Testing the stakeholders’ partnership in a tourism waste management network: an ERGM approach

The exponential random graph model (ERGM) is an effective approach for testing the dynamic and local processes of a network. This paper explores the structure of stakeholders’ partnerships in a tourism waste management network using high-order dependency ERGMs based on relational data obtained from a field survey in Motuo County, China. The results reveal that (1) the network has many edges, indicating a tight network; (2) the geometrically weighted edge distribution shows a high transitive effect of the network; (3) the structural effect is more significant than the attribute effect; (4) there is a good agreement between the simulation results and observations, suggesting a tourism waste network with close connections and collaborative division of labor. These findings indicate that different groups of stakeholders have been extensively involved in tourism waste management in Motuo County. The edgewise shared partners formed by stakeholders of different groups increase the information transmission efficiency of the network. The results have implications for tourism waste management, specifically for promoting sustainability transitions via network governance.


Introduction
Tourism waste has attracted widespread attention from scenic area managers and researchers. Tourism waste management is essential for preventing environmental pollution in tourist destinations. Moreover, the tourism industry can be seen as a test site for waste management and recycling within the scope of its products and services (Tansel 2019). Some tourist destinations have focused on creating "garbage-free scenic areas"; moreover, plans for waste classification and recycling management have been developed, and intelligent technologies for waste monitoring have been introduced. However, tourism waste management requires improvement. The main reason is that managing tourism waste is much more difficult than managing domestic waste.
First, the composition of tourism waste is relatively simple, mainly including packaging materials, containers, disposable kitchen waste of various foods and beverages, and other daily necessities discarded by tourists during their trips (Xu et al. 2021a, b).
Second, although the components of tourism waste are relatively simple, the stakeholders involved in tourism waste management are more complex than those involved in domestic waste (Pongsakornrungsilp and Pongsakornrungsilp 2021). Therefore, the network of stakeholders involved in tourism waste management is more complicated (Kernel 2005).
Third, the amount of tourism waste is uncertain in time and space. The pressure on tourism waste treatment in scenic spots is higher during the peak tourist season (Caponi 2022). Meanwhile, the amount of waste depends on the popularity of scenic spots, increasing the difficulty of waste management. Finally, tourist destinations relying on natural resources, such as national parks, are often in places with low accessibility, an underdeveloped economy, and limited infrastructure. Therefore, environmental protection education for stakeholders is more important than the end treatment of tourism waste.
An increasing number of studies have been conducted on waste management and recycling; however, we have limited insight into the effect of cooperative intervention on the tourism industry. Tourism waste management should focus on prevention and collaborative actions with residents and tourism stakeholders (Hayati et al. 2020;Obersteiner et al. 2021). In other words, the management of tourism waste cannot be performed by a single interested party. Instead, this work involves various stakeholders with different responsibilities to formulate tourism waste management policies and coordinate, communicate, and allocate waste management tasks. Since different stakeholders have different interests, their participation in tourism waste management takes different forms, influencing the effectiveness of waste management. Therefore, researchers must conduct a systematic review of the network structure of tourism waste management to obtain a holistic understanding of the current situation and to diagnose problems. In addition, analyzing the network structure of tourism waste management is valuable for formulating targeted policy suggestions to realize the front-end control of waste .
Generally, the stakeholders of tourism destinations include local residents, tourists, managers of tourism enterprises, government departments, and other non-governmental organizations (Byrd 2007;Roxas et al. 2020;Shafiee et al. 2021). Evaluating the efficiency of waste management from the perspective of stakeholders involved in tourism waste is required due to the comprehensive nature of the tourism industry and tourism economic activities. In addition, network governance theory provides essential theoretical support for urban waste management analysis (e.g., Wang 2015;Tan and Sun 2019;Cramer 2022). Moreover, the exponential random graph model (ERGM), which has emerged in recent years, has attracted increasing interest from scholars because it considers the impact of multiple internal structural factors and external drivers on network generation (Xu et al. 2021a, b). This model assumes that network relationships are generated by stochastic processes (Shen et al. 2022). Thus, the impact of various levels of network structural parameters on the network can be modeled (Wang et al. 2016;Shi and Li 2019).
However, there is a lack of studies on using ERGM for analyzing the governance of multiplex networks. Some scholars have explored the application of ERGM in tourismrelated research fields, such as community governance of tourist destinations (e.g., Zhang 2020), the attraction of scenic spots (e.g., Hernández et al. 2021), the network structure and the interactions between source and destination markets (e.g., Lozano and Gutiérrez 2018), an interest and trust network of stakeholders in rural tourism communities (e.g., Shi and Li 2019), and the role of tourism in developing trust and understanding between tourists and residents in host areas (e.g., Khalilzadeh 2018). However, tourism waste management has been neglected for a long time. Therefore, this paper analyzes the partnership of stakeholders in tourism waste management using ERGM to propose suitable countermeasures that improve the efficiency of tourism waste management.
A tourism waste management network is complicated. What are the interactions between stakeholders and the underlying logic? What are the network structures characterizing these interactive relationships? What are the inspirations of the network structures for exploring the treatment of tourism waste under multi-stakeholders? To answer these questions, we propose research hypotheses related to structural and attribute effects to describe the intra-and crossorganizational relationships in the tourism waste management network objectively based on high-order dependency ERGMs.
In addition, we conduct a case study of tourism waste using data obtained from a field survey to determine the logical relationships between stakeholders by evaluating different network configurations. The ERGM theory is used to provide targeted policy suggestions for the diversified treatment of tourism waste by integrating network governance theory and tourism science.
The rest of the paper is organized as follows. "Literature review" briefly reviews the research on tourism waste management, ERGM, and network waste management. "Methodology" describes the methodology and data, including the study site, the investigation, data source, hypotheses, and establishment of ERGMs. "Results and discussion" describes the quantitative results and discusses the findings of this study. "Conclusions and policy implications" concludes the paper and provides suggestions for achieving long-term sustainable tourism.

Literature review
Scholars have explored the socio-economic and environmental impacts of tourism waste and the unsustainability of tourism areas resulting from tourism waste from different perspectives using interdisciplinary research. Several researchers have studied effective tourism waste management measures to cope with environmental pollution caused by waste (e.g., Bashir and Goswami 2016;Obersteiner et al. 2021;Tsai et al. 2021;Maione 2021). Previous studies on tourism waste in scenic areas have primarily focused on qualitative descriptive analyses of measures or plans for waste collection and treatment (e.g., Little 2017; Abbas et al. 2021). Moreover, the majority of published studies on tourism waste have mainly considered the environmental impact of tourism waste (e.g., Barber et al. 2014;Murava and Korobeinykova 2016;Abdulredha et al. 2017;Greco et al. 2018;Kularatne et al. 2019;Paiano et al. 2020;Diaz-Farina et al. 2020) and used summary statements to describe the experiences and methods of tourism waste management (Liu et al. 2019). Furthermore, existing studies have focused primarily on a single subject (e.g., Giurea et al. 2018;Hu et al. 2018;Wijanarko 2020). Nevertheless, few studies considered tourism waste management in scenic areas from different stakeholders' perspectives (e.g., del Mar Alonso-Almeida 2012; Kariminia et al. 2012;Ezeah et al. 2015;Phu et al. 2018).
The topic of stakeholders has received some attention in municipal waste management (e.g., Yukalang et al. 2018;McNicholas and Cotton 2019;Tonini et al. 2020). However, most published papers concentrated on initiatives, planning, and design (e.g., Fuldauer et al. 2019;Ikhwan et al. 2019). Network governance theory is closely linked to stakeholders and can guide researchers in analyzing the structure of stakeholder networks since the stakeholders have common interests and a desire to cooperate (Knox and Arshed 2022;Thorpe et al. 2022). In addition, network governance theory has expanded to incorporate elements of inter-organizational relationships. Network governance theory has been used to investigate the structural relationship among organizations (e.g., Barraket et al. 2021;Tao 2022;Zhai et al. 2022).
To date, several quantitative methods have been used in empirical studies, such as social network analysis (SNA) and ERGMs (e.g., Dietrich et al. 2018;Schulz et al. 2018;Zhang et al. 2021a, b). SNA has been widely used for analyzing the structural effects of relational networks; the method has methodological advantages for evaluating network relationships with simple structures or individual, relational networks (e.g., Fonseca 2021; . However, ERGM is more advantageous for evaluating complex network structures consisting of multiple stakeholders. ERGM emphasizes the self-organization of relationships and the connectedness of individuals, i.e., the probability of a relationship depends on the existence of other relationships. This method is applicable to assessing any network relationship problems (Zhang et al. 2021a, b).
Typically, researchers apply social network theory to develop hypotheses utilizing primarily relational, quantitative representations and use ERGM as an analytical framework for the subsequent analyses using parameter estimation and hypothesis testing (e.g., Wang et al. 2016;Kornienko and Rivas-Drake 2022). Moreover, the ERGM can infer the generative mechanisms of the network by examining structures and processes rather than merely analyzing the integral network structure (Sharifnia and Saghaei 2020). Therefore, the ERGM enables the integration of micro and macro scales using simulations. These advantages of the ERGM for analyzing complex networks make it highly suitable for studying the partnerships of stakeholders in tourism waste management.
The ERGM was initially applied in environmental public management and waste management. For example, Tan and Sun (2019) analyzed the network structure of urban household waste separation and recycling using ERGMs and proposed resolving the stakeholder conflicts according to their results. Ghinoi et al. (2020) applied the ERGM to evaluate the drivers of stakeholders' interactions in a food waste management system. Nevertheless, the number of published papers is extremely small; thus, it is evident that using an ERGM in waste management research is in the exploratory stage.
Tourism waste management is still government led in China, especially in the less economically developed regions of western China. This management type may lead to inefficient waste management. Unlike the traditional authoritative and hierarchical model of policy making and management, waste governance emphasizes a collaborative approach to the management of public affairs by all stakeholders. In a highly efficient waste governance model, stakeholders share the benefits, resources, and power, achieving a win-win situation through communication, negotiation, sharing, and cooperation. Empirical work as part of network studies has shown how specific endogenous network characteristics and exogenous organizational attributes drive the networking behavior of stakeholders (Ghinoi et al. 2020). However, to the best of our knowledge, this issue has not been explored in applied studies focused on qualifying the stakeholder's partnership structure in a tourism waste management network based on ERGM.
Therefore, inspired by the ERGM, this paper aims to fill this gap by investigating the configurations to test the characteristics of the stakeholders' partnership in the tourism waste management network and identify the factors influencing the network structure. Motuo County, China, has ended the ticket-free mechanism in 2020 and is exploring how to use part of the ticket income for waste disposal in scenic spots. Our focus on Motuo County as a case study is also motivated by its representativeness of Yarlung Zangbo National Park.

Study area and case study overview
Motuo County (27° 33′ N-29° 55′ N, 93° 45′ E-96° 05′ E) is located in southeastern Tibet, China, at the lower reaches of the Yarlung Tsangpo River and covers an area of 31,394.67 km 2 . The average elevation is 1200 m, the lowest altitude in Tibet, and the region receives abundant rainfall (Tang 2020). Motuo is located at the eastern end of the Himalayas, where the extensive movement of the earth's crust has created a topography dominated by high mountain valleys, resulting in world-famous ecotourism resources. This area has attracted a wide range of tourists. As a result, the tour route from Milin to Motuo is currently one of the more popular routes for tourist groups (Zhu et al. 2021).
In 2000, the State Council of China approved the establishment of the Yarlung Zangbo Grand Canyon Nature Reserve with Motuo as the core area. The Yarlung Zangbo Grand Canyon National Park was established in 2010 in the Tibet Autonomous Region of China. The area of Motuo County is about two-thirds of the total area of the Yarlung Zangbo Grand Canyon National Nature Reserve, representing an important part of the Yarlung Zangbo Grand Canyon National Park. In recent years, the problem of tourism waste in the southeast Tibetan ecotourism area has attracted widespread attention from people from all walks of life. Therefore, it is representative to study tourism waste management in the Motuo region. The location of the study area and representative landscapes are shown in Fig. 1.

Stakeholders involved in tourism waste management
The stakeholders' network analyzed in this paper consists of stakeholders involved in tourism waste management. The tourism stakeholders are typically divided into four main groups: government, community, enterprise, and advocacy groups (e.g., Byrd 2007; Saito and Ruhanen 2017; Deale and Lee 2021; Kc et al. 2021;Ye et al. 2021). Similarly, the main stakeholders involved in waste management are classified as government, enterprise, community, and advocacy groups (e.g., Joseph 2006;Caniato et al. 2014;Kim and Nguyen 2020;Song et al. 2021;Su et al. 2021;Shooshtarian et al. 2022). Based on the information obtained from the field survey, the main stakeholders involved in tourism waste management are listed in Table 1.

Survey and data
The data for this study were obtained from a combination of questionnaires and interviews. The questionnaire was designed based on the authors' previous study and field survey in the study area and drawing on ideas from the published literature on the design of network analysis (e.g., Dredge 2006; Baggio et al. 2010; Van der Zee . The questionnaire consisted of two aspects. The first part was a survey of the respondents' familiarity with and participation in tourism waste management. Five questions (see Table 2) were used to assess the level of knowledge and contribution of the respondents to tourism waste management. A total of 42 respondents were interviewed; their information is presented in Appendix (see Table 6). These stakeholders included the main stakeholders in the 4 main categories and 16 sub-categories (No. A-P). We assigned weights to the respondents based on their responses to these five questions. Table 7 in the Appendix lists the details of the survey content and findings. The aim of the second part of the questionnaire was to investigate the existence of a relationship between two stakeholders in tourism waste management. The values "0" and "1" represent no association/no coordination and association/coordination, respectively. The raw data were weighted and averaged based on the results of the first part of the weight assignment to the respondents.
Then, the weighted relational data matrix was binarized using the average value (0.00772) as the cut-off value (or the threshold). The relational data reflecting the stakeholders' cooperation or association in the existing local tourism waste management network is shown in Table 3.

Hypotheses
Our initial observations of the stakeholders were developed based on fieldwork and the questionnaire collation. The relational data in Table 3 indicate that a network governance model for tourism waste management has already been developed among different groups of stakeholders.
Since an information exchange occurs between two stakeholders in combination with resource sharing with third parties in the network, we propose the following three hypotheses reflecting the partnership among stakeholders with different structures in tourism waste management.

H1
The clusters in our network consist of several triads and edges with multiple shared partners. H2 The clusters in our observed network consist of several dyads and edges with multiple shared partners. H3 The identity effect between dyads, i.e., two actors, results in multiplex ties between different groups.

Method
Unlike most linear and nonlinear analysis methods, the ERGM is theory-driven and emphasizes the structure of the network (Desmarais and Cranmer 2012; Sharifnia and Saghaei 2022; Jang and Yang 2022). A critical theoretical assumption of the ERGM is that networks are selforganizing, i.e., local relationships are interdependent. These local relationship patterns are considered network configurations Stephens et al. 2016).
Researchers use multiple network configurations to predict the emergence of networks. Accordingly, the model parameters describe the significance of the network configurations in the observed relationships. Therefore, the ERGM models a given network based on the local relationship structures, such as dyads and triads (Ulibarri and Scott 2017;Felmlee et al. 2021). Moreover, these network configurations are considered to arise from local social processes in which network actors form links corresponding to their living environment. The ERGM assigns a probability to each graph in the sample, resulting in several graphs (Sharifnia and Saghaei 2022). The following four model types represent four key milestones in the development of the ERGM: the simple random graph model, dyadic independence model, dyadic dependence model, and high-order dependence model (Snijders et al. 2006;Robins et al. 2007;Stephens et al. 2016). The expressions and explanations of these four models are listed in Table 8, Appendix. Erdös and Rényi (1959) proposed a statistical model for obtaining simple random graph features based on Bernoulli's assumption that the "relationships between network members are generated randomly and independently of the relationships between other members." The model compares relational network data with an assumed zero distribution with only one network configuration, the edge. Therefore, the model cannot provide useful information on the network structure. On this basis, the p1 model that reveals the model's out-links, in-links, and mutual relationships was proposed (Holland and Leinhardt 1981;Bi et al. 2021). Subsequently, this model evolved into another expression that could integrate covariates; this model is known as the p* model (Wasserman and Patrision 1996;Shin et al. 2022). However, the p* model has certain limitations, such as an inability to fully capture the structural characteristics of the observed network. Thus, researchers improved curved exponential family (CEF) models and modified the p* model by introducing statistical terms, such as the geometrically weighted degree distribution (GWD), geometrically weighted edgewise distribution (GWESP), and geometrically weighted dyadwise shared partners (GWDSP) (e.g., Hunter 2007;Hermans 2021;Xu et al. 2021a, b;Liu and Ge 2022). Correspondingly, the Markov Chain Monte Carlo (MCMC) estimation method was proposed. It has a realization-dependent conditional assumption and relatively high simulation accuracy (Geyer and Thompson 1992;Pattison et al. 2013). Typically, a parameter estimate of more than twice the standard error indicates significance (Snijders et al. 2006;Robins et al. 2007). Based on the above comparison of the analytical models generated by the ERGM theory at different stages, the highorder dependence models were used in our study. Structural and attribute effects are incorporated into the modeling process in this paper to describe the intra-and cross-organizational relationships in the network of tourism waste governance. Specifically, the GWESP and GWDSP are used to determine pure structural effects, including the information transmission and intermediary effect of the network, respectively. The established models are defined in Eqs. (1) and (2), and the information related to the model configurations is provided in Table 4.
where c is a constant; θ, η, and δ are the parameters of the network's edge, GWESP, and GWDSP configurations, respectively; M(y) is used to estimate the impact of the identity attributes of different groups of stakeholders on the network structure, and ζ is its parameter.

Model estimation and fit
The R Statnet toolkit was used to estimate the parameters and determine of the goodness of fit (GOF) of the proposed model. In addition, the step-by-step model (including model (2) P r (Y = y) = 1 c exp{ L(y) + Dsp(y) + M(y)}   (Table 9). Table 5 lists the results of the ERGMs, including the high-order dependence models defined in Eqs. (1) and (2). Based on the information provided by Goodreau et al. (2008) and Zhang (2020), for determining the value of the decay parameter α, α = 0.1 was used as the initial condition. The value of α was gradually increased until the log-likelihood estimate of the model no longer increased. In this paper, α was 0.25. The significant parameter estimates of model 04 (in Appendix Table 9) are at least two times higher than their standard errors, and the Z-values and P-values are significant. Therefore, it can be tentatively concluded that model 04 has a better fit than model 05 (in Appendix Table 9) for the observed relational data.
Furthermore, model 04 has the lowest values of the Akaike information criterion (AIC), Bayesian information criterion (BIC), and log-likelihood. The results of the last iteration of the MCMC model are shown in Fig. 2. The left side of Fig. 2 shows the statistical terms for each configuration, and the right side of Fig. 2 shows the histograms of the densities of the configurations. The statistical time series of model 04 exhibits random fluctuations around 0, indicating that model 04 has high stability and does not exhibit approximate degradation.
Further, the box plots (Fig. 3) demonstrate the soundness of the above analysis. The black lines of the coefficients of the three configurations, i.e., the edges, GWESP, and identity, are close to the medians, indicating a good fit of the model. Specifically, the black line on the GWESP's The graph of model 04 obtained from the simulation (Fig. 4) was used to evaluate the GOF and the model performance. The results demonstrate that model 04 has high stability and high GOF for all configurations. In addition, the means from the simulation of the edges, GWESP, and identity are 87.91, 98.29, and 34.11, respectively. They are very close to the observed mean values of these three configurations (87, 96.9, and 34, respectively). Therefore, model 04 is suitable for practical applications.

Structural effects
The significance levels of the edge and GWESP configurations are high in model 04, indicating high density and information transmission in the network. Therefore, hypothesis H1 can be confirmed. The practical implications of the results of the empirical analysis are as follows. First, all stakeholders have been involved in tourism waste management in the study area. Moreover, communication among different groups participating in tourism waste management occurs frequently. Second, according to the GWESP configuration, the level of information transmission of the network is significant, indicating efficient information exchange between different actors in the stakeholders' network.
The parameters of the GWDSP configuration in model 05 show a significant positive effect, i.e., an intermediary effect similar to a structural hole occurs in the stakeholders' network of tourism waste management. However, the overall fit of model 05 is poor. Hence, it is possible to conclude that waste management is not controlled or constrained by a few intermediaries. This conclusion further confirms the results presented in the previous paragraph. Therefore, we cannot confirm hypothesis H2.

Attribute effect of the actors: the identity effect
The attribute effect, i.e., the identity effect, can be used to estimate the influence of stakeholders with different identities on the formation of network relationships. In model 04, the attribute effect of the nodes is significant. Therefore, we can confirm hypothesis H3. Hence, it can be concluded that the influence of the stakeholders from different sectors is evident in the tourism waste management network.
However, it is important to note that the structural effect is greater than the attribute effect in this network. In other words, the implementation of tourism waste management does not primarily depend on the identity of the stakeholders. That is, the information transmission driven by administrative power is not apparent. In contrast, the current case network structure is relatively flat. Task allocation and coordination among stakeholders engaged in different jobs are highly shared in the network.

Discussion
In summary, relatively frequent and efficient communication and cooperation occur between different departments in tourism waste management in Motuo County. The main reason is that Motuo County has begun to explore diversified and market-based eco-compensation mechanisms since the end of the admission-free mechanism in 2020 and has placed waste management at the forefront of environmental protection and pollution control. For example, the Motuo County Environmental Protection Bureau, the Housing and Construction Bureau, the Urban Management Bureau, and the Culture and Tourism Bureau have widely publicized the ecological compensation policy to local residents and foreign businesspeople and have made efforts to conduct measurements.
Moreover, Motuo County has begun to use admission revenue to alleviate the financial burden of environmental management and has taken advantage of the press to disseminate practices worthy of emulation to various groups of stakeholders, encouraging and guiding them to participate in waste management. In terms of infrastructure, the county's waste disposal site was completed, ending the practice of using an open landfill.
Therefore, Motuo County should take advantage of the current situation and, more importantly, focus on information transmission in the network, emphasizing the role of residents in tourism waste management, particularly families and residents involved in tourism development. This approach will increase the number of channels to inform the public, creating flat channels of information dissemination and reducing the number of channels through which policies or tasks are passed on by government departments. These measures will increase the efficiency of waste management.
Finally, our results also suggest that it is crucial to policy implementation to enhance the influence of the government in the network. Moreover, it is necessary for the government to continue to advocate for all interest groups to treat tourism waste as a common mission and responsibility. The identity role of NGOs, scientific institutions, and tourists as information disseminators, collectors, and feedback providers should be valued, and their position in the network of tourism waste management should be enhanced.

Concluding remarks
Tourism waste in scenic areas has long plagued tourism managers and has deteriorated the tourism image of destinations. Southeastern Tibet, China, is no exception. Tourism waste is the main constraint to developing ecotourism in the Motuo region of the Yarlung Zangbo National Park. The ERGM model was used to investigate the collaboration and participation of various stakeholders in tourism waste management since it is well suited for this task.
High-order dependence models were used to analyze the network of waste management to ensure the objectivity of the results, which is a highlight and novelty of this paper. Unlike dyadic independence and dependence models, highorder dependence models that use decay coefficients do not exaggerate the influence of k-star and triads. To the best of our knowledge, this paper is the first attempt to establish a high-order dependence model for analyzing network relations in tourism waste management.
A questionnaire survey was conducted of participants in tourism waste management. Two structural effects and one attribute effect were used in the model to analyze the survey data. The results confirmed the validity of the theoretical model (model 04) with the GWESP and identity configurations, indicating that triads were crucial for coordinating and communicating with multiple stakeholders. In addition, the stakeholders' identity had a significant impact on the partnership structure.
This paper highlighted the advantages of the ERGM, i.e., its ability to analyze the structure of a network and clarify the formation of the partnership network, the characteristics of the networks' configuration, and the nodes. Moreover, the ERGM can simultaneously analyze the endogenous and structural effects of the networks. In particular, the stakeholders' partnerships can be visualized by modeling highorder dependency models based on relational data obtained from a field survey.
There is little research on tourism waste management networks. This paper investigated the network structure of tourism waste management to evaluate sustainable waste management involving multiple stakeholders. In addition, network governance theory and a case study were used to improve waste governance in tourist destinations.

Policy implications
The ecological environment is extremely fragile. Therefore, it is necessary to prevent environmental degradation and adverse effects on tourist satisfaction caused by tourist waste. Tourism waste management is a long-term and arduous task. Our results indicate it is crucial to improve multi-cooperative management of tourism waste and prevent tourism poverty by maintaining the multi-cooperative and communication relationship of waste management subjects and establishing an orderly cooperative and interactive community multi-co-governance network. Therefore, it is necessary to formulate reasonable incentive measures to ensure that multiple partners participate in the garbage disposal.
First, it is essential to formulate targeted incentives for tourists. For example, tourists should be encouraged to bring waste generated during their trip back to the designated garbage collection point. This garbage can be exchanged for free meal vouchers, ticket discounts, and other practical items to improve the tourists' enthusiasm for participation.
Second, the local government should take the lead in formulating the competition mechanism and incentive measures for garbage disposal for residents in the surrounding communities of the scenic spot. In addition, the government should change existing garbage disposal methods. It is suggested to try to use incineration to replace landfill disposal.
Third, it is suggested that scenic spots should improve the level of garbage collection and classification, including professional training for the staff in garbage classification and recycling and purchasing garbage storage and treatment equipment.
Last, it is suggested that scenic spots should implement free admission for NGOs to improve their tourism welfare. It is necessary for the government to formulate specific implementation rules for tourism waste management and issue documents advocating civilized travel. There is an urgent need for scenic spots to unite with NGOs to educate and guide tourists using online and offline media platforms. It is meaningful to cooperate with local universities in Tibet to attract college students to participate in tourism activities as tourists and volunteers.
Limited by time and human resources, there is room for further optimization of ERGMs and improving the comprehensiveness of network analysis. First, we focused on only one network. The tourism waste management network should be refined into several networks from different aspects, which would facilitate the analysis of exogenous variables between different networks, providing a more detailed understanding of the problems in tourism waste management. Moreover, the identity effect was considered a main effect, whereas other attribute effects were not addressed. In addition, we did not conduct a cross-sectional comparative analysis of the attribute effect, which was related to the other limitations.

Geometrically Weighted Edgewise Shared Partners (GWESP).
ESPi(y) denotes the number of edges with i shared partners in network y. The other letters are denoted as above.

Geometrically Weighted Dyadwise Shared Partners (GWDSP).
DSPi(y) is the number of dyads that share partners with i; The other letters are the same as above.