Human Health and Environmental Sustainability in Pathogenic

HUMAN HEALTH AND ENVIRONMENTAL SUSTAINABILITY IN PATHOGENIC

LANDSCAPES: FEEDBACKS AND SOLUTIONS

A DISSERTATION

SUBMITTED TO THE DEPARTMENT OF BIOLOGY

AND THE COMMITTEE ON GRADUATE STUDIES

OF STANFORD UNIVERSITY

IN PARTIAL FUFILLMENT OF THE REQUIREMENTS

FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

ISABEL JEAN JONES

AUGUST 2020

This work is licensed under a Creative Commons Attribution- Noncommercial 3.0 United States License. http://creativecommons.org/licenses/by-nc/3.0/us/

This dissertation is online at: http://purl.stanford.edu/fv856hm0269

ii I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Giulio De Leo, Primary Adviser

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Erin Mordecai

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

Stephen Palumbi

Approved for the Stanford University Committee on Graduate Studies. Stacey F. Bent, Vice Provost for Graduate Education

This signature page was generated electronically upon submission of this dissertation in electronic format. An original signed hard copy of the signature page is on file in University Archives.

iii Abstract

Feedbacks between human health, well-being, and environmental sustainability are increasingly well recognized, but evidence on actionable solutions to leverage these feedbacks and improve their outcomes is limited. And yet, the recently adopted United Nations Sustainable Development Goals compel decision makers to achieve health for all people while reducing poverty and conserving nature. This dissertation attempts to fill this urgent research gap and empirically assess feedbacks between human health and the environment from two perspectives: (i) environmental drivers of disease in changing landscapes, and (ii) consequences of human health and well-being on environmental change and conservation outcomes. But first, using synthesis science, we describe the environmental components of the world’s most burdensome human parasitic and infectious diseases, then assess a specific type of ecosystem service – natural enemies of human infections – as a sustainable tool to reduce some communicable diseases. Then, we assess feedbacks between human health and environmental sustainability through two data-rich case studies. First, in the Senegal River Basin, a region hyperendemic for schistosomiasis disease, we assess ecological associations and spatial scales of human disease risk. In this region, schistosomiasis disease, which is transmitted to humans by freshwater snail hosts, can be caused by two species of schistosome parasites (and their two unique snail hosts). The World Health Organization currently recommends that snail control be implemented in endemic areas to curb transmission. However, snail control methods rarely take into consideration the unique transmission dynamics of co-endemic species, which could interfere with control outcomes for one or both species. So here we ask a critical question: what is the spatial scale and ecological correlates of schistosomiasis risk, considering two distinct schistosome species in a co- endemic landscape? We find specific, yet divergent, ecological correlates and spatial scales of risk for each species. Ultimately, these findings can be used to design ecological levers for health (like snail control) that best reflect the unique transmission ecology the targeted species. Second, in rural Borneo, we assess the impact of a 10-year, on-the-ground intervention aiming to improve human health care access in order to curb illegal logging in protected forests, and ask: Is healthcare an effective lever for conservation? Using earth observation, household surveys, and clinic data records, we find that the intervention provided health care access for more than 28,000 patients and averted >27km2 forest loss (~70% reduction compared to controls), thus improving human health and well-being, and simultaneously preserving globally important forest carbon. In conclusion, this dissertation shows that, by recognizing and embracing complex mechanisms driving health-environment feedbacks, we can identify effective ecological levers for human health, and health levers for conservation.

Acknowledgements

Many people helped this dissertation come together, but there are two people that facilitated and supported each and every part: Giulio De Leo and Susanne Sokolow. To my generous advisor and powerhouse mentor, many thanks for your continued personal and academic support, and for always putting my well-being ahead of my work. I also had an excellent dissertation committee that challenged and supported me, so thank you to Erin Mordecai and Steve Palumbi. Andy, many thanks to Steve Luby and Rodolfo Dirzo for joining my oral defense committee. Outside of Stanford, no small amount of guidance, patient teaching, and friendship came from Chelsea Wood, Skylar Hopkins, and Andy MacDonald. Thanks also to the NCEAS SNAPP working group, “Ecological Levers for Health”, for welcoming me into an incredible community of researchers that I was first intimidated by but can now call friends and colleagues. The most rewarding aspect of my research experience was the opportunity to work with researchers, conservation practitioners, health care providers, activists, and community members working on- the-ground, around-the-clock to make real positive change to the daily lives of people, to protect and restore the natural world, and to enact climate change mitigation. To that end, thank you to the Upstream Alliance team in Senegal for supported me during five amazing field trips to the Senegal River Basin, to Health In Harmony in Portland and Alam Sehat Lestari in Borneo for endless inspiration, to Martina Laidemitt and Dr. Sam Loker for welcoming to join in and learn from their incredible long-term field research at Lake Victoria in Kenya, and to PIVOT in Madagascar for generously hosting me for a truly unique research experience in Ifanadiana. Andrea Lund was the best dissertation partner and friend to have along the way – thank you for reminding me that “a good dissertation is a done dissertation”. My friends – the basketball champions in the De Leo Lab, Tim White, Serena Lomonico, A2 Andy Chamberlain & Alex Smith, and Shirel Kahane-Rapport – made life in Monterey fun and interesting, and convinced me to stay put in this magical place as long as possible, which is exactly what I’ll do. Luke Netsch has been a stellar partner and has made Monterey and its beaches and mountains a lovely home for us. My family – Laurel, Robert, Amanda, Hannah, Chris, Kyle, Jean, Emer, Diane – remind me every day what’s most important. Ruth, you are the best.

vi Table of Contents

Abstract ...... iv

Acknowledgements ...... v

Introduction ...... 1

Chapter 1 : Biodiversity and environmental persistence of the world’s most burdensome infectious and parasitic diseases ...... 10

Chapter 2 : Enemy of my enemy: Natural enemies for sustainable control of human infections ...... 28

Chapter 3 : Divergent ecological drivers and spatial scales of risk for S. haematobium and S. mansoni in a co-endemic landscape ...... 63

Chapter 4 : Improving rural health care reduces illegal logging and conserves carbon in a tropical forest ...... 98

Conclusion ...... 156

vii List of Illustrations

Figure 1.1. Associations between duration of infectious stages in the environment, transmission strategies, control strategies, host ranges, and global cases of disease ...... 23 Figure 2.1. Natural enemies of human infectious diseases in the environment ...... 57 Figure 2.2. Reports of mosquito vector resistance to insecticides over time (A) and natural enemies of mosquitoes (B) ...... 58 Figure 3.1. Study location (A); examples of overhead drone imagery and high-resolution image classification into dominant aquatic habitat types (B–C) ...... 88 Figure 3.2. Estimated marginal means for Bulinus spp. (A–B) and Biomphalaria pfeifferi (C–D) snails in three dominant aquatic habitat types ...... 89 Figure 3.3. Visualization of experimental design to explore scale of disease transmission risk (A); correlation between non-emergent vegetation and S. haematobium and S. mansoni infection risk (B–C) ...... 90 Figure 3.4. Effect of site morphological characteristics on S. haematobium and S. mansoni infection risk ...... 91 Figure 4.1. Global overlap of aboveground forest carbon and low health care access (A); deforestation trends in protected areas of Indonesia (B); study location and theory of change ...... 130 Figure 4.2. Synthetic controls analysis on climate impacts of the intervention on forest conservation and logging behavior (A-B); comparison of tree heights and carbon storage (C); dose-response of logging activity to the intervention ...... 131 Figure 4.3. Health impacts of the intervention in terms of clinic access and usage (A); disease diagnoses (B); livelihoods (C); and wealth (D) ...... 132 Figure 4.S1. Average above-ground carbon biomass of mixed forests in tropical, temperate, and boreal regions around the world ...... 140 Figure 4.S2. LiDAR imagery used to calculate tree height in Gunung Palung National Park ...... 141 Figure 4.S3. Locations of districts (‘desa’ administrative units, as of 2017) from where clinic patients derived ...... 142

viii Figure 4.S4. Clinic access and usage, comparing patients from MOU-signing villages to patients from non-MOU-signing villages ...... 143 Figure 4.S5. Cumulative unique clinic patients over time (A); annual time series of disease trends in patients from MOU-signing and non-MOU-signing village ...... 144 Figure 4.S6. Individual engagement with all intervention activities for 36 villages bordering the national park (A–B); description of intervention engagement activity types (C) ...... 145

ix List of Tables

Table 1.1. Summary of the transmission strategies and standard prevention and control strategies associated with 151 parasites and pathogens of humans tracked by the WHO Global Burden of Disease database ...... 24 Table 2.1. Select case studies and observations on natural enemies of human pathogens ...... 59 Table 3.1. Infection prevalence and egg burden (egg intensity) for all participants in the longitudinal cohort study at baseline in 2016, and re-infection in 2017 and 2018 following treatment ...... 92 Table 3.2. Regression table for snail-habitat models associating snail density with freshwater habitat type, water depth, and village location, for both Bulinus spp. snails and Biomphalaria pfeifferi ...... 93 Table 3.3. Results from offshore, deep water snail sampling in non-emergent vegetation ...... 94 Table 3.4. Infection presence and egg burden model results for S. haematobium, comparing models that integrated non-emergent vegetation coverage across different scales ranging from 1m to 120m ...... 95 Table 3.5. Infection presence and egg burden model results for S. mansoni, comparing models that integrated non-emergent vegetation coverage across different scales ranging from 1m to 120m ...... 96 Table 4.1. Results from the synthetic control analyses on park-level forest loss in Gunung Palung National Park ...... 133 Table 4.2. Dose-response of forest change to the intervention: results of a linear mixed effects regression ...... 134 Table 4.S1. Regression table showing the effect of signing an MOU on district-level infectious disease outcomes over time, controlling for average district distance to the clinic ...... 146 Table 4.S2. Regression table showing the effect of signing an MOU on district-level noninfectious disease outcomes over time ...... 147

x Table 4.S3. Regression table showing the effect of signing an MOU on district-level “other” and untracked disease outcomes over time ...... 148 Table 4.S4. Description of household survey population ...... 149 Table 4.S5. Survey findings on 5-year and 10-year impact on household-level births, infant deaths, and livelihoods ...... 150 Table 4.S6. List of all ICD10 codes tracked (or not tracked) for analysis of changes in disease diagnoses at the clinic ...... 151

xii

Introduction

Research towards this dissertation commenced in 2015, in the same week that 193 countries adopted the United Nations’ Sustainable Development Goals (SDGs), 17 ambitious social and environmental objectives to achieve comprehensive health, well- being, and environmental sustainability by 2030 (1). The SDGs are built upon the recognition that human health, poverty alleviation, and environmental sustainability are deeply interconnected, and therefore, so too is the pursuit of their improvement. In response, multi-disciplinary communities of scientists and science-based policy makers have been compelled to work together towards a healthier, and more sustainable future (2–4). Yet, bold action to safeguard people and the planet is challenged when the mechanisms underlying many human health-environment relationships and the evidence to support actionable solutions to achieve integrated goals remain limited (2, 3). Too often, goals for improving human health and preventing environmental degradation have been approached in isolation; where they have been approached in concert, evidence to date suggests that the outcomes of deliberately integrated health, development, and environment solutions are mixed at best, and replete with trade-offs (5). And yet, hope remains: the growing fields of Planetary Health and One Health have recently formalized goals to identify and enact mutually beneficial (or unharmful) interventions to safeguard human health and the ecosystems on which people rely (2, 6). Inherent in these efforts is a systems theory approach to problem-solving, wherein complexity is fully recognized not just as a challenge but as a tool to overcome trade-offs and identify integrated solutions for integrated problems (2, 7). The chapters presented here attempt to embrace complexity and address critical health-environment research gaps. Chapters 1 and 2 explore the ecology of human infectious diseases with environmental components: What parasites and pathogens cause the highest burden of human disease, what are their environmental components, and what ecosystem services (“natural enemies”) might mitigate their impact? Chapters 3 and 4 explore complex feedbacks between human health, poverty, and the environment from two perspectives. First, I assess environmental drivers of a globally important disease of poverty in a profoundly altered landscape: schistosomiasis disease in the Senegal River Basin. Then, I assess the impact of human health and well-being on environmental

1 change, leveraging a 10-year integrated health and conservation intervention in tropical Borneo to ask, can improving human health improve conservation outcomes? In the last two decades, the global burden of human infectious diseases has declined more than 40% (8), reflecting accelerating progress towards health-related SDGs (9). At the same time, global trends like urbanization, agricultural intensification, and land-use change have also accelerated. These processes can rearrange when and where humans encounter disease-causing organisms with important environmental components (10). For many such diseases, strategies to lessen disease burden and prevent new outbreaks benefit from integrated approaches to intervention (11). However, the environmental sources of infectious diseases are often overlooked (12). Therefore, Chapter 1 synthesizes data on complex ecological attributes of 151 parasites and pathogens causing a substantial proportion of the global burden of infectious diseases, and assesses relationships between global burden, environmental persistence, transmission strategies, host ranges, and control strategies. In so doing so, this study reveals diverse and persistent environmental components of the majority of human infectious diseases. Ultimately, this database will serve as an open-source resource to assist researchers and policy makers in predicting how environmental change might mediate risk for specific diseases, and identify environmental targets for risk mitigation. In some systems, environmental change has been linked to infectious disease outbreaks when the foodwebs in which parasites, pathogens, and – where relevant – vectors and non-human hosts are embedded are altered (13). For example, an on-going schistosomiasis epidemic in the Senegal River Basin, further discussed in Chapter 3, has been linked to the regional loss of native predators of the parasites’ freshwater intermediate host snails (14). The loss of this important ecosystem service has led to a novel disease control proposition: restore native predators to reduce human disease. My participation in ongoing field studies to test this solution inspired the overarching question addressed in Chapter 2: How common are natural enemies of human infections, and do examples exist where they have been harnessed for effective disease control? Other examples of disease outbreaks following losses of natural enemies (e.g., (15, 16, 17, 18)) suggest that intact natural enemy communities represent an important ecosystem service that may – to some degree – mitigate human disease risk. Therefore, in Chapter 2, I review examples, opportunities, and challenges to developing natural enemy

2 interventions as part of human infectious disease control strategies. I provide a broad set of examples where natural enemies of human infectious diseases have been implemented or proposed to curb disease burden, and summarize successes, challenges, and recent developments. As part of an interdisciplinary team assessing natural enemy restoration as a tool to reduce schistosomiasis disease, I have spent a substantial amount of time during my graduate studies chest wader-deep in schistosome transmission sites in the Senegal River Basin. The region is well known for sustaining the world’s longest recorded schistosomiasis epidemic, which began after the seasonally dry river basin was dammed and permanently inundated in freshwater to support agricultural development (19). The profound alteration of the Sahelian environment created vast areas of permanent freshwater ecosystems, which were quickly colonized by aquatic vegetation and specific species of freshwater snails that serve as an obligate intermediate host for two species of human schistosomes: S. haematobium and S. mansoni (20). With my waders in the water collecting snails, it was immediately apparent to me that the two species of schistosome host snails exhibited different ecological associations, and therefore, might not respond the same way to environmental intervention strategies aimed at reducing their populations. This is important, because the World Health Organization has recently updated its schistosomiasis control strategy to include snail control in endemic areas, whenever and wherever possible (21). Moving forward, a prerequisite for implementing cost-effective, sustainable snail control is an understanding of the ecological dynamics underlying the relationships between snails, ecological distribution, and human risk. This is the goal of Chapter 3, which attempts to understand the ecological correlates and spatial scale of schistosomiasis risk in the Senegal River Basin. I ask a deceptively simple question (Where are the snails that infect me?) for both co-endemic species of schistosomes that cause disease in humans, attempting to resolve ecological differences that determine human risk and that could compromise environmental control outcomes. To do so, I used fine-scale quantitative data on snail-habitat associations to validate a drone-based assessment of risk across increasing spatial scales at focal transmission sites, integrating ecological, spatial, and epidemiological data. Results suggest that infection risk for the two species show divergent ecological and spatial scales. Findings from Chapter 3 provide novel and nuanced insight into the spatial ecology of an ancient human

3 disease. Pending ongoing studies, findings may also be used to validate methods to use high-resolution satellite or drone imagery to rapidly and inexpensively identify transmission foci, aiding in the design of targeted, cost-effective, sustainable interventions to reduce disease. Considering feedbacks between human health and the environment, Chapters 1 through 3 assess environmental components and drivers of human disease. In Chapter 4, we assess this feedback from the opposing side, and explore how human health drives environmental change and conservation. To do so, we travel to rural Borneo to ask: Is improving human health an effective lever to improve environmental stewardship and conservation outcomes? In too many places, the establishment of protected areas has involved excluding, and thus disenfranchising, local communities that surround them (5, 22). Failure to address the needs of local people, an injustice that reflects lasting colonial attitudes toward local small, rural, and native communities, can then threaten conservation efforts, when communities with few financial alternatives to fulfill needs – like basic health care – illegally extract resources and convert land in order to survive (22, 23). Within this context, Chapter 4 evaluates a human-centered conservation solution – an intervention that has focused on increasing health care access and affordability for rural poor communities living near protected forests with high conservation value. The intervention offered region-wide access to a newly opened clinic, and communities living adjacent to a national park received health care discounts to offset costs historically met through illegal logging. Conservation, education, and alternative livelihood programs were also offered. Using clinic database records, household survey data, and remotely- sensed forest change data to statistically assess intervention outcomes, we found that the intervention increased health clinic access and incentivized usage – with population-wide decreases in several communicable and non-communicable diseases – and simultaneously averted illegal logging in the national park. Results demonstrate that this community-derived solution simultaneously improved health care access for local communities, and sustainably conserved carbon stocks in protected tropical forests. The research presented encompasses synthesis science, field ecological studies, social science, novel earth observation techniques, and ecological and epidemiological statistics to assess feedbacks between human health and environmental sustainability. This dissertation aims to provide researchers and policy makers with new insights and

4 tools to design, implement, and objectively evaluate integrated solutions to improve human health, well-being, and conservation of nature. Findings from empirical case studies exploring complex links between health and nature suggest that ecological levers for human health, and health levers for environmental sustainability, are attainable and actionable goals. Achieving the SDGs within the nascent decade will require unprecedented cross-disciplinary action that embraces complexity, and I hope future readers will find the goals achieved.

Statement on multiple authorship: I am the primary author of the research and production of all chapters. Each chapter was only possible due to valued collaborators, whose full names are listed within each chapter and whose contributions are as follows:

Chapter 1: I.J.J., G.A.D.L, S.H.S, and S.H conceived of the database idea. I.J.J., S.H, C.L.B., N.N., L.K., and J.B. performed literatures searches to collect and curate the data. All authors contributed to drafting and editing the manuscript.

Chapter 2: G.D.L conceived the idea. I.J.J conducted the literature review and wrote the draft manuscript. All authors provided substantial input into ideas, text, and figures, and commented on draft manuscripts.

Chapter 3: I.J.J, G.D.L., A.J.C., S.H.S., and C.L.W conceived of the idea, designed field protocols, and collected data. I.J.J. analyzed the data. All authors contributed to drafting and editing the manuscript.

Chapter 4: NGO staff (Alam Sehat Lestiri and Health in Harmony) designed the intervention and provided clinic and household survey data. I.J.J., S.R.H., A.J.L., and A.J.C curated and analyzed data on health and well-being. A.J.M. obtained and curated, forest change data, and with I.J.J. analyzed forest change data. S.H.S. designed overarching data analysis plan. I.J.J. and S.H.S. drafted the manuscript; all other offers contributed to experimental design and editing.

Chapters are presented in the formats that were required by academic journals.

5 References

1. W. Rosa, Ed., in A New Era in Global Health (Springer Publishing Company, New York, NY, 2017; http://connect.springerpub.com/lookup/doi/10.1891/9780826190123.ap02).

2. S. Whitmee, A. Haines, C. Beyrer, F. Boltz, A. G. Capon, B. F. de Souza Dias, A. Ezeh, H. Frumkin, P. Gong, P. Head, R. Horton, G. M. Mace, R. Marten, S. S. Myers, S. Nishtar, S. A. Osofsky, S. K. Pattanayak, M. J. Pongsiri, C. Romanelli, A. Soucat, J. Vega, D. Yach, Safeguarding human health in the Anthropocene epoch: report of The Rockefeller Foundation–Lancet Commission on planetary health. The Lancet. 386, 1973–2028 (2015).

3. IPBES, “Summary for policymakers of the global assessment report on biodiversity and ecosystem services of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services” (IPBES Secretariat, Bonn, Germany, 2019), p. 60.

4. F. C. J. Berthe, T. Bouley, W. B. Karesh, F. G. Le Gall, C. C. Machalaba, C. A. Plante, R. M. Seifman, “Operational framework for strengthening human, animal and environmental public health systems at their interface” (123023, The World Bank, 2018), pp. 1–152.

5. T. O. McShane, P. D. Hirsch, T. C. Trung, A. N. Songorwa, A. Kinzig, B. Monteferri, D. Mutekanga, H. V. Thang, J. L. Dammert, M. Pulgar-Vidal, M. Welch-Devine, J. Peter Brosius, P. Coppolillo, S. O’Connor, Hard choices: Making trade-offs between biodiversity conservation and human well-being. Biological Conservation. 144, 966–972 (2011).

6. J. Zinsstag, E. Schelling, D. Waltner-Toews, M. Tanner, From “one medicine” to “one health” and systemic approaches to health and well-being. Preventive Veterinary Medicine. 101, 148–156 (2011).

7. T. Lang, G. Rayner, Ecological public health: the 21st century’s big idea? An essay by Tim Lang and Geof Rayner. BMJ. 345 (2012), doi:10.1136/bmj.e5466.

8. WHO, “Global Health Estimates 2016: Disease burden by Cause, Age, Sex, by Country and by Region, 2000-2016” (World Health Organization, Geneva, 2018).

9. N. Fullman et al., Measuring progress and projecting attainment on the basis of past trends of the health-related Sustainable Development Goals in 188 countries: an analysis from the Global Burden of Disease Study 2016. The Lancet. 390, 1423–1459 (2017).

10. E. F. Lambin, A. Tran, S. O. Vanwambeke, C. Linard, V. Soti, Pathogenic landscapes: Interactions between land, people, disease vectors, and their animal hosts. International Journal of Health Geographics. 9, 54 (2010).

6 11. A. Garchitorena, S. H. Sokolow, B. Roche, C. N. Ngonghala, M. Jocque, A. Lund, M. Barry, E. A. Mordecai, G. C. Daily, J. H. Jones, J. R. Andrews, E. Bendavid, S. P. Luby, A. D. LaBeaud, K. Seetah, J. F. Guégan, M. H. Bonds, G. A. De Leo, Disease ecology, health and the environment: a framework to account for ecological and socio-economic drivers in the control of neglected tropical diseases. Philos Trans R Soc Lond B Biol Sci. 372 (2017), doi:10.1098/rstb.2016.0128.

12. J. V. Remais, J. N. S. Eisenberg, Balancing Clinical and Environmental Responses to Infectious Diseases. Lancet. 379, 1457–1459 (2012).

13. K. D. Lafferty, A. P. Dobson, A. M. Kuris, Parasites dominate food web links. PNAS. 103, 11211–11216 (2006).

14. S. H. Sokolow, E. Huttinger, N. Jouanard, M. H. Hsieh, K. D. Lafferty, A. M. Kuris, G. Riveau, S. Senghor, C. Thiam, A. N’Diaye, D. S. Faye, G. A. D. Leo, Reduced transmission of human schistosomiasis after restoration of a native river prawn that preys on the snail intermediate host. PNAS. 112, 9650–9655 (2015).

15. G. Swan, V. Naidoo, R. Cuthbert, R. E. Green, D. J. Pain, D. Swarup, V. Prakash, M. Taggart, L. Bekker, D. Das, J. Diekmann, M. Diekmann, E. Killian, A. Meharg, R. C. Patra, M. Saini, K. Wolter, Removing the threat of diclofenac to critically endangered Asian vultures. PLoS Biology. 4, 0395–0402 (2006).

16. H. S. Young, R. Dirzo, K. M. Helgen, D. J. McCauley, S. A. Billeter, M. Y. Kosoy, L. M. Osikowicz, D. J. Salkeld, T. P. Young, K. Dittmar, Declines in large wildlife increase landscape-level prevalence of rodent-borne disease in Africa. Proceedings of the National Academy of Sciences. 111, 7036–7041 (2014).

17. Hofmeester Tim R., Jansen Patrick A., Wijnen Hendrikus J., Coipan Elena C., Fonville Manoj, Prins Herbert H. T., Sprong Hein, van Wieren Sipke E., Cascading effects of predator activity on tick-borne disease risk. Proceedings of the Royal Society B: Biological Sciences. 284, 20170453 (2017).

18. T. Levi, A. M. Kilpatrick, M. Mangel, C. C. Wilmers, Deer, predators, and the emergence of Lyme disease. PNAS. 109, 10942–10947 (2012).

19. P. Steinmann, J. Keiser, R. Bos, M. Tanner, J. Utzinger, Schistosomiasis and water resources development: systematic review, meta-analysis, and estimates of people at risk. The Lancet Infectious Diseases. 6, 411–425 (2006).

20. R. F. Sturrock, O.-T. Diaw, I. Talla, M. Niang, J.-P. Piau, A. Capron, Seasonality in the transmission of schistosomiasis and in populations of its snail intermediate hosts in and around a sugar irrigation scheme at Richard Toll, Senegal. Parasitology. 123, 77–89 (2001).

21. WHO, Field use of molluscicides in schistosomiasis control programmes: an operational manuel for programme managers. (World Health Organization (WHO), Geneva, Switzerland, 2017).

7 22. William. M. Adams, Biodiversity Conservation and the Eradication of Poverty. Science. 306, 1146–1149 (2004).

23. C. B. Barrett, A. J. Travis, P. Dasgupta, On biodiversity conservation and poverty traps. Proceedings of the National Academy of Sciences. 108, 13907–13912 (2011).

9 Chapter 1

Biodiversity and environmental persistence of the world’s most burdensome

infectious and parasitic diseases

Isabel J. Jones1*, Skylar R. Hopkins2,3, Julia Buck4, Chris LeBoa5, Andrea J. Lund6,

Laura Kwong7, Nicole Nova8, Susanne H. Sokolow1,7, Giulio A. De Leo1,7

1Hopkins Marine Station, Stanford University, Pacific Grove, CA, USA 2National Center for Ecological Analysis and Synthesis, Santa Barbara, CA, USA 3Deptartment of Applied Ecology, North Carolina State University, Raleigh, NC, USA 4Department of Biology and Marine Biology, University of North Carolina Wilmington, Wilmington, NC, USA 5Department of Epidemiology, Stanford University, Stanford, CA 94305, USA 6Emmett Interdisciplinary Program in Environment and Resources, Stanford University, Stanford, CA 94305, USA 7Stanford Woods Institute for the Environment, Stanford University, Stanford, CA 94305, USA 8Department of Biology, Stanford University, Stanford, CA 94305, USA

*Corresponding author: Isabel J. Jones; [email protected]

10 ABSTRACT Many parasites and pathogens that cause a substantial proportion of the global burden of infectious and parasitic disease are environmentally-mediated. Here we provide a novel database on the diverse environmental components and persistence of more than 150 parasites and pathogens tracked by the World Health Organization’s Global Burden of Disease study, characterizing for each, among other attributes: standard control or prevention strategy, dominant transmission pathway, obligate and relevant vertebrate host ranges, vectors, obligate intermediate hosts, and duration of time spent in each component of its life cycle (including in humans). To our knowledge, no previous study has attempted to synthesize environmental persistence of human parasites and pathogens in terms of the duration of spent in different life cycle stages. We find that the duration of infectious organisms spent outside of a human or obligate vertebrate host in the environment is associated with pathogen or parasite transmission strategy and standard control and prevention measures. We also explore relationships between the environmental components of infectious organisms and global cases of disease. We find that while most infectious organisms have some environmental component, the estimated contemporary number of global cases of diseases was positively correlated with an infection’s duration in humans, but negatively correlation with duration in the environment. Last, considering obligate vertebrate host range, infectious organisms exclusive to humans have the highest average number of global cases; otherwise, livestock hosts maintain populations of infectious organisms causing the highest average number of cases, followed by mixed species assemblages, domestic animals, birds, and finally wild animals. This is in contrast to most emerging infectious diseases and zoonoses, like the recently emerged novel coronavirus, SARS-CoV-2, which predominately originate in wildlife sources.

KEYWORDS: Infectious disease, parasitic disease, disease control, disease ecology, Neglected Tropical Disease, Sustainable Development Goals

11 INTRODUCTION

The global burden of human infectious diseases has declined by more than 40% since the turn of the 21st century (1), accelerating progress towards health-related Sustainable Development Goals (SDGs) (2). However, long-term progress for some diseases has been stymied by global land-use change, agricultural intensification, and urbanization (3). These environmental changes rearrange human interactions with infectious agents and their reservoirs, modifying infectious disease dynamics and risk for pathogens with environmentally mediated transmission (4, 5). For many such environmentally mediated diseases, control efforts could be more effective if they combined classical medical prevention and treatment with interventions to control environmental sources of infection (3). However, environmental sources of infection are often overlooked (6). If we better understood the enduring, yet changing, environmental components of the world’s most burdensome infectious and parasitic diseases, we could further avert the billions of infections and millions of deaths they cause annually (7). With a global commitment to achieving SDGs, a new field of action-oriented science has been established: Planetary Health. Planetary Health recognizes that human health, well-being, and environmental change and sustainability are inextricably linked, and it calls for integrated approaches to safeguard people and nature (8). In this context, improving human health will rely on a systems-thinking approach to managing by recognizing and addressing the complex biological, environmental, and social components of disease (9). The environmental components of some common diseases are obvious: the protozoan parasites that cause malaria do so through the bite of an infected mosquito, and vector management is therefore an effective lever to reduce disease (10). For other diseases, environmental transmission routes are sometimes unresolved or overlooked by management strategies. For example, improved water, sanitation, and hygiene (WASH) is a standard strategy to reduce contamination of the environment with human feces and subsequent diarrhea risk, but ignores animal feces, which may represent a major, and unmitigated, source of disease risk in some domestic settings (11). To support the SDGs and Planetary Health agenda, here we present a database that synthesizes the environmental components of a broad diversity of human infectious diseases and standard strategies for their control and prevention, to serve as an

12 interdisciplinary resource for scientists and practitioners interested in sustainable disease management. The recent emergence of a novel coronavirus (SARS-CoV-2) from wildlife has made clear the persistent potential for disease spillover at the animal–human interface (12, 13) and the global burden of disease that can result from such events (14, 15). Many important human diseases have relatively recent wildlife origins, including some (e.g., HIV) that have evolved to be confined to human–human transmission (15), and disease emergence has been associated with broad host ranges and zoonotic transmission pathways (14, 16, 17). Given ongoing risks for disease emergence, characterizing the environmental origins and drivers of emerging and re-emerging diseases has received much attention in recent years (18). However, once a disease is established in the human population, management might have nothing to do with their environmental origins. For example, HIV spilled over from primate origins, but is now confined to humans; as such, its control is focused on human interventions rather than environmental ones. Therefore, here we focused on characterizing the diverse environmental components of established human infections, the global burden of which is nearly 40% environmentally mediated through life cycles that involve continued persistence or amplification in the environment (Sokolow et. al in prep). Many studies attempt to assess the environmental persistence of emerging and established diseases by describing disease-causing organisms’ transmission pathways, abiotic environmental reservoirs, and host ranges and/or vector identities (e.g., (13, 16, 19, 20)). Here, we include in our characterization of environmental persistence the additional element of time: we characterize duration time that an infectious organism spends in environmental components of its life cycle and in humans (or its obligate non- human vertebrate host). Such data are difficult to find, especially for rare and understudied diseases, and to our knowledge, a synthesis of this aspect of environmental persistence of major human infectious diseases is unprecedented. Here we present a database of 151 parasites and pathogens of humans tracked by the World Health Organization’s Global Burden of Disease study (21), for which we characterize and assess relationships between: (i) the persistence of pathogens and parasites in humans and in the environment; (ii) dominant transmission pathways; (iii) obligate and relevant vertebrate host ranges; (iv) contemporary estimates of the global

13 cases of disease; and, (v) standard strategies for control and prevention. Data and sources are made freely available for users (see data availability) and contribute new findings on the environmental components of the world’s most burdensome pathogens and parasites.

PRINCIPAL FINDINGS Relationships between environmental persistence, transmission pathways, and control standards The transmission strategies and environmental components of major human infections are diverse (Table 1). While more than half (52%) of all diseases examined here exclusively require humans as an obligate host, more than 80% are environmentally mediated, and just 10% (primarily STDs) have no potential to survive outside of a human host for more than one day. Even among parasites and pathogens acquired directly through human-to-human transmission, there is considerable variability in environmental persistence outside the human host (i.e., duration of free-living stages, in vectors, or in intermediate hosts) (Fig. 1a). Further considering relationships between environmental persistence and transmission pathways, direct human-to-human transmission pathways (i.e., via fomites or droplets, sexual transmission, or normal flora/opportunistic infections), have the lowest average persistence in the environment outside human or obligate vertebrate hosts (Fig. 1a), followed by parasites and pathogens spread through fecal-oral transmission pathways. Foodborne pathways, which include trophically-acquired infections (resulting from direct consumption of animals), have the longest environmental duration (Fig. 1a). Standard strategies for control and prevention are associated with the duration of an infectious organism’s environmental persistence outside a human or vertebrate definitive host (Fig. 1b). As would be expected, pathogens managed through behavior or lifestyle modification (largely to prevent pathogens transmitted via sexual or close contact) have the lowest average environmental duration. Those managed by vector control tend to have shorter average environmental durations than many other parasites and pathogens with obligate environmental stages, likely because many invertebrate vectors, like mosquitoes, are short-lived in the environment (e.g., adult female Anopheles mosquitoes, which transmit malaria, rarely live longer than one to two weeks in nature, and will be infective for only a few days after parasite’s extrinsic incubation period is

14 complete (22)). Unsurprisingly, pathogens that are maintained in reservoirs and targeted for control via reservoir management, have the longest average persistence in the environment.

Relationships between global cases of disease, persistence in humans and the environment, and vertebrate host range The current estimated global cases of human disease caused by parasites and pathogens in this database ranges from 0-100 cases annually to over 1 billion cases. As expected, parasites and pathogens that predominately circulate in the environment (and do not require humans to maintain their population) cause lower global cases than parasites and pathogens of humans (Fig. 1c). Of the predominately human diseases, 69% have environmental transmission pathways (human-environment-human transmission) (Fig. 1c). Even so, the minimum number of putative global cases is positively correlated with the persistence time of the infection in humans (linear mixed effects model, estimate for logarithm of duration of infection in humans: 0.86, SE=0.16, p<0.001), but uncorrelated with total persistence in the environment (estimate for logarithm of duration of infection in environment: -0.43, SE=0.22, df=135, p=0.061). Previous studies have suggested that the majority of emerging diseases are derived from wildlife (13). Here, more than 70% of diseases in this database have been observed in other non-human vertebrate hosts (obligate or non-obligate). However, diseases with obligate wildlife host ranges are associated with the lowest average global cases (Fig. 1d). Livestock, on the other hand, is associated with high global cases of disease, second only the burden caused by exclusively human diseases, followed by mixed species assemblages of obligate hosts (i.e., some combination of livestock, poultry, domestic animals, birds, and wild animals), then domestic animals, then birds, and finally, wildlife (Fig. 1d).

Methods Database construction Since 1990 the World Health Organization and collaborating organizations have attempted to generate comparative data on the global burden of diseases and injuries updated overtime. Here we used the recently updated Global Burden of Disease (GBD)

15 estimates for 2000–2016 as a foundation to identify major categories of infectious and parasitic diseases of global health importance, for which disability-adjusted life years (DALYs) and years of life lost due to disability (YLDs) are calculated (1, 21). We used ICD-10 codes provided for “Infectious and parasitic” Global Health Estimates (GHE) (21) to identify causative agents of disease to species, but excluded diseases categorized in “other” causes and any ICD-10 codes associated with unspecified causes of disease. This resulted in 151 parasites and pathogens. We also include the recently emerged SARS-CoV-2 in the database, but exclude it in descriptive statistics. We then used this list to characterize dominant transmission strategies for each disease, obligate and relevant non-human host ranges, and to estimate the amount of time each infectious organism spends in human, animal, and environmental reservoirs during the course of its life cycle (distinguishing incubation periods from duration of infectiousness). Transmission strategies include: human-to-human; normal flora/opportunistic; fecal-oral; food-borne; water-borne; vector-borne; directly zoonotic (i.e. direct contact between humans and wild or domestic vertebrates); and, sapronotic (saprophages, or free-living organisms that consume dead plant and animal biomass, infect humans opportunistically (20)). Diseases acquired fecal-orally are often characterized as human-to-human transmission, but we decided to distinguish the fecal- oral strategy from direct transmission via respiratory droplets and close, intimate contact. For diseases that can be transmitted via multiple routes, we include the strategy assumed to cause the greatest number of infections. For example, Vibrio cholerae is sometimes considered a sapronotic or waterborne disease (20), but because many large outbreaks are associated with fecal-oral transmission within households or communities (23, 24), we classify it as a fecal-oral pathogen. Vertebrate host ranges represent only the known range of non-human vertebrate hosts, and are therefore likely to expand for less well-studied diseases. Vertebrates are characterized as obligate hosts when humans are incidental or dead-end hosts; ii) relevant hosts when vertebrates maintain parasite and pathogen populations independently from humans (and, therefore, can contribute to disease in certain contexts at the animal-human interface and may compromise long-term control); or, iii) irrelevant if animals are incidental hosts or are not known to confer any risk to human disease.

16 Duration was characterized as time (in days) a single infectious organism spends in the following contexts: in vertebrate hosts (including humans); as free-living stages in abiotic environments; or in obligate vectors or intermediate hosts. We distinguished duration of incubation periods from infectious periods (though they may overlap). For many parasites and pathogens, human (or vertebrate) infectious periods can be cyclical or interspersed with dormant stages; for consistency, we estimate duration of infectiousness as the total average time that an infection lasts, even though this may overestimate the actual period spent infectious. For these and other reasons, the total time an infectious organism spends in various stages of its life cycle does not represent its generation time. Moreover, disease duration can be highly variable in humans and, especially, in environmental reservoirs depending on host body condition, temperature, humidity, and other physical and biological factors. For this reason, we provide a description of duration according to data found in literature sources. We also collected data on the current number of cases each infectious organism causes in humans. Because source estimates of global cases are highly variable, sometimes outdated, and often difficult to find, here we conservatively estimated cases using a categorical variable on a logarithmic scale (i.e., 0–100, 101–1000, 1001–10,000, etc.). Therefore, we reduced the chances of assigning an incorrect or underestimate of global cases, although rare diseases and neglected diseases that lack centralized reporting systems, or diseases that do not typically require medical treatment are likely underestimated. Finally, we included data on the gold standard control strategy recommended by global health organizations for each disease or, for rare diseases with no public health strategy, for closely related diseases. Control strategies include: (i) behavior or lifestyle change; (ii) vaccination; (iii) water, sanitation, and hygiene (WASH); (iv) vector control; (v) reservoir host treatment (e.g., animal vaccination); (vi) integrated human and environmental control (e.g., mass drug administration (MDA) and environmental control); and, (vii) no clear standard. Where possible, data was acquired from the World Health Organization, the Centers for Disease Control and Prevention, and health communications resources provided by other academic, national, and global health institutions (i.e., (25)). Broad literature searches of peer-reviewed scientific publications (conducted between 2017 and 2020) were used to collect data on the ecology of rare diseases and to estimate duration of

17 time spent in human and environmental reservoirs (incubation and infectious periods); types of peer-reviewed sources used include publications on laboratory experiments, modeling exercises, and reviews. The life cycles, host ranges, environmental persistence, and control strategies of many diseases are still being investigated, and data presented here is likely to change as more knowledge is generated. Therefore, we encourage users to consider this database as a first step in synthesizing data on the life cycle biodiversity and environmental persistence of parasites and pathogens causing a substantial burden of human infectious disease, and to further investigate specific diseases of interest.

Descriptive statistics We used linear mixed effects models to make the following comparisons (i) duration of infectious stages outside vertebrate obligate hosts with primary transmission strategy, (ii) duration of infectious stages outside vertebrate obligate hosts with gold standard control and prevention practices, (iii) global cases (minimum value of logarithmic range) with transmission pathways, and (iv) global cases with obligate vertebrate host range. Both outcome variables (duration of infectious stages and global cases) were log-scaled. Random effect terms were included for the 11 major disease categories (Tuberculosis, STDs excluding HIV, HIV/AIDS, Diarrheal diseases, Childhood-cluster diseases, Meningitis, Encephalitis, Hepatitis, Parasitic and vector diseases, Intestinal nematode infections, Leprosy) excluding ‘Other infectious diseases’ within which pathogens and parasites are clustered according to the WHO GBD database, and for the disease-causing organisms’ clade (e.g., virus, bacterium, protozoa, helminth). We used ANOVA tests to look for differences in means of the outcome variables of interest according to factorial predictors assessed. All analyses were performed in the statistical computing software, R (26) and models were built using the ‘lme4’ package (27).

Database availability Alongside publication, all data will be deposited for download in Dryad (https://datadryad.org/stash).

Acknowledgements This research was conducted by the Ecological Levers for Health expert working group supported by the Science for Nature and People Partnership (SNAPP), a collaboration of The Nature Conservancy, the Wildlife Conservation Society, and the National Center for Ecological Analysis and Synthesis (NCEAS) at the University of California, Santa Barbara. IJJ was supported by a National Science Foundation Graduate Research Fellowship (#1656518). NN was supported by the Stanford Data Science Scholarship. AJL was supported by a James and Nancy Kelso Fellowship through the Stanford Interdisciplinary Graduate Fellowship program.

19 References

1. WHO, “Global Health Estimates 2016: Disease burden by Cause, Age, Sex, by Country and by Region, 2000-2016” (World Health Organization, Geneva, 2018).

2. N. Fullman et al., Measuring progress and projecting attainment on the basis of past trends of the health-related Sustainable Development Goals in 188 countries: an analysis from the Global Burden of Disease Study 2016. The Lancet. 390, 1423–1459 (2017).

3. A. Garchitorena, S. H. Sokolow, B. Roche, C. N. Ngonghala, M. Jocque, A. Lund, M. Barry, E. A. Mordecai, G. C. Daily, J. H. Jones, J. R. Andrews, E. Bendavid, S. P. Luby, A. D. LaBeaud, K. Seetah, J. F. Guégan, M. H. Bonds, G. A. De Leo, Disease ecology, health and the environment: a framework to account for ecological and socio-economic drivers in the control of neglected tropical diseases. Philos Trans R Soc Lond B Biol Sci. 372 (2017), doi:10.1098/rstb.2016.0128.

4. Patz Jonathan A., Daszak Peter, Tabor Gary M., Aguirre A. Alonso, Pearl Mary, Epstein Jon, Wolfe Nathan D., Kilpatrick A. Marm, Foufopoulos Johannes, Molyneux David, Bradley David J., null null, Unhealthy Landscapes: Policy Recommendations on Land Use Change and Infectious Disease Emergence. Environmental Health Perspectives. 112, 1092–1098 (2004).

5. N. L. Gottdenker, D. G. Streicker, C. L. Faust, C. R. Carroll, Anthropogenic Land Use Change and Infectious Diseases: A Review of the Evidence. EcoHealth. 11, 619–632 (2014).

6. J. V. Remais, J. N. S. Eisenberg, Balancing Clinical and Environmental Responses to Infectious Diseases. Lancet. 379, 1457–1459 (2012).

7. WHO, “WHO methods and data sources for global causes of death 2000-2016,” Global Health Estimates Technical Paper (WHO/HIS/HSI/GHE/2016.3, World Health Organization, Geneva, 2018), (available at http://www.who.int/healthinfo/global_burden_disease/GlobalCOD_method_2000_ 2016.pdf).

8. S. Whitmee, A. Haines, C. Beyrer, F. Boltz, A. G. Capon, B. F. de Souza Dias, A. Ezeh, H. Frumkin, P. Gong, P. Head, R. Horton, G. M. Mace, R. Marten, S. S. Myers, S. Nishtar, S. A. Osofsky, S. K. Pattanayak, M. J. Pongsiri, C. Romanelli, A. Soucat, J. Vega, D. Yach, Safeguarding human health in the Anthropocene epoch: report of The Rockefeller Foundation–Lancet Commission on planetary health. The Lancet. 386, 1973–2028 (2015).

9. T. Lang, G. Rayner, Ecological public health: the 21st century’s big idea? An essay by Tim Lang and Geof Rayner. BMJ. 345 (2012), doi:10.1136/bmj.e5466.

10. The malERA Consultative Group on Vector Control, A Research Agenda for Malaria Eradication: Vector Control. PLOS Medicine. 8, e1000401 (2011).

20 11. A. Ercumen, A. J. Pickering, L. H. Kwong, B. F. Arnold, S. M. Parvez, M. Alam, D. Sen, S. Islam, C. Kullmann, C. Chase, R. Ahmed, L. Unicomb, S. P. Luby, J. M. Colford, Animal Feces Contribute to Domestic Fecal Contamination: Evidence from E. coli Measured in Water, Hands, Food, Flies, and Soil in Bangladesh. Environ. Sci. Technol. 51, 8725–8734 (2017).

12. M. E. El Zowalaty, J. D. Järhult, From SARS to COVID-19: A previously unknown SARS- related coronavirus (SARS-CoV-2) of pandemic potential infecting humans – Call for a One Health approach. One Health. 9, 100124 (2020).

13. L. H. Taylor, S. M. Latham, M. E. J. Woolhouse, Risk factors for human disease emergence. Phil. Trans. R. Soc. Lond. B. 356, 983–989 (2001).

14. C. Kreuder Johnson, P. L. Hitchens, T. Smiley Evans, T. Goldstein, K. Thomas, A. Clements, D. O. Joly, N. D. Wolfe, P. Daszak, W. B. Karesh, J. K. Mazet, Spillover and pandemic properties of zoonotic viruses with high host plasticity. Sci Rep. 5, 14830 (2015).

15. N. D. Wolfe, C. P. Dunavan, J. Diamond, Origins of major human infectious diseases. Nature. 447, 279–283 (2007).

16. S. Cleaveland, M. K. Laurenson, L. H. Taylor, Diseases of humans and their domestic mammals: pathogen characteristics, host range and the risk of emergence. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences. 356, 991–999 (2001).

17. M. E. J. Woolhouse, S. Gowtage-Sequeria, Host Range and Emerging and Reemerging Pathogens. Emerg Infect Dis. 11, 1842–1847 (2005).

18. R. K. Plowright, C. R. Parrish, H. McCallum, P. J. Hudson, A. I. Ko, A. L. Graham, J. O. Lloyd-Smith, Pathways to zoonotic spillover. Nature Reviews Microbiology. 15, 502–510 (2017).

19. A. M. Kuris, The Global Burden of Human Parasites: Who and Where are They? How are They Transmitted? Journal of Parasitology. 98, 1056–1064 (2012).

20. A. M. Kuris, K. D. Lafferty, S. H. Sokolow, Sapronosis: a distinctive type of infectious agent. Trends in Parasitology. 30, 386–393 (2014).

21. WHO, “WHO methods and data sources for global burden of disease estimates 2000-2016” (Global Health Estimates Technical Paper WHO/HIS/IER/GHE/2018.4, World Health Organization, Geneva, 2018).

22. Centers for Disease Control and Prevention, Malaria (2020), (available at https://www.cdc.gov/parasites/malaria/index.html).

23. V. A. Meszaros, M. D. Miller-Dickson, F. B.-A. Junior, S. Almagro-Moreno, C. B. Ogbunugafor, Direct transmission via households informs models of disease and intervention dynamics in cholera. PLOS ONE. 15, e0229837 (2020).

21 24. J. D. Clemens, G. B. Nair, T. Ahmed, F. Qadri, J. Holmgren, Cholera. The Lancet. 390, 1539–1549 (2017).

25. A. R. Spickler, “Technical Factsheets” (The Center for Food Security and Public Health, 2020), (available at http://www.cfsph.iastate.edu/DiseaseInfo/factsheets.php).

26. R Core Team, R: The R Project for Statistical Computing (R Foundation for Statistical Computing, Vienna, Austria, 2018; https://www.r-project.org/).

27. D. Bates, M. Mächler, B. Bolker, S. Walker, Fitting Linear Mixed-Effects Models Using lme4. J. Stat. Soft. 67 (2015), doi:10.18637/jss.v067.i01.

22 Figures and Tables

Figure 1. A) Distributions of duration of infectious stages outside primary vertebrate hosts (but including vertebrates as obligate intermediate hosts) according to primary transmission strategy; plot and associated analysis exclude normal flora/opportunistic pathogens and directly zoonotic pathogens, which were data limited, and sapronoses, which persist indefinitely in abiotic reservoirs. B) Distributions of duration of infectious stages outside primary vertebrate hosts according to standard strategies for disease prevention and control; plot and associated analysis exclude sapronoses, which persist indefinitely in abiotic reservoirs. C) Distributions of the minimum estimated global cases according to obligate transmission pathways, categorized as direct human-to-human transmission (e.g., STDs), human-to-human transmission with obligate vertebrate stages (e.g., soil-transmitted helminths), and environment-to-human transmission (e.g., rabies virus). D) Distributions of the minimum estimated global cases according to obligate vertebrate host ranges; mixed category includes some combination of livestock, poultry, domestic animals, and wild animals (including birds).

23 Table 1. Summary of the transmission strategies and standard prevention and control strategies associated with 151 parasites and pathogens of humans tracked by the WHO Global Burden of Disease database. We also include the recently emerged novel coronavirus, SARS-CoV-2, for comparison. Key for transmission strategy: H-H: human-to-human; NF/O: normal flora/opportunistic; F-O: fecal-oral; FB: foodborne; WB: waterborne; VB: vectorborne; Z: directly zoonotic; S: sapronotic.

Disease category DALYs DALYs Trans. % (n) Obligate % (n) Vector or % (n) Env. stage % (n) Gold % (n) 2016 2000 strategy vert. Int. host outside standard host(s) host control Tuberculosis 51643 70474 H-H 50.0 (1) None 50.0 (1) None 100 (2) None 100 (2) Behavior/ 100 (2) Lifestyle Z 50.0 (1) Livestock 50.0 (1) STDs excluding HIV a. Syphilis 8635 10988 H-H 100 (1) None 100 (1) None 100 (1) None 100 (1) Behavior/ 100 (1) Lifestyle b. Chlamydia 1298 1109 H-H 100 (1) None 100 (1) None 100 (1) None 100 (1) Behavior/ 100 (1) Lifestyle c. Gonorrhoea 1477 1112 H-H 100 (1) None 100 (1) None 100 (1) None 100 (1) Behavior/ 100 (1) Lifestyle d. Trichomoniasis 198 154 H-H 100 (1) None 100 (1) None 100 (1) None 100 (1) Behavior/ 100 (1) Lifestyle e. Genital herpes 221 168 H-H 100 (1) None 100 (1) None 100 (1) None 100 (1) Behavior/ 100 (1) Lifestyle f. Other STDs 962 827 H-H 100 (5) None 100 (5) None 100 (5) None 100 (5) Behavior/ 80.0 (4) Lifestyle Vaccine 20.0 (1) HIV/AIDS 59951 90629 H-H 100 (2) None 100 (2) None 100 (2) None 100 (2) Behavior/ 100 (2) Lifestyle Diarrheal diseases 81743 157556 F-O 28 (14) Birds 8.0 (4) None 96.0 (48) Cyst 24.0 (12) No clear 16.0 (8) standard FB 4.0 (2) Domestic 6.0 (3) Vertebrate 4.0 (2) Freeliving 74.0 (37) Vaccine 18.0 (9) animals FB/WB 54.0 (27) Livestock 14.0 (7) Spore 2.0 (1) WASH 66.0 (33) H-H 14.0 (7) Mixed 24.0 (12) None 48.0 (24) Childhood-cluster diseases a. Whooping cough 894 1636 H-H 100 (2) None 100 (2) None 100 (2) Freeliving 100 (2) No clear 50.0 (1) standard

Vaccine 50.0 (1) b. Diphtheria 121 656 H-H 100 (1) None 100 (1) None 100 (1) Freeliving 100 (1) Vaccine 100 (1) c. Measles 7957 57524 H-H 100 (1) None 100 (1) None 100 (1) None 100 (1) Vaccine 100 (1) d. Tetanus 3989 19787 S 100 (1) None 100 (1) None 100 (1) Spore 100 (1) Vaccine 100 (1) Meningitis 20277 34891 H-H 83.3 (5) None 100 (6) None 100 (6) Freeliving 66.7 (4) No clear 33.3 (2) standard NF/O 16.7 (1) None 33.3 (2) Vaccine 50.0 (3) WASH 16.7 (1) Encephalitis 6354 9515 VB 100 (13) Birds 53.8 (7) Mosquito 69.2 (9) 100 (13) Reservoir 7.7 (1) None Tx Mixed 7.7 (1) Tick 30.8 (4) Vaccine 7.7 (1) Wild 38.5 (5) Vector 84.6 (11) animals control Hepatitis a. Acute hepatitis A 516 1189 F-O 100 (1) None 100 (1) None 100 (1) Freeliving 100 (1) Vaccine 100 (1) b. Acute hepatitis B 4698 4533 H-H 100 (1) None 100 (1) None 100 (1) Freeliving 100 (1) Vaccine 100 (1) c. Acute hepatitis C 104 93 H-H 100 (1) None 100 (1) None 100 (1) Freeliving 100 (1) Behavior/ 100 (1) Lifestyle d. Acute hepatitis E 2147 3289 F-O 100 (1) None 100 (1) None 100 (1) Freeliving 100 (1) WASH 100 (1) Parasitic and vector diseases a. Malaria 37369 67855 VB 100 (5) None 80.0 (4) Mosquito 100 (5) None 100 (5) Vector 100 (5) control Wild 20.0 (1) animals b. African 203 2127 VB 100 (1) None 100 (1) Fly 100 (1) None 100 (1) Vector 100 (1) Trypanosomiasis control c. Chagas disease 252 329 VB 100 (1) Mixed 100 (1) Kissing Bug 100 (1) None 100 (1) Vector 100 (1) control d. Schistosomiasis 2543 4292 WB 100 (11) Livestock 27.3 (3) Mollusc 100 (11) Larva 100 (11) Integrated 100 (11) Wild 27.3 (3) animals None 45.5 (5) e. Leishmaniasis 1069 1146 VB 100 (14) Mixed 78.6 (11) Sandfly 100 (14) None 100 (14) Vector 100 (14) control Wild 14.3 (2) animals

25 None 7.1 (1) f. Lymphatic 1186 1937 VB 100 (3) None 100 (3) Mosquito 100 (3) None 100 (3) Vector 100 (3) filariasis control g. Onchocerciasis 962 1371 VB 100 (1) None 100 (1) Fly 100 (1) None 100 (1) Vector 100 (1) control h. Taeniasis 1912 3362 FB 50.0 (1) None 100 (2) Vertebrate 100 (2) Egg 100 (2) WASH 100 (2) F-O 50.0 (1) i. Echinococcosis 687 1093 FB 100 (2) Mixed 100 (2) Vertebrate 100 (2) Egg 100 (2) Reservoir 100 (2) Tx j. Dengue 3100 1144 VB 100 (1) None 100 (1) Mosquito 100 (1) None 100 (1) Vector 100 (1) control k. Trachoma 245 263 VB 100 (1) None 100 (1) Fly 100 (1) None 100 (1) Behavior/ 100 (1) Lifestyle l. Yellow fever 739 1068 VB 100 (1) None 100 (1) Mosquito 100 (1) None 100 (1) Vector 100 (1) control m. Rabies 1571 3146 Z 100 (1) Mixed 100 (1) None 100 (1) None 100 (1) Vaccine 100 (1) Intestinal nematode infections a. Ascariasis 1433 2497 SB 100 (1) None 100 (1) None 100 (1) Egg 100 (1) Integrated 100 (1) b. Trichuriasis 337 465 SB 100 (1) None 100 (1) None 100 (1) Egg 100 (1) Integrated 100 (1) c. Hookworm 1682 1771 SB 100 (6) Domestic 50.0 (3) None 100 (6) Larva 100 (6) Integrated 100 (6) disease animals Wild 16.7 (1) animals None 33.3 (2) d. Foodborne 1084 1166 SB 16.7 (1) Mixed 16.7 (1) Mollusc 16.7 (1) Egg 50.0 (3) Integrated 16.7 (1) trematodes F-O 16.7 (1) None 33.3 (2) None 50.0 (3) Larva 50.0 (3) No clear 66.7 (4) standard FB 66.7 (4) Wild 50.0 (3) Vertebrate 33.3 (2) WASH 16.7 (1) animals Leprosy 407 571 H-H 100 (1) None 100 (1) None 100 (1) Free-living 100 (1) Behavior/ 100 (1) Lifestyle COVID19 NA NA H-H 100 (1) None 100 (1) None 100 (1) Free-living 100 (1) No clear 100 (1) standard

27 Chapter 2

Enemy of my enemy: Natural enemies for sustainable control

of human infections

Isabel J. Jones1*, Susanne H. Sokolow 1,2, Giulio A. De Leo1

1Hopkins Marine Station of Stanford University, 120 Oceanview Boulevard, Pacific

Grove, CA, USA 93950

2 Marine Science Institute, University of California, Santa Barbara, Santa Barbara, CA

93106

*Corresponding Author: Isabel J. Jones; [email protected]

28 ABSTRACT

Many agents of human infectious diseases spend a significant portion of their life cycle in environments where ecological interactions with natural enemies influence disease transmission. Here we discuss opportunities and challenges to developing natural enemy interventions as part of human infectious disease control strategies. We provide a broad set of examples where natural enemies of human infectious disease have been used or proposed to curb disease burden, and summarize successes, challenges, and recent developments. When used in complement to medical interventions, natural enemy solutions have the potential to reduce reliance on chemicals and drugs by targeting environmental sources of infection and re-infection, and may help reduce selection for insecticide-resistant vectors and drug-resistant pathogens. There may be opportunities where conserving, restoring, or augmenting specific natural enemies, such as endangered species or those with economic value (such as in fisheries), aligns human health goals with conservation goals and food security. Despite these promises, evidence-based examples of successful use of natural enemies for disease control in the environment are very rare. However, given that human diseases susceptible to natural enemy intervention affect the health of billions of people, developing successful and cost-effective natural enemy solutions may have an important role to play in global health.

29 1. Introduction The war against citrus pests in ancient Chinese orchards was not won by eliminating insect life, but by cultivating it. A strategy still in use today, farmers introduced yellow citrus ants, voracious predators of beetles, flies, and hymenopteran crop pests, to orchards to protect fruit from pest damage (1). This is the earliest documented example of biological control, or pest control using natural enemies. Active management of these natural “enemies of our enemies” is a key component of integrated pest management (IPM), an approach to achieve economically rational and environmentally sound pest management, around which commercial industries worth billions of dollars have emerged (2). Like agricultural pests, many parasites and pathogens that cause disease in humans spend a considerable amount of their life cycle in environments where they are exposed to natural enemies (Figure 1). Discovery of some natural enemy interactions with disease-causing organisms have led to world-changing technologies, including important biopesticides and therapeutic drugs. For instance, discovery of a fungal metabolite first isolated from Staphylococcus aureus-killing mould in 1928 eventually led to mass-production of Penicillin (3). Though uncommon in the current era of Western clinical medical practice, humans have long exploited natural enemies to prevent or cure disease. Millenia before Penicillin was isolated, Egyptian healing traditions involved direct application of moldy bread (actively producing antibacterial metabolites) to surface skin wounds – a practice that extended, in various forms and across centuries, to Asia, Europe, Canada and the United States (4). With adoption of modern germ theory in the 19th century, scientific understanding about disease origins and environmental transmission routes grew, and rapid advances in medicine, sanitation and hygiene, and public health infrastructure dramatically reduced global infectious disease morbidity and mortality (5). The age of antibiotic discovery (with drugs largely derived from “natural scaffolds” developed in soil bacteria and fungi (6)) ushered in the modern era of clinical disease intervention, which, even for many diseases with long-lasting environmental reservoirs, dominates global disease control strategies today (7). Despite notable attempts to control the environmental reservoirs of diseases and exciting advances in biotechnology and drug development, many environmentally mediated diseases – including malaria, diarrhea, and neglected tropical diseases (NTDs) –

30 remain among some of the most significant causes of morbidity and mortality worldwide (8). NTDs alone infect over a billion people, predominantly the world’s poorest and most vulnerable populations. Other diseases have emerged or re-emerged in recent decades, including West Nile virus, Zika, plague, avian influenza, and Lyme disease, among others (9–11). Effective vaccines have not yet been developed for most environmentally- mediated infectious diseases, and many disease-causing pathogens, parasites, and vectors evolve at rates rapid enough to jeopardize long-term interventions that rely on drug distribution or chemical vector control alone (11). Indeed, extensive use of synthetic antibiotics, antifungals, insecticides, and other chemicals has led to a worrying rise in multidrug-resistant bacteria and fungi, and insecticide-resistant vectors, since the mid- 20th century (6, 12). New (or reinvigorated) tools to complement clinical interventions and further prevent infections are badly needed. Improved access to Water, Sanitation and Hygiene (WASH), habitat modification, and chemical-based pest and vector control are all – in the right context – highly effective methods to limit environmental disease transmission and contamination. Here we specifically focus on the less studied option of natural enemies to fight environmentally- mediated diseases of public health importance. In Section 2 we review the ecological mechanisms by which natural enemies function. In Section 3 we provide a broad set of examples where natural enemies of human infectious disease (predators, competitors, and parasites of free-living stages or non-human hosts) have been implemented or proposed to curb human disease burdens, summarizing successes, challenges, and recent developments. Because the global mandate to fight malaria and other mosquito-borne diseases has garnered so much attention, we dedicate special attention to the history and future directions of some prominent mosquito natural enemies in Section 4. Finally, in Section 5 we discuss major challenges that may hinder natural enemy effectiveness and practicality in real-world settings.

2. Ecological interactions between natural enemies and disease-causing organisms 2.1 Predation and parasitism Virtually all organisms serve as food or parasitic hosts for other organisms. Predators consume and can limit free-living disease agents, or infected and susceptible disease hosts involved in transmission (i.e., vectors, non-human intermediate or definitive

31 hosts, reservoir hosts). Generalist predators (non-specialist predators with wide prey - breadth) – whether single species or species assemblages – can chronically suppress prey numbers, leading to stable, limited prey populations. Their role in disease control has often been appreciated only retrospectively, after their removal and consequently, unintentional disease outbreaks (13). For example, overfishing in Lake Malawi resulted in a loss of generalist fish predators of schistosome-transmitting snails; schistosomiasis disease outbreaks followed in people, revealing the potential role of generalist snail predators as important natural enemies of human schistosomes (14). Generalist predators are not limited to the animal kingdom: carnivorous fungi, which entrap and consume prey, have shown strong predatory action against free-living stages of taeniasis (tapeworm), roundworm, pinworm, hookworm, schistosomiasis, and fascioliasis in laboratory settings (15–18). Specialist predators, on the other hand, follow density- dependent dynamics closely linked to prey numbers, and may become limited when prey is scarce, but are advantageous when non-target impacts of natural enemies are a concern (19). Specialist predators that limit human diseases have been identified, such as the East African Jumping Spider that selectively preys on Anopheles mosquitoes (20), but experimental evidence of successful control of human diseases with specialist predators is still scarce in scientific literature. Intraguild predation occurs when predators consume prey items with which they also compete for shared resources. This can take the form of predation on juvenile stages of competitors, and also includes concomitant predation of a prey item’s own parasites or micropredator vectors that the predator competes with for host resources. A strong predatory effect of a natural intraguild Culex mosquito predators on other mosquitoes, including in the Aedes, Anopheles, and Culex genus (disease vectors for dengue, malaria, and Japanese encephalitis, respectively) has been observed in semi-field conditions in Sri Lanka (21). Specialist parasites and pathogens (including hyperparasites, i.e., parasites of parasites) are commonly used for pest control in agriculture, and are actively investigated as tools for human disease control ((19) & Supplementary Table 1). In agriculture, parasitoids (insects that lay eggs within and ultimately kill their host) are the most widely used natural enemy for insect control, and tend to share the following general characteristics: body size similar to that of its host; spatial and temporal synchrony with

32 the host; and narrow diet breadth (22). Commercially available pathogens used against insect disease vectors include the bacteria Bacillus thuringiensis israelensis (Bti), which is effective against numerous mosquito vector species (23–25), and the fungus Beauveria bassiana, which has been tested for use against mosquitoes, black flies, sand flies, and ticks (26–32). Bacteriophages (viruses of bacteria), the most abundant microorganisms on Earth, are currently being tested against infectious bacteria including food-borne bacterial pathogens, cholera, and shigella (33–37)

2.2 Competitive exclusion and decoy hosts Competition for resources can limit disease hosts and vectors in the external environment, or limit parasite and pathogen infections within hosts or vectors. For example, biologically diverse rodent communities in the western United States support a lower Sin Nombre hantavirus abundance, potentially because resource competition limits the abundance of the main hantavirus reservoir, Peromyscus maniculatus (38). Competition amongst trematodes that use the same snail species as obligate intermediate hosts has been shown to limit human schistosome infections within snails (39–42). In addition to limiting the abundance of a pathogen’s hosts or vectors through direct competition, some competitors (or non-competitor co-occurring species) can limit infections in primary disease hosts by serving as ‘decoy hosts’, non-hosts that absorb infectious agents and limit transmission success to the primary host, a process alternately termed encounter-reduction (43). For example, schistosome-competent snails co-exist with non-competent snail species in freshwater environments, and those non-competent hosts may incidentally absorb larval stages schistosomes, thus limiting snail infection prevalence in the target snail species (39). In natural environments with complex community structures, the mechanisms underlying the decoy effect are highly dynamic and difficult to predict (43).

3. Natural enemies of human pathogens Humans receive health benefits from natural enemies all the time. On human skin, mucosa, and in the gut, communities of beneficial microbes can suppress the proliferation of harmful bacteria (44). For example, mouse model studies suggest that disruption of communities of commensal microbes in our intestines likely facilitate colonization or

33 expansion of pathogens like Salmonella enterica serovar Typhimurium and Clostridium difficile via resource competition, and the production of antimicrobial compounds (45). In the environment, free-living disease agents, vectors, and non-human hosts face diverse natural enemies, largely operating without human intervention (Figure 1, Figure 2b). Dung beetles, for instance, can impact the abundance of helminths and protozoa in soil (46), and freshwater crustaceans, which are ubiquitous in many aquatic environments, graze on diverse bacteria including pathogens like Escherichia coli and Campylobacter jejuni (47, 48). Natural enemy types, like their targets, are taxonomically and functionally diverse. Previous reviews of natural enemies of some disease vectors, including mosquitoes, ticks, flies, and snails, reveal a wide range of biological control agents and interactions, including microbial pathogens (fungal, bacterial, and viral) and larger- bodied predators and competitors of vectors (Table 1) (49–59). Like in agriculture, the most used natural enemies for human disease are organisms that can be manipulated for commercial exploitation, including: microbial products used as medical treatment in the human body; microbial products broadcast into the environment for vector and pathogen control; and, freshwater predators of disease vectors that can be grown in aquaculture facilities, transported, and stocked where needed. As such, they are able to be reared in mass quantity, have long-lasting viability, can be distributed widely, and, in some cases, patented. Biological control for conservation, which relies on policy or community agency to secure species, habitat, or ecosystem management, is among the least explored options for human disease control, yet promises positive outcomes for people and nature.

3.1 Natural enemies within the human body – probiotics and phages Understanding species interactions in the human gut microbiome has helped guide clinical interventions to restore or augment beneficial microbial species that naturally suppress disease-causing ones (60, 61). Probiotics are consumable forms of live bacteria and yeasts known to act as natural enemies against many gut pathogens, and are commercially available worldwide. When administered in adequate amounts, randomized control trials (RCTs) compiled in the Cochrane Database of Systematic Reviews suggest that probiotics can shorten the duration of acute diarrhea in children under age 5 (62),

34 reduce the severity of acute upper respiratory tract infections (URTIs) (63), and prevent Clostridium difficile-associated diarrhea (CDAD) in high-risk groups (64). Phages (naturally occurring viruses that infect and kill bacteria and archea) were first reported from the stools of patients recovering from Shigella in 1917, and were immediately thereafter used to treat dysentery in human patients during the early part of the 20th century (65). Commercial research for phage therapy in the Western world dwindled in the 1930s when broad-spectrum antibiotics became widely available. In the former Soviet Union, however, phages continued to be investigated and through the 1970s were standard treatment for diarrhea, typhoid, dysentery, burns, and wound infections (65). With the recent escalation of antimicrobial resistance, bacteriophages have reemerged as a promising natural enemy option for human and animal diseases. Recent advances in synthetic biology that allow for genetic modification of phages to enhance their specificity, persistence, and safety may soon make phage therapy routine (66, 67).

3.2 Natural enemies for environmental intervention – bacteria, fungi, and phages Two of the most widely used natural enemies in natural environments are Bacillus thuringiensis subsp. israelensis (Bti) and Lysinibacillus sphaericus (formerly Bacillus sphaericus) (23). Bti is a naturally occurring, spore-forming soil bacterium that produces larvicidal toxins when consumed by mosquitoes, blackflies, and fungus gnats, and is the most commonly used inoculative biopesticide worldwide in many environments (e.g., wells, ponds, rice fields, tires, rain puddles, etc.). Notably, when faced with widespread insecticide-resistance in onchocerciasis-transmitting blackfly populations, the Onchocerciasis Control Programme (OCP) of West Africa turned to commercially available Bacillus thuringiensis subsp. israelensis (Bti) to enhance control of Simulium spp. blackfly larvae. The program ultimately reduced onchocerciasis in 16 participating countries through a combination of vector control and drug distribution (23). However, some evidence of insect resistance to the bacterium’s primary toxin has been reported, and the full range of non-target impacts, while generally understood to be minimal, are still being unraveled (68). Microbial communities in the midguts of blood-feeding insect vectors (i.e., mosquitoes, triatomine “kissing bugs”, tsetse flies, sand flies) are actively being targeted

35 by naturally occurring antagonistic microbes that may compromise parasite establishment and reduce vectorial capacity (69). The potential and speed at which resistance to such natural enemy tools can develop remains unknown, and could hinder long-term efficacy (69). Beyond bacteria, insect-killing fungi have proven amenable to pest and vector control. Metarhizium brunneum and Baeuveria bassiana are two commercially available entomopathogenic fungi that occur naturally in soil and can induce high mortality in adult Anopheles spp. mosquitoes (malaria vectors), Ixodes spp. black-legged ticks (Lyme disease vectors), and larval Phlebotomus sand flies (leishmaniasis vectors) (26–32). The spore-forming filamentous fungi germinate when in contact with the cuticle of certain insects or arachnids. Hyphae penetrate the cuticle and grow within the body cavity and eventually kill the host within a few days (Figure 1). Fungal spores can be inoculated on surfaces like bed nets for malaria control or broadcast through spraying into the environment. For example, field trials assessing M. brunneum F52 (commercially available as Met52, Novozymes Biological, Franklinton, NC, USA) to control Ixodes scapularis ticks in suburban settings in the northeastern United States showed comparable efficacy to chemical acaricides, and follow-up studies found no discernable reductions in non-target taxa (70, 71). However, host specificity of various fungal strains and the potential ecological impacts of altering non-target species abundances are still being explored. Commercial application of bacteriophages has also been used to limit food- and water-borne pathogens. Host-specific lytic bacteriophages or phage-derived endolysins (enzymes that lyse bacteria) have been identified, and in some cases commercialized, for use in food supply systems to destroy food-borne pathogens like Salmonella, Campylobacter, Listeria, Staphylococcus, and Vibrio spp. found in foods (plants, animal products, and packaging materials) (72). Augmentation of phages in natural water bodies contaminated with human pathogens has also been considered (Figure 1): there is observational evidence that phages can reduce V. cholerae in natural environments (36, 73), but, to our knowledge, no one has tested this in a clinically or epidemiologically relevant setting, or has fully explored a full range of potential and unwanted side effects.

3.3 Natural enemies for integrated disease control and food production

36 Not all commercially available natural enemies are microscopic, and some may have the added benefit of being valued food products for human consumption as well as disease control agents. For example, fish that can be grown in aquaculture facilities have been tested as a tool to control trematode (flatworm) infections. Controlling snail intermediate hosts with fish predators can limit human risk for waterborne schistosomiasis and foodborne trematodiases. In addition to threatening human health, trematode infections in fish exert an economic burden on fish farmers: heavy infections in fish can reduce their survival and marketability. Mixed evidence suggests that introducing exotic juvenile black carp (Mylopharyngodon piceus) that prey on snails may reduce trematode transmission to fish, and subsequently reduce food-borne disease risk to humans (74, 75). Under this scheme, exotic black carp, commercially available and valued for consumption, would be repeatedly stocked and removed to maintain individuals of the optimum size at which snail predation is maximized (76). Black carp have also been stocked in rice paddies to control mosquitoes in southern India and China, and showed increased rice yields during field experiments in both regions and a correlation (unestablished causation) with reduced malaria transmission after many years of stocking fish in China (77, 78). In settings where mosquito-borne disease and schistosomiasis are co-endemic, the scheme could theoretically limit mosquito and snail populations in rice fields at the same time. A downside of this approach is that black carp have invasive potential and would therefore need to be tightly controlled to avoid their spread. Restoration of edible aquatic snail predators has been proposed to control schistosomiasis in sub-Saharan Africa (79). Where schistosomiasis outbreaks are linked to overfishing or land-use change (14, 80–82), aquaculture of edible fish and crustaceans could be a direct route to restore important predator-prey dynamics to interrupt disease transmission, while simultaneously providing income-generating food products for local communities (81, 83).

3.4 Conservation of natural enemies for people and nature In some cases, the value of natural enemies has been realized only after their populations declined and human health was compromised. In India, for instance, evidence suggests that widespread losses of vulture populations coincided with a rise in

37 the number of feral dogs – as vulture populations decline, they may release food resources for feral dogs that also scavenge for carrion (84). This occurred in a country that harbors one third of the world’s human rabies cases (85), the majority of which are transmitted by dog bites, raising concerns about possible links between vulture declines in India and human rabies increases (84). Fears have also been raised regarding the loss of vultures and risk of other bacterial pathogens, including anthrax, which can proliferate in carcasses that would otherwise be efficiently cleared by vultures (86). In eastern Africa, experimental evidence suggests that conservation of large- bodied herbivores (e.g., zebra, giraffe, elephant, and gazelle) which compete with rodents for food and keep rodent populations low, can indirectly control human disease risk. Systematic increases were observed in the density of small-mammals (i.e., rodents) infected with zoonotic pathogens (e.g., bartonellosis, primarily transmitted by flea vectors) in defaunated areas, as compared to paired control areas with abundant wild herbivores (87). Thus, conservation of large herbivores might be an effective tool for preventing human disease outbreaks, but more evidence is needed (87, 88). In temperate North America, Lyme disease (Borrelia burgdorferi), an emerging zoonotic pathogen, is influenced by a complex interplay between small mammals (i.e., rodents, on which nymphal ticks feed and become infected), their predators (i.e., small predators including fox, bobcat, raccoon, and opossum), food resources (acorns), and important reproductive hosts for adult ticks (deer)(89). Some studies have suggested that small predators – like the red fox – could be conserved to reduce rodents (e.g.,(90, 91)), and top predators – like wolves – could offer deer control. Ticks have their own wild predators, including a diverse array of bird species, such as turkeys, guineafowl, and oxpeckers (92). However, there is currently no evidence that birds can reduce nymphal tick abundance enough to meaningfully reduce human disease risk. Moreover, some birds may feed on certain tick species that transmit one disease, while also serving as an important reproductive host to tick species that transmit other diseases, thus redistributing but not eliminating tickborne human disease risk (92).

4. Natural enemies of mosquitoes Given the immense global burden of mosquito-borne disease, it is unsurprising that natural enemies of medically important mosquitoes (species in the Anopheles, Aedes,

38 and Culex genus) and their pathogens have been extensively reviewed and trialed (93– 95). Here we do not attempt to systematically review the literature on mosquito natural enemies; rather, we provide insight into (i) the best studied strategy of the past – the use of larvivorous fish and crustaceans to control aquatic mosquito larvae, and (ii) the most active area of research currently – manipulation of the endosymbiont Wolbachia for mosquito population suppression and modification. Once mosquitoes were identified as vectors for several diseases around the close of the 19th century, mosquito eradication campaigns in the United States began in earnest, and ultimately limited (and in some places, locally eliminated) Yellow Fever, malaria, dengue (since re-emerged), and other arboviruses by the mid-20th century, and have kept new outbreaks under control. The tools used by control districts have evolved, and were initially based on habitat modification (ditching and draining of larval habitats), larvicidal oil, and natural enemies (Gambusia affinis); then, nearly exclusively on widespread insecticide application; and again more recently on integrated control (including insecticides, biopesticides, and natural enemies) (96, 97). For all of these vector control strategies, a major gap in understanding how effects of natural enemies on vectors actually translates into effects on human disease. Despite these advances, half of the world’s population remains at risk for acquiring mosquito-borne diseases (98, 99). In resource-poor contexts, chemical-based control can be prohibitively expensive, environmentally damaging (though decreasingly so), and of dwindling efficacy due to the dramatic rise in insecticide-resistant mosquito populations worldwide (Figure 2a). Thus, there is the need for new vector control options that can be tailored for use in appropriate socio-ecological settings.

4.1 Larvivorous predators as natural enemies for mosquito control Predators of mosquito larvae and pupae in natural environments are abundant, and include fish, copepods, amphibians, odonate (dragonfly and damselfly) instars, water bugs, and even other mosquitoes (Table 1, Figure 2b) (93, 100, 101). Adult mosquitoes, too, have a broad range of understudied natural enemies, including birds, bats, arachnids, and arthropods (Table 1, Figure 2b). The use of larvivorous fish for mosquito control has been the most widely studied and distributed mosquito natural enemy solution for decades. More than 300 species of fish have been used in some context for mosquito

39 control (99), but the mosquitofish (Gambusia affinis) has been the mostly widely used. Mosquitofish can establish self-sustaining populations and consume large numbers of mosquito larvae, and thus, are credited with assisting in the eradication of malaria (transmitted by Anopheline mosquitoes) in numerous locations (e.g., (102)). For this reason, they have been introduced globally from their native ranges in southern and eastern North America, including in other parts of the USA, for more than 100 years to control mosquitoes. However, a recent Cochrane Review suggests that evidence for a causal relationship linking introductions of mosquitofish, adult Anopheles populations, and human malaria incidence is lacking (99). In addition, mosquitofish are now considered one of the most invasive fish species worldwide, with significant impact on native fauna, as they consume a wide range of native, non-target insect, fish, and amphibian eggs wherever they are introduced (103). Even so, mosquitofish are still widely used: permits to obtain and use mosquitofish are currently offered in many California counties to reduce the population size of Culex mosquitoes, regional vectors for West Nile Virus (e.g., (104)). Larvivorous fish and other predators are also widely considered for the control of Aedes mosquitoes to curb the global spread of dengue virus. World Health Organization documentation of potential biological control agents of Aedes mosquitoes for dengue control currently includes fish (guppies) and predatory copepods (crustaceans) (105) (Figure 2b). A series of promising studies carried out over several years in Vietnam showed that Mesocyclops copepods introduced to mosquito breeding containers can reduce larval and adult mosquito density and dengue seroprevalence in humans (106). Such a large-scale intervention has not been replicated in other locations, however, and success in the Vietnamese communities may not translate to locations without environmental conditions conducive to copepod survival; discrete and accessible vector breeding sites; and, significant community involvement. Idiosyncratic results of mosquito control programs using larvivorous fish and copepods may indicate that a ‘one size fits all’ approach is unrealistic. Limitations to larval source management – whether by natural enemies or chemical larvicides – are exacerbated when breeding sites are inaccessible, widespread, ephemeral, or highly variable (107). In areas of high disease burden, near complete elimination of vector populations may be required before substantial reductions in disease transmission to

40 humans is achieved (108). Even so, small-scale and well-designed disease control approaches using natural enemies should not be discounted because they can still have substantial impacts on local community health and economies.

4.2 Manipulating mosquito or parasite populations with Wolbachia endosymbionts The best studied mosquito endosymbionts are strains of Wolbachia pipientis, a rickettsia-like bacterium that naturally infects a wide range of insects. Theoretically, Wolbachia can help fight mosquito-borne diseases via two mechanisms: (i) population suppression via release of Wolbachia-infected males that produce sterile offspring when mated with wild-type females, reducing the vector population over many generations (outcomes similar to those using irradiated sterile insect techniques (109)) , and (ii) population replacement via release of male and female mosquitoes infected with a vertically-transmissible Wolbachia strains that confer resistance to specific pathogens (Figure 2b) (98). The release of Wolbachia-containing mosquitoes into the environment has been ongoing for nearly a decade. In 2016, ‘sterile’ male (non-biting) Aedes aegypti and Aedes albopictus infected with Wolbachia were released in Fresno, California in pursuit of population suppression for the first time in the United States (110, 111). However, this is, by design, a self-limiting strategy, as the lethal gene disappears from the population if not regularly replenished; therefore it requires a great deal of resources and effort to maintain control through repeated release of large numbers of genetically manipulated mosquitoes (113). Mosquito population replacement schemes are also being rolled out at large scales across the world. Following promising pilot studies in 2011, 12 countries across South Asia, Oceana, and the Americas have now adopted population replacement strategies for control of dengue, Zika, and chikungunya in Aedes mosquitoes (111). Wolbachia can suppress pathogen development for viruses transmitted by Aedes and for malaria in Anopheles mosquitoes. However, Wolbachia infection in Anopheles can sometimes increase vector competence, depending on vector species, which would severely challenge the dissemination of this solution into natural environments where transmission occurs (112). And, because this strategy is self-sustaining, they are considered more controversial and risky (113).

41 5. Limitations and challenges of natural enemies for human disease control With the exception of solutions that have been developed for safe commercial application, the effectiveness, feasibility, and cost-effectiveness for most natural enemy solutions previously discussed remain unclear. Natural enemies of human infections might reduce reliance on antibiotics or chemical insecticides, but microbial enemies like bacteria, fungi, and phages can exert evolutionary pressure towards resistance, too. Integrating evolutionary ecology into the design and forecasting of natural enemy efficiency will help navigate such risks. Natural enemies that have been successfully harnessed for commercial use (e.g., probiotics, bacteriophages, Bti, entomopathogenic fungi, and, increasingly, Wolbachia infected vectors) share some common features: they tend to be highly specific, environmentally hardy, scalable (i.e., easy to produce in mass quantity, and transport and where and when needed). Many of the proposed solutions discussed in previous sections do not yet meet these criteria, and more research is needed to overcome formidable challenges. Although it is tempting to condemn natural enemies because of the cautionary tales of classical biological control gone wrong (i.e., the introduction of exotic enemies that later caused unexpected nontarget effects or invasions of unwanted species, like the introduction of the infamous cane toad (114) in Australia), these examples are comparably rare when considering integrated pest and disease management successes, such as in agriculture. Nonetheless, a thorough understanding of unintended natural enemy invasion or escape risk should be a major research priority when testing efficacy and considering implementation. Recent technological advances in bioengineering natural enemies, like monosex aquaculture of freshwater crustaceans that help control schistosomiasis disease, could ensure they are unable to reproduce without management, making them safer (115, 116). While this may add to intervention costs, as their use as a biological control agent requires repeated introduction of an otherwise unsustainable population, it eliminates the environmental danger that exotic species can pose if they became established. Long-term, such a solution could become a cost-effective complement to medical interventions if integrated with aquaculture production(83). Investment in natural enemy solutions is currently impeded by lack of evidence, both for identifying a causal link between natural enemies and epidemiological outcomes,

42 and for identifying specific contexts in which solutions can be effective. Randomized- controlled-trials (RCTs) have been used to test biomedical and behavioral interventions for many infectious diseases, and could be applied for biological control tools, too (REF). However, they are often considered impractical and cost prohibitive for large scale environmental interventions, and other forms of evidence, like observational studies and natural experiments, are more common. Moving forward, confidence in biological control as an effective and realistic tool for human health will grow if we apply the same standards and investments in these tools as we do for biomedical and behavioral ones.

6. Conclusions Despite the unknowns, natural enemies for human disease control offer a promising strategy to bridge ecological disease control with goals to protect the environment, provide nutrition, or fight poverty, aligning well with the United Nation’s 2030 Agenda for Sustainable Development (117). This is especially true for resource- poor populations where interactions between disease, environmental degradation, and poverty perpetuate complex feedbacks that make long-term disease control and environmental sustainability especially challenging (118, 119). The burden of diseases with environmental transmission weighs heavily upon the world’s most vulnerable populations. Human health, environmental change, and poverty are intimately linked. It is at this confluence that natural enemies may be most important to consider, since they could enhance disease control efficiency when used to complement medical care alone. Implementing natural enemy solutions at a scale capable of interrupting disease transmission to at-risk populations has proven challenging, but successful use of natural enemies promises better health and environmental outcomes in a rapidly changing world.

43 Acknowledgements: We thank Armand Kuris for early conversations and inspiration on natural enemies of parasites and parasite ecology, and Kate Lamy for her artwork. Funding: I.J.J. is supported by the National Science Foundation Graduate 399 Research Fellowship Program, DGE – 1656518. I.J.J, S.H.S., and G.D.L. have been supported by NSF CNH grant no. 1414102, the Bill and Melinda Gates Foundation, NIH grant no. 1R01TW010286-01, and the SNAP-NCEAS-supported working group ‘Ecological levers for health: Advancing a priority agenda for Disease Ecology and Planetary Health in the 21st century’.

Conflict of Interest: The authors declare no conflicts of interest.

44 References 1. H. T. Huang, P. Yang, The Ancient Cultured Citrus Ant. BioScience. 37, 665–671 (1987).

2. S. E. Naranjo, P. C. Ellsworth, G. B. Frisvold, Economic Value of Biological Control in Integrated Pest Management of Managed Plant Systems. Annu. Rev. Entomol. 60, 621–645 (2015).

3. R. Hare, New light on the history of penicillin. Med. Hist. 26, 1–24 (1982).

4. M. Wainwright, Moulds in ancient and more recent medicine. Mycologist. 3, 21– 23 (1989).

5. Institute of Medicine (US) Committee for the Study of the Future of Public Health, A History of the Public Health System (National Academies Press (US), 1988; https://www.ncbi.nlm.nih.gov/books/NBK218224/).

6. E. D. Brown, G. D. Wright, Antibacterial drug discovery in the resistance era. Nature. 529, 336–343 (2016).

7. J. V. Remais, J. N. S. Eisenberg, Balance between clinical and environmental responses to infectious diseases. The Lancet. 379, 1457–1459 (2012).

8. WHO, Global Health Estimates 2015: Disease burden by Cause, Age, Sex, by Country and by Region, 2000-2015. (2016).

9. A. M. Kilpatrick, S. E. Randolph, Drivers, dynamics, and control of emerging vector-borne zoonotic diseases. The Lancet. 380, 1946–1955 (2012).

10. A. S. Fauci, D. M. Morens, Zika Virus in the Americas — Yet Another Arbovirus Threat. N. Engl. J. Med. 374, 601–604 (2016).

11. D. M. Morens, G. K. Folkers, A. S. Fauci, The challenge of emerging and reemerging infectious diseases. Nature. 430, 242–249 (2004).

12. M. C. Fisher, N. J. Hawkins, D. Sanglard, S. J. Gurr, Worldwide emergence of resistance to antifungal drugs challenges human health and food security. Science. 360, 739–742 (2018).

13. R. S. Ostfeld, R. D. Holt, Are predators good for your health? Evaluating evidence for top-down regulation of zoonotic disease reservoirs. Front. Ecol. Environ. 2, 13–20 (2004).

14. J. R. Stauffer, H. Madsen, K. McKaye, A. Konings, P. Bloch, C. P. Ferreri, J. Likongwe, P. Makaula, Schistosomiasis in Lake Malawi: Relationship of Fish and Intermediate Host Density to Prevalence of Human Infection. EcoHealth. 3, 22–27 (2006).

45 15. F. R. Braga, A. Ricardo, J. M. Araújo, R. O. Carvalho, J. Victor, Predatory activity of the nematophagous fungi Duddingtonia flagrans, Monacrosporium thaumasium and Artrobotrys robusta on Strongyloides stercoralis. 43, 588–590 (2010).

16. L. N. Frassy, F. R. Braga, A. e Silva, J. V. de Araujo, S. R. Ferreira, L. G. de Freitas, Destruction of Toxocara canis eggs by the nematophagous fungus Pochonia chlamydosporia. Rev. Soc. Bras. Med. Trop. 43, 102–104 (2010).

17. F. De Souza Maia Filho, J. Nunes Vieira, M. E. Aires Berne, F. E. Stoll, P. Da Silva Nascente, L. Pötter, D. I. Brayer Pereira, Fungal ovicidal activity on toxocara canis eggs. Rev. Iberoam. Micol. 30, 226–230 (2013).

18. F. R. Braga, J. V. de Araujo, Nematophagous fungi for biological control of gastrointestinal nematodes in domestic animals. Appl. Microbiol. Biotechnol. 98, 71–82 (2014).

19. W. O. C. Symondson, K. D. Sunderland, M. H. Greenstone, Can generalist predators be effective biocontrol agents. Annu. Rev. Entomol. 47, 561–594 (2002).

20. S. N. Surendran, P. J. Jude, A. C. Thavaranjit, T. Eswaramohan, M. Vinobaba, R. Ramasamy, PREDATORY EFFICACY OF CULEX (LUTZIA) FUSCANUS ON MOSQUITO VECTORS OF HUMAN DISEASES IN SRI LANKA. J. Am. Mosq. CONTROL Assoc. 29, 168–170 (2013).

21. C. S. Straub, D. L. Finke, W. E. Snyder, Are the conservation of natural enemy biodiversity and biological control compatible goals? Biol. Control. 45, 225–237 (2008).

22. L. Regis, M. H. Silva-Filha, C. Nielsen-LeRoux, J. F. Charles, Bacteriological larvicides of dipteran disease vectors. Trends Parasitol. 17 (2001), pp. 377–380.

23. U. Fillinger, B. G. J. Knols, N. Becker, Efficacy and efficiency of new Bacillus thuringiensis var. israelensis and Bacillus sphaericus formulations against Afrotropical anophelines in Western Kenya. Trop. Med. Int. Health. 8, 37–47 (2003).

24. T. G. Floore, Mosquito larval control practices: Past and present. J. Am. Mosq. CONTROL Assoc. 22, 527–533 (2006).

25. E. J. Scholte, B. G. J. Knols, W. Takken, Infection of the malaria mosquito Anopheles gambiae with the entomopathogenic fungus Metarhizium anisopliae reduces blood feeding and fecundity. J. Invertebr. Pathol. 91, 43–49 (2006).

26. M. Farenhorst, A. Hilhorst, M. B. Thomas, B. G. J. Knols, Development of Fungal Applications on Netting Substrates for Malaria Vector Control. J. Med. Entomol. 48, 305–313 (2011).

46 27. J. George, N. E. Jenkins, S. Blanford, M. B. Thomas, T. C. Baker, Malaria Mosquitoes Attracted by Fatal Fungus. Plos One. 8 (2013), doi:10.1371/journal.pone.0062632.

28. G. P. Kaaya, Laboratory and field evaluation of entomogenous fungi for tick control. Ann. N. Y. Acad. Sci. 916, 559–564 (2000).

29. V. L. Hornbostel, R. S. Ostfeld, M. a Benjamin, Effectiveness of Metarhizium anisopliae (Deuteromycetes) against Ixodes scapularis (Acari: Ixodidae) engorging on Peromnyscus leucopus. J. Vector Ecol. J. Soc. Vector Ecol. 30, 91–101 (2005).

30. E. K. K. Fernandes, V. R. Elias Pinheiro Bittencourt, Entomopathogenic fungi against South American tick species. Exp. Appl. Acarol. 46, 71–93 (2008).

31. S. S. A. Amora, C. M. L. Bevilaqua, F. M. C. Feijo, M. A. Silva, R. Pereira, S. C. Silva, N. D. Alves, F. A. M. Freire, D. M. Oliveira, Evaluation of the fungus Beauveria bassiana (Deuteromycotina: Hyphomycetes), a potential biological control agent of Lutzomyia longipalpis (Diptera, Psychodidae). Biol. Control. 50, 329–335 (2009).

32. S. Dobhal, G. Zhang, L. M. Ma, M. Arif, Screening of biological control agents from fresh produce against foodborne human pathogens. Phytopathology. 102, 30 (2012).

33. S. M. Faruque, I. Bin Naser, M. J. Islam, a S. G. Faruque, a N. Ghosh, G. B. Nair, D. a Sack, J. J. Mekalanos, Seasonal epidemics of cholera inversely correlate with the prevalence of environmental cholera phages. Proc. Natl. Acad. Sci. U. S. A. 102, 1702–1707 (2005).

34. J. W. Jun, S. S. Giri, H. J. Kim, S. K. Yun, C. Chi, J. Y. Chai, B. C. Lee, S. C. Park, Bacteriophage application to control the contaminated water with Shigella. Sci. Rep. 6 (2016), doi:10.1038/srep22636.

35. E. J. Nelson, J. B. Harris, J. G. Morris, S. B. Calderwood, A. Camilli, Cholera transmission: the host, pathogen and bacteriophage dynamic. Nat. Rev. Microbiol. 7, 693–702 (2009).

36. J. Zhang, Z. Li, Z. Cao, L. Wang, X. Li, S. Li, Y. Xu, Bacteriophages as antimicrobial agents against major pathogens in swine: a review. J. Anim. Sci. Biotechnol. 6 (2015), doi:10.1186/s40104-015-0039-7.

37. C. A. Clay, E. M. Lehmer, S. St. Jeor, M. D. Dearing, Sin Nombre virus and rodent species diversity: A test of the dilution and amplification hypotheses. PLoS ONE. 4 (2009), doi:10.1371/journal.pone.0006467.

38. P. T. Johnson, P. J. Lund, R. B. Hartson, T. P. Yoshino, Community diversity reduces Schistosoma mansoni transmission, host pathology and human infection risk. Proc. R. Soc. B Biol. Sci. 276, 1657–1663 (2009).

47 39. M. R. Laidemitt, L. C. Anderson, H. J. Wearing, M. W. Mutuku, G. M. Mkoji, E. S. Loker, Antagonism between parasites within snail hosts impacts the transmission of human schistosomiasis. eLife. 8, e50095 (2019).

40. Y. Sulieman, T. Pengsakul, Competitive Capability of Exorchis sp (Trematoda: Cryptogonimidae) Against Schistosoma japonicum (Trematoda: Schistosomatoidae) on Larval Development within the Intermediate Host Snail, Oncomelania hupensis. Egypt. J. Biol. PEST CONTROL. 23, 43–47 (2013).

41. C.-T. Tang, M.-K. Lu, Y. Guo, Y.-N. Wang, J.-Y. Peng, W.-B. Wu, W.-H. Li, B. C. Weimer, D. Chen, Development of larval Schistosoma japonicum blocked in Oncomelania hupensis by pre-infection with larval Exorchis sp. J. Parasitol. 95, 1321–5 (2009).

42. D. W. Thieltges, K. T. Jensen, R. Poulin, The role of biotic factors in the transmission of free-living endohelminth stages. Parasitology. 135, 407–426 (2008).

43. L. Dethlefsen, M. McFall-Ngai, D. A. Relman, An ecological and evolutionary perspective on human–microbe mutualism and disease. Nature. 449, 811–818 (2007).

44. K. M. Ng, J. A. Ferreyra, S. K. Higginbottom, J. B. Lynch, P. C. Kashyap, S. Gopinath, N. Naidu, B. Choudhury, B. C. Weimer, D. M. Monack, J. L. Sonnenburg, Microbiota-liberated host sugars facilitate post-antibiotic expansion of enteric pathogens. Nature. 502, 96–99 (2013).

45. E. Nichols, S. Spector, J. Louzada, T. Larsen, S. Amezquita, M. E. Favila, Ecological functions and ecosystem services provided by Scarabaeinae dung beetles. Biol. Conserv. 141, 1461–1474 (2008).

46. J. W. McMahon, F. H. Rigler, Feeding Rate of Daphnia Magna Straus in Different Foods Labeled with Radioactive Phosphorus1. Limnol. Oceanogr. 10, 105–113 (1965).

47. M. Schallenberg, P. J. Bremer, S. Henkel, A. Launhardt, C. W. Burns, Survival of Campylobacter jejuni in Water: Effect of Grazing by the Freshwater Crustacean Daphnia carinata (Cladocera). Appl. Environ. Microbiol. 71, 5085–5088 (2005).

48. D. W. Jenkins, Pathogens, parasites and predators of medically important arthropods. Bull. World Health Organ. 30, 1–150 (1964).

49. L. Kamareddine, The biological control of the malaria vector. Toxins Basel. 4, 748–767 (2012).

50. M. Samish, J. Rehacek, Pathogens and predators of ticks and their potential in biological control. Annu. Rev. Entomol. 44 (1999), pp. 159–182.

48 51. M. Samish, H. Ginsberg, I. Glazer, Biological control of ticks. Parasitology. 129 Suppl, S389–S403 (2004).

52. T. E. Erlanger, J. Keiser, J. Utzinger, Effect of dengue vector control interventions on entomological parameters in developing countries: a systematic review and meta-analysis. Med Vet Entomol. 22, 203–221 (2008).

53. V. Lardans, C. Dissous, Snail control strategies for reduction of schistosomiasis transmission. Parasitol. Today. 14 (1998), pp. 413–417.

54. B. Fried, J. Pointier, P. David, P. Jarne, Biomphalaria Snails and Larval Trematodes. Biomphalaria Snails Larval Trematodes, 215–38 (2011).

55. T. A. Garrard, K. E. Ross, M. J. Taylor, Biocontrol for environmental control of the human parasite, Strongyloides stercoralis: A necessary next step. Biol. Control. 100 (2016), pp. 25–28.

56. J. Keiser, M. F. Maltese, T. E. Erlanger, R. Bos, M. Tanner, B. H. Singer, J. Utzinger, Effect of irrigated rice agriculture on Japanese encephalitis, including challenges and opportunities for integrated vector management. Acta Trop. 95 (2005), pp. 40–57.

57. L. A. Lacey, B. K. Orr, in American Journal of Tropical Medicine and Hygiene (1994), vol. 50, pp. 97–115.

58. L. A. Lacey, C. M. Lacey, The medical importance of riceland mosquitoes and their control using alternatives to chemical insecticides. J. Am. Mosq. Control Assoc. Suppl. 2, 1–93 (1990).

59. R. L. Koretz, M. Rotblatt, Complementary and alternative medicine in gastroenterology: the good, the bad, and the ugly. Clin. Gastroenterol. Hepatol. Off. Clin. Pract. J. Am. Gastroenterol. Assoc. 2, 957–967 (2004).

60. Z. Kassam, C. H. Lee, Y. Yuan, R. H. Hunt, Fecal Microbiota Transplantation for Clostridium difficile Infection: Systematic Review and Meta-Analysis. Am. J. Gastroenterol. 108, 500–508 (2013).

61. G. B. Aponte, C. A. B. Mancilla, N. Y. C. Pariasca, R. A. R. Galarza, Probiotics for treating persistent diarrhoea in children. Cochrane Database Syst. Rev. (2010), doi:10.1002/14651858.CD007401.pub2.

62. Q. Hao, B. R. Dong, T. Wu, Probiotics for preventing acute upper respiratory tract infections. Cochrane Database Syst. Rev. (2015), doi:10.1002/14651858.CD006895.pub3.

63. J. Z. Goldenberg, C. Yap, L. Lytvyn, C. K.-F. Lo, J. Beardsley, D. Mertz, B. C. Johnston, Probiotics for the prevention of Clostridium difficile‐associated diarrhea in adults and children. Cochrane Database Syst. Rev. (2017), doi:10.1002/14651858.CD006095.pub4.

49 64. N. Chanishvili, in Advances in Virus Research, M. Łobocka, W. Szybalski, Eds. (Academic Press, 2012; http://www.sciencedirect.com/science/article/pii/B9780123944382000013), vol. 83 of Bacteriophages, Part B, pp. 3–40.

65. R. M. Dedrick, C. A. Guerrero-Bustamante, R. A. Garlena, D. A. Russell, K. Ford, K. Harris, K. C. Gilmour, J. Soothill, D. Jacobs-Sera, R. T. Schooley, G. F. Hatfull, H. Spencer, Engineered bacteriophages for treatment of a patient with a disseminated drug-resistant Mycobacterium abscessus. Nat. Med. 25, 730 (2019).

66. F. L. Nobrega, A. R. Costa, L. D. Kluskens, J. Azeredo, Revisiting phage therapy: new applications for old resources. Trends Microbiol. 23, 185–191 (2015).

67. L. A. Lacey, Bacillus thuringiensis serovariety israelensis and bacillus sphaericus for mosquito control. J. Am. Mosq. Control Assoc. 23, 133–163 (2007).

68. W. R. Shaw, F. Catteruccia, Vector biology meets disease control: using basic research to fight vector-borne diseases. Nat. Microbiol. (2018), , doi:10.1038/s41564-018-0214-7.

69. A. Bharadwaj, K. C. Stafford, Evaluation of Metarhizium anisopliae Strain F52 (Hypocreales: Clavicipitaceae) for Control of Ixodes scapularis (Acari: Ixodidae). J. Med. Entomol. 47, 862–867 (2010).

70. I. R. Fischhoff, F. Keesing, R. S. Ostfeld, The tick biocontrol agent Metarhizium brunneum (= M. anisopliae) (strain F52) does not reduce non-target arthropods. PLOS ONE. 12, e0187675 (2017).

71. J. Bai, Y.-T. Kim, S. Ryu, J.-H. Lee, Biocontrol and Rapid Detection of Food- Borne Pathogens Using Bacteriophages and Endolysins. Front. Microbiol. 7 (2016), doi:10.3389/fmicb.2016.00474.

72. C. A. Silva-Valenzuela, A. Camilli, Niche adaptation limits bacteriophage predation of Vibrio cholerae in a nutrient-poor aquatic environment. Proc. Natl. Acad. Sci. 116, 1627–1632 (2019).

73. N. M. Hung, N. V Duc, J. R. Stauffer Jr., H. Madsen, Use of black carp (Mylopharyngodon piceus) in biological control of intermediate host snails of fish- borne zoonotic trematodes in nursery ponds in the Red River Delta, Vietnam. Parasit. Vectors. 6 (2013), doi:10.1186/1756-3305-6-142.

74. J. H. Clausen, H. Madsen, P. T. Van, A. Dalsgaard, K. D. Murrell, Integrated parasite management: Path to sustainable control of fishborne trematodes in aquaculture. Trends Parasitol. 31 (2015), pp. 8–15.

75. F. Ben-Ami, J. Heller, Biological Control of Aquatic Pest Snails by the Black Carp Mylopharyngodon piceus. Biol. Control. 22, 131–138 (2001).

50 76. T. J. Victor, B. Chandrasekaran, R. Reuben, Composite fish culture for mosquito control in rice fields in southern India. Southeast Asian J. Trop. Med. Public Health. 25, 522–527 (1994).

77. N. Wu, G. H. Liao, D. F. Li, Y. L. Luo, G. M. Zhong, The advantages of mosquito biocontrol by stocking edible fish in rice paddies. Southeast Asian J. Trop. Med. Public Health. 22, 436–442 (1991).

78. P. J. Hotez, M. Alvarado, M.-G. Basáñez, I. Bolliger, R. Bourne, M. Boussinesq, S. J. Brooker, A. S. Brown, G. Buckle, C. M. Budke, H. Carabin, L. E. Coffeng, E. M. Fèvre, T. Fürst, Y. A. Halasa, R. Jasrasaria, N. E. Johns, J. Keiser, C. H. King, R. Lozano, M. E. Murdoch, S. O’Hanlon, S. D. S. Pion, R. L. Pullan, K. D. Ramaiah, T. Roberts, D. S. Shepard, J. L. Smith, W. A. Stolk, E. A. Undurraga, J. Utzinger, M. Wang, C. J. L. Murray, M. Naghavi, The Global Burden of Disease Study 2010: Interpretation and Implications for the Neglected Tropical Diseases. PLoS Negl. Trop. Dis. 8, e2865 (2014).

79. N. T. Halstead, C. M. Hoover, A. Arakala, D. J. Civitello, G. A. De Leo, M. Gambhir, S. A. Johnson, N. Jouanard, K. A. Loerns, T. A. McMahon, R. A. Ndione, K. Nguyen, T. R. Raffel, J. V. Remais, G. Riveau, S. H. Sokolow, J. R. Rohr, Agrochemicals increase risk of human schistosomiasis by supporting higher densities of intermediate hosts. Nat. Commun. 9, 837 (2018).

80. S. H. Sokolow, E. Huttinger, N. Jouanard, M. H. Hsieh, K. D. Lafferty, A. M. Kuris, G. Riveau, S. Senghor, C. Thiam, A. N’Diaye, D. S. Faye, G. a. De Leo, Reduced transmission of human schistosomiasis after restoration of a native river prawn that preys on the snail intermediate host. Proc. Natl. Acad. Sci. 112, 9650– 9655 (2015).

81. S. H. Sokolow, I. J. Jones, M. Jocque, D. La, O. Cords, A. Knight, A. Lund, C. L. Wood, K. D. Lafferty, C. M. Hoover, P. A. Collender, J. V. Remais, D. Lopez- Carr, J. Fisk, A. M. Kuris, G. A. De Leo, Nearly 400 million people are at higher risk of schistosomiasis because dams block the migration of snail-eating river prawns. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 372 (2017), doi:10.1098/rstb.2016.0127.

82. C. M. Hoover, S. H. Sokolow, J. Kemp, J. N. Sanchirico, A. J. Lund, I. J. Jones, T. Higginson, G. Riveau, A. Savaya, S. Coyle, C. L. Wood, F. Micheli, R. Casagrandi, L. Mari, M. Gatto, A. Rinaldo, J. Perez-Saez, J. R. Rohr, A. Sagi, J. V. Remais, G. A. De Leo, Modelled effects of prawn aquaculture on poverty alleviation and schistosomiasis control. Nat. Sustain. 2, 611–620 (2019).

83. G. Swan, V. Naidoo, R. Cuthbert, R. E. Green, D. J. Pain, D. Swarup, V. Prakash, M. Taggart, L. Bekker, D. Das, J. Diekmann, M. Diekmann, E. Killian, A. Meharg, R. C. Patra, M. Saini, K. Wolter, Removing the threat of diclofenac to critically endangered Asian vultures. PLoS Biol. 4, 0395–0402 (2006).

84. India’s ongoing war against rabies. Bull. World Health Organ. 87, 885–964 (2009).

51 85. A. Markandya, T. Taylor, A. Longo, M. N. Murty, S. Murty, K. Dhavala, Counting the cost of vulture decline-An appraisal of the human health and other benefits of vultures in India. Ecol. Econ. 67, 194–204 (2008).

86. H. S. Young, R. Dirzo, K. M. Helgen, D. J. McCauley, S. A. Billeter, M. Y. Kosoy, L. M. Osikowicz, D. J. Salkeld, T. P. Young, K. Dittmar, Declines in large wildlife increase landscape-level prevalence of rodent-borne disease in Africa. Proc. Natl. Acad. Sci. 111, 7036–7041 (2014).

87. H. Young, D. McCauley, R. Dirzo, C. Nunn, M. Campana, B. Agwanda, E. Otarola-Castillo, R. Pringle, K. Veblen, D. Salkeld, K. Stewardson, R. Fleischer, E. Lambin, T. Palmer, K. Helgen, Interacting effects of land use and climate on rodent-borne pathogens in central Kenya. Philos. Trans. R. Soc. B Biol. Sci. (2017), doi:10.1098/rstb.2016.0116.

88. R. S. Ostfeld, T. Levi, F. Keesing, K. Oggenfuss, C. D. Canham, Tick-borne disease risk in a forest food web. Ecology. 99, 1562–1573 (2018).

89. Hofmeester Tim R., Jansen Patrick A., Wijnen Hendrikus J., Coipan Elena C., Fonville Manoj, Prins Herbert H. T., Sprong Hein, van Wieren Sipke E., Cascading effects of predator activity on tick-borne disease risk. Proc. R. Soc. B Biol. Sci. 284, 20170453 (2017).

90. T. Levi, A. M. Kilpatrick, M. Mangel, C. C. Wilmers, Deer, predators, and the emergence of Lyme disease. Proc. Natl. Acad. Sci. 109, 10942–10947 (2012).

91. R. S. Ostfeld, A. Price, V. L. Hornbostel, M. A. Benjamin, F. Keesing, Controlling Ticks and Tick-borne Zoonoses with Biological and Chemical Agents. BioScience. 56, 383–394 (2006).

92. G. Benelli, C. L. Jeffries, T. Walker, Biological Control of Mosquito Vectors: Past, Present, and Future. Insects. 7 (2016), doi:10.3390/insects7040052.

93. E. A. McGraw, S. L. O’Neill, Beyond insecticides: new thinking on an ancient problem. Nat. Rev. Microbiol. 11, 181–193 (2013).

94. M. B. Thomas, Biological control of human disease vectors: a perspective on challenges and opportunities. BioControl. 63, 61–69 (2018).

95. G. M. Patterson, Looking Backward, Looking Forward: The Long, Torturous Struggle with Mosquitoes. Insects. 7 (2016), doi:10.3390/insects7040056.

96. J. B. Gerberich, M. Laird, World Health Organization, “An annotated bibliography of papers relating to the control of mosquitos by the use of fish (Revised and enlarged to 1965) / by John B. Gerberich, Marshall Laird.” (World Health Organization, 1966), (available at https://apps.who.int/iris/handle/10665/65337).

97. H. A. Flores, S. L. O’Neill, Controlling vector-borne diseases by releasing modified mosquitoes. Nat. Rev. Microbiol. 16, 508–518 (2018).

52 98. D. P. Walshe, P. Garner, A. A. Adeel, G. H. Pyke, T. R. Burkot, Larvivorous fish for preventing malaria transmission. Cochrane Database Syst. Rev. (2017), doi:10.1002/14651858.CD008090.pub3.

99. S. Munga, N. Minakawa, G. Zhou, O.-O. J. Barrack, A. K. Githeko, G. Yan, Effects of Larval Competitors and Predators on Oviposition Site Selection of Anopheles gambiae Sensu Stricto. J. Med. Entomol. 43, 221–224 (2006).

100. K. Raghavendra, P. Sharma, A. P. Dash, Biological control of mosquito populations through frogs: Opportunities & constrains (2007).

101. I. Tabibzadeh, G. Behbehani, R. Nakhai, Use of Gambusia Fish in the Malaria Eradication Programme of Iran. Bull. World Health Organ. 43, 623–626 (1970).

102. T. H. M. and M. N. Nico, L., P. Fuller, G. Jacobs, M. Cannister, J. Larson, A. Fusaro, “Gambusia affinis” (Gainesville, FL, 2017).

103. Greater Los Angeles County Vector Control District, Mosquitofish. Gt. Los Angel. Cty. Vector Control Dist. (2016), (available at https://www.glacvcd.org/vector- information/mosquitofish/).

104. Biological control. World Health Organ. (2019), (available at http://www.who.int/denguecontrol/control_strategies/biological_control/en/).

105. A. Lazaro, W. W. Han, P. Manrique‐Saide, L. George, R. Velayudhan, J. Toledo, S. R. Ranzinger, O. Horstick, Community effectiveness of copepods for dengue vector control: systematic review. Trop. Med. Int. Health. 20, 685–706 (2015).

106. L. S. Tusting, J. Thwing, D. Sinclair, U. Fillinger, J. Gimnig, K. E. Bonner, C. Bottomley, S. W. Lindsay, Mosquito larval source management for controlling malaria. Cochrane Database Syst. Rev. (2013), doi:10.1002/14651858.CD008923.pub2.

107. G. F. Killeen, J. I. Githure, J. C. Beier, Short report: entomologic inoculation rates and Plasmodium falciparum malaria prevalence in Africa. Am. J. Trop. Med. Hyg. 61, 109–113 (1999).

108. M. Q. Benedict, A. S. Robinson, The first releases of transgenic mosquitoes: an argument for the sterile insect technique. Trends Parasitol. 19, 349–355 (2003).

109. K. V. Brown, Google Has a Plan to Eliminate Mosquitoes Around the World. Bloomberg (2018), (available at https://www.bloomberg.com/news/features/2018- 11-28/how-to-kill-mosquitoes-google-and-verily-have-a-plan).

110. Verily Life Sciences (2019), (available at https://verily.com/).

111. Eliminate Dengue - Our Challenge. World Mosq. Program (2019), (available at http://www.eliminatedengue.com/program).

53 112. B. Weiss, S. Aksoy, Microbiome influences on insect host vector competence. Trends Parasitol. 27, 514–522 (2011).

113. B. L. Phillips, G. P. Brown, M. Greenlees, J. K. Webb, R. Shine, Rapid expansion of the cane toad (Bufo marinus) invasion front in tropical Australia. Austral Ecol. 32, 169–176 (2007).

114. A. Savaya-Alkalay, G. A. De Leo, E. Aalto, T. Levy, O. Rosen, R. Manor, E. D. Aflalo, E. Tricarico, A. Sagi, The IAG gene in the invasive crayfish Procambarus clarkii – towards sex manipulations for biocontrol and aquaculture. Manag. Biol. Invasions (Accepted).

115. N. Shpak, R. Manor, E. D. Aflalo, A. Sagi, Three generations of cultured prawn without W chromosome. Aquaculture. 467, 41–48 (2017).

116. W. Rosa, Ed., in A New Era in Global Health (Springer Publishing Company, New York, NY, 2017; http://connect.springerpub.com/lookup/doi/10.1891/9780826190123.ap02).

117. C. N. Ngonghala, M. M. Pluciński, M. B. Murray, P. E. Farmer, C. B. Barrett, D. C. Keenan, M. H. Bonds, Poverty, Disease, and the Ecology of Complex Systems. PLoS Biol. 12, 1–9 (2014).

118. C. N. Ngonghala, G. A. De Leo, M. M. Pascual, D. C. Keenan, A. P. Dobson, M. H. Bonds, General ecological models for human subsistence, health and poverty. Nat. Ecol. Evol. 1, 1153–1159 (2017).

119. A. Savaya Alkalay, O. Rosen, S. H. Sokolow, Y. P. W. Faye, D. S. Faye, E. D. Aflalo, N. Jouanard, D. Zilberg, E. Huttinger, A. Sagi, The Prawn Macrobrachium vollenhovenii in the Senegal River Basin: Towards Sustainable Restocking of All- Male Populations for Biological Control of Schistosomiasis. PLoS Negl. Trop. Dis. 8 (2014), doi:10.1371/journal.pntd.0003060.

120. S. S. Chiotha, K. R. McKaye, J. R. Stauffer, Use of indigenous fishes to control schistosome snail vectors in Malaŵi, Africa. Biol. Control. 1, 316–319 (1991).

121. G. M. Mkoji, B. N. Mungai, D. K. Koech, E. S. Loker, Experimental control of the schistosome-transmitting snail Biomphalaria pfeifferi by the ampullariid snail Pila ovata. Ann. Trop. Med. Parasitol. 92, 65–72 (1998).

122. B. V Hofkin, D. K. Koech, J. Ouma, E. S. Loker, The North American Crayfish Procambarus clarkii and the Biologica Control of Schistosome-Transmitting Snails in Kenya: Laboratory and Field Investigations. Biol. Control. 1, 183–187 (1991).

123. M. J. Chimbari, Enhancing schistosomiasis control strategy for Zimbabwe: Building on past experiences. J. Parasitol. Res. 2012 (2012), , doi:10.1155/2012/353768.

54 124. G. I. Giraldo-Calderón, S. J. Emrich, R. M. MacCallum, G. Maslen, S. Emrich, F. Collins, E. Dialynas, P. Topalis, N. Ho, S. Gesing, G. Madey, F. H. Collins, D. Lawson, P. Kersey, J. Allen, M. Christensen, D. Hughes, G. Koscielny, N. Langridge, E. L. Gallego, K. Megy, D. Wilson, B. Gelbart, D. Emmert, S. Russo, P. Zhou, G. Christophides, A. Brockman, I. Kirmitzoglou, B. MacCallum, T. Tiirikka, K. Louis, V. Dritsou, E. Mitraka, M. Werner-Washburn, P. Baker, H. Platero, A. Aguilar, S. Bogol, D. Campbell, R. Carmichael, D. Cieslak, G. Davis, N. Konopinski, J. Nabrzyski, C. Reinking, A. Sheehan, S. Szakonyi, R. Wieck, VectorBase: An updated Bioinformatics Resource for invertebrate vectors and other organisms related with human diseases. Nucleic Acids Res. 43, D707–D713 (2015).

125. D. Duffy, R. Downer, C. Brinkley, The effectiveness of helmeted guineafowl in the control of the deer tick, the vector of Lyme disease. Wilson Bull. 104, 342–345 (1992).

126. K. M. Ng, J. A. Ferreyra, S. K. Higginbottom, J. B. Lynch, P. C. Kashyap, S. Gopinath, N. Naidu, B. Choudhury, B. C. Weimer, D. M. Monack, J. L. Sonnenburg, Microbiota-liberated host sugars facilitate post-antibiotic expansion of enteric pathogens. Nature. 502, 96–99 (2013).

127. D. I. Tsitsigiannis, M. Dimakopoulou, P. P. Antoniou, E. C. Tjamos, Biological control strategies of mycotoxigenic fungi and associated mycotoxins in Mediterranean basin crops. Phytopathol. Mediterr. 51, 158–174 (2012).

128. J. M. Wagacha, J. W. Muthomi, Mycotoxin problem in Africa: Current status, implications to food safety and health and possible management strategies. Int. J. Food Microbiol. 124, 1–12 (2008).

129. D. Duval, R. Galinier, G. Mouahid, E. Toulza, J. F. Allienne, J. Portela, C. Calvayrac, A. Rognon, N. Arancibia, G. Mitta, A. Théron, B. Gourbal, A Novel Bacterial Pathogen of Biomphalaria glabrata: A Potential Weapon for Schistosomiasis Control? PLoS Negl. Trop. Dis. 9 (2015), doi:10.1371/journal.pntd.0003489.

130. J. P. Pointier, A. Theron, D. Imbert-Establet, G. Borel, Eradication of a Sylvatic Focus of Schistosoma mansoni Using Biological Control by Competitor Snails. Biol. CONTROL. 1, 244–247 (1991).

131. F. Frandsen, N. O. Christensen, Effect of Helisoma duryi on the survival, growth, and cercarial production of Schistosoma mansoni infected Biomphalaria glabrata. Bull. World Health Organ. 55, 577–580 (1977).

132. J. Gómez Pérez, M. Vargas, E. A. Malek, Displacement of Biomphalaria glabrata by Thiara granifera under natural conditions in the Dominican Republic. Mem. Inst. Oswaldo Cruz. 86, 341–347 (1991).

55 133. H. K. Lim, D. Heyneman, Intramolluscan Inter-trematode Antagonism: A Review of Factors Influencing the Host-parasite System and its Possible Role in Biological Control. Adv. Parasitol. 10, 191–268 (1972).

134. W. B. ROWAN, Daily periodicity of Schistosoma mansoni cercariae in Puerto Rican waters. Am. J. Trop. Med. Hyg. 7, 374–381 (1958).

135. W. B. Knight, L. S. Ritchie, F. Liard, J. Chiriboga, Cercariophagic activity of guppy fish (Lebistes reticulatus) detected by cercariae labeled with radioselenium (75Se). Am. J. Trop. Med. Hyg. 19, 620–625 (1970).

136. N. Christensen, Schistosoma mansoni: Interference with cercarial host-finding by various aquatic organisms. J. Helminthol. 53, 7–14 (1979).

137. P. Nansen, F. Frandsen, The interaction of some environmental factors influencing Schistosoma mansoni cercarial host-finding. J. Helminthol. 54, 203–205 (1980).

138. G. Bowatte, P. Perera, G. Senevirathne, S. Meegaskumbura, M. Meegaskumbura, Tadpoles as dengue mosquito (Aedes aegypti) egg predators. Biol. Control. 67, 469–474 (2013).

139. A. Munoz-Pedreros, C. Gil, J. Yanez, J. R. Rau, Raptor habitat management and its implication on the biological control of the Hantavirus. Eur. J. Wildl. Res. 56, 703– 715 (2010).

56 Figures

Figure 1. Natural enemies of human infectious diseases in the environment. Many human infectious disease agents (outlined in orange) spend a significant portion of their life- cycle in the environment, as free-living pathogens and parasites, and in vectors, intermediate hosts, and non-human reservoir hosts. Infectious stages in the environment are subject to natural regulation by predators, competitors, pathogens, and hyperparasites (outlined in green). (A) Larvivorous fish have long been used for the control of mosquito vectors; several known predators are edible and offer dual benefits of disease control and aquaculture for nutrition or income. (B) Schistosomiasis intermediate host snails can be regulated by a variety of molluscivorous predators, including freshwater prawns, crayfish, fish, and ducks (14, 81, 120–124). (C) Phage predation on V. cholerae in aquatic reservoirs may naturally interrupt epidemics, supporting investigation towards phage- based environmental interventions (34, 36). (D) Daphnia spp. freshwater crustaceans are efficient and voracious grazers of aquatic bacteria, strongly influencing bacterial communities; studies show that Daphnia graze Escheria coli and Campylobacter jejuni (47, 48) (E) Vultures consume carcasses that can be important breeding grounds for diverse pathogens, including anthrax, brucellosis, and tuberculosis (84, 86). (F) Copepods are voracious predators of larval Aedes mosquito vectors and have been used to control dengue virus. (G) Beauveria bassiana is a ubiquitous and commercially available fungus that induces mortality in mosquitoes, ticks (pictured), and other vectors.

Figure 2. (A) The number of countries reporting mosquito vector resistance to one or more major classes of insecticides used globally in the fight against mosquito-borne diseases. Data was obtained from the Vectorbase database(125); resistance data was filtered to include resistance reported as percent mortality, and resistance was conservatively defined as mosquito mortality less than 95%. (B) Natural predators of mosquito larvae and adults (green) are widespread, and include various species of bats, arthropods, birds, amphibians, fish, and crustaceans (copepods) (see also Table 1); currently. Attempts are being made to alter wild mosquito populations with naturally occurring strains of Wolbachia bacteria, which can (i) inhibit development of arboviruses in Aedes spp. and, potentially, malaria in Anopheles spp., or (ii) lead to sterility when artificially infected males mate with wild, uninfected females (see text and Table 1 for details).

58 Table 1. Select case studies and observations on natural enemies of human pathogens.

Type Natural enemy Human disease Evidence and status Reference(s)

Probiotics Infectious When administered in adequate amounts, a review of randomized control trials suggest that probiotics can shorten the (61, 63) diarrhea duration of acute diarrhea in children under age 5, and Clostridium difficile-associated diarrhea in high-risk groups of

adults and children.

Probiotics Upper respiratory A review of randomized control trials suggest that probiotic administration may reduce the severity of acute upper (62) infections respiratory tract infections and alleviate the reliance on antibiotics to contain infections. Bacteriophage Infectious In the 20th century phages were used as standard treatment for diarrhea (including Shigella and typhoid); they are (64, 66) thehuman body diarrhea currently being investigated as an alternative therapy to antibiotics (Shigella, Natural enemies applied to to applied enemies Natural Typhoid) Pathogenic bacteria Onchocerciasis Soon after its discovery, the Onchocerciasis Control Programme (OCP) of West Africa used commercially available (22) Bacillus thuringiensis subsp. israelensis to successfully control insecticide-resistant Simulium spp. blackfly larvae in rivers as part of long-term integrated disease control; the OCP reduced onchocerciasis in 16 participating countries and represents one of the most successful large-scale control programs to date. The Simulium Control Program in Brazil has also successfully used Bti for blackfly control. Pathogenic bacteria Malaria and Bacillus thuringiensis (Bti) and Bacillus sphaericus (Bs) have been commercially developed for mosquito control and (22–24) arboviruses implemented worldwide to control a diverse array of mosquitoes (incl. Anopheles spp., Culex spp., Aedes spp.) and associated diseases, with minimal to no impact on non-target organisms; however, resistance to the microbial larvicide has been documented. Bacteriophage Food-borne Bacteriophages have been phages have been used successfully in food supply systems for rapid detection or (71) pathogens elimination of food-borne pathogens like Salmonella, Campylobacter, Listeria, Staphylococcus, and Vibrio spp.

Entomopathogenic Various disease Metarhizium brunneum and Baeuveria bassiana are two commercially available entomopathogenic fungi that occur (25–31) fungi vectors naturally in soil and can induce high mortality in adult Anopheles spp. mosquitoes (malaria vectors), Ixodes spp. black- (mosquitoes, legged ticks (Lyme disease vectors), and larval Phlebotomus sand flies (leishmaniasis vectors) ticks, sandflies) Molluscivorous fish Fish-borne Aquaculture nursery ponds can be intense transmission sites for fish-borne trematodes from snails to fish, and (73–75). trematodiases subsequently to humans; field trials show that stocking ponds with juvenile black carp, an omnivorous predator, may reduce snail densities and limit transmission, thus reducing human risk for fish-borne trematodiases. Various snail Schistosomiasis Schistosomiasis intermediate host snails can be regulated by a variety of molluscivorous predators; interventions and (14, 80, 119– predators: field trials involving snail control by freshwater prawns, crayfish, fish, and ducks have been successful to varying 123)

Natural enemies applied to the environment the to applied enemies Natural crustaceans, fish, degrees. Certain species, including Macrobrachium freshwater prawns, are widely used in aquaculture, providing a fowl potential benefit of income and protein sources to communities in endemic areas. Larvivorous fish Malaria and Aquaculture of edible fish in rice fields might simultaneously reduce mosquito vector abundance, provide fish protein (76, 77) arboviruses for nutrition and commerce, and improve rice yields; evidence derives from field studies in southern India and China, where black carp were stocked in rice paddies to control, with increased rice yields during field experiments in both regions and a correlation (unestablished causation) with reduced malaria transmission after many years of stocking fish in China.

59 Larvivorous fish Malaria Mosquitofish (Gambusia affinis) can establish self-sustaining populations and consume large numbers of mosquito (98, 101) larvae, and are credited with assisting in the eradication of malaria (transmitted by Anopheline mosquitoes) in numerous locations. For this reason, they have been introduced globally from their native ranges in southern and eastern North America for more than 100 years to control mosquitoes. However, a recent Cochrane Review suggests that evidence for a causal relationship linking introductions of mosquitofish, adult Anopheles populations, and human malaria incidence is lacking. Predatory copepod Dengue A series of promising studies carried out over several years in Vietnam showed that Mesocyclops copepods introduced (105) to Aedes mosquito breeding containers can reduce larval and adult mosquito density and dengue seroprevalence in humans. Wolbachia Arboviruses Strains of the insect endosymbiont, Wolbachia pipientis, can cause sterility in mosquitoes when artificially infected (97, 109, endosymbiont males mate with wild, uninfected females; this sterile insect technique is currently intensely researched and, since 110) 2016, deployed for large-scale testing in California, overseen by Alphabet Inc.’s Verily Life Sciences.

Wolbachia Malaria and Vertically transmissible strains of the insect endosymbiont, Wolbachia pipientis, can interfere with arbovirus and, to a (97, 111, endosymbiont arboviruses less clear degree, malaria development in infected female mosquitoes; efforts to replace wild mosquito populations 112) with Wolbachia-infected populations is underway across the world to combat viruses transmitted by Aedes mosquitoes, including dengue, Zika, and chikungunya.

Vulture populations Mixed zoonotic Vultures compete with pathogens for host tissue (intraguild predation). In India, declines in vulture populations has led (83, 85) pathogens to increased volume of uneaten livestock carcasses, important breeding grounds for diverse human pathogens including anthrax. Carcass consumption by vultures also limit livestock parasites and pathogens including brucellosis and tuberculosis. Vulture populations Rabies Declines in vulture populations correlate with increases in feral dog populations in India, likely as competition for (85) carrion eased; over 48,000 rabies-associated deaths from dog bites occurred over a time period during which vultures were almost entirely lost, suggesting that vulture conservation and breeding management could indirectly reduce feral dog populations. Various large-bodied Rodent-borne In eastern Africa, experimental evidence suggests that conservation of large-bodied herbivores (e.g., zebra, giraffe, (86) herbivores disease elephant, and gazelle) which compete with rodents for food and keep rodent populations low, can indirectly control

human disease risk. Systematic increases were observed in the density of small-mammals (i.e., rodents) infected with zoonotic pathogens (Bartonella spp. transmitted by flea vectors) in cropland and in experimentally defaunated areas, as compared to paired control areas with abundant wild herbivores Various rodent Tick-borne Small predators (e.g., red fox) may regulate populations of rodents and small mammals that serve as tick hosts and (89, 90) predators disease reservoirs of tick-borne diseases, including Lyme disease. management

Various tick Lyme disease Helmeted guineafowl have been shown to reduce adult blacklegged tick density in small experimental plots in (91, 125) predators northeastern United States, supporting anecdotal evidence that guineafowl are effective for control of ticks and Lyme disease. Other avian tick predators include wild turkeys and oxpeckers.

Various snail Schistosomiasis Outbreaks of schistosomiasis have been linked to loss of important snail predators (fish, crustaceans) due to (81) predators overharvesting and land-use change; for example, dams block essential migration paths of Macrobrachium freshwater

Conservation of natural enemies: species, habitat, and ecosystem ecosystem and habitat, species, enemies: natural of Conservation prawns, voracious snail predators, and reversing this population could enhance biodiversity conservation and preserve ecosystem services including disease control and food and livelihood provision.

60 Commensal gut Enteric pathogens Mouse model studies suggest that undisturbed communities of commensal microbes in our intestines can limit (126) microbes colonization by pathogens like Salmonella enterica serovar Typhimurium and Clostridium difficile via resource competition and the production of antimicrobial compounds. Antagonistic bacteria Foodborne Biological control of fungal mycotoxoses acquired through food (i.e., Fusarium spp. and Aspergillus spp.) by (127, 128) disease antagonistic bacteria and yeast isolates, and competing fungus, has been shown in laboratory and field trials. Dung beetle Soil-transmitted Observational and manipulative field studies involving livestock suggest that dung beetles reduce abundance of Summarized helminthes, infective helminthes and protozoa in soil; other studies show that beetles are capable of passing large quantities of by (45) enteric pathogens human feces, which could theoretically reduce the abundance of enteric pathogens in the environment. Freshwater Waterborne Daphnia spp. freshwater crustaceans are efficient and voracious grazers of aquatic bacteria, strongly influencing (46, 47) crustaceans enteric pathogens bacterial communities; studies show that Daphnia graze Escheria coli and Campylobacter jejuni. Bacteriophage Cholera Observational evidence suggests that phages can reduce or limit Vibrio cholerae outbreaks in natural aquatic (35, 72) environments; however, this has not been tested in a clinically or epidemiologically relevant setting, and the speed at which target cholera strains might evolve resistance under phages’ selective pressure is unknown. Bacteriophage Shigella A virulent bacteriophage was isolated from water contaminated with Shigella in North Korea; the bacteriophage shows (34)

lytic activity with potential for biocontrol of antibiotic-resistant Shigella flexneri and S. sonnei in contaminated water. Bacterial pathogen Schistosomiasis A novel microbial pathogen, Candidatus Paenibacillus glabratella, was isolated from an infected laboratory population (129) of Biomphalaria glabrata snails (likely horizontally infected from field collected snails); laboratory experiments show massive mortality (nearly 90%) of snails through horizontal transmission and 20x reduced hatching of exposed eggs. Nematophagous Helminthiases Laboratory experiments have shown strong predatory action of nematophagous fungi on helminths, including taeniasis (15–18) fungi (tapeworm), roundworm, pinworm, hookworm, schistosomiasis, and fascioliasis. Competitor snails Schistosomiasis Field and laboratory studies show that competitor snails can reduce schistosome-transmitting snail growth and (130–132) reproduction long-term, and can act as 'decoy hosts', reducing the number of free-swimming infective stages via absorption by non-competent hosts; however, use of beneficial invaders must carefully consider ecological effects and other parasites carried by competitor snails. Competitor Schistosomiasis Exclusion of Schistosoma parasite infections establishing in compatible snail hosts can occur via competition for host (38, 40, 41, trematodes tissue with antagonistic non-human parasites; laboratory studies show that augmenting the diversity of non-human 133) trematodes could reduce schistosome infections in snails.

Other underexplored natural enemies natural underexplored Other Cercarial predators Schistosomiasis Free-swimming stages of schistosoma parasites (cercariae) are susceptible to predation by aquatic predators; (134–137) invertebrates and fish have been shown to prey directly on schistosome larvae.

Predatory mosquito Malaria and Strong predatory effect of a natural intraguild Culex spp. mosquito predator on Aedes, Anopheles, and Culex disease (20) arboviruses vectors (dengue, malaria, and Japanese encephalitis) has been observed in semi-field conditions in Sri Lanka.

Various predators Malaria Laboratory studies showed that Anopheles gambiae mosquitoes laid fewer eggs in breeding containers conditioned (99) and competitors with various predators or competitors than in control containers, revealing that community structure may impact oviposition attempts. Amphibians Dengue Strong predatory effect of tadpoles from 5 widely distributed frog genera observed on Aedes aegypti mosquito eggs; (138) the frog species distribution among ephemeral and human habitats reflects the highly adaptable Aedes aegypti distribution, suggesting that biological control of mosquitoes by amphibians may be ubiquitous. Amphibians Malaria and Frogs have been broadly observed to prey on various mosquito species, including Anopheles mosquito vectors and e.g., (100) arboviruses Aedes arbovirus vectors.

Raptor populations Rodent-borne Raptor populations around the world have, in general, been negatively impacted by habitat loss and human impacts; (139) disease conservation and augmentation of species known to be important rodent predators could lead to reduced incidence of Hantavirus and other human diseases with rodent reservoirs.

62 Chapter 3

Divergent ecological drivers and spatial scales of risk for S. haematobium and

S. mansoni in a co-endemic landscape

Isabel J. Jones1, Andrew J. Chamberlin1, Skylar Hopkins2,3, Nicolas Jouanard4, Armand

M. Kuris5, Kevin D. Lafferty6, Andrea J. Lund7, Raphaël Ndione4, Gilles Riveau4, Simon

Senghor4, Anne-Marie Schacht4, Chelsea L. Wood8, Susanne H. Sokolow1,9, and Giulio

A. De Leo1,9

1Hopkins Marine Station, Stanford University, Pacific Grove, CA, USA

2National Center for Ecological Analysis and Synthesis, Santa Barbara, CA, USA

3Deptartment of Applied Ecology, North Carolina State University, Raleigh, NC, USA

4Biomedical Research Center EPLS, Saint Louis, Senegal

5Department of Ecology, Evolution, and Marine Biology, University of California, Santa Barbara, CA,

USA

6Western Ecological Research Center, United States Geological Survey

7Emmett Interdisciplinary Program in Environment and Resources, Stanford University, Stanford, CA,

USA

8School of Aquatic and Fishery Sciences, University of Washington, Seattle, WA, USA

9Stanford Woods Institute for the Environment, Stanford University, Stanford, CA, USA

*Corresponding author: Isabel J. Jones; [email protected]

Keywords: schistosomiasis, infectious disease control, neglected tropical disease, disease ecology, parasite ecology, Schistosoma haematobium, Schistosoma mansoni, snail control, transmission control

63 Abstract

Schistosomiasis affects more than 200 million people annually, and disease can be caused by more than one species of schistosome parasite at the same time. Risk for acquiring infection is determined by the distribution of specific freshwater snail species, which serve as obligate intermediate hosts for specific parasites. To curb the global burden of disease, the World Health Organization has recently updated its strategy to include snail control in endemic areas. However, knowing when and where to target snails for cost- effective control requires a better understanding of the spatial distribution of human risk in complex freshwater landscapes, especially where multiple schistosomes and snail host species co-occur. Here we assess ecological associations and spatial scales of schistosomiasis risk in the Senegal River Basin, a region hyperendemic for both Schistosoma haematobium and S. mansoni. We use fine-scale field data on the distribution of the parasites’ unique snail hosts to validate and scale a drone-based assessment of disease risk across increasing spatial scales in discrete water-contact sites adjacent to 16 villages, where disease risk was measured as annual re-infection rates and egg burden accumulation in >1400 local school-aged children. Our findings suggest divergent ecological associations and spatial scales of S. haematobium and S. mansoni risk. We found that S. haematobium infection risk correlated with the local extent of snail habitat in water access sites, and that infection risk integrated across a large spatial scale. S. mansoni risk was associated with a small scale of risk and specific morphological features of water contact that may stabilize water flow. Findings improve our potential to rapidly identify schistosomiasis transmission foci, and to design environmental interventions that best reflect the unique biology and transmission dynamics of specific schistosome species and their snail hosts.

64 INTRODUCTION Schistosomiasis affects more than 200 million people in sub-Saharan Africa (1), and a substantial number of cases may be caused by infection with more than one species of schistosome parasite. Indeed, the geographic distributions of Schistosoma haematobium and S. mansoni, the major causes of urogenital and intestinal schistosomiasis, respectively, frequently overlap (2). Schistosomes infect humans through a complex life cycle involving freshwater snails, and schistosome–snail species pairs are tightly co-evolved. Therefore, dual risk for waterborne infection by S. haematobium and S. mansoni, or by other less common schistosome species, is determined by the spatial distributions of their intermediate freshwater snail hosts. Fortunately, a single drug – praziquantel – is able to kill most mature infections of any schistosome species in humans, and mass distribution of praziquantel to at-risk populations in endemic areas has been the backbone of global schistosomiasis control strategies since 2001 (3), reaching more than 60% of school-aged children requiring treatment in 2018 (1). This one-size-fits-all-schistosomes drug treatment has proven critical to reducing the burden of chronic and heavy infections in humans. But rates of re-infection and new infection remain high, and the total number of people infected today is little changed from levels observed nearly two decades ago (1, 4). Disconcertingly, large-scale mass drug administration (MDA) trials across several endemic countries have revealed the existence of “persistent hotspots”, or transmission settings where burden remains high even after several years of MDA; in such high-transmission settings, unaddressed sources of transmission risk in the environment may drive disease persistence (5, 6, 6, 7). Recognizing that MDA alone is not a silver bullet for schistosomiasis elimination, in 2012 the World Health Assembly updated recommendations for schistosomiasis control, suggesting that environmental intervention should accompany MDA in endemic areas (8). In 2017, the Assembly further urged member states to develop strategic plans to implement snail control interventions alongside MDA where possible (9), reembracing a strategy that had been successful in many parts of the world before praziquantel became widely available (10, 11). However, whereas preventative chemotherapy relies on a solution – praziquantel – that is effective against adult worms of all major human schistosome species, environmental control strategies may need to be adapted to the

65 unique ecologies of different schistosomes and their intermediate snail hosts (12). And in co-endemic settings, specific environmental interventions may have differential outcomes on the control of co-occurring schistosome species if differences in snail and parasite ecology are not accounted for. Therefore, here we present novel insights into how snail and parasite ecology translates to human disease risk by asking a deceivingly simple question: Where are the snails that infect people? We explore two aspects of schistosome transmission ecology essential to effectively monitor human risk and target environmental interventions: ecological correlates of village-level human disease outcomes, and spatial scales of transmission risk in local aquatic environments. Environmental transmission risk has often been assessed at two very different scales: (i) at the snail’s scale, through monitoring of snail presence/absence and infection prevalence in local water bodies near at-risk communities (13), and (ii) at broad geographic scales, using remote-sensed environmental data to map endemic areas at regional or national scales (14, 15). At the snail’s scale, the biology and ecology of schistosomes and their intermediate host snails have been extensively studied over many decades. However, the presence of snails in the environment often fails to correlate with local infection outcomes in humans (16), obscuring efforts to predict where and when infection will occur. This could reflect a failure to monitor snails at a scale relevant to long-term human disease outcomes – snails can be extensively distributed and found in diverse habitat types across endemic areas (9), and snails that contribute to human disease can move across large spatial scales overtime. Risk mapping at broad geographic scales, on the hand, can help governments and health agencies allocate disease control resources to at-risk populations (17), but are often too course to identify environmental transmission foci where infection occurs and environmental interventions need to be targeted. Better understanding common characteristics related to transmission risk at a scale relevant to snail-to-human transmission could help to ensure that environmental interventions (or water-contact behavior modification) are implemented at appropriate scales to be most cost-effective, comprehensive, and incur no undue harm to the environment, especially in settings where resources are limited. To better understand and monitor transmission potential in the environment, we sought to address the following questions in a region of northern Senegal hyperendemic for both S. haematobium and S. mansoni: (1) do Bulinus spp. snails, the intermediate host

66 for S. haematobium, and Biomphalaria pfeifferi snails, the intermediate host for S. mansoni, have different habitat associations in the freshwater ecosystems where they co- occur? (2) Are human infection burdens for S. haematobium and S. mansoni associated with the same or different habitat features? And, (3) is the spatial scale at which habitat features are associated with human infection burden different for infections with S. haematobium versus S. mansoni (suggesting similar or different spatial scales of transmission for the two species)? Our study builds on previous research from this group which found that estimates of remotely sensed Bulinus spp. snail habitat are better at predicting S. haematobium risk than are manual snail counts (16). However, we were unable to make similar associations for S. mansoni risk using snail counts or remote-sensed snail habitat, which led us to suspect that risk for the co-occurring species might respond to different ecological predictors. Furthermore, while our prior findings suggested that precision mapping of snail habitat near villages, which can be implemented relatively easily and inexpensively using unmanned aerial vehicles (drones) or high-resolution satellite imagery, could in the future be used to identify transmission foci, we still do not know what the spatial scale of transmission risk is. In other words, we do not know whether long-term human transmission risk from the environment is restricted to nearshore areas where human activity occurs, or across larger spatial scales where snails move and reproduce over time. To address our questions, we used fine-scale quantitative data on snail–habitat associations to validate a drone-based assessment of the ecological correlates and spatial scales of S. haematobium and S. mansoni infection risk to humans. This longitudinal study focuses on the associations between human infection patterns in >1400 children at 16 villages in the lower Senegal River Basin, and the associated ecological features of village-level water-contact sites. Using a model comparison approach to assess risk while considering the (i) unique ecological features of water-contact sites and (ii) the distribution of snail habitat measured at scales varying from a 1–120m vicinity of shorelines, we found differential ecological correlates and spatial scales of risk for each species. Our findings suggest that specific ecological features and scales of transmission risk can be identified for S. haematobium and S. mansoni, but with divergent outcomes. Elucidating such differences is critical to ensure that disease control measures aiming to

67 interrupt transmission from snails to humans are not thwarted by a one-size-fits-all approach, but rather recognize and embrace the complexity of schistosome transmission ecology in multi-parasite, multi-host systems. Further, the methodological advances presented here can be used evaluate and scale new tools to identify schistosomiasis foci in dynamic landscapes using relatively inexpensive technology, facilitating the design of targeted disease control interventions.

METHODS 1. Study villages All parasitological data collection, ecological field sampling, and overhead unmanned aerial vehicle (drone) imaging took place between 2016 and 2018 at 16 villages in the Senegal River Basin of northwestern Senegal (Fig. 1). Data were collected as part of a longitudinal cohort study on regional schistosomiasis transmission ecology and biological control, and the criteria for village selection from among 701 candidate villages was described previously (16). Notably, all leaders of selected villages reported non-zero baseline S. haematobium and S. mansoni prevalence in 2016 and all villages were located in close proximity to freshwater and contained between one and four distinct water access sites where daily water-contact activities, like collecting water, washing, animal care, swimming, and fishing, took place. One village was located on the Senegal River, four on the nearby Lampsar River, one on the N’Galam River, and ten on a large lake, Lac de Guiers (Fig. 1); population sizes ranged from 400 to 1,852 people. Because this study was conducted while a parallel manipulative experiment was running, some village-time combinations were excluded from the snail–habitat association analysis in 2017-2018 (described below). And, because of limited drone ecological data available for some villages in 2017, some village-year combinations were excluded from analysis on ecological associations with human infection patterns (described below).

2. Parasitological surveys At baseline in 2016, nearly 1400 school-aged children across 16 villages were enrolled in the study after informed consent from participants and their parent or guardian was obtained. In 2016, all children were screened for S. haematobium and S. mansoni infection presence and egg burden according to standard parasitological sampling

68 protocols (described in (18) for S. haematobium and in (19) for S. mansoni). About 50% of participants were female. Infected individuals were offered treatment to clear infection (40 mg/kg praziquantel). The process was repeated in 2017 and again in 2018, resulting in two follow-up years for which annual re-infection rates and egg burden was recorded for both schistosome species. Because some high resolution satellite or drone imagery was unavailable across all spatial scales assessed (up to 120m from shorelines) in some villages in 2017, here we assessed re-infection patterns in a subset 1046 children across 12 villages in 2017, and 1220 children across 16 villages in 2018. All sample collection and processing was conducted by the Biomedical Research Center Espoir Pour La Santé in Saint-Louis, Senegal. The study received approval from the National Committee of Ethics for Health Research from the Republic of Senegal (protocol no. SEN14/33) as well as the Institutional Review Boards of Stanford University (protocol no. 32196) and the University of California, Santa Barbara (protocol no. 19-170676).

3. Field surveys on snail abundance and habitat associations in water access sites Between May 2016 and February 2018, we conducted seasonal snail sampling at all 32 water access sites distributed across the 16 study villages. We sampled three times per year, once in each of the three dominant climatic seasons of the Sahel: dry, warm spring; wet, hot summer; and dry, cool winter. Some villages are not included in this analysis, to exclude potential effects of an ongoing manipulative field experiment carried out in a subset of villages; 14 villages (30 water access sites) were sampled in spring, and 11 villages (24 water access sites) were sampled in summer and winter. In the Senegal River Basin, the two schistosome species use different snail species as obligate intermediate hosts: S. haematobium (and S. haematobium/S. bovis hybrid species (20)) use Bulinus globosus and B. truncatus, while S. mansoni uses Biomphalaria pfeifferi (21). Vegetation (or lack thereof) in the aquatic ecosystems where snails are found can be characterized by three general types: non-emergent vegetation (Ceratophyllum spp., Ludwigia spp., Potamogeton spp., others), emergent vegetation (predominately Typha spp. and Phragmites spp.), and water/mud. Bulinus spp. snail density differs between these habitat types (16), and so here we wanted to estimate and compare the density of both Bulinus and Biomphalaria snails in each habitat type separately. During each field visit, we exhaustively sampled for snails within quadrats (76.2-cm length × 48.26-cm

69 width × 48.26-cm height) at each water access site, where sampling locations were randomly stratified across the three habitat types in proportion to the sampling area each habitat type covered within that water access site. The proportional coverage of dominant habitat types (and the number of quadrats to be sampled in each type) at each site was predetermined using satellite or drone imagery collected during a previous site visit, and verified by two independent technicians upon arrival. Before arriving at a site, we used ArcGIS to distribute sequentially numbered random points across each site area and loaded their locations onto a GPS device; in the field, we then chose sampling locations by navigating to those points in numeric order, until the required number of points per habitat type was sampled. At each quadrat, we recorded water depth, all vegetation to family or genus, vegetation biomass, and counts for all schistosome competent and non-competent snail species recovered. Details on the full methods, including methods on pre-determination of habitat coverage and sampling stratification, can be found in (though described with reference only to Bulinus spp. snails, Biomphalaria pfeifferi snails were sampled concurrently). All snails recovered in each quadrat were labeled by quadrat ID and later screened for schistosome infection via live parasite shedding and dissection in the laboratory.

4. Field surveys on snail abundance in offshore habitats To ensure safety for all field technicians, quantitative field surveys on snail density and habitat associations described above were restricted to water depths <100cm, which typically limited sampling to areas within 10m from shore. Previous efforts from this research group aimed to correlate quantitative snail surveys with human infection outcomes, with inconsistent outcomes – snail counts did correlate with S. haematobium infection probability (16), but remote-sensed area of snail habitat (non-emergent vegetation) was a better predictor. Meanwhile, overhead drone imagery of village water access sites and surrounding aquatic ecosystems revealed (i) vast areas of snail habitat beyond the areas where it was safe for technicians to sample, and (ii) unique morphological characteristics of water access sites that could influence transmission, like water access site size and the degree of enclosure. So, we set out to test whether we could detect the scale at which remote-sensed snail habitat best correlates to human infection

70 outcomes, and if certain morphological characteristics influence infection outcomes (described in the next sections). Before quantifying snail habitat (non-emergent vegetation) across large scales, we wanted to confirm the presence of intermediate host snails in offshore habitats. In July-August 2018, we conducted quantitative snail surveys at one village on the Senegal River (two water access sites), two villages on the Lampsar River (three water access sites), and one village on Lac de Guiers (one water access site) (Fig. 1a). Prior to sampling, we used ArcGIS to generate three sampling polygons of varying distance from the shorelines at each water access site: (1) a nearshore sampling polygon using the predefined sampling polygons from the previous sampling efforts described above; (2) a medium-distance sampling polygon extending 75m from the outside of the near-shore polygon; and (3) a far-distance sampling polygon extending 40-132m from the outer edge of the medium polygon. Where sites were nearby, far-distance polygons needed to be adjusted to avoid overlap; this is the reason for variability in far-distance polygons across sites. Random points (latitude/longitude) were generated and distributed within each polygons. In the field, we sampled five points (following numeric order of the randomly distributed points) in nearshore areas (within 30m) using the same protocols described in the previous section. In the medium-distance and far-distance polygons, we used a boat to sample 10 points per polygon per water access site, and recorded water depth, wind speed, surface water flow, vegetation species, vegetation mass, and snail count. Snails were later screened for infection in the laboratory as described above.

5. Satellite- and drone-based habitat mapping and quantification Beginning in January 2017, we introduced drones to field sampling. A DJI Phantom 4 quadcopter equipped with a 1/2.3 inch CMOS sensor and 12.4-megapixel camera was used to map each water access site at each snail sampling period. To map the freshwater ecosystems within and outside water access sites, we flew the drone in an overlapping grid at a fixed elevation, assuring 60-80% image overlap. We later merged the images and generated a high-resolution, georeferenced orthomosaic geotiff file using the open source software OpenDroneMap. Drone imagery data were then imported into the image analysis software eCognition (Trimble Inc., Sunnyvale, CA, USA), and we classified the

71 imagery into land, water/mud, non-emergent floating/submerged vegetation, and emergent vegetation (Fig. 1b,c) using object-based image analysis (OBIA). Using this approach, pixels are first grouped together into “objects” based on a heterogeneity score derived from parameters of scale, shape, compactness, and weights assigned to each image band (red, green, blue) (22). Before classification, each image was segmented into objects that were small enough in size to distinguish between the functional classes under study using the multiresolution segmentation algorithm in eCognition. We then visually labeled a subset of objects to use as training samples for machine learning (ML) classifiers. Several ML classifiers were tested (Bayes, SVM, Decision Tree, Random Trees, and KNN) and their respective effectiveness was assessed visually. For most images, the k-nearest neighbor (KNN) algorithm produced the most accurate results. Some image distortion occurred because of reflected sunlight and required manual correction of some objects using additional drone data and field sampling data as validation. As the drone was not a part of our field protocols during the first two sampling periods in spring and summer 2016, we used high-resolution satellite imagery to measure snail habitat in 2016. DigitalGlobe Foundation (now Maxar Technologies) provided access to WorldView-2 (WV-2) imagery of our study sites from June 29, 2016 (15 lake sites) and July 4th 2016 (9 river sites), in between our spring and summer field sampling periods. WV-2 multispectral (MS) imagery consists of 8 spectral bands at 2m spatial resolution including near-infrared, which improves image classification accuracy between biological and physical surface features (23). We pansharpened the MS imagery using the supplied 0.5m resolution panchromatic data and ArcGIS Pro to generate 0.5m resolution MS imagery, which was possible to visually interpret into the 4 classifications by the eCognition image analyst using the same methods described above for drone imagery.

6. Quantifying snail habitat across scales Using ArcGIS Pro, we created a digitized shoreline for each water access site using the boundary between land and water-access area; shorelines extended until they reached the emergent vegetation that forms the edges of water access sites. From the shoreline midpoint, multiple circular, overlapping disk buffers were generated at 15m increments

72 up to 120m (Fig. 1b,c). Where buffer zones spatially overlapped with adjacent water access sites, the overlapping areas were split and equally distributed between the two sites so that no buffer areas overlapped. We then overlaid the multiple disk buffers on the classified satellite and drone imagery obtained during July-August field campaigns and measured the exact amount of non-emergent vegetation integrated across the 1m, 5m, and each 15m buffer distance. The timing at which we measured non-emergent vegetation corresponds to peak transmission season for schistosomiasis in this region and directly follows peak vegetation and snail population growth in the spring (Sturrock et al., 2001).

7. Characterizing site morphology We hypothesized that morphological characteristics of each water point might impact schistosome transmission efficiency. Specifically, we suspected that the degree of site circumscription, or whether or not a discrete water access site was nearly or fully enclosed by emergent vegetation and therefore largely disconnected from the main channel of a river or open expanses of the lake, might influence snail density and larval parasite transport. We also suspected that water access site size – proxied as the linear length of a water access site’s shoreline – might influence human–water contact intensity (i.e., larger sites are used more intensely than smaller ones) and transmission efficiency (i.e., smaller sites may have a higher concentration of snails or free-swimming larvae). Last, based on consensus of personal observations that non-emergent habitat tends to accumulate along the emergent “walls” of water access sites, we suspected that the amount of this edge habitat (linear length of sites from shore to open water) could also influence snail density. To test these hypotheses, we used overhead drone imagery obtained in July-August 2018 to record: (i) site circumscription (yes versus no), (ii) linear length of the shorelines, and (iii) linear length of edge habitat at each water access site. We also measured the distance between the edges of water access sites at their opening, where discrete water access sites meet the main river channel or open lake water. For villages with more than one water access site, we estimated a weighted mean for each of these variables, where the average values across water access sites were weighted by water access site size ([(shoreline distance + opening distance)/2]*length). This follows the assumption that larger sites are used more intensely by local villagers.

8. Statistical analysis We were interested in (1) determining predictable habitat associations for the S. haematobium and S. mansoni competent host snails, Bulinus spp. and Biomphalaria pfeifferi, respectively; (2) using classified drone imagery to quantify snail habitat in and measure morphological features of water access sites (e.g., size and shape), to assess ecological features that correlate with human infection patterns; and, (3) determining the spatial scale at which models associating village-level snail habitat and human infection patterns fit best. To address our first goal, we used zero-inflated negative binomial generalized linear mixed effects models (ZINB GLMM) with a log link (package glmmTMB (24)) to test for differences in uninfected and infected Bulinus spp. and Biomphalaria pfeifferi snail abundance across the three dominant habitat types: non-emergent vegetation, emergent vegetation, and water/mud. Zero-inflation was assumed to be constant. We used data on snail abundance collected during quadrat-level sampling in all 6 field sampling periods across all 16 villages, to model four separate response variables: infected and uninfected Bulinus spp. and Biomphalaria pfeifferi snail counts, per quadrat. In addition to habitat type, we included water depth at each sampling quadrat as a covariate. Because water flow regimes are considerably different on lake versus river settings, and because snail-habitat associations may differ accordingly, we included village location (lake versus river) as an interaction term with habitat type. To account for our nested sampling design and repeated measures, we included a term for random intercepts of the field sampling period (1 through 6), within water access site, and within village. We used a model comparison approach (Akaike Information Criterion, AIC (25)) to derive the final model formulation used for each response variable, after comparing the previously described random intercept model to models including season within the nested structure, and to models without a zero-inflation term. Finally, we performed model residual diagnostics using the DhARMA package in R (26), and ensured lack of spatial- autocorrelation in model residuals using Moran’s I statistic in R. We computed estimated marginal means for snail counts in each habitat type and contrasts between all groups using the emmeans package in R (27).

74 To address our second and third goals, we used binomial GLMMs (logit link, package lme4 (28)) to model the probability that an individual was re-infected with S. haematobium or, separately, S. mansoni post-treatment with praziquantel in 2017 and 2018, and negative binomial GLMMs (log link, package glmmTMB (24))) to model individual S. haematobium and S. mansoni annual egg burden accumulation. All models on infection presence and egg burden controlled for demographic variables (age, sex, and village population size) and village location (lake versus river setting). Because infection in the same children was assessed over time, the data include multiple observations for each individual, within each village. Therefore, we included random effects terms for individual ID within village. Our key variables of interest were metrics of exposure to snail habitat across scales from nearshore up to 120m from shore, and morphological characteristics of water access sites that may influence human-water contact, water flow, and snail and larval parasite movement. For the latter, we included static variables for village-level shoreline access (distance, m), size of water access sites’ opening to open water in the river channel or lake (m), linear distance emergent vegetation “walls” representing the extent of edge habitat between emergent and non-emergent vegetation (m), and degree of water access sites’ circumscription, or whether or not a water access site was fully enclosed by emergent vegetation or connected to open water in the river channel or lake. To evaluate individual exposure to snail habitat across scales, we incorporated the total area of non-emergent vegetation in peak transmission season (spring-summer) measured across 15m increments between 15m and 120m from shore. Using ArcGIS Pro, we created an additional set of two smaller buffers at 1m and 5m distances, to test whether or not a simple measure of snail habitat very close to shore would be a sufficient indicator of transmission risk rather than measuring vegetation at greater distances. This resulted in 10 models for each response variable, that differed only in the variable for non-emergent habitat, which integrated across space at different scales. We then used a model comparison approach (AIC) to evaluate model fit, and, if relevant, to determine the spatial scale at which the data fit the model best. All continuous variables were centered and scaled.

RESULTS 1. Schistosomiasis prevalence and egg burden

75 In general, Schistosoma haematobium prevalence in school aged children across all villages in all years was higher than S. mansoni prevalence, but S. mansoni egg burden (geometric mean egg burden for positive infections) exceeded that of S. haematobium (Table 1). As we expected, drug treatment in 2016 led to substantial reductions in high- intensity infections, but re-infection rates and geometric mean egg intensity for both species were nonetheless high in 2017 and 2018; reinfection exceeded 66.0% for S. haematobium in both years, and was 13.8 in 2017 and 25.0% in 2018 for S. mansoni (Table 1).

2. Snail-habitat associations in water access sites Bulinus spp. snails and Biomphalaria pfeifferi snails exhibited similar habitat associations, but the strength of those associations differed based on whether water access sites were located on a river or a lake setting. In general, more snails of both genera were found in water access sites located on the river, and Bulinus spp. snails were more abundant than Biomphalaria pfeifferi in both lake and river settings (Fig. 2, Table 2). Considering habitat type, pairwise contrasts of the estimated marginal means for snail counts in each habitat type showed that both snail genera were found at their highest densities in non-emergent vegetation, followed by emergent vegetation, and with low densities in water/mud (Fig. 2). The exception to this trend was that Biomphalaria pfeifferi snail density did not differ between non-emergent vegetation and emergent vegetation in water access sites on a lake setting (pairwise contrast of estimated marginal means: p=0.992). We found that Bulinus spp. snail density increased with increasing sampling depth (zero-inflated negative binomial GLMMs, Estimate=0.405, SE=0.079, p<0.001, Table 2), but we found no response to depth for Biomphalaria pfeifferi within water access sites.

3. Snail abundance in nearshore and offshore habitats Field surveys on snail density in nearshore and offshore non-emergent vegetation revealed dense aggregations of snails in offshore, deep-water habitats, which are typically inaccessible for manual snail sampling (Table 3) and thus often overlooked in terms of transmission potential. The densest aggregations of Bulinus spp. snails on rivers and the lake were found in non-emergent vegetation >75m from water access site shorelines, with

76 fewer (but many) snails recovered in habitat in river channels or open lake water >135m or more from water access point shorelines. The lowest Bulinus spp. snail densities were found nearshore. In contrast, Biomphalaria pfeifferi snail density did not vary across distances and never exceeded 52 snails/m2, while Bulinus spp. density reached >1000 snails/m2 in lake habitats >75m.

4. Associating S. haematobium infection in humans with ecological correlates and spatial scale Six of ten models describing S. haematobium infection probability given non-emergent habitat integrated over various spatial scales fell within dAIC <10 (dAIC, or delta AIC, is the relative AIC difference between the model with the lowest AIC and each other model): models incorporating habitat at 1m, 5m, 15m, 90m, 105m, and 120m (Table 4). All six models found that human infection was correlated to the area of non-emergent habitat in village water access sites, but surprisingly, the correlation was strong and negative nearshore and strong and positive offshore (GLMM with logit link; non- emergent vegetation 5m scaled estimate= - 0.381, SE=0.082, p<0.001; 90m scaled estimate = 0.437, SE=0.096, p<0.001, Fig. 3b). At intermediate distances (30m to 75m), models integrating vegetation at 30m and 45m suggested no correlation between S. haematobium probability and non-emergent vegetation, but the relationship was strong and positive at 60m and 75m. Across all models, the length of water access site edge habitat and water access site size (shoreline access) were positively correlated with infection probability (Fig. 4a, Table 4). The effect of circumscription, or whether or not a discrete water access site was nearly or fully enclosed by emergent vegetation, differed in villages located on river vs. lake settings: circumscription had no effect on the river, but was negatively correlated with infection on the lake (i.e., enclosed sites had lower infection probability) (Fig. 4b, Table 4). S. haematobium egg burden data was best fit when integrating snail habitat at 90m (dAIC <12 for 105m and 120m; dAIC >60 for models between 1m and 30m). At this scale, the area of non-emergent vegetation had a strong positive correlation with egg burden (negative binomial GLMM with a log link, scaled estimate = 0.793, 0.085, p<0.001), but, unlike infection probability models, edge habitat, water access site size,

77 and circumscription had no effect (Table 4). Consistent with infection probability models, children living in villages on the lake did have a higher egg burden than children living in villages on the river. Younger children had higher egg burdens than older children, and boys had higher infection probability and egg burden than girls.

5. Associating S. mansoni infection in humans with ecological correlates and spatial scale The spatial scale at which non-emergent vegetation was measured had no effect on model fit for S. mansoni infection probability or egg burden when integrated across scales 15m or more from shore. However, models fit at 1m and 5m (dAIC <3) suggested a negative correlation between non-emergent habitat and infection risk (GLMM with a logit link non-emergent vegetation 1m scaled estimate = - 0.489, SE=0.125, p<0.001; Table 5). Rather, morphological features of water access sites influenced infection risk (Table 5). Specifically, water access site size (shoreline distance) was negatively correlated with S. mansoni infection probability (GLMM with a logit link 1m scaled estimate= - 0.661, SE=0.334, p=0.048; Fig. 4c, Table 5); and, the interaction between circumscription and village location revealed a positive correlation between site circumscription and S. mansoni risk on lake settings (scaled marginal estimate = 2.695, SE=1.10, p=0.014; Fig. 4d, Table 5), but no such relationship was found on river settings. Considering egg burden, non-emergent vegetation was negatively correlated with burden nearshore (1- 5m), and the response to circumscription varies in villages on the river vs. the lake (Table 5). Finally, males had higher S. mansoni infection probability and egg burden than females.

6. Summary of divergent ecological correlates and spatial scales of risk for S. haematobium and S. mansoni Models attempting to describe ecological predictors and spatial scales of co-occurring S. haematobium and S. mansoni risk diverged in several important ways. In terms of spatial scale of risk considering snail habitat (non-emergent vegetation), S. haematobium infection prevalence and egg burden were both positively correlated with the local extent of non-emergent vegetation in village water access points at areas 60m or more from shorelines, and model fit maximized at 90m. S. mansoni infection prevalence or risk was

78 not correlated with non-emergent vegetation at any scale above 15m. In terms of water access site morphology, S. haematobium infection prevalence was positively correlated with water access site size, whereas S. mansoni infection prevalence was negatively correlated with water access site size. Site circumscription, or the degree of enclosure by emergent vegetation “walls” that delineate water access sites, impacted risk for both schistosomes on lake settings, but in different ways: circumscription was negatively correlated with S. haematobium infection prevalence on the lake, and positively correlated with S. mansoni infection prevalence. Risk for acquiring S. haematobium was elevated in villages located on a lake setting, but risk for S. mansoni was not correlated with location on a lake or river setting. Overall, males were more likely to acquire infection and have higher egg burdens than girls for both schistosome species.

DISCUSSION Here, we show that human infection risk for S. haematobium and S. mansoni in a region co-endemic for both species corresponds to different ecological predictors and spatial scales of focal transmission. Neglecting the unique ecology of the two parasites and their snail hosts, or implementing environmental interventions, like snail control, at a scale that does not reflect a species’ transmission potential, could impede environmental control efforts. In some cases, the unique ecologies of the two genera could lead to unexpected trade-offs for human disease risk in co-endemic landscapes,, wherein activities to reduce the transmission of one species could lead to increased risk for the other. Identifying the unique features underlying risk for each species could help improve our ability to identify and eliminate transmission foci. We found that risk for S. haematobium was associated with a much broader spatial scale affecting village-level transmission risk than that for S. mansoni. Village- level S. haematobium infection prevalence and egg burden was strongly influenced by the local availability of suitable Bulinus spp. snail habitat, which we measured across scales ranging from 1m to 120m from shorelines using high resolution satellite and drone imagery. Infection was highest in villages with large water access sites and more snail habitat (both non-emergent vegetation and aggregated edge habitat alongside emergent vegetation). Comparison of models that incorporated non-emergent habitat across different spatial scales suggested that annual infection risk integrates over a spatial range

79 of about 90m. These findings were supported by quantitative snail sampling, which showed that Bulinus spp. snail density increased with sampling depth and was highest in non-emergent vegetation (“snail habitat”) beyond 75m from shorelines. S. mansoni infection risk, on the other hand, was highest in villages with small water access sites. The area of non-emergent snail habitat was not associated with infection, and so the scale at which snail habitat was measured was irrelevant. Rather, morphological characteristics of water access sites (e.g., shoreline access size and site circumscription) that probably influence water flow and snail-parasite interactions were the strongest correlates of infection risk. Indeed, previous studies on snail distributions in the Senegal River Basin found that Biomphalaria pfeifferi was most abundant in vegetation in smaller man-made habitats sheltered from wave action and strong water flow, while Bulinus spp. were preferentially found in larger man-made and natural habitats (29). Site circumscription, or the degree of enclosure by emergent vegetation “walls” that delineate water access sites (i.e., Fig. 2b,c) influenced both S. haematobium and S. mansoni infection risk, but only in villages located on Lac de Guiers (versus villages on river settings). However, whereas circumscription was negatively associated with S. haematobium risk, it was positively associated with S. mansoni risk, further suggesting that S. haematobium transmission is greater in large sites with high connectivity to offshore snail habitats, whereas S. mansoni transmission is more efficient in small sites with limited water flow. The observed associations between S. haematobium infection risk and egg intensity with water access site size and the local extent of available of snail habitat reflect risk integrated across space and time (16). Bulinus spp. snail abundance, distribution, and infection prevalence is highly clustered and heterogenous in space and time . Therefore, annual worm accumulation in humans (and egg deposition leading to human-to-snail transmission) is the consequence of exposure to potentially contaminated water over hours, days, weeks, and months. This is especially true for communities that rely on water contact for daily needs and livelihoods (30). Here we extend these findings and show that the association between human burden and snail habitat correlates is strongest when habitat correlates are integrated across areas 90m or more from shore. We use snail habitat as a proxy for susceptible and infected snails at this distance from shore,

80 but our quantitative snail sampling efforts confirmed that Bulinus spp. snails are abundant in non-emergent vegetation across this large scale, and from a study in Zimbabwe more than 30 years showed that the prevalence of infected snails was highest within the vicinity of water access sites at about 60m (31). In our study area, non-emergent vegetation can extend for hundreds of meters outside water access sites and support dense patches of Bulinus spp. snails. Over time, fragments of offshore vegetation can move into water access sites via wind, water flow, and mechanical transport on fishing boats and nets (32), thus importing snails from a large spatial sphere of influence that ultimately determines long-term S. haematobium infection risk, even while snail-to-human and human-to-snail transmission likely occurs at a finer scale (32). Mechanisms underlying the relatively smaller spatial scale of S. mansoni transmission may be related to differing biological features of S. mansoni miracidia and cercaria, and an overall lower abundance Biomphalaria vs. Bulinus spp. snails at our study sites. Across our study sites, Biomphalaria pfeifferi snails were in general less abundant than Bulinus spp. snails, and their density did not vary with distance from shorelines (whereas Bulinus spp. increased with distance). Correspondingly, fewer children enrolled in the study were re-infected with S. mansoni than with S. haematobium (4.8x and 2.8x lower prevalence in 2017 and 2018, respectively). Given low Biomphalaria snail density and, presumably, low S. mansoni egg deposition into local water bodies, small site size and circumscription might benefit S. mansoni transmission efficiency by concentrating miracidia–snail and cercaria-human contact rates (33). While small site size might also benefit S. haemabotium transmission efficiency, we hypothesize that over time, the benefit that large sites provide in terms of providing more space for non-emergent snail habitat and connection to Bulinus-dense off-shore habitats outweighs this benefit. Small site size and circumscription might also differentially benefit S. mansoni cercarial human skin penetration because they tend to shed from snails for shorter daily durations than S. haematobium (34), thus increasing the cercaria-human contact rates. Transmission risk for both species had one surprising correlate in common: the amount of non-emergent vegetation nearshore (within 1-15m) to water access sites was negatively correlated with infection risk (Table 4, Table 5). This contrasts with the common understanding that transmission typically occurs very close to shore. Field

81 surveys in Lake Volta, for example, suggested that the proximate risk for snail-to-human S. haematobium transmission – measured by the density of snails with mature infections in water access sites – was highest within the first few meters of shorelines (32). However, while dense nearshore vegetation might serve as snail habitat, it can also mechanically block the mobility of miracidia and cercaria (35–37), thus reducing snail and human infection success. In addition, nearshore vegetation may impact human water contact behavior in a different way than it does innate schistosome transmission efficiency, if water access sites with more water flow and less vegetation are preferentially used for cleaning, gathering water, and other daily activities. In the Senegal River Basin, infection risk for both S. haematobium and S. mansoni is high, but the snail and parasite ecology that drives transmission risk to humans is very different. The World Health Organization is currently encouraging member states to deploy schistosomiasis transmission control ‘wherever possible’ (4), but there is no one- size-fits-all solution for environmental intervention to interrupt transmission for all schistosome species in all settings, like praziquantel provides for the simultaneous elimination of co-infecting adult schistosomes in humans. To facilitate ecologically relevant intervention design, in this study we aimed to expand basic understanding of schistosome transmission ecology and spatial scales of risk. We also developed methods to validate and scale an aerial assessment of risk at the village level, where transmission occurs. Ultimately, these findings can be used to develop criteria and remote-sensing methods to rapidly identify transmission foci and deploy cost-effective interventions in complex transmission landscapes.

82 Acknowledgements We are grateful to all laboratory and field staff at Espoir Pour La Santé, for all scientific and logistical support. We are also grateful for all study participants and their communities for welcoming us to conduct field surveys. Funding: IJJ was supported by a National Science Foundation Graduate Research Fellowship (#1656518). IJJ was supported by IJJ, GADL, SHS, CLW, and AC were supported by a grant from the Bill and Melinda Gates Foundation (OPP1114050), a grant from the National Institutes of Health (R01TW010286), and a grant from the National Science Foundation (1414102). IJJ, GADL, SHS, and AC were also supported by a 2018 Environmental Venture Program grant from the Stanford University Woods Institute for the Environment and a GDP SEED grant from the Freeman Spogli Institute at Stanford University. CLW was supported by the Michigan Society of Fellows at the University of Michigan, a Sloan Research Fellowship from the Alfred P. Sloan Foundation, a grant from the National Science Foundation (OCE-1829509), and a University of Washington Innovation Award.

83 References:

1. WHO, “Schistosomiasis and soil-transmitted helminthiases: number of people treated in 2018,” Weekly Epidemiological Record (50, World Health Organization, Geneva, Switzerland, 2019).

2. D. G. Colley, A. L. Bustinduy, W. E. Secor, C. H. King, Human schistosomiasis. Lancet. 383, 2253–2264 (2014).

3. N. C. Lo, D. G. Addiss, P. J. Hotez, C. H. King, J. R. Stothard, D. S. Evans, D. G. Colley, W. Lin, J. T. Coulibaly, A. L. Bustinduy, G. Raso, E. Bendavid, I. I. Bogoch, A. Fenwick, L. Savioli, D. Molyneux, J. Utzinger, J. R. Andrews, A call to strengthen the global strategy against schistosomiasis and soil-transmitted helminthiasis: the time is now. The Lancet Infectious Diseases. 17, e64–e69 (2017).

4. “Schistosomiasis: progress report 2001 - 2011, strategic plan 2012 - 2020” (WHO/HTM/NTD/PCT/2013.2, World Health Organization, Geneva, Switzerland, 2013).

5. N. Kittur, C. H. King, C. H. Campbell, S. Kinung’hi, P. N. M. Mwinzi, D. M. S. Karanja, E. K. N’Goran, A. E. Phillips, P. H. Gazzinelli-Guimaraes, A. Olsen, P. Magnussen, W. E. Secor, S. P. Montgomery, J. Utzinger, J. W. Walker, S. Binder, D. G. Colley, Persistent Hotspots in Schistosomiasis Consortium for Operational Research and Evaluation Studies for Gaining and Sustaining Control of Schistosomiasis after Four Years of Mass Drug Administration of Praziquantel. The American Journal of Tropical Medicine and Hygiene. 101, 617–627 (2019).

6. M. W. Mutuku, M. R. Laidemitt, B. R. Beechler, I. N. Mwangi, F. O. Otiato, E. L. Agola, H. Ochanda, B. Kamel, G. M. Mkoji, M. L. Steinauer, E. S. Loker, A Search for Snail-Related Answers to Explain Differences in Response of Schistosoma mansoni to Praziquantel Treatment among Responding and Persistent Hotspot Villages along the Kenyan Shore of Lake Victoria. The American Journal of Tropical Medicine and Hygiene. 101, 65–77 (2019).

7. J. W. Walker, N. Kittur, S. Binder, J. D. Castleman, J. M. Drake, C. H. C. Jr, C. H. King, D. G. Colley, Environmental Predictors of Schistosomiasis Persistent Hotspots following Mass Treatment with Praziquantel. The American Journal of Tropical Medicine and Hygiene. 102, 328–338 (2020).

8. “Accelerating work to overcome the global impact of neglected tropical diseases – A roadmap for implementation” (WHO/HTM/NTD/2012.1, World Health Organization, Geneva, Switzerland, 2012).

9. WHO, Field use of molluscicides in schistosomiasis control programmes: an operational manuel for programme managers. (World Health Organization (WHO), Geneva, Switzerland, 2017).

84 10. S. H. Sokolow, C. L. Wood, I. J. Jones, K. D. Lafferty, A. M. Kuris, M. H. Hsieh, G. A. De Leo, To Reduce the Global Burden of Human Schistosomiasis, Use ‘Old Fashioned’ Snail Control. Trends in Parasitology. 34, 23–40 (2018).

11. S. H. Sokolow, C. L. Wood, I. J. Jones, S. J. Swartz, M. Lopez, M. H. Hsieh, K. D. Lafferty, A. M. Kuris, C. Rickards, G. A. De Leo, Global Assessment of Schistosomiasis Control Over the Past Century Shows Targeting the Snail Intermediate Host Works Best. PLoS Negl Trop Dis. 10 (2016), doi:10.1371/journal.pntd.0004794.

12. D. Rollinson, S. Knopp, S. Levitz, J. R. Stothard, L.-A. Tchuem Tchuenté, A. Garba, K. A. Mohammed, N. Schur, B. Person, D. G. Colley, J. Utzinger, Time to set the agenda for schistosomiasis elimination. Acta Tropica. 128, 423–440 (2013).

13. C. H. King, R. F. Sturrock, H. C. Kariuki, J. Hamburger, Transmission control for schistosomiasis – why it matters now. Trends in Parasitology. 22, 575–582 (2006).

14. C. Simoonga, J. Utzinger, S. Brooker, P. Vounatsou, C. C. Appleton, A. S. Stensgaard, A. Olsen, T. K. Kristensen, Remote sensing, geographical information system and spatial analysis for schistosomiasis epidemiology and ecology in Africa. Parasitology. 136, 1683–1693 (2009).

15. Y. Walz, M. Wegmann, B. Leutner, S. Dech, P. Vounatsou, E. K. N’Goran, G. Raso, J. Utzinger, Use of an ecologically relevant modelling approach to improve remote sensing-based schistosomiasis risk profiling. Geospat Health. 10 (2015), doi:10.4081/gh.2015.398.

16. C. L. Wood, S. H. Sokolow, I. J. Jones, A. J. Chamberlin, K. D. Lafferty, A. M. Kuris, M. Jocque, S. Hopkins, G. Adams, J. C. Buck, A. J. Lund, A. E. Garcia- Vedrenne, E. Fiorenza, J. R. Rohr, F. Allan, B. Webster, M. Rabone, J. P. Webster, L. Bandagny, R. Ndione, S. Senghor, A.-M. Schacht, N. Jouanard, G. Riveau, G. A. D. Leo, Precision mapping of snail habitat provides a powerful indicator of human schistosomiasis transmission. PNAS. 116, 23182–23191 (2019).

17. Y.-S. Lai, P. Biedermann, U. F. Ekpo, A. Garba, E. Mathieu, N. Midzi, P. Mwinzi, E. K. N’Goran, G. Raso, R. K. Assaré, M. Sacko, N. Schur, I. Talla, L.-A. T. Tchuenté, S. Touré, M. S. Winkler, J. Utzinger, P. Vounatsou, Spatial distribution of schistosomiasis and treatment needs in sub-Saharan Africa: a systematic review and geostatistical analysis. The Lancet Infectious Diseases. 15, 927–940 (2015).

18. G. Riveau, A.-M. Schacht, J.-P. Dompnier, D. Deplanque, M. Seck, N. Waucquier, S. Senghor, D. Delcroix-Genete, E. Hermann, N. Idris-Khodja, C. Levy-Marchal, M. Capron, A. Capron, Safety and efficacy of the rSh28GST urinary schistosomiasis vaccine: A phase 3 randomized, controlled trial in Senegalese children. PLoS Negl Trop Dis. 12 (2018), doi:10.1371/journal.pntd.0006968.

19. N. Katz, A. Chaves, J. Pellegrino, A simple device for quantitative stool thick-smear technique in Schistosomiasis mansoni. Revista do Instituto de Medicina Tropical de Sao Paulo. 14, 397–400 (1972).

85 20. B. L. Webster, O. T. Diaw, M. M. Seye, J. P. Webster, D. Rollinson, Introgressive Hybridization of Schistosoma haematobium Group Species in Senegal: Species Barrier Break Down between Ruminant and Human Schistosomes. PLoS Negl Trop Dis. 7, e2110 (2013).

21. L. a. T. Tchuenté, V. R. Southgate, A. Théron, J. Jourdane, A. Ly, B. Gryseels, Compatibility of Schistosoma mansoni and Biomphalaria pfeifferi in Northern Senegal. Parasitology. 118, 595–603 (1999).

22. M. Baatz, A. Schäpe, Multiresolution Segmentation: an optimization approach for high quality multi-scale image segmentation, 12 (2000).

23. R. S. DEFRIES, J. R. G. TOWNSHEND, NDVI-derived land cover classifications at a global scale. International Journal of Remote Sensing. 15, 3567–3586 (1994).

24. M. Brooks E., K. Kristensen, K. Benthem J. ,van, A. Magnusson, C. Berg W., A. Nielsen, H. Skaug J., M. Mächler, B. Bolker M., glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. The R Journal. 9, 378 (2017).

25. K. P. Burnham, D. R. Anderson, K. P. Huyvaert, AIC model selection and multimodel inference in behavioral ecology: some background, observations, and comparisons. Behav Ecol Sociobiol. 65, 23–35 (2011).

26. F. Hartig, DHARMa: residual diagnostics for hierarchical (multi-level/mixed) regression models (2016; https://cran.r- project.org/web/packages/DHARMa/vignettes/DHARMa.html).

27. R. Lenth, emmeans: Estimated Marginal Means, aka Least-Squares Means (2020; https://CRAN.R-project.org/package=emmeans).

28. D. Bates, M. Mächler, B. Bolker, S. Walker, Fitting Linear Mixed-Effects Models Using lme4. J. Stat. Soft. 67 (2015), doi:10.18637/jss.v067.i01.

29. R. F. Sturrock, O.-T. Diaw, I. Talla, M. Niang, J.-P. Piau, A. Capron, Seasonality in the transmission of schistosomiasis and in populations of its snail intermediate hosts in and around a sugar irrigation scheme at Richard Toll, Senegal. Parasitology. 123, 77–89 (2001).

30. A. J. Lund, M. M. Sam, A. B. Sy, O. W. Sow, S. Ali, S. H. Sokolow, S. Bereknyei Merrell, J. Bruce, N. Jouanard, S. Senghor, G. Riveau, D. Lopez-Carr, G. A. De Leo, Unavoidable Risks: Local Perspectives on Water Contact Behavior and Implications for Schistosomiasis Control in an Agricultural Region of Northern Senegal. Am J Trop Med Hyg. 101, 837–847 (2019).

31. M. E. Woolhouse, S. K. Chandiwana, Spatial and temporal heterogeneity in the population dynamics of Bulinus globosus and Biomphalaria pfeifferi and in the epidemiology of their infection with schistosomes. Parasitology. 98 ( Pt 1), 21–34 (1989).

86 32. R. K. Klumpp, K. Y. Chu, Ecological studies of Bulinus rohlfsi, the intermediate host of Schistosoma haematobium in the Volta Lake. Bull World Health Organ. 55, 715–730 (1977).

33. K. Y. Chu, J. A. Vanderburg, R. K. Klumpp, Transmission dynamics of miracidia of Schistosoma haematobium in the Volta Lake. Bulletin of the World Health Organization. 59, 555–560 (1981).

34. R. F. Sturrock, in Schistosomiasis (Imperial College Press, 2001).

35. N. Ø. Christensen, Schistosoma mansoni: interference with cercarial host-finding by various aquatic organisms. Journal of Helminthology. 53, 7–14 (1979).

36. D. W. Thieltges, K. T. Jensen, R. Poulin, The role of biotic factors in the transmission of free-living endohelminth stages. Parasitology. 135, 407–426 (2008).

37. J. E. Welsh, J. van der Meer, C. P. D. Brussaard, D. W. Thieltges, Inventory of organisms interfering with transmission of a marine trematode. Marine Biological Association of the United Kingdom. Journal of the Marine Biological Association of the United Kingdom; Cambridge. 94, 697–702 (2014).

87 Figures and Tables

Figure 1. Study location and design. A) Location of 16 study villages (encompassing 32 distinct water access sites) in the Senegal River Basin; 6 villages are located on river settings, and 10 villages are located adjacent to Lac de Guiers. B–C) Overhead drone imagery (left) and classified non-emergent vegetation (red), emergent vegetation (green), water (blue), and land (tan) superimposed by sampling bands used to integrate and compare model fit across scales, for a village on a river setting (B) and a village Lac de Guiers (C).

Figure 2. Estimated marginal means (and 95% CI) for Bulinus spp. snails and Biomphalaria pfeifferi in three dominant habitat types in water access sites on river and lake setting. Note that each axis scale is different. The highest density of Bulinus spp. snails on river (A) and lake (B) settings were found in non-emergent vegetation, followed by emergent vegetation, then water/mud; density was higher in general on the river than on the lake. Biomphalaria pfeifferi densities in river (C) and lake (D) settings were in general lower than Bulinus spp., but showed similar habitat associations, except that there was no difference in density between emergent and non-emergent habitat in water access sites on a lake setting. Brackets and asterisks indicate pairwise contrasts of snail density in each habitat type. N.S. = non-significant, **p < 0.01, ***p < 0.001.

Figure 3. (A) Individuals performing daily chores (foreground) and playing (background) in a typical water access site in the Senegal river basin; overlaid is a visual of the scales at which we measured risk (red bands) and the distribution of non-emergent snail habitat in relation to human activity. Effect of the extent of non-emergent vegetation integrated across a 90m area from shorelines on S. haematobium (B) and S. mansoni (C) risk. Points represent village-level re-infection prevalence, and the lines (where present) represent statistically significant linear associations. The area of non-emergent vegetation was positively correlated with S. haematobium risk (GLMM with a logit link, scaled estimate = 0.437, SE = 0.01, p<0.001) on river and lake settings. No such correlation between non-emergent vegetation area and S. mansoni risk was found on lake or river settings for S. mansoni.

90 A S. haematobium C S. mansoni River Lake Village prevalence Village

B Water access site shoreline (m) D Water access site shoreline (m) Village prevalence Village

Water access point circumscription Water access point circumscription

Figure 4. Effect of water access site size (linear distance of water access site shorelines) (A), and water access site circumscription (B) on village-level S. haematobium re- infection risk. Solid lines represent model predictions for infection probabilities and points represent village-level infection prevalence in children. Effects of water access site size (linear distance of water access site shorelines) (C), and water access site circumscription (D) on village-level S. mansoni re-infection risk. The association between water access site size and reinfection is not separate for villages located on river versus lake because the effect of site size did not differ according to location.

91 Table 1. Infection prevalence and egg burden (egg intensity) for all participants in the longitudinal cohort study at baseline in 2016, and re-infection in 2017 and 2018 following treatment. Egg intensity is the geometric mean egg count for all positive infections (>1 egg/10ml urine or >1 egg/g feces), and % heavy infections is the percent of all participants that are considered to harbor high intensity infections (>50 eggs/ml urine for S. haematobium and >200 eggs/g feces for S. mansoni).

S. haematobium

Egg % heavy Year N Prevalence (%) intensity intensity 2016 (baseline) 1395 76.8 15.4 24.9 2017 (rate) 1171 66.3 14.8 0.0 2018 (rate) 1397 68.8 13.9 0.0 S. mansoni

Egg % heavy Year N Prevalence (%) intensity intensity 2016 (baseline) 1395 34.2 65.5 5.0 2017 (rate) 1171 13.8 39.3 0.0 2018 (rate) 1397 25.0 43.1 0.0 co-infection

% heavy Year N Prevalence (%) intensity 2016 (baseline) 1395 29.9 2.2 2017 (rate) 1171 11.9 0.0 2018 (rate) 1397 19.9 0.0

92 Table 2. Regression table for snail-habitat models (ZINB GLMM) associating snail density (snail counts per sampling quadrat) in water access sites with freshwater habitat type, water depth, and village location (lake vs. river), for both Bulinus spp. snails and Biomphalaria pfeifferi. Coefficients for change in snail density according to habitat type are shown for non-emergent vegetation and open water with emergent vegetation as the reference category. Bolded p-values represent statistically significant factors. *p < 0.05, **p < 0.01, ***p < 0.001.

Biomphalaria Bulinus spp. pfeifferi Intercept 0.543 -1.333 (0.699) (0.916) Floating/Submerged 1.864 *** 1.368 *** vegetation (0.301) (0.381) Open water -1.097 *** -1.244 ** (0.278) (0.387) Location: Lake -1.759 * -4.065 ** (vs. River) (0.888) (1.459) Depth (scaled) 0.405 *** 0.163 (0.079) (0.114)

Interaction: -1.325 *** -1.062 Floating/Submerged vegetation X Lake (0.384) (0.568) Interaction: Open water -1.536 *** -1.621 * X Lake (0.384) (0.708) N obs. 2538 2540 logLik -2245.716 -783.314 AIC 4515.433 1590.628 *** p<0.001; **p<0.01; *p< 0.05

93 Table 3. Results from offshore, deep water snail sampling in non-emergent vegetation for three villages located on river settings and one village located on a lake setting. Snails were sampled in three categories of distance from shorelines: near (water access site areas surrounded by emergent vegetation walls, to edge of open river channel or lake water), medium (edge of near polygon to 75m beyond), and far (edge of medium polygon to 40–130m beyond).

Bulinus spp. River Lake Distance from shore Snail density SE Snail density SE near 12.1 4.1 4.7 1.4 medium 107.7 29.9 290.7 124.4 far 59.4 20.0 111.8 50.2 Biomphalaria pfeifferi River Lake Distance from shore (m) Snail density SE Snail density SE near 4.5 1.7 5.3 4.1 medium 4.5 2.5 3.3 3.3 far 5.9 2.2 3.9 2.0

94 Table 4. Infection presence (binomial GLMM) and egg burden (negative binomial GLMM) model results for S. haematobium, comparing models that integrated non- emergent vegetation coverage across different scales ranging from 1m to 120m. Shown for infection presence models are all models within 10 dAIC. The model with the lowest AIC is shown for egg burden models. All predictors were centered and scaled. Bolded p- values represent statistically significant factors. *p < 0.05, **p < 0.01, ***p < 0.001.

Egg burden: Infection presence: 1m 5m 15m 90m 105m 120m 90m

(Intercept) -0.529 -0.511 -0.470 -0.006 -0.106 -0.158 1.677 **

(0.387) (0.398) (0.345) (0.396) (0.368) (0.359) (0.651)

Age -0.019 -0.023 -0.039 -0.016 -0.019 -0.020 -0.315 ***

(0.061) (0.061) (0.061) (0.061) (0.061) (0.061) (0.058)

Sex: Male 0.478 *** 0.480 *** 0.481 *** 0.485 *** 0.483 *** 0.484 *** 0.540 ***

(0.111) (0.112) (0.112) (0.112) (0.112) (0.112) (0.114)

Population -0.281 * -0.248 -0.274 * -0.220 -0.253 -0.265 * -0.020

(0.143) (0.148) (0.124) (0.143) (0.131) (0.128) (0.252)

Year: 2018 0.669 *** 0.552 *** 0.212 0.331 ** 0.374 ** 0.385 *** 0.284 **

(0.150) (0.126) (0.116) (0.115) (0.115) (0.116) (0.098)

Area -0.347 *** -0.381 *** -0.358 *** 0.437 *** 0.367 *** 0.365 *** 0.793 *** floating/submerged vegetation (0.094) (0.082) (0.078) (0.101) (0.090) (0.086) (0.091)

Edge habitat (m) 0.760 ** 0.754 ** 0.585 * 0.854 ** 0.806 ** 0.780 ** 0.744

(0.270) (0.276) (0.235) (0.272) (0.252) (0.247) (0.426)

Shoreline distance 0.812 *** 0.803 *** 0.803 *** 0.707 *** 0.720 *** 0.713 *** 0.536 (m) (0.198) (0.204) (0.173) (0.197) (0.183) (0.179) (0.337)

Access point -0.399 -0.376 -0.194 -0.725 * -0.634 * -0.587 * -0.573 opening (m) (0.279) (0.288) (0.251) (0.289) (0.265) (0.259) (0.475)

Average 0.941 1.047 1.234 0.212 0.428 0.560 1.312 circumscription (0.769) (0.796) (0.671) (0.788) (0.723) (0.703) (1.369)

Location: Lake 2.037 *** 2.137 *** 2.323 *** 1.577 *** 1.642 *** 1.670 *** 1.713 *

(0.428) (0.441) (0.378) (0.437) (0.406) (0.397) (0.742)

Interaction: -3.453 *** -3.578 *** -3.210 *** -2.494 *** -2.592 *** -2.634 *** -2.434 Location (Lake) X circumscription (0.754) (0.781) (0.650) (0.751) (0.695) (0.680) (1.334)

N obs. 2262 2262 2262 2262 2262 2262 2262 logLik -1128.998 -1124.803 -1125.092 -1125.960 -1127.410 -1126.501 -8134.225

AIC 2285.996 2277.606 2278.185 2279.920 2282.820 2281.003 16298.450

*** p < 0.001; ** p < 0.01; * p < 0.05.

95 Table 5. Infection presence (binomial GLMM) and egg burden (negative binomial GLMM) model results for S. mansoni, comparing models that integrated non-emergent vegetation coverage across different scales ranging from 1m to 120m. Shown for infection presence models are all the two models within 10 dAIC (1m and 5m scales), and the model integrating vegetation at 90m for comparison. The two models within 10 dAIC are shown for egg burden models. All predictors were centered and scaled. Bolded p-values represent statistically significant factors. *p < 0.05, **p < 0.01, ***p < 0.001.

Egg burden: Infection presence: 1m 5m 90m 1m 5m

(Intercept) -2.326 *** -2.228 *** -2.031 ** 0.436 0.533

(0.644) (0.652) (0.681) (1.146) (1.169)

Age -0.042 -0.048 -0.050 0.035 0.044

(0.069) (0.069) (0.069) (0.139) (0.140)

Sex: Male 0.507 *** 0.501 *** 0.496 *** 0.938 *** 0.902 ***

(0.135) (0.135) (0.134) (0.271) (0.271)

Population 0.305 0.343 0.286 0.628 0.707

(0.252) (0.257) (0.266) (0.440) (0.452)

Year: 2018 0.631 *** 0.333 * 0.214 0.594 0.113

(0.174) (0.143) (0.137) (0.322) (0.269)

Area -0.489 *** -0.403 *** 0.121 -0.884 *** -0.747 *** floating/submerged vegetation (0.125) (0.114) (0.136) (0.223) (0.199)

Edge habitat (m) 0.580 0.505 0.426 1.212 1.077

(0.423) (0.428) (0.447) (0.741) (0.755)

Shoreline distance -0.661 * -0.672 * -0.681 -0.802 -0.854 (m) (0.334) (0.339) (0.351) (0.581) (0.595)

Access point 0.253 0.252 0.086 0.034 0.091 opening (m) (0.468) (0.476) (0.498) (0.815) (0.836)

Average -0.358 -0.251 -0.563 -1.343 -1.009 circumscription (1.330) (1.353) (1.419) (2.324) (2.385)

Location: Lake -0.298 -0.158 -0.278 -0.673 -0.424 (0.723) (0.734) (1.269) (1.298) (0.776)

Interaction: 3.075 * 3.109 * 3.642 ** 4.876 * 4.847 * Location (Lake) X circumscription (1.320) (1.342) (1.396) (2.291) (2.350)

N obs. 2251 2251 2251 2251 2251

logLik -872.329 -873.770 -879.409 -3555.852 -3556.672

AIC 1772.659 1775.540 1786.818 7141.704 7143.344

Chapter 4

Improving rural health care reduces illegal logging and conserves carbon in a

tropical forest

Isabel J. Jones1*, Andrew J. MacDonald2, Skylar Hopkins3, Andrea J. Lund4, Zac Yung-

Chun Liu1, Nurul Ihsan Fawzi5, Mahardika Putra Purba5, Katie Fankhauser6, Andrew J.

Chamberlin1, Monica Nirmala7, Arthur G. Blundell8, Ashley Emerson9, Jonathan

Jennings9, Lynne Gaffikin10, Michele Barry11, David Lopez-Carr12, Kinari Webb11,

Giulio A. De Leo1,13, Susanne H. Sokolow1,14

1 Hopkins Marine Station, Department of Biology, Stanford University, 120 Oceanview Boulevard, Pacific

Grove, CA, USA 2 Department of Biology, Stanford University & Earth Research Institute, University of California, Santa Barbara, CA, USA 3 National Center for Ecological Analysis and Synthesis, Santa Barbara, CA, USA 4 Emmett Interdisciplinary Program in Environment and Resources, Stanford University, Stanford, CA, USA 5 Alam Sehat Lestari, Sukadana, West Kalimantan, Indonesia 6 Oregon Health and Science University, Portland, OR, USA 7 Department of Global Health and Population, Harvard T. H. Chan School of Public Health, Boston, MA, USA 8 Natural Capital Advisors, New Orleans, LA 9 Health In Harmony, Portland, OR, USA 10 Department of Obstetrics and Gynecology & Center for Innovation In Global Health, Stanford University, Stanford, CA, USA 11 Center for Innovation in Global Health, Stanford University, Stanford, CA, USA 12 Bren School of Environmental Science & Management University of California Santa Barbara, Santa Barbara, CA, USA 13 Woods Institute for the Environment, Stanford University, Stanford, CA, USA 14 Marine Science Institute, University of California, Santa Barbara, Santa Barbara, CA, USA

*Corresponding author: Isabel J. Jones; [email protected] Keywords: Planetary health; natural climate solutions; human health; tropical forests; conservation

Abstract

Tropical forest loss currently exceeds forest gain, leading to a net greenhouse gas emission that exacerbates global climate change. This has sparked scientific debate on how to achieve natural climate solutions. Central to this debate is whether sustainably managing forests and protected areas will deliver global climate mitigation benefits, while ensuring local peoples’ health and well-being. Here we evaluate the 10-year impact of a human-centered solution to achieve natural climate mitigation through reductions in illegal logging in rural Borneo: an intervention aimed at expanding health care access and use for communities living near a national park, with clinic discounts offsetting costs historically met through illegal logging. Conservation, education, and alternative livelihood programs were also offered. We hypothesized that this would lead to improved health and well-being, while also alleviating illegal logging activity within the protected forest. We estimated that 27.4 km2 of deforestation was averted in the national park over a decade (~70% reduction in deforestation compared to a synthetic control, permuted p=0.038). Concurrently, the intervention provided health care access to more than 28,400 unique patients, with clinic usage and patient visitation frequency highest in communities participating in the intervention. Finally, we observed a dose response in forest change rate to intervention engagement (person-contacts with intervention activities) across communities bordering the park: the greatest logging reductions were adjacent to the most highly engaged villages. Results suggest that this community-derived solution simultaneously improved health care access for local and indigenous communities and sustainably conserved carbon stocks in a protected tropical forest.

Introduction Tropical forests lose more than 100 trees every second, altering landscapes and impacting livelihoods, health, biodiversity, and climate change (1). Across the tropics, forest loss now exceeds forest gain, leading to a net carbon emission from some of the most important natural carbon stocks in the world (2). Averting further forest loss is an important natural climate solution and a high priority for science, management, and policy from local to global scales (3, 4).

In biodiverse, carbon-rich tropical forests, the establishment of protected areas benefits both conservation and climate mitigation goals, but often involves excluding, and thus disenfranchising, local communities that surround protected areas (5, 6). Failure to address the needs of local people can in turn lead to unsustainable forest use, when communities with few alternatives illegally extract resources and convert land in order to survive (6, 7). Another major hypothesized driver of poverty is lack of access to high- quality, affordable health care, which can lead to vicious cycles of poor health and expanding out-of-pocket costs, further incentivizing poor families to rely on unsustainable resource use, like illegal logging, in order to raise cash to meet critical health care needs (8). Within this context, this study examines whether providing rural health care incentivizes reductions in illegal logging by local and indigenous communities living around a national park in Indonesian Borneo (Fig. 1A), thereby improving health and well-being and conserving biodiversity and globally-important carbon storage.

Indonesia contains some of the most carbon-dense forests in the world (Fig. S1)

(9), with the island nation representing only 1.4% of the world’s land area, but 3.6% of natural forest cover (10). A 2011 moratorium on new logging concessions was implemented to reduce total emissions from deforestation in Indonesia (10), but, at the

100

same time, illegal logging was estimated to represent as much as 61% of all logging activity (11). Protected areas, which cover 12% of Indonesia’s land mass (12), may be key to preserving remaining carbon stocks, biodiversity, and forest ecosystem services

(13). However, top-down law enforcement has proven insufficient to prevent conservation threats to protected areas, like illegal encroachment and fire (14). Indeed, despite protected status, more than 60% of lowland forests within protected areas in

Borneo’s West Kalimantan region were lost to illegal logging between 1985 and 2001

(15, 16), and this trend continues to accelerate across the region (Fig. 1B).

To better understand local attitudes towards forests, conservation, and drivers of illegal logging in West Kalimantan, Indonesia, a non-profit organization conducted more than 400 hours of focus groups between 2005-2007 with nearly 500 community representatives (community leaders, farm group leaders, religious leaders, teachers, women’s group leaders, and interested community members). The open-ended conversations identified access to affordable, high quality health care as a major basic need, and lack of access to health care a potential driver of illegal logging in 23 districts near Gunung Palung National Park (GPNP) (Fig. 1C) (17). This corresponds to a broader concern in Indonesia, which ranks in the lowest one third of countries in terms of coverage of essential health services (Fig. 1A) (18), and where many populations contend with unmet water, sanitation, and hygiene needs, high maternal and infant mortality, and high burdens of infectious and noninfectious diseases (19). In response to local health care needs and conservation implications, the non-profit established a local health clinic in 2007 in close partnerships with the district government and the national park management. Clinic services and alternative payment options (e.g., barter options including seedlings and manure used in conservation activities) were available to anyone

101

seeking care, complementing the limited healthcare available provided by the government. With the support of the National Park management, Memorandum of

Understanding (MOU) agreements were signed by the non-profit and 21 of 23 districts

(‘desa’ administrative units), representing 73 villages (‘dusun’ administrative units), near

GPNP to participate in the health care–conservation exchange intervention. Through the intervention, clinic discounts were given to villages based on community-wide reductions in illegal logging activity, as reported by community liaisons and through monitoring of logging trail activity. At the request of community members in intervention areas, conservation programs, educational programs, and alternative livelihood trainings were also facilitated periodically in partnership with government entities. In 2007, 2012, and

2017, household surveys were conducted in random households across districts bordering the park to assess self-reported changes in well-being, knowledge, attitudes, behaviors, and livelihoods, in order to adapt intervention activities accordingly to meet community needs.

Here we use more than 10 years of de-identified patient records from the health clinic coupled with remotely-sensed earth observation data to test the hypothesis that a multi-sector health care–conservation intervention can simultaneously improve human health care access and disincentivize illegal deforestation in a carbon-rich tropical forest.

First, we use a synthetic controls approach to compare park-level forest loss rates in

GPNP before versus after the intervention began in 2007, compared to all terrestrial

Indonesian IUCN Category II National Parks as potential controls. MOU-signing villages bordered the national park and were thus non-randomly assigned, precluding analysis as a randomized controlled trial. However, we were able to use clinic patient records collected between 2008 and 2018 to compare health care access, usage, and diagnosis trends

102

between MOU signing and non-MOU signing patient groups in an ex post facto, quasi- experimental study design. We tested whether, near GPNP, (i) the clinic increased health care usage for patients from villages with signed MOUs and villages without signed

MOUs, and (ii) trends in disease diagnoses changed over the same time period for patients from villages with and without signed MOUs. To establish the plausibility of a causal relationship between the intervention and conservation outcomes, forest change data, clinic usage data, and records on village-level engagement with the intervention programs (i.e., periodic education and livelihoods programs) were then used to test for a dose-response of village-level forest loss within the national park to village-level intensity of engagement with the intervention and its associated programs. Lastly, select responses from self-reported household survey data collected by the intervention team in

2007, 2012, and 2017 were used to assess changes in household livelihoods and income over the intervention period, and to gain further insight into potential mechanisms driving conservation–health linkages.

Results

Intervention impact on forest change

Remotely sensed forest loss rates (1) in a synthetic controls analysis (20, 21), were significantly lower within districts intersecting the local national park, GPNP, compared to a synthetic control assembled from districts in 32 control parks across

Indonesia during the post-intervention period from 2008 to 2018, as compared to a pre- intervention period in 2001 to 2007 (Estimate: 69.8% reduction of forest loss, 90% CI:

50.8 – 81.4, p=0.003; Table 1, Fig. 2A). This translates to an estimated 27.4 km2 of forest loss averted post-intervention to 2018 (90% CI: 19.9 – 32.0 km2) (Table 1, Fig. 2A). This finding was robust to a number of alternative data subsets defining the “donor pool” of

103

possible control parks to be included in the synthetic control (Table 1), as well as to comparison of the intervention effect with 500 permuted ‘placebo’ treatment regions made up of randomly assigned units from the pool of control districts (Estimate: 69.8% reduction, permuted-p=0.038, 90% CI: 26.3 – 83.7; Fig. 2B) (21). Correspondingly, household responses collected through monitoring and evaluation activities during 2007

(baseline), 2012, and 2017 demonstrated a strong and significant reduction in self- reported illegal logging activity: the number of adult males who reported logging inside

GPNP during the intervention period compared to baseline declined (generalized linear regression with logit link, Estimate: 68.8% reduction, 95% CI: 60.8 – 75.5, p<0.001; Fig.

2A), as did the number of households reported to rely on logging as a primary income source (generalized linear regression with logit link, Estimate: 90.6% reduction, 95% CI:

83.4 – 95.2, p<0.001; Fig. 3C).

Intervention impact on forest carbon

Using published carbon equations parameterized specifically for Borneo (22) along with LiDAR-estimated canopy heights in the local national park, GPNP, in 2014

(Fig. 2C, Fig. S2), the effect size of 69.8% reduction in annual forest loss was estimated to equate to a cumulative 0.59 Teragrams (Tg) of aboveground carbon loss averted (90%

CI: 0.27–1.13 Tg). Based on the maximum trade value of $30 per tonne of CO2 realized on the European Emissions Trading System (23), the gross value of the total carbon loss averted in GPNP on the European carbon market would have been approximately $65.3 million USD in 2019. The estimate of aboveground carbon loss averted is conservative, because (i) the LiDAR flight data in GPNP covered mixed and some previously burned forest types, and the derived average vegetation height (27 m) is much lower than the tallest canopy height recorded in GPNP (71 m) (Fig. 2B), and (ii) the relationship

104

between canopy height and aboveground carbon is convex and non-linear, suggesting that averaging across 30m x 30m pixels consistently underestimates the true carbon density

(22). The carbon value of the intervention impact demonstrates a theoretical monetary return that would more than offset intervention costs if a carbon market were accessible to interventions aiming to couple rural health programs with forest conservation in a similar way.

Intervention impact on health clinic usage diagnoses

Overall, 28,462 unique patients visited the clinic at least once over the study period from 2008-2018. Most patients came from districts located on the periphery of

GPNP that signed MOUs to participate in the intervention, but a substantial fraction of patients (42%) came from districts without MOUs, who sometimes traveled many hours or days, to use clinic services (Fig. S3). Clinic affordability (MOU status and associated discounts) and accessibility (estimated travel time to the clinic) jointly influenced two metrics of clinic usage: probability of clinic use, measured as the proportion of a district’s population that used the clinic at least once, and individual patient visitation frequency.

Patients with shorter travel times to the clinic were more likely to use the clinic (Poisson

GLMM with population size offset and district random effect, Estimate= -1.14, SE=0.17, p<0.0001) and visited the clinic more frequently (negative binomial GLMM with district random effect, Estimate= -0.180, SE=0.035, p<0.0001) (Fig. S4). At the same time, controlling for distance, signing of an MOU (and receiving clinic discounts) increased clinic use: a larger proportion of MOU-signing district populations used the clinic (on average 27.8% versus 2.76%, Estimate=1.93, SE=0.36, p<0.0001; Fig. 3A, Fig. S4A), and individual patients from MOU-signing districts visited the clinic 33% more often, on average (2.4 visits over 10 years versus 1.8 visits; negative binomial GLMM with district

105

random effect, Estimate=0.284, SE=0.073, p<0.0001; Fig. 3A, Fig. S4B). Patients that visited more than 2-3 times were usually returning repeatedly for health care related to a chronic health condition, such as epilepsy, emphysema, or hypertension. Overall, the clinic usage statistics confirm that, controlling for distance effects on clinic usage, signing an MOU to participate in the intervention incentivized increased use of health care services at the clinic. Even so, patients without MOUs represented more than 40% of all patient visits (Fig. S5A), likely because non-cash payment options like exchange of tree seedlings, manure, handicrafts, or labor made service affordable.

Time trends in disease outcomes based on diagnoses at the clinic

De-identified diagnosis records from more than 61,000 unique doctor visits recorded during 2008-2018 showed improvements in many health outcomes for MOU and non-MOU patient populations. We found significant declines over time in diagnosed cases of malaria, tuberculosis, childhood-cluster diseases, neglected tropical diseases

(NTDs), chronic obstructive pulmonary disease (COPD), and diabetes (Fig. 3B). The only preventable and treatable diseases considered here that increased over time were lower and upper respiratory infections (Fig. 3B, Fig. S5B). The increase in diagnosed cases of respiratory diseases regionally may have been related to region-wide fire activity that spiked in 2015 (24). Increases in upper respiratory infections might also be due to increased care-seeking for more minor illnesses as trust was built between communities and the program. Time trends in diagnosed cases of disease were consistent whether or not district-level distance to the clinic was included as a covariate (Fig. 3B).

Regional diagnosis records from government clinics were not available to use as controls against which to compare temporal trends at the intervention clinic, but trends in diagnosis of several diseases departed from population-based prevalence estimates

106

published by independent Global Burden of Disease studies for West Kalimantan during the same time period. For example, tuberculosis and COPD showed an upward trend regionally (25), whereas tuberculosis declined strongly in the clinic population in our study from 2008-2018, when the intervention’s health clinic oversaw TB-DOTS treatment (i.e., directly observed treatment short course for tuberculosis) for all regionally-diagnosed patients (including in local government facilities).

Comparing disease outcomes in patients from MOU-signing versus non-MOU districts

We were unable to assess whether time trends in patient diagnoses were attributable to the increased health care access and use available through the clinic

(beyond government care available to all individuals), because it would have been unethical to withhold access to any patient. As a result, we lacked a matched control (or

“no clinic access”) group, but were able to statistically compare time trends in diagnoses among patients from MOU-signing districts versus those in patients from districts without MOUs, to test whether community health outcomes benefited from clinic discounts associated with the intervention. Controlling for the distance between the clinic and patients’ home districts, we found few differences, indicating largely equitable health outcomes in terms of change in the proportion of patient diagnoses across all diseases

(Tables S1-S3). The few exceptions included cases of lower respiratory infections (LRI) and upper respiratory infections (URI), which increased across all patient populations over the ten year study period, but increased significantly less in MOU-signing patient populations (LRI Estimate= -0.499, SE=0.222, p=0.0025; URI Estimate= -0.650,

SE=0.222, p=0.0036), as did cases of dental diseases (Estimate= -0.877, SE=0.167, p<0.0001). In contrast, neglected tropical disease diagnoses increased more in the MOU group than the no-MOU group in the ten-year intervention period (Estimate=0.675,

107

SE=0.253, p=0.0076), a trend largely driven by an increase in leprosy diagnoses in the

MOU group over time (Estimate=0.864, SE=0.354, p=0.015). This may signify true increases in leprosy rates in MOU-signing districts, or may signify increased health seeking behavior for rare and difficult to treat diseases, like leprosy, in MOU-signing populations as compared to non-MOU populations.

Self-reported well-being and livelihood impacts

Household surveys were conducted by the intervention team in 2007, 2012, and

2017 (see Table S4 for survey demographic information). Between 2007 and 2017, annual birth rates and infant death rates declined significantly, and although the measurement method used in this survey is not directly comparable to standardized

USAID Demographic and Health Survey (DHS) methods, these declines are consistent with substantial regional declines apparent in DHS data for the same region (Table S5)

(26, 27). As illegal logging declined as a livelihood in the ten-year intervention period, the decline did not correspond with significant changes in unemployment, as employment increased in other sectors (Fig. 3C). Monthly household income across all surveyed households in all districts was unchanged from 2012 to 2017 (t-test, p=0.28), but after adjusting income for change in Purchasing Power Parity (PPP) (28), median household

PPP-adjusted income decreased by 2.6% (t-test, p=0.001 Fig. 3D). However, in low- income settings, national-level PPP adjustment of income may not accurately represent wealth, which might be better estimated by asset-ownership (29). Additionally, household perceptions of neighborhood wealth were not significantly changed over time

(Fig. 3D).

Dose-response of the intervention’s effect on deforestation

108

We found evidence of a dose-response across 36 villages (‘dusun’ administrative units) with an access area >0.30 km2 inside GPNP, whereby forest loss declined with increasing intervention engagement (engagement was defined as the sum of recorded person-contacts across all intervention activities, including clinic patient visits, forest liaisons meetings, conservation education activities, livelihoods training, and a number of other smaller programs, Fig. S6). Comparing average forest loss rates over time (in 3 time periods: the pre-intervention period in 2002-2006, the first 5 years of the intervention in

2008-2012, and the most recent 5 years of the intervention in 2013-2017), forest loss rates near highly engaged villages decreased significantly (-0.15% ± 0.048%, p=0.007), while forest loss rates near medium engaged villages did not change (0.06% ± .042%, p=0.147), and forest loss rates near the least engaged villages showed an increasing trend (0.16 ±

0.085%, p=0.067; Fig. 2D). There was also a dose-response in the probability that any 30 m2 forested pixel was lost across the entire intervention period: controlling for slope; elevation; distance to nearest river, road, and park edge; and logging pressure (forest loss) outside the park, we found that highly engaged villages’ access areas inside GPNP lost significantly fewer forest pixels compared to that lost in low-engaged villages’ access areas

(Estimate= -0.85, SE=0.013, p<0.0001; Table 2), whereas medium-engaged villages’ access areas lost equivalent forest pixels to low-engaged villages’ access areas (Estimate=

-0.0087, SE=0.012, p=0.46). GPNP forest loss also decreased with average elevation of forest in GPNP access areas (Estimate= -1.83, SE=0.75, p=0.015; Table 2) and increased with logging pressure outside of the park (Estimate=0.11, SE=0.0073, p<0.0001; Table 2).

The dose-response of the intervention effect is consistent with a causal association between

109

the intervention – consisting of expanded health care access and use, plus livelihood, education, and conservation programs – and ultimate deforestation outcomes.

Discussion

Our results offer objective evidence that increasing access to affordable, high quality health care as part of a comprehensive conservation intervention – in this case, to rural communities with limited resources and income options living near a densely- forested national park in Indonesia – benefits both conservation and human health. In addition, community members self-reported that the intervention was working: by 2012, more than 97% of surveyed households surrounding the park indicated that they believed the intervention was reducing illegal logging. Further insight into mechanisms by which the intervention was reducing illegal logging was gained in 2017 via a household survey question asking, “which programs are most helpful” to stop logging in GPNP. Among the subset of households that interacted with intervention programs, roughly half identified health care discounts alone or in combination with other intervention activities

(representing the plurality of responses) as the most important incentive to reduce illegal logging in the park, roughly one quarter identified livelihood programs alone or in combination with other activities (including health care) as most important, while only a few (6%) indicated that the intervention is not effective at reducing illegal logging.

Further investigation is required to establish whether this approach may be effective in other tropical forest parks where high tree cover, high poverty, and lack of access to affordable, high-quality health care fuel illegal logging and forest loss, even within protected areas (30).

Early and continued collaboration with local communities, who identified mechanisms driving linked health–environment problems and potential regional

110

solutions, was essential to the intervention’s multi-sector success. Globally, about 35% of protected areas are traditionally owned, managed, used, or occupied by indigenous and local communities, yet the perspective and guidance of indigenous peoples and local communities are rarely considered in the design of conservation and climate mitigation programs (31). As a result, many interventions have had negative consequences for local communities that rely on natural resources for subsistence (31). Incentive-based conservation approaches, developed to integrate community development and conservation, have had mixed success, as benefits are not always distributed equitably or do not reflect community needs (32). In contrast, we found that community leadership in the design and implementation of a conservation intervention focusing on pressing health and well-being needs resulted in strong positive benefits to local communities as well as to global conservation goals.

This work demonstrates an actionable framework for aligning cross-sectoral goals. Frameworks such as this are urgently needed to advance effective policy efforts aimed at achieving the United Nations’ Sustainable Development Goals (SDGs) (33–35).

Here we evaluated outcomes related to conservation (Life on Land, SGD 15) and health

(Good health and well-being, SDG 3) resulting from an intervention that actually addressed several additional goals, including Climate action (SDG 13), Decent work and economic growth (SDG 8), and Partnerships for the goals (SDG 17). Because the SDGs are deeply interconnected, there is both opportunity and urgency to address multiple targets at once. This intervention offers a case study of how programs can be designed, implemented, and evaluated to address health and conservation goals simultaneously.

The forest carbon results reported here do not include measures of belowground carbon conservation in mineral soils or peatland, the latter of which stores more carbon

111

than aboveground forest biomass in Borneo (36) and is particularly vulnerable to carbon loss and subsidence following deforestation events (37). We also do not include measures of forest regrowth in preserved areas or previously degraded areas being restored through intervention activities (38), which undoubtedly amplified carbon storage and sequestration benefits of the intervention. Furthermore, over the long-term, preserving and restoring forest-related ecosystem services might also benefit human health by reducing the risk of waterborne diarrheal disease (39), lowering heat stress (40), or reducing vectors of malaria and arboviruses (41). Measuring these longer-term effects of ecosystem integrity on human health remains an important goal for future linked conservation and public health interventions.

A more nuanced assessment of how health care–conservation exchange programs influence disease occurrence is another important future direction for research. In the context of this study, clinic health records offered a rich dataset on more than 1250 unique ICD-10 codes detected in the patient population (Table S6). However, ethical and logistical constraints prevented the establishment of a control group for evaluating health care outcomes: denying health care access to certain individuals was antithetical to the intervention’s goal to improve health and well-being, and measuring disease occurrence for hundreds of ICD-10 codes in a group of non-patient individuals was unrealistic.

Therefore, we cannot fully account for the contribution of 10 years of regional improvements in government health systems and infrastructure development. However, by comparing clinic usage and diagnosis trends in MOU-signing patients receiving clinic discounts versus non-MOU signing patients, we established that the intervention incentivized increased health-seeking behavior (with higher clinic usage in MOU-signing districts) and led to potential benefits for common diseases including respiratory

112

infections and dental diseases. In the future, statistical inference on health outcomes of improved clinic access and affordability could be achieved through (i) close collaboration with existing health care facilities that serve distinct patient populations in order to generate comparable datasets, and/or (ii) adherence to pragmatic experimental designs, such as stepped wedge cluster randomized trials, that randomly expose groups to an intervention incrementally, thus generating a varying number of control groups at each time point while eventually exposing all groups to the intervention (42).

Deforestation in tropical rainforests has doubled since 2008 (1). Tropical Asia contains some of the most carbon-dense forests in the world (Fig. S1), and Indonesia has ranked consistently among the top countries for forest loss worldwide, not far behind much larger and wealthier countries including Brazil, Russia, and the USA (1).

Meanwhile, human health and livelihoods are intimately linked to environmental change.

Here we show that amid this challenging context, local community stewards are both critical actors in, and beneficiaries of, integrated conservation and health solutions.

Where health care access is limited and the conservation value of tropical forests is high, reducing rural health care gaps through conservation–health interventions may offer a synergistic means to enhance health and well-being benefits to local communities, while simultaneously conserving critical forest carbon and biodiversity resources.

Materials and Methods

Intervention impact on forest change

We examined the effect of the intervention on rates of forest loss in Gunung

Palung National Park (GPNP) compared to other national parks across Indonesia, using an ex post facto research design (non-randomized control groups designated after the fact) and a synthetic controls analysis, with nearly two decades of earth observation data

113

quantifying forest change. ‘Annual forest loss’ and ‘total forest cover’ were downloaded from the Hansen Global Forest Change dataset (1) (version 1.6, 2000-2018; accessible through the Google Earth Engine data repository (43), 30m pixel resolution). Area of forest lost by year was extracted by district-level administrative unit (‘desa’) in Google

Earth Engine (GEE) for all districts whose boundaries intersected Gunung Palung

National Park (GPNP) and all other Indonesian national park boundaries.

We used time varying and time invariant characteristics to match treated units

(districts in GPNP) to untreated units (districts in other parks) to assemble the weighted synthetic control group. Time varying characteristics included forest lost inside and outside park boundaries, total forest inside and outside park boundaries, forest fires inside and outside park boundaries, and human population density inside and outside park boundaries. Time invariant characteristics included area of the district inside and outside park boundaries, area of the focal park, marine area of the focal park (to capture information indicating a coast adjacent or a primarily marine park), year established as a

National Park, and average slope within the park (to capture ease of logging access).

Total forest cover estimates inside national park boundaries were included to represent total forest available to log, and total forest cover estimates outside national park boundaries were included as a proxy for logging pressure outside the park.

To estimate changes in human population density during the evaluation period, we extracted population density estimates from WorldPop (www.worldpop.org) by district and year from 2000-2018, both within and adjacent to park boundaries, using the

Google Earth Engine platform (43). The effect of forest fires was controlled for using the

MODIS Burned Area Monthly Global data product (500m), which provides the burn

114

status of each 500m pixel at a monthly resolution. The park characteristics were acquired from the World Database on Protected Areas (12).

We ran synthetic controls models using three different data subsets defining the

“donor pool” of possible control units. These subsets included (i) all districts in terrestrial

(non-marine) parks established before 2001 (i.e., dropping any parks that are designated as marine only parks and those that were established after the start of the deforestation dataset in 2001); (ii) all districts in all terrestrial parks (i.e., dropping only entirely marine parks, but ignoring year of establishment); and (iii) all districts in all National Parks in

Indonesia (i.e., the most inclusive group of possible control districts in National Parks).

Models were run using annual data using 2001-2007 as the pre-intervention period (since the intervention was not expected to lead to immediate changes in deforestation rates in mid-2007, when the intervention started), and 2008-2018 as the post-intervention period.

In each model, p-values and 90% confidence intervals (for a one-tailed lower test) were calculated using a standard normal sampling distribution and Taylor series linearization to estimate the variance and produce confidence intervals (21). In addition, p-values and confidence intervals were also calculated using 500 permuted “placebo” treatment groups for comparison with the estimated effect for the actual treatment group to satisfy a more robust set of assumptions and to generate a more robust and conservative estimate of the sampling distribution (21). These “permutations” are placebo tests in that they randomly assign districts in the “control” group to the placebo treatment group, the synthetic controls model is re-run, and the magnitude of the placebo treatment result is compared to the actual treatment group result (21). All models were run using the ‘microsynth’ package in R (44, 45) following established methods outlined in

Robbins et al. (21).

115

Intervention impact on forest carbon

We evaluated the intervention impact in terms of forest loss rates, and then estimated the quantity of forest carbon conserved from the quantity of forest loss averted estimated by the synthetic controls analysis. Following Jucker et al. (22), we estimated forest carbon stocks in GPNP using canopy heights derived by LiDAR (Light Detection and Ranging), accessed from NASA and Oak Ridge National Laboratory (ORNL)

Distributed Active Archive Center (DAAC) (Fig. S2, see SI text for conversion equations and details) (46). Then we used the effect size of 69.8% forest-loss-rate reduction to estimate the total carbon stock (in Tg C ha-1) conserved in the period from 2008 to 2018.

This method is conservative, as using the mean pixel height of the LiDAR flight is an underestimation of the average TCH of trees targeted by illegal loggers, who target the largest and most valuable trees. Even so, carbon densities were high, in part, because this region of Southeast Asia contains forests with some of the largest and most carbon-dense trees in the world (Fig. S1).

Intervention impact on health clinic usage and diagnoses

The ASRI medical center opened in July 2007 and remains open. Our analyses consider the period from 2008-2018, beginning with the first full year of data and ending with the last full year of data before comprehensive evaluation began. For all patients who visited the ASRI medical center, patient records included only a unique ID to maintain patient privacy, the date of the visit, the patient’s home village (‘dusun’ administrative unit) and district (‘desa’ administrative unit), the patient’s age, and the

116

diagnosis (ICD10 code) given during the visit and/or the reason for visiting (e.g., medical check-up, fill prescription).

We were interested in understanding how increases in clinic affordability

(discounts provided to patients from MOU-signing villages/districts) impacted clinic usage, while controlling for clinic accessibility (estimated mean travel time from the patients’ district to the clinic). We quantified clinic usage at the district-level in two ways: (i) the frequency of patient visits, defined as the number of visits per patient (i.e., the number of unique months that a patient occurred in the database, see SI text for details); and (ii) the proportion of the population in each district that used the clinic at least once during the evaluation period. Clinic access was estimated for each district as the mean travel time (in minutes) from 10 randomly distributed points in each district, weighted by the point’s population density (see SI text for details).

To understand how patient-level visit frequency was affected by affordability

(MOU status) and access (travel time), we ran a negative binomial generalized linear mixed model with a random effect for district that quantified how the number of visits per patient varied with MOU status and the estimated travel time (in minutes) from the patient’s district to the clinic. Next we used a Poisson generalized linear model to quantify how the number of unique patients per district varied with MOU status and estimated travel time to the clinic, where the 2018 population size in each district was included as an offset. For this analysis, we excluded patients from unknown districts

(recorded only as “far” in the patient records) because district-level population size for this group of patients could not be determined. For both clinic usage analyses, we were unable to estimate travel time for one island district (Pelapis), and therefore excluded it

117

from analysis. Details on how travel time was estimated for the “far” (Fig. S3) districts are available in SI.

Overall, 1,255 unique ICD10 codes (47) were applied to at least one patient between 2008 and 2018. Before analyzing how the proportion of unique patients that received a diagnosis (ICD10 code) of a particular disease changed over the intervention period, we classified 824 ICD10 codes that contributed to the most common disease categories (see SI text for details) into the following groups to be tracked: childhood- cluster diseases, chronic obstructive pulmonary disease (COPD), dental disease, diabetes, diarrheal diseases, heart disease, liver disease, lower respiratory infections, upper respiratory infections, malaria, malnutrition, neglected tropical diseases (NTDs), trauma, and tuberculosis. Other unspecified diseases were grouped and appear in the text and figures as “untracked” (see Table S6 for a full list of the 824 ICD10 codes tracked), and the unique patients to which the untracked ICD10 codes were assigned are included in the total patient population.

To track changes in disease occurrence in the patient population, we estimated the proportion of unique patients that received a diagnosis of for each disease in each district, annually. Each patient only counts towards one instance of any particular disease per year, and the denominator is the total unique patients that received any diagnosis in a year. For each disease, we used a binomial generalized linear model with a logit link

(which yields coefficients that can be exponentiated to derive odds ratios) and a random effect for district to quantify differences in the proportion of disease diagnoses over time

(early, 2008-2009 vs. late, 2017-2018) for two populations: patients from districts with and without MOUs. The model was run with and without controlling for district distance

118

to the clinic, with nearly identical outcomes. The full time series showing changes in the period prevalence of each disease in the patient population are shown in Fig. S5.

We also tested whether MOU status impacted disease outcomes among patients over time, while controlling for a patient’s average distance (in minutes) from district to clinic (see SI for details on calculating travel time). To do so, we used binomial generalized linear models with a logit link, using probability of a diagnosis of each disease as the outcome, and including scaled travel time and an interaction term for time by MOU status as predictors, and a random effect for district to control for repeated unmeasured variables.

Self-reported well-being and livelihood impacts

At baseline in 2007 and at follow ups in 2012 and 2017, detailed household surveys were conducted in all districts surrounding the national park (see SI text for details on surveyor selection and training). Within each district, a list of all households in the villages was provided by village heads and from that list, ~10% of households (and, correspondingly, ~10% of the total population of ~60,000 people) were randomly selected for participation. In total, 1,348, 1,498, and 1,379 households were surveyed in

2007, 2012, and 2017, respectively (see Table S4 for details on the surveyed household demographics). At each timepoint, the survey instrument contained modules for: demography (age, sex, births, etc.), health, wealth, and perceptions of wealth (income, designation of the household or neighborhood as ‘wealthy’, ‘average’, or ‘poor’), livelihoods (including logging activity and other occupations), and perceptions of nature, natural resources, conservation, and the intervention.

119

From demographic characteristics, including ages, gender, births and deaths in surveyed households, we calculated average annual infant mortality rates, defined as average annual infant deaths per 1,000 live births among household women in the 3-year period preceding the survey, and average annual births, defined as average annual births per household women ages 19-59 (Table S5). For reference, we also extracted infant mortality rates (IMR) and general fertility rates (GFR) in a 5-year period from USAID

Demographic and Health Survey data for Indonesia in 2007 and 2017 (26, 27). We report

IMR and GFR for the province of West Kalimantan and for ‘rural’ West Kalimantan, which we expect to be a more accurate representation of the rural communities in and around GPNP (Table S5). From reported incomes and household perceptions of neighborhood wealth, we calculated the average monthly income at the household level and the proportion of households that felt neighborhood wealth had increased, decreased, or remained the same at each timepoint. We also calculated the proportions of households that reported members engaged in various livelihoods (logger, fisher, farmer, civil servant, company employee, unemployed, and ‘other’). Last, to infer mechanisms by which the intervention may have reduced illegal logging rates, we calculated household responses to a 2017 survey question that specifically asked what intervention programs provide the strongest incentive to help stop logging inside GPNP.

For reported livelihoods and perceptions of neighborhood wealth, we quantified change over time using generalized linear models with binomial error distributions and logit links. For livelihoods, we compute the percent change the proportion of households reporting each livelihood as a primary income source from the log odds that it livelihood changed over time. We used t-tests to test for a change in average monthly income and

Purchasing Power Parity (PPP)-adjusted monthly income over time. Due to slight

120

variations in survey question wording only 5-year comparison data (2012 vs 2017) were available for monthly incomes and perceptions of neighborhood wealth.

Dose-response of the intervention’s effect on deforestation within GPNP After demonstrating a significant correlation between the intervention and pre- to post-intervention forest loss trajectories in GPNP compared to a synthetic control, we tested whether there was any evidence of a dose-response relationship within GPNP, among villages (‘dusun’) with varying levels of engagement with the intervention programs (including use of the health clinic) and forest loss rates. To answer this question, we quantified engagement effort as cumulative person-contacts (i.e., number of contacts with persons reached by all program activities associated with the intervention from 2007 to 2017, allowing for repeated contacts with the same individuals over time) achieved through: the health care intervention (ASRI clinic visits, mosquito net distribution), conservation programs (Community Conservation Liaisons or “Forest

Guardian” Program, Chainsaw Buyback Program, and Reforestation Program), alternative livelihood trainings (Organic Agriculture Program, Goats for Widows

Program, Green Kitchen Program), and education activities (ASRI Kids Program,

Community Education Program) (Fig. S6). Engagement effort was not distributed evenly across all villages and was predominated by frequent engagement with community liaisons for the intervention as well as doctor-patient contacts at the clinic (Fig. S6).

Variation in engagement across the participating villages intersecting GPNP allowed us to test for evidence of a dose-response of intervention effort on deforestation within different access areas nearest each village around the park.

We used a k-means clustering algorithm to bin engagement (cumulative person- contacts in each village across all of the intervention programs from 2007 to 2018 (Fig.

121

S6) into ‘low’, ‘medium’, and ‘high’ categories (R package ‘classInt’ (48)). We examined the effect of cumulative engagement effort on the proportion of forest lost in each village’s access area in the national park (number of remotely sensed 30m2 pixels lost out of total forested pixels remaining). Village-level access areas inside GPNP were determined by a local team that mapped the parts of each village that extended into

GPNP and represented that village’s typical access area for illegal logging. Ultimately, 36 villages bordering GPNP with logging access areas >0.30km2 were included in the analysis. As in the synthetic controls analysis, forest change data was obtained from the

Hansen Global Forest Change dataset (version 1.6, 2000-2018, 30m pixel resolution (1)).

Total forest cover by village was estimated annually by subtracting forest loss in each subsequent year from total remaining forest cover in the previous year (e.g. forest cover in 2002 = forest cover in 2000 – forest loss in 2001 – forest loss in 2002).

Changes in forest loss rates associated with engagement were estimated in two ways. First, for each engagement category (‘high’, ‘medium’, and ‘low’) we estimated change in average annual forest loss rates over time, from before the intervention (2002 to 2006) to forest loss rates during the first 5 years (2007-2012) to the most recent 5 years

(2013-2017) using a mixed effects linear model with time as a predictor and controlling for and nested random effects of village within district repeated over time. Next, we fit a mixed effects linear model with binomial error distribution and a logit link to estimate the probability that, in the last 5-year period of the intervention as compared to the 5-year period before the intervention, any 30 m2 forested pixel in a medium- or highly-engaged village’s access area within GPNP was lost, compared to the probability that any 30m2 forested pixel was lost in a low-engaged village’s access area. We controlled for village population size (supplied by village leaders in 2017), proportion of forest lost within

122

village boundaries outside of the park (as a proxy for outside logging pressure), average slope, aspect, and elevation of pixels inside the park, average distance of pixels to the nearest river, road, and park boundary, and nested random effects of village within district. We did not include data on fires because fire activity was relatively low in GPNP during the time period under consideration.

123

Acknowledgments

We express gratitude to the communities near and around Gunung Palung National Park that helped design and implement and continue to participate in the ongoing intervention. We thank the dedicated medical and conservation staff and volunteers at Alam Sehat Lestari (ASRI) and Health In Harmony, as well as patients that visited the clinic. We thank the local and national Indonesian government for collaboration on all aspects of this work, including the Gunung Palung National Park staff, the Kayong Utara Regency government, and the Department of Health. We thank Campbell Webb for early contributions on methodology and data interpretation, Erin Mordecai and Steve Palumbi for feedback, and this manuscript’s Editor and Reviewers for helpful suggestions. Lastly, we thank Made By We for assisting in the design of figures. Funding: IJJ was supported by a National Science Foundation Graduate Research Fellowship (#1656518) and a Seed Grant from the Center for Innovation and Global Health of Stanford University; AJM was supported by a National Science Foundation Postdoctoral Research Fellowship in Biology (#1611767); AJL was supported by a James and Nancy Kelso Fellowship through the Stanford Interdisciplinary Graduate Fellowship (SIGF) program at Stanford University; MN was supported by the Fulbright Foreign Student Program; GADL and SHS were partially supported by a Stanford University FSI-SEED Global Poverty and Development Initiative grant and a 2019 Stanford University CIGH seed grant. GADL, SHS and IJJ were supported by a SNAPP-NCEAS working group. SHS, KW, DLC, AE, and GDL were supported by a UC Global Health Institute Planetary Health Center for Expertise seed grant.

124

References

1. M. C. Hansen, P. V. Potapov, R. Moore, M. Hancher, S. A. Turubanova, A. Tyukavina, D. Thau, S. V. Stehman, S. J. Goetz, T. R. Loveland, A. Kommareddy, A. Egorov, L. Chini, C. O. Justice, J. R. G. Townshend, High-Resolution Global Maps of 21st-Century Forest Cover Change. Science. 342, 850–853 (2013).

2. A. Baccini, W. Walker, L. Carvalho, M. Farina, D. Sulla-Menashe, R. A. Houghton, Tropical forests are a net carbon source based on aboveground measurements of gain and loss. Science. 358, 230–234 (2017).

3. B. W. Griscom, J. Adams, P. W. Ellis, R. A. Houghton, G. Lomax, D. A. Miteva, W. H. Schlesinger, D. Shoch, J. V. Siikamäki, P. Smith, P. Woodbury, C. Zganjar, A. Blackman, J. Campari, R. T. Conant, C. Delgado, P. Elias, T. Gopalakrishna, M. R. Hamsik, M. Herrero, J. Kiesecker, E. Landis, L. Laestadius, S. M. Leavitt, S. Minnemeyer, S. Polasky, P. Potapov, F. E. Putz, J. Sanderman, M. Silvius, E. Wollenberg, J. Fargione, Natural climate solutions. PNAS. 114, 11645–11650 (2017).

4. Intergovernmental Panel on Climate Change, Global warming of 1.5°C (2018; http://www.ipcc.ch/report/sr15/).

6. William. M. Adams, Biodiversity Conservation and the Eradication of Poverty. Science. 306, 1146–1149 (2004).

7. C. B. Barrett, A. J. Travis, P. Dasgupta, On biodiversity conservation and poverty traps. Proceedings of the National Academy of Sciences. 108, 13907–13912 (2011).

8. A. Garchitorena, S. H. Sokolow, B. Roche, C. N. Ngonghala, M. Jocque, A. Lund, M. Barry, E. A. Mordecai, G. C. Daily, J. H. Jones, J. R. Andrews, E. Bendavid, S. P. Luby, A. D. LaBeaud, K. Seetah, J. F. Guégan, M. H. Bonds, G. A. De Leo, Disease ecology, health and the environment: a framework to account for ecological and socio-economic drivers in the control of neglected tropical diseases. Philos Trans R Soc Lond B Biol Sci. 372 (2017), doi:10.1098/rstb.2016.0128.

9. J. A. Lutz, T. J. Furniss, D. J. Johnson, S. J. Davies, D. Allen, A. Alonso, K. J. Anderson‐Teixeira, A. Andrade, J. Baltzer, K. M. L. Becker, E. M. Blomdahl, N. A. Bourg, S. Bunyavejchewin, D. F. R. P. Burslem, C. A. Cansler, K. Cao, M. Cao, D. Cárdenas, L.-W. Chang, K.-J. Chao, W.-C. Chao, J.-M. Chiang, C. Chu, G. B. Chuyong, K. Clay, R. Condit, S. Cordell, H. S. Dattaraja, A. Duque, C. E. N. Ewango, G. A. Fischer, C. Fletcher, J. A. Freund, C. Giardina, S. J. Germain, G. S. Gilbert, Z. Hao, T. Hart, B. C. H. Hau, F. He, A. Hector, R. W. Howe, C.-F. Hsieh,

125

Y.-H. Hu, S. P. Hubbell, F. M. Inman‐Narahari, A. Itoh, D. Janík, A. R. Kassim, D. Kenfack, L. Korte, K. Král, A. J. Larson, Y. Li, Y. Lin, S. Liu, S. Lum, K. Ma, J.-R. Makana, Y. Malhi, S. M. McMahon, W. J. McShea, H. R. Memiaghe, X. Mi, M. Morecroft, P. M. Musili, J. A. Myers, V. Novotny, A. de Oliveira, P. Ong, D. A. Orwig, R. Ostertag, G. G. Parker, R. Patankar, R. P. Phillips, G. Reynolds, L. Sack, G.-Z. M. Song, S.-H. Su, R. Sukumar, I.-F. Sun, H. S. Suresh, M. E. Swanson, S. Tan, D. W. Thomas, J. Thompson, M. Uriarte, R. Valencia, A. Vicentini, T. Vrška, X. Wang, G. D. Weiblen, A. Wolf, S.-H. Wu, H. Xu, T. Yamakura, S. Yap, J. K. Zimmerman, Global importance of large-diameter trees. Global Ecology and Biogeography. 27, 849–864 (2018).

10. J. Busch, K. Ferretti-Gallon, J. Engelmann, M. Wright, K. G. Austin, F. Stolle, S. Turubanova, P. V. Potapov, B. Margono, M. C. Hansen, A. Baccini, Reductions in emissions from deforestation from Indonesia’s moratorium on new oil palm, timber, and logging concessions. Proceedings of the National Academy of Sciences. 112, 1328–1333 (2015).

11. S. Lawson, L. MacFaul, Royal Institute of International Affairs, Illegal logging and related trade: indicators of the global response (Chatham House, London, 2010).

12. UNEP-WCMC, Protected Area Profile for Indonesia from the World Database of Protected Areas, July 2019 (2019), (available at www.protectedplanet.net).

13. N. S. Sodhi, L. P. Koh, B. W. Brook, P. K. L. Ng, Southeast Asian biodiversity: an impending disaster. Trends in Ecology & Evolution. 19, 654–660 (2004).

14. D. L. A. Gaveau, M. Kshatriya, D. Sheil, S. Sloan, E. Molidena, A. Wijaya, S. Wich, M. Ancrenaz, M. Hansen, M. Broich, M. R. Guariguata, P. Pacheco, P. Potapov, S. Turubanova, E. Meijaard, Reconciling Forest Conservation and Logging in Indonesian Borneo. PLOS ONE. 8, e69887 (2013).

15. L. M. Curran, S. N. Trigg, A. K. McDonald, D. Astiani, Y. M. Hardiono, P. Siregar, I. Caniago, E. Kasischke, Lowland Forest Loss in Protected Areas of Indonesian Borneo. Science. 303, 1000–1003 (2004).

16. B. A. Margono, P. V. Potapov, S. Turubanova, F. Stolle, M. C. Hansen, Primary forest cover loss in Indonesia over 2000–2012. Nature Clim Change. 4, 730–735 (2014).

17. K. Webb, J. Jennings, D. Minovi, A community-based approach integrating conservation, livelihoods, and health care in Indonesian Borneo. The Lancet Planetary Health. 2, S26 (2018).

18. Global Burden of Disease Collaborative Network, “Global Burden of Disease Study 2017 (GBD 2017) Health-related Sustainable Development Goals (SDG) Indicators 1990-2030” (Institute for Health Metrics and Evaluation (IHME), Seattle, United States, 2018).

126

19. World Health Organization, State of health inequality: Indonesia (2017).

20. A. Abadie, A. Diamond, J. Hainmueller, Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California’s Tobacco Control Program. Journal of the American Statistical Association. 105, 493–505 (2010).

21. M. W. Robbins, J. Saunders, B. Kilmer, A Framework for Synthetic Control Methods With High-Dimensional, Micro-Level Data: Evaluating a Neighborhood- Specific Crime Intervention. Journal of the American Statistical Association. 112, 109–126 (2017).

22. T. Jucker, G. P. Asner, M. Dalponte, P. G. Brodrick, C. D. Philipson, N. R. Vaughn, Y. A. Teh, C. Brelsford, D. F. R. P. Burslem, N. J. Deere, R. M. Ewers, J. Kvasnica, S. L. Lewis, Y. Malhi, S. Milne, R. Nilus, M. Pfeifer, O. L. Phillips, L. Qie, N. Renneboog, G. Reynolds, T. Riutta, M. J. Struebig, M. Svátek, E. C. Turner, D. A. Coomes, Estimating aboveground carbon density and its uncertainty in Borneo’s structurally complex tropical forests using airborne laser scanning. Biogeosciences. 15, 3811–3830 (2018).

23. D. Burtraw, M. Themann, Pricing Carbon Effectively: Lessons from the European Emissions Trading System, 24.

24. P. Crippa, S. Castruccio, S. Archer-Nicholls, G. B. Lebron, M. Kuwata, A. Thota, S. Sumin, E. Butt, C. Wiedinmyer, D. V. Spracklen, Population exposure to hazardous air quality due to the 2015 fires in Equatorial Asia. Sci Rep. 6, 37074 (2016).

25. Global Burden of Disease Collaborative Network, “Global Burden of Disease Study 2017 (GBD 2017) Disability-Adjusted Life Years and Healthy Life Expectancy 1990-2017” (Institute for Health Metrics and Evaluation (IHME), Seattle, United States, 2018).

26. “Statistics Indonesia - Badan Pusat Statistik - BPS - and Macro International. 2008. Indonesia Demographic and Health Survey 2007. IDIR51FL.DTA, IDBR51FL.DTA. Calverton, Maryland, USA: BPS and Macro International. ICF [Distributor], 2008.”

27. “National Population and Family Planning Board - BKKBN, Statistics Indonesia - BPS, Ministry of Health - Kemenkes, and ICF. 2018. Indonesia Demographic and Health Survey 2017. IDIR71FL.DTA, IDBR71FL.DTA. Jakarta, Indonesia: BKKBN, BPS, Kemenkes, and ICF. ICF [Distributor], 2018.”

28. International Comparison Program (ICP). World Bank, (available at https://www.worldbank.org/en/programs/icp).

29. L. D. Howe, J. R. Hargreaves, S. R. Huttly, Issues in the construction of wealth indices for the measurement of socio-economic position in low-income countries. Emerg Themes Epidemiol. 5, 3 (2008).

127

30. B. Fisher, T. Christopher, Poverty and biodiversity: Measuring the overlap of human poverty and the biodiversity hotspots. Ecological Economics. 62, 93–101 (2007).

31. IPBES, “IPBES Assessment Guide Summary” (Secretariat of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services, Bonn, Germany, 2018), p. 12.

32. A. Spiteri, S. K. Nepalz, Incentive-Based Conservation Programs in Developing Countries: A Review of Some Key Issues and Suggestions for Improvements. Environmental Management. 37, 1–14 (2006).

33. W. Rosa, Ed., in A New Era in Global Health (Springer Publishing Company, New York, NY, 2017; http://connect.springerpub.com/lookup/doi/10.1891/9780826190123.ap02).

34. S. Whitmee, A. Haines, C. Beyrer, F. Boltz, A. G. Capon, B. F. de Souza Dias, A. Ezeh, H. Frumkin, P. Gong, P. Head, R. Horton, G. M. Mace, R. Marten, S. S. Myers, S. Nishtar, S. A. Osofsky, S. K. Pattanayak, M. J. Pongsiri, C. Romanelli, A. Soucat, J. Vega, D. Yach, Safeguarding human health in the Anthropocene epoch: report of The Rockefeller Foundation–Lancet Commission on planetary health. The Lancet. 386, 1973–2028 (2015).

35. IPBES, “Summary for policymakers of the global assessment report on biodiversity and ecosystem services of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services” (IPBES Secretariat, Bonn, Germany, 2019), p. 60.

36. M. Warren, K. Hergoualc’h, J. B. Kauffman, D. Murdiyarso, R. Kolka, An appraisal of Indonesia’s immense peat carbon stock using national peatland maps: uncertainties and potential losses from conversion. Carbon Balance and Management. 12, 12 (2017).

37. A. M. Hoyt, E. Chaussard, S. S. Seppalainen, C. F. Harvey, Widespread subsidence and carbon emissions across Southeast Asian peatlands. Nature Geoscience. 13, 435–440 (2020).

38. N. I. Fawzi, V. N. Husna, J. A. Helms, Measuring deforestation using remote sensing and its implication for conservation in Gunung Palung National Park, West Kalimantan, Indonesia. IOP Conference Series: Earth and Environmental Science. 149, 012038 (2018).

39. D. Herrera, A. Ellis, B. Fisher, C. D. Golden, K. Johnson, M. Mulligan, A. Pfaff, T. Treuer, T. H. Ricketts, Upstream watershed condition predicts rural children’s health across 35 developing countries. Nat Commun. 8, 811 (2017).

40. Y. J. Masuda, B. Castro, I. Aggraeni, N. H. Wolff, K. Ebi, T. Garg, E. T. Game, J. Krenz, J. Spector, How are healthy, working populations affected by increasing

128

temperatures in the tropics? Implications for climate change adaptation policies. Global Environmental Change. 56, 29–40 (2019).

41. N. D. Burkett-Cadena, A. Y. Vittor, Deforestation and vector-borne disease: Forest conversion favors important mosquito vectors of human pathogens. Basic and Applied Ecology. 26, 101–110 (2018).

42. K. Hemming, T. P. Haines, P. J. Chilton, A. J. Girling, R. J. Lilford, The stepped wedge cluster randomised trial: rationale, design, analysis, and reporting. BMJ. 350, h391–h391 (2015).

43. N. Gorelick, M. Hancher, M. Dixon, S. Ilyushchenko, D. Thau, R. Moore, Google Earth Engine: Planetary-scale geospatial analysis for everyone. Remote Sensing of Environment. 202, 18–27 (2017).

44. M. Robbins, S. Davenport, microsynth: Synthetic Control Methods with Micro- And Meso-Level Data (2019; https://CRAN.R-project.org/package=microsynth).

45. R Core Team, R: The R Project for Statistical Computing (R Foundation for Statistical Computing, Vienna, Austria, 2018; https://www.r-project.org/).

46. L. Melendy, S. C. Hagen, F. B. Sullivan, T. R. H. Pearson, S. M. Walker, P. Ellis, Kustiyo, A. K. Sambodo, O. Roswintiarti, M. A. Hanson, A. W. Klassen, M. W. Palace, B. H. Braswell, G. M. Delgado, Automated method for measuring the extent of selective logging damage with airborne LiDAR data. ISPRS Journal of Photogrammetry and Remote Sensing. 139, 228–240 (2018).

47. World Health Organization, International statistical classification of diseases and related health problems. (World Health Organization, Geneva, Fifth., 2016), vol. 1.

48. R. Bivand, classInt: Choose Univariate Class Intervals (2018; https://CRAN.R- project.org/package=classInt).

129

Figures and Tables

Deforestation Trends in Terrestrial National Parks B of Indonesia (2001-2018) Key 100% Low health care access (≤30th percentile WHO UHC Index) High forest carbon density (≥100 Mg C/ha) 90% Both low access to health care and high carbon density Study Region 80% Remaining in 2000 out of Forest Cover Forest of out Percent Forest Loss Forest Percent 20002005 2010 2015 Indonesia National Parks C Study Area

Low access to Illegal logging quality, afford- PERPETUATES in carbon-rich

PROBLEM able health care tropical forest parks

Engaging communities identifies most pressing needs and creates a shared goal to reduce illegal logging ENGAGEMENT

Providing quality Reduced health care INCENTIVIZES illegal logging and alternative activities

HYPOTHESIS livelihoods

Local Community benefits from affordable health care and livelihood improvement Global Community benefits from preservation of forests and OUTCOMES above ground carbon

Figure 1. Cross sector global health and forest conservation needs. A) Maps of global forest above ground carbon density and the WHO Universal Health Coverage: tropical areas, particularly Africa and Asia have high forest cover and low health care coverage; inset: B) Forest loss (resulting from deforestation and forest degradation) accelerates over time across all 32 terrestrial IUCN Category II National Parks (12) established before 2001 in Indonesia (boxplots; forest change data: Hansen et al. 2013 (1)) ; C) Study site and approach: locations of IUCN Category II National Parks (12) in Indonesia, with the intervention park highlighted, and an outline of the problems and hypotheses, addressed in this analysis, along with hypothesized outcomes that were tested empirically through objective earth observation health clinic records.

130

Difference between groups A Treatment (GPNP) B *** Synthetic Control 7.8% 10 20

1% % men >19yrs old logging in GPNP in logging 2007 2017

0 (sqkm/year) (sqkm/year) Average annual forest loss rates Average Average annual forest loss rates Average

0 -20 2001 2018 2001 2018 Years Years pre-intervention pre-intervention

80 1,000 MgC/ha 70m N.S. *** Tallest recorded canopy height in GPNP 30% Increase Decrease

54m Mean canopy height of tallest tropical tree species 0%

27m Mean canopy height

in GPNP1MgC/ha intervention pre-to-post -20%

Tree canopy height (m) Tree 9m Low Medium High (3-2700 (2701-7470 (7471-12112

Typical oil palm 0 rates deforestation in change Percent person person person (15 yr plantation) 0 1,000 contacts) contacts) contacts) Carbon storage capacity Engagement effort (MgC/ha) (person-contacts)

Figure 2. Climate impacts. A) As the number of adult men (age 19 or older) who report logging as a primary livelihood declined between 2007 and 2017 based on survey data (inset), a synthetic controls analysis using remotely sensed earth observation data on forest change verified that forest loss rates in Gunung Palung National Park (green solid line, GPNP, the intervention park) were significantly lower than a synthetic control from which the counterfactual (black dashed line) for forest loss after the onset and ramp up of the intervention in GPNP (red vertical dashed line) was estimated; B) The difference between forest loss in GPNP (green) and the 500 “placebo” synthetic control treatments (gray) made up of random permutations of the sampling units – the dotted black line on the x-axis represents no difference between treatment and placebo groups; C) Forest loss rates were converted to estimates of above ground carbon biomass preserved, using average tree height calculated from a publicly available LiDAR-derived dataset and locally-calibrated wood density equations (see Methods for details); D) Quantitative outcomes showing a dose-response in forest loss rates in Gunung Palung National Park to intervention effort: Engagement (see Methods for details) was binned into low-, medium- , and high-engagement categories based on person-contacts across many intervention activities in each village, for 36 villages bordering GPNP with signed MOUs; changes in average forest loss rates (+/- 1 SE) from the 5 year interval before the intervention (2002- 2006) to the last 5 years of the intervention (2013-2017). • p < 0.10; *** p < 0.001; N.S.: not significant.

131

Patient visitation frequency ABHealth care access and use *** Malaria MOU Infectious Average population % of total no MOU Tuberculosis Childhood-cluster diseases NTDs 2 *** Diarrheal disease 40% Lower respiratory infection Upper respiratory infection COPD Non-infectious Diabetes Trauma

Health Impact Health Malnutrition Heart disease

Average visits perpatient Average Liver disease 0 0 Dental disease YesNo Yes No 0.01 0.1 1 5 20 MOU? Odds Ratio

Perception of Household monthly C D neighborhood income (PPP-adjusted 100 wealth USD)

*** N.S. 100% *** 0 2012 *** 2017 Livelihood impact % change 2007 to 2017 2007 to % change

-100 *** 0 2012 2017 $55 $403 $2981

fisher oyed other farmer logging

company employee civil servant unempl Job Category Years

Figure 3. Health impacts. A) Individual visitation frequency (left, average visits/patient to the health clinic during the study period) and health care use (right, measured as the percent of the district population that were recorded at least once during the study period as patients at the clinic), among patients from districts that signed an MOU and thus received discounts on care, and those that did not; partial responses to MOU status are shown after controlling for distance effects (travel time to the clinic); B) Change in odds of disease diagnoses from clinic patient records (presented as odds ratios for MOU and non-MOU patient populations (controlling for distance effects), comparing odds of diagnosis in 2008-2009 vs 2017-2018 with 95% CIs, see Methods); C) Change in primary livelihoods including self-reported logging (proportion of households, 95% CIs) from 2007 to 2017; D) Change in reported perceptions of neighborhood wealth (left, where most responses are “average”) and mean purchasing-power-parity-(PPP)-adjusted household monthly incomes (right), as reported from household surveys at 5 year and 10 year follow up periods (2012 vs 2017). N.S.: not significant, ***p < 0.001.

132

Table 1. Results from the synthetic control analyses on park-level forest loss in Gunung Palung National Park compared to a counterfactual derived from 3 subsets of Indonesian IUCN Category II National Park controls: all non-marine parks established prior to 2001, all non-marine parks, and all parks. The first two columns provide estimates of forest loss (km2) in the treated region following the intervention and loss in the synthetic control region. p-values and confidence intervals are calculated from a standard normal sampling distribution and Taylor series linearization. A permuted p-value and confidence interval was calculated using 500 permuted “placebo” treatment groups to satisfy a more robust set of assumptions and generate a more conservative estimate of the sampling distribution (Robbins et al. 2017 (21)). In both cases, the confidence intervals do not contain 0, and based on a lower-tailed, one-sided hypothesis test, the null hypothesis that there is no intervention effect is rejected (Robbins et al. 2017 (21)).

Forest Forest Permuted Loss, Loss, % p-value N. N. N. Model p-value Treated Control Change [90% CI] Obs. District Parks [90% CI] (km2) (km2)

0.038 Non-marine 0.003 Parks, Est. 11.891 39.30 -69.75 27,702 1,539 32 [-83.7, - before 2001 [-81.4, -50.8] 26.3]

0.013 0.062 Non-marine 11.891 28.36 -58.1 36,738 2,041 44 Parks [-74.0 -32.4] [-78.3, -1.6]

0.013 0.080 All Parks 11.891 28.36 -58.1 40,320 2,240 52 [-74.0, -32.4] [-80.6, 0]

133

Table 2. Dose-response of forest change to the intervention: results of a linear mixed effects regression of forest loss within Gunung Palung National Park over time and the effect of engagement level of each village with the intervention’s programs and activities (see Fig. S6 for details on engagement activities and quantification of engagement levels). Log-odds is presented for centered and scaled predictors. The effect of interest is the interaction of engagement level with year, with log-odds estimates representing the outcome in villages with that engagement level as compared to outcomes in low- engagement villages (as a comparison group). Bolded p-values represent statistically significant factors.

Log-Odds CI p Intercept -0.12 -8.40 – 8.16 0.977

Population -0.47 -1.15 – 0.20 0.171 Forest Lost Outside 0.11 0.09 – 0.12 <0.001

Average elevation -1.83 -3.31 – -0.36 0.015

Average slope 1.7 -0.68 – 4.08 0.162

Distance to Nearest river 0.47 -0.48 – 1.42 0.335

Distance to nearest road -0.12 -1.00 – 0.76 0.792

Distance to park edge 0.03 -0.60 – 0.65 0.936

Medium engagement -0.02 -0.88 – 0.83 0.955

High engagement 0.80 -0.14 – 1.74 0.096

Year 0.34 0.32 – 0.35 <0.001

Interaction terms estimating engagement effect:

Medium engagement*Year -0.01 -0.03 – 0.01 0.456

High engagement*Year -0.85 -0.88 – -0.83 <0.001 Random effects σ2 3.29

!00: Village 0.83

Marginal R2 0.134

Conditional R2 0.308 108 obs. Num. obs. 36 villages

134

Supplementary Information

Additional notes on materials and methods Intervention impact on forest change and forest carbon Synthetic controls statistical approach Synthetic controls is a generalization of the difference-in-differences (DID) approach that provides a rigorous method for identifying comparison units by constructing a “synthetic” control that is a weighted combination of many untreated units (1–3). Similar to traditional DID approaches, synthetic controls uses longitudinal data with repeated observations through time from treated and untreated groups, and constructs a counterfactual (i.e., what would have happened in the absence of treatment) assembled from a weighted combination of the untreated units in order to estimate the effect of the treatment (1–3).The weights are calculated using “matching” variables to maximize the similarity between the synthetic control and treatment units. Thus, in contrast to the ad hoc manual selection of controls by a researcher in DID methods, synthetic controls represents an objective and formalized approach for selection of controls (1–3). By maximizing the similarity of treatment and control units, the assumptions required in the DID approach (e.g., parallel trends pre-treatment) is strengthened, and the method is applicable and feasible in the absence of individual untreated units that are acceptably similar to the treatment unit. Finally, in matching on observed characteristics between treated and untreated units, synthetic controls may also perform better in matching on unobservable characteristics, though this cannot be formally verified (2, 3).

Estimating aboveground forest carbon density Aboveground carbon density (ACD, in Mg C ha-1) was estimated using top-of- canopy height (TCH, in m), basal area (BA, in m ha-1), and wood density (WD, in g cm-3) as inputs; BA and WD can be estimated directly from TCH. Therefore, we estimated ACD using LiDAR-derived estimates of average canopy height in GPNP (a conservative proxy for average TCH) to convert the area of averted forest loss to biomass of conserved carbon. The conversion equations provided by Jucker et. al (4) are as follows:

135

ACD = 0.567 TCH0.553BA1.081WD0.186 BA = 1.112 TCH WD = 0.385 TCH0.097

We took 100 randomly selected 30 m2 forested areas along the LiDAR flight path, each containing 900 LiDAR data pixels (in 1 m/pixel resolution) to build a histogram of average pixel height (Fig. S2). From this we estimated the mean TCH and then used the equations above to compute ACD. Using the average ACD per hectare in GPNP and the effect size of 69.8% forest-loss-rate reduction, we then estimated the total aboveground carbon stock (in Tg C ha-1) conserved in the period from 2008 to 2018.

Intervention impact on health clinic access, usage, and diagnoses Quantifying clinic usage: For accounting reasons, the clinic records included the method that the patient used to pay for their visit (e.g., handicrafts, seedlings, cash), and the patient database thus included multiple entries for most patient visits. Therefore, to conservatively estimate the number of patient visits, we assumed that unique patients came for a medical reason a maximum of one time per month. This undercounts the visits for patients needing frequent medical care for ongoing conditions (e.g., tuberculosis), but avoids overcounting visits and associated diagnoses for the many patients who returned one or more times after a visit only to pay. To estimate the probability of patient visits for each district, we calculated the cumulative number of unique patients from each district that used the clinic from 2008- 2018, divided by the estimated district population size in 2018, as reported by WorldPop (www.worldpop.org). For the 29 districts that were represented by at least one clinic patient of known origin, 2018 population sizes varied from 150 to 8,179 individuals, and the estimated proportion of the district population that used the clinic at least once ranged from 0.18% to 74.27%. For the district nearest to the clinic, more unique patients were recorded than there were people in the population, likely because of occasional usage by non-resident family members of district residents. For this district, the percentage of the population using the clinic was capped at 100%.

136

ICD10 codes represented in the dataset There were 824 diagnoses (ICD10 codes) recorded in the clinic patient records. Of those, there were 21 ICD10 codes that each made up 1-4% of the total diagnoses. They were, in descending order: acute upper respiratory infections (J06), essential (primary) hypertension (I10), dental caries (K02.1), Dyspepsia (K30), encounter for supervision of other normal pregnancy (Z34, Z34.0, Z34.8), myalgia (M79.1), respiratory tuberculosis (A16), gastro-esophageal reflux disease (K21), gastritis and duodenitis (K29, K29.7), necrosis of pulp (K04.1), tension-type headache (G44.2), pneumonia (J18.9), urinary tract infection (N39.0), emphysema (J43), age-related cataract (H25), unspecified fever (R50.9), peptic ulcer (K27), and asthma (J45). A full list of the 824 ICD10 codes tracked is presented in Table S6.

Estimating average district distance to clinic: Travel time estimates for each district were calculated using ArcGIS Pro 2.4 (Esri, Redlands, CA). The clinic database included records for patients originating from 29 named districts. A substantial fraction of the patient population (42%), however, were from unnamed districts reported as “far” (Fig. S3). To estimate average travel times for patients from the 29 named districts and the “far” group of patients, we first had to determine the geographic extent of unnamed districts from where patients traveled to use the clinic. To do so, we measured the maximum straight-line distance from the clinic to the edge of any of the 29 named districts. We then created a circular buffer using that maximum distance as the radius, and the health clinic as the center point, thereby encompassing all named districts and a large area of other, unnamed districts. All unnamed districts that were within or the majority of which intersected the circular buffer were determined as the most reasonable districts from where “far” patients derived. This resulted in selection of 132 districts that represented the most likely patient population in the “far” patient group. Polygons for all districts were downloaded from Humanitarian Data Exchange (https://data.humdata.org/dataset/indonesia-administrative-boundary- polygons-lines-and-places-levels-0-4b) as Indonesian Level 4 administrative boundaries (‘desa’), as of 2017. ArcGIS Pro 2.4 (Esri, Redlands, CA) was used for all geoprocessing. We used the Network Analysis toolset in ArcGIS Pro to estimate average travel time (in minutes) to the clinic from each district. The average travel time was calculated

137

as a weighted mean from 10 random points within each district, weighted by population density of that point. To arrive at this estimate we generated 10 random points per district (excluding any district areas within Ganung Palung National Park), with each point a minimum of 100 meters apart. Using the Generate Origin Destination Cost Matrix service tool within the Network Analysis toolset in ArcGIS Pro, we calculated travel time in minutes from each point to the clinic. To generate each point’s population weight, a 100 m circular buffer was created around each point, and the mean value of the WorldPop 2018 raster cells intersecting that buffer was obtained. For five of the non-MOU signing districts, travel time estimates for 1 random point were not resolved. The travel time tool described above excludes points if the points fall greater than 20 km from the nearest street. These five districts are on a large, sparsely populated island that lacks a completely connected road network in publicly available street network data. However, road network data does exist on parts of the island, so to estimate travel times for these districts, we created a new destination point within 5 km of the points which did return travel time estimates. Using the Generate Origin Destination Cost Matrix tool, travel times were estimated from the resolved random points to this new point, and they were added together to generate a complete set of data for each district. Additionally, we exclude one named non-MOU-signing district, Pelapis, from statistical analysis on clinic use because we lacked confidence in the ArcGIS-estimated travel times generated. The district encompasses a small group of islands west of the health clinic and requires ferry passage to reach the Borneo mainland, and its exclusion did not alter statistical outcomes. Last, there were six districts in the “far” group which did not yield travel time estimates for randomly generated points; these were all on the northwest coast at the edge of the “far” districts, and were excluded from analysis.

Household surveys assessing well-being and livelihood impacts Surveyor selection and training protocol In 2007 and 2012 the detailed household surveys were conducted by teams of local nursing students from Bethesda Hospital and Ketapang government nursing school. In 2017, some additional surveyors from the Ketapang government nursing school served as additional enumerators. The nursing students underwent training in survey

138

administration for 4 days prior to each survey. Training included instruction on how to avoid leading questions, how to identify and avoid potential bias, and the importance of random sampling. Enumerators were all required to conduct mock administrations of the survey with the trainers before the survey instrument was rolled out. Surveys were conducted by teams of 2 nurses, where 1 nurse was enumerator and the other nurse was reporter. None of the surveyors worked either previously or in the future for the non- governmental organization that ran the ASRI clinic.

139

Supplementary Information Figures and Tables

Fig. S1. Average above-ground carbon biomass of mixed forests in tropical, temperate, and boreal regions around the world, from Lutz et al. 2018 (5).

140

Fig. S2. LiDAR imagery used to calculate average and standard error of tree height in Gunung Palung National Park. A) Examples of individual canopies exceeding 30 m diameter (and, therefore, exceeding the pixel size at which forest change was estimated) circled in red. B) Histogram of vegetation height derived from sampling LiDAR imagery. C) Positioning of the LiDAR flight over GPNP.

141

Fig. S3. Locations of districts (‘desa’ administrative units, as of 2017) from where clinic patients derived. MOU-signing districts are shown in green, and are nearest Gunung Palung National Park (green hatched). Non-MOU-signing districts are shown in red and blue, where red represents districts recorded in clinic records, and blue represents patients from unnamed “far” district.

142

25 Visits per patient

0 No Yes MOU? A

100 MOU no MOU, known district no MOU, unknown district 75

0 Percent of population that used clinic 0 50 100 150 200 Travel time to clinic (minutes)

25 Visits per patient

0 No Yes MOU? Fig. S4. Clinic usage and access. A) Clinic usage, shown as the proportion of a district’s population that used the clinic at least once, decreased as access, or average distance (in 100 MOU minutes) of a district to the clinic,no MOU,increased. known district MOU-signing districts are shown in red, while non-MOU-signing districtsno are MOU, unknownshown district in blue. The patient population from unnamed “far”75 districts is shown as a black diamond because the underlying population size, and therefore proportion of population using the clinic, is estimated differently than named districts50 (see SI text for details); B) The number of times a unique patient visited the clinic for general doctor visits was, on average, more than 30% higher for patients from MOU-signing districts. Each point designates an individual patient, and the black 25 points indicate the means for all patients.

0 Percent of population that used clinic 0 50 100 150 200 Travel time to clinic (minutes)

143

Fig. S5. Individual engagement with the clinic and observed changes in disease. A) Cumulative unique clinic patients over time, among two patient populations: those living in villages/districts with MOUs (red) and the option to participate in intervention activities, and those living in villages/districts without MOUs (blue). B) Annual time series of the proportion of diagnoses for a particular disease in the population of unique patients utilizing the clinic, for infectious and non-infectious diseases in MOU-signing (mostly nearby to the clinic) vs non-MOU-signing populations (mostly far away from the clinic).

144

Fig. S6. Individual engagement with all intervention activities for 36 villages (‘dusun’) that border GPNP and have logging access areas within the park >0.30 km2. A) Distribution of cumulative engagement effort, where each line is a unique village included in the analysis over time; logging rates in villages in high- and medium-engaged villages are compared with logging rates in low-engaged villages to assess whether the intervention resulted in a ‘dose-response’ of changes in village-level logging rates to engagement intensity. B) Distribution of cumulative engagement effort across participating villages, stratified across each type of intervention activity. C) Description of intervention engagement activity types, measured as person-contacts to calculate engagement effort in each village with a signed MOU and recorded per village from comprehensive activity log books.

145

Table S1. Regression table showing the effect of signing an MOU on district-level infectious disease outcomes over time, controlling for average district distance to the clinic. For each disease category, the MOU effect was estimated from an interaction term of MOU and time (early, 2008-2009, versus late, 2017-2018), where each observation represents the proportion of patients assigned a particular disease diagnosis among all unique patients in a district, at two time periods. Log-odds, which can be exponentiated to derived odds ratios, are shown. CIs are based on a two-sided hypothesis test, where the null hypothesis of no MOU effect is rejected at 5% significance.

Infectious disease change, 2008-2009 to 2017-2018 Lower Upper Childhood- Diarrheal respiratory Malaria NTD Tuberculosis respiratory cluster diseases disease infections infections Intercept -4.28*** -4.03*** -4.51*** -3.72*** -5.34*** -2.64*** -4.32*** (0.52) (0.43) (0.20) (0.46) (0.20) (0.35) (0.32) Time: after -1.82* 0.09 1.28*** -2.41*** -1.38*** -1.56*** 1.92*** (0.76) (0.21) (0.19) (0.60) (0.22) (0.18) (0.20) MOU status -0.84 0.32 0.92*** -0.52 -0.27 0.35 0.45 (0.73) (0.48) (0.26) (0.62) (0.29) (0.38) (0.37) Distance (min, -0.75* -0.22 0.12 -0.40 -0.05 0.22 -0.26* scaled) (0.30) (0.18) (0.08) (0.31) (0.12) (0.13) (0.13) Interaction: 0.93 0.01 -0.50* -1.81 0.67** -0.05 -0.68** Time X MOU (0.80) (0.24) (0.22) (0.93) (0.25) (0.22) (0.22) status

Observations 166 166 166 166 166 166 166 Log Likelihood -102.02 -211.59 -229.81 -87.61 -177.64 -265.83 -267.66 Akaike Inf. Crit. 216.04 435.17 471.62 187.21 367.27 543.67 547.33 Bayesian Inf. 234.71 453.84 490.29 205.89 385.95 562.34 566.00 Crit. Note: *p<0.05; **p<0.01; ***p<0.001

146

Table S2. Regression table showing the effect of signing an MOU on district-level noninfectious disease outcomes over time, controlling for average district distance to the clinic. For each disease category, the MOU effect was estimated from an interaction term of MOU and time (early, 2008-2009, versus late, 2017-2018), where each observation represents the proportion of patients assigned a particular disease diagnosis among all unique patients in a district, at two time periods. Log-odds, which can be exponentiated to derived odds ratios, are shown. CIs are based on a two-sided hypothesis test, where the null hypothesis of no MOU effect is rejected at 5% significance.

Non-infectious disease change, 2008-2009 to 2017-2018 COPD Diabetes Heart Disease Malnutrition Intercept -3.59*** -2.87*** -3.81*** -6.35*** (0.41) (0.35) (0.18) (0.98) Time: after -1.06*** -0.61*** 0.28 -0.12 (0.20) (0.12) (0.15) (0.67) MOU status 0.54 -0.34 0.18 -0.30 (0.46) (0.39) (0.24) (1.27) Distance (min, scaled) 0.26 0.02 0.17* 0.10 (0.15) (0.15) (0.09) (0.43) Interaction: Time X MOU status 0.24 0.18 0.19 0.67 (0.24) (0.16) (0.20) (0.85)

Observations 166 166 166 166 Log Likelihood -227.35 -226.28 -211.52 -57.31 Akaike Inf. Crit. 466.71 464.56 435.05 126.63 Bayesian Inf. Crit. 485.38 483.23 453.72 145.30 Note: *p<0.05; **p<0.01; ***p<0.001

147

Table S3. Regression table showing the effect of signing an MOU on district-level “other” and untracked disease outcomes over time, controlling for average district distance to the clinic. For each disease category, the MOU effect was estimated from an interaction term of MOU and time (early, 2008-2009, versus late, 2017-2018), where each observation represents the proportion of patients assigned a particular disease diagnosis among all unique patients in a district, at two time periods. Log-odds, which can be exponentiated to derived odds ratios, are shown. CIs are based on a two-sided hypothesis test, where the null hypothesis of no MOU effect is rejected at 5% significance.

Other and untracked disease change, 2008-2009 to 2017-2018 Dental Liver Anemia Trauma Untracked disease Disease Intercept -5.06*** -3.40*** -6.66*** -4.29*** -1.32*** (0.43) (0.20) (0.53) (0.33) (0.04) Time: after -0.40 2.21*** 0.25 0.63* -0.16*** (0.42) (0.15) (0.49) (0.28) (0.03) MOU status -0.15 0.45 0.66 -0.21 -0.04 (0.60) (0.23) (0.72) (0.44) (0.06) Distance (min, scaled) -0.13 -0.43*** 0.49 -0.45** 0.01 (0.22) (0.08) (0.25) (0.16) (0.02) Interaction: Time X MOU status 0.66 -0.89*** -0.03 -0.47 0.02 (0.49) (0.17) (0.67) (0.33) (0.04)

Observations 166 166 166 166 166 Log Likelihood -107.34 -330.58 -63.57 -143.28 -417.99 Akaike Inf. Crit. 226.68 673.17 139.15 298.56 847.99 Bayesian Inf. Crit. 245.36 691.84 157.82 317.23 866.66 Note: *p<0.05; **p<0.01; ***p<0.001

148

Table S4. Number of districts (‘desa’ administrative unit), households surveyed, and individuals within households, including demographic frequencies at baseline (2007), 5- year (2012) and 10-year (2017) time points. Demographic frequencies are expressed as the number of individuals (N) with the percent (%) of individuals in each category in parenthesis.

Year 2007 2012 2017 N. District (Desa) 24 23 23 N. Households 1348 1498 1379 Sex and household age distribution (total number, proportion) Female, all ages 3103 (50.1) 3063 (48.5) 3031 (50.2) Under 1 year 142 (2.3) 115 (1.8) 117 (1.9) 1-5 years 829 (13.4) 632 (10.0) 536 (8.7) 6-18 years 1748 (28.2) 1791 (28.2) 1591 (26.0) 19-59 years 3253 (52.5) 3519 (55.5) 3415 (55.7) 60 or more years 225 (3.6) 274 (4.3) 443 (7.2) Total 6197 6345 6126

149

Table S5. Survey findings on 5-year and 10-year impact on household-level births, infant deaths, and livelihoods. Results of regressions estimate change from baseline to after the intervention in average annual birth rate (births per 1000 women ages 19-59) and infant mortality (average annual infant deaths per 1000 live births). As a reference, we report general fertility rates (GFR) and infant mortality rates (IMR) (with SE) measured through Indonesia Demographic and Health Surveys (5, 6) in 2007 and 2017 for West Kalimantan province (where all communities surveyed are located), and for rural populations within West Kalimantan, which are likely an appropriate match to the rural households involved in the household surveys.

Birth rate and infant mortality from household surveys Units 2007 2017 % decline Birth rate Births per 1,000 123 56 -54.5% women 19-59 (n=1642) (n=1695) Infant Deaths per 1,000 76 25 -67.2% Mortality births (n=605) (n=285) General fertility rate (GRF) and infant mortality rate (IMR) from Demographic and Health Surveys West Kalimantan Detail 2007 2017 % decline General Births per 1,000 128.6 89.5 -30.4% Fertility Rate women 15-44 (SE=6.1) (SE=5.8) Infant 36-month period 47.1 13.1 -72.3% Mortality Rate (SE=21.9) (SE=8.9) West Kalimantan: Rural Detail 2007 2017 % decline General Births per 1,000 127.2 77.7 -38.9% Fertility Rate women 15-44 (SE=11.0) (SE=7.2) Infant 36-month period 53.6 9.7 -81.9% Mortality Rate (SE=35.5) (SE=6.4)

150

Table S6. List of all ICD10 codes tracked (or not tracked) for analysis of change in proportion of unique clinic patient visits to a general doctor resulting in a particular diagnosis. Shown are 824 ICD10 codes that contributed to diagnosis of a disease category accounting for 0.05 to 11.3% of all diagnoses between 2008 and 2018, and an additional 578 untracked ICD10 codes that appeared in the patient diagnosis records during the same time period.

Statistical analysis Specific disease Untracked category category Tracked ICD10 code ICD10 codes Anemia Anemia D46.4,D50,D50.0,D50.9,D61.9,D64,D64.8,D64.9 Childhood- cluster diseases Pertussis A37,A37.0,A37.8,A37.9 Childhood- cluster diseases Mumps B26 Childhood- cluster diseases Measles B05 Childhood- cluster diseases Tetanus A35 Chronic obstructive COPD pulmonary disease J43,J44,J44.1 Dental disease Dental caries K02.0,K02.1,K02.2,K02.3,K02.8,K02.9,K04.6,K04.7,K12.2 Dental disease Peridontitis K04.0,K04.1,K04.2,K04.4,K04.9,K05.1,K05.2,K05.4,K05.6 K00.6,K00.7,K01.0,K01.1,K03.8,K06.1,K06.9,K07,K07.3,K07.6,K08.3,K08.9,K09.0,K09.1,K09.2,K11.2,K11.4,K11.5,K11.6,K11.7,K Dental disease Other dental disease 12,K12.0,K12.1,K13.0 Diabetes mellitus Diabetes type 2 E10,E11,E11.4,E11.5,E11.6,E14 Diarrheal disease Diarrhoea A04,A05,A05.9,A07.1,A08,A08.0,A08.3,A08.4,A09,K52,K52.8,K52.9,K59.1,R19.7,A03,A03.0,A03.3,A03.9 Diarrheal disease Amoebiasis A06,A06.0,A06.4,A06.8 K75.0 Heart Disease Hypertension I11,I11.9 I10,I15,I15.2 Coronary artery Heart Disease disease I20,I20.0,I20.9,I21,I23.8,I24,I24.9,I25,I25.1,I25.2,I25.9,R07.4 Heart Disease Heart failure I11.0,I27.9,I50,I50.0,I50.1,I50.9,J81,O90.3

Heart Disease Heart disease NOS I05.0,I07.1,I31.3,I33,I34.0,I34.1,I34.9,I35.0,I36.1,I39,I42.0,I42.8,I44.0,I44.1,I45.0,I47,I47.2,I48,I49.3,I49.9,R00.0,R00.2,R01,R01.1 Heart Disease Pericarditis I30,I30.9 Liver Disease Hepatitis B15,B17.2,B18,B94.2,K71.2,K73.9,K75.9 Liver Disease Other liver disease K72,K74,K74.0,K74.6,K76.0,K76.6,K76.9

151

Lower respiratory infections Pneumonia A31.0,J12,J15,J15.9,J17,J18,J18.0,J18.1,J18.9,J22,P23 Lower respiratory infections Bronchitis J20,J21,J40,J41.0 Malaria Malaria B54,B50,B51,B51.9 Malnutrition Malnutrition E44,E46,E51.9,E52,E53.9,E56.1,E58 Statistical analysis Specific disease Untracked category category Tracked ICD10 codes ICD10 codes NTD Chikungunya A92 NTD Dengue fever A90,A91 NTD Filariasis B74,B74.9 Leprosy (Hansen's NTD disease) A30,B92 NTD Leptospirosis A27.0 NTD Scabies B86 NTD Trachoma A71 NTD Typhoid A01.0,A01.4 NTD Worm infections B65.9,B69.0,B73,B76.0,B76.9,B77,B79,B80,B82,B83.9,B85.2 S04.1,S05.0,S05.1,S05.2,S05.9,T15,T15.0,T15.1,S09.2,T16,W85,W86,W92,S00,S00.2,S00.4,S00.5,S00.9,S01.0,S01.4,S01.5,S02,S02.2 ,S02.4,S02.5,S02.6,S06,S06.0,S06.2,S06.5,S06.8,S09,S09.9,S13.4,S14.2,S20.0,S20.2,S20.8,S22,S22.3,S29,S30,S30.1,S32.0,S32.8,S33, S33.5,S40.0,S40.9,S43.7,S44,S46.0,S51,S52.6,S60,S61,S61.0,S62.2,S63.7,S64.2,S70.0,S72,S80,S81,S81.7,S82,S82.7,S83,S83.2,S83.4, S83.6,S84.1,S86.0,S90,S90.9,S91,S92.0,S93.4,S96,T00.9,T06,T07,T09.0,T14.0,T14.1,T14.3,T14.4,T21,T21.1,T23,T23.1,T25,T26,T30, Trauma Trauma T31.1,K08.1,M12.5,M17.3,M24.3,M84.0,M84.1,S04.9,T17,T17.1,T17.2,T18,T79.3,T79.7,V99,W45

Tuberculosis Tuberculosis A15,A15.0,A15.3,A16,A16.0,A16.5,A17.0,A17.1,A18,A18.2,A18.3,A18.4,B90,M49.0,M90.0,Z20.1 Upper respiratory Upper respiratory infections tract infection J00,J06,J06.9

152

A46,A49,B07,B08.1,B08.4,B24,D56.9,D59.3,D68.0,D69.1,D69.3,D69.9,D73.1,E16.2,E23.0,E23.2,E24,E28.2,E66.0,E66.9,E73,E73.9,E74.2,E78,E79.0,E83. 1,E84,E86,E87,E87.1,E87.5,E87.6,E87.8,F05,F13.0,F15.3,F15.9,F23.8,F23.9,F25.1,F42,F50.9,F52.2,F52.4,F52.5,F52.9,F70,F72,F79,F83,F84.0,F88,F89,F9 0.0,F95.0,F98.2,G11.1,G12.2,G20,G24.5,G24.9,G25.0,G25.5,G25.8,G35,G47.0,G47.3,G47.9,G51.3,G52.9,G56.2,G61.0,G70.0,G71,G72.3,G82.2,G82.3,G83 .0,G83.4,G90.0,G91,G93.1,G93.3,G93.9,G95,G95.2,G95.8,G96.9,G99.0,G99.8,H01.0,I26,I28.0,I62.1,I72.9,I73.1,I77.6,I78.0,I80.0,I82.9,I83,I83.9,I87.1,I87. 2,J05.0,J33,J38,J39.0,J47,J68.0,J70.4,J84.1,J84.9,J90,J93.8,J94.1,J98.0,J99.8,K13.2,K14,K14.1,K14.6,K20,K22.0,K22.2,K22.5,K22.9,K23,K31.8,K31.9,K50 ,K56,K56.0,K56.1,K56.6,K56.7,K57,K57.3,K57.9,K58.9,K60,K60.2,K60.3,K61.0,K62.2,K62.8,K63.5,K63.9,K65.0,K66.1,K80.2,K82.9,K85,K86.1,K90,K9 0.1,K92.0,K92.9,L10.9,L12.1,L30.1,L30.5,L40.3,L42,L44.0,L51.9,L53.9,L56.0,L60,L60.0,L60.8,L62,L63,L65.9,L72,L72.0,L72.1,L72.9,L73.9,L74,L74.0,L 74.1,L74.3,L75.2,L80,L81.0,L81.9,L82,L84,L85.3,L85.8,L89,L90.5,L90.6,L91.0,L91.8,L92,L92.3,L92.9,L93,L93.0,L94.0,L95.8,L97,L98,L98.0,M.20,M00. 9,M21.3,M22,M22.2,M23.4,M25,M25.0,M31.6,M32,M33.1,M34,M35.3,M40.0,M41,M43.1,M43.6,M45,M48.0,M50,M50.1,M51.1,M51.9,M54.5,M60,M62 .3,M62.6,M62.8,M65.3,M65.4,M67.4,M70.2,M70.4,M70.5,M70.8,M71.2,M71.5,M72.2,M72.6,M75.0,M75.1,M75.3,M75.4,M75.8,M76.3,M76.5,M76.6,M7 7.0,M77.1,M77.5,M77.9,M79.1,M79.2,M79.3,M79.7,M81,M86,M86.9,M89.9,M94.0,M99.5,N08.2,N08.3,N13,N13.3,N16,N25.1,N28,N28.9,N31,N31.9,N3 2.0,N32.9,N35,N39.3,N39.4,N39.9,N40,N41,N43,N43.3,N44,N45,N46,N47,N48.1,N48.5,N48.8,N48.9,N61,N63,N64.4,N64.5,N64.8,N70,N71,N71.0,N71.9, N72,N73,N75.0,N76.0,N76.8,N77,N77.1,N80,N81,N81.6,N83.2,N84.9,N85.0,N85.8,N87,N89.8,N94,N94.0,N95,N95.1,N95.8,N95.9,N97,O009,O01,O02.0, O08,O36.4,O41.0,O68,O70.0,O80,P12.0,P50.9,P93,P95,Q85.0,Q85.1,Q99,R02,R03.0,R04.0,R04.2,R06.1,R06.6,R07.1,R07.3,R09.1,R10,R10.0,R10.2,R10.3 ,R10.4,R11,R13,R14,R16,R16.1,R17,R18,R19.0,R19.6,R25.2,R27.0,R27.8,R29.0,R33,R35,R42,R47.0,R49,R50,R50.9,R52,R52.2,R52.9,R53,R55,R57.9,R5 9.0,R60,R60.9,R61,R62.0,R62.8,R62.9,R63.0,R64,R65.2,R68.8,R73.0,R73.9,R81,R82.3,R98,T37.9,T44.0,T48.6,T57.1,T59.8,T60,T60.3,T63.0,T63.2,T63.6, T63.8,T65,T67.0,T75.3,T78.2,T79.9,T81.9,T88,U89.9,W56,X54,Y40.3,Y44.0,Y49,Z41.2,Z99.0,O34.2,H00.1,H01.1,H02.4,H02.6,H02.8,H04.0,H04.1,H04.3 ,H04.5,H05.2,H11.1,H11.2,H11.3,H16.0,H16.4,H17,H17.8,H17.9,H18.2,H18.7,H27.1,H27.9,H33.0,H34,H35.9,H36,H36.8,H44.5,H44.9,H47,H50.9,H53,H5 Untracked 3.2,H53.4,H53.6,H54,H54.7,H57,H57.1,H57.9,Z01.0,H61.2,H68.1,H69.9,H70,H72.0,H73.9,H81,H90,H93.1,H93.3,H93.9,Z41.3,J45,J45.0,J45.9,J45.99,J46,J diagnoses 66.8,W55,W57,X20,X23,Z30,Z30.9

153

SI References

1. A. Abadie, J. Gardeazabal, The Economic Costs of Conflict: A Case Study of the Basque Country. American Economic Review. 93, 113–132 (2003).

2. A. Abadie, A. Diamond, J. Hainmueller, Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California’s Tobacco Control Program. Journal of the American Statistical Association. 105, 493–505 (2010).

3. M. W. Robbins, J. Saunders, B. Kilmer, A Framework for Synthetic Control Methods With High-Dimensional, Micro-Level Data: Evaluating a Neighborhood- Specific Crime Intervention. Journal of the American Statistical Association. 112, 109–126 (2017).

4. T. Jucker, G. P. Asner, M. Dalponte, P. G. Brodrick, C. D. Philipson, N. R. Vaughn, Y. A. Teh, C. Brelsford, D. F. R. P. Burslem, N. J. Deere, R. M. Ewers, J. Kvasnica, S. L. Lewis, Y. Malhi, S. Milne, R. Nilus, M. Pfeifer, O. L. Phillips, L. Qie, N. Renneboog, G. Reynolds, T. Riutta, M. J. Struebig, M. Svátek, E. C. Turner, D. A. Coomes, Estimating aboveground carbon density and its uncertainty in Borneo’s structurally complex tropical forests using airborne laser scanning. Biogeosciences. 15, 3811–3830 (2018).

5. J. A. Lutz, T. J. Furniss, D. J. Johnson, S. J. Davies, D. Allen, A. Alonso, K. J. Anderson‐Teixeira, A. Andrade, J. Baltzer, K. M. L. Becker, E. M. Blomdahl, N. A. Bourg, S. Bunyavejchewin, D. F. R. P. Burslem, C. A. Cansler, K. Cao, M. Cao, D. Cárdenas, L.-W. Chang, K.-J. Chao, W.-C. Chao, J.-M. Chiang, C. Chu, G. B. Chuyong, K. Clay, R. Condit, S. Cordell, H. S. Dattaraja, A. Duque, C. E. N. Ewango, G. A. Fischer, C. Fletcher, J. A. Freund, C. Giardina, S. J. Germain, G. S. Gilbert, Z. Hao, T. Hart, B. C. H. Hau, F. He, A. Hector, R. W. Howe, C.-F. Hsieh, Y.-H. Hu, S. P. Hubbell, F. M. Inman‐Narahari, A. Itoh, D. Janík, A. R. Kassim, D. Kenfack, L. Korte, K. Král, A. J. Larson, Y. Li, Y. Lin, S. Liu, S. Lum, K. Ma, J.-R. Makana, Y. Malhi, S. M. McMahon, W. J. McShea, H. R. Memiaghe, X. Mi, M. Morecroft, P. M. Musili, J. A. Myers, V. Novotny, A. de Oliveira, P. Ong, D. A. Orwig, R. Ostertag, G. G. Parker, R. Patankar, R. P. Phillips, G. Reynolds, L. Sack, G.-Z. M. Song, S.-H. Su, R. Sukumar, I.-F. Sun, H. S. Suresh, M. E. Swanson, S. Tan, D. W. Thomas, J. Thompson, M. Uriarte, R. Valencia, A. Vicentini, T. Vrška, X. Wang, G. D. Weiblen, A. Wolf, S.-H. Wu, H. Xu, T. Yamakura, S. Yap, J. K. Zimmerman, Global importance of large-diameter trees. Global Ecology and Biogeography. 27, 849–864 (2018).

6. “Statistics Indonesia - Badan Pusat Statistik - BPS - and Macro International. 2008. Indonesia Demographic and Health Survey 2007. IDIR51FL.DTA, IDBR51FL.DTA. Calverton, Maryland, USA: BPS and Macro International. ICF [Distributor], 2008.”

7. “National Population and Family Planning Board - BKKBN, Statistics Indonesia - BPS, Ministry of Health - Kemenkes, and ICF. 2018. Indonesia Demographic and Health Survey 2017. IDIR71FL.DTA, IDBR71FL.DTA. Jakarta, Indonesia: BKKBN, BPS, Kemenkes, and ICF. ICF [Distributor], 2018.”

154

155

Dissertation conclusion

This dissertation starts with a few simple questions: Where and for how long do globally important parasites and pathogens persist in environmental reservoirs; What (or who) are those environmental reservoirs?; and, What ecosystem services (natural enemies) might mitigate their spread? Answers to these questions, pursued through synthesis science and literature reviews, are relevant to better understand basic infectious disease ecology and to help researchers and policy makers identify sustainable environmental targets for disease control, in an embrace of ecological complexity as a source for opportunity. Ultimately, however, global goals to simultaneously improve human health and well-being, and to curb ongoing trends towards environmental destruction, biodiversity loss, and climate change demand transformative action grounded in evidence. To this end, this dissertation moves on to more complex questions probing feedbacks between human health and environmental sustainability within the framework of two case studies. In the Senegal River Basin, we ask: What ecological correlates drive the world’s longest recorded schistosomiasis epidemic, and at what scale? These questions were addressed in the context of an ongoing manipulative experiment assessing environmental restoration as an ecological lever to improve human health, and improve the implementation of ecological levers by identifying where they are needed most. Then, in rural-poor tropical forests of Borneo, we ask: Is improving human health and well-being an effective lever to improve conservation of globally important forests? We find that ecological levers for health, and health levers for conservation can be used to improve health and protect the environment at the same time. In Senegal, we developed methods to identify species-specific ecological correlates and spatial scales of schistosomiasis disease risk from the environment in a complex multi-pathogen, multi- host schistosome system. Findings directly support the design and implementation of important ecological levers for reducing schistosomiasis disease burden, which is a major goal of World Health Organization to reduce disease for nearly 800 million people at risk in endemic areas. In Borneo, we found that over 10 years, an on-the-ground intervention to improve human health and well-being and to, consequently, reduce illegal logging in

156

protected tropical forests provided healthcare for more than 28,000 people and averted more than 27 km2 (~70%) of illegal logging. The complex, interlinked health and environment challenges facing the world are unprecedented. So too are the scientific and technological capacities to address these challenges. This dissertation humbly borrows methods from ecology, earth observation, epidemiology, and social sciences in an effort to advance progress towards a sustainable future for people and nature.

157