By Hongbo Yang A DISSERTATION Submitted t o Michigan State University By dissertation v vi Andrés Viña , Sue Nichols, William Taylor, Shuxin Li, Jindong Zhang, Yue Dou, Fang Wang, Zhiqiang Zhao, Ying Tang, Min Gon Chung, Thomas Connor, Jacqueline Hulina, Anna Herzberger, Molly Good, So - Jung Youn, Kelly Kapsar, Ciara Hovis, and Yingjie Li I would also like to thank my parents and sisters for their unconditional support s and love . In particular, I would like to thank my beloved wife , Yan Zhang , who always stands by me, trusts me, encourages me, and feels proud of me. Any achievement that I made from the compl etion of this dissertation is equally theirs. Finally, I gratefully acknowledge the financial support from the National Science Foundation, the Environmental Science and Policy Program, William W. and Evelyn M. Taylor Endowed Fellowship program , , and . vii TABLE OF CONTENTS LIST OF TABLES ................................ ................................ ................................ .......................... x LIST OF FIGURES ................................ ................................ ................................ ..................... xiii CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 1 1.1 Background ................................ ................................ ................................ ........................... 2 1.2 Research Objectives ................................ ................................ ................................ .............. 5 1.3 Study Area ................................ ................................ ................................ ............................. 7 CHAPTER 2 REVEALING PATHWAYS FROM PAYMENTS FOR ECOSYSTEM SERVICES T O SOCIOECONOMIC OUTCOMES ................................ ................................ .... 15 Abstract ................................ ................................ ................................ ................................ ..... 16 2.1 Introduction ................................ ................................ ................................ ......................... 17 2.2 Con ceptual Framework ................................ ................................ ................................ ....... 19 ................................ .................... 22 2.3.1 Demonstration site and PES programs ................................ ................................ ......... 22 2.3.2 Specification of framework components ................................ ................................ ...... 23 2.3.3 Hypothesized pathways ................................ ................................ ................................ 24 2.4 Materials and Methods ................................ ................................ ................................ ........ 26 2.4.1 Household surveys and measurements ................................ ................................ ......... 26 2.4.2 Estimating the effects of different pathways ................................ ................................ 28 2.4.3 Estimating net effects of PES programs on household income ................................ .... 29 2.5 Results ................................ ................................ ................................ ................................ . 30 2.6 Discussion ................................ ................................ ................................ ........................... 33 CHAPTER 3 UNCOVERING THE HIDDEN COST OF CONSERVATION ........................... 38 Abstract ................................ ................................ ................................ ................................ ..... 39 3.1 Introduction ................................ ................................ ................................ ......................... 40 3.2 Materials and Methods ................................ ................................ ................................ ........ 43 3.2.1 Study area ................................ ................................ ................................ ..................... 43 3.2.2 Study design ................................ ................................ ................................ ................. 43 3.2.3 Data collection ................................ ................................ ................................ .............. 45 3.2.4 Estimating the impact of GTGP on crop damage ................................ ......................... 48 3.2.5 Estimating the forgone crop revenue due to GTGP - induced crop damage .................. 50 3.3 Results ................................ ................................ ................................ ................................ . 51 3.4 Discussion ................................ ................................ ................................ ........................... 53 viii CHAPTER 4 FEEDBACK OF TELECOUPLING: THE CASE OF A PAYMENTS FOR ECOSYSTEM SERVICES PROGRAM ................................ ................................ ...................... 57 Abstract ................................ ................................ ................................ ................................ ..... 58 4.1 Introduction ................................ ................................ ................................ ......................... 60 4.2 Materials and Methods ................................ ................................ ................................ ........ 65 4.2.1 Study area ................................ ................................ ................................ ..................... 65 4.2.2 Data collection ................................ ................................ ................................ .............. 67 4.2.3 Modeling household willingness to participate in future GTGP ................................ .. 71 4.3 Results ................................ ................................ ................................ ................................ . 73 4.4 Discussion ................................ ................................ ................................ ........................... 76 4.5 Conclusion ................................ ................................ ................................ ........................... 80 CHAPTER 5 COMPLEX EFFECTS OF TELECOUPLINGS ON FOREST DYNAMICS: AN AGENT - BASED MODELING APPROACH ................................ ................................ .............. 81 Abstract ................................ ................................ ................................ ................................ ..... 82 5.1 Introduction ................................ ................................ ................................ ......................... 83 5.2 Methods ................................ ................................ ................................ ............................... 86 5.2.1 Study area ................................ ................................ ................................ ..................... 86 5.2.2 Model design ................................ ................................ ................................ ................ 89 5.2.3 Model validation ................................ ................................ ................................ ........... 99 5.2.4 Simulation experiments ................................ ................................ .............................. 100 5.3 Results ................................ ................................ ................................ ............................... 101 5.3.1 Model validation results ................................ ................................ ............................. 101 5.3.2 Forest and household dynamics under different s cenarios ................................ ......... 102 5.4 Conclusion and Discussion ................................ ................................ ............................... 105 CHAPTER 6 CHANGES IN HUMAN WELL - BEING AND RURAL LIVELIHOODS UNDER NATURAL DISASTERS ................................ ................................ ................................ ........... 110 Abstract ................................ ................................ ................................ ................................ ... 111 6.1 Introduction ................................ ................................ ................................ ....................... 112 6.2 Materials and Methods ................................ ................................ ................................ ...... 115 6.2.1 Study area ................................ ................................ ................................ ................... 115 6.2.2 Characterizing livelihood changes ................................ ................................ .............. 118 6.2.3 Measuring human well - being changes ................................ ................................ ....... 120 6.2.4 Modeling the linkages between changes in livelihood and human well - being .......... 122 6. 3 Results ................................ ................................ ................................ ............................... 124 6.3.1 Livelihood changes after the earthquake ................................ ................................ .... 124 6.3.2 Human well - being changes after the earthquake ................................ ........................ 126 6.3.3 Linkages between changes in livelihoods and human well - being .............................. 129 6.4 Discussion ................................ ................................ ................................ ......................... 133 6.5 Conclusions ................................ ................................ ................................ ....................... 137 ix CHAPTER 7 CONCLUSIONS ................................ ................................ ................................ .. 139 APPENDICES ................................ ................................ ................................ ............................ 145 APPENDIX A SUPPORTING INFORMATION FOR CHAPTER 2 ................................ ... 146 APPENDIX B SUPPORTING INFORMATION FOR CHAPTER 3 ................................ .... 156 APPENDIX C SUPPORTI NG INFORMATION FOR CHAPTER 6 ................................ .... 179 REFERENCES ................................ ................................ ................................ ........................... 184 x LIST OF TABLES ................................ ......................... ................................ ................................ ................................ ................................ ....................... ............................ ................................ ................................ ................................ ................................ ......... ................................ ................................ ................................ ................. ................................ ................................ ................................ ....... ................................ .............. ................................ ................................ ....... ................................ ................................ ................................ ................................ ..................... ................................ . ............................. xi ................................ ............................... ............................. ................................ .. ................................ ................................ ................................ ........................ .............................. ................................ ................................ ................................ ............ ................................ ................................ ................................ ... ................................ ................................ ................................ ................................ ..................... ................................ ................................ ................................ ..................... ................................ ................................ ................................ ..................... ................................ ................................ ................................ ............... ................................ .......................... xii ................................ ................................ ................................ ..... ................................ ................................ ................................ ..................... ................................ ................................ ................................ ..................... ................................ ................................ ................................ ................................ .......... ................................ ................................ ................................ ................................ ..................... ................................ ................. ............ ................................ ................................ ............ xiii LIST OF FIGURES ................................ ......................... .................. ................................ ............. ................................ ................................ ........... ................................ ..................... ................................ ... xiv ................................ ................................ ................................ ............... ................................ ......... ................................ ................................ ................................ .................... ................................ ................................ ................................ ............ ................................ ................................ ................................ ................................ ............... ................................ ............................. ................. .............. ............. xv ............. ................................ ................................ ................................ ................................ .. ................................ .............................. ................................ .... ................................ ................................ .............................. ....................... 1 CHAPTER 1 INTRODUCTION 2 1.1 3 4 5 1.2 Wolong hereafter 6 We estimated this impact by comparing damage on remaining cropland parcels close to the former crop land enrolled in the program with their counterparts that are far from the former cropland using the matching method . 7 1.3 To achieve my objectives, I chose Wolong as my study area (Fig . 1. 1 ) . Wolong is ideal for this research for several reasons. First, long - term socioeconomic data from household survey s and environment data from field measurement s ha ve been accumulated almost yearly since 1998. The rich datasets and indigenous knowledge lay a good foundation for systematic experiment design and causal inference regarding the complex effects of telecouplings . Second, the system components and telecouplings in Wolong are similar to those in many other areas around the world. Therefore, results and lessons drawn from this study may help guide future research and management in many other systems. Third, telecouplings in Wolon g have experienced significant changes in recent decades. Before the 2000 s, Wolong was a rural area with few connection s with outside world ( An et al. 2001 ) . Since the early 2000 s, a series of development and conservation programs [ e.g., Cash - cropping Development Program, Tourism Development Plan, Natural Forest Conservation Program (NFCP) , Grain - to - Green Program (GTGP) , and Grain - to - Bamboo Program (GTBP)] have been implemented and significantly increased the 8 linkage between Wolong and the outside world . O n May 12, 2008, a devastating earthquake struck Wolong. The earthquake and its associated landslides caused extensive damage to the forest , houses, and infrastructure , including the main road th at connect s Wolong and the outside world . Transportation of goods became difficult and tourism in the reserve was disrupted. All these conservation programs, development projects, and the natural disaster led to dramatic change s in the telecouplings . Cha nges in telecouplings in turn significantly shaped human - nature interactions in Wolong (e.g., switch from fuelwood to electricity for energy) . The dramatic changes in telecouplings and associated human - nature interactions , again, provide an excellent The information on panda habitat and its change was obtained from the published results in ( Ouyang et al. 2008 ) . 9 oppor tunit y to address my research objectives. The following are more detailed descriptions of Wolong. Wolong (102 o o o o established in 1963 and expanded to its current size of 2,000 km 2 in 1975 . It was designed mainly for th e protection of the icon ic species, g iant p anda ( ) ( Liu et al. 1999 ). The elevation in Wolong varies dramatically from 1,150 m to 6,250 m, forming a wide range of micro - environmental conditions that can harbor a large diversity of plant and animal species ( Schaller et al. 1985 ) . Wolong is home to about 1 04 wild giant pandas and more than 6000 species of plants and other animals such as red panda and golden monkey ( China Ministry of Forestry and World Wildlife Fund 1989 , Lai et al. 2003 , Sichuan Forestry Administration 2015 ) . The natura l forests here are mainly composed of evergreen and deciduous broadleaf forests at lower elevations and subalpine coniferous forests at higher elevations, with understory composed of bamboo species such as umbrella and arrow bamboo ( Schaller et al. 1985 , Reid and Hu 1991 , Taylor and Qin 1993 ) . Other natural land cover types include grassland/shrubland above the tree line ( about 3000 m) and alpine meadows at even high er elevations (> 4000 m). Besides the diverse species of plants and animals, Wolong is also home to about 4 , 900 residents, most of whom belong t o Tibetan and Qiang ethnic minorities ( Yang 2013 ) . The Reserve is managed by the Wolong Administration Bureau, which is hierarchically structured with two townships under its governance the Wolong Township and the Gengda Township (Fig . 1 . 1 ) ( Lai et al. 2003 ) . In each township there are three villa ges, each of which is composed of a number of village groups, the smallest administrative unit in China ( Lai et al. 10 2003 ) . Before the 2000 s, the reserve ha d few connection s with the outside world . Local livelihoods rel ied primarily on subsistence - based agricultural acti vities , such as crop production and livestock husbandry , that could generate tiny income for local households. The average annual income per capita in 1990 was only 470 yuan (72 USD, 1 USD = 6.6 yuan as of June 2016 ) ( Lai et al. 2003 ) . As population and number of househo ld s grew in Wolong, the impacts of human activities on the environment became increasingly extensive and intensive ( Liu et al. 2001 ) . Before the 2000s, f uelwood in Wolong was the major energ y source for cooking meals, cooking pig fodder , and heating houses during winter ( An et al. 2002 ) . Cutting trees for fuelwood generated serious impacts on giant panda habitat ( An et al. 2002 ) . Al though electricit y was available, local households were reluctant to switch from fuelwood to electricity to avoid increase d household expenses ( An et al. 2002 ) . In the mid - 1990s, local communities annually consumed around 1 1 ,000 m 3 of wood and t he forest coverage drastically decreased from 52% in 1965 to 35% in 2001 ( Liu et al. 2001 , Viña et al. 2007 , Yang et al. 2013b ) , which had resulted in severe degradation of panda habitat ( Liu et al. 2001 ) . Widespread poverty made illegal logging and poaching common then ( Lai et al. 2003 ) . For example, more than 69 endangered wild animals (including giant pandas) were killed by traps between 1975 and 1983 ( Lai et al. 2003 ) . T o mitigate the degradation of giant panda habitat, the Wolong Administration Bureau adopted new strateg ies that integrated poverty alleviation and socioeconomic development in its management efforts since the late 1990s ( Lai et al. 2003 ) . As part of these strategies , 11 Wolong invested intensively in improv ing the transportation condition s and promoting the development of cash - cropping. In 1999, with an i nvestment of about 100 million yuan (15.4 million USD as of June 2016 ), a provincial highway connecting the reserve with the outside was completed and greatly improved the accessibility of the reserve ( Lai et al. 2003 ) . I n 1995, with the assistance of Wenchuan Farming Burea u , an off - season cabbage was successfully introduced into the reserve and was widely planted since 1998 ( Lai et al. 2003 ) . S pecial subsidies for buying chemical fertilizers and farming utensils were also provided to farmers ( An et al. 2002 ) . S elling cabba ge to the outside market contributed significantly to the increase of household income. During the 1990s, the per capita income tripled from 470 yuan in 1990 to 1 , 414 yuan in 2000 ( Lai et al. 2003 ) , of which about 40 percent came from the cabbage - cropping. I ncrease in house in turn may ha ve promote d their investment i n livestock, which further diversified their income sources. For example, o ur data indicate the number of livestock in Wolong has doubled from 666 in 1998 to 1232 in 2007. Since the early 2000s , Wol ong received substantial financial support from the central government (e.g., the State Forestry Administration) and implemented several PES programs to incentivize local households to reduce their negative impacts on the ecosystem. By the end of 200 3 , abo ut 6 0% ( 449 ha) cropland were converted to vegetated land under two PES programs : t he GTGP and the GT B P . The GTGP paid local farmers to convert their cropland to forest land and the GTBP paid local farmers to convert their cropland to bamboo land ( Liu et al. 2008b ) . Another important PES program implemented in Wolong is the NFCP . It started in 2001 and paid local households to monitor forests to prevent deforestatio n activities . 12 Besides selling agricultural products to outside market and implementing PES programs, Wolong witnessed a n evident labor shift from on - farm to off - farm income activities made available via two other telecouplings: nature - based tourism ( touri sm based on natural attractions in rural areas ) and labor migration ( rural laborers out - migrate to cities for temporary employment s ) . Nature - based t ourism in Wolong started in the middle 1980s. However, b ack then , the management in Wolong generally followe - and - style , which trie d to strictly protect natural resources by separating humans from ecosystems . The reserve was only open to special visitors such as scientists, government officials , or small groups of visitors approved by the govern ment ( Lai et al. 2003 ) . It was not until the 1990s that the reserve started to open to the general public. I n 2002 , a tourism development plan was proposed and approved by the state government ( He et al. 2008 ) . Tourism in Wolong entered a rapid development stage and became an impo rtant alternative income source for l ocal farmers . The total number of tourists has increased by 10 - fold from about 20,000 in 1996 to about 200,000 in 2006 with about 31 % of households in the reserve directly benefited from the tourism industry in Wolong ( Liu et al. 2012 ) . Meanwhile , a t he stunning rural - urban disparity attracted a rapid rise of labor migrants from rural areas to urban centers ( Liang 2001 ) . In Wolong, the percentage of household s with labor migrants has doubled from 12% in 2004 to 24% in 2009 ( Chen et al. 2012a ) . A p revious study ( Chen et al. 2012a ) in Wolong indicat e s that labor migration may have reduce d energy needs of households and labor for activities that may disturb the ecosystem (e.g., collection of fuelwood and medicin al plant s ). As compare d 13 with their counterparts, household s with labor migrants collected m uch less fuelwood ( Chen et al. 2012a ) . Due to the impact s of the above socioeconomic and political factors, a forest transition occur red in Wolong. The forest cover in Wolong recovered from 35% in 2001 to 37% in 2007. Like many other biodiverse areas in the world , Wolong is subjected to frequent seismic activities ( Zhang et al. 2014 ) . On May 12, 2008, the catastrophic Wenchuan E arthquake (Ms 8.0 , the most devastating in China since the 1950s ) struck the southwest China ( Zhang et al. 2014 ) . Wolong is within the area of high seismicity and near the epicenter of the ear thquake. The earthquake killed 129 people in Wolong, with 6 people missing, and 35 people seriously injured ( Wang 2013 ) . The earthquake also caused e xtensive damage to most building s and the main road that connect s the reserve with the outside world . In resp onse, a series of reconstruction programs were implemented by the government to help restore the ecological and soci o economic systems in Wolong (e.g., about 2.18 billion yuan has been invested by central government alone ( Aba Admi nistration 2016 ) ). Plans call for reconstruction of damaged infrastructure, relocation of household s from sloping areas with high risk to earthquake - related landslides to flat land , and promotion of the redevelopment of the interrupted tourism industry . Despite the great efforts in post - disaster reconstruction, the condition of the main road that holds the key for the recovery of nature - based tourism and sale of local agricultur al products remain ed poor because of the repeated damage by landslides after t he earthquake from 2008 to 2015 . In 2012 , the third round of road reconstruction project was started. To avoid the impact of landslides, the new road goes through tunnels in area s 14 susceptible to impacts of landslides. The project lasted for four years and was completed in 2016 . It may have substantial impacts on the telecouplings (e.g., nature - based tourism) by changing the links between Wolong and the outside world. Because all the main chapters (i.e., Chapter 2 to 6) have been published in or have been p repared for peer - reviewed journals, more i nformation about the study area such as details on different conservation and development policies are provided in each of the main chapters. 15 CHAPTER 2 REVEALING PATHWAYS FROM PAYMENTS FOR ECOSYSTEM SERVICES TO SOCIOECONOMIC OUTCOMES , 16 17 2.1 18 19 2.2 20 21 22 2.3 2.3.1 We used two PES programs implemented in Wolong Nature Reserve (Wolong herea fter) to demonstrate the operationalization of the framework. Wolong is a flagship protected area in southwest China established mainly for protecting the giant panda ( Liu et al. 2016a ) . about 4,900 local residents , living in around 1,200 households ( Liu et al. 2016a ) . The local resi dents mainly rely on crop production and livestock husbandry ( Wang 2013 ) . Since the early 2000s, working in the local tourism industry and out - migrating to work in cities have become important income sources for some local households ( Chen et al. 2012a , Liu et al. 2012 ) . As part of the effort to address the rapid degradation of panda habitat due to human 23 activities (e.g., agricultural expansion, timber harvesting, and fuelwood collection), two PES programs [the Grain - to - Green Program (GTGP, which is a national program and one of the worl - to - Bamboo Program (GTBP, a local program to grow bamboo on cropland for feeding pandas in captivity and for attracting tourists)] have been implemented in Wolong since 2000 and 2002, respectively ( Liu et al. 2016a ) . Under these programs, local households received payments annually from the government based on the amount of cropland they converted to forest land or bamboo land (See text description and Table S 1 in Supplementary Materials for details). As a national conservation program, the GTGP also pays land owners to plant trees on barren land in some regions, but in Wolong, only cropland has been enrolled into the GTGP ( Wang 2013 ) . 2.3.2 24 2.3.3 (e.g., Chen et al. 2012a , Liu et al. 2012 ) the GTGP and the GTBP provoked changes in each of these three livelihood activities, which then affected the income. s previous studies ( Chen et al. 2012a , Liu et al. 2012 ) indicate, we first hypothesized that all the th ree livelihood activities can increase income. W hypothesized that both the GTGP and the migration. This is because previous studies ( Liu et al. 2008a , Lin and Yao 2014 ) show that cropland reduction due to conservation policies could release rural labor fro m crop production and promote the shift from on - farm to off - farm activities such as working in the local tourism industry or out - migrating for jobs in cities. Furthermore, we hypothesized that both the GTGP and the GTBP had negative impacts on crop product ion because participating households converted parts of their cropland to forest or bamboo land. These hypothesized linkages can form two - step pathways through which the GTGP and the GTBP affected these three livelihood activities, which in turn affect non payment income . For example, the GTGP may affect nonpayment income through the pathway in which the GTGP promotes participation in tourism, which then increases nonpayment income. In addition, we hypothesized that these three livelihood activiti es were linked and the linkages among them constituted longer pathways through which the GTGP and the GTBP 25 affected . For example, we hypothesized that tourism negatively affected crop production. This is because tourism activities (e.g., operating a restaurant) are often labor intensive ouseholds that participate in the local tourism industry may have less labor available for farming activities and thus may maintain less land for crop 26 production. We also hypothesized that tourism had a negative influence on labor migration. Althou gh both tourism and labor migration have the potential to increase rural household income, rural migrant workers in cities may lack health insurance coverage, face substantial educational expenses for their children, experience discrimination from urban re sidents, and suffer from high stress and depression . Therefore, local tourism jobs in Wolong are often more attractive than migrant jobs in cities. If a household has access to jobs in the local tourism industry, it is less likely to have labor migrants working in cities. These hypothesized linkages among different livelihood activities, pl us the aforementioned linkages between livelihood activities and the other two components (i.e., PES programs and nonpayment income ), constitute additional pathways through which PES programs affect household income . For instance, the GTGP may a ffect nonpayment income through the pathway in which the GTGP promotes participation in tourism, which then decreases crop production and in turn nonpayment income . To obtain reliable estimates of these hypothesized pathways, we considered a broad set of c ontrol variables to characterize the demographic (e.g., household size), socioeconomic (e.g., social ties to government), and biophysical (e.g., distance to the main road) features of local households (Table S 2). 2.4 2.4.1 27 In 1999 (before the PES programs were implemented), our research team conducted the first household survey in Wolong to collect data covering demogra phic (e.g., household size, birth year, gender, and education level) and socioeconomic (e.g., income sources, cropland area, and expenditures) information of individual households in 1998. Two hundred and twenty households (about 20% of the total in Wolong ) were randomly selected for surveys with strata based on administrative groups (the smallest administrative unit in China). These households sampled in 1999 were revisited for data collection in 2006, when the PES programs had already been implemented for several years. Besides collecting similar demographic and socioeconomic activities, labor migration, the GTGP, and the GTBP in previous years. Eighteen household s were missing from the survey in 2006 for various reasons such as deaths, migration to outside areas, or temporarily working outside Wolong during the survey period. As a result, in this study, we used data from a total of 202 households surveyed in both years to examine the pathways through which PES programs affected household income. In addition to the household surveys, we measured the location of each household using a Global Positioning System device and calculated the distance of each household to t he main road using the software ArcGIS 10.2 (ESRI Inc., California, USA). The survey instruments and data collection procedures we used in this study were reviewed and approved by the Institutional Review Boa rd of Michigan State 28 University (https://hrpp.ms u.edu/). In this study, we measured the GTGP and the GTBP at household level with the proportions of cropland a household converted to forest land under the GTGP and bamboo land under the GTBP, respectively. We measured crop production in 2005 with the am ount of cropland devoted to it. We measured labor migration and participation in the tourism industry in 2005 with two binary variables that indicate whether the household had members temporarily out - migrate to cities for jobs or had members working in the local tourism industry, respectively. Additionally, physical conditions that are commonly found to be relevant to household income or the livelihood activities ment ioned above. Descriptive statistics of all these variables are shown in Table S 2. 2.4.2 We tested the hypothesized linkages among PES programs, related livelihood activities, and non - payment household income using structural equation modeling method ( Bollen and Noble 2011 ) . Structural equation modeling is statistically unbiased and has been widely us ed in statistical inference literature ( Bollen and Noble 2011 ) . 29 ( 2. 1) where is the vector of endogenous variables, rep resenting variables explained by the model. is the vector of exogenous variables in the model (i.e., variables not explained by the model). is the vec tor of error terms. is the coefficient matrix describing the effects of endogenous variables on endogenous variables. is the coefficient matrix descri bing the effects of exogenous variables on the endogenous variables. is the number of endogenous variables. is the number of exogenous variables. Since some endogenous variables are dichotomous (tourism participation and labor migration), we obtained the path coefficients in the model with the robust weighted least square estimator. We used a set of validation indices to test how well the data support the hypothesized pathways. All values of these indice s indicate our empirical data supported the hypothesized pathways well (Table S 3). After obtaining path coefficients (Table S 4), we calculated the effect of each pathway through which the GTGP or the GTBP affected the nonpayment income (Table S 5). Ba sed on that, we further calculated the net effects transmitted through each observed livelihood activity in this study (crop production, tourism participation, and labor migration), the effect transmitted through unspecified processes (i.e., effect capture d by the unspecified pathway), and their total (Table 2. 1). We conducted the statistical modeling and analyses using Mplus 7 ( Muthén and Muthén 2012 ) . 2.4.3 Based on estimates of the total effect on nonpayment income and information on the direct payments, we calculated the net economic effect per unit area (mu) of cropland enrolled in the 30 GTGP and the GTBP with the following equ ation: 2.5 31 32 Livelihood activity/Processes Descriptions Coefficients GTGP GTBP Crop Production Cropland devoted for crop production in 2005 - 0.664 *** - 0.563 *** Tourism Participation Whether the household has a member who directly par ticipated in tourism activities in 2005: 1. Yes; 0. No 0.058 0.142 Labor Migration Whether the household had labor migrants in 2005: 1. Yes; 0. No 0 .0 0 9 0.048 Other unspecified processes Other livelihood activities that are not observed in this study, and/or other dimensions of the observed activities (i.e., crop production, tourism participation and labor migration) that are not captured by their proxies above. - 0.006 - 0.477 Total The sum of all the effects transmitted through all the three liveliho od activities and other unspecified processes. - 0.602 * - 0.850 ** he effect transmitted through unspecified processes is represented by the coefficient of the unspec ified pathway. 33 2.6 Poverty eradication and ecosystem conservation are among the major goals being targeted by the 2030 Agenda for Sustainable Development of the United Nations ( United Nations 2015 ) . To achieve these goals, scientists, policy makers, and conservation practitioners need a better understanding of the underlying pathways through which conservation policies succeed or fail in generating desirable outcomes ( Ferraro and Hanauer 2014 ) . Our study here illustrates that integrated analysis of the linkages among PES programs, livelihood activities, and socioeconomic outcomes can help reveal the pat hways from PES programs to socioeconomic outcomes. In contrast to the positive effects of PES programs on income found in many other studies (e.g., Liu et al. 2010 , Lin and Yao 2014 ) , we found that the negative effects of the GTGP and the payments were counted. These negat ive net effects on household income occurred perhaps because, as time went by, the fixed payments of the GTGP or GTBP failed to cover the growing gap between their positive and negative effects on income through different pathways. From 2000 to 2003, house holds in Wolong enrolled a large portion (about 66% on average) of their cropland into these two programs. However, the price of agricultural products in China has increased dramatically since 2004 ( Lu et al. 2014 ) . Therefore, the strength of the pathway through which these programs negatively affect income by reducing crop production was 34 participating households to find off - farm employment in the local tourism industry or in cities were small, though these livelihood activities can significantly increase household income (Fig. 3). Therefore, the gap between the negative effect on income due to forgon e crop production and the positive effects on income through promoting off - farm employment increased after the implementation of the GTGP and GTBP. However, the fixed payment levels of these programs did not consider the possible changes in the opportunity costs borne by participating households and thus failed to cover the growing cost of lost crop production in the later years of our study period. Based on the understood pathways, conservation practitioners may be able to identify the obstacles to improv ing the socioeconomic performance of PES programs and design effective management strategies accordingly. Our study results in Wolong show that both the GTGP and the GTBP had weak effects on promoting participation in the local tourism industry or labor mi gration. One major reason might be that the local households have limited access to the benefits brought by tourism development. For example, evidence from previous studies in Wolong ( He et al. 2008 , Liu et al. 2012 ) and other areas ( Kiss 2004 ) suggest a large portion of tourism revenu e often goes to tourism development companies and the government. Local communities often receive only a small share of the benefits brought by tourism (< 4% in Wolong) ( He et al. 2008 ) . In addition, alth ough China has witnessed a dramatic increase of labor migrants (from only 2 million in the early 1980s to more than 150 million in 2010 ( Rush 2011 ) ), many barriers that hinder labor migration remain. The major barriers include lack of skills, 35 unequal educational opportunities for children of migrant workers in cities, and administrative restrictions on the shift from rural re sidence to urban residence ( Li 2011 ) . Due to these barriers, participating households may be unable to effectively utilize the payments and surplus labor made available by PES programs to participate in these off - farm livelihood activities. Therefore, management interventions that help overcome these barriers (e.g., providing training to participating households to develop new skills and o ffering equal opportunities for migrant workers in urban areas) should be considered to increase the benefits participating households could obtain from these off - farm livelihoods, and ultimately improve the socioeconomic outcomes of these PES programs. Ot herwise, higher payments should be offered to local households to cover the associated losses from participating these programs, though it may put a heavier financial burden on governments. Like any other conservation polic y ffects often vary across space and time. With a better understanding of the underlying pathways, we may be better positioned to explain and anticipate the socioeconomic outcomes of PES programs in different contexts. For example, PES programs similar to th e GTGP and the GTBP have been widely implemented around the world , such as the Conservation Reserve Program in the United States , the Permanent Cover Program in Canada ( McMaster and Davis 2001 ) , the Common Agricultural Policy in Europe , Pagos de Servicios Ambientales program in Costa Rica ( Pagiola 2008 ) , and payment s for afforestation programs in Bolivia ( Asquith et al. 2008 ) and Ecuador ( Wunder and Albán 2008 ) . Land owners participating in these programs receive payment to convert their cropland to vegetative land. Therefore, 36 pathways as identified in our demonstration case may be applicable to explain the socioeconomic outcomes of these PES programs (e.g., the se programs may also negatively affect income through reducing crop production and positively affect income by prompting them to seek alternative livelihoods). In addition, a better understanding of the pathways may help anticipate the dynamics of PES pro - growing demand for laborers. In a number of coastal cities in China, many factories have been struggling with labor shortages in recent y ears ( Zhan and Huang 2013 ) . Meanwhile, the Chin ese government has implemented a series of policies favorable for labor migrants to work in cities (e.g., reform of the existing urban - biased residence registration system) ( Fan 2008 ) . These changes may help rural households find off - f arm employments in cities, and thus enhance the socioeconomic outcomes of the GTGP and the GTBP which have released many rural laborers from farming activities. As urbanization continues at an increasing speed in the developing world ( Cohen 2006 ) , similar trend in PES Although our framework was developed for analyzing the socioeconom ic outcomes of PES programs, it can be easily adapted for the analysis of other conservation policies (e.g., protected areas) that also have complex socioeconomic effects by affecting different livelihood activities. Ultimately, to improve the socioeconomi c outcomes of conservation policies, it is necessary to develop more elaborate theories (e.g., metacoupling theory that integrates human - nature interactions across space ( Liu 2017a ) ) to guide conservation practices that will enhance positive 37 outcomes while mitigating n egative ones. It is our hope that the framework proposed and its operationalization in this study will contribute to the construction of such theories and a socioec onomic outcomes. Armed with such theories and knowledge, scientists, policy makers and conservation practitioners may be able to better use conservation tools for achieving Sustainable Development Goals. 38 CHAPTER 3 UNCOVERING THE HIDDEN COST OF CONSERVA TION , 39 40 3.1 41 42 43 3.2 3.2.1 3.2.2 44 45 3.2.3 46 47 48 3.2.4 49 50 3.2.5 51 3.3 52 53 3.4 54 55 56 57 CHAPTER 4 FEEDBACK OF TELECOUPLING: THE CASE OF A PAYMENT S FOR ECOSYSTEM SERVICES PROGRAM , Xiaodong Chen, 58 59 60 4.1 61 62 63 64 65 4.2 4.2.1 66 In Wolong, cropland parcels close to forest are often susceptible to crop damage by wild animals, such as . Therefore, they can generate less economic benefit than cropland farther from forest and constitute the majority of the cropland enrolled in the GTGP ( Chen et al. 2010 ) . 67 4.2.2 68 69 Variables Description Mean (SD) Outcome Participation Whether the household is willing to participate in the future GTGP in the hypothesized scenario: 1, Yes; 0, No. 0.63 (0.48) Scenario attributes Crop damage intensity The crop damage intensity assumed in the hypothesized scenario. 0.29 (0.16) Program payment The payment lev el of the future GTGP assumed in the hypothesized scenario. (Yuan) 1005.54 (408.78) Social norm in the future GTGP assumed in the hypothesized scenario. 0.34 (0.27) Characteristics of interv iewee Gender The gender of the interviewee: Male, 1; Female, 0. 0.60 (0.49) Age The age of the interviewee. (Year) 50.34 (12.66) On - farm laborer Whether the main income activity the interviewee is involved in is farming: Yes, 1; No, 0. 0.62 (0.47) Hou sehold demographic and economic conditions Household size The number of members in the household. 4.48 (1.41) Education The average education level of household members. (Year) 6.56 (2.65) Stable off - farm employment The number of household members with an off - farm job that will last at least one year. 0.74 (0.89) Farming income The log - transformed income obtained from agricultural production. (Yuan) 7.15 (3.82) Income The log - transformed gross household income. (Yuan) 10.89 (0.88) Characteristics of cropland parcels owned by the household Cropland area 3.03 (3.33) Cropland renting Whether the household has cropland currently rented to other households: Yes, 1; No, 0. 0.17 (0.38) Max distan ce to road The maximum distance of the cropland parcels owned by the household to the main road. (m) 404.11 (536.79) 70 71 4.2.3 72 73 4.3 74 ; 75 76 4.4 77 78 79 80 4.5 81 CHAPTER 5 COMPLEX EFFECTS OF TELECOUPLINGS ON FOREST DYNAMICS: AN AGENT - BASED MODELING APPROACH 82 83 5.1 from only 2 million in the early 1980s to more than 150 million in 2010 ( Rush 2011 ) . Meanwhile, there has been a rapidly g rowing 84 85 86 5.2 5.2.1 87 88 89 5.2.2 90 91 The demographic submodel simulates dynamics of persons and households. In our model, individual p erson s and h ousehold s are hierarchically connected with each other (i.e., a h ousehold agent consists of a number of p erson agents ) . The demographic profile of each household agent was modeled by simulating life histories of individual person agents. Major events of individual persons include birth, marriage, aging, and death. Maj or household events include : [1] household formation that may occur when young adults get married, [2] change in household size when there are new member s coming or old member s leaving, and [3] household dissolution when there are no members left. Each hou sehold has a specific location in the landscape and its behavior is based on its attributes , including household size, number of laborers, cropland, and whether it is a tourism or a labor migration household . Household behavior is also constrained by envir onmental conditions like elevation and distance to the main road. 92 Our demographic submodel was largely adopted from the models developed in previous studies ( An et al. 2002 , An et al. 2003 , An et al. 2005 , Chen et al. 2014 ) and was initialized with data from an agricultural ce nsus conducted in Wolong in 1996. The data include age, gender, and marital status of each household member, kinship relations among household members, and the amount of cropland. In 1996, there were 4053 residents in Wolong distributed in 892 households. The geocoded locations of households ( An et al. 2002 , An et al. 2003 , An et al. 2005 , An et al. 2006 , Chen et al. 2014 ) . In 1999, our research team conducted the first household survey in Wolong to collect data covering the demographic (e.g., household size, birth year, gender, and education level) and the socioeconomic (e.g., income sources, cropland area, and fuelwood collection ) information of individual households in 1998 ( An et al. 2001 ) . A total of 220 households (about 20% of all 93 households in Wolong) were randomly selected for survey with strata based on administrative groups (the smallest administrative unit in China) . These households sample d in 1999 were revisited in 2006 to collect their information in the previous year (2005). There were 18 households missing from the 2006 survey due to various reasons such as deaths, migration to outside areas, or temporarily working outside Wolong durin g the survey period. Models Va riables Labor migration Tourism participation Coefficient (SE) Coefficient (SE) Tourism participation - 1.47 (0.56) ** - Household size - 0.19 (0.16) 0.26 (0.14) The number of adult (age > 18) household members 1.04 (0.20) *** - 0.12 (0.16) Average ag e of adult household members - 0.013 (0.029) - 0.012 (0.023) The maximum school years of adult household members - 0.084 (0.075) 0.26 (0.14) *** Log transformed distance to main road (m) 0.076 (0.13) - 0.20 (0.10) Township (Gengda: 1; Wolong: 0) - 0.33 (0 .42) 0.33 (0.34) Cons tant - 2.13 (1.68) - 2.85 (1.35) * 94 95 96 - become or stop being a tourism or a labor migration household - over time i s summarized in Fig. 5 .3 97 5 .2.2.3 Landscape submodel The landscape submodel simulates forest dynamics with specific consideration of household fuelwood collection, establishment of new households, and other environme ntal conditions (e.g., elevation and slope). Our simulation focuses on a 6 km - buffer region around all households (Fig. 98 5 .1) because almost all deforestation activities in the study area happen ed within the distance of 6 km from the households ( Linderman et al. 2005 ) . The total area of the simulated natural la ndscape is 553 km 2 . The l andscape is represented in our model as a digital of 90 90 m cells . Each cell has a set of attributes including elevation, slope, aspect, and forest status (forest or nonforest) . The elev ation, aspect, and slope were obtained based on a digital elevation model derived from a topographic map ( Liu et al. 2001 ) . The forest cover information of the landscape cells was initialized with a published binary forest (forest/nonforest) map derived from Landsat Thematic Mapper images acquired in 1997 ( Liu et al. 2001 ) . The Models Parameters Deforestation Forest recovery Coefficient (SE) Coefficient (SE) Elevation (100 m) - 0.008 (0.014) - 0.008 (0.011) Slope (degree) 0.001 (0.006) - 0.009 (0.006) Aspect (Parker scale ( Parker 1982 ) ) - 0.054 (0.008) *** 0.064 (0.01) *** Distance to forest edge (m) - 0.019 (0.001) - 0.014 (0.001) *** Fuelwood impact (m / m) (2) 0.031 (0.008) *** - 0.009 (0.008) Total fuelwood ( m ) 0.20 (0.003) *** - 0.023 (0.003) *** Cons tant 347.46 *** 1.792 *** 99 classification of the satellite images was performed using unsupervised digital classification based on ISODATA technique ( Jensen and Lulla 1987 ) and was validated using ground - truth ing . The accuracy of the forest cover map is about 80% ( Liu et al. 2001 , An et al. 2005 ) . Landscape cells may experience deforestation (from forest to nonforest) or forest recovery (from nonforest to forest). The fo rest change of each cell i s determined by empirical models obtained from a previous study in the reserve ( Chen et al. 2014 ) . According to this study, the slope, aspect, distance to forest edge, and impacts of fuelwood collection by local households (Table 5 .3). Fuelwood collection has a significant positive ef fect on forest loss ( p < 0.001) and a significant negative effect on forest recovery ( p < 0.001) (Table 5 .3). At every time step, we calculated the deforestation probability for each forest cell and recovery probability for each non - forest cell to determin e their forest status (forest or nonforest). For a detailed description of the construction and validation of these forest change models, please refer to the cited study ( Chen et al. 2014 ) . 5.2.3 In this study, we validate d the agent - based model by comparing the simulated landscape , demograph y, and telecoupling - related statuses with the corresponding observed patterns at the whole Wolong level . For the demographic submodel, we calibrated it with the 1996 agricultural census data and ran it for 10 years. To consider the influence of stochastic processes in our model, we used the mean results from 20 runs for validation. We compared the simulated m ean 100 population size and mean number of households with that obtained from the 2006 household registration data . For the telecoupling submodel, we compared the simulated percentages of tourism households and labor migration households in 2005 with the obser ved values from our household survey data. If the difference between observed and simulated values is less than the observed mean yearly change (change in the observ ed values divided by the number of years between the observations) , we considered the model simulation as having good validity. We validated the impacts of tourism and labor migration on fuelwood collection, and impacts of fuelwood collection on forest dynamics together by comparing the simulated forest distributions in 2007 with a published em pirical forest cover map in 2007 ( Viña et al. 2011 ) . This 2007 forest cover map was derived from a digital classification of the imagery of Landsat Thematic Mapper. The map was validated using ground truth data and has an accuracy of 82.6% ( Viña et al. 2011 ) . The comparison between simulated and actual maps was performed using a receiver operating characteristic (ROC) curve ( Hanley and McNeil 1982 ) with a random sample of 5000 pixels (2500 forest pixels and 2500 nonforest pixels) from the empirical forest cover map as the validation dataset. We used the area under the ROC curve (AUC) as a measure of the accuracy of the simulated forest maps. The values of AUC ranges from 0 to 1, where a value of 1 indicates perfect accuracy, while a value of 0.5 implies that the accuracy is no better than a random guess ( Araújo et al. 2005 ) . 5.2.4 101 5.3 5.3.1 102 5.3.2 Factors Observed value Observed mean yearly change Model m ean Difference between model mean and observed value |Difference|< observed mean yearly change Population in 2006 4504 45 4487 17 Yes Household number in 2006 1197 31 1176 21 Yes Tourism households in 2005 (%) 31.2% 3.1% 28.9% 2.3% Yes Labor migration households in 2005 (%) 21.7% 2% 22.2% - 0.5% Yes 103 104 105 5.4 106 ( Li 2011 ) . Therefore, management interventions that hel p overcome these hardships (e.g., offering equal job opportunities for migrant worker s) should be considered to increase the benefits labor migrant could obtain from this off - farm livelihood. The increase in benefit farmers could obtain from labor migratio n may promote tourism households to also have labor migrants and enhance the labor shift from on - farm to off - farm activities . 107 ( Kiss 2004 ) 108 109 in an increasingly telecoupled Anthropocene . 110 CHAPTER 6 CHANGES IN HUMAN WELL - BEING AND RURAL LIVELIHOODS UNDER NATURAL DISASTERS , 111 112 6.1 113 114 115 6.2 6.2.1 Our study area is Wolong in southwestern China (102 o o o o where the Nature Reserve was designed mainly for the conservation of giant pandas ( locates within one of the overlapped regions between earthquake prone zone and global biodiversity hotspots in China. The information on panda habitat and its change was obtained from the published results in ( Ouyang et al. 2008 ) . 116 ) (Fig. 6 . 1). The reserve was established in 1963 and expanded to its current size of 2,000 km 2 in 1975 ( Liu et al. 2016a ) . It about 4,900 local residents , living in around 1,200 households ( Liu et al. 2016a ) . Wolong is ideal for this researc h for several reasons. First, Wolong is within a region susceptible to natural disasters and was seriously affected by the Wenchuan Earthquake (Fig. 6 . 1). he reserve lies on the Longmen Mountain fault and has been subjected to frequent seismic activities ( Zhang et al. 2014 ) . Since 1933, there have been ten earthquakes with the magnitude of 7.0 Ms or higher occurred around this region, including the most recent Jiuzhaigou Earthquake occurred in 2017, Lushan Earthquake occurred in 2013, and the Wenchuan Earthquake occurred in 2008 ( Xu et al. 2013 , Zhang et al. 2014 , Lei et al. 2017 , Yang et al. 2017b ) . The epicenter of 2008 Wenchuan Earthquake was only 2 km Wolong among the areas most seriously affected by the earthquake . The E arthquake and its associated landslides killed severe damage to local infrastructure and facilities , including residential houses, hospitals, schools, hotels, and the main road that connect s Wolong to the outside world 117 Second , household livelihoods in Wolong before the earthquake share many common features with other rural areas around the world. In Wolong, as in many other rural areas, crop produ ction (e.g., growing cabbage, corn and potato) and livestock husbandry (e.g., rearing cattle or yaks) are important livelihood strategies ( Liu et al. 2016a ) . Meanwhile, the rich natural resources in Wolong made it a famous tourism destination. The development of nature - based tourism in the 2000s benefited many local households by bringing off - farm job opportunities ( He et al. 2008 , Liu et al. 2012 , Yang et al. 2015 ) . In recent decades in China, a widening rural - urban disparity of job o pportunities has attracted a rapidly growing number of farmers from rural areas to urban centers ( Rush 2011 ) . Wolong has n ot been an exception, with a growing number of households having members out - migrate to cities for temporary jobs ( Chen et al. 2012a ) . rural areas around the world, meth ods and findings from this study may guide research and management not only in Wolong, but also many other places around the world ( Kramer et al. 2009 , Pulido - Fernandez et al. 2015 ) . Finally, our research team has been conducti ng long - term interdisciplinary research on coupled human and natural systems in Wolong since late 1990s. This lay s a n essential foundation for examining the linkages between changes in livelihoods and human well - being after the earthquake ( Liu et al. 2016a ) . For example, the detailed household information collected before and after the earthquake constitutes an excellent dataset for characterizing the changes of local liveli hoods . In addition, Yang et al. (2013) developed an survey - based approach to quantify 118 human well - being, which offers a feasible way to evaluate human well - being changes of local households ( Yang et al. 2013a ) . 6.2.2 In this study, we focus on four major types of livelihood activities that changed after the eart hquake in Wolong, including local off - farm labor, crop production, labor migration (temporary out - migration to work in cities), and livestock husbandry. Since household members often make joint or coordinated decisions regarding livelihood affairs, all dat a characterizing livelihood changes were collected at the household level. We used household survey data collected in Wolong in 2007, 2010, and 2015. It contains detailed d emographic (e.g., household ) and socio economic (e.g., cropland area, number of livestock , livestock selling prices, and income sources) information of local households at three important time steps: 2007 (just before the earthquake), 2009 (soon after the earthquake) and 2014 (six years after t he earthquake), respectively. We conducted these surveys in the form of face - to - face interviews. During these interviews, we selected household heads or their spouses as interviewees because they usually irs. Before performing the formal surveys, we conducted pretests In total , 199, 2 87 , and 245 randomly sampled households completed our formal survey s, with a response rate of 93%, 95%, and 96%, respectively. 119 Table 6 Variables Description Mean (SD) Livestock husbandry 09 The number of liv estock (as measured by equivalent number of sheep) raised in 2009 . 1.898 (8.624) Change in livestock husbandry Change in the number of livestock from 2009 to 2014. 4.056 (18.71) Socioeconomic and demographic characteristics Human well - being 09 Overall human well - being index value in 2009. 0.363 (0.15) Change in well - being 07~09 Change in the overall human well - being index value from 2007 to 2009. - 0.247 (0.166) Basic materials 09 Sub - index value of basic material in 2009. 0.374 (0.215) Change in bas ic materials 07~09 Change in sub - index value of basic material from 2007 to 2009. - 0.235 (0.247) Security 09 Sub - index value of security in 2009. 0.189 (0.128) Change in security 07~09 Change in sub - index value of security from 2007 to 2009. - 0.449 (0.20 8) Health 09 Sub - index value of health in 2009. 0.432 (0.179) Change in health 07~09 Change in sub - index value of security from 2007 to 2009. - 0.232 (0.186) Social relations 09 Sub - index value of social relations in 2009. 0.682 (0.144) Change in social relations 07~09 Change in sub - index value of social relations from 2007 to 2009. - 0.033 (0.085) Freedom of choice and action 09 Sub - index value of freedom of choice and action in 2009. 0.322 (0.173) Change in freedom 07~09 Change in sub - index value of f reedom of choice and action from 2007 to 2009. - 0.136 (0.153) Total income 09 Log - transformed gross income in 2009. (Yuan b ) 10.033 (1.391) Household size 09 The number of members in the household in 2009 . 4.796 (1.525) Change in household size Househo ld house size change from 2009 to 2014. - 0.215 (1.626) Laborers 09 The number of members involved in income - earning activities in 2009. 3.387 (1.496) Variables Description M ean (SD) Outcome variables Well - being change Change in the overall human well - being index value from 2009 to 2014. 0.271 (0.182) Change in basic material Change in value of sub - index representing basic materials for good life from 2009 to 2014. 0.32 ( 0.279) Change in security Change in value of sub - index representing security from 2009 to 2014. 0.382 (0.191) Change in health Change in value of sub - index representing health from 2009 to 2014. 0.225 (0.186) Change in social relations Change in value o f sub - index representing social relations from 2009 to 2014. 0.013 (0.081) Change in freedom Change in value of sub - index representing freedom of choice and action from 2009 to 2014. 0.195 (0.191) Livelihood activities and their changes Labor work insi de 09 The number of laborers earned income through working in local off - farm sectors in 2009. 1.102 (0.775) Change in labor work inside Change in the number of laborers working in local off - farm sectors from 2009 to 2014. - 0.054 (1.089) Labor work outsid e 09 The number of laborers earned income through working outside the reserve in 2009. 0.409 (0.739) Change in labor work outside Change in the number of laborers working outside the reserve from 2009 to 2014. 0.317 (1.081) Crop production 09 The area o (Mu a ) 3.491 (3.228) Change in crop production 2009 to 2014. (Mu) - 0.481 (3.016) 120 The househol d livelihood information in 2007, 2009 , and 2014 comprises an excellent dataset to characterize livelihood changes after the Wenchuan Earthquake. We operationalized local off - farm labor in Wolong as the number of household member(s) working in local off - fa rm sectors (e.g., construction, operating restaurants). Crop production was operationalized as the average amount of cultivated cropland owned by each local household. Labor migration was operationalized as the number of labor migrants in each household, w hile the livestock husbandry was operationalized as the average number of livestock raised by each household. 6.2.3 Along with the basic household information collected in 2015, w e used a published survey Table 6 Variables Description Mean (SD) Change in laborers Change in the number of labore rs from 2009 to 2014. - 0.183 (1.718) The average schoolyears of laborers. (Year) 5.979 (3.037) education 2009 to 2014. 1.164 (4.21) The gender of t he respondent in our survey (0, female; 1, male) 0.602 (0.491) The schoolyears of the respondent. (Year) 5.688 (3.560) 121 instrument (Table S 6 . 1) to collect data for quantifying human well - being of each surveyed household in 2007, 2009 and 2014. In collecting retrospective information, we followed standard practices of life history calendars to enhance re l l accuracy ( Freedman et al. 1988 , Axinn et al. 1999 ) . In total, 244 households randomly sampl ed in 2015 completed our human well - being survey, with a response rate of 96%. 6 . 122 The survey instrument we used (Table S 6 . 1) was designed based on the framework of human well - being proposed in the Millennium Ecosystem Assessment ( MA 2005 , Yang et al. 2013a ) . Human well - being encompasses five interrelated dimensions: basic material for good life, security, health, good social relations, and freedom of choice and action ( MA 2005 ) . Fig. 6 . 2 presents the structural relations between the overall human well - being index and its five sub - indices represent ing each of the five dimensions. For each sub - index, our survey instrument includes a set of questions to generate indicators to construct it (Table S 6 . 1). We used confirmatory factor analysis with Mplus, version 7 ( Muthén and Muthén 2012 ) to estimate the overall human well - being index and its five sub - indices. We evaluated the validity of these human well - being indices using a set of criteria (Table S 6 . 2). The validation results i ndicate that the overall index and sub - indices of human well - being have high reliability (Table S 6 . 2). To allow cross - year comparisons, we normalized the overall index and the sub - indices to the range from 0 to 1 using maximum - minimum normalization method as suggested by Yang et al. (2015). A higher value of the index value suggests higher satisfaction of corresponding human needs. More technical details regarding construction of the indices, validation, and application can be found in previous studies (e.g., Yang et al. 2013a , Yang et al. 2015 ) . 6.2.4 One of our major goals is to evaluate the relationship between livelihood changes after the earthquake and the recovery of human well - being. Previous work has examined the impact of the earthquake per se ( Yang et al. 2016c ) ; our emphasis is what happened during the recovery period, that is, the year following the earthquake (2009) to six years after the earthquake (2014). 123 We used changes in the overall human well - being index and its five sub - indices between 2009 and 2014 as our measures of recovery outcomes . We hypothesize that the human well - being changes after the earthquake are affected by th e mix of livelihood activities of households in 2009 and their changes during the recovery period (2009 to 2014). To develop indicators characterizing different livelihood changes (Table 6 . 1), we compiled panel data using socioeconomic information on hous eholds in 2009 and 2014. In total, there are186 households surveyed in both years. With these panel data, we constructed linear regression models to relate changes in overall human well - being index and its five sub - indices between 2009 and 2014 to changes in household livelihood activities during the same period as well as their values in 2009. To control for potential confounding effects, our models included some other socioeconomic and demographic factors that may affect human well - being changes (Table 6 . 1). Similar to livelihood activities, some of these factors (e.g., number of laborers in a household) may change during the recovery period (i.e., 2009 to 2014). We thus included variable measuring these socioeconomic and demographic conditions in 2009 a nd their changes between 2009 and 2014 in our models (Table 6 . 1). Our models also included the changes in human well - being indices between 2007 and 2009 as independent variables because the short - term impact of the earthquake on human well - being may have l egacy effect on the long - term recovery. The general form of the models can be given as (6 .1) 124 where refers to the vector of changes in overall human well - being index and the sub - indices between 20 09 and 2014; refers to the vector of corresponding indices in 2009; represents the vector of changes in corresponding indices between 2007 and 2009; and represent the vectors of livelihood activity variables in 2009 and their changes between 2009 and 2014 respectively; and represent the vectors of other socioeconomic and demographic variables in 20 09 and their changes between 2009 and 2014 respectively; is the vector of intercept; are the vectors of coefficients to be estimated; is the vector of error term, in which each error term is assumed to be normally distributed with a mean of zero. We conducted the modeling analyses using Stata 13.1 (StataCorp, College Station, Texas, USA). 6.3 6.3.1 6 . 125 6 . 6 . 126 6 . 6.3.2 6 . 127 128 Variables Coefficients Robust standard error Livelihood activities and their changes Labor work inside 09 0.0620 ** 0.0201 Change in labor work inside 0.0333 * 0.0132 Cropland production 09 0.00641 0.0036 Change in crop production 0.0074 0.0044 Labor work outside 09 0.0256 0.0230 Change in labor work outside - 0.0150 0.0138 Livestock husbandry 09 - 0.00008 0.0001 Change in livestock husbandry - 0.0003 * 0.0002 Socioeconomic and demographic characteristics Human well - being 09 - 0.6121 *** 0.0874 Change in well - being 07~09 - 0.1517 0.0913 Total income 09 - 0.0124 0.0081 Household size 09 - 0.0378 * 0.0147 Change in household size - 0.0255 * 0.0114 Laborers 09 0.0374 0.0191 Change in laborers 0.0385 ** 0.0134 0.0025 0.0062 Change in labo - 0.0008 0.0050 0.0005 0.0226 0.0063 0.0034 Constant 0.4973 *** 0.1006 129 6.3.3 6 . 6 . 6 . 6 . 6 . 130 Variables Coefficients (Robust standard error) Basic materials Security Health Social relations Freedom Livelihood activities and their changes Labor work inside 09 0.0800 ** (0.0291) 0.0146 (0.0224) 0.0548 ** (0.021) 0.0113 (0.0105) 0.06 07 ** (0.0232) Change in labor work inside 0.0484 * (0.0189) - 0.0022 (0.0157) 0.0322 * (0.0132) 0.0082 (0.0071) 0.0313 * (0.0152) Crop production 09 0.0084 (0.0049) 0.0065 (0.0044) 0.0045 (0.0039) - 0.0044 (0.0022) 0.0051 (0.0048) Change in crop production 0.0057 (0.0055) 0.0122 * (0.0049) 0.005 (0.0051) 0.0007 (0.0034) 0.0072 (0.0049) Labor work outside 09 0.0741 * (0.0320) - 0.0116 (0.0270) - 0.0005 (0.0244) - 0.0124 (0.0129) 0.0281 (0.0258) Change in labor work outside 0.0276 (0.0196) - 0.0084 (0.0175) - 0. 0314 * (0.0145) - 0.0035 (0.008) - 0.0329 * (0.0164) Livestock husbandry 09 - 0.00002 (0.00009) - 0.0003 ** (0.0001) 0.00005 (0.0001) - 0.0001 (0.0001) - 0.0001 (0.0001) Change in livestock husbandry - 0.0003 (0.0002) - 0.0005 (0.0002) - 0.0002 (0.0002) 0.0001 (0. 0001) - 0.0004 ** (0.0002) 131 Table 6 Variables Coefficients (Robust standard error) Basic materials Security Health Social relations Freedom Socioeconomic and demographic characteristics Index value 09 a - 0.8031 *** (0.0776) - 0.5046 *** (0.0 971) - 0.4387 *** (0.0812) - 0.0066 (0.0548) - 0.5261 *** (0.0751) Index value change 07~09 b - 0.0839 (0.0745) - 0.1916 ** (0.07) - 0.2584 ** (0.0917) - 0.273 ** (0.1003) - 0.0335 (0.1096) Total income 09 - 0.0197 (0.0132) - 0.0158 (0.0086) - 0.0133 (0.0087) - 0.0053 (0.0043) - 0.0009 (0.009) Household size 09 - 0.0365 (0.0199) - 0.0413 * (0.0179) - 0.0391 * (0.0163) - 0.0217 * (0.0106) - 0.0353 * (0.0165) Change in household size - 0.0262 (0.017) - 0.0348 * (0.0139) - 0.0297 * (0.0115) - 0.0127 (0.0068) - 0.0148 (0.0128) Laborer s 09 0.0201 (0.0273) 0.045 * (0.0216) 0.0447 * (0.0213) 0.0146 (0.0135) 0.035 (0.0201) Change in laborers 0.0247 (0.0189) 0.0351 * (0.0164) 0.0437 ** (0.014) 0.0016 (0.0075) 0.0376 ** (0.0143) 0.0034 (0.0104) - 0.0089 (0.0071) 0.0057 (0.0 061) 0.0034 (0.0028) 0.0041 (0.0073) education - 0.0020 (0.0086) - 0.0075 (0.0061) 0.0025 (0.0048) 0.0022 (0.0027) 0.000003 (0.0058) - 0.0101 (0.0336) 0.0249 (0.0271) - 0.0024 (0.0228) 0.0109 (0.0129) - 0.0028 (0.0255) education 0.0067 (0.0057) 0.0047 (0.0045) 0.0029 (0.0035) - 0.0023 (0.002) 0.0102 * (0.004) Constant 0.7038 *** (0.1501) 0.6002 *** (0.1205) 0.4131 *** (0.1117) 0.1094 (0.0672) 0.2648 * (0.1032) R 2 0.565 0.350 0.503 0.253 0.401 132 6 . 6 . 6 . 6 . 6 . 133 6 . 6.4 6 . 134 6 . 135 6 . 136 137 6.5 138 139 CHAPTER 7 CONCLUSIONS 140 141 142 143 can help to construct a collective base of evidence about the effects of telecouplings that is necessary to develop elaborate and gener alizable theories to guide management practices. For example, it would be worthwhile to adapt the conceptual framework we proposed for analyzing PES programs to analyze other telecouplings (e.g., nature - based tourism ) , evaluate the spillover and feedback e ffect s of the GTGP we detect ed in Chapter 3 and Chapter 4 in other parts of the world where the similar program s have been implemented, investigate the effects of a great diversity of interactions among other telecouplings building off of Chapter 5, and ex plore a holistic picture of the effects 144 of telecoupling change s caused by natural disaster s on other economic and environmental outcomes expanding on Chapter 6 . While assessing effects of telecouplings on rural area s as demonstrated in this dissertation is important , i t is also crucial to evaluate how changes in rural areas in turn affect other places for a more comprehensive understanding of telecouplings. For example, we found nature - based tourism, labor migration, and PES program s all showed positive imp acts on the forest recovery in Wolong, it would be worthwhile to evaluate how the se impacts in turn affected the outflows of ecosystems services from Wolong to other region s. In conclusion, in the face of the sustainability challenges in an increasingly te lecoupled world , improved understanding of the complex effects of telecouplings on rural areas is crucial for effective management. This dissertation ma d e novel contributions to the understanding of telecoupling s by explicitly demonstrat ing that how the co mplex effect s of telecoupling would occur, how they could be quantified, and how knowledge on them could be leveraged to improve the socioeconomic and ecological outcomes in rural areas. The findings and methods from t his dissertation hold the promises to provide much need ed information and tools for future exploration of the complexities of telecouplings across different settings around the world. It is my hope that th is dissertation research together with other telecoupling studies will jointly contribute to the development of elaborated theories (e.g., complexity and metacoupling theory ( Waldrop 1993 , Liu et al. 2007a , Liu 2017b ) ) and a collective base of evidence about the complex effects of telecouplings that can guide the management of human - nature interactions for sustainable development in the telecoupled Anthropocene . 145 APPENDICES 146 147 Wolong Nature Reserve (102 o o o o area in Southwest China ( Fig. S 1). It was established in 1963 and mainly designed for the protection of the world - famous giant panda ( Liu et al. 2016a ) . The reserve is managed by the Wolong Admin istration Bureau, which is hierarchically structured with two townships under its governance - Wolong Township and Gengda Township (Fig. S 1) ( Liu et al. 2016a ) . There are about 4,900 residents in Wolong, most of whom are farmers ( Liu et al. 2012 ) . We chose Wolong as our study area for several reasons. First, situated within one of the top 25 global biodiversity hotspots and being part of the UNESCO World Heritage system, Wolong has special ecological importance. It provides a major sanctuary to 10% of the world's wild giant panda population (total=1,864) and more than 6000 species of other animals and plants, some of which are also endangered (e.g., red panda and golden monkey) ( Wang 2013 ) . Second, long - term socioeconomic data from household surveys h ave been collected in Wolong almost yearly since 1998 by our research team, together with a variety of biophysical data ( Liu et al. 2013b , Liu et al. 2016a ) . The rich datasets and indigenous knowledge lay a good foundation for systematic research design and statistical inference regarding the effects of PES programs on socioeconomic outcomes. Finally, Wolong has been ex periencing significant socioeconomic changes since the implementation of PES programs at the early 2000s. Before 2000, Wolong was a remote area with limited connection to the outside ( Liu et al. 2016a ) . The poverty rate in Wolong at that time 148 was as high as 35% ( Liu et al. 2012 , Liu et al. 2013b ) . The vast majority of local residents were primarily involved in subsistence - based agricultural activities such as growing potatoes and corn. As population size and the number of households rapidly increased, human activities (e.g., farmland expansion, fuelwood collection, and timber harvesting) caused serious degradation of wildlife habitat ( Liu et al. 2001 , Liu et al. 2016a ) . Although the reserve administra tion had implemented several policies before 2000, such as a logging ban and limiting the sites for fuelwood collection, they were not effective mainly due to the lack of alternative income sources and incentives for eco - friendly behaviors ( Liu et al. 2016a ) . As part of the effort to address this critical issue, two PES programs [i.e., the Grain - to - Green Program (GTGP) and the Grain - to - Bamboo Program (GTBP)] have been implem ented in Wolong since the early 2000s (Table S 1). With financial support from the central government (e.g., the State Forestry Administration), the Wolong Administrative Bureau paid the local households annually based on the amount of cropland they conve rted to forest land under the GTGP or bamboo land under the GTBP. More than 60% of the cropland in 1998 was converted to vegetated land under these two PES programs from 2000 to 2003, and has been kept vegetated since ( Wang 2013 ) . Meanwhile, with the politic al and financial support from the government and investments from outside tourism companies, tourism facilities (e.g., roads and hotels) in Wolong were greatly improved. The annual number of tourist visits in Wolong increased more than tenfold in a decade, from 20,000 in 1996 to 235,500 in 2006. In 1998, only about 4% of local households directly benefited from local tourism industry ( Liu et al. 2012 ) . This number increased to around 31% by the period from 2005 to 2007 ( Liu et al. 2012 ) . In addition, as the price of agricultural products 149 increased, selling cash crops (e.g., off - seasonal cabbage) to outside markets gradually became one of the major income sou rces for local households ( Wang 2013 ) . The sale of cash crops constituted as much as 39% of local household income in 2005 ( Liu et al. 2013b ) . Like many other rural areas in China, Wolong experienced an increase in labor migration in the 2000s. The proportion of households with labor migrants nearly doubled from 12% in 2003 to 22% in 2005 ( Chen et al. 2012a , Liu et al. 2013b ) . The average household income significantly increased from 5,000 yuan in 1998 to more than 25,000 yuan in 2007 ( Liu et al. 2013b ) (1 yuan = 0.122 USD in 2005). 150 ( Liu et al. 2016a ) 151 Variables Description Mean (SD) Socioeconomic outcome Nonpayment income Log - transformed incom e obtained from sources other than payments from the GTGP and the GTBP in 2005 9.59 (0.84) PES programs GTGP The proportion of cropland enrolled in the GTGP 0.5 63 (0.1 70 ) GTBP The proportion of cropland enrolled in the GTBP 0.10 0 (0.1 51 ) Livelihood ac tivities Tourism participation Whether the household has member(s) who directly participated in tourism activities in 2005: 1. Yes; 0. No 0.31 (0.46) Labor migration Whether the household has labor migrant(s) in 2005: 1. Yes; 0. No 0.22 (0.41) Crop p roduction The area of cropland (in mu * ) owned by the household for cropping in 2005 3.07 (1.99) Control variables Number of l aborer s Number of laborers in the household 2.65 (1.30) Household size Number of people in the household 4.14 (1.39) Average age Average age of adults (age >18) in the household 42.03 (9.28) Adult education level Maximum education level (in years) of non - student adults in the household 7.65 (3.27) Social t ies to g overnment Whether the household has a member or immediate rela tive working in local governments: 1. Yes; 0. No 0.12 (0.32) Social ties to local leaders Whether the household has a member or immediate relative being a village or group leader: 1. Yes; 0. No 0.18 (0.39) Log (distance to the main road) Log - transforme d distance between the household and the main road 2.77 (1.58) Log (Income 98) Log - transformed total income in 1998 8.31 (0.92) Labor migration (03) Whether the household had labor migrant(s) in 2003: 1. Yes; 0. No 0.12 (0.47) Township Which township th e household is located in: 0 = Wolong township; 1 = Gengda township 0.45 (0.50) * 152 153 154 An a rrow in a p athway represents a linkage through which . Pathway Type Pathway Effect GTGP GTBP Specified P athway PES Program Crop production Nonpayment income - 0.13 - 0. 09 8 PES Program To urism Participation Nonpayment income 0.024 0.051 PES Program Labor Migration Nonpayment income 0 .0 02 0.008 PES Program Tourism Participation Labor Migration Nonpayment income - 0.0 0 6 - 0.0 1 3 PES Program Tourism Participation Crop produ ction Nonpayment income - 0. 006 - 0.013 Unspecified pathway PES Program Nonpayment income - 0.001 - 0.083 155 156 157 158 159 160 161 162 163 164 165 0.12 (0.22) 66.82 (52.45) 19.98 (10.09) 404.67 (599.46) 57.28 (74.44) 0.68 (0.47) 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 ; ; ; * * * 181 Table S 6 * * * * 182 Table S 6 183 Vali dation statistics Rules indicating good validation Model validation results Ratio of Chi - Square to df ( ) < 3 1.61 CFI (Comparative Fit Index) > 0.95 0.977 TLI (Tucker - Lewis Index) > 0.95 0.972 RMSEA (Root Mean Square Error of Approximation) < 0.07 0.03 S RMR ( Standard Root Mean Square Residual) < 0.08 0.030 184 REFERENCES 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202