https://orcid.org/0000-0003-0572-6268
Figures
Abstract
Mesocarnivores play a key role in ecosystem dynamics through the regulation of prey populations and are sensitive to environmental changes; thus, they are often considered good model organisms for conservation planning. However, data regarding the factors that influence the habitat use of threatened small wild felids such as the Andean tiger cat (Leopardus tigrinus pardinoides) are scarce. We conducted a two-year survey with 58 camera trap stations to evaluate the determinants of Andean tiger cat habitat use in three protected areas in the Middle Cauca, Colombia. We developed site occupancy models and found that Andean tiger cat habitat use increased with leaf litter depth at intermediate elevations and far from human settlements. Through conditional cooccurrence models, we found that Andean tiger cat habitat use was invariant to the presence of prey or potential intraguild competitors and killers/predators, but its detectability increased when they were present and detected. This suggests that Andean tiger cats may be more likely to be detected in sites with high prey availability. We found that Andean tiger cats preferred sites with deep leaf litter, which is a particular feature of cloud forests that provides suitable conditions for ambush hunting and hiding from intraguild enemies. Our results showed that Andean tiger cats avoided human settlements, which may minimize potential mortality risks in those areas. Moreover, the restricted use of middle elevations by Andean tiger cats suggested that they could be used as a sentinel species to track the effects of climate change since their suitable habitat is likely to be projected upward in elevation. Future conservation actions must be focused on identifying and mitigating human-related threats close to the Andean tiger cat habitat while preserving microhabitat conditions and the existing networks of protected areas.
Citation: Cepeda-Duque JC, Andrade-Ponce G, Montes-Rojas A, Rendón-Jaramillo U, López-Velasco V, Arango-Correa E, et al. (2023) Assessing microhabitat, landscape features and intraguild relationships in the occupancy of the enigmatic and threatened Andean tiger cat (Leopardus tigrinus pardinoides) in the cloud forests of northwestern Colombia. PLoS ONE 18(7): e0288247. https://doi.org/10.1371/journal.pone.0288247
Editor: Luca Nelli, University of Glasgow, UNITED KINGDOM
Received: January 20, 2023; Accepted: June 22, 2023; Published: July 10, 2023
Copyright: © 2023 Cepeda-Duque et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data and scripts from this study are available in the OSF repository (https://osf.io/hsnwu/).
Funding: The project was funded by the Small Wild Cat Conservation Foundation, the Mohamed bin Zayed Species Conservation Fund (#222528123), and the Research Grant “Proyecto Semilla” in the 2021-1 call for funding of the research category: Master’s and Doctoral student projects, provided by the Sciences Faculty of Los Andes University (INV-2020-104-2042). Tadeu de Oliveira received a research productivity fellowship from Universidae Estatal do Maranhão. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Small wild felids are considered keystone organisms for biodiversity maintenance, as they control the population growth of a wide range of small vertebrates that may include seed predators, invasive species, or disease vectors [1, 2]. The abundance and distribution of small wild felids are often mediated either by habitat quantity and quality or by interspecific interactions with prey and competitors, including humans [3]. Greater habitat structural complexity has been associated with increased hunting success and more effective avoidance of intraguild competitors/predators [4–6]. The spatiotemporal availability of prey represents a gradient of opportunities to supply metabolic requirements and successfully raise offspring [7]. Dominant carnivores may reduce the fitness of mesocarnivores through prey competition, intraguild predation/interspecific killing (nonnutritional benefit), or both [8–11]. However, partitioning resources (e.g., space, time, or food) is the most common way for small wild felids to coexist with superior competitors and killers/predators [8, 9]. Small wild felids are further constrained by human activities, such as deforestation and direct persecution for poultry depredation, and by the invasion of their habitats by domestic species, such as dogs [11, 12].
The Andean tiger cat (Leopardus tigrinus pardinoides) is a small mesocarnivore (weight: ca. 2.2 kg, length: 70-80 cm; Fig 1A) distributed in northwestern South America [13]. Tiger cats mainly hunt small mammals, reptiles, and birds [2]. Little is known about the basic natural history of these animals, especially the spatial and temporal interactions of Andean tiger cats with sympatric carnivores and their prey [14, 15]. They are found in small and large forest remnants [15, 16] and in pristine and disturbed areas [17]. The Andean tiger cat is categorized as vulnerable, and its populations are declining across its geographic range worldwide [17] and specifically in Colombia [18]. Andean tiger cat occurrence in Colombia is thought to be mainly constrained by both natural (inter-Andean valleys below 1500 m asl) and anthropogenic barriers (human settlements and roads) [19, 20]. The latter are specifically associated with the legacy of an intensive human footprint that has impacted the northwestern Andes region since the 1970s, a period of rampant population and economic growth in Colombia [21]. Conversely, it has been suggested that shielding the habitat of this forest-dwelling felid with integrative networks of public and private protected areas could promote the conditions for its persistence [16]. Nevertheless, important knowledge gaps regarding which drivers are relevant to infer the habitat use of the Andean tiger cat remain in the region and must be filled to strengthen conservation actions.
Andean tiger cat in its conservation stronghold, the Andean cloud forest of the Middle Cauca (A), and protected areas where this research was conducted. The Mesenia-Paramillo Nature Reserve – MPNR (B), the Cuchilla de San Juan Integrated Management District – CSJIMD (C) and the Campoalegre Soil Conservation District – CSCD (D). Represented are the main types of land covers and the Risaralda department, the target national land division unit. Layers of land cover types, elevation, and human settlements were obtained from their original sources, which are available online [38–40].
Camera traps are a well-stablished tool for studying relevant population parameters such as abundance, density, and occupancy of a diverse range of cryptic or elusive species [22]. In this context, occupancy has been used as a robust method for determining the role of habitat features, interspecific interactions, and human activities in the occurrence of poorly understood species [23]. In Colombia, there have been a growing number of studies assessing the effects of different abiotic and biotic predictors on the occupancy of small wild felids, such as ocelots (Leopardus pardalis), jaguarundis (Herpailurus yagouaroundi), margays (L. wiedii), and Andean tiger cats [15, 16, 24–26]. Only two occupancy studies involving the elusive Andean tiger cat in mountain cloud forests have been carried out using single and multispecies approaches. Notably, the conditional effect of food and risk availability, at least in some form of cooccurrence between Andean tiger cats and both small mammals and other sympatric carnivores, respectively, is still overlooked. However, it could be a relevant driver of its habitat use as seen in other species of small wild felids [27–29]. We used camera traps in conjunction with occupancy and cooccurrence modeling to determine how the habitat use of Andean tiger cats is modulated by different habitat conditions and by potential competitors, intraguild killers/predators and prey. To answer these questions, we surveyed three protected cloud forests comprising a key location in the Middle Cauca River Basin (hereafter Middle Cauca) in Colombia.
Hypotheses
Increased survey effort has been linked to higher detectability of other neotropical small wild felids [5, 30]. Detectability may also vary among camera brands due to differences in the camera features, such as the time delay of the passive infrared (PIR) sensor [31]. It has been suggested that the use of cameras with shorter PIR time delays increases the probability of detecting fast-movement animals [22]. Thus, we expected that Andean tiger cat detectability would increase with a greater sampling effort [5] and shorter PIR time delays.
The structural complexity of the habitat is the degree of both horizontal and vertical spatial heterogeneity present in the forest structure [32]. We expected that the habitat use of Andean tiger cats would increase in sites with more complexity in the vertical and horizontal forest structure dimensions since they represent greater availability of prey and refuge [4, 6]. We also expected Andean tiger cat habitat use and detectability to decrease in sites with a steeper slope given its related energetic investment in movement [33, 34].
Since they are forest-dwelling felids, we expected that Andean tiger cats would negatively respond to forest fragmentation and landscape heterogeneity but positively respond to forest amount and patch cohesion [6, 16]. We expected that a greater human footprint legacy in the study areas would negatively affect the habitat use of Andean tiger cats, as has been previously found for other neotropical mammals [35]. Proximity to human settlements and roads relates to detrimental activities for Andean tiger cats, such as unregulated tourism, roadkill, illegal hunting, and encounters with dogs, thus, we expected greater habitat use for the species in sites distant from these features [20, 36].
Small mammals are a pivotal energy source for tiger cats, representing nearly half the biomass of their prey [2]. We expected that Andean tiger cats would use sites with greater small mammal occupancy to increase hunting success [3, 27]. We also expected that Andean tiger cats would avoid sites where other carnivores of greater body mass might be present to reduce potential encounters that could promote intraguild killing or competition [37]. Theory predicts that in cases where resources are abundant (e.g., prey), dominant carnivores tend to exclude subordinate carnivores, whereas if resources are scarce, the dominant carnivore is excluded, but the subordinate thrives through specific adaptations [10].
Materials and methods
Ethics statement
We used noninvasive methods for data gathering, and hence, approval from institutional or animal ethics committees was not needed. Our field team was verbally authorized by the local communities and specific landowners to survey the study areas.
Study area
The camera trap survey was conducted in three different protected areas from the Middle Cauca, which is a river basin located between the Western and Central Cordilleras in the Northern Andes of Colombia. The Mesenia-Paramillo Nature Reserve (MPNR) and Cuchilla del San Juan Integrated Management District (CSJIMD) are located on the eastern slope of the Western Cordillera (Fig 1B, 1C). The Campoalegre Soil Conservation District (CSCD) is located on the western slope of the Central Cordillera (Fig 1D) and is the largest reserve surveyed with an area of 211.3 km2, followed by CSJIMD with 117.9 km2 and MPNR with 35 km2. The average annual temperature in the study areas ranges from approximately 15–23°C, and the average annual rainfall is 1,900 to 3,100 mm [41]. The most representative native vegetation cover in the region is the northwestern Andean cloud forest. These forests present stunted and gnarled canopies with high cover of epiphytes and an understory permanently covered by abundant leaf litter, adventitious roots, and fallen trunks [42]. In the surrounding zones of the protected areas, it is common to find large expanses of pastures for cattle ranching, eucalyptus and pine plantations, and abandoned plots with secondary vegetation and crops (mainly coffee and avocado) [15, 43]. There is unmanaged tourism as well as illegal logging and hunting both within and around CSCD and CSJIMD, with both protected areas facing a lack of personnel to control this kind of human activity. In contrast, the MPNR is the only protected area that has trained forest rangers that regulate tourist entry and prevent hunting and logging within its limits. Consequently, the MPNR has a greater proportion of cloud forests in contrast to other existing covers in the studied landscape (80%) compared to CSJIMD (67%) and CSCD (66%), which are currently under expansion due to ongoing restoration efforts.
Sampling design
We set 21 camera traps in MPNR, 22 in CSJIMD and 23 in CSCD inside their native forest covers, which were arranged in a 1 km2 grid with an average spacing of 962 ± 84 m following previous study designs on space use by neotropical small felids [30, 37, 44, 45]. Excluding the damaged or stolen cameras, we were able to collect data from 58 sampling stations for subsequent analysis (Table 1). Prior to installation, we removed the vegetation within a strip in front of the camera’s detection field that was 1 m wide and 5 m long to avoid ‘ghost photos’ caused by wind-blown vegetation [46]. At each station, a single camera was attached at 30 cm high along natural trails favored by felids, without the use of lures [36]. We used several brands and models of passive infrared camera traps for each protected area (APEMAN ® Trail Cam H45, Blaze Video ® A262 Animal Cam, Browning ® Strike Force Apex and Strike Force Gen 5, Bushnell ® Trophy Cam HD, Trophy Cam 119717CW, Trophy Cam Agressor, Reconyx ® Hyperfire, and Victure ® HC300). We set the cameras to take sequences of three photos with the minimum delay possible (0–5 s depending on the brand). The sampling effort and the number of cameras varied between reserves (Table 1), and the average sampling duration for each camera location was 130 ± 24 days. The MPNR sampling period was from August 2020 to February 2022, the CSJIMD was from December 2021 to June 2022, and the CSCD was from February to November 2021.
Covariates.
We included covariates on the microhabitat and landscape scales based on a priori biological hypotheses of their effects on the habitat use of Andean tiger cats (S1 Table). We quantified the number of days in which the cameras were active to estimate the potential effects of the survey effort on Andean tiger cat detectability [47]. To minimize the biases associated with the variation in the camera trap functionality, we included the minimum PIR time delay of each brand as a categorical covariate conditioning detectability [48].
Landscape scale.
We describe landscape structure using a 30 m resolution land cover classification product that fixes the error induced by cloudiness in the study area [38]. We created buffer areas with a 500 m radius using the location of each camera trap as the area centroid to quantify class-derived landscape metrics (S1 Table). Even though 500 m seems to be a small buffer radius for making landscape-scale inferences in the occupancy of small wild felids under some environmental contexts [37]. The area defined by this buffer is projected in a two-dimensional instead of a three-dimensional surface. This issue is usually ignored in studies involving mobile organisms immersed in landscapes with vertical relief [34]. In particular, felids may have greater constraints on their habitat use in three-dimensional (e.g., valleys and hills in the highland cloud forests) rather than two-dimensional (e.g., plains and lowland rainforests) environments, as the former have a greater surface area [49]. Thus, according to the broken topography inside the 500 m buffer of each site surveyed, the minimum home range size (0.9 km2) reported for lowland members of the tigrinus complex [37] and recent evidence on the occupancy of other small wild felids elsewhere [6, 50, 51]. We might expect that a 500 m buffer would be an accurate spatial scale for modeling the effects of landscape predictors on the habitat use of the Andean tiger cat, at least in our study system.
We estimated four landscape metrics reflecting landscape heterogeneity (Shannon’s index), forest amount, edge density and forest patch cohesion index using the LecoS plugin of QGIS 3.16 software [52]. We constructed a Pearson’s correlation matrix to inspect for collinearity issues (r > |0.65|) among landscape structure covariates. We noted that forest amount showed a strong positive correlation with patch cohesion index (r = 0.83) and a strong negative correlation with landscape heterogeneity (r = -0.92) and edge density (r = -0.86). Given the correlation structure we found in our landscape metrics, we decided to summarize their effects with a single landscape structure covariate (LS) after applying principal component analysis [16]. According to Kaiser-Kupferman’s criterion, the most important linear combination was determined using the first principal component, as it presented eigenvalues > 1 [4, 16, 53]. The negative values of the first principal component represented a positive gradient of amount and cohesion of forest patches, while the positive values reflected greater habitat heterogeneity and forest edge density. Further details on this covariate are provided in the supplementary material (S1 Fig).
To determine whether the habitat use of Andean tiger cats increased monotonically with increasing elevation or if it decreased when elevation peaked value, we also tested the linear and quadratic effects of elevation [35]. The Legacy Human Footprint Index (LHFI) combines the effects of land use intensity and time to intervention in the landscape at a resolution of 300 m and we obtained the average value for each station with 500 m buffers [21]. From cartographic information provided by the Departamento Administrativo Nacional de Estadistica – DANE GeoPortal (https://geoportal.dane.gov.co/), we calculate the Euclidean distance between each sampling station and the nearest roads and human settlements using the QGIS 3.16 plugin NNJoin.
Site-scale.
As an indicator of understory vertical complexity, leaf litter depth was determined using a measuring tape and a rod that was slowly driven into the soil at a 90° angle until it reached the rhizosphere [54, 55]. To quantify the horizontal microhabitat structure, we used a 50 x 50 cm wooden square composed of a nylon net with 100 grids to measure the proportion of canopy, herbaceous, and leaf litter cover at each station [56]. Canopy height is a covariate that often represents the vertical complexity in the canopy [57], and we measured it with a Nikon Prostaff 550 laser. We measured the slope with a Suunto PM-5/360 PC clinometer as an indicator of the microhabitat topographic inclination. For each sampling station, we recorded measurements at the point where the camera was installed and at 5 m in each cardinal direction [6]. We averaged the measurements obtained for each microhabitat covariate and used the resulting value as input for subsequent modeling [6].
We scaled all the numerical covariates to facilitate direct comparison between model estimates and improve parameter estimation through the following formula: raw data - average/standard deviation [47, 58]. Subsequently, to minimize bias in the modeling, we removed numerical covariates with multicollinearity or uniformity problems (S2 Fig), which were detected via graphical inspection and a correlation Pearson coefficient [4, 47]. We found a strong positive correlation between elevation and distance to roads, as well as a strong negative correlation between herbaceous cover and leaf litter cover and between LHFI and distance to human settlements (S2 Fig). The final set of covariates used in the occupancy models for the Andean tiger cat included landscape structure, protected area, canopy height, canopy cover, leaf litter cover, leaf litter depth, slope, distance to human settlements and elevation. For the modeling of detection probability, the final set of covariates included herbaceous cover, slope, detection delay of the passive infrared sensor and survey effort of the camera traps.
Occupancy modeling
We considered the occasions to be independent and pooled the pictures of each station over a 24 h period. For the design of the detection/nondetection histories, we collapsed the survey into weekly periods to reduce zero-inflation in the matrix and improve model convergence [35]. In the occupancy modeling, we defined each camera-trap station as the sampling unit [23]. Since the movement patterns of the target species are larger relative to the area covered by the detection field of our sampling units, we interpreted the results of our occupancy and detection estimates as habitat use and intensity of use, respectively [23].
To test the relationships between the habitat use of the Andean tiger cat and the selected covariates, we built single-species-single-season occupancy models, establishing two hierarchically linked processes: ecological and observational. The first represents the occupancy probability (ψij), defined as the probability of the species occurring or using a site during a sampling season [23]. This parameter is represented as a latent variable with Bernoulli distribution and descriptor function of the form zi∼Bernoulli (ψ), where Ζ is the actual occupancy value at each site i with probability of occurrence ψ [59]. The observation process represents the imperfect detection of the species and is represented by the probability function pij |Ζi∼Bernoulli(ψi*pij), where pij is the observation at each site i at each visit j if site i is occupied [23]. These parameters can be modeled as a function of covariates in a logistic regression of the form logit (ψi) = β0 + β1xi0 for occupancy or logit (pi) = α0 + α1xi for detection [59], where β0 and α0 are intercept values from which the model starts and xi is the covariate(s) measured for each station i.
To determine the most influential covariates affecting the occupancy of Andean tiger cats, we used a Bayesian modeling framework [60]. We fit each model with uninformative priors and ran three Markov Chain Monte Carlo (MCMC) samplers of 30,000 iterations, from which the first half was discarded as burn-in. Additionally, we calibrated the models with an acceptance sample (adapt_delta) of 0.99 for the adaptation period of the MCMC sampling [61]. A value of 0.99 for adapt_delta means that the algorithm will explore the parameter space more thoroughly during adaptation, which can lead to better quality posterior samples [62]. We used the expected log pointwise predictive density (elpd) to compare and rank the models, as it is a measure of model predictive accuracy [58]. The elpd reflects how the model predicts future data, assuming that it came from the same distribution as the observed data [63]. To assess model support relative to the top model, we also computed the pairwise differences in elpd (Δelpd) between models and the top model. We performed model selection using the leaving one out (LOOIC) cross-validation information criterion on the estimates of the posterior distribution [63]. LOOIC removes one data point at a time from the dataset, fits the model to the remaining data, and then uses the model to predict the value of the removed data point. This process is repeated for each data point to obtain an estimate of the model’s predictive accuracy [64]. We considered a model to have less support than the top model if its absolute difference in elpd was greater than the standard error of that difference [63]. We then performed sequential-by-selection submodeling using the p-first approach [65]. In the first stage, we modeled the detection process by fixing the individual effect of its associated covariates while keeping ψ constant [35]. In the second stage, we modeled ψ by individually fixing its corresponding covariates while fixing the model for p with its most relevant covariate, as previously determined [65].
To test for unaccounted variation in the relationship between the selected covariates at each study location (e.g., MPNR, CSCD, and CSJIMD) and the occupancy of Andean tiger cats, we included the former as a random effect in the same modeling structure. However, this model did not perform better than models without the random effect (S2 Table). Recent evidence suggests that, when models include a random effect with fewer than five factorial levels, it is convenient to return to a model with fixed effects to prevent inflated variances [66]. We inferred the influence of the supported covariates on the top-ranked occupancy and detection models by inspecting the β coefficients and their 95% credibility intervals [60]. We looked for positive or negative trends in the relevant covariates of each model by determining if the credibility intervals had non-zero overlap and positive or negative values, respectively [4]. We assessed model convergence by inspecting the Gelman-Rubin Statistic (R-hat); R-hat<1.1 indicates that the models are acceptable [67]. We tested the models for overdispersion via a posterior predictive check using the MacKenzie-Bailey test for the most parsimonious models [23]. We computed this test for the real data and for the simulated dataset in each iteration of the MCMC sampling [60]. If the model showed a good fit, the proportion of draws in which the simulated estimator is greater than the real estimator should be greater than 0.05 (Bayesian P value > 0.05) [59, 60]. Bayesian occupancy models were built using the ubms package [60] version 1.1.0 in R software. To assess spatial autocorrelation, we applied Moran’s I index to the residual component of the top-ranked models for the occupancy of the Andean tiger cat and visually inspected spline correlograms for correlation patterns across the distance unit [68]. We obtained the model residuals for the occupancy component defined by Wright et al. [69]. The spatial autocorrelation analysis was performed using the nfc package [70] version 1. 3. 0.
Cooccurrence modeling
To model the effect of the occupancy and detection of other species on the occupancy and detection of Andean tiger cats, we implemented conditional cooccurrence models [71] using the ΨBa parameterization (Table 2). This is an asymmetric scenario established between two interacting species, where the occupancy and detectability of one species is assumed to be conditional on the occupancy and detectability of another species given some a priori criteria [10, 71]. To assess whether Andean tiger cat occupancy and detection were related to that of prey, we assigned all detected small mammals as the dominant species (ΨA) and the Andean tiger cat as the subordinate species (ΨB). In the context of the spatial cooccurrence between Andean tiger cats and sympatric carnivores, we designated the former as the subordinate (ΨB) due to its small body size, restricted geographic range, and previous evidence of competitive exclusion with sympatric felids [44, 72].
The dominant intraguild members were identified according to body mass differences, trophic overlap, and the sample size obtained to improve parameter estimation of the data-hungry cooccurrence models [67, 72–74]. The cooccurrence of Andean tiger cats against other potentially dominant intraguild members was analyzed in two different ways. First, we analyzed the cooccurrence of Andean tiger cats conditional on tayras as potential intraguild competitors. In some areas, tayras exhibited high dietary overlap with tiger cats and similar body mass ratios [9, 74] and, in our context, had an appropriate sample size compared to any other intraguild species to model conditional parameters. Second, we analyzed the cooccurrence of Andean tiger cats conditional on a category of potential intraguild killers/predators (e.g., pumas, ocelots, dogs, and Andean bears). Overall, the selected species exceed by far the body mass ratio required to induce a safety kill over an Andean tiger cat [9] but where combined together due to their low sample size to improve model convergence. We recognize that this coarse taxonomic resolution hinders us from discriminating fine-scale differences in the intensity of the cooccurrence outcome for intraguild killers/predators. Nevertheless, the selected species have the potential to limit the occupancy of Andean tiger cats either by direct predation or killing to avoid encounters during the use of a particular resource [37, 74]. Therefore, irrespective of the kind of antagonism, we might expect a nonrandom response in the use of a given site by Andean tiger cats as a function of the presence of other carnivore species.
Once we defined the conditional state of each taxon, we built models assuming that the occupancy of ΨB was (ΨBA ≠ ΨBa) or was not (ΨBA = ΨBa) affected by the presence of ΨA. Furthermore, we modeled all detection parameters within both independent (pA = rA, pB = rBA = rBa) and dependent (pA ≠ rA, pB ≠ rBA ≠ rBa) cooccurrence states. We performed model selection based on the Akaike information criterion (AIC) corrected for small samples (AICc), and the best models were retained according to a conventional parsimony threshold (ΔAICc < 2) [75].
As with occupancy models, transformed β coefficients were used to infer either positive or negative relationships between covariates and conditional parameters [4]. Covariates from the best single occupancy models were included in the cooccurrence models to account for habitat preferences in imperfect detection. If the best models retained a conditional parameter structure (ΨBA ≠ ΨBa or rBA ≠ rBa), we estimated the species interaction factor (SIF) and species detection factor (DIF). The SIF (Φ) represents the probability that two species of interest cooccur nonrandomly in the landscape and is calculated as follows: Φ = (ΨA ΨBA)/ΨA (ΨA ΨBA + (1- ΨA) ΨBa [76]. If φ > 1, the two species tend to coexist more than expected by chance; if φ < 1, the species tend to evade one another; and if φ = 1, the species exist independently, and there is no evident pattern of evasion or aggregation [71]. The DIF (δ) represents the codetectability, in other words, the probability of two species being nonrandomly codetected, and is calculated as follows: δ = rBA/(rA rBa). If δ ≠ 1, then the detection probabilities of both ΨB and ΨA are conditional on the other, as in the SIF. Conditional cooccurrence models were built using the wiqid package [77] version 0. 3. 3. through R Software.
Results
After 10,689 trap/nights in a sampling area of 74.1 km2, we obtained 129 independent detections of Andean tiger cats under a 7-day occasion framework at 38 sites (54% - MPNR, 36% - CSJIMD, 10% - CSCD), 118 independent detections of tayras at 39 sites, 73 independent detections of potential intraguild killers/predators at 27 sites (61% - MPNR, 27% - CSJIMD, 12% - CSCD) and 463 independent detections of prey at 49 sites (55% - MPNR, 22% - CSJIMD, 23% - CSCD). Of the records obtained for intraguild killers/predators, 33% corresponded to Andean bears (9 sites), 27% to dogs (9 sites), 26% to pumas (11 sites), and the remaining 14% to ocelots (8 sites). We found that 77%, 76%, and 72% of the sites with Andean tiger cat presence also manifested presence of tayras, prey, and potential intraguild killers/predators, respectively.
Occupancy modeling
Hereafter, we present the results obtained for the three protected areas together, yet separate estimates can be found in the supplementary material (S2 Table). The null model showed that there was a 15% chance of obtaining a weekly detection of the Andean tiger cat if present (p = 0.15, SD = 0.02, CI = 0.12–0.17) and an average occupancy of 68% (Ψ = 0.68, SD = 0.06, CI = 0.55–0.80). The top-ranked model showed that there was 13% chance of obtaining a weekly detection of the Andean tiger cat if present (p = 0.13, SD = 0.02, CI = 0.09–0.18) and an average occupancy of 73% (Ψ = 0.73, SD = 0.08, CI = 0.56–0.88).
The top-ranked model for Andean tiger cat detection included the effects of the camera trap’s PIR time delay (Table 3). Detection was higher for the species at sites that had cameras with a PIR time delay ≤ 0.6 seconds (Fig 2A). The top-ranked models for Andean tiger cat occupancy included the effect of leaf litter depth, quadratic elevation, and distance to human settlements (Table 3). Leaf litter depth had a positive and marked effect on the occupancy of Andean tiger cats (β = 1.38, SD = 0.55, CI = 0.41–2.57, R-hat = 1) (Fig 2B). The occupancy of Andean tiger cats reached a maximum between 2,000 and 3,000 m asl, rapidly decreasing above or below this threshold (β = -0.99, SD = 0.41, CI = -1.85 – -0.21, R-hat = 1) (Fig 2C). Distance to human settlements had a positive but weak effect on the occupancy of Andean tiger cats (β = 0.81, CI = 0.0372–1.71, R-hat = 1) (Fig 2D). Goodness-of-fit tests performed to identify the best models showed a marginal fit and signs of underdispersion (S3 Fig). Moreover, we did not find any sign of spatial autocorrelation in Andean tiger cat occupancy (S4 Fig).
The detection of the Andean tiger cat was predicted against the time delay in seconds (A) of the passive infrared sensor. The occupancy of the Andean tiger cat was predicted against the covariates representing (B) leaf litter depth, (C) quadratic elevation, and (D) distance to human settlements. The orange shading represents the 95% credibility intervals. PIR is defined as the passive infrared sensor of the camera trap brands.
Cooccurrence modeling
The cooccurrence modeling procedure revealed a random cooccurrence pattern of Andean tiger cats with their prey, tayras and intraguild killers/predators (S3 Table). However, the Andean tiger cat detection probability was higher when prey was present and detected (rBA = 0.14, CI = 0.11–0.18) than when it was undetected (rBa = 0.10, CI = 0.08–0.13). This was also the case when comparing the Andean tiger cat detection probability with that of intraguild competitors (Fig 3). Andean tiger cat detectability was lower when tayras were present but undetected (rBa = 0.11, CI = 0.09–0.14) than when they were present and detected (tayras: rBA = 0.29, CI = 0.21–0.39). In addition, there was no evidence to support that Andean tiger cat occupancy or detectability was conditional on that of its intraguild killers/predators (S3 Table). However, the detectability of intraguild killers/predators increased at sites with presence of both Andean tiger cats and intraguild killers/predators (rA = 0.09, CI = 0.07–0.12), compared to sites where only intraguild killers/predators were present (pA = 0.07, CI =0.03–0.13).
Coefficients of the parameters estimated by the conditional cooccurrence models of Andean tiger cats against (A) potential prey, (B) tayras, and (C) potential intraguild killers/predators. Vertical bars represent 95% confidence intervals. Parameter abbreviations: pA; detection probability of intraguild competitors/killers/predators and prey given Andean tiger cat absence, pB; detection probability of Andean tiger cats given the absence of intraguild competitors, killers/predators and prey, ΨA; occupancy probability of intraguild competitors, killers/predators and prey, ΨBa; occupancy probability of Andean tiger cats given intraguild competitors, killers/predators and prey absence, ΨBA; occupancy probability of Andean tiger cats given intraguild competitors, killers/predators and prey presence, rA; detection probability of intraguild competitors, killers/predators and prey given the presence of both species, rBa; detection probability of Andean tiger cats given the presence of both species and the nondetection of intraguild competitors, killers/predators and prey, and rBA; detection probability of Andean tiger cats given the presence of both species and the detection of intraguild competitors, killers/predators and prey.
The SIF was 1.0 when comparing the Andean tiger cat occupancy conditional on the presence of either prey, tayras, or intraguild killers/predators, suggesting that Andean tiger cats are using the sites regardless of their presence. Nevertheless, the DIF was positive when comparing Andean tiger cat detection conditional on the detection of prey (δ = 1.41), and tayras (δ = 3.10). None of the relevant covariates for predicting the occupancy of Andean tiger cats could explain the patterns of cooccurrence with their prey, tayras, and killers/predators (S3 Table).
Discussion
This is the first detailed study assessing the habitat requirements for Andean tiger cats in highland ecosystems across its geographic range. The habitat use of the Andean tiger cat was mainly affected by leaf litter depth, elevation, and, to a lesser extent, the distance from human settlements. We did not find support for conditional cooccurrence between Andean tiger cats and either prey or intraguild competitors and killers/predators, but we did find that their detectability increased when both prey and tayras were present and detected. Overall, we found that Andean tiger cats occurred in more than half of the sampling units in the studied protected areas but were imperfectly detected in the surveys. Our results agreed with some previous estimates of occupancy and detection for the tiger cat species complex from the cloud forests of the Western Cordillera to the Atlantic Forest and the Caatinga drylands (S4 Table).
The covariate most relevant for the Andean tiger cat detectability was the PIR time delay, with PIR = 0.6 being the shortest found in the camera brands used. This finding aligns with both experimental and empirical evidence, which suggests that camera brands with short time delays are optimal to detect carnivores, as they tend to be fast-moving species [48, 67]. Thus, we suggest using faster PIR camera brands in future surveys to improve Andean tiger cat detectability. Contrary to the trends observed for other neotropical small felids, we found that the camera trap effort showed no relevant effect on the detectability of the Andean tiger cat [5, 30, 61].
We found a nonlinear response of Andean tiger cats to elevation and a greater use of sites located at approximately 2,000 to 3,000 m asl. Elevation was previously reported as an influential driver of habitat suitability for Andean tiger cats in Colombia, with more suitable habitats at higher elevations [78]. We observed a greater use of sites at middle elevations, which could be related to prey availability likely represented by the assemblages of small mammals [79, 80]. In Andean cloud forests, the diversity of small mammals tends to be greater at middle elevations, as there is greater species turnover and productivity [81]. Moreover, we found a striking correlation of the sampling sites more distant to roads with high elevations, which may also affect the habitat use of the Andean tiger cat. In the Northern Andes of Colombia, areas at high elevations are less accessible than those at low elevations which might reduce the pressures related to encounters with humans, domestic animals, or vehicles [21, 82]. The constrained access to forested areas at higher elevations may also discourage forest loss while improving habitat conditions for small felids, as has been found in mountain forests of Sumatra [83].
Meanwhile, climate change could affect range-restricted species at middle elevations because Andean ecosystems are moving to higher elevations as an effect of increased environmental warming [84, 85]. Under environmental warming, cloud forests are expected to suffer an altitudinal attrition of their cloud formations, which will compromise the integrity of their bottom-up processes [85]. As a consequence, species with restricted niches will be more susceptible to range shifts and contractions due to resource depletion and increased competition against migrant species from the lowlands [85]. Notably, by the 2050s, climatic forecasts under the “business as usual scenario” indicate that the current members of the tigrinus complex will suffer the greatest range contraction of any neotropical small felid, losing 86.2% of its actual range [86]. Regardless, the authors suggest that the conflictive taxonomy of the tigrinus complex could prevent any conclusion on their imminent risks and urge a fair diagnosis between the highland and lowland forms [86]. More research is needed to confirm whether the geographic range of the Andean tiger cat would suffer a similar or different fate due to the warming of the inter-Andean valleys from northwestern Colombia.
Contrary to our expectations, the landscape structure aspects we measured within the protected areas did not influence Andean tiger cat habitat use. It has been suggested that greater landscape heterogeneity increases prey availability and, consequently, the habitat use of small wild felids [87, 88]. However, the flexible use of disturbed areas by these organisms depends on the landscape context since the availability of conserved neighboring habitats is required [16]. Our results aligned with a study conducted in the highlands of Ecuador, which found no effects of habitat loss or fragmentation on the habitat use of pumas, Andean bears, Andean foxes (Lycalopex culpaeus), and strip hog-nosed skunks (Conepatus semistriatus). This study suggested that, regardless of anthropogenic landscape transformation, the high Andes ecosystems are already heterogeneous, comprising forests, shrubs, and grasslands (paramos), which could explain the widespread carnivore occurrence in this landscape [89]. Nonetheless, plasticity in trophic preferences may enable some species to adapt more easily than others in heterogeneous landscapes [16]. For example, omnivorous species (e.g., foxes, Andean bears) are able to use more resources than strict carnivorous species, which is favorable in terms of increased permeability to the existing conditions in the matrix [16].
We found a positive effect between the habitat use of Andean tiger cats and the distance to human settlements, which was consistent with the trends found for lowland tiger cat populations [90]. However, in Colombia, Andean tiger cats have been reported near human settlements and even in peri-urban areas in major cities [78, 91–93]. These observations are commonly based on a few indirect or bycatch records of camera traps and tracks, with no explicit information relating Andean tiger cat populations with proximity to urban areas. Accounting for imperfect detection, we confirmed that Andean tiger cats negatively respond to human settlements, at least for our study sites, which belonged to a central portion of one of the highest populated and economically active regions in Colombia, the Middle Cauca. Therefore, any prior suggestions of Andean tiger cat tolerance to occupy more anthropized environments should be interpreted with caution. A key issue is that our models were based on the overall conditions of the three protected areas sampled, so we suggest precluding any extrapolation of these results to other locations with different conditions, either inside or outside the region. There is evidence suggesting that near human settlements, the mortality risk for tiger cats increases due to interactions with humans in the form of vehicle collisions [20], retaliation in poultry depredation and illegal fur trade [17]. Close to human settlements, light and noise pollution may also affect the space use of wild felids by introducing fear of encounters with humans [94].
We found a strong positive correlation between the sampling sites closer to human settlements and the LHFI. Cloud forests of the middle Cauca proximal to human settlements are known to be differentiated from more distant forests by having a more intense land use change and a longer history of interventions [82]. In particular, the economic balance of industries such as cattle, coffee, and avocado farming relies mainly on the presence of roads and the proximity to intermediate and large population zones [95]. This suggests that Andean tiger cats are negatively impacted by other synergic anthropogenic factors not considered here that may prevail in the region, which also requires expanding our current modeling framework. The occupancy of other small felids in South America has been found to decrease with a greater abundance of livestock and dogs [29, 50]. Including the individual effect of such factors as potential drivers of Andean tiger cat habitat use would improve our discernment of its response to human disturbance.
We found that Andean tiger cat habitat use was positively related to leaf litter depth, which is generally indicative of a mature healthy cloud forest. Mature cloud forest understories have abundant leaf litter and prominent adventitious roots coupled with slow decomposition rates and low nutrient concentrations, respectively [81]. The vertical structural complexity of this microhabitat may promote the habitat use of Andean tiger cats by providing them food and shelter [96]. Deep leaf litter can provide an ideal ambush site, as small mammals tend to seek these conditions for thermoregulation, feeding, or nesting [55, 96]. A recent study showed that leaf litter depth had a strong positive effect on the habitat use of two sympatric species of short-tailed opossums (Monodelphis spp.) in the eastern Amazon [55]. Deep leaf litter was also positively related to the abundance of three terrestrial (Thomasomys laniger, Neomicroxus bogotensis, Cryptotis cf. tamensis), and two arboreal (Marmosops caucae, Rhipidomys fulviventer) small nonvolant mammals in an oak forest in the Eastern Andes of Colombia [96]. We also postulate that the vertical complexity in the understory may provide cavities and galleries of various sizes that can be used by Andean tiger cats and other small carnivores to avoid intraguild killing. Slope showed no effect on Andean tiger cat habitat use and detectability. Carnivores such as Andean bears and pumas tend to prefer sites with little slope, such as ridges or streams, to facilitate their movement [97, 98]. We postulate that Andean tiger cats do not deviate from this overall pattern of carnivore habitat use in rugged terrain.
Conditional cooccurrence models showed that Andean tiger cat habitat use was not affected by the presence of prey, but Andean tiger cat detectability increased when prey was present and detected. In other words, Andean tiger cats used areas that had more prey more frequently. A plausible explanation for this pattern is that Andean tiger cats use sites with greater intensity of use by prey to increase hunting encounters, which is expected for an opportunistic predator [3]. In temperate ecosystems, small carnivores face a metabolic trade-off between maximizing food intake and avoiding heat loss, as environmental temperatures tend to reach extremely low limits [99]. Preliminary evidence on the activity patterns of Andean tiger cats in the surveyed localities also supported this relationship, since there was a marked overlap between Andean tiger cat and small mammal activity.
No support for spatial segregation was found in the habitat use of Andean tiger cats regarding the presence or absence of tayras and potential intraguild killers/predators. Nevertheless, the abundance of the cat’s de facto potential killers/predators was low and hence would not be expected to lead to meaningful behavioral changes [36, 100]. Tayras, the most abundant sympatric carnivore, exhibited, in fact, a greater likelihood of being a competitor rather than an actual killer of Andean tiger cats, as they showed high dietary overlap with small mammals but fell below the minimum body mass required for a “safety” intraguild killing [9, 74]. Thus, it is expected that habitat structure and temporal segregation would act as the modulating factors driving coexistence in this highland environment, as intraguild killers/predators are rare and potential competitors are mostly diurnal and do not depend only on meat to survive. Additionally, being highly territorial, felids use a complex system of scent marking with their urine and feces to delimit areas of exclusive and common use [101]. Hence, we postulate that Andean tiger cats could evade confrontation by recognizing areas recently marked by other sympatric carnivores, but this hypothesis needs to be further explored.
In lowland neotropical forests, ocelots tend to suppress populations of smaller mesocarnivore species, such as tiger cats, through the ocelot/pardalis effect [37, 100]. When ocelot populations reach numbers above a certain threshold, small wild felids decline either due to intraguild killing or competitive interference [27, 36, 37, 100]. Nonetheless, we suggest that in cloud forests from the Middle Cauca, Andean tiger cats may escape from the ocelot effect given the rarity of this mesopredator at higher elevations [5, 25]. This spatial segregation has been found on a horizontal rather than a vertical scale across the Amazon basin for northern tiger cats. In the Guianan savanna shield, northern tiger cats were segregated and used more open and drier habitats even when ocelots occurred at low numbers (0.029 ind./km2) and were restricted to some gallery forests and forest patches within savannas [100].
Conservation implications
The results of this study may help to develop conservation strategies for Andean tiger cat populations thriving in the vanishing cloud forests from the Middle Cauca because our study design represented nearly 67% of their presumed elevational range (1,500–4,800 m asl) in Colombia [19]. A common paradigm when focusing felid research and conservation efforts is to prioritize large species over the needs and threats of smaller ones because of their charisma and cultural value to society [33, 102]. For a long time, the highland forms of the tigrinus species complex were not the global, national, and local focus of conservationists because they are small, elusive cats living in harsh conditions, making them difficult to study [103]. Mountain forests worldwide share a common pattern of greater landscape degradation in the lowlands [33]. Most of the degraded lands in the Northern Andes of Colombia are in the lowland dry forests of the Cauca and Magdalena basins, but this trend has been changing since the last century [104]. The increased landscape degradation upward from the mountains together with climate change could lead to important contractions of the Andean tiger cat’s remaining habitat. We need to examine whether Andean tiger cats may experience vertical migrations across their geographic range in the future to resist the rise in temperature and habitat fragmentation. We highlight the potential of this highland felid to be a sentinel species on the impacts of climate and land-use change on the Andean cloud forest, as has been previously proposed for other small carnivores [105].
Supporting species persistence in forest-dwelling predators necessarily involves granting suitable forest conditions on the long run [106]. Endangered Pacific stone martens (Martes caurina), for instance, specialize in seeking logs and hollows at live trees in forests to establish resting and denning sites, with a particular tendency to use structures of trees with larger diameters [106]. Our results demonstrate that leading the leaf litter cover to gradually accumulate may improve the quality of the habitat available for the Andean tiger cat, which could be considered in future forest management actions interested in benefiting this threatened felid. Indeed, deep leaf litters are regarded as a key trait of cloud forests worldwide [42], providing important resources for a diversity of taxa [55, 96].
The highest habitat use for Andean tiger cats was obtained in the MPNR, a nature reserve established in 2007 by a local NGO after the end of the armed conflict in the Northwestern Andes [107]. The MPNR protects a key corridor for the Tatamá-Orquídeas national park system and is considered a biodiversity refuge for multiple biological groups [107]. This protected area is currently expanding to new land, and there is a cloud forest restoration program that would be key for the long-term conservation of Andean tiger cats. Nearly 10 km south of the MPNR, the CSJIMD has not suffered the same fate as the MPNR because its administrative policies allow the exploitation of forest resources within the reserve. The lack of government restrictions to prevent the transformation of the Andean cloud forest to land for cultivating avocado crops and cattle pastures in the highlands of the CSJIMD may lead to the disappearance of this important biological corridor. CSJIMD was the site with the second highest habitat use estimates for Andean tiger cats; therefore, it is necessary to prioritize the implementation of more restrictive policies to curb the deforestation that this protected area is currently suffering. To guarantee the long-term conservation of Andean tiger cat populations in this part of the Middle Cauca, it will also be necessary to determine whether there is enough connectivity between the CSJIMD and the MPNR. In the future, we hope that the general public will likely recognize the potential of the Andean tiger cat as a flagship species of the Andean cloud forest.
Concluding remarks
Our results showed that in our study areas, Andean tiger cat habitat use was mainly modulated by microhabitat structure, elevation, and codetectability with prey and tayras. Additionally, there was a greater probability of finding Andean tiger cats far from human settlements and of detecting them using mainly cameras with the shortest PIR time delays. There was no evidence of competitive exclusion on the habitat use of Andean tiger cats, likely due to the low abundance of intraguild killers/predators in this system, as predicted by the ocelot effect theory. Indeed, the species was more detectable when its intraguild predators were present. The protected areas currently studied may be considered effective barriers against the effects of fragmentation and forest loss on the habitat use of Andean tiger cats in the region. Decision makers should halt the progress of the deforestation frontiers both inside and around protected areas while preserving the vertical structure of the forest ground for the conservation of this mountain felid.
Supporting information
S1 Fig. The results of the principal component analysis performed to summarize four landscape metrics in a 500 m buffer to model the effects of landscape structure on the habitat use of the Andean tiger cat.
The proportion of variance explained by each component (A), the weight of each covariate on each component (B), and the coordinate space of the first and second components (C) with a comparison between protected areas (D). The three protected areas in which we obtained the landscape metrics were the Mesenia-Paramillo Nature Reserve (MPNR), the Campoalegre Soil Conservation District (CSCD) and the Cuchilla del San Juan Integrated Management District (CSJIMD). The metrics were: Forest_cov; cloud forest amount, Forest_PCI; patch cohesion index, SHANNON; landscape heterogeneity and EDGE; forest edge density. Dim1 (first component) represents more heterogeneous sites with greater forest edge (positive values) to more forested sites with greater cohesion among patches (negative values).
https://doi.org/10.1371/journal.pone.0288247.s001
(DOCX)
S2 Fig. Pairwise Pearson’s correlation coefficients for the original set of covariates selected to explain the habitat use of the Andean tiger cat in three protected cloud forests of northwestern Colombia.
The final set of numerical covariates used in the occupancy modeling of the Andean tiger cat included LS; landscape structure, AD; canopy height, CD; canopy cover, CH; leaf litter cover, PH; leaf litter depth, slope, DIST_POP; distance to human settlements and elev; elevation. For the modeling of detection probability, the final set of numerical covariates included CF; herbaceous cover, slope, and DAYS; survey effort of the camera traps.
https://doi.org/10.1371/journal.pone.0288247.s002
(DOCX)
S3 Fig.
Posterior predictive checks obtained for the occupancy models incorporating the effect of leaf litter depth (A), quadratic elevation (B), and distance to human settlements (C). The Bayesian P value was calculated by simulating 500 datasets under the MacKenzie-Bailey Goodness of Fit approach and comparing the proportion of simulated zeros against the proportion of observed zeros in the dataset for each model. Matching proportions (red line in the middle of the histogram) and Bayesian P values > 0.05 indicate good model fit.
https://doi.org/10.1371/journal.pone.0288247.s003
(DOCX)
S4 Fig. Moran’s I spatial correlograms for the residuals in the occupancy component of the three top-ranked models used to predict the habitat use of Andean tiger cats in cloud forests of the Middle Cauca, Colombia.
https://doi.org/10.1371/journal.pone.0288247.s004
(DOCX)
S1 Table. List of covariates and their effect on the habitat use and detection of the Andean tiger cat in cloud forests of the Middle Cauca, Colombia.
Ψ; effect on habitat use, p; effect on detection, +; positive effect, -; negative effect, +/-; effect dependent on the type of factor, GIS; geographic information system.
https://doi.org/10.1371/journal.pone.0288247.s005
(DOCX)
S2 Table. Estimates of the posterior distribution of the occupancy and detection parameters for the Andean tiger cat, including the protected areas as random effects.
Considering each protected area individually, the null model for CSCD yielded a 6% chance of detecting the Andean tiger cat if present (p = 0.06, SD = 0.02, CI = 0.02–0.11) and an average occupancy of 63% (Ψ = 0.63, SD = 0.16, CI = 0.32–0.95). In the CSJIMD, the null model showed an average detection of 15% (p = 0.15, SD = 0.02, CI = 0.11–0.19) and an average occupancy of 71% (Ψ = 0.71, SD = 0.11, CI = 0.47–0.92). The null model for MPNR showed an average detection of 18% (p = 0.18, SD = 0.01, CI = 0.14–0.22) and an average occupancy of 85% (Ψ = 0.85, SD = 0.08, CI = 0.65–0.96). Abbreviations: ELPD; expected predictive accuracy, K; number of parameters, ELPDΔ; difference in expected predictive accuracy between any model and the best model, SE; standard error of the difference in predictive accuracy.
https://doi.org/10.1371/journal.pone.0288247.s006
(DOCX)
S3 Table. Estimates of the conditional cooccurrence models evaluating the effects of prey, tayras, and intraguild killers/predators on the habitat use of the Andean tiger cat in three protected cloud forests from the Middle Cauca, Colombia.
Abbreviations: AICc; Akaike Information Criterion Value, df; degrees of freedom, AICΔ; difference of the AIC of each model with the best model, and AICw; relative support value of each model.
https://doi.org/10.1371/journal.pone.0288247.s007
(DOCX)
S4 Table. Locations, survey length, sampling effort (number of days of activity), type of region, sampling area and estimations of occupancy and detection probabilities for other studies carried out on the tigrinus species complex throughout South America.
https://doi.org/10.1371/journal.pone.0288247.s008
(DOCX)
Acknowledgments
This paper is a contribution made by the “Wild Cats Americas” conservation and research program and by the “Tiger Cat Conservation Initiative” (TCCI). We are also grateful for the support provided by Bioconservancy and Wild Nature Foundation during the project. Many people helped with fieldwork throughout the years across the three protected areas studied, particularly several undergraduate students from the team of the “Andean Tiger Cat Conservation Project” (ATCCP), Fabio Alberto Agudelo and Victor Puerta from Fundación Vida Silvestre, and Juvenal Sierra. We also thank Christian Osorio, Lester Fox-Rosales, and Roland Kays for their valuable insights and comments regarding earlier versions of the manuscript. The main author is thankful to Juan Camilo Botero, a wildlife photographer who provided an image of the Andean tiger cat in its natural habitat. Finally, we thank the reviewers for helping improve the quality of the manuscript.
References
- 1. Roemer GW, Gompper ME, Valkenburgh B van. The ecological role of the mammalian mesocarnivore. Bioscience 2009;59:165–73.
- 2. Tortato MA, Oliveira-Santos LGR, Moura MO, de Oliveira TG. Small prey for small cats: the importance of prey-size in the diet of southern tiger cat Leopardus guttulus in a competitor-free environment. Stud Neotrop Fauna Environ 2021. https://doi.org/10.1080/01650521.2021.1902202.
- 3. Santos F, Carbone C, Wearn OR, Rowcliffe JM, Espinosa S, Moreira MG, et al. Prey availability and temporal partitioning modulate felid coexistence in Neotropical forests. PLoS One 2019;14. pmid:30861045
- 4. Cruz P, Iezzi ME, de Angelo C, Varela D, di Bitetti MS, Paviolo A. Effects of human impacts on habitat use, activity patterns and ecological relationships among medium and small felids of the Atlantic Forest. PLoS One 2018;13. pmid:30067785
- 5. Wang B, Rocha DG, Abrahams MI, Antunes AP, Costa HCM, Gonçalves ALS, et al. Habitat use of the ocelot (Leopardus pardalis) in Brazilian Amazon. Ecol Evol 2019;9:5049–62. https://doi.org/10.1002/ece3.5005.
- 6. Lombardi J V., Tewes ME, Perotto-Baldivieso HL, Mata JM, Campbell . Spatial structure of woody cover affects habitat use patterns of ocelots in Texas. Mamm Res 2020;65:555–63.
- 7. Bogoni JA, Graipel ME, De Castilho P V., Peroni N. Development of predatory behaviours in young southern tigrinas (Leopardus guttulus). Mammalia 2017;81:421–4.
- 8. Linnell JDC, Strand O. Interference interactions, co-existence and conservation of mammalian carnivores. Divers Distrib 2000;6. https://doi.org/10.1046/j.1472-4642.2000.00069.x.
- 9. de Oliveira T, Pereira J. Intraguild Predation and Interspecific Killing as Structuring Forces of Carnivoran Communities in South America. J Mamm Evol 2014;21. https://doi.org/10.1007/s10914-013-9251-4.
- 10. Robinson QH, Bustos D, Roemer GW. The application of occupancy modeling to evaluate intraguild predation in a model carnivore system. Ecology 2014;95:3112–23. https://doi.org/10.1890/13-1546.1.
- 11. Sévêque A, Gentle LK, López-Bao J v., Yarnell RW, Uzal A. Human disturbance has contrasting effects on niche partitioning within carnivore communities. Biological Reviews 2020;95:1689–705. pmid:32666614
- 12. Bianchi R, Olifiers N, Riski LL, Gouvea JA, Cesário CS, Fornitano L, et al. Dog activity in protected areas: behavioral effects on mesocarnivores and the impacts of a top predator. Eur J Wildl Res 2020;66. https://doi.org/10.1007/s10344-020-01376-z.
- 13. Kitchener A, Breitenmoser-Würsten C, Eizirik E, Gentry A, Werdelin L, Wilting A, et al. A revised taxonomy of the Felidae. Cat News 2017;11:1–80.
- 14. Ramírez-Mejía AF, Sánchez F. Non-Volant mammals in a protected area on the Central Andes of Colombia: New records for the Caldas department and the Chinchiná River Basin. Check List 2015;11. https://doi.org/10.15560/11.2.1582.
- 15. Bonilla-Sánchez A, Gómez-Ruíz DA, Botero-Cañola S, Rendón-Jaramillo U, Ledesma-Castañeda E, Solari S. Riqueza y monitoreo de mamíferos en áreas protegidas privadas en Antioquia, Colombia. Mastozool Neotrop 2020;27. https://doi.org/10.31687/saremmn.20.27.2.0.11.
- 16. Bedoya-Durán MJ, Murillo-García OE, Branch LC. Factors outside privately protected areas determine mammal assemblages in a global biodiversity hotspot in the Andes. Glob Ecol Conserv 2021;32. https://doi.org/10.1016/j.gecco.2021.e01921.
- 17. Payán E, de Oliveira T. Leopardus tigrinus. IUCN Red List 2016. https://doi.org/https://dx.doi.org/10.2305/IUCN.UK.2016-2.RLTS.T54012637A50653881.en.
- 18. Rodríguez-Mahecha JV, Landazábal Mendoza C, Nash SD. Libro rojo de los mamíferos de Colombia. Serie Libros Rojos de Especies Amenazadas de Colombia 2006.
- 19. Castelló JR, Kitchener AC, Sliwa A. Felids and Hyenas of the World. 2020. https://doi.org/10.1515/9780691211862.
- 20. Delgado-V CA. Muerte de mamíferos por vehículos en la vía del Escobero, Envigado (Antioquia), Colombia. Actualidades Biológicas Biol 2007;2987.
- 21. Ayram CA, Etter A, Díaz-Timoté J, Rodríguez Buriticá S, Ramírez W, Corzo G. Spatiotemporal evaluation of the human footprint in Colombia: Four decades of anthropic impact in highly biodiverse ecosystems. Ecol Indic 2020;117. https://doi.org/10.1016/j.ecolind.2020.106630.
- 22. Rovero F, Zimmermann F, Berzi D, Meek P. “Which camera trap type and how many do I need?” A review of camera features and study designs for a range of wildlife research applications. Hystrix 2013;24. https://doi.org/10.4404/hystrix-24.2-6316.
- 23. MacKenzie DI, Nichols JD, Royle JA, Pollock KH, Bailey LL, Hines JE. Occupancy Estimation and Modeling: Inferring Patterns and Dynamics of Species Occurrence: Second Edition. 2017. https://doi.org/10.1016/C2012-0-01164-7.
- 24. Boron V, Xofis P, Link A, Payan E, Tzanopoulos J. Conserving predators across agricultural landscapes in Colombia: Habitat use and space partitioning by jaguars, pumas, ocelots and jaguarundis. ORYX 2020;54. https://doi.org/10.1017/S0030605318000327.
- 25. García-R S, Botero-Cañola S, Sánchez-Giraldo C, Solari S. Habitat use and activity patterns of Leopardus pardalis (Felidae) in the Northern Andes, Antioquia, Colombia. Biodiversity 2019;20. https://doi.org/10.1080/14888386.2019.1590235.
- 26. Pardo LE, Campbell MJ, Cove M V., Edwards W, Clements GR, Laurance WF. Land management strategies can increase oil palm plantation use by some terrestrial mammals in Colombia. Sci Rep 2019;9. https://doi.org/10.1038/s41598-019-44288-y.
- 27. Moreno-Sosa AM, Yacelga M, Craighead KA, Kramer-Schadt S, Abrams JF. Can prey occupancy act as a surrogate for mesopredator occupancy? A case study of ocelot (Leopardus pardalis). Mammalian Biology 2022;102:163–75. https://doi.org/10.1007/s42991-022-00232-8.
- 28. Dias DM, Massara RL, de Campos CB, Rodrigues FHG. Feline predator–prey relationships in a semi-arid biome in Brazil. J Zool 2019;307. https://doi.org/10.1111/jzo.12647.
- 29. Huaranca JC, Valdivia CE, Novaro AJ, Lucherini M. Traditional livestock activities modify the spatial behavior of small wildcats in the high Andes. J Nat Conserv 2022;70:126303. https://doi.org/10.1016/j.jnc.2022.126303.
- 30. Harmsen BJ, Saville N, Foster RJ. Long-term monitoring of margays (Leopardus wiedii): Implications for understanding low detection rates. PLoS One 2021;16. https://doi.org/10.1371/journal.pone.0247536.
- 31. Welbourne DJ, Claridge AW, Paull DJ, Lambert A. How do passive infrared triggered camera traps operate and why does it matter? Breaking down common misconceptions. Remote Sens Ecol Conserv 2016;2:77–83. https://doi.org/10.1002/rse2.20.
- 32. Zenner EK. Does old-growth condition imply high live-tree structural complexity? For Ecol Manage 2004;195:243–58. https://doi.org/10.1016/j.foreco.2004.03.026.
- 33. Tian J, Zou Q, Zhang M, Hu C, Khattak RH, Su H. Spatial and temporal differentiation are not distinct but are covariant for facilitating coexistence of small and medium-sized carnivores in Southwestern China. Glob Ecol Conserv 2022;34. https://doi.org/10.1016/j.gecco.2022.e02017.
- 34. Monterroso P, Sillero N, Rosalino LM, Loureiro F, Alves PC. Estimating home-range size: When to include a third dimension? Ecol Evol 2013;3. pmid:23919170
- 35. Semper-Pascual A, Decarre J, Baumann M, Camino M, di Blanco Y, Gómez-Valencia B, et al. Using occupancy models to assess the direct and indirect impacts of agricultural expansion on species’ populations. Biodivers Conserv 2020;29. https://doi.org/10.1007/s10531-020.
- 36. de Oliveira T, Lima BC, Fox-Rosales L, Pereira RS, Pontes-Araújo E, de Sousa AL. A refined population and conservation assessment of the elusive and endangered northern tiger cat (Leopardus tigrinus) in its key worldwide conservation area in Brazil. Glob Ecol Conserv 2020;22. https://doi.org/10.1016/j.gecco.2020.e00927.
- 37.
de Oliveira T, Tortato M, Silveira L, Kasper CB, Mazim F, Lucherini M, et al. Ocelot ecology and its effect on the small-felid guild in the lowland neotropics. In: Macdonald D, Loveridge A, editors. Biology and conservation of wild felids, Oxford University Press; 2010, p. 559–80.
- 38. González-González A, Clerici N, Quesada B. A 30 m-resolution land use-land cover product for the Colombian Andes and Amazon using cloud-computing. International Journal of Applied Earth Observation and Geoinformation 2022;107. https://doi.org/10.1016/j.jag.2022.102688.
- 39.
United States Geological Service. Shuttle Radar Topography Mission (SRTM) Non-Void Filled n.d. https://doi.org/https://doi.org/10.5066/F7K072R7.
- 40.
Departamento Administrativo Nacional de Estadística - DANE. https://geoportal.dane.gov.co/servicios/descarga-y-metadatos/descarga-mgn-marco-geoestadistico-nacional/2022.
- 41. Bedoya-Gómez BD, Dossman-Gil MA, Marín-Fernández J. Valoración ecológica de los servicios ecosistémicos prestados por el suelo en fincas cafeteras en Belén de Umbría, Colombia. Revista de Ciencias Ambientales 2021;55. https://doi.org/10.15359/rca.55-1.8.
- 42. Fahey TJ, Sherman RE, Tanner EVJ. Tropical montane cloud forest: Environmental drivers of vegetation structure and ecosystem function. J Trop Ecol 2016;32. https://doi.org/10.1017/S0266467415000176.
- 43. Cepeda–Duque JC, Gómez–valencia B, Alvarez S, Gutiérrez–sanabria DR, Lizcano DJ. Daily activity pattern of pumas (Puma concolor) and their potential prey in a tropical cloud forest of colombia. Anim Biodivers Conserv 2021;44. https://doi.org/10.32800/abc.2021.44.0267.
- 44. Oliveira-Santos LGR, Graipel ME, Tortato MA, Zucco CA, Cáceres NC, Goulart FVB. Abundance changes and activity flexibility of the oncilla, Leopardus tigrinus (Carnivora: Felidae), appear to reflect avoidance of conflict. Zoologia 2012;29:115–20. https://doi.org/10.1590/S1984-46702012000200003.
- 45. Fox-Rosales LA, de Oliveira TG. Interspecific patterns of small cats in an intraguild-killer free area of the threatened Caatinga drylands, Brazil. PLoS One 2023;18:e0284850. pmid:37083683
- 46. Hofmeester TR, Rowcliffe JM, Jansen PA. A simple method for estimating the effective detection distance of camera traps. Remote Sens Ecol Conserv 2017;3. https://doi.org/10.1002/rse2.25.
- 47. Dias D de M, Lima Massara R, de Campos CB, Henrique Guimarães Rodrigues F. Human activities influence the occupancy probability of mammalian carnivores in the Brazilian Caatinga. Biotropica 2019;51. https://doi.org/10.1111/btp.12628.
- 48. Palencia P, Vicente J, Soriguer RC, Acevedo P. Towards a best-practices guide for camera trapping: assessing differences among camera trap models and settings under field conditions. J Zool 2022;316:197–208. https://doi.org/10.1111/jzo.12945.
- 49. Farhadinia MS, Heit DR, Montgomery RA, Johnson PJ, Hobeali K, Hunter LTB, et al. Vertical relief facilitates spatial segregation of a high density large carnivore population. Oikos 2020;129. https://doi.org/10.1111/oik.06724.
- 50. Massara R L., Paschoal AM de O, L. Bailey L, F. Doherty P, Hirsch A, G. Chiarello A. Factors influencing ocelot occupancy in Brazilian Atlantic Forest reserves. Biotropica 2018;50. https://doi.org/10.1111/btp.12481.
- 51. Pasa JB, Arrais RC, Massara RL, Pereira G, de Azevedo FCC. Factors influencing the habitat use by ocelots in one of the last large Atlantic Forest remnants in southeastern Brazil. Ecol Evol 2021;11. pmid:33976836
- 52. Jung M. LecoS - A python plugin for automated landscape ecology analysis. Ecol Inform 2016;31:18–21. https://doi.org/10.1016/j.ecoinf.2015.11.006.
- 53. Huaranca JC, Valdivia CE, Torrez AR, Flores E, Novaro AJ. Effect of traditional livestock grazing on abundance of small mammals and activity of mountain vizcacha in the high central semi-arid Andes. J Arid Environ 2022;206:104846. https://doi.org/10.1016/j.jaridenv.2022.104846.
- 54. Alves F, López-Iborra GM, Stojanovic D, Silveira LF. Habitat selection by the endangered Red-billed Curassow (Crax blumenbachii) in an Atlantic forest remnant. Emu 2017;117. https://doi.org/10.1080/01584197.2017.1326010.
- 55. Ardente NC, Ferreguetti ÁC, Gettinger D, Leal P, Martins-Hatano F, Banhos A, et al. Habitat use by two sympatric species of short-tailed opossums (Didelphidae: Monodelphis) in an area in eastern Amazonia. J Mammal 2021;102. https://doi.org/10.1093/jmammal/gyab047.
- 56. Freitas SR, Cerqueira R, Vieira M v. A device and standard variables to describe microhabitat structure of small mammals based on plant cover. Brazilian Journal of Biology 2002;62. pmid:12659030
- 57. August P v. The role of habitat complexity and heterogeneity in structuring tropical mammal communities. Ecology 1983;64. https://doi.org/10.2307/1937504.
- 58. Nguyen T V., Wilting A, Niedballa J, Nguyen A, Rawson BM, Nguyen AQH, et al. Getting the big picture: Landscape-scale occupancy patterns of two Annamite endemics among multiple protected areas. Conserv Sci Pract 2022;4. https://doi.org/10.1111/csp2.620.
- 59.
Kéry M, Royle JA. Applied Hierarchical Modeling in Ecology: Analysis of distribution, abundance and species richness in R and BUGS. vol. 1. London, UK: Academic Press; 2015. https://doi.org/10.1016/c2015-0-04070-9.
- 60. Kellner KF, Fowler NL, Petroelje TR, Kautz TM, Beyer DE, Belant JL. ubms: An R package for fitting hierarchical occupancy and N-mixture abundance models in a Bayesian framework. Methods Ecol Evol 2022;13. https://doi.org/10.1111/2041-210X.13777.
- 61. Calderón AP, Louvrier J, Planillo A, Araya-Gamboa D, Arroyo-Arce S, Barrantes-Núñez M, et al. Occupancy models reveal potential of conservation prioritization for Central American jaguars. Anim Conserv 2022. https://doi.org/10.1111/acv.12772.
- 62. Carpenter B, Gelman A, Hoffman MD, Lee D, Goodrich B, Betancourt M, et al. Stan: A probabilistic programming language. J Stat Softw 2017;76. pmid:36568334
- 63. Vehtari A, Gelman A, Gabry J. Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC. Stat Comput 2017;27:1413–32. https://doi.org/10.1007/s11222-016-9696-4.
- 64. Gelman A, Hwang J, Vehtari A. Understanding predictive information criteria for Bayesian models. Stat Comput 2014;24:997–1016. https://doi.org/10.1007/s11222-013-9416-2.
- 65. Morin DJ, Yackulic CB, Diffendorfer JE, Lesmeister DB, Nielsen CK, Reid J, et al. Is your ad hoc model selection strategy affecting your multimodel inference? Ecosphere 2020;11. https://doi.org/10.1002/ecs2.2997.
- 66. Oberpriller J, de Souza Leite M, Pichler M. Fixed or random? On the reliability of mixed‐effects models for a small number of levels in grouping variables. Ecol Evol 2022;12. https://doi.org/10.1002/ece3.9062.
- 67. Dechner A, Flesher KM, Lindell C, de Oliveira TV, Maurer BA. Determining carnivore habitat use in a rubber/forest landscape in Brazil using multispecies occupancy models. PLoS One 2018;13. https://doi.org/10.1371/journal.pone.0195311.
- 68. Andrade-Ponce GP, Mandujano S, Dáttilo W, Farías-González V, Jiménez J, Velásquez-C K, et al. A framework to interpret co-occurrence patterns from camera trap data: The case of the gray fox, the bobcat, and the eastern cottontail rabbit in a tropical dry habitat. J Zool 2022. https://doi.org/10.1111/jzo.13002.
- 69. Wright WJ, Irvine KM, Higgs MD. Identifying occupancy model inadequacies: can residuals separately assess detection and presence? Ecology 2019;100. pmid:30932179
- 70. Bjornstad N, Cai J. Title Spatial Covariance Functions. 2022.
- 71. Richmond O, Hines JE, Beissinger SR. Two-species occupancy models: a new parameterization applied to co-occurrence of secretive rails. Source: Ecological Applications 2010;20:2036–46. pmid:21049888
- 72. Monterroso P, Díaz-Ruiz F, Lukacs PM, Alves PC, Ferreras P. Ecological traits and the spatial structure of competitive coexistence among carnivores. Ecology 2020;101. pmid:32333382
- 73. Davis CL, Rich LN, Farris ZJ, Kelly MJ, Di Bitetti MS, Blanco Y Diet al. Ecological correlates of the spatial co‐occurrence of sympatric mammalian carnivores worldwide. Ecol Lett 2018;21:1401–12. pmid:30019409
- 74. Zuercher GL, Owen RD, Torres J, Gipson PS. Mechanisms of coexistence in a diverse Neotropical mammalian carnivore community. J Mammal 2022;103:618–38. https://doi.org/10.1093/jmammal/gyac003.
- 75. Burnham K, Anderson D. Model Selection and Multimodel Inference: A Practical Information‐Theoretic Approach. 2nd edn. Springer, Berlin. Bayesian Data Analysis in Ecology Using Linear Models with R, BUGS, and STAN 2002.
- 76. Mackenzie DI, Bailey LL, Nichols JD. Investigating species co-occurrence patterns when species are detected imperfectly. Journal of Animal Ecology 2004;73. https://doi.org/10.1111/j.0021-8790.2004.00828.x.
- 77. Meredith M. wiqid: Quick and Dirty Estimates for Wildlife Populations 2022. R Package Version 0.3.3, https://cran.r-project.org/web/packages/wiqid/wiqid.pdf.
- 78. González-Maya JF, Zárrate-Charry DA, Arias-Alzate A, Lemus-Mejía L, Hurtado-Moreno AP, Vargas-Gómez MG, et al. Spotting what’s important: Priority areas, connectivity, and conservation of the Northern Tiger Cat (Leopardus tigrinus) in Colombia. PLoS One 2022;17:e0273750. https://doi.org/10.1371/journal.pone.0273750.
- 79. Sanchez-Giraldo C, Diaz-Nieto JF. Dynamics of species composition of small non-volant mammals from the northern Cordillera Central of Colombia. Mammalia 2015;79. https://doi.org/10.1515/mammalia-2014-0018.
- 80. Gómez-Valencia B, López-Arévalo HF, Gómez-Laverde M. Pequeños mamíferos no voladores de la Reserva Rio Blanco y aporte de localidades para la identificación de patrones ecológicos en los andes colombianos. Caldasia 2021;43. https://doi.org/10.15446/caldasia.v43n2.85463.
- 81. Bruijnzeel LA, Kappelle M, Mulligan M, Scatena FN. Tropical montane cloud forests: State of knowledge and sustainability perspectives in a changing world. Tropical Montane Cloud Forests: Science for Conservation and Management, 2011. https://doi.org/10.1017/CBO9780511778384.074.
- 82. Etter A, van Wyngaarden W. Patterns of Landscape Transformation in Colombia, with Emphasis in the Andean Region. AMBIO: A Journal of the Human Environment 2000;29:432–9. https://doi.org/10.1579/0044-7447-29.7.432.
- 83. Haidir IA, Kaszta Ż, Sousa LL, Lubis MI, Macdonald DW, Linkie M. Felids, forest and farmland: identifying high priority conservation areas in Sumatra. Landsc Ecol 2021;36. https://doi.org/10.1007/s10980-020-01146-x.
- 84. Larsen T, Brehm G, Navarrete H, Franco P, Gomez H, Meno JL, et al. Range shifts and extinctions driven by climate change in the tropical Andes: synthesis and directions. Climate Change and Biodiversity in the Tropical Andes 2011.
- 85. Foster P. The potential negative impacts of global climate change on tropical montane cloud forests. Earth Sci Rev 2001;55.
- 86. Vale MM, Lorini ML, Cerqueira R. Neotropical wild cats susceptibility to climate change. Oecologia Australis 2015;19. https://doi.org/10.4257/oeco.2015.1901.05.
- 87. Fox-Rosales LA, de Oliveira TG. Habitat use patterns and conservation of small carnivores in a human-dominated landscape of the semiarid Caatinga in Brazil. Mammalian Biology 2022;102:465–75. https://doi.org/10.1007/s42991-022-00245-3.
- 88. Fleschutz MM, Gálvez N, Pe’er G, Davies ZG, Henle K, Schüttler E. Response of a small felid of conservation concern to habitat fragmentation. Biodivers Conserv 2016;25. https://doi.org/10.1007/s10531-016-1118-6.
- 89. Zapata-Ríos G, Branch LC. Mammalian carnivore occupancy is inversely related to presence of domestic dogs in the high Andes of Ecuador. PLoS One 2018;13. pmid:29489855
- 90. Marinho PH, Bezerra D, Antongiovanni M, Fonseca CR, Venticinque EM. Estimating occupancy of the Vulnerable northern tiger cat Leopardus tigrinus in Caatinga drylands. Mamm Res 2018;63:33–42. https://doi.org/10.1007/s13364-017-0330-4.
- 91. Liévano Latorre LF, López Arévalo HF. Comunidad de mamíferos no voladores en un área periurbana andina, cundinamarca, Colombia. Acta Biolo Colomb 2015;20. https://doi.org/10.15446/abc.v20n2.43477.
- 92. Jiménez-Alvarado JS, Moreno-Díaz C, Alfonso AF, Giordano A, Vela-Vargas IM, Gómez-Hoyos DA, et al. Ciudades biodiversas: mamíferos medianos de la Reserva Forestal Protectora Bosque Oriental de Bogotá D. C., Colombia. Mammalogy Notes 2017;4. https://doi.org/10.47603/manovol4n1.37-41.
- 93. Pineda-Guerrero A. Nuevo registro de Leopardus tigrinus (Carnivora: Felidae) en la Reserva Forestal Protectora Bosque Oriental de Bogotá, Colombia. Mammalogy Notes 2018;5. https://doi.org/10.47603/manovol5n1.15-17.
- 94. Lewis JS, Logan KA, Alldredge MW, Bailey LL, Vandewoude S, Crooks KR. The effects of urbanization on population density, occupancy, and detection probability of wild felids. Ecological Applications 2015;25:1880–95. pmid:26591454
- 95. Rodríguez Eraso N, Armenteras-Pascual D, Alumbreros JR. Land use and land cover change in the Colombian Andes: dynamics and future scenarios. J Land Use Sci 2013;8:154–74. https://doi.org/10.1080/1747423X.2011.650228.
- 96. Vargas-Ramírez LC, Colmenares-Pinzón JE, Serrano-Cardozo VH. Microhabitat use by an assamblage of small non volant mammals in an oak forest in the Cordillera Oriental of Colombian Andes. Rev Mex Biodivers 2018;89:479–87. https://doi.org/10.22201/ib.20078706e.2018.2.1933.
- 97. Dunford CE, Marks NJ, Wilmers CC, Bryce CM, Nickel B, Wolfe LL, et al. Surviving in steep terrain: A lab-to-field assessment of locomotor costs for wild mountain lions (Puma concolor). Mov Ecol 2020;8. https://doi.org/10.1186/s40462-020-00215-9.
- 98. Morrell N, Appleton RD, Arcese P. Roads, forest cover, and topography as factors affecting the occurrence of large carnivores: The case of the Andean bear (Tremarctos ornatus). Glob Ecol Conserv 2021;26. https://doi.org/10.1016/j.gecco.2021.e01473.
- 99. Zub K, Szafrańska PA, Konarzewski M, Redman P, Speakman JR. Trade-offs between activity and thermoregulation in a small carnivore, the least weasel Mustela nivalis. Proceedings of the Royal Society B: Biological Sciences 2009;276. https://doi.org/10.1098/rspb.2008.1936.
- 100. de Oliveira T, Fox-Rosales LA, Paemelaere EAD, de Barros Ferraz KMPM. The dominant mesopredator and savanna formations shape the distribution of the rare northern tiger cat (Leopardus tigrinus) in the Amazon. Sci Rep 2022;12:18697. https://doi.org/10.1038/s41598-022-21412-z.
- 101.
Sunquist M, Sunquist F. Wild Cats of the World. 1st ed. Chicago: University of Chicago Press; 2002.
- 102. Dickman AJ, Hinks AE, Macdonald EA, Burnham D, Macdonald DW. Priorities for global felid conservation. Conservation Biology 2015;29. pmid:25864434
- 103. Giordano A. The Oncilla: A new conservation perspective. Wild Felid Monitor 2012;5:16–16.
- 104. Etter A, McAlpine C, Possingham H. Historical patterns and drivers of landscape change in Colombia since 1500: A regionalized spatial approach. Annals of the Association of American Geographers 2008;98. https://doi.org/10.1080/00045600701733911.
- 105. Marneweck CJ, Allen BL, Butler AR, Do Linh San E, Harris SN, Jensen AJ, et al. Middle-out ecology: small carnivores as sentinels of global change. Mamm Rev 2022;52:471–9. https://doi.org/10.1111/mam.12300.
- 106. Delheimer MS, Moriarty KM, Munro HL, Early DA, Hamm KA, Green RE. Structural complexity characterizes fine‐scale forest conditions used by Pacific martens. J Wildl Manage 2023;87. https://doi.org/10.1002/jwmg.22388.
- 107. Ocampo-Peñuela N, Pimm SL. Setting Practical Conservation Priorities for Birds in the Western Andes of Colombia. Conservation Biology 2014;28. pmid:25065287