Humans have fundamentally transformed ecosystems and shaped the distribution of biodiversity on Earth, mainly through agricultural activities (1), which is expected to continue in the coming years. Food demand is forecasted to increase to meet the needs of the rising human population, estimated to reach 9.73 billion people by 2064 (2). The agricultural intensification and expansion, and the expected continued urbanization and industrialization needed to meet global consumption (3) represent a major threat to biodiversity and conservation in the coming decades. These trends have caused high global concerns in terms of increasing risks of emerging zoonotic disease (EZD) (4). Recent studies have evaluated potential approaches to balance the need for food production and the need for biodiversity conservation and its derived ecosystem services (5, 6). These studies have increased our ability to foresee potential consequences of future land-use trajectories and to develop adapted planning tools, particularly within the framework of the UN Sustainable Development Goals [SDGs; (7)]. Among relevant planning tools, scenario modeling allows to examine how land-use changes may influence biodiversity, food production and provision of ecosystem services under different pathways of future human development and policy choices (8).

With the COVID-19 pandemic, there has been a renewed recognition of the public health risks which arise from direct and indirect interactions between humans and wildlife (4). Indeed, around 70% of Emerging Zoonotic Diseases (EZD), and almost all recent pandemics (e.g., MERS- CoV, Ebola), have been associated with the increased interactions among wildlife, domestic animals, and humans (9), as a result of deforestation, urbanization, and intensification of agricultural systems (3, 10–12). Global changes in the mode and intensity of land-use conversion are expanding the hazardous interactions between people, domestic animals, and the wildlife reservoirs of zoonotic diseases; indeed, wildlife species that harbor higher pathogen loads are more likely to occur in human-managed ecosystems (11). When the rates of habitat conversion are high, the rate of transmission of pathogens between the species inhabiting conserved habitats and those inhabiting converted habitats is expected to be higher (13, 14). Although the mechanisms and landscape contexts that influence the events and rates of transmission of pathogens between wildlife species and humans are not completely understood yet, some operational concepts based on a landscape approach have been recently proposed by Plowright et al. (15), including: 1. the “land-use induced spillover” concept which implies that land-use change can lead to spillover events through a series of complex steps, involving pathogen infection in wildlife and shedding of the pathogen by person-to-person transmission; and 2. the “landscape immunity” concept which implies that some ecological conditions can, when combined, maintain and strengthen the immune function of wild species within a particular ecosystem, while preventing the conditions that lead to high pathogen prevalence and shedding.

In this context, tropical areas are particularly prone to land-use induced spillover, as compared to regions in other latitudes, since they have a larger number and diversity of animal species, as well as a greater number and diversity of their associated pathogens, and they are experiencing rapid land-use changes. This fast land-use conversion is increasing the potential for novel interactions between domestic animals, humans and wildlife, along with their associated pathogens. Indeed, many wild animals cannot survive a drastic reduction in the amount of habitat within a landscape, which could limit access to food and other resources (16, 17), and those that can adapt will increasingly occupy anthropogenic areas, thus increasing the risk of EZD. Evidence also suggests that EZD is related to (i) land-use changes in a given area, (ii) changes in the interaction networks of different species driven by native vegetation loss, (iii) local hunting practices, (iv) changes in behavior of pathogen vectors, and (v) the sharing of newly altered landscapes by people, domestic and wild animals, and insect vectors of disease (15, 18–20). Therefore, it is vital to understand how different land-use scenarios might affect species distributions. This can then aid in the identification of potential future EZD risks, which can be used to inform landscape management policies (16, 21).

Based on these outlined assumptions and concepts, we explored potential land-use scenarios to reduce EZD risks using a landscape modeling approach. It is important to highlight that our study was designed to only address strategies for cases where the dynamics of pathogen transmission among species are poorly known and the concepts of land-use induced spillover (LUIS) and landscape immunity (LI) can be used very broadly. First, we projected changes in the native vegetation cover, as a proxy for habitat, under three potential scenarios: (a) business as usual (BAU), (b) avoidance of LUIS, and (c) LI. We then evaluated the effects of the proportion of native vegetation cover on the occupancy probability of mammalian species (including groups that are likely to harbor human-shared pathogens, such as Artiodactyla and Rodentia), and we analyzed how the different land-use scenarios might affect the distribution of species in the landscape, and thus the risk of EZD. We developed our analysis on a region at the interface between the Brazilian Cerrado and the Atlantic Forest, areas recognized as biodiversity hotspots facing strong agribusiness expansion and infrastructural development (22). The rapid land-use changes in this region can lead to local extinctions, increased human-wildlife conflict, and result in loss of ecological services. Therefore, our study area can serve as a good model to understand how land-use changes could be related to the emergence of EZD and how landscape modeling frameworks such as the one presented can aid in identifying strategies that reconcile the potentially conflicting objectives of supporting both agricultural production and biodiversity conservation in tropical regions.

An unprecedented number of studies have been undertaken to reconcile efforts in food production, safeguarding of nature, and protection of public health [e.g., (4, 39–41)]. However, to date, there has been little to identify an approach to harmonize these different, and sometimes conflicting targets using empirical data. Our results showed that combining information on current and modeled land-use changes may be a strategy to reduce EZDs outbreaks based on appropriate biodiversity management. Proposed future scenarios based on empirical information on land-use change allowed us to explore and assess the implications of managing EZDs, considering both assumptions from the “land-use induced spillover” and the “landscape immunity” frameworks introduced by Plowright et al. (15), with implications for biodiversity protection, agriculture, and restoration in the region, particularly for regions under limited known pathogen dynamics landscape, such as the Bodoquena Plateau in Brazil.

Our results showed how changes in the amount of vegetation cover might influence the occupancy probability of different species of mammals across deforestation gradients. For most of these particular species their category of zoonotic disease status is unknown, nonetheless, and according to several authors (see (38) for a review), aspects related to their order might inform on their potential status such as rodents known to be the group with the larger proportion of zoonotic diseases per species with at least one potential zoonotic virus. This information, together with the assumption proposed by Faust et al. (13), in which an intermediate proportion of habitat converted may have an impact on the risk of transmission on the whole community, allows us to draw some suggestions about the relevance of using land-use change scenarios as a strategy to reduce the potential of zoonotic disease outbreaks.

The implementation of any EZD strategy proposed here would reduce native vegetation lost by 2050. Under the BAU scenario, there would be an increase in the proportion of the landscape with native vegetation cover under the occupancy threshold values, implying a restoration effort of 846 km². Nonetheless, our results have clear practical implications of the implementation of the ALUIS scenario and the LI scenario. For example, our study highlights the need for an increase in restoration targets under the ALUIS and LI scenarios that are similar to the restoration efforts needed in the BAU scenario, but slightly higher (965.2 and 1,007.6 km² respectively).

Planning and managing landscapes to manage the populations of pathogen hosts and vectors, as we proposed here, can be considered an area-based management strategy for preventing future zoonotic epidemics (39). The ALUIS scenario assumes that pathogen transmission between species inhabiting conserved and converted habitat is highest when rates of habitat conversion are intermediate (13–15). In our case, in the intermediate category of native vegetation coverage (i.e., between 60 and 15%) most mammalian species show a variation on the probability of occupancy across the gradient of remaining vegetation cover. It implies that around this level of native vegetation coverage, we can see the most interactions (co-occurrence) between forest-dependent (losers) and habitat generalist (winners) species (Figure 5). Although controversial, a potential consequence of this pattern is that more species with fast-paced life-history strategies will replace those with slower-paced life histories as host communities become fragmented or disturbed, and biodiversity loss is fast or abrupt. This could result in an increase in host exposure to parasites by shifting host behavior (36). However, an ongoing question is how to best define this intermediate category. In the case of managing the dilution effect, a key strategy might be to conserve the whole community of hosts and vectors in a landscape. Here a key question, with a long history in conservation ecology is: how much native vegetation should be conserved to maintain the whole community and ecological processes in a given area? Both the ALUIS and the LI scenarios require information about the response of hosts, vectors, and associated pathogens to land-use changes in a region which are not available for most tropical regions, including the region we studied. To overcome this challenge, we expanded the idea of using an ecological threshold as a limit of land-use changes (21, 42). In this context, we propose the inclusion of levels of host thresholds to land-use changes as proxies to define ALUIS and LI scenarios, in other words “safe operating landscapes for EZD avoidance” (43).

Thus, for defining “safe operating landscapes for EZD avoidance” in any region, development at the landscape scale should involve a process of co-design of plans with all the relevant stakeholders, and with specific objectives and thresholds designated (44). As such, the scenarios proposed in our study should not be seen as the most plausible scenarios, but instead encompass a number of potential projected futures (45) that should be collectively discussed, considering levels of risks, and cost-benefits for people, animals and ecosystems, in the coming years. With this in mind, the costs of the interventions and strategies based on different “thresholds of the hosts” could be more rationally assessed. This assessment must include not only the direct costs of an intervention (e.g., restoration), but also changes in future income (opportunity costs) for different stakeholder groups, as well as analysis of cost-benefits for monitoring and prevention actions (e.g., avoiding deforestation) (46). Each scenario implies different strategies, including governance, trade-offs between agricultural lands and protected areas, different cost-benefits for relevant stakeholders, and uncertainties. For example, in our study case, the BAU scenario will clearly result in larger areas for food production in the region, however, it would also result in a larger decrease in the surface of native vegetation, resulting in levels below the thresholds allowing the occurrence of selected indicator species (e.g., more than 846 km² will be below the threshold for the most sensitive species such as P. tajacu). This will decrease the possibility of a dilution effect and increase chances of spillover. On the other hand, if decision-makers choose to go towards the ALUIS scenario, the restoration required to reach the proposed native vegetation thresholds would be around 965 km² which clearly will imply economic costs, while in the LI scenario, the value will be 4.3% more. Thus, the benefits and costs to different stakeholder groups (farmers, politicians, tourists, the environment) in the future are highly variable, depending on the desired outcomes, and thus the desired scenario. However, decisions about which scenario to select as a target should be determined with inputs from a range of actors in the region.

Area-based management as a single strategy is not enough to build “safe operating landscapes for EZD avoidance” (15, 39). It is critical to consider the linkages that connect food production, conservation, and pathogens, including new relationships between wild and domestic animals, hunting activities, agricultural practices, trade, transport, markets, and food consumption patterns. In fact, we should also consider the socioeconomic and environmental interactions over distances (telecoupling process) as relevant aspect on the implementation of safe operating spaces. Petrovan et al. (39) proposed more than 150 recommendations for preventing future zoonotic epidemics. Many of them directed to farmers, the tourism sector, and governments. For instance, hunting is a context-dependent variable which is not currently evaluated in Bodoquena Plateau, but given its clear importance for potentially driving future outcomes in terms of EZD, it should be considered in “safe operating landscapes for EZD avoidance” plans for our study area. Other strategies that should be contemplated include: (i) developing certification for registered farms that demonstrate enforcement of guidelines on safe production standards; (ii) introducing legislation to reduce and control the spread of animal agriculture and overlap/proximity to tropical forest/undisturbed ecosystems; (iii) conduct risk assessment at the wildlife–livestock–human interfaces to inform the type of emergency response or longer-term planning for prevention and control of zoonotic pathogens. However, it is important to highlight that the implementation of any strategy to reduce the risk of zoonotic epidemics should be decided collectively and discussed by the different sectors, such as cattle ranchers, and should be more closely studied, monitored and reported. For example, environmental police and state government programs could adopt a preventive approach, informing the public about the health of wildlife species killed through illegal hunting or found dead on the roadside. Moreover, monitoring the movement and behavior of animals can bring information on the frequency of invasion of cultivated lands and, consequently, the likelihood of increased contact with humans (13). For example, in the Bodoquena Plateau, peccaries are the most frequent wildlife species seen in crop fields around protected areas and are generally responsible for some crop damage. Understanding interspecies contact is key for studies of landscape spillover, and these will be reflected in seasonal and interannual epidemics (47). Therefore, determining the frequency of human or livestock interactions with different wildlife taxa, and the health status of these species, could be seen as a key strategy to guide policy interventions to minimize public health risks and safeguard animal health in the future. The SISS-Geo (sissgeo.lncc.br), a free platform, available on smartphones and on the web, for monitoring the health of wild animals in natural, rural and urban environments, is a promising system in the direction to improve EZD monitoring in Brazil. To integrate and operationalize these ideas, we also strongly support the suggestion by Patz et al. (12) about the importance of creating “Centers of excellence in ecology and health research and training,” based on a network of international and regional universities and/or research institutes, with close links to the surrounding communities. More broadly, we agree with many studies which suggest the implementation of the “One Health” approach to integrate all health centers and health professionals [e.g., (48, 49)] in a collective effort. For the region of the Bodoquena Plateau, we also suggest these initiatives should explicitly include agriculture as a critical issue, and thus suggest the implementation of: “Centers of excellence in biodiversity and ecological services, food production, and health research and training.”

In terms of modeling and datasets, we made a number of assumptions that should be taken into account in future studies. First, some responses of zoonotic disease to land-use changes could be idiosyncratic (50), and not all species present clear thresholds respond non-linearly to land-use change. In these cases, the scenarios and management strategies that should be considered, the model built for each species separately, as well as the limits for acceptable risk, should be decided based on information and perception of risks from relevant stakeholders and societal contexts. Second, in terms of biodiversity responses to landscape dynamics, our model only considers the thresholds for native vegetation cover. The functional role of the landscape matrix, potential corridors, and the habitat configuration in shaping biodiversity, remains largely unknown in the region. Agricultural practices are another aspect to be considered because they dramatically modify the landscape and have an effect on species’ occupancy (26). Third, our scenario approach does not evaluate global demands for food, changes in the demand for certain types of food, such as plant-based products, which can drive important changes in the agricultural systems. This evaluation is significant for future studies because international food demands can play a critical role in the dynamic of land-use changes in regions, which can be particularly affected by commodity production, such as soy and beef (51). Fourth, our modelling process does not explicitly consider human social interactions, adaptation, and changes in people’s behavior and movements across the landscapes, which are key for the future of this region (52), which in 2022 received more than 200 thousand tourists.

The original contributions presented in the study are included in the article/supplementary files. Further inquiries can be directed to the corresponding author.

FOR and JMO-Q conceived and formulated the idea. BB, FVN, FOR and AG developed the mathematical models. CCS, IM, ANA, AGO, AVN, SPGB and JMO-Q data collection. FVN, BB, AG and JMO-Q analyzed the data. All authors contributed to the article and approved the submitted version.