Article Cite This: Environ. Sci. Technol. XXXX, XXX, XXX−XXX
pubs.acs.org/est
Effectiveness of a Household Energy Package in Improving Indoor Air Quality and Reducing Personal Exposures in Rural China Jill Baumgartner,*,†,‡,§ Sierra Clark,‡ Ellison Carter,§,∥ Alexandra Lai,⊥ Yuanxun Zhang,# Ming Shan,∇ James J. Schauer,⊥ and Xudong Yang∇ †
Institute for Health and Social Policy, McGill University, Montreal, Quebec H3A 1A3, Canada Department of Epidemiology, Biostatistics, & Occupational Health, McGill University, Montreal, Quebec H3A 1A2, Canada § Institute on the Environment, University of Minnesota, Minneapolis, Minnesota 55108, United States ∥ Department of Civil & Environmental Engineering, Colorado State University, Fort Collins, Colorado 80521, United States ⊥ Environmental Chemistry and Technology, University of Wisconsin-Madison, Madison, Wisconsin 53706, United States # College of Resources and Environment, University of the Chinese Academy of Sciences, Beijing 100049, P. R. China ∇ Department of Building Science, Tsinghua University, Beijing 100084, P. R. China
Downloaded via KEAN UNIV on July 26, 2019 at 12:34:34 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
‡
S Supporting Information *
ABSTRACT: We evaluated whether an energy package comprising a low-polluting semigasifier cookstove with chimney, water heater, and pelletized biomass fuel would improve air pollution in China. We measured the stove use, 48-h air pollution exposures (PM2.5, black carbon), and kitchen concentrations (PM2.5, black carbon, carbon monoxide, nitrogen oxides) for 205 women, along with ambient PM2.5. Over half (n = 125) were offered the energy package after baseline assessment, forming “treated” and “untreated” groups, and we repeated the measurements up to 3 occasions over 18months. Kitchen carbon monoxide did not change, and nitrogen oxides increased in summer but decreased in winter for both groups. Summer geometric mean exposures and kitchen concentrations of PM2.5 and black carbon decreased by 24−67% in women who received the energy package, but greater reductions (48−70%) were observed in untreated homes, likely due to increased use of gas stoves. After adjusting for differences in outdoor PM2.5, receiving the energy package was associated with decreased winter exposures to PM2.5 (−46%; 95% CI: −70, −2) and black carbon (−55%; −74, −25) and the summer increases were smaller (PM2.5: 8%; −22, 51 and black carbon: 37%; −12, 113). However, PM2.5 exposures remained 1.5−3 times higher than those of health-based international air pollution targets.
■
INTRODUCTION Household air pollution from cooking and heating with solid fuel (i.e., biomass and coal) stoves is a widespread environmental exposure, responsible for an estimated 1.6 million premature deaths in 2017.1 Solid fuel stoves also significantly contribute to outdoor air pollution, accounting for up to a third of ambient fine particulate matter (PM2.5) in China and other parts of Asia.2 Decades of household energy intervention programs aimed to decrease these environmental and health impacts by introducing alternative biomass stoves and fuels that were intended to replace the traditional stoves. However, the success of these programs in reducing air pollution was hindered by a variety of factors.3 Many replacement biomass stoves still emitted high levels of air pollutants into homes and communities.4,5 Single-purpose stoves failed to meet the many energy needs of households, including cooking different foods, space heating, water boiling, and even lighting, leading to continued use of traditional stoves.6−8 © XXXX American Chemical Society
Gas-fueled and electric-powered stoves are the lowest polluting;9 however, issues of accessibility, affordability, and convenience have slowed their adoption and use in more remote and poor regions.10 In regions with abundant biomass, international organizations and governments have promoted the distribution of advanced-combustion biomass stoves as near-term, low-polluting household energy interventions.11−13 Semigasifier stoves burn biomass fuel but mimic gas stoves by separating the combustion processes of the biomass fuels from the generated gas, creating a gas-like adjustable flame.14 Semigasifiers have consistently outperformed other biomass stoves in laboratory 15−18 and in-home field tests of emissions,19,20 and several models are classified into the Received: Revised: Accepted: Published: A
April 5, 2019 June 21, 2019 June 25, 2019 June 25, 2019 DOI: 10.1021/acs.est.9b02061 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
outdoors. Locally available hardwood was processed into pellets at a nearby factory and delivered to homes as needed.31 Distribution of the energy package was supported by the Ministry of Science and Technology and the Ministry of Agriculture as part of a national energy demonstration program.
highest two performing stove categories based on the ISO International Workshop Agreement (IWA) performance targets (ISO/TR 19867−1:2018), with emissions performances that approach gas stoves. If the promising emissions performances of semigasifier stoves were to successfully translate into “real world” performance in communities, then the air pollution and population health benefits would be substantial.21 Comprehensive evaluations of semigasifiers and other advanced combustion biomass stoves under conditions of actual use are essential for determining whether these promising technologies warrant further development and inclusion among clean household energy options. Yet very few studies, most of which were experimental in nature or observational but without comparison groups to control for secular trends, have assessed the air pollution impacts of advanced combustion biomass stoves.19,22−28 We evaluated whether a government-sponsored energy package intervention comprised of a low-polluting semigasifier stove with chimney, water heater, and biomass pellet fuel would improve indoor air quality and personal exposures in rural Chinese women. To our knowledge this is the first assessment of a household energy package and the first study to longitudinally evaluate the indoor air quality and personal exposure impacts of a semigasifier chimney stove that burns pelletized fuel. Study Location and Household Energy Practices. We conducted our study in 12 villages that were located in Beichuan County, Sichuan, China (+31.814°, + 104.457°) (Figure S1 of the Supporting Information, SI). We chose this location for the study because a planned, governmentsupported rural energy demonstration program allowed us to assess the “real world” implementation of a household energy intervention. The study region was ∼6.2 km2, and homes ranged in elevation from 1000 to 1450 m above sea level. At baseline, all households used traditional wood-burning chimney stoves with semi-enclosed (three-sided) combustion chambers and attached chimneys that vented smoke outdoors. Most households (79%) additionally used electric induction or gas stoves. Natural ventilation through opened windows and doors was common, although some households also used an electric fan to ventilate during cooking. Detailed information about the study location and energy use practices is published elsewhere.29,30 Energy Package: Multipurpose Semigasifier Chimney Stove and Processed Biomass Fuel. The energy package intervention was composed of a high-performing semigasifier cookstove and water heater, chimney, and supply of processed biomass fuel that was provided to homes at no cost. The semigasifier stove’s forced draft design uses a small fan to control the supply of air into the combustion chamber, which achieves more efficient combustion. It was iteratively designed by engineers at Tsinghua University (Beijing) through a series of laboratory and in-home emissions assessments, where it achieved high performance in thermal efficiency (ISO Tier 3) and stove emissions of PM2.5 (ISO Tier 4).16,17 The stove was designed to be compatible with a range of local cooking practices and with rural economic structures in China, and to address household energy needs beyond cooking (i.e., an attached tank to heat water for drinking or washing). Its unique features included an automatic ignition system, an adjustable flame that could achieve different levels of firepower, a pellet feeder with a hand crank that enabled refuelling during use, and a custom fit aluminum chimney that vented smoke
■
METHODS AND MATERIALS Study Design and Participants. We prospectively evaluated the stove use and air pollution impacts of the energy package. The energy program was preplanned, but we leveraged its structure to design a controlled before-and-after intervention study within it. In 2014, we enrolled 205 Chinese women from 204 households in 12 villages into the study. Women were invited to participate if they lived in villages where the energy program was being implemented, regularly used traditional biomass stoves, and were not pregnant. Most eligible households had only one adult woman resident, but in the case of multiple eligible women, we selected the oldest. We enrolled only women because they were the main stove users and were not typically smokers, whereas most men (>60%) smoked. Among women who declined participation (n = 75), most did so because they worked outside of the villages. Between May 2014 and February 2017, staff traveled to participants’ homes on up to five occasions to assess energy behaviors and measure air pollution. The energy package was distributed by local technicians and officials who were independent of this study, but with whom we coordinated on the timing of distribution. Women in half of the villages received the intervention package after baseline assessment (“treated”), while women in the remaining villages served as comparison homes (“untreated”) to account for secular trends and reduce potential spillover effects from smoke generated by homes without the energy package. In one village, only half of the participants received the energy package because stove supply ran out and thus the remaining participants in that village were considered untreated, resulting in 125 treated homes with the energy package and 80 untreated homes without the package. Baseline and postintervention data collection campaigns were conducted in the winter and summer to capture seasonal differences in stove use. Post-intervention measurements occurred at least 6 months after homes received the energy package to give them adequate time to adjust to using it. Untreated homes were offered the energy package at the end of our study, approximately 12−18 months after the treated homes (Figure S1), and 55 of untreated homes (69%) accepted the stove. Reasons for not accepting the package at the end of the study were temporary relocation to a nearby city (n = 15) or they not interested (n = 10). Detailed information about the study design and participants is published elsewhere.29 Study protocols were approved by ethical review boards at all investigator institutions. Outdoor, Kitchen, and Personal Exposures to PM2.5 and Black Carbon. Women’s 48-h integrated exposures to PM2.5 were measured using Harvard Personal Exposure Monitors (PEM) with greased impaction surfaces32 that held 37 mm PTFE filters (2.0-μm pore size, Zefluor, Pall Corp, U.S.A.) and were attached downstream from pumps (Apex Pro, Casella, U.K.) operated at 1.8 lpm (±10%). Staff verified the pump flow rates before and after sampling using a calibrated rotameter. The air monitors were placed inside small waistpacks that participants were instructed to wear at all times B
DOI: 10.1021/acs.est.9b02061 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
of exposure and internal batch calibration of the tube markings, according to methods described previously.40 Kitchen concentrations of nitrogen oxides (NOx) and nitrogen dioxide (NO2) were measured using Ogawa Passive Samplers (Ogawa, U.S.A.; Pompano Beach, FL, U.S.A.) and chemically treated sample collection pads. Immediately after measurement, the collection pads were immersed into deionized water and stored in a −30 °C freezer prior to analysis. The NOx and NO2 samples were extracted from the Ogawa collection pads and analyzed for absorbance using spectrophotometric analysis at the University of the Chinese Academy of Science in Beijing.41 Analysis of two separate sample collection pads yields concentrations for combined NOx species and NO2. The difference between the two is the NO concentration (ppb). Details on the analysis procedures and field-blank correction are provided elsewhere.29 We limited our reporting to NOx because over half of the NO2 samples were below the limit of detection and the remaining samples were just above, resulting in high levels of uncertainty. Other Study Measurements. Field staff administered questionnaires to assess household demographic information, socioeconomic status (using an asset index), kitchen ventilation practices (frequency of opening windows and doors), stove and fuel use patterns, and the presence of smokers in the home (no/yes). Questions were read to participants in the local dialect of Mandarin-Chinese and their responses were recorded directly onto the survey. Kitchen volume (m3) and air exchange rates were estimated when measurement conditions allowed, using methods described elsewhere.33 Real-time outdoor temperature and relative humidity were measured throughout the study using a local meteorological station. Finally, we assessed each participant’s level of activity by measuring the number of steps taken over 48-h by placing pedometers inside of the waistpacks that held the air samplers. Detailed information on these measurements is published elsewhere.29,33,42 Statistical Analysis. Season-specific summary statistics for air pollution concentrations were calculated for treated and untreated households at baseline and postintervention. We evaluated whether relevant socio-demographic, behavioral, environmental and housing structural factors were associated with intervention status (treated versus untreated), time (preversus postintervention), or air pollution concentrations using two-sample difference-in-means tests, χ2 measures of association, and by visual inspection of data plots. Variables evaluated included those previously shown to impact household air pollution28,29 including participant characteristics (e.g., age, education, secondary occupation, and pedometer-measured steps during exposure assessment), household characteristics (e.g., kitchen volume, socioeconomic status, number of meals cooked per day during air monitoring), ventilation practices (e.g., use of a kitchen fan, frequency of opening windows and doors), and environmental factors (e.g., ambient temperature, outdoor air pollution, environmental tobacco smoke). A difference-in-difference approach was used compare the changes in air pollution outcomes (exposure: PM2.5, black carbon; kitchen: PM2.5, black carbon, CO) in homes that received the energy package relative to the changes in air pollution in homes that did not receive the package.43 Since the sample was clustered at two levels (participants nested within villages), the associations between the energy package and dependent air pollution variables were assessed using multivariable mixed effects regression models with restricted
but could place within 1−2 m while sitting, sleeping, or bathing. Compliance was assessed through household visits by study staff and by placing small pedometers (Omron HJ321 TriAxis, Omron, Japan) inside of the packs that recorded daily steps. In instances when the pumps failed before capturing ±20% of 24- or 48-h (7% of attempted measurements), we estimated exposures using season-specific prediction models that were developed from kitchen air pollution measurements, as described elsewhere.29 Kitchen PM2.5 was measured simultaneously with personal exposures at 1 min intervals using colocated laser photometers (DustTrack 8520, TSI Inc., U.S.A.) and gravimetric instruments including a calibrated pump and either a PEM or GK2.05 SH cyclone (Mesa Laboratories, U.S.A.) holding 37 mm PTFE filters. The monitors were placed on kitchen surfaces that were approximately 1.5 m above the ground, away from windows and doors, and 1−2 m from the main household stove in a location that would not interfere with daily activities. Staff photographed the measurement location on the first visit so that it could be replicated in subsequent visits. In the 7% of gravimetric kitchen measurements that failed to capture 48-h ± 20%, we used the integrated laser photometer measurement that was corrected using separate regressions of time-weighted 48-h real-time PM2.5 and integrated PM2.5 for each DustTrak and in each season. Details on air pollution measurements and related quality control measures are published elsewhere.29,33 Outdoor PM2.5 was measured at 1 min intervals throughout the study using a laser photometer (DustTrack Model 8530, TSI Inc., U.S.A.) that was equipped with an in-line cassette and PTFE filter. The monitor was placed on top of a two-story building at a central location where, as demonstrated in an earlier study, air pollution concentrations were representative of the other study villages.29,33 The outdoor gravimetric samples were collected at intervals of 24-, 48-, 120-, and 168-h, depending on air quality and perceived risk of filter overloading. Field blank filters were collected for 7−10% of all samples. All filter samples and blanks were weighed before and after sampling using a microbalance (MT 5, Mettler-Toledo Inc., U.S.A.), after being conditioned in a climate controlled room (23 ± 2 °C; 40 ± 5% relative humidity), and then analyzed for black carbon using an optical transmissometer (Sootscan OT21, Magee Scientific, U.S.A.).34 Measurement of both PM2.5 mass and black carbon is an important scientific contribution given the increasing evidence that its composition matters for health35 and climate,36 and can vary by stove combustion conditions and fuel.37,38 All filter analyses took place at the University of WisconsinMadison. Detailed information on laboratory methods and quality control procedures are published elsewhere.29,39 Kitchen Measurements of CO and NOx. In a random subsample of households (15−20%) in each campaign, we used passive samplers to measure 48-h integrated kitchen concentrations of carbon monoxide (CO), nitrogen oxides (NOx) and nitrogen dioxide (NO2). For CO, passive diffusion tubes (CO 50/a-D 50−600 ppm-h; Draeger Safety; Conyers, GA, U.S.A.) were placed next to the PM2.5 inlet. The tubes change color with exposure to CO, and the length of the stain is proportional to the concentration. At the end of each 48-h sampling period, three staff members independently measured the extent of color change in millimeters (mm) twice and reported the average of their two readings. The mm reading was converted to a ppmv concentration based on the duration C
DOI: 10.1021/acs.est.9b02061 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Article
Environmental Science & Technology
Table 1. Baseline Socio-Demographic and Household Characteristics for Women by Intervention Status and for Women without Post-Intervention Measurementc treated (n = 125)a
untreated (n = 80) age (years) 52.3 (49.6, 55.2) ethnicity Han 61 (76%) Qiang 19 (24%) socio-economic status highest (top 33%) 21 (26%) middle (middle 33%) 25 (31%) lowest (lowest 33%) 34 (43%) education none 8 (10%) primary school completion 59 (74%) secondary school completion 2 (3%) no response 11 (13%) number of inhabitants in the home 4.0 (3.7, 4.3) lives with at least one smoker? yes 48 (60%) number of daily steps (median and IQR) summer 8474 (5323, 13391) winter 4370 (2046, 7761) self-reported stove use events during 48-h monitoring period summer 5.6 (5.3, 5.8) winter 4.8 (4.3, 5.3) kitchen volume (m3) 43.0 (41.7, 44.3) estimated effective air exchange rate in the kitchen (h−1)b summer 16.7 (13.9, 19.4) winter 15.7 (13.3, 18.1)
no post-intervention measurement (n = 42)
51.6 (49.6, 53.6)
51.6 (47.7, 55.5)
100 (80%) 25 (20%)
34 (81%) 8 (19%)
47 (38%) 43 (34%) 35 (28%)
10 (24%) 12 (28%) 20 (48%)
21 (17%) 93 (74%) 7 (6%) 4 (3%) 4.1 (3.9, 4.3)
6 (14%) 36 (86%) 0 (0%) 0 (0%) 3.9 (3.5, 4.4)
75 (60%)
23 (55%)
8441 (5302, 12220) 4746 (2417, 7216)
7870 (3994, 10295) 3250 (1959, 6204)
5.6 (5.4, 5.7) 4.9 (4.5, 5.2) 42.4 (41.4, 43.3)
5.7 (5.4, 6.0) 4.9 (4.1, 5.7) 41.9 (39.8, 43.9)
18.7 (16.6, 20.9) 14.9 (13.1, 16.6)
19.1 (14.4, 23.7) 13.5 (11.4, 15.6)
a
No statistically significant differences (p-value
