Throughout the tropics, rapid human population growth has driven extensive land conversion, primarily fueled by resource-based economies in developing countries (Beaudrot et al., 2016). This conversion is largely due to agriculture, roads, and extractive activities, resulting in habitat loss and increased hunting pressure (Espinosa et al., 2014; Di Minin et al., 2016). Significant land-use change has occurred in Amazonian rainforests of Brazil and Peru, particularly near newly established road networks branching from the Interoceanic Highway (Chávez Michaelsen et al., 2013).

Subsistence hunting and slash-and-burn agriculture further affect mammal survival and distribution in tropical forests (Naughton-Treves et al., 2003; Peres, 2001). Human activities increase mortality through hunting, roadkill, and disrupted predation cues, altering species’ foraging and activity patterns (Mendes et al., 2020). Wildlife abundance is often more affected by hunting patterns than by habitat size or forest type (Peres, 2000). Increased hunting accessibility reduces game species and apex predators like jaguars (Panthera onca), whose densities are significantly lower in accessible areas (Espinosa et al., 2018). Predator declines closely track prey declines due to habitat loss and hunting (Boron et al., 2019; Fuller and Sievert, 2001). Fine-scale habitat use of jaguars and pumas (Puma concolor) is best explained by prey availability, with larger prey determining predator carrying capacity (Palomares et al., 2016; Santos et al., 2019; Terborgh and Estes, 2010). Predicting wildlife spatial patterns requires integrating geographic, ecological, and anthropogenic factors.

Degradation can have complex biological effects beyond overt habitat loss. Even Amazonian forests with intact canopy cover may become “empty”—that is, functionally depauperate—through unsustainable hunting (Benítez-López et al., 2019). A pan-tropical synthesis shows that more than half of nominally intact forests exhibit significant mammal declines, indicating that defaunation can precede, and be more pervasive than, deforestation itself. This underscores the need to incorporate hunting pressure into assessments of ecosystem condition, not just visible indicators of land-use change.

Defaunation, driven by human activity, exacerbates biodiversity loss in tropical ecosystems (Beaudrot et al., 2016). Benítez-López et al. (2017) estimated that mammal abundance is, on average, 83% lower in hunted areas than in undisturbed sites. Persistent exploitation by human hunters has caused population declines and near-extinctions of large-bodied mammals, particularly in tropical ecosystems (Bush et al., 2015; Jorge et al., 2013). Defaunation can have far-reaching effects on ecosystem structure and function, including disruptions in dispersal mutualisms and biomass decline (Peres et al., 2016; Gil-Sánchez et al., 2021).

We use the term “persecuted species” to describe those facing high levels of human exploitation—intensive hunting for bushmeat, illegal wildlife trade, or retaliatory killings due to human-wildlife conflict, including real or perceived threats to humans, livestock, or agricultural production. This classification, informed by regional reports, expert interviews, field observations, and a synthesis of literature and expert opinion (e.g., Beck–King et al., 1999; Jędrzejewski et al., 2017; Morcatty et al., 2020; Peres, 2000; Reyna-Hurtado and Tanner, 2007), differentiates these species from those experiencing lower hunting pressure. Examples include apex predators like jaguars and ocelots (Leopardus pardalis), often targeted for their pelts or seen as threats to livestock, and high-market-value prey species such as lowland tapirs (Tapirus terrestris) and white-lipped peccaries (Tayassu pecari).

Despite their ecological importance, the impacts of less visible threats like hunting and wildlife extraction remain poorly studied, particularly outside protected areas (Espinosa et al., 2018). Wildlife populations are typically higher in protected regions, underscoring their value for biodiversity (Benítez-López et al., 2017). However, monitoring wildlife in human-modified landscapes is crucial because ecological processes can degrade over time without apparent signs (Jorge et al., 2013). To address this challenge, we applied an occupancy modeling framework that accounts for imperfect detection while incorporating both environmental and anthropogenic covariates to estimate site use probability.

Peru has become a hotspot for wildlife trade in Latin America, with widespread exploitation and poorly regulated exportation of live animals and parts (D’Cruze et al., 2021). Community-based monitoring highlights how subsistence hunting and market demand for wild felids threaten local wildlife populations, with regional differences in hunting practices shaping local impacts (Valsecchi et al., 2023). The most commercially valuable bushmeat species include white-lipped peccary, collared peccary (Pecari tajacu), brocket deer (Mazama spp.), lowland tapir, paca (Cuniculus paca), and agouti (Dasyprocta spp.) (Espinosa et al., 2014). Hunters’ prey choice often correlates with biomass and ease of capture, particularly as communities transition from traditional subsistence hunting to market-driven activities (Bodmer, 1995). Prior to CITES implementation in 1975, overexploitation across Latin America led to the harvest of 228,376 jaguar skins and 17,301 ocelot skins, decimating populations (Mena et al., 2021). Felid declines impact prey populations and trigger broader changes in ecosystem structure (Sandom et al., 2017).

Our study focused on the occupancy and probability of use of 17 mammal species within an unprotected landscape in Peru. Las Piedras, an unprotected river basin in Madre de Dios, Peru, provides a critical lens into the effects of moderate anthropogenic disturbance—characterized by selective logging, small-scale agriculture, and dispersed hunting pressure, but without the large-scale deforestation, industrial agriculture, or road density typical of high-impact frontiers—on Amazonian mammal communities. Unlike heavily fragmented frontiers or strictly protected parks, this region retains large tracts of forest under sustained pressure from nearby settlements, making it an ideal setting to study early-stage defaunation and species-specific responses. We incorporated environmental and anthropogenic covariates—such as distances between camera trap stations and settlements and agricultural areas—into an occupancy modeling framework that accounts for imperfect detection. Although the Las Piedras region experiences moderate deforestation compared to other areas in Madre de Dios, we hypothesized that mammal occupancy would still be shaped by anthropogenic threats. We predicted that highly persecuted species—particularly the two most persecuted cats (jaguar and ocelot) and the six prey species of the highest market value (lowland tapir, white-lipped peccary, brocket deer, collared peccary, paca, and agouti)—would show the strongest responses to human disturbances. Our goal was to assess whether moderate human impact is accelerating defaunation in this biodiversity hotspot, potentially affecting ecosystem interactions across human-modified Amazonian landscapes. Although our study region is often described as a mosaic of Indigenous and non-Indigenous lands, our sampling was restricted to areas surrounding two non-Indigenous settlements—Puerto Nuevo and Puerto Lucerna. These communities differ in their cultural histories and land-use strategies, which are indirectly captured in our modeling via covariates such as proximity to agriculture and a land-use index.

Out of the 17 mammal species studied, we recorded a total of 14,849 detections across all sampling grids. The most frequently captured species were agouti (1469 captures) and brocket deer (1604 captures), while the least frequently captured were white-lipped peccary (26 captures) and jaguarundi (32 captures). Sample occasions ranged from 50 to 100 days across study sites. Model-averaged mean occupancy ranged from 0.17 (Brazilian rabbit) to 0.93 (brocket deer), and mean detection probabilities varied between 0.054 (giant armadillo and white-lipped peccary) and 0.40 (agouti and brocket deer). Model-averaged coefficients for occupancy from the top model(s) for all species are provided in Table 3.

Model-averaged coefficients for persecuted species revealed distinct responses to environmental and anthropogenic covariates, highlighting key drivers of occupancy across species (Table 3). Jaguars were negatively associated with proximity to agriculture (β = -1.58, SE = 0.96, p = 0.10) and positively associated with mixed-use areas (β = 1.90, SE = 0.87, p = 0.03), suggesting some tolerance for moderate levels of human disturbance in matrix habitats. Lowland tapirs showed a significant decrease in occupancy within 500 m of agricultural areas (β = -0.99, SE = 0.48, p = 0.04), likely reflecting the combined effects of hunting pressure and reduced habitat quality near farmland.

Environmental predictors, with the exception of prey index, were more often included in the top model overall than anthropogenic covariates. Distance to river was the most commonly included covariate in the top model across species (included for 12 out of 17 species). Puma, ocelot, collared peccary, Dasypus armadillo, and Brazilian rabbit had lower occupancy with increasing distance from river (p<0.05); giant armadillo and giant anteater showed the same pattern (p<0.1). Macro-habitat preferences varied across species. Ocelots had higher occupancy in floodplain forest over terra firme forest (β = -1.19, SE = 0.47, p = 0.01) as did lowland tapir (β = -1.42, SE = 0.45, p < 0.01), which may be due to denser vegetation cover in floodplain areas that support these species. In contrast, Dasypus armadillos had increased occupancy in terra firme forest (Dasypus armadillos: β = 0.71, SE = 0.36, p = 0.05), which may offer refuge from hunting pressure concentrated in lower-elevation floodplain habitats.

EVI, a measure of vegetation greenness, was also an important covariate in determining occupancy probability. EVI had a significant positive effect on the occupancy of giant anteaters (β = 0.76, SE = 0.26, p <0.01) and lowland tapirs (β = 0.36, SE = 0.17, p = 0.03). Margay occupancy also increased with higher EVI values (β = 0.35, SE = 0.19, p = 0.08), although this relationship was not as strong.

We found that year entered the top models for a minority of species and showed no consistent increasing or decreasing trend. Year was the only parameter retained in agouti top models and occupancy probability was lowest in year 2022 (p<0.05). Margay occupancy was lowest in 2018 (p< 0.05), while small armadillo was highest in 2018 (p<0.05). Opossum had the highest occupancy in years 2019 and 2020 and the lowest in years 2015 and 2017; however only area B was sampled in 2015 and area C in 2017, which may reflect a difference in the features of the areas as opposed to a direct effect of year. Jaguar included year in the top model, but none of the years differed significantly in occupancy probability.

Area was included in the top model set for 5 of 17 species. For green acouchi, area was the only parameter retained in top models; with area C having significantly lower occupancy probability than other areas (p<0.05). For opossum, Areas B and C had significantly lower occupancy than Area A (p < 0.05), which is consistent with our results for the year effect. Similarly, we found collared peccaries had significantly lower occupancy Areas B and D than Area A (p < 0.05). Paca had lower occupancy in Area B (p < 0.10), while margay models included area but showed no significant effects.

For some species, covariates did not yield significant results, but these patterns still offer insights into their ecological preferences and responses to human activity. White-lipped peccaries were recorded in only 26 instances, and none of the covariates significantly explained their site use. This low detection rate and small sample size limit the interpretability of these results, suggesting potentially low population density or movement patterns that reduce detectability. Similarly, jaguarundis showed no clear association with macrohabitat or anthropogenic covariates, except for a strong relationship with small prey abundance (β = 0.56, SE = 0.23, p = 0.02). Other species, such as Brazilian rabbits and brocket deer, exhibited varied responses to mixed-use and agricultural areas, indicating a complex interplay between tolerance to disturbance and habitat preference.

Detection results (final wording). Detection probability varied widely among species and was strongly influenced by trail type; trail placement was a significant predictor for 10 species (p < 0.05). Large felids were much more likely to be photographed on human trails/roads: jaguar (β = 0.89, SE = 0.21, p < 0.01), puma (β = 1.74, SE = 0.18, p < 0.01), and ocelot (β = 1.47, SE = 0.11, p < 0.01). Several prey/generalist species also showed higher detection on human trails—lowland tapir, agouti, and common opossum (all p < 0.05). In contrast, paca, white-lipped peccary, and brocket deer were more frequently detected on wildlife trails (i.e., lower detection on human trails; p < 0.01). These patterns are consistent with many taxa using linear openings as movement routes, while certain prey avoid larger, human-made paths.

The number of camera nights operable also had a significant impact on detection probability for 3 species. Longer operability periods increased detection for green acouchis (β = 0.20, SE = 0.11, p = 0.07). Conversely, operable nights had a negative coefficient for collared peccary (β = -0.16, SE = 0.05, p< 0.01) and tapir (β = -0.18, SE = 0.05, p<0.01). Often, a longer camera deployment leads to more detections; however, there are some cases when animals may be detected less due to changes in movement over time or other factors. The relatively low detection probabilities for species such as white-lipped peccaries and giant armadillos highlight the challenge of monitoring these species using camera traps, as their low densities and wide-ranging behavior may reduce the likelihood of encounters with cameras. Across all 17 species, model-averaged detection probabilities ranged from 0.05 for white-lipped peccary to 0.40 for agouti, reflecting substantial interspecific variation in detectability.

Our study revealed that mammal occupancy patterns in the Peruvian Amazon are shaped by both anthropogenic and environmental factors, with species responding differently based on their ecological roles and sensitivity to human activities. Persecuted species can act as reliable indicators of ecosystem degradation, guiding conservation in human-modified landscapes. Some showed significant responses to human disturbance, while others were more influenced by proximity to rivers and macrohabitat type. These findings underscore the importance of using indicator species to monitor ecosystem health, particularly in regions where subsistence hunting and habitat fragmentation have synergistic impacts on forest vertebrates (Peres, 2001). Persecuted species offer valuable insights into the intensity and spatial extent of anthropogenic pressures, providing an early-warning tool for conservation efforts in highly impacted landscapes.

Lowland tapirs showed a significant negative relationship with agricultural proximity, consistent with studies linking agricultural expansion to reduced habitat quality through hunting pressure, habitat loss, and restricted water access (Burs et al., 2023; Medici and Desbiez, 2012). Tapirs are frequently targeted by subsistence hunters near agricultural areas, where spotlighting is particularly effective. Forest fragmentation also reduces the abundance of herbaceous plants crucial to their diet (Williams-Linera, 1990). However, their occupancy in mixed-use areas suggests they may persist in less-disturbed matrix habitats with forest connectivity and essential resources.

Jaguars showed a higher probability of using mixed-use areas but avoided agricultural sites. These results suggest that jaguars may use mixed-use areas for movement or as secondary habitat (McBride and Thompson, 2018), while agricultural areas may pose higher risks due to habitat loss and increased human-wildlife conflict (Amador et al., 2013). Agricultural plots often lack sufficient vegetation cover and prey availability, and the presence of livestock without protection increases the likelihood of retaliatory killing by farmers. Mixed-use areas, despite some human disturbance, may retain enough forest cover and prey species to support jaguar persistence (Morato et al., 2018). Similar to findings from other studies, our results emphasize the need for proactive conflict mitigation measures in areas bordering protected forests (Foster et al., 2010; Zeilhofer et al., 2014). Although we did not explicitly include community identity as a covariate in our models, several anthropogenic variables—such as distance from settlements, proximity to agriculture, and a mixed-use land classification—reflect the spatial influence of local land-use practices. Puerto Nuevo is characterized by low-intensity, dispersed hunting and timber harvesting, while Puerto Lucerna engages more intensively in agriculture-driven forest clearing. These differences help explain spatial patterns in human pressure across our study areas and are captured by our land-use index, which distinguishes between protected and mixed-use zones. We also did not include interaction terms among covariates or species in the modeling framework; however, interactions—particularly between land use, proximity to settlements, and river corridors—may reveal more complex habitat relationships and should be considered in future work.

Environmental variables, particularly proximity to rivers and macrohabitat type, significantly influenced species occupancy. Puma and ocelot occupancy were higher near rivers, likely due to prey availability and access to water, similarly Dasypus armadillos and collared peccary occupancy decreased with increasing distance from rivers, reflecting their reliance on riverine habitats for forage and water (Keuroghlian and Eaton, 2008). Seasonal inundations in floodplain forests are critical for determining ungulate habitat use (Bodmer, 1990), with similar patterns observed across Neotropical systems (Boron et al., 2019; Dias et al., 2019). Given the ecological importance of riverine habitats for both wildlife and human communities, future conservation strategies should prioritize protecting these areas from further encroachment.

Macrohabitat type also played a key role in shaping species distributions. Ocelots preferred floodplain forests, which may offer increased prey availability and denser understory for cover (Di Bitetti et al., 2010). Although ocelots avoided agricultural areas, they appeared to tolerate moderate disturbance in mixed-use zones, demonstrating their adaptability in fragmented landscapes. These findings align with research in the Peruvian Amazon showing that ocelots shift their activity patterns temporally to avoid human presence without significantly altering their spatial distribution (Zwicker and Gardner, 2024). Similarly, Dasypus armadillos favored terra firme forests, consistent with their preference for well-drained soils (Gonçalves et al., 2022). Jaguarundis were the only predator species positively associated with small prey availability, suggesting a potential prey dependence. Although their cryptic nature and low number of detections complicate definitive conclusions, their habitat associations underscore the importance of diverse ecosystems for supporting predator-prey dynamics and maintaining overall biodiversity.

The strong responses of highly persecuted species—such as jaguar, lowland tapir, white-lipped peccary, brocket deer, collared peccary, and agouti—to anthropogenic pressures demonstrates their sensitivity to ecosystem degradation. Jaguar, lowland tapir, and brocket deer avoidance of heavily altered areas reflects broader patterns of tropical forest degradation, where the loss of apex predators and large herbivores can disrupt ecological balance and trigger cascading effects. Jaguars, as apex predators, can reflect broader ecosystem changes, while brocket deer and lowland tapir offer insights into prey availability and hunting pressure. The loss of large carnivores like jaguars can trigger significant trophic cascades, reshaping entire ecosystems and altering prey dynamics (Ripple et al., 2014). Their persistence is critical for maintaining ecological balance and biodiversity and they can aid in identifying areas of concern and guiding targeted conservation efforts. Integrating persecuted species data into land-use planning not only helps prioritize critical areas for intervention but also enhances community-led initiatives by providing tangible, locally relevant conservation targets. Similar to the use of large mammals as indicators of global human impacts (Morrison et al., 2007), these persecuted species act as sentinels of ecosystem health in tropical forests. Each of these species displayed significant responses to anthropogenic or environmental covariates, providing early-warning signals of ecosystem degradation. Lowland tapirs and white-lipped peccaries are particularly valuable indicators due to their sensitivity to habitat disturbance and hunting pressure (Fragoso, 2004; Medici and Desbiez, 2012).

White-lipped peccaries were recorded at very low rates and we had no detections in 2018 or 2019. Though a number of covariates were included in the top model for occupancy probability, none were considered significant. This could be attributed to factors such as low population density, recent recolonization following a decline, or wide-ranging movement patterns that reduce detectability in camera trap grids. Previous studies have documented large-scale movements and temporary population declines driven by hunting pressure, disease, and natural cycles (Fragoso, 2004; Fragoso et al., 2022). Our results suggest that more data are required to tease apart factors influencing their occupancy; however, we did find that white-lipped peccaries were less likely to be detected on trails, indicating they are less likely to use human made paths. The variation in detection of white-lipped peccaries, including years with no detections, highlights the importance of long-term monitoring to distinguish short-term fluctuations from long-term trends. Understanding white-lipped peccary population dynamics is crucial given their role as ecosystem engineers and key prey for top predators.

By monitoring these species, conservation practitioners can detect early signs of defaunation and habitat degradation. Protecting riverine habitats and establishing buffer zones around agricultural areas will help mitigate human-wildlife conflict and reduce hunting pressure. Community involvement in conservation initiatives is key to ensuring long-term success. Programs that engage local communities in wildlife monitoring and sustainable land management can help build local capacity and increase stewardship. Developing policies that integrate scientific findings into land-use planning and conflict mitigation can further support these efforts. Collaborative approaches that incorporate both ecological and social perspectives will be essential for balancing conservation priorities with the needs of local communities.

Our study highlights the complex relationships between human activities, habitat characteristics, and mammal occupancy in the Peruvian Amazon. By focusing on persecuted species as human impact indicators, we contribute to a growing body of evidence supporting the use of sensitive species to monitor ecological change and human influence. This approach provides an effective framework for addressing biodiversity loss while fostering sustainable management practices. Targeted strategies addressing specific anthropogenic pressures, coupled with adaptive management and long-term monitoring, will help maintain resilient and biodiverse tropical ecosystems. Future research should continue to refine this framework, incorporating both ecological and social dimensions to balance conservation goals with community needs. Given the global significance of tropical forests for biodiversity and climate resilience, coordinated international efforts are critical to protect these landscapes before further irreversible losses occur.