1. Introduction
As globalization continues, rural areas have been increasingly connected to the rest of the world through telecouplings—socioeconomic and environmental interactions over distances (Liu et al. 2013a, 2015a). As a result, human–nature interactions in rural regions, which were primarily driven by local socioeconomic and biophysical conditions, are now increasingly affected by drivers operating at a distance, such as the growing demand for ecosystem services and labor in cities. This new anthropogenic trend has generated profound impacts on many global sustainability issues such as deforestation (Liu 2014), biodiversity loss (Dou et al. 2018), energy security (Fang et al. 2016), and climate change (Liu et al. 2015b).
Among the telecouplings that link rural areas and other places, rural–urban labor migration (rural residents outmigrate to cities for temporary employments) and nature-based tourism (tourism based on the natural attractions of rural areas) are two globally common and increasingly influential ones (Ratha et al. 2015; Pulido-Fernandez et al. 2015). Factors such as urban economic growth, enlarging rural–urban disparity, and development of transportation networks have been driving a large number of laborers from rural areas to seek job opportunities in cities, especially in developing countries (Ratha et al. 2015). In China alone, the number of rural–urban labor migrants had increased from only 2 million in the early 1980s to more than 150 million in 2010 (Rush 2011). Meanwhile, there has been a rapidly growing demand for visiting the natural and cultural landscapes of rural areas, mostly by residents from cities. For decades, many rural areas around the world have been practicing nature-based tourism (Chung et al. 2018). For example, in the late 1990s, about 80% of nature reserves in China had developed nature-based tourism (Li and Han 2001). Several provinces in southwest China (e.g., Yunnan and Sichuan), where one of the global biodiversity hotspots is located (Myers et al. 2000), have designated nature-based tourism as one of the major sources of their economic growth (Liu 2012).
Previous studies (e.g., Chen et al. 2012a; Pulido-Fernandez et al. 2015; Dai et al. 2012) suggest that labor migration and nature-based tourism can have substantial impacts on the human and natural systems in rural areas. Many rural residents who traditionally relied on subsistence agricultural livelihoods are now shifting to off-farm economic opportunities made possible by these two telecouplings (Kramer et al. 2009). This labor shift substantially mitigated the negative impacts of local farmers on ecosystems (Liu et al. 2012; Yang et al. 2013; Fox 2016). For example, previous studies (Chen et al. 2012a; Cao et al. 2009) suggest that the remittances sent back by labor migrants or revenue from tourism development, have promoted the change of rural energy consumption from fuelwood to electricity, and reduced the deforestation by farmers. With an increasing awareness of the importance of tourism and labor migration, there have been many studies evaluating their effects on socioeconomic and environmental outcomes in rural regions (e.g., Chen et al. 2012a; Liu et al. 2012; He et al. 2008; J. Liu et al. 2016).
However, like studies on other telecouplings, the effects of tourism and labor migration were often evaluated separately and their interaction was often ignored. While both tourism and labor migration have the potential to reduce the negative impacts of local households on the environment (Chen et al. 2012a; Liu et al. 2012), tourism development may limit the growth of labor migration. Previous studies (e.g., Wong et al. 2007) show that rural migrant workers in cities often lack good health insurance coverage, face substantial educational expenses for their children, experience discrimination from urban residents, and suffer high stress and depression due to social displacement. Therefore, local tourism jobs are usually more attractive to rural residents than migrant jobs in cities. Although income opportunities related to nature-based tourism are often seasonal (Cuccia and Rizzo 2011), a household may be less likely to utilize its surplus labor for temporary employments in cities if it can benefit from local tourism development. As a result, the possible positive environmental impacts from labor migration in a rural area with a tourism industry might be smaller than in other rural areas where the tourism industry does not exist.
In this study, we integrated information from different sources and developed an agent-based model to simulate the evolution of tourism and labor migration and their effects on forest dynamics in China’s Wolong Nature Reserve (hereinafter Wolong). We used the agent-based model because it has a unique capacity to consider the heterogeneity and complex interactions of the human and natural components (e.g., households and forest landscape) involved in telecoupling processes. Although empirical modeling approaches like regression models can yield useful insights about single-step or multistep human–nature interactions, such knowledge alone can rarely lead to in-depth understanding of long-term dynamics of coupled human and natural systems (CHANS) (An et al. 2014, 2005). Agent-based modeling can address this limitation by offering an effective way to integrate findings from empirical models with other information to simulate system dynamics over a long period and under alternative scenarios that are hard to be empirically observed (Ligmann-Zielinska and Jankowski 2007; Chen et al. 2012b; Dou et al. 2020).
After calibrating and validating our agent-based model, we used it as a scenario-envisioning laboratory to evaluate the effects of tourism and labor migration by comparing the forest dynamics under different conditions. Specifically, we simulated and evaluated forest dynamics under the following scenarios: 1) tourism and migration are not present in the model, 2) only tourism is present in the model, 3) only migration is present in the model, 4) tourism and migration are present in the model but their interaction is ignored, and 5) tourism, migration, and their interaction are all included in the model.
2. Methods
a. Study area
Wolong (30°45′–31°25′N, 102°52′–103°24′E) was established in 1963 and expanded to its current size of 2000 km2 in 1975 (Liu et al. 1999) (Fig. 1). It provides sanctuary to over 100 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; Sichuan Forestry Administration 2015). The natural forests in Wolong are mainly composed of evergreen and deciduous broadleaf forests at lower elevations, 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). Besides the diverse species of plants and animals, Wolong is also home to about 4900 local residents (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 (Lai et al. 2003).
Wolong Nature Reserve in Southwest China.
Citation: Earth Interactions 26, 1; 10.1175/EI-D-20-0029.1
Before the 2000s, the reserve had few connections with the outside world. Local livelihoods relied primarily on subsistence-based agricultural activities like cropping and livestock husbandry (Zhang et al. 2018). The average annual income per capita in 1990 was only CNY 470 (USD 72; USD 1 = CNY 6.6 as of June 2016) (Lai et al. 2003). As the human population and the number of households grew, local human activities caused serious degradation of panda habitat by the early 2000s (Liu et al. 2001, Bradbury et al. 2014). Of all the human threats, cutting trees for fuelwood by local households was a major one (Bearer et al. 2008). Fuelwood was a major energy source for cooking pig fodder, cooking meals, and heating houses during winter (An et al. 2002). Although electricity was available, local households were reluctant to switch from fuelwood to electricity, in aversion to increased household expenses (An et al. 2002). By the mid-1990s, local households consumed around 11 000 m3 of wood annually and contributed to the rapid shrinkage of forest cover from 52% (1070 km2) in 1965 to 35% (706 km 2) in 2001 (Yang et al. 2013; Viña et al. 2007; Liu et al. 2001).
This trend of net forest loss started to change since 2001 (Tuanmu et al. 2016). Between 2001 and 2007, the forest cover in Wolong recovered from 35% (706 km2) to 37% (799 km2) (Yang et al. 2013). Several socioeconomic and political factors may have contributed to this forest transition. One of them is the development of tourism. Although tourism has existed in Wolong since the late 1980s, the number of tourist visits then was low, and few local people benefited from it (Liu et al. 2012). In 2002, a tourism development plan was formally approved by the provincial and central government (He et al. 2008). Since then, tourism in Wolong has entered a rapid development stage and became an important alternative source of income for local households. The number of tourist visits increased by a factor of 10 from about 20 000 in 1996 to about 200 000 in 2006. In 2005, about 30% of local households directly benefited from tourism activities like selling bacon to tourists and working as tour guides (Liu et al. 2012). Before 2008, when the Wenchuan Earthquake interrupted tourism development in Wolong (W. Liu et al. 2016), tourism season in Wolong often started in May and ended in October.
Meanwhile, as China’s economy has grown rapidly in its cities, the stunning rural–urban disparity attracted a rapid rise of labor migrants from rural areas to urban centers (Liang 2001; Li 2011). In Wolong, the percentage of households with labor migrants has doubled from 12% in 2004 to 24% in 2009 (J. Liu et al. 2016; Liu et al. 2013b). Like many other parts of the world, almost all labor migrants from Wolong find only temporary employments in cities (Chen et al. 2012a). They return to their home villages whenever needed (e.g., in planting or harvesting seasons) and rarely shift to be permanent urban residents (Fan 2008).
In addition to tourism development and labor migration, several conservation policies have been implemented in Wolong since the early 2000s, including the Grain-to-Green Program (GTGP) (also called Grain-for-Green Program) started in 2000, Natural Forest Conservation Program (NFCP) started in 2001, and Grain-to-Bamboo Program (GTBP) started in 2002 (Liu et al. 2008). Under the NFCP, local households receive payment to monitor the forest for preventing illegal timber harvesting, while under the GTGP and GTBP, local households receive payments to convert their cropland to forestland or bamboo land. Of them, the NFCP was specifically designed to reduce deforestation (Viña et al. 2016) and is believed to be the major policy that has contributed to the reductions of deforestation in Wolong after the early 2000s (Yang et al. 2013).
b. Model design
Agent-based models are composed of agents and their environment (Dou et al. 2019; An and Liu 2010). In our model, agents are individual persons and households, and landscape in Wolong is their shared environment. Agents and their environment together are treated as a CHANS, which is connected with other systems (cities) through two telecouplings: tourism and labor migration (Fig. 2). Local households affect forest dynamics mainly through collecting fuelwood. The establishment of new households often lead to small-scale forest clearing (An et al. 2006) and constitutes the other pathway affecting forest dynamics (Fig. 2).
Conceptual framework of the model for simulating the effects of tourism and labor migration on forest dynamics in the coupled human and natural systems of Wolong.
Citation: Earth Interactions 26, 1; 10.1175/EI-D-20-0029.1
The amount of fuelwood collected by each household is determined by its attributes, such as household size, cropland area, and whether being a tourism household and/or labor migration household. In this study, a household is named a tourism household if it directly benefits from local tourism activities. A household is named labor migration household if it has one or more members who outmigrate to cities for temporary employment. Tourism and labor migration reduce fuelwood collection by local households if the households have member(s) directly benefiting from local tourism industry or outmigrating to work in cities. Although a household can be a tourism household and a labor migration household simultaneously, tourism participation by a household reduces its probability to have labor migrant(s). This interaction is manifested in the evolution of households’ attributes (being a tourism and/or a labor migration household) over time.
The interactions among local households, telecouplings, and forest dynamics were implemented in three integrated submodels: a demographic submodel, a telecoupling submodel, and a landscape submodel. Households in this study were modeled as autonomous agents that can interact with each other and with the forest. We parameterized the model using data and findings from different sources such as population and agricultural censuses, household interviews, satellite imagery, and published journal articles (e.g., An et al. 2001; Chen et al. 2014). The agent-based model was developed using the Java programming language on the Swarm platform (Minar et al. 1996). Below are detailed descriptions of each of the submodels.
1) Demographic submodel
The demographic submodel simulates dynamics of persons and households. In our model, individual persons and households are hierarchically connected with each other (i.e., a household agent consists of a number of person 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. Major household events include 1) household formation that may occur when young adults get married, 2) change in household size when there are new members coming or old members leaving (Yu and Liu 2007), and 3) household dissolution when there are no members left. Each household has a specific location in the landscape and its behavior is based on its attributes, including household size, number of laborers, area of cropland, and whether it is a tourism or a labor migration household. Household behavior is also constrained by environmental conditions like elevation and distance to the main road.
Our demographic submodel was largely adopted from the models developed in previous studies (An et al. 2002; Chen et al. 2014; An et al. 2003, 2005) and was initialized with data from an agricultural census 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 the household’s cropland. In 1996, there were 4053 residents in Wolong distributed in 892 households. The geocoded locations of households were measured with a GPS receiver in 2002 (An et al. 2002). Details about the modeling of the events of household and person agents can be found in the cited studies (An et al. 2002; Chen et al. 2014; An et al. 2003, 2005, 2006).
2) Telecoupling submodel
The telecoupling submodel was designed to simulate the interactions between households and forest under five different telecoupling scenarios (see section 1). Telecouplings affect the households’ status of whether having member(s) work in the local tourism industry or outmigrate to work in cities. This status in turn affects the amount of fuelwood collected by households and forest dynamics. We estimated the probabilities of becoming a tourism household or a labor migration household using household survey data collected in 1999 and 2006.
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 households in Wolong) were randomly selected for survey with strata based on administrative groups (the smallest administrative unit in China). These households sampled 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 during the survey period. In 1998, tourism households and labor migration households accounted for 2.7% and 3.9% of all households, respectively. In 2005, those figures increased to 31% and 22%, respectively.
Using the survey data of the 202 randomly sampled households in 2006, we modeled local households’ participation in tourism and labor migration using logistic regression models (Table 1). When modeling the participation in labor migration, we included tourism participation status (1, yes; 0, no) as a predictor as suggested by a previous study in Wolong (Yang et al. 2018). This is because local tourism jobs in Wolong are often more attractive than migrant jobs in cities. Rural migrant workers in cities may lack health insurance coverage, face substantial educational expenses for their children, and suffer from high stress and depression (Cuccia and Rizzo 2011). Therefore, if a household has access to jobs in the local tourism industry, it is less likely to have labor migrants working in cities. Our participation models (Table 1) predict the probabilities of tourism and labor migration households after the development of tourism and labor migration for multiple years. Because only a few households were tourism households or labor migration households in 1998, we approximated the annual probabilities of becoming a tourism or labor migration household by dividing the estimated probabilities by seven years (1998–2005). A higher participation probability of a household indicates it has a larger potential to have one or more members to work in local tourism industry or outmigrate to work in cities.
Logistic model estimations of labor migration and tourism participation by households in Wolong. Significance: plus sign indicates p < 0.1 and one, two, and three asterisks indicate p < 0.05, p < 0.01, and p < 0.001, respectively.
In reality, tourism or labor migration households may stop their involvement in those activities for various reasons (e.g., the laborers in the household are getting too old). However, almost all tourism and labor migration households in 1998 remained the same in 2005. We therefore did not have enough observations to develop empirical models to predict the probability of a household exits the status of being a tourism household or labor migration household. In our agent-based model, we used the minimum predicted probability of all the 63 tourism households surveyed in 2006 (0.06) as the threshold below which a tourism household exits the status of being a tourism household. As time goes by, a tourism household’s attributes may change and have a predicted probability less than this threshold. If this happens, the household’s status changes from being a tourism household to a nontourism household. Similarly, we determined the threshold (0.01) for labor migration households to exit the status of being a labor migration household.
Fuelwood collection by each household without considering impacts of tourism, labor migration, and conservation policy, was determined according to a previous study in Wolong (An et al. 2001), which modeled fuelwood collection as a function of household size, presence or absence of senior people in the household, and farmland area. Because all households in the study area enrolled in the NFCP in 2001, we did not have a control group of households to accurately estimate the impact of the NFCP on fuelwood collection. We approximated this impact using the drastic reduction in average household fuelwood collection that occurred after 2001 when the NFCP started. Of the 220 households surveyed in 1998, 189, 200, and 215 of them were revisited in 2001, 2002, and 2003 with their fuelwood collection information recorded. Before 2002, the average fuelwood collection by each household was around 12 861.5 kg (12 763 kg in 1998 and 12 960 kg in 2001). In 2002, the average fuelwood collection drastically reduced to be around 8618.1 kg (8576.5 kg in 2002 and 8659.7 kg in 2003). We used the difference in the mean fuelwood collections before and after 2001, 4243.4 kg, as the impact of the NFCP on annual household fuelwood collection. In our simulations, this impact on households’ fuelwood collection takes effect after 2001.
According to the results of a previous study in Wolong (Chen et al. 2012a), the impact of labor migration on household fuelwood collection was set to be 1827 kg per year. If a household starts to have member(s) outmigrate to work in cities, we deducted its fuelwood collection by this amount. We estimated the impact of tourism participation on household fuelwood collection by comparing the fuelwood collection of tourism and nontourism households in 2005 using the matching approach (Rubin 1973). For each tourism household, the matching approach finds a counterpart nontourism household with similar attributes, including the number of adults, household size, distance to the main road, and maximum education level of adult household members. On average, a tourism household collected 1708 kg less fuelwood than a nontourism household per year (Table 2). Therefore, if a nontourism household in our model changed to be a tourism household, we deducted its annual fuelwood collection by 1708 kg.
The impact of tourism participation on household fuelwood collection estimated using the matching approach. A superscript of three asterisks indicates statistical significance at the 0.001 level.
The decision process of each household’s status—become or stop being a tourism or a labor migration household—over time is summarized in Fig. 3. If a household is not a tourism household, we calculated its probability to be a tourism household based on its attributes at this time step using the logistic model in Table 1. If a household is already a tourism household at the current time step, we compared its participation probability with the threshold (0.06) to judge whether it is still eligible to be a tourism household. If the tourism household’s participation probability is less than the threshold, its status will be changed from being a tourism household to be a nontourism household. Similarly, if a household is not a labor migration household, we determined whether it can become a labor migration household from its probability to have labor migrants. If a household has already been a labor migration household, we evaluated its eligibility at this time step by comparing its participation probability with the threshold probability (0.01). Only labor migration households with predicted probabilities larger than this threshold can maintain their status of being labor migration households. Households’ statuses for labor migration and tourism participation were then used to calculate their fuelwood collection.
Decision process for each household’s status of being a tourism or a labor migration household at each simulation step and their effects on fuelwood collection.
Citation: Earth Interactions 26, 1; 10.1175/EI-D-20-0029.1
3) Landscape submodel
The landscape submodel simulates forest dynamics with specific consideration of household fuelwood collection, establishment of new households, and other environmental conditions (e.g., elevation and slope). Our simulation focuses on a 6-km buffer region around all households (Fig. 1) because almost all deforestation activities in the study area happened within the distance of 6 km from the households (Linderman et al. 2005a). The total area of the simulated natural landscape is 553 km2. The landscape is represented in our model as a digital “world” consisting of 90 m × 90 m cells. Each cell has a set of attributes including elevation, slope, aspect, and forest status (forest or nonforest). The elevation, 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 classification of the satellite images was performed using unsupervised digital classification based on the iterative self-organizing data analysis technique algorithm (ISODATA) technique (Jensen and Lulla 1987) and was validated using ground-truthing data (Linderman et al. 2005b). The accuracy of the forest cover map is about 80% (An et al. 2005; Liu et al. 2001).
Summary of the logistic model estimations of forest gain and forest loss from the previous study in Wolong (Chen et al. 2014). Significance: A superscript of three asterisks indicates p < 0.001. Because these models were built on the basis of observed forest change for six years (1994–2000), the annual forest change (gain or loss) probabilities of landscape cells are the estimated probabilities using the models divided by 6 as suggested by (Chen et al. 2014).
At every time step, we calculated the deforestation probability for each forest cell and recovery probability for each nonforest cell to determine 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).
c. Model validation
In this study, we validated the agent-based model by comparing the simulated landscape, demography, 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 mean population size and mean number of households with those 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 observed 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 observed 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 empirical 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. The predicted probability of being forest of the sample pixels was calculated by averaging simulated binary forest maps of the 20 runs. We used the area under the ROC curve (AUC) as a measure of the overall accuracy of the simulated forest maps. The values of AUC range from 0 to 1, where a value of 1 indicates perfect accuracy and a value of 0.5 implies that the accuracy is no better than a random guess (Araújo et al. 2005).
d. Simulation experiments
After validating our model, we simulated the dynamics of households and forest under five different scenarios to evaluate the effects of tourism and labor migration: 1) without tourism and labor migration, 2) only with tourism, 3) only with labor migration, 4) with both tourism and labor migration but ignoring their interaction, and 5) with both tourism and labor migration including their interaction effect. When running scenario 1, we ignored the impacts of tourism and labor migration by setting all households’ probabilities being tourism and labor migration households to be zero throughout the simulations. When running scenario 2, we ignored labor migration by setting the probability of labor migration to be zero for all households. Similarly, when running scenario 3, we set tourism participation probability to be zero for all households. When running scenario 4, we ignored the interaction between tourism and labor migration by setting the coefficient of the negative impact of tourism participation on the probability of labor migration to be zero. When running scenario 5, tourism, labor migration, and their interaction all took effect. In these simulations, the numbers of tourism and labor migration households in 1996 were assumed to be zero. We ran all simulations for 34 years (from 1996 to 2030). Because the landscape submodel was calibrated using the 1997 forest cover map, it started running one year later than the demographic and telecoupling submodels.
3. Results
a. Model validation results
Our validation results (Table 4) indicate our model performs accurately. The difference between the mean predicted human population and observed human population in 2006 was 17, which was less than the observed mean yearly population change (45 yr−1) from 1996 to 2006. The predicted number of all households was 1176, which was 21 less than the observed value (n = 1197) and the difference is less than the mean annual change (31 yr−1). The predicted percentages of tourism households (28.9%) and labor migration households (22.2%) were close to their observed values (31.2% and 21.7%, respectively) in 2005. The differences between observed percentages of tourism and labor migration households in 2005 and simulated means of them (2.3% and 0.5%) were all less than the observed mean yearly changes (3.1% yr−1 and 2% yr−1) from 1998 to 2005. The example simulated forest map in 2007 was also close to the empirical forest cover map (Fig. 4). The AUC value of the simulated maps (n = 20) is 0.743, indicating good simulation accuracy.
Comparison of simulated and observed forest distribution in 2007: (a) simulated forest cover in 2007, (b) observed forest cover in 2007 derived from classification of Landsat imagery, and (c) confusion matrix that shows the consistency between the simulated and observed forest cover based on results of 5000 random pixels. The overall accuracy value is 0.72.
Citation: Earth Interactions 26, 1; 10.1175/EI-D-20-0029.1
Comparisons of model predictions of population size, numbers of all households, tourism households, and labor migration households with observed values. The observed mean yearly change is calculated by dividing the observed change by the number of years between the observations.
b. Forest and household dynamics under different scenarios
As expected, both tourism and labor migration have contributed to the forest recovery that occurred after 2001 (Fig. 5). In all the five simulation scenarios, the total forest area decreased between 1996 and 2001, and then started to recover at a gradually decreasing rate. Under the scenario without considering the effects of tourism and labor migration (scenario 1), the predicted forest area in 2030 is 361.2 km2. Under the scenario only with tourism (scenario 2) or only with labor migration (scenario 3), the forest areas in 2030 are 387.4 and 385.1 km2, respectively. The difference in the 2030 forest area between scenario 1 and scenario 2 is 26.2 km2, which represents the cumulative effect of tourism development on forest dynamics throughout our simulation period (1996–2030). The difference in the 2030 forest area between scenario 1 and scenario 3 is 23.9 km2, which represents the cumulative effect of labor migration on forest dynamics from 1996 to 2030.
Dynamics of the forest areas simulated under five different scenarios: 1) without tourism and labor migration, 2) with tourism only, 3) with labor migration only, 4) with both tourism and labor migration but without considering their interaction, and 5) with both tourism and labor migration and with considering their interactions. Lines and corresponding shading represent the means and standard deviations of the results from 20 runs, respectively.
Citation: Earth Interactions 26, 1; 10.1175/EI-D-20-0029.1
The development of tourism reduced the number of labor migration households by 22% (Fig. 6). Under the scenario that did not consider the negative impact of tourism participation on the probability of labor migration (scenario 4), the number of labor migration households in 2030 is predicted at 675 (42.2% of the total), while under the scenario that considered this negative impact (scenario 5), the number of labor migration household in 2030 is 554 (34.6% of the total) (Fig. 6). The difference in the number of labor migration households in 2030 under scenario 4 and scenario 5 is 121, which represents the cumulative effect of tourism development on growth of labor migration throughout our simulation period from 1996 to 2030.
Simulated dynamics of the numbers of labor migration households from 1996 to 2030 under scenarios with (scenario 5) and without (scenario 4) considering the negative impact of tourism participation on the probability of labor migration. Lines and corresponding shading represent the means and standard deviations of the results from 20 runs, respectively.
Citation: Earth Interactions 26, 1; 10.1175/EI-D-20-0029.1
This interaction between tourism and labor migration has an evident impact on forest dynamics (Fig. 5). Under the scenario with both tourism and labor migration but without considering their interaction (scenario 4), the forest area in 2030 is 407.1 km2, which is 3.5 km2 higher than that under the scenario that considered this interaction effect (scenario 5). In short, development of tourism decreases labor migration in the area, which subsequently negatively affects forest cover.
4. Conclusions and discussion
Our agent-based model provides an efficient way to integrate the information from empirical statistic models and other sources to evaluate the impacts of different telecouplings on environmental outcomes at the landscape level over a long period of time. Our results demonstrated that telecouplings can interact and generated evident impact on the forest dynamics in rural areas. By analyzing labor migration and tourism in tandem, we show that the interaction between these two telecouplings significantly attenuate their positive impact on forest recovery across the landscape. While both tourism and migration increase forested area, their interaction results in a lower forest gain. Using the Wolong case study, we argue that potentially related telecouplings should be evaluated jointly rather than separately to reveal their actual effects on socioeconomic and environmental outcomes.
We note that our estimation of tourism’s long-term effect on labor migration may be conservative. This is because we only observed the influence of tourism on individuals staying in the area rather than migrating to cities. Therefore, we did not include the potential effect of tourism on labor migration by attracting labor migrants to come back to only work in the local tourism industry. We hypothesize that, with the inclusion of this attraction effect of tourism on labor migration, the reduction of reforestation would be even more pronounced relative to the results in Fig. 5. We did not observe this effect and include it in our current agent-based model perhaps because the tourism development in Wolong was at its early stage and this attraction effect had not been evident yet. As the tourism industry is recovering from the impact of the Wenchuan earthquake in 2008, future studies in Wolong and other places should also evaluate and consider this negative impact in their analyses.
Results from this study have important implications for management of tourism and labor migration. For example, like Wolong, many rural areas implemented tourism development programs with substantial investment and support from governments (Zhao et al. 2021; Yang et al. 2021). To maximize the efficiency of tourism development programs in providing environmental benefits, these programs may target rural areas where the level of labor migration is low to avoid limiting the positive environmental effect of labor migration. On the other hand, labor migration policies in urban settings may play an important role in mitigating the negative effect of tourism on labor migration. This negative impact occurs mainly because labor migrants in cities often have to confront many difficulties (Li 2011). Therefore, management interventions that help overcome these hardships (e.g., offering equal job opportunities for migrant workers) should be considered to increase the benefits labor migrants could obtain from this off-farm livelihood. The increase in benefits farmers could obtain from labor migration may promote tourism households to also have labor migrants and enhance the labor shift from on-farm to off-farm activities.
In our model, we only considered the impact of tourism on forest through reducing fuelwood collection because tourism development in Wolong remained at its early stage and did not generate other evident impacts on forest (W. Liu et al. 2016; Liu 2012). Although nature-based tourism is widely perceived to be clean and nonconsumptive because it relies on existing natural, cultural, and historical resources, unregulated tourism development can cause serious degradation of ecosystems (Dai et al. 2012). The actual impacts of future tourism development in Wolong on forest dynamics will depend on its design and management. Besides avoiding direct disturbances into the forest (e.g., clearing forest for tourism infrastructure development), we suggest that future development of tourism should also increase the share of benefit local households could obtain from it. Economic leakage (i.e., tourism revenue flowing to outside investors or managers rather than locals) is a common issue that plagues the development of tourism in many rural areas around the world (Kiss 2004). Previous studies (He et al. 2008; Liu 2012) in Wolong also found that only a small fraction of revenue from tourism development (<5%) went to the local community. This issue may have constrained the impact of tourism participation on fuelwood collection because less income would be available for local households to afford the energy shift from fuelwood to cleaner energy like electricity.
Although the telecoupling interaction illustrated in this study is antagonistic, that is, one telecoupling weakens the other, synergetic interactions also commonly exist among telecouplings. For example, the panda loan is another important telecoupling linking Wolong and other places (Liu et al. 2015a). Every year the China Conservation and Research Center for the Giant Panda, a panda-breeding center and tourism site in Wolong, loans captive pandas to zoos inside and outside China. The panda loans have significantly increased the media exposure of Wolong. For example, around 20% of all media reports found in LexisNexis about Wolong are related to panda loans (Liu et al. 2015a). The spread of information about Wolong may have in turn boosted the tourist visits to Wolong. About 24% of visitors to the Wolong breeding center in 2005 expressed that they had previously read media reports on Wolong and 29% of them saw television program about Wolong before the visit (Liu et al. 2015a). This indicates that a synergetic interaction effect may exist between the telecouplings of panda loan and tourism. Currently, neither synergetic nor antagonistic interactions among telecouplings have been well studied (Liu et al. 2013a; Kapsar et al. 2019). They deserve more investigations in the future to improve the understandings of the dynamics of telecouplings and their effects on socioeconomic and ecological outcomes.
Our study also illustrates that agent-based models are useful tools to understand interrelated effects of telecouplings. Human–nature interactions are often complex and vary across different settings (Liu et al. 2007, 2013a). Agent-based models provide flexible tools to effectively integrate empirical knowledge, findings, and data from different sources to characterize the heterogeneities and interactions of the human and natural components in a CHANS. This lays a foundation to understand dynamics of human–nature interactions under telecouplings across space and time. With a validated agent-based model, we can further explore the trajectories of the system dynamics under different telecoupling scenarios that cannot be observed empirically.
We note that our model mainly focuses on simulating the processes operating within Wolong. We did not specifically consider factors associated with tourism and labor migration in other places because our study aims to understand how labor migration and tourism jointly affect forest dynamics in Wolong (and not other places). We framed our study as a telecoupling research to underscore the fact that Wolong is telecoupled with other systems, and our study highlights that telecoupling flows can affect forest covers at the landscape level through influencing agent behaviors. Future research can build upon our study to include other systems and answer other questions. For instance, the nature-based tourism in Wolong may shape urban sustainability through affecting the environmental awareness and behaviors of tourists from cities. A future study may integrate such results with our findings under the framework of telecoupling and assess the possible synergy between sustainability in Wolong and cities.
Like all other models, the agent-based model is a simplified representation of the real world. For example, some of the life history events of person agents in our model such as death, childbirth, and marriage were simplified as stochastic processes. However, modeling the key dynamic interactions using the agent-based model helps us to improve the understanding of the complexities of long-term effects of telecouplings (e.g., nonlinearity; Figs. 5 and 6). We hope that the perspectives and methods proposed in this study can be useful for investigating the effects of telecouplings in Wolong and other CHANS around the world. With improved understanding of telecouplings, policy makers and scientists may be able to develop effective strategies to manage telecouplings for maximizing their positive effects and mitigating their negative effects in an increasingly telecoupled Anthropocene.
Acknowledgments.
We thank all of the students and postdoctoral associates as well as collaborators and households who collected and helped to collect the data used in this study. This work was supported by funding from the Smithsonian Institution, U.S. National Science Foundation (Grant 1924111, SMA 1416730); Michigan AgBioResearch, Michigan State University; the Key Laboratory of Southwest China Wildlife Resources Conservation (Grant XNYB19-01), and the National Natural Science Foundation of China (Grants 41571517 and 42001228). Any opinions, findings, and conclusions or recommendations expressed in this paper are those of the authors(s) and do not necessarily reflect the views of the funding agencies.
REFERENCES
An, L., and J. G. Liu, 2010: Long-term effects of family planning and other determinants of fertility on population and environment: Agent-based modeling evidence from Wolong Nature Reserve, China. Popul. Environ., 31, 427–459, https://doi.org/10.1007/s11111-010-0111-3.
An, L., J. G. Liu, Z. Y. Ouyang, M. Linderman, S. Q. Zhou, and H. M. Zhang, 2001: Simulating demographic and socioeconomic processes on household level and implications for giant panda habitats. Ecol. Modell., 140, 31–49, https://doi.org/10.1016/S0304-3800(01)00267-8.
An, L., F. Lupi, J. G. Liu, M. A. Linderman, and J. Y. Huang, 2002: Modeling the choice to switch from fuelwood to electricity: Implications for giant panda habitat conservation. Ecol. Econ., 42, 445–457, https://doi.org/10.1016/S0921-8009(02)00126-X.
An, L., A. G. Mertig, and J. G. Liu, 2003: Adolescents leaving parental home: Psychosocial correlates and implications for conservation. Popul. Environ., 24, 415–444, https://doi.org/10.1023/A:1023694924954.
An, L., M. Linderman, J. Qi, A. Shortridge, and J. Liu, 2005: Exploring complexity in a human-environment system: An agent-based spatial model for multidisciplinary and multiscale integration. Ann. Assoc. Amer. Geogr., 95, 54–79, https://doi.org/10.1111/j.1467-8306.2005.00450.x.
An, L., G. He, Z. Liang, and J. Liu, 2006: Impacts of demographic and socioeconomic factors on spatio-temporal dynamics of panda habitat. Biodiversity Conserv., 15, 2343–2363, https://doi.org/10.1007/s10531-004-1064-6.
An, L., A. Zvoleff, J. Liu, and W. Axinn, 2014: Agent-based modeling in coupled human and natural systems (CHANS): Lessons from a comparative analysis. Ann. Assoc. Amer. Geogr., 104, 723–745, https://doi.org/10.1080/00045608.2014.910085.
Araújo, M. B., R. G. Pearson, W. Thuiller, and M. Erhard, 2005: Validation of species–climate impact models under climate change. Global Change Biol., 11, 1504–1513, https://doi.org/10.1111/j.1365-2486.2005.01000.x.
Bearer, S., M. Linderman, J. Y. Huang, L. An, G. M. He, and J. Q. Liu, 2008: Effects of fuelwood collection and timber harvesting on giant panda habitat use. Biol. Conserv., 141, 385–393, https://doi.org/10.1016/j.biocon.2007.10.009.
Bradbury, M., M. N. Peterson, and J. G. Liu, 2014: Long-term dynamics of household size and their environmental implications. Popul. Environ., 36, 73–84, https://doi.org/10.1007/s11111-014-0203-6.
Cao, S. X., B. L. Zhong, H. Yue, H. S. Zeng, and J. H. Zeng, 2009: Development and testing of a sustainable environmental restoration policy on eradicating the poverty trap in China’s Changting County. Proc. Natl. Acad. Sci. USA, 106, 10 712–10 716, https://doi.org/10.1073/pnas.0900197106.
Chen, X., K. A. Frank, T. Dietz, and J. Liu, 2012a: Weak ties, labor migration, and environmental impacts: Toward a sociology of sustainability. Organ. Environ., 25, 3–24, https://doi.org/10.1177/1086026611436216.
Chen, X., F. Lupi, L. An, R. Sheely, A. Viña, and J. G. Liu, 2012b: Agent-based modeling of the effects of social norms on enrollment in payments for ecosystem services. Ecol. Modell., 229, 16–24, https://doi.org/10.1016/j.ecolmodel.2011.06.007.
Chen, X., A. Viña, A. Shortridge, L. An, and J. Liu, 2014: Assessing the effectiveness of payments for ecosystem services: An agent-based modeling approach. Ecol. Soc., 19, 7, https://doi.org/10.5751/ES-05578-190107.
China Ministry of Forestry and World Wildlife Fund, 1989: Conservation and management plan for giant pandas and their habitat. WWF Rep., 157 pp.
Chung, M. G., T. Dietz, and J. G. Liu, 2018: Global relationships between biodiversity and nature-based tourism in protected areas. Ecosyst. Serv., 34A, 11–23, https://doi.org/10.1016/j.ecoser.2018.09.004.
Cuccia, T., and I. Rizzo, 2011: Tourism seasonality in cultural destinations: Empirical evidence from Sicily. Tour. Manage., 32, 589–595, https://doi.org/10.1016/j.tourman.2010.05.008.
Dai, L., Y. Wang, B. J. Lewis, D. Xu, L. Zhou, X. Gu, and L. Jiang, 2012: The trend of land-use sustainability around the Changbai Mountain Biosphere Reserve in northeastern China: 1977–2007. Int. J. Sustainable Dev. World Ecol., 19, 369–377, https://doi.org/10.1080/13504509.2012.675599.
Dou, Y., R. F. B. Silva, H. Yang, and J. Liu, 2018: Spillover effect offsets the conservation effort in the Amazon. J. Geogr. Sci., 28, 1715–1732, https://doi.org/10.1007/s11442-018-1539-0.
Dou, Y., and Coauthors, 2019: Land-use changes across distant places: Design of a telecoupled agent-based model. J. Land Use Sci., 14, 191–209, https://doi.org/10.1080/1747423X.2019.1687769.
Dou, Y. , and Coauthors, 2020: Land-use changes in distant places: Implementation of a telecoupled agent-based model. J Artif. Soc. Soc. Simul., 23, 11, https://doi.org/10.18564/jasss.4211.
Fan, C. C., 2008: Migration, hukou, and the city. China Urbanizes: Consequences, Strategies, and Policies, World Bank, 65–89.
Fang, B., and Coauthors, 2016: Energy sustainability under the framework of telecoupling. Energy, 106, 253–259, https://doi.org/10.1016/j.energy.2016.03.055.
Fox, J., 2016: Community forestry, labor migration and agrarian change in a Nepali village: 1980 to 2010. J. Peasant Stud., 45, 610–629, https://doi.org/10.1080/03066150.2016.1246436.
Hanley, J. A., and B. J. McNeil, 1982: The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology, 143, 29–36, https://doi.org/10.1148/radiology.143.1.7063747.
He, G., and Coauthors, 2008: Distribution of economic benefits from ecotourism: A case study of Wolong Nature Reserve for giant pandas in China. Environ. Manage., 42, 1017–1025, https://doi.org/10.1007/s00267-008-9214-3.
Jensen, J. R., and K. Lulla, 1987: Introductory digital image processing: A remote sensing perspective. Geocarto Int., 2, 65, https://doi.org/10.1080/10106048709354084.
Kapsar, K. E. , and Coauthors, 2019: Telecoupling research: The first five years. Sustainability, 11, 1033, https://doi.org/10.3390/su11041033.
Kiss, A., 2004: Is community-based ecotourism a good use of biodiversity conservation funds? Trends Ecol. Evol., 19, 232–237, https://doi.org/10.1016/j.tree.2004.03.010.
Kramer, D. B., G. Urquhart, and K. Schmitt, 2009: Globalization and the connection of remote communities: A review of household effects and their biodiversity implications. Ecol. Econ., 68, 2897–2909, https://doi.org/10.1016/j.ecolecon.2009.06.026.
Lai, B., S. Zhou, H. Zhang, A. Luo, and Q. Luo, 2003: Development History of Wolong Nature Reserve. Sichuan Science and Technology Press, 162 pp.
Li, W., and N. Han, 2001: Ecotourism management in China’s nature reserves. Ambio, 30, 62–63, https://doi.org/10.1579/0044-7447-30.1.62.
Li, Y., 2011: Urban-rural interaction in China: Historic scenario and assessment. China Agric. Econ. Rev., 3, 335–349, https://doi.org/10.1108/17561371111165770.
Liang, Z., 2001: The age of migration in China. Popul. Dev. Rev., 27, 499–524, https://doi.org/10.1111/j.1728-4457.2001.00499.x.
Ligmann-Zielinska, A., and P. Jankowski, 2007: Agent-based models as laboratories for spatially explicit planning policies. Environ. Plann., 34B, 316–335, https://doi.org/10.1068/b32088.
Linderman, M. A., L. An, S. Bearer, G. M. He, Z. Y. Ouyang, and J. G. Liu, 2005a: Modeling the spatio-temporal dynamics and interactions of households, landscapes, and giant panda habitat. Ecol. Modell., 183, 47–65, https://doi.org/10.1016/j.ecolmodel.2004.07.026.
Linderman, M. A., S. Bearer, L. An, Y. Tan, Z. Ouyang, and J. G. Liu, 2005b: The effects of understory bamboo on broad-scale estimates of giant panda habitat. Biol. Conserv., 121, 383–390, https://doi.org/10.1016/j.biocon.2004.05.011.
Liu, J., 2014: Forest sustainability in China and implications for a telecoupled world. Asia Pac. Policy Stud., 1, 230–250, https://doi.org/10.1002/app5.17.
Liu, J., Z. Ouyang, W. W. Taylor, R. Groop, Y. Tan, and H. Zhang, 1999: A framework for evaluating the effects of human factors on wildlife habitat: The case of giant pandas. Conserv. Biol., 13, 1360–1370, https://doi.org/10.1046/j.1523-1739.1999.98418.x.
Liu, J., M. Linderman, Z. Y. Ouyang, L. An, J. Yang, and H. M. Zhang, 2001: Ecological degradation in protected areas: The case of Wolong Nature Reserve for giant pandas. Science, 292, 98–101, https://doi.org/10.1126/science.1058104.
Liu, J., and Coauthors, 2007: Complexity of coupled human and natural systems. Science, 317, 1513–1516, https://doi.org/10.1126/science.1144004.
Liu, J., S. X. Li, Z. Y. Ouyang, C. Tam, and X. D. Chen, 2008: Ecological and socioeconomic effects of China’s policies for ecosystem services. Proc. Natl. Acad. Sci. USA, 105, 9477–9482, https://doi.org/10.1073/pnas.0706436105.
Liu, J. , and Coauthors, 2013a: Framing sustainability in a telecoupled world. Ecol. Soc., 18, 26, https://doi.org/10.5751/ES-05873-180226.
Liu, J., W. McConnell, and J. Luo, 2013b: Wolong Household Study (China). Inter-University Consortium for Political and Social Research, accessed 12 January 2018, https://doi.org/10.3886/ICPSR34365.v1.
Liu, J. , and Coauthors, 2015a: Multiple telecouplings and their complex interrelationships. Ecol. Soc., 20, 44, https://doi.org/10.5751/ES-07868-200344.
Liu, J., and Coauthors, 2015b: Systems integration for global sustainability. Science, 347, 1258832, https://doi.org/10.1126/science.1258832.
Liu, J., V. Hull, W. Yang, A. Viña, X. Chen, Z. Ouyang, and H. Zhang, 2016: Pandas and People: Coupling Human and Natural Systems for Sustainability. Oxford University Press, 304 pp.
Liu, W., 2012: Patterns and impacts of tourism development in a coupled human and natural system. Ph.D. dissertation, Michigan State University, 176 pp.
Liu, W., C. A. Vogt, J. Luo, G. He, K. A. Frank, and J. Liu, 2012: Drivers and socioeconomic impacts of tourism participation in protected areas. PLOS ONE, 7, e35420, https://doi.org/10.1371/journal.pone.0035420.
Liu, W., C. A. Vogt, F. Lupi, G. He, Z. Ouyang, and J. Liu, 2016: Evolution of tourism in a flagship protected area of China. J. Sustainable Tour., 24, 203–226, https://doi.org/10.1080/09669582.2015.1071380.
Minar, N., R. Burkhart, C. Langton, and M. Askenazi, 1996: The swarm simulation system: A toolkit for building multi-agent simulations. Santa Fe Institute Working Paper 1996-06-042, 12 pp.
Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B. da Fonseca, and J. Kent, 2000: Biodiversity hotspots for conservation priorities. Nature, 403, 853–858, https://doi.org/10.1038/35002501.
Parker, A. J., 1982: The topographic relative moisture index: An approach to soil-moisture assessment in mountain terrain. Phys. Geogr., 3, 160–168, https://doi.org/10.1080/02723646.1982.10642224.
Pulido-Fernandez, J. I., D. Ward-Perkins, and O. V. Krukova, 2015: Nature-based tourism. Tourism in Russia: A Management Handbook, Emerald, 181–230.
Ratha, D., S. Yi, and S. R. Yousefi, 2015: Migration and development: The Asian experience. Routledge Handbook of Immigration and Refugee Studies, Routledge, 284–301.
Reid, D. G., and J. Hu, 1991: Giant panda selection between Bashania fangiana bamboo habitats in Wolong Reserve, Sichuan, China. J. Appl. Ecol., 28, 228–243, https://doi.org/10.2307/2404127.
Rubin, D. B., 1973: Matching to remove bias in observational studies. Biometrics, 29, 159–183, https://doi.org/10.2307/2529684.
Rush, A., 2011: China’s labour market. RBA Bull., September Quarter, Reserve Bank of Australia, Sydney, Australia, 29–38, https://www.rba.gov.au/publications/bulletin/2011/sep/pdf/bu-0911-4.pdf.
Schaller, G. B., J. Hu, W. Pan, and J. Zhu, 1985: The Giant Pandas of Wolong. University of Chicago Press, 298 pp.
Sichuan Forestry Administration, 2015: The Survey Report of the Fourth Giant Panda Census in Sichuan (in Chinese). Sichuan Science and Technology Press, 265 pp.
Taylor, A. H., and Z. S. Qin, 1993: Bamboo regeneration after flowering in the Wolong Giant Panda Reserve, China. Biol. Conserv., 63, 231–234, https://doi.org/10.1016/0006-3207(93)90717-F.
Tuanmu, M. N., A. Viña, W. Yang, X. D. Chen, A. M. Shortridge, and J. G. Liu, 2016: Effects of payments for ecosystem services on wildlife habitat recovery. Conserv. Biol., 30, 827–835, https://doi.org/10.1111/cobi.12669.
Viña, A., and Coauthors, 2007: Temporal changes in giant panda habitat connectivity across boundaries of Wolong Nature Reserve, China. Ecol. Appl., 17, 1019–1030, https://doi.org/10.1890/05-1288.
Viña, A., X. Chen, W. J. McConnell, W. Liu, W. Xu, Z. Ouyang, H. Zhang, and J. Liu, 2011: Effects of natural disasters on conservation policies: The case of the 2008 Wenchuan Earthquake, China. Ambio, 40, 274–284, https://doi.org/10.1007/s13280-010-0098-0.
Viña, A., W. J. McConnell, H. B. Yang, Z. C. Xu, and J. G. Liu, 2016: Effects of conservation policy on China’s forest recovery. Sci. Adv., 18, e1500965, https://doi.org/10.1126/sciadv.1500965.
Wong, D. F. K., C. Y. Li, and H. X. Song, 2007: Rural migrant workers in urban China: Living a marginalised life. Int. J. Soc. Welfare, 16, 32–40, https://doi.org/10.1111/j.1468-2397.2007.00475.x.
Yang, H., W. Yang, J. Zhang, T. Conner, and J. Liu, 2018: Revealing pathways from payments for ecosystem services to socioeconomic outcomes. Sci. Adv., 4, eaao6652, https://doi.org/10.1126/sciadv.aao6652.
Yang, H., Q. Y. Huang, J. D. Zhang, M. Songer, and J. G. Liu, 2021: Range-wide assessment of the impact of China’s nature reserves on giant panda habitat quality. Sci. Total Environ., 769, 145081, https://doi.org/10.1016/j.scitotenv.2021.145081.
Yang, W., 2013: Ecosystem services, human well-being, and policies in coupled human and natural systems. Ph.D. dissertation, Michigan State University, 265 pp.
Yang, W., W. Liu, A. Viña, J. Luo, G. He, Z. Ouyang, H. Zhang, and J. Liu, 2013: Performance and prospects of payments for ecosystem services programs: Evidence from China. J. Environ. Manage., 127, 86–95, https://doi.org/10.1016/j.jenvman.2013.04.019.
Yu, E., and J. G. Liu, 2007: Environmental impacts of divorce. Proc. Natl. Acad. Sci. USA, 104, 20 629–20 634, https://doi.org/10.1073/pnas.0707267104.
Zhang, J. D., T. Connor, H. B. Yang, Z. Y. Ouyang, S. X. Li, and J. G. Liu, 2018: Complex effects of natural disasters on protected areas through altering telecouplings. Ecol. Soc., 23, 17, https://doi.org/10.5751/ES-10238-230317.
Zhao, Z. Q., and Coauthors, 2021: Synergies and tradeoffs among Sustainable Development Goals across boundaries in a metacoupled world. Sci. Total Environ., 751, 141749, https://doi.org/10.1016/j.scitotenv.2020.141749.
