ArticlePDF Available

Loss of Sunda clouded leopards and forest integrity drive potential impacts of mesopredator release on vulnerable avifauna

Authors:

Abstract and Figures

Amongst the unintended consequences of anthropogenic landscape conversion is declining apex predator abundance linked to loss of forest integrity, which can potentially re-order trophic networks. One such re-ordering, known as mesopredator release, occurs when medium-sized predators, also called mesopredators, rapidly increase in abundance following the decline in apex predator abundance, consequently reducing the abundance of mesopredator prey, notably including terrestrial avifauna. We examine the cascading impacts of declining Sunda clouded leopard abundance, itself consequent upon a reduction in forest integrity, on the mesopredator community of Sabah, Malaysia, to determine whether the phenomenon of mesopredator release is manifest and specifically whether it impacts the terrestrial avifauna community of pheasants and pittas. To explore this trophic interaction, we used a piecewise structural equation model to compare changes in the relative abundance of organisms. Our results suggest that loss of forest integrity may have broad impacts on the community and trigger mesopredator release, the two acting additively in their impact on already vulnerable species of terrestrial avifauna: a result not previously documented in tropical systems and rarely detected even on a global scale. The limiting effect that the Sunda clouded leopard has on the Sunda leopard cat could illuminate the mechanism whereby mesopredator release impacts this system. Both Bulwer's pheasant and pittas appear to be significantly impacted by the increase in Sunda leopard cats, while the great argus pheasant shows similar compelling, although not statistically significant, declines as Sunda leopard cats increase. The inverse relationship between Sunda clouded leopards and Sunda leopard cats suggests that if a mesopredator release exists it could have downstream consequences for some terrestrial avifauna. These results suggest the under-studied interface between mammalian carnivores and avifauna, or more broadly species interactions in general, could offer important conservation tool for holistic ecosystem conservation efforts.
Content may be subject to copyright.
Heliyon 10 (2024) e32801
Available online 11 June 2024
2405-8440/© 2024 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
Research article
Loss of Sunda clouded leopards and forest integrity drive potential
impacts of mesopredator release on vulnerable avifauna
Darwin S. Mayhew
a
,
*
, Andrew J. Hearn
c
, Olivier Devineau
a
, John D.C. Linnell
a
,
b
,
David W. Macdonald
c
a
Department of Forestry and Wildlife Management, Inland Norway University of Applied Sciences - Campus Evenstad, Anne Evenstads Vei 80, 2480,
Koppang, Norway
b
Norwegian Institute for Nature Research, Vormstuguveien 40, 2624, Lillehammer, Norway
c
WildCRU, Department of Biology, University of Oxford, Recanati-Kaplan Centre, Tubney House, Abingdon Rd, Tubney, OX13 5QL, United
Kingdom
ARTICLE INFO
Keywords:
Forest integrity
Mesopredator release
Bird
Structural equation model
Trophic cascade
Oil palm plantations
ABSTRACT
Amongst the unintended consequences of anthropogenic landscape conversion is declining apex
predator abundance linked to loss of forest integrity, which can potentially re-order trophic
networks. One such re-ordering, known as mesopredator release, occurs when medium-sized
predators, also called mesopredators, rapidly increase in abundance following the decline in
apex predator abundance, consequently reducing the abundance of mesopredator prey, notably
including terrestrial avifauna. We examine the cascading impacts of declining Sunda clouded
leopard abundance, itself consequent upon a reduction in forest integrity, on the mesopredator
community of Sabah, Malaysia, to determine whether the phenomenon of mesopredator release is
manifest and specically whether it impacts the terrestrial avifauna community of pheasants and
pittas. To explore this trophic interaction, we used a piecewise structural equation model to
compare changes in the relative abundance of organisms. Our results suggest that loss of forest
integrity may have broad impacts on the community and trigger mesopredator release, the two
acting additively in their impact on already vulnerable species of terrestrial avifauna: a result not
previously documented in tropical systems and rarely detected even on a global scale. The
limiting effect that the Sunda clouded leopard has on the Sunda leopard cat could illuminate the
mechanism whereby mesopredator release impacts this system. Both Bulwers pheasant and pittas
appear to be signicantly impacted by the increase in Sunda leopard cats, while the great argus
pheasant shows similar compelling, although not statistically signicant, declines as Sunda
leopard cats increase. The inverse relationship between Sunda clouded leopards and Sunda
leopard cats suggests that if a mesopredator release exists it could have downstream consequences
for some terrestrial avifauna. These results suggest the under-studied interface between
mammalian carnivores and avifauna, or more broadly species interactions in general, could offer
important conservation tool for holistic ecosystem conservation efforts.
* Corresponding author. Inland Norway University of Applied Sciences - Campus Evenstad, Anne Evenstads Vei 80, 2480, Koppang, Norway.
E-mail addresses: mayhewdarwin@gmail.com (D.S. Mayhew), andrew.hearn@biology.ox.ac.uk (A.J. Hearn), john.linnell@nina.no
(O. Devineau), olivier.devineau@inn.no (J.D.C. Linnell), david.macdonald@biology.ox.ac.uk (D.W. Macdonald).
Contents lists available at ScienceDirect
Heliyon
journal homepage: www.cell.com/heliyon
https://doi.org/10.1016/j.heliyon.2024.e32801
Received 2 June 2024; Received in revised form 8 June 2024; Accepted 10 June 2024
Heliyon 10 (2024) e32801
2
1. Introduction
The dynamics of ecological communities in forests remain poorly understood, with cryptic species being amongst those lost at the
highest rates to deforestation or agricultural conversion in tropical forest systems [1,2]. The ecological impacts of degraded forest
integrity, and their consequences, have become critical to understanding biodiversity loss in ecological networks [3,4]. Forest integrity
loss not only has direct impacts at varying trophic levels but can also result in cascading impacts between trophic levels [5]. Apex
predators are of particular interest in this context as their disappearance, often triggered by external effects such as landscape con-
version, can lead to trophic cascades through the food web that can alter fundamental ecosystem functions [68].
One mechanism by which these consequences can occur is mesopredator release, in which the decline in, or disappearance of, an
apex predators population results in a population increase of small to medium-sized predators (mesopredators) which in turn causes a
decline in populations of the latters prey [912]. This mechanism is considered to occur at higher frequencies when mesopredators
utilize prey also utilized by the apex predator and when the body-mass ration of apex predators to mesopredators is on average be-
tween 2 and 5.4 as to ensure the risk-reward trade-off of interspecic killing is in the apex predators favor [11]. While mesopredator
release has been thoroughly documented in intraclass systems, it can also impact prey at the intersection of terrestrial and avian
communities through nest depredation, direct predation, or a combination of the two [13,14]. An early documentation of this effect of
mesopredator release revealed how fragmentation of the Southern Californian landscapes led to a decline in coyotes (Canis latrans)
causing a mesopredator release of domesticated cats (Felis catus) that had cascading impacts on their avian prey [13].
As extinction rates amongst rare, specialized, and large-bodied species of tropical forest-dwelling birds are disproportionately high,
mesopredator release may be a causal link between similar declines in tropical apex predators and a reduction in terrestrial forest-
dwelling birds [1518]. The loss of birds is of concern as their decline can have various impacts on ecosystem services including
changes to seed dispersal, pollination, carrion consumption, nutrient cycling, and populations of invertebrates or vertebrates, of which
some are relevant as pests [19]. Globally the loss of native birds has been attributed to, among other factors, increases in domestic cat
(Felis catus) abundance which we believe could be analogous to the effects from mesopredator release of small native felids on tropical
bird species [20,21].
To explore mesopredator release at the intersection of the tropical bird and mammal communities, we investigated the impact of
declining forest integrity on the vertebrate community dynamics of Sabah, Malaysia. As one of the most biodiverse places in the world,
it has experienced rapid conversion of primary tropical forest to oil palm plantations that has broadly impacted forest integrity and
biodiversity, including the loss of apex predators, notably the Sunda clouded leopard (Neofelis diardi), thereby creating conditions
likely to prompt mesopredator release [4,22,23]. Previous studies have, using camera traps, primarily investigated population sizes,
movement, and demographics of feline carnivores with emphasis on the Sunda clouded leopard. Unfortunately, detailed dietary and
interaction data of these species is limited due to the climate of the region and challenge of capturing individuals making grounded
claims about individual predation between species difcult.
Luckly, camera-trapping has proven an effective means of simultaneously monitoring some groups of terrestrial birds, such as
pheasants, in addition to medium-sized and large mammals [24]. This facilitates an investigation of mesopredator release through the
following species relevant to our hypothesis: Sunda clouded leopard (12.025.2 kg); [apex predator], Sunda leopard cat (Prionailurus
javanensis; 1.72.9 kg); [mesopredator], great argus pheasant (Argusianus argus; 1.591.7 kg); [prey], Bulwers pheasant (Lophura
bulweri; 0.911.8 kg); [prey], crested reback pheasant (Lophura ignita; 1.62.6 kg); [prey], and the pitta family consisting of six
species (Family: Pittidae; 0.0420.21 kg; specic species included: black-crowned pitta (Erythropitta ussheri), Bornean banded-pitta
(Hydrornis schwaneri), blue-banded pitta (Erythropitta arquata), blue-headed pitta (Hydrornis baudii), Western hooded pitta (Pitta
sordida), and giant pitta (Hydrornis caeruleus)); [prey]. Hereafter we refer to Sunda clouded leopards as clouded leopards and Sunda
leopard cats as leopard cats, but these should not be confused with the mainland species of Neofelis nebulosa and Prionailurus bengalensis
respectively.
Our goal is to better understand the intraguild interaction of the two felid species, and how those dynamics might impact avifauna
through the cascading effects of possible mesopredator release in this relatively undocumented ecosystem. To do this we predicted 1)
loss of forest integrity would be associated with decreased abundance of apex predators, 2) decrease in the abundance of apex
predators would be associated with increased abundance of mesopredators suggesting mesopredator release, 3) an increase in mes-
opredator abundance would be associated with a subsequent decline in pheasant and pitta abundance.
2. Methods
2.1. Study design
Across the Malaysian state of Sabah on the island of Borneo camera traps were deployed between May 2007 and December 2021,
with all but one camera grid initiated before January 2014. The cameras were located along roads, game trails, or ridgelines, between
0 and 1600 m in elevation at approximately regular 1-km intervals to form camera grids [2527]. A total of eleven grids consisted of
1579 camera stations totaling 498 independent stations (Table A1). At each station, two cameras were deployed (Totaling 996
cameras deployed) ca. 30 cm off the ground, facing one another, to capture both sides of photographed animals and to increase
probability of detection. Six grids were in relatively intact lowland or lower montane forest. Two grids were placed in a mix of
selectively logged lowland forest, fragmented plantations, and mangroves, and three in oil palm plantations [26]. However, as
detection rates of our target species are non-uniform across study areas, the subdivision of the original grids helped account for the
potential spatial variation in abundance and increase the strength of the general linear models that compose our subsequent piecewise
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
3
Fig. 1. Camera Trapping Locations
Camera-trapping grids are depicted as orange and blue dashed outlines in the full Sabah map. Full sub-grids are indicated by color with numbers
indicating the sub-grid number in all the full grids included in A.: 1. Danum Palm, Danum Valley, IJM, Kinabatangan, Malua, Sepilok, Ulu Segama
Tabin; B: Crocker; and C: Tawau, Sabah Softwoods. Forest integrity was based on the Grantham et al. [29] map of anthropogenic modication of
forests with green areas representing the most intact habitat and brown areas the most degraded. Numbers indicate the sub-grid number within each
color-coded grid ranging from 1 to 5 sub-grids (e.g. Danumn Valley) with some grids only having one sub-grid or in other words an unbroken
original grid (e.g. Ulu Segama). (For interpretation of the references to color in this gure legend, the reader is referred to the Web version of
this article.)
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
4
structural equation model. Therefore, we used a constrained k-means clustering algorithm to subdivide the eleven grids into thirty
sub-grids, each containing at least fteen camera traps [28]. Four camera locations were removed post subdivision as they were stolen
or broken while in the eld resulting in no data being collected.
2.2. Estimation of forest integrity
To represent the anthropogenic impacts of land use change, we used Grantham et al.‘s [29] forest loss integrity index, hereafter
referred to as forest integrity, which is a measure of deviation from the natural state calculated for forest conditions in 2019. After
visually comparing the forest integrity GIS layer to satellite imagery from data collection years, we concluded that land use had not
changed signicantly at the spatial scale we considered for the estimation of species abundances. Average forest integrity values were
calculated for each sub-grid area (see Table A2), which was determined by calculating a minimum convex polygon (MCP) around each
set of camera-trap stations that make up their respective sub-grid, plus an additional 100-m buffer added to each MCP (see Fig. 1) in
QGIS [30]. However, as forest integrity was specied to range from 0 (low) to 10 (high), missing forest integrity values (as calculated
by Grantham et al. [29]) were truncated to zero. This was done because missing forest integrity cell values in the original raster le
were based on forest cover of less than 5 m in height; we interpreted this as equivalent to the poorest habitat possible for
forest-dependent species.
2.3. Detectability-corrected abundance estimates
Following Cunningham et al. [31], we calculated detectability-corrected estimates of relative abundance derived from
presence-absence and group counts to be incorporated into the next step of our analysis. We rst used the camtrapR package [32] in the
R environment [33] to collapse photographic observations of all species into 60-min intervals to ensure both temporal independence of
observations and to match older camera data to newer data resulting in 57,283 useable photo records. Given the complex nature of
Sabahan ecosystems and to ensure estimates were sufciently robust, we limited our analyses to ecologically relevant species with
enough camera observations to derive abundance estimates and pooled all six pitta species into a family group. We then created
detection histories for each camera-trap location (N =498) to facilitate species abundance models.
Since carnivores (clouded leopard and leopard cat) and pittas were mostly observed as single individuals, we considered these data
to be presence/absence, and we used the Royle-Nichols abundance model [34] within the unmarked package [35] to estimate their
abundance. On the other hand, pheasants (great argus, crested reback, and Bulwers pheasant) were observed in groups of varying
sizes, leading us to estimate abundance using the N-mixture model [36] also within the unmarked package [35], with group size
dened by the largest number of individuals observed in any one photo within the 60-min interval.
Fig. 2. An a-priori piecewise SEM
Forest integrity has declined rapidly due to oil palm plantations and related human inuences across Southeast Asia. This gure depicts our a-priori
piecewise structural equation model and the potential restructuring of the Sabah felid community following forest integrity loss and its subsequent
effects on the terrestrial avifauna species community. Red lines indicate hypothetical negative relationships, blue lines indicate hypothetical positive
relationships, and drop shadows represent predicted correlated error with a species own squared values. (For interpretation of the references to color
in this gure legend, the reader is referred to the Web version of this article.)
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
5
Fig. 3. A nal piecewise SEM
Our nal piecewise structural equation model showing forest integrity and Sunda clouded leopard abundance have a trophic cascading effect on
abundance of Sunda leopard cats and subsequently on terrestrial avifauna species. Nodes are our species of interest and the forest loss integrity index
extracted from Grantham et al.‘s [29] study on global forest integrity. All solid lines represent signicant pathways from our most parsimonious SEM
at an alpha level of
α
=0.05, with blue lines representing positive relationships, red lines representing negative relationships, purple lines rep-
resenting downward parabolic relationships, and yellow lines representing upward parabolic relationships. The dotted line represents a retained
non-signicant pathway between Sunda leopard cats and great argus pheasants. The dashed line between pittas and Bulwers pheasants represents a
specied correlated error between variables. A second correlated error exists between Sunda clouded leopards and its own squared values that is not
depicted here. Line thickness correlates with coefcient. P-values and coefcients for each species modeled in our SEM are listed in Table 1. (For
interpretation of the references to color in this gure legend, the reader is referred to the Web version of this article.)
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
6
Both the Royle-Nichols model [34] and the N-mixture model [36] include a sub-model for detection, and a (latent) sub-model for
abundance (see Cunningham et al. [31]; Fiske & Chandler [35]; Nakashima [37]; Royle [36]; Royle & Nichols [34] for more details on
these models). We modeled detection as a function of effort (i.e., number of days a given camera station was active), and of pre-
sence/absence of forest roads and ridge lines at the individual camera station [34] (see Table A3). We modeled abundance as a function
of our specic spatial sample units (i.e., 30 sub-grids) which was derived from splitting the original eleven study sites. These models
produce an estimate of abundance for each of the 30 sub-grids by either exploiting the link between detection probability and
abundance as with the Royle-Nichols model [34] or by using repeated count data as with the N-mixture model [36]. We only
considered the 3 detection covariates listed to avoid the strong assumptions with respect to ne scale landscape change inherent to
deriving more covariates from 10-years old remote sensing data under dense canopy cover of our study area [26,32,38].
2.4. Piecewise structural equation model
To assess potential cascading effects in the trophic network in relation to forest integrity, we built a regression model for each of the
6 focal species. We based our focal species models on abundance estimates for each sub-grid from the previous models (N =30), which
we combined into a piecewise structural equation model (SEM) tted with the piecewiseSEM package in R [39]. Prior to comparing
abundance estimates we multiplied all numeric variables by one hundred and rounded the results to allow for the use of negative
binomial models without altering the estimates. For all species we used generalized linear models with a negative binomial distri-
bution. Owing to the limited number of sub-grids, we did not have sufcient observations to include the original camera grids as a
varying intercept to account for the structure of the original study design. Starting from an a priori SEM model (Fig. 2), consisting of 15
pathways, we used an AIC-based stepwise model reduction to remove insignicant pathways (
α
=0.05), until only signicant
pathways remained with one exception (see Cunningham et al. [31] or Gordon et al. [40], for a similar approach). The retention of the
relationship between leopard cats and great argus pheasants in our model was non-signicant based on P-values but was retained
based on a less than two-point change in deltaAIC, a visual inspection of the paired data which displayed a relationship paralleling that
of other terrestrial bird species retained based on P-values, and the objective of this study being an exploratory/descriptive one [41].
We assessed the overall t of the nal SEM using Shipleys test of d-separation [42,43], which tests whether all unconnected variables
are conditionally independent. To account for pathways that were not conditionally independent and unspecied in our a-priori
model, we specied these relationships as partially correlated to account for the effects of covariance. These relationships included one
pathway between Bulwers pheasant abundance and pitta abundance as well as a pathway between clouded leopards abundance and
the squared values of clouded leopards abundance. The inclusion of these relationships in our model was not necessary for the Fishers
C statistic to have a P >0.05; however, their inclusion helped account for all possible relevant connections (see Table A4 for a list of
independence claims from the d-separation test).
Table 1
a
Model Coefcient (SE) P-value
Sunda clouded leopard; GLM
(Intercept) 2.54 (0.78) 0.003 **
Mean Forest Integrity 0.003 (0.001) 0.01*
Sunda leopard cat; GLM
(Intercept) 6.28 (0.32) <2e-16 ***
Sunda clouded leopard 0.02 (0.005) 0.0002***
I (Sunda clouded leopard^2) 0.006 (0.002) 0.002 **
Great Argus Pheasant; GLM
(Intercept) 0.64 (1.01) 0.54
Sunda leopard cat 0.0007 (0.0007) 0.33
Mean Forest Integrity 0.008 (0.001) 2.62e-7 ***
Bulwers Pheasant; GLM
(Intercept) 8.07 (2.27) 0.002 **
Sunda clouded leopard 0.12 (0.02) 3.77e-5 ***
I (Sunda clouded leopard^2) 0.04 (0.008) 7.78e-5 ***
Sunda leopard cat 0.016 (0.004) 0.0005 ***
Mean Forest Integrity 0.006 (0.003) 0.02 *
Pitta Family; GLM
(Intercept) 1.90 (0.96) 0.06
Sunda leopard cat 0.01 (0.003) 0.0001 ***
Mean Forest Integrity 0.003 (0.001) 0.02 *
Bulwers Pheasant/Pitta Family, CE 0.50 (N/A) 0.0028 **
Sunda clouded leopard/
a
(Sunda clouded leopard^2), CE 0.66 (N/A) 0.000 ***
a
Piecewise structural equation models results of the local estimates for each general linear model that compose the global model.
Models were built using general linear models (GLM) and specied correlated error relationships (CE). Estimates are standardized and P-
values are marked as signicant at an
α
=0.05.
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
7
Fig. 4. Species abundance relationship models
These key pathways from our SEM depict how (a) the abundance of our apex predator the Sunda clouded leopard is tied to the forest loss integrity
index and (b) how that decline subsequently results in an increase in abundance of the Sunda leopard cat, a mesopredator. In turn cascading
negative impacts from increased Sunda leopard cat abundance appear to affect (c) great argus pheasants abundance, (d) Bulwers pheasant
abundance, and (e) the pittasabundance. Points denote detectability corrected measures of abundance for each species in each of the 30 sub-grids
using the Royle-Nichols model of abundance to estimate felid species and the N-mixture model of abundance to estimate terrestrial avifauna. Each
graph has a grey line (±95 % CI) indicating the respective general linear model used for each species or functional group. Forest integrity is
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
8
3. Results
3.1. Detectability-corrected abundance estimates
Mean forest integrity ranged from 184 (low) to 994 (high). Estimates of relative abundance at each of the 30 sub-grids obtained
from the N-mixture and Royle-Nichols models ranged from 0 to 272 for clouded leopards, 18 to 1775 for leopard cats, 0 to 1393 for
great argus pheasants, 0 to 302 for Bulwers pheasants, 0 to 3005 for crested rebacks, and 0 to 131 for the pitta family. All estimates
and associated standard errors are presented in Table A2.
3.2. Piecewise structural equation model
Our nal piecewise Structural Equation Model (Fig. 3, Table 1) included eleven pathways and two partially correlated connections
(Table A5). Forest integrity positively corresponded to four pathways including all species abundance estimates aside from leopard
cats and crested rebacks but was the only predictor for clouded leopard abundance (Fig. 4a.). Clouded leopard abundance corre-
sponded to two pathways including one quadratic relationship with leopard cat abundance (Fig. 4b.) and one quadratic relationship
with Bulwers pheasants. Leopard cats corresponded negatively with the abundance of three species of the terrestrial avifauna
including great argus pheasant (Fig. 4c), Bulwers pheasants (Fig. 4d), and pittas (Fig. 4e), although only the two pathways with
Bulwers pheasant and pitta abundance had signicant relationships with leopard cat abundance.
4. Discussion
We provide some evidence for mesopredator release for the rst time in the Bornean avifaunal community involving endemic
felids, and we achieve this using, also for the rst time in this system, a piecewise structural equation model. The community dynamics,
and guild structure of felids in Borneo, and more widely in Southeast Asia, are very poorly understood. It is an important insight into
their ecology that our results suggest an inverse parabolic relationship between abundances of the clouded leopards and leopard cats.
Our ndings also make a case that disruptive cascading impacts of landscape conversion to oil palm plantations change the meso-
predator and ground bird community through the reduction in clouded leopard abundance as a result of a possible mesopredator
release of the leopard cat.
It is highly probable that clouded leopard populations have declined over the last century, likely triggered by the decrease in forest
integrity related to the increase in oil palm plantations [23]. Conversely, leopard cats are considered oil palm adapters’’, using
human-dominated landscapes to avoid predation, competition from other felid species, and/or to take advantage of the heterogeneity
in the landscape for food or shelter [44]. Our results add to these hypotheses by suggesting that Sunda leopard cats ourish in oil palm
plantations partly because of the absence of clouded leopards, which leads to an increase in their abundance in this habitat. Notably
our ndings depict an inverse parabolic relationship between the abundance of clouded leopards and leopard cats but no relationship
between forest integrity and leopard cat abundance which we would expect if habitat preference were driving this relationship.
This relationship between clouded leopards and leopard cats is possibly linked to our small sample size, which we suspect, given the
prevailing negative relationship, would smooth out as sample size increases highlighting an overall negative relationship between the
two species. In subsequent testing the removal of the outlying point driving the parabolic relationship resulted in a signicant negative
relationship. However, given the small sample size and lack of tools in the SEM, such as general additive models (GAM), the quadratic
function was used to increase exibility in our model to capture the non-linear relationship caused by the impact of statistical outliers
and/or any potentially unintended abiotic or biotic factors we may not have included in the model. In an attempt to offer a biological
explanation for the inverse parabolic relationship presented here we hypothesize this may represent scenarios under which high
resource density, increased landscape heterogeneity, or some unmeasured form of spatiotemporal disturbance would impact one or
both species abundance [45]. The outlying nature of this point could plausibly have been derived from land conversion and logging,
affecting predator abundance in Ulu Segama region at the time of data collection resulting in high co-abundance of felid species in our
model. The initiation of the 20072009 forest management plan that saw a spike in removed vegetation and planting/rehabilitation
efforts may explain this datum point as an outlier in our study. However, further research would be needed to conrm how such
management practices impact both individual species and/or ecosystem networks. Thus, our ndings prompt the question of whether
the mechanism resulting in the parabolic relationship between abundances of clouded leopards and leopard cats is solely due to
mesopredator release or whether, at least to some extent, it is due to a shift in carrying capacity, population structure, or other factors
that contribute to changes in abundances that could also contribute to the relationship between these species across their spatio-
temporal range.
Presuming, subject to further conrmatory research, that mesopredator release is one of the casual mechanisms of the species
relationships we identify in our piecewise SEM, we propose the following hypothesis as to how mesopredator release might affect our
study species. We speculate that leopard cats, in the absence of clouded leopards, either directly prey on terrestrial birds and/or create
indicated for each point on a scale from low integrity (Blue) to high integrity (Yellow). The size of the points is representative of the number of
cameras in each sub-grid location with the smallest points representing the minimum camera number of fourteen and the largest a maximum of
twenty-three. Forest integrity scale dened by Grantham et al. [29] low (600); medium (>600) and high integrity (960). (For interpretation of
the references to color in this gure legend, the reader is referred to the Web version of this article.)
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
9
a landscape of fear altering the relative fecundity of prey [46]. However, crested reback pheasants unlike the other prey species
showed no signicant response to either forest integrity or increases in mesopredator abundance, which we hypothesize is evidence of
a non-uniform response to mesopredator abundance from prey species in this study [11]. It is possible that specic life-history traits or
average prey body weight, insofar as this probably affects handling capacity of predators, determines response to the impact of
increased leopard cat abundance [47]. Alternatively, the release of mesopredators might not be uniform across leopard cat de-
mographics resulting in older/larger individuals persisting longer. In a parallel case involving domestic cats in Australia anecdotal
evidence suggested these traits permit predation of dangerous prey including brushtailed possum (Trichosurus vulpecula),
black-headed monitor (Varanus tristis), and domestic chicken (Gallus gallus domesticus) [48].
Among the mesopredators for which sample sizes in a camera-trapping study were too small to include in this analysis, the Bornean
bay cat (Catopmua badia) and marbled cat (Pardofelis marmorata) could help further explain how the loss of clouded leopards impacts
the interaction between the mesopredator and avifaunal communities. The bay cat displays similar temporal activity to those of birds,
suggesting it may be a terrestrial bird specialist [26]. Novel methodology such as Bayesian co-abundance modeling, combined with
both camera trapping and audio detection equipment, may provide future opportunities to investigate these more cryptic community
interactions, especially if they can be incorporated into community models such as ours [49]. Broadly, alongside the global focus on
invasive domestic cats as a major detriment to native fauna, we emphasize the parallel phenomenon whereby even native meso-
predator species may have broad impacts on ecosystems in the absence of apex predators [50].
The use of forest integrity, as the single broad environmental predictor driving shifts in community dynamics, allowed us to account
for ne-scale continuous change that any number of available categorical variables could not have done. As a holistic measure of
deviation from the natural forest state,forest integrity allowed our model to account for both the observed and inferred ecological
effects of forest loss without overburdening or distorting our results [29]. While our investigation of forest integrity matched closely
our experience in the eld, we are mindful that the dates of our data collection, and the creation of the forest integrity index in 2019,
are not a perfect match in the context of the rapid rate of forest change in the region [4,22]. However, despite its shortcomings, we
believe forest integrity remains a useful proxy for the overall, and possibly indirect, impact of anthropogenic activities and landscape
changes on the ecosystem [51].
In highlighting potential mesopredator release, we acknowledge that the complexity of this interaction, and the underlying me-
chanics, necessitate further research. Furthermore, the use of piecewise SEMs, while powerful for building broad ecosystem-level
snapshots, is still a relatively new method, that also merits further investigation and improvements. Nonetheless, our analysis pro-
vides insight into an ecosystem at a broad scale, highlighting previously unsuspected relationships and hopefully motivating the
deeper exploration required to understand a system of this complexity.
Funding
These analyses are based on camera-trapping surveys principally funded by the Darwin Initiative, Recanati-Kaplan Foundation,
Robertson Foundation, and Sime Darby Foundation, with additional funding from the Clouded Leopard Project, the Felidae Conser-
vation Fund, Houston Zoo, HG Wills International Trust for Nature Conservation, Panthera, the Dr. Holly Reed Conservation Fund of
Point Deance Zoo and Aquarium, and Wild About Cats. Publication was funded by Høgskolen i Innlandet (Inland Norway University
of Applied Science).
Ethics statements
The Economic Planning Unit of Malaysia, Sabah Biodiversity Council, Sabah Parks, Sabah Forestry Department, Sabah Wildlife
Department and Yayasan Sabah reviewed all sampling procedures and approved permits for the work conducted. We applied non-
invasive methods for data gathering and hence approval from an Institutional Animal Care and Use Committee or equivalent ani-
mal ethics committee was not required.
Data statement
The data that has been used is condential. Due to the sensitive nature of the species included in this research and the potential
threat poaching poses to their populations, the raw data for this paper is publicly unavailable. Collaborative inquiries regarding data
access may be possible at the discretion of the WildCRU team.
CRediT authorship contribution statement
Darwin S. Mayhew: Writing review & editing, Writing original draft, Visualization, Software, Methodology, Formal analysis,
Conceptualization. Andrew J. Hearn: Writing review & editing, Project administration, Methodology, Investigation, Funding
acquisition, Data curation. Olivier Devineau: Writing review & editing, Validation, Software. John D.C. Linnell: Writing review
& editing, Supervision. David W. Macdonald: Writing review & editing, Supervision, Project administration, Funding acquisition.
Declaration of competing interest
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
10
inuence the work reported in this paper.The following represent public funding sources from charitable foundations or government
agencies:
Andrew J. Hearn reports nancial support was provided by Darwin Initiative.
David W. Macdonald reports nancial support was provided by Recanati-Kaplan Foundation.
David W. Macdonald reports nancial support was provided by Robertson Foundation.
Andrew J. Hearn reports nancial support was provided by Sime Darby Foundation.
Andrew J. Hearn reports nancial support was provided by Clouded Leopard Project.
Andrew J. Hearn reports was provided by Felidae Conservation Fund.
Andrew J. Hearn reports nancial support was provided by Houston Zoo.
David W. Macdonald reports nancial support was provided by HG Wills International Trust for Nature Conservation.
David W. Macdonald reports nancial support was provided by Panthera Corp.
Andrew J. Hearn reports nancial support was provided by the Dr. Holly Reed Conservation Fund of Point Deance Zoo and
Aquarium.
Andrew J. Hearn reports nancial support was provided by Wild About Cats.
We thank Danum Valley Management Committee, Sabah Parks, Sabah Forestry Department, Sabah Wildlife Department, Yayasan
Sabah, the Economic Planning Unit and the Sabah Biodiversity Centre for permission to conduct research. We thank Sam Cushman,
Carol Sartor, and Morten Odden for insightful comments.
The Economic Planning Unit of Malaysia, Sabah Biodiversity Council, Sabah Parks, Sabah Forestry Department, Sabah Wildlife
Department and Yayasan Sabah reviewed all sampling procedures and approved permits for the work conducted. We applied non-
invasive methods for data gathering and hence approval from an Institutional Animal Care and Use Committee or equivalent ani-
mal ethics committee was not required. If there are other authors, they declare that they have no known competing nancial interests
or personal relationships that could have appeared to inuence the work reported in this paper.
Acknowledgements
Our camera-trap surveys in Sabah were kindly funded by the Darwin Initiative, the Recanati-Kaplan Foundation, the Robertson
Foundation, and the Sime Darby Foundation, with additional funding from the Clouded Leopard Project, the Felidae Conservation
Fund, Houston Zoo, HG Wills International Trust for Nature Conservation, Panthera, the Dr. Holly Reed Conservation Fund of Point
Deance Zoo and Aquarium, and Wild About Cats. We are indebted to our dedicated staff who helped us collect data over many years
and under arduous conditions. We thank Danum Valley Management Committee, Sabah Parks, Sabah Forestry Department, Sabah
Wildlife Department, Yayasan Sabah, the Economic Planning Unit and the Sabah Biodiversity Centre for permission to conduct
research. We thank Sam Cushman, Carol Sartor, and Morten Odden for insightful comments.
Appendix
Table A1
b
Grid Name Number Cameras Collection Dates Number of Sub-grids Number Camera in Sub-grids
Crocker 35 June 6th, 2011 to Feb. 7th, 2012 2 18, 17
Danum Palm 23 March 10th, 2009 to July 8th, 2009 1 23
Danum Valley 79 Feb. 24th
,
2012 to Oct. 6th, 2012 5 16, 17, 15, 15, 16
IJM 33 May 25th, 2011 to Aug. 18th, 2011 2 19, 14*
Sepilok 35 Feb. 9th, 2011 to May 24th, 2011 2 20, 15*
Kinabatangan 66 July 25th, 2010 to Dec. 17th, 2010 4 18, 16, 15, 17
Malua 38 July 11th, 2008 to Feb. 11th, 2009 2 17, 21
Tabin 75 Aug. 19th, 2009 to April 26th, 2010 5 15, 15, 15, 15, 15
Tawau 77 Oct. 31st, 2012 to Jan 18th, 2014 5 15, 15, 16, 15, 16
Sabah Softwoods 15 March 25th, 2021 to July 11th, 2021 1 15*
Ulu Segama 22 May 25th, 2007 to Oct. 17th, 2007 1 22
All 498 May 25th, 2007 to July 11th, 2021 30 Minimum of fteen
*Two traps stolen or broken from this sub-grid.
b
Camera trapping grids were named after their relative areas with varying dates of collection. The number of cameras also varied as available areas to
survey and funding varied. Sub-grids were randomly selected using K-mean clustering with equal sized clusters developed by Bradley et al. [28].
Camera grids were broken into sub-grids with a minimum of at least fteen cameras at each to give most sub-grids somewhat equal camera numbers
relative to the smallest camera grids while optimizing the total number of sub-grids possible from each grid. Four camera locations were removed post
subdivision as they were stolen or broken while in the eld resulting in no data being collected.
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
11
Table A2
c
Camera Grid Sub-grid Sub-
grid
Size
Mean
Forest
Integrity
Mean Forest
Integrity
Standard
Deviation
Sunda
Clouded
Leopard
Sunda
Clouded
Leopard
Standard
Error
Sunda
Leopard
Cat
Sunda
Leopard
Cat
Standard
Error
Great
Argus
Pheasant
Great Argus
Pheasant
Standard
Error
Bulwers
Pheasant
Bulwers
Pheasant
Standard
Error
Crested
Fireback
Crested
Fireback
Standard
Error
Pitta
Family
Pitta
Family
Standard
Error
Crocker Crocker 1 18 916 85 113 37 97 34 193 42 199 46 166 73 84 30
Crocker Crocker 2 17 839 158 61 28 58 29 489 76 35 20 0 0 31 18
Danum Palm Danum Palm 17 596 406 0 0 1775 174 28 14 0 0 3005 538 0 0
Danum Valley Danum Valley
1
16 986 4 204 58 66 30 415 67 302 65 339 105 131 45
Danum Valley Danum Valley
2
17 994 1 87 42 18 19 445 78 234 63 1007 173 79 35
Danum Valley Danum Valley
3
15 979 32 203 91 56 40 1393 162 225 77 846 170 99 48
Danum Valley Danum Valley
4
15 943 42 148 74 174 72 1176 140 232 70 402 83 37 22
Danum Valley Danum Valley
5
16 978 28 237 83 29 28 1159 145 128 55 1130 222 88 40
IJM IJM 1 19 270 173 0 0 648 158 0 0 0 0 855 389 0 0
IJM IJM 2 16 184 222 0 0 1259 305 0 0 0 0 0 0 0 0
Kinabatangan Kinabatangan
1
18 387 226 27 28 53 31 0 0 0 0 552 127 10 10
Kinabatangan Kinabatangan
2
16 574 275 26 26 91 53 13 13 0 0 1189 224 12 12
Kinabatangan Kinabatangan
3
15 404 158 272 120 24 24 0 0 0 0 337 86 26 19
Kinabatangan Kinabatangan
4
17 669 214 66 46 22 22 127 42 0 0 902 197 0 0
Malua Malua 1 17 932 32 27 20 201 48 0 0 0 0 178 76 9 9
Malua Malua 2 21 948 59 58 29 127 31 0 0 0 0 86 37 0 0
Sepilok Sepilok 1 20 664 124 0 0 38 27 254 54 0 0 117 69 29 17
Sepilok Sepilok 2 15 689 137 0 0 113 58 469 90 0 0 0 0 0 0
Tabin Tabin 1 15 927 33 128 53 91 42 595 88 0 0 308 65 11 11
Tabin Tabin 2 15 945 17 106 45 147 48 342 79 0 0 546 114 0 0
Tabin Tabin 3 15 894 48 166 68 236 75 335 71 0 0 504 89 0 0
Tabin Tabin 4 15 742 138 89 46 175 64 265 58 0 0 601 103 11 11
Tabin Tabin 5 15 858 54 122 68 18 18 105 36 0 0 350 69 20 15
Sabah Softwoods Sabah
Softwoods
17 28 69 0 0 664 159 0 0 0 0 0 0 0 0
Tawau Tawau 1 15 754 214 132 37 110 34 892 91 186 40 349 106 51 20
Tawau Tawau 2 15 900 7 184 56 44 22 266 47 429 69 354 118 39 17
Tawau Tawau 3 16 862 28 100 32 59 26 525 68 96 30 19 15 0 0
Tawau Tawau 4 15 891 27 118 50 40 23 236 51 0 0 324 63 33 19
Tawau Tawau 5 16 918 5 168 50 71 25 712 77 445 70 0 0 14 10
Ulu Segama Ulu Segama 17 916 72 362 110 334 63 223 76 0 0 609 100 0 0
c
Modied abundance (±SE) measurements derived from the Royle-Nichols and N-mixture abundance models, multiplied by 100, and rounded to the nearest integer for each focal species in each study
site. Camera grids are listed by alphabetical order in addition to sub-grids and sub-grid sizes, forest loss integrity index is the mean (±SD) of each grid multiplied by 100 and rounded to the nearest integer
on a scale from 0 (Low) to 1000 (High) based on Grantham et al. [29].
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
12
Table A3
d
Species Detection Abundance Number of Parameters AICc ΔAICc AICc Weight
Sunda Clouded Leopard: Royle-Nichols
~ Effort +Forest Road ~ Camera Sub-grid 34 1712.81 0.00 0.78
~1 ~1 2 1883.62 170.82 6.3e-38
Sunda leopard cat: Royle-Nichols
~ Effort +Forest Road +Ridge ~ Camera Sub-grid 34 1838.51 0.00 0.54
~ Effort +Forest Road ~ Camera Sub-grid 33 1838.81 0.29 0.46
~1 ~1 2 2327.40 488.88 3.7e-107
Great Argus Pheasant: N-Mixture
~ Effort +Forest Road ~ Camera Sub-grid 33 6107.90 0.00 0.54
~ Effort +Forest Road +Ridge ~ Camera Sub-grid 34 6108.25 0.34 0.46
~1 ~1 2 7210.91 1103.01 1.7e-240
Bulwers Pheasant: N-Mixture
~ Effort ~ Camera Sub-grid 32 1805.21 0.00 0.58
~ Effort +Ridge ~ Camera Sub-grid 33 1805.82 0.61 0.42
~1 ~1 2 2268.23 463.02 1.6e-101
Crested Fireback: N-Mixture
~ Effort +Ridge ~ Camera Sub-grid 34 3332.29 0.00 0.64
~ Effort +Forest Road +Ridge ~ Camera Sub-grid 35 3333.47 1.18 0.36
~1 ~1 3 3808.46 476.17 2.6e-104
Pitta Family: Royle-Nichols
~ Effort ~ Camera Sub-grid 32 697.86 0.00 0.46
~ Effort +Ridge ~ Camera Sub-grid 33 698.95 1.09 0.27
~1 ~1 2 755.25 57.39 1.6e-13
d
Abundance model selection using both the Royle-Nichols model of abundance for Sunda clouded leopard, Sunda leopard cat, and the Pitta family,
and N-mixture model of abundance for great argus pheasant, Bulwers pheasant, and crested reback. Forest road and ridge, short for ridge line, are
both binary covariates that indicate if a camera was or was not placed on one or both landscape structures. Both model types estimate abundance as
well as detection probability. We present models for each species with ΔAIC less than 2, as well as the null model.
Table A4
e
Dependent Variable Independent Variable Test Type Degrees of Freedom Critical Value P-Value
Sunda Leopard Cat Forest Integrity Coefcient 26 0.7260 0.4743
Great Argus Pheasant I (Sunda Clouded Leopard) Coefcient 26 1.4646 0.1550
Pitta Family I (Sunda Clouded Leopard) Coefcient 26 0.6295 0.5345
Sunda Clouded Leopard Crested Fireback Coefcient 27 1.2745 0.2133
Bulwers Pheasant Crested Fireback Coefcient 24 1.4252 0.1670
Sunda Leopard Cat Crested Fireback Coefcient 26 1.1208 0.2726
Great Argus Pheasant Crested Fireback Coefcient 26 0.2442 0.8090
Pitta Family Crested Fireback Coefcient 26 0.2903 0.7739
Great Argus Pheasant Sunda Clouded Leopard Coefcient 26 1.6125 0.1189
Pitta Family Sunda Clouded Leopard Coefcient 26 1.2284 0.2303
Great Argus Pheasant Bulwers Pheasant Coefcient 24 0.9852 0.3343
Pitta Family Great Argus Pheasant Coefcient 26 1.5520 0.1327
f
List of independence claims as the results from the test of d-separation run during the evaluation of our nial piecewise SEM model. Signicant
relationships would be indicated with an asterix however, no signicant relationships remain after specifying the relationship between the pitta
family and Bulwers pheasant along with Sunda clouded leopards and the squared values of Sunda clouded leopards as correlated errors.
Table A5
f
Response Variable Predictor Variable Model Type
Sunda Clouded Leopard Abundance
(CL Abund)
Mean Sub-grid Forest Integrity GLM (Negative Binomial distribution)
Sunda Leopard Cat Abundance (LC
Abund)
CL Abund +I (CL Abund^2) GLM (Negative Binomial distribution)
Great Argus Pheasant LC Abund +Mean Sub-grid Forest Integrity GLM (Negative Binomial distribution)
Bulwers Pheasant CL Abund +I (CL Abund^2) +LC Abund +Mean Sub-grid
Forest Integrity
GLM (Negative Binomial distribution)
Crested Fireback ~1 N/A
Pitta Family LC Abund +Mean Sub-grid Forest Integrity GLM (Negative Binomial distribution)
Sunda Clouded Leopard/I (Sunda Clouded Leopard) Sunda Clouded Leopard/I (Sunda
Clouded Leopard)
Correlated
Error
Bulwers Pheasant/Pitta Family Bulwers Pheasant/Pitta Family Correlated Error
e
Model structures of pathways that compose the nal piecewise structural equation model. Model types were restricted to general linear models only
as the sample size was too small for mixed effect models to be used. These models represent the most parsimonious models for our data given our
limited exibility.
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
13
References
[1] R.J. Morris, Anthropogenic impacts on tropical forest biodiversity: a network structure and ecosystem functioning perspective, Phil. Trans. Biol. Sci. 365 (1558)
(2010 Nov 27) 37093718.
[2] N. Ocampo-Pe˜
nuela, J. Garcia-Ulloa, I. Kornecki, C.D. Philipson, J. Ghazoul, Impacts of four decades of forest loss on vertebrate functional habitat on Borneo,
Frontiers in Forests and Global Change 3 (2020 May 5) 53.
[3] T.A. Gardner, J. Barlow, R. Chazdon, R.M. Ewers, C.A. Harvey, C.A. Peres, N.S. Sodhi, Prospects for tropical forest biodiversity in a human-modied world, Ecol.
Lett. 12 (6) (2009 Jun) 561582.
[4] C.A. Peres, J. Barlow, W.F. Laurance, Detecting anthropogenic disturbance in tropical forests, Trends Ecol. Evol. 21 (5) (2006 May 1) 227229.
[5] A.D. Barnes, K. Allen, H. Kreft, M.D. Corre, M. Jochum, E. Veldkamp, Y. Clough, R. Daniel, K. Darras, L.H. Denmead, N. Farikhah Haneda, Direct and cascading
impacts of tropical land-use change on multi-trophic biodiversity, Nature ecology & evolution 1 (10) (2017 Oct) 15111519.
[6] A. Ordiz, M. Aronsson, J. Persson, O.G. Støen, J.E. Swenson, J. Kindberg, Effects of human disturbance on terrestrial apex predators, Diversity 13 (2) (2021 Feb
9) 68.
[7] A.C. Stier, J.F. Samhouri, M. Novak, K.N. Marshall, E.J. Ward, R.D. Holt, P.S. Levin, Ecosystem context and historical contingency in apex predator recoveries,
Sci. Adv. 2 (5) (2016 May 27) e1501769.
[8] C. Wolf, W.J. Ripple, Rewilding the worlds large carnivores, R. Soc. Open Sci. 5 (3) (2018 Mar 14) 172235.
[9] G.A. Polis, A.L. Sears, G.R. Huxel, D.R. Strong, J. Maron, When is a trophic cascade a trophic cascade? Trends Ecol. Evol. 15 (11) (2000 Nov 1) 473475.
[10] W.J. Ripple, J.A. Estes, O.J. Schmitz, V. Constant, M.J. Kaylor, A. Lenz, J.L. Motley, K.E. Self, D.S. Taylor, C. Wolf, What is a trophic cascade? Trends Ecol. Evol.
31 (11) (2016 Nov 1) 842849.
[11] E.G. Ritchie, C.N. Johnson, Predator interactions, mesopredator release and biodiversity conservation, Ecol. Lett. 12 (9) (2009 Sep) 982998.
[12] G. Takimoto, S. Nishijima, A simple theory for the mesopredator release effect: when does an apex predator protect their shared prey from a mesopredator?
Oikos 2022 (5) (2022 May) e09021.
[13] K.R. Crooks, M.E. Soul´
e, Mesopredator release and avifaunal extinctions in a fragmented system, Nature 400 (6744) (1999 Aug 5) 563566.
[14] L. Saggiomo, V. Bar, B. Esattore, The fox who cried wolf: a keywords and literature trend analysis on the phenomenon of mesopredator release, Ecol. Complex.
48 (2021 Dec 1) 100963.
[15] J.M. Northrup, J.W. Rivers, Z. Yang, M.G. Betts, Synergistic effects of climate and land-use change inuence broad-scale avian population declines, Global
Change Biol. 25 (5) (2019 May) 15611575.
[16] G. Shahabuddin, R. Goswami, M. Krishnadas, T. Menon, Decline in forest bird species and guilds due to land use change in the Western Himalaya, Global
Ecology and Conservation 25 (2021 Jan 1) e01447.
[17] A.R. Styring, R. Ragai, J. Unggang, R. Stuebing, P.A. Hosner, F.H. Sheldon, Bird community assembly in Bornean industrial tree plantations: effects of forest age
and structure, For. Ecol. Manag. 261 (3) (2011 Feb 1) 531544.
[18] D.S. Wilcove, X. Giam, D.P. Edwards, B. Fisher, L.P. Koh, Navjots nightmare revisited: logging, agriculture, and biodiversity in Southeast Asia, Trends Ecol.
Evol. 28 (9) (2013 Sep 1) 531540.
[19] Ç.H. S
¸ekercio˘
glu, G.C. Daily, P.R. Ehrlich, Ecosystem consequences of bird declines, Proc. Natl. Acad. Sci. USA 101 (52) (2004 Dec 28) 1804218047.
[20] S.R. Loss, T. Will, P.P. Marra, The impact of free-ranging domestic cats on wildlife of the United States, Nat. Commun. 4 (1) (2013 Jan 29) 18.
[21] G.L. Sz´
eles, J.J. Purger, T. Moln´
ar, J. Lanszki, Comparative analysis of the diet of feral and house cats and wildcat in Europe, Mammal Research 63 (2018 Jan)
4353.
[22] S.A. Cushman, E.A. Macdonald, E.L. Landguth, Y. Malhi, D.W. Macdonald, Multiple-scale prediction of forest loss risk across Borneo, Landsc. Ecol. 32 (2017
Aug) 15811598.
[23] D.W. Macdonald, H.M. Bothwell, ˙
Z. Kaszta, E. Ash, G. Bolongon, D. Burnham, ¨
O.E. Can, A. Campos-Arceiz, P. Channa, G.R. Clements, A.J. Hearn, Multi-scale
habitat modelling identies spatial conservation priorities for mainland clouded leopards (Neofelis nebulosa), Divers. Distrib. 25 (10) (2019 Oct) 16391654.
[24] T.G. OBrien, M.F. Kinnaird, A picture is worth a thousand words: the application of camera trapping to the study of birds, Bird. Conserv. Int. 18 (S1) (2008 Sep)
S144S162.
[25] A.J. Hearn, J. Ross, H. Bernard, S.A. Bakar, L.T. Hunter, D.W. Macdonald, The rst estimates of marbled cat Pardofelis marmorata population density from
Bornean primary and selectively logged forest, PLoS One 11 (3) (2016 Mar 23) e0151046.
[26] A.J. Hearn, S.A. Cushman, J. Ross, B. Goossens, L.T. Hunter, D.W. Macdonald, Spatio-temporal ecology of sympatric felids on Borneo. Evidence for resource
partitioning? PLoS One 13 (7) (2018 Jul 20) e0200828.
[27] A.J. Hearn, J. Ross, H. Bernard, S.A. Bakar, B. Goossens, L.T. Hunter, D.W. Macdonald, Responses of Sunda clouded leopard Neofelis diardi population density to
anthropogenic disturbance: rening estimates of its conservation status in Sabah, Oryx 53 (4) (2019 Oct) 643653.
[28] P.S. Bradley, K.P. Bennett, A. Demiriz, Constrained k-means clustering, Microsoft Research, Redmond. 20 (2000 May 8), 0):0.
[29] H.S. Grantham, A. Duncan, T.D. Evans, K.R. Jones, H.L. Beyer, R. Schuster, J. Walston, J.C. Ray, J.G. Robinson, M. Callow, T. Clements, Anthropogenic
modication of forests means only 40% of remaining forests have high ecosystem integrity, Nat. Commun. 11 (1) (2020 Dec 8) 5978.
[30] QGIS Development Team, QGIS Geographic Information System, Open Source Geospatial Foundation Project, 2022, 2022, http://qgis.osgeo.org.
[31] C.X. Cunningham, C.N. Johnson, M.E. Jones, A native apex predator limits an invasive mesopredator and protects native prey: tasmanian devils protecting
bandicoots from cats, Ecol. Lett. 23 (4) (2020 Apr) 711721.
[32] J. Niedballa, R. Sollmann, A. Courtiol, A. Wilting, camtrapR: an R package for efcient camera trap data management, Methods Ecol. Evol. 7 (12) (2016 Dec)
14571462.
[33] R Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2020. URL, https://www.R-
project.org/.
[34] J.A. Royle, J.D. Nichols, Estimating abundance from repeated presenceabsence data or point counts, Ecology 84 (3) (2003 Mar) 777790.
[35] I. Fiske, R. Chandler, Unmarked: an R package for tting hierarchical models of wildlife occurrence and abundance, J. Stat. Software 43 (2011 Aug 24) 123.
[36] J.A. Royle, N-mixture models for estimating population size from spatially replicated counts, Biometrics 60 (1) (2004 Mar) 108115.
[37] Y. Nakashima, Potentiality and limitations of N-mixture and Royle-Nichols models to estimate animal abundance based on noninstantaneous point surveys,
Popul. Ecol. 62 (1) (2020 Jan) 151157.
[38] O.R. Wearn, J.M. Rowcliffe, C. Carbone, H. Bernard, R.M. Ewers, Assessing the status of wild felids in a highly-disturbed commercial forest reserve in Borneo
and the implications for camera trap survey design, PLoS One 8 (11) (2013 Nov 4) e77598.
[39] J.S. Lefcheck, piecewiseSEM: piecewise structural equation modelling in r for ecology, evolution, and systematics, Methods Ecol. Evol. 7 (5) (2016 May)
573579.
[40] C.E. Gordon, D.J. Eldridge, W.J. Ripple, M.S. Crowther, B.D. Moore, M. Letnic, Shrub encroachment is linked to extirpation of an apex predator, J. Anim. Ecol.
86 (1) (2017 Jan) 147157.
[41] D. Berner, V. Amrhein, Why and how we should join the shift from signicance testing to estimation, J. Evol. Biol. 35 (6) (2022 Jun 1) 777787.
[42] B. Shipley, A new inferential test for path models based on directed acyclic graphs, Struct. Equ. Model. 7 (2) (2000 Jun 1) 206218.
[43] B. Shipley, Conrmatory path analysis in a generalized multilevel context, Ecology 90 (2) (2009 Feb) 363368.
[44] M.A. Chua, N. Sivasothi, R. Meier, Population density, spatiotemporal use and diet of the leopard cat (Prionailurus bengalensis) in a human-modied succession
forest landscape of Singapore, Mammal Research 61 (2016 Apr) 99108.
[45] L.G. Shoemaker, B.A. Melbourne, Linking metacommunity paradigms to spatial coexistence mechanisms, Ecology 97 (9) (2016 Sep) 24362446.
[46] W.S. Symes, D.P. Edwards, J. Miettinen, F.E. Rheindt, L.R. Carrasco, Combined impacts of deforestation and wildlife trade on tropical biodiversity are severely
underestimated, Nat. Commun. 9 (1) (2018 Oct 3) 4052.
D.S. Mayhew et al.
Heliyon 10 (2024) e32801
14
[47] S.M. Portalier, G.F. Fussmann, M. Loreau, M. Cherif, The mechanics of predatorprey interactions: rst principles of physics predict predatorprey size ratios,
Funct. Ecol. 33 (2) (2019 Feb) 323334.
[48] P.A. Fleming, H.M. Crawford, C.H. Auckland, M.C. Calver, Body size and bite force of stray and feral catsare bigger or older cats taking the largest or more
difcult-to-handle prey? Animals 10 (4) (2020 Apr 17) 707.
[49] Z. Amir, A. Sovie, M.S. Luskin, Inferring predatorprey interactions from camera traps: a Bayesian co-abundance modeling approach, Ecol. Evol. 12 (12) (2022
Dec) e9627.
[50] J.C. Woinarski, A.M. Stobo-Wilson, H.M. Crawford, S.J. Dawson, C.R. Dickman, T.S. Doherty, P.A. Fleming, S.T. Garnett, M.N. Gentle, S.M. Legge, T.
M. Newsome, Compounding and complementary carnivores: Australian bird species eaten by the introduced European red fox Vulpes vulpes and domestic cat
Felis catus, Bird. Conserv. Int. 32 (3) (2022 Sep) 506522.
[51] O.T. Lewis, R.M. Ewers, M.D. Lowman, Y. Malhi, Conservation of tropical forests: maintaining ecological integrity and resilience, Key Topics in Conservation
Biology 2 (2013 Apr 15) 222235.
D.S. Mayhew et al.
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
Predator–prey dynamics are a fundamental part of ecology, but directly studying interactions has proven difficult. The proliferation of camera trapping has enabled the collection of large datasets on wildlife, but researchers face hurdles inferring interactions from observational data. Recent advances in hierarchical co‐abundance models infer species interactions while accounting for two species' detection probabilities, shared responses to environmental covariates, and propagate uncertainty throughout the entire modeling process. However, current approaches remain unsuitable for interacting species whose natural densities differ by an order of magnitude and have contrasting detection probabilities, such as predator–prey interactions, which introduce zero inflation and overdispersion in count histories. Here, we developed a Bayesian hierarchical N‐mixture co‐abundance model that is suitable for inferring predator–prey interactions. We accounted for excessive zeros in count histories using an informed zero‐inflated Poisson distribution in the abundance formula and accounted for overdispersion in count histories by including a random effect per sampling unit and sampling occasion in the detection probability formula. We demonstrate that models with these modifications outperform alternative approaches, improve model goodness‐of‐fit, and overcome parameter convergence failures. We highlight its utility using 20 camera trapping datasets from 10 tropical forest landscapes in Southeast Asia and estimate four predator–prey relationships between tigers, clouded leopards, and muntjac and sambar deer. Tigers had a negative effect on muntjac abundance, providing support for top‐down regulation, while clouded leopards had a positive effect on muntjac and sambar deer, likely driven by shared responses to unmodelled covariates like hunting. This Bayesian co‐abundance modeling approach to quantify predator–prey relationships is widely applicable across species, ecosystems, and sampling approaches and may be useful in forecasting cascading impacts following widespread predator declines. Taken together, this approach facilitates a nuanced and mechanistic understanding of food‐web ecology. Predator–prey dynamics are a fundamental part of ecology, but studying interactions from observational data (e.g., camera trapping) has proven difficult. Here, we developed a Bayesian hierarchical N‐mixture co‐abundance model that is suitable for inferring predator–prey interactions by accounting for excessive zeros and overdispersion in count histories. Our co‐abundance model detected clear positive and negative predator–prey relationships from a large Southeast Asian camera trapping dataset, and we highlight how it overcomes barriers, is widely applicable across species, ecosystems, and sampling approaches, and may be useful in forecasting cascading impacts following widespread predator declines.
Article
Full-text available
A paradigm shift away from null hypothesis significance testing seems in progress. Based on simulations, we illustrate some of the underlying motivations. First, p-values vary strongly from study to study, hence dichotomous inference using significance thresholds is usually unjustified. Second, 'statistically significant' results have overestimated effect sizes, a bias declining with increasing statistical power. Third, 'statistically non-significant' results have underestimated effect sizes, and this bias gets stronger with higher statistical power. Fourth, the tested statistical hypotheses usually lack biological justification and are often uninformative. Despite these problems, a screen of 48 papers from the 2020 volume of the Journal of Evolutionary Biology exemplifies that significance testing is still used almost universally in evolutionary biology. All screened studies tested default null hypotheses of zero effect with the default significance threshold of p = 0.05, none presented a pre-specified alternative hypothesis, pre-study power calculation and the probability of 'false negatives' (beta error rate). The results sections of the papers presented 49 significance tests on average (median 23, range 0-390). Of 41 studies that contained verbal descriptions of a 'statistically non-significant' result, 26 (63%) falsely claimed the absence of an effect. We conclude that studies in ecology and evolutionary biology are mostly exploratory and descriptive. We should thus shift from claiming to 'test' specific hypotheses statistically to describing and discussing many hypotheses (possible true effect sizes) that are most compatible with our data, given our statistical model. We already have the means for doing so, because we routinely present compatibility ('confidence') intervals covering these hypotheses.
Article
Full-text available
Two introduced carnivores, the European red fox Vulpes vulpes and domestic cat Felis catus , have had extensive impacts on Australian biodiversity. In this study, we collate information on consumption of Australian birds by the fox, paralleling a recent study reporting on birds consumed by cats. We found records of consumption by foxes on 128 native bird species (18% of the non-vagrant bird fauna and 25% of those species within the fox’s range), a smaller tally than for cats (343 species, including 297 within the fox’s Australian range, a subset of that of the cat). Most (81%) bird species eaten by foxes are also eaten by cats, suggesting that predation impacts are compounded. As with consumption by cats, birds that nest or forage on the ground are most likely to be consumed by foxes. However, there is also some partitioning, with records of consumption by foxes but not cats for 25 bird species, indicating that impacts of the two predators may also be complementary. Bird species ≥3.4 kg were more likely to be eaten by foxes, and those <3.4 kg by cats. Our compilation provides an inventory and describes characteristics of Australian bird species known to be consumed by foxes, but we acknowledge that records of predation do not imply population-level impacts. Nonetheless, there is sufficient information from other studies to demonstrate that fox predation has significant impacts on the population viability of some Australian birds, especially larger birds, and those that nest or forage on the ground.
Article
Full-text available
The effects of human disturbance spread over virtually all ecosystems and ecological communities on Earth. In this review, we focus on the effects of human disturbance on terrestrial apex predators. We summarize their ecological role in nature and how they respond to different sources of human disturbance. Apex predators control their prey and smaller predators numerically and via behavioral changes to avoid predation risk, which in turn can affect lower trophic levels. Crucially, reducing population numbers and triggering behavioral responses are also the effects that human disturbance causes to apex predators, which may in turn influence their ecological role. Some populations continue to be at the brink of extinction, but others are partially recovering former ranges, via natural recolonization and through reintroductions. Carnivore recovery is both good news for conservation and a challenge for management, particularly when recovery occurs in human-dominated landscapes. Therefore, we conclude by discussing several management considerations that, adapted to local contexts, may favor the recovery of apex predator populations and their ecological functions in nature.
Article
Full-text available
Land use change is the most widespread driver of biodiversity loss in densely populated tropical countries. Biodiversity loss, in turn, results in changes in functional guilds responsible for various forest ecosystem services. It is thus necessary to understand the extent and types of biodiversity loss and functional guild alteration caused by land use change in order to facilitate sustainable land use policies. Here we study the effects of land use change on forest bird species and guilds in a human-dominated landscape in the Western Himalaya, India. We carried out systematic breeding-season surveys in six land use types within moist temperate forest: natural (protected) oak forest, degraded (lightly used) oak forest, lopped (heavily used) oak forest, pine forest, cultivation and built-up sites, in two adjoining landscapes, over two consecutive years. Our study shows moderate to drastic species loss in all modified land use types in comparison to natural oak forest. Species composition in modified land use types diverged significantly from natural oak; this effect was highest in cultivation and built-up sites and least in degraded forests. Compositional change in modified land uses occurred due to partial replacement of forest specialists with commensals and open country species, whereas abundance of forest generalists was relatively constant across the gradient. We also find a steep decline in the abundance of functional guilds such as pollinators, and insectivorous pest controllers in all modified land uses in comparison to natural oak forest. Our results have important implications for conservation in biodiverse mountain landscapes with significant human imprint. In particular, (a) low faunal diversity in monocultures and urban sites (b) drastic (50% loss or more) loss of forest specialists, pollinators and insectivores in degraded forests, monocultures and urbanised sites; and (c) the potential for degraded forest as refugia for forest species, are findings that can be globally applied to land use and conservation planning in human-dominated landscapes.
Article
Full-text available
Many global environmental agendas, including halting biodiversity loss, reversing land degradation, and limiting climate change, depend upon retaining forests with high ecological integrity, yet the scale and degree of forest modification remain poorly quantified and mapped. By integrating data on observed and inferred human pressures and an index of lost connectivity, we generate a globally consistent, continuous index of forest condition as determined by the degree of anthropogenic modification. Globally, only 17.4 million km² of forest (40.5%) has high landscape-level integrity (mostly found in Canada, Russia, the Amazon, Central Africa, and New Guinea) and only 27% of this area is found in nationally designated protected areas. Of the forest inside protected areas, only 56% has high landscape-level integrity. Ambitious policies that prioritize the retention of forest integrity, especially in the most intact areas, are now urgently needed alongside current efforts aimed at halting deforestation and restoring the integrity of forests globally.
Article
Full-text available
Tropical forests are undergoing drastic transformations, putting at risk the species that rely on them. On the island of Borneo, between 1973 and 2015, 50% of the forest was lost, much of this to oil palm and other industries. We explore the impacts of these four decades of forest loss on the functionally connected habitat of 245 forest birds and mammals. First, we map potential suitable habitat in 1973 and 2015 by refining reported ranges by elevation, forest cover and patch size. We find that, on average, these species have lost 28% of habitat within their ranges. We then use graph-theory connectivity models to calculate the functionally connected area for each species, according to their natal dispersal abilities. We find a mean loss of 35% in functionally connected area, revealing the often hidden impacts of deforestation. Losses in functionally connected habitat are largely driven by area of habitat loss, though maximum elevational range limit also explains some of the differences modeled across species, with lowland species being most affected. We present a vulnerability index of threat arising from loss of functionally connected habitat. The spatial distribution of vulnerability index values serves as a tool for setting conservation priorities for forest remnants on Borneo, given that most of the ranges of these species are not protected. We make recommendations for the use of connectivity models to prioritize resources for research and conservation on Borneo and other biodiversity hotspots.
Article
Full-text available
As carnivorans rely heavily on their head and jaws for prey capture and handling, skull morphology and bite force can therefore reflect their ability to take larger or more difficult-to-handle prey. For 568 feral and stray cats (Felis catus), we recorded their demographics (sex and age), source location (feral or stray) and morphological measures (body mass, body condition); we estimated potential bite force from skull measurements for n = 268 of these cats, and quantified diet composition from stomach contents for n = 358. We compared skull measurements to estimate their bite force and determine how it varied with sex, age, body mass, body condition. Body mass had the strongest influence of bite force. In our sample, males were 36.2% heavier and had 20.0% greater estimated bite force (206.2 ± 44.7 Newtons, n = 168) than females (171.9 ± 29.3 Newtons, n = 120). However, cat age was the strongest predictor of the size of prey that they had taken, with older cats taking larger prey. The predictive power of this relationship was poor though (r2 < 0.038, p < 0.003), because even small cats ate large prey and some of the largest cats ate small prey, such as invertebrates. Cats are opportunistic, generalist carnivores taking a broad range of prey. Their ability to handle larger prey increases as the cats grow, increasing their jaw strength, and improving their hunting skills, but even the smallest cats in our sample had tackled and consumed large and potentially ‘dangerous’ prey that would likely have put up a defence.
Article
Predation plays a variety of important roles in structuring ecological communities. The mesopredator release effect occurs when the removal of an apex predator increases the density of a mesopredator, which in turn reduces the density of their shared prey. The mesopredator release effect can pose significant challenges for predator management and biodiversity conservation. Although several mathematical models have proposed specific circumstances that induce the mesopredator release effect, no theory has yet provided general conditions for this effect. Here, we propose a simple mathematical model to clarify the general conditions that induce the mesopredator release effect. The model predicts that the mesopredator release effect will occur when 1) the carrying capacity of the mesopredator exceeds a certain threshold, and 2) the top–down effect of the apex predator is larger on the mesopredator than on their shared prey. These conditions unify those from previous models and match the existing empirical examples. The simplicity of our theory may be useful for developing system‐specific guidelines to control the mesopredator release effect in various ecosystems.
Article
Human activities severely impact the distribution and behaviour of apex predators in numerous terrestrial and aquatic ecosystems, with cascading effects on several species. Mesopredator outbreaks attributable to the removal of an apex predator have often been recorded and described in the literature as “mesopredator release”. During recent decades several examples of the phenomenon have been observed and studied in many different parts of the world. In this paper, we quantitatively reviewed the existing literature on mesopredator release using two software packages (VOSviewer and CiteSpace) to investigate patterns and trends in author keywords through occurrences and temporal analyses, and creating relative network maps. The results showed that even though the general scientific interest in mesopredator release has increased in recent decades, the vast majority of studies focus on canid species, leaving many other species or entire taxa (e.g., reptiles) understudied and under-described. The connection between invasive species and mesopredator release has only recently been more extensively explored and also the effects of apex predators declining in aquatic ecosystems are still only partially investigated. Due to the increasing effect of biological invasions, overfishing, and either the decline or the rise of apex predators in different parts of the world, we expect an even higher increase in interest and number of published documents on the subject. We also encourage widening the research focus beyond canids to include other important taxa.