Research ArticleENVIRONMENTAL STUDIES

Residential solid fuel emissions contribute significantly to air pollution and associated health impacts in China

See allHide authors and affiliations

Science Advances  28 Oct 2020:
Vol. 6, no. 44, eaba7621
DOI: 10.1126/sciadv.aba7621

Abstract

Residential contribution to air pollution–associated health impacts is critical, but inadequately addressed because of data gaps. Here, we fully model the effects of residential energy use on emissions, outdoor and indoor PM2.5 concentrations, exposure, and premature deaths using updated energy data. We show that the residential sector contributed only 7.5% of total energy consumption but contributed 27% of primary PM2.5 emissions; 23 and 71% of the outdoor and indoor PM2.5 concentrations, respectively; 68% of PM2.5 exposure; and 67% of PM2.5-induced premature deaths in 2014 in China, with a progressive order of magnitude increase from sources to receptors. Biomass fuels and coal provided similar contributions to health impacts. These findings are particularly true for rural populations, which contribute more to emissions and face higher premature death risks than urban populations. The impacts of both residential and nonresidential emissions are interconnected, and efforts are necessary to simultaneously mitigate both emission types.

INTRODUCTION

There is growing evidence suggesting that indoor combustion emissions have a notable impact on air quality and human health. A previous estimate indicates that in China, ~30% of ambient air pollution–associated premature deaths in 2010 were attributable to residential emissions (1). A recent study found that in 2012, rural residential emissions contributed to nearly 20% of ambient PM2.5 (e.g., particulate matter with aerodynamic sizes less than 2.5 μm) exposure in China (2). While the importance of residential emissions has been recognized, results from these studies vary greatly, largely because of a lack of detailed and reliable data for residential energy use (2).

Moreover, many results are based on ambient PM2.5 concentrations (1, 2), thus preventing a full understanding of this issue when indoor air exposures are excluded because people spend most of their time indoors (35). According to the latest Global Burden of Disease (GBD) report, 24% of PM2.5-associated premature deaths in China were attributed to household sources/exposures in 2017 (6). Recently, Chen et al. (7) included household heating emissions in their model and obtained a higher value. Recently, detailed firsthand data for energy types and activities, including stove locations, which were obtained from a nationwide residential energy survey (8, 9), have provided a unique opportunity to fully address this issue.

Here, we present the results of sequential modeling to track the contributions of residential energy use to emissions, indoor and outdoor PM2.5 concentrations, PM2.5 exposures, and PM2.5-associated premature deaths. The residential effects are categorized by energy type (e.g., coal, biomass, and clean energy), activity (e.g., cooking and heating), area (e.g., urban and rural), and location (e.g., indoor and outdoor). The term “clean energy” is defined as the natural gas, LPG (liquefied petroleum gas), biogas, and electricity used in residences, which emit relatively lower pollutant levels in households compared to solid fuels. All nonresidential sources are grouped together for comparison. A number of emission reduction scenarios are modeled to evaluate the effectiveness of the mitigation options. The detailed methods are presented in Materials and Methods.

RESULTS

Amplified residential contributions

The contributions of nonresidential and residential activities to energy use, primary PM2.5 emissions, ambient and indoor PM2.5 concentrations, exposures, and premature deaths (in order) are summarized in Fig. 1. For all factors, the total/average values (numbers on the left), detailed residential energy types, activities, and indoor and outdoor PM2.5 concentrations are presented separately. SO2 (sulfur dioxide) emissions are also shown as a representative secondary aerosol precursor in addition to primary PM2.5. Urban and rural populations are addressed individually, and fluxes along the path from energy usage to premature deaths are marked. As shown, the contributions of residential solid fuels are amplified from energy consumption to premature deaths. In 2014, direct residential energy accounted for only 7.5% of total energy consumption in China in terms of joules. Solid fuels accounted for 88% of the total residential energy use. In terms of the percentage of use, solid fuels accounted for 41% of cooking and 82% of heating energy usage in rural China in 2014 (9).

Fig. 1 Relative contributions of residential and nonresidential energy consumption to the emissions of primary PM2.5 and SO2 (as a representative secondary aerosol precursor), indoor/outdoor PM2.5 concentrations, population exposure to PM2.5, and premature deaths associated with PM2.5 exposure.

The emissions of both primary PM2.5 and SO2 are shown. Residential energy sources were further categorized into coal, biomass fuels, and clean energy for cooking and heating. Urban and rural usages are separated as both sources and receptors. The quantified contributions from one step to the next are shown as numbered arrows. Total quantities are shown in the centers of the pie charts. The relative contributions by the residential sector are presented for the individual steps on the right side.

Because the emission factors (EFs; e.g., quantities of pollutants emitted per unit of energy consumed) of primary PM2.5 for residential stoves are orders of magnitude higher than those for other activities (10), a 7.5% residential contribution to energy consumption led to a 27% contribution to primary PM2.5 emissions. Despite the fact that the residential sector contributed relatively low emissions of secondary aerosol precursors such as SO2 (7.5%), the contributions from this sector to PM2.5 concentrations in the air were high. In terms of annual mean values, 23 and 71% of PM2.5 in ambient and indoor air were from this sector, respectively. These high contributions to indoor air pollution are due to emissions coming directly from stoves (11). The results of many field surveys have confirmed this finding and have revealed that PM2.5 concentrations in households using solid fuels can be very high (7, 12). For households using clean energy, indoor PM2.5 is mainly originated from the ambient air and thus leads to positive correlations between indoor and ambient PM2.5 concentrations (13, 14). By classifying the households into two categories (e.g., using solid or clean energy) and incorporating detailed information on stove locations (Materials and Methods), the average indoor PM2.5 concentration in Chinese households (73 ± 34 μg/m3) was determined to be three times that for outdoor air (22 ± 16 μg/m3).

With both the quantification of ambient and indoor air PM2.5 concentrations, the average exposure of the entire population to PM2.5 was calculated as 69 ± 30 μg/m3, which led to 1,150,000 (630,000 to 1,770,000 by using an uncertainty of 95%) premature deaths in 2014. Residential emissions accounted for 68% of the exposure and 67% of premature deaths. This value is much higher than the 24% (270,000 of 1,120,000) estimated by the latest GBD study for household contributions (6) mainly because updated residential energy data, including heating data, were used. Chen et al. (7) used a similar approach but without stove locations or outdoor-to-indoor penetration and obtained a higher estimate of 670,000 to 930,000 (by using an uncertainty of 95%) premature deaths in rural China in 2010. This result is significantly higher than the 43% value for total premature deaths derived by Zhao et al. (15), which is likely due to the different methods (e.g., perturbation-normalization approach versus elimination method) used to calculate premature deaths.

Indoor exposure domination

Because people spend the most time indoors, indoor exposure dominates the total PM2.5-associated premature deaths (91%). This trend was particularly true for rural areas (94%) where PM2.5 levels in the indoor air (95 ± 34 μg/m3) were significantly higher than those in urban indoor air (58 ± 23 μg/m3) because of the strong dependence of rural residents on solid fuels. For total indoor PM2.5 levels, 85 and 55% were from residential emissions in rural and urban areas, respectively. In contrast, ambient PM2.5, which is mainly from nonresidential sources (77%), was higher in urban (45 ± 19 μg/m3) than in rural areas (22 ± 15 μg/m3). The differences between rural and urban areas are shown in Fig. 2 as cumulative frequency distributions of the exposures associated with residential (solid lines) and all (dashed lines) emissions. The percentage of the population with annual mean exposure originating from only residential sources exceeded the Chinese national ambient air standard (16) of 35 μg/m3 by 35 and 84% in urban and rural areas, respectively, which is also the first World Health Organization (WHO) interim target (17). Although these percentages are significantly higher (81 and 95%) if all sources are considered, residential sources contributed to a notable fraction of the total.

Fig. 2 Cumulative distributions of the annual mean population exposures to PM2.5 originating from residential sources and all sources in urban and rural China in 2014.

The WHO air quality guideline (AQG) and interim targets 1 to 3 (IT-1, IT-2, and IT-3) (17), which are also applicable to indoor air, are shown as vertical lines. IT-1 (35 μg/m3) is the same as the Chinese national ambient air standard (16).

All-source PM2.5-associated premature deaths totaled 640,000 (370,000 to 980,000 by using an uncertainty of 95%) and 510,000 (260,000 to 790,000 by using an uncertainty of 95%) in rural and urban areas, respectively. Because the urban population (0.736 billion) is larger than the rural population (0.618 billion), the overall risk, which is defined as the probability of premature death, for the rural population (1.0 × 10−3) was 45% higher than that of the urban population (6.9 × 10−4). If only residential emissions are considered, the risks in the rural and urban areas are 8.5 × 10−4 and 3.4 × 10−4, respectively, and thus show a 60% difference. Although residential emissions are localized, cross-influences between rural and urban areas do occur because of atmospheric transport. The contributions of rural (5.2 ± 3.9 μg/m3) and urban (5.0 ± 6.5 μg/m3) residential emissions to urban ambient PM2.5 levels were nearly the same, which can be explained by the fact that the majority of cities are surrounded by rural villages with high population densities. Therefore, any effort to reduce emissions from rural residential sources would improve air quality in both rural and urban areas.

Premature deaths associated with residential emissions vary extensively in space (Fig. 3) for both urban (the symbols have areas proportional to the population) and rural (shaded background) areas. The most-influenced areas were the Sichuan Basin and the Northern Plain; by contrast, the risks associated with residential emissions in the southeastern coastal provinces were much lower.

Fig. 3 Spatial distributions of premature deaths per square kilometer associated with residential emissions of both primary PM2.5 and secondary PM2.5 precursors at the municipality level.

The number of premature deaths per square kilometer in rural areas is shown with shaded blue backgrounds, and that in urban areas is shown as shaded circles; the areas of the symbols are proportional to the urban populations.

High contributions of biomass fuels and cooking

Among the 770,000 (430,000 to 1,170,000 by using an uncertainty of 95%) total premature deaths associated with residential solid fuels in China, coal (52%) and biomass fuels (48%) contributed equally if model uncertainties are considered. Notable spatial variations in the relative contributions of the two fuel categories are shown in Fig. 4 (A and B). The patterns for coal were similar between urban and rural areas and showed significant correlations (P < 0.01). Relatively high percentages of coal use occurred in areas with rich coal reserves, such as Shanxi, Guizhou, Inner Mongolia, and Xinjiang. The highest biomass fuel contribution in Tibet is due to the extensive use of wood and dung. Other regions with relatively large contributions of biomass fuels are those regions with heavy forest coverage (e.g., northeast and southwest) or agricultural production (e.g., south and northeast).

Fig. 4 Spatial distributions of the relative contributions of residential emissions to total PM2.5 exposures.

Relative contributions of residential coal (A), biomass fuels (B), cooking (C), and heating (D) to total PM2.5 exposures are shown, respectively. Both urban (symbols) and rural (shaded background) areas are shown. Rural areas are shown at the municipality scale, and urban areas are shown as individual cities, for which the symbol areas are proportional to the urban populations. Biomass fuels are not available in urban areas.

Although a rapid transition to clean cooking has occurred in rural China (8), solid fuels were still the major cooking energy source for rural (3330 PJ) households in 2014. Consequently, the contribution of cooking to total PM2.5 originating from residential emissions in indoor air is substantial (37%) but is lower than that from heating (63%). The contributions of cooking and heating to PM2.5 exposure also vary extensively in space (Fig. 4, C and D). Note that cooking and heating are not always completely separate activities in northern China in the winter, which can lead to more uncertainty in the data analysis.

Simultaneous mitigation approach

Because the residential sector contributes approximately two-thirds of the total premature deaths associated with PM2.5, efforts to reduce residential emissions can be effective. Moreover, for households that shift from solid fuels to clean fuels, indoor air will still be affected by penetration of PM2.5 from outside, which largely comes from nonresidential emissions, thus causing the situation to be complicated. The results of reductions in emissions modeling from both sources to various degrees (Materials and Methods) suggest that exposure is not simply proportional to total emissions (fig. S1, A and B). For example, if the residential emissions of PM2.5 are reduced by 20, 60, or 100% and other emissions remain constant, the contributions of ambient PM2.5 to rural indoor air would nonlinearly increase from 17 to 19, 22, or 24 μg/m3, respectively.

Such a relationship is illustrated in fig. S2A, which shows the total exposure as a function of both the residential and nonresidential emission reduction rates. Similar dependencies of the outdoor and indoor PM2.5 concentrations on reduction rates are shown in fig. S2 (B and C). These results clearly demonstrate that to achieve effective overall exposure reductions, both residential and nonresidential emissions should be mitigated simultaneously with an emphasis on the residential sector. Technically, the optimal path to any location on the plane is that which is perpendicular to the contoured isoclines, such as the blue arrow starting from (0%, 0%) in 2014. Therefore, the residential/nonresidential emission reduction ratio of approximately 2.2:1 follows such a route. However, if the outdoor PM2.5 concentrations (e.g., ambient air quality) are targeted, the path is very different and has a heavy emphasis on nonresidential emissions (0.24:1). A detailed analysis of residential energy types, activities, and areas was conducted, and the results are shown in fig. S2 (D to F, respectively). Because coal and biomass fuels contributed nearly equally to PM2.5 exposure, the best way is to co-reduce these emissions at similar rates (e.g., 1.1:1). Compared to cooking, heating should be emphasized (e.g., 1.5:1). Last, more efforts for reducing emissions in rural areas (e.g., 2.4:1) can lead to better results. Note that this is only a technical approach and a realistic policy strategy needs to include socioeconomic aspects, which are often more important. Therefore, a full strategy should also be based on cost-benefit and social feasibility analyses. In addition, the calculations are based on exposures instead of health consequences, and the nonlinear relationship between exposure and health impacts can lead to additional uncertainty.

DISCUSSION

Residential energy usage can cause air pollutant emissions, indoor and ambient air pollution, population exposure, and adverse health impacts associated with PM2.5. Although some individual steps of the process leading from energy use to health consequences have been investigated previously, in this study, a systematic approach connecting energy, emissions, air pollution, and health outcomes with quantification of the contributions of various residential energy types provides consistency in methodology and allows uncertainty analysis. We report that the magnification of the relative contributions of residential solid fuels was quantified and show that 7.5% of energy use in residential sector can lead to 67% of premature deaths. The methodology used has the potential to be applied to assess environmental, health, and climate influences of energy use in other sectors.

Although biomass fuels are extensively used in the rural residential sector, adverse impacts on air quality and health associated with biomass use have been overlooked in comparison with coal (18). One reason is that the use of biomass fuels is relatively evenly distributed in space and across seasons, whereas coal heating is often associated with severe pollution episodes in the heating season in northern China (19). We also determined the relative contributions of indoor and outdoor exposure, cooking and heating, and coal and biomass fuels. Our results indicate that biomass fuels contributed 32% of the overall premature deaths. Similarly, we also revealed the similar importance of cooking and heating in terms of health impacts, whereas the impacts of cooking are not covered by existing mitigation plans (20, 21).

By including both indoor and outdoor exposures, we also distinguished the relative contributions of the two different paths on health impacts and found that indoor exposure is more important than outdoor exposure. By using a systematic approach with an indoor-outdoor exchange incorporated in the modeling, we demonstrated that the relative contribution of indoor exposure (91%) to total premature deaths was previously underestimated. We also found that compared with urban populations, rural populations contributed more to the emissions but were also affected more by the pollution generated.

MATERIALS AND METHODS

Energy and emissions

The study year was 2014, which happened to be the turning point when a series of air pollution mitigation actions were launched and the ambient PM2.5 concentration began to decline in China. Rural residential energy data from 2014 were previously derived (9) by extrapolating the results of a nationwide survey conducted in 2012 (8). Detailed energy types, including coal (e.g., bituminous coal and honeycomb briquette), charcoal, wood (e.g., fuel wood and brush wood), corncobs, crop residues (excluding corncobs), LPG, biogas, and electricity (for heaters, kettles, rice cookers, and induction stoves), were available (9). Urban residential energy data compiled by Shen et al. (22) were adopted. Detailed residential energy types for various areas (e.g., rural and urban) and activities (e.g., cooking and heating) are presented in table S1.

The emissions of primary PM2.5 and PM10 (e.g., particulate matter with an aerodynamic size less than 10 μm), black carbon, organic carbon, CO (carbon monoxide), NH3 (ammonia), SO2, and NOx (nitrogen oxides) from the residential sector in 2014 were calculated by multiplying the quantities of energy consumed by EFs. Values of EFs were from a database used to compile the Peking University (PKU) emission database (23). Emissions from all other sectors were directly extracted from the PKU emission database and distinguished between urban and rural areas based on an urban mask developed by Shen et al. (22). The emissions of non-combustion NH3 and non-methane volatile organic compounds were from EDGAR (Emission Database for Global Atmospheric Research) (24) and HTAP (Hemispheric Transport of Air Pollution) (25).

Ambient and indoor PM2.5

The weather research and forecasting model coupled with chemistry (WRF/Chem) version 3.5 (26) was used with a horizontal resolution of 50 km by 50 km and with 5-min time steps to model PM2.5 in the ambient air. The model domain covered mainland China and surrounding areas (e.g., 13°N to 56°N, 67°E to 143°E). The meteorological field data were obtained from the National Centers for Environmental Prediction Final Operational Global Analysis data (27). WRF-modeled meteorological fields were evaluated using observations from the China Earth International Exchange Stations (28) (fig. S3). The modeled PM2.5 concentrations were downscaled using a Gaussian downscaling method (22) based on the PM2.5 emission inventory and wind fields. The downscaled modeled PM2.5 concentrations were evaluated using PM2.5 observations collected at 708 sites in 126 Chinese municipalities (figs. S4 and S5) (29) and PM2.5 observations from sites at the American embassy and consulates (fig. S6) (30). To distinguish the contributions of various sources, a perturbation method was applied by running the model repeatedly with a 20% emission reduction for an individual source in each run as well as a base run using the total emissions from all sources (31, 32). The outputs of the individual runs were normalized to allow the sums of these runs to be equal to those from the base run (31, 32).

The PM2.5 concentrations in indoor air were calculated by categorizing households into those with solid fuel use and those without such use based on the time-sharing fractions of solid fuel consumption and stove locations (8). For the former, the annual mean indoor PM2.5 concentrations were derived from the literature for coal, biomass fuels, and clean energy users in kitchens, living rooms, and other indoor microenvironments for the heating and nonheating seasons (7). For households without solid fuel use, the indoor PM2.5 concentrations were obtained based on the ambient air concentrations and infiltration factors derived on the basis of the method of Xiang et al. (33).

Exposure and premature deaths

Population exposures to PM2.5 were calculated for each grid cell based on the modeled ambient and indoor PM2.5 concentrations, population densities, and average time people spent indoors and outdoors in the heating and nonheating seasons in six regions of China, namely, northern, eastern, southern, northwestern, northeastern, and southwestern China (35, 34); in addition, stove locations were obtained from the same database that was used for residential energy use in 2012 (8). The entire population was classified into four age groups, namely, 0 to 4, 5 to 14, 15 to 64, and >64 years for males and females. The premature deaths associated with PM2.5 exposure were derived for five diseases, namely, acute lower respiratory infections for children, lung cancer, ischemic heart disease, cerebrovascular disease (stroke), and chronic obstructive pulmonary disease, which were based on the latest hazard ratios and the integrated exposure-response (IER) functions established by Cohen et al. (35). Background disease burdens at the provincial level were based on a GBD study (36) and were scaled to the latest GBD background disease burdens in China (37); these were categorized into urban and rural populations (38). A detailed calculation is presented in Supplementary Materials and Methods. To distinguish the contributions of various energy-activity combinations, which are not linearly additive, both the exposures and health consequences, as represented by premature deaths, were quantified separately; the results were then normalized to allow the sum of all individual impacts to equal the total impact. This “normalized marginal” approach can appropriately isolate contributions of the studied source/sector to the others (31, 32, 39); however, its results may be different from those by using other methods (e.g., zeroed-out, source tracking), and the uncertainties in the following analysis should be addressed in the future. Emissions from household electricity use do not occur onsite and were not covered in this study.

To evaluate the benefits of emission reductions associated with residential or nonresidential sources and their interactions, a set of 36 simulations was conducted and the two sources were reduced by 0, 20, 40, 60, 80, and 100% individually. The changes in population exposure, outdoor and indoor PM2.5 concentrations, and penetration of PM2.5 from the ambient air to indoor environment were compared among the scenarios. Similarly, the consequences of residential emission reductions were compared for coal and biomass fuels, cooking and heating, and urban and rural areas. For each comparison, 36 scenarios were simulated with the emissions from the two sources set to 0, 20, 40, 60, 80, and 100% individually. The changes in population exposures were compared to identify the optimal reduction pathways.

Uncertainty analysis

Monte Carlo simulations were performed 1000 times to address model uncertainties from emission calculations to premature deaths. For emissions, the coefficients of variation (CVs) for activity intensities and EFs (log-transformed) were from the PKU inventories (23). For the time people spent indoors, CVs of 5% were used on the basis of the method of Chen et al. (7). For the infiltration factors in indoor/outdoor air exchange and the amount of time windows are opened or closed, the CVs were from the Ministry of Environmental Protection (3). For indoor PM2.5 concentrations in households using solid fuels, the CVs of log-transformed PM2.5 concentrations were derived by Chen et al. (7) on the basis of 1821 observations collected from the literature, and the mean and SD of the CVs for various fuels-microenvironment-season combinations were 14 ± 16%, suggesting that the overall uncertainty of the calculated indoor PM2.5 concentrations and, consequently, exposure and premature deaths was likely from these data sources. More measurements are recommended in the future to improve the estimation. Because Monte Carlo simulations could not be conducted for atmospheric chemical transport modeling because of the high computational load, CVs of 10% were assumed for gridded ambient air PM2.5 (40). The distributions of parameters in the dose-response curves of premature death models were from the IER (35).

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/6/44/eaba7621/DC1

https://creativecommons.org/licenses/by-nc/4.0/

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

REFERENCES AND NOTES

Acknowledgments: Funding: This work is funded by the National Natural Science Foundation of China (grants 41830641, 41821005, 41629101, and 41922057), the Chinese Academy of Science (XDA23010100), and the 111 program (B14001). Author contributions: S.T. designed the study. X.Y. performed the simulations with help from H.S., J.M., and J.L. in model evaluation and validation and Y.C. in the health assessment. X.Y., S.T., G.S., H.S., Y.C., and W.M. performed data analysis with help from Xuejun Wang, Xilong Wang, J.H., Y.W., Q.Z., Y.R., H.X., and W.D. Writing was led by S.T. and X.Y. with substantial inputs from H.C., H.S., G.S., Y.C., and B.L. All authors participated in the interpretation of results and improvement of the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors. The basic emissions data used in the study can be downloaded freely from inventory.pku.edu.cn.

Stay Connected to Science Advances

Navigate This Article