The roles of the metallurgy, nonmetal products and chemical industry sectors in air pollutant emissions in China

To calculate the emissions from each sector, we followed the method in Lei et al (2010) and Liu et al (2010):

$$\begin{equation} E_{i,j,p}^{{\rm production}} = \sum\limits_k {A_{j,k,p} \left[ {\sum\limits_m {X_{j,k,m,p} F_{i,j,k,m} } } \right]} \end{equation} \tag{ 1 }$$

$$\begin{equation} F_{i,j,k,m} = F_{i,j,k,m}^{{\rm unabated}} \sum\limits_n {C_{i,j,k,n} \left( {1 - \eta _{i,n} } \right)} \end{equation} \tag{ 2 }$$

where i is the air pollutant type; j is the economic sector; k is either fuel or product type; m is the technology used for production; n is the removal devise; p is the province; and E_i,j,p^production is the production-based emission of air pollutant i from sector j in province p. A_j,k,p is the activity rate of fuel or product k of sector j in province p; X_j,k,m,p is the share of fuel or production processed by a specific technology m of sector j in province p; F_i,j,k,m is the effective emission factor after adjustment by emission control devises; F_i,j,k,m^unabated represents the unabated emission factor. C_i,j,k,n represents the penetration control technology n for air pollutant i in sector j; and η_i,n represents the removal efficiency of control technology n for air pollutant i.

The values for the share of technology (X_j,k,m,p), the penetration of control technology (C_i,j,k,n) and the removal rate η_i,n were adopted from the Greenhouse Gas-Air Pollution Interactions and Synergies model (GAINS, http://gains.iiasa.ac.at/models/gains_models3.html) under the current legislation V5a scenario (Stohl et al 2015). The unabated emission factors (F_i,j,k,m^unabated) were compiled from the GAINS model and Lei et al (2010). We collected the activity rates of fuel and products from China Emission Accounts and Datasets (CEADs, www.ceads.net/data/energy-inventory/) and the China Industry Yearbook, respectively. The emissions for power and transportation sectors were compiled from the Multi-Resolution Emission Inventory for China (MEIC, www.meicmodel.org/).

The uncertainties of our estimation were originated from the emission factors, the activity levels, the share of technology, the penetration of control technology and the removal rates. The largest uncertainties were due to the wide range of emission factors from industry sectors, as the emissions varied greatly among different enterprises (State Environmental Protection Administration 1996). Another source of the uncertainties was the activity levels, reaching 15%–20% among industry sectors (Intergovernmental Panel on Climate Change 2006). Due to lack of data on the penetration of the control technology, we assumed the penetration rate following the current legislations and laws. Zhao et al (2011) quantified the uncertainties of these parameters, and found that the uncertainties of these production-based PM_2.5, SO₂ and NO_x emissions (within 95% confidence intervals around the central estimates), respectively, varied between (−17%–54%), (−14%–13%) and (−13%–37%). Additionally, we compared our estimates with those from MEIC, PKU-FUEL (http://inventory.pku.edu.cn), 2012 Environmental Statistical Report (Ministry of Environmental Protection 2013) and the work of Wang et al (2014), and we found that our estimated emissions were comparable to these studies.

The 2012 MRIO database (Mi et al 2017) was used to calculate the portions of MN&C outputs were driven by consumption behaviors. From the MRIO table, we derived the following equations to estimate the driving force of a sector (Meng et al 2016):

$$\begin{equation} L = \left( {I - A} \right)^{ - 1} \end{equation} \tag{ 3 }$$

$$\begin{equation} C_{j^\prime j}^{p^\prime p} = L_j^p Y_{j^\prime}^{p^\prime} \left( {j \ne j^\prime} \right) \end{equation} \tag{ 4 }$$

$$\begin{equation} {\rm CDF}_{j^\prime j}^{p^\prime p} = C_{j^\prime j}^{p^\prime p} /x_j^p \end{equation} \tag{ 5 }$$

where L is the Leontif inverse representing the total (direct and indirect) input coefficient; I is the identity matrix; A is the direct input coefficient matrix; C_j'j^p'p is the output of sector j in province p driven by sector j' in province j; L_j^p is the row vector of sector j in province p; Y_j'^p' is a vector representing the final demand of sector j' in province p', while the values for the final demand of the other sectors are set to zero; CDF_j'j^p'p, defined as consumption-based driving force, measures the contribution of sector j' in province p' to the output of sector j in province p using consumption-based accounting (figure 1); and x_j^p is the output of sector j in province p.

Zoom In Zoom Out Reset image size

To calculate the pollutant emissions from the upstream sectors of a given sector, we adopted the PBL method introduced by Sonis et al (2010). PBL initially quantifies the output (excluding the output from sector j) that is induced by sector j and can then be extended to calculate the emissions of upstream sectors induced by sector j using the following equations (Skelton et al 2011):

$$\begin{equation} h_{i,j,p} = E_{i,j,p}^{{\rm production}} /x_j^p \end{equation} \tag{ 6 }$$

$$\begin{equation} E_{i,j,p}^{{\rm PBL}} = h_i^* \left( {I - A_j^* } \right)^{ - 1} A_j^p x_j \end{equation} \tag{ 7 }$$

where we define E_i,j,p^PBL as PBL emissions of pollutant i from sector j in province p; A_j^p is the column vector of all intermediate purchases made by sector j in province p; A_j^p' is the matrix of intermediate purchases, with purchases by and from j set to zero; h_i is the intensity vector of direct emissions of air pollutant i per unit output, and its element h_i,j,p is the intensity of direct emissions of air pollutant i per unit output from sector j in province p; and h_i,j,p^⁎ is the intensity vector of direct emissions of air pollutant i per unit output, with the emissions intensity for sector j in province p set to zero.

We then used PBL to calculate the share of the output from sector j in province p driven by sector j' in province p'.

$$\begin{equation} L_{j^\prime}^{p^\prime*} = \left( {I - A_{j^\prime}^{p^\prime*} } \right)^{ - 1} \end{equation} \tag{ 8 }$$

$$\begin{equation} {\rm PBL}_{j^\prime j}^{p^\prime p} = L_j^{p*} A_j^{p^\prime*} x_{j^\prime}^{p^\prime} \end{equation} \tag{ 9 }$$

$$\begin{equation} {\rm PDF}_{j^\prime j}^{p^\prime p} = {\rm PBL}_{j^\prime j}^{p^\prime p} /x_j^p \end{equation} \tag{ 10 }$$

where PBL_j'j^p'p is the output of sector j in province p driven by sector j' in province p' and PDF_j'j^p'p is defined as a pure backward linkage-based driving force, representing the share of the output from sector j in province p driven by sector j' in province p' according to the PBL method.

To combine the emissions from sector j and its upstream sectors, we define the term output-based accounting, which represents the emissions from all sectors that are embodied in the total output x_j of sector j. As shown in figure 1, the output-based accounting consists of two components: (a) the emissions from all sectors, excluding sector j, that are embodied in the total output of sector j and (b) emissions from the production of sector j. The output-based accounting E_i,j,p^Output can then be calculated with the following equation:

$$\begin{equation} E_{i,j,p}^{{\rm Output}} = E_{i,j,p}^{{\rm PBL}} + E_{i,j,p}^{{\rm Production}}. \end{equation} \tag{ 11 }$$

Figure 2 shows that the output-based emissions of MN&C were significantly higher than the production-based emissions due to high PBL emissions. The importance of CISs in emission mitigation increased greatly when considering emissions from upstream sectors. On average, the CIS output-based primary PM_2.5, SO₂ and NO_x emissions in 30 provinces were higher than their provincial production-based emissions by a factor of 7.7. 3.5 and 6.2, respectively. Additionally, PBL identified the CIS in Shandong as a large source of emissions, whereas according to the production-based accounting, the CIS in Hubei was the largest emitter of air pollutant emissions. After the addition of emissions from upstream sectors, output-based air pollutant emissions from the CIS in Shandong were 111.5 Gg primary PM_2.5, 665.0 Gg SO₂ and 694.5 Gg NO_x, which were twice the output-based emissions in Hubei.

Zoom In Zoom Out Reset image size

MSs were major emitters of primary PM_2.5 and SO₂ emissions and indirectly induced substantial quantities of emissions in their upstream sectors. For example, PBL primary PM_2.5 and SO₂ emissions from the MS in Jiangsu (i.e. 130.2 Gg PM_2.5 and 491.6 Gg SO₂) almost reached the production-based emissions of the Hebei MS, which was the largest production-based emitter. Furthermore, MSs were not major sources of NO_x emissions, but indirectly caused large quantities of NO_x emissions in their upstream sectors. The PBL NO_x emissions of MSs in 30 provinces were 6 to 237-fold greater than their provincial production-based emissions.

NPSs were an important sources of direct primary PM_2.5, SO₂ and NO_x, and PBL emissions further increased the contribution of NPSs to SO₂ and NO_x emissions. For most provinces, output-based emissions doubled the production-based SO₂ and NO_x emissions. Using this output-based accounting, the NPS in Shandong became the largest source of primary PM_2.5, SO₂ and NO_x emissions, which was not the case when using the production-based accounting.

Figure 3 shows the upstream sectors from which the PBL emissions were emitted. For MN&C, a common pattern was that PBL emissions were primarily originated from coal mining and power sectors. In detail, power sectors contributed more than 20% of the PBL primary PM_2.5 emissions and nearly half of the PBL SO₂ and NO_x emissions. Coal mining sectors were particularly critical for PBL emissions from CISs, covering 27%, 43% and 24% of PBL primary PM_2.5, SO₂ and NO_x emissions, respectively. Additionally, transport sectors were a major source of PBL NO_x emissions and accounted for approximately 20% of NO_x emissions from MN&C. Figure 3 also indicates that much of PBL emissions were embodied in the interregional trade. For example, MSs from other provinces covered approximately 20% and 15% of the local MS PBL primary PM_2.5 and SO₂ emissions, respectively. These findings indicate that the same amounts of emissions were embodied in the interregional trade of metallurgy outputs since local metallurgy emissions were not counted in the provincial PBL emissions. Similarly, we found that 20% of the PBL primary PM_2.5 emissions from NPSs were embodied in the interregional trade. To further explore the role of trade in upstream emissions, we traced the interregional flows of the PBL emissions in the next section.

Zoom In Zoom Out Reset image size

Figure 4 shows that economically developed regions (eastern coastal provinces) commonly tend to import products embodying primary PM_2.5 emissions from less developed provinces (detailed information on the geographical locations and economic conditions of the 30 provinces is given in figure S1 available at stacks.iop.org/ERL/13/084013/mmedia in the supporting information). Over 75% of the total PBL emissions in the MSs of Jiangsu, Zhejiang, Guangdong, Chongqing and Beijing were originated from other inland provinces (figure 4(a)). For NPSs, the pattern was similar to MSs. More than 50% of the PBL primary emissions in eastern provinces (e.g. Jiangsu, Zhejiang and Guangzhou) and megacities (Beijing, Tianjin, Shanghai and Chongqing) were not sourced locally. Figure 4(c) shows that the PBL emissions from CISs were not heavily influenced by interregional trade compared to the PBL emissions from MSs and NPSs, generating over half of the PBL emissions locally in most provinces (except in Jiangsu, Beijing, Shaanxi and Hainan).

Zoom In Zoom Out Reset image size

Figure 4(a) also reveals that the MS outputs drove high primary PM_2.5 emissions embodied in trade. Large trade flows of primary PM_2.5 emissions were primarily sourced from Hebei. In Hebei, more than 7 Gg of primary PM_2.5 emissions were embodied in the exported outputs to the MSs in each of Jiangsu, Zhejiang, Shandong and Henan province. Inner Mongolia was another important source of the traded PM_2.5 emissions, exporting outputs embodying over 13 Gg and 7 Gg of primary PM_2.5 emissions to Hebei and Jiangsu, respectively. The major driver of these large primary PM_2.5 trade flows was the MS in Jiangsu, causing large PM_2.5 emissions in Hebei, Henan, Anhui, Inner Mongolia and Yunnan. Collectively, the output of the MS in Jiangsu caused 99.6 Gg PM_2.5 emissions in the rest provinces, of which 34.4 Gg (34.5%) was emitted in Hebei and Inner Mongolia.

In the trade flows induced by the outputs of NPSs, the dominant source of the traded primary PM_2.5 emissions was Henan (figure 4(b)). The NPS outputs in Jiangsu, Zhejiang, Anhui, Hunan, Shaanxi and Guangdong each stimulated more than 3 Gg primary PM_2.5 emissions in Henan. The NPS in Jiangsu was the primary destination of the traded primary PM_2.5 emissions, causing 42.9 Gg primary PM_2.5 emissions in other provinces. The trade flow from Henan to Jiangsu involved the largest 16.6 Gg primary PM_2.5 emissions. Additionally, Anhui exported intermediate inputs embodying 6.0 Gg primary PM_2.5 emissions to the MS in Jiangsu.

We found that the outputs of CISs generally induced less traded primary PM_2.5 emissions than those of MSs and NPSs (figure 4(c)). The largest trade flow embodied only approximately 3 Gg primary PM_2.5. The major source of primary PM_2.5 emissions embodied in trade was Shanxi. For example, CISs in Hebei, Shandong, Zhejiang and Jiangsu were collectively responsible for approximately 8 Gg of primary PM_2.5 emissions in Shanxi. Unlike the case in figures 4(a) and (b), there were two major destinations of the traded primary PM_2.5 emissions. The CIS in Jiangsu aggregately drove 24.7 Gg primary PM_2.5 emissions, and the CIS in Shandong drove 18.8 Gg of primary PM_2.5 emissions in other provinces.

The trade flows of SO₂ and NO_x are shown in the supporting information (figures S2 and S3), and both show a pattern similar to that of primary PM_2.5.

Figure 5 compares the PDF and CDF of MN&C downstream industries. For many industrial sectors, the CDF was nearly zero, but these sectors had high values of PDF. This difference occurred because outputs from industrial sectors were primarily used as intermediate inputs.

Zoom In Zoom Out Reset image size

For downstream sectors of MSs, the PDF was generally higher than the CDF. The 99th percentile value was 0.03 for the PDF, twice the 99th percentile of the CDF (0.014). Furthermore, PDF maximally increased to 0.8, while CDF reached only 0.6 at the maximum. According to the consumption-based accounting, local construction sectors were the largest drivers to the MS in each province (table S1). However, using the PBL method, we identified several new driving sectors in addition to construction sectors. In detail, the PDF of general machinery sectors, transport equipment sectors and electrical equipment sectors reached maxima of 0.44, 0.40 and 0.27, respectively (table S1). However, the CDF of these sectors was less than 0.1.

Construction sectors contributed greatly to the outputs from NPSs whether evaluating by PDF or CDF. As shown in figure 3, the highest values of PDF and CDF in the 30 provinces were 0.85 and 0.84, respectively. For most provinces, the total outputs and consumption of construction sectors exerted approximately the same influence on NPSs (table S2). One exception was Jiangsu, where the PDF of construction sectors was 0.46, but the CDF only reached 0.24.

Downstream sectors of CISs had lower PDF and CDF values than those of MSs and NPSs because the highest values of PDF and CDF nearly reached 0.25 and 0.20, respectively. Construction sectors and other services sectors were recognized as major drivers to CISs according to CDF (table S3). However, PBL identified diverse drivers to CISs, including the MSs, agriculture, electronic equipment, textile, clothing and food processing sectors, which were mostly ignored by consumption-based accounting. In particular, the total output of agriculture sectors was the largest driver to the CISs in 5 provinces. For example, the CDF of the Shaanxi agriculture sector was 0.07, but the PDF was 0.25.

Figure 6(a) shows that MSs in Hebei, Jiangsu and Shandong had diversified major PBL-based drivers, including electrical equipment, construction, metal products and general machinery sectors. The PDF of these sectors all had a value greater than 0.1. In further detail, the output of the Hebei construction sector drove 20% of the local MS output, the output of the Jiangsu electrical equipment sector drove 27% of the local MS output, and the Shandong general machinery sector drove 44% of the local MS output. Additionally, other provinces strongly drove the MS output in Hebei (e.g. Jiangsu and Zhejiang). Specifically, outputs of the MSs in Jiangsu and Zhejiang contributed 7.7% and 4.1% of the Hebei MS outputs, respectively.

Zoom In Zoom Out Reset image size

Anhui, Henan and Shandong were the three provinces that emitted large output-based air pollutant emissions from the NPSs, and the outputs of local construction sectors were the dominant drivers of these emissions (figure 6(b)). In Anhui, Henan and Shandong, the PDF of local construction sectors was approximately 0.4. Construction activities in Jiangsu, Zhejiang and Shanghai also largely drove the NPSs in Anhui and Henan. For example, the PDF values of construction sectors from Jiangsu, Zhejiang and Shanghai to the NPS in Anhui were 0.16, 0.03 and 0.02, respectively.

Hubei, Jiangsu and Shandong were important output-based air pollutant emitters of CISs (figure 6(c)). In contrast to MSs and NPSs, CISs did not have a dominant driver that outweighed any other sector. Several different sectors aggregately drove the production of CISs in these three provinces. The outputs of food processing, transportation equipment, electronic equipment and agriculture sectors all accounted for approximately 10% of local CIS outputs.

The parameters affecting the uncertainties of PBL emissions are the intensity vector of air pollutants h and the direct input coefficient matrix A. The parameters affecting the consumption-based outputs were the final demand Y and the direct input coefficient matrix A. The output of each sector was directly derived from the statistical data, and thus we assume their values are robust. Based on equations (4) and (7), we find that the consumption-based outputs varied linearly with the final demand, so did the PBL emissions with the intensity vector. This linear relationship indicates that 1% percent change of the final demand or the intensity vector would result in 1% change of the consumption-based outputs or PBL emissions.

Tables S4–S12 in the supporting information show the sensitivities of PBL emissions to the direct input coefficient matrix. We uniformly changed all elements in the coefficient matrix, and investigate the overall effects on PBL emissions. A non-linear relationship is observed between the PBL emissions and the coefficient matrix. If the coefficient matrix increases or decreases by 5%, the PBL emissions would correspondingly increase or decrease by 10%–15%. We also did sensitivity tests of the coefficient matrix to the consumption-based outputs in tables S13–15, and found similar results for the PBL-based emissions.

Furthermore, we did a Monte Carlo analysis with 10 000 simulations to quantify the uncertainties of the PBL emissions. Based on the method in Lin et al (2014), we assumed that relative errors of all parameters followed the normal distribution. Details on parameters' setting are given in table S16, and the results of Monte Carlo simulations for the 30 provinces are shown in tables S17–19. We have also done Monte Carlo simulations for the consumption-based outputs of the 30 provinces, and the results can be seen in table S20. Generally, a factor of −30%–35% uncertainties are implicated from these Monte Carlo simulations.