Unraveling human drivers behind complex interrelationships among sustainable development goals: a demonstration in a flagship protected area

. The transformational potential of the United Nations’ 2030 Agenda for Sustainable Development Goals (SDGs) lies in effective efforts to reconcile the conflicts and maximize the synergies among the interrelated SDGs. Previous research on the interrelationships among SDGs often focused on depicting the degree to which different goals reinforce or hamper each other; however, drivers behind these interrelationships have rarely been evaluated. We developed a novel approach to unraveling the impact of human activities on the complex trade-offs and synergies among SDGs. We used the approach to assess the impacts of four globally common livelihoods, including cropping, local off-farm labor work, labor migration, and livestock husbandry, on the interrelationships among SDG 1 (no poverty), SDG 3 (enhance human well-being), and SDG 15 (protect life on land) in a demonstration site. The results show that our approach can be very useful in informing coherent governance and facilitating progress toward SDGs across social, economic, and environmental dimensions simultaneously.


INTRODUCTION
To facilitate progress toward sustainability across social, economic, and environmental dimensions simultaneously, the member states of the United Nations adopted the 2030 Agenda for Sustainable Development in 2015, which aims to achieve 17 Sustainable Development Goals (SDGs) by 2030 (United Nations 2015). The 17 SDGs are "an integrated and indivisible whole" of global objectives designed to catalyze coherent governance and avoid "sustainability solutions" in one system causing deleterious effects in others (Colglazier 2015). But operationalizing the 2030 Agenda on the ground is far from straightforward. The grand challenges facing humanity, from poverty, water scarcity, and food insecurity to climate change and biodiversity loss, are closely intertwined (Griggs et al. 2017. The linked challenges, however, have often been managed in silos and critical interrelationships among them are largely ignored (Zhao et al. 2021), often with counterproductive consequences for sustainability (Fader et al. 2018, Wong andvan der Heijden 2019). For example, biofuels were proposed as part of the solution to CO 2 emissions from burning fossil fuels (Rulli et al. 2016). Despite its potential to mitigate climate change (promoting SDG 13), the biofuel approach in many cases unintentionally threatens biodiversity and increases water and food shortages (undermining SDG 2, SDG 6, and SDG 15) because a large amount of land and water was diverted for biofuel production (Renzaho et al. 2017, Pörtner et al. 2021).
Knowledge of the SDG interrelationships and drivers behind them is key to addressing such challenges (Guerry et al. 2015, Nilsson et al. 2016, 2018. Armed with such knowledge, policy makers may identify and strengthen actions that facilitate progress toward different goals simultaneously and avoid unintended trade-offs (Xu et al. 2020). However, the nature and strengths of the interrelationships among SDGs are largely context specific and rely on the development strategies chosen to pursue them (Moallemi et al. 2020). This makes the evaluation of interrelationships among SDGs and the drivers behind them challenging (Tosun and Leininger 2017) and encourages a growing call for understanding the complex interrelationships among different goals (McGowan et al. 2019).
Evaluations of SDG interrelationships are still nascent (Fu et al. 2019). Previous studies about SDG interrelationships often focused on the pattern of the relationships using expert knowledge and syntheses of the literature. For example, Nilsson et al. (2016) proposed a rating system to depict the extent to which different SDGs are linked to each other based on expert knowledge of the possible influence of gain in one goal on the other goals. Tosun and Leininger et al. (2017) evaluated the linkages among SDGs based on overlaps among the descriptions of the 169 specific targets of the 17 SDGs. McGowan et al. (2019) used a formal systems analysis approach and quantitatively assessed the relationships among the 17 SDGs, also based on expert knowledge. Certainly, it is important to characterize patterns of the interrelationships among SDGs. But to understand why these linkages occur is also essential; that is we have to understand the drivers of the SDGs to identify where common drivers lead to the emergence of positive or negative linkages across SDGs. Such linkages can facilitate or hamper achieving the goals as a whole. A poor understanding of the drivers can often result in puzzling SDG interrelationships that beg for explanations (McGowan et al. 2019) and failures in identifying ways to maximize the reinforcing relationships among the goals and minimize the conflicting ones.
We offer an analysis of how common forms of human livelihoods impact several SDGs, thus showing how human activities drive the emergence of trade-offs and synergies across SDGs in a particular context, the Wolong Nature Reserve (Wolong hereafter;. Our results help explain important dynamics in the system we study. The four livelihoods we consider, (1) cropping (Linderman et al. 2005a), (2) local off-farm labor work (Zhang et al. 2018), (3) labor migration (temporary outmigration to work in cities, [Yang 2018]), and (4) livestock husbandry (Wang et al. 2021), are very common across the globe (Carter et al. 2014. Because of this commonality, the kinds of interactions we find in our study are also likely to occur beyond our study area. We consider three SDGs: SDG 1 (no poverty), SDG 3 (enhance human well-being), and SDG 15 (protect life on land; see Indicators for SDGs). On the basis of the findings from our study site, we provide suggestions on how to harmonize the livelihood impacts on the SDGs in Wolong. Although details will vary across contexts, we hope our methods and results can serve as a working model that can be tailored for studying drivers behind complex SDG interrelationships in other contexts (e.g., Carter et al. 2015).

Conceptual framework
Our approach is an example application of existing frameworks for studying human-nature interactions as coupled human and natural systems (CHANS; Liu et al. 2007) and metacoupling (Liu 2017). We conceptualize a place as a CHANS in which humans interact with nature ( Fig. 1). In the CHANS, we focus on three interdependent components that are important for the coherent management of the SDGs: human activities, SDG interrelationships, and policy making and governance (Fig. 1). Human activities are various actions performed by people to meet their needs based on assets available to them: financial, natural, human, physical, and social resources (Scoones 2009, Dietz 2015. Each of the human activities causes a set of impacts across the SDGs, which shape the interrelationships among the goals. Understanding how SDG interrelationships (e.g., synergies and trade-offs) emerge as a result of human activities can inform policy making and governance to regulate human activities so as to achieve progress toward different goals simultaneously (Fig.  1).
The human activities and their impacts on SDG interrelationships that occur in the focal location are not happening in isolation. Different places are increasingly connected by the flow of information, energy, people, organisms, and capital (Liu 2017). As a result, the interactions among human activities, SDGs, and governance within the focal CHANS (intracoupling; Liu 2017) are often affected by the interactions between the focal system and adjacent or distant systems (intercoupling; Liu 2017). Understanding of the intercouplings between the focal system and others helps us to account for the influences of socioeconomic and environmental interactions over distances (e.g., international trade) on the interactions within the system , Dou et al. 2020, Zhao et al. 2020, Tromboni et al. 2021; Fig. 1).

Quantifying the effects of human activities on SDG interrelationships
Interrelationships among SDGs can be viewed as an assembly of bilateral linkages among the goals. The bilateral linkages describe the extent to which the progress towards one goal may be Fig. 1. The conceptual framework for unraveling the human drivers behind interrelationships among SDGs. Different human activities generate varying effects on SDGs and shape the interrelationships among the goals. An improved understanding of the impacts of human activities on SDG interrelationships informs policymaking and governance, which in turn regulate the human activities to minimize conflicts among goals. This interaction process within focal coupled human and natural systems (intracoupling) affects, and is affected by, adjacent or distant human-nature systems through intercouplings.
positively or negatively associated with the progress towards the other. Human activity often impacts more than one goal and shapes the linkages across these goals. For example, transitioning agricultural land for biofuel production could positively impact clean energy provision (SDG 7) while negatively affecting food security (SDG 2) and biodiversity (SDG 15;Renzaho et al. 2017, Pörtner et al. 2021), contributing to a trade-off linkage between SDG 7 and SDG 2 as well as between SDG 7 and SDG 15. We propose a five-point scale to score the effect of a human activity on the bilateral linkage between a pair of goals to reflect the capability of the activity to simultaneously facilitate the goals. Because the impacts of an activity on different goals (e.g., impact on poverty and impact on wildlife conservation) are often not comparable, this scoring is based on the nature of the impacts (positive, negative, or neutral [no significant effect]) of human activity on the goals and does not consider the magnitude of the impacts. The possible combination of the nature of the impacts on a pair of goals can be positive-positive, positive-neutral, positive-negative, neutral-neutral, negative-neutral, and negativenegative. Those six different combinations correspond to six different types of effects of an activity on the bilateral linkage between a pair of goals: synergy (positive effect on both goals), gain-no change (positive effect on one goal, neutral effect on the other), trade-off (positive effect on one goal, negative on the other), no change-no change (neutral effect on both), loss-no change (negative effect on one goal, neutral effect on the other), and loss-loss (negative effects on both goals; Fig. 2). It can be useful to score those six different types of effects on a bilateral Ecology and Society 27(3): 15 https://www.ecologyandsociety.org/vol27/iss3/art15/ Fig. 2. Illustration (left) and descriptions (right) of the six types of effects of human activity on a bilateral linkage between two Sustainable Development Goals (SDGs): synergy, gain-no change, trade-off, no change-no change, lossno change, and loss-loss. The type of the effect of a human activity on a linkage between two SDGs is defined by the impacts of human activity on the goals. We ranked and scored different types of effects on bilateral linkage based on their capabilities to facilitate the achievement of the two goals at the ends of the linkage simultaneously.
linkage ranging from the highest (synergy, scoring +2) to the lowest (loss-loss, scoring -2) to reflect their capacity to facilitate the achievement of the two goals at the ends of the linkage simultaneously (Fig. 2). This scoring system is different from the rating system of SDG interrelation proposed by Nilsson et al (2016). Our system aims to rate the effect of an activity on an interlinkage between two goals while Nilsson's system rates the linkage itself.
The overall impact of human activity on the interrelationships among multiple SDGs is viewed as the sum of its effects on each of the bilateral linkages among the goals. We measured the overall impact by a coherence index, which is calculated by summing the scores of its effects on all the bilateral linkages among the goals. The total number of bilateral linkages among N SDGs equals N×(N-1)/2. Therefore, the coherence index value of an activity on the interrelationships among N SDGs equals the sum of the scores of its effects on the N×(N-1)/2 bilateral linkages. For example, if the impacts of a human activity on the three linkages among three SDGs are trade-off (0), synergy (+2), and gain-no change (+1), respectively, its coherence index value would be +3 (0 + 2 +1). A higher coherence index value of a human activity indicates it has a larger capability to facilitate progress toward the goals as a unified whole.

EMPIRICAL ANALYSIS
We operationalized our approach in Wolong. In this "proof of concept" analysis, we evaluated the impacts of the four most important human activities that constitute the key local livelihood strategies: cropping, local off-farm labor work, labor migration, and livestock husbandry. We examined the impact of each livelihood on the interrelationships among SDG 1 (no poverty), SDG 3 (enhance human well-being), and SDG 15 (protect life on land) to demonstrate our approach. Those three SDGs represent major economic, social, and ecological sustainability goals in Wolong and are directly or indirectly related to the other 14 SDGs. We note that the four livelihoods evaluated in this demonstration study can have impacts on other SDGs. We did not evaluate those impacts on other goals because of data limitations.
Wolong is an ideal site for demonstrating our approach. The human community, ecosystems, and the interactions between them in Wolong form a prototypical CHANS , Yang et al. 2018b, 2018c. Wolong is a flagship protected area in Sichuan Province, Southwest China ( Fig. 3; Viña et al. 2008). It is designated primarily for the protection of giant pandas, an icon of global conservation, and an umbrella species whose habitats provide sanctuary for many other sympatric species (Linderman et al. 2005b, Li andPimm 2016). Besides rich biodiversity, Wolong is also home to 4933 human residents (Yang et al. 2018a). Like many other places, Wolong is confronting the sustainability challenge of balancing the needs for socioeconomic development and biodiversity conservation . Previous research there (e.g., Chen et al. 2012, Yang et al. 2013a, Yang et al. 2018a) has explicated some of the humanecosystem dynamics and thus provides background for our work.

Intercouplings and livelihoods in Wolong
Wolong is connected with other places via intercouplings that shape livelihoods and their impacts on interrelationships among SDGs (Fig. 3). First, households in Wolong have access to outside markets where they sell crops and livestock (e.g., cabbage, radish,

Fig. 3.
Wolong Nature Reserve (Wolong) connected with other places via intercouplings and giant panda occurrence probability at Hetaoping region in Wolong before and after livestock encroachment: (A) distribution of panda habitats at Hetaoping in Wolong; (B) the occurrence probability of giant pandas before the encroachment of livestock; and (C) the occurrence probability of giant pandas after the encroachment of livestock. The occurrence probability maps were adapted from . and sheep), a key part of local livelihoods based on cropping and livestock husbandry (Yang et al. 2013b. Second, as a flagship protected area, Wolong is a famous tourist destination and the site of many infrastructure investments from the central government. The flow of tourists and investments to Wolong provide off-farm labor work opportunities that contribute to the livelihoods of local residents (Liu et al. 2012, Yang et al. 2018a). Third, a growing number of households have members who out-migrate to cities for temporary jobs and send remittance back to Wolong communities, and this also influences our target SDGs (Chen et al. 2012, Yang et al. 2022).
We quantified household livelihood strategies using data from household surveys conducted from 2009 to 2014 (Appendix 1, A1.1-A1.2 and Fig. A1). We measured the local off-farm labor as the number of household members working in local off-farm sectors, cropping as the amount of cropland cultivated by the household, labor migration as the number of labor migrants in each household, and livestock husbandry as the number of livestock raised by each household. To make different types of livestock (sheep, yak, cattle, and horses) comparable, we followed methods from a previous study (Yang et al. 2018a) and measured the livestock number using the equivalent number of sheep based on the ratios of their average selling prices obtained from our survey.

Indicators for SDGs
Although the United Nations has developed 232 indicators for the SDGs (United Nations 2018), these indicators are mostly designed for measuring SDGs at the national level. Many of them can be difficult to operationalize at micro levels (e.g., for households). For example, some of the United Nations indicators measure the progress toward SDGs as an increase in government expenditure to achieve the goals, a logic that cannot be directly applied to our household-level analyses in Wolong. Therefore, we used three measurements appropriate to our analyses as the indicators for the three SDGs: annual household income (SDG 1), human well-being (SDG 3), and giant panda habitat suitability (SDG 15).
For each SDG, there are a couple of specific targets to reflect different dimensions of the SDG. The indicators we chose to measure for SDG 1, SDG 3, and SDG 15 can reflect progress to multiple targets of each goal. We chose annual household income as an indicator for SDG 1 because increasing household income can address or reflect progress in achieving specific targets of SDG 1, such as reducing the number of people living in poverty (Target 1.1, Target 1.2), and improving access to social protection systems and economic resources (Target 1.3, Target 1.4). The median annual household income in the community in 2014 is modest at 53,324 Yuan ($7465 as of 2014), so income increases can have important impacts on these targets. Human well-being is measured by a composite index constructed using a surveybased approach that has been used in several studies in rural communities (Yang et al. 2013a(Yang et al. , 2018a. The instrument (Table A1.1) is based on the human well-being framework proposed in the Millennium Ecosystem Assessment (Millennium Ecosystem Assessment 2005). It includes five dimensions: basic material for good life, security, health, good social relations, and freedom of choice and action (Millennium Ecosystem Assessment 2005). These dimensions match key targets of SDG 3, such as combating diseases (Target 3.3), reducing premature mortality (Target 3.4), and enhancing security (Target 3.6). The measure has been calculated for each household using data from our survey. More technical details regarding the construction, validation, and application of the index can be found in previous studies (Yang et al. 2013a(Yang et al. , 2018a.
We chose giant panda habitat suitability as an indicator for SDG 15 because giant panda is an umbrella species for protection of thousands of species sharing the same habitat (Li and Pimm 2016). Changes in giant panda habitat suitability can therefore serve as a barometer for important targets of SDG 15 in this region (Xu et al. 2017), such as the integrity of local ecosystems (Target 15.1, 15.4) and effectiveness of conservation management (Target 15.2, 15.5). We followed previous studies in Wolong (Yang et al. 2013c, Chen et al. 2014) and measured household influence on giant panda habitat suitability using the amount of fuelwood collected by each household because fuelwood collection is a major pathway by which local communities degrade panda habitat (Bearer et al. 2008). Of the four livelihoods considered, livestock husbandry may generate a sizable impact on panda habitat through an extra pathway in addition to affecting https://www.ecologyandsociety.org/vol27/iss3/art15/ household fuelwood collection. Previous studies in Wolong (Hull et al. 2014 and other nature reserves  show that livestock can have a major impact on giant panda habitat suitability by encroaching into core habitats, competing with pandas for space and food sources, and driving pandas from highly suitable habitats to less suitable areas. To capture this impact, we used changes in giant pandas' occurrence probability in their habitats before and after livestock encroachment at Hetaoping area in Wolong as the second indicator of the impact of livestock husbandry on giant panda habitat suitability. A decrease in occurrence probability of pandas in their core habitats after the encroachment of livestock indicates that livestock generated a negative impact on panda habitat suitability.

Quantifying the impacts of livelihoods on SDGs
To evaluate the impacts of different livelihoods on the three SDGs at household level, we compiled a panel dataset using socioeconomic information on households in 2009 and 2014 (Appendix 1, A1.1). In total, there were 186 households surveyed in both years. With the panel data, we constructed linear regression models to relate changes in annual household income, human well-being index, and fuelwood collection between 2009 and 2014 to changes in household livelihoods during the same period. To control for potential confounding effects, our models included some other socioeconomic and demographic factors that may affect changes in the three SDGs (Table A2.1). Similar to livelihood activities, some of these factors (e.g., number of laborers in a household) may change during the study period (2009 to 2014); therefore, we included variables measuring these socioeconomic and demographic conditions in 2009 and their changes between 2009 and 2014 in our models (Table A2.1). The general form of the models can be given as respectively; β 0 and β 1 are intercept and coefficient for the initial status of the SDG indicators, respectively; β 2 -β 5 are the vectors of other coefficients to be estimated; ε is the error term. In essence, this is a fixed-effect model where change in an SDG for the household is predicted based on initial SDG level and household characteristics and changes in household characteristics over time (Table A2.1). Standard regression diagnostics of linear regression assumptions revealed no concerns; variance inflation factors were all below 10. The sample size for our regression models is 186.
The size of the sample meets the rule of thumb of at least 10 observations per independent variable to detect reasonable-sized effects, but we acknowledge that it may not be large enough to detect small effects. Our study uses existing data to demonstrate the application of the proposed approach. Future research that includes new data collection could usefully deploy procedures, such as power analysis, to ensure the size of the sample is large enough to estimate target effect sizes. We performed the regression analyses in R (R Development Core Team 2020) using the package "rms" (Harrell Jr 2016).
To evaluate the impact of livestock encroachment on giant panda habitat suitability, we analyzed panda occurrence probability change in the Hetaoping area, which is a roughly 30 km² area in Wolong (Fig. 3), with more than 20 pandas living there . As a core habitat of giant pandas that has been invaded by livestock, Hetaoping is an ideal site to study the impact of livestock encroachment on habitat suitability. We obtained from a previous study the occurrence probability maps before (year 2012) and after (year 2014) the encroachment of livestock . To understand the giant panda occurrence probability change across the landscape, we mapped the distribution of giant panda habitat at Hetaoping using an integrated biophysical model that combines elevation, slope, and forest cover , Xu et al. 2017. The elevation and slope were derived from a 30 m SRTM digital elevation model. The forest cover was obtained from the digitization of a 0.65 m resolution Google Earth imagery in 2014.

Livelihood impacts on SDGs
Our regression results show that local off-farm labor work had significant positive effects on SDG 1 (no poverty), SDG 3 (enhance human well-being), and SDG 15 (protect life on land). The change in the number of laborers with off-farm jobs inside the reserve was positively related to changes in the human wellbeing index (  Labor migration positively affected SDG 1, but had little influence on SDG 3 and SDG 15. Change in labor migration was positively related to change in annual household income (Coef. = 0.53, p < 0.001, CI = [0.37, 0.69]). For households that had one more laborer who participated in labor migration, the percent increase in household income between 2009 and 2014 was higher than their counterparts by 69.7 percentage points. Change in labor migration was negatively related to change in fuelwood collection and human well-being, but neither of those coefficients were significant (p > 0.05).
The livestock husbandry impeded SDG 3, compromised SDG 15, and contributed little to SDG 1. Change in livestock husbandry was negatively associated with change in human well-being (Coef. = -0.0004, p < 0.05, CI = [-0.0008, -0.00005]). An additional livestock animal owned by a household would decrease household well-being index value by 0.0004. The association between livestock husbandry and annual household income was not statistically significant (p > 0.1; Table 1). Livestock husbandry did not show a significant influence on household fuelwood collection, but we found that the encroachment of livestock degraded the panda habitat (Fig. 3). After the encroachment of livestock into Hetaoping, the average panda occurrence probability decreased in suitable habitat area from 0.55 to 0.49 (p < 0.001, paired t-test) and increased in marginally suitable area from 0.26 to 0.32 (p < 0.001, paired t-test). This change pattern suggests that livestock encroachment had driven giant pandas from highly suitable habitats to marginally suitable habitats.

Livelihood impacts on SDG interrelationships
Based on the impacts on the three SDGs, we assessed each livelihood's effect on the interrelationships among SDG 1 (no poverty), SDG 3 (enhance human well-being), and SDG 15 (protect life on land; Fig. 4). Local off-farm work generated synergetic effects on all linkages among the three goals, with the highest coherence index value of +6. Labor migration caused two gain-no change effects and one no change-no change effect on the three linkages among the SDGs, with a coherence index value of +2. Cropping contributed little to the coherence among the SDGs. It led to two loss-no change and one no change-no change effect on the three linkages among the SDGs, with a coherence index value of -2. The most negative impact on the SDG interrelationships was found in livestock husbandry. It caused one loss-loss and two loss-no change effects on the linkages among the three SDGs, with a coherence index value of -4. Fig. 4. Summary of the livelihood impacts on Sustainable Development Goal (SDG) 1, SDG 3, and SDG 15, and on the interrelationships among them. The numbers in the parentheses next to the symbols for impacts on SDG bilateral linkage are the scores for the types of impact (Fig. 2). The total impact of each livelihood on the interrelationships among SDGs was measured by the coherence index (COI), which reflects the livelihood's capability to facilitate progress toward the three goals simultaneously.
These results suggest that local off-farm employment can produce positive synergies among the three SDGs we examined, as can, to a lesser degree, labor migration to cities that send remittances back. In contrast, livestock husbandry seems to cause negative linkages among the goals because it slows progress toward SDG 3 and SDG 15. Of course, these results are specific to the context in which we conducted our study. In other regions, the same livelihood strategies could have different effects on the SDGs and drive different linkages among them.

DISCUSSION
Understanding the effects of human livelihood activities on SDGs and their interrelationships can help policy makers identify obstacles to achieving progress toward the goals. For example, our results show that livestock husbandry in Wolong impeded the achievement of all the three SDGs. An important factor driving the rapid livestock expansion in Wolong is an incentive policy of the local government, that provides interest-free loans to households to raise more livestock (Hull et al. 2014. Despite the government's good intentions of boosting household income and well-being, our findings show that livestock expansion actually did the opposite. To make matters worse, free-roaming livestock encroached into giant panda habitat because the pasture land in Wolong was not sufficient to support the rapidly growing number of livestock (Hull et al. 2014. To avoid the continuation of the unexpected loss-loss impact on SDG 3 and SDG 15 and enhance the synergies among the goals, we suggest that livestock expansion should be discouraged rather than incentivized in Wolong. Because different places are increasingly connected by intercouplings (Liu 2017), factors beyond the local system can have important influences on SDG interrelationships. For example, our results show that labor migration positively affected household income but showed no effect on promoting household well-being. A possible reason for this is that labor migrants in cities often confront many hardships (e.g., poor education resources for their children; Qin 2010). Therefore, policies in cities that help to overcome the hardships confronting labor migrants (e.g., investing more to provide quality education to children of labor migrants) can be considered to help turn the effect of labor migration on human well-being to positive and diminish the counterproductive parts of the interrelationships among SDGs.
In different political, geographic, and temporal settings, the interactions among human activities, governance, and SDG interrelationships will likely be different. For example, in many other places, livestock is fed in fenced areas that is not important for wildlife and therefore may not cause much impact on wildlife habitat, generating different effects on SDG interrelationships. Another example is the positive impact of local off-farm jobs on wildlife habitats (SDG 15) in Wolong. Tourism development and infrastructure building, the main sources of off-farm jobs in Wolong, are strictly constrained within a small residential area (< 0.15% of the total area of the reserve) so that possible https://www.ecologyandsociety.org/vol27/iss3/art15/ detrimental impacts on wildlife habitats from tourism development and infrastructure building are limited. However, poorly planned tourism development and associated infrastructure building, noise pollution, and irresponsible behavior of tourists have generated substantial negative impacts on wildlife habitats in some cases (Zhong et al. 2011). Human activities and SDG interrelationships in the same place may also change across time. For example, a new road connecting Wolong to the outside was completed in 2016  and is expected to generate substantial impacts on the intercouplings that link Wolong and other places (e.g., attract more tourist visitations), change local livelihoods (e.g., more people involved in tourism-related businesses), and thus reshape the SDG interrelationships. More research is needed to understand the impact of human activities on SDG interrelationships across space and time, and to assist the design of effective efforts to achieve progress toward different SDGs simultaneously. The evaluation approach and demonstration study presented here lay a good foundation for conducting similar research across different spatial and temporal settings. Many methods and insights from our previous studies conducted in Wolong have been applied to many other countries around the world (e.g., Liu et al. 2007, An et al. 2014) and at different times (e.g., Tuanmu et al. 2011).
Compared to studies evaluating the impacts on a single sustainability goal, it is often more challenging to reveal the effects of human activities on interrelationships between multiple SDGs. As a pioneering effort to tackle the challenges, our study has two limitations. First, we did not reveal the full range of the effects of human activities on SDG interrelationships. With the increased complexity of more goals, more data and analyses are required to assess the effects of human activities on the SDGs, as well as the interrelationships among them. In our demonstration study, for example, labor migration might also affect SDG 11 (sustainable cities and communities) by changing the demographic profile in both rural and urban areas. Our demonstration study did not assess those impacts because of data limitations. Second, weighing the positive impacts of human activities against their negative impacts on SDGs remains a challenge because impacts on different SDGs are often not comparable. The approach we proposed classifies and scores the effects of human activities on SDG interrelationships to measure their potential to achieve multiple goals simultaneously. This scoring approach is based on the nature of the impacts of human activities on SDGs and does not consider the magnitude of the impacts. Results from the approach can assist, but cannot replace, the necessary decision-making process to involve stakeholders and weigh the benefits against costs from certain activities. For example, agricultural expansion often generates a negative impact on wildlife habitats, but is essential for promoting many rural households' income (Socolar et al. 2019). Using our approach may identify this trade-off effect of agriculture expansion on the linkage between the goals of wildlife habitat conservation and poverty reduction, but cannot measure or judge whether it is worthwhile to pursue the economic benefits at the cost of some biodiversity loss. As is often the case, science has to be complemented with consideration of values in making decisions (Dietz 2013). The government, communities, and other stakeholders need to weigh the positive and negative impacts of cropping and jointly plan future cropping strategies to balance the needs of wildlife habitat conservation and human well-being.

CONCLUSION
We presented an approach to understanding how human drivers shape the complex interrelationships among SDGs. Our method moves from the focus of previous research on patterns of SDG interrelationships to interrogating the drivers that shape those patterns. We have focused on the livelihood strategies of households in a local community located in an area of global significance for biodiversity. This bottom-up approach complements top-down (international and national level) analyses tracking changes and patterns of interrelationships among SDGs. We believe this bottom-up approach can help policy makers, resource managers, and other stakeholders to design more effective strategies to unlock the transformational potential of the 2030 Agenda.
Responses to this article can be read online at: https://www.ecologyandsociety.org/issues/responses.   (Yang et al. 2013). Table A2 Summary of variables included in the linear regression models that relate changes in SDG indicators (human well-being index, household income, and fuelwood collection) to livelihood changes and other socioeconomic factors (sample size = 186).

A1 Household surveys
The data for evaluating the impacts of the livelihoods on SDG intercorrelations in Wolong at the household level were all derived from survey data. Since household members often make joint or coordinated decisions regarding livelihood affairs, data characterizing livelihoods and SDGs were mainly collected at the household level. We collected household survey data in Wolong in 2010 and 2015. It contains detailed demographic (e.g., household size, members' age, education, and occupation) and socioeconomic (e.g., cropland area, number of livestock, livestock selling prices, and income sources) information of local households in 2009 and 2014 respectively. In the surveys, we included questions (Table A1) to collect information for the construction of human well-being index. 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 have the best knowledge about their households' affairs.
Before performing the formal surveys, we conducted pretests to assess respondents' comprehension of our survey questions and how difficult they were to answer. Based on interviewees' responses in pretests, we iteratively revised our survey instruments to ensure that interviewees understood and were able to answer our questions correctly. In total, 287 and 245 households completed our formal surveys, with a response rate of 95%, and 96%, respectively.
Of the 287 households randomly sampled in 2010, 186 households were revisited in 2015. We asked the survey questions using Mandarin Chinese. The survey instruments and data collection procedures we used in this study were reviewed and approved by the Institutional Review Board of Michigan State University (https://hrpp.msu.edu/).

A2 Changes in livelihoods in Wolong
We observed that cropping in Wolong decreased from 2009 to 2014 (Fig. A1). The average amount of cropland cultivated by each household decreased from 3.4 mu (1 mu = 0.067 ha) in 2009 to 2.9 mu in 2014. About 39% of the cropland loss was caused by the lasting (or postdisaster) impacts of landslides after the 2008 Wenchuan Earthquake and the rest (61%) was caused by land appropriation for infrastructure construction.   There are many opportunities to meet neighbors and work on solving community problems Q4.4*: How active do you think your household members in your community groups or village or township? (Options: 1. Very inactive; 2. inactive; 3. Neither inactive nor active; 4. Active; 5. Very active) Q4.5: Do you agree that people here look out mainly for the welfare of their own families and they are not much concerned with village/neighborhood welfare? Q4.6*: Suppose someone in your village/neighborhood had something unfortunate happen to them, such as a family member's sudden death, there are always some others would be ready to help Q4.7: Overall, you are satisfied with your household's social relationships with others Freedom of choices and actions Q5.1*: Do you think that your household members are always treated equally without regard to gender, race, language, religion, political beliefs, socioeconomic status and more? (Options: 1. Never; 2. Seldom; 3. Sometimes; 4. Often; 5. Always) Q5.2: Your household has affordable access to quality and nutritious food for an enjoyable life Q5.3: Your household has affordable access to quality medical care Q5.4: Your household has affordable access to quality education Q5.5: Your household has affordable access to spacious and quality house Q5.6*: It is difficult to find a satisfied job Q5.7: How often do you feel that you want to help others but limited by your socioeconomic or physical conditions that you cannot help them? Q5.8: Overall, you are satisfied with your freedom of choice and actions Notes: Unless response options are specified after indicator contents, the options are designed in the five-category Likert scale (i.e., strongly disagree, mildly disagree, unsure, mildly agree, and strongly agree). *: Indicators not included in the final confirmatory factor analysis due to low variation or internal consistency with other indicators in the same categories.

Outcome variables
Well -

Livelihoods and their changes
Local off-farm labor work in 2009 The number of laborers earned income through working in local off-farm sectors in 2009.