Host-, Environment-, or Human-Related Effects Drive Interspecies Interactions in an Animal Tuberculosis Multi-Host Community Depending on the Host and Season

Abstract

In many Mediterranean ecosystems, animal tuberculosis (TB), caused by Mycobacterium bovis, is maintained by multi-host communities in which cattle and different wildlife species establish interaction networks contributing to M. bovis transmission and persistence. Most studies have addressed wildlife–cattle disease-relevant interactions, focusing on reservoir hosts, while disregarding the potential contribution of the so-called accidental hosts and/or neglecting wildlife–wildlife interactions. In this work, we aimed to characterise interspecies interactions in an endemic TB risk area and identify the ecological drivers of interaction patterns regardless of the pre-attributed role of host species on TB epidemiology. For that purpose, spatial–temporal indirect interactions between wildlife mammals and cattle, and between different wildlife species, were investigated through camera trapping. Second, five ecological hypotheses potentially driving species pair interactions in the wet and dry seasons were tested covering water and control sites: human presence (H1), landscape composition (H2), topography (H3), weather (H4), and natural food and water resources (H5). Wild boar (Sus scrofa), red deer (Cervus elaphus), and red fox (Vulpes vulpes) were the wildlife species mostly involved in indirect interactions. We found that indirect wildlife–cattle interactions were more frequent than wildlife interactions and, for certain species pairs, interaction rates were higher in the wet season in both wildlife–cattle and wildlife groups. Natural food and water resources (H5) was the most supported hypothesis that influenced the abundance of wildlife–cattle interactions, with positive effects during the dry season and negative effects during the wet season. In contrast, the abundance of indirect interactions between wildlife species was mainly supported by the human disturbance hypothesis (H1), with negative effects exerted on the dry season and variable effects on the wet season. Other tested hypotheses also influenced wildlife–cattle and wildlife–wildlife interactions, depending on the season and host species. These results highlight that indirect interactions, and thus conditions potentially favouring the transmission of M. bovis in shared environments, are determined by different ecological backgrounds.

Full text

Research Article
Host-, Environment-, or Human-Related Effects Drive
Interspecies Interactions in an Animal Tuberculosis Multi-Host
Community Depending on the Host and Season
Eduardo M. Ferreira ,
1,2,3
Mónica V. Cunha ,
4,5
Elsa L. Duarte ,
1,6
Renata Gonçalves,
3
Tiago Pinto ,
1,2,3
António Mira ,
1,3
and Sara M. Santos
1,2,3
1
MED—Mediterranean Institute for Agriculture,
Environment and Development and CHANGE—Global Change and Sustainability Institute, University of Évora, Mitra,
Évora 7006-554, Portugal
2
IIFA—Institute for Advanced Studies and Research, University of Évora, Vimioso Palace, Évora 7002-554, Portugal
3
Conservation Biology Lab, Department of Biology, University of Évora, Évora, Portugal
4
Centre for Ecology, Evolution and Environmental Changes (cE3c) and CHANGE—Global Change and Sustainability Institute,
Faculty of Sciences, University of Lisbon, Lisbon, Portugal
5
Biosystems and Integrative Sciences Institute (BioISI), Faculty of Sciences, University of Lisbon, Lisbon, Portugal
6
Department of Veterinary Medicine, University of Évora, Mitra, Évora 7006-554, Portugal
Correspondence should be addressed to Eduardo M. Ferreira; ferreiraeduardo.mr@gmail.com
Received 4 January 2024; Revised 4 May 2024; Accepted 20 May 2024
Academic Editor: Andrew Byrne
Copyright ©2024 Eduardo M. Ferreira et al. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
In many Mediterranean ecosystems, animal tuberculosis (TB), caused by Mycobacterium bovis, is maintained by multi-host
communities in which cattle and different wildlife species establish interaction networks contributing to M. bovis transmission
and persistence. Most studies have addressed wildlife–cattle disease-relevant interactions, focusing on reservoir hosts, while
disregarding the potential contribution of the so-called accidental hosts and/or neglecting wildlife–wildlife interactions. In this
work, we aimed to characterise interspecies interactions in an endemic TB risk area and identify the ecological drivers of
interaction patterns regardless of the pre-attributed role of host species on TB epidemiology. For that purpose, spatial–temporal
indirect interactions between wildlife mammals and cattle, and between different wildlife species, were investigated through camera
trapping. Second, five ecological hypotheses potentially driving species pair interactions in the wet and dry seasons were tested
covering water and control sites: human presence (H1), landscape composition (H2), topography (H3), weather (H4), and natural
food and water resources (H5). Wild boar (Sus scrofa), red deer (Cervus elaphus), and red fox (Vulpes vulpes) were the wildlife
species mostly involved in indirect interactions. We found that indirect wildlife–cattle interactions were more frequent than
wildlife interactions and, for certain species pairs, interaction rates were higher in the wet season in both wildlife–cattle and
wildlife groups. Natural food and water resources (H5) was the most supported hypothesis that influenced the abundance of
wildlife–cattle interactions, with positive effects during the dry season and negative effects during the wet season. In contrast, the
abundance of indirect interactions between wildlife species was mainly supported by the human disturbance hypothesis (H1), with
negative effects exerted on the dry season and variable effects on the wet season. Other tested hypotheses also influenced
wildlife–cattle and wildlife–wildlife interactions, depending on the season and host species. These results highlight that indirect
interactions, and thus conditions potentially favouring the transmission of M. bovis in shared environments, are determined by
different ecological backgrounds.
Wiley
Transboundary and Emerging Diseases
Volume 2024, Article ID 9779569, 19 pages
https://doi.org/10.1155/2024/9779569

1. Introduction
Wildlife–livestock interfaces are physical spaces where wild-
life and domestic species can overlap in space and time, along
with humans, and where they can potentially interact [1, 2].
Human activities (e.g., agricultural, husbandry practices,
deforestation and industry) have been causing marked trans-
formations in habitats (e.g., encroachment into natural areas,
habitat fragmentation), shaping these interfaces [3, 4, 5]. With
the loss of natural habitats due to anthropogenic land-use
changes, many wildlife species are forced to live in close prox-
imity to those interfaces. In addition, hunting activities have
been leading to a notable overlap of large game hunting areas
with cattle extensive farming in several regions [6, 7]. Such
changes have profound effects on species interactions and
thereby increase the risk of pathogen transmission and the
(re)emergence of multi-host diseases [4, 8, 9].
Pathogens shared by wildlife and cattle that are of eco-
nomic and public health concern are considered an increas-
ing problem worldwide [10, 11, 12, 13]. In the last decades,
various studies have been addressing wildlife–cattle interac-
tions in the context of multi-host diseases, including animal
tuberculosis (TB), covering different eco-epidemiological sce-
narios [14]. Animal TB is mainly caused by Mycobacterium
bovis and is a globally distributed zoonosis, affecting cattle
and a wide range of wild mammals [15, 16, 17, 18]. The
negative economic impacts of TB on cattle are related to
premature culling of animals, animal trade restrictions, rejec-
tions at slaughterhouses, and costly eradication plans when
mandatory [19, 20]. Wildlife maintenance hosts, which vary
across ecosystems, hamper eradication efforts via pathogen
spilling back to cattle [17, 21, 22]. Transmission may not only
occur when a susceptible host comes into close contact with
an infected host (direct interaction: same location and time)
but also when animals contact asynchronously through con-
taminated environments (indirect interaction: shared space
use in different time frames) [15, 16, 23]. In this sense, defin-
ing these spatial–temporal interactions between mammal
hosts is of major importance for understanding TB transmis-
sion [24, 25, 26]. This has been recognised as a critical step
towards knowing where and when control actions should be
prioritised [27, 28, 29].
Local and global studies have previously shown that direct
interactions between wildlife hosts and cattle are scarce; in
contrast, indirect interactions involving shared environments
occur more frequently [14, 30, 31, 32]. Although explored in
fewer studies, similar trends have been observed between dif-
ferent wildlife species, with indirect interactions being more
frequent [33, 34]. Opportunities for indirect interactions
among wildlife at the wildlife–cattle interface are of particular
concern in systems where M. bovis circulates in multi-host
communities along ecosystem boundaries, potentially favour-
ing pathogen transmission [35, 36]. This is the case in Medi-
terranean ecosystems (Iberian Peninsula), where M. bovis is
able to infect multiple domestic (cattle, pigs, and goats) and
wildlife hosts (ungulates and carnivores) that occur in sym-
patry [37, 38, 39, 40].
In Mediterranean ecosystems, the availability and distribu-
tion of water and food resources are deemed important for
animal aggregation and subsequent interspecies interactions
[6, 41, 42], with summer–autumn periods promoting increased
disease-relevant interactions [28, 41]. Some studies have exam-
ined the effect of host attributes (e.g., animal density; [6]), as
well as of the environment and landscape contexts (e.g., land
cover; [43]) on patterns of interactions between TB hosts at the
wildlife–cattle interface. However, the relative importance of
different ecological factors, and how they contribute to regulate
interspecies interaction patterns in multi-host communities,
remains poorly understood [44]. Moreover, multifaceted
studies that also focus on non-reservoir hosts in the host–
space–time axes and/or beyond the classic wildlife–cattle
binomen are lacking. Considering accidental hosts and their
interactions could help reconstruct missing links in M. bovis
transmission chains, either among wildlife populations or
from wildlife to cattle. Therefore, a community-based perspective
when targeting complex multi-host TB systems is crucial [25, 36]
to identify potential host species and to typify the group of animal
interactions that most likely contribute to TB maintenance within
the community [45].
In Portugal, red deer (Cervus elaphus) and wild boar (Sus
scrofa) have been recognised as the most TB-relevant wildlife
hosts, with reports of environmental contamination of natu-
ral substrates (soil and water bodies) in areas where wildlife
TB is highly prevalent [46, 47, 48, 49, 50]. In this work, we
aimed to increase global understanding of spatial–temporal
indirect interaction patterns within a multi-host mammal
community (cattle and wildlife: red deer, wild boar, red fox
(Vulpes vulpes), and badger (Meles meles)), focusing on a
high prevalence TB area within a Mediterranean agroforestry
system of Southern Portugal.
Specifically, we aimed to:
(i) typify the interaction patterns between cattle and
wildlife, and between wildlife species, and discuss
these patterns in relation to pathogen transmission
risk,
(ii) compare the interaction rates between wildlife–cattle
and between wildlife groups in the dry and wet sea-
sons, and
(iii) evaluate the potential effect of a set of 18 ecological
factors related to human disturbance, landscape com-
position, topography, weather, and natural resources
on both wildlife–cattle and wildlife–wildlife interac-
tions in the dry and wet seasons.
2. Materials and Methods
2.1. Study Area. This study was conducted in Barrancos,
located in Southeast Portugal (Alentejo region), close to
the Spanish border (38˚08′N; 6˚59′W) (Figure 1). This area
is considered a hotspot for TB in cattle and is included in the
official epidemiological TB risk area where special measures
(a mandatory veterinary examination of carcasses to search
for TB-compatible lesions) apply to hunted big game species
2 Transboundary and Emerging Diseases

(red deer and wild boar) [48, 51, 52]. Ungulates are abundant
in the region (wild boar density =3–4 individuals/km
2
; red
deer density =4–8 individuals/km
2
) [53]. Barrancos is an
important Montado region (i.e., woodland, a savannah-like
open tree forest) with extensively cattle breeding in sympatry
with wildlife (e.g., big game). Herd TB prevalence was esti-
mated at 1.83%for the Alentejo region in 2022 [54]. A local
study specifically conducted in Barrancos in 2014–2015
points towards a TB prevalence of 3.1%and 1.8%for red
deer and wild boar, respectively [55]. While official numbers
are remarkably lower, a meta-regression and systematic
review analyses estimated the pooled TB prevalence at a
national scale as 27.5%and 13.3%for the red deer and
wild boar, respectively [56].
The study area (SA) has a Mediterranean climate, with
mild and wet winters and hot and dry summers. Mean
annual temperature ranges from 5 to 14°C during the winter
(January), and from 15 to 34°C during the summer (July)
(Beja; 1981–2010; [57]). During this study period, the mean
temperature in January was 8.9°C and in July was 25.5°C.
The average annual precipitation is 555 mm, concentrated
between October and May. The topography is characterised
by gentle to moderate undulating relief, with altitude ranging
between 160 and 350 ma.s.l. The landscape is dominated by
holm oak (Quercus rotundifolia)Montado, with varying tree
and shrub density (Agro: holm oak stands with low or absent
shrub cover due to grazing and other pastoral activities; For-
est: holm oak stands or mixed woodland patches with high
shrub cover) (Figure 1). Other less representative land cover
types include olive groves and few shrub and agricultural
area mosaics.
2.2. Study Design. We used camera-trapping to assess
spatial–temporal patterns of interactions involving wildlife–cattle
and wildlife–wildlife species over a year (from April 2021 to April
2022). Besides cattle, we used as target species the TB reservoir
hosts described for Portugal (red deer and wild boar [17, 53]),
and two other susceptible species that occur in the region: the red
fox and the badger [38, 58].
We selected five free-ranging adjoining farms with simi-
lar management practices, comprising an area of ~3,048 ha
(farm size ranging from 148 to 980 ha), with an average of
136 adult cows per farm. A 1 km grid was overlaid on the SA
[59, 60]. One camera was installed on each 1 km
2
cell, to assure
spatial independence of sampling sites and land cover repre-
sentativeness. From this grid, we first selected key sites (water
and supplementary food sites; [24])—known as important
aggregation points between species—prioritising sites located
in different grid cells [61], and an even distribution across
farms. The remaining empty cells were defined as control sam-
pling sites, and camera-traps were placed on their centroids. A
total of 38 sampling sites (hereafter called camera sites; Figure 1)
were defined: three food sites for cattle (hay feeders); 16 water
sites (natural water sources and water trough) and 19 control
sites (without any water sources or supplementary food, e.g.,
forest animal path). Minimum distance between camera sites
averaged 686 m (range: 350–1,300 m).
Each camera site consisted of a single camera-trap (Busn-
hell Trophy Cam HD Aggressor or Reconyx Hyperfire)
placed 30–50 cm above the ground, attached to trees or arti-
ficial stakes. At water and food sites, the cameras were facing
towards areas highly used by cattle and wildlife to maximise
the detection of interaction between different species. At
Portugal
Alentejo
Barrancos municipality
0 1 2 km
Camera sites
Control
Food
Water
Land use
Agro
Forest
Other land use types
Study area limits
FIGURE 1: Study area location in Barrancos region, Portugal, showing camera sites and main land uses.
Transboundary and Emerging Diseases 3

control sites, we prioritised animal trails or other areas (e.g.,
resting sites) potentially used by cattle and wildlife in suitable
habitats. No bait of any kind was used. We programmed
cameras to operate 24 hr a day, taking three sequential pic-
tures per trigger with a 30-s delay between consecutive trig-
gers [24, 61]. On average, every 10–15 days, we visited camera
sites for battery and memory card replacement.
2.3. Data Coding and Processing. Pictures recorded by each
camera were individually classified by visual observation. The
following information was recorded in an Excel database:
camera coordinates, camera site type(water, food, or control),
target species (cattle, red deer, wild boar, red fox, and badger),
and number of individuals (minimum number of individuals
recorded in each picture). In addition, date–time of picture
capture were retrieved using the open-source software Exif-
Tool [62]. An independent observation of the same species
(hereafter called “detections”) was considered at a given cam-
era site when pictures were taken at least 15min apart
[6, 24, 63].
If cattle were unable to reach a given camera site in a
certain period (due to cattle grazing rotation and manage-
ment), that period from that camera was excluded from
analyses. We assumed that the fences were permeable to
wildlife [6], as confirmed in the field and several times in
the camera pictures. The three camera sites initially classified
as food sites had no food for long periods of time, and thus
were excluded from further analyses.
2.4. Definition and Estimation of Interactions. An indirect
interaction was defined as the detection of one species at a
given camera site, following the detection of another species
within a pre-established critical time window, CTW, related
to estimated M. bovis’s environmental survival time. A CTW
of 3 days for the dry season (June–September) and of 12 days
for the wet season (October–May) was assumed, following
the procedures of Kukielka et al. [24] and Cowie et al. [33],
applied in a similar eco-environmental context (Figures S1
and S2). A direct interaction was defined whenever indivi-
duals of different species were captured in the same picture
[14], although it was not analysed in this study due to the
much lower number of observations recorded.
The number of indirect interactions was calculated for
each camera site and month, discriminated by species pairs.
The species pairs considered in this study are composed of
the combinations of the five target species and are divided
into two groups: the wildlife–cattle group includes four spe-
cies pairs: BT_CE (cattle—red deer), BT_SS (cattle—wild boar),
BT_MM (cattle—badger), and BT_VV (cattle—red fox); and
the wildlife group includes six species pairs: CE_MM (red deer
—badger), CE_SS (red deer—wild boar), CE_VV (red deer—
red fox), SS_MM (wild boar—badger), VV_MM (red fox—bad-
ger), and VV_SS (red fox—wild boar). For each species pair and
camera site, we calculated monthly rates of indirect interactions
(RatesInt) as a function of the number of interactions (nr of
interactions) per time (RatesInt =nr of interactions/time),
adapted from Ferreira et al. [14] study. Time was expressed as
a proportion, corresponding to active camera days (days
when cameras were operational and recording without any
interference) divided by the number of days in a given month.
We summarised RatesInt by species pairs and seasons (indirect
interactions/month/camera), computing RatesInt means along
with the corresponding standard errors. Generalised linear mod-
els (GLM) were used to inspect potential differences in RatesInt
between wildlife–cattle and wildlife groups, across seasons.
2.5. Human, Landscape, and Environment Predictors. To
address the third objective, we defined a total of 18 eco-
environmental predictors that influence the abundance of
the target species and thus may influence species interaction
patterns. These predictors were arranged according to five
ecological hypotheses that might regulate species interac-
tions: (H1) human disturbance (n=4 predictor variables);
(H2) landscape composition (n=5); (H3) topography (n=3);
(H4) weather (n=3); and (H5) natural food and water
resources (n=3) (Table 1).
We estimated human disturbance (H1) for each camera
site through the total number of days with human records
(visually extracted from pictures); and through the Euclidean
distance of camera sites to the nearest houses, to hunting
sites (stand sites for hunting, where baiting is placed nearby
for attracting wildlife) and road density metrics of unpaved
roads (length of roads/total area within a given neighbour-
hood) in the SA (Quantum GIS v. 3.0.3; [69]). For landscape
composition-related predictors (H2), we computed the pro-
portion of land cover, considering the main land uses (agro-
forest and forest) occurring in the SA; the Shannon landscape
diversity index, and the Euclidean distance of camera sites to
forest edges. Those metrics were obtained from the Corine
Land Cover (2018) dataset (European Union, Copernicus
Land Monitoring Service, European Environment Agency)
and were retrieved from the “landscapemetrics”Rpackage
[70]. In addition, tree cover density was derived from the
Tree Cover Density (2018) dataset (Copernicus Land Monitor-
ing Service, European Environment Agency) (Table 1).
Regarding topographic predictors (H3), we estimated
elevation from a 30-m digital elevation model (DEM) and
derived terrain ruggedness index and slope from the DEM
using Quantum GIS v. 3.0.3. Weather-based predictors (H4;
i.e., Rain and Temp) were obtained from data collected at a
local weather station. Lastly, for H5, the water content
(Water_cont) was visually estimated based on the area cov-
ered by standing water (using some marks in situ to retrieve
estimates) during field work visits throughout the sampling
period. The typology of each camera site—Station_site (con-
trol or water)—was used as a categorical variable. The nor-
malised difference vegetation index (NDVI) was derived
from the LANDSAT 8 image collection (level 2, Tier 1),
with a 30 m spatial resolution, and processed in Google Earth
Engine [71]. The NDVI has shown a high correlation with
vegetation biomass and dynamics in various ecosystems
worldwide. Several authors have used NDVI to assess vege-
tation productivity—representing resource quantity and
quality—and the dynamics of habitat use by wild mammals,
including ungulates and carnivores [72, 73, 74]. For this
reason, we used NDVI as a proxy for natural food availabil-
ity. We only retained high-quality images with ≤5%of cloud
4 Transboundary and Emerging Diseases

TABLE 1: Study hypotheses and description of the eco-environmental predictors used for modelling interspecies interactions.
Hypothesis Inclusion rationale Prediction Predictor acronym Description
H1. Human disturbance
Wildlife species tend to show a
spatial–temporal avoidance of humans and to
humans-related activities, which in turn may
influence patterns of interspecies interactions
[64, 65, 66]
We expect a negative association between
human disturbance and abundance of
interspecies interactions. We also expect that a
greater human presence may also imply a
greater presence of domestic species which
increases the likelihood of wildlife–cattle
interactions
Dist_houses Distance of camera sites to the nearest artificial
houses/facilities (m)
Dist_hunt Distance of camera sites to the nearest hunting
site (m)
DensRoad Density of unpaved roads within 100, 250, and
500 m spatial scales around camera sites
Human_days Number of days with occurrence of humans (a
proxy for human presence)
H2. Landscape composition
The occurrence and distribution of species
depend on their habitat requirements, and thus
landscape context may be a key driver for
interspecies interactions [6, 56, 65]
We predict that landscape composition is the
most important mechanism driving
interspecies interactions. We expect a positive
relationship between forest and heterogeneous
areas and wildlife interactions; and a positive
relationship between agro-dominated areas and
wildlife–cattle interactions
Agro
Percentage of agroforest land (holm oak stands
with low or absent shrub cover due to grazing
and other pastoral activities) within 100, 250,
and 500 m spatial scalesaround camera sites (%)
Forest
Percentage of forest (holm oak stands or mixed
woodland patches with high shrub cover)
within 100, 250, and 500m spatial scales
around camera sites (%)
TreeD
Proportion of tree cover density within 100,
250, and 500m spatial scales around camera
sites (%)
Dist_edgeF Distance of camera sites to the nearest edge of
forest patches (m)
Shidi
Shannon’s landscape diversity index within
100, 250, and 500 m spatial scales around
camera sites
H3. Topography
Terrain features are important drivers that
regulate species co-occurrence and thus
influence shared space among host species
[64, 67]
We expect a negative relationship between
topography-based predictors and species
interactions
Altitude Terrain altitude within 100, 250, and 500 m
spatial scales around camera sites
Rugg Terrain ruggedness index within 100, 250, and
500 m spatial scales around camera sites
Slope Topographic slope within 100, 250, and 500 m
spatial scales around camera sites
H4. Weather
Weather conditions shape species activity and,
in turn, can drive interactions among hosts
across space and time gradients [24, 60, 68]
We predict that weather conditions exert
positive or negative effects on interspecies
interactions, being species-specificand season
dependent
Temp Minimum monthly temperature (°C), used in
the wet season
Temp Maximum monthly temperature (°C), used in
the dry season
Rain Total monthly accumulated precipitation (mm)
H5. Natural food and water resources
Food and water resources can facilitate species
aggregation, thus being an important factor
shaping spatial and temporal patterns of
interactions between mammal host species
[24, 28]
We predict that food-rich areas, along with
water abundance, have a positive influence on
interspecies interactions, particularly during
the dry season
Water_cont
Water content at each camera site (mean
monthly water area size; m
2
), calculated by
visual estimation in the field
Station_site Typology of the camera sites: control sites and
water sites
NDVI
Normalised Difference Vegetation Index within
100, 250, and 500 m spatial scales around
camera sites
Transboundary and Emerging Diseases 5

cover considering the whole SA (more details are available in
[75]). For the missing data in our time series (a 3-month gap,
non-consecutive months), we used images from the month
before and after (time interpolation; [76]) to estimate the
NDVI values [77].
A multi-scale approach was carried out to cover a wide
range of scales and thus maximise potential responses with
the target species [78]. Continuous predictors not based on
distances (Dens_roads, TreeD, Altitude, Rugg, Slope, and
NDVI) were stacked in a 30 m spatial resolution multi-raster
layer. We then applied the following spatial scales of analysis:
90, 240, and 510 m focal-radius moving window as a proxy
for 100, 250, and 500 m neighbourhood scales of analysis
around camera sites. Mean was used to summarise the raster
values within each spatial scale A similar procedure (in terms
of scales) was applied to Agro, Forest, and Shidi using a
spatial resolution basis of 10 m, and thus a focal-radius mov-
ing window of 100, 250, and 500 m (Table 1).
We also estimated the relative abundance index of each
target species (e.g., RAI), discriminated by camera site and
season, to be used as a proxy of animal density in the model-
ling process. Animal abundance was calculated as the num-
ber of detections of each species in a month/(number of
active camera days/number of days of a given month).
2.6. Modelling: Hypotheses Explaining Interspecies Indirect
Interactions. Interaction analyses were conducted separately
for each species pair, and for the dry and wet seasons, allow-
ing the identification of potential differences in the effects of
predictors driving interactions between seasons. As pre-
modelling procedures, we checked for outliers and inspected
collinearity among variable predictors. Pairwise Spearman
correlations were calculated among all predictors to check
for multicollinearity. Numeric predictors with skewed distri-
butions were transformed (square-root, logarithmic, or arc-
sine) to approach normality and to reduce the influence of
extreme values [79]. In addition, all continuous predictors
were standardised, allowing comparisons of their strength in
the modelling process.
We fitted the response variable—number of species
interactions—to generalised linear mixed models (GLMM)
with a Poisson or negative binomial family distribution and
log link (package “glmmTMB”[80]), using camera site as a
random factor because each camera site was sampled repeat-
edly through time. The log of the number of active camera-days
was used as offset in the models to integrate sampling effort
between camera sites over time [24]. This procedure avoided
transforming count data (log-transformed data or RatesInt), as
recommended by Zuur et al. [79] and O’Hara and Kotze [81].
The five ecological hypotheses (H1–H5) were indepen-
dently evaluated [82], first through simple models, testing
one predictor at a time. These simple models always included
the abundances of each species (RAI) involved in a given spe-
cies pair interaction as fixed predictors, since higher host abun-
dance increases interaction levels [14]. Then, if more than one
predictor was informative within a hypothesis, a multivariate
model was built for each hypothesis with all informative
predictors.
Model example: species pair AB|season
Number of interactions~animal abundance (A)+animal
abundance (B)+predictor X+random (1|camera site), offset
(log (camera days)), family (Poisson/negative binomial).
A predictor variable was considered informative when:
(1) the 95%confidence intervals (CI: 95%)of the variable
coefficient being tested did not include zero; and (2) a
deltaAICc >2(ΔAICc; Akaike’s information criterion adjusted
for small sample sizes) was obtained when comparing the
tested model with the reference model (without the specific
predictor; [79, 83, 84]). If highly correlated informative pre-
dictors (r>0.7) were identified, we only retained the one
producing a lower AICc to be included in the multivariate
model. This procedure also involved comparing multiple
scales for a given predictor. Multivariate models were built
with all possible combinations of the informative predictors of
each hypothesis, always keeping animal abundance (RAI) in
all competing models, and limiting each model to a maximum
of four predictors to avoid model instability. We selected the
best multivariate model for each hypothesis using AICc. Mod-
els having a ΔAICc <2 are considered equally supported.
When several models had ΔAICc <2 : 1) all associated pre-
dictors were included in a single best multi-model [85] if≤
four predictors were selected; (2) all models within ΔAICc <2
of the top-ranked models were retained for interpretation,
otherwise.
The dredge function (R package “MuMIn”[86]) was used
for model selection. Once we identified all the best models for
the hypotheses tested (H1–H5), we again ran the models with
a restricted maximum likelihood (REML). Since it is impor-
tant to assess model adequacy [87, 88], models were evaluated
and validated using diagnostic tools (normality, outliers, and
zero inflation) available in the “DHARMa”package [89].
3. Results
We obtained a total of 15,537 detections of cattle and target
wild mammal species over 6,170 effective trap days across the
35 camera sites (mean =176 Æ61 sd trap days per camera
site) during the study period. Cattle were the most frequently
detected species (66.8%;n=10,379). Red fox (10.5%;n=
1,631), red deer (8.6%;n=1,335), and wild boar (7.3%;n=
1,141) were detected in similar numbers and were widespread
in the SA (detection in >85%of camera sites). The badger
occurred at lower rates (2.5%;n=382), although it was also
widespread in the SA (detection in >75%of camera sites).
3.1. Wildlife–Cattle and Wildlife Species Interactions. Wild-
life–cattle indirect interactions represented 52.7%(n=3,619)
of the interaction data (only 0.1%(n=7) were direct interac-
tions involving cattle). The wildlife species that were most
frequently involved in these interactions were the red fox
(BT_VV; mean RatesInt: wet season =6.1 and mean RatesInt:
dry season =4.5), followed by the wild boar (BT_SS; mean
RatesInt: wet season =4.8 and mean RatesInt: dry season =
2.8) and red deer (BT_CE; mean RatesInt: wet season =4.5
and mean RatesInt: dry season =2.5). The badger (BT_MM;
mean RatesInt: wet season =1.6 and mean RatesInt: dry
6 Transboundary and Emerging Diseases

season =1.5) interacted less frequently with cattle (Figure 2(a)).
Interactions with cattle involving the three most detected spe-
cies (red fox, wild boar, and red deer) occurred in all farms, at
more than 80%of camera sites during the wet season, and at
30%–60%of camera sites in the dry season. Interaction rates
were significantly higher in the wet season for the pairs
BT_VV (GLM; coef: wet season =0.361, CI: 95%(0.050;
0.672)), BT_SS (GLM; coef: wet season =0.304, CI: 95%
(0.024; 0.585)), and BT_CE (GLM; coef: wet season =0.441,
CI: 95%(0.167; 0.714)).
Indirect interactions between wildlife represented 46.8%
(n=3,210) of the interaction data (only 0.4%(n=25) were
direct interactions). The wildlife species pairs most frequently
interacting were CE_SS (mean RatesInt: wet season =3.6 and
mean RatesInt: dry season =2.2), CE_VV (mean RatesInt: wet
season =3.3 and mean RatesInt: dry season =2.7), and
VV_SS (mean RatesInt: wet season =3.4 and mean RatesInt:
dry season =2.3) (Figure 2(b)). Indirect interactions between
the three main species (red fox, wild boar, and red deer)
occurred at more than 80%of camera sites during the wet
season, and at 40%–60%of camera sites during the dry season.
Interaction rates were significantly higher in the wet season
for the pairs CE_SS (GLM; coef: wet season =0.283, CI: 95%
(0.031; 0.535)), CE_VV (GLM; coef: wet season =0.302, CI:
95%(0.045; 0.559)), and VV_SS (GLM; coef: wet season =
0.297, CI: 95%(0.038; 0.556)).
3.2. RatesInt between Wildlife–Cattle and Wildlife Groups.
Globally, interaction rates (RatesInt)were higher in the wet
season for both wildlife–cattle and wildlife groups when
compared to the dry season. The mean interaction rates of
the wildlife–cattle group were 1.8 and 1.6 times significantly
higher than the wildlife rates for the dry and wet seasons,
respectively (GLM dry season; coef wildlife: −0.156, CI: 95%
(−0.285; −0.0269); GLM wet season; coef wildlife: −0.269,
CI: 95%(−0.354; −0.184)).
3.3. Ecological Hypotheses Driving Species Interactions. All
models were fitted with a Poisson family distribution. The
predictors Slope, Rugg, Agro, and Forest were not used simul-
taneously in the same model due to multicollinearity pro-
blems. Locations with high terrain ruggedness had also
higher slope (rs =0.99) and low percentage of Agro (rs =
−0.73). On the other hand, locations with high percentage
of Agro had low percentage of Forest (rs =−0.74). Model
residual patterns revealed a good to adequate fit of most of
the models to the data (Figures S3, S4, S5, and S6: DHARMa
diagnostic plots showing residual, dispersion, and zero-
inflation fits of the tested models). Four of the five
ecological hypotheses tested were significantly associated
with abundance of wildlife–cattle interactions, covering one
to three species pairs, depending on the hypothesis (Table 2
and Figure 3(a)). Three of the five ecological hypotheses tested
0
BT_CE BT_MM BT_SS BT_VV
1
2
3
4
5
6
7
RatesInt (indirect interactions/month)
Widlife–cattle
Dry season
Wet season
ðaÞ
0
1
2
3
4
5
6
7
RatesInt (indirect interactions/month)
Wildlife
CE_MM CE_SS CE_VV SS_MM VV_MM VV_SS
Dry season
Wet season
ðbÞ
FIGURE 2: Weighted means and standard errors of RatesInt (indirect interactions/month) summarised by species pairs and seasons and
displayed by animal group ((a) wildlife–cattle; (b) wildlife). Species pair acronyms are (BT_CE) cattle—red deer; (BT_MM) cattle—badger;
(BT_SS) cattle—wild boar; (BT_VV) cattle—red fox; (CE_MM) red deer—badger; (CE_SS) red deer—wild boar; (CE_VV) red deer—red
fox; (SS_MM) wild boar—badger; (VV_MM) red fox—badger; and (VV_SS) red fox–wild boar.
Transboundary and Emerging Diseases 7

TABLE 2: Summary of the hypotheses (H) tested and predictors (highlighted in bold) significantly related to wildlife–cattle species pair interactions.
Species pair Season H Model id Null model AICc Model ref AICc Model AICc DeltaAIC Predictors Coeff. CI: 95%IRR
BT_CE Wet H2 BTCE_wH2 1,110.9 601.6 598.1 3.5
BT abundance 0.945 (0.792; 1.098) 2.573
CE abundance 1.021 (0.901; 1.141) 2.776
Forest_100 0.175 (0.023; 0.326) 1.191
BT_CE Wet H4 BTCE_wH4 1,110.9 601.6 597.0 4.6
BT abundance 0.871 (0.726; 1.015) 2.388
CE abundance 1.065 (0.944; 1.186) 2.902
Temp −0.088 (−0.155; −0.020) 0.916
BT_CE Wet H5 BTCE_wH5 1,110.9 601.6 594.7 6.9
BT abundance 0.937 (0.794; 1.080) 2.552
CE abundance 1.090 (0.973; 1.207) 2.974
Station_site: water −0.361 (−0.595; −0.127) 0.697
BT_CE Dry H2 BTCE_dH2 248.7 165.5 161.6 3.9
BT abundance 0.626 (0.246; 1.005) 1.870
CE abundance 1.363 (0.949; 1.778) 3.908
TreeD_100 −0.868 (−1.659; −0.077) 0.420
BT_CE Dry H4 BTCE_dH4 248.7 165.5 160.9 4.6
BT abundance 1.071 (0.561; 1.580) 2.917
CE abundance 1.362 (0.936; 1.788) 3.902
Rain 0.314 (0.061; 0.568) 1.369
BT_CE Dry H5 BTCE_dH5 248.7 165.5 151.3 14.2
BT abundance 0.855 (0.402; 1.309) 2.352
CE abundance 1.248 (0.852; 1.643) 3.483
NDVI_500 0.198 (−0.074; 0.471) 1.219
Water_cont 0.878 (0.367; 1.388) 2.405
BT_SS Wet H1 BTSS_wH1 1,042.8 629.2 624.6 4.6
BT abundance 0.743 (0.611; 0.875) 2.103
SS abundance 1.013 (0.904; 1.123) 2.754
Human_days −0.180 (−0.321; −0.039) 0.836
BT_SS Wet H2 BTSS_wH2 1,042.8 629.2 624.7 4.5
BT abundance 0.796 (0.662; 0.929) 2.216
SS abundance 0.998 (0.896; 1.100) 2.713
Agro_100 −0.141 (−0.247; −0.035) 0.868
BT_SS Wet H5 BTSS_wH5 1,042.8 629.2 625.7 3.5
BT abundance 0.707 (0.580; 0.833) 2.027
SS abundance 1.052 (0.945; 1.159) 2.863
NDVI_100 −0.088 (−0.162; −0.014) 0.915
BT_SS∗Dry —— — — — — — — — —
BT_VV Wet H1 BTVV_wH1 1,386.5 688.3 683.4 4.9
BT abundance 0.985 (0.841; 1.129) 2.678
VV abundance 1.013 (0.908; 1.119) 2.755
Human_days −0.110 (−0.193; −0.026) 0.896
BT_VV Wet H5 BTVV_wH5 1,386.5 688.3 681.9 6.4
BT abundance 0.983 (0.841; 1.125) 2.671
VV abundance 0.999 (0.897; 1.102) 2.717
NDVI_500 −0.073 (−0.137; −0.008) 0.930
Station_site: water −0.308 (−0.571; −0.045) 0.735
BT_VV Dry H1 BTVV_dH1 287.8 180.3 175.9 4.4
BT abundance 0.752 (0.490; 1.013) 2.120
VV abundance 1.551 (1.215; 1.887) 4.715
DensRoad_250 −0.384 (−0.682; −0.086) 0.681
BT_VV Dry H4 BTVV_dH4 287.8 180.3 171.8 8.5
BT abundance 1.026 (0.672; 1.380) 2.789
VV abundance 1.550 (1.179; 1.921) 4.711
Rain 0.296 (0.111; 0.481) 1.344
8 Transboundary and Emerging Diseases

TABLE 2: Continued.
Species pair Season H Model id Null model AICc Model ref AICc Model AICc DeltaAIC Predictors Coeff. CI: 95%IRR
BT_VV Dry H5 BTVV_dH5 287.8 180.3 173.1 7.2
BT abundance 0.966 (0.643; 1.289) 2.627
VV abundance 1.651 (1.262; 2.040) 5.211
NDVI_100 0.360 (0.124; 0.596) 1.433
BT_MM Wet H2 BTMM_wH2 651.1 371.3 368.5 2.8
BT abundance 0.920 (0.703; 1.137) 2.508
MM abundance 1.174 (1.037; 1.312) 3.236
Agro_100 −0.206 (−0.391; −0.020) 0.814
BT_MM∗Dry —— — — — — — — — —
For each species pair and season, we provided the best model according to the model’s AICc (Akaike’s information criterion adjusted for small sample sizes). The AICc of the reference model and the null model are
also provided. DeltaAICc (ΔAICc) was obtained between the reference model and each best model for a given hypothesis. The coefficients (Coeff.) and corresponding 95%confidence intervals (CI: 95%) for each
tested predictor are presented. Incidence rate ratios (IRR) are reported as exponentiated results. ∗(asterisk) was used to mark species pairs and seasons for which we did not find asignificant association with the
tested hypotheses.
Transboundary and Emerging Diseases 9

were significantly associated with the abundance of wildlife
interactions, covering from one to four species pairs, depending
on the hypothesis (Table 3 and Figure 3(c)). Wildlife–cattle
interactions were most related to natural food and water
resources hypothesis (H5) (Figure 3(b)), while wildlife
interactions were often associated with human disturbance
hypothesis (H1) (Figure 3(d)).
3.3.1. Modelling: Wildlife–Cattle Interactions. The number of
wildlife–cattle interactions, involving the red fox and wild
boar, increased in areas with a lower human presence during
the wet season (H1, models: BTVV_wH1 and BTSS_wH1;
Table 2). Additionally, in this season, interactions encom-
passing the red deer, wild boar, and badger increased in more
forested areas (e.g., areas with low Agro cover; H2, models:
BTCE_wH2, BTSS_wH2, and BTMM_wH2). More interac-
tions between cattle and red deer were associated with low-
temperature periods (H4, model: BTCE_wH4). The higher
abundance of interactions, covering red deer, red fox, and
wild boar, occurred in areas where natural resources are less
abundant (i.e., control sites and less productive areas (NDVI))
(H5, models: BTCE_wH5, BTVV_wH5 and BTSS_wH5). Dur-
ing the dry season, wildlife–cattle interactions increased in areas
with lower road densities, as evidenced for the red fox (H1,
model: BTVV_dH1), and in areas with lower tree cover, in the
case of the red deer (H4, model: BTCE_dH2). Rain had a positive
influence on the abundance of wildlife–cattle interactions (H4,
models: BTCE_dH4 and BTVV_dH4), and interactions were
more frequent in sites with higher water content and in more
productive areas, for carnivores and ungulates, such as the red
fox and the red deer, respectively (H5, models: BTVV_dH5 and
BTCE_dH5). Overall, animal abundance had a strong effect size
in all models: with one-point increase in animal abundance
(wildlife or cattle), number of interactions would be expected
Negative
Positive
0
1
2
3
4
0
1
2
3
4
Number of species pairs
Dry
Wet
Wildlife–cattle
H1 H2 H3 H4 H5
ðaÞ
Human_days
2
DensRoad
1
Agro
2Forest
1
TreeD
1
Rain
2
Temp
1
NDVI
3Station_site
2
Water_cont
1
H1
H2 H4
H5
(–)
(–)
(–)
(–)
(–)
(–)
(+)
(+)
(+)
(±)
ðbÞ
Negative
Positive
0
1
2
3
4
0
H1 H2 H3 H4 H5
1
2
3
4
Number of species pairs
Dry
Wet
Wildlife
ðcÞ
Dist_houses
2
DensRoad
2
Human_
days
1
Shidi
1
Rain
1
Temp
1
H1 H2
H4
(–) (–)
(±)
(+)
(–)
(+)
ðdÞ
FIGURE 3: Number of species pairs influenced by ecological hypotheses regarding indirect interactions, displayed by wildlife–cattle (a) and
wildlife (c) groups and considering sampled seasons. For each hypothesis, the sign of the coefficient effect is shown (positive, negative, or
null). Treemaps show the number of times the tested predictors, underlying ecological hypotheses, were associated with species pair
interactions, displayed by wildlife–cattle (b) and wildlife (d) groups.
10 Transboundary and Emerging Diseases

TABLE 3: Summary of the hypotheses (H) tested and predictors (highlighted in bold) significantly related to wildlife species pair interactions.
Species pair Season H Model identifier Null model AICc Model ref AICc Model AICc DeltaAIC Predictors Coeff. CI: 95%IRR
CE_SS Wet H1 CESS_wH1 928 539.7 537.8 2.0
CE abundance 0.740 (0.632; 0.847) 2.095
SS abundance 0.788 (0.672; 0.903) 2.198
Dist_houses 0.083 (0.001; 0.164) 1.086
CE_SS Wet H4 CESS_wH4 928 539.7 537.1 2.6
CE abundance 0.747 (0.638; 0.856) 2.111
SS abundance 0.752 (0.635; 0.870) 2.122
Temp −0.082 (−0.157; −0.007) 0.921
CE_SS∗Dry —— — — — — — — — —
CE_VV Wet H1 CEVV_wH1 906.1 529.9 527.4 2.5
CE abundance 0.933 (0.832; 1.033) 2.542
VV abundance 0.892 (0.784; 1.000) 2.440
DensRoad_100 −0.116 (−0.224; −0.008) 0.891
CE_VV∗Dry —— — — — — — — — —
CE_MM∗Wet —— — — — — — — — —
CE_MM Dry H4 CEMM_dH4 205.4 90.4 83.7 6.7
CE abundance 0.757 (0.295; 1.218) 2.131
MM abundance 1.043 (0.736; 1.350) 2.838
Rain 0.455 (0.118; 0.791) 1.575
VV_SS Wet H1 VVSS_wH1 859.4 546.1 543.8 2.3
VV abundance 0.837 (0.708; 0.966) 2.310
SS abundance 0.920 (0.788; 1.051) 2.509
Human_days −0.161 (−0.318; −0.004) 0.851
VV_SS Wet H2 VVSS_wH2 859.4 546.1 538.7 7.4
VV abundance 0.855 (0.737; 0.972) 2.350
SS abundance 0.929 (0.807; 1.051) 2.532
Agro_100 −0.091 (−0.191; 0.008) 0.913
Shidi_100 0.108 (0.024; 0.192) 1.114
VV_SS Dry H1 VVSS_dH1 251.2 136.3 132.3 4.0
VV abundance 1.220 (0.838; 1.602) 3.387
SS abundance 1.054 (0.762; 1.345) 2.869
Dist_houses −0.376 (−0.708; −0.043) 0.687
SS_MM∗Wet —— — — — — — — — —
SS_MM Dry H1 SSMM_dH1 173.4 94.3 89.2 5.1
SS abundance 1.111 (0.676; 1.545) 3.036
MM abundance 1.065 (0.811; 1.319) 2.900
DensRoad_100 −0.635 (−1.122; −0.147) 0.530
VV_MM∗Wet —— — — — — — — — —
VV_MM∗Dry —— — — — — — — — —
For each species pair and season, we provided the best model according to the model’s AICc (Akaike’s information criterion adjusted for small sample sizes). The AICc of the reference model and the null model are
also provided. DeltaAICc (ΔAICc) was obtained between the reference model and each best model for a given hypothesis. The coefficients (Coeff.) and corresponding 95%confidence intervals (CI: 95%) for each
tested predictor are presented. Incidence rate ratios (IRR) are reported as exponentiated results. ∗(asterisk) was used to mark species pairs and seasons for which we did not find a significant association with the
tested hypotheses.
Transboundary and Emerging Diseases 11

to increase by an average IRR of 2.93 (sd =0.58), holding all
variables constant. Ecological predictors, linked to the study
hypotheses, had a lesser pronounced effect (positive predictors:
average IRR =1.49, sd =0.46; negative predictors: average IRR
=0.79, sd =0.15).
3.3.2. Modelling: Wildlife–Wildlife Interactions. During the wet
season, wildlife interactions involving ungulates increased at
longer distances to houses (H1, model: CESS_wH1; Table 3),
and in areas with lower road densities, for the species pair
CE_VV(H1,model:CEVV_wH1).Humandisturbance,
through human presence, also had a negative effect on the
abundance of wildlife interactions in this season: in this case
between wild boar and red fox (H1, model: VVSS_wH1).
Furthermore, wildlife interactions—encompassing VV_SS
and CE_SS species pairs—increased in areas with higher
landscape diversity (H2, model: VVSS_wH2) and when the
temperature was lower (H4, model: CESS_wH4). In the dry
season, wildlife interactions also increased as a function of
low road densities, specifically for the SS_MM species pair
(H1, model: SSMM_dH1), while interactions between the
red fox and wild boar increased at reduced distances
from houses (H1, model: VVSS_dH1). Furthermore, wildlife
interactions—involving badger and red deer—increased in
rainy periods (H4, model: CEMM_dH4). Overall, with a
one-point increase in animal abundance (wildlife), the num-
ber of interactions would be expected to increase by an aver-
age IRR of 2.52 (sd =0.39), holding all variables constant.
Ecological predictors, linked to the study hypotheses, had a
lesser pronounced effect size. Positive predictors had an
average IRR of 1.26 (sd =0.27), while predicators exhibiting
a negative relation with the number of wildlife interactions
had an average IRR of 0.80 (sd=0.149), meaning that a one-
point increase in a given predictor would be expected to result
in a decrease in the rate ratio for the number of interactions.
4. Discussion
Pathogen transmission at shared interfaces is a heteroge-
neous and dynamic process, significantly dependent on spa-
tial and temporal processes. Despite being overlooked in
certain TB risk areas, characterising spatial–temporal varia-
tion in interaction patterns, addressing all relevant hosts, is
essential to properly understand pathogen transmission dynam-
ics in complex animal communities.
We demonstrated that (1) wildlife–cattle and wildlife
indirect interactions occur frequently. All the target species
contributed to the network of disease-relevant interactions
yet, wild boar, red deer, and red fox were the wildlife hosts
mostly involved in indirect interactions across seasons. Regard-
less of the group considered, speciespairinteractionsweregen-
erally higher in the wet season; (2) the rates of indirect
interaction involving wildlife–cattle were higher than the inter-
actions between wildlife species, in both seasons; (3) several
hypotheses influenced indirect interactions, although responses
differed among groups and seasons. Wildlife–cattle interactions
were more frequently related to the natural food and water
hypothesis (H5), while wildlife indirect interactions were more
associated with the human disturbance hypothesis (H1).
4.1. Wildlife–Cattle and Wildlife Interaction Patterns. Inter-
species direct interactions were rare, as previously documen-
ted in other studies [31, 33, 63]. This highlights that even
generalist species, with similar ecological requirements, tend
to partition resource use and habitat exploitation spatially
and temporally [59, 60, 90]. On the other hand, wildlife–cattle
and wildlife indirect interactions were frequent and wide-
spread throughout the study area. Such results are consistent
with previous findings reported in Mediterranean ecosystems,
supporting the hypothesis that M. bovis transmission (and
other multi-host pathogens with similar excretion routes) is
mainly indirect through contaminated shared environments
[23, 36, 91]. Agroforestry systems like Montado—known as
Dehesa in Spain—are highly complex structures often consid-
ered as high nature value farming systems, supporting high
levels of biodiversity [92]. Human activities (e.g., hunting
interests), along with other ecological and social factors,
have been shaping these interfaces, promoting a notable overlap
between wildlife (e.g., big game hunting) and cattle farming.
Consequently, Montado interfaces have become increasingly
interconnected, requiring improved management practices, as
shared space is expected to favour interspecies disease transmis-
sion. Indeed, the long-term excretion and viability maintenance
of Mycobacterium tuberculosis complex bacteria (MTBC) in
environmental substrates [50] increase animal exposure risk,
particularly in animal aggregation areas that are asynchronously
used by different species. In Mediterranean Spain, host species
richness has been correlated with increased community compe-
tence to maintain and transmit MTBC, oppositely to other epi-
demiological settings where biodiversity could favour a “dilution
effect”and moderate pathogen transmission [93].
Wild boar, red deer, and red fox were the wildlife hosts more
frequently involved in indirect interactions, as shown in previous
studies conducted in similar environments [6, 91, 94, 95]. The
positive relationship between wildlife/cattle abundance and the
number of interactions is notable, with significant effects
observed in all tested models. This pattern is compatible with
a density-dependent mechanism, a hypothesis previously sug-
gested in the context of animal interactions within disease sys-
tems [96], including TB [6, 14]. Thus, higher interaction events
involving ungulates and red foxes are expected, as they are
more abundant in our study area. On the other hand, the
low number of indirect interactions involving badgers could
be related to their lower local abundance, in contrast to other
Iberian environments (e.g., Asturias, Northern Spain) and
other European TB contexts (e.g., UK), known to have higher
badger population densities and where significant shared space
between badgers, cattle, and other wild mammals has been
documented [91, 97]. From an epidemiological perspective,
these results highlight that reservoir hosts (wild boar and red
deer) potentially play a key role in disease transmission in the
study region and should therefore receive increased attention
[53]. Wild boar has been identified as a TB maintenance host in
most study sites across the Iberian Peninsula. In the context of
multi-pathogen networks (study conducted in Spain), wild
boar is considered as the key and most connected species of
the system community, bridging several hosts relevant to the
epidemiology of MTBC [45, 53]. Also, TB prevalence in wild
12 Transboundary and Emerging Diseases

boar and the red deer was considered an important factor
positively linked to TB in cattle farms of Iberian regions
[98, 99]. Nevertheless, additional research (e.g., pathogen
excretion patterns and burden) is needed, including for other
non-reservoir hosts, given their potential to indirectly interact
with various species, as the case of the red fox in our study. The
red fox is a generalist carnivore that can exploit a variety of
habitats, including farm-related sites [100], and was recognised
as a spillover host in certain regions [101]. However, despite
recent insights about MTBC environmental contamination in
the Iberian Peninsula [50, 102], the relative importance of cer-
tain TB hosts—including the red fox—to environmental con-
tamination remains poorly understood in TB risk areas.
The higher rates of interactions during the wet season
may be due to different factors (e.g., species-specific beha-
viours, animal density; [6, 28]) but are mostly driven by two.
First, the higher availability and abundance of resources dur-
ing the wet season (e.g., autumn). While summer periods
tend to drive species aggregation around spatially limited
resources (e.g., water sites), the wet season is characterised
by high availability and abundance of natural food and water
sites. This could attract species to new areas, resulting in
indirect shared space across landscapes, which can be signif-
icant when considering common and generalist species as in
the case of red deer, wild boar, and red fox. Second, in our
study area, cattle are confined to fewer grazing plots during
the dry season when compared to the wet season. This may
also be a plausible explanation for the lower rates of inter-
actions involving cattle in the dry season (less sites where
animal hosts may engage), and contradicts other studies that
referred to a generalised increase in indirect interactions in
dry periods (but also in autumn periods) [24, 33].
4.2. Differences between Wildlife–Cattle and Wildlife Interaction
Rates. Wildlife–cattle indirect interaction rates were almost two
times higher than wildlife interaction rates in both seasons.
Triguero-Ocaña et al. [95] have also found that wildlife–cattle
interactions involving red deer, fallow deer (Dama dama), and
wild boar were more frequent than interactions between wildlife
species. Such patterns could be related to how species partition
resources across the landscape and to species-specificbehaviour
traits, which may differ between the two groups. The response of
wildlife to cattle presence (e.g., behavioural effects) can be het-
erogeneous when considering different animal species and land-
scape contexts [103]. Although some studies have shown that
cattle presence had a negative influence on space use by some
carnivore host species (e.g., badger and red fox; [64, 104, 105]),
others have shown that cattle presence was positively associated
with wildlife occurrence, namely for the wild boar and red deer
in agroforestry areas [6, 63]. Regarding the spatial–temporal
profiles of wildlife species, some studies demonstrated that even
habitat-generalist carnivores (e.g., red fox and badger) may
exhibit contrasting habitat preferences at a small-scale in agro-
forestry systems [59]; and mesocarnivore co-occurrence is lim-
ited by landscape homogeneity [67], a trait observed to some
extent in our study area. In addition, species (e.g., ungulates)
can segregate in terms of space and time to avoid competitive
and agonistic encounters [60]. Therefore, in Mediterranean
ecosystems characterized by multifunctional landscapes, inter-
species avoidance through shared resources between cattle and
wildlife should be smaller [106] than between nocturnal wild-
life species with more similar activity rhythm periods, sizes, and
diets [94, 107]. In turn, animal co-occurrence patterns may
dictate indirect interaction between hosts through shared
environments, and thus having considerable influence on ani-
mal TB epidemiology.
4.3. Ecological Hypotheses Driving Wildlife–Cattle and Wildlife
Indirect Interactions. The abundance of natural food and water
resources (H5) markedly influenced wildlife–cattle indirect
interaction patterns, particularly those involving red deer,
wild boar, and red fox. Our results indicate that, in the wet
season, wildlife–cattle interactions increased in less productive
areas (e.g., forested areas with high shrub cover), and around
control sites; while during the dry season, wildlife–cattle indi-
rect interactions were associated with more productive areas
andoccuredsignificantly more at sites with higher water con-
tent. Water and food resources (natural and artificial) have
been previously identified as key components, highly used by
both cattle and wildlife at shared interfaces, thereby favouring
interspecies transmission of M. bovis [24, 31, 32]. The seasonal
patterns evidenced in our work may be related to changes in
resource availability and abundance throughout the year. In
the wet season (mainly autumn and early winter), acorns
(important for ungulates) and pastures (important for cattle,
ungulates, and carnivores) are abundant in the study area and
more water sites are available. Oppositely, water and natural food
resources tend to be scarce and more spatially limited in the dry
season. Given that, in the wet season, although lower levels of
wildlife–cattle interactions are expected at specific sites (due to
the use of different resources), spatial co-occurrence between
cattle and wildlife continues to take place outside key resource
areas in different habitats, as documented in other studies [91].
On the other hand, highly productive natural food areas and
water sites become more attractive to numerous animal hosts
in the dry season. This leads to spatial aggregation of hosts at
specific sites, increasing the probability of indirect interactions
around key resources, as shown in previous studies [31, 41].
The tested hypotheses also revealed that the wildlife–cattle
interactions increased in areas with low human presence
(H1), more dense vegetation (H2; e.g., Forest), and in periods
of low temperature (H4) during the wet season; and, during
the dry season, wildlife–cattle interactions increased in areas
with lower road densities (H1), in more open areas (H2; i.e.,
less tree cover) and during rainy periods (H4). The effect of
land use [41] and human disturbance (e.g., hunting effects;
[63]) on species interactions have previously been suggested
in other Mediterranean areas. In addition, weather effects
(H4) can also play a role in interactions involving cattle, since
wildlife movement behaviour on farms can be affected by
temperature and rain (e.g., red fox and badger; [108]). Overall,
our results indicate that the critical conditions for animal
interactions, depending on the season, are shaped by several
ecological components. This highlights the importance to
consider a broad range of different ecological factors when
determining when and where disease transmission can occur.
Transboundary and Emerging Diseases 13

Effects associated with human disturbance hypothesis
(H1) were observed for wildlife interactions as well, which
have been largely understudied in the context of TB until
now. During the wet season, wildlife interactions were nega-
tively related to road density and human presence, and posi-
tively related to the distance to houses. In the dry season,
lower road densities and increased distances from houses
were also found to be key conditions where transmission
of M. bovis may be favoured between wildlife species (i.e.,
high rates of indirect interactions). Studies have demon-
strated that wildlife occurrence is strongly affected by differ-
ent anthropogenic factors, such as roads (e.g., ungulates and
carnivores in relation to dirt roads; [109, 110]), human pres-
ence (e.g., ungulates; [66]), or even human settlements (e.g.,
carnivores; [111]). We hypothesised that in the study area,
wildlife species (both carnivores and ungulates) tend to avoid
unpaved roads—they are frequently used by local workers
and hunters throughout seasons—and areas close to houses
(particularly interactions involving the red deer). By adopt-
ing such behaviours, species reduce the probability of distur-
bance, which, as expected, results in lower abundance of
indirect interactions through common space use in those
areas. In the dry season, the higher probability of wildlife
interactions in areas close to houses could be explained by
the characteristic behaviour of the species involved, namely
the red fox and wild boar. These are opportunistic species
that can take advantage of resources close to human settle-
ments when those resources are scarce elsewhere, as docu-
mented in other Mediterranean areas and habitats [64]. This
may also explain why wildlife indirect interactions involving
those species increase in more heterogeneous areas (H2), but
in this case, evidenced during the wet season when various
resources are often available across different habitats. Finally,
models showed that wildlife interactions were influenced by
weather conditions (H4; ungulates in relation with tempera-
ture and red deer and badger in relation to rain). We
hypothesised that during the wet season, species home range
could increase as a function of temperature, as documented
for ungulates and some carnivores [68, 112]. As a result, this
can lead species to use different spatial resources, likely
reducing the abundance of interactions under these circum-
stances. On the other hand, species can boost their activity
during the dry season in rainy periods (very infrequent
events), which could be linked to increased prey activity
and/or immediate water availability, for instance. Because
resources are more limited in the dry season, such patterns
can result in negligible spatial segregation, and thus probably
increase indirect interactions between wildlife species, par-
ticularly at specific resource sites (e.g., water sites).
Overall, improving our understanding of the ecological
and environmental drivers underlying disease-relevant inter-
actions at the wildlife–cattle interfaces is likely to provide
valuable insights into the real nature of pathogen transmis-
sion events. This knowledge can help refine and guide effec-
tive control actions in risk areas wherein disease still persists.
Currently, TB surveillance in wildlife in Portugal almost
exclusively relies in veterinary inspection of hunted large
game animals in specific areas with endemic circulation of
M. bovis. Moreover, conventional biosecurity measures can
be particularly difficult to implement in animal extensive
production systems, posing a considerable challenge for con-
trolling multi-species pathogens. Still, additional preventive
measures could be considered for disrupting M. bovis trans-
mission chains. One example could involve implementing
selective fencing and gating systems in specific areas where
wildlife and cattle frequently share space, and where increas-
ing interaction rates are expected (e.g., water sites in the dry
season; [113]). Data from the present study may guide future
actions as it could help refine disease risk maps, which pres-
ently mainly rely on data from disease breakdowns in cattle
herds. Furthermore, wildlife densities—given their role in
our study—should be closely monitored, along with environ-
mental sampling to assess contamination of natural sub-
strates, particularly in areas highly used by different hosts.
4.4. Study Limitations and Future Perspectives. We identified
three main aspects that should be further scrutinised by
researchers in the multi-host TB context: first, in our study,
the even distribution of cameras across the landscape, encom-
passing different land uses, enhances the representation of
features influencing animal detection proportionally to their
availability. However, this does not eliminate the overall
detection bias arising from the landscape, which can influence
the field of vision of camera traps (e.g., reduced detection field
in dense environments compared to open areas). Future stud-
ies on interaction patterns should integrate new tools (e.g.,
occupancy models) to address imperfect detection of indivi-
duals. Additionally, exploring animal-based metrics (e.g., via
REM—random encounter model) that consider the collective
viewsheds of a camera array could improve animal detection
rates and related estimates (e.g., interactions) across varying
spatial gradients and external drivers [59, 114]; second, host
behaviour may determine the relative importance of a host
within a disease system. Even if not very abundant, the beha-
vioural repertoire could favour an increased contact with
other hosts through shared environments [12, 115]. For
instance, certain risk behaviours (e.g., wallowing, drinking)
can promote frequent and prolonged contact with various
infection sources and affect infection outcome and excretion
patterns per host. This topic needs further research as it
remains poorly understood in the Iberian context; third, since
indirect transmission depends on M. bovis survival time in
environments, the use of CTW is crucial for generating reli-
able estimates. However, as M. bovis can survive for extended
periods, depending on climate, substrates, and others [116],
important questions arise: where should the baseline (CTW,
in time axis) be established in a given context? Should the
infectious period be based on the average environmental per-
sistence of M. bovis? Should we examine the frequency of
indirect interactions that occurred within a plausible range
of CTW’s, according to hosts, to better define baselines?
Should different CTW estimates based on M. bovis survival
be considered across various substrates associated with sam-
pling sites? [34]. Progress in addressing these important
questions has been made, with a few studies pioneering the
implementation of CTW’s through different approaches to
14 Transboundary and Emerging Diseases

define host interactions [31, 63]. Adopting similar frame-
works, with environmental survival as a gold-standard metric,
will improve integration and comparison of results across
studies. Nevertheless, researchers will also benefit from stud-
ies exploring multiple CTW’s as a function of interaction
gradients, as well as the definition of CTW’s according to
sampling spatial conditions [33]. This is key to developing
general theory on this topic, also applicable to other infectious
diseases at the wildlife–cattle interface.
5. Conclusions
This is the most comprehensive study carried out in Portugal
focusing on species indirect interactions in an endemic TB
context, and identifying the most likely key ecological factors
driving these interactions across shared environments. Our
study confirmed that the availability of natural food and
water was a main driver of wildlife–cattle interactions, while
wildlife indirect interactions were more associated with human
disturbance. However, other ecological hypotheses influenced
indirect interaction patterns, suggesting that the conditions
favouring the complex transmission of M. bovis are determined
by multiple factors, depending on the host species and season.
Future studies should combine interaction data with the extent
of environmental contamination with M. bovis to properly
assess transmission risk in multi-host communities. Further-
more, the composition and structure of multi-host communi-
ties determining complex interaction patterns in space-time
axes should also be considered when establishing priority mea-
sures for disease control in shared environments.
Data Availability
The data associated with this research are available from the
corresponding author upon reasonable request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Authors’Contributions
Eduardo M. Ferreira, Elsa L. Duarte, Mónica V. Cunha,
António Mira, and Sara M. Santos conceived the study.
Eduardo M. Ferreira collected the data, analysed the data,
and wrote the first manuscript draft. Eduardo M. Ferreira,
Renata Gonçalves, and Tiago Pinto processed the data.
Eduardo M. Ferreira and Sara M. Santos developed the anal-
ysis protocol. All authors contributed substantially to revi-
sions and gave final approval for publication.
Acknowledgments
We thank landowners, farmers, and game managers for
engaging in this collaborative research project. A special
thanks to Noudar Nature Park (EDIA, S.A.) for their collab-
oration and invaluable support for this work. This study was
funded by the Portuguese Foundation for Science and Tech-
nology (FCT), namely with two PhD grants (SFRH/BD/146037/
2019; DOI: https://doi.org/10.54499/SFRH/BD/146037/2019) to
Eduardo M. Ferreira and (2020.04581.BD) to Tiago Pinto;
Eduardo M. Ferreira was also financed by MED (DOI: https://
doi.org/10.54499/UIDB/05183/2020) and CHANGE (DOI:
https://doi.org/10.54499/LA/P/0121/2020) funds to PhD stu-
dents. This work was also funded by FCT within the scope of
MOVERCULOSIS project (2022.06014.PTDC)—combining
animal behaviour and movement to assess the influence of
wildlife–livestock interactions on the spatiotemporal trans-
mission risk of animal tuberculosis (https://doi.org/10.
54499/2022.06014.PTDC). We also acknowledge strategic
funding from FCT to cE3c and BioISI Research Units (DOI:
https://doi.org/10.54499/UIDB/00329/2020 and DOI: https://
doi.org/10.54499/UIDB/04046/2020, respectively), to MED
Research Unit (UIDP/05183/2020 and UIDB/05183/2020),
and to the Associate Laboratory CHANGE (DOI: https://
doi.org/10.54499/LA/P/0121/2020).
Supplementary Materials
Additional supplementary information includes tempera-
ture/rainfall comparisons between Ciudad real (Spain) and
Barrancos (Portugal), and DHARMa diagnostic plots of the
tested models. (Supplementary Materials)
References
[1] J. M. Hassell, M. Begon, M. J. Ward, and E. M. Fèvre,
“Urbanization and disease emergence: dynamics at the
wildlife-livestock–human interface,”Trends in Ecology &
Evolution, vol. 32, no. 1, pp. 55–67, 2017.
[2] A. Caron, J. A. Barasona, E. Miguel, J. Michaux, and M. De
Garine-Wichatitsky, “Characterization of wildlife-livestock
interfaces: the need for interdisciplinary approaches and a
dedicated thematic field,”in Diseases at the Wildlife—
Livestock Interface, pp. 339–367, Springer, Cham, 2021.
[3] O. Venter, E. W. Sanderson, A. Magrach et al., “Sixteen years
of change in the global terrestrial human footprint and
implications for biodiversity conservation,”Nature Commu-
nications, vol. 7, no. 1, pp. 1–11, 2016.
[4] R. J. White and O. Razgour, “Emerging zoonotic diseases
originating in mammals: a systematic review of effects of
anthropogenic land-use change,”Mammal Review, vol. 50,
no. 4, pp. 336–352, 2020.
[5] F. Meurens, C. Dunoyer, C. Fourichon et al., “Animal board
invited review: risks of zoonotic disease emergence at the
interface of wildlife and livestock systems,”Animal, vol. 15,
no. 6, Article ID 100241, 2021.
[6] R. Carrasco-Garcia, J. A. Barasona, C. Gortazar, V. Montoro,
J. M. Sanchez-Vizcaino, and J. Vicente, “Wildlife and
livestock use of extensive farm resources in south central
Spain: implications for disease transmission,”European Journal
of Wildlife Research, vol. 62, no. 1, pp. 65–78, 2016.
[7] A. J. Carpio, M. Apollonio, and P. Acevedo, “Wild ungulate
overabundance in Europe: contexts, causes, monitoring and
management recommendations,”Mammal Review, vol. 51,
no. 1, pp. 95–108, 2021.
[8] B. A. Jones, D. Grace, R. Kock et al., “Zoonosis emergence
linked to agricultural intensification and environmental change,”
Proceedings of the National Academy of Sciences of the United
States of America, vol. 110, pp. 8399–8404, 2013.
[9] K. A. Alexander, C. J. Carlson, B. L. Lewis et al., “The
ecology of pathogen spillover and disease emergence at the
human-wildlife-environment interface,”in The Connections
Transboundary and Emerging Diseases 15

Between Ecology and Infectious Disease, pp. 267–298,
Springer, Cham, 2018.
[10] C. Gortazar, L. A. Reperant, T. Kuiken et al., “Crossing the
interspecies barrier: opening the door to zoonotic pathogens,”
PLoS Pathogens, vol. 10, no. 6, Article ID e1004129, 2014.
[11] C. Gortazar, I. Diez-Delgado, J. A. Barasona, J. Vicente, J. De
La Fuente, and M. Boadella, “The wild side of disease control
at the wildlife-livestock-human interface: a review,”Frontiers
in Veterinary Science, vol. 1, Article ID 27, 2015.
[12] J. P. Webster, A. Borlase, and J. W. Rudge, “Who acquires
infection from whom and how? Disentangling multi-host and
multimode transmission dynamics in the ‘elimination’era,”
Philosophical Transactions of the Royal Society B: Biological
Sciences, vol. 372, no. 1719, Article ID 20160091, 2017.
[13] J. M. Hassell, T. Newbold, A. P. Dobson et al., “Towards an
ecosystem model of infectious disease,”Nature Ecology &
Evolution, vol. 5, no. 7, pp. 907–918, 2021.
[14] E. M. Ferreira, E. L. Duarte, M. V. Cunha, A. Mira, and
S. M. Santos, “Disentangling wildlife-cattle interactions in
multi-host tuberculosis scenarios: systematic review and
meta-analysis,”Mammal Review, vol. 53, no. 4, pp. 287–302,
2023.
[15] S. D. Fitzgerald and J. B. Kaneene, “Wildlife reservoirs of
bovine tuberculosis worldwide: hosts, pathology, surveil-
lance, and control,”Veterinary Pathology, vol. 50, no. 3,
pp. 488–499, 2013.
[16] C. Gortázar, A. Che Amat, and D. J. O’Brien, “Open questions
and recent advances in the control of a multi-host infectious
disease: animal tuberculosis,”Mammal Review, vol. 45, no. 3,
pp. 160–175, 2015.
[17] A. C. Pereira, A. C. Reis, B. Ramos, and M. V. Cunha,
“Animal tuberculosis: impact of disease heterogeneity in
transmission, diagnosis, and control,”Transboundary and
Emerging Diseases, vol. 67, pp. 1828–1846, 2020.
[18] B. Ramos, A. C. Pereira, A. C. Reis, and M. V. Cunha,
“Estimates of the global and continental burden of animal
tuberculosis in key livestock species worldwide: a meta-
analysis study,”One Health, vol. 10, Article ID 100169, 2020.
[19] A. Caminiti, F. Pelone, G. La Torre et al., “Control and
eradication of tuberculosis in cattle: a systematic review of
economic evidence,”The Veterinary Record, vol. 179, no. 3,
pp. 70–75, 2016.
[20] J. Zinsstag, E. Schelling, F. Roth, and R. R. Kazwala,
“Economics of Bovine Tuberculosis,”in Mycobacterium
Bovis Infection in Animals and Humans, C. O. Thoen,
J. H. Steele, and M. J. Gilsdorf, Eds., pp. 68–83, Wiley-
Blackwell, Boston, Massachusetts, USA, 2008.
[21] L. A. L. Corner, “The role of wild animal populations in the
epidemiology of tuberculosis in domestic animals: how to
assess the risk,”Veterinary Microbiology, vol. 112, no. 2–4,
pp. 303–312, 2006.
[22] M. V. Palmer, “Mycobacterium bovis: characteristics of wildlife
reservoir hosts,”Transboundary and Emerging Diseases,vol.60
Suppl 1, pp. 1–13, 2013.
[23] A. R. Allen, T. Ford, and R. A. Skuce, “Does Mycobacterium
tuberculosis var. bovis survival in the environment confound
bovine tuberculosis control and eradication? a literature
review,”Veterinary Medicine International, vol. 2021,
Article ID 8812898, 19 pages, 2021.
[24] E. Kukielka, J. A. Barasona, C. E. Cowie et al., “Spatial and
temporal interactions between livestock and wildlife in South
Central Spain assessed by camera traps,”Preventive Veterinary
Medicine, vol. 112, no. 3-4, pp. 213–221, 2013.
[25] M. Q. Wilber, K. M. Pepin, H. Campa et al., “Modelling
multi-species and multi-mode contact networks: implications
for persistence of bovine tuberculosis at the wildlife-livestock
interface,”Journal of Applied Ecology, vol. 56, no. 6,
pp. 1471–1481, 2019.
[26] G. F. Albery, L. Kirkpatrick, J. A. Firth, and S. Bansal,
“Unifying spatial and social network analysis in disease
ecology,”Journal of Animal Ecology, vol. 90, no. 1, pp. 45–
61, 2021.
[27] A. Yang, M. Q. Wilber, K. R. Manlove et al., “Deriving
spatially explicit direct and indirect interaction networks
from animal movement data,”Ecology and Evolution, vol. 13,
no. 3, Article ID e9774, 2023.
[28] R. Triguero-Ocaña, J. A. Barasona, F. Carro, R. C. Soriguer,
J. Vicente, and P. Acevedo, “Spatio–temporal trends in the
frequency of interspecificinteractions between domestic and
wild ungulates from Mediterranean Spain,”PLOS One,
vol. 14, no. 1, Article ID e0211216, 2019.
[29] B. H. Hayes, T. Vergne, M. Andraud, and N. Rose,
“Mathematical modeling at the livestock-wildlife interface:
scoping review of drivers of disease transmission between
species,”Frontiers in Veterinary Science, vol. 10, Article ID
1225446, 2023.
[30] M. Böhm, M. R. Hutchings, and P. C. L. White, “Contact
networks in a wildlife- livestock host community: identifying
high-risk individuals in the transmission of bovine TB among
badgers and cattle,”PLoS ONE, vol. 4, no. 4, Article ID
e5016, 2009.
[31] M. J. Lavelle, S. L. Kay, K. M. Pepin, D. A. Grear, H. Campa,
and K. C. VerCauteren, “Evaluating wildlife-cattle contact
rates to improve the understanding of dynamics of bovine
tuberculosis transmission in Michigan, USA,”Preventive
Veterinary Medicine, vol. 135, pp. 28–36, 2016.
[32] E. L. Campbell, A. W. Byrne, F. D. Menzies et al.,
“Interspecific visitation of cattle and badgers to fomites: a
transmission risk for bovine tuberculosis?”Ecology and
Evolution, vol. 9, no. 15, pp. 8479–8489, 2019.
[33] C. E. Cowie, M. R. Hutchings, J. A. Barasona, C. Gortázar,
J. Vicente, and P. C. L. White, “Interactions between four
species in a complex wildlife: livestock disease community:
implications for Mycobacterium bovis maintenance and
transmission,”European Journal of Wildlife Research, vol. 62,
pp. 51–64, 2016.
[34] A. Payne, S. Philipon, J. Hars, B. Dufour, and E. Gilot-Fromont,
“Wildlife interactions on baited places and waterholes in a
French area infected by Bovine Tuberculosis,”Frontiers in
Veterinary Science,vol.3,pp.1–11, 2017.
[35] B. Borremans, C. Faust, K. R. Manlove, S. H. Sokolow, and
J. O. Lloyd-Smith, “Cross-species pathogen spillover across
ecosystem boundaries: mechanisms and theory,”Philosophi-
cal Transactions of the Royal Society B: Biological Sciences,
vol. 374, Article ID 20180344, 2019.
[36] C. Gortázar, J. de la Fuente, A. Perelló, and L. Domínguez,
“Will we ever eradicate animal tuberculosis?”Irish Veterinary
Journal, vol. 76, no. Suppl 1, Article ID 24, 2023.
[37] E. L. Duarte, M. Domingos, A. Amado, and A. Botelho,
“Spoligotype diversity of Mycobacterium bovis and Mycobac-
terium caprae animal isolates,”Veterinary Microbiology,
vol. 130, no. 3-4, pp. 415–421, 2008.
[38] N. Santos, M. Correia-Neves, V. Almeida, and C. Gortázar,
“Wildlife tuberculosis: a systematic review of the epidemiol-
ogy in Iberian Peninsula. In: Cunha MLRS (ed) epidemiology
insights,”InTech, Rijeka, 2012.
16 Transboundary and Emerging Diseases

[39] J. Vicente, J. A. Barasona, P. Acevedo et al., “Temporal trend
of tuberculosis in wild ungulates from Mediterranean Spain,”
Transboundary and Emerging Diseases, vol. 60 Suppl 1,
pp. 92–103, 2013.
[40] A. C. Reis, L. C. M. Salvador, S. Robbe-Austerman et al.,
“Whole genome sequencing refines knowledge on the population
structure of Mycobacterium bovis from a multi-host tuberculosis
system,”Microorganisms,vol.9,no.8,pp.1–15, 2021.
[41] J. A. Barasona, M. C. Latham, P. Acevedo et al., “Spatiotem-
poral interactions between wild boar and cattle: implications
for cross-species disease transmission,”Veterinary Research,
vol. 45, no. 1, Article ID 122, 2014.
[42] A. C. Abrantes, P. Acevedo, J. Martínez-Guijosa, J. Serejo,
and M. Vieira-Pinto, “Identification and evaluation of risk
factors associated to Mycobacterium bovis transmission in
southeast hunting areas of central Portugal,”Galemys,
Spanish Journal of Mammalogy, vol. 31, pp. 61–68, 2019.
[43] R. Triguero-Ocaña, E. Laguna, S. Jiménez-Ruiz et al., “The
wildlife-livestock interface on extensive free-ranging pig farms
in central Spain during the “montanera”period,”Transbound-
ary and Emerging Diseases, vol. 68, no. 4, pp. 2066–2078, 2021.
[44] A. R. Allen,R. A. Skuce, and A. W. Byrne, “Bovine tuberculosis
in Britain and Ireland—a perfect storm? The confluence of
potential ecological and epidemiological impediments to
controlling a chronic infectious disease,”Frontiers in Veterinary
Science, vol. 5, Article ID 109, 2018.
[45] P. Barroso, D. Relimpio, J. A. Zearra et al., “Using integrated
wildlife monitoring to prevent future pandemics through one
health approach,”One Health, vol. 16, Article ID 100479,
2023.
[46] M. Vieira-Pinto, J. Alberto, J. Aranha et al., “Combined
evaluation of bovine tuberculosis in wild boar (Sus scrofa)
and red deer (Cervus elaphus) from central-east Portugal,”
European Journal of Wildlife Research, vol. 57, pp. 1189–
1201, 2011.
[47] N. Santos, C. Santos, T. Valente, C. Gortázar, V. Almeida,
and M. Correia-Neves, “Widespread environmental contam-
ination with Mycobacterium tuberculosis complex revealed by
a molecular detection protocol,”PLoS ONE, vol. 10, no. 11,
Article ID e0142079, 2015.
[48] N. Santos, T. Nunes, C. Fonseca et al., “Spatial analysis of
wildlife tuberculosis based on a serologic survey using dried
blood spots, Portugal,”Emerging Infectious Diseases, vol. 24,
no. 12, pp. 2169–2175, 2018.
[49] J. Aranha, A. C. Abrantes, R. Gonçalves, R. Miranda,
J. Serejo, and M. Vieira-Pinto, “Gis as an epidemiological tool
to monitor the spatial–temporal distribution of tuberculosis
in large game in a high-risk area in Portugal,”Animals,
vol. 11, Article ID 2374, 2021.
[50] A. C. Pereira, D. Pinto, and M. V. Cunha, “Unlocking
environmental contamination of animal tuberculosis hot-
spots with viable mycobacteria at the intersection of flow
cytometry, PCR, and ecological modelling,”The Science of the
Total Environment, vol. 891, Article ID 164366, 2023.
[51] M. V. Cunha, M. Monteiro, P. Carvalho, P. Mendonça,
T. Albuquerque, and A. Botelho, “Multihost tuberculosis:
insights from the Portuguese control program,”Veterinary
Medicine International, vol. 2011, Article ID 795165, 2011.
[52] DGAV, “Tuberculose em Caça Maior, Edital 1/2011,”2011,
https://www.dgav.pt/wp-content/uploads/2021/04/EDITAL-
No.-1-TUBERCULOSE-EM-CACA-MAIOR.pdf.
[53] N. Santos, E. F. Colino, M. C. Arnal et al., “Complementary
roles of wild boar and red deer to animal tuberculosis
maintenance in multi-host communities,”Epidemics, vol. 41,
Article ID 100633, 2022.
[54] DGAV, “Dados dos planos sanitários de erradicação /2022,”
2023, https://www.dgav.pt/wp-content/uploads/2023/11/DA
DOS-PLANOS-SANITARIOS-DE-ERRADICACAO_2022_
f.pdf.
[55] A. Costa, Contribution to the epidemiological study of
tuberculosis in large game animals and cattle in the municipalities
of Moura and Barrancos, M.S. thesis, Lusofona University,
Lisbon, 2015.
[56] A. C. Reis, B. Ramos, A. C. Pereira, and M. V. Cunha, “The
hard numbers of tuberculosis epidemiology in wildlife: a
meta-regression and systematic review,”Transboundary and
Emerging Diseases, vol. 68, no. 6, pp. 3257–3276, 2021.
[57] IPMA, “Instituto Português do Mar e da Atmosfera, IP,
Portugal. Normal Climatológica—Beja 1981-2010,”2023,
https://www.ipma.pt/bin/file.data/climate-normal/cn_81-10_
BEJA.pdf.
[58] A. C. Matos, L. Figueira, M. H. Martins, M. L. Pinto,
M. Matos, and A. C. Coelho, “New insights into Mycobacte-
rium bovis prevalence in wild mammals in Portugal,”
Transboundary and Emerging Diseases, vol.63, no.5,
pp. e313–e322, 2016.
[59] G. Curveira-Santos, T. A. Marques, M. Björklund, and
M. Santos-Reis, “Mediterranean mesocarnivores in spatially
structured managed landscapes: community organisation in
time and space,”Agriculture, Ecosystems &Environment,
vol. 237, pp. 280–289, 2017.
[60] M. Zanni, F. Brivio, S. Grignolio, and M. Apollonio,
“Estimation of spatial and temporal overlap in three ungulate
species in a Mediterranean environment,”Mammal Research,
vol. 66, pp. 149–162, 2021.
[61] R. Triguero-Ocaña, J. Vicente, P. Palencia, E. Laguna, and
P. Acevedo, “Quantifying wildlife-livestock interactions and their
spatio-temporal patterns: is regular grid camera trapping a
suitable approach?”Ecological Indicators,vol.117,ArticleID
106565, 2020.
[62] P. Harvey, “EXIFtool,”2022, http://www.sno.phy.queensu.
ca/%7Ephil/exiftool/.
[63] J. Martínez-Guijosa, A. López-Alonso, C. Gortázar,
P. Acevedo, M. Torres, and J. Vicente, “Shared use of
mineral supplement in extensive farming and its potential for
infection transmission at the wildlife-livestock interface,”
European Journal of Wildlife Research, vol. 67, Article ID 55,
2021.
[64] M. Alexandre, D. Hipólito, E. Ferreira, C. Fonseca, and
L. M. Rosalino, “Humans do matter: determinants of red fox
(Vulpes vulpes) presence in a western Mediterranean landscape,”
Mammal Research, vol. 65, pp. 203–214, 2020.
[65] E. Laguna, J. A. Barasona, J. Vicente, O. Keuling, and
P. Acevedo, “Differences in wild boar spatial behaviour
among land uses and management scenarios in Mediterra-
nean ecosystems,”The Science of the Total Environment,
vol. 796, Article ID 148966, 2021.
[66] L. M. Rosalino, D. Teixeira, C. Camarinha et al., “Even
generalist and resilient species are affected by anthropic
disturbance: evidence from wild boar activity patterns in a
Mediterranean landscape,”Mammal Research, vol. 67,
pp. 317–325, 2022.
[67] P. Linck, F. Palomares, N. Negrões et al., “Increasing
homogeneity of Mediterranean landscapes limits the co-
occurrence of mesocarnivores in space and time,”Landscape
Ecology, vol. 38, no. 12, pp. 3657–3673, 2023.
Transboundary and Emerging Diseases 17

[68] I. M. Rivrud, L. E. Loe, and A. Mysterud, “How does local
weather predict red deer home range size at different
temporal scales?”Journal of Animal Ecology, vol. 79, no. 6,
pp. 1280–1295, 2010.
[69] QGIS Development Team, “QGIS geographic information
system. Open-source geospatial foundation,”2022, http://
qgis.org.
[70] M. H. K. Hesselbarth, M. Sciaini, K. A. With, K. Wiegand,
and J. Nowosad, “landscapemetrics: an open-source Rtool to
calculate landscape metrics,”Ecography, vol. 42, no. 10,
pp. 1648–1657, 2019.
[71] N. Gorelick, M. Hancher, M. Dixon, S. Ilyushchenko,
D. Thau, and R. Moore, “Google Earth Engine: planetary-
scale geospatial analysis for everyone,”Remote Sensing of
Environment, vol. 202, pp. 18–27, 2017.
[72] J. T. Kerr and M. Ostrovsky, “From space to species:
ecological applications for remote sensing,”Trends in Ecology
&Evolution, vol. 18, no. 6, pp. 299–305, 2003.
[73] N. Pettorelli, J. M. Gaillard, A. Mysterud et al., “Using a
proxy of plant productivity (NDVI) to find key periods for
animal performance: the case of roe deer,”Oikos, vol. 112,
no. 3, pp. 565–572, 2006.
[74] M. J. Santos, L. M. Rosalino, M. Santos-Reis, and S. L. Ustin,
“Testing remotely-sensed predictors of meso-carnivore habitat
use in Mediterranean ecosystems,”Landscape Ecology,vol.31,
no. 8, pp. 1763–1780, 2016.
[75] T. Pinto, S. M. Santos, A. Mira, and N. D. Sillero,
“Importance of water availability for amphibian roadkill in
a Mediterranean landscape,”Biodiversity and Conservation,
vol. 32, no. 7, pp. 2513–2537, 2023.
[76] X. Li, W. Zhu, Z. Xie et al., “Assessing the effects of time
interpolation of NDVI composites on phenology trend
estimation,”Remote Sensing, vol. 13, no. 24, Article ID 5018,
2021.
[77] J. Oeser, M. Heurich, C. Senf, D. Pflugmacher, and
T. Kuemmerle, “Satellite-based habitat monitoring reveals
long-term dynamics of deer habitat in response to forest
disturbances,”Ecological Applications, vol. 31, no. 3,
Article ID e02269, 2021.
[78] V. Arroyo-Rodríguez, M. Martínez-Ruiz, J. Bezerra,
C. Galán-Acedo, M. San-José, and L. Fahrig, “Does a species’
mobility determine the scale at which it is influenced by the
surrounding landscape pattern?”Current Landscape Ecology
Reports, vol. 8, no. 1, pp. 23–33, 2023.
[79] A. F. Zuur, E. N. Ieno, N. Walker, A. A. Saveliev, and
G. M. Smith, Mixed Effects Models and Extensions in Ecology
with R, Springer, New York, USA, 2009.
[80] M. E. Brooks, K. Kristensen, K. J. van Benthem et al.,
“glmmTMB balances speed and flexibility among packages
for zero-inflated generalized linear mixed modeling,”The R
Journal, vol. 9, no. 2, pp. 378–400, 2017.
[81] R. B. O’Hara and D. J. Kotze, “Do not log-transform count
data,”Methods in Ecology and Evolution, vol. 1, no. 2,
pp. 118–122, 2010.
[82] A. F. Vallejo-Vargas, D. Sheil, A. Semper-Pascual et al.,
“Consistent diel activity patterns of forest mammals among
tropical regions,”Nature Communications, vol. 13, no. 1,
Article ID 7102, 2022.
[83] K. P. Burnham and D. R. Anderson, Model Selection and
Multimodel Inference: A Practical Information—Theoretic
Approach, Springer- Verlag, New York, USA, 2002.
[84] P. A. Stephens, S. W. Buskirk, G. D. Hayward, and C. M. Del
Rio, “Information theory and hypothesis testing: a call for
pluralism,”Journal of Applied Ecology, vol. 42, no. 1, pp. 4–
12, 2005.
[85] J. E. Humphrey, A. Haslem, and A. F. Bennett, “Housing or
habitat: what drives patterns of avian species richness in
urbanized landscapes?”Landscape Ecology, vol. 38, no. 8,
pp. 1919–1937, 2023.
[86] K. Barton, “MuMIn: multi-model inference (version 1.47.1),”
2022.
[87] R. Mac Nally, R. P. Duncan, J. R. Thomson, and J. D. L. Yen,
“Model selection using information criteria, but is the “best”
model any good?”Journal of Applied Ecology, vol. 55, no. 3,
pp. 1441–1444, 2018.
[88] A. T. Tredennick, G. Hooker, S. P. Ellner, and P. B. Adler, “A
practical guide to selecting models for exploration, inference,
and prediction in ecology,”Ecology, vol. 102, no. 6,
Article ID e03336, 2021.
[89] F. Hartig, “DHARMa: residual diagnostics for hierarchical
(multi-level/mixed) regression models,”R package version
0.4.6, 2022.
[90] D. F. Teixeira, G. Ares-Pereira, C. Camarinha et al., “Effect
of anthropic disturbances on the activity pattern of two
generalist mesocarnivores inhabiting Mediterranean forestry
plantations,”Biodiversity and Conservation, vol. 32, no. 4,
pp. 1251–1270, 2023.
[91] L. D. Varela-Castro, I. A. Sevilla, A. Payne, E. Gilot-Fromont,
and M. Barral, “Interaction patterns between wildlife and
cattle reveal opportunities for mycobacteria transmission in
farms from north-eastern Atlantic Iberian Peninsula,”
Animals, vol. 11, no. 8, Article ID 2364, 2021.
[92] T. Pinto-Correia, N. Ribeiro, and P. Sá-Sousa, “Introducing
the montado, the cork and holm oak agroforestry system of
Southern Portugal,”Agroforestry Systems, vol. 82, no. 2,
pp. 99–104, 2011.
[93] J. A. Barasona, C. Gortázar, J. de la Fuente, and J. D. Vicente,
“Host richness increases tuberculosis disease risk in game-
managed areas,”Microorganisms, vol. 7, no. 6, Article ID
182, 2019.
[94] A. Payne, S. Chappa, J. Hars, B. Dufour, and E. Gilot-
Fromont, “Wildlife visits to farm facilities assessed by camera
traps in a bovine tuberculosis-infected area in France,”
European Journal of Wildlife Research, vol. 62, no. 1, pp. 33–
42, 2016.
[95] R. Triguero-Ocaña, B. Martínez-López, J. Vicente,
J. A. Barasona, J. Martínez-Guijosa, and P. Acevedo,
“Dynamic network of interactions in the wildlife-livestock
interface in Mediterranean Spain: an epidemiological point of
view,”Pathogens, vol. 9, no. 2, pp. 1–16, 2020.
[96] K. Manlove, M. Wilber, L. White et al., “Defining an
epidemiological landscape that connects movement ecology
to pathogen transmission and pace-of-life,”Ecology Letters,
vol. 25, no. 8, pp. 1760–1782, 2022.
[97] J. A. Drewe, H. M. O’Connor, N. Weber, R. A. McDonald,
and R. J. Delahay, “Patterns of direct and indirect contact
between cattle and badgers naturally infected with tuberculo-
sis,”Epidemiology and Infection, vol. 141, no. 7, pp. 1467–
1475, 2013.
[98] N. P. LaHue, J. V. Baños, P. Acevedo, C. Gortázar, and
B. Martínez-López, “Spatially explicit modeling of animal
tuberculosis at the wildlife-livestock interface in Ciudad Real
province, Spain,”Preventive Veterinary Medicine, vol. 128,
pp. 101–111, 2016.
[99] C. Herraiz, Jín Vicente, C. Gortázar, and P. Acevedo, “Large
scale spatio-temporal modelling of risk factors associated
18 Transboundary and Emerging Diseases

with tuberculosis exposure at the wildlife-livestock interface,”
Preventive Veterinary Medicine, vol. 220, Article ID 106049,
2023.
[100] B. A. Tolhurst, A. I. Ward, and R. J. Delahay, “A study of fox
(Vulpes vulpes) visits to farm buildings in southwest England
and the implications for disease management,”European
Journal of Wildlife Research, vol. 57, pp. 1227–1230, 2011.
[101] C. Richomme, E. Réveillaud, J. L. Moyen et al., “Mycobacte-
rium bovis infection in red foxes in four animal tuberculosis
endemic areas in France,”Microorganisms, vol. 8, no. 7,
Article ID 1070, 2020.
[102] J. Martínez-Guijosa, B. Romero, J. A. Infantes-Lorenzo et al.,
“Environmental DNA: a promising factor for tuberculosis
risk assessment in multi-host settings,”PLoS ONE, vol. 15,
no. 5, Article ID e0233837, 2020.
[103] J. C. Huaranca, A. J. Novaro, and C. E. Valdivia, “Effects of
livestock grazing on biodiversity: a meta-analysis on three
trophic levels,”Journal for Nature Conservation, vol. 66,
Article ID 126126, 2022.
[104] E. M. Mullen, T. MacWhite, P. K. Maher, D. J. Kelly,
N. M. Marples, and M. Good, “Foraging Eurasian badgers
Meles meles and the presence of cattle in pastures,”Do
Badgers Avoid Cattle? Applied Animal Behaviour Science,
vol. 144, pp. 130–137, 2013.
[105] R. Woodroffe, C. A. Donnelly, C. Ham et al., “Badgers prefer
cattle pasture but avoid cattle: implications for bovine
tuberculosis control,”Ecology Letters, vol. 19, no. 10,
pp. 1201–1208, 2016.
[106] J. Carvalho, R. T. Torres, P. Acevedo et al., “Propagule
pressure and land cover changes as main drivers of red and
roe deer expansion in mainland Portugal,”Diversity and
Distributions, vol. 24, pp. 551–564, 2018.
[107] M. Vilella, M. Ferrandiz-Rovira, and F. Sayol, “Coexistence of
predators in time: effects of season and prey availability on
species activity within a Mediterranean carnivore guild,”
Ecology and Evolution, vol. 10, no. 20, pp. 11408–11422,
2020.
[108] D. T. O’Mahony, “Multi-species visit rates to farmyards:
implications for biosecurity,”Veterinary Journal, vol. 203,
no. 1, pp. 126–128, 2015.
[109] M. D’Amico, S. Périquet, J. Román, and E. Revilla, “Road
avoidance responses determine the impact of heterogeneous
road networks at a regional scale,”Journal of Applied Ecology,
vol. 53, pp. 181–190, 2016.
[110] R. Pita, R. Morgado, F. Moreira, A. Mira, and P. Beja, “Roads,
forestry plantations and hedgerows affect badger occupancy
in intensive Mediterranean farmland,”Agriculture, Ecosys-
tems and Environment, vol. 289, Article ID 106721, 2020.
[111] B. A. Tolhurst, R. J. Delahay, N. J. Walker, A. I. Ward, and
T. J. Roper, “Behaviour of badgers (Meles meles) in farm
buildings: opportunities for the transmission of Mycobacte-
rium bovis to cattle?”Applied Animal Behaviour Science,
vol. 117, pp. 103–113, 2009.
[112] M. T. Main, R. A. Davis, D. Blake, H. Mills, T. S. Doherty, and
T. Dutta, “Human impact overrides bioclimatic drivers of red
fox home range size globally,”Diversity and Distributions,
vol. 26, no. 9, pp. 1083–1092, 2020.
[113] J. A. Barasona, K. C. VerCauteren, N. Saklou, C. Gortazar,
and J. Vicente, “Effectiveness of cattle operated bump gates and
exclusion fences in preventing ungulate multi-host sanitary
interaction,”Preventive Veterinary Medicine, vol. 111, no. 1-2,
pp. 42–50, 2013.
[114] N. A. Gilbert, J. D. J. Clare, J. L. Stenglein, and
B. Zuckerberg, “Abundance estimation of unmarked animals
based on camera-trap data,”Conservation Biology:The Journal
of the Society for Conservation Biology, vol. 35, pp. 88–100,
2020.
[115] E. R. Dougherty, D. P. Seidel, C. J. Carlson, O. Spiegel,
W. M. Getz, and K. Lafferty, “Going through the motions:
incorporating movement analyses into disease research,”
Ecology Letters, vol. 21, no. 4, pp. 588–604, 2018.
[116] A. E. Fine, C. A. Bolin, J. C. Gardiner, and J. B. Kaneene, “A
study of the persistence of mycobacterium bovis in the
environment under natural weather conditions in Michigan,
USA,”Veterinary Medicine International, vol. 26, Article ID
765430, 2011.
Transboundary and Emerging Diseases 19