Linkages between conservation and health

Conservation, biodiversity, and land and resource governance programs employ a variety of interventions to reduce habitat conversion and forest degradation, including (1) to improve the enabling environment for conservation, (2) to change behavior and reduce the threat of deforestation or degradation, and (3) to relieve direct stress on species and ecosystems through strengthened land, water and wildlife management. Common conservation and biodiversity interventions include protected area management, conservation enterprises, payment for ecosystem services, reforestation, law enforcement, wildlife demand reduction and behavior change campaigns, conservation planning, education and training, and institution strengthening, as well as market-based and direct economic payment schemes (Although not directly related to land-use change, another increasingly important category of interventions for biodiversity and zoonoses objectives center on wildlife trade. These interventions aim to halt or reduce the farming, killing, butchering, trade, transport, and consumption of live and freshly killed birds and mammals in provincial towns and metropolitan cities, while not undermining the health and well-being of subsistence hunters.). Similarly, land and resource governance interventions include land-use planning, natural resource management, clarification of rights, policy and legal reform, awareness raising, and many others. Reforestation is an increasingly popular development intervention, and is the subject of a growing number of current and planned research and evaluation initiatives. In 2021 at COP26 in Glasgow, world leaders committed to conservation and restoration of forests by 2030. Increasingly, private corporations and donors have committed to planting trees and restoring forests.

Theoretical and lab-in-the-field evidence strongly suggests that avoided deforestation will improve ecosystem health and have flow down benefits for human health. Correspondingly, forests and wetlands provide ecosystem services that help maintain water quality, and there is a small but growing body of evidence about the impact of improved watershed health on human health outcomes. Studies have shown that forested watersheds may reduce water-borne sediments and contaminants, and thus improve raw water quality in ways that moderate the effectiveness of water quality treatments. In a multi-country study, Herrera et al. 2017 find that higher upstream tree cover is associated with a lower probability of diarrheal disease. Thus, deforested watersheds could undermine the effectiveness of water, sanitation, and hygiene interventions—with the implication that improving forest conditions can have a positive effect on related health outcomes [4, 5].

Correspondingly, ecological restoration, through natural forest replanting versus monoculture tree plantations, is increasingly presented as a nature-based and cost-effective solution to mitigate climate change, biodiversity loss, and emergence of novel zoonotic pathogens [6]. Ecologists have shown that restoration of degraded forest habitat, habitat creation, or actions to improve connectivity between isolated fragments can change the size, structure, and density of wildlife populations [7] and the movement of individual animals [8], which suggests cascading implications for pathogen transmission. Forest restoration could also benefit human health by increasing biodiversity (in particular species that do not amplify viral zoonotic disease risk), lowering disease prevalence in reservoir populations, and reducing reservoir host-human contacts and hazard.

Similarly, studies have highlighted human population density, mammalian diversity, active conversion to land cover, and recent loss of forest as important drivers of zoonotic disease emergence [9, 10]. Zoonotic pathogen spillover is especially associated with the conversion of tropical broadleaf forests [9–12]. Land-use change and habitat conversion effect many of the risk factors for zoonotic disease spillover such as biodiversity loss, changes in the distribution of zoonoses host species, and increased human-wildlife contact. Land-use change alters the interface between recipient human hosts and reservoir hosts, and can lead to increases in cross-host exposure to viruses. Land conversions can increase stress and reduce immune responses among wildlife, increasing their pathogen load. Intensification of certain agricultural practices can increase resources for certain animal hosts of zoonotic diseases and decrease resources for others. Even the structure of farms impacts viral spillover risk, as smallholder farmers increase their forest access as a result of more interface or contact points.

Conversion of natural habitat to agriculture or other land uses leads to biodiversity loss [13], changes in the distribution and abundance of zoonotic disease host species [14], and increased exposure, frequency, and intimacy of contact between wildlife [15], humans, and domesticated species [16]. Thus, there is a strong reason to expect if deforestation is prevented, drivers of zoonotic disease risk such as contact between humans and wildlife and infection prevalence in wildlife will be reduced. Improving biodiversity of habitats can serve as a disease buffer, and virus shedding by wildlife hosts will be reduced in areas where they have less stress and sufficient nourishment.

Limited counterfactual evidence

However, significant evidence gaps exist about the impact of development interventions to improve forest outcomes, along with the downstream effects of improved ecosystem health on human health outcomes [17]. There are a number of conservation interventions that have little to no rigorous empirical evidence about intervention effectiveness on conservation and biodiversity outcomes, including interventions focused on alternative livelihoods, forest restoration, wildlife demand reduction, resource protection, conservation enterprises, eco-tourism, among others [3]. Thus, we are missing an understanding of whether most conservation initiatives even have the expected ‘first-order’ effects. In most cases data is not collected on biophysical and health outcomes that would enable rigorous study of whether and how environmental outcomes link to health outcomes.

Impact evaluations offer an approach to rigorously evaluate the effectiveness of conservation programs or policies. Impact evaluations rely on a counterfactual to measure the causal impact of a program, or the difference in outcomes caused by a program or intervention and not by other external factors. With the exception of the program of interest, the counterfactual areas should be as similar to the treatment areas as possible. The designation of an evaluation as an impact versus performance evaluation ultimately depends on the validity of the control group or counterfactual.

Impact evaluations employ experimental and quasi-experimental methods to identify treatment effects. Experimental approaches measure the causal impact of programs through randomized assignment (e.g., randomized control trials (RCT); (2) quasi-experimental measure causal impacts but without randomization (e.g., difference-in-difference [DID], regression discontinuity, and statistical matching).

Impact evaluations can be prospective where the research design is embedded in the intervention. The most rigorous prospective method for constructing the counterfactual is through random assignment (i.e. RCT). Impact evaluations can also be retrospective, in which the control is constructed after the intervention has begun or concluded and the opportunity for pre-intervention baseline data collection has passed).

Finally, impact evaluations can be designed to measure causal impact for a combination or ‘package’ of interventions or seek to isolate the impact of one or more interventions. Conservation programs typically include a bundle of interventions not easily disentangled, such resource protection, alternative livelihoods, tenure, and knowledge/outreach raising.

RCTs are an experimental approach to impact evaluation (While some methods are deemed “quasi-experimental” because they attempt to replicate a randomized design, RCTs are experimental because they are based on random assignment.). In this approach, random assignment, such as through a coin toss or random number generator, determines who may participate in the program so that those assigned to participate in the program are, on average, the same as those who are not. RCTs are a powerful tool for providing robust evidence regarding program effects, as well as for testing causal theories to inform program decisions and to justify resource allocations.

Quasi-experimental methods utilize a counterfactual group that is not determined through a randomized process. The comparison group is purposefully selected or constructed to create the best and most credible comparison for treatment areas. Thus, quasi-experimental methods ultimately represent data-driven methods for evaluating the causal effect of a program; a large-scale data collection effort and econometric methods must be employed to ensure that selection bias between treatment and counterfactual groups is minimized. In theory, a well-designed quasi-experimental method can be a powerful statistical tool to minimize selection bias between treatment and control groups. However, they require stronger assumptions than randomized selection, and there are several methodological limitations because of bias in the selection of treatment areas.

Quasi-experimental methods are more common for conservation and land and resource governance interventions. For evaluating forest condition outcomes, projects apply a matching approach to develop synthetic controls of forest pixels. Comparison areas and settlements are identified from non-intervention areas that are matched on key biophysical and human population characteristics.

A counterfactual approach is critical for generating evidence about whether interventions achieve their expected objectives as well as identifying potential unintended, negative consequences that can undermine the goals of the project. In 2006, a working group led by the Center for Global Development argued that impact evaluations were underutilized in international development for three main reasons: (1) high up-front costs and long-term, diffuse benefits (public goods problem); (2) pressure to design and procure programs (action bias); and (3) difficulties for funders and other stakeholders to adapt in response to evidence contrary to expectations or perceived ‘truths’ (confirmation, inertia, and optimism biases). Despite these challenges, many development sectors have embraced impact evaluations, including global health, education, and, increasingly, tenure and governance [18].

Coined by Ferraro (2009), ‘implementation science’ is a more recent term that encompasses the impact evaluation approach. Implementation science focuses on interventions that lack rigorous evidence to determine whether these interventions are working, why, and for whom. It aims to overcome some of the challenges noted above that are faced by proponents of impact evaluations by putting the focus on using rigorous evidence to improve program effectiveness. The term moves away from perceptions of judgment or academic aloofness denoted by ‘evaluation science’ [1]. By pairing ‘implementation’ (action) with ‘science’, ‘implementation science’ seeks to bridge the artificial divide between researchers and practitioners committed to achieving positive outcomes for conservation and biodiversity.

Yet, the application of an implementation science approach through counterfactual-based research remains rare in conservation [2]. Evidence on the effectiveness of conservation strategies is vastly inadequate, with poor design, lack of scope, and too few examples [19]. RCTs have not been applied to the evaluation of conservation interventions and are limited in research on land and resource governance [20]. There is little data on biodiversity outcomes. Many studies involve the simple monitoring of indicators or case studies, and there is an over reliance on self-reported behavioral indicators [21]. These approaches are insufficient to establish cause-and-effect relationships [22, 23].

Recent reviews have found very few counterfactual studies on the impact of tenure interventions on biodiversity conservation, forest condition, and forest conservation, and none on wildlife trafficking [24]. Unfortunately, there is even less rigorous empirical evidence about the effectiveness of wildlife demand reduction campaigns. Studies on effectiveness of demand reduction lack rigor, with selection, data quality and design challenges. While there have been more recent advances to understand the impact on demand reduction such as before-after study designs using survey data, the lack of robust research designs make it difficult to draw insights to inform future efforts [25]. Finally, at present, no impact evaluations have been completed on the effectiveness of interventions to mitigate the risk of zoonotic disease spillover.

Wildlife conservation through the protection and restoration of ecosystems has the potential to reduce the risk of zoonotic disease spillover [26–28]. However, this concept has yet to be demonstrated on a large scale in a real-world setting, and evidence specific to the outcomes from programs designed to support forest restoration is currently absent. A 2012 review of research on land use, land use change, and infectious disease found 302 papers published on the subject [12]. However, more than one third of these papers were themselves reviews, and the underlying studies remain fairly poor due to the challenges of obtaining data at the wildlife-human interface. Most studies are cross-sectional surveys of land use classes. Only two studies [29, 30] focus on reforestation and zoonoses risk. Although reforestation or restoration are core components of many countries’ climate change mitigation commitments, the implications for zoonoses risk are unclear, with results likely dependent on the type of reforestation and biodiversity species richness.

Counterfactual or causal studies are not prioritized in the conservation space, relative to other development sectors, for a variety of reasons [24]. Measuring impact on biodiversity is thought to be methodologically challenging and expensive [1, 21]. Frequently cited challenges include: large variability in ecological outcomes; long time lags between intervention and ecological response, programs with multiple interventions, complex spillover effects (e.g., due to species movement); large spatial scales of environmental processes; and data constraints, including a lack of reliable biodiversity measures. Studies of the wildlife trade face similar challenges, including multifaceted drivers of demand that complicate the study of consumption, lack of data on actual behavioral change, and delayed responses for long-term behavior change.

Overall, despite the strong theoretical foundations for a One Health approach, very few studies have evaluated the impacts of integrated conservation and health programs using a rigorous counterfactual framework. There is a significant knowledge gap in our understanding of (1) what interventions will most effectively support improved forest outcomes and biodiversity, and (2) whether and how targeting those interventions will impact health outcomes and the risk of zoonotic disease spillover. These gaps limit our ability to design effective evidence-based programs.

Methods—Case studies from Madagascar, Zambia, and West Africa

The insights from this paper are built on lessons learned from three case studies where researchers and practitioners are collaborating to put implementation science for conservation and health into practice. Specifically, these case studies are drawn from USAID’s Health, Ecosystems, and Agriculture for Resilient Thriving Societies (HEARTH) Activity portfolio, which engages private sector partners to collaboratively implement integrated development programs that aim to conserve high-biodiversity landscapes and improve the well-being and prosperity of communities that depend on them. Across the HEARTH portfolio, USAID has made a strong commitment to rigorous monitoring, evaluation, research, and learning, which has provided the necessary foundation for impact evaluations to be designed and conducted.

The first case study of Health In Harmony’s work in Madagascar is not a HEARTH activity, but it has informed USAID’s approach to integrated One Health programming. The second and third case studies are from designing impact evaluations for two HEARTH programs, which were selected for impact evaluations based on a variety of factors, including buy-in from key stakeholders, and after undergoing an in-depth assessment of the feasible evaluation design options. Overall, these case studies highlight an implementation science approach to conservation, biodiversity, and integrated conservation and health programs, and all three provide examples of counterfactual-based studies to ground the theory in the previous section.

In addition to the three specific cases highlighted in this paper, the authors’ understanding of implementation science has benefited greatly from almost a decade of designing and conducting impact evaluations for integrated and cross-sectoral programs in the fields of tenure, governance, health, among others. The paper also draws heavily on feedback and findings from workshop series, seminars, and reviews commissioned by USAID over the past two years related to improving implementation science for integrated conservation and health programs, including the evaluation feasibility assessments for the two HEARTH case studies, a presentation of lessons learned from designing cross-sectoral impact evaluations at USAID’s internal Environmental and Natural Resource Management (ENRM) speaker series in 2024, and a panel at the 2023 annual conference of the American Evaluation Association (AEA). This includes feedback from donor partners, government contractors, conservation organizations, local NGOs, and academics.

The case studies for this paper were selected because—at present—they are the only known examples of integrated conservation, climate, and health projects that adopt an implementation science approach. The authors of this manuscript represent researchers and practitioners who are collaborating on the two HEARTH cases that are funded by USAID, and through the process of preparing the assessments, webinars, and presentations mentioned above have created space for the authors to collaboratively reflect on their experiences with the case studies and identify lessons learned relevant for a broader audience of donors, practitioners, and evaluators. The methods that were used to develop this manuscript drew loosely on the “pause and reflect” approach taken at USAID to support adaptive management of programs, and are distinct from methods that are required for designing and conducting counterfactual-based research and evaluation. The case studies of focus are as follows:

Early MEL engagement and due diligence

An implementation science approach should be integrated into program design at project conceptualization. Independent third-party MEL experts should be embedded in design teams at the concept stage (activity start-up workshops) as well as throughout the design and procurement process to assist in the development of MEL plans and their implementation, as well as during inflection points (pause and reflect workshops) to adapt the program in response to evidence and data. Engagement at the project design stage can help manage expectations early on for implementing partners and other stakeholders regarding the time and resources needed for an effective learning approach.

Evaluation design teams typically conduct a due diligence scoping trip to the program area well after a program is procured. In parallel with local qualitative data collection, the evaluation experts on the design team should conduct a literature review to reduce information asymmetries and constrain systematic human biases in program design. The findings from this due diligence, along with baseline data, then inform evaluation technical feasibility and adaptive management of the program. However, due diligence conducted earlier in the design process could inform not only technical feasibility but design of the program itself, saving the need to adapt after the fact. Indeed, during the initial stages of program procurement, many development interventions do not have detailed implementation details, as workplans are at a high level. But those more detailed, community-level implementation plans are necessary in order to identify the feasibility of various evaluation designs, as well as assess critical program design components, such as weaknesses in the theory of change and problems of co-location for integrated programs.

Include research language in the program scope of work

An effective way to promote change is to include appropriate language about the integration of a counterfactual approach into program implementation design and roll-out in the legal agreements for funding recipients. The program scope of work should clearly signal to offerors that a program may be subject to an evaluation and will be required to incorporate evaluation findings in work planning and pause-and-reflects to adapt the program as needed. To the extent possible, MEL specialists should work with donors on providing design specifics to facilitate rigorous research and evaluation as part of a program scope of work.

Not every program should be subjected to an evaluation. But for those programs for which it makes strategic sense to use evaluation methods, the evaluation team can help develop a MEL plan in partnership with the program implementers. In this way the underlying program benefits from evaluation findings through adaptive management and the evaluation team leverages M&E data collected by the program. To manage expectations, the solicitation should include as a technical requirement the evaluation-implementer partnership and the requirement of adaptive management in response to evaluation data.

Localization

There are significant information asymmetries between many policy makers and practitioners and local people and communities [35]. As early as possible in the design process local communities should be engaged to overcome information asymmetries and improve program design. This can be achieved through a co-production or co-creation process in collaboration with local beneficiaries. Due diligence should help build the interventions from the local context, based on listening sessions with the relevant beneficiaries. Additionally, there remains a tremendous need to improve the quality of the data captured through implementing partners’ MEL systems. This requires setting up the necessary MEL protocols early in the program design process in collaboration with the research and evaluation team. In some cases, this might also require improving or supporting the technical skills for some implementing partners. It also requires contract and budget considerations, as local partners much have the logistical capacity and budget support that is necessary to produce high quality research and to consistently collect, manage, and share data. In the context of conservation, donors and evaluation teams are supporting local capacity and networks for a more sustainable approach to monitoring forest conditions and biophysical conditions, using free and accessible tools, such as Global Forest Watch tool. This supports monitoring throughout the lifetime of the activity.

Among donors, there is clear demand for more localization and distribution of more of the workshare for MEL to local firms. Traditionally, development evaluations are very extractive—firms or universities are hired to collect data—but analysis and reporting is completed almost entirely in the Global North, despite growing capacity in the Global South. However, MEL firms are pivoting to an approach that attempts to delegate parts of analysis, such as qualitative coding, descriptive statistics and contextual analysis, to local firms in low and middle income countries. One major challenge to this is the tension between the timeline for local capacity building and the quality expectations from donors. Some solutions could include relaxing some quality or “polishing” requirements for research products produced through localization, supporting timelines that enable additional time to train, review, conduct quality assurance checks, etc., or expand budgets to facilitate significant capacity building while holding to typical timelines and the production of ‘polished’ products.

Interrogate the theory of change

The project design team should refine the theory of change based on the qualitative data from due diligence as well as the literature review results. An implementation science approach adds value by strengthening a program’s theory of change and connecting interventions to the evidence base during the feasibility stage. Baseline data is used to challenge assumptions and promote more effective, adaptive programming. The theory of change should be specific to the very local context of the program, including the interventions to be implemented. Part of the theory of change should cover how the treatment units are expected to respond to the resources offered by the program and thereby change their baseline perceptions, incentives, norms, or behaviors.

For the Zambia EKNA program, as part of adaptive management, the theory of change is continually revisited throughout activity design and implementation, including pause and reflect workshops that have participation from local community members and that were informed by baseline needs assessments with local communities. In the case of the Zambia EKNA evaluation, the research team worked closely with the implementing partner and USAID to evaluate the validity of the theory of change based on feedback and findings from the adaptive management process. This process revealed several weaknesses in the project design that the implementing partner corrected. For example, one issue encountered upon community mapping for baseline data collection was a lack of geographic co-location for certain conservation and health interventions that were expected to work in concert as part of an integrated program. The adaptation was to change the implementation plans to include co-location, greatly increasing the learning potential about the benefits of integrated programming. Another challenge that was uncovered during the field scoping and interviews with potential beneficiaries was the risk of unintended negative consequences of sustainable agriculture interventions leading to increased land clearing. The adaptation was the introduction of requirements to prohibit land clearing.

Consider multiple evaluation design options

The counterfactual-based evaluation approach to integrated conservation and health can take on multiple forms. The case studies highlight that quasi-experimental methods can be considered, in combination with a mix of impact and rigorous performance evaluation methods for a single program. In the case of landscape level projects (including many conservation and biodiversity projects), a whole of project RCT is obviously not feasible. However, specific activities or interventions within a project’s strategic approach might be amenable to evaluation through an RCT. This could include many community or individual level activities, such as alternative livelihoods, behavioral change, or knowledge and awareness raising interventions designed to reduce wildlife trafficking.

Another form is to approach the human health component as an ‘add-on’ to explore as part of a planned evaluation of a conservation, biodiversity, or wildlife demand project. This is highlighted by the RESTORE case study where health is not part of the intervention package but the evaluation will explore a number of environmental and health linkages. In the case of projects seeking to understand the influence of land-use change on zoonotic spillover risk, this would not require fundamentally different programming—it can instead focus on collecting additional data sources and measuring additional data sources and outcomes for interventions related to land-use change or wildlife demand reduction. This means that—if there is sufficient funding—modules to measure health outcomes can be tacked onto a conservation or biodiversity intervention where we anticipate improvements to ecosystem health. This could also include adding on modules to measure zoonotic risk factors, in cases where we expect improvements in habitats for wildlife with a greater risk of zoonotic spillover.

Interdisciplinary research team

Interdisciplinary research teams are needed, including human health professionals, epidemiologists, wildlife disease ecologists, forest ecologists, veterinarians, and social scientists. This is an especially clear theme in many recent policy and synthesis papers on zoonoses. A long term counterfactual-based research program to identify causal relationships between land-use conversion and human health outcomes requires a diverse team of researchers and policy professionals to integrate sociological, economic, and ecological drivers.

Mixed-methods designs

As per best practices with other prospective large-scale research efforts, an implementation science approach to conservation and health programming should be designed as a mixed-method study. A mixed-method study integrates two or more research methods, usually drawing on both quantitative and qualitative data sources. Generally, mixed methods evaluations can provide a deeper understanding of why change is or is not occurring and capture a wider range of perspectives. Mixed-method evaluations may use multiple evaluation designs, for example incorporating both randomized controlled trial experiments and case studies. The qualitative strategy serves two primary purposes, 1) to add a social context within which to situate the statistics, and 2) to add depth to the overall study and the descriptive data. Quantitative data answers “what” and “whether” programs had an impact, whereas the qualitative data addresses critical questions about “how” and “why” interventions are having an effect.

Leverage innovative and existing data collection

Implementation science requires comprehensive data collection—and this will have significant cost implications for monitoring conservation, biodiversity, and health outcomes. To help mitigate costs, research and evaluation efforts should leverage existing data sources from implementing partners, along with the increasing availability of high-resolution satellite imagery. Moreover, citizen science—which involves the data collection and analysis by local communities and the general public—could provide an innovative source of cost-effective and higher frequency data. Building off of the advanced monitoring that many conservation partners are using, both implementing partners and research teams can consider leveraging on-going or planned monitoring as part of research efforts. This includes the utilization of camera traps and remote-sensed data (and machine learning applications to process and analyze the data), efforts at zoonotic monitoring and surveillance, increased field observation coupled with field-appropriate data collection technologies (e.g. SMART monitoring), and drone flyovers or aerial surveys (where possible). If these monitoring initiatives are increasingly a part of the standard conservation programming, then USAID can focus on funding similar monitoring in comparison sites to reduce research costs while increasing the rigor of the research.

Some new technologies and datasets require advanced skills for data processing and analysis. Increased collaboration with universities that have ecology labs with the skills and expertise to conduct and analyze biophysical data can provide a more cost-effective model. Universities are interested in high quality research and data collection opportunities, along with training and publication opportunities for students and faculty.

Prioritizing geographies for zoonosis

If the focus is on understanding and mitigating the risk of zoonotic spillover, it would make the most sense to target limited resources on areas with the highest risk of zoonotic spillover. Not all projects in all geographies make sense for applying an implementation approach to studying land-use conversion and zoonotic spillover. The heterogeneity in zoonotic host and pathogen distributions, as well as landscape conversion, creates opportunities for focusing resources and research at times and areas at highest risk.

Tropical areas are high priorities for biodiversity conservation: ecosystems in these regions support diverse species assemblages, often with endemic species found nowhere else. In some countries, these ecosystems face increasing deforestation pressure that destroys or degrades habitat and leads to biodiversity loss. Biodiversity in tropical forests harbors a high diversity of pathogens, including many viruses, bacteria, and parasites that can infect humans. The disturbance of these ecosystems, primarily through deforestation of tropical broadleaf forests, has been linked to zoonotic outbreaks.

In particular, one option is to focus on tropical areas within regions highlighted as high risk for zoonotic emergence: including, but not limited to, Central American tropical forests, coastal West Africa, the African Great Lakes Region, India, and eastern China [9]. These include landscapes with certain types of animals (e.g., bats, rodents, primates, ungulates, and certain carnivores) that are the source for the majority of zoonotic pathogens [34] In general, if the focus in on studying zoonotic spillover risk, priority geographies and interventions can be selected based on the following considerations:

• The probability or risk of animal-to-human pathogen emergence is concentrated in regions where key animal hosts (e.g. bats, primates, and rodents), pathogens, human behavior and ecological conditions align to increase likelihood of spillover.
• Sites can prioritize fragmented edge regions and areas undergoing intermediate levels of land conversion (e.g. degraded forests) given that the risk of land use-induced zoonotic spillover is particularly high in areas undergoing intermediate levels of land conversion.
• Areas where there is greater human community interaction with wild ecosystems. Close contact between wildlife and humans is a prerequisite for spillover. Interventions can consider a focus on areas where wildlife markets exist, communities engage in pervasive hunting, and/or forest/edge areas with frequent forest entry/activities within forests by community members.
• Areas that are already slated for a forest conservation or biodiversity impact evaluation and ideally also overlapping with a relevant health intervention (see the HIH case description above for an example).
• Areas where zoonosis surveillance and monitoring are already occurring. Given the cost and complexity of wildlife and health data collection efforts for zoonosis, donors can consider leveraging and/or supporting other projects where research is already underway.

Personal bias and counterproductive incentives

Personal bias and counterproductive incentives continue to pose major challenges. Adapting (or abandoning) a program that one designed and championed for years is hard. Traditionally, funders have emphasized ‘success’ rather than learning to improve program design and implementation. Funders as well as implementing partners have incentives to reduce risk in conservation programs. This risk mitigation incentive may in some circumstances result in weak MEL that consists of input and output indicators—but lacks the ability to effectively assess key outcomes. The current development context does not foster a culture of deliberate, rigorous evidence-based design that embraces uncertainty and mixed or null results. To avoid the misallocation of scarce resources in environment programming, it is absolutely critical to understand why a null result or negative impact is observed and adapt programs and policies accordingly.

Common misconceptions of counterfactual methods lead to under investment and use of implementation science. While there is always a commitment to “evidence”, this does not translate to a commitment for rigorous evidence. Common misconceptions include a belief that implementation science is too costly, too slow to produce results, and will not lead to adaptive management on meaningful timescales given the scale of the conservation and climate crises.

Stronger evidence should be used to actively challenge design and promote real adaptation. However, once programs are designed—and processes set in place for implementation—there is path dependence and a resistance to altering course, even in the face of baseline or contextual data that challenges baseline assumptions. This is a significant challenge. One practical solution would be to include criteria for adaptive management as a performance indicator in contracts. This would prompt more attention from donors and implementing partners if positive performance is bench marked against active adaptations and course corrections following data collections and pause and reflect sessions. Instead of making adaptation seems like a ‘failure’, this would reframe the process as a positive—similarly incentivizing the active use of baseline data.

Methodological challenges

Research designs need to contend with long time lags between intervention and ecological response, complex spatial spillover effects (including from species movement), bundled interventions, and large spatial scales of environmental processes. Many standard threats to integrity would need to be considered, including the historical legacy of prior interventions, multiple bundled interventions, and selection bias. Certain interventions will be more likely to achieve the statistical power required for an impact evaluation.

Many of these challenges also apply to those encountered in governance and other sectors that have managed to adapt impact evaluation designs to the realities of those sectors. Conservation and the link between governance, conservation, and global health could follow suit. A long running program is not necessarily an impediment to a rigorous research agenda but rather an opportunity for rigorous research to add value.

Survey-based measures are appropriate for answering particular questions. In some cases, credible measures to uncover conservation and health linkages will need to move beyond self-reported behavioral indicators. For example, survey experiments, such as list and conjoint experiments, are increasingly used as an option to reduce bias for survey-based measures. The suitability of the method depends on the question being asked.

Ideal study designs will require human and wildlife biological sampling. Moreover, developing appropriate measures of the risk of zoonotic spillover is especially challenging. Appropriate measures require integration of multiple fields, including biology, virology, conservation science, anthropology, and more. Risk of emergence can be measured as spatiotemporal variation in (1) likelihood of human-wildlife contact (including contract with livestock or other bridge hosts), and (2) reservoir hazard or likelihood of infection given a contact (a function of disease prevalence and intensity and reservoir host abundance) [36, 37].

Better integration of M&E

Despite progress, the EKNA and RESTORE case studies highlight that there is room for improvement in terms of better integrating the M and E in M&E. Although these two projects were slated for an implementation science approach, the evaluations are barely mentioned in the MEL plan. Also, MEL plans are often focused only on the monitoring indicators they need to report to donors, rather than the full suite of rich data that is often collected, but which only the implementing partners are aware of and have access to. For example, in the EKNA Zambia case, an extensive household M&E survey was conducted that included a number of food security, socio-economic status, and human well-being indicators, however, only a fraction of those indicators are included in the MEL plan. Even in cases where indicators are being collected in addition to those required by the contract, there are challenges with integrating those into evaluation efforts. This could be driven by the extra time and resources required to report out on additional indicators—and incentives to focus on a more limited set of indicators for performance and accountability.

Cost and logistics

Although feasible, a comprehensive impact evaluation that includes all of the measures necessary to measure the linkages between environment and human health outcomes has cost implications and is logistically challenging. In the context of interventions to affect land-use change and forest condition, a comprehensive study will need to take into consideration measures for forest condition, biodiversity, wildlife health and human health. And, as discussed in previous sections, although there have been improvements in conservation and biodiversity evaluations over the past ten years, we still have a long way to go to implementing successful and comprehensive evaluations in these fields.

Priority outcomes (and not just intermediate or related outcomes) often take a longer time to manifest than in most other sectors for two reasons. First, biophysical changes often take longer themselves (not always, but in most theories of change). Second, the logic we are looking at in cross sectoral programs requires a link between human outcomes leading to biophysical (or vice versa) which then takes even longer to occur. This long time period reduces the benefits of evaluation and makes it more challenging both methodologically (maintaining controls) and logistically (time frames of donors).

Integrated conservation and health research will require a 10+ year time horizon, with data collected at key points versus the current 5-year timeline that coincides with the program cycle for many donor-funded projects. There are long time lags between interventions and ecological response. Primary biophysical outcomes of interest—including impacts on forest regrowth, biodiversity and impacts on human well-being from improved watershed health—will take a longer time period to materialize than the standard donor program cycle. In addition, long-term data on well-being, livelihoods, and health will answer important questions about the sustainability of program impacts.

Although some biodiversity and conservation outcomes will take decades to observe, behavioral change will be observable within one to two years after interventions and provides an indication of future program efficacy. For zoonosis, depending on the species, adult annual mortality and host viral dynamics (e.g., propensity to shed virus and viral load), might be observable within 10 years.

Long term research time frames have long been used for health and education programming and are increasingly being used by donors for sectors such as land and resource tenure and democracy and governance. Using publicly available geospatial datasets, which are increasingly available for a wider variety of outcomes and at a higher resolution, are one way to reduce costs and track forest condition outcomes across a 10-year period. As is currently becoming a standard/best practice for conservation impact evaluations, research designs in this field should propose project endline, as well as long-term follow-up data collection and analysis about five years after the end of the activity. Earlier rounds of data collection can examine key short-term behavioral outcomes in cases where enough time will not have passed to measure a meaningful change in more distal biophysical indicators. For example, if there is no evidence that people are changing threat reduction behavior at endline, it is unlikely that long-terms impacts would be achieved.

Research translation and data dissemination

Beyond counterfactual and rigorous research, we need more consideration of ways to improve research translation and dissemination. Improved research translation and dissemination will help promote broader use of the data generated through partnerships with universities, local researchers, and others.

The Zambia EKNA case study shows the use of baseline data to help inform current and future agricultural interventions. It also facilitated rich discussions about local health issues and potential solutions at pause and reflect sessions.

However, there are numerous cases of a lack of adaptive management taking place, including cases where baseline data is not examined or cases where the project fails to adjust programming despite findings from baseline data collection that indicate that the planned intervention does not address the source of the development challenge of interest. Using baseline data to challenge assumptions, perceived truths, or conventional wisdom cuts against the grain of systematic human biases. There must be dedicated time and space for stakeholders to interrogate the baseline data and the project’s theory of change to determine which assumptions hold or not and where adaptive management may be needed.