Spatiotemporal differentiation and dynamics simulation of China’s industrial soot and dust emissions

Abstract

To address the challenge of balancing industrial development and environmental protection in China, we explore the spatiotemporal evolution characteristics, driving mechanisms, and emission reduction paths of industrial soot and dust emissions using an interdisciplinary methodological framework. Results show that national ISDE follows a trajectory of stable fluctuation–linear increase–fluctuating decrease–fluctuating increase–linear decrease, with the spatial pattern shifting from “high in the east, low in the west” to “national convergence + local targeted governance”. The ISDE centroid moves along a path of southward –northward –northwestward. Guangdong, Jiangsu, and Zhejiang are core nodes in the inter-provincial ISDE correlation network, with strong linkages between eastern coastal provinces, economic hubs, and adjacent regions. The permanent population, secondary industry employment, crude oil consumption, and coal consumption are dominant drivers for most provinces, with heterogeneous factors across regions. Multi-scenario simulation for Shanxi indicates that the optimal path (Path 7) is high-speed GDP and population growth and controlling the low-speed development of industrial structure, industrial production demand, and energy consumption intensity.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-42409-y.

Introduction

As a key driver of national economic development, industrial activities are inherently accompanied by industrial soot and dust emissions (ISDE). These emissions exacerbate air pollution, pose risks to residents’ health^1–3, exert potential adverse impacts on the global ecological environment⁴, and may also lead to security conflicts⁵, and have become a critical bottleneck restricting the improvement of ecological environment quality and the protection of public health⁶. As the world’s largest developing country, China’s ISDE is deeply intertwined with its industrialization and urbanization process^7,8, meanwhile, the heterogeneity in energy consumption structure, economic development level, and governance technology capabilities among regions has led to complex and dynamic spatiotemporal distribution characteristics of ISDE⁹. Against this backdrop, accurately analyzing the spatiotemporal evolution of ISDE, clarifying the interaction mechanisms of multiple factors, and developing scientifically feasible emission reduction pathways are not only critical to China’s achievement of its “dual carbon” goals, but also core to advance responsible consumption and production (UN Sustainable Development Goal 12, SDG 12) and safeguarding human well-being.

Current research on industrial pollution primarily centers on investigating the distribution patterns and evolutionary trends of pollutants¹⁰, and dissecting interaction mechanisms¹¹. Methodologically, mainstream approaches include the Environmental Kuznets curve¹², the Logarithmic Mean Divisia Index¹³, and others¹⁴. Notably, the spatiotemporal distribution of ISDE features regionality and complexity, while the interactions and response dynamics among its influencing factors further exacerbate this complexity¹⁵. While these existing studies have greatly promoted our understanding of industrial pollution, there remain areas where research can be further deepened to address unresolved questions. For instance, in the analysis of influencing factors, much of the current work focuses on examining linear correlations, leaving the nonlinear interaction mechanisms across multi-dimensional systems underexplored. The majority of studies adopt a static analytical perspective¹⁶, which limits the effective characterization of dynamic spillover effects that may arise from regional interactions or long-term system evolution. Although a limited number of studies have attempted to apply system dynamics models to simulate the interactive relationships between pollution and the socioeconomic system^17,18, such efforts are still insufficient to fully address the aforementioned research gaps.

To bridge these gaps, we integrate complex network analysis (CNA), grey relational analysis (GRA) and system dynamics (SD) simulation into a single “Temporal Evolution - Influencing Mechanism - Dynamics Simulation” pipeline. Specifically, we reveal the dynamic characteristics of pollution spillover effects, identify dominant influencing factors and design 9 future development paths to simulate ISDE future scenarios. This integrated chain has three aspects of value. First, the spatiotemporal module based on CNA exploits 30-year provincial panel data to reveal the evolutionary law of ISDE during China’s industrial transition. It also offers an empirical benchmark for evaluating the effectiveness of long-term governance policies. Second, the mechanism module enhanced by GRA quantifies non-linear interactions and region-to-region spillovers. It compensates for the linear-bias weakness of conventional econometric models. Third, the scenario module driven by SD delivers nine differentiated emission-reduction pathways. It provides a quantitative basis for region-specific policy packages that can accelerate the sustainable transformation of the industrial sector.

Data and method

Research area

The spatial research area of this study includes 31 provinces. Due to data availability limitations, Hong Kong, Macao, and Taiwan regions are not included. The temporal scope is from 1992 to 2030, where the period 1992–2020 involves the simulation of historical data, and 2020–2030 focuses on predictive simulation, based on the previous simulation results and combined with government policies, different future development scenarios are constructed to predict the emission trend.

Data sources

ISDE data are derived from the National Bureau of Statistics of China (https://data.stats.gov.cn). A small number of missing data in the statistics are supplemented by linear interpolation using the least squares method. Data pertaining to the influencing factors of ISDE are primarily collected from multiple authoritative statistical sources, including the China Statistical Yearbook, provincial statistical yearbooks, development statistical bulletins, and key data bulletins.

Research framework

This study develops an integrated “what-why-how” research framework to analyze China’s ISDE, bridging characteristics analysis with actionable policy implications (Fig. 1). Specifically, based on the panel data spanning 1992–2021, spatial centroid transfer, standard deviational ellipse, and complex network analysis are employed to systematically characterize the spatiotemporal heterogeneity of ISDE from three dimensions: temporal evolution stages, spatial distribution patterns, and inter-provincial correlation networks. Furthermore, 17 key indicators are selected from five subsystems (population, economy, energy, industry, and technological policy), and GRA is adopted to quantify the nonlinear driving mechanisms, thereby clarifying the common laws and regional disparities of dominant influencing factors across provinces. Finally, a systematically validated SD model is constructed, with three development scenarios (low-speed, baseline, and high-speed) and nine combined paths designed to simulate the evolutionary trends of ISDE during 2020–2030, providing quantitative support for the formulation of region-targeted emission reduction policies.

Fig. 1.

Research framework.

Research methods

Spatial centroid transfer and standard deviational ellipse

To quantify the spatiotemporal and directional characteristics of ISDE, we jointly employ spatial centroid transfer and standard deviational ellipse (SDE) analysis. Centroid transfer tracks the directional displacement of the ISDE gravity center over time, while SDE summarizes the centrality, azimuth, dispersion, and shape of the emission pattern by fitting an ellipse defined by its center, azimuth, major semi-axis, and minor semi-axis¹⁹.

The computational formulas are as follows:

1

2

3

4

Where: represents the coordinates of the center point, θ is the azimuth angle, and are the coordinate deviations respectively along the x-axis and y-axis, and denote the longitude and latitude coordinates of the center point of each province, respectively, and represent the coordinate deviations from the center point of each province to the center of the ellipse, is the ISDE value of each province.

Complex network analysis

To reveal the dynamic spillover effects and inter-provincial interaction mechanisms of ISDE and to overcome the static-analysis limitation, we construct a provincial ISDE correlation network using complex network analysis, which simultaneously quantifies system-wide relationships, visualizes inter-provincial links, and identifies key nodes and tightly-knit communities^20,21. Community detection and centrality metrics are then applied to pinpoint emission-linked regions (see Appendix. S1.1 for detailed procedures).

Grey relational analysis

Grey relational analysis is a specific application of grey system theory in system analysis²². Its principle is based on the data sequence of factors, applying mathematical methods to study the geometric correspondence between factors and quantitatively determine the priority of independent variables relative to dependent variables. This method can overcome the uncertainties, such as lack of information and limited data (Detailed methods are provided in Appendix. S1.2).

System dynamics modeling

The system of influencing factors for ISDE has characteristics of multiple variables and non-linearity, with complex causal relationships among variables. System dynamics can accurately depict the complex interactions and feedback mechanisms between various elements within the system^23,24. When facing insufficient data or difficulties in quantification, system dynamics models can combine limited data with causal relationships for simulation analysis through functions such as table functions and initial function assignment. The construction of a System Dynamics model typically follows the five steps shown in Fig. 2.

Fig. 2.

Modeling steps for ISDE based on system dynamics.

Results

Characteristics of ISDE

This section analyzes the temporal variation trends and distribution characteristics of China’s ISDE from 1992 to 2021, covering three core aspects: the total national ISDE, the total ISDE of each province, and the temporal characteristics of total ISDE.

Spatiotemporal characteristics of ISDE

The temporal variation of China’s total ISDE during 1992–2021 can be roughly categorized into five distinct stages (Fig. 3). The first stage (1992–1997) was characterized by stable fluctuations of total ISDE, with emissions hovering around 15 million tons. The second stage (1998–2006) marked the peak period of ISDE over the 30-year study: though the overall trend was fluctuating downward, emissions first reached the highest historical level (25 million tons) in 1998, remaining in a higher range compared to the previous stage. The third stage (2006–2013) exhibited a steady decrease in total ISDE, with no significant intermittent fluctuations in the overall downward process. The fourth stage (2012–2017) showed a fluctuating upward trend in total ISDE, with emissions increasing intermittently during this period. The fifth stage (post-2018) presented a linear decrease in total ISDE, with total emissions continuing to decline steadily.

Fig. 3.

Total ISDE in China.

The provincial ISDE in China from 1992 to 2021, classified using the Jenks Natural Breaks Optimization method, are detailed in Appendix. S2.1. During 1992–1997, Sichuan and Liaoning were the provinces with relatively high ISDE. The period 1998–2006 witnessed the most severe ISDE, with high-emission provinces mainly concentrated in Hebei, Henan, Shanxi, Shandong, Sichuan, and other regions. From 2007 to 2015, Hebei maintained a relatively high level of emissions. After 2016, the national emission trend slowed down overall, and the emissions of all provinces showed a linear decrease. By 2020 and 2021, the ISDE of most provinces in China had dropped below 238,000 tons.

The centroid transfer curve of China’s ISDE

To analyze the centroid transfer curve of China’s ISDE over the past three decades, annual emissions were used as weights; centroid transfer analysis was employed to calculate the centroids and transfer paths from 1992 to 2021, and standard deviational ellipse analysis was conducted at 3-year intervals using the 1st standard deviation (68.27%). Results are presented in Fig. 4.

Fig. 4.

The centroid transfer curve of China’s ISDE.

Overall, the centroid of China’s ISDE was mainly distributed in Henan, Shaanxi, and Shanxi provinces (Table 1). The movement proceeds in three phases: first southward, then northward, and finally northwestward. During 1992–1999, it moved southward, with its latitude gradually shifting from 34.72°N to the southernmost(33.43°N). During 2000–2013, it moved northward and reached the northernmost point (113.21°E, 35.31°N) in 2013. During 2014–2021, it began to move northwestward, with this movement becoming most obvious from 2016 (112.72°E, 33.99°N) onward and eventually reaching the north-westernmost point (109.43°E, 36.42°N) in 2021.

Table 1.

The centroid coordinates of China’s ISDE.

Years	Coordinates	Years	Coordinates	Years	Coordinates
1992	(114.00°E,34.72°N)	2002	(113.45°E,33.95°N)	2012	(112.55°E,35.46°N)
1993	(113.41°E,34.06°N)	2003	(113.33°E,33.87°N)	2013	(112.58°E,35.52°N)
1994	(113.95°E,34.18°N)	2004	(113.25°E,34.00°N)	2014	(113.29°E,35.31°N)
1995	(113.57°E,34.03°N)	2005	(113.56°E,34.40°N)	2015	(113.45°E,35.11°N)
1996	(113.73°E,34.34°N)	2006	(113.50°E,34.34°N)	2016	(112.72°E,33.99°N)
1997	(113.94°E,34.74°N)	2007	(113.59°E,34.47°N)	2017	(112.45°E,34.36°N)
1998	(113.35°E,33.49°N)	2008	(113.54°E,34.41°N)	2018	(112.27°E,34.03°N)
1999	(113.71°E,33.43°N)	2009	(113.50°E,34.33°N)	2019	(111.89°E,34.03°N)
2000	(113.43°E,33.62°N)	2010	(113.05°E,34.60°N)	2020	(110.03°E,35.49°N)
2001	(113.40°E,33.91°N)	2011	(113.21°E,35.31°N)	2021	(109.43°E,36.42°N)

Data for the standard deviational ellipse analysis is available in Appendix. S2.2. No obvious expansion or contraction pattern was observed in the main scope of the standard deviational ellipse. The longitude of the ellipse center ranged approximately from 110°E to 114°E, and the latitude roughly from 33°N to 35°N, with a certain degree of fluctuation. Specifically, the center coordinates moved northwestward from 1992 to 1996, began shifting northeastward in 2000, and then exhibited a partial northwestward movement again afterward. The lengths of the major and minor axes fluctuated significantly, indicating inherent instability and variability in the spatial distribution of the ISDE. Additionally, the azimuth varied substantially across years, yet its primary directions were consistently concentrated around 55° and 80°. In summary, over the past 30 years, the center position of the ISDE has experienced certain changes and fluctuations, while the overall variation trend remains indistinct, and its spatial distribution demonstrates a certain level of instability.

Phased grade characteristics of ISDE

To gain in-depth insights into the interannual variations of ISDE across provinces and link these variations to the evolution of China’s industrial development stages, the 1992–2021 study period was divided into three phases, and the provincial emissions were classified into five grades, using the Jenks Natural Breaks Optimization method (Fig. 5).

Fig. 5.

Phased grade characteristics of provincial ISDE.

In the phase 1 (1992–2001), the ISDE grades of provinces were concentrated in high-emission areas and upper-middle emission areas, with 10 provinces in the former and 7 in the latter. Jilin, Jiangxi, and Chongqing belonged to middle emission areas, while 11 provinces were categorized as lower-middle emission or low-emission areas. In the phase 2 (2002–2011), ISDE generally increased, with severe pollution even emerging in some regions. The number of high-emission areas remained 10, though Guangxi and Jiangsu were replaced by Heilongjiang and Inner Mongolia; the number of upper-middle emission areas decreased to 2, while the counts of middle emission areas and lower-middle emission areas increased, and Ningxia was reclassified from a low-emission area to a lower-middle emission area. In the phase 3 (2012–2021), provincial ISDE grades were concentrated in lower-middle emission areas (14 provinces), with only 3 high-emission areas (Hebei, Inner Mongolia, Shanxi) and 1 upper-middle emission area (Liaoning) remaining. Notably, Beijing, Tianjin, Shanghai, Hainan and Qinghai, maintained their status as low-emission areas throughout the entire study period. The phase classification depicts the spatial restructuring process of China’s ISDE from the extensive expansion pattern of “high emissions in the east and low in the west” (1990s) to the green transition pattern of “national convergence with local targeted governance” (2010s).

Spatial correlation characteristics of complex networks

ISDE in a region not only affects the ecological security and residents’ livelihoods of that region but also exerts impacts on other regions through spillover effects. The phased grade characteristics above reflect the static spatial differences of ISDE across provinces, but they fail to reveal the dynamic spillover effects and mutual influence between regions. To further clarify the inter-provincial connection mechanisms behind the spatial distribution, we construct a provincial ISDE correlation network using complex network analysis (Fig. 6). Based on the overall statistical properties of complex network analysis, community detection is conducted, and node importance evaluation is performed (Appendix. S1.1).

Fig. 6.

Complex network of China’s ISDE.

The Degree, Weighted Degree, Clustering Coefficient, Closeness Centrality, Betweenness Centrality, and Eigenvector Centrality of each node are presented in Appendix S2.3. In this network, the top-degree nodes are Guangdong, Jiangsu, and Zhejiang (degree = 23), and the bottom-degree ones are Inner Mongolia and Xinjiang (degree = 1). This reflects that the former three are the most critical nodes (most edges to other provinces), while the latter two are less important (fewest edges). The degree values of network nodes are mainly concentrated between 15 and 23, suggesting that most nodes have a relatively high level of connectivity.

Among the nodes, Liaoning, Jiangxi, Jiangsu, Shaanxi, Shanxi, and Shandong demonstrate relatively high edge weights in their inter-provincial connections. In contrast, Xinjiang maintains low edge weights in its connections with most other provinces; Yunnan, Xizang, Qinghai, Ningxia, Inner Mongolia, Gansu, and Fujian also exhibit low edge weights in their links with a large number of provinces.

Notably, high edge weights are concentrated in three types of inter-provincial connections: first, among eastern coastal provinces such as Guangdong, Zhejiang, and Jiangsu; second, among economic hub municipalities and provinces including Beijing, Tianjin, Chongqing, and Shanghai; third, among western provinces like Xinjiang, Qinghai, and Gansu. Additionally, high edge weights are observed between some geographically adjacent provinces, with Guangdong, Zhejiang, and Jiangsu serving as typical examples.

Furthermore, structural indicators of complex network characteristics, such as average degree, average weighted degree, average clustering coefficient, and average path length, were calculated (Table 2). This network demonstrates high density, featuring abundant inter-node connections alongside strong connection strengths. It exhibits a high clustering tendency, with nodes inclined to form aggregated subgroups characterized by tight internal linkages.

Table 2.

Characteristic structure of complex networks.

Measure parameters	Parameter values
Average degree	15.87
The average weighted degree	288.98
The average clustering coefficient	0.79
The average path length	1.63

Community detection was performed using Gephi modularity (Table 3). Modularity measures the quality of community structure in a network: a value closer to 1 indicates a better community structure. In this network, the modularity result is approximately 0.47, which shows that there is a certain degree of community structure in the network, with relatively close connections within communities, while some inter-community connections also exist.

Table 3.

Results of community detection.

Community division	Provinces
Community 1	Shaanxi, Beijing, Shanghai, Hubei, Chongqing, Jiangsu, Jilin, Shandong, Hainan
Community 2	Heilongjiang, Liaoning, Henan, Hunan, Sichuan, Guangxi, Tianjin, Hebei
Community 3	Guangdong, Fujian, Anhui, Jiangxi, Guizhou, Yunnan, Ningxia, Gansu, Inner Mongolia, Xizang
Community 4	Zhejiang, Shanxi, Qinghai, Xinjiang

Analysis of influencing factors of ISDE

This section constructs an indicator system for ISDE, which consists of 17 indicators categorized into 5 subsystems. Additionally, the grey relational analysis method is applied to identify the dominant factors influencing ISDE across various provinces.

Selection and construction of influencing factor indicators

The selection of influencing factor indicators must adhere to the principles of scientificity, systematicness, and feasibility. It is essential to ensure the objectivity and validity of selected indicators, avoid overly correlated or redundant indicators, and ensure that indicators are reliable, easily accessible, and quantifiable. Based on previous studies^25–28, we select 17 indicators and organize them into the 5 subsystems to form the ISDE influencing factor indicator system (Table 4). Detailed reasons for indicator selection are provided in the Appendix. S3.

Table 4.

Influencing factors indicator system of ISDE.

Subsystems	Index	Unit	Abbreviation
Population growth	Permanent population	10k persons	PP
	Employed population in industry	10k persons	EPI
	Gross domestic product per capita	10k persons	GDP
Economic development	Industry added value	100 million RMB	IAV
	Number of foreign enterprises	household	NFE
	Foreign investment	million USD	IFE
	Natural gas consumption	100 million km3	NGC
Energy consumption	Coal consumption	10k tons	CC
	Crude oil consumption	10k tons	COC
	Thermal power generation	100 million kWh	TPG
Industrial production	Cement output	10k tons	CO
	Crude steel output	10k tons	CSO
	Pig iron output	10k tons	PIO
Technology policy	Technical level of governance	%	TLG
	Efforts of environmental governance	%	EEG
	Domestic patent authorization	Item	DPA
	Technology market turnover	100 million RMB	TMT

Due to data source limitations of the National Bureau of Statistics of China, the data for the indicators of “governance technology level” and “environmental governance intensity” cover the period from 2007 to 2021.

Analysis results of provincial leading factors of ISDE

Based on the constructed index system for influencing factors, index data were subjected to preprocessing via mean normalization for dimensionless treatment, and Grey Relational Analysis was employed to analyze the influencing factors of ISDE in 31 provinces, yielding the correlation degree and correlation order between ISDE and each influencing factor in each province (Appendix. S4). The top four dominant factors are relatively consistent across most provinces in China, in the sequence of PP, EPI, COC, and CC: PP exhibits the highest correlation degree in most provinces, EPI serves as the second-ranked dominant factor, COC and CC demonstrate a significant positive correlation with ISDE, and TPG, CO, CSO, and TLG also exert significant effects on such emissions.

Nevertheless, discrepancies in dominant factors exist across certain provinces. Specifically, the primary dominant factor varies by region: Beijing is dominated by CSO, Fujian by NFE, Shanghai by CC, Xinjiang by COC, and Xizang by DPA. For the secondary dominant factor, Guangxi, Hainan, and Tianjin are influenced by NFE; Yunnan and Zhejiang by PP; Beijing by CC; Fujian by PP; and Xizang by CO. As for the tertiary dominant factor, Guizhou, Hubei, Hunan, Sichuan, and Zhejiang are driven by CC; Beijing and Shanghai by PIO; Fujian and Shanxi by CSO; Guangdong and Shandong by CO; and Guangxi and Hainan by EPI. As the correlation order of indices declines, the dominant factors across provinces show a trend of diversification. For instance, the fourth dominant factor is NGC for Hainan, Ningxia, and Xinjiang; CSO for Anhui, Guizhou, and Chongqing; CO for Beijing, Gansu, and Zhejiang; and TPG for Heilongjiang, Qinghai, and Sichuan.

System dynamics modeling

Among high emission provinces, Shanxi Province not only has the highest ISDE volume but also its dominant influencing factors (PP, EPI, CC) are consistent with the national generality identified in Sect.  3.2.2. This consistency ensures that the modeling results of Shanxi can provide a reference for other regions, so Shanxi is selected as a case study to analyze the interaction of influencing factors and simulate future emission scenarios. To this end, after calibrating the system dynamics model against Shanxi’s ISDE for 1992–2020, we constructed 9 future development paths based on policies to simulate emission trajectories up to 2030.

Definition of system boundaries for influencing factors of ISDE

The research boundaries of the system for influencing factors of ISDE are defined to cover 5 subsystems: the population growth subsystem, economic development subsystem, energy consumption subsystem, industrial production subsystem, and technological policy subsystem. The geographical boundary is Shanxi Province, and the time boundary spans from 1992 to 2020, with an annual step size for simulation. The base year is 2020; the model was validated using data from 1992 to 2020, and emissions from 2020 to 2030 were predicted.

Causal loop analysis of each subsystem for ISDE

Causal loops are not only important tools for analyzing system structure and feedback relationships but also serve as a key foundation for mapping the stock and flow diagram of a dynamic system. In this section, causal loops were constructed for each of the 5 subsystems within the system of influencing factors of ISDE (Appendix. S5).

Stock-flow model of the ISDE influencing factors system

Based on the analysis of influencing factors and causal loops, a total of 64 key variables were selected in this study (Appendix. S6.1), including 8 state variables, 9 rate variables, 39 auxiliary variables, and 8 constants. Using these variables, a stock and flow diagram for ISDE was constructed (Fig. 7). The basic parameters and function equations of the model are provided in Appendix. S6.2.

Fig. 7.

Stock-flow model.

Model validation

To ensure the rational structure of the constructed model, the built stock-flow model for ISDE was subjected to intuitive structure validation, robustness validation, and historical behavior validation, with an analysis of the simulation results.

Intuitive structure validation

The stock-flow diagram of the ISDE influencing factors system was refined through repeated discussions, ensuring the rational selection of model variables and the logical consistency of causal relationships. Meanwhile, the rationality of the model structure and the consistency of dimensions are essential prerequisites for the model to be valid. Relevant data underwent preliminary processing to ensure dimensional consistency, which was further confirmed by the model passing the dimensional consistency check when run on the Vensim platform. Additionally, the model passed the boundary check, further verifying both dimensional consistency and model rationality.

Robustness validation

To verify the robustness of the ISDE model, GDP and Cement Output were selected and simulations were conducted for these variables using three different step sizes (Fig. 8). The small differences in simulation results under different step sizes confirm the robustness of the model.

Fig. 8.

Robustness validation.

Historical behavior validation

With the period from 1992 to 2020 as the validation time frame, the historical data of variables were compared with the data simulated by the model in terms of errors to verify the effectiveness of the ISDE model. The representative variables mainly validated include four indicators: GDP, Permanent Population, Coal Consumption, and Cement Output. The error between the simulated values and the actual values is within ± 3% (Appendix. S6.3). Most of our model’s historical fittings yield errors within ± 5%, with only a point approaching the ± 10% range, reflecting relatively satisfactory fitting results (Fig. 9).

Fig. 9.

True and simulated values of ISDE.

Future scenario setting and simulation

We establish 3 scenarios: high-speed development, baseline development, and low-speed development. 9 key influencing indicators are adjusted, including GDP, permanent population, industrial structure, cement output, crude steel output, pig iron output, coal consumption intensity, crude oil consumption intensity, and natural gas consumption intensity. Through the combination of 3 different scenarios of these indicators, 9 different development paths are constructed to simulate ISDE from 2020 to 2030.

Future scenario setting

With relevant government documents for key influencing indicators (Appendix. S7), future development scenarios are set (Table 5).

Table 5.

Parameter settings of key indicators for future scenario prediction.

Key indicators	Low speed	Baseline	High speed
GDP	4.30%	5.10%	5.90%
Permanent population	1.59‰	2.09‰	2.59‰
Industrial structure	− 0.92%	− 1.12%	− 1.32%
Coal consumption intensity	− 5.47%	− 5.97%	− 6.47%
Crude oil consumption intensity	− 3.49%	− 3.99%	− 4.49%
Natural gas consumption intensity	− 2.31%	− 2.51%	− 2.71%
Cement output	− 1.58%	− 2.08%	− 2.58%
Crude steel output	− 0.97%	− 1.40%	− 1.83%
Pig iron output	− 0.97%	− 1.40%	− 1.83%

After setting the change parameters of key influencing factor indicators, future scenario predictions are conducted for GDP, Permanent Population, Industrial Structure, Cement Output, Crude Steel Output, Pig Iron Output, Coal Consumption Intensity, and Natural Gas Consumption Intensity (Fig. 10).

Fig. 10.

Future scenario prediction results of key parameters.

By combining the possible different development scenarios of each indicator, 9 different future development paths are set (Table 6). Here, since Cement Output, Crude Steel Output, and Pig Iron Output all belong to Industrial Production Demand, their development paths are controlled to remain synchronized and replaced by the Industrial Production Demand indicator. Similarly, Coal Consumption Intensity, Crude Oil Consumption Intensity, and Natural Gas Consumption Intensity all belong to Energy Consumption Intensity, so their development paths are controlled to remain synchronized and replaced by the Energy Consumption Intensity indicator. Paths 1–3 see GDP and population controlled to maintain the low speed development scenario, with the scenarios of Industrial Structure, Industrial Production Demand, and Energy Consumption Intensity remaining synchronized and set to low speed, baseline, and high speed development scenarios respectively; Paths 4–6 have GDP and population controlled to maintain the baseline development scenario, with the scenarios of Industrial Structure, Industrial Production Demand, and Energy Consumption Intensity remaining synchronized and set to low-speed, baseline, and high speed development scenarios respectively; Paths 7–9 follow the same logic, with GDP and population controlled to maintain the high speed development scenario, and the scenarios of Industrial Structure, Industrial Production Demand, and Energy Consumption Intensity remaining synchronized and set to low-speed, baseline, and high speed development scenarios respectively.

Table 6.

Setting of Future Scenario Combinations.

Future development paths	GDP	Permanent population	Industrial structure	Industrial production demand	Energy consumption intensity
Path 1	L	L	L	L	L
Path 2	L	L	B	B	B
Path 3	L	L	H	H	H
Path 4	B	B	L	L	L
Path 5	B	B	B	B	B
Path 6	B	B	H	H	H
Path 7	H	H	L	L	L
Path 8	H	H	B	B	B
Path 9	H	H	H	H	H

L refers to the low-speed development scenario, B refers to the baseline development scenario, and H refers to the high-speed development scenario.

Future scenario simulation and prediction

Based on the analysis of future scenario paths in the above table, the ISDE from 2020 to 2030 is simulated (Fig. 11).

Fig. 11.

Future emission scenarios under different paths.

As shown in the figure, the overall trends of the 9 future scenario paths are consistent: Paths 1–3 exhibit the smallest emission reduction trend, Paths 4–6 exhibit a moderate emission reduction trend, and Paths 7–9 exhibit the largest emission reduction trend. Among the three groups of key influencing factors, the emission reduction effect is better when GDP and population are in the high-speed development scenario. Within Paths 1–3, Path 1 achieves the best emission reduction effect specifically, under the premise of controlling GDP and population, the emission reduction effect is optimal when Industrial Structure, Industrial Production Demand, and Energy Consumption Intensity are all in the low-speed development scenario. This same optimal configuration is mirrored in Paths 4–6 and 7–9.

Discussion

Balancing economic growth and ecological governance

This study reveals that the evolution of ISDE in China follows a distinct phased characteristic of “rapid growth - partial control - comprehensive governance - in-depth improvement.” We argue that this trajectory is a vivid manifestation of the dynamic game between economic kinetic energy conversion and ecological environment governance.

The phases of this dynamic game are vividly reflected in China’s historical emissions trajectory. The initial period (1992–2003) exemplifies an era where economic development demands consistently held primacy. Although the state formulated environmental policies, their impact was muted by either insufficient implementation or the overwhelming priority of growth. This tension culminated during the 2004–2008 period, which was coincided with the most severe pollution²⁹. Although environmental protection policies slowed emission growth in some areas, the overall trend remained upward³⁰. The years 2009–2015 marked a critical turning point for structural emission reduction. China intensified environmental governance and elevated ecological management to an important agenda. This top-down intensification of governance led to a significant and structural drop in total ISDE, particularly in the very eastern regions that previously had the highest emissions. Subsequently, as China’s economic development entered a new normal (2016–2021), the industrial structure was continuously optimized and emerging industries gradually developed, driving a sustained overall decline in ISDE.

Placing China’s experience in an international context reveals a unique and compressed narrative of industrial pollution governance. While ISDE is a universal challenge of industrialization¹, nations have forged divergent governance paradigms. Initially, China’s trajectory mirrored the “pollute first, then control” model characteristic of the United States’ historical development. However, the defining feature of China’s path is not the similarity of its starting point, but the unprecedented velocity and top-down nature of its recent transition. Unlike the European Union’s emphasis on gradual, cross-national regulatory coordination or Japan’s iterative legal upgrades through its Air Pollution Control Law, China’s post-2009 shift towards “comprehensive governance” represents a decisive, state-led intervention. This approach effectively synthesizes elements from other models, such as Japan’s focus on total quantity management and the EU’s coordinated control of pollutants, but implements them at a scale and pace that are arguably without precedent. This unique blend of governance paradigms and rapid, centralized execution offers valuable insights for other developing nations.

Multiple factors leading to regional ISDE differences

This study reveals that the regional differences in ISDE are the result of the combined effects of multiple factors. This aligns with the consensus in existing literature on multi-faceted driving mechanisms^25–28, including the path dependence of economic structure and industrial transition, the gradient difference effect of energy structure and technology diffusion, and the institutional differences in the implementation of environmental regulations. However, our analysis supplements and deepens this understanding in several key aspects, particularly in explaining recent dynamic shifts and their underlying mechanics.

A significant finding of this study is the rebound of ISDE in northwestern China during 2016–2021 (particularly in Xinjiang, Qinghai, Ningxia, and Inner Mongolia), which is consistent with the observations of Tang et al.³¹. We propose that this trend serves as strong evidence for the “pollution haven” hypothesis within contemporary China. This hypothesis, established by Markusen³², posits that stringent environmental policies can drive pollution-intensive industries to relocate to regions with relatively lax oversight. Our research provides new empirical weight to this theory. We find that national air pollution control policies have likely catalyzed the relocation of heavily polluting enterprises to these western, less-developed regions, thereby forming new pollution havens. This specific dynamic of industrial transfer resonates strongly with the findings of Song et al.³³. The “lax oversight” that enables such transfers is reflected in the implementation of environmental regulations; a difference previously noted by Revesz³⁴. Moreover, the level of governance technology and scientific innovation also leads to the differentiation of ISDE. The widespread application of high-tech in the southeastern regions helps to improve the effect of dust governance; in contrast, western provinces lag behind, facing prominent information asymmetry issues, which further exacerbates the differences in regional governance.

Moreover, a clear divergence is observed. Eastern provinces such as Guangdong, Jiangsu, and Zhejiang (core nodes in the ISDE correlation network, Sect.  3.1.4) have relocated high-pollution industries via the ‘making room for new industries to gain a blue sky’ strategy. Their high edge weights in the network reflect strong economic and technological links, which provide the necessary conditions for industrial structure upgrading—including access to advanced green technologies and market resources for high-value-added sectors. This supports the shift to high-tech and service sectors, thereby boosting the share of tertiary industry, cutting coal consumption (a key ISDE source), and reducing ISDE intensity. This mechanism is consistent with the observed trend in core nodes: while maintaining high network influence, their local ISDE has declined steadily since 2016 (Sect.  3.1.1). In contrast, resource-based provinces in central and western China, trapped in path dependence, still rely on energy-intensive and heavy chemical industries as pillars. This situation is exacerbated by fiscal pressures; consistent with the view of Cheng & Zhu³⁵, performance assessments and a high dependency on resource taxes worsen the “high emissions, low transition” vicious cycle. Consequently, their persistently high coal consumption keeps ISDE density far above eastern levels.

Limitations of the modeling

While this study employs a SD model to conduct in-depth analysis and scenario simulations of ISDE in Shanxi Province, several limitations should be acknowledged regarding model application and policy inference.

Regarding data sensitivity and structural assumptions, the model’s parameters are primarily calibrated based on historical data and existing policies. Consequently, future policy shifts or unexpected events, such as abrupt population changes, economic restructuring, or technological breakthroughs, may introduce uncertainty into the prediction results. Furthermore, the current framework treats macroeconomic variables (e.g., GDP and population) as exogenous drivers and assumes synchronized adjustments across different pathways. This approach inevitably simplifies the complex endogenous feedback mechanisms within the economic-energy system and may not fully capture the intricate non-linear trade-offs between variables. Future research could integrate more sophisticated models to better characterize these dynamic coupling effects.

Regarding the regional applicability of the findings, while Shanxi, as a key province for ISDE governance in China, offers important reference value for other regions, different regions exhibit distinct economic, social, and environmental conditions, which lead to variations in the adaptability of emission reduction policies. For instance, economically advanced eastern coastal regions might prioritize market mechanisms and economic incentives, whereas central and western regions with relatively lagging development may rely more on government investment and technical support. Therefore, when formulating emission reduction policies, full consideration should be given to regional differences and policy adaptability, and the “one-size-fits-all” approach should be avoided.

Conclusion

This study employs a multi-method approach to comprehensively investigate the spatiotemporal evolution characteristics, driving factors, and emission reduction paths of China’s ISDE during the period 1992–2021. The main conclusions are as follows:

Nationally, the temporal pattern of ISDE presents a trajectory of “stable fluctuation–linear increase–fluctuating decrease–fluctuating increase–linear decrease” over the past three decades, reflecting the dynamic adaptation of industrial emissions to China’s economic transformation and environmental governance. Regionally, the spatial pattern of ISDE has transformed from the 1990s’ “high east, low west” extensive model to the 2010s’ “national coordinated development + local targeted governance” green pattern—marking a critical shift from scale-driven to quality-oriented emission management. The westward shift of the emission centroid reflects the spatial spillover of industrial activities, while the spatial correlation network identifies eastern hubs and adjacent clusters as core nodes. These findings collectively emphasize the need for both region-specific strategies and coordinated inter-regional regulation.

All influencing factors subsystems (population, economy, energy, industry, and policy) correlate significantly with ISDE, though dominant factors vary by province. While population expansion, energy consumption (particularly coal and crude oil), and industrial activity are the primary drivers, technological regulation and thermal power generation also exert non-negligible impacts. These findings verify that ISDE is driven by the nonlinear synergy of multiple factors rather than single-subsystem dominance.

Multi-scenario simulations for Shanxi identify the optimal strategy as balancing high-speed socio-economic growth with strict control over industrial structure and energy intensity. Broadly, the study confirms that optimizing the industrial structure, curbing energy intensity, and enhancing technological regulation capacity are the most effective measures for decoupling economic growth from ISDE, and this path offers a scalable framework for resource-based regions facing similar development dilemmas globally.

Based on the spatial heterogeneity and dynamic interplay between economic development and ecological governance revealed in this study, we recommend adopting differentiated policy approaches across China’s regions. Eastern provinces should prioritize green manufacturing innovation and regional emission trading, while western provinces require targeted support for clean energy infrastructure and ecological compensation mechanisms. Such region-specific strategies are essential to achieving coordinated, inclusive, and effective industrial emission reduction.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

All authors have made substantial contributions to this work: Shouwei Li designed the overall research framework, collected and processed data, and drafted the initial manuscript; Guoming Zhang (corresponding author) supervised the research process, guided the analytical methods, and revised the manuscript and supplementary materials. All authors reviewed the manuscript. We confirm that there are no conflicts of interest among the authors, and neither the manuscript nor the supplementary materials have been published previously, nor are they under consideration for publication in any other journal.

Funding

This research was supported by the Inner Mongolia Autonomous Region Science and Technology Major Special Project “Demonstration of Comprehensive Governance of Wind Sand Dunes at Yabulai Mountain and Construction of Windbreak and Sand-fixing Forest and Grass Belts on the Desert Edge” (2024JBGS0009).

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Competing interests

The authors declare no competing interests.