ://
doi.org/10.1198/01621 45050 00000015
Dorazio, R. M., Royle, J. A ., Söderström, B., & Glimskär, A. (20 06).
Estimating species richness and accumulation by modeling species
occurrence and detectability. Ecolog y, 87(4), 842–854. https ://doi.or
g/10.1890/0012-9658(2006)87[842:ESRAA B]2.0.CO;2
Drouilly, M., Clark, A., & O'Riain, M. J. (2018). Multi-species occupancy
modelling of mammal and ground bird communities in rangeland in
the Karoo: A case for dryland systems globally. Biological Conservation,
224, 16–25. https ://doi.org/10.1016/j.biocon.2018.05.013
Duckworth, J. W., Batters, G., Belant, J. L., Bennett, E. L ., Brunner, J.,
Burton, J., … Wirth, R. (2012). Why South-East Asia should be the
world's priority for averting imminent species extinctions, and a call
to join a developing cross-institutional programme to tackle this
urgent issue. Surveys and Perspectives Integrating Environment and
Society, 5, 2.
Duckworth, J. W., Timmins, R., Chutipong, W., Gray, T. N. E., Long, B.,
Helgen, K., … Willcox, D. H. A. (2016). Arctony x collaris. The IUCN Red
List of Threatened Species, 2016: e.T70205537A45209459.
Espinosa, S., Branch, L. C., & Cueva, R. (2014). Road development and
the geography of hunting by an Amazonian indigenous group:
Consequences for wildlife conservation. PLoS ONE, 9(12), e114916.
https ://doi.org/10.1371/journ al.pone.0114916
Facelli, J. M., & Pickett, S. T. (1991). Plant litter: Its dynamics and effects
on plant community structure. The Botanical Review, 57(1), 1–32.
h t t p s : / / d o i . o r g / 1 0 . 1 0 0 7 / B F 0 2 8 5 8 7 6 3
Ganas, J., & Lindsell, J. A . (2010). Photographic evidence of Jentink's
duiker in the Gola Forest Reserves. Sierra Leone. African Jour nal
of Ecology, 48(2), 566–568. https ://doi.org/10.1111/j.1365-2028.
20 09.01137. x
Gariepy, T. D., Lindsay, R., Ogden, N., & Gregory, T. R. (2012). Identif ying
the last supper: Utility of the DNA barcode library for bloodmeal
identification in ticks. Molecular Ecology Resources, 12(4), 646–652.
https ://doi.org/10.1111/j.1755-0998.2012.03140.x
Gelman, A., Carlin, J. B., Stern, H. S., & Rubin, D. B. (2004). Bayesian data
analysis. Boca Raton, FL: Chapman and Hall.
Gerland, P., Raftery, A. E., Ševčíková, H., Li, N., Gu, D., Spoorenberg, T.,
… Bay, G. (2014). World population stabilization unlikely this centur y.
Science, 346(6206), 234–237.
Giacomini, H. C., & Galetti, M. (2013). An index for defaunation.
Biological Conservation, 163, 33–41. https ://doi.org/10.1016/j.
biocon.2013.04.007
GIMP team (2017). GNU image manipulation program. Retrieved from
https ://www.gimp.org
Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., & Moore,
R. (2017). Google earth engine: Planetary-scale geospatial analysis
for everyone. Remote Sensing of Environment, 202, 18–27. https ://doi.
org/10.1016/j.rse.2017.06.031
Goular t, F. V. B., Cáceres, N. C ., Graipel , M. E., Tort ato, M. A., Ghizoni,
I. R. Jr, & Oliveira-Santos, L. G. R. (2009). Habitat selection by
large mammals in a southern Brazilian Atlantic Forest. Mammalian
Biology, 74(3), 182–190. https ://doi.org/10.1016/j.mambio.2009.
02.006
Gray, T. N., Hughes, A. C., Laurance, W. F., Long, B., Lynam, A. J., O'Kelly,
H., … Wilkinson, N. M. (2018). The wildlife snaring crisis: An insidious
and pervasive threat to biodiversit y in Southeast Asia. Biodiversity
and Conservation, 27 (4), 1031–1037. https ://doi.org/10.10 07/
s10531-017-1450-5
Guisan, A., Tingley, R., Baumgartner, J. B., Naujokaitis-Lewis, I., Sutcliffe,
P. R., Tulloch, A. I. T., … Buckley, Y. M. (2013). Predicting species dis-
tributions for conservation decisions. Ecology Letters, 16(12), 1424–
1435. ht tp s ://doi.org/10.1111/ele.12189
Harrison, R. D. (2011). Emptying the forest: Hunting and the extirpation
of wildlife from tropical nature reserves. BioScience, 61(11), 919–924.
https ://doi.org/10.1525/bio.2011.61.11.11
Harrison, R. D., Sreekar, R., Brodie, J. F., Brook, S., Luskin, M., O'Kelly,
H., … Velho, N. (2016). Impacts of hunting on tropical forests in
Southeast Asia. Conservation Biology, 30(5), 972–981. https ://doi.
org /10.1111/cobi.12785
Hijmans, R. J. (2019). raster: Geographic data analysis and modeling. R
pac ka ge version 2.8-19. Retrieved fro m ht tps ://CRAN .R-pr oje ct.org/
packa ge=raster
Kalan, A. K., Mundry, R., Wagner, O. J., Heinicke, S., Boesch, C., &
Kühl, H. S. (2015). Towards the automated detection and occu-
pancy estimation of primates using passive acoustic monitoring.
Ecological Indicators, 54, 217–226. https ://doi.org/10.1016/j.ecoli
nd.2015.02.023
Koerner, S. E., Poulsen, J. R., Blanchard, E. J., Okouyi, J., & Clark, C.
J. (2017). Vertebrate community composition and diversity de-
clines along a defaunation gradient radiating from rural villages
in Gabon. Journal of Applied Ecology, 54(3), 805–814. https ://doi.
org /10.1111/1365-2664.12798
Kümpel, N. F., Milner-Gulland, E. J., Cowlishaw, G. U. Y., & Rowcliffe, J.
M. (2010). Assessing sustainability at multiple scales in a rotational
bushmeat hunting system. Conser vation Biology, 24 (3), 861–871.
https ://doi.org/10.1111/j.1523-1739.2010.01505.x
Laurance, W. F., Clements, G. R., Sloan, S., O'Connell, C. S., Mueller, N. D.,
Goosem, M., … Arrea, I. B. (2014). A global strategy for road building.
Nature, 513 (7517), 229. https ://doi.org/10.1038/natur e13717
Li, X., Bleisch, W. V., & Jiang, X. (2018). Using large spatial scale cam-
era trap data and hierarchical occupancy models to evaluate species
richness and occupancy of rare and elusive wildlife communities
in southwest China. Diversity and Distributions, 24(11), 1560–1572.
https ://doi.org/10.1111/ddi.12792
Long, B. (2005). Identification of priority areas for integrated conservation
management in Quang Nam Province, Vietnam. (unpublished doctoral
thesis). Cambridge, UK.
Long, B., Hoang, M., & Truyen, T. (2005). A Conservation Assessment
of Quang Na m Province, Central V ietnam. Tam Ky, Vietnam: WWF
Indochina and Quang Nam Forest Protection Department.
|
439
TILKER E T aL.
MacKenzie, D. I., Nichols, J. D., Lachman, G. B., Droege, S., Andrew Royle,
J., & Langtimm, C. A. (2002). Estimating site occupancy rates when
detection probabilities are less than one. Ecology, 83(8), 2248–2255.
https ://doi.org/10.1890/0012-9658(2002)083[2248:ESORW
D]2.0.CO;2
MacKenzie, D. I., Nichols, J. D., Royle, J. A., Pollock, K. H., Bailey, L., &
Hines, J. E. (2017). Occupancy estimation and modeling: Inferring
patterns and dynamics of species occurrence. London, UK: Academic
Press.
MacMillan, D. C., & Nguyen, Q. A. (2014). Factors influencing the illegal
harvest of wildlife by trapping and snaring among the Katu ethnic
group in Vietnam. Oryx, 48(2), 304–312. https ://doi.org/10.1017/
S 0 0 3 0 6 0 5 3 1 2 0 0 1 4 4 5
Martin, E. H., Cavada, N., Ndibalema, V. G., & Rovero, F. (2015). Modelling
fine-scale habitat associations of medium-to-large forest mammals
in the Udzungwa Mountains of Tanzania using camera trapping.
Tropical Zoology, 28 (4), 137–151. https ://doi.org/10.1080/03946
975.2015.1078568
Mathai, J., Sollmann, R., Meredith, M. E., Belant, J. L., Niedballa, J.,
Buckingham, L., … Wilting, A. (2017). Fine-scale distributions
of carnivores in a logging concession in Sarawak, Malaysian
Borneo. Mammalian Biology, 86, 56–65. https ://doi.org/10.1016/j.
mambio.2017.04.002
Matusch, T. (2014). Islands of felicit y?–The effect of land cover changes
in and around protected areas: A case study of Bach Ma National
Park, Vietnam. American Journal of Environmental Protection, 3(3),
1 5 2 –1 6 1 . h t t p s : / / d o i . o r g / 1 0 . 1 1 6 4 8 / j . a j e p . 2 0 1 4 0 3 0 3 . 1 7
Meyfroidt, P., & Lambin, E. F. (2008). Forest transition in Vietnam and
its environmental impacts. Global Change Biology, 14(6), 1319–1336.
https ://doi.org/10.1111/j .1365-2486.200 8.01575. x
Moore, J. E., & Swihart, R. K. (2005). Modeling patch occupancy by forest
rodents: incorporating detectability and spatial autocorrelation with
hierarchically structured data. The Journal of Wildlife Management,
69(3), 933–949.
Niedballa, J., Sollmann, R., Courtiol, A., & Wilting, A. (2016). camtrapR:
An R package for efficient camera trap data management. Meth ods in
Ecology and Evolution, 7(12), 1457–1462.
O'Connor, B., Secades, C., Penner, J., Sonnenschein, R., Skidmore,
A., Burgess, N. D., & Hut ton, J. M. (2015). Ear th observation as a
tool for tracking progress towards the Aichi Biodiversity Targets.
Remote Sensing in Ecolog y and Conservation, 1(1), 19–28. https ://doi.
org/10.1002/rse2.4
Pettorelli, N., Wegmann, M., Skidmore, A., Mücher, S., Dawson, T. P.,
Fernandez, M., … Geller, G. N. (2016). Framing the concept of satel-
lite remote sensing essential biodiversity variables: Challenges and
future directions. Remote Sensing in Ecology a nd Conservation, 2(3),
122–131. https ://doi.org/10.1002/rse2.15
Plummer, M. (2003). JAGS: A program for analysis of Bayesian graphi-
cal models using Gibbs sampling (Vol. 124 (125), p. 10). Proceedings
of the 3rd International Workshop on Distributed Statistical
Computing.
Plummer, M. (2018). rjags: Bayesian graphical models using MCMC. R pack-
age version 4–8. Retrieved from https ://CRAN.R-proje ct.org/packa
ge=rjags
QGIS Development Team. (2016). QGIS geographic information system.
Open Source Geospatial Foundation Project. http://qgis.osgeo.org
R Core Team (2018). R: A language and environment for statis tical com put-
ing. Vienna, Austria: R Foundation for Statistical Computing.
Raloff, J. (1999). Rarest of the rare: Remote-camera images and dung-
heap data give a portrait of Vietnam's rhinos. Science News, 156(10 ),
153–155. https ://doi.o rg /10. 2307/4011765
Rao, M., Myint, T., Zaw, T., & Htun, S. (2005). Hunting patterns in tropi-
cal forests adjoining the Hkakaborazi National Park, north Myanmar.
Oryx, 39(3), 292–300. https ://doi.org/10.1017/S0030 60530
5000724
Redford, K. H. (1992). The empty forest. BioScience, 42(6), 412–422.
htt ps ://doi.org/10.2307/1311860
Richardson, R. T., Bengtsson-Palme, J., & Johnson, R. M. (2017).
Evaluating and optimizing the performance of software commonly
used for the taxonomic classification of DNA metabarcoding se-
quence data. Molecular Ecology Resources, 17(4), 760–769. ht tps ://
doi.org/10.1111/1755- 0998.12628
Ripple, W. J., Abernethy, K., Betts, M. G., Chapron, G., Dirzo, R., Galetti,
M., … Young, H. (2016). Bushmeat hunting and extinction risk to the
world's mammals. Royal Society Open Science, 3(10), 160498. https ://
doi.org/10.1098/rsos.160498
Rodgers, T. W., Xu, C. C., Giacalone, J., Kapheim, K. M., Saltonstall, K.,
Vargas, M., … Jansen, P. A. (2017). Carrion fly-derived DNA me-
tabarcoding is an effective tool for mammal surveys: Evidence from
a known tropical mammal community. Molecular Ecology Resources,
17(6), e133–e145. https ://doi.org/10.1111/1755-0998.12701
Rodríguez, J. P., Brotons, L., Bustamante, J., & Seoane, J. (2007). The ap-
plication of predictive modelling of species distribution to biodiver-
sity conservation. Diversity and Distributions, 13(3), 243–251. https ://
doi.org/10.1111/j.1472-4 642. 20 07.0 0356.x
Rosa, I. M., Smith, M. J., Wearn, O. R., Purves, D., & Ewers, R. M.
(2016). The environmental legacy of modern tropical deforesta-
tion. Current Biology, 26(16), 2161–2166. https ://doi.org /10.1016/j.
cub.2016.06.013
Sandalj, M., Treydte, A. C., & Ziegler, S. (2016). Is wild meat a luxury?
Quantif ying wild meat demand and availability in Hue, Vietnam.
Biological Conservation, 194, 105–112.
Schnell, I. B., Bohmann, K., Schultze, S. E., Richter, S. R., Murray, D.
C., Sinding, M. H. S., … Gilbert, M. T. P. (2018). Debugging di-
versity–a pan-continental exploration of the potential of ter-
restrial blood-feeding leeches as a ver tebrate monitoring tool.
Molecular Ecology Resources, 18(6), 1282–1298. https ://doi.
org /10.1111/1755-0998.12912
Schnell, I. B., Sollmann, R., Calvignac-Spencer, S., Siddall, M. E., Douglas,
W. Y., Wilting, A., & Gilbert, M. T. P. (2015). iDNA from terrestrial
haematophagous leeches as a wildlife surveying and monitoring
tool–prospects, pitfalls and avenues to be developed. Frontiers in
Zoology, 12(1), 24. https ://doi.org/10.1186/s12983-015-0115-z
Schnell, I. B., Thomsen, P. F., Wilkinson, N., Rasmussen, M., Jensen, L. R.,
Willerslev, E., … Gilbert, M. T. (2012). Screening mammal biodiversity
using DNA from leeches. Current Biology, 22(8), R262–R263. https ://
doi.org/10.1016/j.cub.2012.10.014
Schuette, P., Wagner, A. P., Wagner, M. E., & Creel, S. (2013). Occupancy
patterns and niche partitioning within a diverse carnivore commu-
nity exposed to anthropogenic pressures. Biological Conservation,
158, 301–312. https ://doi.org/10.1016/j.biocon.2012.08.008
Skidmore, A. K., Pettorelli, N., Coops, N. C., Geller, G. N., Hansen, M.,
Lucas, R., … Wegmann, M. (2015). Environmental science: Agree on
biodiversity metrics to track from space. Nature News, 523( 7561),
403. https ://doi.org/10.1038/523403a
Sollmann, R., Mohamed, A., Niedballa, J., Bender, J., Ambu, L., Lagan, P.,
… Wilting, A. (2017). Quantifying mammal biodiversity co-benefits in
certified tropical forests. Diversity and Distributions, 23(3), 317–328.
https ://doi.org/10.1111/ddi.1253 0
Somervuo, P., Koskela, S., Pennanen, J., Henrik Nilsson, R., & Ovaskainen,
O. (2016). Unbiased probabilistic taxonomic classification for
DNA barcoding. Bioinformatics, 32(19), 2920–2927. https ://doi.
org/10.1093/bioin forma tics/btw346
Steinmetz, R., Chutipong, W., Seuaturien, N., Chirngsaard, E., &
Khaengkhetkarn, M. (2010). Population recovery patterns of
Southeast Asian ungulates after poaching. Biological Conservation,
143 (1), 42–51. https ://doi.org/10.1016/j.biocon.2009.08.023
Sterling, E. J., & Hurley, M. M. (2005). Conserving biodiversity in Vietnam:
Applying biogeography to conservation research. Proceedi ngs of the
California Academy of Sciences, 56(1), 98–113.
440
|
TILKER ET aL.
Tilker, A. (2014). A survey of eastern areas of Xe Sap National Protected
Area, Lao PDR, for Saola and other large u ngulates; final report to Global
Wildlife Conservation and the Saola Working Group. Vientiane, Lao
PDR: World Wildlife Fund Greater Mekong Program.
Tilker, A., Abrams, J. F., Mohamed, A ., Nguyen, A . N., Wong, S. T.,
Sollmann, R., … Wilting, A. (2019). Habitat degradation and indis-
criminate hunting differentially impact faunal communities in the
Southeast Asian tropical biodiversity hotspot. Communications
Biology, 2(1), 1–11. https ://doi.org/10.1038/s42003-019-0640-y
Timmins, R. J., Duckworth, J. W., Robichaud, W., Long, B., Gray, T. N.
E., & Tilker, A. (2016). Muntiacus vuquangensis. The IUCN Red List of
Threatened Species, 2016: e.T44703A22153828.
Tobler, M. W., Carrillo-Percastegui, S. E., Pitman, R. L., Mares, R., & Powell,
G. (2008). An evaluation of camera traps for inventor ying large-and
medium-sized terrestrial rainforest mammals. Animal Conservation,
11(3), 169–178. https ://doi.org/10.1111/j.1469-1795.2008.00169.x
Tobler, M. W., Hartley, A. Z., Carrillo-Percastegui, S. E., & Powell, G. V.
(2015). Spatiotemporal hierarchical modelling of species richness
and occupancy using camera trap data. Journal of Applied Ecology,
52(2), 413–421. https ://doi.org/10.1111/1365-2664.12399
Tordoff, A., Timmins, R., Smith, R., & Vinh, M. K. (2003). A Biological
Assessment of the Central Truong Son L andscape. Central Truong Son
Initiative. Report No. 1. Hanoi, Vietnam: WWF Indochina.
Torres, P. C., Morsello, C., Parry, L., Barlow, J., Ferreira, J., Gardner, T.,
& Pardini, R. (2018). Landscape correlates of bushmeat consump-
tion and hunting in a post-frontier Amazonian region. Environmental
Conservation, 45(4), 315–323. https ://doi.org/10.1017/S0376 89291
7000510
Ushio, M., Fukuda, H., Inoue, T., Makoto, K., Kishida, O., Sato, K., … Miya,
M. (2017). Environmenta l DNA enables detection of terres trial mam-
mals from forest pond water. Molecular Ecology Resources, 17(6) , e63–
e75. https ://doi.o rg /10.1111/1755-0998 .12690
Van Etten, J. (2017). R package gdistance: Distances and routes on geo-
graphical grids. Journal of Statistical Sof tware, 76 (3), 1–21.
Van Song, N. (2003). Wildlife trading in Vietnam: Why it flourishes.
Singapore: Economy and Environment Program for Southeast Asia,
International Development Research Centre Regional Office for
Southeast and East Asia, Singapore.
Vit t, L. J., & Caldwell, J. P. (1994). Resource utilization and guild structure
of small vertebrates in the Amazon forest leaf litter. Journal of Zoolo gy,
234(3), 463–476. https ://doi.org/10.1111/j.1469-7998.1994.tb048
60.x
Walston, J., Robinson, J. G., Bennett, E. L., Breitenmoser, U., da Fonseca,
G. A., Goodrich, J., … Leader-Williams, N. (2010). Bringing the tiger
back from the brink—the six percent solution. PLoS Biolog y, 8(9),
e10004 85.
Weiskopf, S. R., McCarthy, K. P., Tessler, M., Rahman, H. A., McCarthy,
J. L., Hersch, R., … Siddall, M. E. (2018). Using terrestrial haemato-
phagous leeches to enhance tropical biodiversity monitoring pro-
grammes in Bangladesh. Journal of Applied Ecology, 55(4), 2071–2081.
https ://doi.org/10.1111/1365-26 64.13111
Whitfield, J. (1998). Zoology: A saola poses for the camera. Nature,
396(6710), 410. https ://doi.org/10.1038/24726
Wilcove, D. S., Giam, X., Edwards, D. P., Fisher, B., & Koh, L. P. (2013).
Navjot's nightmare revisited: Logging, agriculture, and biodiversity in
Southeast Asia. Trends in Ecology & Evol ution, 28(9), 531–540. htt ps ://
doi.org/10.1016/j.tree.2013.04.005
Wilkie, D. S., Bennett, E. L., Peres, C. A., & Cunningham, A. A. (2011). The
empty forest revisited. Annals of the New York Academy of Sciences,
1223(1), 120–128.
Wilkinson, N. (2016). Report on effects of five years of snare removal pa-
trols on snaring in the T hua Thien Hue - Quang Nam Saola Landscape:
An analysis of data collected by Fore st Guard patrols. Hanoi, Vietnam:
WWF CarBi project.
Wilkinson, N. (2017). Conserving the unknown: Decision-making for the
critically endangered Saola Pseudoryx nghetinhensis in Vietnam. (un-
published doctoral thesis). Cambridge, UK.
WWF (2017). Feasibility study repo rt: CarBi Phase II. (unpublished doc-
toral thesis). Hanoi, Vietnam.
BIOSKETCHES
The Biodiversity Dynamics team (https ://ecolo gical-dynam
ics-izw.com/team-3-biodi versi ty-dynam ics/) is interested in
developing standardized biodiversity survey and monitoring ap-
proaches for the threatened mammal communities of South-East
Asia. We combine data from camera-trapping and environmental
DNA with state-of-the-art analytical techniques to provide in-
sights into the ecology and distribution of threatened mammal
species. We work with local partners to ensure that our scien-
tific findings are transferred to on-the-ground conservation ac-
tion. We have a strong focus on the ecology and conservation
of endemic species in the Annamites ecoregion of Vietnam and
Laos. WWF-Vietnam and WWF-Laos works to protect biodiver-
sity in the Annamites by strengthening protected area manage-
ment effectiveness, promoting sustainable resource use through
community-based conservation models and maintaining and
restoring forest areas. CRES provides leadership in the imple-
mentation of strategic activities on biodiversity conservation in
Vietnam and conducts multi- and inter-disciplinary research in
biodiversity and environmental issues.
Author contributions: A.T. and A.W. conceived the ideas; A.T. and
A.N. collected the data; A.T., J.F.A., L.H., J.A ., J.L. and R.S. ana-
lysed the data; B.M.R., H.A.N.Q., F.G, T.N. and M. L. assisted the
field surveys; A.T., J.F.A. and A.W. led the writing; and all authors
commented on and reviewed the manuscript.
SUPPORTING INFORMATION
Additional supporting information may be found online in the
Supporting Information section.
How to cite this article: Tilker A, Abrams JF, Nguyen A, et al.
Identifying conservation priorities in a defaunated tropical
biodiversity hotspot. Divers Distrib. 2020;26:426–440. h t t p s : / /
doi.org/10.1111/ddi.13029
Available via license: CC BY
Content may be subject to copyright.
