Content uploaded by Sarah Sherburne
Author content
All content in this area was uploaded by Sarah Sherburne on May 06, 2022
Content may be subject to copyright.
289
RAFFLES BULLETIN OF ZOOLOGY 2022
Javan mongoose (Herpestes javanicus) abundance and spatial ecology
in a degraded dry dipterocarp forest
Sarah Sherburne1*, Wyatt Joseph Petersen1, Marnoch Yindee2, Tommaso Savini1 & Dusit Ngoprasert1
Abstract. Southeast Asia is rich in small carnivore species, but baseline information on these species is frequently
lacking. Many of the region’s remaining forests are degraded, which can drastically change ecosystem function
and structure. The Javan mongoose is a small generalist carnivore with a wide distribution across Southeast Asia,
whose population, home-range size, and micro-habitat selection are poorly known. We investigated each within
a degraded forest fragment in Northeast Thailand using a multimethod approach involving camera trap and radio
telemetry data. We found mongoose abundance was positively associated with dry dipterocarp forest (DDF) and
has a negative relationship with basal area of small trees (diameter at breast height < 10 cm). Across our entire
study site, we found a mean abundance of 1.10 animals per sampling station (SE 0.30 95% CI 0.65–1.92) and
within the DDF we found 3.04 animals per station (SE 0.75 95% CI 1.87–4.96). The mean home-range size for
two males was 1.86 km2 and for one female was 0.27 km2. Availability of termite mounds with entry holes was our
top model for den site selection. Prey availability did not affect micro-habitat selection by mongoose, presumably
due to an even distribution of small mammals across the DDF. Mongoose selected for areas with low numbers
of small trees, indicating an avoidance of closed forest environments. Our ndings indicate that Javan mongoose
select for open dry forest and can tolerate moderate forest degradation.
Key words. home-range estimate, Auto-correlated Kernel Density Estimation, habitat selection, den site selection,
radio telemetry
RAFFLES BULLETIN OF ZOOLOGY 70: 289–304
Date of publication: 6 May 2022
DOI: 10.26107/RBZ-2022-0013
http://zoobank.org/urn:lsid:zoobank.org:pub:C227D6E3-A0B5-41DC-B7E3-23C0E9A91057
© National University of Singapore
ISSN 2345-7600 (electronic) | ISSN 0217-2445 (print)
INTRODUCTION
Southeast Asia has a disproportionately high level of small
carnivore (< 16 kg) species when compared to other regions,
according to the International Union for Conservation of
Species (IUCN) (Marneweck et al., 2021). Habitat changes
such as deforestation for agricultural use impacts small
carnivores species worldwide, and Southeast Asia has among
the fastest worldwide deforestation rates, with only 35%
of old-growth forest left as of 2015 and many remaining
forests being both degraded and fragmented (Sodhi et al.,
2010; Wilcove et al., 2013; Taubert et al., 2018; Estoque
et al., 2019; Grantham et al., 2020). Deforestation degrades
habitat by altering forest structure and ecosystem functions
and can affect habitat choice and use by its inhabiting
species, as the ability of species to choose ‘ideal’ habitats is
diminished (Delciellos et al., 2017; Vanbianchi et al., 2017).
Small carnivores show a variable response to degradation;
most species are negatively affected (Mudappa et al., 2007;
Marneweck et al., 2021) while some show the ability
to adapt to it (Deuel et al., 2017; Ramesh et al., 2017).
Unfortunately, basic ecological data are not available for
many small carnivore species (Brooke et al., 2014), therefore
their response to habitat degradation is not easily predicted.
The Javan mongoose (Herpestes javanicus) is found
throughout Southeast Asia, with recorded distributions
ranging from Indonesia to the western edge of China
(Chutipong et al., 2016). The IUCN considers the Javan
mongoose to be of Least Concern because of its perceived
abundance, apparent widespread distribution, and the fact
that there are recorded occurrences in degraded habitats and
agricultural areas (Duckworth et al., 2010; Chutipong et al.,
2016). However, data on the species’ actual distribution
are limited to anecdotal sightings and non-targeted camera
trap surveys mainly within protected areas (Chutipong
et al., 2014). As a result, no detailed status assessments
have been undertaken to date and little is known about the
species’ distribution, ecology, or abundance throughout
its wide range. This lack of assessment is partly due to a
recent taxonomic separation from the extensively introduced
small Indian mongoose (Herpestes auropunctatus), which
has been widely studied in island environments due to its
deleterious ecological effects on native species within its
introduced range (Lowe et al., 2004; Veron & Jennings,
2017). However, the ndings from these island studies are
not necessarily transferable to native populations, as the
Conservation & Ecology
Accepted by: Norman Lim T-Lon
1Conservation Ecology Program, King Mongkut’s University of Technology Thonburi,
49 Thakham, Bangkhuntien, Bangkok, 10150 Thailand; Email: sr.sherburne@gmail.
com (*corresponding author)
2Akkhraratchakumari Veterinary College, Walailak University 222 Thaiburi, Thasala
District, Nakhon Si Thammarat 80160, Thailand
290
Sherburne et al.: Javan mongoose abundance and spatial ecology
Fig. 1. Map and location of Sakaerat Biosphere Reserve with camera trap stations used to estimate Javan mongoose (Herpestes javanicus)
abundance in 2017. Prey grid stations were used to calculate yearly averaged rodent biomass from January 2017 to November 2017.
291
RAFFLES BULLETIN OF ZOOLOGY 2022
ecological conditions in island environments are often very
different from native environments. For example, density
estimates for small Indian mongoose average 51–64 times
higher in introduced ranges, and reproductive features are
signicantly different (Owen & Lahti, 2020).
Studies on the Javan mongoose, and indeed on most
Asian mongooses, are sparse in their native range, and
more information is needed to understand if this species
is threatened by habitat degradation. This study aims to
investigate Javan mongoose ecology in a forest fragment to
inform conservation planning. We considered three different
aspects of its ecology: a) abundance, b) home-range size, and
c) micro-habitat selection. Abundance was chosen because
low abundance is correlated with high risk of extinction
(Mace et al., 2008) and it allows for comparisons between
habitats. Home-range size and micro-habitat selection was
investigated because they can inform habitat requirements
within human modied landscapes (Hinton et al., 2016; Filla
et al., 2017). We investigated home-range size via radio
telemetry and predicted that it would be larger in males than
females, as Javan mongoose show sexual dimorphism, which
can indicate polygamy in mammals and greater ‘ranging’
for mating access (Simberloff et al., 2000; McPherson &
Chenoweth, 2012). To inform our micro-habitat selection we
recorded vegetation characteristics, prey availability, and den
abundance in selected and available habitats. We collected
prey abundance data as it can be a strong predictor for
mesocarnivores and is most accurate when measured directly
as opposed to using habitat with suitable characteristics
for high prey density as a surrogate (Wolff et al., 2015).
Selection for high prey abundance must be weighed against
exposure to predation and competition for high resource
areas, so whether individuals select for or against it can
be informative (Case & Bolger, 1991; Suraci et al., 2016).
Den sites and their relative availability can also factor
in micro-habitat selection by small carnivores, since den
availability can decrease predation and improve reproductive
success through safe rearing of young (Tannerfeldt et al.,
2003; Moehrenschlager et al., 2007). We predicted that on
the micro-habitat level, mongoose would select areas with
a higher relative biomass of prey resources and a larger
number of potential den sites to decrease predation risk and
increase reproductive success.
METHODS
Study Site. The study was conducted at Sakaerat Biosphere
Reserve (14.44–14.55°N, 101.88–101.95°E) in northeastern
Thailand. Elevation ranges 250–762 metres above sea level.
Total forested area is 148 km2, with dry dipterocarp forest
(DDF) comprising 11%, and the other two major forest types
being dry evergreen forests (54%) and plantation (33%)
(Petersen et al., 2019). The DDF is bordered by agricultural
areas to the east and closed canopy dry evergreen forest and
reforested areas to the west. It is split in the southern portion
by a segment of dry evergreen forest (Fig. 1). Our site has
a tropical climate characterised by a wet and dry season.
We dened the wet season as May to October (average
rainfall of 920 mm) and the dry season as November to April
(average rainfall of 220 mm) for the reserve (Khamcha et
al., 2018). DDF is burned yearly during the dry season by
the reserve’s staff. The reserve is now a protected zone, but
prior to 1977, much of the original forest had been converted
to plantations and elds and undergone severe deforestation,
and the remaining forest is still recovering (Ongsomwang &
Sutthivanich, 2014). The reserve is bordered on the southern
side by a major highway, on the northern side by a reservoir,
and on the eastern and western sides by agricultural areas.
Camera trapping. Sixty camera traps (SG565, Scout Guard)
were deployed across Sakaerat Biosphere Reserve to sample
small carnivore species from January 2017 to October 2017
covering 65 km2 of major forest types. Traps were deployed
in a stratied random manner by forest type, with a minimum
distance of 1 km between camera-trap stations (dry evergreen
forest: 28 traps covering 34 km2; dry dipterocarp forest: 16
traps covering 16 km2; reforested areas: 16 traps covering 15
km2, Fig. 1). Cameras were secured to the closest available
tree to the site coordinates, 3 metres from the target zone
and 45 cm above the ground. Cameras were left active 24
hours a day. Fish oil scent lures were placed in 325 ml
aluminum cans buried ush to the ground in the focal area
of each camera. Scent lures were used because they have
been shown to increase the probability of detection for
mesopredators without affecting abundance estimates (Gerber
et al., 2012; Ferreras et al., 2018; Holinda et al., 2020).
Cameras were revisited once per month for battery, SD card,
and lure replacement. Mongoose detections were considered
notionally independent when the time between photographs
was ≥ 30 minutes (O’Brien et al., 2003). We constructed a
binary detection history (i.e., detect/not-detect) using daily
occasions (0000–2359 h). Trap nights were dened as the
number of daily occasions where cameras were functioning.
Mongoose abundance estimations. We estimated mongoose
abundance using the Royle-Nichols model (RN model),
which utilises repeated observations of unmarked animals in
detected/non-detected data (Royle & Nichols, 2003; Kalle et
al., 2014; Paolino et al., 2018). The model has two parameters
to be estimated: 1) mean abundance per sampling station
(λ), which can be interpreted as the number of mongoose
home ranges overlapping one of our camera traps, and 2)
the probability of detecting a species at a sampling station
(p) (Royle & Nichols, 2003; Nakashima, 2020). The RN
model calculates abundance as an analogue to uneven
detection probability, while incorporating the effects of both
site-specic covariates and observed (sampling) covariates.
All analysis was done in program R 4.0.5 (R Core Team,
2020). Abundance estimates were made using the ‘unmarked’
package (Fiske & Chandler, 2011). The number of days after
each scent lure refuelling was used as a sampling covariate
to account for differences in detection probability across
sampling days. We subset our camera trap data to the rst
180 occasions to meet the closure assumption. To model
differences in forest structure that could affect abundance,
we collected small tree basal area, large tree basal area, and
canopy cover percentage (Table 1). These three abundance
covariates were collected at ten-metre plots centred on each
292
Sherburne et al.: Javan mongoose abundance and spatial ecology
Table 1. Descriptions of covariates used in micro-habitat selection models, den site selection models, and abundance estimation models.
Habitat variables were collected from 15-metre radius plot for micro-habitat selection variables and 10-metre radius plots for abundance
variables. Plots were centred on radio tracked mongoose locations or camera trap stations, respectively. Prey variables were collected
from prey grids and averaged (Figs. 1, 2).
Micro-habitat and den site selection variables
Variables Name Description and Measurement Unit
Mean basal area of small trees avgBasalSmall Mean basal area for small trees (DBH ≤ 10 cm) (cm2)
Mean basal area of large trees avgBasalLarge Mean basal area for large trees (DBH > 10 cm) (cm2)
Shrub Count shrubCount Number of shrubs < 3 cm DBH and ≥ 50 cm height
Termite Mound with Entry termiteAvailable Number of termite mounds with entry holes
Termite Mound without Entry termiteNonavailable Number of termite mounds without entry holes
Holes holes Number of dug holes ≥ 4 cm diameter
Small tree count smallTreeCount Number of trees (DBH ≤ 10 cm)
Large tree count largeTreeCount Number of trees (DBH > 10 cm)
Total tree count totalTrees Sum of all trees in plot
Average Rodent Biomass avgRodentBiomass Average mass of large rodent species for grid multiplied by large rodent species
density for grid + average mass of small rodent species for grid multiplied
by small rodent species density for grid. Averaged per habitat zone. (g/ha)
Pitfall invertebrate mass insectPitfall Sum of dry mass for invertebrates collected by pitfall trap per habitat zone (g)
Total prey totalPrey Sum of pitfall invertebrate mass, average rodent biomass and sweep invertebrate
mass (g)
Sweep net invertebrate mass insectSweep Sum of dry mass for invertebrates collected by sweep netting per habitat zone (g)
FAI FAI Average basal large multiplied by large tree count (cm2)
Abundance estimation variables
Average rodent biomass rAvg Averaged bimonthly rodent biomass (g/ha) value over wet and dry season
small trees basal area basalS Average basal area for all small trees (DBH < 10 cm) within each camera-trap
station’s two micro-habitat plots, multiplied by the density of trees (DBH <
10 cm) within the same plots. Scaled to m2/ha.
large trees basal area basalL Average basal area for all large trees (DBH ≥ 10 cm) within each camera-trap
station’s two micro-habitat plots, multiplied by the density of trees (DBH ≥
10 cm) within the same plots. Scaled to m2/ha.
Canopy cover cc Percent canopy cover
Road road Distance to nearest road (m)
Edge edge Distance to edge of Sakaerat Biosphere Reserve (m)
Oil oil Date from deployment of sh oil scent lure to replacement
camera trap station. To model how disturbance might affect
abundance, we calculated the distance to roads and the edge
of Sakaerat Biosphere Reserve’s boundary for each point.
Covariates were tested for pairwise correlation using the
‘corplot’ package (Wei & Simko, 2021). Covariates with
correlation values greater than 0.5 were not included within
the same model. Continuous data were standardised using
the ‘scale’ function (mean = 0, sd = 1) in R. We tested the
default upper summation index for latent abundance (K =
25) to ensure that it will not affect parameter estimates by
comparing approximate small sample size corrected Akaike
Information Criterion (AICc) values between K=25, K=50
and K=100. We found that the default value of 25 was
sufcient for our data.
Best tting models for abundance were selected using AICc
(Burnham et al., 2011). To test the goodness of t for our
RN models, we followed the MacKenzie & Bailey (2004)
methodology of parametric bootstrapping using three tests:
sum-of-squared errors, chi-square, and Freeman–Tukey
(Paolino et al., 2018). We generated 1,000 bootstrapped
resamples and tested the fit of the most parameterised
(global) model with our detection covariate (sh oil) and our
293
RAFFLES BULLETIN OF ZOOLOGY 2022
Fig. 2. Forest types in Sakaerat Biosphere Reserve: A, dry dipterocarp forest; B, dry evergreen forest
A B
abundance covariates (basalS+rAvg+basalB+edge+road) in
the models. We also calculated an overdispersion ratio ()
for our global model to test for possible overdispersion in
our detection history (Burnham & Anderson, 1998). In order
to ensure our detections were independent among samples,
we tested for spatial autocorrelation in our data using the
Moran’s I test in the ‘ape’ package (Paradis & Schliep,
2019). We created an inverted distance matrix of our camera
trap locations and used relative abundance index (number
of independent photographs per station) as our value. With
a P-value of > 0.05, the hypothesis that there is signicant
spatial correlation is not supported. We estimated mean
abundance for our trapping area by predicting λi for each
of our camera trap stations based on estimates from our top
model using the ‘predict’ function in R. We then calculated
the mean of overall stations.
Trapping and radio tracking. Mongoose trapping was
carried out from December 2019 to January 2021. A total
of 11 large live traps (121 × 50 cm) and 34 (24 × 9.5; 31
× 10 cm) small live traps were used to capture mongooses,
for a total of 110 trap sites. Traps were left in place for a
minimum of 7 days unless disturbed by dogs, in which case
traps were moved immediately. Trapping sites were within the
DDF and placed according to our camera trap data for high
mongoose detection within Sakaerat. In addition, we selected
sites based on recent mongoose sightings, the presence of
tracks, and high quantity of potential den sites such as termite
mounds. Live traps were baited with a chicken carcass and
left open for 24 hours and checked every morning.
For each individual captured, we recorded body length,
body condition, sex, and age class (Appendix 1) before
collaring them with VHF collars model M1-2A from Holohil
Systems limited, with built-in activity sensors. Mongooses
were immobilised using Zoletil general anaesthetic to allow
for measurements and collar tting while minimising stress
(Zoletil the Versatile Anaesthetic, 2021) (see Appendix 2). A
leather harness system was added to help with collar retention.
All VHF collars (31 g) with harnesses were approximately
ve percent or less of the animal’s mass (Sikes & Gannon,
2011). Age class was determined by the condition of the teeth
(worn teeth indicating old age), descended testes for males,
and condition of the nipples in females. Post-processing,
individuals were kept within the live trap until fully recovered
from sedation, and then released at their capture location.
Collared mongooses were located up to two times a day.
Radio tracking was done on foot using a two-element antenna
and an ATS receiver model R410. Due to tall grasses in
the DDF (Fig. 2), visual sightings of collared mongoose
were rare. Locations for individuals were determined using
triangulation, with three lines of intersection from a distance
of 20 metres or less. We allowed a minimum of three hours
to pass between tracks to decrease temporal autocorrelation
between points. For each location we recorded habitat type,
observer distance, and whether the collar’s activity sensor
was activated.
If a collar’s activity sensor did not indicate movement
during the tracking, we approached to see if the mongoose
was sheltering in a den site. We waited for inactivity to
ensure actual selection of a den, and to minimise disruption
if the mongoose was foraging. The type of den, number of
entrances, entrance dimensions, and number of days the den
was visited were all recorded.
Home-range estimation. Home-range estimations were
calculated using Auto-correlated Kernel Density Estimation
(AKDE) to incorporate potential temporal or positional
autocorrelation (Fleming et al., 2015; Noonan et al., 2019).
AKDE analysis was done using the R package ‘ctmm’
(Calabrese et al., 2016). We used a variogram to plot a semi-
variance function, which visualised autocorrelation in our data
by plotting the average square displacement unit (hectares)
over a time lag period (months). We then applied a variety
of continuous-time stochastic movement models to each
individual’s telemetry data. Models tested include Ornstein–
Uhlenbeck motion (Brownian motion within a home-range),
Ornstein–Uhlenbeck with foraging, and independent identical
distribution (no autocorrelation) (Fleming et al., 2014). We
compared AICc values to determine the best tting movement
294
Sherburne et al.: Javan mongoose abundance and spatial ecology
mode model for each individual (Burnham et al., 2011). All
home-ranges were calculated at 95% utilisation distribution
and its 95% condence interval. We also calculated a 50%
utilisation distribution as a proxy for an individual’s ‘core
area’. All analysis was completed in program R 4.0.5 (R
Core Team, 2020).
Micro-habitat selection. We collected vegetation variables
in areas used by mongooses and compared them to available
locations to determine the micro-habitat selection of
radio-collared mongooses. Used locations were chosen by
randomly selecting radio telemetry locations. We collected
micro-habitat data on 50 used and 100 available locations for
each of our male mongooses, and 40 used and 80 available
for the female. For available locations, we collected a plot
100 metres to the north and south of each used location.
Aside from characteristics of den sites, identical variables
were collected for both general micro-habitat plots and den
site plots.
At each location we established a 15-m radius circular plot.
We collected ground cover percentage, diameter at breast
height (DBH) for trees ≥ 3 cm DBH, count of shrubs and
saplings < 3 cm in diameter and ≥ 50 cm in height, and
potential den site type and count (Table 1). Shrubs less
than 50 cm height were at the same height as the primary
groundcover (Arundinaria pusilla) and were considered
functionally groundcover and not included in shrub count.
Trees were subset in the analysis as ‘small’ or ‘large’, large
being > 10 cm DBH and small between 3–10 cm DBH.
Ground cover, understorey (shrub and sapling) count, and den
site availability were recorded as an indication of potential
cover and refuge site availability. Number of trees and
DBH were recorded to indicate undisturbed forested area
versus open or edge habitat, as tree size varies within each.
To compare habitat features across the DDF, we labelled
micro-habitat plots as plots near the edge (< 400 metres from
edge) or core DDF (≥ 400 metres from edge).
We ran logistic regression models using the ‘lme4’ package
(Bates et al., 2015) in R to determine mongoose selection
based on micro-habitat features and prey availability. Prior
to data analysis, we tested for outliers, correlation among
variables, and standardised the continuous variables (see
above). We tested the inclusion of mongoose ID as a random
effect on our variables and found no change in the outcome,
while decreasing the parsimony of the model and therefore
we chose not to include mongoose ID as a random effect.
Top models were selected using AICc values. We validated
our top models by calculating the area under the receiver
operator characteristic curve (AUC) for model averaged
predictions in the R package ‘PresenceAbsence’ (Freeman
& Moisen, 2008). Model averaging was done using the
‘AICcmodavg’ package (Mazerolle, 2020).
Prey biomass. A recent study by Subrata et al. (2021) provided
evidence that rodents and arthropods occur frequently (83%
and 100% frequency of occurrence, respectively) in the
Javan mongoose’s diet. We used biomass of these groups
as a proxy for prey availability of mongoose, and averaged
prey biomass was used as a covariate in our micro-habitat
selection analysis.
Rodent prey biomass for abundance estimation was sampled
during four periods in 2017 (February–March, April–May,
July–August, September–October) at 15 sites (60 sessions
in total; 7 sites in dry evergreen; 4 in dry dipterocarp; 4
in reforested). We used Sherman live traps (7.62 × 8.89 ×
22.86 cm) placed on the ground. At each of the 15 sites,
we arranged 25 live traps in a 5 × 5 grid with 20 m spacing
between traps. Sites were sampled once every two months,
with each bimonthly session lasting seven consecutive trap
nights. Traps were baited with peanut butter and checked
once a day in the morning. Captured animals were uniquely
marked with an ear tag, weighed, and then released at their
capture site. The mean mass of rodents captured each session
was multiplied by the session’s density estimate to obtain
session biomass. Yearly averaged biomass estimates from
each small mammal trapping session were assigned as a
covariate to the nearest 4 camera-trap stations within the same
habitat type. For further details, see Petersen et al. (2019).
Prey biomass for micro-habitat selection was collected
between October–December 2020. Prey grids were selected
to represent habitat used by radio tracked mongoose, which
was within the DDF. Our study area contains only a small
fragmented portion of DDF surrounded by dry evergreen
forest and agricultural areas (Fig. 1). We determined that
edge type is therefore likely to affect prey distribution. We
placed our prey grids in three different sections within the
DDF, representing DDF adjacent to dry evergreen edge
(DDF-dry evergreen edge), core DDF habitat, and DDF
adjacent to agricultural areas (DDF-agricultural edge) (Fig.
3). All prey grids were sampled for eight consecutive days
using 25 live traps in a 5 × 5 grid with 20 m spacing between
traps. Prey grids were spaced ≥ 500 metres from one another.
Edge grids were placed < 200 metres from their respective
edge type to ensure that edge effect was represented. Core
DDF zone grids were ≥ 400 metres from any other habitat.
For rodent capture, Sherman live traps (9.5 cm × 24 cm; 31
cm × 10 cm) were baited with peanut butter and checked
once per day (12 rodent sites total; 3 in DDF-agricultural
edge; 4 in the core DDF; 5 in the DDF-dry evergreen edge).
Rodents were sampled using mark-recapture methods. For the
rst capture, each rodent was identied to the species-level,
sexed, weighed, and uniquely marked with an identifying
ear tag before release. All traps were placed on the ground
and shaded with vegetation.
Rodent density (individuals/hectare) was calculated using
the ‘secr’ package in R programming (Efford, 2021).
Spatially explicit capture recaptures (SECR) allows for the
calculation of density while accommodating heterogeneous
probability of capture for individuals (Borchers, 2012). We
plotted rst captured mass of each individual by species and
separated them into large (> 50 g average) and small (< 50 g
average) groups. These groups were analysed separately to
accommodate potentially different behavioural responses
and home range sizes. Species groups were modelled in a
295
RAFFLES BULLETIN OF ZOOLOGY 2022
Fig. 3. Map of Sakaerat Biosphere Reserve with radio tracked (December 2019 to January 2021) Javan mongoose (Herpestes javanicus)
home ranges and prey grids (PG). 95% utilisation contours (U.C) for male (M6, M1) and female (F1) mongooses are labelled in the legend.
50% U.C are solid line circles within each individual’s home range. Prey grids collected ground-dwelling invertebrate mass as well as rodent
biomass within the DDF (October to December 2020). Stars indicate where only ground-dwelling invertebrates were collected. Triangles
indicate areas where sweep netting for invertebrates occurred in addition to sampling for rodent biomass and ground-dwelling invertebrates.
296
Sherburne et al.: Javan mongoose abundance and spatial ecology
multi-session analysis, with each grid representing a separate
session. For each session and group, we calculated sigma
(σ, a parameter of movement derived from distance between
individual’s recapture locations) and g0 (the probability of
capture at an individual’s activity centre), to derive estimated
density.
To accommodate behavioural responses to trapping that may
affect g0, we incorporated potential lingering or transient ‘trap
shy’ and ‘trap happy’ responses into our models. Models
tested included: 1) a global lingering change in behaviour after
rst capture (b); 2) a global change in behaviour depending
on capture during the previous occasion (B); 3) a trap-specic
lingering response post initial capture (bk); 4) a transient
trap-specic behaviour response post capture (Bk) (Efford,
2021). AICc scores were compared for all models to select
the top-ranked model (Appendix 3). Density per grid for
small and large species groups were estimated separately,
multiplied by average mass for each group, and summed to
calculate biomass per grid. Rodent biomass was used as a
micro-habitat variable in selection analysis by calculating
the average mass per habitat zone and assigning that zone’s
mass to micro-habitat plots within it.
Invertebrate biomass was collected using pitfall traps and
sweep netting (15 pitfall sites: 5 in DDF-agricultural edge;
4 in core DDF; 6 in DDF-dry evergreen edge; 6 sweep net
sites: 2 in each zone) (Fig. 3). With the exception of three
grids, invertebrates and rodent biomass were collected at the
same location. All pitfall grids followed the same layout as
rodent trapping girds, 80-metre length transects in a ve by
ve trap gird, with traps spaced 20 metres apart. A total of
25 pitfall traps per grid were placed. Sweep net collection
took place walking alongside pitfall grids and was conducted
within the same transects, with collection not exceeding the
grid’s layout. The collection took place upon arrival to the
prey grid. For every 10 m along the sweep netting transect,
10 uninterrupted sweeps with a net were completed to collect
ying insects. For pitfall traps, 15.7 cm deep plastic cups
were dug ush to the ground, with two cups nesting within
each other for easy removal. All invertebrates captured by
either method were euthanised by being placed in a freezer
overnight, and all large invertebrates were then stored in
a 90% ethanol solution. Sweep net insects were stored
in a freezer until drying. Dry weight of invertebrates was
obtained by oven drying at 57°C for 24 hours. Gastropods
were then deshelled and dried for an additional 5.5 hours
till completely dry. Summed dry biomass (g) per habitat
zone was calculated.
Den site selection. Den sites used by tracked mongoose
were recorded and analysed separately from general micro-
habitat plots. For variables collected and methods used to
analyse den site selection, refer to the micro-habitat selection
section (see above, and Table 1). For each den site used by
mongoose, we collected den type, number of entry holes, and
diameter of entry holes. Potential den sites were categorised
as dug holes or termite mounds. Termite mounds with entry
points and those without were categorised separately as
‘non-available’ (no entry holes) and ‘available’ (with entry
holes). This allowed us to differentiate whether mongooses
were simply selecting for termite rich areas or specically for
available den sites. Holes in termite mounds were assessed
visually and deemed as viable entry points if they were the
width of our smallest captured mongoose head or greater
(diameter ≥ 4 cm).
RESULTS
Mongoose abundance. In the 8,867 trap nights considered
for this study, we obtained 354 independent detections of
Javan mongoose, 348 of which were within the DDF. For a
full list of all mammals captured over the course of the study,
see Petersen et al. (2019: appendix A). Our averaged rodent
biomass was highest in dry evergreen forest and lowest in
reforested area (dry evergreen = 458.3 g/ha, SE = 13.7; DDF
= 287.5 g/ha, SE = 56.5; reforested area = 178.8 g/ha, SE =
22.6). We tested the addition of ‘oil’ as a covariate affecting
detection probability and found that it decreased parsimony
without adding to the model and so we chose not to include
it (Table 2). For our parametric bootstrapped goodness of
t tests, the three generated P-values were all > 0.05. For
our overdispersion ratio, our value was 0.55, suggesting the
model adequately t with the Poisson assumption. We ran our
Moran’s I test on data collected from traps within the DDF
where our detections were grouped. We found that our data
did not show signicant spatial autocorrelation (p > 0.05).
Our best ranked model for mean abundance per sampling
station (λ) was incorporating additive effects of basal area for
both large and small trees and distance to Sakaerat Bioreserve
edge (Table 2). Our λmean value for this model was 1.10 (SE
0.30, 95% CI 0.65–1.92). The greatest negative inuence
on abundance within this model was average basal area for
trees with a diameter < 10 cm (β = -0.79, SE 0.28, 95% CI
-1.34 to -0.24), followed by distance to edge (β = -0.70, SE
0.21, 95% CI -1.12 to -0.28) and then average basal area
for trees with a diameter ≥ 10 cm (β = -0.52, SE 0.26, 95%
CI -1.03 to -0.02). The probability of detection for our top
model was 0.038 (SE 0.131, 95% CI 0.030–0.048). Because
most (98%) of our observations were in DDF habitat, we
calculated DDF specic λmean using our top model. Our DDF
specic λmean was 3.04 (SE 0.75, 95% CI 1.87–4.96) and the
range was 0.78–7.23.
Home-range estimation. We trapped a total of seven Javan
mongoose over the course of the study, six males and one
female (Appendix 1). Four males removed their radio collars
within one to two days of release. A backpack system was
added to help retain collars. Spatial data were collected for
two adult males, M1 and M6, and one young adult female,
F1, over the course of our tracking period (2 January 2019
to 14 September 2020). We collected 97 locations for M1,
194 locations for F1, and 101 locations for M6. M1 was
not tracked from 9 February 2020 to 2 July 2020 due to a
collar malfunction which led to data loss until recapture
and collar replacement. Both M1 and F1 were predated by
reticulated pythons (Malayopython reticulatus) during the
tracking session (M1, 7 September 2020; F1, 31 March
297
RAFFLES BULLETIN OF ZOOLOGY 2022
Table 2. Model selection for Javan mongoose (Herpestes javanicus) abundance (λ) estimation and probability of detection (p) at Sakaerat
Biosphere Reserve from January to June 2017, based on Royle-Nichols model. Denitions for covariates can be found in Table 1.
Models KAICc ΔAICc wi
λ~basalL + basalS, p~1 5 2061.145 0.000 0.985
λ~basalL + basalS, p~1 4 2070.304 9.158 0.010
λ~basalS, p~1 3 2072.099 10.954 0.004
λ~cc+basalL, p~1 4 2078.338 17.193 0.000
λ~cc, p~1 3 2078.912 17.767 0.000
λ~edge, p~1 3 2080.395 19.250 0.000
λ~rAvg+cc, p~1 4 2080.465 19.319 0.000
λ~edge + road, p~1 4 2080.624 19.479 0.000
λ~cc+road, p~1 4 2080.757 19.612 0.000
λ~basalL, p~1 3 2110.377 49.232 0.000
λ~road, p~1 3 2112.222 51.077 0.000
λ~rAvg, p~1 3 2119.94 58.795 0.000
λ~constant, p~1 2 2122.554 61.409 0.000
K is the number of parameters included in the model, AICc is Akaike’s Information Criteria corrected for small sample size, ΔAICc is
the difference in AICc values, and wi is the Akaike weight.
2020). M6’s radio telemetry collar fell off at the end of the
tracking period and was recovered.
For our two male mongooses, the most suitable movement
mode for our data was independent identically distributed
points, indicating little autocorrelation. For the female,
Ornstein–Uhlenbeck anisotropic (random motion within a
home-range) movement modes were deemed the best t.
Based on these movement modes, we chose a traditional
kernel density estimator for our two males and an AKDE for
the female in order to compensate for any autocorrelation.
M1 had an estimated home-range area of 1.77 km2 with
a 95% CI of 1.24–2.38 km2, and a core area of 0.34 km2.
M6 had an estimated home-range of 1.95 km2 with a CI
of 1.59–2.35 km2, and a core area of 0.34 km2. F1 had an
estimated home-range area of 0.27 km2 with CI of 0.22–0.32
km2, and a core area of 0.06 km2 (Fig. 3). Ninety-ve percent
of all home-range locations were within habitat classied
as dry dipterocarp forest. Non-DDF locations were in edge
habitat bordering DDF, except for one location of M1 in
dry evergreen forest (Fig. 3).
Micro-habitat selection. We found that both number of
small trees and total tree count were negatively associated
with mongoose micro-habitat selection (small tree count β
= -0.45, SE 0.15, 95% CI -0.77 to -0.17; total tree count β
= -0.36, SE 0.13, 95% CI -0.63 to -0.12), based on lowest
AICc score (Table 3). Our area under the curve score for
our model averaged prediction based on our two top models
for micro-habitat selection was 0.52.
The average rodent biomass was highest in core DDF habitat
(Core DDF = 215.86 g/ha, SE = 19.47; DDF-dry evergreen
edge = 162.27 g/ha, SE = 32.84; DDF-agricultural edge =
160.38 g/ha, SE = 44.49). Arthropod mass was similar in
both core DDF (112 g; sweep = 18.1 g) and DDF-agricultural
edge (114 g; sweep = 14.7 g), but lowest in the DDF-dry
evergreen edge (36.3 g; sweep = 20.9 g). We used a one-
way ANOVA to compare rodent density between habitat
sections in the DDF, and found that rodent biomass did
not signicantly differ across the DDF (ANOVA F(2,9) =
0.932, p = 0.43).
We found that small tree count varied signicantly within the
DDF (Kruskal-Wallis test, chi-squared = 26.51; degrees of
freedom = 2, p < 0.001). Specically, a Pairwise Wilcoxon
test found that median small tree count was signicantly
lower in the area of the DDF closest to the dry evergreen
edge (2 trees/plot) than the agricultural edge (5 trees/plot)
and core DDF zone (5 trees/plot).
Den site selection. We found a total of 36 den site locations
from our three mongooses, with two additional sites from
a male mongoose (M5) tracked for only two days before
his collar fell off. The majority of our den sites were in
(available) termite mounds, with only 18% of our dens being
dug holes. Den sites had on average 1.7 entrances, a mean
width of 11.34 cm, and a mean height of 10.38 cm. Den
sites were occupied for an average of 1.6 days, with 30%
of sites reused at least once, and one site reused by F1 ten
times. The vast majority of our den site reuse was by F1,
with a 36% site reuse rate. M1 revisited one den site three
298
Sherburne et al.: Javan mongoose abundance and spatial ecology
Table 3. Model selection for Javan mongoose (Herpestes javanicus) micro-habitat selection at Sakaerat Biosphere Reserve from December
2019 to January 2021, based on logistic regression analysis. Denitions for covariates can be found in Table 1.
Variables KAICc ΔAICc wi
smallTreeCount 2 435.028 0.000 0.631
totalTrees 2 436.675 1.647 0.277
largeTreeCount 2 441.415 6.387 0.026
shrubCount 2 442.104 6.387 0.018
Constant 1 443.992 8.964 0.007
termiteAvailable 2 444.458 9.430 0.006
FAI 2 445.084 10.056 0.004
avgBasalSmall 2 445.232 10.203 0.004
totalPrey 2 445.583 10.555 0.003
avgBasalSmall + avgBasalLarge +shrubCount 4 445.598 10.570 0.003
termiteNonavailable 2 445.605 10.576 0.003
avgRodentBiomass 2 445.641 10.702 0.003
insectPitfall 2 445.730 10.702 0.003
insectSweep 2 445.852 10.824 0.003
avgBasalLarge 2 445.966 10.938 0.003
holes 2 446.014 10.985 0.003
avgBasalSmall + avgBasalLarge 3 447.264 12.236 0.001
termiteNonavailable + termiteAvailable + holes 4 447.764 12.735 0.001
K is the number of parameters included in the model, AICc is Akaike’s Information Criteria corrected for small sample size, ΔAICc is
the difference in AICc values, and wi is the Akaike weight.
times, but otherwise did not reuse sites. M6 only reused a
site once. However, site reuse could be underreported, as
radio tracking did not typically last more than one hour
after sunset when mongoose could still be active. For our
den site selection, we found that termite mounds with entry
holes were positively associated with mongoose detection
(β = 0.69, SE 0.23, 95% CI 0.27–1.17), and that it was the
top model for den site selection (Table 4). Our area under
the curve score for our top model was 0.71.
We found that total number of termite mounds varied
signicantly between habitat zones within the DDF (Kruskal-
Wallis test, chi-squared = 23.3; degrees of freedom = 2,
p < 0.001). Specically a Pairwise Wilcoxon test found that
median termite mound count was signicantly lower in the
area of the DDF closest to the agricultural edge (1 mound/
plot) and when compared to core DDF (2 mounds/plot)
and DDF adjacent to dry evergreen edge (2 mounds/plot).
DISCUSSION
Our research shows the importance of DDF for Javan
mongoose in our study site, as 98% of all of our camera trap
detections for Javan mongoose were within DDF. However,
as our results are specic to a degraded forest fragment, any
comparison to populations inhabiting primary forest should
be approached with caution. In addition, while our study
clearly shows a preference for more open environments at
both the macro- and micro-habitat levels, we are hesitant to
say that Javan mongoose are purely open forest specialists
based on this study alone, since there are some sparse records
of them in evergreen habitat and we only examine one site
(Chutipong et al., 2014). Future Javan mongoose studies
in other areas of the species range will likely conrm or
disprove whether they are specialists.
Our top abundance model showed small tree basal area had
the greatest negative effect on mongoose abundance, with
distance to edge of the bioreserve having a slightly less,
though still quite large, negative effect. Basal area of large
trees had the weakest negative effect in our top model.
However, these results should be interpreted cautiously as
almost all of our detections were from the DDF and all three
of the aforementioned covariates were strongly associated
with DDF habitat (Petersen et al., 2019). Specically, DDF
in our study site had fewer large trees than dry evergreen,
and fewer small trees than either dry evergreen or plantation
(Oliver et al., 2019; Petersen et al., 2019). Likewise, the
camera traps in the DDF were disproportionately closer to
299
RAFFLES BULLETIN OF ZOOLOGY 2022
Table 4. Model selection for Javan mongoose (Herpestes javanicus) den site selection at Sakaerat Biosphere Reserve from December 2019
to January 2021, based on logistic regression analysis. Denitions for covariates can be found in Table 1.
Variables KAICc ΔAICc wi
termiteAvailable 2 135.547 0.000 0.578
termiteNonavailable + termiteAvailable + holes 4 136.462 0.914 0.366
constant 1 144.149 8.602 0.008
smallTreeCount 2 144.22 8.673 0.008
termiteNonavailable 2 144.74 9.193 0.006
totalTrees 2 145.194 9.647 0.005
averageBasalLarge 2 145.715 10.168 0.004
averageBasalSmall 2 145.775 10.227 0.003
largeTreeCount 2 145.906 10.359 0.003
insectSweep 2 145.978 10.431 0.003
insectPitfall 2 146.031 10.483 0.003
FAI 2 146.066 10.519 0.003
avgRodentBiomass 2 146.221 10.674 0.003
holes 2 146.222 10.675 0.003
shrubCount 2 146.223 10.676 0.003
avgBasalSmall + avgBasalLarge 3 147.518 11.971 0.001
avgBasalSmall + avgBasalLarge + shrubCount 4 149.656 14.109 0.000
K is the number of parameters included in the model, AICc is Akaike’s Information Criteria corrected for small sample size, ΔAICc is
the difference in AICc values, and wi is the Akaike weight.
the edge of the bioreserve compared to those in other habitats
as the DDF was restricted to the bioreserve’s edge (Fig. 1).
Because this is the rst study on Javan mongoose abundance
and spatial ecology, we use the closest living relative to the
Javan mongoose (Veron et al., 2007), the grey mongoose
(Herpestes edwardsii) as a metric for comparison. Grey and
Javan mongoose both appear to have similar selection for
open dry forests, so their ecology may be comparable (Kalle
et al., 2013, 2014; Bajaru et al., 2020). Grey mongoose
mean abundance, estimated using the RN model in a non-
fragmented forest over two years, was lower (λmean = 0.34
[95% CI 0.026–0.654] to 0.68 [95% CI 0.006–1.366]; Kalle
et al., 2014) but otherwise similar to our overall estimates for
Javan mongoose in a forest fragment (λmean = 1.10, 95% CI
0.65–1.92). However, grey mongoose estimates were closer
to Javan mongoose estimates in closed forests (λmean = 0.36,
95% CI 0.18–0.78) than those in the DDF (λmean = 3.04,
95% CI 1.97–4.96). This is possibly due to grey mongoose
abundance being clustered in smaller patches of open dry
habitat, similar to our species. This would cause lambda
averaged across large forest matrixes to appear misleadingly
low. In addition, grey mongoose abundance could be higher
within fragmented forests, as is the case for some mongoose
species (Mudappa et al., 2007).
Our home-range estimations suggest a difference between
male and female area utilisation, with the female using only
14.5% of the average male’s home-range size, although
our low sample size means we cannot ascertain the
representativeness of our observations compared to either the
population or the species as a whole (Fig. 3). Nevertheless,
this observation appears to support our prediction that
males would use a larger area than females, presumably to
increase reproductive access by having their range overlap
with several females (Emlen & Oring, 1977; Greenwood,
1980). The sex-based difference in home-range size may also
contribute to our male trapping bias, as a larger home-range
allows for overlap with several trapping stations, while a
female’s smaller range may not encompass traps unless they
are relatively near the individual’s activity centre. Sexual
dimorphism in home-range size of small mustelid species
results in similar trapping bias towards males (Buskirk
& Lindstedt, 1989). Similarly, the short-tailed mongoose
(Herpestes brachyurus) also shows larger home-ranges in
males (mean = 2.33 km2) than females (mean = 1.32 km2)
(Jennings et al., 2010). Home-range data on other Asian
mongoose species is lacking, with the only available data
for grey mongoose based on one individual and 21 locations
(0.155 km2), and therefore probably underestimated (Kumar
& Umapathy, 1999).
300
Sherburne et al.: Javan mongoose abundance and spatial ecology
In congruence with our abundance results, our micro-
habitat selection data shows that individual mongooses
were associated with areas with low numbers of small trees
(Table 3). This indicates a selection within DDF for more
open areas with less small woody regrowth. It may also
imply a preference for mature forests, which tend to have
larger widely spaced trees (Hamer et al., 2003; Watson et
al., 2004). We found signicantly less small trees and more
termite mounds in DDF habitat closer to the interior of the
reserve (DDF-dry evergreen edge) than in DDF along its
agricultural edge. This could be due to re suppression
by reserve staff, who patrol the edges of the DDF during
yearly burns and discourage high re intensity near the
edge of the reserve to prevent spread to agricultural areas.
Reduced burning in DDF can increase tree density and
decrease macrotermitinae species richness (Davies, 1997;
Wood, 2012). Another cause might be historical harvesting
for charcoal production near the edge of the reserve, since
charcoal harvesters often remove larger trees (Aabeyir et al.,
2016) which would create space for a greater density of small
trees once the reserve was protected. Likely several additive
factors contribute to the difference in microhabitat structure.
Javan mongoose did not select for microhabitats with higher
den site counts as we predicted, but termite mounds were
overwhelmingly selected as den sites by mongoose over
dug holes, and hollow termite mounds were our top model
for den site specic selection (Table 4). Rather than select
for termite mounds on the micro-habitat scale, our tracked
mongoose may have selected on a home-range level for
sections of the DDF where den sites were more available.
However, we cannot be certain if that is the case.
Neither overall prey biomass nor rodent specic biomass
appeared to inuence mongoose micro-habitat selection,
so our prediction that mongoose would select for areas
with higher prey biomass was not supported (Table 3).
However, we did not nd a signicant difference in rodent
biomass between different DDF habitat zones, suggesting
the availability of rodents may have been relatively uniform
across the small DDF patch we investigated. Moreover,
mongoose are generalist predators (Subrata et al., 2021), and
therefore have the ability to shift their diets in response to
resource availability, so they may simply have prioritised
other more abundant prey when rodents or arthropod
abundance were low, rather than adjust their range to match.
Finally, the low predictive ability of our micro-habitat
selection model (AUC 0.52) is likely a reection of the
relatively homogenous landscape of the DDF selected for
by mongoose.
The status of Javan mongoose populations in other natural
open habitat or in agriculture is not known. Understanding the
level of dependence Javan mongoose have on open habitat is
urgent for conservation, as open habitats are underrepresented
in Thailand’s current protected area system (Tantipisanuh &
Gale, 2013) and are often mislabelled as degraded forests
of low conservation priority (Ratnam et al., 2016). Open
forests also face unique threats such as invasion by woody
species (Stevens et al., 2017) compounded by inappropriate
re management strategies (O’Connor et al., 2014; Moura et
al., 2019) and a lack of traditional large herbivores (Daskin
et al., 2016). If Javan mongoose are as selective to open
habitat as our study suggests, similar open habitats across
Southeast Asia, such as pine savannahs or mixed teak forest,
may serve as reservoirs for the species. Camera trap surveys
specic to open forest habitat will likely provide valuable
insight on the status of current Javan mongoose populations.
ACKNOWLEDGEMENTS
Thank you to Sakaerat Biosphere Reserve’s director Surachit
Waengsothorn. We would also like to thank Jirayut Iemnok,
Kanoktip Somsiri, Phonnapa Luekhamhan, Marisa Pringproh,
Shannon Thrasher, Samantha Jones, and Dayna Levine for
their assistance in the eld. Thanks to Diane Turner, Jack
Sherburne, and Mimi Sherburne for their support. Funding
was provided by Thailand’s National Science and Technology
Development Agency (grant number P-17-50347) and the
Association of Tropical Biology Conservation.
LITERATURE CITED
Aabeyir R, Adu-Bredu S, Agyare WA & Weir MJ (2016) Empirical
evidence of the impact of commercial charcoal production
on Woodland in the Forest-Savannah transition zone, Ghana.
Energy for Sustainable Development 33: 84–95.
Bajaru S, Pal S, Mrugank P, Patel P, Khot R & Apte D (2020)
A multi-species occupancy modeling approach to access the
impacts of land use and land cover on terrestrial vertebrates
in the Mumbai Metropolitan Region (MMR), Western Ghats,
India. PLoS ONE, 15(10): e0240989.
Bates D, Mächler M, Bolker B & Walker S (2015) Fitting linear
mixed-effects models using lme4. Journal of Statistical Software,
67(1): 1–48.
Borchers DA (2012) A non-technical overview of spatially explicit
capture–recapture models. Journal of Ornithology, 152(2):
435–444.
Brooke ZM, Bielby J, Nambiar K & Carbone C (2014) Correlates
of research effort in carnivores: body size, range size and diet
matter. PLoS ONE, 9(4): e93195.
Burnham KP & Anderson DR (1998) Practical Use of the
Information-Theoretic Approach. In: Model Selection and
Inference. Springer, New York, pp. 75–117.
Burnham KP, Anderson DR & Huyvaert KP (2011) AIC model
selection and multimodel inference in behavioral ecology:
some background, observations, and comparisons. Behavioral
Ecology and Sociobiology, 65: 23–35.
Buskirk SW & Lindstedt SL (1989) Sex biases in trapped samples
of Mustelidae. Journal of Mammalogy, 70(1): 88–97.
Calabrese JM, Fleming CH & Gurarie E (2016) ctmm: an R
package for analyzing animal relocation data as a continuous
time stochastic process. Methods in Ecology and Evolution,
7(9): 1124–1132.
Case TJ & Bolger DT (1991) The role of interspecic competition
in the biogeography of island lizards. Trends in Ecology &
Evolution, 6(4): 135–139.
Chutipong W, Tantipisanuh N, Ngoprasert D, Lynam AJ, Steinmetz
R, Jenks KE, Grassman LI, Tewes M, Kitamura S, Baker
MC, McShea W, Bhumpakphan N, Sukmasuang R, Gale
GA, Harish FK, Treydte AC, Cutter P, Cutter PB, Suwanrat
S, Siripattaranukul K, Hala-Bala Wildlife Research Station,
Wildlife Research Division & Duckworth JW (2014) Current
301
RAFFLES BULLETIN OF ZOOLOGY 2022
distribution and conservation status of small carnivores in
Thailand: a baseline review. Small Carnivore Conservation,
51: 96–136.
Chutipong W, Duckworth JW, Timmins R, Willcox DHA & Ario A
(2016) Herpestes javanicus. The IUCN Red List of Threatened
Species, 2016: e.T70203940A45207619. https://dx.doi.
org/10.2305/IUCN.UK.2016-1.RLTS.T70203940A45207619.
en (Accessed 3 November 2021).
Daskin JH, Stalmans M & Pringle RM (2016) Ecological legacies
of civil war: 35 year increase in savanna tree cover following
wholesale large mammal declines. Journal of Ecology, 104(1):
79–89.
Davies RG (1997) Termite species richness in re-prone and re-
protected dry deciduous dipterocarp forest in Doi Suthep-Pui
National Park, northern Thailand. Journal of Tropical Ecology,
13(1): 153–160.
Delciellos AC, Suzy ER & Vieira MV (2017) Habitat fragmentation
effects on ne-scale movements and space use of an opossum in
the Atlantic Forest. Journal of Mammalogy, 98(4): 1129–1136.
Deuel NR, Conner LM, Miller KV, Chamberlain MJ, Cherry
MJ & Tannenbaum LV (2017) Habitat selection and diurnal
refugia of gray foxes in southwestern Georgia, USA. PLoS
ONE, 12(10): 1–12.
Duckworth JW, Timmins RJ & Tizard T (2010) Conservation status
of Small Asian Mongoose Herpestes javanicus (E. Geoffroy
Saint-Hilaire, 1818) (Mammalia: Carnivora: Herpestidae) in
Lao PDR. Rafes Bulletin of Zoology, 58(2): 403–410.
Efford MG (2021) Secr: Spatially Explicit Capture-recapture
Models. R Package Version 4.4.5. http://CRAN.R-project.org/
package=secr (Accessed 1 November 2021).
Emlen ST & Oring LW (1977) Ecology, sexual selection, and the
evolution of mating systems. Science, 197(4300): 215–223.
Estoque RC, Ooba M, Avitabile V, Hijioka Y, DasGupta R, Togawa
T & Murayama Y (2019) The future of Southeast Asia’s forests.
Natural Communication, 10(1): 1–12.
Ferreras P, Francisco DT & Monterroso P (2018) Improving
mesocarnivore detectability with lures in camera-trapping
studies. Wildlife Research, 45(6): 505–517.
Filla M, Premier J, Magg N, Dupke C, Khorozyan I, Waltert M,
Bufka L & Heurich M (2017) Habitat selection by Eurasian
lynx (Lynx lynx) is primarily driven by avoidance of human
activity during day and prey availability during night. Ecology
and Evolution, 7(16): 6367–6381.
Fiske I & Chandler R (2011) unmarked: An R package for tting
hierarchical models of wildlife occurrence and abundance.
Journal of Statistical Software, 43(10): 1–23.
Fleming CH, Calabrese JM, Mueller T, Olson KA, Leimgruber
P & Fagan WF (2014) From fine-scale foraging to home
ranges: a semivariance approach to identifying movement
modes across spatiotemporal scales. The American Naturalist,
183(5): E154–E167.
Fleming CH, Fagan WF, Mueller T, Olson KA, Leimgruber P &
Calabrese JM (2015) Rigorous home range estimation with
movement data: a new autocorrelated kernel density estimator.
Ecology, 96(5): 1182–1188.
Freeman EA & Moisen G (2008) PresenceAbsence: An R package
for presence-absence model analysis. Journal of Statistical
Software, 23(11): 1–31.
Gerber BD, Karpanty SM & Kelly MJ (2012) Evaluating the
potential biases in carnivore capture–recapture studies associated
with the use of lure and varying density estimation techniques
using photographic-sampling data of the Malagasy civet.
Population Ecology, 54(1): 43–54.
Grantham HS, Duncan A, Evans TD, Jones KR, Beyer HL, Schuster
R, Walston J, Ray JC, Robinson JG, Callow M, Clements
T, Costa HM, DeGemmis A, Elsen PR, Ervin J, Franco P,
Goldman E, Goetz S, Hansen A, Hofsvang E, Jantz P, Jupiter
S, Kang A, Langhammer P, Laurance WF, Lieberman S,
Linkie M, Malhi Y, Maxwell S, Mendez M, Mittermeier R,
Murray NJ, Possingham H, Radachowsky J, Saatchi S, Samper
C, Silverman J, Shapiro A, Strassburg B, Stevens T, Stokes
E, Taylor R, Tear T, Tizard R, Venter O, Visconti P, Want S
& Watson JEM (2020) Anthropogenic modication of forests
means only 40% of remaining forests have high ecosystem
integrity. Nature Communications, 11(1): 5978.
Greenwood PJ (1980) Mating systems, philopatry and dispersal
in birds and mammals. Animal Behaviour, 28(4): 1140–1162.
Hamer KC, Hill JK, Benedick S, Mustaffa N, Sherratt TN & Maryati
MT (2003) Ecology of butteries in natural and selectively
logged forests of northern Borneo: the importance of habitat
heterogeneity. Journal of Applied Ecology, 40(1): 150–162.
Hinton JW, Proctor C, Kelly MJ, Van Manen FT, Vaughan MR
& Chamberlain MJ (2016) Space use and habitat selection by
resident and transient red wolves (Canis rufus). PLoS ONE,
11(12): e0167603.
Holinda D, Burgar JM & Burton AC (2020) Effects of scent lure
on camera trap detections vary across mammalian predator and
prey species. PLoS ONE, 15(5): e0229055.
Jennings AP, Zubaid A & Veron G (2010) Home ranges, movements
and activity of the short-tailed mongoose (Herpestes brachyurus)
on Peninsular Malaysia. Mammalia, 74: 43–50.
Kalle R, Ramesh T, Qureshi Q & Sankar K (2013) Predicting
the distribution pattern of small carnivores in response to
environmental factors in the Western Ghats. PLoS ONE,
8(11): e79295.
Kalle R, Ramesh T, Qureshi Q & Sankar K (2014) Estimating
seasonal abundance and habitat use of small carnivores in
the Western Ghats using an occupancy approach. Journal of
Tropical Ecology, 30(5): 469–480.
Khamcha D, Corlett RT, Powell LA, Savini T, Lynam AJ & Gale
GA (2018) Road induced edge effects on a forest bird community
in tropical Asia. Avian Research, 9(1): 1–13.
Kumar A & Umapathy G (1999) Home range and habitat use
by Indian grey mongoose and small Indian civets in Nilgiri
Biosphere Reserve, India. In: Hussain SA (ed.) ENVIS
Bulletin: Wildlife and Protected Areas, Mustelids, Viverrids
and Herpestids of India, Wildlife Institute of India: 87–91.
Lowe S, Browne M, Boudjelas S & Poorter MD (2004) 100 of
the world’s worst invasive alien species. A selection from the
global invasive species database. Invasive Species Specialist
Group, Auckland, New Zealand, 11 pp.
Mace GM, Collar NJ, Gaston KJ, Hilton Taylor C, Akçakaya HR,
Leader Williams N, Milner Gulland EJ & Stuart SN (2008)
Quantication of extinction risk: IUCN’s system for classifying
threatened species. Conservation Biology, 22(6): 1424–1442.
MacKenzie DI & Bailey LL (2004) Assessing the fit of site-
occupancy models. Journal of Agricultural, Biological, and
Environmental Statistics, 9(3): 300–318.
Marneweck C, Butler AR, Gigliotti LC, Harris SN, Jensen AJ,
Muthersbaugh M, Newman A, Saldo EA, Shute K, Titus KL,
Yu SW & Jachowski DS (2021) Shining the spotlight on small
mammalian carnivores: global status and threats. Biological
Conservation, 255: 109005.
Mazerolle MJ (2020) AICcmodavg: Model selection and multimodel
inference based on (Q)AICI. R Package Version 2.3-1.
https://cran.r-project.org/package=AICcmodavg (Accessed 8
November 2021).
McPherson FJ & Chenoweth PJ (2012) Mammalian sexual
dimorphism. Animal Reproduction Science, 131(3–4): 109–122.
Moehrenschlager A, List R & Macdonald DW (2007) Escaping
intraguild predation: Mexican kit foxes survive while coyotes
and golden eagles kill Canadian swift foxes. Journal of
Mammalogy, 88(4): 1029–1039.
302
Sherburne et al.: Javan mongoose abundance and spatial ecology
Moura LC, Scariot AO, Schmidt IB, Beatty R & Russell-Smith
J (2019) The legacy of colonial fire management policies
on traditional livelihoods and ecological sustainability in
savannas: impacts, consequences, new directions. Journal of
Environmental Management, 232: 600–606.
Mudappa D, Noon BR, Kumar A & Chellam R (2007) Responses
of small carnivores to rainforest fragmentation in the southern
Western Ghats, India. Small Carnivore Conservation, 36: 18–26.
Nakashima Y (2020) Potentiality and limitations of N mixture and
Royle Nichols models to estimate animal abundance based on
noninstantaneous point surveys. Population Ecology, 62(1):
151–157.
Noonan MJ, Tucker MA, Fleming CH, Akre TS, Alberts SC,
Ali AH, Altmann J, Antune PC, Belant JL, Beyer D, Blaum
N, Bohning-Gaese K, Cullen Jr L, de Paula RC, Dekker J,
Drescher-Lehman JD, Farwig N, Fichtel C, Fischer C, Ford
A, Goheen JR, Janssen R, Jeltsch F, Kauffman M, Kappeler
PM, Koch F, LaPoint S, Markham AC, Medici EP, Morato
RG, Nathan R, Oliveira-Santos LGR, Olson KA, Pattersoh BD,
Paviolo A, Ramalho EE, Roesner S, Schabo D, Selva N, Sergiel
A, da Silva X, Spiegel O, Thompson P, Ullmann W, Zieba
F, Zwijacz-Kozica T, Fagan WF, Mueller T & Calabrese JM
(2019) A comprehensive analysis of autocorrelation and bias in
home range estimation. Ecological Monographs, 89(2): e01344.
O’Brien TG, Kinnaird MF & Wibisono HT (2003) Crouching
tigers, hidden prey: Sumatran tiger and prey populations in a
tropical forest landscape. Animal Conservation, 6: 131–139.
O’Connor CD, Falk DA, Lynch AM & Swetnam TW (2014)
Fire severity, size, and climate associations diverge from
historical precedent along an ecological gradient in the Pinaleño
Mountains, Arizona, USA. Forest Ecology and Management,
329: 264–278.
Oliver K, Ngoprasert D & Savini T (2019). Slow loris density in a
fragmented, disturbed dry forest, nort east Thailand. American
Journal of Primatology, 81(3): e22957.
Ongsomwang S & Sutthivanich I (2014) Integration of remotely
sensed data and forest landscape pattern analysis in Sakaerat
Biosphere Reserve. Suranaree Journal of Science and
Technology, 21(3): 233–248.
Owen MA & Lahti DC (2020) Rapid evolution by sexual selection
in a wild, invasive mammal. Evolution, 74(4): 740–748.
Paolino RM, Royle JA, Versiani NF, Rodrigues TF, Pasqualotto N,
Krepschi VG & Chiarello AG (2018) Importance of riparian
forest corridors for the ocelot in agricultural landscapes. Journal
of Mammalogy, 99(4): 874–884.
Paradis E & Schliep K (2019) ape 5.0: an environment for modern
phylogenetics and evolutionary analyses in R. Bioinformatics,
35: 526–528.
Petersen WJ, Savini T, Steinmetz R & Ngoprasert D (2019) Periodic
resource scarcity and potential for interspecic competition
inuences distribution of small carnivores in a seasonally dry
tropical forest fragment. Mammalian Biology, 95(1): 112–122.
R Core Team (2020) R: A language and environment for statistical
computing. R Foundation for Statistical Computing, Vienna,
Austria. https://www.R-project.org/ (Accessed 8 November
2021).
Ramesh T, Kalle R & Downs CT (2017) Space use in a South
African agriculture landscape by the caracal (Caracal caracal).
European Journal of Wildlife Research, 63(1): 1–11.
Ratnam J, Tomlinson KW, Rasquinha DN & Sankaran M (2016)
Savannahs of Asia: antiquity, biogeography, and an uncertain
future. Philosophical Transactions of the Royal Society B:
Biological Sciences, 371(1703): 20150305.
Royle JA & Nichols JD (2003) Estimating abundance from repeated
presence–absence data or point counts. Ecology, 84(3): 777–790.
Sikes RS & Gannon WL (2011) Guidelines of the American Society
of Mammalogists for the use of wild mammals in research.
Journal of Mammalogy, 92(1): 235–253.
Simberloff D, Dayan T, Jones C & Ogura G (2000) Character
displacement and release in the small Indian mongoose,
Herpestes javanicus. Ecology, 81(8): 2086–2099.
Sodhi NS, Posa MRC, Lee TM, Bickford D, Koh LP & Brook
BW (2010) The state and conservation of Southeast Asian
biodiversity. Biodiversity and Conservation, 19(2): 317–328.
Stevens N, Lehmann CER, Murphy BP & Durigan G (2017) Savanna
woody encroachment is widespread across three continents.
Global Change Biology, 23(1): 235–244.
Subrata SA, Siregar SRT, André A & Michaux JR (2021) Identifying
prey of the Javan mongoose (Urva javanica) in Java from
fecal samples using next-generation sequencing. Mammalian
Biology, 101(1): 63–70.
Suraci JP, Clinchy M, Dill LM, Roberts D & Zanette LY (2016)
Fear of large carnivores causes a trophic cascade. Nature
Communications, 7(1): 1–7.
Tannerfeldt M, Moehrenschlager A & Angerbjörn A (2003) Den
Ecology of Swift, Kit and Arctic Foxes: A Review. In: Sovada
M & Carbyn L (eds.) Ecology and Conservation of Swift
Foxes in a Changing World. Canadian Plains Research Center,
University of Regina, pp. 167–181.
Tantipisanuh N & Gale GA (2013) Representation of threatened
vertebrates by a protected area system in southeast Asia: the
importance of non-forest habitats. Rafes Bulletin of Zoology,
61(1): 359–395.
Taubert F, Fischer R, Groeneveld J, Lehmann S, Müller MS, Rödig
E, Wiegand T & Huth A (2018) Global patterns of tropical
forest fragmentation. Nature, 554(7693): 519–522.
Vanbianchi CM, Melanie AM & Karen EH (2017) Canada lynx
use of burned areas: conservation implications of changing re
regimes. Ecology and Evolution, 7(7): 2382–2394.
Veron G & Jennings AP (2017) Javan mongoose or small Indian
mongoose–who is where? Mammalian Biology, 87(1): 62–70.
Veron G, Patou ML, Pothet G, Simberloff D & Jennings AP
(2007) Systematic status and biogeography of the Javan and
small Indian mongooses (Herpestidae, Carnivora). Zoologica
Scripta, 36(1): 1–10.
Watson JEM, Whittaker RJ & Dawson TP (2004) Habitat
structure and proximity to forest edge affect the abundance
and distribution of forest-dependent birds in tropical coastal
forests of southeastern Madagascar. Biological Conservation,
120(3): 311–327.
Wei T & Simko V (2021) R package “corrplot”: Visualization of
a Correlation Matrix. (Version 0.88). https://github.com/taiyun/
corrplot (Accessed 8 November 2021).
Wilcove DS, Giam X, Edwards DP, Fisher B & Koh LP (2013)
Navjot’s nightmare revisited: logging, agriculture, and
biodiversity in Southeast Asia. Trends in Ecology and Evolution,
28: 531–540.
Wolff PJ, Taylor CA, Heske EJ & Schooley RL (2015) Habitat
selection by American mink during summer is related to hotspots
of craysh prey. Wildlife Biology, 21(1): 9–17.
Wood TF (2012) Fire ecology of the dry dipterocarp forests of
South West Cambodia. Cambodian Journal of Natural History,
2012(1): 64–74.
Zoletil the Versatile Anaesthetic (2011) Virbac Animal Health
India PVT LTD. https://in.virbac.com/les/live/sites/virbac-
in/les/predened-les/PDF%20documents/Zoletil%20-%20
Wildlfe%20dosage%20guidelines.pdf (Accessed 23 August
2021).
303
RAFFLES BULLETIN OF ZOOLOGY 2022
APPENDIX
Appendix 1. Morphological data for Javan mongoose (Herpestes javanicus) trapped within Sakaerat Biosphere Reserve.
ID Sex Age
class
Tail length
(cm)
Body
length (cm)
Mass
(g)
Circumference
of neck (cm)
Circumference
of skull at
mandible (cm)
Hind paw
width (cm)
Front paw
width (cm)
M1 male adult 35.5 45.0 1055 14.1 15.1 2.3 2.2
M2 male adult 33.0 30.0 1130 14.2 14.9 1.3 1.2
M3 male young
adult
34.6 30.5 720 12.5 13.2 1.9 1.9
F1 female young
adult
38.0 32.0 600 10.2 11.4 1.2 1.2
M4 male adult 35.2 33.5 710 11.7 14.4 2.0 2.4
M5 male adult 33.1 36.4 1250 15.5 16.8 2.5 2.5
M6 male adult 31.5 41.2 1100 14.5 15.9 2.4 2.2
Appendix 2. Immobilisation dosing and details of Javan mongoose (Herpestes javanicus) trapped within Sakaerat Biosphere
Reserve.
We anaesthetised Javan mongoose (Herpestes javanicus)
using Zoletil™ 100 for injection (tiletamine and zolazepam).
The addition of 5 ml diluent produces a solution containing
the equivalent of 50 mg tiletamine base, 50 mg zolazepam
base, and 57.7 mg mannitol per millilitre. This solution has
a pH of 2 to 3.5 and is recommended for deep intramuscular
injection. The doses used in this study were 5.0 mg/kg
intramuscular injections (personal recommendation by Doctor
Marnoch Yindee).
Average anaesthesia duration of each individual lasted 60–180
minutes longer than recommended by Virbac pharmaceutical
company (20–60 min). We recommend for future researchers
to reduce the dose to 4.4 mg/kg as recommended for black-
footed mongoose (Bdeogale nigripes) (Zoletil the Versatile
Anaesthetic, 2021). We prepared Doxapram and Yohimbine
HCI as an antidote of tilemine and zolazepam in case of
accidental overdose.
304
Sherburne et al.: Javan mongoose abundance and spatial ecology
Appendix 3. Spatially explicit capture-recapture models for small and large mass species group rodent densities. Behavioural effects tested:
~b, change in behaviour after rst capture; ~B, change in behaviour depending on capture previous occasion; ~bk, trap-specic lingering
response post initial capture; ~Bk, trap-specic response post capture.
Small Mass Species (< 50 g)
variables KAICc ΔAICc wi
D~session, g0~bk, sigma~1 13 2515.00 0.00 1.00
D~session, g0~b, sigma~1 13 2527.07 12.07 0.00
D~session, g0~Bk, sigma~1 13 2541.35 26.35 0.00
D~session, g0~B, sigma~1 13 2543.31 28.31 0.00
D~session, g0~1, sigma~1 12 2563.13 48.13 0.00
D~session, g0~session, sigma~session 30 2576.59 61.59 0.00
D~1, g0~1, sigma~1 3 2610.19 95.19 0.00
Large Mass Species (> 50 g)
D~session, g0~Bk, sigma~1 13 2823.71 0.00 0.94
D~session, g0~b, sigma~1 13 2830.63 6.92 0.03
D~session, g0~bk, sigma~1 13 2830.68 6.97 0.03
D~session, g0~B, sigma~1 13 2847.27 23.56 0.00
D~session, g0~1, sigma~1 12 2849.13 25.42 0.00
D~1, g0~1, sigma~1 3 2869.55 45.84 0.00
D~session g0~session sigma~session 30 2870.70 46.99 0.00
K is the number of parameters included in the model, AICc is Akaike’s Information Criteria corrected for small sample size, ΔAICc is
the difference in AICc values, and wi is the Akaike weight.
