Enhancing the Efficiency of Enterprise Shutdowns for Environmental Protection: An Agent-Based Modeling Approach with High Spatial-Temporal Resolution Data

Qi Zhou , Shen Qu , Miaomiao Liu , Jianxun Yang , Jia Zhou , Yunlei She , Zhouyi Liu , Jun Bi

Engineering ›› 2024, Vol. 42 ›› Issue (11) : 295 -307.

PDF (2357KB)
Engineering ›› 2024, Vol. 42 ›› Issue (11) :295 -307. DOI: 10.1016/j.eng.2024.02.006
Research
Enhancing the Efficiency of Enterprise Shutdowns for Environmental Protection: An Agent-Based Modeling Approach with High Spatial-Temporal Resolution Data
Author information +
History +
PDF (2357KB)

Abstract

Top-down environmental policies aim to mitigate environmental risks but inevitably lead to economic losses due to the market entry or exit of enterprises. This study developed a universal dynamic agent-based supply chain model to achieve tradeoffs between environmental risk reduction and economic sustainability. The model was used to conduct high-resolution daily simulations of the dynamic shifts in enterprise operations and their cascading effects on supply chain networks. It includes production, consumption, and transportation agents, attributing economic features to supply chain components and capturing their interactions. It also accounts for adaptive responses to daily external shocks and replicates realistic firm behaviors. By coupling high spatial-temporal resolution firm-level data from 18 916 chemical enterprises, this study investigates the economic and environmental impacts of an environmental policy resulting in the closure of 1800 chemical enterprises over three years. The results revealed a significant economic loss of 25.8 billion USD, ranging from 23.8 billion to 31.8 billion USD. Notably, over 80% of this loss was attributed to supply chain propagation. Counterfactual analyses indicated that implementing a staggered shutdown strategy prevented 18.8% of supply chain losses, highlighting the importance of a gradual policy implementation to prevent abrupt supply chain disruptions. Furthermore, the study highlights the effectiveness of a multi-objective policy design in reducing economic losses (about 29%) and environmental risks (about 40%), substantially enhancing the efficiency of the environmental policy. The high-resolution simulations provide valuable insights for policy designers to formulate strategies with staggered implementation and multiple objectives to mitigate supply chain losses and environmental risks and ensure a sustainable future.

Graphical abstract

Keywords

Agent-based model / Supply chain network / Economic sustainability / Environmental policy

Cite this article

Download citation ▾
Qi Zhou, Shen Qu, Miaomiao Liu, Jianxun Yang, Jia Zhou, Yunlei She, Zhouyi Liu, Jun Bi. Enhancing the Efficiency of Enterprise Shutdowns for Environmental Protection: An Agent-Based Modeling Approach with High Spatial-Temporal Resolution Data. Engineering, 2024, 42 (11) : 295-307 DOI:10.1016/j.eng.2024.02.006

登录浏览全文

4963

注册一个新账户 忘记密码

1. Introduction

Effective environmental management is urgently needed in China to address the growing environmental challenges. However, the ability to adequately evaluate the impact of environmental policies is limited. Current policy evaluations are typically based on direct costs and benefits [1], [2], which do not consider the broader implications for the supply chain network. As a result, policymakers may underestimate the policy effects and fail to realize the synergistic management of economic and environmental risks. Following the national policy, many provinces and cities in China have recently implemented rectification and shutdown plans for chemical parks and enterprises to mitigate environmental risks. For example, after the “3.21” chemical plant explosion in Jiangsu Province, the Jiangsu Provincial Government implemented an environmental policy requiring closing 60% of chemical parks and more than 1000 chemical enterprises in the province.

The closure of chemical enterprises resulted in a dual effect on the supply chain. First, the shutdown of chemical enterprises has shrunk the production of downstream customers because of a shortage of raw materials [3], such as butanol, glyphosate, acetic acid, and urea. Second, it also has caused a loss of orders sent to upstream suppliers due to a lack of demand. The combination of the two effects could potentially cause small region- or industry-specific shocks, significantly impacting the entire supply chain network [4], [5], [6]. For example, in the first half of 2021, approximately 50 chemical enterprises in Jiangsu Province closed during their first year of operation, and nearly 50% lasted less than six months. Moreover, since the beginning of 2023, chemical enterprises have announced periodic maintenance or production cutbacks. Over 50 chemical enterprises have announced shutdowns for maintenance or production reductions to cope with supply chain instability, impacting a production capacity exceeding 20 million tonnes. Due to large-scale and rapid changes in the operating status of enterprises under top-down environmental management, high-temporal resolution models are required to evaluate the supply chain risks and balance environmental and economic consequences.

Commonly used models for measuring supply chain risks, such as the input-output (IO) model or computable general equilibrium (CGE) model [7], [8], [9], [10], are typically used for interannual simulations, although they cannot capture the daily changes of the environmental economic system under external shocks. For instance, the IO model assumes a fixed technology level (fixed direct consumption coefficient matrix) and does not consider the possibility of unaffected enterprises increasing their output to supplement the production loss of affected enterprises. As a result, the IO model underestimates the elasticity of the economic system and overestimates the losses. Similarly, the CGE model balances demand and supply based on the price [11] and is more suited for analyzing longer time steps (1-10 years) [12], making it less effective for characterizing daily changes in the environmental economic system. Moreover, previous studies that evaluated the operating status of enterprises focused on natural disasters or public health events [4], [13], [14] and utilized solely small-scale questionnaire survey data [15], [16], [17], [18]. These approaches cannot capture the high-frequency events of enterprise entries and exits and their combined impacts on the supply chain network [4], [19], [20].

The focus in recent years has been the development of daily-scale assessment models for the dynamic simulation of economic systems, considering complex socio-economic structures and supply chain networks in various industries. Notable examples, such as the adaptive IO model [12], [21], [22], [23], Acclimate [24], [25], [26], [27], [28], and network dynamical model [5], have overcome the limitations of simple agent behavior and long simulation steps of the static IO and CGE models. A comprehensive summary of the literature on related models is presented in Table 1 [4], [5], [12], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32]. However, it should be noted that previous agent-based modeling efforts have not investigated the synergies and tradeoffs between environmental management and supply chain resilience, particularly in implementing firm-level environmental policies in developing countries such as China. Therefore, there is a significant need to develop an agent-based model that can address the balance of environmental, social, and economic factors in the supply chain network using firm-level operating data with a finer temporal resolution.

This study coupled a dynamic agent-based supply chain model with high spatial-temporal resolution firm-level data on environmental risks and operating information to investigate the daily dynamics of the economic and environmental consequences resulting from enterprise closures due to environmental policies. The environmental risk was assessed using environmental penalties and safety violations related to air pollution, water pollution, solid waste pollution, sudden risk, and safety production. We focused on a local policy issued after the “3.21” chemical plant explosion in Jiangsu Province, China, requiring the closure of chemical enterprises to verify the model’s feasibility and examine the shock propagation through supply chains. A case study of 18 916 chemical enterprises in a city-level multi-regional supply chain network was used to evaluate the daily economic and environmental consequences of the policy in the country from January 1, 2019 to December 31, 2021. Sensitivity analyses and validations were conducted to verify the model’s robustness. Scenario analyses were performed to conduct multi-objective policy optimization, considering environmental protection, production safety, and economic output.

The paper is structured as follows. Section 2 provides an overview of a universal model framework for evaluating the economic and environmental consequences of enterprise shutdowns. Section 3 presents the case study focusing on an environmental regulation policy targeting enterprises in the chemical industry in Jiangsu Province, China. Section 4 presents the simulation results of the case study, including model validation, multi-objective policy optimization, and the discussion. Section 5 provides a summary of the key findings and concludes the paper.

2. Model

2.1. Model framework

We developed a universal model framework to assess the day-to-day dynamics of the economic and environmental consequences of enterprise shutdowns due to environmental policies. First, a standardized enterprise database was constructed using data crawlers and geographic encoding methods. This process involved big data mining and text analysis to determine the shutdown and rectification dates and the shutdown durations for the enterprises. Second, a dynamic agent-based complex network model comprising production, consumption, and transportation agents was established. This model incorporated the daily dynamic changes in the operational status of enterprises to assess the evolving economic risk in sectors and regions resulting from the enterprise shutdowns and rectifications. Subsequently, an evaluation framework for environmental risk indicators was formulated to analyze the daily evolution of environmental risk in sectors and regions. Lastly, a multi-scenario parameter sensitivity analysis and multi-objective policy optimization were conducted. The universal framework serves as a robust support system for the design of environmental policies.

2.2. Enterprise shutdown information mining

Multiple enterprise databases, such as the national enterprise credit information publicity database, the China energy statistics database, the second national pollution source survey databases, and others, were used to collect information from diverse perspectives.

A flowchart of the data sources and mining procedures is illustrated in Fig. S1 in Appendix A. All databases were matched and combined using the enterprises’ names and unified social credit codes. The longitudes and latitudes of the enterprises were obtained using the geocoding function of the Baidu Map open platform [33] based on the enterprises’ addresses. During data cleaning, the missing economic information of some enterprises was filled in using the inferred linear relationship between the economic production capacity and the paid-in capital (or registered capital when the paid-in capital is missing). The closure of an enterprise and the shutdown date were inferred based on the operating status and historical records of changes in industrial and commercial information. The date when an enterprise was removed from the list of its industries was considered the shutdown date. Since the rectification dates were difficult to infer using the database, it was assumed that enterprises would rectify randomly during the study period. According to the measures for restricting production and stopping production [34], the production restriction cannot exceed three months. Thus, it was assumed that the rectification lasted two months in the main analysis, and other durations were used in the scenario analyses. The data crawler was implemented in the Rstudio platform (R version 4.1.2 and Rstudio version 2021.9.1.372) using the Rselenium and rvest packages to collect enterprise information.

2.3. Agent-based complex network modeling

The data on the enterprises’ operating conditions obtained from multi-source data mining was used to calculate the daily production losses and their propagation caused by the downsizing of the production capacity due to enterprise shutdown using a dynamic agent-based supply chain model. The losses were the direct results of the reduction in daily production capacity of the studied industry (first-order effects) or the indirect results of economic dependencies in the network (higher-order effects). An agent-based model for simulating the evolution of an IO system can be a monetary or physical model and be applicable to one or multiple regions. The model can be used to determine the supply chain impacts of external shocks (such as shortage of raw materials and/or loss of production capacity), and the modeled scenarios can unfold at relatively fine temporal scales (such as days). An agent-based model includes production, consumption, and transportation agents. It is used to simulate economic flows in the regions of the network and describe their dynamic interactions [4], [26], [21]. The model can capture the adaptive responses of various entities to daily external shocks and the realistic behaviors of firms during the year, including inventory management, order and supplier adjustments, and the transportation of goods across regions.

Fig. 1 is a schematic diagram of the agent behaviors in the dynamic agent-based complex network model. The production agents represent the regional production sectors, the consumption agents represent the regional final demand sectors, and the transportation agents reflect the time required for transporting products between regional agents. The model can depict small perturbations day by day. All agents seek to regain their initial productivity as quickly as possible after a perturbation. There are generally more than a thousand agents and more than a million pairs of agents. Executing these equations for a complex network environment is a formidable challenge. Therefore, the matrix forms of the equations were used. The model is detailed in Supplementary Text S1 in Appendix A. The model parameters are listed in Table S1 in Appendix A. The model was coded on the MATLAB R2021a platform.

The proportions of production capacity reductions (θ) in the directly affected sectors and regions were calculated using Eqs. (1), (2), (3). They consisted of two parts: the partially recovered perturbation before the shutdown and the new perturbation per day. The external shocks θ were input into the dynamic agent-based model to simulate the value-added losses caused by enterprise shutdown.

X t r , s = ∑ n x n r , s × 1 − α n , t
X t P r , s = ∑ n x n r , s × 1 − α n , t × List n
θ t r , s = 1 − Δ t τ θ × θ t − 1 r , s + X t P r , s − X t − 1 P r , s X t r , s

where r is the index of the directly affected region; s is the index of the directly affected sectors; n is the index of enterprises in region r and sector s; t is the daily index during the study period; X t r , s is the total production capacity of sector s in region r on day t; x n r , s is the initial production capacity of enterprise n in region r and sector s; α n , t is the proportion of production capacity reduction on day t for enterprise n; X t P r , s is the total production capacity of listed enterprises in sector s in region r on day t; List n indicates whether an enterprise n is listed. A value of 1 indicates inclusion, θ t r , sand 0 denotes exclusion; θ t r , s is the cumulative percentage reduction in production capacity in sector s in region r on day t caused by the shutdown or rectification of listed enterprises; Δt is the length of the time step, 1/365; τ θ is the characteristic time for reconstruction, 1/4.

The parameters θ are input into the dynamic agent-based supply chain model to simulate the daily direct and indirect economic risks associated with the enterprise shutdown and improvement policy. The equations for calculating the direct and indirect economic losses are defined as follows:

DirectLoss t r , s = VA ¯ t r , s × θ t r , s
IndirectLoss t r , s = VA ¯ t r , s × 1 − X a r , s | θ t r 1 , s 1 , r 2 , s 2 , … , r R , s S X t r , s

where DirectLoss t r , s represents the direct economic loss for sector s in region r on day t; IndirectLoss t r , s represents the indirect economic loss for sector s in region r on day t; VA ¯ t r , s denotes the economic value added generated by production agent for sector s in region r on day t under steady-state conditions; assuming there are total R regions and S sectors, X a r , s | θ t r 1 , s 1 , r 2 , s 2 , . . . , r R , s S represents the production output of production agent for sector s in region r on day t under the joint effects of external shocks θ t from r=1,...,R regions and s=1,...,S sectors, which is obtained from the agent-based model.

2.4. Environmental risk assessment

Official data indicate various reasons for regulatory measures resulting in enterprise shutdowns. These reasons typically include inadequacies in environmental protection, unsafe production practices, fire control deficiencies, non-compliance with industrial policies, inefficient resource utilization, and other unspecified reasons. Environmental protection and production safety are pivotal factors in policy design. We quantified these aspects and used the environmental penalty per unit of production capacity ( P n), the proportion of safety violations ( V n), and an integrated risk index ( Risk n) to determine the environmental and safety risks of the enterprises regarding air pollution, water pollution, solid waste pollution, sudden environmental risk, and safety production.

First, P n was calculated. It is the number of environmental penalties imposed on the enterprise for violating environmental standards or regulations divided by the enterprise’s total production value. Therefore, this indicator provides a comprehensive insight into the enterprise’s environmental risk level regarding air pollution, water pollution, solid waste pollution, and sudden environmental risks. These penalties are recorded and disclosed through legally effective administrative penalty decisions. The penalty records were extracted from the environmental and administrative penalty records in the national enterprise credit information publicity database. Table S2 in Appendix A lists examples of penalty records for enterprises fined due to environmental violations.

Second, V n was calculated. It is the number of violations found during inspections divided by the total number of inspections of the enterprise. Regulatory authorities perform routine and unscheduled safety inspections of enterprises to improve safety management and prevent accidents. These inspections cover a wide range of areas, including hazardous waste management, emergency management, the integrity of environmental impact assessments, online monitoring compliance, and other aspects. Thus, this indicator comprehensively describes the environmental risk of enterprises during production. The outcomes of the safety inspections were obtained from the records of the national enterprise credit information publicity database. Examples of these records are listed in Table S3 in Appendix A.

Third, Risk n was calculated by weighting P n and V n for each enterprise, as shown in Eq. (6). The weights were allocated using the mean square errors of the two indicators, reflecting their relative importance in the index. In theory, a higher mean square error of an indicator corresponds to greater variability. This index comprehensively assesses the enterprise’s environmental risks regarding air pollution, water pollution, solid waste, sudden environmental risks, and safety production.

Risk n = P n × σ P n σ P n + σ V n + V n × σ V n σ P n + σ V n   P n , V n , Risk n ∈ [ 0 , 1 ]

where n is the index of an enterprise; P n and V n ranging from 0 to 1; σ P n is the mean square error of P n, and σ V n is the mean square error of V n. Risk n integrated risk index of P n and V n ranging from 0 to 1.

An expectation model was used to forecast the counterfactual risk of enterprises that stopped operating halfway through the study period ( T target), assuming the continuous operation during the T target. The expectations for the two indicators P n and V n were calculated using complete data records from enterprises that remained active during the T target and the preceding time period ( T previous) using the Eq. (7). The bootstrap method was used to perform random sampling with replacement of 50% of the records in the datasets 500 times. The model was applied repeatedly to predict the expected values for each bootstrap datasets, and the 95% confidence intervals were estimated.

E Y | X i = ∑ j E Y j | X i × p Y j | X i i = 0 , 1 ; j = 0 , 1

where T target refers to the study period; T previous refers to the period preceding the T target with an equivalent duration; X 0 represents the indicator equal to 0 during period T previous; X 1 represents the indicator not equal to 0 during T previous; Y 0 represents the indicator equal to 0 during T target; ; Y1 represent represents the indicator not equal to 0 during T target; p Y j | X i is the probability of Y = Y j when X = X i; E Y j | X i is the expected value of Y = Y j when X = X i; E Y 1 | X 0 was calculated by averaging the values of Y 1 in the group; E Y 1 | X 1 was calculated by fitting a linear regression using sample points from the group. The expected value was weighted by the production capacity. Finally, the expected E Y | X i was assigned to enterprises that stopped operating during T target to indicate their environmental risks.

2.5. Scenario analysis

Based on the results in 2.3 Agent-based complex network modeling, 2.4 Environmental risk assessment, we aggregated the daily dynamics of the economic and environmental performances of the enterprises. Then, scenario analyses were performed to evaluate the economic and environmental risks after adjusting the scope of the affected enterprises and the implementation parameters of the shutdown policy. Key factors influencing the policy’s efficiency were identified by comparing the results. This analysis facilitated the identification of critical risk nodes of enterprises, sectors, and regions, offering actionable insights to perform proactive management of economic and environmental risks.

3. Case study

We conducted a case study to illustrate how the user inputs can be incorporated into the universal model framework for practical applications.

3.1. Policy background

On March 21, 2019, a major explosion occurred in a chemical plant in Xiangshui County, Jiangsu Province, China (“3.21” chemical plant explosion), causing at least 78 deaths, 640 injuries, and extensive property loss amounting to around 284.3 million USD [35]. The investigation found the company responsible for the accident because it did not comply with laws and regulations regarding environmental protection and production safety in dealing with chemical waste. One month after the Xiangshui incident, the Jiangsu government issued the Improvement Plan for the Safety and Environmental Protection of Chemical Industries in Jiangsu Province (the Improvement Plan), requiring a significant reduction in chemical product plants in environmentally sensitive areas and populated urban areas within 1 km of the Yangtze River and outside the chemical parks [36]. The Improvement Plan announced reform suggestions for 3943 chemical enterprises in Jiangsu; 1728 enterprises were required to shut down and exit from the market, 552 had to shut down temporarily and perform rectifications within the time limits, 1587 had to rectify within the time limits, and 76 had to relocate or reorganize.

3.2. Risk assessment

This study focused on chemical enterprises coded C251 (refined petroleum products manufacturing industry) and C26 (chemical manufacturing industry), referring to the guidebook of the Industrial Classification for National Economic Activities (GB/T 4754—2017) [37]. Information from 18 916 enterprises was collected from five enterprise databases: ① The national enterprise credit information publicity database provided the industrial, commercial, and historical operating information of the enterprises. ② The China energy statistics database and the Improvement Plan enterprise database provided economic information for the enterprises. ③ The second national pollution source survey database provided the 5-digit industry type of the enterprises. ④ The Improvement Plan enterprise database provided reform suggestions for 3943 chemical enterprises in Jiangsu. Following data combination and data mining outlined in Section 2.2, the enterprises’ operating status and shutdown date were inferred by examining their historical records of industrial and commercial information.

Based on the detailed information on enterprises’ operating conditions, the dynamic agent-based supply chain model was used to evaluate the daily production losses and their propagation caused by the downsizing of production capacity under the Improvement Plan. The latest Chinese city-level multi-regional IO (MRIO) table for 2017 obtained from the Carbon Emission Accounts and Datasets [38] was used to determine the supply chain network in China. The MRIO represents the annually averaged monetary transactions between 42 industry sectors (and final demand) in 313 regions in China. The refined petroleum products manufacturing industry (C251) and chemical manufacturing industry (C26) are subsets of the petroleum processing sector (coded 11 in MRIO) and chemical production sector (coded 12 in MRIO), respectively. The proportion of production capacity decline (α) of the enterprise is listed in Table 2 [34]. Its validation is described in the scenario analyses.

Official data state six reasons for including enterprises in the Improvement Plan: 35.2% of enterprises were included because of inadequate environmental protection, 35.6% for unsafe production, 12.1% for inadequate fire control, 1.5% for not meeting industrial policies, 0.9% for inefficient resource utilization, and 14.7% for other reasons. P n, V n, and Risk n were evaluated for periods 2016-2018 and 2019-2021 to assess the environmental and safety risks of the enterprises. The weight of P n was 0.48, and that of V n was 0.52. The counterfactual risk indicators of the enterprises during 2019-2021 were predicted using a simplified expectation model and all data records from enterprises that remained active from 2016 to 2021.

3.3. Scenarios and simulations

Scenario analyses (Table 3) were performed to validate the model and improve the policy design. The business as usual (BAU) scenario describes the real-world effects of the Improvement Plan. A sensitivity analysis was conducted for scenario S1 to evaluate the impact of the rectification setting. The length of the rectification period was 30, 60, and 90 days, and the production capacity reduction for these periods was 10%, 20%, and 30%, respectively. Since the Improvement Plan should reduce the risks of chemical production, this study explored the economic impact of different policy designs adjusted according to the environmental and safety performances of enterprises. Scenarios S2, S3, and S4 were designed based on the environmental and safety risk indicators to determine the policy effects after the adjustment by easing regulations, tightening regulations, and recompiling the Improvement Plan enterprise list. The timing of the shutdown and rectification of the enterprises has a substantial effect on economic losses resulting from the policy. The basic assumption is that if different enterprises centralize shut down in a short time, this will cause higher impacts on the supply chain network. Scenarios S5 and S6 were designed to examine the impact of dispersing shutdown timing and centralizing shutdown timing respectively.

4. Results and discussion

4.1. Enterprise shutdown

Fig. 2(a) illustrates the types and operating conditions of the 18 916 enterprises at the end of 2021. The analysis excluded 26.8% of enterprises that shut down before 2019. Of the 3434 enterprises listed in the Improvement Plan, 99.4% are ordinary companies producing chemical products. Only 0.6% are small-scale individually-owned businesses. At the end of 2021, 52.2% of the enterprises on the Improvement Plan list had dissolved, with 25.2% directly withdrawing from the market and 56.4% choosing to transform to other industries, such as wholesale, retailing, and scientific research services.

Fig. 2(b) shows the daily changes in the number of two types of chemical enterprises based on the firm-level data. The triangles are the official statistics of the number of ordinary chemical companies from the China Basic Statistical Units Yearbook. The declining line matches the official statistical data. The number of ordinary companies decreased significantly from 12 739 at the beginning of 2019 to 8334 at the end of 2021. In contrast, the number of individually-owned businesses, which only accounted for a small proportion of the market (approximately 8%), increased during this period.

Fig. 3(a) depicts a map of the locations of chemical enterprises operating in Jiangsu at the end of 2021. Although the southern part of Jiangsu is more economically advanced and urbanized than the northern part, more chemical enterprises are located in the south. The reason is that water resources are critical for chemical production, and the southern part of Jiangsu has more abundant water resources [39]. For example, the Yangtze River, which is China’s most abundant water resource, flows through the southern part of Jiangsu. Thus, the Yangtze River provides convenient transportation for raw materials and finished chemical production products throughout China.

Fig. 3(b) provides detailed information on the distribution of chemical enterprises in 13 cities in Jiangsu. The grey triangles and the blue dots illustrate the number of chemical enterprises in 2019 and 2021, respectively. The colors of the line segments denote the percentage decrease in the number of enterprises from 2019 to 2021. The results show more chemical enterprises in the southern cities than in the northern ones. After the promotion of the Improvement Plan and the Yangtze River Basin Protection Action at the beginning of 2021, the number of chemical enterprises in densely populated areas and 1 km from the Yangtze River decreased substantially. As a result, the number of chemical enterprises in all cities in Jiangsu decreased by more than 25% from 2019 to the end of 2021. Wuxi experienced the most significant decrease (43.6%). Additionally, the number of enterprises located 1 km from the Yangtze River decreased by more than 35%. The remaining enterprises were closing down or converting to nonproduction enterprises. Moreover, 57.4% of chemical enterprises were located in urban areas, accounting for only 22.4% of Jiangsu. The number of chemical enterprises in urban areas has decreased by more than 30% in recent years.

The regulation of chemical enterprises in Jiangsu Province is critical due to its significant role in China’s chemical industry and the potential risks associated with this industry [40]. Chemical plants have hundreds or thousands of pieces of hazardous production equipment [41], [42]. An accident in a chemical plant may trigger a chain of accidents, resulting in safety, environmental, and economic problems [39]. Jiangsu is a highly developed province and has a significant chemical production capacity. Many chemical enterprises are located along the Yangtze River in Jiangsu. The province has the highest frequency of chemical accidents. In accordance with China’s environmental risk management regulations and the Yangtze River Protection Law, regulating the production of chemical enterprises in Jiangsu Province is an important task. Thus, the results of this study have practical significance and provide basic data and specific suggestions for the governance of Jiangsu chemical enterprises.

4.2. Economic losses

The detailed information on enterprises’ operating conditions obtained from multi-source data mining was used to calculate the daily production capacity losses and their propagation through the supply chain network using the dynamic agent-based supply chain model. These losses were the direct result of reducing the daily production capacity of the chemical industry in Jiangsu (first-order effect) or the indirect result due to economic dependencies in the network (higher-order effect). The model accounts for both effects, allowing for a comprehensive assessment of the economic impacts of the environmental management policy.

Fig. 4(a) shows the simulated daily dynamics of value added in Jiangsu. The red line provides a good fit in the first two years of the simulation but underestimates the value added in the third year. The relative differences between the simulated and actual value-added are 0.19%, 0.10%, and 2.58% for the three years. The relatively high discrepancy in the third year may be attributed to external shocks affecting the supply chain or alterations in the direct consumption coefficient matrix. The validation of the simulation results was conducted for the whole country of China and the 13 cities in Jiangsu, as shown in Fig. S2 in Appendix A. The results consistently demonstrate a good fit between the daily simulations and the statistical values collected from the yearbook, with a relative difference ranging from 0.17% to 2.30% for the whole country of China and from 0.1% to 10.4% for the respective cities.

Fig. 4(b) shows the simulated daily value-added losses due to the Improvement Plan. An upward trend is observed due to the amplification of economic risks caused by daily shocks resulting from the implementation of firm shutdowns and rectifications. These shocks can lead to indirect losses propagating throughout the supply chain, potentially causing a cumulative increase in total economic losses over time. The total production losses from 2019 to 2021 were estimated at 25.8 billion USD, equivalent to 0.07% of the national gross domestic product (GDP). The decomposition of these losses shows that the contribution of higher-order losses to the total loss was significant, ranging from 40.0% to 90.8%. Thus, a small external shock can lead to a sharp rise in higher-order losses.

Interannual simulations were conducted using the model by adjusting the time step to one year. The results show that the total direct economic losses were 1.1 times the losses in the daily model, whereas the indirect economic losses were 3.1 times the losses in the daily model. This result indicates that the relative difference between the simulated and actual value-added was 90%. These findings underscore the significant disparities between the two models in capturing the dynamics of economic losses. This study demonstrates the unique contribution of the daily model in capturing short-term fluctuations, considering the potential influence of adaptive behavior, and providing a more realistic representation of the economic impact of external shocks. The results underscore the importance of high-resolution, daily modeling for understanding economic losses in a dynamic and adaptable economic system.

Fig. 4(b) shows two sharp peaks of higher-order losses at the end of 2019 and 2020. They were caused by the shutdown or rectification of large-scale enterprises in Xuzhou, Wuxi, Zhenjiang, and Yangzhou (Fig. S3 in Appendix A). The two peaks occurred at the end of the year. The likely reason is that environmental plans are usually implemented annually. Thus, many enterprises tend to delay their closure until the end of the year, leading to a significant surge in higher-order losses before the Chinese Spring Festival. It is suggested that regulatory measures for chemical enterprises be implemented at different times to mitigate sudden disruptions in the supply chain. Specifically, it is proposed to shift the policy deadline from the end to the middle of the year to stagger the effects from enterprises that withdraw from the market. The real-time simulations provide valuable insights to obtain early warning of supply chain disruptions and inform policy design for effective response planning.

Fig. 5(a) shows a map of the city-level total value-added losses as a percentage of the regional GDP. Jiangsu, where the Improvement Plan was implemented, experienced the most substantial economic losses. In addition, due to the dependencies in the supply chain network, the impact of the Improvement Plan in Jiangsu spread across China. Northwest China suffered the highest percentage of losses after Jiangsu, whereas the coastal areas of eastern China had relatively lower losses. These findings reveal the different internal stabilities of the supply chain networks.

Fig. 5(b) plots the top 20 cities with the highest total value-added losses. The 13 cities in Jiangsu ranked in the top 13. The first-order losses directly caused by the Improvement Plan accounted for 10.2% to 47.4% of all losses. Xuzhou had the highest higher-order losses, indicating a high dependence of the local industry on chemical products. The total losses in Jiangsu accounted for 81.5% of the national losses and 0.5% of the GDP of Jiangsu.

Fig. 5(c) plots the top 20 sectors with the highest total value-added losses. The chemical production sector (coded 12) incurred the most significant losses of 6.2 billion USD, accounting for 24.2% of the total losses across the country. The petroleum processing sector (coded 11) suffered losses of 0.7 billion USD, accounting for 2.7% of the total losses across the country. The top four sectors with the largest higher-order losses included chemical production, wholesale and retailing, transport and storage, and agriculture. These results indicate that the initial losses caused by the decline in the production capacity of the chemical industry propagated throughout the supply chain in various stages of production, sales, and use.

4.3. Environmental and safety performance

The official data suggest six reasons for including enterprises in the Improvement Plan. Inadequate environmental protection and unsafe production are the dominant factors accounting for 35.2% and 35.6% of all cases, respectively. We analyzed these factors using indicators, including P n and the V n, to compare the performance of ordinary companies. Table 4 lists the environmental, safety, and economic performance of enterprises in different subgroups during 2016-2018. The results show that the P n and V n of enterprises that had shut down were 65.52 and 1.51 times that of enterprises still operating, respectively. Similar results were observed when comparing enterprises listed and not listed on the Improvement Plan list. The P n and V n of listed enterprises were 16.53 and 1.53 times that of not listed enterprises. Besides, the differences in enterprises were also reflected in the illegal discharge of air, water, and solid waste, violation of regulations of environmental impact assessment and safe production, and inadequate information disclosure.

The decisions on reforming enterprises are based on their environmental safety performance and production capacity. The enterprises with the highest P n and V n in each subindustry group were usually required to shut down, followed by those required to temporarily shut down and rectify (Table 4 and Table S2 in Appendix A). Some enterprises with high environmental penalties were not required to shut down because of their important role in the economic market. Enterprises with high economic output and low P n were more likely to be required to rectify, whereas those with low economic output and high P n were more likely to be required to shut down. The number of accidents was lower in large-scale enterprises than in small- and medium-sized enterprises, consistent with previous study [40]. The likely reason is that large enterprises have a more comprehensive safety management system and more abundant resources.

A simplified expectation model was used to predict the counterfactual risk of enterprises that had shut down during 2019-2021. Based on data records from enterprises that remained active during 2016-2018 (T1) and 2019-2021 (T2), results show that P n of enterprises that were previously fined during T1 increased by 8.97 times during T2 compared to those that were not previously fined. Besides, V n of enterprises that were previously fined during T1 increased by 4.74 times during T2 compared to those that were not previously fined. These findings suggest the reliability of using enterprises’ past environmental behaviors to predict their future risks.

4.4. Multi-objective policy design

The results of the scenario analyses are shown in Fig. 6. The red curve in Fig. 6(a) represents the daily changes in the total economic losses as a percentage of the GDP in the BAU scenario, and the gray area represents the results of the sensitivity analyses (S1) of the rectification duration and proportion of α. The results show that the economic losses in the sensitivity analysis ranged from 23.8 billion to 31.8 billion USD, indicating that the influences of these parameters on economic losses were relatively small, ranging from −7.8% to 23.3%. The economic losses were more sensitive to α, which contributed more than 90% to the changes. The reason may be that the loss occurs primarily in the early stage of the reduction in the enterprise's production capacity. Subsequently, the entities in the economic system adapt to external shocks through adaptive behaviors, such as replenishing inventory, adjusting the order share of upstream suppliers, using spare production capacity, adjusting production technology, and others.

The results of the easing regulation scenario (S2) and the strengthening regulation scenario (S3) in Figs. 6(b)-(c) show that adjusting the implementation strength considering environmental and safety performance significantly affects economic losses. The loss curves of S2 and the BAU scenario were similar, while the loss amount showed significant differences. The first-order losses of the S2 and S3 scenarios were 84.3% and 123.3% of the BAU scenario, respectively, indicating that strengthening regulations caused higher direct losses. The higher-order losses of S2 and S3 scenarios were 77.9% and 328.0% of the BAU, respectively, indicating that external shocks were amplified in the economic system due to strengthening governance and control (Fig. 6(g)). This finding suggests that policymakers should adjust the strength of regulations to avoid abrupt changes to the supply chain network.

The risk reductions are shown in Figs. 6(d)-(f). The production capacity-weighted indicators, including Pn, the Vn, and the Riskn, declined over time. The three indicators in the BAU scenario declined 55.56% (ranging from 48.39% to 61.57%), 39.18% (32.80% to 44.85%), and 51.70% (ranging from 45.63% to 56.93%), respectively. The risk reduction between scenario S2 (easing the regulation) and BAU was 49.94%, 37.24%, and 46.91%, equivalent to 89.9%, 95.0%, and 90.7% of the BAU scenario. The risk reduction between scenario S3 (strengthening the regulation) and BAU was 66.85%, 59.80%, and 65.17%, equivalent to 120.3%, 152.6%, and 126.0% of the BAU scenario.

These proportions were consistent with the changes in first-order economic losses (Fig. 6(g)). In addition, scenario S4 showed that recompiling the Improvement Plan enterprise list while considering environmental, safety, and economic objectives led to a first-order economic loss of 68.9% of the BAU scenario, and the higher-order economic loss was 77.9% of the BAU (Fig. 6(g)). The reduction of the three risk indicators between scenarios S4 and BAU was 80.84% (ranging from 78.86% to 82.28%), 49.38% (ranging from 45.98% to 52.41%), and 73.44% (ranging from 71.33% to 75.09%), equivalent to 146% (ranging from 142% to 148%), 126.0% (ranging from 117% to 134%), and 142.0% (ranging from 138% to 145%) of the BAU scenario, indicating high risk reduction with low economic losses (Figs. 6(d)-(f)). Therefore, the multi-objective policy design is feasible to improve environmental, safety, and economic outcomes substantially.

The results of the two scenarios (S5 and S6) for adjusting the implementation timing indicated that the timing of shutdown and rectification were important factors affecting total economic losses. The first-order losses in scenarios S5 and S6 were the same as in the BAU scenario, but the higher-order losses deviated significantly from the BAU scenario. When the implementation timing was evenly dispersed throughout the study period (S5), the higher-order loss was 81.2% of the BAU scenario (Fig. 6(g)). In contrast, when the implementation timing was centralized (S6) in a shorter period, such as 90 days in scenario S6.1, 180 days in scenario S6.2, and 365 days in scenario S6.3, the higher-order losses were 18.6, 15.6, and 6.6 times that of the BAU scenario, respectively (Fig. 6(g)). When policymakers prepare to shut down or rectify enterprises, it is suggested that these actions be implemented in batches to avoid massive economic losses caused by abrupt changes to the supply chain network in a short period, such as at the end of the year. This finding highlights the importance of carefully planning the timing of policy implementation to achieve the desired outcomes while minimizing economic losses.

Many studies have used carbon emissions as an indicator of the environmental performance of supply chains [43], [44], [45], [46], [47]. In contrast, this study incorporated environmental risk assessments. This novel perspective enabled a more in-depth analysis of the multifaceted dynamics of environmental management in supply chains. Considering both carbon emissions and environmental risk indicators, such as Pn and the Vn, broadens the scope of environmental evaluation.

Effective environmental governance requires the collaboration of multiple departments with diverse objectives. For example, environmental management aims to improve environmental quality, economic management pursues stabilization of the economy, and emergency management focuses on reducing accidents. It is noteworthy that different departmental objectives may conflict, presenting a significant challenge to effective governance. In this context, the proposed model considers the requirements of multiple departments and optimizes multiple objectives encompassing environmental improvement, safe production, and economic viability. This innovative approach is instrumental in achieving effective and efficient environmental governance by balancing the contrasting objectives.

The results of the scenario analyses demonstrate that the proportion of production capacity decline and the timing of policy implementation are critical to reducing economic losses. They reveal a rising trend in higher-order losses as environmental governance and control measures become more stringent, indicating the amplification of external shocks in the supply chain network. Furthermore, the cost amplification effect may be more pronounced at the firm level [4], [19]. Thus, policymakers should focus on balancing the implementation strength and timing of policies to avoid large-scale market impacts.

4.5. Implications

This study coupled a dynamic agent-based supply chain model with high spatial-temporal resolution firm-level data from 18 916 enterprises to investigate the economic and environmental consequences of regulating chemical enterprises for environmental protection in Jiangsu, China. This integration highlights the development and amplification of economic and environmental risks due to daily small shocks resulting from firm shutdowns and rectifications. The results emphasize the need for refined environmental governance strategies that consider the complex interdependencies between economic and environmental systems and suggest that these strategies can lead to sustainable economic growth while mitigating environmental risks.

This study established a universal dynamic agent-based supply chain network model to capture the adaptive responses of various entities to daily external shocks. More than 13 000 agents and theoretically over 170 000 000 pairs of agents were analyzed. The matrix forms of equations were employed to reduce computational complexity. Traditionally, the supply-constrained IO model has been used to model supply-side shocks in economic systems [48], [49]. Unlike the agent-based approach, it cannot capture the realistic behaviors of firms within a year, including inventory management, order and supplier adjustments, and the transportation of goods across regions.

Our analysis showed that 82.2% of the losses stemmed from the economic dependencies in the network, reflecting the loss propagation across the supply chain at various stages of production, sales, and use. Counterfactual analyses indicated that implementing a staggered shutdown of enterprises prevented 18.8% of supply-chain losses, highlighting the importance of implementing policies in stages rather at one time to prevent abrupt supply-chain losses. Besides, the multi-objective policy design considering environmental, safety, and economic impacts could reduce economic losses and environmental risks. The analytical approach and real-time simulation provide an effective and efficient early warning measure to formulate environmentally and economically friendly policies toward a sustainable future.

Our universal model enables “virtual experiments”, in which a large number of agents interact in a complex network, allowing policymakers to observe the evolution of the environmental-economic system under various scenarios. This unique capability indicates the model’s applicability to assessing economic and environmental consequences related to supply chain disruptions, as well as environmental risks, natural disasters, and policy scenarios. For example, evaluation models have been used to analyze the effects of California wildfires [50], heat stress [27], the Japan earthquake [4], river floods [28], and corona virus disease 2019 (COVID-2019) [32]. These applications demonstrate the versatility of these modeling approaches in addressing real-world challenges. It is imperative to conduct additional case studies of various scenarios and real-world situations to validate the model’s applicability, efficiency, robustness, and capacity to assist in effective decision-making in various domains.

4.6. Limitations

This study has some limitations that should be addressed. First, the data contained some missing values. Second, the timing of rectification was based on certain assumptions. Third, it was assumed that the shutdown of an enterprise would occur immediately rather than a gradual withdrawal from the market. Fourth, the supply-chain network was constructed based on MRIO data instead of plant-level supply chain data, potentially overlooking detailed supply-chain dynamics due to a more integrated but generalized representation. The 18 916 enterprises were integrated into 42 sectors and 313 regions. This integration was imperative because the MRIO table is currently the best dataset characterizing the connections in the supply chain. Moving forward, we aim to investigate more detailed supply chain network databases that provide more insight into the connections between enterprises.

We addressed these limitations as follows. First, data cleaning and processing were conducted to address missing and incorrect data. For example, the missing economic information of some enterprises was filled in using the inferred linear relationship between the economic production capacity and the paid-in capital (or registered capital when paid-in capital is missing). Additionally, when data records were incomplete due to enterprise shutdowns during 2019-2021, a simplified expectation model was used to predict the counterfactual risk indicators. Second, sensitivity analysis was performed, indicating that the assumptions regarding the rectification timing and length did not alter the main findings. Third, this study used optimal data to accurately depict the shutdown conditions for a large number of enterprises. Additionally, the model considered the gradual mitigations of external shocks.

5. Conclusions

This study analyzed the complex relationship between environmental policies and economic repercussions, emphasizing the significance of considering tradeoffs between economic losses and environmental risks. A universal dynamic agent-based supply chain model was proposed to simulate the daily evolution of environmental risks and economic states under daily shocks due to firm shutdowns and rectifications. The model is unique due to its matrix form to deal with the large computational complexity and its ability to capture the adaptive responses of entities facing daily external shocks. Unlike traditional supply-constrained IO models, the proposed model can simulate the realistic behaviors of firms during a year, incorporating inventory management, order adjustments, supplier interactions, and cross-regional goods transportation.

The findings underscore the immense economic losses incurred as a result of regional policies, revealing an estimated production loss of 25.8 billion USD from 2019 to 2021, equivalent to 0.07% of the national GDP. It is noteworthy that more than 80% of this loss can be attributed to the cascading effect in the supply chain. The model exhibited high accuracy; the relative differences between the simulated and actual value-added were 0.19%, 0.10%, and 2.58% for the three years, respectively. The effectiveness of the multi-objective policy design is another pivotal discovery, suggesting a reduction in economic losses of approximately 29% and a decrease in environmental risks of about 40%. This finding underscores the pivotal role of synergistic governance in achieving economic stability and environmental sustainability.

Moreover, the model has high generalization ability and versatility, making it applicable to various scenarios. Policymakers can leverage their capabilities to conduct “virtual experiments” to observe the evolution of the environmental-economic system under various conditions. Further case studies should be conducted to validate the model’s applicability in diverse real-world scenarios.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (52200228 and 72022004), the China Postdoctoral Science Foundation (2022M721817), and the National Key Scientific Research Project (2021YFC3200200).

Compliance with ethics guidelines

Qi Zhou, Shen Qu, Miaomiao Liu, Jianxun Yang, Jia Zhou, Yunlei She, Zhouyi Liu, and Jun Bi declare that they have no conflict of interest or financial conflicts to disclose.

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.eng.2024.02.006.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]

D. Sofia, F. Gioiella, N. Lotrecchiano, A. Giuliano. Cost-benefit analysis to support decarbonization scenario for 2030: a case study in Italy. Energ Policy, 137 (2020), p. 111137
[2]

W. Liu, J. Zhang, T. Yang. Will China’s household coal replacement policies pay off: a cost-benefit analysis from an environmental and health perspective. J Clean Prod, 357 (2022), p. 131904
[3]

J.N. Barrot, J. Sauvagnat. Input specificity and the propagation of idiosyncratic shocks in production networks. Q J Econ, 131 (3) (2016), pp. 1543-1592
[4]

H. Inoue, Y. Todo. Firm-level propagation of shocks through supply-chain networks. Nat Sustain, 2 (9) (2019), pp. 841-847
[5]

C. Shughrue, B.T. Werner, K.C. Seto. Global spread of local cyclone damages through urban trade networks. Nat Sustain, 3 (8) (2020), pp. 606-613
[6]

D. Acemoglu, V.M. Carvalho, A. Ozdaglar, A. Tahbaz-Salehi. The network origins of aggregate fluctuations. Econometrica, 80 (5) (2012), pp. 1977-2016
[7]

T. Wiedmann, M. Lenzen. Environmental and social footprints of international trade. Nat Geosci, 11 (5) (2018), pp. 314-321
[8]

Y. Xia, Y. Li, D. Guan, D.M. Tinoco, J. Xia, Z. Yan, et al. Assessment of the economic impacts of heat waves: a case study of Nanjing. China. J Clean Prod, 171 (2018), pp. 811-819
[9]

M. Baghersad, C.W. Zobel. Economic impact of production bottlenecks caused by disasters impacting interdependent industry sectors. Int J Prod Econ, 168 (2015), pp. 71-80
[10]

L. Tan, X. Wu, Z. Xu, L. Li. Comprehensive economic loss assessment of disaster based on CGE model and IO model—a case study on Beijing “7.21 Rainstorm”. Int J Disast Risk Re, 39 (2019), p. 101246
[11]

W. Xie, A. Rose, S. Li, J. He, N. Li, T. Ali. Dynamic economic resilience and economic recovery from disasters: a quantitative assessment. Risk Anal, 38 (6) (2018), pp. 1306-1318
[12]

S. Hallegatte. Modeling the role of inventories and heterogeneity in the assessment of the economic costs of natural disasters. Risk Anal, 34 (1) (2014), pp. 152-167
[13]

J. Kim, J. Kim, Y. Wang. Uncertainty risks and strategic reaction of restaurant firms amid COVID-19: evidence from China. Int J Hospit Manag, 92 (2021), p. 102752
[14]

R. Eyre, F. De Luca, F. Simini. Social media usage reveals recovery of small businesses after natural hazard events. Nat Commun, 11 (1) (2020), p. 1629
[15]

M.A. Cole, R.J.R. Elliott, T. Okubo, E. Strobl. Pre-disaster planning and post-disaster aid: examining the impact of the great east Japan earthquake. Int J Disast Risk Re, 21 (2017), pp. 291-302
[16]

J. Lee. Business recovery from Hurricane Harvey. Int J Disast Risk Re, 34 (2019), pp. 305-315
[17]

Y. Isoda, S. Masuda, S.I. Nishiyama. Effects of post-disaster aid measures to firms: evidence from Tohoku University earthquake recovery firm survey 2012-2015. J Disaster Res, 14 (8) (2019), pp. 1030-1046
[18]

N. Altay, A. Ramirez. Impact of disasters on firms in different sectors: implications for supply chains. J Supply Chain Manag, 46 (4) (2010), pp. 59-80
[19]

F. Henriet, S. Hallegatte, L. Tabourier. Firm-network characteristics and economic robustness to natural disasters. J Econ Dyn Control, 36 (1) (2012), pp. 150-167
[20]

C. Colon, S. Hallegatte, J. Rozenberg. Criticality analysis of a country’s transport network via an agent-based supply chain model. Nat Sustain, 4 (3) (2021), pp. 209-215
[21]

S. Hallegatte. An adaptive regional input-output model and its application to the assessment of the economic cost of Katrina. Risk Anal, 28 (3) (2008), pp. 779-799
[22]

N. Ranger, S. Hallegatte, S. Bhattacharya, M. Bachu, S. Priya, K. Dhore, et al. An assessment of the potential impact of climate change on flood risk in Mumbai. Clim Change, 104 (1) (2011), pp. 139-167
[23]

S. Hallegatte, N. Ranger, O. Mestre, P. Dumas, J. Corfee-Morlot, C. Herweijer, et al. Assessing climate change impacts, sea level rise and storm surge risk in port cities: a case study on Copenhagen. Clim Change, 104 (1) (2011), pp. 113-137
[24]

R. Bierkandt, L. Wenz, S.N. Willner, A. Levermann. Acclimate—a model for economic damage propagation. Part 1: basic formulation of damage transfer within a global supply network and damage conserving dynamics. Environ Syst Decis, 34 (4) (2014), pp. 507-524
[25]

L. Wenz, S.N. Willner, R. Bierkandt, A. Levermann. Acclimate—a model for economic damage propagation. Part II: a dynamic formulation of the backward effects of disaster-induced production failures in the global supply network. Environ Syst Decis, 34 (4) (2014), pp. 525-539
[26]

C. Otto, S.N. Willner, L. Wenz, K. Frieler, A. Levermann. Modeling loss-propagation in the global supply network: the dynamic agent-based model acclimate. J Econ Dyn Control, 83 (2017), pp. 232-269
[27]

L. Wenz, A. Levermann. Enhanced economic connectivity to foster heat stress-related losses. Sci Adv, 2 (6) (2016), p. e1501026
[28]

S.N. Willner, C. Otto, A. Levermann. Global economic response to river floods. Nat Clim Chang, 8 (7) (2018), pp. 594-598
[29]

H. Inoue, Y. Todo. Propagation of negative shocks across nation-wide firm networks. PLoS One, 14 (3) (2019), p. e0213648
[30]

H. Krichene, H. Inoue, T. Isogai, A. Chakraborty. A model of the indirect losses from negative shocks in production and finance. PLoS One, 15 (9) (2020), p. e0239293
[31]

H. Inoue, Y. Todo.The propagation of economic impacts through supply chains: the case of a mega-city lockdown to prevent the spread of COVID-19. PLoS One, 15 (9) (2020), p. e0239251
[32]

D. Guan, D. Wang, S. Hallegatte, S.J. Davis, J. Huo, S. Li, et al. Global supply-chain effects of COVID-19 control measures. Nat Hum Behav, 4 (6) (2020), pp. 577-587
[33]

Baidu. Baidu Map open platform [Internet]. Beijing: Baidu Maps; 2022 Mar 1 [cited 2023 Aug 22]. Available from: https://lbsyun.baidu.com/index.php?title=webapi/guide/webservice-geocoding.
[34]

China’s Ministry of Ecology and Environment. Measures for restricting production and stopping production [Internet]. Beijing: Ministry of Ecology and Environment; 2014 Dec 19 [cited 2023 Aug 22]. Available from: https://www.mee.gov.cn/gkml/hbb/bl/201412/t20141224_293385.htm.
[35]

The State Council of the People’s Republic of China. Cause of deadly chemical plant blast in east China revealed. Beijing: The State Council of the People’s Republic of China; 2019 Nov 15 [cited 2023 Aug 22]. Available from: http://english.www.gov.cn/news/topnews/201911/15/content_WS5dce9c8bc6d0bcf8c4c1729f.html.
[36]

The Jiangsu Provincial People’s Government. Improvement plan for the safety and environmental protection of chemical industries in Jiangsu Province. Jiangsu: The Jiangsu Provincial People’s Government; 2019 May 15 [cited 2023 Aug 22]. Available from: http://safety.nanjing.gov.cn/zwfw/shfw/whpgl/201908/t20190823_1634045.html.
[37]

The State Bureau of Quality and Technical Supervision. GB/T 4754-2017: Industrial classification for national economic activities. Chinese standard. Beijing: The State Bureau of Quality and Technical Supervision; 2017. Chinese.
[38]

H. Zheng, J. Tobben, E. Dietzenbacher, D. Moran, J. Meng, D. Wang, et al. Entropy-based Chinese city-level MRIO table framework. Econ Syst Res, 34 (4) (2022), pp. 519-544
[39]

F. Wang, J. Wang, J. Ren, Z. Li, X. Nie, R.R. Tan, et al. Continuous improvement strategies for environmental risk mitigation in chemical plants. Resour Conserv Recycling, 160 (2020), p. 104885
[40]

C. Chen, G. Reniers. Chemical industry in China: The current status, safety problems, and pathways for future sustainable development. Safety Sci, 128 (2020), p. 104741
[41]

T. Zeng, G. Chen, Y. Yang, P. Chen, G. Reniers. Developing an advanced dynamic risk analysis method for fire-related domino effects. Process Saf Environ, 134 (2020), pp. 149-160
[42]

C. Chen, G. Reniers, N. Khakzad. Cost-benefit management of intentional domino effects in chemical industrial areas. Process Saf Environ, 134 (2020), pp. 392-405
[43]

J. Jemai, B.D. Chung, B. Sarkar. Environmental effect for a complex green supply-chain management to control waste: a sustainable approach. J Clean Prod, 277 (2020), p. 122919
[44]

M. Ullah, I. Asghar, M. Zahid, M. Omair, A. AlArjani, B. Sarkar. Ramification of remanufacturing in a sustainable three-echelon closed-loop supply chain management for returnable products. J Clean Prod, 290 (2021), p. 125609
[45]

B. Sarkar, M. Sarkar, B. Ganguly, L.E. Cárdenas-Barrón. Combined effects of carbon emission and production quality improvement for fixed lifetime products in a sustainable supply chain management. Int J Prod Econ, 231 (2021), p. 107867
[46]

A. Sepehri, U. Mishra, B. Sarkar. A sustainable production-inventory model with imperfect quality under preservation technology and quality improvement investment. J Clean Prod, 310 (2021), p. 127332
[47]

B. Sarkar, M. Tayyab, N. Kim, M.S. Habib. Optimal production delivery policies for supplier and manufacturer in a constrained closed-loop supply chain for returnable transport packaging through metaheuristic approach. Comput Ind Eng, 135 (2019), pp. 987-1003
[48]

E. Dietzenbacher. In vindication of the Ghosh model: a reinterpretation as a price model. J Regional Sci, 37 (4) (1997), pp. 629-651
[49]

S. Qu, S. Liang, M. Konar, Z. Zhu, A.S.F. Chiu, X. Jia, et al. Virtual water scarcity risk to the global trade system. Environ Sci Technol, 52 (2) (2018), pp. 673-683
[50]

D. Wang, D. Guan, S. Zhu, M.M. Kinnon, G. Geng, Q. Zhang, et al. Economic footprint of California wildfires in 2018. Nat Sustain, 4 (3) (2021), pp. 252-260

Rights & permissions

THE AUTHOR

PDF (2357KB)

Supplementary files

2024-42-11-S295

11557

Accesses

0

Citation

Detail
Sections
Recommended

/

〈 〉