Characteristics of PM_2.5 and Its Reactive Oxygen Species in Heating Energy Transition and Estimation of Its Impact on the Environment and Health in China—A Case Study in the Fenwei Plain

Household solid fuels (i.e., coal and biomass) are widely used for residential heating in rural areas of northern China in winter (Feng et al., 2021a). The topic has attracted international attention because of the increasing prominence of air pollution caused by solid fuel combustions (He et al., 2021; Gu et al., 2023). In 2018, the total domestic coal consumption was ~80 million tons, with approximately half being contributed by dispersed coal combustion in rural areas of China (Department of Energy Statistics, National Bureau of Statistics, 2019). The annual agricultural biomass energy utilization in 2017 in China was ~35 million tons of standard coal (National Energy Administration, 2017). Because of the low combustion efficiency and lack of emission control, domestic coal and biomass combustions are the main sources of fine particulate matter (PM_2.5, PM with an equivalent aerodynamic diameter of ≤ 2.5 μm) emission (Lu et al., 2010; Yun et al., 2020), which can lead to serious haze pollution, health impairment, and regional and global climate changes (Pope and Dockery, 2006).

PM_2.5 is a complex mixture composited of organic compounds, inorganic ions, metals, and biological elements (Pope and Dockery, 2006). The influences of PMs are not limited to mass concentration but also chemical composition (Lewné et al., 2007; Hu et al., 2008; Ku et al., 2017). Lewné et al. (2007) reported that there are positive correlations between organic carbon (OC) and element carbon (EC), and induction of cardiopulmonary diseases and increase in mortality. Hu et al. (2008) found that the water-soluble inorganic ions in PM_2.5 can lead to cell death. Toxicological experiments have supported that PM_2.5-bound elements, such as lead (Pb), manganese (Mn), and cadmium (Cd) could induce a variety of human dysfunctions and diseases (Ku et al., 2017). Personal respiratory exposure to PM_2.5 is associated with increased health risks such as bronchial and cardiovascular diseases and mutagenic and carcinogenic effects (Feng et al., 2022). However, the underlying mechanisms of the health risks associated with the different components of PM_2.5 are unknown. One of the pathophysiological mechanisms is oxidative stress (OS) (Hellack et al., 2017). The relationship between different PM_2.5 components and OS is a current popular topic in this field.

OS is caused by the excessive production of reactive oxygen species (ROS), such as superoxide (•O₂⁻), hydrogen peroxide (H₂O₂), and hydroxyl radical (•OH) (Lakey et al., 2016). Previous studies have found that there are strong associations between specific components of PM_2.5 and ROS, proving negative impacts on human health (Samara, 2017; Lei et al., 2022). Metals of iron (Fe), copper (Cu), and Mn have been verified to be the key components in PM_2.5 to induce the ROS generation (Verma et al., 2014). Bioavailable ionizable transition metals participate in a redox-cycling reaction by generating superoxide and hydroxyl radicals (Bevans et al., 2013). To our best knowledge, there are limited studies on the demonstrations of variation of ROS in PM_2.5 caused by different types of household solid fuels. However, household fuels have a short contact distance, a long respiratory exposure time, and obvious environmental and health effects (He et al., 2021; Zhang et al., 2021). So, it is urgent to investigate the oxidation potential and human damage caused by PM_2.5 exposure from burning different fuels.

The National Development and Reform Commission of China established “The 5-year Winter Clean Heating Plan (WCHP) for Northern China from 2017 to 2021” to reduce air pollution (National Development and Reform Commission of China, 2017). Some scholars have evaluated the positive effects of the transition to clean energy in China (Sun et al., 2019; Li et al., 2021a; Zhou et al., 2022). However, due to the high cost of clean heating, scattered settlements in rural areas, shortage of clean energy supply (such as natural gas), and energy security, the large-scale use of natural gas, electricity, and wind energy cannot be rapidly achieved, especially in rural northwestern China (Feng et al., 2021b). Promotion of the uses of environmental-friendly treated solid fuels, such as biomass charcoal, semi-coke, and charcoal-based briquettes can significantly decrease the emissions of sulfur dioxide (SO₂), nitrogen dioxide (NO₂), and PM at a reasonable cost (Zhang et al., 2020). Shen and Xue (2014) reported that substitution of the use of raw coal by the use of briquettes could reduce PM emission by factors ranging from 12.9 ± 10.6 g kg^–1 to 6.37 ± 6.59 g kg^–1 (P = 0.035). Clean coal is a low-cost implementable and an alternative solution, but it has not been systematically evaluated for its PM_2.5 emission reduction and impact on human health.

Tongchuan consists of abundant mineral resources (e.g., coal) and is the first city to excavate coal resources in the Fenwei Plain. It has been selected as the national pilot city to investigate energy conservation for clean heating and air pollution reduction policy in China since 2019. The use of clean coal is a current practice in rural regions, while the clean energy transformation is the priority action in Tongchuan. The objectives of this study are to: 1) investigate the reductions of PM_2.5 and its carbonaceous fraction, water-soluble inorganic ions, and elements in household indoor and personal exposure (PE) of homemakers after the partial replacement of traditional coal and biomass with clean coal; 2) explore the correlations between ROS activities and components of PM_2.5; and 3) estimate the health benefits of the clean energy transition in Tongchuan. This study provides a quantitative analysis of the effects of clean energy substitution and could be used as a basis for governmental guidelines on residential energy consumption in rural areas of northwestern China.

The sampling campaign was conducted at Shibao village (109.34°N, 35.45°E) in Yijun District of Tongchuan city, which is at the northern edge of the Fenwei Plain in northwestern China (Fig. 1). The ambient air temperature is usually very low in winter (i.e., daily average temperature of 1°C–9°C) in the rural areas of Tongchuan, leading to strong demands of residential heating. Clean coal began to replace traditional solid fuels in rural households in 2019. Before this, coal (e.g., lumped bituminous) and biomass (e.g., wood branch) were the main heating fuels in this village. The clean coal used here is a coal product made largely from semi-coke. This material is produced in a high-pressure forming and drying process through the mixing of additives such as a sulfur-fixing agent, binder, and combustion aid. The sulfur content is less than 0.5%, volatile content is less than 20%, and the calorific value is higher than 27 MJ kg^–1 [Table S1 in in the electronic supplementary material (ESM)] (Zhang et al., 2021).

Figure 1.  Sketch map of the sampling site and photos of the PM_2.5 sampling indoor and personal exposure.

In this study, 18 households were selected (the selection process and criteria are described in ESM) and divided into three groups based on heating fuels used, including lump coal (Fig. S1a in the ESM), biomass (Fig. S1b), and clean coal (Fig. S1c) groups. Six households were involved in each fuel type group to collect PM_2.5 indoors and corresponding PE samples from homemakers. The architectural features of these 18 households are common and thus represented in the Fenwei Plain. Their living room and bedroom were connected and located in the same indoor environment, while all were equipped with traditional stoves for heating in winter (Zhang et al., 2021). The typical house structures of the 18 households in this study are shown in Fig. S2 in the ESM. The heating stoves used in the coal-burning households in this study are uniformly provided by the local government. The lump coal and clean coal were burned on the coal stove in the bedroom, whereas biomass fuels were burned on the stove in the kitchen that is connected to a brick bed in the bedroom. The different types of stoves are indeed used with different fuels (i.e., coal and biomass), but the heating stoves for the same fuel type group are the same. The households use about one ton of coal for heating throughout the winter, and we do not have accurate quantitative statistics on the amount of coal and biomass consumption for each household. The outdoor environments are almost totally unaffected by high-intensity pollution sources such as traffic, garbage dumps, farms, and factories. Only a healthy homemaker was selected for each household for the PE study, who had no history of smoking and was aged between 37 and 57 years (average of 48.2 years). Moreover, all selected homemakers stayed indoors for more than 18 hours per day during the sampling period. Detailed information about the selected homemakers and their households is listed in Table S2 in the ESM.

This study was conducted between 25 December 2019 and 13 January 2020. For each household, the 24-h integrated indoor and PE PM_2.5 samples were collected synchronously from 1200 LST (local standard time, LST = UTC + 8) to 1200 LST the next day consecutively.

A MiniVol PM_2.5 portable air sampler (AirMetrics, Springfield, OR, USA) was used for the indoor sampling. The PM_2.5 was collected on a pre-fired (780°C, 3 h) 47-mm quartz fiber filter (QM/A, Whatman, Little Chalfont, UK) at a flow rate of 5 L min^–1. The indoor sampler was placed in the bedroom approximately 1.5 m from the ground, consistent with the breathing zone of the participants.

The PE PM_2.5 samples were collected by the selected participants using a personal environmental monitoring PM_2.5 sampling device (SKC Company, Chino, CA, USA). During the sampling period, the sampling head was worn near the breathing area of the participants (less than 15 cm apart from the nose) and connected to the pump through a tube. The PE PM_2.5 samples were collected on 37-mm quartz fiber filters (QM/A, Whatman, Little Chalfont, UK) at a flow rate of 10 L min^–1 with the same sampling duration as the indoor sampling.

A total of 36 pairs of indoor and PE PM_2.5 samples were collected in this study. Identical samplers were used to collect the field blanks of indoor and PE sampling. The blank values were used to correct the actual account in the real samples. All sampling pumps used in this study were calibrated using flow meters before and after the sampling. All filter samples were transported to the laboratory in a cooler and stored in a freezer at −4°C.

After equilibrating at a temperature of 20°C–23°C and relative humidity (RH) of 35%–45% for 24 h, the filters were weighed using a Sartorius ME5-F electronic microbalance (Goettingen, Germany) with a sensitivity of ±1 μg to obtain the mass concentrations of PM_2.5. All filters (blanks and samples) were weighed at least twice or until the two weighing results were less than the allowable error (the difference between the two weighing results of the blank and sample filters was 15 μg and 20 μg, respectively).

OC and EC on a 0.5-cm² punch in each filter were analyzed with a Desert Research Institute Model 2001 thermal/optical carbon analyzer (Atmoslytic, Calabasas, CA, USA) using the Interagency Monitoring of Protected Visual Environment (IMPROVE_A) thermal/optical reflectance protocol. The protocol defines OC = OC1 + OC2 + OC3 + OC4 + optically detected pyrolyzed carbon (OP), EC = EC1 + EC2 + EC3 − OP, and total carbon (TC) = OC + EC. The explanation of parameters, detailed analytical method, and detection limits have been described in Cao et al. (2003).

Four anions (F⁻, Cl⁻, ${\rm {SO}}_4^{2-}$, and ${\rm{NO}}_3^{-}$) and five cations (Na⁺, NH₄⁺, K⁺, Mg²⁺, and Ca²⁺) in the water extracted from a one-quarter filter were determined through ion chromatography (Dionex-600, Sunnyvale, CA, USA). The AS11-HC anion column and a CS12A cation column were used for separation. Methodological details and information on quality assurance and quality control are described in Shen et al. (2009a).

A total of 10 elements (K, Fe, Cr, Zn, Mn, Ni, Cu, As, Cd, and Ba) on the filters were quantified using an inductively coupled plasma atomic emission spectrometry (ICP-AES) (ICPE-9000, Shimadzu, Japan). A small portion of the filter (2.5 cm²) was cut into small pieces and transferred to a polytetrafluoroethylene (PTFE) container in a microwave digester for acid treatment (4.5 mL nitric acid, 1.5 mL hydrochloric acid, and 1 mL hydrofluoric acid). After a subsequent heating step to remove acid, the solution was fixed to 10 mL with 2% nitric acid, and then the sample was filtered through a 0.22-μm pore-diameter membrane. The heating program for microwave digestion and detailed experimental procedures of sample pretreatment are described in Wang et al. (2022). The limit of detections (LODs) of K and Fe are 0.20 μg L^–1 and 0.15 μg L^–1, respectively, and the remaining elements are all 0.05 μg L^–1.

ROS activity induced by PM_2.5 was measured using a 2,7-dichlorofluorescein (DCFH). The DCFH reagent was prepared by first dissolving a solid 2,7-dichlorofluorescein-diacetate probe (DCFH-DA) in dimethyl sulfoxide (DMSO) (Solarbio, Beijing, China) to obtain a 100-μM DCFH-DA mixture, which was then mixed with 0.1 N sodium hydroxide (NaOH) in 1:4 ratio and incubated for 30 min. A 2-cm² PM_2.5 filter was dissolved with 8 mL Milli-Q water, then placed in an ultrasonic water bath for 1 h, then shaken for 1 h by using a mechanical shaker, and finally filtered with a 0.22-μm pore-diameter membrane to obtain a PM_2.5 extract. The 100-μL PM_2.5 extract was inoculated into a 96-well plate, then 100 μL DCFH was added in each well, and finally, Hank's balanced salt solution (HBSS) was added to a total of 250 μL. After 10-min incubation at 37°C, the fluorescence intensity of the samples was measured at 485 nm and 530 nm by the microplate reader (Synergy LX, Biotek, VT, USA). The measured light intensity in the sample was accounted with the calibration curve of standard H₂O₂ prepared at concentrations ranging from 50 μM to 1500 μM. The ROS activity units are defined as nM H₂O₂ m^–3. The test was repeated three times with a precision of < 20% and subtracted from the field blank to obtain the corrected ROS activity.

Questionnaires and activity logs were collected from each homemaker who participated in the study. The questionnaire was used to record basic information (such as age, permanent household population, and housing area), living habits (such as passive smoking, drinking, heating, and cooking habits), and living environment (such as roads, factories, and farmland). The hourly activity logs (except bedtime) were used to assess the timing and activities of the participants in each microenvironment. Based on the results of the questionnaires and Table S2, we found that the house structure, housing area, and room temperature (13°C–14°C) experienced by each participant during the sampling period were similar. Details of the questionnaires and time–activity diaries can be referenced in Feng et al. (2021a, 2022) and Xu et al. (2018).

In this study, Excel 2019 (Microsoft, Redmond, WA, USA) was used for data input and sorting, and the results were statistically analyzed with SPSS 19.0 (IBM, Armonk, NY, USA). An independent-sample t-test was used to compare the data between groups. The relationship between PM_2.5 chemical compositions and ROS activity was examined using the Spearmen correlation coefficient. Statistically significant differences were indicated by P < 0.05 and P < 0.01, respectively.

The mortality data in Tongchuan in 2018 was collected from the death cause surveillance system of Shaanxi Province (Chinese Center for Disease Control and Prevention, 2019), and the female population in Tongchuan in 2018 was retrieved from the Shaanxi Statistical Yearbook (Shaanxi Bureau of Statistics and Survey Office of the National Bureau of Statistics in Shaanxi, 2019). The attributable deaths (including all-cause, cardiovascular, and respiratory disease causes) relevant to PM_2.5 exposure in the lump coal, biomass, and clean coal groups were estimated by the following equations:

In Eqs. (1) and (2), ΔY is the number of deaths prevented by the clean coal substitution in Tongchuan and Y₀ is the three types of disease deaths per 100 000 population. In this study, Y₀ is 616.77 for total disease deaths (all-cause deaths), 151.35 for cardiovascular, and 39.62 for respiratory disease deaths in Shaanxi Province in 2018 (Chinese Center for Disease Control and Prevention, 2019); β is derived from the relative risk (RR) associated with a 10 μg m^–3 increment of PM_2.5 concentration based on the exposure-response relationship in the study of Cao et al. (2012), calculated by Eq. (2); ΔPM_2.5 is the difference in PE PM_2.5 concentrations between different fuel groups, such as from the lump coal to the clean coal group (ΔPM_2.5 of 33.3 μg m^–3) and from the biomass to clean coal group (ΔPM_2.5 of 44.7 μg m^–3); POP is the local female population in Tongchuan in 2018, which is 385 503 in the study.

The concentrations of PM_2.5, TC, OC, and EC in this study are listed in Table 1 and Table S3 in the ESM. The mean indoor PM_2.5 mass concentrations were 307 ± 110 μg m^–3, 216 ± 113 μg m^–3, and 173 ± 76.6 μg m^–3 for the lump coal, biomass, and clean coal groups, and the PE PM_2.5 mass levels were 198 ± 77.1 μg m^–3, 210 ± 131 μg m^–3, and 165 ± 61.5 μg m^–3, respectively. The clean coal group exhibits the lowest PM_2.5 mass concentrations in the indoor and PE samples. Compared with the lump coal and biomass groups, the PM_2.5 concentrations in the clean coal group were 43.6% and 20.0% lower in the indoor samples, respectively, and 16.8% and 21.3% lower in the PE samples, respectively. A significant difference is noted between the clean coal and lump coal groups for the indoor samples (P = 0.000), while no significant difference was observed among other pairs. These results indicate that the use of clean coal for heating could significantly reduce the levels of indoor PM_2.5 in winter. The indoor concentrations of the three fuel groups all exceed the daily PM_2.5 guideline of 50 μg m^–3 in the latest Standards for Indoor Air Quality in China (State Administration for Market Regulation and Standardization Administration of the People's Republic of China, 2022). This means that indoor air pollution such as PM_2.5 is still severe in rural households of northern China, even if clean coal was being used. The PE PM_2.5 in the biomass group is the highest among the three groups, while the PE PM_2.5 in the lump coal group is 45% lower than indoors. This may be because lump coal can be maintained for a relatively long burning time after a single fueling. In comparison, biomass and clean coal have a smaller volume which requires frequent fueling to maintain heating demand. Therefore, homemakers who use lump coal would be less frequently exposed to the high direct emission source, even though the indoor concentration would still elevate.

Figure 4.  (a) Reduction of PM_2.5 concentration after clean coal substitution; (b) Estimated reductions of total, cardiovascular, and respiratory mortality after clean coal substitution in Tongchuan.

Numerous studies have reported that the particles emitted from coal and biomass burning are dominated by carbonaceous aerosols (Sun et al., 2017). For the lump coal group, the TC concentrations in indoor and PE samples are 1.8 times and 1.3 times, respectively, those of the clean coal group (Table 1). A similar trend is also observed in the comparison between the clean coal and biomass groups. The clean coal used in this study is mainly made of semi-coke, which minimizes the formation of volatile organic matter (Table S1). The TC constituted the largest and comparable proportions in both indoor and PE PM_2.5 of the three groups, ranging from 41.3% to 47.5%. It could be preliminarily concluded that clean coal exhibits a limited effect on the proportion of carbonaceous aerosol in PM_2.5, even though its absolute value is relatively lower than that of the other two fuel groups.

Figure S3 in the ESM illustrates the contributions of the eight thermal carbon fractions to TC in the indoor and PE PM_2.5 samples. OC is dominated by OC1 and OC3 in all three fuel groups. EC1 is the largest component of EC for all samples, with a mass concentration of approximately 30 times EC2 and EC3. The high contribution of EC1 is attributable to the typically low combustion temperature used in residential facilities, while EC2 and EC3 are mostly formed in high-temperature combustion equipment, such as diesel engines (Sun et al., 2017; Zhang et al., 2020). In addition, among the three fuel groups, the high proportions of OP and EC1 (EC1 + OP) in TC are indicators of incomplete combustion (Zhang et al., 2020). The highest proportion of EC1 + OP in TC (62.8%) is seen in the indoor samples of the biomass group, in comparison with 56.7% and 53.8% for the lump coal and clean coal, respectively. This could be explained by the incomplete combustion due to the large volume of biomass and uneven heat distribution over the stove at the low heating temperature (Hong et al., 2017).

The total water-soluble inorganic ions accounted for a large proportion of PM_2.5 mass, with an average value of 23% (in a range of 20%–28%) for the three fuel groups. The proportions are consistent with those reported in a previous study (Sillapapiromsuk et al., 2013). ${\rm{SO}}_4^{2-}$ dominates the total quantified inorganic ions, accounting for 32%, 25%, and 20% in the indoor samples, and 30%, 24%, and 22% in the PE samples of the lump coal, biomass, and clean coal groups, respectively (Fig. 2). ${\rm{SO}}_4^{2-}$ is mainly produced from SO₂ emitted as part of the exothermic oxidation process of coal combustion (Li et al., 2021b). Its concentrations and proportions in the lump coal group are significantly higher than those of the indoor (P < 0.001) and PE (P = 0.001) samples of the clean coal group. Moreover, the proportion of ${\rm{SO}}_4^{2-}$ in PE samples of the clean coal group (22%) is slightly lower than the biomass group (24%). Both results represent that clean coal, in which a sulfur fixation agent was added, emitted less inorganic sulfur.

Figure 2.  Water-soluble inorganic ion distribution in indoor and PE PM_2.5 of the three fuel groups.

Ions K⁺ and Cl⁻ are the two majors emitted in biomass combustion (Shen et al., 2009b). The proportion of K⁺ in the indoor samples is 5% and 2% higher than that of the lump coal and clean coal groups, respectively, while the proportion of Cl⁻ is 7% and 5% higher than that of the lump coal and clean coal groups, respectively. For the PE samples, the proportions of K⁺ and Cl⁻ in the biomass group are 10% and 13% higher than those of the lump coal group, and 8% and 13% higher than those of the clean coal group, respectively. Higher K⁺ and Cl⁻ proportions in the biomass group are ascribed to their enrichment from herbaceous plants (Lindberg et al., 2016). It should be noted that the potential influence of soil dust or marine sources on these two ions can be ignored in the studied area, which is thousands of kilometers away from the deserts and ocean.

Figure S4 in the ESM compares the measured elemental compositions of PM_2.5 of the three fuel groups. K is the most dominant element in all samples. In between, the biomass group shows the highest K concentrations in both indoor and PE samples, 5%–20% higher than those of the lump coal and clean coal groups. K is widely used as a tracer for biomass burning (Zíková et al., 2016). Moreover, the concentrations of Fe in the PE samples of the lump coal, biomass, and clean coal groups are 3.0 times, 1.3 times, and 2.7 times their corresponding indoor levels, respectively, which is attributed to the exposures to local crustal origins, such as windblown dust and re-suspension of road dust (Xu et al., 2021). The outdoor physical activities of the subjects could lead to a relatively higher PE concentration of Fe.

Indoor and PE concentrations of As in the lump coal and clean coal groups are both higher than in the biomass group. As is relatively light and exists as fly ash in the gas phase in coal combustion, leading to easy adsorption on the surfaces of particles (Duan et al., 2012). The patterns of Zn and Cd in the indoor and PE samples among the three groups are similar, with the highest levels in the lump coal group, followed by the clean coal and biomass groups. Compared with that in the lump coal group, the concentration of Ba in the clean coal group is 39.5% lower in the indoor samples (P < 0.05) and 25.9% lower in the PE samples (P > 0.05), indicating that household heating energy transition can effectively reduce the level of Ba in PM_2.5.

ROS activities of the indoor and PE PM_2.5 are shown in Table 2. The values provide an overview of the oxidation potential resulting from PM_2.5 emitted from different household solid fuel combustions. The differences of ROS activities are obvious among the three fuel groups. For the clean coal group, the indoor ROS concentration (0.79 ± 0.16 nM H₂O₂ m^–3) is 41% lower and 10% lower than the lump coal (1.33 ± 0.70 nM H₂O₂ m^–3) (P < 0.05) and biomass (0.88 ± 0.22 nM H₂O₂ m^–3) (P > 0.05) groups, respectively. The results demonstrate a lower potential of oxidative damage and its associated health risk on the human respiratory system in the Fenwei Plain for clean coal compared to the other two solid-fuel groups. As shown in Table 1, the indoor and PE PM_2.5 concentration ratios (1.0–1.6) are close between each fuel group. However, greater ROS activity distinctions are observed between the indoor and PE values, with ratios of 3.3–4.4. It is reasonable to speculate that chemical components play major roles in the ROS, instead of the mass concentrations. Therefore, it is necessary to deduce their relationship in the following section.

The relationships between chemical composition and ROS in the indoor and PE samples are elaborated using a Spearman rank correlation method (Fig. 3). The carbonaceous aerosol and its sub-fractions mostly exhibit weak correlations (no statistical significance) with ROS activity for both types of fuel groups. However, OC4 is strongly correlated with ROS in the PE lump coal (R = 0.89, P < 0.01) and PE clean coal (R = 0.77, P < 0.01) groups, as well as being moderately correlated in the biomass group (R = 0.41, P < 0.05). According to a previous study, OC4 is dominated by secondary organic aerosol formation with low volatile and high molecular weight compounds, which can provide a synergistic ROS generation effect between multiple metals and organic compounds (Yu et al., 2018).

Figure 3.  Correlations between chemical components in PM_2.5 and ROS in different groups (*P < 0.05; **P < 0.01).

Ion F⁻ and ROS activity exhibit strong correlations in both the lump coal (R = 0.72, P < 0.01) and clean coal (R = 0.75, P < 0.01) groups, possibly attributed to the high emissions of F⁻ from coal combustion (Chen et al., 2014). Moreover, the highest correlation between Na⁺, an indicator of solid fuel combustion (Liu et al., 2021), and ROS is also found in the lump coal group (R = 0.71, P < 0.01). Relatively weak correlations between other ions and ROS were seen in this study, resulting from the water-soluble ions not directly showing an effect on the formation of ROS. But by increasing the solubility and oxidative potential of metals and metalloids, they can generate free radicals (Fang et al., 2017).

Elements Cu and Mn are both positively correlated with ROS in all groups: lump coal (R = 0.92, P < 0.01 for Cu and R = 0.92, P < 0.01 for Mn), biomass (R = 0.69, P < 0.01 for Cu and R = 0.70, P < 0.01 for Mn), and clean coal (R = 0.81, P < 0.01 for Cu and R = 0.61, P < 0.01 for Mn). Transition metals such as Fe, Cu, and Mn have incomplete inner valence d-sub-shell, and their electrons of internal 3d orbitals can be available for chemical bonding (Bondy, 2016). Therefore, these metals form a variety of valence states under physiological conditions, and the flux between these valences constitutes the basis for their ROS production (Shuster-Meiseles et al., 2016). Fe and Cu have been theorized to serve as catalysts in Fenton reactions by acting as electron mediators in the oxidation/reduction of hydrogen peroxide to hydroxyl and hydroperoxyl radicals (Valko et al., 2005). Dissimilar to Fe and Cu, Mn is the most stable in its lower valence form (Mn²⁺), and this may account for its ability to act either in a pro- or anti-oxidant manner (Bondy, 2016). However, this is unable to explain the negative correlations between Fe and ROS in all three fuel groups, even the significant negative correlations in the lump coal and clean coal groups. The negative correlations are probably caused by some unidentified interactions between Fe and organic compounds, such as the antagonism of metals and quinones (Yu et al., 2018).

The correlation between Ba and ROS is strong in all fuel groups, in a descending order of lump coal (R = 0.95, P < 0.01), clean coal (R = 0.84, P < 0.01), and biomass (R = 0.82, P < 0.01) groups. Even though a full toxicological mechanism of Ba is still unclear, it might be associated with oxidative stress induction and ROS production (Elwej et al., 2016). As and Cd are carcinogens with extensive toxic effects, and their toxicities may be largely due to their abilities in coupling with sulfhydryl groups, consequently causing oxidative stress (Xi et al., 2010). In this study, the correlations between both As and Cd, and ROS are strong, with P < 0.01.

It is discovered that the correlations between chemical compositions of PM_2.5 and ROS are quite different among the three fuel groups. Characteristic components including As, Cd, Ba, Ni, Cu, Mn, OC4, Zn, Cr, F^-, and Na⁺ are strongly correlated with ROS generation in the lump coal group. Elements As, Cd, Ni, Ba, Zn, and Mn are strongly correlated to ROS in the biomass group. Although K⁺ represents biomass burning to a particular extent, it shows a weak correlation with the ROS activity, possibly implying that its contribution to health hazards in biomass burning is insignificant. In the clean coal group, Ni, Cd, As, Ba, Cu, OC4, Zn, F⁻, and Mn are strongly associated with ROS levels. However, although most carbonaceous aerosols exhibit negative correlations with ROS, positive correlations are seen in the lump coal group. This means that chemical species in PM_2.5 that can indirectly stimulate ROS generation, such as organic compounds, showing the restricted oxidative potential or chemical components affect particle-induced ROS production through combined synergistic or antagonistic effects. These two mechanisms have been proven and are completely contradictory (Yu et al., 2018), which makes it tougher to know the mechanism of particle-induced ROS generation in the atmosphere.

The oxidative stress response is considered the most plausible mechanism that results in adverse effects of PM_2.5 on human health such as aging and disease (Lee et al., 2014). Compared with fixed indoor sampling, portable PE sampling can better reflect a person’s real inhalation and thus lead to more accurate results in the health assessment. According to the results shown in Table 2, for the PE samples, the ROS activity in the clean coal group is 20% lower than that in the lump coal group (P > 0.05), indicating that the use of clean coal offers lower human health risks. The levels of ROS induced by the PE are comparable between the clean coal and biomass groups. The level of oxidative stress induced by the emissions from biomass burning was found to be relatively low in this study.

The health benefits of the use of clean coal can also be assessed with the results shown in a previous epidemiological study. Owing to a lack of updated PM_2.5 epidemiological findings on the local populations in the rural Fenwei Plain, we used the results obtained from all areas (including urban and rural regions) for calculation (Cao et al., 2012). Cao et al. (2012) reported that a 10 μg m^–3 increase in 1-day lagged PM_2.5 is associated with a 0.2% (95% confidence interval [CI]: 0.1%, 0.3%), 0.3% (0.1%, 0.4%), and 0.4% (0.2%, 0.6%) increase in total, cardiovascular, and respiratory mortality, respectively. In addition, they found an almost linear relationship without a threshold for the three kinds of mortalities based on the exposure–response curve of PM_2.5 obtained by using the Poisson regression model. This indicates that their results are appropriate for assessing the PM_2.5 effect on mortality in this study.

According to the data shown in Table 1, the concentrations of PE PM_2.5 in the lump coal and biomass groups are 33.3 μg m^–3 (95% CI: 19.4, 47.3) and 44.7 μg m^–3 (95% CI: 0.17, 89.4), respectively, which are much higher than the concentration of the clean coal group (Fig. 4a), while the three mortality types (i.e., total, cardiovascular, and respiratory mortality) decreased by averages of 3.3 times (95% CI: 1.9, 4.7) and 4.5 times (95% CI: 0.01, 8.9) of 10 μg m^–3, respectively, in comparison to the clean group. Figure 4b shows that the use of clean coal in rural households in Tongchuan during the sampling period led to estimated decreases of 0.66% (95% CI: 0.4%, 1.0%), 1.0% (95% CI: 0.6%, 1.4%), and 1.3% (95% CI: 0.8%, 1.9%) in total, cardiovascular, and respiratory mortality, respectively, compared with lump coal when other conditions remained invariant. For the biomass group, the three mortality types after the clean coal transformation decreased 0.89% (95% CI: 0.0%, 1.8%), 1.3% (95% CI: 0.0%, 2.7%), and 1.8% (95% CI: 0.0%, 3.6%), respectively (Fig. 4b). Therefore, the promotion of using clean coal for heating could significantly reduce the mortality potential resulting from PM_2.5 exposure.

Moreover, we further used the previous epidemiological conclusions from meta-analyses and cohort studies (Chen et al., 2017) to consolidate the health benefits of clean energy transformation. Table 3 clearly shows that if the current ordinary solid fuel groups switch to using clean coal, the female deaths from all-cause, cardiovascular, and respiratory disease could be reduced by 16, 6, and 3, respectively, for the lump coal group, and 22, 8, and 3, respectively, for the biomass group in Tongchuan based on 2018 data.

PM_2.5 mass concentration and its components have different adverse effects on health. On one hand, we were evaluating the health benefits of clean coal based on the premise of PM_2.5 concentration. On the other hand, we introduced ROS to evaluate the impact of exposure to PM_2.5 components from different fuel combustions and found that some key chemical components of PM_2.5-bound transition elements, such as Cd, Ni, Zn, and Mn, have a greater impact on health. Future research could be directed at quantifying ROS and its health effects, although this is difficult to do at present.

In this study, the mass concentrations of PM_2.5 and its bounded chemical species in indoor and PE samples were used to evaluate the environmental and health effects of the clean energy transition for household heating in the rural Tongchuan, Fenwei Plain, northwestern China. The PM_2.5 concentration levels of both indoor and PE samples in the clean coal group are lower than those of the lump coal and biomass groups, demonstrating a remarkable PM_2.5 exposure reduction. From the health benefits perspective, the ROS activity of PE PM_2.5 in the clean coal group is 20% lower than that in the lump coal group, indicating that the use of clean coal can reduce the oxidation of exposed PM_2.5. Moreover, after the substitution of traditional solid fuels with clean coal, the total, cardiovascular, and respiratory mortality could be reduced by 0.66%–1.8%.

However, ROS was analyzed by filter-based approach in this study. Due to the reactivity and short lifetime of ROS (e.g., free radicals or some peroxides), the results may be underestimated by such an offline method (Zhang et al., 2022). Overall, the promotion of clean coal as a heating fuel in rural households in Tongchuan can assist to improve the air quality in both indoor and outdoor environments, as well as the health of the residents. However, for a long-term environmental conservation plan, cleaner energies such as electric power, natural gas, and solar energy should gradually be applied to complete the replacement of the use of solid fuels in rural regions of China.

Acknowledgements. This research was supported by the National Natural Science Foundation of China (Grant Nos. 41877376 and 41877404) and the open fund of the State Key Laboratory of Loess and Quaternary Geology, Institute of Earth Environment, Chinese Academy of Sciences (SKLLQG2110).

Electronic supplementary material: Supplementary material is available in the online version of this article at https://doi.org/10.1007/s00376-022-2249-1.