Analyzing the Impacts of Inter-Provincial Trade on the Quantitative and Spatial Characteristics of Six Embodied Air Pollutants in China Through Multi-Scenario Simulation

Abstract

Inter-provincial trade is accompanied by the transfer of embodied pollution emissions, leading to emissions leakage, thereby hindering the sustainable development of society. Therefore, it is imperative to analyze the characteristics of embodied pollutant emission and spatial transfer driven by inter-provincial trade. In this study, the quantitative and spatial characteristics of the six main embodied pollutants (i.e., SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀) were analyzed by a hypothetical extraction method (HEM) and complex network analysis (CNA) under an input–output analysis (IOA) framework. Then, the row arrange series (RAS) method was employed to simulate the impacts of varying levels of trade intensity, economic growth rate, and technological progress on embodied pollutants and spatial-transfer characteristics. The major findings are as follows: (i) the increase in inter-provincial trade led to a corresponding rise in embodied pollutant emissions due to the relocation of production activities towards provinces with higher emission intensity. Excessive responsibility was assumed by provinces such as Shanxi and Hebei, engaging in production outsourcing for reducing pollutants. (ii) The macro direction of pollutant transfer paths was from the resource-rich northern and central provinces towards the trade-developed southern provinces. Sectors in the transfer path, such as the industry sectors of Shanxi, Guangdong, Henan, and the transport sector of Henan, exhibited high centrality and dominated pollutant transfer activities in the network. (iii) The industry sector, characterized by substantial energy consumption, was the predominant emitter of all pollutant production-based emissions, accounting for more than 40% of total emissions. This study is conducive to analyzing the impacts of inter-provincial trade on embodied pollutant emissions and developing emissions reduction policies considering equitable allocation of emissions responsibilities from both production and consumption perspectives.

1. Introduction

Reckless energy consumption and pervasive air pollution have been concomitant with the growth of the economy and amelioration of human welfare. The frequent occurrence of global warming, acid rain, enhanced nitrogen deposition, and haze has harmed the natural environment and public health, thereby hindering the sustainable development of society [1,2]. Air quality deterioration is the fifth highest health-related factor causing disease in humans [3]. For example, exposure to PM_2.5 was associated with a variety of circulatory and respiratory diseases, causing millions of deaths [4,5]. To combat severe air pollution, China has implemented significant measures to control and prevent air pollution, such as the implementation of the Action Plan for the Prevention and Control of Air Pollution (2013–2017) and the Three-Year Action Plan for Winning the Blue-sky Defense Battle (2018–2020). To target the primary sources of pollutant emissions, the government has introduced the Fourteenth Five-Year Plan for Renewable Energy Development (2021–2025) and the Fourteenth Five-Year Plan for Industrial Green Development (2021–2025), which are aimed at promoting high-quality growth in the industrial and renewable energy sectors. In general, the aforementioned measures primarily targeted the abatement of pollutants originating from production activities, ensuring the sustainable development of society.

In the past years, numerous scholars have conducted a series of studies on emission inventories, source apportionments, and cost-effective mitigation strategies primarily from the production perspective [6,7,8]. For example, Tong et al. (2018) [7] assessed the evolution of coal-fired power plants and associated SO₂, NO_X, and PM_2.5 emissions in China from 2010 to 2030. Li et al. (2019) [6] estimated VOC emissions from China’s anthropogenic sources during 1990–2017 and investigated the main drivers behind the trends. These studies were crucial for identifying direct emission sources and guiding the implementation of cleaner production technology. However, these studies predominantly focused on pollutant emissions from production activities, neglecting the embodied emissions induced by the supply chain. This oversight resulted in the unfair assignment of emission reduction responsibilities, which would further decrease emission reduction efficiency [9].

Convenient transportation, technological advancements, economic development, and national industrial layout have contributed to the gradual strengthening of inter-provincial trade in China. From an environmental perspective, the rapid expansion of inter-provincial trade exacerbates the transfer of embodied emissions, leading to emissions leakage. Emissions leakage refers to the phenomenon where the reduction of pollutant emissions in one region or country is offset by an increase in emissions in another region or country [10]. Wealthier regions with stronger environmental laws may relocate industries to regions with more flexible environmental regulations, specializing in producing pollution-intensive products to fulfill the needs of other regions [11]. Therefore, inter-provincial trade has exacerbated the issue of pollutant emissions leakage and inequitable environmental accountability.

In order to assess emissions leakage induced by inter-provincial trade, consumption-based accounting (CBA) using a Leontief input–output analysis (IOA) has been widely adopted. This accounting method calculates the pollutant emissions of final products throughout the entire supply chain and attributes emission responsibilities to the final consumers [9]. Numerous studies have utilized CBA to calculate embodied pollutant emissions and determine emission responsibilities at global, regional, and industrial levels [12,13,14]. Globally, extensive research has been conducted to investigate the emission characteristics resulting from multinational trades [10,11,15,16]. With the rapid expansion of inter-provincial trade in China, numerous studies have identified inter-provincial transfer paths of embodied pollutant emissions in China to allocate reduction responsibilities among various provinces [1,9,17,18]. Zheng and Xu (2020) [18] found that most provinces with dominant embodied PM_2.5 emissions were located on the east coast of China, outsourcing emissions to Central and Western China. At industrial levels, CBA has revealed the concealed interrelationship of embodied pollutant emissions across sectors [19,20,21]. In general, characterizing regions and sectors as direct emitters or ultimate consumers in CBA is conducive to the formulation of targeted emission reduction policies. Existing research in CBA assessing sectors’ contributions to emission reduction has primarily focused on the current situation, neglecting the impact of changes in trade intensity.

Scenario analysis and model simulation based on the IOA facilitate the assessment of both embodied emissions and the impact of external factors, such as policy changes and economic development, on embodied emissions, which are beneficial for developing pollutant reduction policies. In the past years, most studies had primarily analyzed the impacts of factors such as economic growth [22,23,24], industrial structure [25], energy intensity [26], and electricity demand [27,28] on the characteristics of pollutant emissions, with the limited studies disclosing the intricate relationship between trade scenarios and emission characteristics. A small number of studies simulating trade scenarios have been conducted. For example, Maeno et al. (2022) [29] estimated the impact of the newly formed automotive supply chain structure on global CO₂ emissions by extracting sectors of the Japanese automotive supply chain. Li et al. (2022) [30] proposed a non-trade scenario in which Germany abstained from trading with the rest of the world, demonstrating how Germany’s participation in international trade contributed to carbon reduction. Xu et al. (2020) [31] calculated various sustainable development indicators under a no-trade scenario by adopting foreign technology to add the total embodied emissions imports back to domestic emissions. However, the aforementioned trade scenarios represented extreme cases and solely focused on non-trade mode, making it challenging to quantify the impact of different levels of trade intensity on pollutant emissions. At the same time, a further processed calculation on the measured embodied emissions underestimated the radiation impact of trade changes in a region on the overall economy.

Establishing reasonable assumptions is crucial for simulating changes in the trade environment and policies. Various methods have been proposed to simulate and forecast diverse scenarios based on input–output tables, with the row arrange series (RAS) method being widely employed by numerous researchers for the compilation of input–output tables, as well as for conducting scenario forecasting and simulation studies [22,25,32,33]. For example, Chen et al. (2023) [25] adopted the RAS method to simulate future economic growth and industrial structure to analyze the development and internal structure of economic sectors in 2025. Wang et al. (2021) [33] employed the RAS method to develop multiple scenarios based on direct CO₂ reduction and final demand mitigation in various industries. Generally, the utilization of the RAS method was more prevalent in the analysis of economic growth and industrial structure, while it was less frequently employed in simulating trade structure and trade intensity. The scenario model based on input–output tables reconstructed by the RAS method quantified different levels of externalities such as the intensity of inter-provincial trade. It provided a more objective evaluation of the radiation impact of trade changes in a region on the embodied emissions of the overall economy.

The analysis of sectoral emission linkages is crucial for assessing a sector’s connection with the rest of the economy. The Classical Multiplier Method (CMM) is the traditional approach to measure sectoral linkage [34]. It merely assesses sectoral linkage by computing a simple technical coefficient, which is represented by the column sum in Leontief for backward linkage and the row total in the Ghosh model for forward linkage [35]. The hypothetical extraction method (HEM) is used to extract a sector from the economic system and analyze its economic contribution by comparing the actual economic system to a hypothetical economic system in which the sector has been extracted [36,37]. The HEM analyzes the relative magnitude of a sector’s impact on the rest of the sectors, considering the technological levels for all sectors [38]. It has been widely accepted by scholars and applied to the linkage analysis of pollutant emissions. For example, He et al. (2017) [39] calculated inter-industrial air pollutant emissions using the HEM and found that the power and gas sectors contributed to SO₂ emissions and the nonmetal products sector to soot emissions. Li et al. (2022) [38] identified Ningxia and Qinghai as key provinces and industry, transport, and construction as key energy conservation sectors in China.

However, the HEM is inadequate for investigating the characteristics of the trade network and identifying key sectors. The development of complex network analysis (CNA) provides an efficient method for this purpose. CNA abstracts economic entities as nodes in a network and constructs a series of indicators to analyze network characteristics and identify key sectors [40,41,42]. Numerous recent studies have combined complex network analysis with the input–output method to analyze the structural characteristics of pollutant networks. For example, Huo et al. (2022) [43] clarified the spatial association network of provincial building carbon emissions and their influential drivers in China. Huang et al. (2022) [44] detected carbon communities in networks and examined the effect of community structure on sectors’ carbon emissions.

In general, numerous studies have extensively studied the embodied emissions resulting from interregional trade, providing valuable practical and theoretical insights. However, there still exist certain research gaps. Firstly, the trade scenarios depicted in previous studies illustrated extreme and simplistic cases, making it difficult to quantify the influence of varying levels of trade intensities on pollutant emissions. Secondly, previous studies often focused exclusively on emissions of individual pollutants or CO₂, disregarding the analysis of multi-pollutants and impeding comprehensive co-control measures for major air pollutants. Consequently, they failed to provide adequate information for effectively managing the co-control of primary air pollutants. Third, the CMM can assess sectoral linkages by calculating simple technical coefficients, which is insufficient for analyzing the relative magnitude of a sector’s impact and investigating the trade network. Addressing these research gaps is essential for exploring the influence mechanisms of trade on inter-provincial embodied pollutant emissions.

In order to bridge the above research gaps, the contributions of this study were as follows: (i) This study employed the HEM under the 2017 multi-regional input–output (MRIO) model to calculate the embodied pollutant emissions of SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀ from the consumption and production sides and quantify the transfer paths of embodied pollutant emissions. (ii) CNA was employed to describe the overall network characteristics of all pollutants and identify the key transfer media sectors. (iii) The RAS method was adopted to simulate the impacts of varying levels of trade intensity, economic growth rate, and technological progress on embodied pollutant emissions and spatial transfer characteristics.

2. Method

2.1. Leontief Model

The IO model designed by Leontief quantifies the interconnections between sectors and regions. The basic form of the IO model is as follows:

$$X=AX+F$$

(1)

$$X={(I-A)}^{-1}Y$$

(2)

where $X$ is the total output vector of $x_i$, representing the output of economic sector $i$; $I$ represents the identity matrix; $A$ denotes the intermediate demand matrix, with $a_{ij}={\displaystyle \frac{x_{ij}}{x_j}}$ representing the required total output $i$ to produce one unit of $j$; ${(I-A)}^{-1}$ is the Leontief inverse matrix; and $F$ denotes the final demand vector of $f_i$.

Under the above framework, the embodied emissions in CBA can be calculated as follows:

$$C=E{(I-A)}^{-1}\widehat{F}$$

(3)

where $C$ denotes the vector of embodied emissions driven by the final demand; $E=[e_1{e}_2\dots e_n]$ represents the row vector of emission intensity, which indicates emissions per unit of output $e_i$ in the economic sector $i$; and $\widehat{F}$ is the diagonal vector of $F$.

2.2. Emissions Prediction Based on the Row Arrange Series Method (RAS)

The RAS method is widely used for adjusting and estimating input–output tables. The processes of the RAS method include setting the target values for the rows and columns. This method then utilizes the proportional relationship between the actual values and the predetermined target values, enabling matrix rebalancing through iteration [45]. The formulas are as follows:

$$A_1=\widehat{R}{A}_0\widehat{S}$$

(4)

$$\widehat{R}{A}_1\widehat{X}\widehat{S}I=U$$

(5)

$$I^{\mathrm{T}}\widehat{R}{A}_1\widehat{X}\widehat{S}=V$$

(6)

where $A_1$ is the technical coefficient matrix of the target year; $A_0$ is the technical coefficient matrix of the base year; $R$ and $S$ are row and column iteration coefficients, respectively; $\widehat{X}$ is the diagonal matrix of total outputs; $I$ is a column vector with all elements equaling to 1; and $U$ and $V$ represent the column vector of the total outputs and the row vector of the total inputs in the iteration process, respectively.

This study improves the RAS method by setting the inter-provincial trade component of the matrix to a target value. For example, in the scenario of a 10% strengthening of inter-provincial trade, the inter-provincial trade component of the matrix is multiplied by 1.1 and extracted. The extracted elements of the matrix are set to zero and the target values are subtracted from this component. Then, the standard RAS method’s iterative procedure is performed. Due to the nature of the RAS method, the zero cell remains updated to zero and then the extracted portion of inter-provincial trade is refilled into the zeroed cell.

2.3. Use of the Hypothetical Extraction Method (HEM) for Emission Linkage Analysis

The principle of calculating the embodied emissions linkage with the hypothetical extraction method (HEM) involves comparing emission changes between the original economic system and a hypothetical economy in that the target sectors have been extracted [46]. The entire economic system is divided into the target (to be extracted) block $S$ and the rest of the economic blocks $-S$. The emissions of the entire economic system are as follows:

$$C=\left[\begin{array}{c}{C}_s\\ C_{-s}\end{array}\right]=\left[\begin{array}{cc}{E}_s& 0\\ 0& E_{-s}\end{array}\right]\left[\begin{array}{cc}{L}_{s,s}& L_{s,-s}\\ L_{-s,s}& L_{-s,-s}\end{array}\right]\left[\begin{array}{c}{F}_s\\ F_{-s}\end{array}\right]$$

(7)

where $C=\left[\begin{array}{c}{C}_s\\ C_{-s}\end{array}\right]$ is the vector of total emissions of the target block and the rest of the economy; $\left[\begin{array}{cc}{E}_s& 0\\ 0& E_{-s}\end{array}\right]$ is a diagonal matrix of emissions intensities; $\left[\begin{array}{cc}{L}_{s,s}& L_{s,-s}\\ L_{-s,s}& L_{-s,-s}\end{array}\right]$ is the Leontief inverse matrix; and $\left[\begin{array}{c}{F}_s\\ F_{-s}\end{array}\right]$ is the vector of final demand.

Assume that the external linkages of the target sector are extracted and excluded from the import and export activities of the system. The emissions of the extracted economic system are as follows:

$$C^{\prime }=\left[\begin{array}{cc}{E}_s& 0\\ 0& E_{-s}\end{array}\right]\left[\begin{array}{cc}{\left(I-A_{s,s}\right)}^{-1}& 0\\ 0& {\left(I-A_{-s,-s}\right)}^{-1}\end{array}\right]\left[\begin{array}{c}{F}_s\\ F_{-s}\end{array}\right]$$

(8)

The change in emissions which reflects the impact of the extracted sector is as follows:

$$\Delta C=C-C^{\prime }=\left[\begin{array}{cc}{E}_s& 0\\ 0& E_{-s}\end{array}\right]\left[\begin{array}{cc}{L}_{s,s}-{\left(I-A_{s,s}\right)}^{-1}& L_{s,-s}\\ L_{-s,s}& L_{-s,-s}-{\left(I-A_{-s,-s}\right)}^{-1}\end{array}\right]\left[\begin{array}{c}{F}_s\\ F_{-s}\end{array}\right]$$

(9)

The total embodied emissions linkage was further decomposed into four components and are represented as follows:

$$I{E}_s=E_s{(I-A_{s,s})}^{-1}{F}_s$$

(10)

$$M{E}_s=E_s[L_{s,s}-{(I-A_{s,s})}^{-1}]F_s$$

(11)

$$FL{E}_s=E_s{L}_{s,-s}{F}_{-s}$$

(12)

$$BL{E}_s=E_{-s}{L}_{-s,s}{F}_s$$

(13)

$$N{T}_s=FL{E}_s-BL{E}_s$$

(14)

Internal emissions ($I{E}_s$) refer to the emissions generated by sector s to satisfy their own final demand $F_s$. Mixed emissions ($M{E}_s$) refer to the emissions generated when a part of products in sector s are purchased by sector −s as intermediate inputs and then repurchased by sector s. Forward-linkage emissions ($FL{E}_s$) refer to the emissions generated by sector s to meet the final demand of sector −s, reflecting the net outflow of embodied emissions from sector s. Backward-linkage emissions ($BL{E}_s$) refer to the emissions generated by sector −s to meet the final demand of sector s, reflecting the net inflow of embodied emissions to sector s. Net emissions transfer ($N{T}_s$) is the difference between $FL{E}_s$ and $BL{E}_s$.

2.4. Complex Network Analysis

Based on the sectors, this study constructs the following complex network to further understand the relationships within the transfer of embodied emissions among sectors in China. The provincial sectors are visualized as nodes of the network. The edges of the network represent embodied emissions transfers.

2.4.1. Overall Network Characteristics

Network density measures the closeness of inter-provincial links in the embodied emission transfer network. The higher the density of the overall network, the closer the connections between the members of the network [47]. The formula is as follows:

$$D={\displaystyle \frac{L}{n\times (n-1)}}$$

(15)

The embodied emission transfer network in this study is a directed network comprising n nodes. The theoretical maximum number of transfer paths in the network is n(n−1). The actual number of transfer paths included in the network is L.

The path length is calculated as the minimum number of edges between two nodes. The average path length of the network is defined as the mean of the shortest path lengths between any two nodes, assessing the connectivity of a network. The average path length formula is as follows:

$$L={\displaystyle \frac{1}{n(n-1)/2}}{\displaystyle \sum d_{ij}}$$

(16)

where $d_{ij}$ denotes the shortest path length from nodes i to j and n is the number of nodes in the network.

The clustering coefficient is defined as the proportion of the number of edges among neighboring nodes in the potential maximum number of edges among them. The average clustering coefficient formula is as follows:

$$C={\displaystyle \frac{1}{n}}{{\displaystyle \sum }}_1^n{\displaystyle \frac{E_i}{\frac{q_i(q_i-1)}{2}}}$$

(17)

If node i has $q_i$ adjacent nodes, the theoretical maximum possible number of edges is $q_i(q_i-1)/2$. $E_i$ is the actual number of edges in these adjacent nodes.

2.4.2. Centrality

Centrality is an indicator of the significance of each sector within the embodied emissions transfer network [47]. Sectors with high centrality have greater power and influence. It can be assessed by betweenness centrality and closeness centrality.

Betweenness centrality measures the node’s ability to connect other nodes, serving as a bridge. A greater betweenness centrality indicates a stronger control of the sector over other sectors in the network. The betweenness centrality can be calculated as follows:

$$B{C}_i={\displaystyle \frac{{\displaystyle \sum _j^n{\displaystyle \sum _k^{n}g}}{(i)}_{jk}}{g_{jk}}}$$

(18)

where $B{C}_i$ is the betweenness centrality of node i. $g_{jk}$ represents the total amount of the shortest path between node j and node k. $g{(i)}_{jk}$ indicates the number of occurrences of node i in the shortest paths between node j to node k.

Closeness centrality measures the average distance between a node and other nodes in the network. A higher closeness centrality means a closer relationship with other sectors. The closeness centrality can be calculated as follows:

$$C{C}_i={\displaystyle \frac{n-1}{{\displaystyle \sum _j^n{d}_{ij}}}}$$

(19)

where $C{C}_i$ is the closeness centrality of node i. $d_{ij}$ represents the shortest distance between node i and node j.

2.5. Study Area

With China’s rapid economic development and prosperous trade, the air pollution and emissions leakage issues have become the bottleneck impeding social development. More attention has been paid to seek equitable allocation of emissions responsibilities among provinces for sustainable development. The macro direction of the pollutant emission transfer was from the resource-rich northern and central provinces to the trade-developed southern provinces of China. The northern provinces such as Shandong, Shanxi, and Hebei are resource-rich and resource-processing. The southern provinces such as Jiangsu and Zhejiang are economically developed and tended to outsource pollutant-intensive production and service activities to other regions, thereby exacerbating the emissions leakage problem.

2.6. Data Source

The 42-sector MRIO table in 2017 was obtained from China Emission Accounts and Datasets (CEADS). The standard MRIO tables have only been updated to 2017 and the 2017 table is widely adopted by many of the lasted studies in embodied emissions [20,48]. PM_2.5 and PM₁₀, as well as their precursor gases (including SO₂, NO_X, and VOC) and catalysts (CO) were selected as the main pollutants, considering adverse effects on the natural environment and human health. Numerous studies have utilized the above pollutants to calculate embodied emissions and determine emission responsibilities [21,49,50,51]. Direct pollutant emission data were obtained from the Multi-resolution Emission Inventory for China (MEIC), developed by Tsinghua University. The MEIC divided industries into five sectors, excluding the service sector. This study estimated emissions from the service sector based on its share of total emissions in the Chinese Environmentally Extended Input-Output (CEEIO) database developed by Michigan University. Due to the disparity in sector classification between the MRIO table and the emission inventory, the 42 sectors in the MRIO table were aggregated into 5 sectors [52]. Sectors numbered from one to five represent the electric power, industry, transport, agriculture, and service sectors, respectively. The abbreviation of the provinces is presented in

Table 1. Abbreviations for China’s 31 provinces in 2017.

Province	Abbreviation	Province	Abbreviation
Beijing	BJ	Hubei	HB
Tianjin	TJ	Hunan	HN
Hebei	HE	Guangdong	GD
Shanxi	SX	Guangxi	GX
Inner Mongolia	IM	Hainan	HI
Liaoning	LN	Chongqing	CQ
Jilin	JL	Sichuan	SC
Heilongjiang	HL	Guizhou	GZ
Shanghai	SH	Yunnan	YN
Jiangsu	JS	Tibet	TB
Zhejiang	ZJ	Shaanxi	SN
Anhui	AH	Gansu	GS
Fujian	FJ	Qinghai	QH
Jiangxi	JX	Ningxia	NX
Shandong	SD	Xinjiang	XJ
Henan	HA

. To simplify the expression, “The transport sector of Beijing” will be abbreviated as “BJ3” and the expression of the sector in other provinces follows this convention.

2.7. Scenario Design

Table 2. Scenario settings of trade intensities, economic growth, and technological improvement.

	Scenarios	Setting
Trade scenario	ST0	Current trade intensity (2017)	Current trade mode
	ST1	Reduction scenarios	Non-trade
	ST2		50% trade reduction
	ST3		30% trade reduction
	ST4		10% trade reduction
	ST5	Strengthening scenarios	10% trade strengthening
	ST6		30% trade strengthening
	ST7		50% trade strengthening
Economic development scenario	SE1	Economic growth was 4.0%
	SE2	Economic growth was 4.5%
	SE3	Economic growth was 5.0%
Technology improvement scenario	SI1	The rate of decline of pollutant intensity is determined according to the data for 2017–2020	SE1–SI1	Combining SE1 and SI1
			SE2–SI1	Combining SE2 and SI1
			SE3–SI1	Combining SE3 and SI1
	SI2	The rate of decline of pollutant intensity is designed to decline at 2% per year faster than SI1.	SE1–SI2	Combining SE1 and SI2
			SE2–SI2	Combining SE2 and SI2
			SE3–SI2	Combining SE3 and SI2

presents the detailed scenario settings. In detail, twelve scenarios covering seven trade scenarios, three levels of economic growth, and two levels of technology progress were designed to simulate the impact of trade intensity, economic development, and technology improvement on embodied pollutants emissions, respectively. Among them, seven trade scenarios (i.e., ST₁, ST₂, ST₃, ST₄, ST₅, ST₆, ST₇) covering the four trade reduction scenarios of ST₁ (non-trade), ST₂ (reduction by 50%), ST₃ (reduction by 30%), and ST₄ (reduction by 10%) and three trade strengthening scenarios of ST₅ (increased by 10%), ST₆ (increased by 30%), and ST₇ (increased by 50%) were set up to simulate changes in the inter-provincial trade intensity. Economic growth was set at an average growth rate of 4.0%, 4.5%, and 5.0%, and the predicted time was 2030. According to Chen et al., 2023 [25], the proportions of the electric power sector, industry sector, transport sector, agriculture sector, and service sector would be 2.4%, 47.6%, 6%, 7%, and 37% of the total economic output, respectively. Trade intensity was employed to reflect technology improvement scenarios. SI₁ was designed to maintain the existing annual decline rates of trade intensities (reducing by 87.9%, 96.7%, 96.38%, 97.33%, 93.36%, and 93.91% each year for SO₂, NO_X, CO, VOC, PM_2.5 and PM₁₀, respectively) based on pollutant emissions data collected in the MEIC database from 2017 to 2020. Meanwhile, SI₂ was designed to decline at 2% per year faster than SI₁.

Figure 1 presents the input–output analysis (IOA) framework combining the RAS, HEM and CAN methods. First, under the multi-regional input–output (MRIO) model framework, the HEM was employed to calculate the embodied pollutant emissions of SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀ from the production and consumption perspectives, as well as quantify the inter-provincial transfer paths of embodied emissions in China. In detail, this study individually extracted each sector of the 31 provinces in China for a total of 155 extractions, quantifying the inter-provincial sectoral linkages of embodied pollutants. Second, CNA was adopted to describe the overall network characteristics and identify the key transfer media sectors. Lastly, the RAS method was employed to design multiple scenarios to analyze the impacts of inter-provincial trade, economic development, and technology improvement on embodied pollutant emissions. In detail, the RAS method took the sum of the intermediate use of the target value as the row vector and the sum of the intermediate input of the target value as the column vector, adopting the bilateral proportion adjustment algorithm to adjust the row and column vector coefficients so that the intermediate matrix was rebalanced as the updated IO table [53]. The characteristics of embodied pollutant emissions and spatial transfers were then analyzed based on these updated IO tables under different scenarios. The detailed calculation steps can be seen in Figure 1.

3. Results

3.1. Provincial Embodied Pollutant Emissions Derived by PBA and CBA

Two dominant methods exist for measuring pollutant emissions: production-based accounting (PBA) and consumption-based accounting (CBA). PBA refers to the emissions produced by manufacturing activities and CBA attributes emission responsibilities to the final consumers or consuming sectors. According to Equation (7), the provincial embodied emissions of the six pollutants (i.e., SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀) were calculated using PBA and CBA in 2017. In Figure 2, several major regularities are revealed as follows:

First, the characteristics of resource-rich and resource-processing were the primary reasons for the high pollutant emissions in most production-based provinces. As a result, Shandong, Shanxi, Hebei, and Inner Mongolia represented typical production-based provinces. For example, Shanxi had the second largest production-based SO₂ emissions and the fourth largest production-based PM_2.5 and PM₁₀ emissions. This was because Shanxi possessed abundant mineral resources and its industry was dominated by coal mining and washing, which was characterized by high emissions and low added value. Moreover, there was a relatively large difference between PBE and CBE in Shanxi. Its production-based SO₂ emissions were 60% higher than that of CBE. It reflected the fact that Shanxi was more important as a supplier and direct emitter than as a final consumer. The results also indicate that Hebei had the largest CO emissions and the second largest NO_X, PM_2.5, and PM₁₀ emissions from a production-based perspective. The results of PM_2.5 and PM₁₀ emissions were similar to SO₂ and NO_X, due to the characteristics of PM_2.5 and PM₁₀ as the secondary transformation products of pollutants like dust, soot, SO₂, NO_X, and VOC. The transport sector and metals smelting and processing industry in Hebei were well-developed, which was the major source of SO₂, NO_X, VOC, and CO emissions. Specifically, the transport sector in Hebei accounted for 34% and 19% of NO_X and CO emissions, respectively.

Second, consumption-oriented provinces tended to outsource pollutant-intensive production and service activities to other regions, thereby exacerbating the emissions leakage problem resulting from inter-provincial trade [18]. These provinces relied on the resources and manufacturing capabilities of other provinces. Guangdong, Jiangsu, Zhejiang, and Beijing were typical consumption-based provinces. For example, Guangdong had the second largest consumption-based embodied SO₂, CO, PM_2.5, and PM₁₀ emissions. Its consumption-based embodied SO₂, CO, PM_2.5, and PM₁₀ emissions were 37%, 40%, 51%, and 50% higher than their PBE, which indicated that Guangdong was more important as a final consumer than as a supplier or a producer. In detail, the imported embodied SO₂ emissions in Guangdong were 297% higher than the exported embodied SO₂ emissions. This is mainly because the characteristics of the booming economy, resource endowment, and insufficient productivity urged the local authorities to outsource pollutant-intensive and energy-intensive production and services activities to other regions and import a large amount of products and intermediate products from other regions to meet the enormous final demand. The remarkable aspect lays in the excessive responsibility assumed by the provinces engaging in production outsourcing for reducing pollutants, which indirectly exacerbated the environmental governance imbalance. Moreover, consumption-based embodied SO₂ emissions in Beijing, though not large in total emission, were 730% higher than PBE. The primary reason can be attributed to Beijing’s proactive efforts in promoting industrial transformation and limiting the development of industries characterized by high-pollutant emissions and high energy consumption over the past year. Consequently, Beijing was compelled to import goods and services from other provinces, thereby mitigating the exposure to air pollution resulting from direct production activities. However, other provinces (e.g., Hebei) experienced more exposure to air pollution caused by primary supply, final sales, and consumption originating from Beijing. In conclusion, inequity exists in PBA of emission reductions and it is imperative to ensure equitable allocation of responsibilities among provinces, taking into account a comprehensive assessment of emissions from both production and consumption perspectives. This will contribute to a more equitable and effective approach to mitigating pollutant emissions.

Third, certain provinces demonstrated significantly high levels of pollutant emissions from both production-based and consumption-based perspectives, such as Shandong, Henan, and Jiangsu. Take Shandong for example; it had the largest production-based embodied SO₂, NO_X, VOC, PM_2.5, and PM₁₀ emissions and had the largest consumption-based embodied SO₂, PM_2.5, and PM₁₀ emissions. The reasons might be as follows: (i) the energy mix in Shandong province was predominantly composed of coal, leading to the release of substantial quantities of SO₂, NO_X, PM_2.5, and PM₁₀ emissions. (ii) The province of Shandong possessed a relatively comprehensive industrial supply chain, necessitating substantial imports of intermediate products, products, and services from other provinces. (iii) The transport industry in Shandong province held a significant share, serving as the primary source of emissions for CO, NOx, PM_2.5, PM₁₀, and VOC. (iv) The emission intensity of pollutants in Shandong province was relatively high, indicating enhancing pollutant removal technology and eliminating outdated production capacity are effective measures for reducing pollutant emissions. For example, the emission intensity of SO₂ was 0.044 g/Yuan, surpassing that of economically comparable provinces such as Guangdong and Jiangsu. China’s policies for reducing pollutant emissions primarily focused on production-related measures, such as accelerating the coal desulfurization project, promoting the development of renewable energy, implementing advanced pollutant removal and production technologies, and so on. Nevertheless, the government should consider the allocation of responsibilities for reducing pollutant emissions from both the production and consumption perspectives. From a consumption perspective, the Chinese government can exert influence over consumption behaviors through the implementation of environmentally friendly labels on products from key sectors, such as transport equipment, construction, and so on [54,55]. In addition, the government should consider reducing taxes and offering subsidies to manufacturers of environmentally friendly products. Furthermore, the government can impose taxes on intermediate products based on embodied pollutant emissions, thereby incentivizing final producers to select intermediate products with lower upstream emissions [55,56]. Finally, the government can provide subsidies to major emitters to bolster the adoption of advanced production technologies [55,56].

Figure 3 shows the pollutant emission intensity of different provinces from the import, export, and local sides. Several findings are listed as follows: (i) Local emission intensity characteristics resembled exported characteristics because both originate from local production activities. (ii) The average value of exported emission intensity for production-oriented provinces such as Shanxi (3.92 × 10⁻² g SO₂/Yuan), Hebei (17.3 × 10⁻² g CO/Yuan), and Ningxia (8.28 × 10⁻² g NO_X/Yuan) was generally higher than that of the imported emission intensity. These provinces were dominated by resource extraction and heavily polluting industries. Therefore, the reduction of reliance on heavily polluting industries and the enhancement of pollutant removal efficiency is imperative for these provinces. (iii) The distribution pattern of import scatter points exhibited an inverse relationship with that of export. The average value of imported emission intensity for consumption-oriented provinces such as Guangdong (1.68 × 10⁻² g VOC/Yuan) and Hubei (2.19 × 10⁻² g NO_X/Yuan) were generally higher than that of the export and local emissions intensities. This is because consumption-oriented provinces mainly rely on imports from other provinces, particularly those with high emission intensity, to produce fewer pollutants and face less pressure to export goods and services. Consequently, their local and import emission intensities were considerably lower than that of production-oriented provinces.

3.2. Sectoral Embodied Pollutant Emissions

Figure 4 presents the sector-level primary pollutant (SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀) emissions of different Chinese provinces from production and consumption perspectives. From the production perspective, the industry sector was the predominant contributor to all six types of pollutant emissions, accounting for 63%, 43%, 60%, 78%, 75%, and 76% of SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀ emissions, respectively. The primary reason for this phenomenon can be attributed to the inherent characteristics of the industry sector, which is characterized by substantial energy consumption and significant emissions. Take Hebei as an example; the industry sector contributed 73% of CO emissions. The primary reason for this is that Hebei served as one of the prominent steel production bases in China, making up 24.5% of national values [57]. Therefore, enhancing production technology, eliminating outdated production capacity, and facilitating the transformation and upgrading of the steel industry is imperative for achieving pollutant reduction in Hebei.

The electric power sector also played an important role, ranking as the second largest contributor to SO₂, PM_2.5 and PM₁₀ emissions. The predominant reliance on coal in China’s power generation structure resulted in the emission of a substantial number of pollutants. Moreover, this sector provided indispensable energy inputs for nearly all final demands and industries. For example, the electric power sector in Inner Mongolia and Ningxia accounted for 43% and 52% of total SO₂ emissions. These provinces possessed abundant coal resources and played a significant role in exporting electricity to other regions, thereby shouldering an unreasonable and excessive burden of mitigating production-based pollutants. Therefore, the government should actively accelerate the development and utilization of clean energy such as wind and solar power through policy support and investments.

The transport sector was the second largest contributor to NO_X, CO, and VOC emissions, accounting for 36%, 30%, and 19%, respectively. The main reason lays in the rapid growth of freight and passenger transport in the context of industrialization and urbanization. As a result, high emissions in the industry sector often accompany high emissions in the transport sector. For instance, the provinces of Shandong and Henan served as crucial industrial bases, exhibiting the highest levels of production-based NO_X emissions within the transport sector. These findings highlight the pressing necessity to tackle air pollution resulting from the surge in vehicular growth. It is imperative for governmental authorities to actively promote the adoption of new energy vehicles. However, this will significantly increase the demand for electricity, shifting pollutant emissions from the transport sector to the power sector. Therefore, achieving coordinated governance across sectors is critical to reducing emissions across all sectors.

From the consumption-based perspective, the industry sector accounted for the majority of pollutant emissions; the service sector contributed approximately 10%, while the electric power and agriculture sectors made negligible contributions. Take Beijing for example; the service sector accounted for 40–50% of all six types of pollutant emissions. This suggests that the service sector plays a more critical role as an ultimate consumer rather than a direct emitter and should share responsibility for emission reduction.

3.3. Inter-Provincial Embodied Pollutants Transfer Paths

Figure 5a–f illustrate the top 10 embodied pollutant transfer paths of China’s inter-provincial trade in 2017. As a result, the inter-provincial trade has made a significant contribution to the emissions of embodied pollutants. Inter-provincial trade resulted in 4 Mt, 13 Mt, 41 Mt, 10 Mt, 2 Mt, and 3 Mt of SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀, which accounted for 59%, 45%, 49%, 40%, 44%, and 44% of total emissions, respectively.

From the macroscopic characteristics, the key transfer paths were relatively dispersed and demonstrated a consistent trend from the resource-rich northern and central provinces to the trade-developed southern provinces of China [58]. As shown in Figure 5, the coexistence transfer paths of embodied pollutant emissions were from Hebei to Zhejiang, Hebei to Guangdong, Henan to Zhejiang, Henan to Guangdong, and Shandong to Jiangsu. For example, Henan to Guangdong was the largest transfer path for NO_X and VOC emissions, the second-largest transfer path for PM₁₀ emissions, and the third-largest transfer path for CO and PM_2.5 emissions. From the sectoral perspective, the industry sector constituted the primary source of embodied pollutant emissions in the main transfer paths. According to Figure 6d, taking the path of Henan to Guangdong for example, the industry sector accounted for the high quantity of VOC emissions, amounting to 44 kt and representing 59% of the overall VOC emissions. The inter-provincial trade of industrial products contributed to a high level of pollutant emissions within the transport sector. According to Figure 6b, HN3 exported the largest amount of embodied NO_X to GD2 (29 kt), significantly surpassing the transfer path from HN2 (19 kt). Policymakers can focus on common transfer paths and key sectors to achieve synergistic control of embodied pollutant emissions.

In addition, the transfer paths between geographically adjacent provinces, such as Shanxi to Hebei and Hebei to Beijing, as well as Shandong to Jiangsu, exhibited substantial embodied pollutant emissions. For example, the largest transfer paths of SO₂, PM_2.5, and PM₁₀ were from Shanxi to Hebei (43 kt; 20 kt; and 27 kt). Geographically adjacent provinces often benefit from various trade advantages such as reduced transportation costs and times, enhanced supply chain flexibility, and expanded market access.

3.4. Complex Network Analysis of China’s Inter-Provincial Pollutants Transfer Paths

Based on the results from the HEM, this study employed CNA to visually present the network structure and position of each industry of the different provinces for SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀. If the embodied pollutants’ transfer paths of all provinces and sectors are displayed, the original network presents a fully connected network, making it difficult to highlight the key characteristics of provinces and sectors. According to Wang, 2020 [59], removal of the edges that are lower than the average weight and the corresponding node allows for better analysis the characteristics of the pollutants transfer network. Figure 7 demonstrates the embodied pollutant emissions in the top 100 flows to analyze the network structure more clearly. The nodes indicate five sectors in 31 provinces. The size of the nodes represents the degree of the nodes. The color of the nodes indicates the nodes within the same community. The edges between nodes indicate pollutant transfer paths between the sectors. The thickness of the edge indicates the weight of pollutant transfer paths. Identifying overall network characteristics and key sectors with high centrality could help trace the origins of the emissions and make effective and targeted mitigation strategies for embodied pollutant emissions in China’s inter-provincial trade.

3.4.1. Overall Network Structure

This study calculated overall network characteristic indicators using complex network theory, and the obtained main indicators of complex networks of different pollutants are shown in

Table 3. Network structure of embodied pollutant emissions.

Pollutant	Network Density	Clustering Coefficient	Average Path Length	Betweenness Centrality		Closeness Centrality
				Sector	Value	Sector	Value
SO2	0.107	0.799	2.279	GD2	2130	SX2	0.675
NOX	0.109	0.68	2.31	HA2	1651	HA3	0.674
CO	0.103	0.763	2.316	HA2	1626	HE2	0.668
VOC	0.111	0.828	2.23	GD2	1454	GD2	0.653
PM2.5	0.101	0.755	2.25	GD2	1546	SX2	0.698
PM10	0.099	0.757	2.28	GD2	1611	SX2	0.691

. The network density for the six pollutants was relatively similar, with values of approximately 0.1. This indicates insufficient connections between sectors, leading to the concentration of embodied pollutants transfer paths in a few provinces. As shown in Figure 7, there is a significant concentration of transfer paths in Henan, Hebei, Guangdong, and Zhejiang. Nevertheless, the networks exhibited a high clustering coefficient and short average path length. In detail, about 2.3 transfer paths were required for one sector to connect with another sector, and over half of a sector’s neighbors were interconnected. The majority of provinces and sectors can establish connections rapidly through key provinces and sectors. The results indicated that implementing energy-saving technologies and emission reduction policies in key provinces (i.e., Henan, Hebei, and Guangdong) will substantially reduce pollutant emissions nationwide.

3.4.2. Centrality Analysis

Betweenness centrality measures the node’s ability to connect other nodes, serving as a bridge. The sector exhibiting a high betweenness centrality possesses a greater capacity to exert influence over the flow of information and resources among other sectors. As in Figure 7, GD2 had the highest betweenness centrality for SO₂, VOC, PM_2.5, and PM₁₀, and HN2 had the highest betweenness centrality for CO and NO_X. This was primarily due to the robust economic and industrial foundation of these provinces, in conjunction with their advantageous geographical locations and convenient transportation infrastructure. These provinces not only consumed intermediate products from resource-rich provinces (e.g., Shanxi, Inner Mongolia, and Shandong) for production but also supplied final products to other provinces (e.g., Zhejiang and Jiangsu). In addition, the majority of the shortest transfer paths between sectors traversed through industrial sectors. Therefore, the government should prioritize attention to the industrial sectors, particularly in Guangdong and Henan, when formulating and implementing policies on emission reduction and energy-saving technologies policies.

Closeness centrality measures the average distance between a node and other nodes in the network. A higher closeness centrality means a closer relationship with other sectors. As in Table 3, SX2, HN3, HE2, and GD2 had a high closeness centrality, indicating that the embodied pollutant emissions of those sectors can be transferred more quickly and efficiently throughout the network. Meanwhile, import and export activities in those sectors were the least susceptible to being interfered with and influenced by other sectors. These central provinces (i.e., Shanxi, Hebei, and Henan) were situated in the heartland of China and are characterized by their close proximity to other provinces. As in Figure 7b, the NOx network exhibited high closeness centrality, which was mainly due to the prevalence of e-commerce and the rapid development of the transport sector. This suggests that the transport sector holds a prominent status in the NOx network, enabling it to influence the NOx emissions of other sectors. Therefore, promoting low-emissions transports, such as electric vehicles, boosting public transportation, and raising public awareness about low-emissions travel, represent effective strategies for mitigating NOx emissions.

3.5. Scenarios Analysis

3.5.1. Analysis of the Impact of Trade Intensity on Pollutant Emissions

Based on the RAS method, this study rebalanced input–output tables for seven scenarios to simulate changes in inter-provincial trade intensity (Table 1).

Table 4. Emissions and network characteristics of six embodied pollutants under different trade scenarios.

Pollutant	Scenarios of Trade	Total Emissions (kt)	Clustering Coefficient	Average Path Length	Betweenness Centrality		Closeness Centrality
					Sector	Value	Sector	Value
SO2	ST0	9423	0.799	2.279	GD2	2130	SX2	0.675
	ST1	9274
	ST2	9366	0.787	2.495	GD2	1984	SX2	0.639
	ST3	9392	0.791	2.361	GD2	2399	SX2	0.662
	ST4	9413	0.802	2.3	GD2	2224	SX2	0.672
	ST5	9431	0.804	2.25	GD2	2136	SX2	0.694
	ST6	9444	0.803	2.2	GD2	1998	SX2	0.713
	ST7	9423	0.812	2.17	GD2	2007	SX2	0.813
NOx	ST0	21,349	0.68	2.31	HA2	1651	HA3	0.674
	ST1	20,754
	ST2	21,096	0.66	2.547	HA2	2594	HA3	0.631
	ST3	21,205	0.664	2.43	HA2	2119	HA3	0.663
	ST4	21,302	0.671	2.35	HA2	1850	HA3	0.667
	ST5	21,391	0.689	2.27	JS2	1588	HA3	0.678
	ST6	21,471	0.687	2.2	JS2	1517	HA3	0.709
	ST7	21,547	0.681	2.15	JS2	1284	HA3	0.714
CO	ST0	82,543	0.763	2.316	HA2	1626	HE2	0.668
	ST1	80,735
	ST2	81,813	0.755	2.465	HA2	2293	HE2	0.648
	ST3	82,139	0.747	2.36	HA2	1847	HE2	0.661
	ST4	82,417	0.76	2.32	GD2	1912	HE2	0.668
	ST5	82,654	0.768	2.22	HE2	1482	HE2	0.702
	ST6	82,851	0.769	2.19	HA2	1535	HE2	0.719
	ST7	83,013	0.78	2.16	GD2	1389	HE2	0.723
VOC	ST0	25,523	0.828	2.23	GD2	1454	GD2	0.653
	ST1	25,271
	ST2	25,431	0.806	2.4	GD2	1777	GD2	0.624
	ST3	25,475	0.817	2.32	GD2	1527	GD2	0.642
	ST4	25,510	0.822	2.26	GD2	1360	GD2	0.649
	ST5	25,534	0.826	2.21	GD2	2043	GD2	0.619
	ST6	25,547	0.826	2.16	GD2	1155	SD2	0.691
	ST7	25,548	0.838	2.11	GD2	1051	SD2	0.712
PM2.5	ST0	4674	0.755	2.25	GD2	1546	SX2	0.698
	ST1	4546
	ST2	4625	0.741	2.438	GD2	2162	SX2	0.641
	ST3	4648	0.733	2.348	GD2	1843	SX2	0.657
	ST4	4666	0.746	2.273	GD2	1623	SX2	0.68
	ST5	4681	0.755	2.239	GD2	1523	SX2	0.702
	ST6	4692	0.758	2.213	GD2	1378	SX2	0.711
	ST7	4700	0.758	2.17	SX2	1567	SX2	0.738
PM10	ST0	6888	0.757	2.28	GD2	1611	SX2	0.691
	ST1	6710
	ST2	6821	0.75	2.454	GD2	2166	SX2	0.638
	ST3	6852	0.745	2.353	GD2	1634	SX2	0.656
	ST4	6877	0.751	2.299	GD2	1638	SX2	0.672
	ST5	6897	0.759	2.26	GD2	1590	SX2	0.698
	ST6	6912	0.757	2.21	GD2	1464	SX2	0.699
	ST7	6922	0.763	2.18	SX2	1552	SX2	0.728

and

Table 5. (a) Transfer path ranking of SO₂ in various trade scenarios; (b) transfer path ranking of NOx in various trade scenarios; (c) transfer path ranking of CO in various trade scenarios; (d) transfer path ranking of VOC in various trade scenarios; (e) transfer path ranking of PM_2.5 in various trade scenarios; and (f) transfer path ranking of PM₁₀ in various trade scenarios.

(a) Transfer path ranking of SO2 in various trade scenarios
Scenarios of trade	Rank	Starting	Destination	Value
ST0	1	SX2	HE2	26
	2	SX2	JS2	25
	3	SX2	ZJ2	22
ST2	1	SX2	HE2	17
	2	SX2	ZJ2	15
	3	SX2	JS2	15
ST3	1	SX2	HE2	21
	2	SX2	JS2	19
	3	SX2	ZJ2	18
ST4	1	SX2	HE2	25
	2	SX2	JS2	23
	3	SX2	ZJ2	21
ST5	1	SX2	HE2	28
	2	SX2	JS2	26
	3	SX2	ZJ2	23
ST6	1	SX2	HE2	30
	2	SX2	JS2	29
	3	SX2	ZJ2	24
ST7	1	SX2	HE2	31
	2	SX2	JS2	31
	3	SX2	ZJ2	25
(b) Transfer path ranking of NOx in various trade scenarios
Scenarios of trade	Rank	Starting	Destination	Value
ST0	1	HA3	JS2	38
	2	HE2	ZJ2	30
	3	HE2	GD2	29
ST2	1	HA3	JS2	23
	2	HA3	ZJ2	20
	3	HA3	GD2	19
ST3	1	HA3	JS2	30
	2	HA3	ZJ2	24
	3	HA3	GD2	23
ST4	1	HA3	JS2	36
	2	HE2	ZJ2	28
	3	HA3	ZJ2	27
ST5	1	HA3	JS2	41
	2	HE2	ZJ2	32
	3	HE2	GD2	32
ST6	1	HA3	JS2	45
	2	HE2	GD2	36
	3	HE2	ZJ2	36
ST7	1	HA3	JS2	49
	2	HE2	GD2	39
	3	HE2	ZJ2	39
(c) Transfer path ranking of CO in various trade scenarios
Scenarios of trade	Rank	Starting	Destination	Value
ST0	1	HE2	ZJ2	215
	2	HE2	GD2	212
	3	HA2	GD2	160
ST2	1	HE2	ZJ2	127
	2	HE2	GD2	120
	3	HA2	GD2	99
ST3	1	HE2	ZJ2	166
	2	HE2	GD2	159
	3	HA2	GD2	127
ST4	1	HE2	ZJ2	200
	2	HE2	GD2	195
	3	HA2	GD2	150
ST5	1	HE2	ZJ2	230
	2	HE2	GD2	228
	3	HE2	JS2	172
ST6	1	HE2	GD2	257
	2	HE2	ZJ2	256
	3	HE2	JS2	195
ST7	1	HE2	GD2	284
	2	HE2	ZJ2	279
	3	HE2	JS2	216
(d) Transfer path ranking of VOC in various trade scenarios
Scenarios of trade	Rank	Starting	Destination	Value
ST0	1	HA2	GD2	44
	2	JS2	GD2	39
	3	SD2	ZJ2	39
ST2	1	HA2	GD2	27
	2	HA2	ZJ2	24
	3	JS2	ZJ2	23
ST3	1	HA2	GD2	35
	2	HA2	ZJ2	30
	3	JS2	ZJ2	30
ST4	1	HA2	GD2	42
	2	JS2	GD2	36
	3	JS2	ZJ2	36
ST5	1	HA2	GD2	27
	2	HA2	ZJ2	24
	3	JS2	ZJ2	23
ST6	1	HA2	GD2	52
	2	SD2	JS2	48
	3	SD2	ZJ2	47
ST7	1	HA2	GD2	56
	2	SD2	JS2	54
	3	SD2	ZJ2	53
(e) Transfer path ranking of PM2.5 in various trade scenarios
Scenarios of trade	Rank	Starting	Destination	Value
ST0	1	SX2	HE2	14
	2	SX2	JS2	13
	3	SX2	ZJ2	12
ST2	1	SX2	HE2	9
	2	SX2	ZJ2	8
	3	SX2	JS2	8
ST3	1	SX2	HE2	12
	2	SX2	JS2	10
	3	SX2	ZJ2	10
ST4	1	SX2	HE2	14
	2	SX2	JS2	13
	3	SX2	ZJ2	11
ST5	1	SX2	HE2	15
	2	SX2	JS2	14
	3	SX2	ZJ2	13
ST6	1	SX2	HE2	16
	2	SX2	JS2	16
	3	SX2	ZJ2	13
ST7	1	SX2	HE2	17
	2	SX2	JS2	17
	3	SX2	ZJ2	14
(f) Transfer path ranking of PM10 in various trade scenarios
Scenarios of trade	Rank	Starting	Destination	Value
ST0	1	SX2	HE2	20
	2	SX2	JS2	18
	3	SX2	ZJ2	16
ST2	1	SX2	HE2	12
	2	SX2	ZJ2	11
	3	SX2	JS2	11
ST3	1	SX2	HE2	16
	2	SX2	JS2	14
	3	SX2	ZJ2	14
ST4	1	SX2	HE2	18
	2	SX2	JS2	17
	3	SX2	ZJ2	15
ST5	1	SX2	HE2	21
	2	SX2	JS2	20
	3	SX2	ZJ2	17
ST6	1	SX2	HE2	22
	2	SX2	JS2	22
	3	HA2	GD2	18
ST7	1	SX2	HE2	23
	2	SX2	JS2	23
	3	HA2	GD2	20

illustrate the emissions, transfer paths, and network characteristics of six pollutants under different scenarios. The results indicated that the increase in inter-provincial trade intensity would lead to a corresponding rise in pollutants. For example, comparing ST₅ with ST₀, a 10% strengthening of inter-provincial trade resulted in a 0.08%, 0.2%, 0.13%, 0.04%, 0.15%, and 0.13% increase in the total embodied emissions of SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀. A similar study was conducted by Yang et al., 2019 [60], concluding that inter-provincial trade in China contributed to an increase in SO₂ emissions, which is consistent with this study’s findings. The reason behind this phenomenon lies in the fact that as inter-provincial trade increases, there is a tendency for production activities to relocate towards provinces with higher emission intensity. As other scholars have examined, the findings that increased trade exacerbates pollutant emissions apply equally to international studies [16,61,62]. For example, some production activities such as raw materials and intermediate goods production are relocating from China and India to other developing countries, and the embodied emissions have surged [16].

From the province perspective, production-based provinces such as Shanxi were the victims of the strengthening of trade in the environmental aspect. For example, economic output of Shanxi increased from 2049 billion under ST₀ to 2078 billion under ST₅, but Shanxi has endured a greater environmental burden. Shanxi’s share of national production-based emissions for SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀ increased from 8.02%, 4.38%, 4.52%, 2.54%, 6.75%, and 6.37% under ST₀ to 8.13%, 4.44%, 4.57%, 2.57%, 6.82%, and 6.43% under ST₅. This is mainly because the intensification of inter-provincial trade has resulted in Shanxi assuming a greater share of production tasks. In contrast, Zhejiang’s share of national emissions from a production perspective would decrease, but its share of national emissions from a consumption perspective for SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀ emissions would increase by 0.11%, 0.06%, 0.06%, 0.01%, 0.12%, and 0.12%. This was mainly because Zhejiang, as a typical consumption-based province, outsourced pollutant-intensive and energy-intensive production activities to other regions and imported a large number of products and intermediate products from other regions to meet the enormous final demand. The results indicated that the escalation of inter-provincial trade has further exacerbated the issue of pollutant leakage and inequitable environmental accountability. Therefore, it is imperative to ensure a fair distribution of responsibilities among provinces considering both production-based and consumption-based perspectives, which would contribute to a more equitable and effective approach to reduce pollutant emissions. However, from the perspective of sector, changes in trade have a minimal impact on the sector.

3.5.2. Analysis of the Influence of Trade Intensity on Embodied Pollutants’ Transfer Paths

As in Table 5, changes in inter-provincial trade have exerted a remarkable influence on the embodied pollutants’ transfer paths. For example, the significance of the transfer path from HE2 to GD2 became more prominent with the strengthening of the trade intensity. In comparing scenarios ST₆ and ST₇ to ST₀, a 30% and 50% strengthening of trade would result in the transfer path from HE2 to GD2 increasing by over 20%, becoming the largest inter-provincial transfer of CO and surpassing the transfer path from HB2 to JS2. Meanwhile, the embodied pollutants transfer paths from HE2 to GD2 would increase by over 50% under ST₆ and ST₇ compared to ST₂, becoming the second largest transfer path of NO_X. More attention should be given to the potential environmental risks arising from the emergence of transfer paths with high emissions due to changes in inter-provincial trade.

3.5.3. Analysis of the Impacts of Trade Intensity on Pollutants’ Network Characteristics

The characteristics of the pollutants’ networks would undergo significant changes in response to variations in inter-provincial trade. Several major findings can be summarized as follows:

(i)

The average clustering coefficient gradually increased with the strengthening of trade, while the average path length decreased. Moreover, the connections between provinces in embodied pollutants transfer paths became progressively closer. For example, the average clustering coefficient of SO₂ gradually increased from 0.787 under ST₂ to 0.812 under ST₇. The average path length of SO₂ gradually decreased from 2.495 in ST₂ to 2.17 in ST₇. This suggests that the interdependence between sectors would significantly increase with the strengthening of trade. Therefore, the government should adopt a more comprehensive approach to formulate emission reduction policies, ensuring concerted efforts are made to reduce emissions across diverse regions and sectors.

(ii)

The study reveals that alterations in inter-provincial trade exerted a substantial influence on betweenness centrality. Under the scenarios ST₂, ST₃, and ST₄, the betweenness centrality of GD2 and HA2 was found to be the highest. However, the roles of GD2 and HA2 as pivotal points declined with the strengthening of inter-provincial trade. For example, the betweenness centrality of HA2 decreased by 45%, dropping from 2594 in ST₂ to 1405 in ST₅ within the NO_X network. In contrast, trade reallocated resources and strengthened the centrality of JS2. Under ST₅, ST₆, and ST₇, JS2 surpassed HA2 as having the largest betweenness centrality for NO_X. Similarly, under the scenario ST₇, SX2 surpasses GD2 as having the largest betweenness centrality in the PM_2.5 and PM₁₀ networks. As a result, greater attention should be directed towards provinces and sectors that serve as pivotal nodes in the transfer of pollution with changes in inter-provincial trade. Therefore, the government should scientifically assess the demand for upstream and downstream products in intermediary sectors such as HA2 and SX2. Subsidies should be provided to these sectors to develop cleaner production technologies.

(iii)

The results indicate that most closeness centrality remained relatively stable despite changes in trade. It could be found that the sectors like SX2, HA2, and HE2 that ranked high according to closeness centrality indicators did not change largely, indicating the stable central status of these sectors. Due to the strong industrial foundation and superior geographical location, these sectors can directly provide or receive pollutants, with limited reliance on intermediate departments.

3.6. Analysis of the Influence of Economy and Technology

In order to simulate the impact of economic development and technological improvement on embodied pollutants emissions, different scenarios of economic growth and technological progress were designed (as in Table 1). Figure 8 illustrates the embodied pollutant emissions from various sectors in each province in 2030 under six future scenarios. The results indicated an overall downward trend in SO₂, PM_2.5, and PM₁₀ emissions, while NO_X, CO, and VOC emissions exhibited an overall upward trend under SI₁ from 2017 to 2030. For example, under SE₁–SI₁, SO₂, PM_2.5, and PM₁₀ emissions would decrease by 70%, 37%, and 34% from 2017 to 2030, respectively, while NO_X, CO, and VOC emissions would increase by 14%, 7%, and 14%. There are two reasons for the decrease in SO₂, PM_2.5, and PM₁₀ emissions. First, the share of industry in the national economy would decline from 2017 to 2030. Second, desulfurization and dust removal technologies have effectively reduced emissions in recent years. The increase in NOx and CO emissions might be attributed to the future transport industry being poised to play an increasingly significant role in the national economy, as it was the primary source of NOx and CO emissions. Moreover, the increase in economic growth rate would lead to an increase in pollutant emissions. For example, under SE₂–SI₁, associated with 5% economic growth in 2030, pollutant emissions would be 6.4% higher than those under SE₁–SI₁ with 4.5% economic growth.

Reducing emission intensity could effectively decrease pollutant emissions. Under SE₁–SI₂, emissions of SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀ would decrease by 78%, 13%, 19%, 12%, 53%, and 50% from 2017 to 2030. This is mainly due to the promotion of clean energy and the progress of emission mitigation technology, contributing to a decrease in emission intensity throughout the planning horizon. Compared with the scenarios of SE₁–SI₁ and SE₁–SI₂, this result indicated that a 2% decrease in pollutant intensity would result in a reduction of 724 kt, 5807 kt, 21,081 kt, 6921kt, 720 kt, and 1116 kt of SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀ emissions. Even with the rapid economic development leading to increased energy consumption and pollutant emissions in the future, emission mitigation technology can still effectively reduce pollutant emissions. For example, under SE₃–SI₂, emissions of SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀ would decrease by 75%, 1%, 8%, 1%, 46%, and 43% when comparing 2017 to 2030. Developing emission reduction technologies can effectively balance the tradeoff between economic development and environmental emissions. It is noteworthy that the changes in economy and technology have minimal impacts on the shares of pollutant emissions between different provinces and sectors.

4. Conclusions

This study establishes an input–output analysis (IOA) framework combining the RAS method, the HEM, and CNA to assess the characteristics of China’s inter-provincial transfer of embodied emissions under different trade scenarios. In detail: (i) the HEM under the IOA framework was adopted to calculate the provincial and sectoral embodied pollutant emissions of SO₂, NO_X, CO, VOC, PM_2.5, and PM₁₀ from the production and consumption perspectives and quantify the transfer paths of embodied pollutant emissions. (ii) CNA was employed to describe the overall network characteristics of all pollutants and identify the key transfer media sectors. (iii) Based on the RAS method, twelve scenarios were designed to simulate the impacts of trade intensity, economic development, and technology improvement on embodied pollutants emissions and emission characteristics.

The main conclusions are summarized as follows:

(1)

The macro direction of emissions transfer paths was from the resource-rich northern and central provinces towards the trade-developed southern provinces. Complex network analysis revealed that GD2 and HN2 had the highest betweenness centrality, while SX2, HN3, HE2, and GD2 had high closeness centrality. The results imply that implementing energy-saving technologies and emission reduction policies in key provinces (i.e., Henan, Hebei, and Guangdong) will substantially reduce pollutant emissions nationwide.

(2)

The increase in inter-provincial trade intensity would lead to a corresponding rise in pollutants due to the relocation of production activities towards provinces with higher emission intensity. The escalation of inter-provincial trade has further exacerbated the issue of pollutant leakage and inequitable environmental accountability. Therefore, when formulating national pollution emission reduction policies, emission reduction tasks should be reasonably distributed among provinces and attention should be paid to the impact of inter-provincial trade to prevent outsourcing more pollution to the central and northern provinces.

(3)

Inter-provincial trade contributed to the transfer of embodied emissions, leading to emissions leakage. Production-based provinces and consumption-based provinces should adopt different emission reduction policies. For example, emission reduction policies for production-based provinces could be implemented, such as accelerating the coal desulfurization project, promoting the development of renewable energy, and implementing advanced pollutant removal and production technologies. From a consumption perspective, the Chinese government can exert influence over consumption behaviors through the implementation of environmentally friendly labels on products, imposing taxes on intermediate products based on embodied pollutant emissions, and allocating subsidies towards major emitters to bolster the adoption of advanced production technologies.

(4)

From the sector perspective, the industry sector, characterized by substantial energy consumption, was the predominant emitter of all pollutant production-based emissions, accounting for more than 40% of total emissions. From the consumption-based perspective, the industry sector accounted for the majority of pollutant emissions and the service sector contributed approximately 10%, while the electric power and agriculture sectors made negligible contributions.

(5)

Rapid economic development would increase energy and resource consumption, consequently leading to higher emissions. Developing emission reduction technologies can effectively balance the tradeoff between economic development and environmental emissions.

Compared with previous studies, this research employed the RAS method to further simulate the IOA of specific trade scenarios to reflect the impact of inter-provincial trade on pollutant emissions. It is worth mentioning that there are some limitations in terms of methods and results. First, the original 42 sectors in the input–output table were merged into five aggregated sectors based on MEIC and CEEIO emission inventory due to incomplete data for refined sectors. This data processing method inevitably created errors because it ignored the differences in pollutant emission intensity in some aggregated sectors. Compiling the more detailed sector–scale pollutant emissions inventory is essential to improve the accuracy. In addition, more detailed scenarios will be modeled in the future under the National Economic Circle Strategy.