Environ. Sci. Technol. 2011, 45, 882–889
Pollutant Concentrations within Households in Lao PDR and Association with Housing Characteristics and Occupants’ Activities L . M O R A W S K A , * ,† K . M E N G E R S E N , † H. WANG,† F. TAYPHASAVANH,‡ K. DARASAVONG,‡ AND N. S. HOLMES† International Laboratory for Air Quality and Health, Queensland University of Technology, GPO Box 2434, Brisbane Queensland, 4001, Australia, and Ministry of Health, P.O. Box 1232, Vientiane, Lao PDR
Received July 8, 2010. Revised manuscript received November 21, 2010. Accepted November 30, 2010.
The paper presents the results of a study conducted to investigate indoor air quality within residential dwellings in Lao PDR. Results from PM10, CO, and NO2 measurements inside 167 dwellings in Lao PDR over a five month period (December 2005-April 2006) are discussed as a function of household characteristics and occupant activities. Extremely high PM10 and NO2 concentrations (12 h mean PM10 concentrations 1275 ( 98 µg m-3 and 1183 ( 99 µg m-3 in Vientiane and Bolikhamxay provinces, respectively; 12 h mean NO2 concentrations 1210 ( 94 µg m-3 and 561 ( 45 µg m-3 in Vientiane and Bolikhamxay, respectively) were measured within the dwellings. Correlations, ANOVA analysis (univariate and multivariate), and linear regression results suggest a substantial contribution from cooking and smoking. The PM10 concentrations were significantly higher in houses without a chimney compared to houses in which cooking occurred on a stove with a chimney. However, no significant differences in pollutant concentrations were observed as a function of cooking location. Furthermore, PM10 and NO2 concentrations were higher in houses in which smoking occurred, suggestive of a relationship between increased indoor concentrations and smoking (0.05 < p < 0.10). Resuspension of dust from soil floors was another significant source of PM10 inside the house (634 µg m-3, p < 0.05).
1. Introduction A number of indoor emission sources have been identified in developing countries, including combustion sources such as cooking and heating (predominantly from biomass burning, often in open fires, but also using coal stoves), tobacco smoking (1), and other sources such as dust resuspension and the penetration of outdoor pollutants to indoors. The multiplicity of factors which have an impact on indoor pollutant concentrations in developing countries are such that predictions of concentrations based on studies * Corresponding author phone: +61 (7) 3138 2616; fax: +61 (7) 3138 9079; e-mail:
[email protected]. † Queensland University of Technology. ‡ Ministry of Health. 882
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 45, NO. 3, 2011
conducted in other countries or climatic zones, for houses of different construction or using different fuels, are usually impossible. One developing country where no studies on indoor pollution concentrations have been conducted, and for which hardly data exists on outdoor concentrations, is the Lao People’s Democratic Republic (PDR). Lao PDR is a landlocked tropical country located in the center of the Southeast Asian peninsula, with mountains and plateaus covering approximately 80% of the country. A large part of the country is covered with forest, and therefore, wood is by far the dominant fuel used for cooking and also for heating during winter (but since it is a tropical country the need for heating is nowhere near as high as in more temperate and colder climates). There are substantial parts of the country where people do not have access to electricity, and poverty and poor health are among the most significant problems of the country. Life expectancy remains low (62 years in 2008) (2), and, in general, health indicators compare unfavorably by regional standards. Poverty leads to health problems, while health problems lead to further poverty, and one of the major factors expected to contribute to poor health is indoor air pollution. At the time of this study, there was no quantitative information available on indoor air pollutant characteristics in dwellings in Lao PDR, nor had any previous studies investigated, in a direct or indirect manner, the presence or concentration of pollutants in indoor air. However, there were some hypotheses on relative concentrations of pollutants resulting from indoor cooking and heating during the cold season as well as on a number of aspects relating to house design (for example, dirt versus covered floor), household structure, and life style of the occupants, which could have direct or indirect impact on the presence and characteristics of pollutants originating from indoor sources. In particular, an important feature of house design is location of the kitchen in relation to the rest of the house, especially its sleeping areas. It was found that in rural areas there are two main types of traditional house designs: (i) dwellings where the cooking area was located inside the house and the sleeping areas were in an immediate vicinity to the stove used for cooking and heating or (ii) dwellings where the kitchen located outside the main house. Typically, neither type of house had inner walls (partition between different house areas was provided by curtains made out of fabric), and therefore it could only been speculated as to what impact these partitions had on indoor air quality. Another important factor for consideration was tobacco smoking, and although no quantitative information was found on the prevalence of tobacco smoking in Lao PDR, it appeared that a large proportion of both male and female adults and young people smoked, with large pipes often used in rural areas. Considering these large gaps in knowledge, the aim of this study was to investigate indoor pollutant concentrations within dwellings in Lao PDR as well as associations between indoor air quality and housing characteristics and household activities. The study did not, however, aim to investigate the indoor/outdoor relationship or to apportion the sources of the individual pollutants. This paper discusses the methodology used in the study, compares the average indoor pollutant concentrations with World Health Organisation (WHO) air quality guidelines, and analyzes the results of pollutant measurements in relation to housing characteristics and occupant activity. 10.1021/es102294v
2011 American Chemical Society
Published on Web 12/20/2010
2. Methodology 2.1. Study Location. Measurements were conducted between 23rd December 2005 and 14th April 2006 inside 199 dwellings in nine districts of Lao PDR. These districts were selected as part of a stratified random sample and provided a wide representation of ethnic groups, wide range of housing characteristics, high prevalence of respiratory illness, accessibility and adequate staff resources (Phonhong, Mad, Feuang, Thoulakhom, Kasy, and Vangvieng districts in Vientiane province and Bolikhanh, Khamkeut, and Pakkading districts in Bolikhamxay province). For each province, a list of hospitals, health centers, and villages and the village population within each district was obtained from the National Statistics Centre. A total of 20 health centers were selected as follows: 12 from Vientiane province (Thoulakhom (3), Phonhong (1), Kasy (1), Vangvieng (3), Feuang (3), Mad (1)) and 8 from Bolikhamxay province (Bolikhanh (2), Pakkading (2), Khamkeut (4)). For each health center (or group of centers), four children admitted for ARI (acute respiratory illness), aged 1-4 years sequentially from first October 2005, were enrolled in the study. No more than two such children were enrolled from any one village. One to two children were randomly selected from the same village (in a different house), matched by age, ethnic group, and location of kitchen. The houses in which these children resided were included in the study. Wherever possible, houses with a kitchen inside and houses with a separate kitchen were chosen within each district. Measurements were made during the dry season from December to April (corresponding to winter and spring), encompassing a graduation of heating frequency within the households over the period. It should be stressed, however, that in Lao, which is a tropical country, there is in fact little variation in average temperature, with a difference in average temperature between the seasons of about 6 °C (http:// www.climatetemp.info/laos/). Therefore, the variation of temperature between the seasons is not expected to have a major impact on pollutant concentrations. In addition, both were dry seasons, so outdoor pollutant levels were not affected by rain during any of the measuring campaigns, which would have resulted in lowering the general outdoor background. During the winter (November to February) measurements were made in Vangvieng, Kasy, and Mad within Vientiane; with measurements in Bolikhamxay (Bolikhanh, Khamkeut, and Pakkading) from January to February and in Vientiane (Thoulakhom, Phonhong, and Feuang) from March to April. 2.2. Pollutants and Instrumentation. Particle mass (PM10), nitrogen dioxide (NO2), and carbon monoxide (CO) were measured within the houses, similar to many other air quality studies in developing countries, due to their emission from wood burning stoves and resuspension within the households. Since the major source of sulfur dioxide (SO2) is fossil fuels, which are not common within Lao PDR, SO2 concentration was not measured in this study. The lack of electricity, especially within the rural areas, meant it was necessary to measure concentrations using battery operated pumps that were recharged daily back at the Department of Health offices. These pumps were connected to a cyclone (PM10), long duration color detection CO tubes, and NO2 adsorption tubes. Active sampling methods were used to measure NO2 and CO concentrations since they are typically more accurate than passive sampling methods (3). Measurements in the dwellings were performed between 06:00-18: 00, in order to include all major indoor sources of pollutant generation and also to ensure that the sampling time was within the capacity of portable batteries. Detailed timeactivity profiles were not recorded; however, it is customary in Lao that three meals are cooked per day. All of the measurements only took place once in each house over a
12 h period. The location of instruments was chosen based on the following constraints: 1) representative of the main living area of the house; 2) away from the sources; 3) away from the windows and the doors; and 4) safe for the instrumentation and people. In order to keep the ventilation for the houses as close as possible to typical conditions, the residents were asked to keep the doors and windows opened according to their usual practice. It should be noted that during the day, the windows and doors of the houses were typically left open, regardless of the season. Ventilation was typically high in timber houses anyway, due to their construction, which left large, visible gaps between the beams or sticks from which the walls were constructed. 2.2.1. Measurement of PM10. Particle mass sampling was performed following the procedure outlined by the US Department of Health and Labor, in the Occupational Safety and Health Administration Technical Manual (see Supplementary 1 for details of method validation for the purpose of this study). The measurement of particle mass of respirable particles was performed using an Escort ELF sampling pump connected to a cyclone (MSA) containing preweighed 0.8 µm pore PVC filters (MSA) (see Supplementary 3 Figure 1S). The flow rate through the filter was measured using a Digital calibrator (MSA), and valves were adjusted until a flow rate of 1.7 L min-1 was obtained. Filter cassettes were stored and equilibrated in a desiccator (20 °C, 65%RH) for 24 h after sampling. The filters were then removed and weighed on a microbalance (Mettler Toledo), with an upper limit of 3 g and 0.002 mg readability. 2.2.2. Measurement of NO2. Following the NIOSH method 6014 (4), NO2 was sampled by adsorption of the NO2 onto two consecutive molecular sieve tubes (SKC 224-40-02) treated with triethanolamine and extracted with a diluted triethanolamine solution (50 mL) which reacted with NO2 to form nitrite and nitrate ions. The sum of nitrite and nitrate ions was measured as NO2 concentration using a UV-vis spectrometer (Thermo Spectronic) at λ ) 540 nm after the completion of color development with Griess-Saltzman reagent. The method was validated for the purpose of this study (see Supplementary 1 for the details). The two ends of each adsorption tube were broken immediately prior to sampling and connected together via a short length of plastic tubing. The open end of the breakthrough tube was inserted into one port of the Gemini twin port sampler and the flow rate through the absorbent tubes adjusted to 0.2 L min-1. The sensitivity of the NO2 measurement method was dependent on that of the Griess-Saltzman reagent (5). 2.2.3. Measurement of CO. CO concentrations were measured using long duration color detection tubes designed for measurement of time weighted average concentrations. A measured volume of gas (or air) is drawn through a tube which contains chemicals that change in color in response to the presence of a specific target gas (or range of gases) present in the sample. By knowing the volume of gas or air sampled, the amount of color change, read on a linear scale on the colorimetric gas detection tube, can be translated into an assessment of the level of gas present, described as a percentage of the total air or in parts per million (ppm). 2.3. Experimental Set Up. Within the dwellings in Lao PDR the filters and detector tubes were located about 1.5 m above the ground and connected in parallel to a pump using a T-piece and Gemini twin tube sampler (see Supplementary Figure 1S). Pollutant concentrations were measured within the living area in each dwelling for a period of twelve hours, from 06:00 to 18:00, using the methodologies described above. During the sampling period the instruments were left unattended to minimize the interference with the normal daily activities occurring within the dwelling. Following sampling the NO2 tubes and filter cassettes were sealed and returned to the International Laboratory for Air Quality and VOL. 45, NO. 3, 2011 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
883
TABLE 1. Arithmetic Mean, Standard Error of the Mean and Range (Values in Parentheses) of PM10, NO2, and CO Concentrations in the Two Provinces pollutant
Bolikhamxay
Vientiane
WHO guideline (averaging time period)
PM10
1183 ( 99
1275 ( 98
(µg m-3) NO2a (µg m-3) CO (ppm)
(147-3030) 561 ( 45 (142-1837) 0.490 ( 0.059(0.045-1.976)
(43-4577) 1210 ( 94 (168-4904) 0.430 ( 0.032 (0.002-1.667)
50 µg m-3 (24 h mean) 150 µg m-3 (24 h mean,5% increase of short-term mortality) 200 µg m-3 (1 h mean) 10 ppm (8 h mean)
a
The concentrations measured by the tubes have been multiplied by 1.92 based on the comparison between the NOx analyzer and NO2 tubes at QUT (see Supplementary 1).
Health (ILAQH), Queensland University of Technology (QUT), in Brisbane, Australia, for analysis, while the TWA CO color detector tubes were read in Lao PDR and the concentration calculated from the total volume of air sampled through the tube. 2.4. Other Information. A survey was administered to all residents in the houses in which the air quality measurements were performed. The survey comprised of questions regarding the house characteristics, number of people in the dwelling, number of rooms and materials used in the dwelling construction, cooking location, ventilation, and heating and cooking methods as well as occurrence of smoking within the dwelling and conduct of other pollution generation activities. 2.5. Statistical Analysis. All statistical analyses (correlation, univariate and multivariate ANOVA, and general linear regression) were conducted using Microsoft Excel and a statistical analysis software package, SPSS (SPSS). A 5% significance level was used for all statistical tests. After conducting Kolmogorov-Smirnov goodness-of-fit tests and inspecting the histograms of the data, the data were analyzed using linear regression, with the natural logarithm of the pollutant concentration forming the response, and location and dwelling characteristics forming the explanatory variables. GLM and ANOVA were used to determine the equality or otherwise of the mean concentrations with respect to the covariates. Based on the ANOVA results, differences in mean particle concentration within the provinces between the different dwelling characteristics/activities were analyzed using either Tukey’s post hoc multiple comparison method for the cases where equal variances was assumed or Dunnett’s C method for groups in which equal variances were not assumed. Pearson correlation coefficients were calculated to assess the pair wise linear relationships between the three pollutants under different cooking conditions. In the following, crude effects refer to associations that have been calculated without adjustment or account for other covariates, and adjusted effects refer to associations obtained from a regression involving multiple covariates. This differentiation is not stated where the distinction is obvious.
3. Results 3.1. Measurements Inside the Dwellings. Of the 199 houses in which measurements were made, approximately 84% of the data (167 houses) were analyzed (exclusions were due to technical problems, including inability to unambiguously match the tube or filter with the corresponding data sheet, and, subsequent to inspection, were considered to be missing at random). Due to concerns regarding the integrity of some of the measurement data, for example occasions where the recorded flow rates varied by greater than 10% between individual readings, two separate data sets were created. The first included all the measurements, while the second excluded data from measurements in which the individual readings varied by greater than 10%. Unless significant differences 884
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 45, NO. 3, 2011
were observed between the correlations for the two different data sets only the results of the complete data set are discussed. 3.1.1. Distribution of Concentrations. KomolgorovSmirnov analysis of the data sets showed that the natural logarithm of the measurement data displayed a better goodness of fit for a normal distribution. Thus all the analyses in this study were conducted using loglinear regression. 3.1.2. Indoor Pollutant Concentrations. Table 1 summarizes pollutant concentrations measured in Lao PDR. Mean indoor PM10 and NO2 concentrations were significantly higher in Vientiane (PM10 ) 1275 ( 98 µg m-3, 95% CI 1081-1469 µg m-3; NO2 ) 1210 ( 94 µg m-3, 95% CI 1023-1396 µg m-3) than in Bolikhamxay (PM10 ) 1183 ( 99 µg m-3, 95% CI 984-1382 µg m-3; NO2 ) 561 ( 45 µg m-3, 95% CI 471-651 µg m-3), while CO concentrations were significantly higher in Bolikhamxay (CO ) 0.430 ( 0.032 ppm, 95% CI 0.367-0.494 ppm) compared to Vientiane (CO ) 0.490 ( 0.059 ppm, 95% CI 0.372-0.609 ppm). 3.1.3. Pollutant Concentrations in the Different Provinces and Villages. Box and whisker plots of the PM10, NO2, and CO concentrations in the nine districts are shown in Figure 1a), b), and c). ANOVA analysis confirmed that some mean pollutant concentrations differed significantly between districts. The mean NO2 concentrations in Phonhong and Vangvieng were significantly higher (p < 0.05) than all other districts except Mad and Kasy; the mean CO concentration in Kasy was significantly lower (p < 0.05) than all other districts except Mad; and the mean PM10 concentration in Phonhong was significantly lower (p < 0.05) than all other districts except Mad, Pakkading, and Thoulakhom. 3.1.4. Household Activities. ANOVA analysis showed that PM10 and NO2 concentrations were significantly higher in Vientiane in houses that used wood (1343 ( 108 µg m-3 and 1263 ( 100 µg m-3, respectively) compared to electricity (377 ( 109 µg m-3, p < 0.05 and 234 ( 37 µg m-3, p < 0.05, respectively). However, it should be noted that over 90% of houses used wood as a dominant fuel in this area, and only five households used electricity as cooking fuel. In addition, the PM10 concentration was significantly higher (p ) 0.003) across all houses in which cooking occurred on open stoves without chimneys (1267 ( 74 µg m-3) compared to those that had chimneys (349 ( 145 µg m-3), pointing out a relationship between stove type, fuels and elevated indoor pollutant concentrations (see Figure 4 and Supplementary Figure 2S). Furthermore, NO2 concentrations displayed a strong linear correlation with CO concentrations (p < 0.001 in Bolikhamxay) and PM10 concentrations (p < 0.032) for households with wood burning stoves and those households with an open stove without a chimney indicating that cooking is also a significant source of PM10, NO2, and CO. However, somewhat surprisingly, no significant difference was observed between pollutant concentrations as a function of cooking location (cooking in living/sleeping room PM10 ) 1303 ( 213 µg m-3; NO2 ) 999 ( 174 µg m-3; CO ) 0.497 ( 0.096 ppm; cooking in separate room PM10 ) 1183 ( 105 µg m-3; NO2
( 116 µg m-3; NO2 ) 942 ( 148 µg m-3; CO ) 0.475 ( 0.060 ppm in houses with reported nonsmokers). 3.2. Association between the Air Quality and House Characteristics. Pollutant (PM10, NO2, and CO) concentrations inside the house as a function of materials used in the construction of the dwelling (roof, floor, internal and external walls) were analyzed using the methodology described above for the total data set as well as each province independently (see Supplementary Table 1S and 2S for details). PM10 concentration was significantly higher (p < 0.05) in houses with soil floors, 1503 ( 131 µg m-3, compared to houses with nonsoil floors, 869 ( 87 µg m-3, as shown in Figure 2. This suggests that approximately 634 µg m-3, which is equivalent to 42% of the total PM10 concentration within houses with soil floors, may be associated with the resuspension of dust from the floor in soil houses. NO2 concentrations within houses with brick walls were significantly higher (brick external wall NO2 ) 1968 ( 742 µg m-3; brick internal wall NO2 ) 1560 ( 530 µg m-3) than in houses using other types of materials, shown in Figure 3 a) and b). There is no clear explanation for this, and further studies should investigate the role of the brick building shell in affecting pollutant transport and retention within the house. For example, in several cases, while the sleeping and living areas were constructed of brick, they were immediately next to a separate (in theory) cooking area, constructed of timber, which was fully open, to allow for airflow. The mean CO concentration was significantly higher in houses that were not heated during the cold season (CO ) 0.700 ( 0.121 ppm in houses that were not heated; CO ) 0.427 ( 0.029 ppm in houses that were heated, p ) 0.024). However, the measurements were conducted over several months during which the temperature changed somewhat from the start of the campaign to the end. Furthermore, since no record was made of whether the heating was used at the time of the measurements, nor the regularity or otherwise of heating within the majority of households during the cold season, it is impossible to assess the relationship between heating and pollutant concentrations. One possible explanation for the higher CO concentrations in houses that were not heated during the cold season is that these houses could have reduced ventilation and so increased indoor pollutant concentrations could result from reduced dispersion to the outdoors. FIGURE 1. Box and whisker plots of a) PM10; b) NO2; and c) CO concentrations in the nine districts. ) 973 ( 110 µg m-3; CO ) 0.404 ( 0.036 ppm; cooking in separate building PM10 ) 1109 ( 109 µg m-3; NO2 ) 1034 ( 122 µg m-3; CO ) 0.475 ( 0.063 ppm; cooking outside PM10 ) 1739 ( 284 µg m-3; NO2 ) 1055 ( 219 µg m-3; CO ) 0.509 ( 0.070 ppm) in Bolikhamxay. NO2 and CO concentrations were correlated for all cooking locations: cooking in the living/ sleeping room (p ) 0.013), in a separate room (p ) 0.001), or in a separate building (p ) 0.036). Furthermore, PM10 and NO2 concentrations were correlated (p ) 0.002) in dwellings in which the kitchen was in a separate building. The findings indicate that a relationship exists between the pollutant concentrations but suggests that other household characteristics and activities, in addition to cooking influence the concentrations and affect the relationship. In particular, while these locations were visually separate, the small distance between them and the partitions which are often so transparent that one can easily see though them made the transport of pollutants between them very quick and efficient. Mean pollutant concentrations of PM10 and NO2 were typically higher in dwellings in which smoking occurred, with p-values of 0.058 and 0.078, respectively (PM10 ) 1335 ( 91 µg m-3; NO2 ) 1032 ( 77 µg m-3; CO ) 0.439 ( 0.030 ppm in houses with reported smokers compared to PM10 ) 1041
4. Discussion In this section we discuss the associations between the measured pollutant concentrations and housing characteristics and household activities and compare them to those identified in other studies as well as with the WHO guidelines. The overwhelming use of wood and open stoves without chimneys for cooking meant that it was very difficult to determine differences between pollutant concentrations inside the dwellings as a function of cooking fuel and stove type, due to the small number of houses using alternative sources. However, PM10 concentration was significantly higher in Vientiane in households without a chimney, which is consistent with the emission of pollutants directly into the dwelling. This was expected as lower concentrations have been observed in other studies homes with vented stoves (6-8). Too few houses used stoves with chimneys in Bolikhamxay to allow such an analysis for this province. The higher PM10 concentrations in dwellings in which food was cooked using wood and in houses without chimneys, coupled with the absence of significantly higher CO concentrations in houses as a function of cooking fuel and stove type suggests that incomplete combustion was not a major contributor to pollutant concentrations in the houses. No significant difference in the pollutant concentrations in the main living area in dwellings with a separate kitchen was VOL. 45, NO. 3, 2011 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
885
FIGURE 2. Box and whisker plots of natural logarithm of the PM10 concentrations in Vientiane and Bolikhamxay.
FIGURE 3. Box and whisker plots of natural logarithm of the NO2 concentrations in combined Vientiane and Bolikhamxay as a function of a) external and b) internal walls. observed, which is in contrast to previous studies (9, 10). Also, Balakrishnan et al. observed significantly higher PM10 concentrations in dwellings without partitions compared to those with partitions and dwellings with outdoor kitchens (10). Furthermore, they observed a strong correlation between living room and kitchen concentrations for dwellings with 886
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 45, NO. 3, 2011
indoor kitchens. Likewise, the studies by Jin et al. (9, 11, 12) observed significantly lower respirable suspended particle (RSP) concentrations in the living area for dwellings with a separate kitchen and in which cooking was done with kerosene or gas compared to wood burning stoves. However, in contrast to the above, and in line with this study, Parikh
FIGURE 4. Grouped 2-D scatter plots of concentrations of PM10, NO2, and CO within households in LAO PDR as a function of a) cooking location and b) stove type. et al. observed similar PM10 concentrations, up to 2000 µg m-3, at the stove during cooking using biomass burning to those two meters from the stove (13), suggesting that indoor PM10 concentrations would be relatively uniform throughout the dwelling in houses with relatively unobstructed airflows. Our analysis suggested that while there were a number of different reasons for the variation in pollutant concentrations between the dwellings with different constructions, the strong correlation between CO and NO2 for open stoves without chimneys, cooking location, and fuel used for cooking indicates that cooking is a significant determinant of pollutant concentrations within the dwellings. This is consistent with the above referenced studies which pointed out to cooking as a significant source of emission within the dwelling; however, it also shows that there were a number of contributing factors such as fuel type, stove type, and cooking location that complicate statistical analyses of their contribution to pollutant concentrations. The contribution from heating (also using wood as fuel) was less clear, most likely due to the much lower prevalence and regularity of heating, occurring predominantly at night time during the winter, if at all, considering the still relatively high temperature during the day in winter. In contrast, some studies identified heating as a significant source of indoor emissions (14-16). Glasius et al. observed increased particle levels (PM2.5) and combustion gases in both the residential
area and at a background site in the evening and night (17). It was often impossible to distinguish the contribution of heating and cooking, since the same source is often used for both cooking and heating; however, Fischer et al. (14) concluded that indoor mean 24-h levels of particle mass and CO levels were dominated by heating as opposed to cooking sources. Several other studies have shown the relationship between seasons and indoor concentration of pollutants. For example, Edwards et al. (8) observed significantly higher PM4 concentrations in households using unvented stoves in the winter compared to the summer, due to increased fuel use and reduced ventilation during the winter. Since the results were not affected by the exclusion of houses in which smoking occurred, they concluded that particle mass emissions from cooking were substantially higher than from smoking. However, Naeher et al. (18) observed no consistent trend across all pollutants in all locations for the two seasons. Smoking is another potential source of indoor emissions, though not as significant as cooking. The frequency of cigarette smoking in the households, however, was not recorded during the sampling period, and, therefore, the correlation between the number of cigarettes smoked and pollutant concentrations could not be established. The substantially higher mean pollutant concentrations for PM10 and NO2 in dwellings in which smoking occurred VOL. 45, NO. 3, 2011 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
887
(p ) 0.058 and 0.078, respectively) are suggestive of a relationship between increased PM10 and NO2 concentration and smoking. PM10 and NO2 concentrations were correlated in both Vientiane and Bolikhamxay in households where smoking occurred, which is consistent with the suggestive p values. Correlations were observed between NO2 and CO concentrations in Bolikhamxay, both in houses of smokers and those without smokers. Other variables analyzed included the number of people in the dwelling, number of rooms, and materials used in the dwelling construction. No significant differences were observed regarding the number of people or number rooms within the dwelling, which is consistent with observations from Balakrishnan et al. (10). Considerably fewer significant correlations were observed between pollutant concentrations as a function of housing characteristics in Vientiane, suggesting that there is a wider range of sources in Vientiane. In order to reduce exposure to indoor air pollution in houses that use wood burning as a source of cooking fuel, stove design needs to be improved to prevent pollution emissions in the vicinity of the stove. For example, instead of open fire stoves, enclosed stoves with chimneys are recommended, as these would exhaust the pollution directly outside the houses. In addition, ventilation of the cooking areas should also be improved, particularly if the area is small. For example, stoves should be placed as close to windows and doors as possible, and chimneys should be installed in cooking areas. In Table 1, the mean PM10 and NO2 concentrations measured over 12 h in each province are compared with the 24 h WHO guideline values (19). While the actual 24 h concentrations in the dwellings are expected to be lower than the measured 12 h concentrations (since, in the absence of activities during the night, the concentrations are lower than during the day), the measured PM10 concentrations are well over an order of magnitude higher than the 24 h WHO health guideline value and more than six times higher than the concentration predicted by the WHO to cause a 5% increase in short-term mortality (19), which is also consistent with the concentrations observed in other developing countries (12, 20). The mean 12 h NO2 concentrations in Vientiane and Bolikhamxay were between 2.8 and 6.0 times greater than the recommended 1 h WHO guideline (200 µg m-3), respectively, with the highest concentrations over an order of magnitude higher than the guideline values. In both provinces the mean and maximum CO concentrations measured over the twelve hour period were considerably lower than the guideline values, consistent with other studies (12, 18). Also consistent with other air quality studies in developing countries, the major indoor pollutants of concern in Lao PDR are particles and NO2. A number of significant differences were observed as a function of housing characteristics and activities, fuel type, and stove type. No significant differences were observed in pollutant concentrations as a function of cooking location or in houses occupied by smokers, although p values were suggestive of a relationship in the latter case. Although cooking and the materials used in the construction of the house contributed to the highly elevated PM10 and NO2 concentrations within the houses, it is likely that outdoor sources, such as widespread burning of forests, that occur throughout Lao PDR as well as local industries, such as brick making or salt production, that use open wood burning ovens contribute to the elevated indoor concentrations. While it is likely that these outdoor sources exist, they were not measured as part of the study. Furthermore, these significant outdoor emission sources are likely to cause a substantial variation in the pollutant concentrations within the houses and could explain the lack of correlations of the pollutants 888
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 45, NO. 3, 2011
with the other expected sources, in contrast to previous studies (18, 21). Since the majority of particles emitted from combustion sources are in the ultrafine range (
