Human activities have substantially modified natural ecosystems (Dirzo et al., 2014), leading to ecological consequences, such as changes in resource availability, species distribution, and interspecific interactions (Liu et al., 2007). In many cases, anthropogenic influence can exert a more pronounced effect on wildlife populations (e.g. changes in distribution or abundance; Wolff et al., 2019; Newbold et al., 2020) and communities (e.g. changes in species richness, evenness, and trophic structure; Roemer et al., 2009; Magioli et al., 2019) than the natural environmental conditions (Nagy-reis et al., 2017).

Wild mammals have been severely affected by human activities, primarily through overexploitation and habitat loss and degradation, which pose significant threats to their survival (Galetti et al., 2021). These impacts are particularly evident in tropical forests (Dirzo et al., 2014) such as the Atlantic Forest, a biodiversity hotspot (Ribeiro et al., 2011). In this biome, most of the remaining fragments are highly defaunated, with declining populations of medium- and large-sized mammals (Bogoni et al., 2018). This context may exacerbate shifts in species dominance and alter trophic cascades within forest fragments (Roemer et al., 2009), with distinct disturbance regimes driving different tendencies in the species assemblages. For instance, the extirpation of top predators aligned with increasing food supply in agricultural fields may drive an increase in large herbivores abundance, consequently impacting plant assemblages in forest fragments; but overhunting and habitat loss may restrain this group, leading to the ecological release of smaller herbivores, usually seed predators (Pires and Galetti, 2023).

Mesocarnivores – predators weighing less than 15 kg and occupying an intermediate position in the trophic web (Roemer et al., 2009) – are another functional group of medium-sized mammal that may benefit from top predator extirpation and from resource surplus in anthropogenic matrices (Pires and Galetti, 2023). These species contribute to key ecological functions, including predation of invertebrates and small vertebrates, and primary and secondary seed dispersal (Cazetta and Galetti, 2009). Although mesocarnivores are considered resilient and expected to increase in abundance with the removal of larger competitors (Estes et al., 2011), 91% of small carnivorans are threatened by habitat loss – particularly from agricultural expansion (85%) – as well as by vehicle collisions and retaliatory hunting throughout their distribution (Marneweck et al., 2021). Human disturbances can have both negative and positive impacts on mesocarnivores. Changes in the distribution and abundance of mesocarnivore population can serve as indicators of environmental change, emphasizing their effectiveness as sentinel species (Marneweck et al., 2021).

One key conservation strategy to slow down populations’ decline and mitigate human impacts, is the establishment and maintenance of protected areas (PAs), which can safeguard the most sensitive species. In this study, we investigated differences in the occupancy of mesocarnivore assemblages in two PAs under different management regimes – strict protection and sustainable use – in the northeast Atlantic Forest of Brazil. Combining camera-trapping and hierarchical community occupancy models, we estimated mesocarnivore species richness at the PA scale and assessed the influence of fine-scale environmental and anthropogenic factors on species-specific occupancy patterns. We empirically tested expectations that mesocarnivore occupancy is positively associated with i) strict levels of PA protection and larger size (Gray et al., 2016), ii) flat terrains (O’Malley et al., 2024) and iii) mature and better-preserved forest patches (Regolin et al., 2017), while being iv) negatively related to anthropogenic stressors (access roads and PAs limits; Goulart et al., 2009; Nagy-reis et al., 2017). However, species responses might be trait-dependent, including dependence on forest habitats, or being vulnerable or benefiting from human disturbance.

We obtained 1310 independent records of 11 mesocarnivore species over a total of 13,609 trap-days (Table 1, Appendix 3), all expected to occur in the study region, except Leopardus pardalis (Magioli et al., 2023). Mesocarnivore species richness was higher in PNPB (11) than in RPPNEV (9), as estimated from the first-stage model with all species included. Two mesocarnivore species were exclusively detected in PNPB: Conepatus semistriatus and Galictis cuja, albeit with only one detection each.

Detection probability estimates at the community level was 0.045 (Global mean across all species, areas, and surveys), although there was considerable interspecific and temporal variation (Appendix 4a). Five mesocarnivore species - C. semistriatus, G. cuja, Procyon cancrivorus, Herpailurus yagouaroundi, and Leopardus guttulus – were detected few times (<5) in each area, resulting in detection probabilities below 0.1 throughout all surveys and unprecise occupancy estimates (Appendix 4b). As such, four species were considered in the second-stage model for inferences on mesocarnivore distribution – Cerdocyon thous, Didelphis aurita, Eira barbara, and Nasua nasua – and data for all Leopardus species were combined at the genus level, including detections with uncertain phenotypic differentiation, and modeled as a distinct functional species.

The occupancy probability of the species did not show obvious differences between PAs (Fig. 2), and we observed a high degree of interspecific and temporal variation across surveys, which was often asymmetric between areas. Nasua nasua was more widespread in RPPNEV during all surveys (Ψ > 0.75). Didelphis aurita had higher occupancy in PNPB in surveys 2 and 3 (Ψ = 0.788 and 0.750, respectively), E. barbara showed an increased occupancy in PNPB during surveys 3 and 4 (Ψ = 0.573 and 0.679, respectively), and Leopardus sp. had higher occupancy in PNPB in surveys 2 and 4 (Ψ = 0.661 and 0.431, respectively) and in RPPNEV during survey 3 (Ψ = 0.612).

Fig. 2.

Species-specific occupancy probability estimated in the second stage of the model in each survey: survey 1 (August–October/2018), survey 2 (April–June/2019), survey 3 (August–October/2019), and survey 4 (August–October/2021), conducted in Pau Brasil National Park (PNPB) and Private Reserve of Natural Heritage Estação Veracel (RPPNEV), southern Bahia, Brazil. Points above the diagonal line indicate greater species occupancy in the RPPNEV (y-axis), while those below it indicate higher species occupancy in the PNPB (x-axis). The error bars depict 95% Bayesian credible intervals. Species codes: crdt- Cerdocyon thous; ddla-Didelphis aurita; erbr- Eira Barbara; lsp-Leopardus sp.; nsns-Nasua nasua.

No clear occupancy response to site-scale covariates was observed at the community level, with mesocarnivore species often exhibiting contrasting associations with environmental and disturbance factors. Cerdocyon thous occupancy was strongly associated with more disturbed forest sites, i.e. with a lower basal area (βBASAL = −0.97, [range: −1.85 to −0.19]; Fig. 3A), closer to roads (βROAD = −0.57 [−1.33 to 0.05]; Fig. 3C). Conversely, N. nasua occupancy moderately increased in sites with higher basal area (βBASAL = 0.42 [−0.12 to 1.04]; Fig. 3A) and that of D. aurita moderately increased further away from roads (βROAD = 0.41 [−0.04 to 0.90]; Fig. 3C). The occupancy of N. nasua and Leopardus sp. was positively associated withthe proximity to PA limits (βLIMITS = −0.51, [−1.45 to 0.05] and βLIMITS = −0.28 [−0.70 to 0.09], respectively; Fig. 3D). None of the species had their occupancy substantially influenced by terrain slope (Fig. 3B), and E. barbara did not respond to any covariate. Detection probability for most mesocarnivores was strongly associated with camera-trap placement (Fig. 3E), with bimodal responses across species. While detectability of N. nasua and E. barbara was higher in the forest (βCAMPLACE = −0.38 [−0.68 to −0.08] and βCAMPLACE = −0.98 [−1.77 to −0.19], respectively), the detection of C. thous and Leopardus sp. increased in roads or trails [βCAMPLACE = 3.78 [2.89–4.72] and β CAMPLACE = 0.19 [−0.13 to 0.49], respectively).

Fig. 3.

Effect size of site-scale environmental and disturbance covariates on occupancy (A-D) and detectability (E) probabilities for the five selected species in Pau Brasil National Park (PNPB) and Private Reserve of Natural Heritage Estação Veracel (RPPNEV), southern Bahia, Brazil. Points are posterior medians and error bars represent 95% Bayesian credible intervals (BCI). Colors indicate the direction of the effect (negative in red, positive in green, no effect in gray), while the shade of red and green colors reflect the probability of an effect, calculated as the proportion of the posterior with the same sign as the mean: darker tones indicate strong support (≥0.95) and lighter tones indicate moderate support (≥0.90). Vertical lines depict mean (solid line) and 95% BCI (dashed lines) community effects.

Our findings revealed that the mesocarnivore community structures in PNPB and RPPNEV are complex and heterogeneous. Both PAs support a diverse range of species, with PNPB exhibiting greater species richness. Overall, species detectability was low, and occupancy showed strong interspecific and temporal variation. There were no uniform response to the environmental and anthropogenic factors at the assemblage level l; with contrasting patterns emphasizing the importance of species-specific evaluations. In general, environmental factors induced stronger but fewer responses in site occupancy by mesocarnivores, while anthropogenic factors seemed to cause greater variation in habitat use, potentially facilitating species occupancy. Notably, there was a tendency for species to occupy the borders of the PAs, likely benefiting from foraging opportunities in adjacent agricultural areas.

Despite the relatively lower abundance of carnivores compared to other trophic groups (Newbold et al., 2020), the low detectability observed in our study compared to other areas in the Brazilian Atlantic Forest, such as Iguacu National Park (da Silva et al., 2018) and Vale Natural Reserve (Wolff et al., 2019), suggests low population abundances within the studied areas (Burns et al., 2019). The detectability variation in response to camera trap location was consistent with other studies in the Atlantic Forest (Di Bitetti et al., 2010). The higher species richness in PNPB is likely due to its larger size—which supports more habitats, greater resource availability, and larger populations (Connor and McCoy, 2001)—rather than its legal protection status, as the sustainable use PA has an intensive surveillance system, a pattern already observed in southern Bahia (Magioli et al., 2021b). Furthermore, the higher species occupancy in PNPB underscores the critical role of protecting large areas for mesocarnivore conservation in the Atlantic Forest.

The temporal variation in species-specific occupancy across both PAs is particularly intriguing for a relatively stable tropical forest. Such differences can be partly explained by low precision of occupancy estimations (a season's credible interval overlaps the mean of other seasons), but may also reflect natural population fluctuations combined with human interference. Notably, the most variable occupancy estimation was observed for D. aurita, the smallest species among those investigated, whose life cycles are shorter and reproductive rates are higher (Emmons and Feer, 1997), traits often associated with faster population dynamics.

Our study highlights the complex and heterogeneous mesocarnivore assemblage supported by the studied PAs, but also stresses the effect of anthropogenic factors on species occupancy. The low detectability and occupancy of the mesocarnivores in our study compared to other areas in the Atlantic Forest implies either naturally low abundance within these PAs, or other limiting factors impacting these populations. The absence of assemblage-wide responses to environmental and anthropogenic factors underscores the importance of species-specific strategies for species conservation. Interestingly, some species tended to use the edges of the PAs, suggesting the use of agricultural areas as supplementary food resources and part of their habitat (Magioli et al., 2019). These findings, coupled with the absences of the apex predator (P. onca) and the largest mesopredator (L. pardalis), along with and the presence of species typically associated with open areas (C. semistriatus), suggest that anthropogenic disturbances in both areas have shaped the structure of local mesocarnivore assemblages.

Marília Marques: Writing - original draft, Visualization, Formal analysis, Data curation. Marcelo Magioli: Writing - review & editing, Supervision, Data curation. Pedro Monterroso: Writing - review & editing, Supervision, Formal analysis. Gonçalo Curveira-Santos: Writing - review & editing, Formal analysis. Camila Righetto Cassano: Writing - review & editing, Supervision.