The Amazon Basin is home to diverse freshwater turtle species (Buhlmann et al., 2009), which face ever-increasing anthropogenic pressures, including overexploitation, deforestation, urbanization, and hydroelectric dams (Bárcenas-García et al., 2021; Fagundes et al., 2021, 2018; Johns, 1987; Smith, 1979). The yellow-spotted river turtle, Podocnemis unifilis, Troschel, 1848 is one of the most widely distributed and abundant Amazonian turtles (TTWG, 2021); however, its populations have declined significantly in recent decades due to overexploitation and habitat loss (Escalona et al., in press, Páez et al., 2015; TTWG, 2021). A modeling analysis of P. unifilis populations predicted severe (≥50%) and rapid (<50 years) future losses across 60% (5.3 M km²) of its pan-Amazonian range (Norris et al., 2019). It is currently assessed as Vulnerable (A1acd) by the IUCN Red List (TFTSG, 1996), pending publication of the most recent assessment of Vulnerable (A2cd+4bcd) (Escalona et al., in press)

Effective nesting area protection is essential for the integrated conservation management of P. unifilis and, likely other, Amazonian freshwater turtles (Andrade et al., 2022; Norris et al., 2019; Páez et al., 2015). Nest site selection is crucial for the fitness of P. unifilis, influencing egg-hatching success, hatchling development, and population demography (Refsnider and Janzen, 2010; Escalona et al., 2009; Iverson, 1991; Norris et al., 2019). Although predominantly aquatic, P. unifilis lay their eggs on land close to the water’s edge (Thorbjarnarson et al., 1993). These terrestrial nesting areas are susceptible to anthropogenic threats (Fagundes et al., 2021; Steen et al., 2012), and are prone to habitat degradation and fragmentation due to the close associations with human needs and resource use (Leal et al., 2016; Paula et al., 2018). Furthermore, the spatial and temporal predictability of nesting facilitates exploitation, with females and their eggs targeted during the nesting season (Pezzuti et al., 2010; Norris and Michalski, 2013; Smith, 1979). Nesting areas can, therefore, represent an evolutionary trap for P. unifilis (Quintana et al., 2019). The environmental cues females use to select nesting areas do not account for human-induced threats, which in turn makes populations vulnerable to overexploitation (Quintana et al., 2019; Fagundes et al., 2021).

Assessing the vulnerability of nesting areas is necessary to develop more effective conservation strategies for Amazonian turtles including P. unifilis (Fagundes et al., 2021). Vulnerability is the degree to which a system, subsystem, or system component is likely to experience harm due to exposure to a hazard, either a perturbation or stressor (Turner et al., 2003). Spatial analysis of species occurrence and/or habitat use helps to identify areas that are most vulnerable to anthropogenic activities (Kukkala and Moilanen, 2013; Ippolito et al., 2010). This is often used in the risk-hazard model, where the vulnerability component called exposure is frequently evaluated to quantify the overlap between the variable of interest and hazards/threats (Auber et al., 2022; Ippolito et al., 2010). Predictions can then be made to estimate the degree to which vulnerable areas are likely to change when exposed to multiple threats to inform management and conservation actions (Turner et al., 2003; Ippolito et al., 2010). By identifying critical habitats and factors contributing to vulnerability, conservationists can prioritize areas for protection and restoration efforts to prevent local extinction and facilitate the recovery of impacted populations (Norris et al., 2019).

Remote sensing has been used to identify and map freshwater turtle nesting area vulnerability in the Brazilian Amazon (Fagundes et al., 2021). While such broad-scale assessments provide much relevant information, more local scale data is needed to generate insight at the level of river basins (Ippolito et al., 2010). This is especially true for P. unifilis, which uses nesting areas that can be too small and/or too close to canopy forests to be detected by remote sensing (D. Norris pers. obs.). Local-scale information is therefore necessary to develop more effective management plans for riparian habitats and other important areas for turtle nesting, as well as mitigation or compensation measures for environmental impacts.

Here we use data on P. unifilis nesting areas upstream of a recently operational hydroelectric dam to assess their vulnerability to anthropogenic activities. We asked (a) what is the level of nesting area vulnerability, (b) which variables determine nesting area vulnerability, and (c) where should different conservation strategies be prioritized?

Vulnerability varied substantially across the nesting areas, ranging from 0.01 to 0.81, increasing along an upstream to downstream gradient ( Figure 3 ). Vulnerability varied significantly among zones (Kruskal-Wallis, P< 0.0001). The post-hoc tests indicated that differences were significant among all zones (adjusted P< 0.0001 for all pairwise comparisons).

The differences in vulnerability were associated with different threats among zones ( Table 2 ). The least vulnerable zones were those without commercial activity and further away from houses located along the Falsino River, representing 37% of the overall river length surveyed. Commercial activity and distance to the nearest house were important in determining the difference between Zones 2 and Subzone 3.1, which had similar distances to the town and hydroelectric dam ( Table 2 ). Distance to the nearest house was also relevant to determine the difference in vulnerability between the upstream Subzone 3.2 and the downstream Subzone 3.1 ( Table 2 ).

The first three components of the PCA accounted for 97.4% of the variance in the data (56.8, 24.3, and 16.4% for components 1, 2, and 3, respectively). Commercial activity (r = 0.97) and distance to the nearest house (r = 0.88) were most strongly associated with components 1 and 2, respectively ( Supplementary Figures S3.1, S3.2 ). The only variable that was not strongly correlated with any of the first three components was forest loss in a 500m radius area (r = 0.33, 0.33, and 0.47, respectively). The other five variables were strongly associated with component 3 (r > 0.80).

HCPC identified six clusters that differed in nesting area average size and vulnerability, three of them with high conservation priority ( Table 3 ). The highest priority covered a quarter of the river stretch studied (24.4%, 48.2 km) and was separated into three distinct spatial clusters. Subzones 3.1 and 3.2 were presented mid to large size and low-vulnerability clusters, while large nesting areas with high vulnerability occurred downstream, close to both the town and the dam ( Figure 4 ).

As expected, there was a clear gradient of nesting area vulnerability relative to anthropic impacts. The proximity to both a town and a hydroelectric dam generated surprisingly high vulnerability immediately upstream for a region with such high environmental protection. Urbanization and hydroelectric dams have numerous, often synergistic, impacts on freshwater turtles (Lovich et al., 2018; Bárcenas-García et al., 2021; Chaves et al., 2021; Conway-Gomez, 2007). The vulnerability gradient we found reflects the well-documented limitations of terrestrial protected areas to adequately conserve aquatic species (Anderson et al., 2019; Nogueira et al., 2021; Mollmann et al., 2022). Indeed, the protected areas in the study region appear to have limited capacity to conserve aquatic species such as turtles (Norris et al., 2018).

Our findings, which identify commercial activity and proximity to human infrastructure such as towns and hydroelectric dams as drivers of nesting area vulnerability, align with a broad consensus in the literature on Amazonian chelonians (Sousa et al., 2024; Fagundes et al., 2021, 2018; TTWG, 2021). While we employed a fine-scale, multi-factor vulnerability index, other research on P. unifilis and its congeners present similar conclusions. Our finding that vulnerability increased near human settlements is explained by empirical results from other regions, which documented unsustainable harvesting of eggs and females as a direct consequence of human proximity and urban demand for turtle meat and eggs (Mayor et al., 2024; Chaves et al., 2021; Conway-Gomez, 2007; Pezzuti et al., 2010). Furthermore, the risks we identified are reflected at an evolutionary scale in population genetic studies of Podocnemis species. Anthropogenic impacts and landscape features, including natural rapids (barriers analogous to dams), can restrict gene flow and are associated with significant population structure, demonstrating the potential long-term consequences of the vulnerability we mapped at a local scale (Dos Santos et al., 2016; Agostini et al., 2024; Cárdenas-Barrantes et al., 2024). Thus, by capturing threats invisible from space our vulnerability predictions complement broader scale assessments (Fagundes et al., 2021, 2018; Sousa et al., 2024).

Commercial activities and proximity to human habitations increased vulnerability both among and within the zones/subzones. Remote sensing provides increasingly fine-scale models of land-use and land-cover change (Pettorelli et al., 2018; Kissling et al., 2018). However, variables such as the presence of commercial activity and distance to the nearest house are still difficult to obtain via remote sensing, as they are almost undetectable (Peres et al., 2006). The relevance of these variables, which were not included in previous remote sensing studies (Fagundes et al., 2021), shows the importance of more detailed field data to inform vulnerability assessments along Amazonian rivers. Although fieldwork-based publications are declining (Ríos-Saldaña et al., 2018), integrating field data to complement remote sensing approaches is vital to generate robust scientific evidence.

By integrating results on nesting area size, vulnerability, and geographic location, it was possible to identify and prioritize conservation and management options. While all high-priority clusters included relatively large nesting areas, recovery and restoration actions are most urgently needed in the most downstream cluster to mitigate losses associated with the dam and the nearby town (Bárcenas-García et al., 2022; Norris et al., 2018). These actions should include the creation of new nesting areas along 15 km of river currently devoid of nesting areas due to flooding by the hydroelectric dam.

Although scientific evidence can inform solutions and actions to conserve vulnerable nesting areas, the greatest challenge is to act and implement them. For example, the hydroelectric power plant operator needs to act to reduce vulnerability and mitigate impacts associated with the dam. Evidence documenting the effects of the hydroelectric dam on nesting areas was presented to the relevant authorities in 2018 (MPAP, 2023). To date, this process is stalled, and neither the state environmental agency nor the company operating the dam has acted to create and/or restore turtle nesting areas. While enforcement actions appear to be ineffective, community-based actions may generate viable alternatives to counteract threats (Andrade et al., 2022; Campos-Silva et al., 2018; Páez et al., 2015).

Community-based management of nesting areas could turn vulnerability into resilience (Norris et al., 2020, 2018). Areas near Porto Grande, for example, are potentially ideal for environmental education and community-based management actions, which have been successful for the conservation of Podocnemis turtles outside of protected areas (Campos-Silva et al., 2018; Rivera et al., 2021; Andrade et al., 2022). Conservation strategies may include buffer zone establishment, community engagement, and sustainable land-use practices to mitigate the impact of human activities (Andrade et al., 2022; Bárcenas-García et al., 2021; Steen et al., 2012; Campos-Silva et al., 2018). More specifically, fishing and gravel extraction could be restricted in zones or subzones with important nesting sites. Buffer zones could also be established around nesting sites to reduce human disturbance.

The index we used is adaptable to local contexts across the species’ range. For example, while we used distance to the nearest house as a proxy for human activity, other variables like distance to settlements or number of residents could better reflect harvest pressure in different areas. Further studies are needed to test and refine the index across the species range, assessing its generalizability and identifying context-specific modifications.

Several caveats should be considered when interpreting our results. First, a more detailed understanding of how anthropogenic impacts affect P. unifilis populations is crucial. Currently, there is a lack of data on adult population dynamics and the effects of threats like deforestation and dams, hindering comprehensive impact assessments (Páez et al., 2015). This information is crucial for the development of effective mitigation or compensation measures in the context of infrastructure projects. Second, grouping continuous values into classes (e.g., low to very high) remains subjective. Future studies should evaluate the sensitivity of vulnerability assessments to these classifications. Third, some of the variables that we included reflected the local context. It may be necessary to include additional variables to adequately reflect the threats in other areas. The inclusion of additional social data would strengthen planning and stakeholder involvement, e.g., reporting wildlife crimes and the participation of local communities in conservation and development programs (Pimid et al., 2022; Bennett et al., 2017; Turner et al., 2003). Finally, we did not use variable weights due to a lack of robust evidence on the relative importance of different threats. The relative importance of different threats could likely depend on factors such as the ecosystem, geographic location, and socio-economic context, among others. Although there is a lack of data for the study region, expert opinion or meta-analyses could provide valuable information to enable a weighted approach in future assessments.

Our findings provide evidence of the local scale vulnerability of turtle nesting areas and identify priority clusters for conservation actions. Clusters provided a spatially explicit baseline to inform the choice of priority nesting areas for local scale population studies. Conservation actions should initially be focused on maximizing nesting success by reducing losses due to flooding (Norris et al., 2020) and unsustainable harvest levels (Quintana et al., 2019), and then advance to include environmental education to reduce hunting of adults (Páez et al., 2015). Integrating turtle nesting area assessments in state-wide biodiversity conservation planning could benefit broader conservation strategies for the region’s human well-being, minimizing conflicts between stakeholders.