LIU Haimeng, LIU Huaming, CHENG Yi,＊

a Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing, 100101, China

b College of Resources and Environment, University of Chinese Academy of Sciences, Beijing, 100049, China

c Department of Business Administration, University of Granada, Granada, 18011, Spain

Keywords：

Environmental quality

Environmental

perception

Regional sustainable

development

Inter-organizational

coupling

Human-environment

system

Air pollution

Water pollution

Soil pollution

Big data

A B S T R A C T

The relationship between stakeholders and the environment influences sustainable development and human wellbeing. To illustrate the multistakeholder perceptions of environmental pollution in China, we interpreted a feedback loop in the perception-behavior-environment nexus from the perspective of the coupled human-environment system, measured the differences of environmental perceptions among five stakeholders (the public,government, media, companies, and scientists) and regions (including 31 provinces, autonomous regions, and municipalities in China, with exceptions of Taiwan of China, Hong Kong of China, and Macao of China due to a lack of data) using big data, and made a comparison between the perceptions and the actual pollution situation. The results showed that the five stakeholders exhibited similar perceptions of environmental pollution at the national scale,with air pollution being of most concern, followed by water pollution and soil pollution. There were significant spatial differences in environmental perceptions. All stakeholders in the developed regions in eastern China paid relatively high attention to environmental issues, while those in the northwestern regions paid much less attention. There existed a mutual influence and interaction among the different stakeholders. More attention should be paid to air pollution in Xinjiang Uygur Autonomous Region and Ningxia Hui Autonomous Region, water pollution in Hainan Province, Inner Mongolia Autonomous Region, Heilongjiang Province, and Jilin Province,and soil pollution in Hainan Province, Fujian Province, and Jilin Province.This paper provides a research paradigm on multi-stakeholder environmental perceptions based on big data, and the results provide a background reference for regional environmental governance.

1. Introduction

With the rapid development of urbanization and industrialization, many developing countries are facing serious conflicts between social development and environmental protection (Fang et al., 2016; Burki et al., 2021). Serious environmental pollution causes a loss of human health and wellbeing as well as affects sustainable development at the national and regional levels. For example, 4.51×106deaths globally were attributed to outdoor air pollution in 2019(IHME, 2021), and around 80% of illnesses in developing countries were attributed to unsafe drinking water and waterborne diseases (SDWF, 2017). The perceptions and preferences of people regarding environmental issues affect the behavior and form of governance, and these aspects have a remarkable impact on pollution, industrial emissions,and remediation measures (Islam et al., 2014). The misunderstandings and preferences for environmental issues by policymakers, researchers, and the public may aggravate the global crisis (Rai and Henry, 2016; Abbott et al., 2019).Therefore, research on environmental perception and collaborative participation of different stakeholders concerning environmental governance is extremely important (Baird et al., 2019; Cheng et al., 2021). Polycentric or multistakeholder governance is the new path that needs to be taken if we are to solve the many environmental and climatic problems in today’s world (Newell et al., 2012; Jordan et al., 2015). The basis of polycentric bottom-up governance is to understand the ideas of different stakeholders. Thus, it is necessary to highlight and address multi-stakeholder perceptions for environmental protection and governance. Meanwhile, many types of big data in the internet era that reflect people’s preferences and cognition bring opportunities for illustrating and addressing this issue.

Study on the relationships between multi-stakeholders and the environment needs to be an interdisciplinary research. Multi-stakeholder partnerships are increasingly utilized to implement global sustainable development initiatives in fields such as climate change, biodiversity, and natural disaster prevention (Pattberg and Widerberg,2016). Geographers usually explore the emphasis of environmental perception on spatial decision-making from the perspective of behavioral geography (Amedeo and Golledge, 2004). Inter-organizational coupling was proposed to tackle the coupled human and natural systems, and this concept provides a new idea for conducting research on collaborative environmental governance (Liu et al., 2007; Liu et al., 2020a). The concept of polycentric governance was introduced by economists. It comprises multiple governing authorities at diff erent scales; these authorities do not stand in a hierarchical relationship to each other, but are engaged in self-organization and mutual adjustment (Ostrom,2010; Heikkila et al., 2018). This approach is widely used in many fields, such as coral reef governance (Morrison,2017) and telecoupled resource systems (Oberlack et al., 2018). By means of game theory, many researchers have analyzed the complex relationships existed among different actors in environmental governance, including government, enterprises, farm operators, and citizens (Chen et al., 2019; Chhipi-Shrestha et al., 2019; Ghorbani Mooselu et al., 2019). In addition, many studies have explored environmental perception from the viewpoint of a particular kind of stakeholder (Anderson et al., 2007; Maione et al., 2021).

Most previous studies in China have focused more on stakeholders’ perceptions of air pollution. Du et al. (2019a)introduced the concept of group negotiation equilibrium to determine the equilibrium schemes for reducing air pollution from the perspective of conflict resolution. Zhang et al. (2020) constructed a mixed strategy model of the multi-stakeholders, including the central government, local government, enterprises, and the public, to study the game behavior of haze governance. Huang and Yang (2018) examined public risk perception and policy support related to air pollution problem in China. Huang (2018) illustrated the relative influence of the use of the media and air pollution perception on public trust in local governments. Wang and Cheng (2020) analyzed the effects of media representations on Chinese citizens’ attitudes and beliefs about air pollution. Wang and Jia (2021) examined the influence of social media on environmental governance, particularly its effect on mitigating PM2.5 pollution. In contrast, there are few studies on the environmental perceptions of water and soil pollution. Several related articles have investigated nonpoint source pollution in water conservation zones (Du et al., 2019b), some heavy pollution projects (Cong et al.,2021), and risk perception in soil pollution (Zhou et al., 2020).

Although multi-stakeholder perceptions of environmental pollution have been studied from different aspects, there are still some knowledge gaps. Some studies have focused on certain environmental issues or regions, but comparative research between various environmental problems and different regions is lacking. Most studies in China focus on air pollution and lack multi-stakeholder perception research concerning water and soil pollution, as well as heterogeneity analysis. Also, previous research relies heavily on questionnaires and interviews, thus it is difficult to obtain largescale and multi-regional sample data. Besides, few studies have compared the actors’ perception of the environment with the actual state of the environment.

Considering the above limitations of past research, this paper proposes several potential innovations. First, based on a wide variety of internet big data, we explored the perceptions of the public, government, media, companies, and scientists on air, water, and soil pollution. Second, due to the large area of China, the regional differences of multistakeholders’ perceptions were analyzed. Third, based on the measured data of pollutants, we further compared the attention of different stakeholders and the actual quality of the environment.

The research approach is as follows. To begin with, a theoretical analysis framework of multi-stakeholder interactions on environmental pollution was proposed from the perspective of a coupled human-environment system.Then, based on the Baidu Index, the government website, China National Knowledge Internet (CNKI), and other big data platforms, we measured the levels of attention of the public, government, media, companies, and scientists on environmental pollution. Further, data mining, heatmaps, geographic visualization, and other methods were used to analyze the levels of attention of different stakeholders to air, water, and soil pollution at the national and regional scales, and the differences and correlations among different levels of attention were also analyzed. Finally, by comparing the actual environmental pollution situation with the levels of attention of stakeholders, we obtained the cognitive bias of each stakeholder in the different regions of China, and proposed some implications for the findings.Overall, this paper will be a research paradigm of multi-stakeholder environmental perception based on big data, and the results can provide a background reference for decision-making concerning regional environmental protection and sustainable development.

2. Theoretical analyses

Environmental problems are complex, nonlinear, dynamic, and multi-scale in nature, reflecting the imbalance of development between human beings and nature (Fang et al., 2019, 2021; Liu et al., 2022a). People perceive and interact with the environment in different ways. Due to similar values, goals, and backgrounds, people can be divided into different groups, such as the public, government, media, companies, scientists, Non-Governmental Organizations(NGOs), etc. These groups are all, to a greater or lesser extent, stakeholders of environmental pollution. To solve environmental problems, it is necessary to deal with the relationships among these stakeholders. All stakeholders collectively determine the direction of evolution of the regional environment. To analyze the complex interactions among multi-stakeholders, we refered to some theories and concepts, including the coupled human-environment system and inter-organizational coupling in the natural sciences (Liu et al., 2007; Wang et al., 2018), environmental perception (Tuan, 1990; Steg et al., 2013), consciousness-behavior theory (Guagnano et al., 1995), and the cooperative game (Xue et al., 2015) in the social sciences.

This issue will now be illustrated in the macro framework of the coupled human-environment system (Fig. 1).People’s environmental perception and consciousness determine their behaviors and this then affect the environmental quality (Saarinen and Sell, 1980; Hou et al., 2021a); also, these effects have both positive and negative aspects.Further, based on environmental psychology research, the environmental quality could influence people’s experiences, perceptions, and wellbeing (Steg et al., 2013). In turn, the surrounding environment is also one of the crucial motivations for people, and this could directly affect the various behaviors of people, such as the mode of travel and economic activity (Deckers, 2018). Moreover, these behaviors could also strengthen the peoples’ cognition and stereotypes (Hou et al., 2021b), which are a direct reflection of environmental quality. Thus, the people’s perception, behavior, and environmental quality could interact with each other to form a feedback loop (Fig. 1a).

Fig. 1. Multi-stakeholder interactions in the coupled human-environment system. (a), perception-behaviorenvironment nexus; (b), multi-stakeholder interactions and feedbacks.

Inter-organizational coupling refers to the complex game strategy where there is a trade-off between selfdevelopment and environmental protection among the different organizations and stakeholders (Liu et al., 2020a).The behavior of each stakeholder has its own internal drivers. Specifically, citizens seek to maximize their wellbeing;the government seeks to maintain social justice and security for its citizens; companies seek to maximize profits; the media pursue opportunities for effective communication; and scientists seek to explore cause and effect mechanisms to support the sustainable development of society. Under the influence of these distinct interest preferences, the relationships and feedbacks between various stakeholders and the environmental system may well be very different,and these differences will become more diverse over time and space (Fig. 1b).

The internal driving forces and external interactions with other stakeholders determine environmental attitudes and behaviors (Guagnano et al., 1995). In the face of human living environment, the interaction of these interest groups is a cooperative game. Interdisciplinary research on collaborative networks demonstrates which actors are involved,with whom they collaborate, and how they are tied to the structures of the eco-environmental systems may have profound implications on the actors’ abilities to address different types of environmental problems (Bodin, 2017).Synergies and trade-offs would both need to exist in this collaborative network to achieve environmental sustainability(Akinyi et al., 2021), as embodied by the dotted line in Figure 1b. Some examples of synergies include: increasing public concern for the environment can effectively promote the local government to pay more attention to solving environmental problems through increased investment and changing the types of energy supply and the industrial structure (Wu et al., 2020); media attention concerning environmental events could facilitate an improved efficiency in governmental management and improve the public’s environmental awareness and self-protection ability (Gao et al., 2019). Some examples of trade-offs include: companies need to balance economic benefits and environmental responsibilities in order to meet the government’s environmental regulations and media supervision (Henriques and Sadorsky, 1999); the government often makes trade-offs among economic growth, public opinion, and environmental protection (Pradhan et al., 2017); under the influence of the “tragedy of the commons” (Hardin, 1968), the public might weigh their own interests for the common good of the environment. The key to a sustainable development of the coupled human-environment system is to balance the needs of each stakeholder and realize effective cooperation.The premise to achieving this goal is to deeply understand the perceptions and preferences of multi-stakeholders,which is also the main purpose of this study.

3. Data and methodology

As defined by the Oxford English Dictionary, the term perception means “an idea, a belief or an image you have as a result of how you see or understand something” (Soanes and Stevenson, 2004). The term attention is an important part of human perception, and the focus and concentration of one’s consciousness are the essential elements (Eysenck and Keane, 2020). To some extent, a person’s perception of an object can be measured on the basis of the level of attention (McBride and Cutting, 2017). Thus, this paper considers environmental perception through the attention of different stakeholders about environmental pollution and problems.

Questionnaires and interviews have mainly been used previously to measure the perceptions of different people(Du et al., 2019b; Maione et al., 2021). However, both of them have many shortcomings. First, the sample size may be very limited, usually dozens to thousands of samples. Second, the geographical coverage is limited, and many studies only focus on one city or one region. Third, questions concerning the key interests of the investigators may not be answered truthfully. To avoid these deficiencies, we tried to illustrate multi-stakeholder perceptions of environmental pollution based on big data. In recent years, multivariate big data have been widely used in studies on the eco-environment, risk management, human behavior, and other aspects (Wu et al., 2016; Guo et al., 2020; Wang et al., 2021; Yin et al., 2021). Based on multiple internet big data platforms, we crawled relevant keywords and used data mining approaches to study the attention given by different people to environmental pollution in 31 provinces,autonomous regions, and municipalities of China. It should be noted that Taiwan of China, Hong Kong of China, and Macao of China were not analyzed due to a lack of data. These big data were generated by internet users that accounted for more than half of the country’s population, covering all 31 provinces, autonomous regions, and municipalities in China, thus being timely and giving better objectivity and dynamic information. The specific data and methods are outlined in the next section. We selected all the data from 2018 to 2021 based on availability, and the spatial dimension included data at the national and regional scales.

3.1. Data preparation and preprocessing

3.1.1. Attention data for different stakeholders

“Baidu” is the most popular search engine in China. There are approximately 0.795×109Chinese internet users who search on-line, and 70% of them used “Baidu” to search in 2021 (China Internet Network Information Center,2021). The Baidu search volume is provided as a weighted index, which is available to the public as the Baidu Search Index (http://index.baidu.com) (Yin et al., 2021). Based on this index, we can detect the public’s concerns about different types of environmental pollution, temporal trends, and regional differences. The Baidu Media Index is also an open service provided by “Baidu”, reflecting the attention, reporting, and reprinting of specific keywords by various media on the internet (Wang and Loo, 2019). Based on this index, we can realize how much attention Chinese media pay to diverse environmental events and pollution issues and incidents. Using these two indices, we searched air pollution (haze), water pollution, and soil pollution from 2018 to 2021, and obtained the attention data of the public and media for different pollutants nationwide and in each region.

The governmental perception of all types of environmental pollution is reflected mainly in various official documents or public reports, including policies, planning, regulations, service information, conference reports,administrative penalties, etc. Therefore, the frequency of different keywords appearing in the headlines of government web pages in 2018-2021 was adopted to reflect governmental attention, such as in the central government website(http://www.gov.cn/) and the Hubei Provincial People’s Government website (https://www.hubei.gov.cn/).

From a macro perspective, more companies provide products or services for a certain pollutant, indicating that the companies in this region are more concerned about this kind of pollution. “Qichacha” is a website that aggregates various kinds of information about almost all registered companies in China, including name, legal representative,address, type, business scope, etc. (https://www.qcc.com/). Based on the database of this website, we crawled the data reflecting the corporate attention on environmental pollution in each region (province, autonomous region, or municipality). We screened each company’s scope for business. The keywords for identification included: “air pollution”+“air purification”; “water pollution”+“water purification”; and “soil pollution”+“soil remediation”. Then,the number of relevant companies in a particular region was summed, giving an entry for corporate attention for the region on the corresponding environmental problem.

The scientists’ perception to various environmental pollution issues may be reflected in their research outputs. The CNKI is the largest platform for unified retrieval, online reading, and downloading of literature resources in China(https://www.cnki.net/). It is a Chinese database that integrates journals, theses, conference papers, newspapers,books, patents, and standards, and contains the academic publications of most Chinese researchers. By searching the keywords for each pollutant “+” the names of the regions (provinces, autonomous regions, and municipalities), it is possible to obtain the scientists’ attention to different pollution problems in the different regions of China.

3.1.2. Air, water, and soil pollution data

PM2.5 is the most serious air pollution issue in China (Liu et al., 2022b). The regional PM2.5 data for 2018-2021 were based on the annual mean PM2.5concentration of 1509 environmental monitoring stations across China merged by calculating the average value via ArcGIS software. These monitoring data came from the Qingyue Data(https://data.epmap.org). The water pollution data were derived from the Water Quality Index published by the Institute of Public and Environmental Affairs (IPE) (http://www.ipe.org.cn/). This index is a comprehensive assessment of surface water, drinking water sources, and groundwater, with a higher index commensurate with more serious water pollution.

The soil pollution of farmland is serious in China. Research has shown that 22% of farmland soils in China are polluted to varying degrees (Zeng et al., 2019). China consumes nearly one-third of the world’s fertilizer, and the pesticide usage per unit area is 2.5 times the world average, hence the combined load has a great influence on soil pollution (Chen et al., 2014; Lu et al., 2015; Zhang and Wang, 2020). Given that the Chinese government has not released comprehensive data on soil pollution, this paper used the intensity of fertilizer and pesticide application to indirectly reflect the soil pollution situation (Delang, 2017). The intensity of fertilizer application refers to the amount of fertilizer applied per unit area of crops, and the intensity of pesticide application refers to the amount of pesticide used per unit area of crops. We standardized the data usingZ-score normalization and then calculated the sums of the two indicators giving the same weighting to each source (Fang et al., 2015). Relevant data were from the China Statistical Yearbook (National Bureau of Statistics, 2018-2021a), the China Rural Statistical Yearbook (National Bureau of Statistics, 2018-2021b), and the annual statistical reports of the provinces, autonomous regions, and municipalities.

3.2. Correlation analysis

There are many specific methods for correlation analysis. Considering that the purpose of this study is to identify the linear correlation between two types of continuous variables, we finally selected Pearson’s correlation analysis.The formula for the Pearson’s correlation coefficient (r) is as follows:

where,xandyrepresent two types of variables; andandare the average values of the two variables. The absolute value ofrreflects the strength of the correlation. Whenr＞0, it indicates that the two variables are positively correlated; whenr＜0, it indicates that the two variables are negatively correlated.

3.3. Heatmap analysis

The heatmap is a common visualization method that is widely used in various big data analysis scenarios on account of the rich color changes and the clear presentation of relevant information (Liu et al., 2020b; Ye et al., 2021). A heatmap can describe the values of the primary variables of interest across two axis variables as a grid of colored squares. The axis variables can be divided into different sections in terms of color, and the color of each cell indicates the values of the major variables. The heatmap is an effective way to show the relationship between two variables. It is, therefore, possible to observe if there are any patterns in the values of one or more variables by observing the color changes of the cells across each axis.

4. Results

4.1. Multi-stakeholder perceptions of environmental pollution at the national scale

The multi-stakeholder perceptions on different environmental issues based on the average of all data from 2018 to 2021 are shown in Figure 2. The five interest groups have similar perceptions to environmental pollution at the national scale; that is, air pollution was of most concern, followed by water pollution, and then soil pollution. Due to the visibility and spread of air pollution, it can be experienced directly and receives widespread attention from all groups. Most stakeholders exhibited more than a 50% level of attention to air pollution, among which the media paid the highest attention, with the value reaching up to 70%. The attention to water pollution was lower than that to air pollution, and the overall value was about 30%. With officials facing performance appraisals and given that companies are required to meet the sewage discharge standards, the government and companies tended to pay more attention to water pollution, accounting for 37% and 32%, respectively. All groups paid less attention to soil pollution, probably because soil pollution was not readily discerned and there was a time lag between the discharge of the pollutant and its deleterious effect. Among the five interest groups, companies paid more attention to soil pollution, an attention value of 27% being recorded.

Fig. 2. Levels of attention by multi-stakeholders for the three major sources of environmental pollution. (a), public;(b), government; (c), media; (d), companies; (e), scientists.

Different stakeholders paid different levels of attention to the different pollution sources. Companies had the most balanced focus on the three types of pollution, which may be a reflection of the actual pollution situation given that many companies are the front-line actors in addressing pollution. Media’s attention to the three types of pollution was the most uneven, with the level of attention to air pollution accounting for 70% and soil pollution only 2%. The proportions of the public, government, and scientists showing concerns for air pollution were 53%, 54%, and 54%,respectively, indicating that these three groups had a similar attention preference with respect to air pollution. In the case of the government, it also paid less attention to soil pollution, with the level of attention being only 9%.

4.2. Spatial differences in the multi-stakeholder perceptions of environmental pollution

There were significant spatial differences in the attention of the different groups to environmental pollution (Fig.3). Overall, all stakeholders in the developed regions in the east of the country paid relatively high attention to pollution, while those in the northwestern regions paid relatively little attention. Based on the consideration of the coefficient of variation values for the attention of the various groups to environmental pollution (Fig. 3f), the regional variation was greatest for companies, followed by the media and then the scientists; there was a relatively small level of regional variation for the public and government.

Fig. 3. Spatial differences in the perceptions of multi-stakeholders with respect to environmental pollution. Note that the above maps are based on the standard map (No. GS (2019) 1823) marked by the Ministry of Natural Resources of the People’s Republic of China, and the base map has not been modified. Taiwan of China, Hong Kong of China,and Macao of China are not included in the scope of research area due to a lack of data.

A survey on the public’s perceptions of environmental pollution showed that, Beijing City attracted the most public attention, followed by the coastal regions such as Jiangsu Province, Guangdong Province, and Shandong Province,and Tibet Autunomous Region least concerned (Fig. 3a). In addition, Henan Province, Sichuan Province, and Shaanxi Province also had relatively high levels of public attention. In terms of preferences, the order of public attention in all regions was “air pollution＞water pollution＞soil pollution”. The level of public attention on air pollution in Beijing,Hebei Province, Shanxi Province, Henan, and Shandong accounted for more than 50%.

From the perspective of local governments, Zhejiang Province paid the highest attention to environmental pollution,followed by Shaanxi, Tianjin City, Hunan Province, and Shandong, while other regions such as Xinjiang Uygur Autonomous Region, Fujian Province, and Qinghai Province paid less attention. There were great differences in the attention preferences among local governments. Most regions in central China and northern China paid the most attention to air pollution. Among them, government of Henan was particularly concerned with air pollution, with the level being up to 82%. Anhui Province, Fujian, Guangdong, Hubei Province, Jilin Province, Jiangsu Province,Qinghai, Tianjin, Tibet, Yunnan Province, and Chongqing City paid the most attention to water pollution. Hunan and Zhejiang paid more attention to soil pollution than air pollution or water pollution (Fig. 3b).

The media paid more attention to urban agglomeration and heavily polluted areas, such as Beijing, Shanghai City,Hebei, Shandong, Tianjin, and Guangdong, while little attention was paid to the Tibetan Plateau (Fig. 3c). Consistent with the public’s perceptions, the order of media attention to environmental pollution in all regions was “air pollution＞water pollution＞soil pollution”. Moreover, the level of attention shown by the media to soil pollution in every region was less than 3%.

The companies in Shandong, Shanghai, Guangdong, and Beijing paid much more attention to environmental pollution, while those in Qinghai, Ningxia Hui Autonomous Region, and Tibet with fewer companies paid correspondingly less attention. Air pollution in most regions was of concern to companies. Meanwhile, companies in Fujian, Zhejiang, Hainan Province, Sichuan, Beijing, and Yunnan were most concerned about water pollution, and those in Inner Mongolia Autonomous Region, Gansu Province, Qinghai, and Xinjiang were most concerned about soil pollution (Fig. 3d).

Scientists paid more attention to densely populated areas and heavily polluted areas, and less attention was given to western regions of the country. The attention paid to Beijing was particularly high, being more than twice that of Shanghai, which ranked the second. Scientists paid the most attention to air pollution in most regions, however, they paid more attention to soil pollution in the case of Guangxi Zhuang Autonomous Region, Guizhou Province, and Yunnan (Fig. 3e).

4.3. Coupled analysis of the attention of multi-stakeholders to the various sources of environmental pollution

Given that the levels of attention of different stakeholders to environmental pollution varied, the logarithm of all the data was computed. Then, thervalues between the pairs were calculated. The value ofrand the 95% confidence ellipse were included in the results; and the bar charts showed the overall distribution of the levels of attention for each stakeholder with respect to the sources of pollution.

4.3.1. Air pollution

With regard to the level of attention to air pollution, each group had certainrvalues, but the degrees of significance were very different between groups. The media was most relevant to the other groups, followed by the public. (Fig.4). The correlations between the government and the other groups were the weakest, with the value ofrlower than 0.5. Scientists were also less relevant to the other groups. Among all the relationships, the correlation of the levels of attention to air pollution between the public and media was the highest (r=0.96), followed by the correlation between the media and companies (r=0.92), and then the correlation between the public and companies (r=0.90).

Fig. 4. Correlations of the levels of attention for different stakeholders (the public, government, media, companies,and scientists) to air pollution.

4.3.2. Water pollution

With respect to the level of attention to water pollution, there was a certain correlation among all stakeholders(Fig. 5). The media and public perceptions were the most consistent (r=0.95). Second, there was also a good consistency between the media and companies (r=0.87). However, the correlations between the government and other groups were weak (r＜0.50). Besides, the correlations of the companies with the public, media, and scientists were relatively high (r＞0.60). Further, the correlations of the scientists with the public, media, and companies were also relatively high (r＞0.60).

Fig. 5. Correlations of the levels of attention for different stakeholders (the public, government, media, companies,and scientists) to water pollution.

4.3.3. Soil pollution

In addition to the government, there was a certain correlation between the levels of attention of the other groups with respect to soil pollution, but the degree of significance varied greatly (Fig. 6). The correlation between the public and media was the highest (r=0.96), while the correlation between the government and companies was the lowest (r=-0.04). The correlations among the public, companies, and media were relatively high (r＞0.80). Scientists exhibited a moderate level of relevance with the other four groups.

Fig. 6. Correlations of the levels of attention for different stakeholders (the public, government, media, companies,and scientists) to soil pollution.

4.4. Comparison between the actual pollution situation and multi-stakeholder perceptions

4.4.1. Air pollution

The heatmap, which reflects the differences between the actual air pollution and multi-stakeholder perceptions at the regional scale, is presented in Figure 7. The colder the color is, the lower the pollution level and the level of attention, andvice versa. Overall, the ranking of the level of attention for each stakeholder to air pollution was similar to the actual pollution situation. Air pollution in Xinjiang was serious, but the level of attention given by stakeholders was not enough; Ningxia was also heavily polluted, but the public, media, and companies paid little attention to air pollution. In addition, although the air quality of Guangdong, Fujian, and Zhejiang was much better, various stakeholders paid more attention to this topic in these regions. Specifically, Guangdong and Zhejiang ranked high in terms of air quality, but their citizens were very concerned about air pollution. From the government perspective, the governments of Anhui, Hubei, and Jiangsu paid relatively little attention to air pollution; although Inner Mongolia,Heilongjiang Province, and Guizhou had better air quality, their governments still paid much attention to air quality.From the perspective of the media, reports on air quality in Guangdong and Zhejiang were more numerous. In terms of the companies, the companies in Tianjin, Shanxi, Xinjiang, and Ningxia paid little attention to air pollution. As far as the scientific research, air pollution research in Hubei, Hunan, Jiangxi, and Xinjiang was insufficient.

Fig. 7. Comparison between the level of actual air pollution (PM2.5) and the level of attention of multi-stakeholder perceptions to air pollution at the regional scale. The regions included 31 provinces, autonomous regions, and municipalities in China, with exceptions of Taiwan of China, Hong Kong of China, and Macao of China due to a lack of data.

4.4.2. Water pollution

The heatmap reflecting the difference between the actual water pollution and multi-stakeholder perceptions at the regional scale is presented in Figure 8. In general, water pollution in Hainan, Inner Mongolia, Heilongjiang, and Jilin was relatively serious, but all groups in these regions paid minimal attention on water pollution. The people who were most concerned about water pollution resided in Guangdong, Jiangsu, and Beijing, although water quality in these regions was not poor. In addition, although water pollution in Hunan, Sichuan, and Zhejiang was not serious,stakeholders paid high attention to water quality in these regions. From the public perspective, water pollution in Tianjin, Shanxi, Hainan, Inner Mongolia, Heilongjiang, Jilin, and Ningxia was serious, but local people paid little attention to it. From the governments’ perspective, Hebei, Hainan, Henan, and Anhui suffered from serious water pollution, but local governments paid little attention to it; although water pollution in Guizhou, Hunan, and Sichuan was not serious, the local governments were very concerned about the water environment. From the media perspective, Hainan, Inner Mongolia, and Ningxia had serious water pollution, but there was not much media coverage on this topic in these regions. In terms of the companies, Tianjin, Shanxi, Inner Mongolia, Heilongjiang, Jilin, and Ningxia suffered from severe water pollution, but there were relatively few water treatment companies in these regions. From the perspective of the scientific study, researchers have not done sufficient work in Hainan,Heilongjiang, Jilin, and Henan.

Fig. 8. Comparison between the level of actual water pollution and the leve of attention of multi-stakeholder perceptions to water pollution at the regional scale. The regions included 31 provinces, autonomous regions, and municipalities in China, with exceptions of Taiwan of China, Hong Kong of China, and Macao of China due to a lack of data.

4.4.3. Soil pollution

The heatmap illustrating the difference between the actual soil pollution and multi-stakeholder perceptions at the regional scale is presented in Figure 9. Soil pollution in Hainan, Fujian, and Jilin was relatively serious, but various interest groups did not pay enough attention to it. Soil pollution was relatively mild in Sichuan and Chongqing,however, stakeholders attached quite a high level of attention to the topic. From the public’s perspective, soil pollution in Hainan, Guangxi, and Jilin was serious, but the public in these regions paid little attention to it. From the government perspective, soil pollution in Hainan, Henan, Hubei, Jiangsu, and Anhui was relatively serious, while local governments in these regions paid little attention to it. From the media perspective, soil pollution in Fujian,Hainan, Guangxi, and Jilin received minimal concern from the media despite its serious levels. In terms of the companies, there were more soil remediation companies in most regions with serious soil pollution, despite this, there were fewer such companies in Hainan and Beijing. From the perspective of the scientists, soil pollution research in Hainan, Fujian, Hubei, Liaoning, and Xinjiang was insufficient.

Fig. 9. Comparison between the level of actual soil pollution and the level of attention of multi-stakeholder perceptions to soil pollution at the regional scale. The regions included 31 provinces, autonomous regions, and municipalities in China, with exceptions of Taiwan of China, Hong Kong of China, and Macao of China due to a lack of data.

5. Discussion

5.1. Policy implications

Environmental problems are major challenges threatening the health, economic prospects, and food and water supply of people across the world (Martin et al., 2021). Human perception and behavior are the crucial driving factors of these problems. If we are to secure significant improvement in environmental quality, an intrinsic knowledge of people’s perceptions and wider value judgments are essential (Bickerstaff and Walker, 2001). The environmental perceptions and value judgment of the above results have many implicationsfor China and the rest of the world.

Environmental governance needs the balance and cooperation of multi-interest organizations, but this kind of game is not a zero-sum game, because environmental improvement is beneficial to all stakeholders. The government should play an active role in the mutual influence and interaction between different stakeholders. Decision-makers need to coordinate the multi-interests among the stakeholders and emphasize the importance of stakeholder consultation in order to meet the needs of local stakeholders. Environmental knowledge among the actors should be strengthened and public consciousness should be changed in order to achieve strong sustainability (Neumayer, 2003), thus avoiding the tragedy of the commons through a frank policy of mutual coercion and cooperation (Hardin, 1998). Companies should improve social responsibility and environmental awareness and develop a long-term vision. Scientists should not only address and pursue “hot topics” but also be the first to anticipate future environmental problems and provide support for policymaking.

Air pollution in China has attracted significant attention in recent years, possibly because it is easier to see and perceive directly. As air quality has improved dramatically in recent years, all stakeholders in China should pay more attention to soil and water pollution. Regionally differentiated policies are a promising instrument to mitigate severe pollution. For specific provinces (autonomous regions and municipalities), more attention should be paid to air pollution in Xinjiang and Ningxia, water pollution in Hainan, Inner Mongolia, Heilongjiang, and Jilin, and soil pollution in Hainan, Fujian, and Jilin.

5.2. Research limitations and prospects

The uncertainty of the results comes mainly from the data. The public perception of the data based on website retrieval may make it difficult to reflect the views of the children and the elderly, who rarely use the internet; these groups are more vunerable to environmental pollution. The government’s concern about environmental pollution needs to be further examined because of the different levels of complexity and openness of information on government websites. The data on companies may be biased as a result of summarizing the number of companies. In spite of the large volume of data, big data have limited attributes; although the data are difficult to pinpoint cause and effect relationships, they are suitable for correlation analysis (Mayer-Schönberger and Cukier, 2013). The data used in this research reflect one aspect of environmental perception. Despite these limitations, the significance of this study is that it provides a research paradigm of multi-stakeholder perceptions on the environment, which can be referenced by the research community and other interested parties.

This paper highlights the multi-stakeholder perceptions of environmental pollution based on data visualization,comparison, and correlation analysis. Why do people in different regions have different perceptions when facing the same environmental problem? Is this related to the education level of people and local economic development? How do we combine big data with traditional survey data and better deduce the interaction mechanisms between the different stakeholders? These questions deserve to be studied further. Environmental stakeholders and environmental perceptions play crucial roles in the coupled human-environment system, but most eco-environmental researchers have not yet realized their importance. In the future, such inter-organizational couplings should be better embedded in the framework of Earth system science, and multi-disciplinary integration initiatives should be strengthened. We should explore both the common characteristics and the regional heterogeneities of multi-stakeholders in order to better describe more precisely the driving effects of people with different backgrounds.

6. Conclusions

With reference to the current literature, the potential contributions and significance of this research are as follows:(1) a feedback loop in the perception-behavior-environment nexus was interpreted from the perspective of the coupled human-environment system; (2) big data were used to reflect the environmental perceptions of different stakeholders instead of conducting interviews and questionnaire-type surveys; (3) the differences of environmental perceptions among the different stakeholder groups and regions (provinces, autonomous regions, and municipalities) were measured; and (4) a comparison was made between actual pollution situation and multi-stakeholder perceptions of the environment.

We found that the five stakeholders (the public, government, media, companies, and scientists) exhibited similar levels of attention with respect to environmental pollution at the national scale, with air pollution being of most concern, followed by water pollution, and then soil pollution. Companies tended to have a more balanced focus on air, water, and soil pollution, while the level of attention of the media was the most variable. There were significant spatial differences in the perceptions of different stakeholders to environmental pollution. All stakeholders in the developed regions of eastern China exhibited relatively high attention, while those in the northwestern regions paid much less attention. Regional variation was greatest for the companies, followed by the media and scientists; it was relatively small for the public and government.

Certain correlations with respect to environmental perception were exhibited by the different stakeholders but with differing degrees of significance, which indicated that there existed a mutual influence and interaction among different stakeholders. The media was most relevant to the other groups, followed by the public. The correlations between government and the other groups were the weakest.

The ranking with respect to the attention of each stakeholder to air pollution was similar to the actual pollution situation. Water pollution in Hainan, Inner Mongolia, Heilongjiang, and Jilin was relatively serious, but all groups in these regions did not pay enough attention to water pollution. Soil pollution in Hainan, Fujian, and Jilin was relatively serious, and the various interest groups did not pay enough attention to this problem. This paper provides a research paradigm of multi-stakeholder environmental perceptions based on big data and gives insights to other related studies.The results provide a background reference for regional environmental governance in China.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This work was financially supported by the National Natural Science Foundation of China (42171210), the Research Funding for the Second Comprehensive Scientific Investigation of the Qinghai-Tibet Plateau(2019QZKK1005), and the Start-up Research Program of Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences (E0V00102YZ).