COMPLEX INTERACTIONS AMONG ECOSYSTEM SERVICES, HUMAN WELL - BEING, AND THEIR LINKAGES TO TELECOUPLING PROCESSES By Min Gon Chung A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Fisheries and Wildlife Doctor of Philosophy Environmental Science and Policy Dual Major 2020 ABSTRACT COMPLEX INTERACTIONS AMONG ECOSYSTEM SERVICES, HUMAN WELL - BEING, AND THEIR LINKAGES TO TELECOUPLING PROCESSES By Min Gon Chung With rapid economic and population growths, the increasing separation between where ecosystem services are needed and from where they are supplied makes managing multiple ecosystem services difficult. It is also c hallenging to strengthen the synergies between such ecosystem service flows and conservation activities as conservation activities can enhance human well - being through the improvements of ecosystem services. Increasing the demands for ecosystem services ac ross regions may accelerate ecosystem service flows yet also damage the basic ability to provide ecosystem services in supply areas. However, little research has holistically examined the environmental and socioeconomic impacts of increasing separation bet ween the supplies of and demands for ecosystem services. To fill these gaps in knowledge, the overall objectives of this dissertation are to examine the intricate interconnections among ecosystem service flows, natural systems, human well - being, and conser vation policies simultaneously. This dissertation research applies the integrated framework of telecoupling (socioeconomic and environmental interactions over distances) to systematically uncover the agents, causes, and effects of dynamic ecosystem service flows across multiple coupled human and natural systems. This research explores telecoupling processes regarding nature - based tourism (cultural service, Chapter 2 and 3), food (provisioning service, Chapter 4), and fresh water (provisioning and regulating service, Chapter 5) as well their interactions with biodiversity, human demands, and conservation policies. The spatial scale of this research is at the global level, as flows of touris ts , food, and fresh water occur across national and regional boundarie s. Chapter 2 shows that protected areas managed strictly for biodiversity conservation have more visitors and species than those managed for mixed use. High population density surrounding protected areas and national income levels are also major socioecono mic factors related to nature - based tourism. Chapter 3 indicates that global tourism networks have become highly consolidated over time and that reduced transaction costs (e.g., language, distance, and visa policies) are more important in attracting intern ational tourists than natural and cultural attractions. Furthermore, cost of living differences between countries decreased in importance over time. International tourist flows are resilient to political instability and terrorism risks. Chapter 4 investiga tes the effects of international food trade on biodiversity hotspots between developed and developing countries. My results show that international food trade may benefit global biodiversity due to the increasingly important role of developing countries wi thout biodiversity hotspots in food exports. Chapter 5 explores how to integrate watershed conservation activities with built infrastructure approaches to sustain freshwater ecosystem services for global cities. My results indicate that wetlands in protect ed areas contribute to sustaining freshwater provisions to global cities. Forests in protected areas complement large dams for sediment reduction and hydropower production for cities, but cities mainly depend on dams for flood mitigation. By assessing ecosystem service flows to people over distances, this research identifies how multiple ecosystem services are managed in order to provide benefits for distant beneficiaries and to whom subsidies (or payments) are paid for biodiver sity and ecosystem service conservation. The integration of the telecoupling framework with ecosystem services provides new perspectives on global sustainability that help with the development of proactive strategies for biodiversity and ecosystem service conservation. iv To Hana v ACKNOWLEDGEMENTS I have met incredible mentors and scholars who have dedicated their support and guidance to me throughout my Ph.D. studies at Michigan State University. I would not have completed my Ph.D. journey without their support and mentorship. and natural systems. Based on his wonderful work, I was able to develop the career of my dreams. Dr. Liu was always on the top of my list when I was searching for Ph.D. advisors. I am indebted to Dr. Liu for his great guidance and mentorship. I have been fortunate to experience his brilliance for six years. He has expanded my knowledge of coupled human and natural systems and ecosyst em services in multiple perspectives. I will bring his lessons with me in my career. I appreciate all your support, patience, and encouragement, Dr. Liu. I would like to express my sincere gratitude to Dr. Thomas Dietz for his careful guidance and encourag ement. His professional expertise and brilliant insights helped me enhance my knowledge and scientific thinking in interdisciplinary research. I am deeply thankful to Dr. Kenneth Frank for his valuable guidance and mentorship in shaping my research. My mot ivation in the field of network analysis was inspired through working with and learning from Dr. Frank. I also owe my appreciation to Dr. Yadu Pokhrel for his knowledge and expertise in freshwater. His expertise and willingness help me frame the final piec e of my dissertation in a sophisticated manner. I would like to thank the members of the Center for Systems Integration and Sustainability for their friendship and collaboration these past six years. In my Ph.D. journey, they helped and supported me to enh ance my academic passions and shape my academic foundation. I also owe a great deal of thanks to my Ph.D. cohort from the Sustainable Michigan vi Endowed Project and colleagues from the Environmental Science and Policy Program. I also thank MSU FW and CSIS st aff (especially Sue Nichols, Jennifer Mitchell, Shuxin Li, James Vatter, and Jill Cruth). I appreciatively acknowledge the financial support of the Environmental Science and Policy Program, the Sustainable Michigan Endowed Project, the National Science Fou ndation, William W. and Evelyn M. Taylor Endowed Fellowship program, FW Graduate Student Organization, NASA - MSU program, and Michigan AgBioResearch. Certainly, I am grateful to my family for their love and support. My parents have always believed in me and encouraged my career decisions. I deeply appreciate my beloved Hana. In my difficulties and setbacks, she has always been there for me, encouraged me, and supported me with her wisdom and kind consideration. Words can hardly express how grateful I am for her love and patience. This dissertation is dedicated to my parents and Hana. vii TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ ......................... ix LIST OF FIGURES ................................ ................................ ................................ ........................ x CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 1 1.1. Background ................................ ................................ ................................ .......................... 2 1.2. Goal and Objectives ................................ ................................ ................................ ............. 4 CHAPTER 2 GLOBAL RELATIONSHIPS BETWEEN BIODIVERSITY AND NATURE - BASED TOURISM IN PROTECTED AREAS ................................ ................................ ............. 5 2.1. Introduction ................................ ................................ ................................ .......................... 7 2.2. Materials and Methods ................................ ................................ ................................ ......... 9 2.2.1. Data ................................ ................................ ................................ ............................... 9 2.1.2. Modeling strategy ................................ ................................ ................................ ....... 13 2.1.3. Regression model ................................ ................................ ................................ ........ 14 2.3. Results ................................ ................................ ................................ ................................ 18 2.3.1 Biodiversity and its conservation strategies had a positive relationship with nature - based tourism ................................ ................................ ................................ ........................ 18 2.3.2 Nature - based tourism was influenced by socioeconomic and environmental drivers . 20 2.4. Discussion ................................ ................................ ................................ .......................... 22 2.4.1 The ro le of biodiversity in nature - based tourism ................................ ......................... 22 2.4.2 Management implications ................................ ................................ ............................ 25 CHAPTER 3 INTERNATIONAL TOURISM DYNAMICS IN A GLOBALIZED WORLD: A SOCIAL NETWORK ANALYSIS APPROACH ................................ ................................ ........ 27 3.1. Introduction ................................ ................................ ................................ ........................ 29 3.2. Literature Review ................................ ................................ ................................ ............... 31 3.2.1. Social network analysis ................................ ................................ ............................... 32 3.2.2. Hypothesized factors affecting international tourism ................................ ................. 35 3.3. Materials and Methods ................................ ................................ ................................ ....... 37 3.3.1. Data collection ................................ ................................ ................................ ............ 37 3.3.2. Cluster analyses ................................ ................................ ................................ .......... 39 3.3.3. Mixed - effects model ................................ ................................ ................................ ... 41 3.4. Results ................................ ................................ ................................ ................................ 44 3.4.1. Consolidated global tourism networks ................................ ................................ ........ 45 3.4.2. Factors related to international tourism ................................ ................................ ...... 46 3.5. Discussion ................................ ................................ ................................ .......................... 52 3.5.1. Reasons behind consolidated global to urism networks ................................ .............. 52 3.5.2. The role of conservation in international tourism ................................ ....................... 53 3.5.3. The impact of policies on international tourism ................................ ......................... 55 3.6. Conclusions ................................ ................................ ................................ ........................ 57 viii CHAPTER 4 RETHINKING INTERNATIONAL FOOD TRADE FOR GLOBAL BIODIVERSITY CONSERVATION ................................ ................................ .......................... 60 4.1. Introduction ................................ ................................ ................................ ........................ 62 4.2. Materials and Methods ................................ ................................ ................................ ....... 64 4.2.1. Biodiversity hotspo t and non - hotspot countries ................................ ......................... 64 4.2.2. Data collection ................................ ................................ ................................ ............ 65 4.2.3. Aggregate analysis ................................ ................................ ................................ ...... 66 4.2.4. Panel data analysis ................................ ................................ ................................ ...... 67 4.3. Results ................................ ................................ ................................ ................................ 69 4.4. Discussion ................................ ................................ ................................ .......................... 78 CHAPTER 5 INTEGRATING BUILT INFRASTRUCTURE AND WATERSHED CONSERVATION TO SUSTAIN FRESHWATER ECOSYSTEM SERVICES FOR GLOBAL CITIES ................................ ................................ ................................ ................................ .......... 80 5.1. Introduction ................................ ................................ ................................ ........................ 82 5.2. Materials and Methods ................................ ................................ ................................ ....... 85 5.2.1. City and watershed selection ................................ ................................ ...................... 85 5.2.2. Freshwater ecosystem services ................................ ................................ ................... 86 5.2.3. Source watershed and city characteristics ................................ ................................ ... 89 5.2.4. Egocentric network analysis ................................ ................................ ....................... 91 5.3. Results ................................ ................................ ................................ ................................ 94 5.4. Discussion ................................ ................................ ................................ .......................... 97 5.4.1. The role of forests and wetlands in protected areas ................................ .................... 97 5.4.2. Conservation strategies for freshwater ES ................................ ................................ .. 99 5.5. Conclusions ................................ ................................ ................................ ...................... 101 CHAPTER 6 CONCLUSIONS ................................ ................................ ................................ .. 103 APPENDICES ................................ ................................ ................................ ............................ 108 APPENDIX A SUPPORTING INFORMATION FOR CHAPTER 2 ................................ ... 109 APPENDIX B SUPPORTING INFORMATION FOR CHAPTER 3 ................................ ... 121 APPENDIX C SUPPORTING INFORMATION FOR CHAPTER 4 ................................ ... 124 APPENDIX D SUPPORTING INFORMATION FOR CHAPTER 5 ................................ ... 136 REFERENCES ................................ ................................ ................................ ........................... 140 ix LIST OF TABLES Table 2.1. Summary of variables regarding nature - based tourism hypothesis ............................. 14 Table 2.2. Descriptive statistics of dependent and independent variables, N=929. ...................... 17 Table 2.3. Summary results of the model averaging predicting annual visitor numbers in PAs. . 19 Tab le 3.1. Odds ratios for cluster analysis and p - value based on simulations followed by mean, median, and 95% Quantile interval of simulations. ................................ ................................ ...... 46 Table 3.2. Phi para meter estimates with median and 95% Quantile intervals. ............................. 50 Table 4.1. Coefficients of panel data analyses in three different periods: 2000 2007, 2008 2015 and 2000 2015. ................................ ................................ ................................ ............................. 77 Table 5.1. Multi - level coefficients predicting four freshwater ecosystem services. .................... 96 Table S2.1. The descriptions of dependent and independent variables. ................................ ..... 109 Table S2.2. List of PAs ( N=929). ................................ ................................ ............................... 111 Table S2.3. Variance Inflation Factor (VIF) for variables used in linear regression. ................. 118 Table S2.4. Unstandardized coefficients from multiple regression model predicting annual visitor numbers except biodiversity and total species in PAs. ................................ ............................... 119 Table S3.1. Odds ratios for cluster analysis and p - value based on simulations followed by mean, median, and 95% Quantile interval of simulations. ................................ ................................ .... 121 Table S4.1. List of hotspot and non - hotspot countries, subdivided into developed and developing countries groups. ................................ ................................ ................................ ......................... 130 Table S4.2. The quantity of food exports and imports for developed and developing countries in 2000, 2015, and 2000 2015 (average annual). ................................ ................................ ........... 131 Table S4.3. The quantity of agricultural area saved (km 2 ) due to food imports. ........................ 132 Table S4.4. The percentages of popu lation, food production, and food trade among high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC), with each group subdivided into developed and developing countries, for 2000, 2015, and 2000 2015 (average annual). ................................ ................................ ................................ ......................... 134 Table S5.1. Descriptions of dependent and independent variables. ................................ ........... 137 Table S5.2. Variance Inflation Factor (VIF) for variables used in multi - level models .............. 139 x LIST OF FIGURES Figure 2.1. 929 PA locations in the world. ................................ ................................ ................... 12 Figure 2.2. Buffer zone variables with 10 - km distance increments across PAs boundaries. A. Population density, B. Agricultural yields, C. Agricultural areas, D. Water supply from upstream PAs ................................ ................................ ................................ ................................ ................ 22 Figure 3.1. Clusters of global tourism networks in (a) 2000 2002 and (b) 2011 2013. The size of each node i ndicates the sum of international tourist arrivals and departures. Red ties indicate tourist flows within the same cluster, and gray ties indicate tourist flows between different h of the interactions between countries based on the number of international tourists. The core countries were located in the center of the cluster maps. ................................ ................................ ................................ .. 46 Figure 3.2. Mean and 95% Highest Posterior Density (HPD) confidence intervals of the coefficients from 2000 2013: (a) the size of protected areas in receiving countries (km 2 ), (b) the size of World Cultural Heritage sites in receiving countries (km 2 ), (c) politic al stability in sending countries (index), (d) political stability in receiving countries (index), (e) visa - free status between sending and receiving countries (visa - free=1), (f) shared language between sending and receiving countries (shared language=1), (g) distances between countries (km), (h) national price level difference between sending and receiving countries (price - level ratio), (i) per capita GDP in sending countries (constant 2010 US $), (j) per capita GDP in receiving countries (constant 2010 US $), (k) population size of sending countries (person), and (l) population size of receiving countries (person). ................................ ................................ ................................ ......................... 49 Figure 3.3. Distributions of (a) the sender effects in 2000 2002, (b) the sender effects in 2011 2013, (c) the receiver effects in 2000 2002, (d) the receiver effects in 2011 2013. Country abbreviations: Australia (AUS), Belgium (BEL), Canada (CAN), Switzerland (CHE), China (CHN), Germany (DE U), Spain (ESP), France (FRA), United Kingdom (GBR), India (IND), Italy (ITA), Japan (JPN), Republic of Korea (KOR), Maldives (MDV), Malaysia (MYS), Netherlands (NLD), New Zealand (NZL), Russian Federation (RUS), Thailand (THA), United States (USA), and So uth Africa (ZAF). ................................ ................................ ........................ 51 Figure 4.1. The quantity of net food trade between developed and developing high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC). Blue indicates net food trade in 2000, red indicates net food trade in 2015, and gray indicates average net annual food trade from 2000 2015. The net amounts of food trade in each group are not linearly increased or decreased over time. The net amounts of food trade in 2000 and 2015 can be lower or higher than those in other mid - years as shown in Figure S4.2. ................................ ................ 70 Figure 4.2. Average annual food flows (Mt/year) from 2000 2015. Food flows between developed and developing high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC). Non - hotspot countries are marked by red, high - hotspot co untries by dark green, and low - hotspot countries by light green. The arc length of an outer circle indicates xi the sum of food exported and imported in each group. The arc length of a middle circle refers to the quantity of food exports. The inner arc length shows the quantity of food imports. Raw data from UN FAO (2018). ................................ ................................ ................................ .................. 7 2 Figure 4.3. Changes in agricultural intensification and agricultural area in high - hotspot count ries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC), with each group subdivided into developed and developing countries. (A) Fertilizer use (tonne/km 2 ); (B) pesticide use (tonne/km 2 ); (C) agricultural water withdrawal (m 3 /km 2 ); (D) agr icultural area change (km 2 ). Raw data from UN FAO (2018). ................................ ................................ ................................ .. 74 Figure 5.1. Egocentric networks between cities (egos) and freshwater source watersheds (alters). Red sq uares indicate cities, and blue circles indicate freshwater source watersheds. Each city has more than one source watershed, and thus they form an egocentric network. ............................. 92 Figure 5.2. Spatial changes in numbers of dams and sizes of PAs from 2000 to 2016 in (A) freshwater source watersheds and (B) hydropower watersheds. Orange indicates an increase in the numbers of dams and size of PAs, red indicates an increase in on ly dams, green indicates an watershed from 2000 to 2016. ................................ ................................ ................................ ...... 97 Figure positive correlation for a given pair, and red indicates negative correlation. Colored correlation coefficients are significant at the p=0.05 level. ................................ ................................ .......... 120 Figure S3.1. Clusters of global tourism networks by country: (a) 2000 2002, (b) 2011 2013, (c) 2000, (d) 2001, (e) 2002, (f) 2003, (g) 2004, (h) 2005, (i) 2006, (j) 2007, (k) 2008, (l) 2009, (m) 2010, (n) 2011, (o) 2012, and (p) 2013. ................................ ................................ ...................... 122 Figure S3.2. Mean and 95% Highest Posterior Density (HPD) confidence intervals of the coefficients from 2 000 2013 in the alternative model: (a) the proportion of protected areas in receiving countries, (b) the proportion of World Cultural Heritage sites in receiving countries, and (g) the number of direct flights between countries. ................................ ............................. 123 Figure S4.1. Global distribution of hotspot countries (high - hotspot countries [HHC], low - hotspot countries [LHC] and non - hotspot countries [NHC]). NA = countries with missing data. Raw da ta from Myers et al. (2000) and Myers (2003) . ................................ ................................ .............. 124 Figure S4.2. The amounts of net food trade between developed and developing countries in high - hotspot countries ( HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC) from 2000 2015. Non - hotspot countries are indicated by red, high - hotspot countries dark green, and low - hotspot countries by light green. ................................ ................................ .......................... 125 Figure S4.3. The percentage of agricultural area in biodiversity hotspots out of total agricultural area. Raw data from Myers et al. (2000), Myers (2003), and Tuanmu and Jetz (2014). ............ 126 Figure S4.4. Number of countries with different percentages of biodiversity hotspots (land area with biodiversity hotspots out of total terrestrial land area). Raw data from Myers et al. (2000) and Myers (2003) . ................................ ................................ ................................ ....................... 127 xii Figure S4.5. Annual food flows (Mt) in 2000. Food flows between developed and developing countries in high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC). Non - hotspot countries are indicated by red, high - hotspot countries by dark green, and low - hotspot countries by light green. The arc length of an outer circle indicates the sum of food exported and i mported in each group. The arc length of a middle circle refers to the amounts of food exported. The inner arc length shows the amounts of food imported. Raw data from UN FAO (2018). ................................ ................................ ................................ ................ 128 Figure S4.6. Annual food flows (Mt) in 2015. Food flows between developed and developing countries in high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC). Non - hotspot countries are indicated by red, high - hotspot countries by dark green, and low - hotspot countries by light green. The arc length of an outer circle indicates the sum of food exported and imported in each group. The arc length of a middle circle refers to the amounts of fo od exported. The inner arc length shows the amounts of food imported. Raw data from UN FAO (2018). ................................ ................................ ................................ ................ 129 Figure S5.1. The locations of cities and watersheds in (A) fr eshwater source watersheds, (B) flood watersheds, and (C) hydropower watersheds. Red indicates cities, and blue indicates watersheds. ................................ ................................ ................................ ................................ .. 136 1 CHAPTER 1 INTRODUCTION 2 1.1. Background The challenges of managing multiple ecosystem services include determining the complex interactions between ecosystem service supply and demand in the context of coupled human and natural systems (Bagstad et al., 2013; Burkhard et al., 2012; Chung and Kang, 2013; Liu et al., 2007) . While rapid economic growth leads humans to become more dependent on ecosystem services (Guo et al., 2010) , high ecosystem service demands may exceed the capacity of ecosystem service supplies worldwide (Burkhard et al., 2012; Liu et al., 2016a) . The demands for ecosystem services in distant locations are met via processes such as food trade, water transfer, or traveling (Chung et al., 2018a; Chung et al., 2019; Liu et al., 2016a) . The increasing separation between where ecosystem services are needed and where they are supplied from makes managing multiple ecosystem services difficult (Burkhard et al., 2014; Chung et al., 2018b; Wei et al., 2017) . Increasing ecosystem service basic abilities to provide those resources (Bagstad et al., 2014; Chung et al., 2018b) . supply or demand separately. In addition, it is a big challenge to strengthen the synergies between ecosystem s ervices and biodiversity conservation (Adams, 2014) . This is because people depend on key ecosystem services such as food production and water in areas that have high biodiversity (Brooks et al., 2014) . Although biodiversity conservation efforts can enhance human well - being through the improvement of ecosystem services (Ferraro et al., 2015) , increasing ecosystem service demand across regions may accelerate ecosystem service flows but damage the basic ability to provide ecosystem services in supply areas (Bagstad et al., 2014) . Unequal distributions of ecosystem 3 service benefits may also cause stakeholder conflict, institutional failure, and environmental degradation (Adams, 2014) . A holistic ap proach is necessary to investigate complex interconnections between the ecosystem service supply and demand areas and to integrate human and natural systems using the framework of telecoupling (Liu et al., 2015b) . Telecoupling is defined as socioeconomic and environmental interactions over distances (Liu et al., 2013) . The telecoupling framework allows the investigation of complex interconnections across distant coupled human and natural systems using five interrelated components: system, flow, cause, effect, and agent (Liu et al., 2013) . Using the telecoupling framework helps determine causes and effects of flows (e.g., movement of energy, information, capital, and natural resources) between sending and receiving systems (Chung and Liu, 2019; Chung et al., 2018b) . Ecosystem service supply, demand, and flow can be integrated with the ecosystem service supply and demand areas, are factors that affect the emergence and dynamics of telecouplings. E are socioeconomic and environmental materials, energy, people, a nd information between systems. are entities that facilitate or prevent telecoupling processes . The integration of the telecoupling framework with ecosystem services can provide new perspectives on global sustainability and biodiversity that will help develop proactive strategies for global conservation. For example, the quantification of ecosystem service flows allows us to develop better management strategies, minimize trade - offs, and maximize synergies by determining the relationships between pr ovisioning areas (e.g., source watersheds or sending systems) and beneficiaries (e.g., urban areas or receiving systems) (Serna - Chavez et al., 2014) . 4 1.2. Goal and Objectives The main goal of this dissertation is to analyze complex interconnections among biodi versity, ecosystem services, and human well - being, and their linkages to telecoupling processes. This dissertation research applies the integrated framework of telecoupling to systematically uncover the agents, causes, and effects of dynamic ecosystem serv ices flows in telecoupled human and natural systems. The specific objectives are to: (1) (2) (3) (4) This research explores telecoupling processes regarding food (provisioning services), nature - based tourism (cultural servic es), and freshwater (provisioning and regulation services). The spatial scale of this research is at the global level because flows of food, tourism, and freshwater occur across national and regional boundaries. By assessing ecosystem service flows to peop le, this research can identify how multiple ecosystem services are managed to provide benefits for distant beneficiaries and to whom subsidies (or payments) are paid for ecosystem services conservation. This dissertation research contributes to a better fo undation for future research and policy to enhance global sustainability. 5 CHAPTER 2 GLOBAL RELATIONSHIPS BETWEEN BIODIVERSITY AND NATURE - BASED TOURISM IN PROTECTED AREAS In collaboration with Thomas Dietz and Jianguo Liu 6 Abstract The relationships between biodiversity conservation and ecosystem services (ES) are widely debated. However, it is still not clear how biodiversity conservation and ES interact with different strategies in and surrounding protected areas (PAs), the cornerstone for biodiversity conservation. Here, we present results from the interplay between biodiversity conservation and nature - based tourism (a cultural ES), while controlling for environmental and socioeconomic factors in and surrounding terrestrial PAs worldwi de. Results indicate that nature - based tourism is more frequent in PAs that are of higher biodiversity, older, larger, more accessible from urban areas and at higher elevation. High population density surrounding PAs and national income levels are also maj or socioeconomic factors related to nature - based tourism. Furthermore, PAs managed mainly for biodiversity conservation have nearly 35% more visitors than those managed for mixed use. Strict management for biodiversity is also associated with increased bio diversity. These results show the importance of biodiversity in addressing nature - based tourism and suggest this interrelationship could be altered by different management strategies used by PAs. 7 2.1. Introduction For more than a century, designating and managing protected areas (PAs) has been done with a goal of allowing current use of biodiversity, usually through tourism, while preserving resources for future generations (Beissinger et al., 2017) . But since the first designation of PAs , there have been conflicts over the appropriate goals in managing such areas (Dietz, 2017b; Joppa and Pfaff, 2010; Liu et al., 2012; Mace, 2014; Talli s and Lubchenco, 2014; Watson et al., 2014) . One goal emphasizes the protection of natural systems and biodiversity (nature for itself) (Mace, 2014) . The other emphasizes the contribution of ecosystem services (ES) from PAs to human well - being (nature for people) (Mace, 2014) . Some PAs are managed with a sharp focus on the sole goal of preserving biodiversity; others are managed with an intent to enhance the provis ion of multiple types of ES. Of course, preservation of natural systems and biodiversity can contribute to cultural ES, including nature - based tourism (Bayliss et al., 2014; Clements and Cumming, 2017) . Additionally, biodiversity may enhance the production of a wide variety of ES beyond just cultural ES (Chung et al., 2015; Smith et al., 2017; Turner et al., 2012) but it is not necessarily the case that managing a PA for biodiversity will optimize overall provision of ES (K arp et al., 2015; Naidoo et al., 2008) . Thus, understanding the relationship between ES and biodiversity is a major challenge for sustainability science (Carpenter et al., 2009; Chan et al., 2006; Graves et al., 2017; Ouyang et al., 2016; Turner et al., 2007) . Two further complexities emerge because PAs are not isolated from the rest of the world. the PA but that affects and is affected by what happens in the PA (DeFries, 2017) . Further, PAs are telecoupled with non - adjacent systems in several ways that influence the supply of and demand for ES (Bagstad et al., 2013; Liu et al., 2016a) . Most visitors to PAs have traveled from 8 distant places to visit them (Liu et al., 2013; Xiao et al., 2017) . PAs may provide water purification that have benefits to people hundreds or thousands of kilometers away, and in turn may be affected by upstream degradation of water quality (Watson et al., 2014) . Agricultural activities surrounding PAs can negatively influence biodiversity conditions in PAs (Bailey et al., 2016; Palomo et al., 2013) . The demand for agricu ltural products from the surrounding PAs may also be local, regional or global (Liu et al., 2015a) . Finally, invasive species, which threaten many PAs may h ave their origins across the globe and climate change, a severe threat to many PAs, has its drivers distributed globally as well (Pimm et al., 2014; Zhong et al., 2015) . For many PAs, one of the most important ES is providing an attractive destination for nature - based tourism, which is both regional and global in origin. Such tourism may be influenced in complex ways by how PAs are managed (Graves et al., 2017; Karp et al., 2015) . In some PAs, managing primarily for biodiversity might discourage nature - based touris m, while in others such management might be compatible with high demand for visits. Agricultural landscape surrounding PAs may provide additional attractions that could either increase or decrease demand for tourism at a PA (Baudron and Giller, 2014; Fleischer et al., 2018; Jie et al., 2013; Liu et al., 2012) . For individual PAs, we can trace plausible paths by which biodiversity conservation strategies change demand for nature - based tourism via environmental and socioeconomic changes in the PA and surrounding areas. But there is little empirical analysis of the overall effects of PA management on tourism demand and supply. To address this gap in the literature, we used data from PAs w orldwide to examine the number of visitors to PAs as a function of the number of species in the PA and the management strategy being used, while controlling for environmental and socioeconomic factors. In addition, we investigated how different 9 conservatio n strategies influence biodiversity and other factors both inside and outside PAs. Our analysis addresses two questions. First, how does biodiversity and nature - based tourism interact in PAs that may be governed by different conservation strategies? Second , which environmental and socioeconomic factors in and surrounding PAs influence visitation to PAs? Our analysis is based on terrestrial PAs that have visitation information between 2000 and 2014. Our results can contribute to a better understanding of how biodiversity and nature - based tourism interact in PAs and how these interactions may be altered by different conservation strategies used by PAs. 2.2. Materials and Methods 2.2.1. Data The dataset was obtained by aggregating data from a number of interna tional institutions, national statistical agencies, online datasets and the gray literature (Table S2 .1). Our key dependent variable was the average annual visitor numbers for each PA. The final dataset contained 929 PAs in 50 countries with the annual vis itor numbers at some point in the period 2000 to 2014 ( Fig u re 2. 1 and Table S2 .2). We calculated visitation as the average annual visitor numbers in each PA over the 15 - year period. The two key independent variables are the management strategy being used at the PA and its biodiversity. Management strategy was operationalized as the IUCN management category. The IUCN management category is based on the primary management objectives of PAs, which should apply to more than 75% of the PA area (Dudley, 2008) . The IUCN category facilitates global assessments across different countries by providing an international standard for classifying management strategies of PAs. The primary objective of categories II - IV is to protect biodiversity (PAs managed for biodive rsity), while categories V - VI are to both protect nature 10 and use natural resources sustainably (PAs managed for mixed use) (Baudron and Giller, 2014; Dudley, 2008; Joppa et al., 2008; Laurance et al., 2012) . For example, Categories II - IV focus on minimizing human activities keeping the system in Categories V - VI allow sustainable use of natural resources (e.g., hunting and/or forestry) to balance interaction between people and nature (Dudley, 2008) . Dividing all PAs into two groups helps to differentiate conservation management practices between those that manage for nature for itself (II - IV) and those that manage for nature and people (V - VI). We divided all 9 29 PAs into two groups (II - IV and V - VI): 677 PAs in Category II - IV were coded 1 and 252 PAs in Category V - VI were coded 0. We excluded marine PAs and PAs which had not been classified into one of the IUCN management categories . PAs in IUCN category Ia and Ib where visitor access is strictly limited were also excluded. To include active management PAs, we selected PAs that were designated and managed at the national or sub - national level. The designated PAs have a long - term commitment to conservation with l egal means (IUCN and UNEP - WCMC, 2017) . Second, biodiversity was operationalized as the number of species of birds, mammals and amphibians within the PA (Jenkins et al., 2013; Pimm et al., 2014) . The biodiversity mapping website (http://biodiversi tymapping.org) provided a global map of species ranges for birds, mammals and amphibians based on data from IUCN (IUCN, 2014) and BirdLife International NatureServe (BirdLife International NatureServe, 2013) . A species range polygon underlies these mapping efforts. We selected mammals, birds and amphibians because these species have most comprehensive data at a global level and because they seem likely to be the species that (Hausmann et al., 2017a; Siikamäki et al., 2015) . The species 11 bitat, elevation limited, and other expert knowledge of the species and its range (IUCN, 2014) maps are the best available global datasets, we note the maps may overestimate species richness as the range of potential distribution tends to be larger than the actual occurrences of the species (Willemen et al., 2015) . All species maps have a spatial resolution of 10km by 10km, based on 2013 upd ated data. We only included native and extant species. We overlaid this species range map with the locations of PAs and extracted species number in each PA by using zonal statistics in ArcGIS (ESRI, 2015) . We included as control variables a number of characteristics of the PA that might influence its attractiveness f or nature - based tourism: size, mean elevation, mean annual temperature, mean annual precipitation and age (years since formal designation). We also controlled for remoteness which was defined as travel time (in minutes) from the nearest major cities (popul ation>50,000) and the percentage of total water supply originated in the PA. Higher percentage of water supply in the PA indicates that the PA has more freshwater resources (WRI, 2015) . Finally our model included dummy variables for the continent in which the PA was located. Data on size, mean elevation, mean annual temperature, mean annual precipitation and travel time from the nearest urban area for each PA was extracted from the appropriate geographic data bases using PA boundaries to develop zonal statistic s in ArcGIS. In addition to the features of the PA itself, we have characterized buffer zones for each PA. Following previous research (Joppa and Pfaff, 2010; Wittemyer et al., 2008) , we specified 10 - km buffer zones around each PA. Capturing activity within the buffer zones is important because the PA and its management may influence conditions within the buffer zones and vice versa. For each 10 - km buffer zone, we extracted population density, agricultural yield and the 12 percentage of agricultural area. We selected only cases with valid values for all variables excluding those PAs for which data for relevant variables were missing. We appreciated that specifying a 10 - km buffer zone is s omewhat arbitrary. To test the sensitivity of our analysis to the size of the buffer zone, we performed a multiple ring buffer analysis in ArcGIS and QGIS (ESRI, 2015; QGIS Developme nt Team, 2014) . We designated 10 - km distance intervals from the PA boundary (0 - km) to 50 - km buffer zones. Then, we extracted numerical values from the PA boundary and each of five rings (0 - 10, 10 - 20, 20 - 30, 30 - 40 and 40 - 50 km) using the spatial dataset. In each PA boundary and ring, we obtained numerical values of environmental and socioeconomic factors (population density), agricultural factors (agricultural yield and agricultural area) and regulating ES (water supply originated in PAs). In the multiple ring buffer analysis, we did not consider agricultural factors within PA boundaries because many PAs prevent people from engaging in agricultural activities (Palomo et al., 2013) . Figure 2 . 1 . 929 PA locations in the world. 13 2.1.2. Modeling strategy Our basic model predicts annual visits to each PA as a function of the species richness of the PA and the management strategy being used, with strategies ranging from strict emphasis on biodiversity protection to more mixed use. We also include a variety of control variables in our regressions to minimize the risk that the effects we estimate for biodiversity and management strategy ar e spurious. We control for features of the PA by including its size, mean elevation, annual mean temperature and precipitation, remoteness and age. We also control for population density within a 10 km buffer zone around the PA and for the affluence (gross domestic product per capita) of the nation in which the PA is located (reliable data on affluence cannot be obtained at a spatial scale corresponding to the 10 km buffer zone). Controls are also included for agriculture in the buffer zone and water supply originated in PAs (agricultural yield, % land area in agriculture and % total water supply originating in PAs ) . We provide a summary of variables regarding nature - based tourism hypothesis (Table 2. 1). Finally, we include dummy variables for continent. This model allows us to address our research questions by examining how biodiversity, management strategy and the characteristics of the PA itself and its buffer zone influence the popularity of a site for nature - based tourism. 14 Table 2 . 1 . Summary of variables regarding nature - based tourism hypothesis Variables Relationships with nature - based tourism Source(s) Species richness More species richness contributes to greater nature - based tourism value. Arbieu et al. (2017); Hausmann et al. (2017a); Siikamäki et al. (2015); (Smith et al., 2017); Willemen et al. (2015) Management strategies PAs managed for biodiversity actively encourage visitors for nature - based tourism. Dudley (2008) Size of PAs Larger size of PAs has more visitors Balmford et al. (2015); Baum et al. (2017) Elevation Geographical attributes such as elevation may in Hausmann et al. (2017b); Kumari et al. (2010) Temperature and precipitation Climate and weather are important factors for visitors (e.g., low humidity and heat stress) Scott et al. (2008a); Verbos et al. (2017) PA remoteness Visitors are reluctant to go remote PAs Balmford et al. (2015); Neuvonen et al. (2010) PA age Visitor numbers increase with PA age Karanth and DeFries (2011); Neuvonen et al. (2010) Population Visitor num bers are higher when there is a higher population density surrounding PAs. Balmford et al. (2015); Ghermandi and Nunes (2013) GDP per capita PAs in high - income countries have more visitor numbers Balmford et al. (2015); Ghermandi and Nunes (2013) Agricultural factor Agricultural landscape surrounding PAs may provide additional attractions and/or food - related activities. Baudron and Giller (2014); Fleischer et al. (2018); Hjalager and Johansen (2013); Jie et al. (2013) Water supply in PAs Plenty of water resources in PAs provide greater attractions (e.g., lakes, streams, waterfalls) Cao et al. (2016); Nyaupane and Chhetri (2009); Reinius and Fredman (2007) 2.1.3. Regression m odel The multiple regression equation for the nature - based tourism model is in the multiplicative form commonly used in the STIRPAT models (STochastic impacts by Regression on Population, Affluence and Technology) of human drivers of environmental change (D ietz, 2017a) : For ease of estimation we used log base e of all except the binary variables, thus: 15 Where Y stands for the average annual visitor numbers in each PA from 2000 to 2014, X 1 is the number of species, X 2 is IUCN management category, X 3 is the area of each PA, X 4 is mean elevation, X 5 is annual mean temperature, X 6 is annual precipitation, X 7 is PA remoteness from major cities, X 8 is PA age, X 9 is population density, X 10 is per capita GDP at the national level, X 11 is agricultural yield, X 12 is the percentage of agricultural area, X 13 is % water supply originated in PAs, X 14 - X 17 are dummy variables for each contin ent (Asia and Oceania, Africa, Europe and North America). E is the error term. Note that in this multiplicative form the unstandardized regression coefficients can be interpreted as elasticities. That is, our estimates indicated that a 1% change in an inde pendent variable is associated with a b% change in the dependent variable, net of all other variables in the model. STIRPAT models have frequently been used to examine non - linearities beyond the log - log form and other specifications when there are theoreti cal arguments to do so. However, since our analysis is an initial exploration of factors related to visitation, we have kept to this rather well known functional form. To account for model selection uncertainty, we used an information theoretic approach f or model averaging. This approach provides robust parameter estimates based on model averaging across the best set of models by information theoretic criteria (e.g., Akaike Information Criterion (AIC)) rather a more traditional approach of selecting the be st fitting model (Galipaud et al., 2014; Grueber et al., 2011) . We first gene rated a candidate model set of 131,072 models to determine the model set for averaging. These models were then ranked based on AICc (AIC for small samples) to avoid overfitting (Grueber et al., 2011) . Models with a smaller AICc are considered to have a better fit. We used a top 2AICc cut - off criterion which results in a s et of three best models. The top 2AICc cut - off criterion indicates that AICc difference between model i and the top - ranked model is less than 2 ( = AICc i AICc top ) 16 (Burnham and Anderson, 2002) . Then, the parameter estimates of the top three models were averaged using Akaike weig hts ( ). The Akaike weights ( ) indicate the relative likelihood of the candidate models with a normalized scale (0 - 1) and provide a way to interpret values as probabilities (Burnham and Anderson, 2002) . Models with a bigger have a smaller . T he percentage of water supply originating in PAs did not appear in the final model as this variable was not in cluded in the top three models developed using the information theoretic approach. Although some variables were not statistically significant, including all variables allow to identify indirect relationships on the annual visitations via biodiversity and g uards against spurious relationships (Table 2. 2). To formally test the indirect impacts of other factors on the annual visitations via biodiversity, we performed the regression of visitor numbers on all other variables except biodiversity. To capture the d ifference of the number of species between PAs primarily managed for biodiversity and PAs managed with more mixed objectives, we also modeled the number of species in PAs as a function of the same independent variables. Since this analysis is secondary to the analysis of tourism, we did not deploy the information theoretical approach to model selection. According to the correlation matrix for the independent variables, 96% of 76 pairs had the value of r less than 0.5 (Fig ure S2 .1). In addition to the correl ation matrix, we examined collinearity using variance inflation factors (VIF) . All VIFs were less than 5, indicating no serious collinearity problems (Table S2 .3). All statistical analyses were performed with R software (R Core Team, 2017) . The information theoretical model averaging approach was deployed using MuMIn package in R. We used the procedures developed by Frank et al. (2013) to examine the robustness of our results. These procedures calculate what proportion of cases in the data set would ha ve to be replaced with null hypothesis cases in order for the 17 significance of a coefficient to drop below a threshold of interest. We used the conventional p=0.05 as our threshold. If a relatively modest proportion of cases would have to be replaced with n ull cases for a coefficient to fall below the p=0.05 threshold then the inference is rather fragile; if a high proportion of cases would have to be replaced the inference is robust. Table 2 . 2 . Descriptive st atistics of dependent and independent variables, N=929. Category Variable Mean Std. Dev Nature - based Tourism Annual visitor numbers in PAs (persons) 367,405 1,793,697 Biodiversity Total species (species) 326.88 172.540 Protected Area IUCN category (II - IV=1) 0.729 0.445 Size of PAs (km 2 ) 860.91 2,640.659 Mean elevation (meter) 825.3 880.661 Annual mean temperature (ºC) 14.449 8.063 Annual precipitation (mm) 1298.898 827.042 PA remoteness (minutes) 360.8 413.191 PA age (year) 38.24 23.095 Demographic Population density § (persons/km 2 ) 140.012 471.987 Economic GDP per capita ¶ (2005 const. $ per capita) 16,127.7 16,342.57 Agricultural factor Agricultural yields § (tonne/km 2 ) 553.9 387.914 Agricultural area § (%) 30.051 24.773 Regulating ES Water supply originated in PAs (%) 13.66 13.827 Region Asia and Oceania 0.378 0.485 Africa 0.097 0.296 Europe 0.231 0.422 North America 0.127 0.333 Latin America 0.167 0.373 § 10 - km buffer zone ¶ Country level data, not PAs level 18 2.3. Results 2.3.1 Biodiversity and its conservation strategies had a positive relationship with nature - based tourism Biodiversity has a positive relationship with the number of annual visitors to PAs (Table 2. 3). Each 1% increase in the number of species is associated with an increase in annual visitors of about 0.87%, indicating that biodiversity is one of the strongest influences on tourism. IUCN management category also has a positive association with the annual visi tors meaning that PAs managed strictly for biodiversity conservation attract more visitors than PAs for mixed use. Validation suggests that these results are relatively robust. To invalidate the inference of a positive relationship of the number of species with the annual visitors, 48% of the estimated effect would have to be due to bias (Frank et al., 2013) . One can interpret this as 48% (or 446 PAs) of the cases in this study would have to be replaced with null hypothesis cases to invalidate the inference. 19 Table 2 . 3 . Summary results of the model averaging predicting annual visitor numbers in PAs. Category Variable Model 1 Model 2 Model 3 Model Averaging Biodiversity Total species (species) 0.879** (0.231) 0.870** (0.231) 0.868** (0.234) 0.874** (0.232) Protected Area IUCN category (II - IV=1) 0.351* (0.166) 0.347* (0.166) 0.348* (0.166) 0.349* (0.166) Size of PA (km 2 ) 0.309** (0.039) 0.309** (0.039) 0.310** (0.039) 0.309** (0.039) Mean elevation (meter) 0.329** (0.058) 0.343** (0.057) 0.331** (0.058) 0.334** (0.058) Annual mean temperature (ºC) - 0.378* (0.161) - 0.341* (0.160) - 0.383* (0.162) - 0.367* (0.162) Annual precipitation (mm) - 0.480** (0.118) - 0.469** (0.118) - 0.483** (0.118) - 0.477** (0.118) PA remoteness (minutes) - 0.236* (0.111) - 0.253* (0.111) - 0.240* (0.112) - 0.242* (0.112) PA age (year) 0.665** (0.117) 0.668** (0.117) 0.663** (0.117) 0.665** (0.117) Demographic Population density § (persons/km 2 ) 0.455** (0.061) 0.469** (0.060) 0.448** (0.066) 0.458** (0.062) Economic GDP per capita ¶ (2005 const. $ per capita) 1.262** (0.086) 1.279** (0.086) 1.266** (0.087) 1.268** (0.087) Agricultural factor Agricultural yields § (tonne/km 2 ) 0.101 (0.059) - 0.094 (0.063) 0.099 (0.060) Agricultural area § (%) - - 0.027 (0.091) 0.027 (0.091) Regulating ES Water supply originated in PAs (%) - - - - Region Asia and Oceania 1.866** (0.223) 1.837** (0.223) 1.856** (0.226) 1.855** (0.224) Africa 0.967* (0.321) 0.867* (0.316) 0.966* (0.321) 0.935* (0.323) Europe 0.685* (0.267) 0.687* (0.268) 0.658* (0.282) 0.681* (0.271) North America 1.233** (0.304) 1.186** (0.303) 1.221** (0.306) 1.216** (0.305) Intercept - 9.757** (1.797) - 9.451** (1.790) - 9.695** (1.810) - 9.648** (1.803) R 2 0.478 0.476 0.478 k 17 16 18 AICc 3969.053 3969.936 3971.041 0.000 0.882 1.988 0.129 0.083 0.048 * P<0.05, ** P<0.001 § 10 - km buffer zone ¶ Country level data, not PAs level Values in parentheses are standard errors 20 2.3.2 Nature - based tourism was influenced by socioeconomic and environmental drivers We find that agriculture surrounding PAs and water supply in PAs do not have a direct relationship with the annual visitor numbers in PAs at the p=0.05 level. Additionally, indirect associations of agriculture and wa ter supply on visitor numbers via biodiversity are not significant (Table S2 .4). P opulation density in 10 - km buffer zones around PAs is positively associated with visitor numbers (P<0.001). We acknowledge that our cross - sectional data cannot disentangle ca usal direction: some people in the buffer zones may also visit the PA (the larger the population, the more visitors) but large numbers of visitors may also encourage local population growth. Per capita GDP has the strongest link to the number of visitors ( P<0.001) presumably because in high - income countries there are more people who can afford nature - based tourism and because PAs in high - income nations may be more desirable destinations since there may be larger budgets for tourist infrastructures (e.g., vi sitor centers), all other things being equal. The characteristics of PAs also influence the annual visitor numbers in PAs. The age and size of PAs positively affect the visitor numbers (P<0.001). Older PAs have had more time to gain recognition, often repr esent the most spectacular areas and may have been preserved in more pristine state than more recent PAs. In addition, PAs with larger sizes attract more nature - based tourists, presumably because large PAs have more natural attractions and habitats for spe cies. While the visitor numbers are positively associated with mean elevation, the visitor numbers are negatively associated with annual mean temperature and annual precipitation. This means that PAs with in a cooler temperature, lower precipitation and hi gher elevation have more visitors. People may visit PAs with high elevation areas to appreciate novel aesthetic views and 21 natural habitats with high biodiversity because these PAs may avoid development pressures, maintain good natural habitat conditions an d often have spectacular scenery. In addition, PA remoteness is negatively associated with the visitor numbers. PAs with good accessibility have more visitors. If PAs are located in the remote areas far from urban areas, people may not be able to afford th e cost and/or time to visit the PAs even if the PAs provide good natural attractions. All regional variables (Asia and Oceania, Africa, Europe and North America) have a significant p - value (P<0.05) when compared with Central and South America, the baseline continent. Net of the controls we have used, PAs in the other four continents have more visitors than those in Central and South America. There are major variations in management goals in PAs, reflected in the IUCN categorization. We find that this categ orization is capturing differences that are important in terms of the amount of biodiversity in a PA, with the PAs primarily managed for biodiversity having 1.05 times more species than the PAs managed with more mixed objectives (Table S2 .4). The nature o f the buffer zone seems to have some correlation with number of visitors, with each 1% increase in population density associated with a 0.45% increase in visits. Agriculture in the buffer zone has no relationships with visitors to PAs. We tested the sensit ivity of our analyses to the size of the buffer zones (Fig ure 2. 2). Population and agricultural variables have the same pattern of effects when measured for larger buffer zones as they do in the 10 - km buffer zone . 22 Figure 2 . 2 . Buffer zone variables with 10 - km distance increments across PAs boundaries. A. Population density, B. Agricultural yields, C. Agricultural areas, D. Water supply from upstream PAs 2.4. Discussion 2.4.1 The role of biodivers ity in nature - based tourism This study examines the relationships of biodiversity and other factors to nature - based tourism and the factors that are associated with biodiversity in PAs. The results demonstrate that biodiversity has a positive relationship with nature - based tourism even when a variety of other factors are controlled: with each 1% increase in biodiversity associates with a 0.87% increase in 23 tourism. Furthermore, management strategies matter: PAs managed primarily for biodiversity protection h ave nearly 1.35 times the visits of those managed for mixed use. And management for biodiversity is associated with higher biodiversity, given the controls for other factors. Thus, we tentatively suggest that producing both biodiversity and nature - based tourism simultaneously is possible given appropriate conservation strategies. That is, biodiversity is compatible with economic development via tourism if pro per strategies are deployed (Oldekop et al., 2016) . More visitors can increase opportunities for local economic developments such as hotels, restaurants and employment opportunities for nature guides (Liu et al., 2012) . Management plans that consi der both biodiversity and local community participation could enhance economic development surrounding PAs and thus provide livelihood benefits to the local residents and reduce economic inequalities (Das and Chatterjee, 2015; Oldekop et al., 2016; Plummer and Fennell, 2009) . Because our data are cross - sectional, we cannot fully d isentangle complex causal loops. Nevertheless, we feel our models capture the dominant interrelationships and lay the groundwork for further research. We have used an information theoretic approach to calculate the average of top models among the set of mo dels. These models assume a linear in the logs functional form and specify no interactions of the form that allow effects to differ across subgroups in our data. But we note that results are fairly robust with regard to such specification errors nearly hal f the cases would have to be invalidated to change our most important inferences and it seems unlikely that a missing specification that powerful has not been suggested in the literature. Of course, further work is required to overcome a lack of global bio diversity data. Although species richness is a crucial factor of nature - based tourism in PAs (Arbieu et al., 2017; Hausmann et al., 2017a; Siikamäki et al., 2015) , the relationship of other aspects of biodiversit y (e.g., evenness 24 and abundance) to nature - based tourism in PAs warrants attention (Graves et al., 2017; Siikamäki et al., 2015) . Further research might fruitfully examine more complex causal feedbacks that we have been able to estimate. For example, it may be that higher biodiversity PAs are given more protective management strategies or that there is some feedback from high visitat ion rates to an emphasis on biodiversity protection policies. We also note that although we have used a well - accepted standard international classification of PA management strategies, we lack data that would allow for detailed comparisons of management st rategies (e.g., targeted species, budgets for tourism). In particular, PAs in high - income countries may have better accessibility with larger budgets for tourist infrastructures (e.g., visitor centers, roads within PAs and campgrounds). The causal feedback s can be complex. For instance, tourist infrastructure can increase the visitor numbers, but construction of tourist facilities, the footprint of the facilities and increased traffic can all be a threat to biodiversity (Daniel et al., 2012) . Several strategies would allow further research to expand on our analyses. There are ongoing efforts for improving global data sets by using social media (Hausmann et al., 2017b; Willemen et al., 2015) and developing global database of protected areas including visitor counts and biodiversity (Dubois et al., 2016; Schägner et al., 2017) . These could all allow for more refined analyses. Data over time deployed as a panel would allow for strong causal inference. And detailed comparative case studies would allow a better understanding of how processes that link tourism, biodiversity and management strategy co - evolve. 25 2.4.2 Management implications ES supply and demand change over temporal and spatial scales (Burkhard et al., 2014; Renard et al., 2015) and so do the interactions between biodiversity and nature - based tourism. Further, these changes are ver y context specific. It follows that effective plans for biodiversity protection would benefit from local community participation (Kovács et al., 2015; Liu et al., 2007; Pleasant et al., 2014) . For example, with the rapid increases of human population and income in many parts of the globe, human demands for food have increased pressure on ecosystems including those in the buffer zone (Til man and Clark, 2014) . The increased human demands have caused unsustainable extraction of natural resources and biodiversity loss in many places (Liu et al., 2016b; Rands et al., 2010 ) . We find PAs managed with mixed uses have higher agricultural yields in the buffer zones than those managed primarily for biodiversity conservation, while the proportion of agricultural areas in the buffer zones does no t differ significantly across management strategies (Fig ure 2. 2B and C). Population density surrounding PAs managed for mixed uses is also higher than those managed primarily for biodiversity conservation (Fig ure 2. 2A). PAs managed primarily for biodiversi ty conservation have higher biodiversity and more water supply as well as lower anthropogenic pressures than those managed for mixed uses. Since anthropogenic pressures in the buffer zones mainly arise from the population density, land suitability for agri culture (e.g., slope, fertility and climate) and the demand for food production with urban development, these pressures could be reduced by more sustainable agricultural activities (Foley et al., 2011) . Because much of the demand for tourism comes from areas distant from PAs, applying integrated conceptual frameworks such as telecoupling (socioeconomic and environmental interactions over distances) can help develop a more holistic and refined analysis of changes in 26 tourism supply and demand and their impacts on biodiversity over various temporal and spatial scales (Liu et al., 2013; Liu et al., 2016a) . From the perspective of the telecoupling framework, nature - based tourism is a telecoupled system with complex interactions among local biodiversity, regional to global origins of nature - based tourism, inte rnational networks discussing and advocating management strategies for PAs and global changes in the supply and demand for ES (Liu et al., 2015a; Liu et al., 2015b) . Disentangling these influences will require careful analysis of their dynamics over time. Here we have taken a first ste p by examining, in particular, how biodiversity, management strategy and the characteristics of the buffer zone surrounding a PA influences tourism and in turn how the buffer zone and management influence biodiversity. 27 CHAPTER 3 INTERNATIONAL TOURISM DYNAMICS IN A GLOBALIZED WORLD: A SOCIAL NETWORK ANALYSIS APPROACH In collaboration with Anna Herzberger, Kenneth Frank, and Jianguo Liu 28 Abstract A complex network of tourism has emerged in the globalized world, but there is little research on the dynamics of global tourism networks and the underlying forces that affect those dynamics. Using international tourism data for 124 countries between 2000 and 2013, we integrated cluster analyses and so cial network models to identify the structure s of global tourism networks and uncover factors affecting changes in international tourist flows. R esults indicate that global tourism networks have become highly consolidated over time and that reduced transac tion costs (e.g., language, distance, and visa policies) are more important in attracting international tourists than natural and cultural attractions. Furthermore, cost of living differences between countries decreased in importance over time. Finally, in ternational tourist flows are resilient to political instability and terrorism risks. Our approach and findings highlight the key strategic factors for decision - making to implement proactive tourism policies. 29 3.1. Introduction Globally, tourism is boomin g, generating complex global networks with expanding economic power that consume s increasingly larger resources (Glaesser et al., 2017; Higham and Mill er, 2018; Song et al., 2017) . The globalization of tourism is increasing the interdependence between sending systems (supply areas, origins, departures) and receiving systems (demand ar eas, destinations, arrivals) worldwide, contributing to socioeconomic and environmental ties across regions (Dwyer, 2015; Gl aesser et al., 2017; van der Zee and Vanneste, 2015; von Bergner and Lohmann, 2014) . The proportion of the world economy occupied by tourism is rapidly increas ing, accounting for approximately 10% of global GDP and employment in 2017 (Scott and Gössling, 2015; World Travel and Tourism Council, 2018) . In addition, annual global tourism consumes approximately 16,700 PJ of ene rgy, 138 km 3 of fresh water, 62,000 km 2 of land, 39.4 Mt of food, and leads to 4.5 Gt of CO 2 emissions (Gössling and Peeters, 2015; Lenzen et al., 2018) . As tourism encourages extensive interactions betwe en human and natural systems (Jones et al., 2016; Liu et al., 2015a) , the tourism sector contains many opportunities to enhance global sustainability regarding job creation, economic growth, and environmental protection (Jones et al., 2016; Scheyvens, 2018; World Tourism Organization, 2018) . These trends raise important questions about the impacts of the growing connectivity and interdependency of globalized tourism networks, yet research has not kept pace with these changes. A holistic conceptualization and quantification is therefore urgently needed. In a globalized world, tourist flows fluctuate in response to a variety of socioeconomic and environmental factors across regions, which complicate tourism management by making supply and demand difficult to predict (Albrecht, 2013; Liu et al., 2015a; Song et al., 2017; van der Zee and Vanneste, 2015; von Bergner and Lohmann, 2014) . Historically, international 30 tourism mostly occurred between high - income countries, but in the mid - 1990s international tourist arrivals increased rapidly in middl e - and low - income countries (Scott and Gössling, 2015) . Some of the tourism to middle and low income countries may have been nature - based and cultural tourism, but the effectiveness of conservation efforts (e.g., protected areas and World Heritage sites) in attracting more inte rnational tourists is uncertain (Cellini , 2011; Cuccia et al., 2016; Patuelli et al., 2013; Yang and Lin, 2011; Yang et al., 2010) . There is also ongoing debate as to whether international tourism is resilient to political instability and terrorism risks (Liu and Pratt, 2017; Saha and Yap, 2013; van der Zee and Vanneste, 2015) . Thus, tourism studies should explore the increased complexity of global tourism networks and how they respond to natural resources and social and political conditions . Until now, quantitative research has been lacking to understand how the dynamics of global tourism networks have changed over time and how these networks affect, and are affect ed by, tourism supply and demand . Social network analysis is a sophisticated way to quantify the network structures of the tourism sector (Albrecht, 2013; Casanueva et al., 2016) . Social network analysis also proves useful for uncovering the drivers of tourist flows in both sending and receiving systems (Albrecht, 2013; Merinero - Rodríguez and Puli do - Fernández, 2016) . However, most tourism stud ies that use social network analysis concentrate on the structural characteristics of personal and organizational networks (e.g., density, centrality, and clusters) in the destinations (Casanueva et al., 2016; van der Zee and Vanneste, 2015) . In addition, although many network models have been developed to estimate both network dependencies (e.g., r eciprocation ) and the drivers of network structures with statistical inference (e.g., standard error s, p - values, or posterior distributions) (Snijders, 2011) , little tourism research applies network models to investigate the environme ntal and socioeconomic drivers of tourism. 31 To fill this research gap, we integrate a social network model with cluster analysis to uncover the network structure of international tourist flows and examine the factor s influencing international tourism. Utilizing longitudinal data, the network model identifies the influence of environmental and socioeconomic factors on international tourism while accounting for statistical dependencies within global tourism network s . We answer two q uestions: (1) How has the network structure of international tourism changed over time?, and (2) Which factors contribute to increased international tourist flows over time? By establishing a theoretical foundation within a social network framework, we qua ntify the spatial and temporal changes of global tourism networks. On a practical front, measuring network dependencies and the factors involved in global tourism networks on both the supply and demand sides provide s valuable insights for researchers, poli cymakers, and stakeholders implementing tourism development and destination management in a globalized world. The next section begins with a literature review of social network analysis in tourism and factors that contribute to international tourism. The t hird section describes the data collection, processing, and network methods. The fourth section presents results from global - level network analyses. The paper concludes with a discussion of the theoretical and practical implications of employing these meth ods for future research and decision - making. 3.2. Literature Review O ur approach is based on an application of network science to describe international tourist flows as a network. This section is a narrative review that covers three topics: 1) the theore tical background of social network analysis in tourism, 2) the application of social network 32 analysis to investigations of the dynamics of global tourism networks, and 3) the environmental and socioeconomic factors of international tourism used in this stu dy . 3.2.1. Social network analysis Social network analysis uses network and graph theory to investigate social structures (Baggio et al., 2010; Otte and Rousseau, 2002; Wasserman and Galaskiewicz, 1994) . Social networks form a relationa l structure of ties (or edges) between actors (or nodes), such as friendships between individuals or trade between countries (Albrecht, 2013; Snijders, 2011) . Similarly, international tourism forms a relational network by connecting the sending system (supply area, origin, departure) to attract ions in the receiving system (demand area, destination, arrival) that is manifest in tourist flows (Albrecht, 2013; Sainaghi and Baggio, 2017) . The use of social network analysis to analyze tourism has grown rapidly over the last two decades (Baggio et al., 2010; Casanueva et al., 2016; Pulido - Fernández and Merinero - Rodríguez, 2018) . Importantly, such approaches allow for the examination of both tourism supply perspectives (Pulido - Fernández and Merinero - Rodríguez, 2018; Sainaghi and Baggio, 2017) and tourism demand perspectives (Money, 2000; Tyler and Dinan, 2001) . However, most tourism literature that uses social network analysis has focused on personal and organizational networks in tourism destinations (tourism supply - side) (Casanueva et al., 2016; van der Zee and Vanneste, 2015) . For example, tourism studies have used social networ k analysis to investigate effects of collaborations among tourism stakeholders (Baggio, 2011; Pulido - Fernández and Merinero - Rodríguez, 2018) , marketing (Bhat and Milne, 2008; Wang and Xiang, 2007) , sustainable tourism (Albrecht, 2013) , and geography (Jin et al., 2017; Lee et al., 2013) in tourism destinations. 33 Additionally, the most com monly used methods of social network analysis in tourism studies are concentrated on investigating static structural network proper ties (e.g., size , density, betweenness, and clusters ) (Baggio et al., 2010; Benckendorff and Zehrer, 2013; Lee et al., 2013; Pulido - Fernández and Merinero - Rodríguez, 2018; Raisi et al., 2017; Scott et al., 2008b) . Although tourism network properties may change significantly over time (Westveld and Hoff, 2011) , few tourism studies have included any quantitative analysis of longitudinal datasets using a social network analysis approach (Baggio and Sainaghi, 2016; Jin et al., 2017) . Recent exceptions include bibliometric network visualizations showing changes in tourism research output over time (Güzeller and Çeliker, 2018; Jiang et al., 2017; L i et al., 2017) . Social network analysis accounts for dependencies among ties between set s of actors (e.g., reciprocity and transitivity) (Snijders, 2011) . For example, i nternational tourism leads to dependence between send ing and receiving countries if two countries have reciprocal tourism flows. Various statistical models have been developed to capture network dependencies between actors (Snijders, 2011) . These statistical network models can estimate parameters to express network structures with statistical inference (e.g., standard errors, p - values, or posterior distributions). The p 2 network model has been shown to yield a robust estimation procedure that accounts for network dependencies associated with common senders and receivers of network ties as well as potential reciprocal relationships between pairs of actors (Hoff, 2005; van Duijn et al., 2004) . The p 2 model parameters are estimated with Bayesian inference based on a Markov Chain Monte Carlo (MCMC) algorithm (Hoff, 2005; van Duijn et al., 2004) . Bayesian inference is a method for statistical inference used to compute the conditional probability of an event after taking into account new evidence or information that the event has occurred (Gamerman and 34 Lopes, 2006) . The MCMC is a mathematical method for generating the probabilit y distribution of a parameter by randomly sampling from a complex probabilistic space (Andrieu et al., 2003) . Social networks also contain temporal dependencies, wherein changes in network ties depend on the earlier structure of network ties (e.g., the evolution of interna tional tourism networks) (Hoff, 2015; Snijders, 2011; Ward and Hoff, 2007) . Longitudinal network data with regular temporal intervals are often referred to as network dynamics (Snijders , 2011) . For longitudinal network data, statistical modeling approaches such as ordinary least squares and generalized linear models risk overestimating the significance of parameters by ignoring network and temporal dependencies with the assumption of independence (Westveld and Hoff, 2011) . But Westveld and Hoff (2011) developed a mixed - effects model to account for both network and temporal dependencies as a stochastic process. The mixed - effects model extended the p 2 model of van Duijn et al. (2004) and Hoff (2005) . This model (1) uses a latent space approach to produce visualizations of the network structure with the latent space positions, (2) develop s a generalized linear modeling framework that allows for continuous data, and (3) outline s a general Bayesian estimation approach for model parameters with the MCMC algorithm (Westveld and Hoff, 2011) . Despite recent developments in s ocial network models, these models have not been much used in tourism studies. With the social network model for longitudinal data, we provide a unique perspective on the dynamics of global tourism networks by quantifying both network and temporal dependencies. We also integrate s ocial network modeling and cluster analysis to examine which environmental and socioeconomic factors influence changes in inter national tourist flows across countries. Thus, the application of social network model s in tourism studies 35 provides a better orientation to understand the processes of tourism development and destination management worldwide. 3.2.2. Hypothesized factors a ffecting international tourism Following previous studies that investigated factors shaping tourism demand (Balmford et al., 2015; Lim, 1997, 1999; Marrocu and Paci, 2013; Peng et al., 2014; Song et al., 2012a; Song and Li, 2008; Witt and Witt, 1995) , the most widely used factors affecting in ternational tourism were considered for inclusion in the social network model regarding the characteristics of sending countries, receiving countries, and their pairs. These factors represent environmental, political, social, economic, and demographic feat ures in both sending and receiving countries. We note that the factors used in tourism demand models may change extensively, depending on the research questions, time periods, methodologies, and selection of countries (Dogru et al., 2017) . B ased on the above literature review, we examine whether transaction costs (e.g., language, ge ogra phic distance, and visa policy ) and demographic forces (e.g., population and income growth) are more important in attracting international tourists than natural and cultural attractions (e.g., protected areas and World Heritage sites) and political stability . First , transaction costs of travel include visa - free status, national price - level difference, shared language, and proximity. I nternational tourists prefer to travel to visa - free countries. Visa restrictions and requirements in de stination countries can have a negative impact on the number of tourist arrivals (Balli et al., 2013; Cheng, 2012; Neumayer, 2010) . Additionally , international tourists prefer to travel to countries that have advantageous prices relative to their ho me countries (Cheng, 2012; Dogru et al., 2017; Saha and Yap, 2013) . There are two types of measurements for price level differences in the tourism demand model: 1) relative prices of the 36 place of origin to the prices in the destination and 2) substitute prices of the destination to the prices in competing destinations (Dogru et al., 2017; Kronenberg et al., 2016) . As a measurement of the price - level differences between countries, relative price standardized by exchange rate s has been found to be more significant than the exchange rate alone (De Vita and Kyaw, 2013; Dogru et al., 2017) . I nternational tourists also prefer to travel to countries that use the same language as their home country. Thus, a shared language between sending and receiving countries plays an essential role in promoting tourist flows (Eilat and Einav, 2004; Khadaroo and Seetanah, 2008) . Finally, international tourists pref er to travel to nearby countries. Greater distances between sending and receiving countries have a negative impact on international tourist flows (Lim, 1999; Patuelli et al., 2014; Yang et al., 2010) . The number of direct flights between countries also contributes to increases in international tourist flows (Lohmann et al., 2009; Rehman Khan et al., 2017) . Second, demographic forces include population size and GDP per capita. P opulation and income per capita are important determinants for international tourist arrivals and departures. Tourism studies typically use real GDP per capita and populat ion as proxies for relative income and market size (Lim, 1997; Peng et al., 2014; Witt and Witt, 1995) . Higher per capita GDP in both sending and receiving countries positively affect international tourist flows (Lim, 1999; Saha and Yap, 2013; Song et al., 2010) . Interna tional tourist flows also increase in sending and receiving countries with higher populations (Khadaroo and Seetanah, 2008; Llorca - Viv ero, 2008; Yang et al., 2010) . Third, many tourism studies have investigated the role of conservation efforts (e.g., protected areas and World Heritage sites) for touris m demand (Song et al., 2012a) . Larger protect ed areas have been found to attract more tourists (Balmford et al., 2015; Chung et al., 37 2018a) . Protected areas are good at attracting nature - based tourists while conserving biodiversity (Balmford et al., 2015; Chung et al., 2015) . Furthermore, nature - based tourism often contributes to the management and conservation of protected areas by provid ing financial resources (Buckley et al., 2015; Buckley et al., 2017) . However, there is an ongoing debate regarding the effectiveness of World Heritage sites in promoting tourist arrivals (Cellini, 2011; Cuccia et al., 2016; Patuelli et al., 2013; Yang and Lin, 2011; Yang et al., 2010) . While some studies show that the presence of World Heritage sites attracts more visitors due to proper management and accessibility (Richards, 2011; Su and Lin, 2014; Yang et al., 2010) , others show that World Heritage sites do not affect the number of tourist arrivals (Cellini, 2011; Cuccia et al., 2016, 2017) . Fourth , empirical research lacks agreement regarding the effects of political instability and terrorism risks on b oth international tourist arrivals and departures (Liu and Pratt, 2017; Saha and Yap, 2013; van der Zee and Vanneste, 2015) . Some studies have found that political instability and terrorism risks (e.g. public violence, riots, civil wars, and military coups) negatively influence international tourist ar rivals (Eilat and Einav, 2004; Llorca - Vivero, 2008; Sah a and Yap, 2013; Sönmez, 1998) . But others have claimed that international tourists are resilient to political instability and terrorism risks (Liu and Pratt, 2017; van der Zee and Vanneste, 2015) . 3.3. Materials and Methods 3.3.1. Data collection Data on international tourist arrivals were obtained from the UN World Tourism Organization (UNWTO). This raw dataset covers over 200 countries from 1995 2013. UNWTO 38 defines visitors to include both tourists (overnight visitors) and excursionists (same - day visitors) (World Tourism Organization, 2016a) . Following UNWTO methods for estimating the number of international tourists, we excluded excursionists prior to selecting 124 countries over the period from 2000 2013 for analysis. The selected countries cover appro ximately 90% of international tourist arrivals in the specified time period. Although the UNWTO data are the best available international tourist arrival datasets, the UNWTO dataset has some weaknesses inherent in how different jurisdictions collect visit or arrival data (World Tourism Organization, 2016a) . When countries did not report international tourist arrivals at national borders (referred to as TF), we supplemented by using other datasets following UNWTO methods: international visitor arrivals at national borders (VF), international tourist arrivals at hotels and similar establishments (THS), or international tourist arrivals at collective tourism establishments (TCE) (World Tourism Organization, 2016b) . To test our hypotheses, we collected data regarding possible factors influencing internati onal tourism: transaction costs of travel, environmental, political, and demographic factors . Transaction costs of travel included visa requirements for tourism, price level ratio to the market exchange rate, shared language, and geographic distances betwe en sending and receiving countries. At the level of the pair of countries, the visa - free score is 1 if a receiving country waives visa requirements for a sending country, including both visa - free and visa - on - arrival entry ( https://www.passportindex.org (The World Bank, 2017) . The price - level differences between countries were calculated by subtracting the price level ratio of each receiving country from each sending country. Countries having a shared language was also included, where if two countries share an official language 39 (e.g., Canada and the United Kingdom), their language factor was 1. Geographic d istances between the centroids of pairs of countries were calculated using GeoDa (Anselin et al., 2006) and remained constant over the study period. The number of direct flight s between countries was obtained from Openflights ( https://openflights.org ). Environmental factors included the size of protected areas in receiving countries (IUCN and UNEP - WCMC, 2017) , restricted to protected areas that are legally and officially designated at the national or sub - national level. Marine protected areas were excluded as well as the International Union for Conservation of Nature (IUCN) Category I protected areas, where tourism is pre vented for strict conservation. Additionally, World Cultural Heritage sites were included as an environmental factor (UNESCO, 2017) . World Natural Heritage sites were excluded to avoid double counting a site. Protected areas and World Cultural Heritage sites were used to represent a cou (Balmford et al., 2015; Chung et al., 2018b; Yang et al., 2010) . Political factors included the index of political stability and the absence of violence and terrorism (The World Bank, 2017) . The index of political stability and the absence of violence and terrorism measures the likelihood of polit ical instability and politically motivated violence, ranging from - 2.5 to 2.5 (The World Bank, 2017) . In both sending and receiving countries, population size was a demographic factor (The World Bank, 2017) . P er capita GDP was included as an a dditional economic factor (The World Bank, 2017) . 3.3.2. Cluster analyses We used Kliquefinder software to identify clusters of countries within global tourism ne tworks (Frank, 1995, 19 96) . The raw data for this analysis consist of the total tourist flows 40 between each pair of countries over a given interval. The algorithm maximizes the odds ratio of flows within clusters relative to between clusters by switching actors among clusters repeatedly. Because countries in the same cluster have a higher probability of sending tourists to each other than countries in different clusters, the Kliquefinder algorithm can identify clusters of countries that can then be investigated to see if they a re focused around income level or other factors such as population or geographic location. To test the statistical significance of the clustering, Kliquefinder is applied to a random redistribution of flows. This is repeated 1,000 times, and the measure of fit is noted to generate a Monte Carlo sampling distribution under the null hypothesis of no clustering in the data (data are generated at random). The observed measure of fit is then compared to the Monte Carlo generated sampling distribution to obtain a p - value. To perform the cluster analysis, we examined the number of international tourists in two different ways: by analyzing the average of the data from the first three years (2000 2002) and the last three years (2011 2013) in the UNWTO datasets, and b y analyzing each year (from 2000 2013) separately. Although some significant political, social, and natural events occurred during the study period (e.g. the events of 9/11, the Indian Ocean tsunami, global financial crisis, the Arab Spring, and Olympics e vents), we consider our analyses to be valid because there are rarely multi - year periods in which a significant political, social, or natural event does not occur somewhere in the world. Furthermore, we believe that the use of both 3 - year averages and sing le - year data accounts for such occurrences. We performed cluster analyses with Kliquefinder for each of the temporal periods, and tested for evidence of clusters in each period. For the results from the analyses of both the 3 - year periods and each individu al year, we used the igraph package in R to visualize the cluster results. In the graphs, we used the number of international tourists to identify the core and peripheral countries in global tourism networks 41 based on the k - core decomposition approach, an i terative approach that determines the most central nodes by consecutively cutting out the least connected nodes in a given network (Barberá et al., 2015) . We also presented the cluster results by country on a global map using ArcGIS (ESRI, 2015) . 3.3.3. Mixed - effects model In addition to the cluster analyses, we used a mixed - effects (including random - and fixed - effects) model for longitudinal tourism network data with the number of international tourist arrivals from a sending country to a receiving country as the dependent variable. The mixed - effects model was developed by Westveld and Hoff (2011) to account for both network and temporal dependencies. Westveld and Hoff (2011) provided R code script that we deployed using the MCMCpack package in R. The results provide means and regression estimates of the factors affecting global tourism networks, as well as evidence of statistical dependencies. By using a generalized linear mo del framework, this model can adopt the gravity approach described in the next paragraph, which models the set of bilateral tourist flows (Khadaroo and Seet anah, 2008; Morley et al., 2014; Westveld and Hoff, 2011; Yang et al., 2010) . Since tourism is a type of trade in services, tourist flows can also be analyzed using the gravity approach for bilateral trade (Cheng, 2012; Eilat and Einav, 200 4; Kimura and Lee, 2006; Morley et al., 2014) . The gravity model has been widely appl ied on both the tourism supply and demand sides over the last decade (Marrocu and Paci, 2013; Morley et al., 2014) . The gravity model of internation al trade can be derived from the Heckscher - Ohlin theory based on international differences in factor endowments (Deardorff, 2007) . Furthermore, Morley et al. 42 (2014) derived a theoretical framework to support the gravity model for bilateral tourist flows by using consumer util ity theory. The gravity model assumes that international tourist flows between sending and receiving transportation costs between countries (e.g., distance) (Eilat and Einav, 2004; Khadaroo and Seetanah, 2008; Witt and Witt, 1995) . Some studies also include some dummy variables (e.g., visa requirements or shared language) in addition to the gravity model (Eilat and Einav, 2004; Neumayer, 2010) , an approach we followed in our study. The basic gravity model for bilateral trade is shown as: (1) where T i j is the amount of trade flows between two regions i and j; m i and m j are the ij is the geographical distance between region i and region j; E ij is a normal distributed error ter m; and B, , , and are coefficients to be estimated. By taking the natural log transformation in equation (1), the basic gravity equation for estimation purposes can be expressed as follows: (2) where is a residual error term. We applied this gravity model to annual international tourist arrivals between 124 countries for each year from 2000 2013. The model for longitudinal tourist flows is: 43 (3) where International Tourists i,j,t is the number of international tourist arrivals from s ending country i to receiving country j at time t ; PA j,t is the size of protected areas in the receiving country at time t; World Heritage j,t is the size of World Cultural Heritage sites in the receiving country at time t; Political Stability i,t and Politi cal Stability j,t are the political stability and absence of violence and terrorism indices for the sending and receiving countries at time t , respectively; Visa Free i,j is the visa - free score between the sending and receiving country; Language i,j is shared language factor between the sending and receiving countries; D i,j,t is the geographic distance from the centroid of country i to the centroid of country j ; Price Level i,j,t is the national price level difference between sending and receiving countries; GD P i,t and GDP j,t are the per capital GDP in the sending and receiving countries at time t , respectively; Pop i,t and Pop j,t denote the population size in the sending and receiving countries at time t , respectively; s i,t is a sender effect; r j,t is a receiver effect; and g i,j,t is a residual error term. The sender (s i,t ) and receiver (r j,t ) random effects measure the average deviations of the levels of tourist arrivals and departures in each country. With these effects, we can identify which countries are the most or least active in global tourism networks. In international tourism, International Tourists i,j,t is the directed flow from sending country i to receiving country j at time t ; thus International Tourists i,j,t is not equal to International Tourists j,i,t . 44 For the sake of clarity, we also used an alternative model, which compared the proportion of protected areas and World Cultural Heritage sites to the total land area of a country instead of compared to the absolute size of protected areas a nd World Cultural Heritage sites. This alternative model also included the number of direct flights between countries instead of the geographic distance. To estimate both models, an MCMC algorithm iterated 11,000 times, and we dropped the first 1,000 itera tions to allow convergence to the stationary distribution. Our model parameters were automatically saved every 10th scan. Then, we calculated means and 95% confidence regions of the parameters using the joint posterior distribution. For 95% confidence regi ons, we used Highest Posterior Density (HPD) interval. 3.4. Results This section presents the results of our global - level network analyses in two parts: cluster analyses and the social network model. The first part of the analyses began by examining the n etwork structure of international tourism in the two temporal periods (2000 2002 and 2011 2013) and in each year (from 2000 2013) individually. To test the sensitivity of our cluster results to the choice of the temporal periods, we examined the network st ructure of international tourism in each year from 2000 2013. The second part of the analyses determined which factors contributed to changes in international tourist flows over time and quantified network and temporal dependencies in global tourism networ ks. 45 3.4.1. Consolidated global tourism networks While global tourism networks from 2000 2002 were divided into eight clusters (Figure 3. 1 A ), the network structure from 2011 2013 had only two clusters (Figure 3. 1 B ). Figure 3. 1 also identified the core and peripheral countries in global tourism networks. The core countries (e.g., USA and western European countries) located in the center played active roles in both tourist arrivals and departures. At the first time point (2000 2002), the largest cluster included 54 countries highlighted by yellow circles in Figure 3. 1 A . All high - income countries were located in this group. The remaining seven cl usters included middle - and low - income countries, grouped by geographic locations (the Caribbean Sea, central and southern America, southern Africa, eastern and western Africa, central Asia, southern Asia, and eastern Europe) (Figure S3. 1 A ). The dominant c luster sent a large number of tourists to countries within the same cluster (red lines in Figure 3. 1 A ) and to the other seven clusters (gray lines in Figure 3. 1 A ). Interestingly, over the period of 2011 2013, the dominant cluster expanded to include 121 co untries. The consolidated cluster contained all countries in our dataset, excepting only Burkina Faso, Niger, and Togo in western Africa (Figure S3. 1 B ). The cluster results for each individual year from 2000 2013 also indicated the same pattern that global tourism networks have become consolidated over time (Figure S3. 1). Specifically, the number of clusters in 2009 was highest (12 clusters) over the 14 - year period, followed by 2004 (11 clusters). These clusters were mainly based on geographic location (Fig ure S3. 1). After 2009, the number of clusters decreased, from nine in 2010 to two in 2012 (Table S3. 1). Informed by the Monte Carlo sampling distribution, we confirmed the existence of clusters in global tourism networks in each time period (Table 3. 1 and Table S3. 1, P<0.001). 46 Figure 3 . 1 . Clusters of global tourism networks in (a) 2000 2002 and (b) 2011 2013. The size of each node indicates the sum of international tourist arrivals and departures. Red ties indicate tourist flows within the same cluster, and gray ties indicate tourist flows between different between countries based on the number of internat ional tourists. The core countries were located in the center of the cluster maps . Table 3 . 1 . Odds ratios for cluster analysis and p - value based on simulations followed by mean, median, and 95% Quantile inter val of simulations . N Odds ratio p - value M ean median 2.5% 97.5% 2000 2002 8 0.792 <0.001 0.596 0.604 0.365 0.636 2011 2013 2 0.828 <0.001 0.561 0.563 0.532 0.591 3.4.2. Factors related to international tourism By using a mixed - effects model, we were able to estimate the effect of each independent variable on international tourist arrivals, as well as of network and temporal dependencies within global tourism networks. Figure 3. 2 shows the mean for each coefficie nt and its 95% HPD confidence intervals from 2000 2013. Regarding receiving countries, the size of protected areas and World Cultural Heritage sites did not have a significant relationship with international tourist flows. From 2000 2013, the coefficients for protected areas and World Cultural Heritage sites changed little and their confidence intervals contained zero (Figures 3. 2 A and 2 B ). In the 47 alternative model, the proportions of protected areas and World Cultural Heritage sites to the total land area were also not statistically significant (their confidence intervals contained 0) between 2000 and 2013 (Figure S3. 2). Regarding sending countries, the coefficients for political stability and absence of violence and terrorism did not shift, and their intervals consistently contained zero (Figure 3. 2 C ). However, with respect to receiving countries, the coefficients of poli tical stability and absence of violence and terrorism declined from 2000 2011 and then shifted upward from 2011 2013 (Figure 3. 2 D ). Third, the coefficients for visa - free score and shared language were positive over the entire study period (Figures 3. 2 E an d 2 F ). There was an increase in the coefficients for visa - free score from 2000 2013. Fourth, international tourists prefer to travel to nearby countries. Geographic distance between sending and receiving countries was negatively associated with the number of international tourists from 2000 2013 (Figure 3. 2 G ). In the alternative model, the number of direct flights between sending and receiving countries also was positively associated with the number of international tourists over time (Figure S3. 2). Fifth , the coefficients for price level difference between sending and receiving countries declined over the study period (Figure 3. 2 H ). The confidence intervals were positive from 2000 2009 but contained zero from 2010 2013. Sixth, in sending and receiving co untries, higher income levels increase the number of both international tourist arrivals and departures. The coefficients for per capita GDP in sending countries increased over time (Figure 3. 2 I ). In receiving countries, the confidence intervals for per ca pita GDP shifted upward (Figure 3. 2 J ). 48 Finally, in both sending and receiving countries, population size was positively associated with the number of international tourists. Over the study period, all population coefficients were positive, and their inter vals were consistently above zero (Figures 3. 2 K and L ). This trend suggests that international tourism between countries with high per capita GDP and rapid population growth was above the global average. In receiving countries, the inferences we would make regarding per capita GDP and population size were more uncertain than for those in sending countries because of the larger confidence intervals over time. 49 Figure 3 . 2 . Mean and 95% Highest Posterior Density (HPD) confidence intervals of the coefficients from 2000 2013: (a) the size of protected areas in receiving countries (km 2 ), (b) the size of World Cultural Heritage sites in receiving countries (km 2 ), (c) political stability in sending countries (index), (d) political stability in receiving countries (index), (e) visa - free status between sending and receiving countries (visa - free=1), (f) shared language between sending and receiving countries (shared language=1), (g) distances be tween countries (km), (h) national price level difference between sending and receiving countries (price - level ratio), (i) per capita GDP in sending countries (constant 2010 US $), (j) per capita GDP in receiving countries (constant 2010 US $), (k) populat ion size of sending countries (person), and (l) population size of receiving countries (person). 50 Phi parameter estimates identified the auto - regressive effect of the previous year on tourist arrivals, departures, and reciprocity of the current year (Table 3. 2). The medians of the posterior distribution of and were 0.998 and 0.003. This means that the number of international tourist departures in the current year highly depended on the level of international tourist departures from the previous yea r. Yet international tourist arrivals in the previous year did not have an impact on the current international tourist departures. In addition, the medians of and are 0.967 and 0.004, respectively. When countries had a high number of internationa l tourist arrivals in the previous year, they also tended to have large international tourist arrivals in the current year. However, the number of international tourist arrivals in the current year did not depend on the number of international tourist depa rtures in the previous year. Finally, the median of was 0.014. This indicates that the level of reciprocity in the previous year may not explain the level of reciprocity in the current year. Table 3 . 2 . Phi parameter estimates with median and 95% Quantile intervals. Parameter Median 2.5% 97.5% 0.998 0.996 0.999 0.003 0.000 0.005 0.004 - 0.009 0.018 0.967 0.958 0.975 0.014 0.013 0.016 In 2000 2002 and 2011 2013, sender and receiver random effects were investigated at the country level (Figure 3. 3). The random effects estimated the deviations of the number of international tourist arrivals from the predicted values by the mixed - effects m odel. The positions of the countries changed slightly from 2000 2002 to 2011 2013. USA, Canada, and Australia played crucial roles as both senders and receivers in global tourism networks, even after accounting for controls in the regression model. From 20 00 2013, China, Spain, and Russia 51 became active tourists - senders while South Africa, India, Malaysia, and Maldives became active tourists - receivers. Over the period of 2011 2013, China and Russia emerged as both important senders and receivers in global to urism networks. Figure 3 . 3 . Distributions of (a) the sender effects in 2000 2002, (b) the sender effects in 2011 2013, (c) the receiver effects in 2000 2002, (d) the receiver effects in 2011 2013. Country abbreviations: Australia (AUS), Belgium (BEL), Canada (CAN), Switzerland (CHE), China (CHN), Germany (DEU), Spain (ESP), France (FRA), United Kingdom (GBR), India (IND), Italy (ITA), Japan (JPN), Republic of Korea (KOR), Maldives (MDV), Malaysia (M YS), Netherlands (NLD), New Zealand (NZL), Russian Federation (RUS), Thailand (THA), United States (USA), and South Africa (ZAF). 52 3. 5 . Discussion 3.5.1. Reasons behind consolidated global tourism networks Using cluster analysis and a mixed - effects model for longitudinal network data, we investigated the flows and factors relating to international tourism. S ocial network analysis helped examine how international tourism connects regions and identify temporal changes in the network struc ture. Results of our cluster analysis show that international tourist flows form a consolidated network over time (Figure 3. 1). Sender and receiver random effects from the mixed - effects model then revealed which countries played increasingly active roles i n the consolidated networks (Figure 3. 3). Another finding of the mixed - effects model may indicate a causal relationship between the changes in global tourism networks in Figure 3. 1 and Figure S3. 1 and the factors in Figure 3. 2. From 2000 2009, the price le vel difference between sending and receiving countries was a major factor of international tourist flows based on the law of demand. This result is consistent with previous studies (De Vita and Kyaw, 2013; Dogru et al., 2017) . However, after 2010, the price level difference became a less important factor for international tourism. This result shows that middle - and lo w - income countries with rapid income and population growth, such as China, increasingly play an important role as sending countries (see also Buckley et al. (2015); Scott and Gössling (2015) ). Despite the price level differences, developing countries send more tourists to both deve loped and developing countries. In sending countries, per capita GDP and population size were the most significant factors for international tourism (Song et al., 2010; Song and Li, 2008; Yang et al., 2010) . Per capita GDP and population size represent the effects of income l evel and market size differences between sending and receiving countries. In the consolidated networks, the roles of these factors 53 in sending countries intensify over time (Lim, 1997; Peng et al., 2014; Witt and Witt, 1995) . In receiving countries, although per capita GDP and population size are significant (Khadaroo and Seetanah, 2008; Saha and Yap, 2013) , the uncertainty of the effects of these factors is high (i.e., large confidential intervals). While over half of interna tional tourists visit high - income countries, increasing arrivals in new destinations such as Malaysia, a middle - to low - income country with a large population, led to the uncertainty of coefficients. International tourist flows are complex and dynamic syst ems affected by many other factors that were not measured in our study. For example, global crisis events such as economic and financial downturns, political instability, terrorist attacks, and natural disasters can affect the size and frequency of interna tional tourist flows (Hall, 2010) . Our results may indicate that glo bal crisis events have dispersed the consolidated global tourism networks, based on geographic locations (Figure S3. 1). The global financial crisis from 2007 2010 may have caused the rapid increase in the number of clusters by weakening the interdependence between distant countries (see also Campos - Soria et al. (2015); Hall (2010) ). In 2004, global tourism networks were separated into 11 clusters, in part because of outbreaks of severe acute respiratory syndrome (SARS) and the Indian Ocean tsunami (see also Hall (2010); Kuo et al. (2008) ). These types of global events may also contribute to the uncer tainty of some coefficients (e.g., political stability variable) in the mixed - effects model. 3.5.2. The role of conservation in international tourism Although nature - based and cultural tourism are the fastest growing sectors in the tourism industry (Newsome et al., 2012; World Tourism Organization, 2015) , the presented results show that efforts to conserve natural and cultural sites were not significant factors contributing to the 54 number of international arrivals in receiving countries. The results from the mixed - effects model for the proportions of protec ted areas and World Cultural Heritage sites show that neither were significant factors between 2000 and 2013. Within a given country, protected areas have varying success in attracting international tourists from different regions and over time (Diefendorf et al., 2012; Su and Lin, 2014) . Some protected areas have higher levels of domestic tourist arrivals than international tourist arrivals (Chung et al., 2018b) , whereas other protected areas attract more international tourists than domestic tourists (Baral et al., 2017) . These varying patterns of international tourist arrivals may have led to an insignificant result in the mixed - effects model. In addition, many protected areas are located at high altitudes, far from the major urban areas from which most international tou rism emanates (Chung et al., 2018a; Joppa and Pfaff, 2009) . The remoteness of protected areas may prevent visits from international tourists (Chung et al., 2018a) . Due to different numbers of international tourist arrivals, decision - makers may need to establish different mana gement plans to increase tourism while protecting the environment effectively. For example, protected areas that successfully attract domestic tourists may lack the transportation infrastructure for international tourists. If decision - makers aim to increas e international tourism, such protected areas will need additional infrastructure investment to increase accessibility from airports or train stations. However, further infrastructure development could have a negative environmental impact, and therefore sh ould be considered as a part of management and conservation strategies. Furthermore, World Cultural Heritage sites were not effective in attracting international tourists in accordance with the findings of Cellini (2011) , Cuccia et al. (2016) , and Cuccia et al. (2017) . This is consis tent with the main purpose of World Cultural Heritage sites, which is not to 55 - (Cellini, 2011; Cuccia et al., 2016; Su and Lin, 2014) . In addition, the increase in international tourist arrivals in middle - and low - income countries that have few W orld Cultural Heritage sites may reduce the attraction of World Cultural Heritage sites for international tourists because over half of World Cultural Heritage sites are based in high - income European countries (Su and Lin, 2014) . Although World Cultural Heritage sites are ineffective for international tourism, there are ongoing efforts to encourage cultural tourism to World Cultural Heritage si tes. In the rapidly globalizing tourism network, one of the major challenges at World Cultural Heritage sites is how to encourage cooperation between the tourism and culture sectors. In 2015, UNWTO and UNESCO organized the first World Conference on Tourism and Culture to initiate the sustainable development of cultural tourism (World Tourism Organization, 2016c) . 3.5.3. The impact of policies on international tourism Visa - free policies can stimulate flows of international touris ts . Between 1980 and 2015, visa openness in middle - and low - income countries increased , with fewer travel requirements than those of high - income countries (World Tourism Organization, 2016d) . The increase in visa openness in middle - and low - income countries may attract more internationa l tourists. Visa - free policies can also support sustainable economic growth because improving visa openness can contribute to an increase of tourism expenditures and create jobs without additional tourism development (Song et al., 2012b; World Tourism Organization, 2016d) . T o maximize the effects of visa openness, receiving countries need to prioritize relaxing their vis a policies for citizens of sending countries with shared language s and short travel distances. 56 Further, international tourists are resilient to political instability and terrorism risks in both sending and receiving countries. This result is consistent wit h Liu and Pratt (2017) and van der Zee and Vanneste (2015) . After 2007, international tourist arrivals in receiving countries show a complicated relationship with political instability and terrorism risks. From 2007 2011, international tourist arrivals were negatively associated with political stability and the absence of violence and terrorism in dex. Over the study period, European countries led this trend, as these European countries decreased in political stability and increased in violence and terrorism risks driven by the global financial crisis following the economic recession (Campos - Soria et al., 2015; The World Bank, 2017) . The effect was a slight decrease in international tourist arrivals in European countries. International tourist arrivals in high - income countries may be more resilient to political instability and terrorism risks than those of middle - and low - income countries (Liu and Pratt, 2017; Llorca - Vivero, 2008) . In middle - and low - income countries, political instability and terrorism risks can lead to significant decreases in international tourism due to riots and wars (Sönmez, 1998) . For example, in 2011, political changes in Middle East ern and North Africa n countries such as Egypt and Yemen led to decreases international tourist arrivals (Avraham, 2015) . As a result, the Arab Spring contributed to the uncertainty of coefficients of political stability and absence of violence and terrorism index. The emergence of the Islamic State in Iraq and Syria (ISIS) and Syrian refugee crisis gen erated terrorism risks and political tensions in both the Middle East and the rest of the world (Khan and Ruiz Estrada, 2016) . Countries that experience such events can have difficulties in tourism management and planning with unpredictable tourism demand (Issa and Altinay, 2006; Saha and Yap, 2013) . Therefore, tourism po licy makers should recognize the impacts of political instability and terrorism risks while 57 planning crisis management strategies for the tourism industry (e.g., restoration of a positive image for international tourists) (Khan and Ruiz Estrada, 2016; Saha and Yap, 2013) . 3. 6 . Conclusions Our study is the first international tourism study to adopt a social network analysis approach t hat quantifies the complex structure of global tourism networks and examines underlying factors over time. The results of our global - level network analyses have several theoretical and practical implications, including identifying emerging countries that n eed tourism policies and providing key strategic factors for tourism development and destination management in each phase of global tourism networks. From a theoretical point of view, our global - level network analyses made a significant contribution to adv ancing the application of social network analysis approach in the tourism field since to date, a limited number of tourism studies have utilized a social network approach to perform a longitudinal quantitative study at a global level. In drawing conclusion s, we should also note the limitations of our study. The most compelling limitation regards the lack of data availability at the global level. For instance, due to the lack of time - series data for the visa - free score and for the number of direct flights between countries, we assumed the same visa policy and the number of direct flights over the period from 2000 2013. Additionally, although our cluster results may indicate that global tourism networks were dispersed following global crisis events (e.g., global financial crisis), we could not detect a causal relationship between global crisis events and changes of network structure in international tourism. Second, it is noted that when using longitudinal network data, it is dif ficult to discern the most important factors because the pattern of each factor is based on variation among years 58 within a country and/or variation among countries. Third, we identified a few countries that were not predicted from the mixed - effects model. For example, although Australia has large geographic separation from other countries, Australia is the center of global tourism networks. This is because international tourism supply and demand have been influenced by many other factors across local, regio nal, and global levels. At the local and regional level s , different key factors for international tourism may require strategies different from our global implications, and therefore destination management should be flexible across regions. Future tourism network research will need to extend our methods to include hierarchical network models and examine hierarchical network structures from global to local levels. Furthermore, future tourism research should evaluate socioeconomic and environmental effects of international tourism as well as the agents that are involved in international tourism, in addition to the tourist flows and factors affecting tourism (causes) reported in this study. The new integrated framework of metacoupling (socioeconomic - environment al interactions within and between adjacent and distant systems such as countries) provides a good foundation for such future efforts as it integrates tourist flows, causes, agents, and effects across different systems (Liu, 2017) . Despite these limitations, on the practical front, quantifying the network structure of international tourism helps explore how international tourist flows are changing in the face of external social, economic, political, and environmental issues. Our clus ter results confirm the consolidation of global tourism networks and identify which countries increasingly contribute to this trend over the past 14 years. Our results support that some global crisis events (e.g., global financial crisis and the Indian Oce an tsunami) may weaken the structure of international tourist flows from consolidated networks to separated networks based on geographic location. This result indicates that social, economic, political, and environmental changes in emerging countries 59 may h ave more significant impacts on other countries in the same cluster than those in other clusters. Policy makers can utilize the results of our cluster analysis to understand the cross - border impacts of tourism development and destination management to attr act more international tourists across countries. Our mixed - effects model identifies key strategic factors for proper tourism development and destination management. In consolidated global tourism networks, results indicate that transaction costs (e.g., sh ared language, geogra phic distance, and visa policy ) are more important in attracting international tourists than natural and cultural attractions (e.g., protected areas and World Cultural Heritage sites). We suggest that middle - and low - income countries t hat increasingly depend on the tourism industry should maintain their political stability and enhance visa - free policies to encourage more international tourist arrivals. In this situation, these countries have put more effort into tourism development such as transportation and accommodation. However, a high degree of tourism development traditionally conflicts with environmental protection. One of the best ways to balance between tourism development and environmental protection is to integrate tourism deve lopment plans into conservation policies. Our results show that conservation efforts (e.g., protected areas) may contribute to balancing the benefits and risks of tourism development for international tourism, and thus avoid over - development in the long ru n. In conclusion, the presented approach and findings provide a better foundation for evidence - based decision - making to implement proactive tourism policies. 60 CHAPTER 4 RETHINKING INTERNATIONAL FOOD TRADE FOR GLOBAL BIODIVERSITY CONSERVATION In collaboratio n with Jianguo Liu 61 Abstract To achieve the United Nations Sustainable Development Goals such as food security and biodiversity, it is essential to identify their interrelationships. It is widely held that developed countries negatively affect biodivers ity in developing countries through importing food. However, through examining comprehensive datasets comprising 300 food items across 160 countries during 2000 2015, o ur results show that developed countries exported more food to developing countries than they imported from developing countries , suggesting that biodiversity in developed countries is also negatively affected by production for exports to developing countries . This is especially the case when developed countries with biodiversit y hotspots exported food to developing countries without biodiversity hotspots. Furthermore, m ost exports from developing countries , especially those with biodiversity hotspots, went to other developing countries instead of developed countries . On the othe r hand, because many developed countries with biodiversity hotspots imported food from developing countries without biodiversity hotspots, such imports might have actually benefited biodiversity in developed countries . Developing countries without biodiver sity hotspots played an increasingly important role as net exporters in international food trade. With increasing attention to food security and biodiversity (e.g., the upcoming Fifteenth Meeting of the Conference of the Parties to the Convention on Biolog ical Diversity), it is time to develop new approaches that help operationalize the post - 2020 global biodiversity framework and achieve relevant UN Sustainable Development Goals by minimizing the negative impacts of global food production and trade on biodi versity hotspots worldwide. 62 4.1. Introduction As the world pursues the ambitious Sustainable Development Goals (SDGs) including food security and biodiversity it is important to understand their interrelationships (Lu et al., 2015; Nilsson et al., 2016; United Nations, 2015) . The 17 SDGs were adopted by the UN - do list for people and planet, and a blueprint for (United Nations, 2015) . Quantitative information on the connections among SDGs is urgently needed to assess whether and how the multiple SDGs can be achieved simultaneously (Nilsson et al., 2016; Xu et al., 2020a; Xu et al., 2020b) . Over the past few decades, i ncreasing food avail ability (a key component of food security) while sustain ing biodiversity is key factors for global sustainability (Carole and Ignacio, 2016; Crist et al., 2017; Delzeit et al., 2017; Foley et al., 2011; Wiedmann and Lenzen, 2018) . Identifying the relat ionships between global food production and trade and biodiversity becomes essential to pursue multiple SDGs, which linked with SDGs 2 ( food security ) and 15 ( biodiversity ) (Nilsson et al., 2016; United Nations, 2015; Wiedmann and Lenzen, 2018) . With continuous population and income growth (Crist et al., 2017; Marques et al., 2019) as well as uneven distribution of food supply and demand, international food trade is essential for ensuring food availability (Porkka et al., 2013) , improving nutrient access (Wood et al., 2018) , and meeting rising food demands (Foley et al., 2011; MacDonald et al., 2015) . Many countries depend on food imports to meet their growing demands (DeFries et al., 2010; Godfray et al., 2010; Moran and Kanemoto, 2017) , but rapid increases of international food trade cause environmental consequences around the world (Crist et al., 2017; Dalin et al., 2017; Lenzen et al., 2012) . Producing food for export s causes land use and land cover change (Chaudhary and Kastner, 2016; DeFries et al., 2010; Delzeit et al., 2017) and exerts pressure on biodiversity in 63 exporting countries (Chaudhary and Kastner, 2016; Green et al., 2019; Lenzen et al., 2012; Moran and Kanemoto, 2017; Tilman et al., 2017) . Biodiversity is distributed unevenly across space (Brooks et al., 2014) . Therefore, the impact of international food trade on biodiversity is highly dependent on the origins of food production (Carole and Ig nacio, 2016; DeFries et al., 2010; Moran and Kanemoto, 2017) . It is widely concluded that importing food from tropical , developing countries to developed countries is worsening biodiversity (Chaudhary and Kastner, 2016; Lenzen et al., 2012; Moran and Kanemoto, 2017) . Although many studies have documented the negative impacts of international food trade on biodiversity in developing countries with rich biodiversity (Chaudhary and Kastner, 2016; DeFries et al., 2010; Lenzen et al., 2012; Moran and Kanemoto, 2017) , little is known about biodiversity implications of food exports from developed countries to developing countries, despite the fact that some developed countries have biodiversity hotspots ar eas with high concentration of biodiversity (Myers, 2003; Myers et al., 2000) , while some developing countries do not . Failing to recognize developed countries with biodiversity hotspots and developing countries without biodiversity hotspots may lead to biased results about the impacts of food trade on biodiversity worldwide. Thus, understanding food trade among countrie s with and without biodiversity hotspots is crucial for uncovering the implications of international food trade for global biodiversity. To address th e fundamental knowledge gap s , we divided 160 countries with relevant data into three categories: high - hots pot, low - hotspot, and non - hotspot countries. Specifically, we identified 64 high - hotspot countries (countries where biodiversity hotspots account for more than 50% of terrestrial lands), 53 low - hotspot countries (biodiversity hotspots < 50%), and 43 non - ho tspot countries (countries with no biodiversity hotspots) ( Figure S4.1, Table S4. 1 ). We also 64 classified the countries in each category as developing ( with low, low - middle, and upper - middle income) and developed ( with high income) according to the World Ban (The World Bank, 2017) . These classifications help analyze food trade among countries with different concentrations of biodiversity and levels of economic development. Our food dataset contains relevant annual information for 300 food items , including 203 crops from 2000 2015. 4 .2. Materials and Methods 4.2.1. Biodiversity hotspot and non - hotspot countries We divided all 160 countries with available data into 117 hotspot countries and 43 non - hotspot countries ( Fig ure S4. 1 , Table S4. 1 ) based on the relevant biodiversity hotspot information (Myers, 2003; Myers et al., 2000) . Hotspot countries are those that contain at least part of a recognized global biodiversity hotspot (Liu et al., 2003; Myers, 2003; Myers et al., 2000) . Biodiversity hotspots are areas with not only a high degree of species richness (hold >= primary, native vegetation) to human disturbance (Myers, 2003; Myers et al., 2000) . In contrast, non - hotspot countries do not include any part of a global biodiversity hotspot. B ecause hotspot countries vary substantially in terms of biodiversity hotspots ( Fig ure S4. 4 ), we classified hotspot countries into 64 high - hotspot countries and 53 low - hotspot countries, in which biodiversity hotspots account for more or less than 50% of terrestrial lands, respectively. High - hotspot countries , low - hotspot countries , and non - hotspot countries also have a number of other differences. For example, in developed and developing high - hotspot countries , 85.2% and 9 3 . 8 % of agricultural area s were located in biodiversity hotspots respectively , where as in developed and developing low - hotspot countries , only 7 . 4 % and 21.8% of total 65 agriculture land were located in a biodiversity hotspot , respectively ( Fig ure S4. 3 ). Since non - hotspot countries have no hotspots, agricultural area in non - hotspot countries was of course not located in a ny biodiversity hotspots. Per capita GDP of developed non - hotspot countries in 2015 ($ 44 , 990 in 2010 - constant USD ) was roughly 1.6 times as high as that of developed high - hotspot countries ($ 26 , 103 ) and developed low - hotspot countries ($ 28 , 306 ). Developing countries had similar per capita GDP across high - hotspot ($4,060 in 20 1 0 - constant USD), low - hotspot ($3,720), and non - hotspot countries ($3,991). Developed high - hotspot countries had the lowest population growth rates (4.5%) during 2000 2015, followed by developed non - hotspot countries (9.3%) and developed low - hotspot countries (11.0%). Population size in developing high - hotspot countries , developing low - hotspot countries , and developing non - hotspot countries increased 2 4 . 5 %, 2 3 . 0 %, and 23 . 9 % during 2000 2015, respectively . In addition, l and size in developed low - hotspot countries in 2015 (3,648,610 km 2 ) was 21.2 times and 7.2 times as large as that of developed high - hotspot countries (172,263 km 2 ) and developed non - hotspot countries (506,390 km 2 ), respect ively. Land size in developing low - hotspot countries in 2015 (1,124,795 km 2 ) was 5.1 times and 2.2 times larger than developing high - hotspot countries (219,132 km 2 ) and non - hotspot countries (516,284 km 2 ), respectively. 4.2.2. Data collection Datasets wer e obtained from the UN FAO, the UN Data, and the World Bank (The World Bank, 2017; UN FAO, 2018; United Nations Statistics Div ision, 2015) . Our database consisted of agricultural, environmental, and socioeconomic data. We selected the time period from 2000 2015 because of data availability. Agricultural datasets were obtained from the UN FAO (UN FAO, 2018) and included information about food production, food trade matrices, agricultural 66 areas, agricultural in tensification (fertilizer application, pesticide use, and water withdrawal), and average dietary energy supply adequacy (a percentage of average dietary energy requirement). The basic food trade unit in this research was the physical volume (metric tonne) of food produced, imported, and exported. This unit was chosen for two reasons. First, the number of countries in the volume dataset was much higher than in the monetary dataset. Second, using the volume of food trade is more appropriate for showing the e xtent to which food trade is linked with agricultural area because the monetary value varies with price fluctuations. Socioeconomic data such as population and per capita GDP came from the World Bank (The World Bank, 2017) . We also used published spatial data for identifying hotspot countries. Hotspot cou ntries were identified according to Conservation International (Myers, 2003; Myers et al., 2000) . 4.2.3. Aggregate analysis In the aggregate analysis, we divided the data collected from the individual countries into three groups of countries (high - hotspot countries, low - hotspot countries, and non - hotspot countries), which were further divided into developed and developing countries. We calculated agricultural intensification and agricultural area change from 2000 2015. By using agriculture and land use data sets, we were able to divide agricultural intensification values by agricultural area in each group. For example, we aggregated the amounts of fertilizer application and agricultural areas in each group. Then, we divided the amounts of fertilizer applicati on by total agricultural areas. We also calculated food trade flows among high - hotspot countries , low - hotspot countries, and non - hotspot countries in each individual year from 2000 2015, as well as annual averages over the same time period. In the FAO food trade matrix dataset (UN FAO, 67 2018) , we used food import matrix data. Food export matrix datasets were only used for filling in the data gaps in food import data. In addition, we used an origin - tracing algorithm to reduce data uncertainty regarding re - exports (Dalin et al., 2017; Kastner et al., 2011) . For example, some countries such as the Netherlands import food products fr om exporting countries and re - export them to other importing countries. The origin - tracing algorithm developed by Kastner et al. (2011) has a basic assumption that food consumption in each country proportionally originates from their domestic production and other countries. The origin of food impor ted can be examined using the bilateral food trade data. This algorithm assigns re - export volumes from intermediate countries to the original exporting country of production (Dalin et al., 2017; Kastner et al., 2011) . We also estimated the amount of land saved due to food imports based on yield and quantity of those imports (Liu, 2014) . The quantity of food imports (tonne) was divided by yield (tonne/km 2 ) in each country. The amount o f land saved by food imports was aggregated with developed and developing countries in high - hotspot countries , low - hotspot countries , and non - hotspot countries . 4.2.4. Panel data analysis To uncover factors affecting food production for domestic supply, food exports, and food imports, we performed panel data analyses in R (R Core Team, 2017) . Panel data analysis allows control for variables in different entities (e.g., countries) over time (Torres - Reyna, 2010) . We selected the random effects model because agricultural, socioeconomic, and environmental differences across countries have some influ ence on the quantity of food production and trade. The random effects model assumes that the each entity has its own error term that is random and 68 not correlated with independent variables in the model (Torres - Reyna, 2010) . An advantage of the random effects model over fixed effects is that time - invariant variables can be included as independent variables. We use d the same value of biodiversity hotspots in our panel data analyses because of the lack of time - series data for biodiversity hotspots. Since food production has two important purposes domestic supply and export we included the quantities of food productio n for domestic supply and export separately. This separation is essential for identifying responsible parties. For example, although the impact of food production on biodiversity is likely the same irrespective of whether the produced food is consumed loca lly or exported, the percentages of food exported out of food produced differed among countries. Exporting countries may shift some responsibilities for biodiversity loss caused by food production for exports to importing countries. To identify the changes in significant factors for food production and trade over time, we constructed random effects models for three different periods (2000 2007, 2008 2015, and 2000 2015). The three random effects models had 160 countries and included 8, 8, and 16 temporal po ints in each panel, respectively. We estimated the amounts of food production for domestic supply, food exports, or food imports as a function of agricultural factors (average dietary energy supply adequacy and total agricultural area), socioeconomic facto rs (per capita GDP and total population), and environmental factors (percentage of biodiversity hotspots) . We performed log transformation on all dependent and independent as the log - log transformation allows to interpret coefficients as an elasticity (Chung et al., 2018a) . We performed the Breusch - Pagan Lagrange multiplier (LM) test to choose between a random effects model and a simple ordinary least squares model (Torres - Reyna, 2010) . The LM test concluded that there are significant differences across countries (existenc e of panel effects) 69 and that our random effects models were more suitable. The random effects model allows to include time - invariant variables that preclude fixed effects . We also tested for heteroscedasticity using the Breusch - Pagan test. Since heterosced asticity was detected in all random effects models, we controlled for heteroscedasticity using a robust covariance matrix estimation (also known as a s andwich estimat or ) (Torres - Reyna, 2010) . We identified multicollinearity problem s using variance inflation factors (VIF). A VIF of 10 indicates a severe multicollinearity problem . All VIF results in our models were less than 4. 4.3. Resul ts Results indicate that developed countries were net food exporters while developing countries were net food importers during 2000 2015 (Fig ure 4. 1 and 4. 2 ). Among the exports from developed countries to developing countries, a lmost all (97.0%) went to hotspot countries. Specifically, developing hotspot countries received 97.8% of the exports from developed hotspot countries to developing countries (Fig ure 4. 2 ). Of exports from developed high - hotspot countries to developing countries, 34.1% and 52.2% went to developing high - hotspot and low - hotspot countries, respectively. 70 Figure 4 . 1 . The quantity of net food trade between developed and developing high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC). Blue indicates net food trade in 2000, red indicates net food trade in 2015, and gray indicates average net annual food trade from 2000 2015. The net amounts of food trade in each group are not linearly increased or decreased over time. The net amounts of food trade in 2000 and 2015 can be lower or higher than those in other mid - years as shown in Figure S4.2. Among the exports from developed low - hotspot countries to developing countries, 44.0% and 54.3% were destinated to high - hotspot and low - hotspot countries. D eveloped low - hotspot countries (e.g., USA and France) were the main contribut ors to international food trade as net ex porters while developing low - hotspot countries (e.g., China and Egypt) played an increasingly important role as net food importers (Fig ure 4. 1 and 4. 2 ). The USA and France accounted for 71 77.9% (158.9 Mt/year during 2000 2015) of food exported from developed low - hotspot countries (203.9 Mt/year), but only 3.3% and 12.2% of terrestrial areas were biodiversity hotspots , respectively, and the agricultural area in biodiversity hotspots accounted for 1.8% and 8.2% of total agricultural area respectively ( Fig ure S4 . 3 ). Agricultural areas in these countries also decreased from 2000 2015 by - 2.1% in the USA and by - 3.6% in France , suggesting less areal impact of agriculture on biodiversity . Food imports from developed low - hotspot countries with the highest quantities of net food exports kept much land from food production in high - hotspot countries . In developing high - hotspot countries , an estimated 202,257 km 2 of agricultural area per year (as big as the combined territories of two high - hotspot countries Cuba and Guatemala ) was saved due to imports from developed low - hotspot countries during 2000 2015 ( Table S4. 3 ) . In developed high - hotspot countries, food imports from developed low - hotspot countries accounted for a saving of 34,650 km 2 of agricu ltural area per year, larger than two thirds of Costa Rica territory. 72 Figure 4 . 2 . Average annual food flows (Mt/year) from 2000 2015. Food flows between developed and developing high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC). Non - hotspot countries are marked by red, high - hotspot countries by dark green, and low - hotspot countries by light green. The arc length of an outer circle indicates the sum of food exported and imported in each group. The arc length of a middle circle refers to the quantity of food exports. The inner arc length shows the quantity of food imports. Raw data from UN FAO (2018) . D eveloping high - hotspot countries were net food importers (1.7 Mt/year of average net annual food imported) from 2000 2015 (Fig ure 4 .1 and 4. 2 ). Such imports are particularly 73 important to reduce agricultural impacts on biodiversity in high - hotspot countries because 78.1% of high - hotspot countries had over 90% of their terrestrial area as biodiversity hotspots ( Fig ure S4. 4 ), and 92.1% of high - hotspot countries hotspots ( Fig ure S4. 3 ). Over the same period, developing high - hotspot countries accounted for 50.3% (96.1 Mt/year during 2000 2015) of total food imports among all high - hotspot countries (190.9 Mt/year) ( Table S4. 2 ). F ood imports from low - hotspot and non - hotspot countries to developing high - hotspot countries accounted for production in roughly 7.6% of annual agricultural ar ea (340,428 km 2 , larger than the territory of Malaysia) in developing high - hotspot countries. D eveloping low - hotspot countries rapidly increased their net food imports during the study period by 514.4% ( Fig ure S4. 2 , Table S4. 2 ). Of exports from developed n on - hotspot countries to developing countries, 93.8% went to developing hotspot counties ( 31.0% and 62.8% to high - hotspot and low - hotspot countries, respectively) . Most exports (56.3%) from developing countries went to other developing countries, rather tha n developed countries (Fig ure 4. 2). Of which, the destinations of 52.3%, 61.7%, and 40.5% of exports from developing high - hotspot, low - hotspot, and non - hotspot countries were developing countries. In other words, developed countries received less than half of the exports from developing hotspot countries and more than half of the exports from developing non - hotspot countries. Among the exports from developing hotspot countries to developed countries, most of them (60.9%) were destinated to developed hotspot countries (Fig ure 4. 2). Of exports from developing high - hotspot countries to developed countries, developed high - hotspot and low - hotspot countries received 18.2% and 48.2%, respectively . Among exports from developing low - hotspot countries to developed cou ntries, almost equal amounts ( 28.5% , 28.2%) went to developed high - hotspot and low - hotspot countries. 74 Figure 4 . 3 . Changes in agricultural intensification and agricultural area in high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC), with each group subdivided into developed and developing countries. (A) Fertilizer use (tonne/km 2 ); (B) pesticide use (tonne/km 2 ); (C) agricultural water withdrawal (m 3 /km 2 ); (D ) agricultural area change (km 2 ). Raw data from UN FAO (2018) . Imports to developed high - hotspot countries (67.0 Mt/year of average net annual food imported during 2000 2015 ) can also help further reduce negative impacts on biodiversity because agricultural intensification in developed high - hotspot countries was higher than in other types of countries (Fig ure 4. 3). For example, fertilizer use per unit in developed high - hotspot countries (6.9 tonne/km 2 ) was 60.5% higher than developing high - hotspot countries (4.3 75 tonne/km 2 ) in 2015 (Fig ure 4. 3A). Pesticide use per unit in developed high - hotspot countries (0.258 tonne /km 2 ) was three times higher than in developing high - hotspot countries (0.066 tonne/km 2 ) in 2015 (Fig ure 4. 3B). In 2014, freshwater withdrawal (212,333.8 m 3 /km 2 ) for agricultural production in developed high - hotspot countries was about twice as high as in developing high - hotspot countries (77,599.7 m 3 /km 2 ) (Fig ure 4. 3C). Therefore, food imports may decrease biodiversity threats in developed high - hotspot countries as, without food imports, more agricultural land in developed high - hotspot countries would be u sed or intensified for domestic food production. D eveloping non - hotspot countries (e.g., Ukraine and Romania) played an increasingly important role as net food exporters ( with an average net food export of 18.8 Mt/ year during 2000 - 2015 ) (Fig ure 4. 1, Fig ure S4. 5 and S4. 6). They exported 1.6% (6.7 Mt) of their food production in 2000, but 9.7% (61.6 Mt) in 2015 ( Table S4. 4 ). Such exports freed much area for production in hotspot countries. For instance , developed high - hotspot countries saved agricultura l areas of 24,339 km 2 (larger than the territory of Belize) per year from 2000 2015 ( Table S4. 3 ). Developing high - hotspot countries saved agricultural areas of 3,270 km 2 (over 60% larger than the territory of Mauritius) in 2000 and increasingly 72,309 km 2 ( approximately the territory of Panama) in 2015 ( Table S4. 3 ). Developing non - hotspot countries had the lowest agricultural intensification and least agricultural area among all types of countries, which suggests food imports from developing non - hotspot cou ntries further reduce biodiversity threats from food production in hotspot countries (Fig ure 4. 3). Fertilizer use per unit land in developing non - hotspot countries (1.0 tonne/km 2 ) was 430% and 330% lower than developing low - hotspot countries (5.3 tonne/km 2 ) and developing high - hotspot countries (4.3 tonne/km 2 ) in 2015 , respectively (Fig ure 4. 3A). Pesticide use per unit in developing non - hotspot countries (0.031 76 tonne/km 2 ) was 300% and 113% lower than developing low - hotspot countries (0.124 tonne/km 2 ) and developing high - hotspot countries (0.066 tonne/km 2 ) in 2015 (Fig ure 4 .3B ). In 2014, freshwater withdrawal (4,680.5 m 3 /km 2 ) for agricultural production in developing non - hotspot countries was 23 times and 17 times lower than developin g low - hotspot countries (107,351.5 m 3 /km 2 ) and developing high - hotspot countries (77,597.7 m 3 /km 2 ) respectively (Fig ure 4 .3C). While agricultural areas in developing non - hotspot countries decreased 0.7% from 2000 2015, agricultural areas in developing low - hotspot countries and developing high - hotspot countries increased 5.6% and 6.0% respectively (Fig ure 4 .3D). Panel d ata analyses for three different time periods (2000 2007, 2008 2015, and 2000 2015) identified factors that affect global food production and trade (Table 4. 1). Countries with larger agricultural areas tended to produce more for both domestic consumption and exports, whereas countries with smaller agricultural areas tended to import more food. This result may indicate that food importers with s maller agricultural areas displaced agricultural land use to food exporters. For instance, developed high - hotspot countries ha d the smallest agricultural areas among the six types of countries (Fig ure 4 .3D ) and would have saved 97,177 km 2 of agricultural a rea (larger than the territory of Portugal ) per year during 2000 - 2015 as net food importers , accounting for roughly 11.1% of their annual agricultural area ( Table S4. 3 ). In addition, the average dietary energy supply adequacy had a positive association with the quantity of food imported. C ountries that had a higher average dietary energy supply imported more food from abroad to meet increases in per capita caloric and p rotein demands. Per capita GDP and population size drove all significant correlated results with food production and trade ( Table 4. 1). 77 Table 4 . 1 . Coefficients of panel data analyses in thr ee different periods: 2000 2007, 2008 2015 and 2000 2015. 2000 2007 2008 2015 2000 2015 Variable Food production for domestic supply Food export Food import Food production for domestic supply Food export Food import Food production for domestic supply Food export Food import Biodiversity hotspot (%) 0.055 (0.037) 0.222* (0.081) 0.085* (0.036) 0.043 (0.033) 0.121 (0.094) 0.096 (0.063) 0.076* (0.038) 0.275* (0.098) 0.123* (0.060) Dietary energy supply adequacy (%) 1.522** (0.297) 2.415** (0.730) 1.197* (0.387) 0.680* (0.281) 1.164 (0.832) 1.693* (0.652) 1.393** (0.249) 1.146 (0.785) 1.861** (0.541) Agriculture area (km 2 ) 0.324* (0.121) 0.486* (0.162) - 0.159* (0.061) 0.313** (0.062) 0.264 (0.136) - 0.259** (0.067) 0.455** (0.114) 0.683** (0.181) - 0.231** (0.064) Population (1,000 persons) 0.719** (0.139) 0.518* (0.195) 1.086** (0.070) 0.718** (0.069) 0.753** (0.170) 1.193** (0.081) 0.577** (0.121) 0.316 (0.223) 1.165** (0.078) GDP per capita (constant $) 0.137** (0.032) 0.673** (0.123) 0.621** (0.064) 0.155** (0.040) 0.595** (0.130) 0.597** (0.084) 0.155** (0.043) 0.849** (0.156) 0.621** (0.087) Intercept - 1.918 (1.393) - 14.943** (3.170) - 5.725** (1.623) 2.056 (1.212) - 7.902* (3.283) - 7.827* (2.836) - 1.663 (1.247) - 10.893* (3.468) - 8.969** (2.163) R - Squared 0.658 0.223 0.487 0.599 0.125 0.312 0.556 0.150 0.344 F - statistics 476.2 71.16 235.78 375.5 35.46 113.85 627.23 89.17 263.38 Values in parentheses are standard errors All variables are log transformation variables ** P<0.001, * P<0.05 78 4.4. Discussion Our research integrating biodiversity hotspots and economic development status provides a new perspective i nternational food imports may benefit developing and developed countries with biodiversity hotspots. By increasing the proportion of food production for exports, developing non - hotspot countries with lower intensification played an increasingly important role as net exporters in international food trade. As threats from agricultural activities vary among species and across space (Brooks et al., 2014; Moran and Kanemoto, 2017) , species - specific analys e s are needed within national boundaries. Worldwide, identifying species - specific relationships with international food trade items would be possible through the analysis of high - resolution data (Green et al., 2019; Moran and Kanemoto, 2017; Wiedmann an d Lenzen, 2018) . Future research efforts are needed to accurately determine causal re lationships among global food production and trade and biodiversity based on high - resolution sub - national and local data over time (Carole and Ignacio, 2016; Green et al., 2019; Moran and Kanemoto, 2017; Wiedmann and Lenzen, 2018) , and to identify the impacts of food producti on for export s in specific locations. With increasing attention to food security and biodiversity conservation (e.g., global assessment of biodiversity and ecosystem services (Díaz et al., 2019) , upcoming meeting of the Conference of the Parties to the Convention on Biological Diversity (UNEP) ), it is time to rethink international food trade by creating more innovative approaches to minimize the negative impacts of global food production and trade on biodiversity in both developed and developing hotspot countries, especially high - hotspot countries . For instance, new international initiatives and agreements are necessary to reduce threats to biodiversity from food production and trade (Brooks et al., 2014; Ehrlich and Harte, 2015; Redford et al., 2015) . Food price s should 79 incorporate the biodiversity cost of producing food (Pe'er et al., 2014) . Earnings from such a price hike c ould be used to mitigate negative impacts on biodiversity. Biodiversity can also benef it from the further development of t echniques with low input but high yield, such as wildlife - friendly farming (Pywell et al., 2015) . Both food importing and exporting countries work ing together to implement new policies and technologies can lower negative impacts on biodiversity while increasing food security. The approaches and findings in this paper provide a foundation for further work incorporating data with higher resolutions to quantify biodiversity impacts, operationalize post - 2020 global biodiversity framework, and achieve United Nations Sustainable Development Goals (e.g., Goal 2 food security and Goal 15 biodiversity conservation ) across multiple scales worldwide . 80 CHAPTER 5 INTEGRATING BUILT INFRASTRUCTURE AND WATERSHED CONSERVATION TO SUSTAIN FRESHWATER ECOSYSTEM SERVICES FOR GLOBAL CITIES In collaboration with Kenneth Frank, Yadu Pokhrel, Thomas Dietz, and Jianguo Liu 81 Abstract Worldwide rapid urbanization demands more freshwater . T his need is conventionally met through the construction of infrastructure. Watershed conservation activities have also increased to provide freshwater ecosystem services , but little research has examined the intricate relationships between built infrastructure and watershed conserva tion activities for provisioning freshwater ecosystem services to global cities. By using egocentric network analysis, this study examines how to integrate built infrastructure approaches with ongoing watershed conservation activities for sustaining four f reshwater ecosystem services (i.e., freshwater provision, sediment regulation , flood mitigation, and hydropower production) to cities. Our results indicate that wetlands in protected areas contribute to sustaining freshwater provision to cities. Forest cov er in protected areas can improve the capacity of large dams for sediment reduction and hydropower production, but cities mainly depend on dams for flood mitigation. Our findings suggest strategic approaches for integrat ing built infrastructure and watersh ed conservation activities to enhance urban water sustainability. 82 5.1. Introduction O ver the past few decades , rapid urbanization causes various water - related problems such as water shortage, water quality, floods, and energy for cities worldwide (McDonald and Shemie, 2014; McDonald et al., 2014; McDonald et al., 2016) . W ith increased urban population and income levels, built (or gray) infrastructure has been rapidly constructed to meet the increased freshwater demand s for cities (Grill et al., 2019; McDonald and Shemie, 2014; McDonald et al., 2014; Vorosmarty et al., 2010) . Built infrastructure defines as the human - engineered construction for water resources such as dams and treatment facilities (Gartner et al ., 2013) . M odifications of natural river systems through built infrastructure increase water security for residential users (Tessler et al., 2015) but cause the loss of freshwater biodiversity, water quality, and habitat degradation (Grill et al., 2019; Michalak, 2016; Palmer, 2010; Vorosmarty et al., 2010) . Since the early 20th century, almost 90% of watersheds providing water to cities have experienced a degradation of their water quality, including increases in nitrogen and phosphorous due to anthropogenic activities (e.g., changes in agricultural land use) (McDonald et al., 2016) . This degraded water quality directl y affects water for drinking and recreation in cities (Michalak, 2016) . Furthermore, potential supplies of freshwater ecosystem services (ES) to c ities have decreased over time and across regions (Dodds et al., 2013) . On the other hand, watershed conservation activities (e. g. , protected areas (PAs) and investments in watershed services (IWS)) have continuously provided provisioning freshwater ES as a part of the Convention on Biological Diversity Aichi Targets and the United Nations Sustainable Development Goals (SDGs) (Harrison et al., 2016; Romulo et al., 2018; Tellman et al., 2018; Visconti et al., 2019) . W atershed conservation activities can potentially help reduce the negative effects of built infrastructure that degrade freshwater biodiversity, damage fisheries, and 83 displace local people (Liu et al., 2016a; Palmer et al., 2015; Ziv et al., 2012) . W atershed conservation areas can provide various freshwat er ES to humans, such as fisheries (Brennan et al., 2019; McIntyre et al., 2016) , the improvement of water quantity and quality (Harrison et al., 2016; Mapulanga and Naito, 2019; Veldkamp et al., 2017; Vörösmarty et al., 2018) , flood regulation (Russi et al., 2013; Tellman et al., 2018) , recreational opportunities (Chung et al., 2018b) , and carbon sequestration (Viña et al., 2016) . Specifically, water storage from forests and wetlands in PAs may increase the capability for freshwater provision, flood protection, and hydropower production in addition to dam storage (Harrison et al., 2016; Moran et al., 2018; Te llman et al., 2018) . The capacity of watershed conser vation areas under IWS can also help meet the increased freshwater demands of cities by maintaining high freshwater ES and biodiversity (Adamowicz et al., 2019; Romulo et al., 2018; Zheng et al., 2 013) . With the rapid increases of PAs and IWS worldwide, the networks of watershed conservation areas may complement built infrastructure by providing various freshwater ecosystem services . Yet, little is known about the relationsh ips between built infrastructure and watershed conservation activities for sustaining freshwater ES to cities . The relationship warrants attention because maintaining the benefits of built infrastructure while conserving healthy freshwater ecosystems is a complex challenge (Grill et al., 2015; Poff and Schmidt, 2016) . These relationships between built infrastructure and watershed conservation activities become more complicated as cities are increasingly reliant on not only surrounding watersheds but also distant watersheds through built infrastructure construction (e.g., dams and canals) (Liu and Yang, 2013; Liu et al., 2016a; McDonald et al., 2016) . Th us , a new strategy is urgently needed to integrate a built infrastructure approach with ongoing watershed conservation activities (i.e., PAs and IWS) (Harrison et al., 2016; Romulo et al., 2018; Zheng et al., 2013) . 84 To achieve sustainable freshwater ES supplies to cities, management strategies should consider a balance between human demands and ecosystem conservation (Lehner et al., 2011; Vorosmarty et al., 2010) . Built infrastructure app roach and watershed conservation activities can be combined because the supply of freshwater ES depends on both protected watersheds and traditional built infrastructure (Green et al., 2015; Vörösmarty et al., 2 018) . However, very limited studies to date have examined the approaches for integrating built infrastructure and watershed conservation activities in source watersheds for sustaining freshwater ES to cities. T herefore, t he goal of this study is to fill this knowledge gap . W e seek to answer two questions: (1) Which built infrastructure and watershed conservation activities have a significant relationship with freshwater ES supplies for global cities? and (2) Which socioeconomic and environmental factors of source watersheds and cities contribute to the changes of freshwater ES supplies to cities? This study focuses on four freshwater ES freshwater provision, sediment regulation , flood mitigation, and hydropower production in source watersheds as t hese four ES flows have exponentially increased to meet IWS influence the provision of freshwater ES for global cities, while controlling for the net of geograp hical factors, watershed characteristics, and city characteristics. This study provides a new perspective on how to combine built infrastructure approaches and watershed conservation activities to sustain freshwater ES supplies for cities while protecting freshwater ecosystems and biodiversity. 85 5.2. Materials and Methods 5.2.1. City and watershed selection We first identified global cities that mainly depend on surface water sources from the City Water Map database (McDonald et al., 2014) . Each selected city has a n average population of over 300,000 people from 2000 to 2010 according to the World Urbanization Prospects data (UNDP, 2015) . For urban extents, we used the Global Administrative Database (GADM) that defines urban admini strative areas (Global Administrativ e Areas, 2018) . For cities not defined in the GADM, we used the global urban extent map from Schneider et al. (2009) based on MODIS satellite d ata. In the USA, the Cartographic Boundary File for urban areas was used to define urban extents (United States Census Bureau, 2017) . For e ach city, we identified three types of source watersheds: (1) freshwater source watersheds (freshwater provision and sediment regulation), (2) flood watersheds, and (3) hydropower watersheds. As freshwater ES are produced in source watersheds and provide b enefits to cities, source watersheds are directly and indirectly connected to cities through the flows of freshwater ES. Source watersheds were designated following the United States Geological Survey (USGS) HydroSheds database (Lehner et al., 2008) . Freshwater source watersheds provide surface water sources to cities. Cities depend on not only the surrounding watersheds but also distant watersheds for freshwater resources (Liu and Yang, 2013; McDonald et al., 2016) . Surface water in freshwater source watersheds is transferred from water intake points to the city. Surface water intake points were obtained from the City Water Map database (McDonald et al., 2014) . Freshwater source watersheds are also watersheds wi th sediment flows affecting freshwater quality in cities. Flood watersheds have a higher elevation than cities, can overlap with urban extent areas, and increase or reduce the flood 86 risks of cities by directly draining surface water to the urban extent are a. Hydropower watersheds generate and provide electricity from hydropower dams to cities and are connected with cities through high voltage power lines within 100 km from the urban extent. High voltage power lines linkages to the cities were obtained from OpenStreetMap ( https://www.openstreetmap.org ). With the different locations and numbers of watersheds for each freshwater ES, the number of cities also varies across freshwater ES. For exploring freshwater prov ision and sediment regulation, we selected 333 cities and 1,198 freshwater source watersheds. We also analyzed a total of 665 flood watersheds across 200 cities for flood mitigation. Finally, we selected 197 cities and 469 hydropower watersheds for hydropo wer pr oduction (Figure S5. 1). 5.2.2. Freshwater ecosystem services Built infrastructure and watershed conservation activities have a variety of impacts on natural ecosystems and ES. For example, hydropower capacity may increase with the number of dams, bu t such expansion causes habitat loss and river fragmentation, and those changes in turn impact freshwater biodiversity and water quality (Grill et al., 2019; Palmer, 2010; Roy et al., 2018) . Watershed c onservation can benefit forest and wetland cover, positively contributing to fisheries and water quality (Bilotta et al., 2012; Brennan et al., 2019; McIntyre et al., 2016; Vörösmarty et al., 2018) . Both built infrastructure and w atershed conservation areas contribute to increasing water regulation (e.g., freshwater provisioning and flood mitigation) (Chen and Olden, 2017; Harrison et al., 2016; Mapulanga and Naito, 2019; Roy et al., 2018; Russi et al., 2013) . - related demands: freshwater provision (Veldkamp et al., 2017) , sediment regulation (Cohen et al., 2014) , 87 flood mitigation (Dottori et al., 2016) , and hydropower production (Byers et al., 2018; Zarfl et al., 2015) (Table S5. 1) . These four ES have flows from source watersheds to cities and are divided into provisioning and regulating ES. Provisioning ES include freshwater provision and hydropower production . Regulating ES are com prised of sediment regulation and flood mitigation . We used global modeling data for freshwater ES, except for hydropower production. These datasets utilized local and regional observational data for to produce their output data . The resulting datasets hav e been widely used in peer - reviewed papers in high - impact journals (Best, 2019; Grill et al., 2019; Smith et al., 2019; Veldkamp et al., 2017; Vorosmarty et a l., 2010) . 5.2.2.1. Freshwater provision In this study, freshwater provisioning that supplies cities refers to the annual average volumes of surface water flowing through a river channel. Surface water is extracted at water intake points and transferred to cities (McDonald et al., 2014) . Freshwater provision data for 200 1 2010 were obtained from phase 2 of the Inter - Sectoral Impact Model Intercomparison Project (ISIMIP2a, http://www.isimip.org ), which provides the daily outputs from five global hydrological models (GHMs): H08 (Hanasaki et al., 2008a, 2008b) , LPJmL (Bonde au et al., 2007; Schaphoff et al., 2013) , MATSIRO (Pokhrel et al., 2012; Pokhrel et al., 2015) , PC R - GLOBWB (van Beek et al., 2011; Wada et al., 2014) , and Water Gap (Müller Schmied et al., 2016) . With these five GHMs, driven by three historical climate forcing datasets (PGFv2 (Sheffield et al., 2006) , GSWP3 (Dirmeyer et al., 2006) , and WFDEI (Weedon et al., 2014) ), we used 15 model combinations to quantify the volumes of surface water supplies from source watersheds to cities. We extracted the annual - in water intake points and calculated median values for the 15 model combinations in each source watershed. 88 The GH Ms account for the most natural surface and sub - surface hydrologic processes relevant for the simulation of water resource availability (e.g., local run - off and upstream discharge) at 0.5° (~50 km) grid cells globally. Human water management activities are also represented by accounting for various sectoral water demands including those for the agriculture (irrigation and livestock), industry (manufacturing and thermal energy), and public (domestic use) sectors under time - varying socioeconomic conditions (e .g., population, GDP, and land - use) (Veldkamp et al., 2017) . 5.2.2.2. Sediment regulation We obtained results from a global suspended sediment flux model b ased on the WBMsed (Cohen et al., 2014; Wisser et al., 2010) . Cohen et al. (2014) provided the amounts of suspended sedim ent flux in a 6 arc - minute (~12 km or 0.1°) grid cell. We extracted the amounts of annual - averaged suspended sediments in water intake points for cities from 2000 to 2010. We concentrated on surface water sources, not groundwater, because built infrastruct ure and watershed conservation activities mainly contribute to changes in surface water quality (e.g., sediment flux and phosphorous pollution) (McDonald et al., 2016; Robin Abell et al., 2017) . Although suspended sediments are crucial to sustain freshwater ecosystems in downstream areas (e.g., creating natural habitats) (Vercruysse et al., 2017) , suspended sediments deteriorate water quality and therefore cause additional costs for urban water treatment (Bilotta and Brazier, 2008; Bilotta et al., 2012; McDonald et al., 2016) . 5.2.2.3. Flood mitigation We used global flood hazard maps with return periods of 100 years to identify the probability of river flood magnitudes over an urban area (Dottori et al., 2016) . These flood 89 haza rd maps show flood extents and depths in a 30 arc - second (~1 km or 0.0083°) grid cell based on hydrological information from the Global Flood Awareness System (GloFAS) (Alfieri et al., 2018; Dottori et al., 2016) . Based on this mo del, we calculated the proportion of flood extent areas to total urban extent areas in each flood watershed. 5.2.2.4. Hydropower production The Global Power Plant Database provides the geolocation of operational hydropower dams above 1 megawatt (MW) capaci ty (Byers et al., 2018) . This database covers approximately 89% of global installed capacity in the hydropower sector (Byers et al., 2018) . This dataset pr ovides point hydropower locations, and we aggregated the installed capacity of the hydropower dams in each hydropower watershed. 5.2.3. Source watershed and city characteristics To examine which characteristics contribute to four freshwater ES flows from source watersheds to cities, we collected data regarding dams, watershed conservation activities, environmental factors, and socioeconomic factors in source watersheds and cities. These data were obtained from international organizations, online databases, and peer - reviewed papers (Table S5. 1). Our indicators are dam density as a measure of built infrastructure, and watershed conservation activities included PAs in source watersheds (IUCN and UNEP - WCMC, 2017) and IWS programs in cities (Romulo et al., 2018) . For each of the three different types of source watersheds (freshwater source, flood, and hydropower), we obtained information on forest and wetland cover in PAs, dam density, irrigation areas, and geographic characteristics of the watersheds. The spa tial boundaries and characteristics of PAs were obtained from the World Database on Protected Areas (IUCN and 90 UNEP - WCMC, 2017) . We selected terrestrial PAs that are legally designated and actively managed at the national or sub - national level. We also included all PAs that were assigned, not reported, or not assigned to the IUCN management category because many countries do not consistently apply or use the IUCN management category (Bingham et al., 2019) . Since many PAs spatially overlap each other (Deguignet et al., 2017; Jones et al., 2018) , we dissolved PA boundaries to avoid double counting problems. Then, we intersected a s ingle PA polygon with (ESR I, 2015) . Forest cover data were obtained from global land cover data that provide the percentage of forest cover with 1 km resolution (Tuanmu and Jetz, 2014) . We tland cover data were collected from the Global Lakes and Wetlands database, which provides global wetland extents at 30 arc - second (~1 km) resolution (Lehner and Döll, 2004) . Then, in each watershed, we calculated the proportion of forest and wetlan d cover in PAs to total watershed areas, respectively. The attributes of dams were obtained from the Global Reservoir and Dam (GRanD) database (Lehner et al., 2011) . This database includes the name, spatial lo cation, construction year, and various characteristics of dams that are higher than 15 m and have a reservoir larger than 0.1 km 3 . To estimate river length, river network data were obtained from the HydroSHEDS at 30 arc - second (~1 km) resolution (Lehner et al., 2008) . With dam numbers and river lengths, we calculated dam density (dams per 100 km of river length) in each watershed. We also included irrigated croplands from the Global Food Sec urity - Support Analysis Data with 1 km resolution (Thenkabail et al., 2016) . Using the size of irrigated croplands, we calculated the proportion of irrigation areas to total watershed areas. Geological characteristics of watersheds included the size of each watershed, geographic distances between cities and watersheds, elevation, and slope. We calcu lated the size of 91 watersheds and geographic distances between the centroids of cities and source watersheds using ArcGIS (ESRI, 2015) . Elevation and slope data in river networks were gathered from Domisch et al. (2015) at 1 km resolution. size of the urban economy, and climatic factor s. IWS program data were collected from Romulo et al. (2018) and Bennett and Ruef (2016) . IWS, a kind of payments for ES, are broader conservation strategies to provide and enhance freshwater ES with incentive - based mechanisms between the beneficiary and provider of watershed services (Huber - Stearns et al., 2015; Romulo et al., 2018) . We included IWS programs that provided freshwater resources to a city in the City Water Map datab ase and had a specific goal for drinking water protection (Bennett and Ruef, 2016; Romulo et al., 2018) . Average annual populations size from 2000 to 2010 were obtained from the World Urbanization Prospects report (UN DP, 2015) . Spatially explicit GDP data in 2010 were obtained from the global dataset of gridded GDP and population scenarios at 0.5° (~50 km) resolution (Murakami and Yamagata, 2019) . Climatic factors (annual mean temperature and annual precipitation) came from the WorldClim database at 1 km res olution (Hijmans et al., 2005) . Since our dataset included spatially explicit data, we extracted variables at the watershed or city level by using zonal statistics in R (R Core Team, 2017) . For example, we extracted the numeric values of elevation and slope in watersheds and GDP and climatic factors in cities. 5. 2.4. Egocentric network analysis We used multi - level models applied to an egocentric network analysis to estimate the contribution of each independent variable to freshwater ES supplies from source watersheds to 92 cities (Wellman and Frank, 2001) . Because cities usually have more than o ne source watershed, they form an egocentric network: ego is the city and alters are the source watersheds. Each tie and source watershed (alter) at the end of that tie is nested in each urban water network and the city (ego) to which that network belongs (Figure 5. 1). Cities (egos) form an egocentric network by environmentally and socioeconomically interacting with source watersheds (alters) that supply freshwater ES to cities. Multi - level network models help examine the effects of the characteristics of e ach ego, its alters, and their ties to freshwater ES flows. We investigated how the characteristics of source watersheds and cities contribute to freshwater ES flows. Figure 5 . 1 . Egocentric networks between cities (egos) and freshwater source watersheds (alters). Red squares indicate cities, and blue circles indicate freshwater source watersheds. Each city has more than one source watershed, and thus they form an egocentric network. The level 1 model includes the effects of the characteristics of alter (i) and tie (i, j), and the level 2 model includes the effects of the characteristics of ego (j). At level 1, we modeled changes in freshwater ES flows as a function of forest cover in PAs, wetland cove r in PAs, dam 93 density, irrigation area, watershed areas, distance from city to watershed, elevation, and slope. coefficients linking changes in flows to characteri stics, , are used as an outcome in the level two model. At level 2, we modeled the intercept in the level 1 model as a function of the IWS multi - level model for the flows of freshwater ES between alter (i) and ego (j) is as follows: Level 1 (Alter and tie): Level 2 (Ego): For example, represents the effect of the presence of the IWS program. The errors at level 1, , are assumed to follow a normal distribution (0, ), and the level 2 errors, , are assumed to follow a normal distribution (0, ). To pursue linearity and normality, we carried out natural log transformations on all variables. Then, we estimated multi - level models in R u sing the restricted maximum likelihood method (Bates et al., 2015; R Core Team, 2017) . We also measu red variance inflation factors (VIF) to check the multicollinearity of our multi - level 94 models. Our VIF results showed that independent variables of multi - level models had no serious multicollinearity problems ( Table S5. 2 ). 5.3. Results Forest cover in PAs of source watersheds had a negative relationship with the amount of sediment flux but was positively associated with hydropower production (Table 5. 1). Watersheds with larger wetland cover in PAs had larger freshwater provisioning. The extent of forests and wetlands in PAs increased sustaining freshwater ES to cities except for flood mitigation. Watersheds with high dam density had low sediment flows and flood risks while having high hydropower production. However, dam density did not ha ve a statistically significant effect on freshwater provisioning to cities. In fact, the estimate of dam density was considerably less than its standard error. Our results indicate that forest covers in PAs complemented dams for sediment reduction and hydr opower production. The proportion of irrigation areas in source watersheds was negatively associated with freshwater provisioning to cities. Geological characteristics in watersheds also contributed to the flows of freshwater ES to cities. Watersheds with larger watersheds and greater distances between watersheds and cities had a positive relationship with more freshwater provisioning, sediment flows, and hydropower production. Watersheds at lower elevations provided fewer freshwater and sediments. Steeper watersheds had larger sediment flows and hydropower production but lower flood risks. City p opulation size was positively associated with freshwater provisioning and sediment flows, while urban GDP was negatively associated with these two ES. In other word s, cities with high population size and low GDP had not only high freshwater provision but also high sediment flows. Cities with higher average temperatures had larger sediment flows and hydropower 95 on with freshwater provision. The presence of IWS programs in cities was not statistically significant with all four freshwater ES, but with a p - value less than 0.1, freshwater provisioning was at the margins of statistical significance (P = 0.058). In add ition, we identified the spatial locations of new large dams and PAs from 2000 to 2016 at the watershed level (Figure 5. 2). We concentrated on freshwater source watersheds and hydropower watersheds, as both dams and PAs positively contributed to sediment r eduction and hydropower production. From 2000 to 2016, new PAs were designated in 34.1% of freshwater source watersheds and 56.1% of hydropower watersheds without new large dam constructions. These watersheds were mainly located in North America and Europe (Figure 5. 2). In the same period, areas in 4.8% of freshwater source watersheds and 2.8% of hydropower watersheds not only received new PA designations but also construct ed new large dams worldwide. However, 2.9% of freshwater source watersheds and 3.8% o f hydropower watersheds constructed large dams without new PA designations, of which approximately two - thirds were located in China and India (Figure 5. 2). China and India did not designate new PAs in 97.3% of freshwater source watersheds and all hydropo we r watersheds over the period 2000 to 2016. In these two countries, 11.9% and 15.6% of their freshwater source watersheds and hydropower watersheds, respectively, constructed large dams without any new PA designations. 96 Table 5 . 1 . Multi - level coefficients predicting four freshwater ecosystem services. Variable Water Supply Sediment Flow Flood Risk Hydro - power Watershed Forest cover in PAs (%) - 0.005 (0.026) - 0.094** (0.034) 0.085 (0.077) 0.241** (0.087) (Level 1) Wetland cover in PAs (%) 0.094* (0.044) 0.112. (0.057) 0.006 (0.104) 0.060 (0.168) Dam density (#/100 km of river length) 0.005 (0.040) - 0.154** (0.052) - 0.606* (0.271) 1.549** (0.340) Irrigation area (%) - 0.073* (0.029) 0.058 (0.037) 0.014 (0.045) - 0.053 (0.082) Watershed area (km 2 ) 0.148** (0.023) 0.309** (0.030) - 0.048 (0.059) 0.268* (0.114) Urban - watershed distance (km) 0.253** (0.053) 0.151* (0.069) - 0.107 (0.115) 0.530* (0.208) Elevation (meter) - 0.114** (0.042) - 0.245** (0.054) - 0.166 (0.135) - 0.275 (0.191) Slope (degree) - 0.032 (0.040) 0.138** (0.051) - 0.214** (0.061) 0.340** (0.121) Urban IWS program (0, 1) - 0.634. (0.333) 0.190 (0.355) - 0.115 (0.347) 0.443 (0.413) (Level 2) Urban population (1,000 persons) 0.218* (0.089) 0.219* (0.096) 0.149. (0.077) 0.008 (0.123) Urban GDP - PPP (2005 const. billion USD) - 0.166* (0.084) - 0.300** (0.090) - 0.055 (0.079) - 0.122 (0.130) Temperature (°C) 0.126 (0.220) 1.326** (0.239) 0.410* (0.183) 0.852* (0.380) Precipitation (mm) 0.980** (0.121) 0.074 (0.131) 0.081 (0.118) - 0.024 (0.177) Intercept - 5.009** (1.072) - 5.487** (1.175) 1.903 (1.509) - 2.240 (2.379) Random effect City (Intercept) 1.651** 1.736** 0.620** 1.179** Residual 0.683** 1.211** 1.040** 1.890** N 1,323 1,323 767 497 Standard errors in parentheses: ** P < 0.01, * P < 0.05, . P < 0.1 97 Figure 5 . 2 . Spatial changes in numbers of dams and sizes of PAs from 2000 to 2016 in (A) freshwater source watersheds and (B) hydropower watersheds. Orange indicates an increase in the numbers of dams and size of PAs, red indicates an increase in only dams, green indicates an increase in only PAs, and blue indicate watershed from 2000 to 2016. 5.4. Discussion 5.4.1. The role of forests and wetlands in protected areas This study examines a possible way to integrate watershed conservation activities (PA and IWS) wi th built infrastructure for sustainably providing four freshwater ES for cities. Forest cover in PAs and dams both provide sediment reduction. In addition, forest PAs increase 98 hydropower production over what would be expected from dams alone. Protected for ests in source watersheds help decrease sediment flows because forest covers reduce soil erosion with tree root systems, high infiltration rates, and low overland flows (Blumenfeld et al., 2009; Gartner et al., 2013; Vercruysse et al., 2017) . High evapotranspiration rates in protected forests can reduce overland runoff and therefore sediment generation and transport (Edwards et al., 2014; Wang et al., 2011; Wu et al., 2014) . Upstream protected forests may a lso enhance the longevity of dams with the reduction of sediment flows to a reservoir (Bilotta and B razier, 2008) . Furthermore, protected forests can provide sustainable water sources for hydropower production by influencing river discharge via rainfall and soil moisture (Moran et al., 2018; Stickler et al., 2013) . Wetland PAs help sustain the amounts of surface water supplies to cities. Pr otected wetlands retain water in wetland soils and vegetation, and the water gradually flows into streams and rivers (Maltby and Acreman, 2011) . But cities mainly depend on dams for flood mitigation. Our results show that PAs can enhance various freshwat er ES supplies for cities, in addition to their main purpose of biodiversity conservation. Thus, integrating built infrastructure conserving natural habitats for biodi versity (Visconti et al., 2019; Vörösmarty et al., 2018; Yang et al., 2019) . Watershed conservation activities for freshwater ES flows could also support the global sustainable development agenda (Vörösmarty et al., 2018) . Integrating the two approaches can have co - benefits for multiple SDGs simultaneously, including freshwater sources (SDG 6, clean water and sanitation), h ydropower production (SDG 7, affordable & clean energy), dams (SDG 9, industry, innovation of infrastructure), cities (SDG 11, sustainable cities & communities), and biodiversity (SDG 15, life on land) (Bhaduri et al., 2016; Garrick et al., 2017; United Nations, 2015; Vörösmarty et al., 2018) . 99 5.4.2. Conservation strategies for freshwater ES This study highlights important implications for new PA designations that help sust ain freshwater ES supplies to cities. As expected, increased urban populations had a positive relationship with the amounts of freshwater provision and sediment loads from source watersheds. However, increased affluence in cities was negatively associated with both freshwater provision and sediment loads, suggesting an ameliorating impact of affluence. Cities with low affluence may have to use low quality of freshwater, partly because of the lack of water infrastructure and conservation activities in their source watersheds (McDonald et al., 2014; Romulo et al., 2018) crucial to reduce sediment flows, because high sediments in source watersheds cause additional costs for urban water treatment (Bilotta and Brazier, 2008; McDonald et al., 2016; Vercruysse et al., 2017) . Such PA designations, however, need to consider other social, economic, and political contexts to avoid potential conflicts with local communities (Dinerstein et al., 2019; K remen and Merenlender, 2018; Symes et al., 2016) . In addi tion, increased temperature tended to increase sediment load and flood risk for global cities by affecting the availability of water resources. For instance, high temperature may increase sediment flows from source watersheds with the reduction of vegetati on cover as well as the loss of ground aggregates (Achite and Ouillon, 2007; Haritashya et al., 2006) . A warmer climate with high temperatures may also raise flood risks worldwide (Hirabayashi et al., 2013) . If this cross - sectional relationship holds with temporal changes, climate changes could exacerbate th ese problems. Our analysis can also help target areas where conservation actions might improve the flows of multiple freshwater ES. Source watersheds, particularly in China and India, have largely 100 focused on dam construction without new PA designations fro m 2000 to 2016. While PAs appear not to change the flood protection services from dams, PA designations in the source watersheds of these dams could add to sediment reduction and hydropower production. Additionally, although our network analyses included g eological characteristics (i.e., watershed size, distance to cities, elevation, and slope) largely to avoid spurious effects, results indicate that these geological characteristics played an important role in freshwater ES supplies. Source watersheds with a larger size, greater distance from cities, and steeper slope have less sediment flow and more hydropower production. Source watersheds with a larger size, greater distance from cities, and lower elevation have more freshwater provision while reducing sed iment flows. Other types of watershed conservation activities would also work for sustaining freshwater ES supplies in highly developed watersheds. Our results show that there was a conflict for freshwater resources between irrigated croplands in water demand. In the context of water resource conflicts, IWS programs might be an alternative to PAs for freshwater provisioning and could possibly balance the negative effects of irrigated croplands to meet the freshwater de mand that comes from population and affluence (Romulo et al., 2018; Zheng et al., 2013) . Many cities adopted IWS programs after experiencing low freshwater prov isioning, partly because of high irrigation withdrawals in source watersheds (McDonald et al., 2016; Romulo et al., 2018; Veldkamp et al., 2017; Zheng et al., 2013) . In source watersheds, irrigation demands from agriculture are driven by food demand beyond city and watershed boundaries (Dalin et al., 2012; Dalin et al., 2017; Soligno et al., 2019) . For instance, in India, which accounts for 6.8% of the global net increase in green leaf areas from 2000 to 2017 , cropland increases contributed to 82% of that net increase (Chen et al., 2019) . Thus, future watershed conservation activities in India might find the IWS approach a useful tool 101 for improving freshwater provisioning, although further research is clearly needed to establish more firmly the effects of IWS programs. In some countries, other t ypes of payments for ES may also have indirect effects on forest conservation program the Natural Forest Conservation Program (NFCP) to conserve and restore fores ts. The NFCP in China has significantly contributed to net increases in forest cover over the past two decades (Chen et al., 2019; Viña et al., 2016) . Since the NFCP bans and monitors illegal logging and harvesting in natural forests (Viña et al., 2016) , conservation and restoration of forests under this program may provide additional freshwater ES to cities (Ouyang et al., 2016) . These watershed conservation activities can be expanded to other regions that experience rapid dam construction and high levels of human intervention without any watershed conservation efforts. Our study has several limitat ions. We note that our multi - level models correlate rather than estimate causal directions among dams, conservation activities, and freshwater ES flows because of the lack of time - series data. These interrelationships may be altered with seasonal changes o f freshwater ES supplies and different changes between dam numbers and conservation activities over time. 5.5. Conclusions This study determines the relationships of built infrastructure and watershed conservation activities with freshwater ES for cities. Our findings and approach provide a new perspective to the link between cities that demand freshwater ES and the watersheds that provide them. From a practical point of view, our analyses suggest ways that watershed conservation activities can 102 enhance the function of built infrastructure in providing sustainable freshwater ES supplies in urban water systems. One of the best ways is to integrate dam construction and conservation policies. Our results may indicate that forest cover in PAs improves the capaci ty of large dams in sediment reduction and hydropower production. Wetland cover in PAs helps provide further freshwater provisioning to cities. In our analysis, dams mainly contribute to flood mitigation in cities, although of course this global pattern un doubtedly has many regional exceptions. In conclusion, our findings capture the role of watershed conservation activities for freshwater ES supplies to global cities as well as their potential provision of additional ES that can enhance the function of lar ge dams. Therefore, we hope this study sets groundwork for future research. 103 CHAPTER 6 CONCLUSIONS 104 By integrating the telecoupling framework with ecosystem services, this dissertation uncovers complex t elecoupling p rocesses of tourism, food trad e, and flows of fresh water with biodiversity, conservation activities , and h uman w ell - b eing across distant places. My research includes these important telecoupling processes regarding the dynamic flows of ecosystem services at the global level. Some of t he greatest challenges revolve around increasing pressures on food production for exports in tropical countries , water demands in agriculture and urban areas, and tou rism in conservation areas that have high cultural heritages and ecological hotspots . The second chapter of this dissertation examines how different conservation strategies affect nature - based tourism in terrestrial protected areas. My results show that protected areas strictly managed for biodiversity protection have 35% more v isitations than those managed for mixed use. In addition, biodiversity is positively associated with the number of nature - based tourists even when other environmental and socioeconomic factors are controlled. Management for biodiversity has a positive rela tionship with the number of species. Therefore, the results imply that enhancing both biodiversity and nature - based tourism together is feasible given suitable conservation strategies. The third chapter of this dissertation adopts a network analysis approa ch that determines the dynamics of global tourism networks and identifies underlying socioeconomic and environmental factors over time. My results confirm the consolidation of global tourism networks and identify which countries have rapidly contributed to this consolidation over the past two decades. In consolidated global tourism networks, my mixed - effects model provides key strategic factors for proper tourism development and destination management . For example, results indicate that transaction costs (e .g., shared language, geogra phic distance, and visa policy ) are more important in attracting international tourists than natural and cultural attractions 105 (e.g., protected areas and World Cultural Heritage sites). My network approach and findings provide a science - based evidence for decision - making to implement proactive tourism policies. The fourth chapter of this dissertation performs comprehensive global analyses to reveal unexpected food trade among developed and developing countries with or without biod iversity hotspots. This research combines biodiversity hotspots and economic development status and provides a new perspective that international food trade may benefit biodiversity conservation in both developing and developed countries with biodiversity hotspots. Additionally, my results indicate that developing countries without biodiversity hotspots played an increasingly important role as net exporters in international food trade. With rising attention to biodiversity conservation and food security, it is time to develop innovative approaches to minimize the negative impacts of global food production and trade on biodiversity hotspots. The fifth chapter of this dissertation determines a feasible way to integrate watershed conservation activities with bu ilt infrastructure approaches to sustain essential freshwater ecosystem services for global cities. This research adopts an egocentric network analysis to investigate the complex relationships of watershed conservation activities and built infrastructure o n the flows of key freshwater ecosystem services from source watersheds to cities worldwide. My results show that forest cover in protected areas complement large dams for sediment reduction and hydropower production. In addition, wetland cover in protecte d areas contributes to provide further freshwater provisioning to cities. However, global cities mainly depend on large dams for flood mitigation. My network approach and findings capture the role of watershed conservation activities for freshwater ecosyst em services to integrate built infrastructure with watershed conservation activities to enhance urban water sustainability. 106 To sum up, this dissertation research can help reduce the negative impacts of ecosystem service flows on sustainable development as my results provide science - based knowledge for implementing cross - border policies and landscape planning that works to achie ve environmental and socioeconomic sustainability. My approaches and findings also suggest the potential benefits of key ecosystem services with different conservation activities to inform policymakers, land managers, and other stakeholders, such as ecotou rists, farmers, and urban residents. The dissertation demonstrates how ecosystem service flows benefit different beneficiaries and those paying subsidies for global ecosystem service conservation. Future research beyond this dissertation can uncover the complex interactions of multiple telecoupling processes (e.g., food - water - energy nexus trade) with diverse ecosystem services across the world. Integrating mul tiple telecouplings simultaneously is more challenging than focusing on a single telecoupling because telecoupling processes interact with each other across spatial and temporal scales (Liu et al., 2018; Liu et al., 2015b) . Quantifying the dynamics of multiple telecouplings with key ecosystem services is essential to develop practical policies and land management strategies for glob al sustainable development. For example, food production for exports surrounding protected areas may threaten charismatic species that play an important role to attract nature - based tourists. Also, the increasing water uses for irrigation in source watersh eds may decrease freshwater provisioning to urban areas. Establishing protected areas in source watersheds may enhance the retention of fresh water, while providing natural attractions for tourists. Transforming the understanding of multiple telecoupling p rocesses and their cross - sectoral interactions with ecosystem services can help decision - makers formulate adaptive management priorities for natural lands in ways that promote target biodiversity while 107 provisioning essential ecosystem services to beneficia ries. This dissertation provides a foundation for future research by integrating the telecoupling framework with ecosystem services at the global level. This dissertation can help with the development of win - win proactive strategies for both biodiversity a nd ecosystem service conservation to enhance global sustainability. 108 APPENDICES 109 APPENDIX A SUPPORTING INFORMATION FOR CHAPTER 2 Table S 2 . 1 . The descriptions of dependent and independent variables. Variable Dataset Unit of Measure Time Period Spatial Extent Reference Link Nature - based Tourism PAs visitor numbers Annual visitation data for PAs person 2000 - 2014 Global Balmford et al. (2015) , National Statistics, Gray literatures http://jo urnals.plos.or g/plosbiology/article? id=10.1371/journal.p bio.1002074 Demographic Population LandScan person (30arc second~1km 2 ) 2002 - 2012 Global, raster Bright et al. (2013) http://web.ornl.gov/sc i/landscan Biodiversity The number of species Birds, Mammals, Amphibians species 2000s Global, raster Pimm et al. (2014) http://biodiversityma pping.org Agricultural factor Agricultural Yields Yields for 175 crops tonne/ km 2 (5arc mi nute~10km 2 ) 2000 Global, raster (Monfreda et al., 2008) http://www.earthstat. org/data - download Agricultural area Cultivated and managed vegetation area proportion (0 - 100, 30arc second~1km 2 ) the early 2000s Global, raster Tuanmu and Jetz (2014) http://www.earthenv. org/landcover.html Regulating ES Upstream Protected Land % total water supply originated in protected land % of total water supply originated in PAs 2000s Global, watershed WRI (2015) http://www.wri.org/r e sources/data - sets/aqueduct - global - maps - 21 - data Economic Income level GDP per capita 2005 USD const. 2000 - 2014 Country level United Nations Statistics Division (2015) http://data.un.org Protected Area PAs management status IUCN PAs management category Category, II - VI 2014 Global IUCN and UNEP - WCMC (2017) http://www.protected planet.net PAs age Subtraction of PAs establishment year from 2014 year 2014 Global IUCN and UNEP - WCMC (2017) http://www.protected planet.net Size of PAs PAs areas square kilometers 2014 Global IUCN and UNEP - WCMC (2017) http://www.protected planet.net 110 Variable Dataset Unit of Measure Time Period Spatial Extent Reference Link Protected Area Elevation Global Multi - resolution Terrain Elevation (GMTED 2010) Elevation (30arc second~1km 2 ) 2010 Global, raster EROS Data Center (2015) http://earthexplorer.us gs.gov https://lta.cr.usgs.gov /GMTED2010 Temperature Annual mean temperature °C (30 arc second~1km 2 ) 1950 - 2000 Global, raster Hijmans et al. (2005) http://www.worldcli m.org Precipitation Annual mean precipitation millimeter (30 arc second~1km 2 ) 1950 - 2000 Global, raster Hijmans et al. (2005) http://www.worldcli m.org PAs remoteness from major cities (>50,000) Travel time to major cities: A global map of Accessibility Time (minutes) (30 arc second~1km 2 ) 2000 Global, raster Nelson (2008) http://forobs.jrc.ec.eu ropa.eu/products/gam /download.php 111 Table S 2 . 2 . List of PAs (N=929). Africa (N=90) Cameroon Mbam et Djerem (II), Nki (II), Waza (II) Republic of the Congo Nouabale - Ndoki (II) Ethiopia Bale Mountains (II) Ghana Bomfobiri (IV), Bui (II), Digya (II), Gbele (VI), Kakum (II), Kalakpa (VI), Mole (II), Owabi (IV), Shai Hills (VI) Kenya Aberdare (II), Amboseli (II), Arabuko Sokoke (II), Hell's Gate (II), Lake Bogoria (II), Lake Nakuru (II), Longonot (II), Meru (II), Mount Kenya (II), Nairobi (II), Samburu (II), Shimba Hills (II), Tsavo East (II), T savo West (II) Madagascar Analamerana (IV), Montagne d'Ambre (II) Namibia Ai - Ais Hot Springs (II), Cape Cross Seal Reserve (IV), Daan Viljoen Game Park (II), Etosha (II), Gross Barmen Hot Springs (III), Hardap Recreation Resort (V), Khaudum (II), Mangetti (II), Mudumu (II), Nkasa Rupara (II), Popa Game Park (III), Skeleton Coast Park (II), Von Bach Recreation Resort (V), Waterberg Plateau Park (II) Rwanda Akagera (II), Nyungwe (IV), Volcans (II) Tanzania Arusha (II), Gombe (II), Katavi (II), Kili manjaro (II), Kitulo (II), Lake Manyara (II), Mahale (II), Mikumi (II), Mkomazi (IV), Ruaha (II), Rubondo (II), Selous (IV), Serengeti (II), Tarangire (II), Udzungwa Mountains (II) Uganda Bwindi Impenetrable (II), Katonga (III), Kidepo Valley (II), Lake Mburo (II), Mgahinga Gorilla (II), Mount Elgon (II), Murchison Falls (II), Queen Elizabeth (II), Rwenzori Mountains (II), Semuliki (II) South Africa Agulhas National Park (II), Augrabies Falls National Park (II), Bontebok National Park (II), Golden G ate Highlands National Park (II), Kalahari Gemsbok National Park (II), Karoo National Park (II), Kruger National Park (II), Mapungupwe National Park (II), Marakele National Park (II), Mokala National Park (II), Mountain Zebra National Park (II), Namaqua Na tional Park (II), Richtersveld National Park (II), Table Mountain National Park (II), Tankwa - Karoo National Park (II), Vaalbos National Park (II) Zambia Kasanka (II), Lavushi Manda (II) 112 Table S2.2 . Asia and Oceania (N=351) UAE Dubai Desert Conservation Reserve (II) Australia Ben Lomond (II), Douglas - Apsley (II), Hartz Mountains (II), Hastings Caves (III), Kakadu National Park (II), Mole Creek Karst (II), Moreton Island (II), Mount Field (II), Purnululu (II), Uluru - Kata Tjuta National Park (II) China Dafengmilu (Jiangsu) (V), Huanglongsi (V), Jiuzhaigou (V), Wolong (V), Wuyishan (V) Indonesia Alas Purwo (II), Baning (V), Bantimurung Bulusaraung (II), Batang Gadis (II), Batu Angus (V), Batu Putih (V), Berbak (II), Betung Kerihun (II), Bogani Nani Wartabone (II), Bromo Tengger Semeru (II), Bukit Baka - Bukit Raya (II), Bukit Barisan Selatan (II), Bukit Dua Belas (II), Bukit Kaba (V), Bukit Serelo (V), Bukit Tiga Puluh (II), Bunder (VI), Camplong (V), Cani Sirenreng (V), Car ita (VI), Cimanggu (V), D. Sicikeh - cikeh (V), Danau Matano (V), Danau Sentarum (II), Grojogan Sewu (V), Gunung Baung (V), Gunung Ciremai (II), Gunung Gede - Pangrango (II), Gunung Guntur (V), Gunung Halimun - Salak (II), Gunung Kelam (V), Gunung Leuser (II ), Gunung Meja (V), Gunung Merapi (II), Gunung Merbabu (II), Gunung Palung (II), Gunung Pancar (V), Gunung Rinjani (II), Gunung Tampomas (V), Holiday Resort (V), Jember (V), Kawah Ijen (V), Kawah Kamojang (V), Kayan Mentarang (II), Kelimutu (II), Kerandang an (V), Kerinci Seblat (II), Klamono (V), Kutai (II), Laiwangi Wanggameti (II), Lejja (V), Lore Lindu (II), Madapangga (V), Malino (V), Mangolo (V), Manupeu Tanadaru (II), Manusela (II), Meru Betiri (II), Minas (Sultan Sarif Hasyim) (VI), Muka Kuning (V), Nanggala III (V), Papandayan (V), Pulau Kembang (V), Punti Kayu (V), Rawa Aopa Watumohai (II), Rimbo Panti (V), Ruteng (V), Sebangau (II), Seblat (V), Semongkat (V), Siberut (II), Sidrap (V), Sorong (V), Sultan Adam (VI), Suranadi (V), Tahura Ir. H. Juanda (VI), Talaga Bodas (V), Telaga Patengan (V), Telogo Warno Pengilon (V), Tesso Nilo (II), Tretes (V), Way Kambas (II) India Bandhavgarh (II), Bandipur (II), Bhadra (IV), Corbett (II), Jaldapara (IV), Kalakad (IV), Kanha (II), Kaziranga (II), Ken Gharial ( IV), Melghat (IV), Mudumalai (IV), Panna (II), Pench (II), Periyar (IV), Rajiv Gandhi (Nagarhole) (II), Ranthambhore (II), Sariska (IV), Satpura (II), Valley Of Flowers (II) Japan Aichikogen (V), Akan (II), Akiyoshidai (V), Aso kuju (V), Bandai asahi (II) , Biwako (V), Chichibu tama kai (V), Chubusangaku (II), Daisetsuzan (II), Echigosanzan - Tadami (II), Hakusan (II), Hayachine (II), Hiba - Dogo - Taishaku (V), Hida - Kisogawa (V), Hyonosen - Ushiroyama - Nagisan (V), Ibi - Sekigahara - Yoro (V), Iriomote (IV), Ishizuchi (V), Kitakyushu (V), Kongo - Ikoma - Kisen (V), Koya - Ryujin (V), Kurikoma (II), Kushiroshitsugen (II), Kyushuchuosanchi (V), Meiji Memorial Forest Minoo (V), Meiji Memorial Forest Takao (V), Minami alps (II), Muroo - Akame - Aoyama (V), Myogi - Arafune - Sakukogen (V) , Nikko (V), Nishichugokusanchi (V), Onuma (V), Shikotsu toya (II), Shiretoko (IV), Sobo - Katamuki (V), Suzuka (V), Tanzawa - Oyama (V), Tenryu - Okumikawa (V), Towada hachimantai (II), Tsurugisan (V), Yaba - Hita - Hikosan (V), Yamato - Aogaki (V), Yatsugatake - Chush inkogen (V), Zao (V) 113 T able S2.2 . South Korea Biseulsan County Park (V), Bogyeongsa County Park (V), Bongmyeongsan County Park (V), Bukhansan (V), Bullyeonggyegok County Park (V), Cheongnyangsan Provincial Park (V), Cheongwansan Provincial Par k (V), Cheonmasan County Park (V), Chiaksan (II), Chilgapsan Provincial Park (V), Daedunsan Provincial Park (V), Daeiri County Park (V), Deogyusan (V), Duryunsan Provincial Park (V), Gajisan Provincial Park (V), Gangcheonsan County Park (V), Gayasan (II), Geumosan Provincial Park (V), Gibaeksan County Park (V), Gobok Provincial Park (V), Gwangyang Baegunsan (IV), Gyoungpo Provincial Park (V), Hwangmaesan County Park (V), Hwawangsan County Park (V), Ipgok County Park (V), Jangansan County Park (V), Jirisan ( II), Juwangsan (II), Maisan Provincial Park (V), Moaksan Provincial Park (V), Mungyeongsaejae Provincial Park (V), Myeongjisan County Park (V), Naejangsan (II), Naksan Provincial Park (V), Namhansanseong Provincial Park (V), Odaesan (II), Sangnim Woods in Hamyang (IV), Seonunsan Provincial Park (V), Seoraksan (II), Sobaeksan (II), Songnisan (II), Taebaeksan Provincial Park (V), Unmunsan County Park (V), Upo Wetland (IV), Valley of Bulyeongsa Temple in Uljin (V), Wolchulsan (II), Woraksan (II) Sri Lanka Bun dala (II), Gal Oya (II), Galway's Land (IV), Horton Plains (II), Kaudulla (II), Lahugala (II), Lunugamwehera (II), Maduru Oya (II), Minneriya (II), Uda Walawe (II), Wasgamuwa (II), Wilpattu (II), Yala East(Kumana) (II) Malaysia Batang Ai (II), Endau Rompin (Johor) (II), Gunong Gading (II), Kubah (II), Lambir Hills (II), Loagan Bunut (II), Niah (II), Taman Negara (II), Tanjong Datu (II), Tawau Hill Park (II) Thailand Bang Lang (II), Budo - Sungai Padi (II), Chae Son (II), Chalearm Rattanakosin (II), Doi Inthanon (II), Doi Khuntan (II), Doi Luang (II), Doi Phaklong (II), Doi Phukha (II), Doi Suthep - Pui (II), Erawan (II), Huai Nam Dang (II), Kaeng Krung (II), Kaeng Tana (II), Kaengkrachan Forest Complex (II), Khao Chamao - Khao Wong (II), Khao Khitchakut (II ), Khao Laem (II), Khao Luang (II), Khao Nam Khang (II), Khao Nan (II), Khao Phanom Bencha (II), Khao Phravihan (II), Khao Pu - Khao Ya (II), Khao Sib Ha Chan (II), Khao Sok (II), Khao Yai (II), Khlong Lamngu (II), Khlong Lan (II), Khlong Wang Chao (II), K huen Si Nakarin (II), Khun Chae (II), Khun Pra Vor (II), Klong Phanom (II), Kuiburi (II), Lansaang (II), Mae Charim (II), Mae Moei (II), Mae Phang (II), Mae Ping (II), Mae Puem (II), Mae Wa (II), Mae Wang (II), Mae Wong (II), Mae Yom (II), Mukdahan (II), N am Nao (II), Nam Phong (II), Namtok Chat Trakan (II), Namtok Huai Yang (II), Namtok Klong Kaew (II), Namtok Mae Surin (II), Namtok Ngao (II), Namtok Phleiw (II), Namtok Sai Khao (II), Namtok Si khid (II), Namtok Yong (II), Ob Luang (II), Pa Hin Ngam (II), Pang Sida (II), Pha Tam (II), Phu Chong - Na Yoi (II), Phu Hin Rong Kla (II), Phu Kao - Phu Phan Kham (II), Phu Kradueng (II), Phu Lan Ka (II), Phu Langka (II), Phu Pa - Yol (Huai Huat) (II), Phu Pha Lek (II), Phu Pha Man (II), Phu Phan (II), Phu Rua (II), Phu Sa Dokbua (II), Phu Soi Dao (II), Phu Toei (II), Phu Wiang (II), Phu Zang (II), Ramkamhaeng (II), Sai Thong (II), Sai Yok (II), Salawin (II), Si Nan (II), Si Phangnga (II), Sri Lanna (II), Sri Satchanalai (II), Ta Phraya (II), Taad Moak (II), Taad Ton (II), Tai Romyen (II), Taksin Maharat (II), Thaleban (II), Tham Pla - Pha Seu (II), Thap Lan (II), Thong Pha Phum (II), Thung Salaeng Luang (II), Wiang Kosai (II) 114 Table S2.2 . Nepal Annapurna (VI), Api - Nampa (VI), Bardia (II), Chitwan (II), D horpatan (VI), Gauri - Shankar (VI), Kanchanjunga (VI), Khaptad (II), Koshi Tappu (IV), Krishnasar (VI), Langtang (II), Makalu - Barun (II), Manaslu (VI), Parsa (IV), Rara (II), Shey - Phoksundo (II), Shivapuri - Nagarjun (II), Suklaphanta (IV) New Zealand Abel T asman (II) Philippines Mount Kitanglad Range (II), Mt. Pulag National Park (II), Puerto Princesa Subterranean River (III) Vietnam Cuc Phuong (II), Phong Nha - Ke Bang (II) Europe (N=215) Bulgaria Centralen Balkan (II), Vitosha (V) Czech Republic Finland Helvetinjarven kansallispuisto (II), Hiidenportin kansallispuisto (II), Isojarven kansallispuisto (II), Kauhanevan - Pohjankankaan kansallispuisto (II), Kolin kansallispuisto (II), Koloveden kansallispuisto (II), Kurjenrahkan kansallispuisto (II), Lauhanvuoren kansallispuisto (II), Leivonmaen kansallispuisto (II), Liesjarven kansallispuisto (II), Linnansaaren kansallispuisto (II), Nuuksion kansallispuisto (II), Paijanteen kansallispuisto (II), Patvinsuon kansallispuisto (II), Petkeljarven kansallispuisto (II), Puurijarven ja Isonsuon kansallispuisto (II), Pyha - Hakin kansallispuisto (II), Repoveden kansallispuisto (II), Rokuan kansallispuisto (II), Salamajarven ka nsallispuisto (II), Seitsemisen kansallispuisto (II), Sipoonkorven kansallispuisto (II), Tiilikkajarven kansallispuisto (II), Torronsuon kansallispuisto (II), Valkmusan kansallispuisto (II) UK Arundel Park (IV), Attenborough Gravel Pits (IV), Aylesbeare C ommon (IV), Berney Marshes & Breydon Water (IV), Blean Woods (IV), Brampton Wood (IV), Brandon Marsh (IV), Brecon Beacons (V), Broads (V), Cairngorms (V), Castle Eden Dene (IV), Clifton Country Park (IV), Clumber Park (V), Coombe Valley Woods (IV), Danbury and Lingwood Commons (V), Dartmoor (V), Dungeness (IV), Elmley (IV), Epping Forest (IV), Exe Estuary (IV), Exmoor (V), Fairburn Ings (IV), Fowlmere (IV), Frampton Pools (IV), Gamlingay Wood (IV), Garston Wood (IV), Geltsdale (IV), Gibside (V), Ham Wall (I V), Havergate Island & Boyton Marshes (IV), Haweswater (IV), Hodbarrow (IV), Ken - Dee Marshes (IV), Lake District (V), Leighton Moss (IV), Loch Lomond and The Trossachs (V), Lochwinnoch (IV), Marshside (IV), Mere Sands Wood (IV), Mid Yare Valley (IV), Minsm ere (IV), Nagshead (IV), Nene Washes (IV), New Forest (V), North Warren (IV), North York Moors (V), Northumberland (V), Northward Hill (IV), Oare Marshes (IV), Ogden Water (IV), Orford Ness (V), Otmoor (IV), Ouse Washes (IV), Parndon Woods & Common (IV), P eak District (V), Pembrokeshire Coast (V), Poole's Cavern and Grin Low Wood (IV), Pulborough Brooks (IV), Queenswood (IV), Radipole Lake (IV), Rye Meads (IV), Sherwood Forest (IV), Snettisham Carstone Quarry (IV), Snowdonia (V), South Downs (V), The Lodge (IV), Titchfield Haven (IV), Titchmarsh (IV), Tudeley Woods (IV), West Sedgemoor (IV), Wolves Wood Reserves (IV), Wood Of Cree (IV), Yorkshire Dales (V) 115 Table S2.2 . Spain Aiguestortes i Estany de Sant Maurici (II), Cabaneros (II), Donana (II), El Teide (II), Garajonay (II), La Caldera de Taburiente (II), Ordesa y Monte Perdido (II), Parque Nacional de Timanfaya (II), Picos de Europa (II), Sierra Nevada (II), Tablas de Daimiel (II) France Causses du Quercy (V), La Narbonnaise en Mediterranee (V ), Volcans d'Auvergne (V) Croatia Krka (II), Paklenica (II), Plitvicka jezera (II), Risnjak (II), Sjeverni Velebit (II) Hungary Aggteleki (II), Balaton - felvideki (V), Bukki (II), Duna - Drava (V), Duna - Ipoly (V), Ferto - Hansagi (II), Hortobagyi (II), Kiskunsagi (II), Koros - Maros (V), Orsegi (V) Italy Parco nazionale dei Monti Sibillini (II), Parco regionale La Mandria (V) Poland Park Narodowy wy (V), Park Narodowy "Bory Tucholskie" (II), Park Narodowy (II), Wielkopolski Park Narodowy (II), Wigierski Park Narodowy (V) Portugal Alvao (V), Arriba Fossil da Costa da Caparica (V), Douro Internacional (V), Estuario do Sado (IV), Estuario do Tejo (IV), Montesinho (V), Paul de Arzila (IV), Paul do Boq uilobo (IV), Peneda - Geres (II), Ria Formosa (V), Sapal de Castro Marim e Vila Real de Santo Antonio (IV), Serra da Estrela (V), Serra da Malcata (IV), Serra de Sao Mamede (V), Serra do Acor (V), Serras de Aire e Candeeiros (V), Sintra - Cascais (V), Tejo Int ernacional (V), Vale do Guadiana (V) Romania Balta Mica a Brailei (V), Bucegi (V), Buila - Vanturarita (II), Calimani (II), Ceahlau (II), Cheile Bicazului - Hasmas (II), Cheile Nerei - Beusnita (II), Comana (V), Cozia (II), Defileul Jiului (II), Domogled - Valea Cernei (II), Geoparcul Dinozaurilor Tara Hategului (V), Geoparcul Platoul Mehedinti (V), Gradistea Muncelului - Cioclovina (V), Lunca Joasa a Prutului Inferior (V), Lunca Muresului (V), Muntii Apuseni (V), Muntii Macinului (II), Muntii Maramuresulu i (V), Piatra Craiului (II), Portile de Fier (V), Putna - Vrancea (V), Retezat (II), Rodna (II), Semenic - Cheile Carasului (II), Vanatori Neamt (V) Russia Kenozersky (II) Slovakia Biele Karpaty (V), Cerova vrchovina (V), Horna Orava (V), Kysuce (V), Mala Fatra (II), Muranska planina (II), Pieninsky (II), Polana (V), Poloniny (V), Slovensky raj (II), Stiavnicke vrchy (V), Strazovske vrchy (V), Tatransky (II) 116 Table S2.2 . Central and South America (N=155) Argentina Baritu (II), Bosques Petrificados (III), Calilegua (II), Campo de los Alisos (II), Chaco (II), El Leoncito (II), El Palmar (II), El Rey (II), Iguazu (II), Lago Puelo (II), Laguna Blanca (II), Laguna de los Pozuelos (III), Lanin (II), Lihue Calel (II), Los Alerces (II), Los Cardones (II), Los Glaciares (II), Mburucuya (II), Nahuel Huapi (II), Perito Moreno (II), Pre - Delta (II), Quebrada del Condorito (II), Rio Pilcomayo (II), San Guillermo (II), Sierra de las Quijadas (II), Talampaya (II) Belize Bermudian Landing Commun ity Baboon Sanctuary (IV) Bolivia Eduardo Avaroa (IV), Madidi (II), Noel Kempff Mercado (II) Brazil Area De Protecao Ambiental Da Chapada Dos Guimaraes (V), Floresta Nacional De Brasilia (VI), Parque Nacional Da Chapada Dos Veadeiros (II), Parque Nacional Da Serra Da Canastra (II), Parque Nacional Da Serra Da Capivara (II), Parque Nacional Da Serra Da Cipo (II), Parque Nacional Da Serra Do Divisor (II), Parque Nacional Da Serra Dos Orgaos (II), Parque Nacional Da Tijuca (II), Parque Nacional Das Em as (II), Parque Nacional De Aparados Da Serra (II), Parque Nacional De Caparao (II), Parque Nacional De Sete Cidades (II), Parque Nacional De Ubajara (II), Parque Nacional Do Itatiaia (II), Parque Nacional Serra Das Confusoes (II) Chile Alerce Andino (II) , Alerce Costero (III), Alto Biobio (IV), Altos de Lircay (IV), Altos de Pemehue (IV), Bellotos El Melado (IV), Bosque Fray Jorge (II), Cerro Castillo (IV), Cerro Nielol (III), Chiloe (II), Conguillio (II), Contulmo (III), Coyhaique (IV), Cueva Del Milodon (III), Dos Lagunas (III), El Morado (III), El Yali (IV), Federico Albert (IV), Futaleufu (IV), Hornopiren (II), Huemules de Niblinto (IV), Huerquehue (II), La Campana (II), Lago Cochrane (IV), Lago Jeinemeni (IV), Lago Las Torres (IV), Lago Penuelas (IV), Laguna del Laja (II), Laguna El Peral (IV), Laguna Parrillar (IV), Laguna Torca (IV), Lahuen Nadi (III), Las Chinchillas (IV), Las Vicunas (IV), Lauca (II), Llanos del Challe (II), Llanquihue (IV), Los Flamencos (IV), Los Queules (IV), Los Ruiles (IV), Ma gallanes (IV), Malalcahuello (IV), Malleco (IV), Mocho - Choshuenco (IV), Nahuelbuta (II), Nalcas (IV), Nevado Tres Cruces (II), Nuble (IV), Pali Aike (II), Pampa del Tamarugal (IV), Pan De Azucar (II), Pichasca (III), Puyehue (II), Queulat (II), Radal Sie te Tazas (IV), Ralco (IV), Rio Clarillo (IV), Rio Los Cipreses (IV), Rio Simpsom (IV), Robleria Cobre Loncha (IV), Salar De Surire (III), Tolhuaca (II), Torres del Paine (II), Vicente Perez Rosales (II), Villarrica (II), Volcan Isluga (II), Yerba Loca (IV) Costa Rica Arenal (II), Bahia Junquillal (estatal) (IV), Barbilla (II), Barra del Colorado (mixto) (IV), Barra Honda (II), Braulio Carrillo (II), Cano Negro (mixto) (IV), Carara (II), Chirripo (II), Golfito (mixto) (IV), Grecia (VI), Guanacaste (II), Igu anita (estatal) (IV), Internacional La Amistad (II), Juan Castro Blanco (II), Las Tablas (VI), Los Santos (VI), Mata Redonda (estatal) (IV), Palo Verde (II), Rincon de la Vieja (II), Rio Macho (VI), Taboga (VI), Volcan Irazu (II), Volcan Poas (II), Volcan Tenorio (II) 117 Table S2.2 . Equador Cajas (II), Cayambe Coca (VI), Chimborazo (IV), Cotacachi Cayapas (VI), Cotopaxi (II), Cuyabeno (VI), El Boliche (V), Limoncocha (VI), Llanganates (II), Parque Lago (V), Podocarpus (II), Sangay (II), Sumaco Na po - Galeras (II), Yacuri (II), Yasuni (II) Peru Bahuaja Sonene (II), Calipuy (III) North America (N=118) Canada Bruce Peninsula National Park of Canada (II), Cape Breton Highlands National Park of Canada (II), Elk Island National Park of Canada (II), Fathom Five National Marine Park of Canada (VI), Fundy National Park of Canada (II), Georgian Bay Islands National Park of Canada (II), Grasslands (III), Gros Morne National Park of Canada (II), Gwaii Haanas National Park Reserve and Haida Heritage s ite (II), Kejimkujik National Park and National Historic site of Canada (II), Mount Revelstoke National Park of Canada (II), Parc National du Canada de la Mauricie (II), Point Pelee National Park of Canada (II), Prince Edward Island National Park of Canada (II), Pukaskwa National Park of Canada (II), Reserve de Parc National du Canada de l'Archipel - de - Mingan (II), Riding Mountain National Park of Canada (II), Terra Nova National Park of Canada (II), Thousand Islands National Park of Canada (II), Waterton La kes National Park of Canada (II) USA Acadia (IV), Agate Fossil Beds (V), Alibates Flint Quarries (V), Allegheny Portage Railroad (II), Aniakchak (V), Apostle Islands (V), Arches (II), Badlands (II), Bandelier (V), Big Bend (II), Big Thicket (V), Black Can yon of the Gunnison (II), Bluestone (V), Bryce Canyon (II), Canyonlands (II), Capitol Reef (II), Carlsbad Caverns (II), Casa Grande Ruins (V), Catoctin Mountain (II), Cedar Breaks (III), Chiricahua (V), City of Rocks (V), Colorado (III), Congaree (II), Cor onado (III), Cowpens (III), Crater Lake (II), Craters of the Moon (III), Cumberland Gap (III), Cuyahoga Valley (II), Death Valley (II), Dinosaur (III), El Malpais (III), El Morro (V), Fort Bowie (III), Fort Pulaski (V), Fort Union Trading Post (III), Fort Washington (V), Gila Cliff Dwellings (V), Glacier (II), Grand Canyon (II), Grand Portage (V), Grand Teton (II), Great Basin (II), Great Sand Dunes (II), Great Smoky Mountains (II), Guadalupe Mountains (II), Hovenweep (V), Indiana Dunes (V), Jean Lafitte Na tional Historical Park and Preserve, Barataria (V), Jewel Cave (V), John Day Fossil Beds (III), Johnstown Flood (III), Joshua Tree (II), Katmai (II), Kenai Fjords (II), Kings Canyon / Sequoia (II), Klondike Gold Rush (V), Lake Chelan (V), Lake Mead (V), La va Beds (III), Little Bighorn Battlefield (V), Mammoth Cave (II), Mesa Verde (II), Missouri (V), Mojave (V), Montezuma Castle (V), Mount Rainier (II), Mount Rushmore (V), Natural Bridges (III), Niobrara (V), North Cascades (II), Olympic (II), Oregon Caves (V), Organ Pipe Cactus (III), Ozark (V), Pecos (V), Pictured Rocks (V), Pipe Spring (V), Rio Grande (V), Ross Lake (V), Saguaro (II), Santa Monica Mountains (V), Scotts Bluff (V), Shenandoah (II), Sleeping Bear Dunes (V), Theodore Roosevelt (II), Theodore Roosevelt Island (II), Timpanogos Cave (V), Tonto (V), Tumacacori (V), Tuzigoot (V), Voyageurs (II), Washington Monument (V), White Sands (V), Wupatki (III), Yosemite (II), Zion (II) (in parentheses) IUCN management categories 118 Table S 2 . 3 . Variance Inflation Factor (VIF) for variables used in linear regression. Category Variable VIF Biodiversity Total species (species) 3. 244 Protected Area IUCN category (I - IV=1) 1.2 44 Size of PA (km 2 ) 1.7 34 Mean elevation (meter) 1. 763 Annual mean temperature ( ºC ) 2.5 07 Annual precipitation (mm) 1 . 908 PA remoteness (minutes) 3.6 25 PA age (year) 1.2 62 Demographic Population density § (persons/km 2 ) 3. 373 Economic GDP per capita ¶ (2005 const. $ per capita) 3. 690 Agricultural factor Agricultur al yields § (to nne/ km 2 ) 1 . 440 Agricultural area § (%) 2. 635 Regulating ES Water supply originated in PAs (%) 1. 344 Region Africa 2. 699 Europe 2. 035 North America 3. 193 Latin America 2.3 51 § 10 - km buffer zone ¶ Country level data, not PAs level 119 Table S 2 . 4 . Unstandardized coefficients from multiple regression model predicting annual visitor numbers except biodiversity and total species in PAs. Category Variable Annual visitors Total species Biodiversity Total species (species) - - Protected Area IUCN category (II - IV=1) 0.392* (0.168) 0.052* (0.023) Size of PA (km 2 ) 0.316** (0.040) 0.008 (0.006) Mean elevation (meter) 0.362** (0.058) 0.035** (0.008) Annual mean temperature ( ºC ) - 0.132 (0.148) 0.289** (0.020) Annual precipitation (mm) - 0.329* (0.111) 0.178** (0.015) PA remoteness (minutes) - 0.225* (0.113) 0.018 (0.016) PA age (year) 0.651** (0.118) - 0.013 (0.016) Demographic Population density § (persons/km 2 ) 0.447** (0.066) - 0.0002 (0.009) Economic GDP per capita ¶ (2005 const. $ per capita) 1.176** (0.085) - 0.104** (0.012) Agricultural factor Agricultur al yields § (tonne/ km 2 ) 0.076 (0.063) - 0.020* (0.009) Agricultural area § (%) 0.077 (0.091) 0.058** (0.013) Regulating ES Water supply originated in PAs (%) 0.059 (0.081) 0.060** (0.011) Region Asia and Oceania 1.894** (0.227) 0.044 (0.031) Africa 1.232** (0.315) 0.306** (0.044) Europe 0.756* (0.283) 0.113* (0.039) North America 1.529** (0.297) 0.357** (0.041) Intercept - 6.251** 3.978** R 2 0.470 0.692 F - statistic 50.61 127.9 DF 912 912 * P<0.05, ** P<0.001 § 10 - km buffer zone ¶ Country level data, not PAs level Values in parentheses are standard errors 120 Figure S 2 . 1 . indicates positive correlation for a given pair, and red indicates negative correlation. Colored correlation coefficients are significant at the p=0.05 level. 121 APPENDIX B SUPPORTING INFORMATION FOR CHAPTER 3 Table S 3 . 1 . Odds ratios for cluster analysis and p - value based on simulations followed by mean, median, and 95% Quantile interval of simulations. N Odds ratio p - value mean median 2.5% 97.5% 2000 2 0.852 <0.001 0.592 0.604 0.346 0.638 2001 7 0.804 <0.001 0.594 0.604 0.345 0.641 2002 9 0.778 <0.001 0.598 0.608 0.351 0.640 2003 3 0.715 <0.001 0.597 0.609 0.348 0.641 2004 11 0.822 <0.001 0.600 0.604 0.546 0.642 2005 2 0.856 <0.001 0.592 0.598 0.539 0.629 2006 2 0.851 <0.001 0.588 0.599 0.347 0.630 2007 9 0.765 <0.001 0.588 0.597 0.344 0.626 2008 9 0.789 <0.001 0.586 0.585 0.555 0.617 2009 12 0.819 <0.001 0.579 0.586 0.370 0.615 2010 9 0.779 <0.001 0.571 0.579 0.357 0.605 2011 2 0.822 <0.001 0.561 0.568 0.466 0.591 2012 2 0.845 <0.001 0.567 0.570 0.537 0.597 2013 5 0.713 <0.001 0.566 0.569 0.533 0.598 122 Figure S 3 . 1 . Clusters of global tourism networks by country: (a) 2000 2002, (b) 2011 2013, (c) 2000, (d) 2001, (e) 2002, (f) 2003, (g) 2004, (h) 2005, (i) 2006, (j) 2007, (k) 2008, (l) 2009, (m) 2010, (n) 2011, (o) 2012, and (p) 2013. 123 Figure S 3 . 2 . Mean and 95% Highest Posterior Density (HPD) confidence intervals of the coefficients from 2000 2013 in the alternative model: (a) the proportion of protected areas in receiving countries, (b) the p roportion of World Cultural Heritage sites in receiving countries, and (g) the number of direct flights between countries. 124 APPENDIX C SUPPORTING INFORMATION FOR CHAPTER 4 Figure S 4 . 1 . Global distribution of hotspot countries (high - hotspot countries [HHC], low - hotspot countries [LHC] and non - hotspot countries [NHC]). NA = countries with missing data. Raw data from Myers et al. (2000) and Myers (2003) . 125 Figure S 4 . 2 . The amounts of net food trade between developed and developing countries in high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC) from 2000 2015. Non - hotspot countries are indicated by red, high - hotspot countries dark green , and low - hotspot countries by light green. 126 Figure S 4 . 3 . The percentage of agricultural area in biodiversity hotspots out of total agricultural area. Raw data from Myers et al. (2000) , Myers (2003) , and Tuanmu and Jetz (2014) . 127 Figure S 4 . 4 . Number of countries with different percentages of biodiversity hotspots (land area with biodiversity hotspots out of total terrestrial land area). Raw data from Myers et al. (2000) and Myers (2003) . 128 Figure S 4 . 5 . Annual food flows (Mt) in 2000. Food flows between developed and developing countries in high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC). Non - hotspot countries are indicat ed by red, high - hotspot countries by dark green, and low - hotspot countries by light green. The arc length of an outer circle indicates the sum of food exported and imported in each group. The arc length of a middle circle refers to the amounts of food expo rted. The inner arc length shows the amounts of food imported. Raw data from UN FAO (2018) . 129 Figure S 4 . 6 . Annual food flows (Mt) in 2015. Food flows between developed and developing countries in high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC). Non - hotspot countries are indicated by red, high - hotspot countries by dark green, and low - hotspot countries by light green. The arc length of an outer circle indicates the sum of food exported and imported in each group. The arc length of a middle circle refers to the amounts of food exported. The inner arc length shows the amoun ts of food imported. Raw data from UN FAO (2018) . 130 Table S 4 . 1 . List of hotspot and non - hotspot countries, subdivided into developed and developing countries groups. High - hotspot Countries (HHC), N=64 Developed, N=14 Antigua and Barbuda, Bahamas, Barbados, Brunei Darussalam, Chile, Cyprus, Greece, Italy, Japan, Malta , New Zealand, Portugal, Saint Kitts and Nevis, Spain Developing, N=50 Albania, Armenia, Azerbaijan, Belize, Cabo Verde, Cambodia, Costa Rica, Cuba, Djibouti, Dominica, Dominican Republic, Ecuador, El Salvador, Ethiopia, Fiji, Georgia, Grenada, Guatemala, Haiti, Honduras, Indonesia, Jamaica, Kyrgyzstan, Lao People's Democratic Republic, Lebanon, Liberia, Madagascar, Malaysia, Mauritius, Mexico, Morocco, Myanmar, Nepal, Nicaragua, Panama, Philippines, Saint Lucia, Saint Vincent and the Grenadines, Samoa, Sao Tome and Principe, Sierra Leone, Solomon Islands, Sri Lanka, Swaziland, Tajikistan, Thailand, Tunisia, Vanuatu, Viet Nam, Yemen Low - hotspot Countries (LHC), N=53 Developed, N=10 Australia, Croatia, France, Israel, Oman, Russian Federation, Sa udi Arabia, Slovenia, Uruguay, United States of America Developing, N=43 Afghanistan, Algeria, Argentina, Bangladesh, Benin, Bolivia, Bosnia and Herzegovina, Brazil, Bulgaria, Cameroon, Cote d'Ivoire, China, Colombia, Egypt, Ghana, Guinea, India, Iran, Ir aq, Jordan, Kazakhstan, Kenya, Macedonia, Malawi, Montenegro, Mozambique, Namibia, Nigeria, Pakistan, Paraguay, Peru, Rwanda, Serbia, South Africa, Sudan, Togo, Turkmenistan, Uganda, United Republic of Tanzania, Uzbekistan, Venezuela, Zambia, Zimbabwe Non - hotspot Countries (NHC), N=43 Developed, N=23 Austria, Belgium, Canada, Czechia, Denmark, Estonia, Finland, Germany, Iceland, Ireland, Kuwait, Latvia, Lithuania, Luxembourg, Netherlands, Norway, Republic of Korea, Slovakia, Sweden, Switzerland, Trinid ad and Tobago, United Arab Emirates, United Kingdom Developing, N=20 Angola, Belarus, Botswana, Burkina Faso, Central African Republic, Chad, Congo, Gabon, Gambia, Guinea - Bissau, Guyana, Hungary, Lesotho, Mali, Mongolia, Niger, Romania, Senegal, Suriname, Ukraine 131 Table S 4 . 2 . The quantity of food exports and imports for developed and developing countries in 2000, 2015, and 2000 2015 (average annual). Year Income level Variables HC NHC Total HHC LHC 2000 Food Export (Mt) 22.41 184.45 88.89 295.75 Developed Food Import (Mt) 86.22 63.43 112.76 262.41 Export - Import (Mt) - 63.81 121.02 - 23.87 33.34 Food Export (Mt) 57.92 95.97 6.71 160.60 Developing Food Import (Mt) 73.57 114.28 6.09 193.94 Export - Import (Mt) - 15.65 - 18.31 0.62 - 33.34 2015 Food Export (Mt) 31.75 248.65 128.91 409.31 Developed Food Import (Mt) 100.30 87.90 156.30 344.50 Export - Import (Mt) - 68.55 160.75 - 27.39 64.82 Food Export (Mt) 139.19 219.64 61.58 420.41 Developing Food Import (Mt) 138.03 332.14 15.05 485.23 Export - Import (Mt) 1.16 - 112.5 46.53 - 64.82 2000 - 2015 Food Export (Mt) 27.89 203.86 99.85 331.60 Developed Food Import (Mt) 94.86 77.05 137.69 309.60 Export - Import (Mt) - 66.97 126.81 - 37.84 21.99 Food Export (Mt) 96.05 156.81 29.01 281.87 Developing Food Import (Mt) 97.75 195.84 10.26 303.86 Export - Import (Mt) - 1.7 - 39.04 18.75 - 21.99 132 Table S 4 . 3 . The quantity of agricultural area saved (km 2 ) due to food imports. Year Food Importers Food Exporters HC NHC, Developed NHC, Developing Total Total agricultural area HHC, Developed HHC, Developing LHC, Developed LHC, Developing 2000 HHC, Developed 4,673 5,616 37,847 17,101 15,702 1,518 82,457 869,410 HHC, Developing 4,261 30,590 220,232 44,496 39,997 3,270 342,846 4,479,316 LHC, Developed 30,135 129,833 313,423 216,360 470,589 15,625 1,175,964 12,696,596 LHC, Developing 11,285 110,620 219,536 195,603 127,711 8,132 672,888 23,110,023 NHC, Developed 8,171 9,638 34,090 14,142 20,186 1,466 87,693 1,365,563 NHC, Developing 3,655 6,325 46,512 121,634 17,623 3,516 199,264 4,503,655 2015 HHC, Developed 7,642 13,362 33,726 29,595 20,468 16,766 121,558 822,858 HHC, Developing 6,940 127,651 264,281 108,924 43,986 72,309 624,090 4,584,203 LHC, Developed 60,598 227,096 521,931 562,989 334,566 249,063 1,956,244 12,104,780 LHC, Developing 13,999 131,471 252,578 408,302 102,763 39,972 949,085 22,857,369 NHC, Developed 11,777 16,167 36,956 21,896 40,557 8,837 136,189 1,336,014 NHC, Developing 7,320 36,897 46,679 159,617 14,337 7,709 272,560 4,520,668 133 Table S 4 . 3 Year Food Importers Food Exporters HC NHC, Developed NHC, Developing Total Total agricultural area HHC, Developed HHC, Developing LHC, Developed LHC, Developing 2000 - 2015 HHC, Developed 6,082 7,903 34,650 24,339 16,270 7,933 97,177 869,410 HHC, Developing 6,005 57,936 202,257 89,437 31,090 18,943 405,668 4,479,316 LHC, Developed 41,103 152,669 362,588 410,813 231,437 213,194 1,411,804 12,696,596 LHC, Developing 10,954 120,458 192,947 259,794 66,367 19,602 670,121 22,907,336 NHC, Developed 9,694 13,464 34,875 21,140 29,522 4,003 112,698 1,365,563 NHC, Developing 6,075 14,651 47,734 109,958 11,456 4,066 193,941 4,503,655 134 Table S 4 . 4 . The percentages of population, food production, and food trade among high - hotspot countries (HHC), low - hotspot countries (LHC), and non - hotspot countries (NHC), with each group subdivided into developed and developing countries, for 2000, 2015, and 2000 2015 (average annual). Year Variables HC NHC, Developed NHC, Developi ng HHC, Developed HHC, Developing LHC, Developed LHC, Developing 2000 Population (1000 persons) 268,720 928,744 545,837 3,588,769 317,928 177,626 Population percentages 4.6% 15.9% 9.4% 61.6% 5.5% 3.1% Food Production (Mt) 655.0 1,704.0 2,852.9 6,832.6 1,171.5 433.5 Food Production percentages 4.8% 12.5% 20.9% 50.1% 8.6% 3.2% Food Export (Mt) 22.4 57.9 184.5 96.0 88.9 6.7 Food Export percentages 4.9% 12.7% 40.4% 21.0% 19.5% 1.5% Food Export/Food Production percentages 3.4% 3.4% 6.5% 1.4% 7.6% 1.6% Food Import (Mt) 86.2 73.6 63.4 114.3 112.8 6.1 Food Import percentages 18.9% 16.1% 13.9% 25.0% 24.7% 1.3% 2015 Population (1000 persons) 280,722 1,156,289 605,638 4,366,135 347,337 220,028 Population percentages 4.0% 16.6% 8.7% 62.6% 5.0% 3.2% Food Production (Mt) 524.1 2,330.9 2,952.2 9,596.2 1,023.0 638.5 Food Production percentages 3.1% 13.7% 17.3% 56.2% 6.0% 3.7% Food Export (Mt) 31.7 139.2 248.7 219.6 128.9 61.6 Food Export percentages 3.8% 16.8% 30.0% 26.5% 15.5% 7.4% 135 Year Variables HC NHC, Developed NHC, Developing HHC, Developed HHC, Developing LHC, Developed LHC, Developing 2015 Food Export/Food Production percentages 6.1% 6.0% 8.4% 2.3% 12.6% 9.7% Food Import (Mt) 100.3 138.0 87.9 332.1 156.3 15.1 Food Import percentages 12.1% 16.6% 10.6% 40.0% 18.8% 1.8% 2000 - 2015 Population (1000 persons) 272,292 1,040,098 574,445 3,973,213 331,919 196,526 Population percentages 4.3% 16.3% 9.0% 62.1% 5.2% 3.1% Food Production (Mt) 616.4 2,254.1 3,052.6 8,747.8 1,144.0 603.3 Food Production percentages 3.8% 13.7% 18.6% 53.3% 7.0% 3.7% Food Export (Mt) 27.9 96.1 203.9 156.8 99.8 29.0 Food Export percentages 4.6% 15.7% 33.2% 25.6% 16.3% 4.7% Food Export/Food Production Percentages 4.5% 4.3% 6.7% 1.8% 8.7% 4.8% Food Import (Mt) 94.9 97.8 77.1 195.8 137.7 10.3 Food Import percentages 15.5% 15.9% 12.6% 31.9% 22.4% 1.7% 136 APPENDIX D SUPPORTING INFORMATION FOR CHAPTER 5 Figure S 5 . 1 . The locations of cities and watersheds in (A) freshwater source watersheds, (B) flood watersheds, and (C) hydropower watersheds. Red indicates cities, and blue indicates watersheds. 137 Table S 5 . 1 . Descriptions of dependent and independent variables. Category Variable Dataset Unit of measure Time period References Link System boundary Urban extents Global Administrative database (GADM) polygon 2018 Global Administrative Areas (2018) http://gadm.org Global urban extents polygon mid - 2000 Schneider et al. (2009) https://www.naturalearthdata.c om/downloads/10m - cultural - vectors/10m - urban - area/ Urban extents in the USA polygon 2016 United States Census Bureau (2017) https://www.cens us.gov/geo/ maps - data/data/cbf/cbf_ua.html Watershed boundaries HydroSHEDS polygon mid - 2000 Lehner et al. (2008) https://www.hydrosheds.org/p age/hydrobasins Flow River length The City Water Map v2.2 point and polygon 2016 McDonald et al. (2014) http://doi.org/10.5063/F1J67D WR HydroSHEDS 30 arc - sec mid - 2000 Lehner et al. (2008) http://www.hydrosheds.org/ Power lines OpenStreetMap power networks line 2000s OpenStreetMap https://www.openstreetmap.or g Freshwater ES Freshwater availability Global Water Modeling, ISMIP m 3 /s, 30 arc - min 2001 - 2010 Veldkamp et al. (2017) https://www.isimip.org/output data Sediment flow Modeled Suspended Sediment in Global Rivers kg/m 3 , 6 arc - min 2000 - 2010 Cohen et al. (2014) http://sdml.ua.edu/datasets - 2 Hydropower production A Global Database of Power Plants MW, point 2000 - 2016 Byers et al. (2018) http://datasets.wri.org/dataset/ globalpowerplantdatabase Flood risk Flood Hazard Maps at Global Scale meter, 30 arc - sec 2000s Dottori et al. (2016) http://data.jrc.ec.europa.eu/col lection/floods Watershed Protected area ProtectedPlanet polygon - 2016 IUCN and UNEP - WCMC (2017) https://www.protectedplanet.n et Forest Global 1 - km Consensus Land Cover proportion (0 - 100), 30 arc - sec 2000 Tuanmu and Jetz (2014) http://www.earthenv.org/landc over.html Wetland Global lakes and wetland s database (GLWD) 30 arc - sec 2000s Lehner and Döll (2004) https://www.worldwildlife.org /pages/global - lakes - and - wetlands - database Dam Global Reservoir and Dam database (GRanD) point - 2016 Lehner et al. (2011) http://globaldamwatch.org/gra nd/ 138 Category Variable Dataset Unit of measure Time period References Link Watershed Irrigation areas The Global Food Security - Support Analysis Data (GFSAD) class, 30 arc - sec 2010 Thenkabail et al. (2016) https://croplands.org/home Elevation and Slope Freshwater environmental variables in EarthEnv meter and degree, 30 arc - sec 2000s Domisch et al. (2015) http://www.earthenv.org/strea ms City Urban population The World Urbanization Prospects persons 2000 - 2010 UNDP (2015) https://esa.un.org/unpd/wup/C D - ROM / Urban GDP Global dataset of gridded GDP and population scenarios 2005 const. billion US$, 30 arc - min 2010 Murakami and Yamagata (2019) http://www.cger.nies.go.jp/gc p/population - and - gdp.html IWS program Investments or Payments in Watersheds Science Program binary (0,1) 2000s Romul o et al. (2018) https://doi.org/10.1038/s41467 - 018 - 06538 - x Temperature and precipitation WorldClim v2 °C and mm, 30 arc - sec 1970 - 2000 Hijmans et al. (2005) http://worldclim.org/version2 139 Table S 5 . 2 . Variance Inflation Factor (VIF) for variables used in multi - level models Variable Water Supply Sediment Flow Flood Risk Hydro - power Watershed Forest cover in PAs (%) 1.166 1.173 1.611 1.462 Wetland cover in PAs (%) 1.160 1.160 1.363 1.202 Dam density (#/100 km of river length) 1.125 1.121 1.105 1.203 Irrigation area (%) 1.088 1.096 1.138 1.149 Watershed area (km 2 ) 4.214 4.352 2.342 1.780 Urban - watershed distance (km) 3.841 4.009 2.226 1.540 Elevation (meter) 4.635 4.446 1.960 1.922 Slope (degree) 3.501 3.397 1.814 1.880 Urban IWS program (0, 1) 1.069 1.075 1.138 1.107 Urban population (1,000 persons) 1.417 1.428 1.424 1.529 Urban GDP - PPP (2005 const. billion USD) 1.519 1.539 1.695 1.815 Temperature (°C) 1.207 1.212 1.155 1.423 Precipitation (mm) 1.111 1.127 1.163 1.277 140 REFERENCES 141 REFERENCES Achite, M., & Ouillon, S. (2007). Suspended sediment transport in a semiarid watershed, Wadi Abd, Algeria (1973 1995). Journal of Hydrology, 343 (3), 187 - 202. doi:10.1016/j.jhydrol.2007.06.026 Adamowicz, W., Calderon - Etter, L., Entem, A., Fenichel, E. P., H all, J. S., Lloyd - Smith, P., . . . Stallard, R. F. (2019). Assessing ecological infrastructure investments. Proceedings of the National Academy of Sciences , 201802883. doi:10.1073/pnas.1802883116 Adams, W. M. (2014). The value of valuing nature. Science, 3 46 (6209), 549 - 551. doi:10.1126/science.1255997 Albrecht, J. N. (2013). Networking for Sustainable Tourism Towards a Research Agenda. Journal of Sustainable Tourism, 21 (5), 639 - 657. doi:10.1080/09669582.2012.721788 Alfieri, L., Cohen, S., Galantowicz, J., Schumann, G. J. P., Trigg, M. A., Zsoter, E., . . . Salamon, P. (2018). A global network for operational flood risk reduction. Environmental Science & Policy, 84 , 149 - 158. doi:10.1016/j.envsci.2018.03.014 Andrieu, C., de Freitas, N., Doucet, A., & Jordan, M. I. (2003). An Introduction to MCMC for Machine Learning. Machine Learning, 50 (1), 5 - 43. doi:10.1023/A:1020281327116 Anselin, L., Syabri, I., & Kho, Y. (2006). GeoDa: An introduction to spatial data anal ysis. Geographical analysis, 38 (1), 5 - 22. doi:10.1111/j.0016 - 7363.2005.00671.x Arbieu, U., Grünewald, C., Martín - López, B., Schleuning, M., & Böhning - Gaese, K. (2017). Large mammal diversity matters for wildlife tourism in Southern African Protected Areas: Insights for management. Ecosystem Services . doi:10.1016/j.ecoser.2017.11.006 Avraham, E. (2015). Destination Image Repair During Crisis: Attracting Tourism During the Arab Spring Uprisings. Tourism Management, 47 , 224 - 232. doi:10.1016/j.tourman.2014.10.0 03 Baggio, R. (2011). Collaboration and Cooperation in a Tourism Destination: A Network Science Approach. Current Issues in Tourism, 14 (2), 183 - 189. doi:10.1080/13683500.2010.531118 Baggio, R., & Sainaghi, R. (2016). Mapping Time Series into Networks as a Tool to Assess the Complex Dynamics of Tourism Systems. Tourism Management, 54 , 23 - 33. doi:10.1016/j.tourman.2015.10.008 Baggio, R., Scott, N., & Cooper, C. (2010). Network science: A Review Focused on Tourism. Annals of Tourism Research, 37 (3), 802 - 827. doi:10.1016/j.annals.2010.02.008 142 Bagstad, K. J., Johnson, G. W., Voigt, B., & Villa, F. (2013). Spatial dynamics of ecosystem service flows: A comprehensive approach to quantifying ac tual services. Ecosystem Services, 4 (0), 117 - 125. doi:10.1016/j.ecoser.2012.07.012 Bagstad, K. J., Villa, F., Batker, D., Harrison - Cox, J., Voigt, B., & Johnson, G. W. (2014). From theoretical to actual ecosystem services: mapping beneficiaries and spatial flows in ecosystem service assessments. Ecology and Society, 19 (2). doi:10.5751/ES - 06523 - 190264 Bailey, K. M., McCleery, R. A., Binford, M. W., & Zweig, C. (2016). Land - cover change within and around protected areas in a biodiversity hotspot. Journal of L and Use Science, 11 (2), 154 - 176. doi:10.1080/1747423X.2015.1086905 Balli, F., Balli, H. O., & Cebeci, K. (2013). Impacts of Exported Turkish Soap Operas and Visa - Free Entry on Inbound Tourism to Turkey. Tourism Management, 37 , 186 - 192. doi:10.1016/j.tourma n.2013.01.013 Balmford, A., Green, J. M. H., Anderson, M., Beresford, J., Huang, C., Naidoo, R., . . . Manica, A. (2015). Walk on the Wild Side: Estimating the Global Magnitude of Visits to Protected Areas. PLoS Biology, 13 (2), e1002074. doi:10.1371/journa l.pbio.1002074 Baral, N., Kaul, S., Heinen, J. T., & Ale, S. B. (2017). Estimating the Value of the World Heritage Site Designation: A Case Study from Sagarmatha (Mount Everest) National Park, Nepal. Journal of Sustainable Tourism, 25 (12), 1776 - 1791. doi:1 0.1080/09669582.2017.1310866 Barberá, P., Wang, N., Bonneau, R., Jost, J. T., Nagler, J., Tucker, J., & González - Bailón, S. (2015). The critical periphery in the growth of social protests. PloS One, 10 (11), e0143611. doi:10.1371/journal.pone.0143611 Bates, D., Machler, M., Bolker, B. M., & Walker, S. C. (2015). Fitting Linear Mixed - Effects Models Using lme4. Journal of Statistical Software, 67 (1), 1 - 48. Retrieved from ://WOS:000365981400001 Baudron, F., & Giller, K. E. (2014). Agriculture and nat ure: Trouble and strife? Biological Conservation, 170 , 232 - 245. doi:10.1016/j.biocon.2013.12.009 Baum, J., Cumming, G. S., & De Vos, A. (2017). Understanding Spatial Variation in the Drivers of Nature - Based Tourism and Their Influence on the Sustainability of Private Land Conservation. Ecological Economics, 140 , 225 - 234. doi:10.1016/j.ecolecon.2017.05.005 Bayliss, J., Schaafsma, M., Balmford, A., Burgess, N. D., Green, J. M. H., Madoffe, S. S., . . . Yu, D. W. (2014). The current and future value of nature - based tourism in the Eastern Arc Mountains of Tanzania. Ecosystem Services, 8 (0), 75 - 83. doi:10.1016/j.ecoser.2014.02.006 Beissinger, S. R., Ackerly, D. D., Doremus, H., & Machlis, G. E. (Eds.). (2017). Science, Conservation, and National Parks : University of Chicago Press. 143 Benckendorff, P., & Zehrer, A. (2013). A Network Analysis of Tourism Research. Annals of Tourism Research, 43 , 121 - 149. doi:10.1016/j.annals.2013.04.005 Bennett, G., & Ruef, F. (2016). Alliances for green infrastructure: State of watersh ed investment 2016. Forest Trends. Retrieved from http://foresttrends.org/releases/p/sowi2016 . Nature Geoscience, 12 (1), 7 - 21. d oi:10.1038/s41561 - 018 - 0262 - x Bhaduri, A., Bogardi, J., Siddiqi, A., Voigt, H., Vörösmarty, C., Pahl - Wostl, C., . . . Osuna, V. R. (2016). Achieving Sustainable Development Goals from a Water Perspective. Frontiers in Environmental Science, 4 (64). doi:10.3389/fenvs.2016.00064 Bhat, S. S., & Milne, S. (2008). Network Effects on Cooperation in Destination Website Development. Tourism Management, 29 (6), 1131 - 1140. doi:10.1016/j.tourman.2008.02.010 Bilotta, G. S., & Brazier, R. E. (2008). Understanding the influence of suspended solids on water quality and aquatic biota. Water Research, 42 (12), 2849 - 2861. doi:10.1016/j.watres.2008.03.018 Bilotta, G. S., Burnside, N. G., Cheek, L., Dunbar, M. J., Grove, M. K., Harrison, C., . . . Davy - Bowker, J. (2012). D eveloping environment - specific water quality guidelines for suspended particulate matter. Water Research, 46 (7), 2324 - 2332. doi:10.1016/j.watres.2012.01.055 Bingham, H. C., Juffe Bignoli, D., Lewis, E., MacSharry, B., Burgess, N. D., Visconti, P., . . . Ki ngston, N. (2019). Sixty years of tracking conservation progress using the World Database on Protected Areas. Nature Ecology & Evolution, 3 (5), 737 - 743. doi:10.1038/s41559 - 019 - 0869 - 3 BirdLife International NatureServe. (2013). Bird Species Distribution Map s of the World. Retrieved from http://datazone.birdlife.org/species/requestdis . Retrieved September 10th, 2018, from BirdLife International Blumenfeld, S., Lu, C., Christophersen, T., & Coates, D. (2009). Water, wetlands and forests. a review of ecological, economic and policy linkages. Paper presented at the Secretariat of the Convention on Biological Diversity and Secretariat of the Ramsar C onvention on Wetlands, Montreal and Gland. CBD Technical Series. Bondeau, A., Smith, P. C., Zaehle, S., Schaphoff, S., Lucht, W., Cramer, W., . . . Smith, B. (2007). Modelling the role of agriculture for the 20th century global terrestrial carbon balance. Global Change Biology, 13 (3), 679 - 706. doi:10.1111/j.1365 - 2486.2006.01305.x Brennan, S. R., Schindler, D. E., Cline, T. J., Walsworth, T. E., Buck, G., & Fernandez, D. P. (2019). Shifting habitat mosaics and fish production across river basins. Science, 36 4 (6442), 783 - 786. doi:10.1126/science.aav4313 144 Bright, E. A., Rose, A. N., & Urban, M. L. (2013). LandScan 1998 - 2012. Retrieved from http://web.ornl.gov/sci/landscan/ . Retrieved September 10th, 2018, from O ak Ridge National Laboratory Brooks, T. M., Lamoreux, J. F., & Soberón, J. (2014). IPBES IPCC. Trends in Ecology & Evolution, 29 (10), 543 - 545. doi:10.1016/j.tree.2014.08.004 Buckley, R., Gretzel, U., Scott, D., Weaver, D., & Becken, S. (2015). Tourism Me gatrends. Tourism Recreation Research, 40 (1), 59 - 70. doi:10.1080/02508281.2015.1005942 Buckley, R., Zhong, L., & Ma, X. (2017). Visitors to Protected Areas in China. Biological Conservation, 209 , 83 - 88. doi:10.1016/j.biocon.2017.01.024 Burkhard, B., Kandzi ora, M., Hou, Y., & Müller, F. (2014). Ecosystem service potentials, flows and demands - concepts for spatial localisation, indication and quantification. Landscape Online, 34 (1), 1 - 32. doi:10.3097/LO.201434 Burkhard, B., Kroll, F., Nedkov, S., & Müller, F. (2012). Mapping ecosystem service supply, demand and budgets. Ecological Indicators, 21 , 17 - 29. doi:10.1016/j.ecolind.2011.06.019 Burnham, K. P., & Anderson, D. R. (2002). Model Selection and Multimodel Inference: A Practical Information - Theoretic Approach . New York, NY: Springer - Verlag New York, Inc. Byers, L., Friedrich, J., Hennig, R., Kressig, A., Li, X., Mccormick, C., & Valeri, L. M. (2018). A Global Database of Power Plants. Retrieved from https://www.wri.org/publication/global - power - plant - database . Retrieved May 25th, 2019, from World Resource Institute Campos - Soria, J. A., Inchausti - Sintes, F., & Eugenio - Martin, J. L. (2015). Understanding Tourists' Economizing Strategies During the Global Economic Crisis. Tourism Management, 48 , 164 - 173. doi:10.1016/j.tourman.2014.10.019 Cao, Y., Wang, B., Zhang, J., Wa ng, L., Pan, Y., Wang, Q., . . . Deng, G. (2016). Lake macroinvertebrate assemblages and relationship with natural environment and tourism stress in Jiuzhaigou Natural Reserve, China. Ecological Indicators, 62 , 182 - 190. doi:10.1016/j.ecolind.2015.11.023 Ca role, D., & Ignacio, R. - I. (2016). Environmental impacts of food trade via resource use and greenhouse gas emissions. Environmental Research Letters, 11 (3), 035012. Retrieved from 10.1088/1748 - 9326/11/3/035012 Carpenter, S. R., Mooney, H. A., Agard, J., Ca pistrano, D., Defries, R. S., Diaz, S., . . . Whyte, A. (2009). Science for managing ecosystem services: Beyond the Millennium Ecosystem Assessment. Proceedings of the National Academy of Sciences of the United States of America, 106 (5), 1305 - 1312. doi:10. 1073/pnas.0808772106 145 Casanueva, C. b., Gallego, A. n., & Garc a - S nchez, M. a. - R. (2016). Social Network Analysis in Tourism. Current Issues in Tourism, 19 (12), 1190 - 1209. doi:10.1080/13683500.2014.990422 Cellini, R. (2011). Is UNESCO Recognition Effecti ve in Fostering Tourism? A Comment on Yang, Lin and Han. Tourism Management, 32 (2), 452 - 454. doi:10.1016/j.tourman.2010.01.018 Chan, K. M. A., Shaw, M. R., Cameron, D. R., Underwood, E. C., & Daily, G. C. (2006). Conservation planning for ecosystem service s. PLoS Biology, 4 (11), e379. doi:10.1371/journal.pbio.0040379 Chaudhary, A., & Kastner, T. (2016). Land use biodiversity impacts embodied in international food trade. Global Environmental Change, 38 (Supplement C), 195 - 204. doi:10.1016/j.gloenvcha.2016.03. 013 Chen, C., Park, T., Wang, X., Piao, S., Xu, B., Chaturvedi, R. K., . . . Myneni, R. B. (2019). China and India lead in greening of the world through land - use management. Nature Sustainability, 2 (2), 122 - 129. doi:10.1038/s41893 - 019 - 0220 - 7 Chen, W., & Ol den, J. D. (2017). Designing flows to resolve human and environmental water needs in a dam - regulated river. Nature Communications, 8 (1), 2158. doi:10.1038/s41467 - 017 - 02226 - 4 Cheng, K. M. (2012). Tourism Demand in Hong Kong: Income, Prices, and Visa Restric tions. Current Issues in Tourism, 15 (3), 167 - 181. doi:10.1080/13683500.2011.569011 Chung, M. G., Dietz, T., & Liu, J. (2018a). Global Relationships between Biodiversity and Nature - Based Tourism in Protected Areas. Ecosystem Services, 34 , 11 - 23. doi:10.1016 /j.ecoser.2018.09.004 Chung, M. G., Herzberger, A., Frank, K. A., & Liu, J. (2019). International Tourism Dynamics in a Globalized World: A Social Network Analysis Approach. Journal of Travel Research, 59 (3), 387 - 403. doi:10.1177/0047287519844834 Chung, M. G., & Kang, H. (2013). A Review of Ecosystem Service Studies: Concept, Approach and Future Work in Korea. Journal of Ecology and Environment, 36 (1), 1 - 9. doi:10.5141/ecoenv.2013.001 Chung, M. G., Kang, H., & Choi, S. - U. (2015). Assessment of Coastal Ecosy stem Services for Conservation Strategies in South Korea. PloS One, 10 (7), e0133856. doi:10.1371/journal.pone.0133856 Chung, M. G., & Liu, J. (2019). Telecoupled impacts of livestock trade on non - communicable diseases. Globalization and Health, 15 (1), 43. doi:10.1186/s12992 - 019 - 0481 - y 146 Chung, M. G., Pan, T., Zou, X., & Liu, J. (2018b). Complex interrelationships between ecosystem services supply and tourism demand: General framework and evidence from the origin of three Asian rivers. Sustainability, 10 (12), 4576. doi:10.3390/su10124576 Clements, H. S., & Cumming, G. S. (2017). Manager strategies and user demands: Determinants of cultural ecosystem service bundles on private protected areas. Ecosystem Services . doi:10.1016/j.ecoser.2017.02.026 Cohen, S., Kettner, A. J., & Syvitski, J. P. M. (2014). Global suspended sediment and water discharge dynamics between 1960 and 2010: Continental trends and intra - basin sensitivity. Global and Planetary Change, 115 , 44 - 58. doi:10.1016/j.gloplacha.2014.01.011 Crist, E., Mora, C., & Engelman, R. (2017) . The interaction of human population, food production, and biodiversity protection. Science, 356 (6335), 260 - 264. doi:10.1126/science.aal2011 Cuccia, T., Guccio, C., & Rizzo, I. (2016). The Effects of UNESCO World Heritage List Inscription on Tourism Destinations Performance in Italian Regions. Economic Modelling, 53 , 494 - 508. doi:10.1016/j.econmod.2015.10.049 Cuccia, T., Guccio, C., & Rizzo, I. (2017). UNESCO Sites and Performance Trend of Italian Regional Tourism Destinations: A Two - Stage DEA Window Analysis with Spatial Interaction. Tourism Economics, 23 (2), 316 - 342. doi:10.1177/1354816616656266 Dalin, C., Konar, M., Hanasaki, N., Rinaldo, A., & Rodriguez - Iturbe, I. (2012). Evolution of the global virtual water trade network. Proceedings of the National Academy of Sciences, 109 (16), 5989 - 5994. doi:10.1073/pnas.1203176109 Dalin, C., Wada, Y., Kastner, T., & Puma, M. J. (2017). Groundwater deplet ion embedded in international food trade. Nature, 543 (7647), 700 - 704. doi:10.1038/nature21403 Daniel, T. C., Muhar, A., Arnberger, A., Aznar, O., Boyd, J. W., Chan, K. M. A., . . . von der Dunk, A. (2012). Contributions of cultural services to the ecosyste m services agenda. Proceedings of the National Academy of Sciences, 109 (23), 8812 - 8819. doi:10.1073/pnas.1114773109 Das, M., & Chatterjee, B. (2015). Ecotourism: A panacea or a predicament? Tourism Management Perspectives, 14 , 3 - 16. doi:10.1016/j.tmp.2015. 01.002 De Vita, G., & Kyaw, K. S. (2013). Role of the Exchange Rate in Tourism Demand. Annals of Tourism Research, 43 , 624 - 627. doi:10.1016/j.annals.2013.07.011 Deardorff, A. V. (2007). Determinants of bilateral trade: Does gravity work in a Neoclassical W orld? In J. A. Frankel (Ed.), The Regionalization of the World Economy : University of Chicago Press. 147 DeFries, R. (2017). The Tangled Web of People, Landscapes, and Protected Areas. In S. R. Beissinger, D. D. Ackerly, H. Doremus, & G. E. Machlis (Eds.), Sci ence, Conservation, and National Parks (pp. pp.227 - 246). Chicago: University of Chicago Press. DeFries, R. S., Rudel, T., Uriarte, M., & Hansen, M. (2010). Deforestation driven by urban population growth and agricultural trade in the twenty - first century. Nature Geoscience, 3 (3), 178 - 181. doi:10.1038/ngeo756 Deguignet, M., Arnell, A., Juffe - Bignoli, D., Shi, Y., Bingham, H., MacSharry, B., & Kingston, N. (2017). Measuring the extent of overlaps in protected area designations. PloS One, 12 (11), e0188681. doi :10.1371/journal.pone.0188681 Delzeit, R., Zabel, F., Meyer, C., & Václavík, T. (2017). Addressing future trade - offs between biodiversity and cropland expansion to improve food security. Regional Environmental Change, 17 (5), 1429 - 1441. doi:10.1007/s10113 - 0 16 - 0927 - 1 Díaz, S., Settele, J., Brondízio, E. S., Ngo, H. T., Agard, J., Arneth, A., . . . Zayas, C. N. (2019). Pervasive human - driven decline of life on Earth points to the need for transformative change. Science, 366 (6471), eaax3100. doi:10.1126/science .aax3100 Diefendorf, S., Barbaria, L., Roberts, N., Fox, F., & Bovarnick, A. (2012). International Guidebook of Environmental Finance Tool. A Sectoral Approach: Protected Areas, Sustainable Forests, Sustainable Agriculture and Pro - Poor Energy . Retrieved fr om New York, USA. Dietz, T. (2017a). Drivers of human stress on the environment in the twenty - first century. Annual Review of Environment and Resources, 42 (1), 189 - 213. doi:10.1146/annurev - environ - 110615 - 085440 Dietz, T. (2017b). Science, Values, and Conf lict in the National Parks. In S. R. Beissinger, D. B. Ackerly, H. Doremus, & G. E. Machlis (Eds.), Science, Conservation, and National Parks (pp. 247 - 274). Chicago: University of Chicago Press. Dinerstein, E., Vynne, C., Sala, E., Joshi, A. R., Fernando, S., Lovejoy, T. E., . . . Wikramanayake, E. (2019). A Global Deal For Nature: Guiding principles, milestones, and targets. Science Advances, 5 (4), eaaw2869. doi:10.1126/sciadv.aaw2869 Dirmeyer, P. A., Gao, X., Zhao, M., Guo, Z., Oki, T., & Hanasaki, N. (20 06). GSWP - 2: Multimodel Analysis and Implications for Our Perception of the Land Surface. Bulletin of the American Meteorological Society, 87 (10), 1381 - 1398. doi:10.1175/bams - 87 - 10 - 1381 Dodds, W. K., Perkin, J. S., & Gerken, J. E. (2013). Human Impact on F reshwater Ecosystem Services: A Global Perspective. Environmental Science & Technology, 47 (16), 9061 - 9068. doi:10.1021/es4021052 148 Dogru, T., Sirakaya - Turk, E., & Crouch, G. I. (2017). Remodeling International Tourism Demand: Old Theory and New Evidence. Tou rism Management, 60 , 47 - 55. doi:10.1016/j.tourman.2016.11.010 Domisch, S., Amatulli, G., & Jetz, W. (2015). Near - global freshwater - specific environmental Scientific Data, 2 , 150073. doi:10.1038/sdata. 2015.73 Dottori, F., Salamon, P., Bianchi, A., Alfieri, L., Hirpa, F. A., & Feyen, L. (2016). Development and evaluation of a framework for global flood hazard mapping. Advances in Water Resources, 94 , 87 - 102. doi:10.1016/j.advwatres.2016.05.002 Dubois, G., Bastin, L., Bertzky, B., Mandrici, A., Conti, M., Saura, S., . . . Graziano, M. (2016). Integrating multiple spatial datasets to assess protected areas: Lessons learnt from the Digital Observatory for Protected Areas (DOPA). ISPRS International Journal of Geo - Information, 5 (12), 242. doi:10.3390/ijgi5120242 Dudley, N. (2008). Guidelines for applying protected area management categories : IUCN. Dwyer, L. (2015). Globalization of Tourism: Drivers and Outcomes. Tourism Recreation Research, 40 (3), 32 6 - 339. doi:10.1080/02508281.2015.1075723 Edwards, D. P., Tobias, J. A., Sheil, D., Meijaard, E., & Laurance, W. F. (2014). Maintaining ecosystem function and services in logged tropical forests. Trends in Ecology & Evolution, 29 (9), 511 - 520. doi:10.1016/j.tree.2014.07.003 Ehrlich, P. R., & Harte, J. (2015). Food security requires a new revolution. International Journal of Environmental Studies, 72 (6), 908 - 920. doi:10.1080/00207233.2015.1067468 Eilat, Y., & Einav, L. (2004). Determinants of International Tourism: A Three - Dimensional Panel Data Anal ysis. Applied Economics, 36 (12), 1315 - 1327. doi:10.1080/000368404000180897 EROS Data Center. (2015). EarthExplorer. Retrieved from http://earthexplorer.usgs.gov . Retrieved September 10th, 2018, from United Sta tes Geological Survey ESRI. (2015). ArcGIS Desktop: Release 10.3.1. Redlands, CA: Environmental Systems Research Institution. Retrieved from http://desktop.arcgis.com Ferraro, P. J., Hanauer, M. M., Miteva, D. A., Nelson, J. L., Pattanayak, S. K., Nolte, C., & Sims, K. R. E. (2015). Estimating the impacts of conservation on ecosystem services and poverty by integrating modeling and evaluation. Proceedings of the National Academy of Sciences, 112 (24), 7420 - 7425. doi: 10.1073/pnas.1406487112 Fleischer, A., Tchetchik, A., Bar - Nahum, Z., & Talev, E. (2018). Is agriculture important to agritourism? The agritourism attraction market in Israel. European Review of Agricultural Economics . doi:10.1093/erae/jbx039 149 Foley, J. A., Ramankutty, N., Brauman, K. A., Cassidy, E. S., Gerber, J. S., Johnston, M., . . . Zaks, D. P. M. (2011). Solutions for a cultivated planet. Nature, 478 (7369), 337 - 342. doi:10.1038/nature10452 Frank, K. A. (1995). Identifying Cohesive Subgroups. Social Net works, 17 (1), 27 - 56. doi:10.1016/0378 - 8733(94)00247 - 8 Frank, K. A. (1996). Mapping Interactions within and between Cohesive Subgroups. Social Networks, 18 (2), 93 - 119. doi:10.1016/0378 - 8733(95)00257 - X Frank, K. A., Maroulis, S. J., Duong, M. Q., & Kelcey, B . M. (2013). What would it take to inferences. Educational Evaluation and Policy Analysis, 35 (4), 437 - 460. doi:10.3102/0162373713493129 Galipaud, M., Gillingham, M. A. F. , David, M., & Dechaume - Moncharmont, F. - X. (2014). Ecologists overestimate the importance of predictor variables in model averaging: a plea for cautious interpretations. Methods in Ecology and Evolution, 5 (10), 983 - 991. doi:10.1111/2041 - 210X.12251 Gamerman , D., & Lopes, H. F. (2006). Markov chain Monte Carlo: Stochastic Simulation for Bayesian Inference . New York: Chapman and Hall/CRC. Garrick, D. E., Hall, J. W., Dobson, A., Damania, R., Grafton, R. Q., Hope, R., . . . Money, A. (2017). Valuing water for s ustainable development. Science, 358 (6366), 1003 - 1005. doi:10.1126/science.aao4942 Gartner, T., Mulligan, J., Schmidt, R., & Gunn, J. (2013). Natural infrastructure. World Resource Institute, 56 , 18. Ghermandi, A., & Nunes, P. A. L. D. (2013). A global ma p of coastal recreation values: Results from a spatially explicit meta - analysis. Ecological Economics, 86 (0), 1 - 15. doi:10.1016/j.ecolecon.2012.11.006 Glaesser, D., Kester, J., Paulose, H., Alizadeh, A., & Valentin, B. (2017). Global Travel Patterns: An Ov erview. Journal of Travel Medicine, 24 (4), 1 - 5. doi:10.1093/jtm/tax007 Global Administrative Areas. (2018). GADM database of Global Administrative Areas, v.3.6. Retrieved from https://gadm.org . Retrieved June 25th, 2019 Godfray, H. C. J., Beddington, J. R., Crute, I. R., Haddad, L., Lawrence, D., Muir, J. F., . . . Toulmin, C. (2010). Food security: The challenge of feeding 9 billion people. Science, 327 (5967), 812 - 818. doi:10.1126/science.1185383 Gössling, S., & Peeters, P. (2015). Assessing Tourism's Global Environmental Impact 1900 2050. Jo urnal of Sustainable Tourism, 23 (5), 639 - 659. doi:10.1080/09669582.2015.1008500 150 Graves, R. A., Pearson, S. M., & Turner, M. G. (2017). Species richness alone does not predict cultural ecosystem service value. Proceedings of the National Academy of Sciences . doi:10.1073/pnas.1701370114 Green, J. M. H., Croft, S. A., Durán, A. P., Balmford, A. P., Burgess, N. D., Fick, S., . . . West, C. D. (2019). Linking global drivers of agricultural trade to on - the - ground impacts on biodiversity. Proceedings of the Nation al Academy of Sciences , 201905618. doi:10.1073/pnas.1905618116 Green, P. A., Vörösmarty, C. J., Harrison, I., Farrell, T., Sáenz, L., & Fekete, B. M. (2015). Freshwater ecosystem services supporting humans: Pivoting from water crisis to water solutions. Gl obal Environmental Change, 34 , 108 - 118. doi:10.1016/j.gloenvcha.2015.06.007 Grill, G., Bernhard, L., Alexander, E. L., Graham, K. M., Christiane, Z., & Catherine Reidy, L. (2015). An index - based framework for assessing patterns and trends in river fragment ation and flow regulation by global dams at multiple scales. Environmental Research Letters, 10 (1), 015001. doi:10.1088/1748 - 9326/10/1/015001 Grill, G., Lehner, B., Thieme, M., Geenen, B., Tickner, D., Antonelli, F., . . . Zarfl, C. (2019). Mapping the wor - flowing rivers. Nature, 569 (7755), 215 - 221. doi:10.1038/s41586 - 019 - 1111 - 9 Grueber, C. E., Nakagawa, S., Laws, R. J., & Jamieson, I. G. (2011). Multimodel inference in ecology and evolution: challenges and solutions. Journal of Evolutionary Biolog y, 24 (4), 699 - 711. doi:10.1111/j.1420 - 9101.2010.02210.x Guo, Z., Zhang, L., & Li, Y. (2010). Increased dependence of humans on ecosystem services and biodiversity. PloS One, 5 (10), e13113. doi:10.1371/journal.pone.0013113 Güzeller, C. O., & Çeliker, N. (20 18). Bibliometric Analysis of Tourism Research for the Period 2007 - 2016. Advances in Hospitality and Tourism Research (AHTR), 6 (1), 1 - 22. Hall, C. M. (2010). Crisis Events in Tourism: Subjects of Crisis in Tourism. Current Issues in Tourism, 13 (5), 401 - 417. doi:10.1080/13683500.2010.491900 Hanasaki, N., Kanae, S., Oki, T., Masuda, K., Motoya, K., Shirakawa, N., . . . Tanaka, K. (2008a). An i ntegrated model for the assessment of global water resources Part 1: Model description and input meteorological forcing. Hydrol. Earth Syst. Sci., 12 (4), 1007 - 1025. doi:10.5194/hess - 12 - 1007 - 2008 Hanasaki, N., Kanae, S., Oki, T., Masuda, K., Motoya, K., S hirakawa, N., . . . Tanaka, K. (2008b). An integrated model for the assessment of global water resources Part 2: Applications and assessments. Hydrol. Earth Syst. Sci., 12 (4), 1027 - 1037. doi:10.5194/hess - 12 - 1027 - 2008 151 Haritashya, U. K., Singh, P., Kumar, N., & Gupta, R. P. (2006). Suspended sediment from the Gangotri Glacier: Quantification, variability and associations with discharge and air temperature. Journal of Hydrology, 321 (1), 116 - 130. doi:10.1016/j.jhydrol.2005.07.037 Harrison, I. J., Green, P. A. , Farrell, T. A., Juffe - Bignoli, D., Sáenz, L., & Vörösmarty, C. J. (2016). Protected areas and freshwater provisioning: a global assessment of freshwater provision, threats and management strategies to support human water security. Aquatic Conservation: M arine and Freshwater Ecosystems, 26 (S1), 103 - 120. doi:10.1002/aqc.2652 Hausmann, A., Slotow, R., Fraser, I., & Di Minin, E. (2017a). Ecotourism marketing alternative to charismatic megafauna can also support biodiversity conservation. Animal Conservation, 20 (1), 91 - 100. doi:10.1111/acv.12292 Hausmann, A., Toivonen, T., Heikinheimo, V., Tenkanen, H., Slotow, R., & Di Minin, E. (2017b). Social media reveal that charismatic species are not the main attractor of ecotourists to sub - Saharan protected areas. Scien tific Reports, 7 (1), 763. doi:10.1038/s41598 - 017 - 00858 - 6 Higham, J., & Miller, G. (2018). Transforming Societies and Transforming Tourism: Sustainable Tourism in Times of Change. Journal of Sustainable Tourism, 26 (1), 1 - 8. doi:10.1080/09669582.2018.1407519 Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A. (2005). Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology, 25 (15), 1965 - 1978. doi:10.1002/joc.1276 Hirabayashi, Y., Mahendran, R., Koirala, S., Konoshima, L., Yamazaki, D., Watanabe, S., . . . Kanae, S. (2013). Global flood risk under climate change. Nature Climate Change, 3 (9), 816 - 821. doi:10.1038/nclimate1911 Hjalager, A. - M., & Johansen, P. H. (2013) . Food tourism in protected areas sustainability for producers, the environment and tourism? Journal of Sustainable Tourism, 21 (3), 417 - 433. doi:10.1080/09669582.2012.708041 Hoff, P. D. (2005). Bilinear mixed - effects models for dyadic data. Journal of th e American Statistical Association, 100 (469), 286 - 295. doi:10.1198/016214504000001015 Hoff, P. D. (2015). Multilinear Tensor Regression for Longitudinal Relational Data. The Annals of Applied Statistics, 9 (3), 1169 - 1193. doi:10.1214/15 - AOAS839 Huber - Stearn s, H. R., Goldstein, J. H., Cheng, A. S., & Toombs, T. P. (2015). Institutional analysis of payments for watershed services in the western United States. Ecosystem Services, 16 , 83 - 93. doi:10.1016/j.ecoser.2015.10.009 Issa, I. A., & Altinay, L. (2006). Imp acts of Political Instability on Tourism Planning and Development: The Case of Lebanon. Tourism Economics, 12 (3), 361 - 381. doi:10.5367/000000006778493664 152 IUCN. (2014). The IUCN Red List of Threatened Species. Retrieved from http://www.iucnredlist.org . Retrieved September 10th, 2018, from IUCN IUCN and UNEP - WCMC. (2017). The World Database on Protected Areas (WDPA). Retrieved from http://www.protectedplanet.net . Retri eved March 14th, 2019, from UNEP - WCMC Jenkins, C. N., Pimm, S. L., & Joppa, L. N. (2013). Global patterns of terrestrial vertebrate diversity and conservation. Proceedings of the National Academy of Sciences, 110 (28), E2602 - E2610. doi:10.1073/pnas.13022511 10 Jiang, Y., Ritchie, B. W., & Benckendorff, P. (2017). Bibliometric Visualisation: An Application in Tourism Crisis and Disaster Management Research. Current Issues in Tourism , 1 - 33. doi:10.1080/13683500.2017.1408574 Jie, G., Carla, B., & Corinne, V. (20 13). Agricultural landscape preferences: Implications for agritourism development. Journal of Travel Research, 53 (3), 366 - 379. doi:10.1177/0047287513496471 Jin, C., Cheng, J., & Xu, J. (2017). Using User - Generated Content to Explore the Temporal Heterogene ity in Tourist Mobility. Journal of Travel Research, 57 (6), 779 - 791. doi:10.1177/0047287517714906 Jones, K. R., Venter, O., Fuller, R. A., Allan, J. R., Maxwell, S. L., Negret, P. J., & Watson, J. E. M. (2018). One - third of global protected land is under i ntense human pressure. Science, 360 (6390), 788 - 791. doi:10.1126/science.aap9565 Jones, P., Hillier, D., & Comfort, D. (2016). The Sustainable Development Goals and the Tourism and Hospitality Industry. Athens Journal of Tourism, 4 (1), 1 - 11. Joppa, L. N., Proceedings of the National Academy of Sciences, 105 (18), 6673 - 6678. doi:10.1073/pnas.0802471105 Joppa, L. N., & Pfaff, A. (2009). High and Far: Biases in the Location of Protecte d Areas. PloS One, 4 (12), e8273. doi:10.1371/journal.pone.0008273 Joppa, L. N., & Pfaff, A. (2010). Global protected area impacts. Proceedings of the Royal Society of London B: Biological Sciences, 278 (1712), 1633 - 1638. doi:10.1098/rspb.2010.1713 Karanth, K. K., & DeFries, R. (2011). Nature - based tourism in Indian protected areas: New challenges for park management. Conservation Letters, 4 (2), 137 - 149. doi:10.1111/j.1755 - 263X.2010.00154.x Karp, D. S., Mendenhall, C. D., Callaway, E., Frishkoff, L. O., Karei va, P. M., Ehrlich, P. R., & Daily, G. C. (2015). Confronting and resolving competing values behind conservation objectives. Proceedings of the National Academy of Sciences, 112 (35), 11132 - 11137. doi:10.1073/pnas.1504788112 153 Kastner, T., Kastner, M., & Nonhebel, S. (2011). Tracing distant environmental impacts of agricultural products from a consumer perspective. Ecological Economics, 70 (6), 1032 - 1040. doi:10.1016/j.ecolecon.2011.01.012 Khadaroo, J., & Seetanah, B. (2008). The Role of Transport Infrastructure in International Tourism Development: A Gravity Model Approach. Tourism Management, 29 (5), 831 - 840. doi:10.1016/j.tourman.2007.09.005 Khan, A., & Ruiz Estrada, M. A. (2016). The Effects of Terrorism on Economic Performance : The Case of Islamic State in Iraq and Syria (ISIS). Quality & Quantity, 50 (4), 1645 - 1661. doi:10.1007/s11135 - 015 - 0226 - 9 Kimura, F., & Lee, H. - H. (2006). The Gravity Equation in International Trade in Services. Review of World Economics, 142 (1), 92 - 121. d oi:10.1007/s10290 - 006 - 0058 - 8 Kovács, E., Kelemen, E., Kalóczkai, Á., Margóczi, K., Pataki, G., Gébert, J., . . . Mihók, B. (2015). Understanding the links between ecosystem service trade - offs and conflicts in protected areas. Ecosystem Services, 12 (Supplem ent C), 117 - 127. doi:10.1016/j.ecoser.2014.09.012 Kremen, C., & Merenlender, A. M. (2018). Landscapes that work for biodiversity and people. Science, 362 (6412), eaau6020. doi:10.1126/science.aau6020 Kronenberg, K., Fuchs, M., Salman, K., Lexhagen, M., & Hö pken, W. (2016). Economic Effects of Advertising Expenditures A Swedish Destination Study of International Tourists. Scandinavian Journal of Hospitality and Tourism, 16 (4), 352 - 374. doi:10.1080/15022250.2015.1101013 Kumari, S., Behera, M. D., & Tewari, H . R. (2010). Identification of potential ecotourism sites in West District, Sikkim using geospatial tools. Tropical Ecology, 51 (1), 75 - 85. Kuo, H. - I., Chen, C. - C., Tseng, W. - C., Ju, L. - F., & Huang, B. - W. (2008). Assessing Impacts of SARS and Avian Flu on International Tourism Demand to Asia. Tourism Management, 29 (5), 917 - 928. doi:10.1016/j.tourman.2007.10.006 Laurance, W. F., Carolina Useche, D., Rendeiro, J., Kalka, M., Bradshaw, C. J. A., Sloan, S. P., . . . Zamzani, F. (2012). Averting biodiversity col lapse in tropical forest protected areas. Nature, 489 (7415), 290 - 294. doi:10.1038/nature11318 Lee, S. - H., Choi, J. - Y., Yoo, S. - H., & Oh, Y. - G. (2013). Evaluating Spatial Centrality for Integrated Tourism Management in Rural Areas Using GIS and Network Anal ysis. Tourism Management, 34 , 14 - 24. doi:10.1016/j.tourman.2012.03.005 Lehner, B., & Döll, P. (2004). Development and validation of a global database of lakes, reservoirs and wetlands. Journal of Hydrology, 296 (1), 1 - 22. doi:10.1016/j.jhydrol.2004.03.028 154 L ehner, B., Liermann, C. R., Revenga, C., Vörösmarty, C., Fekete, B., Crouzet, P., . . . Wisser, D. (2011). High - resolution mapping of the world's reservoirs and dams for sustainable river - flow management. Frontiers in Ecology and the Environment, 9 (9), 494 - 502. doi:10.1890/100125 Lehner, B., Verdin, K., & Jarvis, A. (2008). New Global Hydrography Derived From Spaceborne Elevation Data. Eos, Transactions American Geophysical Union, 89 (10), 93 - 94. doi:10.1029/2008EO100001 Lenzen, M., Moran, D., Kanemoto, K., Foran, B., Lobefaro, L., & Geschke, A. (2012). International trade drives biodiversity threats in developing nations. Nature, 486 (7401), 109 - 112. doi:10.1038/nature11145 Lenzen, M., Sun, Y. - Y., Faturay, F., Ting, Y. - P., Geschke, A., & Malik, A. (2018). The Carbon Footprint of Global Tourism. Nature Climate Change, 8 , 522 - 528. doi:10.1038/s41558 - 018 - 0141 - x Analysis using CiteSpace. International Journal of Hospitality Manag ement, 60 , 77 - 93. doi:10.1016/j.ijhm.2016.10.006 Lim, C. (1997). Review of International Tourism Demand Models. Annals of Tourism Research, 24 (4), 835 - 849. doi:10.1016/S0160 - 7383(97)00049 - 2 Lim, C. (1999). A Meta - Analytic Review of International Tourism De mand. Journal of Travel Research, 37 (3), 273 - 284. doi:10.1177/004728759903700309 Liu, A., & Pratt, S. (2017). Tourism's Vulnerability and Resilience to Terrorism. Tourism Management, 60 , 404 - 417. doi:10.1016/j.tourman.2017.01.001 Liu, J. (2014). Forest sus tainability in China and implications for a telecoupled world. Asia & the Pacific Policy Studies, 1 (1), 230 - 250. doi:10.1002/app5.17 Liu, J. (2017). Integration across a metacoupled world. Ecology and Society, 22 (4), 29. doi:10.5751/ES - 09830 - 220429 Liu, J., Daily, G. C., Ehrlich, P. R., & Luck, G. W. (2003). Effects of household dynamics on resource consumption and biodiversity. Nature, 421 (6922), 530 - 533. doi:10.1038/nature01359 Liu, J., Dietz, T., Carpenter, S. R., Folke, C., Alberti, M., Redman, C . L., . . . Provencher, W. (2007). Complexity of coupled human and natural systems. Science, 36 (8), 639 - 649. doi:10.1126/science.1144004 Liu, J., Hull, V., Batistella, M., deFries, R., Dietz, T., Fu, F., . . . Zhu, C. (2013). Framing sustainability in a te lecoupled world. Ecology and Society, 18 (2), 26. doi:10.5751/ES - 05873 - 180226 155 Liu, J., Hull, V., Godfray, H. C. J., Tilman, D., Gleick, P., Hoff, H., . . . Li, S. (2018). Nexus approaches to global sustainable development. Nature Sustainability, 1 (9), 466 - 4 76. doi:10.1038/s41893 - 018 - 0135 - 8 Liu, J., Hull, V., Luo, J., Yang, W., Liu, W., Viña, A., . . . Zhang, H. (2015a). Multiple telecouplings and their complex interrelationships. Ecology and Society, 20 (3), 44. doi:10.5751/ES - 07868 - 200344 Liu, J., Mooney, H. , Hull, V., Davis, S. J., Gaskell, J., Hertel, T., . . . Li, S. (2015b). Systems integration for global sustainability. Science, 347 (6225), 1258832. doi:10.1126/science.1258832 Liu, J., & Yang, W. (2013). Integrated assessments of payments for ecosystem services programs. Proceedings of the National Academy of Sciences of the United States of America, 110 (41), 16297 - 16298. doi:10.1073/pnas.1316036110 Liu, J., Yang, W., & Li, S. ( 2016a). Framing ecosystem services in the telecoupled Anthropocene. Frontiers in Ecology and the Environment, 14 (1), 27 - 36. doi:10.1002/16 - 0188.1 Liu, W., Vogt, C. A., Luo, J., He, G., Frank, K. A., & Liu, J. (2012). Drivers and socioeconomic impacts of to urism participation in protected areas. PloS One, 7 (4), e35420. doi:10.1371/journal.pone.0035420 Liu, W., Vogt, C. A., Lupi, F., He, G., Ouyang, Z., & Liu, J. (2016b). Evolution of tourism in a flagship protected area of China. Journal of Sustainable Touri sm, 24 (2), 203 - 226. doi:10.1080/09669582.2015.1071380 Llorca - Vivero, R. (2008). Terrorism and International Tourism: New Evidence. Defence and Peace Economics, 19 (2), 169 - 188. doi:10.1080/10242690701453917 Lohmann, G., Albers, S., Koch, B., & Pavlovich, K. (2009). From Hub to Tourist Destination An Explorative Study of Singapore and Dubai's Aviation - Based Transformation. Journal of Air Transport Management, 15 (5), 205 - 211. doi:10.1016/j.jairtraman.2008.07.004 Lu, Y., Nakicenovic, N., Visbeck, M., & Stevan ce, A. - S. (2015). Policy: Five priorities for the UN sustainable development goals. Nature, 520 (7548), 432 - 433. doi:10.1038/520432a MacDonald, G. K., Brauman, K. A., Sun, S., Carlson, K. M., Cassidy, E. S., Gerber, J. S., & West, P. C. (2015). Rethinking a gricultural trade relationships in an era of globalization. Bioscience, 65 (3), 275 - 289. doi:10.1093/biosci/biu225 Mace, G. M. (2014). Whose conservation? Science, 345 (6204), 1558 - 1560. doi:10.1126/science.1254704 Maltby, E., & Acreman, M. C. (2011). Ecosys tem services of wetlands: pathfinder for a new paradigm. Hydrological Sciences Journal, 56 (8), 1341 - 1359. doi:10.1080/02626667.2011.631014 156 Mapulanga, A. M., & Naito, H. (2019). Effect of deforestation on access to clean drinking water. Proceedings of the N ational Academy of Sciences , 201814970. doi:10.1073/pnas.1814970116 Marques, A., Martins, I. S., Kastner, T., Plutzar, C., Theurl, M. C., Eisenmenger, N., . . . Pereira, H. M. (2019). Increasing impacts of land use on biodiversity and carbon sequestration driven by population and economic growth. Nature Ecology & Evolution, 3 (4), 628 - 637. doi:10.1038/s41559 - 019 - 0824 - 3 Marrocu, E., & Paci, R. (2013). Different Tourists to Different Destinations. Evidence from Spatial Interaction Models. Tourism Management, 3 9 , 71 - 83. doi:10.1016/j.tourman.2012.10.009 McDonald, R. I., & Shemie, D. (2014). Urban water blueprint: mapping conservation solutions to the global water challenge . Retrieved from Washington, DC. McDonald, R. I., Weber, K., Padowski, J., Flörke, M., Sch neider, C., Green, P. A., . . . Montgomery, M. (2014). Water on an urban planet: Urbanization and the reach of urban water infrastructure. Global Environmental Change, 27 , 96 - 105. doi:10.1016/j.gloenvcha.2014.04.022 McDonald, R. I., Weber, K. F., Padowski, J., Boucher, T., & Shemie, D. (2016). Estimating watershed degradation over the last century and its impact on water - treatment costs for Proceedings of the National Academy of Sciences, 113 (32), 9117 - 9122. doi:10.1073/pnas.160535 4113 McIntyre, P. B., Reidy Liermann, C. A., & Revenga, C. (2016). Linking freshwater fishery management to global food security and biodiversity conservation. Proceedings of the National Academy of Sciences . doi:10.1073/pnas.1521540113 Merinero - Rodríguez, R., & Pulido - Fernández, J. I. (2016). Analysing Relationships in Tourism: A Review. Tourism Management, 54 , 122 - 135. doi:10.1016/j.tourman.2015.10.010 Michalak, A. M. (2016). Study role of climate change in e xtreme threats to water quality. Nature, 535 (7612), 349. doi:10.1038/535349a Money, R. B. (2000). Social Networks and Referrals in International Organizational Buying of Travel Services. International Journal of Hospitality & Tourism Administration, 1 (1), 27 - 48. doi:10.1300/J149v01n01_03 Monfreda, C., Ramankutty, N., & Foley, J. A. (2008). Farming the planet: 2. Geographic distribution of crop areas, yields, physiological types, and net primary production in the year 2000. Global Biogeochemical Cycles, 22 (1 ), GB1022. doi:10.1029/2007GB002947 Moran, D., & Kanemoto, K. (2017). Identifying species threat hotspots from global supply chains. Nature Ecology & Evolution, 1 , 0023. doi:10.1038/s41559 - 016 - 0023 157 Moran, E. F., Lopez, M. C., Moore, N., Müller, N., & Hyndm an, D. W. (2018). Sustainable hydropower in the 21st century. Proceedings of the National Academy of Sciences, 115 (47), 11891 - 11898. doi:10.1073/pnas.1809426115 Morley, C., Rosselló, J., & Santana - Gallego, M. (2014). Gravity Models for Tourism Demand: Theo ry and Use. Annals of Tourism Research, 48 , 1 - 10. doi:10.1016/j.annals.2014.05.008 Müller Schmied, H., Adam, L., Eisner, S., Fink, G., Flörke, M., Kim, H., . . . Döll, P. (2016). Variations of global and continental water balance components as impacted by climate forcing uncertainty and human water use. Hydrol. Earth Syst. Sci., 20 (7), 2877 - 2898. doi:10.5194/hess - 20 - 2877 - 2016 Murakami, D., & Yamagata, Y. (2019). Estimation of Gridded Population and GDP Scenarios with Spatially Explicit Statistical Downscali ng. Sustainability, 11 (7). doi:10.3390/su11072106 Myers, N. (2003). Biodiversity Hotspots Revisited. Bioscience, 53 (10), 916 - 917. doi:10.1641/0006 - 3568(2003)053[0916:Bhr]2.0.Co;2 Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca, G. A. B., & Ke nt, J. (2000). Biodiversity hotspots for conservation priorities. Nature, 403 (6772), 853 - 858. doi:10.1038/35002501 Naidoo, R., Balmford, A., Costanza, R., Fisher, B., Green, R. E., Lehner, B., . . . Ricketts, T. H. (2008). Global mapping of ecosystem services and conservation priorities. Proceedings of the National Academy of Sciences, 105 (28), 9495 - 9500. doi:10.1073/p nas.0707823105 Nelson, A. (2008). Estimated travel time to the nearest city of 50,000 or more people in year 2000. Retrieved from http://forobs.jrc.ec.europa.eu/products/gam . Retrieved September 10th, 2018, from European Commission Neumayer, E. (2010). Visa Restrictions and Bilateral Travel. The Professional Geographer, 62 (2), 171 - 181. doi:10.1080/00330121003600835 Neuvonen, M., Pouta, E., Puustinen, J., & Sievänen, T. (2010). Visits to national p arks: Effects of park characteristics and spatial demand. Journal for Nature Conservation, 18 (3), 224 - 229. doi:10.1016/j.jnc.2009.10.003 Newsome, D., Moore, S. A., & Dowling, R. K. (2012). Natural area Tourism: Ecology, Impacts and Management (Vol. 58): Ch annel View Publications. Nilsson, M., Griggs, D., & Visbeck, M. (2016). Policy: Map the interactions between Sustainable Development Goals. Nature, 534 (7607), 320 - 322. doi:10.1038/534320a Nyaupane, G. P., & Chhetri, N. (2009). Vulnerability to climate chan ge of nature - based tourism in the Nepalese Himalayas. Tourism Geographies, 11 (1), 95 - 119. doi:10.1080/14616680802643359 158 Quality & Quantity, 41 (5), 673 - 690. doi:10.1007/s 11135 - 006 - 9018 - 6 Oldekop, J. A., Holmes, G., Harris, W. E., & Evans, K. L. (2016). A global assessment of the social and conservation outcomes of protected areas. Conservation Biology, 30 (1), 133 - 141. doi:10.1111/cobi.12568 Otte, E., & Rousseau, R. (2002). Social Network Analysis: A Powerful Strategy, Also for the Information Sciences. Journal of Information Science, 28 (6), 441 - 453. doi:10.1177/016555150202800601 Ouyang, Z., Zheng, H., Xiao, Y., Polasky, S., Liu, J., Xu, W., . . . Daily, G. C. (2016). Improvements in ecosystem services from investments in natural capital. Science, 352 (6292), 1455 - 1459. doi:10.1126/science.aaf2295 Palmer, M. A. (2010). Water resources: Beyond infrastructure. Nature, 467 (7315), 534 - 535. doi:10.1 038/467534a Palmer, M. A., Liu, J., Matthews, J. H., Mumba, M., & D'Odorico, P. (2015). Manage water in a green way. Science, 349 (6248), 584 - 585. doi:10.1126/science.aac7778 Palomo, I., Martín - López, B., Potschin, M., Haines - Young, R., & Montes, C. (2013). National Parks, buffer zones and surrounding lands: Mapping ecosystem service flows. Ecosystem Services, 4 (0), 104 - 116. doi:10.1016/j.ecoser.2012.09.001 Patuelli, R., Mussoni, M., & Candela, G. (2013). The Effects of World Heritage Sites on Domestic Touri sm: A Spatial Interaction Model for Italy. Journal of Geographical Systems, 15 (3), 369 - 402. doi:10.1007/s10109 - 013 - 0184 - 5 Patuelli, R., Mussoni, M., & Candela, G. (2014). Cultural Offer and Distance in a Spatial Interaction Model for Tourism. Economics and Business Letters, 3 (2), 96 - 108. Pe'er, G., Dicks, L. V., Visconti, P., Arlettaz, R., Báldi, A., Benton, T. G., . . . Scott, A. V. (2014). EU agricultural reform fails on biodiversity. Science, 344 (6188), 1090 - 1092. doi:10.1126/science.1253425 Peng, B., S ong, H., & Crouch, G. I. (2014). A Meta - Analysis of International Tourism Demand Forecasting and Implications for Practice. Tourism Management, 45 , 181 - 193. doi:10.1016/j.tourman.2014.04.005 Pimm, S. L., Jenkins, C. N., Abell, R., Brooks, T. M., Gittleman, J. L., Joppa, L. N., . . . Sexton, J. O. (2014). The biodiversity of species and their rates of extinction, distribution, and protection. Science, 344 (6187), 1246752. doi:10.1126/science.1246752 Pleasant, M. M., Gray, S. A., Lepczyk, C., Fernandes, A., Hu nter, N., & Ford, D. (2014). Managing cultural ecosystem services. Ecosystem Services, 8 (0), 141 - 147. doi:10.1016/j.ecoser.2014.03.006 159 Plummer, R., & Fennell, D. A. (2009). Managing protected areas for sustainable tourism: prospects for adaptive co - managem ent. Journal of Sustainable Tourism, 17 (2), 149 - 168. doi:10.1080/09669580802359301 Poff, N. L., & Schmidt, J. C. (2016). How dams can go with the flow. Science, 353 (6304), 1099 - 1100. doi:10.1126/science.aah4926 Pokhrel, Y. N., Hanasaki, N., Koirala, S., Ch o, J., Yeh, P. J. - F., Kim, H., . . . Oki, T. (2012). Incorporating Anthropogenic Water Regulation Modules into a Land Surface Model. Journal of Hydrometeorology, 13 (1), 255 - 269. doi:10.1175/jhm - d - 11 - 013.1 Pokhrel, Y. N., Koirala, S., Yeh, P. J. F., Hanasak i, N., Longuevergne, L., Kanae, S., & Oki, T. (2015). Incorporation of groundwater pumping in a global Land Surface Model with the representation of human impacts. Water Resources Research, 51 (1), 78 - 96. doi:10.1002/2014WR015602 Porkka, M., Kummu, M., Sieb ert, S., & Varis, O. (2013). From food insufficiency towards trade dependency: A historical analysis of global food availability. PloS One, 8 (12), e82714. doi:10.1371/journal.pone.0082714 Pulido - Fernández, J. I., & Merinero - Rodríguez, R. (2018). Destinatio ns' Relational Dynamic and Tourism Development. Journal of Destination Marketing & Management, 7 , 140 - 152. doi:10.1016/j.jdmm.2016.09.008 Pywell, R. F., Heard, M. S., Woodcock, B. A., Hinsley, S., Ridding, L., Nowakowski, M., & Bullock, J. M. (2015). Wildl ife - friendly farming increases crop yield: evidence for ecological intensification. Proceedings of the Royal Society of London B: Biological Sciences, 282 (1816), 20151740. doi:10.1098/rspb.2015.1740 QGIS Development Team. (2014). QGIS Geographic Informatio n System: Open Source Geospatial Foundation Project. Retrieved from http://qgis.osgeo.org R Core Team. (2017). R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.R - project.org/ Raisi, H., Baggio, R., Barratt - Pugh, L., & Willson, G. (2017). Hyperlink Network Analysis of a Tourism Destination. Journal of Travel Research, 57 (5), 671 - 686. doi:10.1177/0047287517708256 Rands, M. R. W., Adams, W. M., Bennun, L., Butchart, S. H. M., Clem ents, A., Coomes, D., . . . Vira, B. (2010). Biodiversity conservation: Challenges beyond 2010. Science, 329 (5997), 1298 - 1303. doi:10.1126/science.1189138 Redford, K. H., Huntley, B. J., Roe, D., Hammond, T., Zimsky, M., Lovejoy, T. E., . . . Cowling, R. M . (2015). Mainstreaming Biodiversity: Conservation for the Twenty - First Century. Frontiers in Ecology and Evolution, 3 , 137. doi:10.3389/fevo.2015.00137 160 Rehman Khan, S. A., Qianli, D., SongBo, W., Zaman, K., & Zhang, Y. (2017). Travel and Tourism Competiti veness Index: The Impact of Air Transportation, Railways Transportation, Travel and Transport Services on International Inbound and Outbound Tourism. Journal of Air Transport Management, 58 , 125 - 134. doi:10.1016/j.jairtraman.2016.10.006 Reinius, S. W., & F redman, P. (2007). Protected areas as attractions. Annals of Tourism Research, 34 (4), 839 - 854. doi:10.1016/j.annals.2007.03.011 Renard, D., Rhemtulla, J. M., & Bennett, E. M. (2015). Historical dynamics in ecosystem service bundles. Proceedings of the Nati onal Academy of Sciences, 112 (43), 13411 - 13416. doi:10.1073/pnas.1502565112 Richards, G. (2011). Creativity and Tourism: The State of the Art. Annals of Tourism Research, 38 (4), 1225 - 1253. doi:10.1016/j.annals.2011.07.008 Robin Abell, Nigel Asquith, Giulio Boccaletti, Leah Bremer, Emily Chapin, Andrea Erickson - Quiroz, . . . Wood, S. (2017). Beyond the Source: The Environmental, Economic and Community Benefits of Source Water Protection. Retrieved from Arlington, VA, USA. Romulo, C. L., Posner, S., Cousins, S., Hoyle Fair, J., Bennett, D. E., Huber - Stearns, H., . . . McDonald, R. I. (2018). Global state and potential scope of investments in watershed services for large cities. Nature Communications, 9 (1), 4375. doi:10.1038/s41467 - 018 - 06538 - x Roy, S. G., Uchi da, E., de Souza, S. P., Blachly, B., Fox, E., Gardner, K., . . . Hart, D. (2018). A multiscale approach to balance trade - offs among dam infrastructure, river restoration, and cost. Proceedings of the National Academy of Sciences, 115 (47), 12069. doi:10.10 73/pnas.1807437115 Russi, D., ten Brink, P., Farmer, A., Badura, T., Coates, D., Förster, J., . . . Davidson, N. (2013). The economics of ecosystems and biodiversity for water and wetlands. IEEP, London and Brussels . Saha, S., & Yap, G. (2013). The Modera tion Effects of Political Instability and Terrorism on Tourism Development. Journal of Travel Research, 53 (4), 509 - 521. doi:10.1177/0047287513496472 Sainaghi, R., & Baggio, R. (2017). Complexity Traits and Dynamics of Tourism Destinations. Tourism Manageme nt, 63 , 368 - 382. doi:10.1016/j.tourman.2017.07.004 Schägner, J. P., Maes, J., Brander, L., Paracchini, M. - L., Hartje, V., & Dubois, G. (2017). Monitoring recreation across European nature areas: A geo - database of visitor counts, a review of literature and a call for a visitor counting reporting standard. Journal of Outdoor Recreation and Tourism, 18 , 44 - 55. doi:10.1016/j.jort.2017.02.004 161 Schaphoff, S., Heyder, U., Ostberg, S., Gerten, D., Heinke, J., & Lucht, W. (2013). Contribution of permafrost soils to t he global carbon budget. Environmental Research Letters, 8 (1), 014026. doi:10.1088/1748 - 9326/8/1/014026 Scheyvens, R. (2018). Linking Tourism to the Sustainable Development Goals: A Geographical Perspective. Tourism Geographies, 20 (2), 341 - 342. doi:10.1080 /14616688.2018.1434818 Schneider, A., Friedl, M. A., & Potere, D. (2009). A new map of global urban extent from MODIS satellite data. Environmental Research Letters, 4 (4), 044003. doi:10.1088/1748 - 9326/4/4/044003 Scott, D., & Gössling, S. (2015). What Coul d The Next 40 Years Hold for Global Tourism? Tourism Recreation Research, 40 (3), 269 - 285. doi:10.1080/02508281.2015.1075739 Scott, D., Gössling, S., & de Freitas, C. R. (2008a). Preferred climates for tourism case studies from Canada, New Zealand and Swede n. Climate Research, 38 (1), 61 - 73. doi:10.3354/cr00774 Scott, N., Cooper, C., & Baggio, R. (2008b). Destination Networks: Four Australian Cases. Annals of Tourism Research, 35 (1), 169 - 188. doi:10.1016/j.annals.2007.07.004 Serna - Chavez, H. M., Schulp, C. J. E., van Bodegom, P. M., Bouten, W., Verburg, P. H., & Davidson, M. D. (2014). A quantitative framework for assessing spatial flows of ecosystem services. Ecological Indicators, 39 (0), 24 - 33. doi:10.1016/j.ecolind.2013.11. 024 Sheffield, J., Goteti, G., & Wood, E. F. (2006). Development of a 50 - Year High - Resolution Global Dataset of Meteorological Forcings for Land Surface Modeling. Journal of Climate, 19 (13), 3088 - 3111. doi:10.1175/JCLI3790.1 Siikamäki, P., Kangas, K., Paas ivaara, A., & Schroderus, S. (2015). Biodiversity attracts visitors to national parks. Biodiversity and Conservation, 24 (10), 2521 - 2534. doi:10.1007/s10531 - 015 - 0941 - 5 Smith, A., Bates, P. D., Wing, O., Sampson, C., Quinn, N., & Neal, J. (2019). New estimat es of flood exposure in developing countries using high - resolution population data. Nature Communications, 10 (1), 1814. doi:10.1038/s41467 - 019 - 09282 - y Smith, A. C., Harrison, P. A., Pérez Soba, M., Archaux, F., Blicharska, M., Egoh, B. N., . . . Wyllie de Echeverria, V. (2017). How natural capital delivers ecosystem services: A typology derived from a systematic review. Ecosystem Services, 26 , 111 - 126. doi:10.1016/j.ecoser.2017.06.006 Snijders, T. A. B. (2011). Statistical models for social networks. Annual Review of Sociology, 37 (1), 131 - 153. doi:10.1146/annurev.soc.012809.102709 162 Soligno, I., Ridolfi, L., & Laio, F. (2019). The globalization of riverine environmental resources through the food trade. Environmental Research Letters, 14 (2), 024020. doi:10.108 8/1748 - 9326/aaf93a Song, H., Dwyer, L., Li, G., & Cao, Z. (2012a). Tourism Economics Research: A Review and Assessment. Annals of Tourism Research, 39 (3), 1653 - 1682. doi:10.1016/j.annals.2012.05.023 Song, H., Gartner, W. C., & Tasci, A. D. A. (2012b). Visa Restrictions and Their Adverse Economic and Marketing Implications Evidence from China. Tourism Management, 33 (2), 397 - 412. doi:10.1016/j.tourman.2011.05.001 Song, H., Kim, J. H., & Yang, S. (2010 ). Confidence Intervals for Tourism Demand Elasticity. Annals of Tourism Research, 37 (2), 377 - 396. doi:10.1016/j.annals.2009.10.002 Song, H., & Li, G. (2008). Tourism Demand Modelling and Forecasting A Review of Recent Research. Tourism Management, 29 (2), 203 - 220. doi:10.1016/j.tourman.2007.07.016 Song, H., Li, G., & Cao, Z. (2017). Tourism and Economic Globalization: An Emerging Research Agenda. Journal of Travel Research , 0047287517734943. doi:10.1177/0047287517734943 Sönmez, S. F. (1998). Tourism, Terror ism, and Political Instability. Annals of Tourism Research, 25 (2), 416 - 456. doi:10.1016/S0160 - 7383(97)00093 - 5 Stickler, C. M., Coe, M. T., Costa, M. H., Nepstad, D. C., McGrath, D. G., Dias, L. C. P., . . . Soares - Filho, B. S. (2013). Dependence of hydropo wer energy generation on forests in the Amazon Basin at local and regional scales. Proceedings of the National Academy of Sciences, 110 (23), 9601 - 9606. doi:10.1073/pnas.1215331110 Su, Y. - W., & Lin, H. - L. (2014). Analysis of International Tourist Arrivals W orldwide: The Role of World Heritage Sites. Tourism Management, 40 , 46 - 58. doi:10.1016/j.tourman.2013.04.005 Symes, W. S., Rao, M., Mascia, M. B., & Carrasco, L. R. (2016). Why do we lose protected areas? Factors influencing protected area downgrading, dow nsizing and degazettement in the tropics and subtropics. Global Change Biology, 22 (2), 656 - 665. doi:10.1111/gcb.13089 Tallis, H., & Lubchenco, J. (2014). Working together: A call for inclusive conservation. Nature, 515 (7525), 27 - 28. doi:10.1038/515027a Tel lman, B., McDonald, R. I., Goldstein, J. H., Vogl, A. L., Flörke, M., Shemie, D., . . . Veiga, F. (2018). Opportunities for natural infrastructure to improve urban water security in Latin America. PloS One, 13 (12), e0209470. doi:10.1371/journal.pone.020947 0 163 Tessler, Z. D., Vörösmarty, C. J., Grossberg, M., Gladkova, I., Aizenman, H., Syvitski, J. P. M., & Foufoula - Georgiou, E. (2015). Profiling risk and sustainability in coastal deltas of the world. Science, 349 (6248), 638 - 643. doi:10.1126/science.aab3574 T he World Bank. (2017). World Development Indicators. Retrieved from https://data.worldbank.org . Retrieved December 30th, 2019, from The World Bank Thenkabail, P., Knox, J., Ozdogan, M., Gumma, M., Congalton, R., Wu, Z., . . . Nagler, P. (2016). NASA Making Earth System Data Records for Use in Research Environments (MEaSUREs) Global Food Security Support Analysis Data (GFSAD) Crop Dominance 2010 Globa l 1 km V001. NASA EOSDIS Land Processes DAAC . doi:10.5067/MEaSUREs/GFSAD/GFSAD1KCD.001 Tilman, D., & Clark, M. (2014). Global diets link environmental sustainability and human health. Nature, 515 (7528), 518 - 522. doi:10.1038/nature13959 Tilman, D., Clark, M ., Williams, D. R., Kimmel, K., Polasky, S., & Packer, C. (2017). Future threats to biodiversity and pathways to their prevention. Nature, 546 , 73. doi:10.1038/nature22900 Torres - Reyna, O. (2010). Getting Started in Fixed/Random Effects Models using R . Ret rieved from Tuanmu, M. - N., & Jetz, W. (2014). A global 1 - km consensus land - cover product for biodiversity and ecosystem modelling. Global Ecology and Biogeography, 23 (9), 1031 - 1045. doi:10.1111/geb.12182 Turner, W. R., Brandon, K., Brooks, T. M., Costanza , R., da Fonseca, G. A. B., & Portela, R. (2007). Global conservation of biodiversity and ecosystem services. Bioscience, 57 (10), 868 - 873. doi:10.1641/b571009 Turner, W. R., Brandon, K., Brooks, T. M., Gascon, C., Gibbs, H. K., Lawrence, K. S., . . . Selig , E. R. (2012). Global biodiversity conservation and the alleviation of poverty. Bioscience, 62 (1), 85 - 92. doi:10.1525/bio.2012.62.1.13 Tyler, D., & Dinan, C. (2001). The Role of Interested Groups in England's Emerging Tourism Policy Network. Current Issue s in Tourism, 4 (2 - 4), 210 - 252. doi:10.1080/13683500108667888 UN FAO. (2018). FAOSTAT Statistics Database. Retrieved from http://www.fao.org/faostat . FAOSTAT Retrieved December 30th, 2019, from FAO UNDP. (2015). Wo rld Urbanization Prospects: The 2014 Revision . New York: United Nations Population Division. UNEP. Post - 2020 global biodiversity framework: discussion paper adopted by the Conference of the Parties, 25 January 2019, CBD/POST2020/PREP/1/1. [cited December 3 1st, 2019]. Available from:, 164 https://www.cbd.int/doc/c/d0f3/aca0/d42fa469029f5a4d69f4da8e/post2020 - prep - 01 - 01 - en.pdf . UNESCO. (2017). World Heritage Li st. Retrieved from http://whc.unesco.org/en/list . Retrieved March 14th, 2019, from United Nations Education, Scientific, and Cultural Organization United Nations. (2015). Sustainable Development Goals. Retriev ed from https://sustainabledevelopment.un.org . Retrieved April 18th, 2019 United Nations Statistics Division. (2015). UNdata Retrieved from http://data.un.org . Retri eved December 30th, 2019, from United Nations United States Census Bureau. (2017). 2016 Cartographic Boundary File, Urban Areas for United States, 1:500,000. Retrieved from https://www.census.gov/geographies/mapping - files/time - series/geo/carto - boundary - file.2016.html . Retrieved June 15th, 2020, from United States Cens us Bureau van Beek, L. P. H., Wada, Y., & Bierkens, M. F. P. (2011). Global monthly water stress: 1. Water balance and water availability. Water Resources Research, 47 (7). doi:10.1029/2010WR009791 van der Zee, E., & Vanneste, D. (2015). Tourism Networks Unravelled; A Review of the Literature on Networks in Tourism Management Studies. Tourism Management Perspectives, 15 , 46 - 56. doi:10.1016/j.tmp.2015.03.006 van Duijn, M. A. J., Snijders, T. A. B., & Zijlstra, B. J. H. (2004). p2: A random effects model with covariates for directed graphs. Statistica Neerlandica, 58 (2), 234 - 254. doi:10.1046/j.0039 - 0402.2003.00258.x Veldkamp, T. I. E., Wada, Y., Aerts, J. C. J. H., Döll, P., Gosling, S. N., Liu, J., . . . Ward, P. J. (2017). Water scarcity hotspots travel downstream due to human interventions in the 20th and 21st century. Nature Communications, 8 , 15697. doi:10.1038/ncomms15697 Verbos, R. I., Altschuler, B., & Brownlee, M. T. J. (2017). Weather studies in outdoor recreation and nature - based tourism: A research synthesis a nd gap analysis. Leisure Sciences , 1 - 24. doi:10.1080/01490400.2017.1325794 Vercruysse, K., Grabowski, R. C., & Rickson, R. J. (2017). Suspended sediment transport dynamics in rivers: Multi - scale drivers of temporal variation. Earth - Science Reviews, 166 , 38 - 52. doi:10.1016/j.earscirev.2016.12.016 Viña, A., McConnell, W. J., Yang, H., Xu, Z., & Liu, J. (2016). Effects of conservation policy on Science Advances, 2 (3). doi:10.1126/sciadv.1500965 Visconti, P., Butchart, S. H. M., Brooks, T. M., Langhammer, P. F., Marnewick, D., Vergara, S., . . . Watson, J. E. M. (2019). Protected area targets post - 2020. Science, 364 (6437), 239 - 241. doi:10.1126/science.aav6886 165 von Bergner, N. M., & Lohmann, M. (2014). Future Challenges for Global Tourism: A Delphi Survey. Journal of Travel Research, 53 (4), 420 - 432. doi:10.1177/0047287513506292 Vorosmarty, C. J., McIntyre, P. B., Gessner, M. O., Dudgeon, D., Prusevich, A., Green, P., . . . Davies, P. M. (2010). Global threats to human water security and riv er biodiversity. Nature, 467 (7315), 555 - 561. doi:10.1038/nature09440 Vörösmarty, C. J., Rodríguez Osuna, V., Cak, A. D., Bhaduri, A., Bunn, S. E., Corsi, F., . . . Uhlenbrook, S. (2018). Ecosystem - based water security and the Sustainable Development Goals (SDGs). Ecohydrology & Hydrobiology, 18 (4), 317 - 333. doi:10.1016/j.ecohyd.2018.07.004 Wada, Y., Wisser, D., & Bierkens, M. F. P. (2014). Global modeling of withdrawal, allocation and consumptive use of surface water and groundwater resources. Earth Syst. D ynam., 5 (1), 15 - 40. doi:10.5194/esd - 5 - 15 - 2014 Wang, Y., & Xiang, Z. (2007). Toward a Theoretical Framework of Collaborative Destination Marketing. Journal of Travel Research, 46 (1), 75 - 85. doi:10.1177/0047287507302384 Wang, Y., Yu, P., Feger, K. - H., Wei, X ., Sun, G., Bonell, M., . . . Xu, L. (2011). Annual runoff and evapotranspiration of forestlands and non - forestlands in selected basins of the Loess Plateau of China. Ecohydrology, 4 (2), 277 - 287. doi:10.1002/eco.215 Ward, M. D., & Hoff, P. D. (2007). Persi stent patterns of international commerce. Journal of Peace Research, 44 (2), 157 - 175. doi:10.1177/0022343307075119 Wasserman, S., & Galaskiewicz, J. (1994). Advances in Social Network Analysis: Research in the Social and Behavioral Sciences (S. Wasserman & J. Galaskiewicz Eds.): SAGE Publications. Watson, J. E. M., Dudley, N., Segan, D. B., & Hockings, M. (2014). The performance and potential of protected areas. Nature, 515 (7525), 67 - 73. doi:10.1038/nature13947 Weedon, G. P., Balsamo, G., Bellouin, N., Gomes , S., Best, M. J., & Viterbo, P. (2014). The WFDEI meteorological forcing data set: WATCH Forcing Data methodology applied to ERA - Interim reanalysis data. Water Resources Research, 50 (9), 7505 - 7514. doi:10.1002/2014WR015638 Wei, H., Fan, W., Wang, X., Lu, N., Dong, X., Zhao, Y., . . . Zhao, Y. (2017). Integrating supply and social demand in ecosystem services assessment: A review. Ecosystem Services, 25 (Supplement C), 15 - 27. doi:10.1016/j.ecoser.2017.03.017 Wellman, B., & Frank, K. (2001). Network capital in a multilevel world: Getting support from personal communities. Social capital: Theory and research , pp.233 - 273. Westveld, A. H., & Hoff, P. D. (2011). A mixed effects model for longitudinal relational an d network data, with applications to international trade and conflict. The Annals of Applied Statistics, 5 (2A), 843 - 872. doi:10.1214/10 - AOAS403 166 Wiedmann, T., & Lenzen, M. (2018). Environmental and social footprints of international trade. Nature Geoscience , 11 (5), 314 - 321. doi:10.1038/s41561 - 018 - 0113 - 9 Willemen, L., Cottam, A. J., Drakou, E. G., & Burgess, N. D. (2015). Using social media to measure the contribution of Red List species to the nature - based tourism potential of African protected areas. PloS O ne, 10 (6), e0129785. doi:10.1371/journal.pone.0129785 Wisser, D., Fekete, B. M., Vörösmarty, C. J., & Schumann, A. H. (2010). Reconstructing 20th century global hydrography: a contribution to the Global Terrestrial Network - Hydrology (GTN - H). Hydrol. Earth Syst. Sci., 14 (1), 1 - 24. doi:10.5194/hess - 14 - 1 - 2010 Witt, S. F., & Witt, C. A. (1995). Forecasting Tourism Demand: A Review of Empirical Research. International Journal of Forecasting, 11 (3), 447 - 475. doi:10.1016/0169 - 2070(95)00591 - 7 Wittemyer, G., Elsen, P., Bean, W. T., Burton, A. C. O., & Brashares, J. S. (2008). Accelerated human population growth at protected area edges. Science, 321 (5885), 123 - 126. doi:10.1126/science.1158900 Wood, S. A., Smith, M. R., Fanzo, J., Remans, R., & DeFries, R. S. (2018). Trade and the equitability of global food nutrient distribution. Nature Sustainability, 1 (1), 34 - 37. doi:10.1038/s41893 - 017 - 0008 - 6 World Tourism Organization. (2015). Tourism at World Heritage Sites - Challenges and Opportunites, International tourism semi nar, Cesme (Izmir), Turkey, 26 March 2013 . Retrieved from Madrid, Spain. World Tourism Organization. (2016a). Methodological Notes to the Tourism Statistics Database, 2016 Edition . Retrieved from Madrid, Spain. World Tourism Organization. (2016b). UNWTO Tourism Highlights, 2016 Edition . Retrieved from Madrid, Spain. World Tourism Organization. (2016c). UNWTO/UNESCO World Conference on Tourism and Culture: Building a New Partnership Siem Reap, Cambodia, 4 - 6 Febru ray 2015 . Retrieved from Madrid, Spain. World Tourism Organization. (2016d). Visa Openness Report 2015 . Retrieved from Madrid, Spain. World Tourism Organization. (2018). Tourism for SDGs. Retrieved from http://tour ism4sdgs.org . Retrieved March 14th, 2019, from World Tourism Organization World Travel and Tourism Council. (2018). Travel and Tourism Economic Impact 2018 - World. Retrieved from https://www.wttc.org/economic - impact/country - analysis/regional - reports . Retrieved March 14th, 2019 167 WRI. (2015). Aqueduct Global Maps 2.1. Retrieved from https:/ /www.wri.org/resources/data - sets/aqueduct - global - maps - 21 - data . Retrieved September 10th, 2018, from WRI Wu, X. - X., Gu, Z. - J., Luo, H., Shi, X. - Z., & Yu, D. - S. (2014). Analyzing forest effects on runoff and sediment production using leaf area index. Journa l of Mountain Science, 11 (1), 119 - 130. doi:10.1007/s11629 - 013 - 2436 - 8 Xiao, L., Haiping, T., & Haoguang, L. (2017). A theoretical framework for researching cultural ecosystem service flows in urban agglomerations. Ecosystem Services, 28 (Part A), 95 - 104. doi :10.1016/j.ecoser.2017.09.014 Xu, Z., Chau, S. N., Chen, X., Zhang, J., Li, Y., Dietz, T., . . . Liu, J. (2020a). Assessing progress towards sustainable development over space and time. Nature, 577 (7788), 74 - 78. doi:10.1038/s41586 - 019 - 1846 - 3 Xu, Z., Li, Y., Chau, S. N., Dietz, T., Li, C., Wan, L., . . . Liu, J. (2020b). Impacts of international trade on global sustainable development. Nature Sustainability . doi:10.1038/s41893 - 020 - 0572 - z Yang, C. - H., & Lin, H. - L. ( 2011). Is UNESCO Recognition Effective in Fostering Tourism? A Comment on Yang, Lin and Han: Reply. Tourism Management, 32 (2), 455 - 456. doi:10.1016/j.tourman.2010.03.011 Yang, C. - H., Lin, H. - L., & Han, C. - C. (2010). Analysis of International Tourist Arriva ls in China: The role of World Heritage Sites. Tourism Management, 31 (6), 827 - 837. doi:10.1016/j.tourman.2009.08.008 Yang, H., Viña, A., Winkler, J. A., Chung, M. G., Dou, Y., Wang, F., . . . Liu, J. (2019). cing deforestation. Environmental Science and Pollution Research . doi:10.1007/s11356 - 019 - 05232 - 9 Zarfl, C., Lumsdon, A. E., Berlekamp, J., Tydecks, L., & Tockner, K. (2015). A global boom in hydropower dam construction. Aquatic Sciences, 77 (1), 161 - 170. do i:10.1007/s00027 - 014 - 0377 - 0 Zheng, H., Robinson, B. E., Liang, Y. - C., Polasky, S., Ma, D. - C., Wang, F. - C., . . . Daily, G. C. (2013). Benefits, costs, and livelihood implications of a regional payment for ecosystem service program. Proceedings of the Natio nal Academy of Sciences of the United States of America, 110 (41), 16681 - 16686. doi:10.1073/pnas.1312324110 Zhong, L., Buckley, R. C., Wardle, C., & Wang, L. (2015). Environmental and visitor management in a thousand protected areas in China. Biological Con servation, 181 , 219 - 225. doi:10.1016/j.biocon.2014.11.007 Ziv, G., Baran, E., Nam, S., Rodríguez - Iturbe, I., & Levin, S. A. (2012). Trading - off fish biodiversity, food security, and hydropower in the Mekong River Basin. Proceedings of the National Academy of Sciences, 109 (15), 5609 - 5614. doi:10.1073/pnas.1201423109