Integrating Species Distribution Models to Identify Overlapping Predator–Prey Conservation Priorities in Misiones, Argentina

Abstract

Misiones province covers < 1% of Argentina’s land area yet harbors > 50% of the country’s biodiversity, with a significant remnant of Atlantic Forest, a global biodiversity hotspot. Approximately 540,000 ha of this native forest is protected, with the remaining areas facing threats from ongoing land conversion, an expanding road network, and a growing rural population. A prior study incorporated noninvasive data on five carnivores into a multifaceted cost analysis to define the optimal location for a multispecies biological corridor, with the goal of enhancing landscape connectivity among protected areas. Subsequent analyses, with an updated framework, emphasized management strategies that balanced human–wildlife coexistence and habitat needs. Building on these efforts, our study applied ecological niche modeling to data located by conservation detection dogs, with genetics used to confirm species identity, and two land-use scenarios, to predict potential distributions of three game species—lowland tapir (Tapirus terrestris), white-lipped peccary (Tayassu pecari), and collared peccary (Pecari tajacu)—that are not only threatened by poaching, road mortality, and habitat loss but also serve as essential prey for carnivores. We assessed the suitability of unique and overlapping vegetation types, within and outside of protected areas, as well as within this multispecies corridor, identifying zones of high conservation concern that underscore the need for integrated planning of predators and prey. These results highlight that ensuring the long-term viability of wildlife across the heterogeneous land-use matrices of Misiones requires going beyond protected areas to promote functional connectivity, restore degraded habitats, and balance human–wildlife needs.

1. Introduction

The survival of apex predators in the Neotropics, such as the jaguar (Panthera onca) and puma (Puma concolor), is tightly linked to the extensive habitats they require, making them critical indicators of ecosystem health [1,2,3]. Moreover, the persistence of these carnivores depends on the availability of their prey, which includes the lowland tapir (tapir; Tapirus terrestris), white-lipped peccary (Tayassu pecari), and collared peccary (Pecari tajacu), that are highly vulnerable to deforestation, habitat fragmentation, poaching, and other anthropogenic pressures [4,5,6,7,8,9]. The threats to carnivores and their prey, including habitat loss, road mortality, human–wildlife conflict, and disease, often operate synergistically, creating a cascading effect that compromises their long-term persistence and ecosystem stability [5,6,10,11,12,13,14,15,16,17]. For instance, the absence of adequate prey can force carnivores to expand their foraging ranges into fragmented or human-dominated landscapes, thereby increasing mortality risks [9,18,19,20,21]. Even the loss of a single prey species may trigger shifts in community structure and function, including density compensation, a key aspect of ecological release, where surviving species expand their realized ecological niche and increase population densities [9,20,22,23,24,25].

The relationship between predators, prey, and habitat is complex, extending beyond the habitat use (i.e., how a species interacts or uses the environment) to its suitability (i.e., how effective an environment is at supporting a population) [26]. Predator–prey interactions have the potential to be enhanced or diminished by various physiological consequences (e.g., microclimate effects on physiological capacities) and biotic interactions (e.g., competition), with the latter considering how the physical environment (i.e., habitat) interacts with population and communities [27,28]. At a fine scale (i.e., home range), predators and prey focus on resource utilization and balancing trade-offs between resource abundance (i.e., food) and risk avoidance (i.e., safety) [29,30]. These trade-offs often differ for predators and prey. For example, while a complex habitat can provide physical refuge for prey, this same scenario can increase predation costs for predators [28,31]. However, these dynamics can shift when expanding scale to the broader landscape, with the introduction of additional factors, such as human disturbance (e.g., roads, human settlements) [29,32]. Understanding shifts in habitat suitability at different scales is fundamental to determining how ecological challenges, including climate change and land modifications, may influence the habitat-species occurrence [26,29,32]. The importance of this understanding extends to habitat restoration efforts, where successful predator–prey coexistence depends on the new environment balancing the costs and benefits of both [28].

Although large, protected areas are ideal for maintaining biodiversity, small reserves and habitat fragments can also contribute to connectivity, if appropriately located [15,16,33]. However, their effectiveness varies across species. For example, the white-lipped peccary tends to avoid heavily converted landscapes, even when forest corridors exist [34]. Ensuring functional connectivity between core habitat patches is therefore essential to support animal movement, genetic flow, and resilience in the face of increasing environmental transformation [6,17,35,36,37]. The urgency of maintaining these connections is further amplified by climate change, which may alter habitat suitability and species distributions [35,38,39,40,41].

In this context, the province of Misiones in northeastern Argentina represents both a conservation opportunity and a challenge. Specifically, although it comprises <1% of Argentina’s territory, it harbors > 50% of the country’s biodiversity and contains one of the largest remaining tracts of Atlantic Forest, a recognized global biodiversity hotspot [42,43,44,45,46]. Despite its small size, Misiones was recently declared the National Capital of Biodiversity (National Law No. 27.494), with the province designating approximately 540,000 ha of native forest as protected areas. However, these protected areas are increasingly threatened by an expanding road network (many located inside), land-use changes, and human encroachment along and outside their borders. To address this, the provincial government created the Green Corridor (Corredor Verde; Provincial Law XVI No 60) in 1999, encompassing about 1.1 million ha of this forest, but weak enforcement continues to leave both habitat and species vulnerable to degradation [13,47,48,49,50,51]. To reverse these trends, DeMatteo et al. [52] modeled a multispecies biological corridor (~400,000 ha) that connected key protected areas in the northern-central (N-C) zone of Misiones by incorporating the spatial needs of five carnivore species: jaguar, puma, ocelot (Leopardus pardalis), southern tiger cat (Leopardus guttulus), and bush dog (Speothos venaticus). Unlike single-species models, multispecies corridors account for diverse ecological requirements and offer broader conservation utility [36,53,54,55,56,57]. Later work by DeMatteo et al. [58] provided management strategies to mitigate human–wildlife conflicts and prioritize areas for protection and restoration, but prey distribution was not integrated into these models.

Our study aimed to fill this gap by generating ecological niche models (ENMs) for three prey species—tapir, white-lipped peccary, and collared peccary—whose ecological roles as seed dispersers, vegetation regulators, and ecosystem engineers are critical to forest regeneration and biodiversity maintenance [4,5,6,7,8,9,59,60,61,62,63,64,65]. Because these models extend beyond predicting prey presence to potential areas of high species richness, it allows them to be a proxy for areas with high prey abundance and resource availability [66]. These potential distributions across the N-C region of Misiones used presence-only data located by conservation detection dogs, with genetics used to confirm species identity, and two alternative land-use layers. Subsequent threshold-defined spatial analyses were used to evaluate habitat suitability and develop management strategies (i.e., areas in need of protection and/or restoration based on importance and ecological role for the species) [31,67], as well as quantify overlaps with the previously modeled multispecies corridor [52,58]. Our results provide insight into areas where prey persistence may be compromised and inform corridor design by integrating both predator and prey spatial requirements, ultimately supporting more effective, evidence-based conservation strategies in Misiones.

2. Materials and Methods

2.1. Study Area

This study was conducted in the province of Misiones, Argentina, located in the Upper Paraná Atlantic Forest ecoregion, one of the most biologically diverse and endangered tropical forests globally [42,45,68,69] (Figure 1). This region is home to a variety of threatened or endangered vertebrates and is increasingly affected by landscape fragmentation resulting from human activities [44,45,46,47,50,69]. Misiones encompasses numerous protected areas, including the UNESCO-designated Yabotí Biosphere Reserve and one of the most significant remnants of continuous forest in the Atlantic Forest ecosystem [42,43,45,47,69].

2.2. Focal Species and Field Surveys

We selected three large-bodied Neotropical mammals—white-lipped peccary, collared peccary, and tapir—as focal species due to their ecological roles, conservation concerns, and sensitivity to habitat disturbance. These species, which are commonly hunted for subsistence and poaching, are important indicators of forest integrity [71,72,73]. All three species are listed under conservation categories by the IUCN and Argentina’s national legislation [72,73,74,75,76,77].

Data were collected during two surveys (2016 & 2018) conducted primarily during the austral winter (May–August) in Misiones, Argentina. The Ministerio de Ecología y Recursos Naturales Renovables (MEyRNR) issued all general permits related to the project in the province, as well as those related to the collection, storage, and analysis of samples from multiple provincial parks.

The 2016 surveys were completed using two detection dogs: one with three years of field experience in Argentina locating scat from three primates [78,79] and one with seven years of field experience in Argentina and the USA finding scat from five carnivores [52,58,80,81,82]. The 2018 surveys were completed with only the latter detection dog. Within Misiones, detection dogs offer a reliable sampling, even in remote or human-dominated landscapes, where camera traps are not feasible [80,81,82]. They also remove the potential bias in population sampling that could occur with the inability to deploy camera traps outside of protected areas, which would affect the breadth and scale at which data could be interpreted [16,83,84,85,86].

Survey tracks consisted of two-lane paved roads, 1–2 lane dirt roads, and a variety of established trails (e.g., machete-cut, illegal hunting, animal). The team surveyed a total of 131 unique routes and walked a total of 908.56 km [mean (SD) = 6.94 (3.71) km; range = 0.21–15.15 km per route; Figure 1], with previous studies estimating the dog covered an average distance 4–6× that walked by humans [80,87]. These surveys were spread over northern-central Misiones to determine focal species presence across the various habitats in the heterogeneous landscape of the multispecies corridor [52,58]. Most surveys (667.86 km; 73.51%) covered unique habitats, including private native forests, small-scale agriculture, monoculture plantations of pine and eucalyptus, small communities with subsistence agriculture, pastures, and human-occupied areas, with the remainder located in protected areas of native forest, with varying degrees of legal protection (Figure 1).

The training of the detection dog, swabbing of scat for genetic analysis, collection-storage of scats, and recording of field data were carried out using the same strategies optimized in previous surveys [81,82,88].

2.3. Genetic Analysis and Species Occurrence Data

Scat swabs were processed using DNA extraction protocols and genetic analyses detailed in [88], with a summary of the genetic procedures provided here. DNA was extracted from one of two independent swabs using a manual extraction method [89]. For samples that presented complications in subsequent analyses (PCR amplification), the second swab was extracted using a Qiagen (Venlo, the Netherlands) DNeasy™ DNA extraction kit following a modified protocol by Vynne et al. [90]. To confirm species identity, a 110-bp fragment (171-bp with primers) of the carnivore-specific region of mitochondrial cytochrome b gene [91], which has been shown to yield better results for degraded samples, was amplified with a modified version of the protocols, plus reagents described by Farrell et al. [91] and Miotto et al. [92]. Amplifications were performed on a Veriti Thermal Cycler System (Applied Biosystems, Waltham, MA, USA) in 25-µL volumes, with purified PCR products sequenced using the ABI (Applied Biosystems, Inc.) PRISM BigDye Terminator v3.1 Cycle Sequencing Kits and analyzed by Macrogen (Seoul, Korea) on ABI 3730 Genetic Analyzer (ABI). Sequences were edited and aligned using BioEdit Sequence Alignment Editor 7.2.6 [93] and compared with reference entries in GenBank using the Basic Local Alignment Search Tool (BLAST 2.10.0) [94] to identify sequences from Neotropical species that had high similarity and closely matched sample sequences (threshold of ≥98% match).

Of the 242 total samples collected, species identity was confirmed in 186 (76.86%;

Table A1. Details of the 186 scat swabs with confirmed species identity. For the 37 tapir, 128 white-lipped peccary (WLP), and 21 collared peccary (CP) samples, the location and zone [north (N) and central (C)] are summed by species. For protected areas, the total area is reported in parentheses.

Location	Zone	Tapir	WLP	CP
P.P. Urugua-í (84,000 ha)	N	20	11
Refugio Privado Aguaraí-mi (3050 ha)	N	1	7	9
Reserva de Vida Silvestre Urugua-í (3243 ha)	N	---	---	2
P.P. Cruce Caballero (522 ha) & Valle del Arroyo Alegría (8000 ha)	C	---	1	---
P.P. Esmeralda (31,569 ha) & Reserva de Biósfera Yabotí (236,313 ha)	C	2	13	---
Reserva Privada Yaguaroundí (400 ha)	C	---	2	---
outside protected areas	N	11	74	8
outside protected areas	C	3	20	2

). While most samples were identified as white-lipped peccary (n = 128; 68.82%), there were 37 (19.89%) tapir and 21 (11.29%) collared peccary [88,95]. In the remaining scat swabs, species identity was not possible due to the low quantity/quality of DNA or urine contamination from animals’ scent marking on the scat. Only genetically confirmed presence points were used in the models, with data from both years (2016 & 2018) combined for each of the three species. Exact sample locations are not reported or displayed at the government’s request, as a precaution to protect these threatened and endangered game species from targeted poaching.

2.4. Modeling the Ecological Niche

We implemented the Maximum Entropy algorithm, a machine learning method, in MaxEnt 3.3.3.k [96] to model the potential distribution of each of the three game species and evaluate habitat suitability. MaxEnt was chosen for its strong performance with presence-only data and small sample size [97], as it has been reported to perform consistently better than other algorithms [97,98,99]. Models were fit using auto features and default parameters for regularization multiplier (1), convergence threshold (1.0 × 10⁻⁵), and background points (10,000) [100,101,102]. For each species, models were tested randomly, withholding 25% of presence localities and running 15 subsample replicates, with the average output used for interpretation [103]. We generated a logistic output, which gives the probability of species presence on a scale of 0 to 1 and has been shown to improve model performance via model calibration of output values and corresponding suitability [100].

Applying threshold or ‘cutoff’ values to the logistic output of MaxEnt enables it to be converted into predictions that allow habitat suitability to be evaluated, with habitat categories defined as suitable (high probability of species’ presence) or unsuitable. Ecologically, values equal to or greater than the defined threshold value can be interpreted to contain cells that are predicted to be at least as suitable as those where the species was identified present. We compared the extrinsic omission rate and proportional predicted area (proxy for commission rate) at several logistic thresholds [e.g., minimum training presence (MTP), fixed cumulative value 1 (FCV1)] [100,104,105], with our criteria to have an omission rate of zero but set lower restrictions on the size of the potential predicted area. Second, the total suitable habitat for each species was subdivided into marginal and optimal habitats using the maximum training sensitivity plus specificity (MTSS), a threshold noted to be most appropriate at defining optimal habitats in presence-only models [104,106,107]. This threshold approach allows for a direct inference of habitat suitability, which allows for areas to be defined based on their importance and ecological role for species [31,67]. Ecologically, a marginal habitat corresponds to values equal to or greater than the binary threshold but less than the MTSS threshold. In contrast, optimal habitat corresponds to values greater than the MTSS threshold.

2.5. Predictor Variables for the Ecological Niche Model (ENM)

We compared two data sets that have a strong fit to the years of data collection (2016 & 2018), which effectively capture forest loss and land conversion across the landscape (Figure 1). Model 1 used a land use derived from a mosaic of Landsat-8 TM (Thematic Mapper) satellite images from 2015 [43]. Model 2 utilized a land use derived from automated classification of satellite images from 2018 [70]. Both layers were used to ensure that the inconsistencies in how intact or altered habitats are defined in the highly heterogeneous habitat outside of protected areas, as found in DeMatteo et al. [58], do not affect comparability in the final model results. Any discrepancy between the two models seems to be an artifact of how the pixels are defined versus an actual shift in habitat quality (e.g., intact versus altered), with the heterogeneous landscape outside of protected areas difficult to distinguish, with a mix of habitats in proximity and intermixed with each other, resulting in comparable results where the definition of intact or altered is consistent even if the define type of habitat varies [58]. Information directly related to road (frequency or distance) was not included in either model, as we believed these data would skew the model since a large portion of sample locations were closely associated with them.

Using ArcGIS Pro (Version 3.3; ESRI Inc., Redlands, CA, USA), a neighborhood analysis, with neighborhood scale set equal to a species-specific home range, was conducted by applying a focal statistic frequency operation to calculate the frequency (i.e., %) of each predictor variable in each land use layer at a 30 m × 30 m resolution (

Table 1. Summary of the predictor variables tested and used in the two species-specific ecological niche models (ENMs). Model 1 used land use derived from a mosaic of Landsat-8 TM (Thematic Mapper) satellite images from 2015 [43] and Model 2 used land use derived from the automated classification of satellite images from 2018 [70]. The list includes all predictor variables used in the development of the species-specific ENMs, as well as whether each variable had a focal statistic frequency operation (%) at a species-specific neighborhood scale applied and if it was ultimately included in the final model.

Variable	Frequency	Final Model
Model 1:
Native forest	yes	yes
Plantations of pine & eucalyptus	yes	yes
Wetlands	yes	no
Pastures	yes	no
Herbaceous crops of tobacco & maize	yes	yes
Shrub crops of tea & yerba mate	yes	yes
Mixed crops with small patches of shrubs & herbaceous plants	yes	yes
Naturally bare ground	yes	no
Water bodies (natural & artificial)	yes	no
Urban and infrastructure	yes	yes
Landscape heterogeneity index	yes	yes
Land use type	no	yes
Model 2:
Native forest	yes	yes
Forest plantations	yes	yes
Wetlands	yes	no
Pastures	yes	yes
Annual agricultural crops	yes	yes
Mix of agriculture and pastures	yes	yes
No vegetation	yes	no
Water	yes	yes
Perennial agricultural crops	yes	yes
Landscape heterogeneity index	yes	yes
Land use type	no	yes

). The landscape heterogeneity index is based on the Shannon–Wiener diversity index of habitat type diversity, where higher heterogeneity values represent strongly anthropogenic landscapes (i.e., a significant number of human-modified landscapes in a small area) [108]. A neighborhood scale of 96 cells × 96 cells was used to represent a home range of approximately 8 km² (8.31 km²) for the tapir, the average value determined by Medici et al. [109], a comprehensive study for tapir across three Brazilian ecosystems, including several habitat types, a breadth of individuals, and substantial duration of monitoring. A neighborhood scale of 145 cells × 145 cells was used to represent a home range of approximately 19 km² (18.92 km²) for the white-lipped peccary, a value determined from a long-term radio telemetry study in fragment of the Atlantic Forest in Brazil [110], which was similar to limited studies in continuous forest [111,112]. In comparison, a neighborhood scale of 49 cells × 49 cells was used to represent a home range of approximately 2 km² (2.16 km²) for the collared peccary [113], which represents the average value determined from an extended study using radio telemetry to evaluate seasonal variation in the rain forest of French Guiana. For peccary, both values are conservative compared to some studies but equal to values found in fragmented and continuous forests like those present in Misiones, Argentina. With all three species, the aim was to adopt a conservative approach, which was believed to provide a more accurate measure given the scale of these analyses.

The initial species-specific ENMs were run with all defined predictor variables, 12 in Model 1 and 11 in Model 2 (Table 1). Evaluation of jackknife tests of variable importance and response curves for individual variables eliminated select variables from each final model, due to either no effect or a negative effect on model performance. Both models eliminated wetlands and areas with no vegetation or naturally bare ground. Model 1 also eliminated pastures and water bodies (natural and artificial). This evaluation used the regularized training gain and test gain generated by MaxEnt, which accounts for dependency among predictor variables and compares the effect of a specific feature by itself, with a model of all features except that single feature. Efficacy of the selected predictor variables was evaluated by MaxEnt jackknife tests using test gain and area under the ROC curve (AUC) on test data, with the latter providing a threshold-independent measure of overall model accuracy [114].

2.6. Potential Species Richness (PSR)

The PSR, which is obtained by combining the three species-specific ENMs, as defined by the species-specific binary logistic threshold of suitable and unsuitable habitat, identifies areas where none (value = 0) or all (value = 3) of the species overlap in habitat defined as suitable [115]. PSR highlights areas where multiple species have suitable habitat and quantifies the number of species. This degree of overlap is used as an additional method to confirm range-restricted species and identify areas where habitat restoration could result in increased PSR. This PSR was also compared with the upper levels of PSR for carnivores (3–5 and 4–5 species; DeMatteo et al. [58]) across the N-C zone of Misiones, as well as within the DeMatteo et al. [52] multispecies corridor, to evaluate their degree of overlap across this heterogeneous and unprotected landscape.

2.7. Identification of Regions in Need of Management Strategies

The three species-specific ENMs with suitable areas defined as marginal and optimal were combined to generate a PSR, which identified areas in need of management strategies for these three species, specifically determining which areas required restoration, protection, or a combined approach. To simplify the analyses and interpretation, game species data were grouped into five classes: 0 species or unsuitable habitat, 1–2 species with marginal habitat in need of restoration, 1–2 species with optimal habitat in need of restoration and protection, 2 species with optimal habitat in need of protection, and 3 species with optimal habitat in need of protection. The overlap of these refined levels of marginal and optimal suitability was compared to the DeMatteo et al. [52] multispecies corridor and its ‘connectors’, which were those areas (26.02%) identified as threatening connectivity across the N-C zone due to their low levels in carnivore PSR and habitat integrity.

3. Results

3.1. Model Performance

Both models demonstrated high discriminatory power with AUC values indicating high accuracy in discriminating areas of species presence and absence [116]. In Model 1, the AUC values were ≥0.90 for test data from tapir [0.96 ± 0.04 (SD)] and white-lipped peccary (0.90 ± 0.02), with the collared peccary ≥ 0.85 for test data (0.85 ± 0.09). In Model 2, all AUC values were ≥0.90 for test data [tapir = 0.95 ± 0.04 (SD), white-lipped peccary = 0.95 ± 0.01, and collared peccary = 0.90 ± 0.05].

3.2. Predicted Habitat Suitability and Habitat Quality

Models were converted to a binary prediction of suitable and unsuitable habitat (Figure 2). In Model 1, FCV1 best fits the defined threshold criteria for tapir (0.0055), white-lipped peccary (0.0189), and collared peccary (0.0407). In Model 2, the results were similar, with FCV1 having the best fit for the defined threshold criteria for tapir (0.0131), white-lipped peccary (0.0183), and collared peccary (0.0735). While MTP also fit the criteria, FCV1 was determined to be the more conservative choice in both models, which was considered the better choice because it identified the maximum potential areas possible, while still maintaining a zero-omission rate for both training and test data.

When binary threshold values were assigned to habitat suitability and the area restricted to the N-C zone of Misiones (i.e., where the surveys occurred), the tapir was identified as the most restricted species (

Table 2. A summary of the proportion (%) of suitable habitat, as determined within each species-specific ecological niche model (ENM), at three scales. The northern-central (N-C) zone ALL refers to the proportion (%) of suitable habitat across the approximately 1.5 million ha in the N-C zone in Misiones, Argentina, including overlap with existing protected areas (approximately 540,000 ha). In contrast, the N-C Zone OUTSIDE PA refers to the percentage (%) of total suitable habitat that excludes areas overlapping with existing protected areas. The corridor relates to the proportion (%) of total suitable habitat captured by the approximately 400,000 ha multispecies biological corridor modeled by DeMatteo et al. [52]. For each species the total suitability, as defined by the binary threshold, plus the division of this category into marginal and optimal habitat, is reported. Each species has two ENMs reported: Model 1 and Model 2.

	ENM	N-C Zone ALL			N-C Zone OUTSIDE PA			Corridor
		Total	Marginal	Optimal	Total	Marginal	Optimal	Total	Marginal	Optimal
Tapir	Model 1	52.73	36.17	16.56	37.91	25.16	12.75	51.52	31.00	20.52
	Model 2	51.31	38.15	13.16	37.11	25.26	11.85	59.34	38.46	20.88
White-lipped Peccary	Model 1	69.80	40.85	28.95	59.08	34.13	24.95	78.01	30.77	47.24
	Model 2	62.59	37.68	24.91	51.91	27.83	24.08	69.31	30.47	38.84
Collared Peccary	Model 1	60.29	14.17	46.12	45.55	17.58	27.97	67.86	20.24	47.62
	Model 2	63.29	16.08	47.21	50.89	20.05	30.84	64.36	17.70	46.66

). While there was a difference between the models, Model 1 (52.73%) had a higher total area of habitat defined as suitable for the tapir compared to Model 2 (51.31%); this difference was minimal (1.42%) (Table 2). In the white-lipped peccary, this pattern of Model 2 (62.59%) having a lower proportion of suitable habitat than Model 1 (69.80%) was more pronounced (7.21%) (Table 2). Unlike the other game species, the collared peccary had higher values in Model 2 (63.29%) compared to Model 1 (60.29%), with a slight difference between the two models (3.00%) (Table 2). When these values were averaged across the three game species, the proportion of suitable habitat in Model 1 (60.94%) was only slightly higher (1.88%) than Model 2 (59.06%) (Table 2).

Despite this range in the proportion of suitable habitat among the three game species in Model 1 (range = 52.73–69.80%) and Model 2 (range = 51.31–63.29%), all had relatively high proportions of forest in their potential distributions, with relatively small spread or difference between the two models (7.59% and 5.90%, respectively) (Figure 3). In both models, tapir and collared peccary had the highest proportions of forest, with slightly lower levels in white-lipped peccary. Forest plantations were the most modified habitat among all three species, with each having lower levels of habitat modified by crops and mixed use (Figure 3). Breaking down these proportions of suitable habitat into those located outside protected areas provides insight into their sensitivity to ongoing landscape changes (Table 2). The average proportion across all three species showed slight variation, with very little difference between the two models (47.51% and 46.64%, respectively). The proportion of suitable area outside protected areas varied between the white-lipped peccary and collared peccary models, with the tapir having similar values in both models (Table 2). The tapir had the lowest proportions of suitable habitat outside of protected areas (37.91% and 37.11%, respectively). In comparison, the white-lipped peccary and collared peccary had higher proportions of suitable habitat outside of protected areas (>50%), except for the collared peccary in Model 1 (45.55%).

When these binary thresholds of suitable habitat were further subdivided into marginal and optimal habitat (Figure 2; Table 2), the tapir (36.17% and 38.15%, respectively) and white-lipped peccary (40.85% and 37.68%, respectively) had higher proportions of marginal habitat in both models, which contrasted with the collared peccary that had higher proportions of optimal habitat in both models (46.12% and 47.21%, respectively). All three species exhibited a consistent pattern across the two models in the proportion of marginal to optimal habitat (Table 2). The patterns were similar when the marginal and optimal habitat in the N-C zone was divided into the proportion located outside of protected areas, with the tapir and white-lipped peccary having higher proportions of marginal habitat located outside of protected areas in both models. In contrast, the collared peccary had higher proportions of optimal habitat outside of protected areas (Table 2).

3.3. PSR

This low degree of variation among the three game species in their proportion (Table 2) and type (Figure 3) of suitable habitat can help explain the PSR or spatial overlap among species (Figure 2 and Figure 4). The three species-specific ENMs captured the overlap and unique characteristics of the three game species, as indicated by the proportion of area occupied by the highest PSR level (3 species), with Model 1 (45.85%) and Model 2 (46.82%) having similar proportions and a difference of only 1% (0.97%) (Figure 4). While Model 2 showed a similar value (44.58%) at the highest PSR level (3 species) outside of protected areas, Model 1 showed a decreased proportion of high species richness (28.80%) resulting in a pronounced difference between the two models (16.78%) (Figure 4). There was a relatively substantial overlap between the PSR of these game species and the five carnivores, with the highest PSR (3 species) of both models aligning with about half of the areas defined as suitable habitat for either 3–5 carnivores (44.63% and 44.47%, respectively) or 4–5 carnivores (43.98% and 42.05%, respectively) [58].

3.4. Comparison with DeMatteo et al. [52] Multispecies Corridor Model

The proportion of suitable habitat captured by the DeMatteo et al. [52] corridor allows insight into the risk still facing each game species, as these areas represent regions lacking formal protection, with all areas intersecting protected areas excluded from these analyses (Table 2). For the tapir, Model 2 (59.34%) had a higher proportion of suitable habitat captured by the corridor than Model 1 (51.52%) (Table 2). The opposite was the case in the white-lipped peccary (78.01% and 69.31%, respectively), with the collared peccary having little difference between the two models (67.86% and 64.36%, respectively) (Table 2). While no single species showed a dramatic difference in the proportion of suitable habitat captured by either model (range across both models = 3.5–8.7%) (Table 2), differences were observed among the species. The proportion of suitable habitat captured by the DeMatteo et al. [52] corridor was the lowest for the tapir in both models. At the same time, the white-lipped peccary had the highest proportion in both models, with a more pronounced difference between Model 1 (26.49%) and Model 2 (9.97%) (Table 2). The proportion of suitable habitat for the collared peccary in both models (67.86% and 64.36%, respectively) was similar to the levels of the white-lipped peccary in Model 2 (Table 2).

While the proportion of marginal habitat captured by the DeMatteo et al. [52] corridor was similar for the tapir and white-lipped peccary, the latter had a higher proportion of optimal habitat within the corridor, comparable to that of the collared peccary (Table 2). Within individual species, the most extensive spread, or difference, between the two models for suitable habitat in the corridor was seen in the tapir (7.82%) and white-lipped peccary (8.7%), with the collared peccary being less pronounced (3.5%) (Table 2). Across the three species, the spread, or difference, between the proportion of marginal habitat captured by the corridor was higher in Model 2 (27.10%) than in Model 1 (10.76%), with this spread less pronounced for optimal habitat in the corridor between the two models (20.76% and 25.78%, respectively) (Table 2).

There was a 7.18% difference in the proportion of the highest PSR levels (3 species) captured by the DeMatteo et al. [52] corridor between the two models (45.99% and 53.17%, respectively) (Figure 4). While the level of habitat considered suitable and captured by the corridor was consistent for the lowest PSR level (1 species), the two models differed at the other levels. Model 2 (24.39%) had a higher proportion of unsuitable habitat (0 species) compared to Model 1 (18.36%), while Model 1 (23.77%) had a higher level of habitat considered suitable at a moderate PSR level (2 species) compared to Model 2 (11.05%). There was a relatively substantial overlap between the PSR of these game species and the five carnivores within the DeMatteo et al. [52] multispecies corridor, with the highest PSR (3 species) in both models aligning with about half of the areas defined as suitable habitat for either 3–5 carnivores (45.94% and 52.16%, respectively) or 4–5 carnivores (45.60% and 51.08%, respectively).

As with DeMatteo et al. [52] and DeMatteo et al. [58], both models showed higher levels of native forest (72.25% and 71.02%, respectively) versus modified habitats occupying the N-C zone (Figure 5). While the proportion of native forest occupying the corridor in both models (79.08% and 77.39%) (Figure 5) was slightly lower than the average proportion of native forest across the three ENMs in the N-C zone for both models (84.88% and 83.63%, respectively) (Figure 3), both had native forest as the most dominant habitat type. While land-use-defined categories varied according to the data used in each model, modified land-use types were grouped into four main categories: forest plantations, agriculture at different scales, pastures, and waterways.

3.5. Identification of Regions in Need of Management Strategies

As with DeMatteo et al. [58], the areas in need of habitat restoration and/or protection were evaluated based on the levels of PSR, as defined by marginal and optimal habitat (Figure 6), which provided the ability to evaluate the quality of habitat for these overlapping species. Across the N-C zone, Model 1 and Model 2 captured comparable totals of suitable habitat (75.08% and 71.87%, respectively), with similar levels of marginal habitat (12.02% and 11.05%, respectively), optimal habitat with 1–2 species (39.06% and 38.77%, respectively), optimal habitat with 2 species (12.70% and 13.38%, respectively), and optimal habitat for 3 species (11.30% and 8.67%) (Figure 7). The optimal habitat was highest across both models for PSR levels of 1–2 species, followed by habitat defined as unsuitable. The total optimal habitat defined as suitable (1–3 species) was similar between Model 1 and Model 2 (63.06% and 60.82%, respectively) (Figure 7).

When the areas in need of habitat restoration and/or protection were narrowed to the area within the multispecies corridor from DeMatteo et al. [52], the model showed that this corridor effectively captured a high proportion of suitable habitat (marginal and optimal) for game species in Model 1 and Model 2 (81.64% and 75.61%, respectively). The area at the lower end of optimal habitat (1–2 species) was found to require habitat restoration and protection (39.49% and 32.79%, respectively), with those levels slightly higher than the area at the high end of optimal habitat (2 or 3 species) that showed in need of habitat protection (31.69% and 32.99%, respectively) (Figure 7). Unfortunately, there was a dramatic decrease in the quantity of suitable habitat captured within the connectors of the DeMatteo et al. [52] corridor in Model 1 and Model 2 (52.63% and 42.67%, respectively), with these levels dropping further for the total amount of optimal habitat captured by the connectors (35.49% and 28.48%, respectively) (Figure 7,Figure 8). Of this optimal habitat, the majority needs habitat restoration and protection (1–2 species; 29.11% and 22.73%, respectively), with <7% (6.38% and 5.75%, respectively) of high PSR (2 or 3 species), with habitat not needing restoration but in need of protection (Figure 7).

4. Discussion

This study provides the first spatially explicit assessment of habitat suitability for three ecologically essential game species, with similar but disparate sensitivities to environmental disturbance—the tapir, white-lipped peccary, and collared peccary—in the N-C zone of Misiones, Argentina. These species are not only targets of illegal hunting but also serve as critical prey for large carnivores, whose persistence depends on prey availability and landscape connectivity [117]. By integrating ENMs with two alternative land-use layers, we identified shared and species-specific patterns of suitable habitat and priority areas, as well as evaluated their alignment with the multispecies carnivore corridor proposed by DeMatteo et al. [52].

4.1. Predicted Habitat Suitability

Of the three species, the large-bodied, solitary tapir was the most restricted, with only half of the N-C zone in Misiones suitable and even lower levels when the region was restricted to outside of the protected areas. In both scenarios, the proportion of marginal versus optimal habitat was higher, suggesting potential habitat flexibility in the tapir in Misiones. This pattern aligns with its ability to adapt to using a variety of habitats, including disturbed and secondary forests, plantations, and agricultural lands [118]. However, the drop in suitability levels outside of protected areas likely reflects other findings of Flesher and Medici [118] that hunting and highways restrict the movement of this species. Therefore, even though suitable habitats may exist in other regions of N-C zone of Misiones, the tapir is likely unable to move to these areas given the ever-expanding human population in rural Misiones and the extensive network of roads that cross the region, with many intersecting protected areas and others being converted to wide, paved highways [11,13]. Consequently, even though our surveys covered a diverse breadth of habitats throughout the N-C zone, if the tapir distribution is restricted to isolated populations in specific localities, it could skew the habitats considered suitable for it in Misiones, Argentina.

In contrast, across the N-C zone of Misiones, the white-lipped peccary had the broadest distribution, with the collared peccary intermediate among the three species. While both species exhibited a similar decline outside of protected areas, they remained approximately equal to or greater than 50% suitable habitat, with the peccary at the lower end of this range. While the white-lipped peccary and tapir had more marginal than optimal habitat in the N-C zone, the collared peccary was the opposite, with more optimal than marginal habitat, a pattern that was paralleled outside of the protected areas. The differences between these two group-living species may reflect their resource partitioning on the landscape scale. Specifically, the white-lipped peccary exhibits large movements in search of abundant but dispersed food resources, whereas collared peccaries utilize dispersed resources on a smaller scale [111,119,120]. While the higher levels of marginal habitat in the white-lipped peccary could align with greater habitat flexibility, it is more likely associated with the species’ need for a larger home range in fragmented or disturbed habitat where food sources are at lower levels [7]. The need to increase levels of optimal habitat for the white-lipped peccary is essential, given that these areas appear to act as important sources for recolonization during natural population cyclicity [25].

The effect decreased levels of habitat suitability outside of protected areas has on all three species is emphasized by the increased pressures in these areas, including agricultural expansion, road proliferation, and poaching, threats widely recognized in the context of the Atlantic Forest [50]. Roads pose dual risks by fragmenting habitats and facilitating human access [11], with recent data from Misiones indicating an increase in wildlife-vehicle collisions in unmitigated stretches [13]. The white-lipped peccary is especially vulnerable due to its large range requirements and susceptibility to hunting [5], while tapir populations risk genetic isolation in fragmented habitats [6]. The collared peccary’s broader tolerance may mask population declines in more sensitive species, underscoring the importance of monitoring across multiple prey taxa.

4.2. PSR and Habitat Quality

Despite the three game species varying in their home range sizes and sociality [7,109,110,113,118,119], all showed similar habitat associations, with all primarily coupled to forested areas. Among modified or altered habitats, all three species had similar averages across both models, with higher proportions of monoculture plantations and only the white-lipped peccary having an association with two unique vegetation types (herbaceous crops or a mix of agriculture and pastures). This strong intersection of the three species at the landscape scale is reflected in our finding that almost half of the N-C zone had a complete overlap of areas considered suitable for all three species—the highest PSR. The fact that this overlap is not higher than half of the N-C zone highlights that the distribution of the three game species extends beyond habitat association and emphasizes their underlying differences in habitat flexibility and sensitivity to human disturbance [7,25,111,118,119,120,121]. This disconnect between similar habitat associations and habitat predicted to be suitable is also reflected in a visual examination of the pattern, or configuration, of the individual species distribution models for the N-C zone. Specifically, a clear pattern is evident, with the tapir exhibiting an extremely patchy and discontinuous distribution in suitable habitats, where these levels decrease in the white-lipped peccary and are substantially lower in the collared peccary. The low overlap among the three species in suitable habitat, with most shared habitat occurring in marginal zones, is a pattern also reported in carnivores [58]. These findings emphasize the need to increase internal connectivity across the N-C zone of Misiones, as well as underscore the importance of an integrated approach that includes predators and their prey when identifying conservation priorities.

4.3. Identification of Regions in Need of Management Strategies

Within the DeMatteo et al. [52] corridor, all three species showed a slightly higher average in total suitable habitat compared to the N-C zone of Misiones. However, while the tapir and collared peccary maintained higher levels of marginal versus optimal habitat, the white-lipped peccary shifted, with the proportion of optimal habitat rising above that of marginal habitat, a pattern opposite to that found in the N-C zone of Misiones or outside of protected areas. The fact that both the white-lipped peccary and collared peccary have high levels of optimal habitat in the corridor, with even the tapir showing a slight increase, suggests that the corridor is capturing the habitat preferences of these three game species. This is supported by the two models, which average a 50% overlap among the suitable habitats for the three species (the highest level of PSR), with almost three-quarters of the total area suitable for between 1–3 species.

The overlap of marginal and optimal habitats among these three game species enables a refined assessment of the quality of these habitats and whether they require restoration, protection, or a combination of both management strategies. Our analyses revealed that most areas designated as suitable within the DeMatteo et al. [52] corridor fall into categories requiring either restoration, protection, or a combined approach. However, there is concern when this evaluation is limited to connectors within the DeMatteo et al. [52] corridor, with these areas defined as the regions that put the long-term connectivity across the N-C zone of Misiones at risk. Within these connectors, only about half were considered suitable habitat, which represents a 30% decrease compared to the larger multispecies corridor. This is lower than that found in carnivores, with about half of the connectors corresponding to marginal habitats for five carnivores, and the remainder to optimal habitats, with varying levels of PSR [58]. This gains relevance considering the knowledge that more than half of this corridor is privately owned, rural populations continue to increase, and these connectors are primarily composed of privately owned parcels less than 100 ha in size [52].

Given the spatial mismatch between species-specific optimal habitats in the corridor and those areas designated as connectors, a layered management approach is essential. Core zones should focus on the strict protection of high-quality habitats, while connector zones require targeted restoration that improves suitability, especially for tapirs. Measures should include antipoaching enforcement, road mitigation structures (e.g., wildlife crossings, speed reduction zones), and community engagement to promote human–wildlife coexistence, aligning with recommendations from DeMatteo et al. [52]. While the high levels of optimal habitat combined with a small home range [113] suggest that the collared peccary could serve as a resilience indicator species, the tapir’s discontinuous distribution and sensitivity to human disturbance [118], and the white-lipped peccary’s high levels of marginal habitat and need for larger home ranges in fragmented habitat [110], suggests they could function as early-warning species for connectivity breakdown. Taken together, these comparisons emphasize the value of integrating prey-based distribution models with carnivore-focused corridor designs. Aligning both perspectives can improve the ecological validity of connectivity planning by ensuring that proposed linkages not only connect landscapes structurally but also maintain functional habitats capable of sustaining predator–prey interactions.

4.4. Limitations of the Study

While our approach provides a robust framework, it is not without limitations that warrant attention. First, the models rely on available occurrence records, which may underrepresent particular species due to detectability biases related to irregular and low population densities across the heterogeneous landscape. Second, land-use layers were treated as static, though rapid agricultural expansion and infrastructure development may quickly alter habitat suitability in the near term. Third, climate change impacts—not addressed here—could further shift species distributions in unpredictable ways. Fourth, our analysis did not explicitly account for hunting pressure, which is a critical factor affecting game species distributions and densities in the region. Addressing these uncertainties will require integrating dynamic land-use and climate models, as well as spatial data on hunting intensity and accessibility, which could provide a more comprehensive assessment of threats and inform management strategies that address both habitat and human pressures. Additionally, it will be necessary to validate ENM outputs with movement and occupancy data.

4.5. Conclusions and Future Directions

Our results underscore the importance of integrating multispecies ENMs with existing corridor designs, which offer a powerful approach for evaluating landscape-scale conservation strategies. In N-C Misiones, the multispecies carnivore corridor effectively captures much of the optimal habitat for key game species, but connectivity gaps—especially for tapirs—require targeted interventions. Conservation planning should therefore prioritize (1) protection and restoration of connector areas, (2) reduction in hunting and road-related threats, and (3) landscape-level coordination that considers prey-predator dynamics.

Our findings emphasize the importance of ongoing efforts in the region to work with landowners in the multispecies corridor of DeMatteo et al. [52] to develop specific management strategies that balance human–wildlife needs [58,122,123,124,125,126,127,128,129,130]. These efforts go beyond habitat restoration and focus on protecting areas of optimal habitat and, therefore, the species that inhabit them. Landscape restoration and the promotion of forest regrowth are management strategies that can help restore connectivity across a complex landscape, especially when strategically placed in areas with high potential for species richness [58]. These efforts, combined with the protection of high-quality or optimal habitat, can optimize the proportion of forest across the landscape, which is important given the high levels of forest associated with carnivores and the prey they depend on.

Ensuring connectivity must be paired with safeguarding animal movement across the landscape [6,17,35,36]. The fact that this is a major factor threatening the long-term survival of biodiversity in Misiones, Argentina, is evident from the patterns seen during the two years (2016 & 2018) of this study. Specifically, there was a dramatic decline in the number of scats located from carnivores (41.77%) and game species (47.06%), which was paired with a large increase in the evidence of illegal hunting [95]. While the MEyRNR reacted to these findings with the formation of a group of provincial park guards tasked with priority towards antipoaching (fauna and flora) patrols (Grupo de Operaciones en Selva or GOS; Provincial Resolution No. 492), there is a desperate need to increase the personnel, logistical, and tactical resources provided.

The threats facing the fauna, flora, and humans that coexist on the landscape are magnified with Misiones showing evidence of climate change in shifting seasonality in rainfall and temperatures, which affect water availability, soil moisture levels, and risk of fire, all of which will inevitably change what habitats are defined as marginal or optimal in their suitability for carnivores and the prey they depend on [131,132,133,134,135]. While none of these findings, or the need for additional understanding, are unique to Misiones, Argentina, given that species expand across borders [134] and the human footprint ignores boundaries [136,137], the approaches taken in this study can be applied to other regions locally, nationally, and internationally.

As human pressures continue to reshape the Atlantic Forest, our approach offers a timely and evidence-based tool to support practical conservation planning and enhanced habitat connectivity in one of South America’s most threatened ecosystems. Future efforts should extend this multispecies prey-predator approach to other portions of the Atlantic Forest (Paraguay and Brazil), especially in transboundary contexts where conservation planning efforts require coordination across political borders. Ultimately, sustaining both carnivores and their prey in Misiones will demand adaptive, cross-sectoral management that balances human livelihoods with biodiversity protection.