Impacts of stove use patterns and outdoor air quality on household air pollution and cardiovascular mortality in southwestern China

Graydon Snidera,b, Ellison Carterc, Sierra Clarka,b, Joy (Tzu Wei) Tsenga, Xudong Yangd, Majid Ezzatie, James J. Schauerf,g, Christine Wiedinmyerh, Jill Baumgartnera,b,i

aInstitute for Health and Social Policy, McGill University, Montréal, QC, CanadabDepartment of Epidemiology, Biostatistics & Occupational Health, McGill University, Montréal, QC, CanadacCivil and Environmental Engineering, Colorado State University, Fort Collins, Colorado, USAdDepartment of Building Science, Tsinghua University, Beijing, ChinaeMRC-PHE Centre for Environment and Health, School of Public Health, Imperial College London, London, UKfEnvironmental Chemistry and Technology Program, University of Wisconsin, Madison, WI, USA gWisconsin State Laboratory of Hygiene, University of Wisconsin, Madison, WI, USAhNational Center for Atmospheric Research, Boulder, Colorado, USAiInstitute on the Environment, University of Minnesota, St. Paul, Minnesota, USA

Journal: Environment InternationalAddress correspondence to Jill Baumgartner at the Department of Epidemiology, Biostatistics & Occupational Health and Institute for Health and Social Policy, McGill University, Montréal, QC, Canada. Phone: 514-398-6688 Email: jill.baumgartner@mcgill.caThe authors declare they have no actual or potential competing financial interests regarding this publication.

ABSTRACT

Background: Decades of intervention programs that replaced traditional biomass stoves with cleaner-burning technologies have failed to meet the World Health Organization (WHO) interim indoor air quality target of 35-μg m-3 for PM2.5. Many attribute these results to continued use of biomass stoves and poor outdoor air quality, though the relative impacts of these factors have not been empirically quantified.

Methods: We measured 496 days of real-time stove use concurrently with outdoor and indoor air pollution (PM2.5) in 150 rural households in Sichuan, China. The impacts of stove use patterns and outdoor air quality on indoor PM2.5 were quantified. We also estimated the potential avoided cardiovascular mortality in southwestern China associated with transition from traditional to clean fuel stoves using established exposure-response relationships.

Results: Mean daily indoor PM2.5 was highest in homes using both wood and clean fuel stoves (122 μg m -3), followed by exclusive use of wood stoves (106 μg m -3) and clean fuel stoves (semi-gasifiers: 65 μg m -3; gas or electric: 55 μg m-3). Wood stoves emitted proportionally higher indoor PM2.5 during ignition, and longer use was not associated with higher indoor PM2.5. Only 24% of days with exclusive use of clean fuel stoves met the WHO’s indoor air quality target, though this fraction rose to 73% after subtracting the outdoor PM 2.5

contribution. Reduced PM2.5 exposure through exclusive use of gas or electric stoves is estimated to prevent 48,000 yearly premature deaths in southwestern China, with greater reductions if local outdoor PM 2.5 is also reduced.

Conclusions: Clean stove and fuel interventions are not likely to reduce indoor PM2.5 to the WHO target unless their use is exclusive and outdoor air pollution is sufficiently low, but may still offer some cardiovascular benefits.

Keywords: biomass, cardiovascular mortality, cookstoves, liquefied petroleum gas, PM2.5, semi-gasifier

1

Highlights:

● Indoor PM2.5 in homes with exclusive use of semi-gasifier, electric or LPG stoves was 36-55% lower than in homes using wood chimney stoves

● Households using wood chimney and clean fuel stoves together had higher indoor PM 2.5 than those only using of wood chimney stoves

● Outdoor PM2.5 contributed an estimated 5-54 g m-3, or 20-44%, of indoor PM2.5

● Most indoor PM2.5 from wood chimney stoves was emitted during lighting; longer duration of use was not associated with higher indoor PM2.5

1. INTRODUCTION

Over 600 million Chinese and 2.8 billion people globally use solid fuel (i.e., biomass and coal) stoves for cooking, space heating, and other energy needs [1]. Traditional solid fuel stoves emit high concentrations of health-damaging pollutants, including particulate matter <2.5 microns in diameter (PM2.5), that contribute to both household and outdoor air pollution [2]. Household air pollution is responsible for an estimated 2.6 million yearly premature deaths, including 1.2 million deaths from ischemic heart disease and stroke, two leading causes of death in China and globally [3].

Replacing traditional biomass stoves with cleaner-burning technologies has the potential to reduce air pollution and its health impacts. Yet decades of efforts to implement cleaner-burning stoves [4, 5] have largely failed to measurably reduce air pollution or meet air quality guidelines [6, 7]. Many studies attribute these disappointing results to poor stove maintenance, low uptake of the new stove, and continued use of traditional stoves, i.e., mixed use or ‘stove stacking’ [8, 9]. Air pollution from local sources including industry, traffic, and neighbours’ solid fuel burning may also influence PM2.5

exposures, and subsequently mask or negate the benefits of clean stove interventions [10-13].

The extent to which stove-use patterns, i.e., frequency and duration of stove use, and outdoor air quality impact the success of stove interventions in reducing indoor PM2.5 is not well quantified [14]. Field studies in China, India, Ghana, and Mexico measured the post-stove intervention changes in stove use, but not the impacts on air pollution [13, 15-18]. A study in Kenya evaluated the indoor PM2.5 in homes using different stoves, but did not assess stove-use duration [19]. Johnson and Chiang ([20]) modeled the indoor PM2.5 impacts of different stove-use duration scenarios, but built their models from laboratory stove tests, which are known to differ considerably from field studies [21]. Further, none of these studies accounted for the contribution of outdoor air pollution, which is regarded as a barrier to clean indoor air in cities, but its contribution in rural areas is poorly understood [22]. Field studies that combine quantitative measures of indoor and outdoor air pollution and household stove-use patterns can provide more realistic insights into which stove intervention programs can meet air quality targets and achieve their intended health benefits.

We investigated the associations between stove-use patterns and indoor PM2.5 in rural Chinese homes, also accounting for outdoor PM2.5 that was measured concurrently. Drawing from our empirical results on indoor PM2.5 for different fuel-stove combinations, we also estimated the potential avoided

2

cardiovascular mortality associated with partial and complete transition to clean fuel stoves in southwestern China. To our knowledge, this is the first empirical study to quantitatively assess the independent and combined impacts of stove-use patterns and same-day outdoor air quality on indoor PM2.5 concentrations.

2. METHODS

2.1 Study location

We conducted the study in 12 villages along the eastern edge of the Tibetan Plateau in Beichuan County, Sichuan Province, China (+31.814°, +104.457°) (Figure S1). We selected this location because of a government-supported rural energy intervention program that provided homes with low-polluting semi-gasifier cookstoves and a supply of pelletized biomass fuel. The region’s temperate climate is characterized by mean temperatures of 20-30°C in June-August and 4-14°C in November-January. The nearest metropolitan area is Mianyang City, about 50 km southeast, which has a population of 5.5 million. Detailed information about the study site is published elsewhere [23, 24].

2.2 Study design

In this study we used the post-intervention rounds of data collected in summer (June - August, 2016) and winter (November 2016 - January 2017) in 150 homes that were enrolled in a stove intervention study. Seasonal measurements were used to capture the increased frequency and duration of stove combustion events in winter, likely due to space heating [25]. In both seasons, field staff traveled to participants’ homes to measure 48-h stove use and kitchen air pollution concentrations, and to administer questionnaires. Ethical review boards at McGill University, the University of Minnesota, Tsinghua University, and the University of Wisconsin-Madison approved this study.

2.3 Housing characteristics and energy-use practices

Housing characteristics and energy use practices in our study homes are described elsewhere [15, 23]. Briefly, houses were one or two stories and constructed of either wooden frames with partial earth exteriors or of brick and cement, and were clustered relatively close together (average distance to nearest neighbour=34 m). Natural ventilation practices (e.g. opening of doors and windows to the outside) were common, resulting in high air exchange rates in kitchens (mean=17 hr-1) [25]. All kitchens had wood-burning chimney stoves that vented outdoors. Most homes (79%) had received a semi-gasifier stove and supply of pelletized biomass fuel approximately 4-7 months before this study’s measurements began. Many (42%) had a liquefied petroleum gas (LPG) and/or electric stove (Figure S2). The wood-chimney stoves rank as Tier 1 (lowest performing group) for air pollution emissions according to the International Organization for Standardization International Workshop Agreement [26], whereas the LPG and electric stoves are classified as low-polluting Tier 4 stoves (highest performing group). The semi-gasifier stove can also perform at the Tier 4 level [27, 28].

2.4 Household questionnaires

3

Primary cooks completed questionnaires on household energy use and ventilation, and the presence of other air pollution sources (e.g., tobacco smokers) at each visit. Questions were adapted from previous energy surveys conducted in China [29] and re-tested prior to implementation.

2.5 Measurement of household stove-use frequency and duration

Stove use was objectively measured for 48-h periods in each season using real-time temperature sensors with a resolution of 1°C (Thermochron iButtons, Models DS1922L/DS1921G, Berkeley Air, USA). Sensors were placed on all stoves used at least once per month and programmed to record temperature every 10-min, a frequency shown to detect all stove-use events [15]. A control sensor was wall-mounted in kitchens, away from heating sources, to distinguish indoor temperature fluctuations from stove use. A previous study showed that 48-h measurement was representative of ‘usual’ stove-use behaviour in that season, and that the presence of air monitoring equipment did not alter stove-use patterns [15].

The number and duration of stove-use events were identified from temperature data using an algorithm adapted from Ruiz-Mercado et al [30]. Briefly, a stove-use event was defined as a time period during which the stove surface temperature exceeded the wall control temperature by at least 5°C and met other conditions of peak shape that distinguished it from room temperature change. To identify stove-use events, we generated time-resolved graphs of 48-h stove and wall control temperatures using an online software package (Plotly 2015) and MATLAB (2012). Each 48-h session was divided into two 24-h periods (SI section S.1). Temperature data were magnified so that the start and end time of each event could be visually identified within a ±5-min window of uncertainty. Duration of the stove use event (minutes) was calculated as the time between the start of a temperature increase to the temperature maxima, given field observations that a sustained decrease in stove temperature indicates the end of active use. Multiple stove-use events within an hour were considered a single event, where duration was calculated as the time between the start of a temperature increase to the temperature maxima of the final peak (Figure S3).

We calculated the following use metrics for each stove type, averaged across homes: 1) Daily mean number of stove events = total number of events divided by the total number of days that the stove was used at least once.2) Daily mean duration of stove events (minutes) = daily duration of use divided by the number of days on which the stove was used at least once.3) Daily mean stove-use duration per event (minutes) = duration of use divided by the number of measured events.

Each household-day was assigned into 1 of 4 stove categories: exclusive wood-chimney, exclusive semi-gasifier, exclusive LPG or electric, or combined use of two or more stoves defined as ‘mixed use’. Exclusive use was defined as days when at least 85% of daily stove-use minutes were attributable to a single stove category. Exclusive LPG and exclusive electric stove-use days were combined for analysis as both stoves emit low levels of pollution and the total number of days with exclusive electric stove use was small (n=2).

2.6 Real-time indoor and outdoor PM2.5 concentrations

4

Coincident with stove-use monitoring, we measured indoor and outdoor PM2.5 at 1-min intervals using DustTrak laser photometers (Model 8520; TSI Inc.; Shoreview, MN). Indoor measurements were conducted in kitchens. Instruments were placed approximately 1.5 m above the ground, away from open windows and doors, and 1-2 m from the main stove in a location that would not interfere with household activities. Outdoor PM2.5 was measured on the rooftop location that was shown to be representative of outdoor air pollution in all study villages [25].

We calibrated each DustTrak monitor with concurrent, 48-h integrated gravimetric PM2.5

measurements using methods reported elsewhere [25]. Days when indoor PM2.5 measurement did not meet our quality control criteria of ±20% of the target 24-h duration due to equipment or battery failure (23 days in winter and 43 days in summer) were excluded from analysis. For 8 non-consecutive days where outdoor PM2.5 was unavailable due to power interruption, we estimated values using a running average of outdoor PM2.5 on the 3 days prior to the missing day.

2.7 Estimating the contribution of outdoor PM2.5 to indoor PM2.5

Mean concentrations of 24-h outdoor and indoor PM2.5 were calculated for each household-day. We estimated the indoor-only contributions to daily indoor PM2.5 by subtracting out the 24-h means of outdoor PM2.5. This subtraction was based on the high temporal correlations between real-time outdoor and indoor PM2.5 during periods without indoor combustion activity (range of Pearson r: 0.86-0.97; Figure S4). Occasionally, this approach resulted in negative 24-h indoor PM2.5 concentrations (summer: n=14, 8% of days; winter: n=9, 8% of days). These values were removed from subsequent analysis after testing for sensitivity of their omission.

2.8 Analysis of stove-use patterns and indoor PM2.5

We evaluated the univariate associations between daily stove-use duration (minutes) and indoor PM2.5

concentrations for each stove-use category using season-specific scatterplots. We also conducted multivariable mixed effects regression models of daily stove-use duration and indoor PM2.5 with random household-specific intercepts to account for the correlation of repeated household measurements [31]. The regression models included variables known to be associated with indoor PM2.5, including kitchen volume (m3), use of a kitchen fan (no/yes), ventilation (frequency of opening windows and doors, categorical), and the presence of smokers in the home (no/yes). Separate models were conducted for each season and stove-use category. We additionally adjusted for outdoor PM2.5

and kitchen air exchange rates in a subset analysis of 380 household-days where both were measured. Indoor PM2.5 concentrations were log-transformed to account for their skewed distribution to higher values. Natural cubic splines with 2-4 degrees of freedom indicated approximately linear associations between natural log-transformed indoor PM2.5 and continuous independent variables. The analysis was conducted in STATA V12 (StataCorp 2011).

2.9 Cardiovascular impact of different stove-use scenariosWe estimated the premature deaths from ischemic heart disease and stroke associated air pollution exposures under different stove-use scenarios based on this study’s empirical results. The analysis was limited to the estimated 108 million rural residents of Chongqing, Guizhou, Qinghai, Sichuan, Tibet, and Yunnan provinces, where homes have environmental conditions [32] and energy use practices [33] that are similar to our study site. To estimate exposures to PM2.5, average indoor PM2.5 values for each

5

stove scenario and season from this study were multiplied by 0.7, which is the ratio of personal exposure to indoor concentration in this region [23]. We generated annual mean exposure by averaging the non-heating (8 months) and heating seasons (4 months), based on heating-related differences in air pollution observed in China by Streets et al. [32]. Estimates of relative risk and premature cardiovascular mortality were calculated using methods adopted from Apte et al. [34]. Briefly, we applied our estimated yearly exposure data for each scenario to published versions of the integrated exposure-response functions (IERs) used in the Global Burden of Disease study to estimate air pollution-related mortality due to stroke and ischemic heart disease [35]. The relative risks from the IERs were combined with provincial-level population data and death rates from the 2014 China Statistical Yearbook to estimate the number of yearly premature deaths attributable to air pollution under different stove scenarios [36].

2.10 Sensitivity analyses

We re-conducted our analysis using a more conservative ‘exclusive’ stove-use categorization where at least 95% of stove-use time was attributable to a single stove. To determine whether stove choice was associated with meal type (i.e. breakfast, lunch, dinner), we divided stove-use categories into three time periods: 7:00–9:00, 12:00–14:00, and 18:00–20:00 and compared these results with the daily average. To evaluate the impact of stove lighting on indoor air quality, we compared kitchen PM2.5

concentrations during the first half versus the second half of stove events in a sub-set of events (averaging n=25 events per stove category). Finally, to test the sensitivity of indoor-only PM2.5 derived from subtracting outdoor PM2.5, we modified outdoor DustTrak calibration slopes by ±10% and reconducted the analysis.

3. RESULTS

3.1 Household stove-use patterns

We measured 496 household-days of stove use and air pollution in 150 homes (Table S1). After excluding days with equipment failure or without stove use (n=52), a total of 444 household-days were included in our analysis (250 in summer and 194 in winter). Exclusive use of wood-chimney stoves was most common (59% of household-days), followed by mixed stove use (24% of household-days) and exclusive use of clean fuel stoves (electric or gas stoves=11%; semi-gasifier stoves=6%) (Figure 1). The average number of stove-use events per day was 2.6. Traditional stoves were used 2.1 times per day, on average, while LPG or electric stoves were used 2.3 times per day. Semi-gasifier stoves were used comparatively less (1.6 times per day, on average across seasons) (Table 1). Patterns of stove use across different mealtimes were similar, indicating that the choice of stove was not dependent on the time of day (Figures S5-S6). Use of a more conservative threshold for exclusive stove use did not appreciably change the results (<2% change in each category).

3.2 Indoor and outdoor PM2.5 concentrations

Daily geometric mean (GM) indoor PM2.5 concentrations were twice as high on days with any wood-chimney stove use compared with exclusive LPG or electric stove use (Table 2). Indoor PM2.5 was highest on days with mixed use (GM: 84 μg m -3 in summer and 177 μg m-3 in winter). In summer,

6

when semi-gasifier stove use was most common, indoor PM2.5 was lower on days with exclusive semi-gasifier stove use compared with days where wood-chimney stoves were used exclusively (GM: 46 μg m-3 versus 72 μg m-3) or with other stoves (GM: 84 μg m-3).

The percentage of household-days meeting the WHO’s interim indoor air quality target of 35 μg m-3

was 20% in summer and 4% in winter (Table 2). In summer, 16% of exclusive wood-chimney days and 15% of mixed-use days met the WHO target. One-third of days with exclusive semi-gasifier or LPG/electric stove use met the target (Figure 2).

Daily outdoor PM2.5 ranged from 5-51 μg m-3 in summer and 27-54 μg m-3 in winter (Table 2). After subtracting the outdoor PM2.5 contribution to indoor PM2.5, the proportion of summer household-days meeting the WHO air quality target increased to 15% for mixed stove-use days and to 39%, 77%, and 89% for exclusive wood-chimney, semi-gasifier, and LPG or electric stove-use days, respectively (Table S2). In winter, the proportions increased to 15% for mixed stove use, 18% for traditional stove use, and 55% for LPG or electric stove use. These results changed by less than 3% after modifying the calibration slope between gravimetric and optical PM by ±10% and after substituting low positive values (random assignment of 0.1-10 μg m-3) on outdoor-subtracted days with negative PM2.5, indicating that our results are robust to error in instrument calibration.

3.3 Relationships between stove-use duration and indoor PM2.5

We observed weak positive correlations between stove-use duration and indoor PM2.5 concentration on days with mixed stove use or exclusive use of wood-chimney or semi-gasifier stoves (Pearson r: 0.12-0.37). On days with exclusive LPG or electric stove use, duration was negatively correlated with indoor PM2.5 (r = -0.53 in summer and -0.13 in winter) (Figure S7). After adjusting for kitchen volume, ambient temperature, household ventilation, and the presence of smokers in the multivariate models, stove-use duration was not associated with indoor PM2.5 on days with exclusive use of wood stoves or mixed stove use (Table 3). In contrast, each additional 10 min of semi-gasifier use was associated with 5.9% (95% CI: 2.7, 9.2; p<0.001) higher indoor PM2.5 and, on days with exclusive use of LPG or electric stoves, an additional 10 min of use was associated with lower PM2.5 (-5.2%, 95% CI: -9.6, -0.9; p=0.02). Further adjustment for kitchen air exchange rates and outdoor PM2.5 did not appreciably change our results (Table S3).

3.4 Estimated avoided cardiovascular mortality with fuel-stove transitions in rural southwestern China

The estimated relative risks of cardiovascular mortality for different stove-use scenarios are presented in Figure 3. Compared with wood-chimney stoves, exclusive use of LPG or electric stoves translated into an estimated reduction of 48,000 yearly premature cardiovascular deaths in southwestern China, of which 7,000 were attributable to ischemic heart disease and 26,000 to stroke. Transitioning to semi-gasifier stoves was estimated to reduce 35,000 cardiovascular deaths in the region. Assuming that outdoor air pollution would also decrease with widespread clean energy transition, subtracting the estimated outdoor air pollution contribution to indoor PM2.5 exposures would further decrease the avoided number cardiovascular deaths by 71,000 or 124,000 with transition to exclusive semi-gasifier or exclusive LPG/electric stove use, respectively. In the case of incomplete transition and subsequent mixed stove use, we estimated a small increase in cardiovascular mortality (Table 4).

7

4. DISCUSSION

In our study, homes exclusively using clean fuel stoves (e.g., semi-gasifiers, LPG, or electric stoves) had lower indoor PM2.5 than homes using traditional biomass stoves, but seldom met the WHO’s indoor air quality target. Homes with mixed use of wood-chimney and clean fuel stoves had higher indoor PM2.5 than those exclusively using traditional stoves. Longer duration of use was associated with higher PM2.5 in homes exclusively using semi-gasifier stoves but not in homes using wood-chimney stoves, and outdoor air pollution was an important barrier to clean indoor air quality for all stove users. Together these findings offer several important directions for future rural energy intervention programs.

Clean stove technologies have served complementary roles, rather than as replacements for traditional stoves, in many settings [8, 37-39]. Our finding that mixed stove use was associated with higher indoor PM2.5 than exclusive use of traditional wood-burning stoves supports previous studies in India [40, 41], and reinforces the need for revised guidance for future stove intervention programs. Many factors explain households’ continued use of traditional stoves as well as their uptake of new stoves, including cooking preferences, user training and support, and effective financing [9, 42]. Household energy transition strategies that move beyond the stove technology and also take into account the local energy environment, households’ multiple purposes of energy use, and energy use behaviours are likely required to achieve more complete displacement of traditional stoves [43].

Duration of wood-chimney stove use was not associated with higher indoor PM2.5, suggesting that gradual reductions in traditional stove use alone may not measurably improve indoor air quality. This finding is consistent with field studies conducted in Nepal and India which found weak or no associations between measured or self-reported duration of biomass stove use and indoor PM2.5 [44, 45]. Our empirical results contrast with a recent modeling study [20], which indicated strong linear associations between stove use duration and indoor PM. Departures of empirical studies from single-zone box modeling results are expected, given that the assumptions of these models (e.g., perfect mixing, constant stove emissions rates, constant ventilation rates) are unlikely to be met by real-world, multi-zonal residences. For example, while we did not directly measure stove emissions in this study, we observed that average PM2.5 during the first half of a traditional stove combustion event was high and variable compared with average PM2.5 over an entire event (SI section S.2). On average in our study, 72% (standard deviation, SD=22%) of indoor PM2.5 from wood-chimney stove use events was emitted in the first half of stove-use activity. Further, our previous study found that natural ventilation is common and air change rates are high and variable in summer and winter (mean±SD=18.1±9.1 hr-1

and 15.2±7.4 hr-1, respectively)[25].

In contrast to traditional stoves, we found strong negative association between stove-use duration and indoor PM2.5 during LPG and electric stove use. This finding may relate to distinct cooking activities associated with this stove category. Shorter duration of LPG and electric stove use (<10 minutes) is likely to involve flash frying, which can produce high levels of cooking emissions [46]. Longer use (>30 minutes) is more likely to involve water-boiling and rice cooking tasks [15], which usually emit very little PM. This result highlights the need to consider the diversity of food preparation techniques and uses in intervention design and development. Future models of household energy intervention scenarios may better reflect ‘real world’ conditions by assigning PM2.5 exposure based on stove activity rather than as a function of usage time.

8

Outdoor PM2.5 was a key determinant of whether indoor PM2.5 met the WHO air quality target. Notably, the outdoor PM2.5 levels in our study were similar to those measured in rural areas of India, Mexico, and Guatemala (23-50 μg m-3) [40, 47, 48] but lower than rural northern China (41-267 μg m-

3) [49-51]. Outdoor air pollution is often considered an urban issue, however our study provides further evidence that outdoor PM is an important contributor to indoor air quality in rural areas as well [52]. Homes exclusively using LPG/electric stoves did not meet the WHO target unless outdoor PM was also low. The IER functions indicate a supra-linear association between PM2.5 and risk of cardiovascular mortality; that is, risk is steep at lower levels of air pollution and converges to a near flat relationship at higher air pollution levels [53]. This shape implies that substantial health benefits will accrue only from interventions that reduce exposures to low levels. Our results support a recent observation that improved coordination between indoor and outdoor air monitoring is important to achieving the full potential health benefits of clean energy interventions [22].

Our analysis was informed by air pollution measurements in homes with exclusive clean fuel use that are located in villages where traditional wood stoves are pervasive. This study and our previous work indicate that a considerable fraction of community PM2.5 in this setting is from household solid fuel burning, and that indoor/outdoor air recirculation can occur within minutes [25]. Residential emissions from solid fuel combustion may account for up to a quarter of China’s outdoor PM2.5 [2]. Based on the exponential nature of the IER curves, significant additional health benefits are achieved with PM2.5

reductions in the lower range. Subtracting outdoor PM2.5 from indoor levels in homes exclusively using clean fuel stoves indicate that cardiovascular mortality reductions could more than double. We expect that, in addition to agricultural burning, motor vehicle traffic, and local industry, local PM2.5 sources also include neighbours’ solid fuel burning. Thus, stove programs implemented at the regional scale, rather than the household scale, should result in reductions in both indoor and outdoor PM2.5, and thus greater health gains.

Our study has several limitations to consider for future work. First, our analysis focused on PM2.5 given its importance for air quality targets and guidelines. However, the chemical composition of PM2.5, which may impact its health and climate impacts [54, 55], can vary by fuel-stove combination. Chemical analysis may also facilitate more accurate quantification of indoor-outdoor PM2.5

relationships compared with the relatively simple subtraction method used in our study, which can help inform air pollution exposure mitigation strategies [2]. Future studies should consider chemical analysis and source apportionment of indoor and outdoor PM2.5. Second, differences in cooking styles, ventilation, and the presence of other air pollution sources across settings may limit the generalizability of our results, particularly to settings that are more heavily impacted by regional PM sources than our rural and mountainous study setting. Our measurements and methods could be applied to future studies in settings with energy, housing and regional PM conditions that differ from ours. Finally, our health analysis was based on several assumptions related to exposure estimation. We did not account for uncertainty in these assumptions or the IER functions, as this is the focus of our future study of the air quality and health impacts of different fuel-stove transitions for China.

Notable strengths of our study include its large sample size of 496 household-days that capture a unique combination of field-based, real-time measurements of stove use and both indoor and outdoor air pollution. The stove patterns analyzed include exclusive and near-exclusive use of multiple clean fuel stoves including semi-gasifiers, which have been promoted as promising rural energy

9

interventions [56], but for which little empirical evidence on their performance exists [57, 58]. Previous studies collected stove-use data without air pollution, measured indoor PM2.5 with self-reported stove use at low temporal resolution (i.e., hours), or modeled the stove use-PM2.5 relationships based on laboratory-assessed stove emissions. Although a small number of studies in rural settings measured outdoor air pollution and hypothesized its impact on stove intervention success [52, 59, 60], our large sample of paired, real-time outdoor and indoor PM2.5 concentrations enabled us to quantify the impact of outdoor PM2.5 on indoor air quality at a household level.

Our empirical results provide a useful synthesis of stove use and air pollution information that is directly relevant to policymakers and clean energy implementers who are considering stove intervention programs. Over 53 countries, including China, have committed to supporting the adoption and implementation of clean stoves and fuels [61]. Our study demonstrates the value of measuring different stove-use patterns and outdoor air pollution concentrations to understand their impacts on intervention effectiveness. Advances in sensing technology and data processing continue to lower barriers to tasks such as merging stove monitoring data with indoor and outdoor air pollution measurements. This study suggests that policy makers, researchers, and other stakeholders could leverage these technological and computational advances to improve evaluation of community-level household energy projects and better discern between interventions that fail to meet indoor air quality targets due to high outdoor air pollution levels versus low stove adoption versus both.

5. CONCLUSIONS

To our knowledge this is the first empirical study to simultaneously measure indoor and outdoor air quality data under different stove-use patterns and quantitatively evaluate their combined and independent impacts on indoor air quality. Indoor PM2.5 was lowest in rural Chinese homes exclusively using LPG or electric stoves, followed by homes exclusively using semi-gasifier stoves. Mixed stove use, predominately a traditional wood chimney stove plus clean fuel stove, resulted in daily PM2.5

concentrations exceeding days with traditional-only cooking. Since cooking times were not associated with daily mean PM2.5 concentrations in homes using traditional stoves, we recommend that intervention stoves should categorically replace, rather than supplement, traditional cooking devices. Incremental fuel replacement methods are at risk of increasing household PM2.5 exposure levels. Outdoor PM2.5 also makes an important contribution to household PM2.5 levels; many homes exclusively using clean fuels would readily meet WHO indoor air quality target in the absence of contributions from outdoor PM. Clean energy programs implemented at the community or regional scales in regions with heavy reliance on traditional biomass stoves are likely to have even greater impacts in air pollution exposures relative to household level intervention.

ACKNOWLEDGEMENTSWe thank our staff and participants in Sichuan and Ajay Pillarisetti and Josh Apte for input on the health model. The manuscript’s contents are solely the responsibility of the grantee and do not necessarily represent the official views of the EPA. The EPA does not endorse the purchase of any commercial products or services mentioned in the publication. The authors declare no financial or competing interests.

10

Funding: This study was supported by the US Environmental Protection Agency [EPA-STAR #83542201] and the Wellcome Trust Sustaining Health program (#103906/Z/14/Z). J.B. was supported by a CIHR New Investigator Award [grant no. 141959].

11

Tables

Table 1: Seasonal stove-use activity for different fuel-stove types* in rural Sichuan homesSummer Winter

Wood chimney

Semi-gasifier LPG or electric

Woodchimney

Semi-gasifier LPG orelectric

Number of household-days measured 136 47 54 99 17 35

Absolute number of stove events recorded 293 71 116 214 29 82

Daily mean (SD) number of stove events** 2.1 (0.1) 1.5 (0.1) 2.2 (0.1) 2.2 (0.1) 1.7 (0.1) 2.4 (0.1)

Daily mean (SD) duration of stove events (min)** 241 (10) 100 (4) 64 (3) 279 (14) 147 (8) 75 (3)

Mean (SD) stove-useduration per event (min) 112 (4) 66 (4) 30 (2) 129 (5) 86 (6) 32 (3)

Peak stove activitytimes (hh:mm)

8:43,12:52,21:17

7:59, 12:04,19:59

8:15, 12:13,19:12

8:48, 12:44, 19:09

8:51,13:00,18:30

7:45, 11:33, 17:51

LPG=liquefied petroleum gas; SD = standard deviation *Categories combine events from household-days with exclusive and mixed-use (i.e., more than one stove type used).**Means only include days with at least one stove-use event.

Table 2: Average daily indoor and outdoor air pollution concentrations (μg m-3) in rural Sichuan, by fuel-stove category and season

Measurement typeNumber of household-

days

Arithmetic mean PM2.5

(95% CI, μg m-3)

Geometric mean PM2.5

(95% CI, μg m-3)

% household-

days <35 μg m-3*

% household-days <35

μg m-3 *, outdoor PM2.5 subtracted

SummerKitchen Wood chimney 94 114 (108,120) 72 (59, 87) 16 39 Semi-gasifier 13 119 (44,194) 46 (20, 101) 38 77 LPG or electric 19 48 (42,54) 38 (26, 56) 32 89 Mixed stove-use 41 111 (102,120) 84 (65, 107) 15 24 All stoves 172 105 (102,108) 62 (53, 73) 20 45Outdoor 172 36 (34,38) 16 (5, 51) - -

WinterKitchen Wood chimney 80 238 (226,250) 155 (125, 192) 5 18 Semi-gasifier** 2 102 92 0 0 LPG or electric 11 85 (73,97) 79 (62, 100) 0 55 Mixed stove-use 20 321 (243,399) 177 (110, 285) 5 15 All stoves 114 237 (228,246) 140 (117, 167) 4 22Outdoor 106 41 (39,43) 38 (27, 54) - -

CI = confidence interval; LPG = liquefied petroleum gas; PM2.5 = particulate matter <2.5 μm in diameter*Percent of household-days at or below the WHO interim 24-h target for household (household or indoor-only) air pollution based on arithmetic 24-h means of household PM2.5. **Our estimates for exclusive semi-gasifier stove use in winter were limited to just two measurement days and should therefore be treated with caution.

12

Table 3: Associations of stove-use duration (minutes) and indoor concentrations of fine particulate matter (ln(PM2.5), μg m-3) by season and stove-use category. Coefficients interpreted as the percent (%) change in indoor PM2.5 per 10-min increase in stove use, based on the log regression. Bold text indicates p≤0.05. Fuel-stove category

Summer Winter Both Seasons

Univariate Multivariable* Univariate Multivariable* Univariate Multivariable*

Wood chimney

N 139 110 102 102 241 183% change(95% CI)

0.4%(-0.7, 1.4)

0.0%(-1.4, 1.4)

0.8%(-0.4, 1.9)

0.9%(-0.3, 2.0)

0.8%(0.0, 1.7)

0.1%(-0.9, 1.0)

p-value 0.51 0.99 0.20 0.13 0.06 0.89

Semi-gasifier

N 45 37 19 19 64 48% change(95% CI)

5.5%(2.4, 8.6)

4.9%(1.3, 8.6)

6.7%(1.8, 11.6)

8.7%(3.5, 13.9)

6.8%(4.0, 9.6)

5.9%(2.7, 9.2)

p-value 0.001 0.009 0.008 0.001 <0.001 <0.001

LPG or electric

N 54 44 37 37 91 81% change(95% CI)

-2.2%(-8.0, 3.7)

-2.4%(-9.6, 4.8)

-3%(-6.4, 0.2)

-3.5%(-6.8, -0.2)

-2.1%(-6.6, 2.3)

-5.2%(-9.6, -0.9)

p-value 0.46 0.51 0.07 0.04 0.35 0.02

Mixed stove-use**

N 56 46 32 22 88 68% change(95% CI)

0.6%(-1.3, 2.6)

0.7%(-1.4, 2.9)

0.0%(-2.3, 2.3)

-0.6%(-2.6, 1.3)

0.8%(-0.8, 2.5)

0.0%(-1.5, 1.4)

p-value 0.52 0.51 0.99 0.54 0.33 0.96CI = confidence interval; LPG=liquefied petroleum gas; N=number of household-days *All models adjusted for kitchen volume, ambient temperature, use of a fan in the kitchen, frequency of opening windows during stove use, and the presence of one or more smokers in the home. ** Mixed-use refers to use of least two stoves during the 24-hour period.

13

Table 4:  Estimated cardiovascular-specific mortality (thousands) and relative %-changes in premature mortality in southwestern China associated with transition from exclusive traditional wood-chimney stove use to alternative stove scenarios*Household stove-use scenario**

PM2.5

(μg m-3) exposure

All cardiovascular

mortality

% changemortality*

**

Stroke mortality

% change mortality

IHDmortality

% change mortality**

*

Mortality (absolute and % reduction) for PM2.5 exposure groups based on stove-use Wood-chimney 70 142 Ref. 77 Ref. 31 Ref. Mixed-use 80 151 +6% 81 +5% 33 +6% Semi-gasifier 43 107 -25% 58 -25% 26 -16% LPG or electric 36 94 -34% 51 -34% 24 -23%Mortality for clean stove groups with outdoor PM2.5 sources subtractedSemi-gasifier 26 71 -50% 36 -53% 20 -35%LPG or electric 7 18 -87% 6 -92% 8 -74%IHD, Ischemic heart disease; PM particulate matter* Mortality is relative to counterfactual exposure level of 7 g m-3, below which we assume no additional health risk. Total deaths are based on rural population estimates of six southwestern Chinese provinces Sichuan (41.2M est. rural pop.), Yunnan (25.7M), Guizhou (19.8M), Chongqing (10.9M), Qinghai (2.7M), and Tibet (2.1M)**Stove-use scenarios are held constant on an annual basis ***Percentages are referenced to wood-chimney stoves

14

Figures

Graphical Abstract:

15

Figure 1: Proportion of household-days in each fuel-stove-use category in summer (250 household-days in 139 homes) and winter (194 household-days in 109 homes). The overlapping areas of circles represent mixed stove-use. Household-days when at least 85% of stove-use minutes were from a single fuel-stove type were assigned to an exclusive use category.

16

a b c

Figure 2. Cumulative distribution of kitchen PM2.5 concentrations for all household-days by season (a) and by fuel-stove type for summer (b), and winter (c). The vertical line (purple) is the WHO’s interim 1 (IT1) target of 35

μg m-3.

17

Figure 3. Relative risks for adult mortality from ischemic heart disease and cerebrovascular disease (i.e., stroke), as reported by Burnett et al. (2014). Vertical lines are the estimated annual PM2.5 exposure levels for adults in each fuel-stove group based on our empirical results.

18

References

1. Bonjour S, Adair-Rohani H, Wolf J, Bruce NG, Mehta S, Prüss-Ustün A, et al. 2013. Solid Fuel Use for Household Cooking: Country and Regional Estimates for 1980–2010. Environ. Health Perspect. 121:784–790; doi:10.1289/ehp.1205987.

2. Chafe ZA, Brauer M, Klimont Z, Van Dingenen R, Mehta S, Rao S, et al. 2014. Household cooking with solid fuels contributes to ambient PM2. 5 air pollution and the burden of disease. Environ. Health Perspect. 122: 1314; doi: 10.1289/ehp.1206340

3. Gakidou E, Afshin A, Abajobir AA, Abate KH, Abbafati C, Abbas KM, et al. 2017. Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. The Lancet. 390: 1345-1422; doi: 10.1016/S0140-6736(17)32366-8.

4. Manibog FR. 1984. Improved Cooking Stoves in Developing Countries: Problems and Opportunities. Annu. Rev. Energy 9: 199–227; doi:10.1146/annurev.eg.09.110184.001215.

5. Kshirsagar MP, Kalamkar VR. 2014. A comprehensive review on biomass cookstoves and a systematic approach for modern cookstove design. Renew. Sustain. Energy Rev. 30: 580–603; doi:http://dx.doi.org/10.1016/j.rser.2013.10.039.

6. Pope D, Bruce N, Dherani M, Jagoe K, Rehfuess E. 2017. Real-life effectiveness of “improved” stoves and clean fuels in reducing PM2.5 and CO: Systematic review and meta-analysis. Environ. Int. 101: 7–18; doi:http://doi.org/10.1016/j.envint.2017.01.012.

7. Quansah, R., Semple, S., Ochieng, C. A., Juvekar, S., Armah, F. A., Luginaah, I et al. (-2017. Effectiveness of interventions to reduce household air pollution and/or improve health in homes using solid fuel in low-and-middle income countries: A systematic review and meta-analysis. Environ. Int. 103: 73-90; doi: https://doi.org/10.1016/j.envint.2017.03.010.

8. Ruiz-Mercado I, Masera O. 2015. Patterns of Stove Use in the Context of Fuel–DeviceStacking: Rationale and Implications. Ecohealth 12: 42–56; doi:10.1007/s10393-0151009-4.

9. Rehfuess EA, Puzzolo E, Stanistreet D, Pope D, Bruce NG. Enablers and barriers to large-scale uptake of improved solid fuel stoves: a systematic review. 2014. Environ. Health Perspect. 122: 120; doi: 10.1289/ehp.1306639.

10. Mortimer K, Ndamala CB, Naunje AW, Malava J, Katundu C, Weston W, et al. 2017. A cleaner burning biomass-fuelled cookstove intervention to prevent pneumonia in children under 5 years old in rural Malawi (the Cooking and Pneumonia Study): a cluster randomised controlled trial. The Lancet. 389: 167–175; doi:10.1016/S0140-6736(16)32507-7.

11. Smith KR, McCracken JP, Weber MW, Hubbard A, Jenny A, Thompson LM, et al. 2011. Effect of reduction in household air pollution on childhood pneumonia in Guatemala (RESPIRE): a randomised controlled trial. The Lancet. 378: 1717–1726; doi:10.1016/S0140-6736(11)60921-5.

12. Secrest MH, Schauer JJ, Carter E, Baumgartner J. 2017. Particulate matter chemical component concentrations and sources in settings of household solid fuel use. Ind. Air. ; 125: 076002; doi:10.1111/ina.12389.

13. Piedrahita R, Dickinson KL, Kanyomse E, Coffey E, Alirigia R, Hagar Y, et al. 2016. Assessment of cookstove stacking in Northern Ghana using surveys and stove use monitors. Energy Sustain. Dev. 34: 67–76; doi:10.1016/j.esd.2016.07.007.

14. Khandelwal M, Hill Jr. ME, Greenough P, Anthony J, Quill M, Linderman M, et al. 2017. Why Have Improved Cook-Stove Initiatives in India Failed? World Dev. 92: 13–27; doi:https://doi.org/10.1016/j.worlddev.2016.11.006.

15. Clark S, Carter E, Shan M, Ni K, Niu H, Tseng J, et al. 2017. Adoption and use of a semi-gasifier cooking and water heating stove and fuel intervention in the Tibetan Plateau. Environ. Res. Lett. 12: 075004; https://doi.org/10.1088/1748-9326/aa751e.

16. Mukhopadhyay R, Sambandam S, Pillarisetti A, Jack D, Mukhopadhyay K, Balakrishnan K, et al. 2012. Cooking practices, air quality, and the acceptability of advanced cookstoves in Haryana, India: an exploratory study to inform large-scale interventions. Glob. Health Action 5: 19016;

19

doi:10.3402/gha.v5i0.19016.17. Ruiz-Mercado I, Canuz E, Smith KR. Temperature dataloggers as stove use monitors (SUMs): Field

methods and signal analysis. 2012. Biomass and Bioenergy. 47: 459-68; doi: https://doi.org/10.1016/j.biombioe.2012.09.003.

18. Pillarisetti A, Vaswani M, Jack D, Balakrishnan K, Bates MN, Arora NK, et al. 2014. Patterns of Stove Usage after Introduction of an Advanced Cookstove: The Long-Term Application of Household Sensors. Environ. Sci. Technol. 48: 14525–14533; doi:10.1021/es504624c.

19. Lozier MJ, Sircar K, Christensen B, Pillarisetti A, Pennise D, Bruce N, et al. 2016. Use of Temperature Sensors to Determine Exclusivity of Improved Stove Use and Associated Household Air Pollution Reductions in Kenya. Environ. Sci. Technol. 50: 4564–4571; doi:10.1021/acs.est.5b06141.

20. Johnson MA, Chiang RA. 2015. Quantitative Guidance for Stove Usage and Performance to Achieve Health and Environmental Targets. Environ. Health Perspect. 123: 820–826; doi:10.1289/ehp.1408681.

21. Roden CA, Bond TC, Conway S, Osorto Pinel AB, MacCarty N, Still D. 2009. Laboratory and field investigations of particulate and carbon monoxide emissions from traditional and improved cookstoves. Atmos. Environ. 43: 1170–1181; doi:http://dx.doi.org/10.1016/j.atmosenv.2008.05.041.

22. Bruce N, Pope D, Rehfuess E, Balakrishnan K, Adair-Rohani H, Dora C. 2015. WHO indoor air quality guidelines on household fuel combustion: Strategy implications of new evidence on interventions and exposure–risk functions. Atmos. Environ. 106:451–457; doi: https://doi.org/10.1016/j.atmosenv.2014.08.064.

23. Ni K, Carter E, Schauer JJ, Ezzati M, Zhang Y, Niu H, et al. 2016. Seasonal variation in outdoor, indoor, and personal air pollution exposures of women using wood stoves in the Tibetan Plateau: Baseline assessment for an energy intervention study. Environ. Int. 94: 449–457; doi:10.1016/j.envint.2016.05.029.

24. Shan M, Yang X, Ezzati M, Chaturvedi N, Coady E, Hughes A, et al. 2014. A feasibility study of the association of exposure to biomass smoke with vascular function, inflammation, and cellular aging. Environ. Res. 135:165–172; doi: 10.1016/j.envres.2014.09.006.

25. Carter E, Archer-Nicholls S, Ni K, Lai AM, Niu H, Secrest MH, et al. 2016. Seasonal and Diurnal Air Pollution from Residential Cooking and Space Heating in the Eastern Tibetan Plateau. Environ. Sci. Technol. 50:8353–8361; doi:10.1021/acs.est.6b00082.

26. IWA. 2012. Guidelines for evaluating cookstove performance. Geneva: International Organisation on Standardisation. 2012.

27. Carter EM, Shan M, Yang X, Li J, Baumgartner J. 2014. Pollutant Emissions and Energy Efficiency of Chinese Gasifier Cooking Stoves and Implications for Future Intervention Studies. Environ. Sci. Technol. 48:6461–6467; doi:10.1021/es405723w.

28. Shan M, Carter E, Baumgartner J, Deng M, Clark S, Schauer JJ, et al. 2017. A user-centered, iterative engineering approach for advanced biomass cookstove design and development. Environ. Res. Lett. 12: 095009; doi: https://doi.org/10.1088/1748-9326/aa804f.

29. Baumgartner J, Schauer JJ, Ezzati M, Lu L, Cheng C, Patz J, et al. 2011. Patterns and predictors of personal exposure to indoor air pollution from biomass combustion among women and children in rural China. Indoor Air. 21:479–488; doi:10.1111/j.1600-0668.2011.00730.x.

30. Ruiz-Mercado I, Canuz E, Walker JL, Smith KR. Quantitative metrics of stove adoption using Stove Use Monitors (SUMs). 2013. Biomass and Bioenergy. 57: 136-48; doi: https://doi.org/10.1016/j.biombioe.2013.07.002.

31. Laird NM, Ware JH. 1982. Random-effects models for longitudinal data. Biometrics 963–974.32. Streets DG, Bond TC, Carmichael GR, Fernandes SD, Fu Q, He D, et al. 2003. An inventory of gaseous

and primary aerosol emissions in Asia in the year 2000. J. Geophys. Res. Atmos. 108: 30; doi:10.1029/2002JD003093.

33. Duan X, Jiang Y, Wang B, Zhao X, Shen G, Cao S, et al. 2014. Household fuel use for cooking and heating in China: Results from the first Chinese Environmental Exposure-Related Human Activity Patterns Survey (CEERHAPS). Appl. Energy. 136:692–703; doi:http://doi.org/10.1016/j.apenergy.2014.09.066.

20

34. Apte JS, Marshall JD, Cohen AJ, Brauer M. 2015. Addressing Global Mortality from Ambient PM2.5. Environ. Sci. Technol. 49:8057–8066; doi:10.1021/acs.est.5b01236.

35. Institute for Health Metrics and Evaluation (IHME). 2013. Global burden of disease study, ambient air pollution risk model 1990–2010. Inst. Heal. Metrics Eval. Available: http://www.healthdata.org/gbd.

36. National Bureau of Statistics-China. China Statistical Yearbook. 2014; Available from: http://www.stats.gov.cn/tjsj/ndsj/2014/zk/indexeh.htm

37. Johnson NG, Bryden KM. 2012. Factors affecting fuelwood consumption in household cookstoves in an isolated rural West African village. Energy 46: 310–321; doi:https://doi.org/10.1016/j.energy.2012.08.019.

38. Masera OR, Saatkamp BD, Kammen DM. 2000. From Linear Fuel Switching to Multiple Cooking Strategies: A Critique and Alternative to the Energy Ladder Model. World Dev. 28: 2083–2103; doi:https://doi.org/10.1016/S0305-750X(00)00076-0.

39. Stanistreet D, Hyseni L, Bashin M, Sadumah I, Pope D, Sage M, et al. 2015. The Role of Mixed Methods in Improved Cookstove Research. J. Health Commun. 20:84–93; doi:10.1080/10810730.2014.999896.

40. Aung TW, Jain G, Sethuraman K, Baumgartner J, Reynolds C, Grieshop AP, et al. 2016. Health and Climate-Relevant Pollutant Concentrations from a Carbon-Finance Approved Cookstove Intervention in Rural India. Environ. Sci. Technol. 50:7228–7238; doi:10.1021/acs.est.5b06208.

41. Kar A, Rehman IH, Burney J, Puppala SP, Suresh R, Singh L, et al. 2012. Real-Time Assessment of Black Carbon Pollution in Indian Households Due to Traditional and Improved Biomass Cookstoves. Environ. Sci. Technol. 46: 2993–3000; doi:10.1021/es203388g.

42. Lewis JJ, & Pattanayak, SK. (2012). Who adopts improved fuels and cookstoves? A systematic review. Environ. Health. Perspect. 120: 637-645; doi: 10.1289/ehp.1104194.

43. Ezzati M, Baumgartner JC. Household energy and health: where next for research and practice?. 2017. The Lancet. 389: 130-2; doi: https://doi.org/10.1016/S0140-6736(16)32506-5.

44. Balakrishnan K, Ghosh S, Ganguli B, Sambandam S, Bruce N, Barnes D. F, & Smith K. R. (2013). State and national household concentrations of PM 2.5 from solid cookfuel use: results from measurements and modeling in India for estimation of the global burden of disease. Environ. Health, 12:77; doi: 10.1186/1476-069X-12-77.

45. Pokhrel AK, Bates MN, Acharya J, Valentiner-Branth P, Chandyo RK, Shrestha PS, et al. 2015. PM2.5 in household kitchens of Bhaktapur, Nepal, using four different cooking fuels. Atmos. Environ. 113: 159–168; doi:http://dx.doi.org/10.1016/j.atmosenv.2015.04.060.

46. Huang W, Baumgartner J, Zhang Y, Wang Y, Schauer JJ. 2015. Source apportionment of air pollution exposures of rural Chinese women cooking with biomass fuels. Atmos. Environ. 104: 79–87;doi:http://dx.doi.org/10.1016/j.atmosenv.2014.12.066.

47. Brauer M, Bartlett K, Regalado-Pined J, & Perez-Padilla, R. (1995). Assessment of particulate concentrations from domestic biomass combustion in rural Mexico. Environ. Sci. Technol. 30: 104-109; doi: 10.1021/es9501272.

48. Naeher LP, Smith, KR, Leaderer BP, Mage D, & Grajeda, R. (2000). Indoor and outdoor PM 2.5 and CO in high-and low-density Guatemalan villages. Journal of Exposure Science and Environmental Epidemiology. 10: 544-551: doi: 10.1038/sj.jea.7500113.

49. Xiang RB, Song JB, Zhu Y. Guang W. (2016). Indoor Air Quality in Kitchens in Rural China. Environ. Eng. Sci. 33: 699-704; doi: https://doi.org/10.1089/ees.2015.0210.

50. Zhao X, Zhang X, Xu X, Xu J, Meng W, Pu W. 2009. Seasonal and diurnal variations of ambient PM2.5 concentration in urban and rural environments in Beijing. Atmos. Environ. 43:2893–2900; doi:http://dx.doi.org/10.1016/j.atmosenv.2009.03.009.

51. Zhu C-S, Cao J-J, Shen Z-X, Liu S-X, Zhang T, Zhao Z-Z, et al. 2012. Indoor and outdoor chemical components of PM2. 5 in the rural areas of Northwestern China. Aerosol Air Qual. Res 12: 1157–1165.

52. Du W, Shen G, Chen Y, Zhuo S, Xu Y, Li X, et al. 2017. Wintertime pollution level, size distribution and personal daily exposure to particulate matters in the northern and southern rural Chinese homes and

21

variation in different household fuels. Environ. Poll. 231: 497-508; https://doi: doi.org/10.1016/j.envpol.2017.08.039.

53. Pope CA, Burnett RT, Krewski D, Jerrett M, Shi Y, Calle EE, Thun MJ. Cardiovascular mortality and exposure to airborne fine particulate matter and cigarette smoke: shape of the exposure-response relationship. 2009. Circulation. 120: 941-948; https://doi.org/10.1161/CIRCULATIONAHA.109.857888.

54. Baumgartner J, Zhang Y, Schauer JJ, Huang W, Wang Y, Ezzati M. 2014. Highway proximity and black carbon from cookstoves as a risk factor for higher blood pressure in rural China. Proc. Natl. Acad. Sci. 111:13229–13234; doi:10.1073/pnas.1317176111.

55. Safai PD, Raju MP, Rao PSP, Pandithurai G. 2014. Characterization of carbonaceous aerosols over the urban tropical location and a new approach to evaluate their climatic importance. Atmos. Environ. 92: 493–500; doi:http://dx.doi.org/10.1016/j.atmosenv.2014.04.055.

56. Smith K, Deng K. 2010. A Chinese national improved stove program for the 21st century to promote rural social and economic development. Energy Policy Res. 1: 23–35.

57. Jetter J, Zhao Y, Smith KR, Khan B, Yelverton T, DeCarlo P, & Hays MD. 2012. Pollutant emissions and energy efficiency under controlled conditions for household biomass cookstoves and implications for metrics useful in setting international test standards. Environ. Sci. 46: 10827-10834; doi: 10.1021/es301693f.

58. Shen G, Tao S, Wei S, Zhang Y, Wang R, Wang B, et al. 2012a. Reductions in Emissions of Carbonaceous Particulate Matter and Polycyclic Aromatic Hydrocarbons from Combustion of Biomass Pellets in Comparison with Raw Fuel Burning. Environ. Sci. Technol. 46:6409–6416; doi:10.1021/es300369d.

59. Yip F, Christensen B, Sircar K, Naeher L, Bruce N, Pennise D, Lozier M, et al. 2017. Assessment of traditional and improved stove use on household air pollution and personal exposures in rural western Kenya. Environment international. 99: 185-191; doi: https://doi.org/10.1016/j.envint.2016.11.015.

60. Du W, Shen G, Chen Y, Zhu X, Zhuo S, Zhong Q, Qi M, Xue C, Liu G, Zeng E, Xing B. 2017. Comparison of air pollutant emissions and household air quality in rural homes using improved wood and coal stoves. Atmos. Environ. 166: 215-223; doi: https://doi.org/10.1016/j.atmosenv.2017.07.029.

61. Rehfuess E, Mehta S, & Prüss-Üstün, A. 2006. Assessing household solid fuel use: multiple implications for the Millennium Development Goals. Environ. Health Perspect. 114: 373-378; doi: 10.1289/ehp.8603.

22

SUPPLEMENTAL INFORMATION

S.1 Defining 24-h measurement periods: 48-h sampling times were divided into two days: the first spanned late-morning of day 1 to 11:59 on day 2, and the second went from noon (12:00) to late-morning of day 3. In instances where combustion events spanned two household-days, the events were assigned to the first day if most use happened before 12:00 on day 2 and to the second day if most use occurred after 12:00.

S.2 Further sensitivity analyses. In a subset of our data (n = 20) in which cooking events were fully resolved (i.e. no overlap from other stove events), we estimated the relative distribution of PM2.5 concentrations derived from stove-use events was different based on stove type. On average in our study, 72% (SD = 22%) of indoor PM2.5 from wood-chimney stove use events was emitted during the initial half of stove-use activity, compared with 53% (SD = 23%) for semi-gasifier stoves and 40% (SD = 17%) for LPG or electric stoves.

Traditional wood stoves were used on 79% of household-days in the summer and increased to 88% in winter. Traditional stoves were frequently used alone (52% household-days in summer and 68% in winter), whereas semi-gasifiers were most often used in combination with the traditional stoves (8% of household-days in summer; 12% in winter) than exclusively (3% in summer; 7% in winter). The number of household-days that used LPG or electric stoves at least once decreased from 28% in the summer to 22% in the winter. Exclusive LPG or electric stove use was 13% in summer and 8% in winter.

Dividing stove-use habits into mealtimes (breakfast, lunch, and dinner) did not strongly affect the percentage of stove-types within each group (Figure S5). Because of the finer division, we expect more occurrences of exclusive use. For example, exclusive traditional wood-chimney stoves remained close to 56-58% in summer and 63-69% in winter. For all fuel types inter-variation of stove choice based on mealtime was low.

23

Supplemental Tables

Table S1: Attempted and completed measurement-days of stove use and kitchen PM2.5 monitoringSummer Winter Both seasons

Number of attempted measurement-days 278 218 496Complete 24-h stove-use measurement* 250 (90%) 194 (89%) 444 (90%)Complete 24-h PM2.5 measurement* 191 (69%) 124 (57%) 315 (64%)Complete 24-h stove-use and PM2.5 measurements

172 (62%) 114 (52%) 286 (58%)

Same as above, plus matched outdoor PM2.5 measurements

172 (62%) 106 (49%) 278 (56%)

*Includes days meeting the criteria of a measurement being ±20% of the 24-h period. Incomplete stove use measurement was due to temperature sensors breaking or falling off of the stoves. Incomplete kitchen PM2.5

measurement was due to equipment failure or power loss.

Table S2: In reference to Table 2, a re-evaluation of the average daily indoor PM2.5 concentrations in rural Sichuan homes after subtracting the estimated outdoor PM2.5 contribution, by fuel-stove category and season.

Stove TypeNumber of household-

days

Mean PM2.5

(95% CI, μg m-3)

Geometric mean PM2.5

(95% CI, μg m-3)

% household-days

<35 μg m-3 *

Summer Wood chimney 94 91 (84,98) 43 (31, 60) 41 Semi-gasifier 13 87 (14.160) 29 (11, 80) 69 LPG or electric 19 8 (1,15) 7 (3, 15) 95 Mixed stove-use 41 82 (74,90) 54 (37, 80) 24 All stove types 172 78 (75,81) 46 (38, 56) 47Winter Wood chimney 72 162 (151,173) 76 (51, 113) 21 LPG or electric 11 35 (21,49) 29 (13, 63) 55 Mixed stove-use 20 276 (198,354) 111 (52, 237) 20 All stove types 106 174 (164,184) 82 (62, 109) 25*Percent of household-days at or below the WHO interim 24-h target for household (indoor) air pollution.

24

Table S3: Associations of stove-use duration (minutes) and indoor concentrations of PM2.5, ln(μg/m3) by season and stove-use category. Coefficients interpreted as the percent (%) change in indoor PM2.5 per 10-min increase in stove use, based on the log regression. Bold text indicates p≤0.05.

Fuel-stove type

Summer Winter Both Seasons

Multivariable with outdoor

PM2.5*

Multivariable with outdoor

PM2.5 and AERs**

Multivariable with outdoor

PM2.5*

Multivariable with outdoor

PM2.5 and AERs**

Multivariable with outdoor

PM2.5*

Multivariable with outdoor

PM2.5 and AERs**

Wood chimney

N 96 66 67 63 141 104Coeff

(95% CI)-0.8%

(-2.7, 01.1)0.4%

(-0.5, 1.4)1.1%

(-0.4, 2.7)1.0%

(-0.4, 2.4)-0.4%

(-1.6, 0.8)0.5%

(-0.5, 1.5)p-value 0.40 0.41 0.14 0.15 0.54 0.32

Semi-gasifier

N 30 20 12 11 37 23Coeff

(95% CI)5.2%

(1.6, 8.9)6.3%

(1.1, 12.7) N too limited N too limited 6.1%(2.5, 9.6)

4.9%(-0.4, 10.3)

p-value 0.005 0.05 0.001 0.07

LPG or electric

N 44 29 23 21 67 50Coeff

(95% CI)-2.2%

(-9.4, 5.0)-5.4%

(-13.2, 2.4)-5.4%

(-11.6, 0.8)-2.3%

(-6.0, 1.4)-6.8%

(-12.5, -1.1)-5.8%

(-11.0, -0.6)p-value 0.56 0.18 0.09 0.23 0.02 0.03

Mixed stove-use*

N 41 25 12 13 53 38Coeff

(95% CI)-0.4%

(-2.7, 2.0)-1.5%

(-3.6, 6.7) N too limited N too limited -0.7%(-2.3, 0.8)

-0.5%(-1.9, 0.9)

p-value 0.77 0.18 0.36 0.50AER, air exchange rate; CI, confidence interval; LPG, liquefied petroleum gas; PM, particulate matter*Models adjusted for kitchen volume, ambient temperature, use of a fan in the kitchen, frequency of opening windows during stove use, the presence of one or more smokers in the home, and outdoor air pollution. **Additionally adjusted for air exchange rates

25

Supplemental Figures

Figure S1. Location of the study site in the eastern Tibetan Plateau (Beichuan, Sichuan)

26

Figure S2. Clockwise from top left: Example of an LPG stove, traditional wood-burning chimney stove, electric induction plate, and semi-gasifier stove that are common in our study villages.

ba

c d

27

Figure S3. Stove-use patterns in four study households. Indoor PM2.5 (black line) with activity overlaid stove-use activity. Stove-use activity peaks are marked by asterisks (*) for wood chimney (blue), LPG (red), and semi-gasifier (orange).

28

Figure S4. Real-time outdoor PM2.5 (green line) and indoor PM2.5 in study households (black line) during periods without of stove use. Associated Pearson correlation coefficients (r) displayed are based on smoothed over 60-min running averages.

29

Figure S5: Proportion of meals in each fuel-stove use category in rural Sichuan homes in summer (339 meal events) and winter (253 meal events) based on objective stove-use monitoring with temperature sensors. The overlapping areas of circles represent mixed stove-use.

30

Figure S6: Comparison of combustion event frequency as determined by real-time kitchen air pollution concentrations (top) and stove-use frequency (bottom) in summer (a) and winter (b).

31

a

b

Figure S7: Relationship between daily minutes of stove use and daily PM2.5 concentrations (μg m-3) for different fuel-stove types in summer (top; a) and winter (bottom; b) including Pearson correlation coefficients (r) and 95% confidence intervals. Lines are fitted to a least-squares linear regression. Black

lines show the modeled relationship between daily minutes of stove use and daily PM2.5 concentrations (μg m-3) for a ventilated wood stove by Johnson and Chiang (2015).

32