Introduction

Noise in megacities is pervasive and originates from a variety of sources, both long-term and occasional1, with the associated damages being increasingly random and unpredictable. These features make noise less amenable to the mandatory governance strategies traditionally used in water and air pollution control. As noise intensity decreases significantly with distance, it leads to disparities among spatially distributed groups, and traditional coarse-grained environmental governance approaches can result in unfairness2. This disparity complicates the reconciliation of interests between the public and the government, as well as among different public groups, thereby hindering effective noise abatement (NA) in megacities. Moreover, complex psychological factors amplify the subjective impacts of noise pollution3. Therefore, focusing solely on the physical characteristics of noise in NA is insufficient, especially in megacities characterized by multicultural societies.

Globally, noise legislation typically aims to protect a quiet environment for residents’ daily lives by integrating NA into the public health ___domain and promoting public participation. Public participation in NA is characterized by co-governance between the government and residents. For example, the United States4 and Germany5 encourage public participation through information disclosure, while France6 and Japan7 have established community-based self-governance.

In China, megacities—continuing to grow in number due to rapid urbanization—are often densely populated with high-rise residential areas that require extensive transportation and other public services, thereby increasing the risk of noise pollution. In 2021, China amended theNoise Pollution Abatement Law (NPAL), signaling a shift toward co-governance. However, public participation in this process is frequently hindered by a lack of expertise or time8, imbalances in interest structures9, and inconsistencies between individual benefits and social welfare10. Addressing these social challenges is essential for effective co-governance in NA.

Existing literature highlights the health impacts of noise pollution, including sleep disturbances11, cardiovascular problems12, and a decline in mental well-being13. In response, technical measures such as the use of advanced acoustic materials14, silencing technologies15,16, and noise pollution mapping17,18 have been widely implemented. However, most people do not desire a completely silent world. An ideal soundscape not only protects residents’ physical and mental health but also fosters economic growth. The International Organization for Standardization (ISO) defines soundscape as the ‘acoustic environment as perceived or experienced and/or understood by a person or people, in context’19. Soundscape, aligned with fundamental human health needs, integrates individual preferences with socially defined physical indicators, offering a new approach to NA. Therefore, a model within the soundscape framework is developed to facilitate a smoother connection between noise-related health impacts and public participation, incorporating residents’ subjective attributes. Although the questionnaire was designed using noise-related expressions for greater public comprehension, the framework extends beyond physical indicators by integrating subjective psychological factors, such as health impact assessments (HIA). This approach establishes an innovative means of measuring the soundscape.

Co-governance refers to a governance model that emphasizes public participation, diversity, and interactive collaboration among multiple stakeholders. The governance of traditional pollution has evolved, transitioning from a focus on the role of a single dominant actor in public administration theory20 to the theory of new public management21, and further to co-governance theory22, which emphasizes the coordination of various stakeholders. In contrast to traditional governance, which often relies on top-down regulatory measures, co-governance fosters more inclusive decision-making processes and dynamic stakeholder engagement. As highlighted in analyses of public participation23, co-governance serves as a mechanism for enhancing communication and incentivizing the substitution of ex-ante governance (preventive measures) for ex-post governance (reactive measures). Examples include pilot programs for household waste sorting and citizen engagement initiatives established by the Chinese government24.

In the case of noise pollution in megacities, where the range of impacts is geographically smaller yet more complex due to the presence of diverse and often conflicting stakeholders, co-governance proves more effective in coordinating subjective, differentiated, and even opposing environmental interests among the public. Therefore, soundscape co-governance (SCG) aims to create an ideal soundscape that promotes health, enables individuals to enjoy a tailored quiet environment, and facilitates multi-agent coordination for effective implementation.

Manufacturing enterprises are currently withdrawing from the centers of megacities, creating a situation that contrasts with the polycentrism described by Elinor Ostrom25. This shift has led to increasingly severe non-industrial noise pollution in megacities, bringing the government-public and intergroup relationships to the forefront of SCG. However, this challenge also presents an opportunity for the sustainable transformation of environmental governance in megacities. In this context, Shenzhen—a pioneer in China’s policy reform and urban planning, and a crucial link between China and the world—has been selected as the case study26. This paper examines the effectiveness of NA in Shenzhen from the residents’ perspective and identifies feasible pathways toward achieving an ideal soundscape within the co-governance framework.

By integrating the concept of SCG with the characteristics of megacities, a theoretical model is developed based on the Framework for Human Health Risk Assessment to Inform Decision Making (FHHRA)27 established by the Environmental Protection Agency (EPA) and the Theory of Planned Behavior (TPB)28. The proposed model refines existing frameworks in two key ways: First, it incorporates expressed damage into the FHHRA to measure residents’ self-experienced and perceived health impacts of noise. Second, drawing on the TPB, which is widely used in behavior prediction and intervention, the model introduces a subjective indicator—HIA—which reflects residents’ evaluation of their health impacts from noise. This indicator, along with attitude, subjective norms, and perceived behavioral control from the original TPB framework, is used to predict individuals’ behavioral intentions to participate in co-governance within megacities, aiming to better adapt the model to the scenario of megacity SCG.

To validate the scientific robustness and applicability of this innovative framework, the theoretical model is integrated with the Extended Theory of Planned Behavior (E-TPB) and applied to the analysis of SCG in megacities. Through survey questionnaires and subsequent econometric modeling, the risk features and corresponding solutions for residents’ participation in SCG are explored. This analysis provides solid empirical support for the scientific foundation and replicability of the megacity SCG framework developed in this study.

The research questions and the marginal contributions of this paper are as follows:

  1. (1)

    How can residents’ subjective experiences and perceptions be integrated to facilitate soundscape measurement: This study introduces residents’ self-assessed health impacts by incorporating HIA as an innovative variable in the questionnaire design, rather than relying solely on instrument-detected decibel levels. This approach aligns with the NA goal of enhancing residents’ experiences within the soundscape. Notably, the inclusion of residents’ self-assessed health impacts marks a departure from traditional methods, complementing the widely used, instrument-detected health impact measures. This distinction is important, as instrument-based measures often overlook the interaction between psychological factors and the physical health of individuals.

  2. (2)

    How can co-governance be constructed using the diverse features of megacities: It tackles the disparities in noise pollution sources, exposure levels, and population sensitivity features, moving beyond the rigid frameworks of traditional environmental governance by implementing differentiated and flexible co-governance approaches.

  3. (3)

    How to identify and overcome challenges in public participation: The paper identifies and addresses the challenges in transforming risk identification into risk solution for public participation. By enhancing co-governance, it serves for achieving an ideal soundscape.

The “Results” section presents the theoretical model, SCG investigation results and correlation, and estimation outcomes. “Discussion” interprets the findings, draws conclusions, and offers recommendations. “Methods” details the model creation, sample processing, and empirical strategies. Supplementary Information provides additional model specifications, data and analysis details.

Results

The Logical Framework and Basic Hypotheses are as follows:

Stage1 (Risk Identification)

Based on the FHHRA developed by EPA, and features of noise, this paper establishes the framework of risk identification under the view of ideal soundscape. Particularly, health effects of self-assessment of the residents are introduced in expressed damages. Therefore, key factors determining the risks of SCG are shown in Fig. 2a including:

  1. (1)

    Source: Urban noise sources can be categorized into industry, transportation, construction, and social life according to NPAL of China, as detailed in Supplementary Table 1. Unlike industrial noise, which can often be effectively mitigated through engineering and technical measures29, noises from transportation, construction, and social activities are intricately embedded within residential areas and are closely linked to the individual behaviors of residents. This integration makes it challenging to gather concrete evidence of pollution from these sources, complicating efforts to regulate and control them30,31.

  2. (2)

    Exposure: Noise levels attenuate rapidly with distance from the source, leading to significant disparities in exposure and experienced damage among residents. These differences in exposure, primarily driven by proximity to noise sources, pose a challenge that is difficult to mitigate through technical or management solutions alone. This spatial variability complicates the development of uniform NA strategies, emphasizing the need for tailored approaches that consider specific exposure conditions.

  3. (3)

    Sensitivity: Even when residents are exposed to the same level of noise or similar health impacts are detected by instruments, the actual health damage they experience can vary significantly. This indicates that although health impacts detected by instruments are widely used as a fundamental basis for environmental standards, their application—whether they exceed the standard thresholds set by Chinese noise regulations—to evaluate the health impacts is an estimation based on high-probability conditions. However, this measurement lacks consideration of individual factors such as personal health status and age, making it insufficient for assessing specific population groups or individuals. This variation is largely due to the particularly sensory nature of noise and its effects on mental well-being32. Known as sensitivity, it highlights the need for targeted measures to address the health impacts as perceived by individuals.

Stage 2 (Risk solution)

This paper examines SCG participation by exploring the relationship between public participation willingness (BI) and behavior (BR) within the framework of the TPB28. Figure 1 illustrates the features of the indicators and their relationship with SCG participation, while Fig. 2b presents the correlations between these indicators.

Fig. 1: Indicator features and the relationship with SCG participation.
figure 1

This figure illustrates the key indicators influencing soundscape co-governance (SCG) participation, emphasizing the interplay between participation willingness and participation behavior. While individuals often assume their participation behavior is driven solely by willingness, other external and internal factors also play a role.

Full size image
Fig. 2: Theoretical model construction progress.
figure 2

This figure presents a comprehensive theoretical model integrating risk identification and the extended Theory of Planned Behavior (E-TPB) to guide public participation in soundscape co-governance. a Outlines the key components of soundscape-related risks, including sources of stressors (e.g., industry, transportation), exposure pathways, health endpoints, affected receptors, and the expressed damage. b Introduces the theoretical framework of the E-TPB, incorporating AT, SN, PBC, and HIA. These factors influence BI and BR in soundscape governance. c Presents a two-stage approach: Stage 1 identifies risk severity and inconsistencies in public perception, while Stage 2 provides solutions for managing risks through participatory decision-making.

Full size image

According to the Self-Determination Theory33, individuals naturally believe that their actions are self-directed, implying that they participate in activities because they want to, not because they are compelled to. This suggests that BR is inherently influenced by BI. However, a complete motivation for participation is also affected by uncontrollable external forces, which may contradict personal intentions, as highlighted by the non-self-determination theory34. Therefore, the relationships between BR and BI can vary significantly under different circumstances, leading to the formulation of hypothesis H1.

H1: BI has a significantly positive impact on BR

The TPB consists of a behavioral model framework that includes the indicators of Attitude (AT), Subjective Norm (SN), and Perceived Behavioral Control (PBC). This study extends the framework by incorporating the indicator of ‘Health Impact Assessment (HIA)’, which represents health-oriented participation and establishes a connection between health impacts and public participation, thereby forming an E-TPB. The paper aims to explore the characteristics of these different indicators and their relationships with BI and BR, providing a multi-dimensional analysis of participation features.

AT stems from an individual’s awareness and influences public participation in SCG through its relationship with BI and BR. According to the TPB, AT is defined as the awareness level of a behavior, shaped by past experiences in participation. This awareness is heavily dependent on the individuals’ preferences and the information they possess.

First, the paramount importance of protecting residents’ health is well recognized as a key environmental value, with defensive behaviors commonly employed to evaluate how residents mitigate health impacts from pollution35. By safeguarding their health interests, AT shapes their willingness and behavior regarding SCG participation.

Second, AT also arises from individuals’ sense of social responsibility, which can lead to what is technically referred to as ‘emotional or hedonistic’ participation36. In this context, when the residents perceive that ideal SCG benefits both the environment and social welfare, their participation willingness, driven by emotional or intrinsic impulses, is likely to increase.

This paper, therefore, explores individuals’ attitudes toward health (AT1) and the environment (AT2) and how these attitudes influence SCG participation at both the self-interest and social welfare levels. This leads to the formulation of hypothesis H2.

H2: AT has a significantly positive impact on BR and BI

SN is typically defined as the perceived social pressure to participate, which can either encourage or inhibit an individual’s willingness to participate. While individual and public interests serve as primary motivators for participation, external factors also play a critical role as prerequisites37. According to Rogers38, the public participates even in less enjoyable activities because they need to interact with others. In the context of public participation in SCG, this interaction includes factors such as policy support (SN1), economic incentives (SN2), law enforcement (SN3), and encouragement from family and friends (SN4). Based on these considerations, we propose hypothesis H3.

H3: SN has a significantly positive impact on BR and BI

PBC refers to the extent to which individuals perceive the difficulty of participation, encompassing factors such as personal abilities (PBC1–PBC4), efficient feedback mechanism (PBC6), and family nexus with environmental occupation (PBC5)39. Zhao et al.40 suggest that the degree to which individuals feel their abilities are being developed influences their level of behavioral motivation. In the context of SCG, residents who believe they are capable of contributing to the ability to improve soundscape (PBC1), ability to learn relevant regulations (PBC2), ability to prevent health impacts caused by noise (PBC3), ability to access feedback channels (PBC4) are more likely to have a proactive attitude towards participation. According to the principles of Responsible Research and Innovation (RRI), the relationships between individuals and stakeholders, along with their participation, play a crucial role in shaping participation willingness and behavior41. Based on these considerations, we propose hypothesis H4.

H4: PBC has a significantly positive impact on BR and BI

When applied to NA, relying solely on the traditional TPB framework without considering health impacts can lead to deviations from the central value of SCG, which is predominantly motivated by health concerns. For the E-TPB developed in this paper, the HIA consists of two components: HIA1, which reflects the self-assessed health impacts, and HIA2, which measures the willingness to pay for reducing these health impacts. This extension allows for a deeper understanding of how health-related considerations drive the public’s participation in SCG.

HIA2, a concept commonly used in environmental economics, involves willingness questions that not only consider participation out of one’s own volition but also acknowledge the potential for free-riding behavior—a scenario that frequently occurs in real-life situations42. Generally, when individuals self-assess and recognize greater health impacts from noise pollution, their BI and BR towards participating in SCG are strengthened32.

By comparing the results of HIA2 with those of HIA1, a more accurate measurement of the health impacts caused by noise pollution can be obtained. This comparison allows for a nuanced understanding of how perceived health damages directly influence public participation in SCG. Based on this rationale, hypothesis H5 is formulated.

H5: HIA has a significantly positive impact on BR and BI

This paper draws on existing measurements43,44,45 and establishes an indicator system detailed in Supplementary Table 2. The variables defined are BR and BI, with BR assessed through the question ‘How do you describe your participation behavior’ and BI through ‘Which participation willingness do you prefer’. The responses for BR are categorized as ‘Endurance’ (1), ‘Complaint’ (2), ‘Negotiations’ (3), and ‘Relocation’ (4).

For HIA variables, HIA1 and HIA2 are measured with distinct scales. HIA1 responses are graded as ‘Dysphoric’ (1), ‘Sleepless’ (2), ‘Daily life bothered’ (3), ‘Ill’ (4), and ‘Inpatient’ (5). HIA2 follows a monetary scale (Yuan) with values assigned as ‘0–100’ (1), ‘100–500’ (2), ‘500–1000’ (3), ‘1000–2000’ (4), and ‘>2000’ (5).

For the remaining variables, this paper utilizes the Likert scale method, categorizing responses into ‘Strongly disagree’ (1), ‘Disagree’ (2), ‘Not sure’ (3), ‘Agree’ (4), and ‘Strongly agree’ (5). This method helps in quantifying the attitudes and perceptions of the participants regarding various aspects of SCG.

This paper establishes a comprehensive theoretical model of SCG decision-making in Chinese megacities (Fig. 2c). In stage 1, the relationship between HIA and BR is examined based on the key factors determining risks—source, exposure, and sensitivity—and their specific characteristics, to identify risk features. The analysis then explores the reasons for the inconsistency between residents’ experience of health damage and the effective expression of their demands.

When HIA and BR are consistent, residents can clearly express both the health impacts they experience and their demands, largely due to information symmetry. This alignment facilitates the achievement of an ideal soundscape. However, when there is inconsistency between HIA and BR, an information gap arises between residents’ experiences and the demands conveyed, rendering the risks difficult to resolve within the co-governance framework.

In such cases, stage 2 analysis must be introduced to eliminate the inconsistency between HIA and BR, by adjusting AT, SN, and PBC.

In considering the realization of an ideal soundscape, the following scenarios are taken into account:

  1. (1)

    Inconsistency between BI and BR: For instance, highly educated residents may have the capability and willingness to participate in SCG, but the high time costs due to work commitments may hinder them from taking action. In this case, adjustments should be made to ensure consistency between BI and BR.

  2. (2)

    Consistency between BI and BR, but inconsistency between HIA and BI: For example, residents affected by noise from square dancing may develop negative perceptions of the social forces involved in SCG, leading to a weakened willingness (BI) despite consistent behavior (BR). In such cases, adjustments should be made to ensure consistency between HIA and BI, strengthening the willingness to participate in SCG.

  3. (3)

    Consistency between BI and BR, but inconsistency between HIA and BR: This often results in a corresponding inconsistency between HIA and BI. Typically, when residents face a prolonged lack of policy support, they may lose confidence in taking action and choose to endure severe health impacts. This leads to a diminished willingness (BI) and a reduced likelihood of participation (BR) in SCG. To address this issue, adjustments should be made to ensure consistency between HIA, BR, and BI, promoting more proactive participation in SCG.

SCG investigation results and correlations are as follows:

This paper utilizes data from a survey on the current status of NA in Shenzhen, conducted by the Shenzhen Ecological Environment Bureau (SEEB) from October to December 2022. The survey aims to identify the features of public participation in NA in Shenzhen, assess the current situation and challenges of SCG, and gather public opinions and suggestions on NA.

Supplementary Table 3 (column ‘Total valid sample’) and Fig. 3a present the basic characteristics of the sample. Among the 626 respondents, 297 hold a bachelor’s degree or higher, reflecting the high educational level in Shenzhen. Additionally, approximately 41.53% of the respondents work less than 40 h per week, while 23.32% work more than 60 h per week, highlighting Shenzhen’s status as a busy and vibrant megacity.

Fig. 3: Descriptive statistics of variables and features in the sample.
figure 3

This figure presents descriptive statistics of variables and features in the sample, analyzing the distribution, correlation, and variation of key factors in the extended Theory of Planned Behavior (E-TPB). a Shows the basic features of the sample. b Presents the percentage of option values, mean, standard deviation, and correlation of variables in the E-TPB framework. c Illustrates the distribution and percentage BR and BI. d Displays the percentage of variable values, mean, and standard deviation of BR and BI under different conditions of source, exposure, and sensitivity. e Provides the frequency, percentage, and cumulative percentage of health impact assessments HIA1 and HIA2. f Presents the percentage of variable values, mean, and standard deviation of HIA1 and HIA2 under different conditions of source, exposure, and sensitivity.

Full size image

Supplementary Table 4 and Fig. 3b and c indicate that most variables follow a normal distribution. Apart from SN1 and PBC6, variables within the same indicator show positive correlations. The negative correlation between SN1 and the other SN variables suggests that policy support alone is insufficient to effectively promote public participation in SCG. Similarly, the negative correlation between PBC6 and the other PBC variables indicates that convenient feedback channels can partially compensate for the lack of government economic incentives and law enforcement, thereby enhancing public participation.

The characteristics of residents vary across groups in terms of the BR-BI relationship (Supplementary Table 5_BR and BI, Fig. 3d). For instance, residents impacted by construction noise tend to have higher public participation willingness (BI), while those affected by industrial noise exhibit higher public participation behavior (BR). This indicates that while the current governance system effectively addresses industrial noise abatement, it has not successfully motivated residents affected by non-industrial noise to convert their willingness into active participation in SCG. Additionally, groups with relatively higher education levels (Group 1, Group 2) generally demonstrate a greater willingness to participate, but their BR is lower compared to Group 3 and Group 4, which have lower education levels.

Risks in public participation are reflected by HIA1 and HIA2, as shown in Fig. 3e and Supplementary Table 5_HIA1 & HIA2. For HIA1, subjectively recognized damages such as ‘dysphoric,’ ‘sleepless,’ and ‘daily life bothered’ account for a significant 94.73% of the responses. In contrast, only 5.28% of respondents reported ‘ill’ and ‘inpatient’ conditions that can be objectively measured. Regarding HIA2, approximately 37.70% of respondents are willing to pay more than 1,000 yuan to reduce health impacts, indicating that the negative experiences caused by noise are highly unacceptable to a substantial portion of the population.

As shown in Fig. 3f, from an exposure perspective, residents living within 0–100 m of noise sources experience the highest levels of HIA1. Although HIA1 decreases as the distance from the noise source increases for those living 100–500 and 500–1000 m away, it rises again for residents living over 1 km from the source, second only to those within 0–100 m. This pattern aligns with existing findings46, which suggest that residents living farther from noise sources generally have higher incomes and education levels, prefer quieter environments, and are more sensitive to external disturbances. Unlike those within 0–100 m, residents living more than 1 km away exhibit high consistency between HIA, BI, and BR, which is advantageous for SCG. From a sensitivity perspective, individuals who spend more time at home perceive more severe HIA1, while HIA2 is higher among those with higher incomes.

SCG estimation results are as follows:

Stage1 (Risk identification)

Figure 4a presents the results of the Lorenz Curve and Gini Coefficient analysis, showing values of 0.8021 for HIA1 and 0.7328 for BR. To account for sampling variability and quantify uncertainty, a non-parametric bootstrap method with 1000 replications was applied. The 95% confidence intervals ranged from 0.7769 to 0.8274 for HIA1, and from 0.7153 to 0.7487 for BR. These results suggest variability in responses, with the relatively narrow confidence intervals indicating stable estimates and low sampling uncertainty (Fig. 4b). Additionally, the Concentration Curve was plotted, and the Concentration Index was calculated to assess the inconsistency between HIA1 and BR across different factors: Source (HIA1: 0.0598, BR: -0.0130), Exposure (HIA1: -0.0785, BR: -0.0471), and Sensitivity (HIA1: 0.0210, BR: 0.1052) (Fig. 4c). A systematic clustering method using Euclidean distance was applied to the Sensitivity indicator, resulting in six clusters based on the ‘iceberg diagram’ and the dendrogram. Through single-factor ANOVA analysis (Supplementary Table 6), the study identified four categories of respondents with distinct sensitivity characteristics (Supplementary Table 7), considering individual as well as family factors. Compared to residents in Group 4, who are older and have lower education levels, residents in Groups 1 and 2 possess higher education levels. Residents in Groups 2 and 3 are younger, while those in Groups 1 and 3 spend more time at home and have higher annual family incomes.

Fig. 4: Identification of Inconsistency in HIA1 and BR.
figure 4

This figure analyzes the inconsistency between health impact assessment (HIA1) and public participation behavior (BR) using inequality and concentration measures. a Presents the Lorenz curve and Gini coefficient, illustrating the distribution inequality of HIA1 and BR. b Shows the distribution results of the non-parametric bootstrap method for HIA Gini coefficient and BR Gini coefficient. c Depicts the concentration curve and concentration index from the perspectives of Source, Exposure and Sensibility. The red dashed line (Absolute Fire Line) serves as the benchmark for perfect equality in distribution. The green circles represent the Lorenz curve or concentration curve of HIA1, illustrating its distribution pattern, while the brown triangles represent the Lorenz curve or concentration curve of BR, showing the distribution of BR. These visual distinctions help highlight the differences in inequality and concentration measures between HIA1 and BR.

Full size image

The regression results of eq. (3) are presented in Table 1, with parallelism tests and model likelihood ratio tests provided in Supplementary Table 8. The analysis identifies that construction and social-life sources (source), residents living within 0–100 m of the source (exposure), and individuals in Groups 1 and 3 (sensitivity) are the groups experiencing higher levels of HIA1. The risks in public participation related to HIA1 and the challenges in improving BR are as follows:

Table 1 Deepening Identification of the Inconsistency between HIA1 and BR
Full size table

Panel A shows that, the probability of experiencing higher levels of HIA1 due to construction and social-life noise sources is 1.96 times and 2.31 times greater, respectively, compared to industrial noise sources. However, the proportions of individuals adopting more positive BR to construction, transportation, and social-life noise are only 34%, 34%, and 44% of those affected by industrial noise.

Panel B shows that residents living within 0-100 meters of the noise source experience a higher probability of HIA1, which is directly associated with their proximity to the noise source. Despite this higher exposure to health impacts, these residents have a lower probability of implementing proactive public participation.

Panel C shows that residents in Group 1 and Group 3 have a higher probability of experiencing more HIA1 compared to those in other groups. This is because longer home time results in greater exposure, and higher-income groups tend to be more health-conscious, making them more sensitive to the health impacts of noise. Although Group 2 consists of middle-aged individuals with good physical fitness and less home time, these residents still consider health protection necessary47. However, residents in Groups 1 and 2, despite having higher levels of education, show significantly less proactive BR. Therefore, the inconsistency between HIA1 and BR is significant.

From the public’s perspective, a significant barrier exists between residents’ experiences of damage and their ability to effectively express their demands, leading to a vicious cycle of ‘more damage but less participation’. The key to successful SCG lies in breaking this cycle by fostering a virtuous cycle of ‘more participation and less damage’.

Stage 2 (Risk solution)

The maximum p-value for the skewness test is 0.1165, and for the kurtosis test, it is 0.7248, neither of which meets the normality assumption. As a result, the maximum-likelihood method was chosen for estimation48. Regression results (Supplementary Table 9) based on Eqs. (6) and (7) reveal significant effects of BI in hypotheses H2-H5, and significant effects of BR in hypotheses H1, H4, and H5.

Regarding AT, the primary factor driving BI is AT1, rather than AT2. When the probability of achieving optimal results is at its peak, every 1% increase in AT1 leads to a 0.19% increase in BI. This finding indicates that SCG should prioritize addressing residents’ health demands as its central goal.

Regarding SN, SN3 significantly promotes BI. When the probability of actual results is maximized, a 1% increase in SN3 leads to a 0.27% increase in BI. This highlights the strong incentive provided by legal assurances for willingness to participate in SCG, emphasizing the need to strengthen legal systems to better encourage public participation.

Regarding PBC, PBC4 and PBC6 are crucial factors. When the probability of actual results is maximized, a 1% increase in PBC6 leads to a significant 0.34% increase in BI. Notably, PBC4 contributes to increases in both BI and BR, highlighting the importance of information disclosure and feedback in SCG. These results further suggest that relying solely on individual efforts within the SCG framework is insufficient to effectively incentivize public participation.

Regarding HIA, higher HIA1 significantly stimulates both BI and BR, underscoring the strong correlation between SCG and public health. However, as HIA2 increases, BR significantly decreases49. This suggests a potential inconsistency between HIA1 and HIA2, where greater financial commitment to mitigating health impacts does not necessarily translate into increased public participation.

Based on the above analysis, an effective pathway toward achieving an ideal soundscape is established, as illustrated in Fig. 5. The robustness check results are demonstrated in Supplementary Table 10.

Fig. 5: Path diagram of the baseline estimation results.
figure 5

This figure presents the path diagram of the baseline estimation results, illustrating the relationships among key variables in the extended Theory of Planned Behavior (E-TPB) framework.

Full size image

Differences between groups are further explored as follows:

To address the barriers arising from the inconsistencies between health damage experience, participation willingness, and participation behavior in SCG, it is essential to bridge the gaps identified in the previous analysis. Equations (6) and (7) are applied across different Source, Exposure, and Sensitivity categories to identify the underlying issues that affect different groups’ choices in participation behavior (Supplementary Table 11) and to enhance the pathway for SCG (Fig. 6).

Fig. 6: Path diagram of SCG optimization of path estimation results.
figure 6

This figure presents the optimized path estimation results for Soundscape Co-Governance (SCG). It illustrates the adjusted relationships among key variables in the extended Theory of Planned Behavior (E-TPB) framework after optimization.

Full size image

Noise pollution sources

Better understanding of policies can significantly enhance residents’ awareness of the governance system, thereby improving both BR and BI (PBC2 → BR/BI in column ‘Ind’, Supplementary Table 11, Panel A). However, for non-industrial noise sources, there is no significant change in BR or BI (PBC2 → BR/BI in columns ‘Const’, ‘Tran’ and ‘Soc’, Supplementary Table 11, Panel A).

In the case of industrial noise, the environmental policies are legally mandated and well-defined, which aligns with the common interests of neighboring residents and encourages participation in SCG. Conversely, non-industrial noise sources lack fixed or continuous polluters, leading to conflicting interests among residents (SN4 → BR/BI in columns ‘Const’, ‘Tran’ and ‘Soc’, Supplementary Table 11, Panel A). This conflict undermines neighborhood relations and complicates efforts to foster effective SCG participation.

Health exposure

In China’s megacities, residential areas often adjoin traffic roads or intermingle with commercial zones50. Residents living on the periphery of residential areas but closer to their core typically experience the greatest health impacts from noise pollution. However, these residents often show insensitivity to policy support or economic incentives (PBC1/SN1/SN2 → BI/BR in column ‘0–100 m’, Supplementary Table 11, Panel B). The availability and convenience of feedback channels also fail to significantly boost their participation. Their willingness to participate in SCG does not correlate with the health damages they endure (PBC4/PBC6/HIA1 → BI/BR in column ‘0–100 m’, Supplementary Table 11, Panel B). In contrast, residents farther from the core (‘100–500’, ‘500–1000 m’, and ‘>1 km (1000–2000 and >2000 m)’) exhibit more proactive participation in SCG, indicating an inconsistency in participation levels between those closer to noise sources and those farther away.

Population sensitivity

Residents with longer home time and higher incomes (Groups 1 and 3) exhibit significantly higher sensitivity to noise, which directly links HIA1 to BR (HIA1 → BR in column ‘Group 1’, ‘Group 3’, Supplementary Table 11, Panel C). In contrast, Group 4 is more actively involved in SCG compared to other groups. This suggests that individuals with higher education levels may be less enthusiastic about public participation due to high opportunity costs associated with work commitments. This lack of enthusiasm stems from an underestimation of the societal health costs of noise pollution, widening the gap between individual health improvements and broader environmental restoration goals.

For individuals with higher education but lower incomes (Group 2), the effects of AT1 and AT2 on BR can even be contradictory (AT1/AT2 → BR in column ‘Group 2’, Supplementary Table 11, Panel C). This highlights that a ‘high-penalty on high-cost’ approach in NA, requiring polluters to compensate affected residents, could enhance SCG participation among highly educated individuals.

Discussion

From the public perspective, SCG aims for consistency between residents’ self-assessed health impacts, their willingness to participate, and their actual participation behavior. In megacities, however, there is a disconnect between the experienced damage and the expression of demands. Residents who report severe health impacts may not actively participate in SCG, which hampers efforts to achieve the ideal soundscape.

To tackle non-industrial noise pollution, Chinese policies are increasingly favoring less mandatory measures like intensified monitoring, economic compensation, and mediation, which emphasize public participation in SCG. However, these approaches can fall short of replacing the need for institutional authority and may overwhelm government departments with ongoing social conflicts. Effective SCG in megacities requires transforming individual-based public participation into community-based participation. This means supplementing compulsory measures with robust social governance that fosters strong neighborhood relations. Unlike previous studies, which often overlook practical governance tools51, building social platforms to facilitate flexible approaches is highlighted (e.g., improved PBC4/PBC6), helping address non-industrial noise issues effectively.

Residents are often unaware of the shift from industrial to non-industrial noise sources (PBC1 → BR/BI, in Supplementary Table 11, Panel A). This finding is consistent with previous studies, which also identified similar patterns in urban areas undergoing industrial transformation52. In 2022, social-life noise complaints accounted for 67.5% of all noise complaints in China, while industrial noise complaints were only 3.1%. This inconsistency originates from insufficient attention from officials and scholars, who typically focus on more prominent environmental issues. Consequently, the public tends to attribute solutions for new noise pollution types to those for industrial sources. This misconception leads residents to view SCG primarily through the lens of environmental protection rather than health improvement (AT1/AT2 → BI/BR, in Supplementary Table 11, Panel A). Following the amended NPAL issued by China in 2021, which prioritizes public health over environmental restoration, Shenzhen’s local practices should urgently adapt to achieve the ideal SCG.

The extensive governance approach to NA often falls short in addressing the diverse health exposures to noise sources, particularly affecting residents on the periphery of residential areas. The Acoustic Environment Quality Standard (GB 3096-2008) groups ‘mixed residential, commercial, and industrial areas, into a single acoustic functional zone (Type 2), which can lead to increased noise pollution for residents in these areas and complicate NA efforts. Unlike water and air quality, where extensive governance is effective due to minimal individual exposure differences, noise pollution presents unique challenges due to significant variations in individual exposure53. This calls for a refined SCG approach that considers these variations and addresses the disparities in health impacts of self-assessment to ensure a fair and effective achievement of the ideal soundscape.

Megacities, characterized by rapid expansion and dense populations, often experience conflicts between residential life and other activities within the same acoustic functional zones, as well as among households within a single residential area54. This complexity necessitates refined approaches to SCG, such as ecological compensation in NA, which helps promote health equity among residents with varying levels of exposure. Residents closer to the ‘core’ of residential areas tend to enjoy better soundscapes and are more proactive in SCG efforts to maintain these conditions.

The 14th Five-year Noise Abatement Action Plan (NAAP) addresses this issue by mandating the disclosure of noise-related information in the purchase contracts for newly built residential properties, thereby introducing soundscape rights from a market perspective. Despite these efforts, the rapid development of megacities complicates the ability to accurately predict future changes in soundscapes, underscoring the need for dynamic and consultative mechanisms in SCG.

Given the varying health impacts and levels of participation willingness among residents, there is a crucial need to align individual beliefs with public values more consistently. Initiatives like public participation funds and community social networks based on exposure represent meaningful attempts to address these disparities, promoting more equitable and effective SCG practices55.

As Chinese cities continue to expand into megacities, an increasing differentiation in population characteristics highlights the integrated governance facilitating the realization of complementary capabilities. Consequently, three groups warrant attention: individuals with higher education levels, those with a strong willingness to engage in public participation, and residents who spend more time at home. These groups can guide others in understanding the core objectives of SCG, promote proactive behavior, and enhance the participation of younger residents who are often occupied with work.

Additionally, addressing non-industrial noise through dialog between different groups, reinforcing the government-public nexus, enhancing NA mechanisms driven by health goals, and establishing penalties based on health impacts are critical for promoting a harmonious social atmosphere. These efforts can also mitigate the potential negative effects of increased diversification on social cohesion in megacities. Drawing on recent Chinese governance practices, we recommend: (1) establishing a co-governance dialog platform to facilitate collaboration between local governments and residents in noise management, similar to the “citizen-initiated government contacts” model recently proposed, and leveraging advancements in information and communication technology to deepen public participation by enabling individuals to connect with authorities56 and (2) expanding co-governance initiatives to include stricter enforcement mechanisms, such as the use of real-time noise sensors and data analytics, as demonstrated in the Municipal Ecological Environment Bureau’s recent regulations. These measures would ensure compliance with health-based noise standards and foster a more collaborative and effective approach to noise management57.

During this research, conventional conceptual and methodological challenges arose, particularly in defining and measuring health-related risks within SCG. Existing frameworks, such as the FHHRA, offered limited consideration of psychological impacts, making it difficult to capture residents’ perceived risks. Integrating the soundscape perspective added theoretical value but introduced further challenges in quantifying subjective experiences. To address this, we developed a survey-based approach with innovative questions, enabling the incorporation of public perceptions into the analysis and supporting a more comprehensive understanding of health risks in megacity soundscape governance.

It should be noted that due to divergent cultural traditions and development paths in megacities, this study focusing on Shenzhen cannot cover all possibilities. Policy innovation, industrial restructuring, demographic changes, and technological advancements will likely reveal additional deficiencies in SCG, warranting further exploration. Future research should include pluralistic cases and comparative studies of SCG in other megacities to deepen the findings of this study.

Methods

Establishment of theoretical model

The risk assessment and management framework, along with the four-step method of quantitative risk assessment outlined in the 1983 report Risk Assessment in the Federal Government: Managing the Process by EPA58, advocates for integrating traditional scientific evaluation with socio-economic and political considerations to determine the most appropriate regulatory actions. The EPA has subsequently developed the FHHRA.

The theoretical model presented in this study is based on the FHHRA proposed by the EPA. However, the FHHRA has limitations when applied to megacities’ SCG, especially with regard to residents’ health impacts. Specifically, the FHHRA considers factors influencing risk metrics, such as the source and its stressors, exposure pathways/routes, receptors, and endpoints27. While this framework offers a health-oriented approach to risk metrics and endpoints, it overlooks the subjectivity and group differences in soundscape-related health, which is particularly significant in megacities. This paper proposes the following optimization strategy for the FHHRA (Fig. 2):

  1. (1)

    Incorporating expressed damage: Expressed damage is introduced to measure residents’ self-experienced and perceived health impacts of noise, as shown in Fig. 2a. Unlike the FHHRA’s endpoint, which focuses solely on physical impacts, expressed damage accounts for the interaction between mental and physical health. To reflect the impact of receptor sensitivity differences on expressed damage, this study further expands the definition of receptors to include those affected psychologically. This includes considering the influences of individual, family, attitude, and behavioral characteristics of the receptors59. Based on this, the positions of receptors and endpoints in the FHHRA are adjusted, and expressed damage is incorporated as the final measure of noise impacts. This adjustment aims to more accurately reflect the entire process and multi-factor characteristics of soundscape impacts.

  2. (2)

    Extending the TPB to develop the E-TPB: To improve the applicability of the enhanced FHHRA in SCG, this study extends the TPB to develop the E-TPB, as shown in Fig. 2b. The TPB is widely used in behavior prediction and intervention, capable of predicting individuals’ behavioral intentions through AT, SN, and PBC. However, it fails to account for the subjectivity in residents’ self-assessment of health impacts from noise, making it difficult to reveal the underlying reasons for residents’ participation in SCG. To address this gap, this study incorporates HIA as a factor influencing the public’s willingness and behavior to participate in SCG, resulting in an E-TPB model better suited for the SCG scenario in megacities.

To verify the scientific validity and applicability of this innovative framework, the study integrates the transition from expressed damage to HIA, coupling the FHHRA with the E-TPB for the analysis of SCG in megacities, as shown in Fig. 2c. The subjectivity and individual differences in noise health risk assessment facilitate the integration of the FHHRA with the E-TPB. Based on the direct collection of residents’ subjective assessments through surveys, this study concretizes the indicators and their connections between each section of the FHHRA and each dimension of the E-TPB. Furthermore, an econometric model is employed to verify the significance of the relationships among these indicators and to identify the factors influencing residents’ participation in SCG in megacities. This ensures the scientific validity and replicability of the megacity SCG analytical framework developed in this study.

The established model can identify risks and propose solutions for SCG (Supplementary Fig. 1). Its applicability is as follows:

  1. (1)

    NA in megacities: NA is a major environmental challenge in China. According to the 2023 Noise Pollution Abatement Report of China by the Ministry of Ecology and Environment, noise disturbances accounted for 59.9% of all environmental pollution complaints in 2022. In eight out of ten megacities in China, the total number of noise complaints was approximately 1,485,000, making up one-third of all noise complaints in prefecture cities.

  2. (2)

    NA beyond environmental standards: NA involves more than merely meeting environmental standards. In this paper, soundscape is defined by residents’ experiences of the acoustic environment, integrated with their self-assessed health effects from noise pollution. Therefore, achieving the ideal soundscape is the ultimate goal of NA.

  3. (3)

    SCG requires public participation and health improvement: SCG requires not only fostering public participation but also improving residents’ health through collaborative efforts. To achieve this, it is crucial to understand why some residents are unwilling or unable to actively participate in SCG. Therefore, the risks associated with SCG are addressed by examining the inconsistency between residents’ experiences of noise-related damage and their levels of participation. This paper proposes governance-based solutions to optimize SCG.

Sampling

This paper utilizes data from a survey on the current status of NA in Shenzhen, conducted from October to December 2022 by the SEEB. The determination of the sample size is based on Eqs. (C1) and (C2) outlined in Supplementary Note 1. As illustrated in Supplementary Figs. 2 and 3, the analysis sample was generated using the following steps:

First, this study employed a two-stage stratified purposive sampling method to ensure the representativeness and generalizability of the sample. Purposive sampling was chosen because it allows for the deliberate selection of participants based on specific criteria, such as the district of residence, which is particularly useful in urban settings like Shenzhen, where population diversity is high. This method is further supported by its use in similar large-scale social surveys60.

The sampling process was divided into two stages. In the first stage, the population was stratified by district, with the smallest sample sizes for each district determined proportionally based on Shenzhen’s total population by district. Within each district, households were randomly selected, and individuals with age-defined independent behavioral capacity were interviewed, which is commonly used internationally. This process resulted in 778 valid questionnaires.

Of these, approximately 12% of the interviewees reported not experiencing noise pollution and were therefore excluded from this study, which focuses on SCG participation to enhance the sound environment and the self-health condition of the respondents. Consequently, 686 valid questionnaires remained. The fact that 88% of respondents reported experiencing noise pollution highlights the extent of this issue among residents of this megacity.

Supplementary Table 3 (column ‘Noise-affected sample’) presents the descriptive statistics of the 686 valid samples obtained from this survey. The statistical characteristics at both the household and individual levels are consistent with data from the Seventh National Census, making the sample representative of Shenzhen’s population with age-defined independent behavioral capacity. After excluding outliers and focusing on respondents aged 70 and below, 626 valid questionnaires were retained for further statistical and model analysis.

According to the World Health Organization, the starting age for considering individuals as elderly is typically 60. However, due to improved medical care in China, most people between the ages of 60 and 70 remain active in their daily lives. For individuals aged above 70, it becomes challenging to distinguish between underlying health conditions and physical damages caused by noise, adding uncertainty to the assessment of noise-related impacts in this age group.

Gini coefficient and Lorenz curve

This paper employs the Gini coefficient and Lorenz curve to quantify the inconsistency between HIA and BR61. The Lorenz curve plots the cumulative percentage of the population, sorted by XHIA1,i (the choice of HIA1 by respondent i) or Yi (the choice of BR by respondent i). The Gini coefficient is measured with Eq. (1):

$$Gini=\frac{B}{A+B}$$
(1)

In this context, A represents the area between the Lorenz curve and the line of absolute fairness, while B represents the area below the Lorenz curve.

To quantify the sampling uncertainty surrounding the Gini coefficient estimate, a non-parametric bootstrap method with 1000 replications was applied. This method is widely used to generate empirical approximations of the sampling distribution for complex statistics without relying on parametric assumptions. In each replication, the dataset was resampled with replacement, and the Gini coefficient was recalculated. This procedure produced an empirical distribution of the Gini coefficient, from which we derived the standard error and 95% confidence intervals using a normal approximation.

Concentration index and concentration curve

To account for potential external factors not captured by the Gini coefficient, it is essential to verify that the inconsistency between HIA and BR persists even when the surveyed residents are grouped based on source (D1), exposure (D2), and sensitivity (D3). To achieve this, the paper introduces the Concentration Index (CI)62, as illustrated in eq. (2):

$$CI=2COV(X,H)/M$$
(2)

In this context, COV represents the covariance, X denotes the rank of the targeted variables sorted by D1, D2 or D3, H corresponds to the associated XHIA1,i or Yi, and M represents the mean of XHIA1,i or Yi. On the concentration curve, the horizontal axis indicates the cumulative percentage of the population sorted by D1, D2 or D3, while the vertical axis shows the cumulative percentage of XHIA1,i or Yi.

Ordered logit model

To isolate the effects of other variables while identifying the key risks associated with SCG using targeted indicators, this paper employs an ordered logit model and formulates Eq. (3). This approach allows for an exploration of the inconsistencies between HIA and BR across different sources, exposure levels, and sensitivities.

$$logit({p}_{k})={\alpha }_{k}+\beta {D}_{m,n,i}+{\gamma }_{j,k}\cdot {Z}_{j,i}+{\varepsilon }_{i}$$
(3)

In this context, Dm,n,i represents the dummy variable, where m = 1 corresponds to source, m = 2 corresponds to exposure, and m = 3 corresponds to sensitivity. The variable n denotes the specific range grouping within each m. To avoid collinearity, when m = 1,2, the variable set Zj,i includes the variables related to sensitivity. Conversely, when m = 3, Zj,i includes variables related to both source and exposure.

logit(pk) = logit(pk1) when HIA is discussed, with pk1 representing the cumulative probability of XHIA1,i taking the smaller k values, as in Eq. (4):

$${p}_{k}^{1}=p({X}_{HIA{1},i}\le k,|,{D}_{j,i})$$
(4)

When discussing BR, the equation logit(pk) = logit(pk2) is used, where pk2 represents the cumulative probability of Yi taking the smaller k values. This relationship is expressed in Eq. (5):

$${p}_{k}^{2}=p\left({Y}_{i}\le k,|,{D}_{j,i}\right)$$
(5)

By comparing the results of Eqs. (4) and (5), the sources of inconsistency between HIA and BR can be identified. This comparison allows for a clearer understanding of where and how these inconsistencies arise, providing insights into the factors contributing to the differences between HIA and BR.

Structural equation model

This paper constructs a Structural Equation Model (SEM) to estimate causal effects63, which is particularly suited for path analysis in the context of SCG. Path analysis, a form of multiple regression analysis, enables the examination of both direct and indirect effects of multiple independent and interrelated variables64. In this study, the SEM framework is divided into a direct effect model (eq. 6) and an indirect effect model (eq. (7)) for path analysis.

$$\begin{array}{l}{Y}_{i}=\alpha +{\beta }_{BAj}\cdot {X}_{BAj{,}i}+{\beta }_{SNj}\cdot {X}_{SNj{,}i}+{\beta }_{PBCj}\cdot {X}_{PBCj{,}i}+{\beta }_{HIAj}\cdot {X}_{HIAj{,}i}\\ \,\,\,\,\,\,\,\,+\,\omega \cdot {u}_{i}+{\gamma }_{k}\cdot {Z}_{j,i}+{\varepsilon }_{i}\end{array}$$
(6)
$$\begin{array}{l}{u}_{i}=\alpha +{\beta }_{BAj}\cdot {X}_{BAj,i}+{\beta }_{SNj}\cdot {X}_{SNj,i}+{\beta }_{PBCj}\cdot {X}_{PBCj,i}\\ \,\,\,\,\,\,\,+\,{\beta }_{HIAj}\cdot {X}_{HIAj,i}+{\gamma }_{k}\cdot {Z}_{j,i}+{\varepsilon }_{i}\end{array}$$
(7)

In this context:

Yi represents the response provided by respondent i for the BR variable.

ui represents the response given by respondent i for the BI variable.

XBAj,i, XSNj,i, XPBCj,i, and XHIAj,i correspond to the responses for variable j by respondent i related to AT, SN, PBC, and HIA, respectively.

Zj,i includes the variables related to sensitivity, which are derived from the responses to the question concerning variable j by respondent i.

The acronyms used in this paper, along with their corresponding explanations, are provided in Supplementary Table 12.