Investigation of time-resolved atmospheric conditions and indoor/outdoor particulate matter concentrations in homes with gas and biomass cook stoves in Nogales, Sonora, Mexico

Abstract

This paper reports findings from a case study designed to investigate indoor and outdoor air quality in homes near the United States–Mexico border. During the field study, size-resolved continuous particulate matter (PM) concentrations were measured in six homes, while outdoor PM was simultaneously monitored at the same location in Nogales, Sonora, Mexico, during March 14–30, 2009. The purpose of the experiment was to compare PM in homes using different fuels for cooking, gas versus biomass, and to obtain a spatial distribution of outdoor PM in a region where local sources vary significantly (e.g., highway, border crossing, unpaved roads, industry). Continuous PM data were collected every 6 seconds using a valve switching system to sample indoor and outdoor air at each home location. This paper presents the indoor PM data from each home, including the relationship between indoor and outdoor PM. The meteorological conditions associated with elevated ambient PM events in the region are also discussed. Results indicate that indoor air pollution has a strong dependence on cooking fuel, with gas stoves having hourly averaged median PM₃ concentrations in the range of 134 to 157 μg m⁻³ and biomass stoves 163 to 504 μg m⁻³. Outdoor PM also indicates a large spatial heterogeneity due to the presence of microscale sources and meteorological influences (median PM₃: 130 to 770 μg m⁻³). The former is evident in the median and range of daytime PM values (median PM₃: 250 μg m⁻³, maximum: 9411 μg m⁻³), while the meteorological influences appear to be dominant during nighttime periods (median PM₃: 251 μg m⁻³, maximum: 10,846 μg m⁻³). The atmospheric stability is quantified for three nighttime temperature inversion episodes, which were associated with an order of magnitude increase in PM10 at the regulatory monitor in Nogales, AZ (maximum increase: 12 to 474 μg m⁻³).

Implications:

Regulatory air quality standards are based on outdoor ambient air measurements. However, a large fraction of time is typically spent indoors where a variety of activities including cooking, heating, tobacco smoking, and cleaning can lead to elevated PM concentrations. This study investigates the influence of meteorology, outdoor PM, and indoor activities on indoor air pollution (IAP) levels in the United States–Mexico border region. Results indicate that cooking fuel type and meteorology greatly influence the IAP in homes, with biomass fuel use causing the largest increase in PM concentration.

Introduction

Since 1997, the body of literature linking the effects of air pollution to human health has grown substantially, with results indicating that exposure to particulate matter (PM) causes damage to the cardiovascular and respiratory systems in humans. Combined estimates from more than 100 research studies that report health effects associated with short-term exposure to PM suggest a 0.4–1.5% increase in the relative risk of all cause mortality for an incremental acute exposure of 10 μg m⁻³ for PM_2.5 or 20 μg m⁻³ for PM₁₀ (aerodynamic diameter less than 10 μm) (Pope and Dockery, 2006). In spite of these studies as well as guidelines issued by the World Health Organization (WHO, 2014) for concentrations of acute exposure to PM_2.5 (25 μg m⁻³) and PM₁₀ (50 μg m⁻³) deemed harmful to human health, current 24-hr averaged National Ambient Air Quality Standards (NAAQS) set by the U.S. Environmental Protection Agency (EPA) are 35 μg m⁻³ and 150 μg m⁻³ for PM_2.5 and PM₁₀, respectively (EPA, 2006a), and those set by the Mexican Secretariat for Environment and Natural Resources (SEMARNAT) are 65 μg m⁻³ and 120 μg m⁻³ (SEMARNAT, 2005).

An evaluation of outdoor air pollution in developing countries found that increased exposure to air pollution, specifically PM_2.5 and ozone, increases the risk of acute respiratory infections (ARI) in young children (Romieu et al., 2002). There have been studies in regions along the United States–Mexico border, where fugitive dust concentrations and automobile emissions are prevalent, designed to investigate asthma rates and respiratory complications (U.S. Geological Survey [USGS], 2006; Staniswalis et al., 2005; Stephen et al., 2003; English et al., 1998). English et al. (1998) investigated asthma hospitalization rates for two border counties in California, concluding that the county with a significantly higher poverty level had asthma hospitalization rates two to three times higher than state and federal rates. A study of respiratory symptoms in two sister cities on the United States–Mexico border, Nogales, AZ, and Nogales, Sonora (Ambos Nogales, meaning both Nogales), by Stephen et al. (2003) found that although the asthma rates on both sides of the border were similar there was an increase in cough, phlegm, and nasal symptoms in Nogales, Sonora, where the concentrations of ambient PM₁₀ were higher. Both studies concluded that asthma cases in communities with lower socioeconomic status (SES) are underreported because of the limited access to preventative health care.

Ambos Nogales is located in complex, mountainous terrain with meteorology favorable to cold air settling on the valley floor causing temperature inversions, where the air on the ground is colder than the air above (Whiteman et al., 2001). In complex terrain, this phenomenon (known as a cold air pool or CAP) can occur daily, forming in the early evening and being destroyed in the morning when surface heating erodes the stable nocturnal layer. During the wintertime, persistent CAPs can form that cause a temperature inversion with weak winds that lasts for days or even weeks, leading to an accumulation of pollutants in the atmosphere (Silcox et al., 2012). Due to a variety of reasons these thermal inversions tend to occur in intermountain regions (Seinfeld and Pandis, 2006; Zhong et al., 2001). Previous studies that investigated the seasonal and diurnal variations of PM in the United States–Mexico border region indicate that PM concentrations are elevated during wintertime and nighttime periods (Ellis, 2002; Choi, Hyde, and Fernando, 2006; Chow and Watson, 2001; Jeon et al., 2001; Kelly et al., 2010; Li et al., 2001; Jaramillo, 2005; Holmes et al., 2009). The wintertime and nighttime increases in PM are associated with these CAP episodes with stable atmospheric conditions and low wind speeds; however, very few studies (e.g., Holmes et al., 2009) provide sufficient meteorological data to quantify the effect of atmospheric stability on pollutant accumulation. Additional information on the characterization of air pollution in the United States–Mexico border region can be found in Holmes et al. (2009), and air quality concerns specific to the Ambos Nogales region are summarized in Holmes et al. (2011).

This paper builds on the work presented in Holmes et al. (2011), which describes the Nogales Indoor/Outdoor Air Quality (AQINO) field study conducted during March 14–31, 2009. The AQINO study was a case study designed to investigate indoor and outdoor PM pollution in Nogales, Sonora, Mexico, during the winter, and to determine the influence of various sources and atmospheric conditions on indoor and outdoor PM. Holmes et al. (2011) compare time-resolved indoor PM concentrations and carbon content for two homes, specifically comparing cooking versus noncooking time periods. Their results indicate that the indoor elemental carbon content was dominated by indoor sources during biomass burning, and by outdoor sources at all other times. The indoor organic carbon content was influenced by both gas and biomass stove use. A correlation between indoor home activities (i.e., cooking and cleaning) and an increase in indoor PM concentrations for the two homes was also shown.

The focus of this paper is on the time-series analysis of the continuous PM concentrations for all six homes and the micrometeorological data collected during the 2-week-long field study. Continuous time-series of indoor PM for each home are compared, and correlated to the outdoor PM measured at each home. Additionally, an investigation of the outdoor PM distribution throughout Nogales, Sonora, is presented, and the findings are related to the atmospheric conditions. Following the results, a discussion that relates these findings to the influence of local sources and meteorological conditions is given.

Experimental Methods

Participant home selection

The candidate households for the case study were selected based on the results from surveys distributed in Nogales, Sonora, six months prior to the AQINO experiment (see Holmes, 2010). The survey was designed to determine potential sources of indoor air pollution (IAP) in each home and to provide information on the health of residents. The researcher who administered the surveys noted that wealthy residents typically did not answer the door; therefore, it is estimated that the highest quintile SES residences did not participate. It is also estimated that the lowest quintile SES did not participate due to safety concerns prohibiting the surveys from being administered in particular neighborhoods. The experiment description and survey, both English and Spanish versions, were approved by the University of Utah Institutional Review Board (IRB) for a human subject study prior to the distribution of the surveys. The participants whose homes were selected for observation in the AQINO field study were compensated for the use of electricity for the duration of the experiment.

Participant homes for the case study were selected from home surveys based on two parameters, the type of cooking fuel and home location in the semi-urban area, including proximity to local sources of outdoor air pollution (i.e., unpaved roads, the main regional highway, industry). An additional criterion for the AQINO study was that only homes void of environmental tobacco smoke were selected. Six homes were selected for the study and located in five different colonias (neighborhoods): Colonia del Sol/Bella Vista (Home 1), Flores Magón/Las Torres (two participant homes, Homes 2 and 4), Lomas de la Fátima (Home 3), El Santuario/Centro (Home 5), and Villas de Nogales/Villa Sonora (Home 6); see Figure 1. Construction varied in each of the six homes and ranged from simple cinder-block construction to well-insulated, tightly sealed homes. The air exchange rate in each home was not measured and none of the homes used mechanical ventilation systems during the winter months (i.e., during the AQINO study). The two homes in Flores Magón (Homes 2 and 4) varied in distance from a major pollution source, the main highway, built to reroute the majority of diesel truck traffic from the downtown area of Nogales through the western edge of the city (solid black line in Figure 1). Two homes (Homes 3 and 5) were located near the border crossing (gray horizontal line in Figure 1) and train tracks near the center of town (black dashed line in Figure 1). The elevation profile for the map cross section corresponding to each home location is also shown in Figure 1. Home elevation and location within the basin are important factors when investigating the meteorological phenomena associated with wintertime temperature inversions and cold air pooling, as discussed in the Introduction.

Figure 1. Left: Satellite image showing locations of the six sampling homes in Nogales, Sonora, Mexico (

) and sonic anemometer/ADEQ sampling (

). Railroad tracks (

) through the center of town and Highway (

) located in western Nogales. Right: Elevation (in meters above sea-level) profiles of east–west map cross section corresponding to each sampling location (indicated by marker).

In addition to home location, cooking fuel was a critical selection criterion because the primary objective of the AQINO study was to compare cooking fuels used in homes on the United States–Mexico border and to determine the impact of biomass (wood) burning on IAP. Of the six homes, two were equipped with gas and biomass stoves (Homes 4 and 1), one an improved biomass stove (Home 2), and three had gas stoves (Homes 3, 5, and 6). Anthropologist Dr. Diane Austin, from the University of Arizona, has collaborators in Nogales, Sonora, who supply residents with energy-efficient stoves as part of her research (Austin et al., 2006). The improved biomass stove home (Home 2) was located through her collaborators. The homes with biomass stoves (Homes 1, 2, and 4) used wood to varying extents for cooking and heating based on economic considerations and availability; therefore, the biomass results from the AQINO experiment cannot be generalized and are home specific. Examples of stoves from AQINO are shown in Figure 2, with a gas stove, biomass stove, and improved biomass stove pictured.

Figure 2. Stoves from the AQINO field study: (left) gas stove, (middle) biomass stove, and (right) improved biomass stove.

Continuous PM concentration measurements

To obtain data at a high enough temporal resolution to investigate the transient behavior of indoor PM, optical particle counters (OPC), GRIMM Technologies, Inc. (Douglasville, GA), model 1.108 aerosol monitors were used to monitor continuous size segregated PM concentrations. The GRIMM monitor measures a size distribution of PM mass concentrations in 15 size bins from PM_0.3 to PM₂₀ by converting number counts to mass concentrations using a constant particle density (2.8 g cm⁻³ for model 1.108). Previous PM mass concentrations from the AQINO study reported in Holmes et al. (2011) were the values directly measured from the OPC. However, the results presented in this paper were adjusted by changing the particle density from 2.8 g cm⁻³ to 1.7 g cm⁻³, the GRIMM established “urban environment” factor (Burkart et al., 2010; Grimm and Eatough, 2009). In addition to particle density OPC measurements are sensitive to different sources due to the change in shape of the particles and index of refraction. During the AQINO study one OPC device was used at each site and therefore the indoor and outdoor concentrations at each home location can be compared. The data from the GRIMM monitor can be summed in postprocessing to obtain PM₁₀ and PM₃ concentrations for comparison with PM₁₀ and PM_2.5 measurements from a regulatory monitoring network. Previous work has shown that when in environmental mode (measurements of PM₁₀, PM_2.5 and PM₁) the GRIMM measures PM₁₀ and PM_2.5 concentrations comparable to PM concentrations measured using a EPA federal reference method during periods with low to moderate relative humidity (RH; i.e., RH < 75%) (Grover et al., 2006; Wilson et al., 2006). The temperature and RH during the AQINO field study are shown in Figure 3; the RH was less than 55% hence there should be minimal impact on the OPC PM measurements due to RH. All of the GRIMM monitors used in the AQINO field study were factory calibrated by GRIMM Technologies, Inc., one week before the experiment started. Additionally, all of the GRIMM monitors were collocated in the field prior to being deployed in homes to verify that they were all reading similar PM concentrations.

Figure 3. Relative humidity (RH) and temperature (T) measured at the Nogales, AZ airport station (KOLS) for the duration of the AQINO field study. Tic marks are every 24 hr and denote 00:00.

One GRIMM monitor was placed in each of the six homes to enable simultaneous collection of indoor and outdoor PM at a sampling rate of 1/6 Hz, with a valve switching system to monitor indoor and outdoor air on a set sampling schedule (3.175 mm inner diameter Tygon tubing and sampling flow rate of 1.2 L per minute). All of the GRIMM monitors were set to sample on the same schedule, with the valve switching system set up to monitor indoor air for 12 min and outdoor air for 8 min (time crystal with a maximum time error of 0.002%). The following time schedule was used for the AQINO study: indoor air: 0, 20, and 40 min after the hour; and outdoor air: 12, 32, and 52 min after the hour. The longer indoor sampling period was selected under the assumption that the outdoor concentrations would change on a slower time scale than the indoor concentrations, due to the short time duration often associated with indoor activities. While care was taken to keep sampling line lengths short to minimize line losses of PM, postexperiment calculations using the Aerocal program (Paul A. Baron, 3-Nov-01 version) indicated line losses of 1.1–1.7%, 4.1–6.1%, 8.8–13%, 39–52%, and 63–77% for PM₁, PM₂, PM₃, PM₇, and PM₁₀. Therefore, the results and discussion will exclude PM data for size bins larger than PM₃; hence it is not possible to investigate the coarse to fine ratios for the AQINO study. A more detailed description of the continuous PM monitoring and valve switching system, including postprocessing steps for the data, is given in Holmes et al. (2011).

Table 1 shows home descriptions for the AQINO field study homes, including the sampling line lengths for indoor and outdoor air. At each home, the tubing for indoor sampling was placed in the “living-room” area, the room closest to the kitchen where people in the home generally congregate for social gatherings. This location was chosen so the measured PM would be representative of resident’s exposures. The outdoor sampling inlets were placed at respirable height to measure ambient PM indicative of the resident’s outdoor exposure. Caution was exercised to keep the inlet away from major PM sources (i.e., direct tailpipe emissions, children’s play area, outdoor pets) and from direct indoor sources; however, cross-contamination of indoor sources of PM to the outdoor measurement inlet was unavoidable. The study participants in each home were asked to keep an activity log to document cooking and cleaning activities every 30 min. Additional experimental constraints and sampling inlet considerations are given in Holmes et al. (2011).

Table 1. Home descriptions for the AQINO field study homes

Home	Occupants	Pets	Fuel	Roads	Garage	Flooring	Indoor tubing	Outdoor tubing	Stove distance
1	Adults: 3 Children: 3	None	C: Wood/Gas H: Wood	N: Unpaved H: Paved	None	Tile/cement	L: 1.22 m H: 0.737 m	L: 1.83 m H: 1.397 m	B: 7.3 m G: 7.6 m
2	Adults: 1 Children: 0	O: Dog, other	C: Wood (‘I’) H: None	N: Unpaved H: Unpaved	None	Cement, dusty	L: 1.52 m H: 0.927 m	L: 1.83 m H: 1.245 m	B: 2.6 m
3	Adults: 3 Children: 0	None	C: Gas H: None	N: Paved H: Unpaved	Attached	Tile	L: 1.22 m H: 0.838 m	L: 1.30 m H: 3.42 m	G: 6.0 m
4	Adults: 1 Children: 3	O: Dog I: Birds	C: Wood/Gas H: Wood	N: Unpaved H: Unpaved	Attached	Tile, dusty	L: 1.22 m H: 0.914 m	L: 1.83 m H: 1.473 m	B: 1.75 m G: 7.9 m
5	Adults: 1 Children: 1	None	C: Gas H: None	N: Paved H: Paved, Bus	Attached	Tile	L: 1.22 m H: 1.524 m	L: 1.22 m H: 1.194 m	G: 4.47 m
6	Adults: 3 Children: 0	O: Dogs	C: Gas H: Gas	N: Paved H: Paved, Bus	None	Tile	L: 1.22 m H: 0.864 m	L: 1.83 m H: 1.600 m	G: 6.5 m

Notes: Pets: outdoor (O), indoor (I); Fuel: cooking (C), heating (H), improved biomass stove (‘I’); Roads: neighborhood (N), road in front of the home (H); Indoor/Outdoor tubing: length (L), height (H); and Stove Distance: gas stove (G), biomass stove (B), distance from indoor sampling inlet.

Atmospheric measurements

One micrometeorology station was installed next to the Arizona Department of Environmental Quality (ADEQ) monitoring equipment on top of the Post Office in Nogales, Arizona (ADEQ in Figure 1). It was equipped with one Campbell Scientific, Inc. (Logan, UT), CSAT-3 three-dimensional sonic anemometer/thermometer (Sonic). While not ideal, this location was chosen for the safety of the equipment and because it was easily accessible. Care was taken to place the Sonic in a location where the rooftop effects would be minimal. The anemometer was set up to measure three components of wind speed and sonic temperature at a frequency of 10 Hz, 14 m above the ground (4 m above the rooftop). The CSAT-3 allowed for the measurement of three-dimensional wind speed and sonic temperature, which are necessary to calculate the turbulence parameters that influence the transport of particulate matter in the atmosphere.

The ADEQ routinely monitors PM₁₀ and PM_2.5 concentrations (8.24 m and 8.37 m above ground, respectively) as part of its regulatory monitoring network for criteria air pollutants in Nogales, AZ. The AQEQ uses a federally approved regulatory monitoring device (EPA, 2006b) for PM concentrations, a beta-attenuation mass (BAM) monitor (Chung et al., 2001; Wilson et al., 2006). The ADEQ also uses an anemometer at this location to collect hourly wind speed and wind direction data and the BAM monitoring devices measure temperature, so meteorology data were available to supplement our sonic anemometer data when the CSAT-3 was not running.

Results

Continuous PM data were collected every 6 sec in each home with a valve switching system to sample indoor and outdoor air. Postprocessing steps were necessary to obtain averages of the indoor (11 min) and outdoor (7 min) samples (Holmes et al., 2011). Hourly averages were calculated from these shorter term time-series for both indoor and outdoor air samples. Home schedule information presented in this section came from activity logs recorded by residents; therefore, in some cases interpretation was required for the activity duration. To evaluate local atmospheric conditions (micrometeorology), sonic data collected at 10 Hz were used, and postprocessing steps were necessary to calculate the mean and turbulent quantities. Mean and averaged quantities were calculated using a 30 min block average, while the fluctuating components were obtained by removing the mean with a linear detrend using a 15 min window (Metzger and Holmes, 2008).

Continuous PM concentration measurements

The mean, median, standard deviation, minimum and maximum concentrations of PM₃ and PM₁ for hourly averaged data at each home are shown in Table 2, including indoor and outdoor concentrations.

From the examination of outdoor PM, it is evident that a large spatial heterogeneity of ambient PM exists throughout the region. For example, even though Homes 2 and 4 are <280 m apart, the dependence of PM concentration on microscale sources is evident from the outdoor PM values in the table (discussed further in the Discussion section). However, Homes 3 and 5 are <770 m apart and have similar median PM values with a difference of 10% and 9.3% for PM₃ and PM₁. The two lowest median outdoor PM concentrations occur at Homes 4 and 6, but due to different physical mechanisms that can be investigated using the minimum and maximum hourly PM values. At Home 4 the maximum hourly PM was 3480 and 2358 μg m⁻³ for PM₃ and PM₁, representative of highly acute elevated PM concentrations. At Home 6, the maximum PM was much lower (877 and 594 for PM₃ and PM₁) with higher minimum values than Home 4 and less variability, indicated by a lower standard deviation at Home 6. This suggests that while the long-term averaged PM concentrations may be similar, there are fewer acute PM episodes.

The values from Table 2 can be related to Figure 4, which shows the hourly averaged indoor and outdoor PM₃ measured at each home starting on March 14, 2009, at 00:00 and ending on April 1, 2009, at 00:00 on a logarithmic ordinate axis. It appears that most of the peak concentrations for indoor PM are correlated to outdoor peaks, but closer investigation reveals that there are several indoor peaks that do not coincide with high outdoor PM. To illustrate the relationship between hourly PM₃ concentrations and the average range of values measured during the experiment, Figure 4 has gray dashed lines indicating the minimum and maximum median outdoor PM₃ values for all six homes during the AQINO study (also listed in Table 2).

Table 2. Indoor and outdoor mean, median, standard deviation (σ), minimum (Min) and maximum (Max) values calculated from hourly PM₃ and PM₁ concentrations (μg m⁻³) for the six home locations in the AQINO field study; and in the last column the Pearson correlation coefficient (R) for indoor and outdoor PM

		Indoor						Outdoor
		Mean	Median	σ	Min	Max	Time of max	Mean	Median	σ	Min	Max	Time of max	R
Home 1: Biomass	PM3	813	501	983	38	7194	19 March, 21:00	1343	770	1523	91	10846	16 March, 22:00	0.70
	PM1	573	345	695	27	4860		909	519	1029	63	7299		0.68
Home 2: Improved	PM3	888	504	1116	72	8551	19 March, 08:00	854	464	1120	49	9411	19 March, 08:00	0.79
biomass	PM1	603	351	756	50	5774		577	314	755	34	6345		0.79
Home 3: Gas	PM3	215	134	360	13	4361	28 March, 14:00	309	216	298	33	2029	20 March, 09:00	0.15
	PM1	152	93	246	10	2928		210	147	201	23	1371		0.16
Home 4: Biomass	PM3	342	163	572	17	4981	17 March, 11:00	198	130	283	28	3480	19 March, 09:00	0.42
	PM1	258	113	484	12	4661		138	89	196	20	2358		0.44
Home 5: Gas	PM3	247	157	335	17	4965	18 March, 11:00	296	195	320	23	3481	15 March, 19:00	0.30
	PM1	173	109	240	13	3621		203	134	217	17	2341		0.29
Home 6: Gas	PM3	213	148	236	20	1814	18 March, 08:00	183	140	135	42	877	19 March, 09:00	0.46
	PM1	151	102	173	14	1448		127	95	93	29	594		0.46

Notes: All Pearson correlation coefficients have p values <0.01.

Figure 4. Indoor and outdoor hourly PM₃ concentrations during the Nogales Indoor/Outdoor Air Quality (AQINO) study at H1: biomass, H2: improved biomass, H3: gas, H4: biomass, H5: gas, and H6: gas. Tic marks are every 24 hr and denote 00:00; gray dashed lines indicate the minimum and maximum median outdoor PM₃ for all homes from Table 2.

Note the high levels of PM₃ in Figure 4 for Homes 1 and 2, which appear to be dominated by the exceedingly high levels of outdoor PM₃ at those locations. At Home 1, there was active road construction taking place throughout the duration of the AQINO study, which resulted in increased traffic on unpaved roads and emissions from road paving. This is evident in the decrease of outdoor PM on Sunday, March 22, indicating that road construction is likely the source of increased outdoor PM. At Home 2, there were several outdoor sources that may have led to elevated ambient PM levels: unpaved roads, close proximity to a local bus route and a primary road in the colonia, several outdoor farm type animals, and shared property with a restaurant having several visitors throughout the day. Similar to Home 1, the outdoor PM at Home 2 also decreased on Sunday, March 22, indicating that reduction in bus route frequency or fewer restaurant visitors may have resulted in less PM.

The lowest levels of indoor PM₃ were measured at Home 6, which is likely a result of the cooking fuel (gas) and the low levels of outdoor PM. With the exception of a local bus route, this location was the furthest from local PM sources and at an elevation approximately 122 m above the valley floor. Both of these factors contribute to the second lowest outdoor PM being measured at Home 6 during the AQINO field study (Table 2).

Among the three biomass stove homes, Home 4 had considerably lower levels of indoor PM than Homes 1 and 2. Upon investigation of the outdoor PM at these homes, it appears that the lower levels at Home 4 are due to the outdoor PM concentrations, where the median outdoor PM₃ at Home 4 was 17% of the Home 1 median concentration and 28% of Home 2. Based on the data in Table 2, the largest difference in indoor concentrations for gas versus biomass stoves occurs in the maximum hourly PM₁.

To investigate the relationship between the indoor and outdoor PM levels the indoor/outdoor (I/O) ratio is calculated for the duration of the AQINO study and plotted as an ensemble average versus time of day. Figure 5 shows the median (black circle), 25th and 75th percentile (gray shaded box), the data within 1.5 times the interquartile (gray lines), and the outliers (x). Note that to plot data from each home with the same I/O range on the y-axis some outliers are missing from the plots. An important initial observation is the low I/O ratio for two biomass stove homes Home 1 and Home 2, while the I/O for Home 6 (gas stove) is greater than 1. This is due to the influence of the outdoor PM on the I/O, where an increase in indoor sources should correlate to an I/O greater than 1. Factors contributing to this low ratio for Homes 1 and 2 include outdoor air penetrating indoors because the homes are not sealed (Table 1). The opposite is true at Homes 3 and 6 where the homes are well sealed and indoor sources include regular cooking and cleaning, resulting in I/O ratios greater than 1. The values shown here must be interpreted with caution because simply evaluating the I/O does not provide an accurate representation of the IAP for homes in the AQINO study. However, useful information can be obtained from these plots. Particularly, the diurnal evolution of the IAP is visible in the homes that maintained a regular daily pattern (Homes 3, 4, 5, and 6). This is particularly present at Home 3 where lunch was cooked on weekday afternoons, Home 4 where wood burning for heat and cooking occurred throughout the day, Home 5 where cleaning activities started at 10:00 or 11:00 several days per week, and Home 6 where breakfast preparation and home heating were done daily starting at 06:00.

Figure 5. Indoor/outdoor PM₃ ratios for the six homes in the AQINO field study: median (black circle), 25th and 75th percentile (gray shaded box), data within 1.5 times the interquartile (gray lines), and outliers (x). Solid black line indicates unity.

Micrometeorology

Figures 6 and 7 show select time periods during the AQINO study chosen to investigate the correlation between outdoor PM and atmospheric conditions. Specifically, the relationship between atmospheric stability and PM accumulation using the Monin–Obuhov (M-O) stability parameter z/L, where z is the measurement height (m) and L is M-O length given by , and T_o, T_s’, g, and κ are the mean absolute surface temperature, turbulent fluctuations of virtual (sonic) temperature, gravitational constant, and the von Karman constant (0.4) (Seinfeld and Pandis, 2006; Stull, 1988). The M-O length is the height (m) where shear and buoyancy-driven turbulence are both important, with mechanical and buoyant mechanisms quantified by the friction velocity (m sec⁻¹), where u’, v’, and w’ are the turbulent fluctuations of the streamwise, transverse, and wall normal velocities (Stull, 1988), and the buoyancy flux (m² sec⁻³). Positive values of z/L indicate a stable atmospheric boundary layer (ABL) and negative an unstable ABL, while zero is a neutral ABL (Mahrt, 1998). A typical stable ABL situation occurs during the presence of weak winds and increasing potential temperatures with height, or conditions that exist during CAP episodes (see Introduction). These stable conditions are known to inhibit the mixing and dilution of pollutants (Fernando, 2010).

Figure 6. PM₃ and meteorological data from March 14–16, 2009, during the AQINO study: (a) mean horizontal wind speed, (b) horizontal wind direction (0 degrees from the north), (c) Monin-Obukhov stability parameter, and (d) hourly averaged outdoor PM concentration from GRIMM monitors at all six homes and the ADEQ BAM monitor (logarithmic PM axis).

Figure 7. PM₃ and meteorological data from March 28–29, 2009, during the AQINO study: (a) mean horizontal wind speed, (b) horizontal wind direction (0 degrees from the north), (c) Monin-Obukhov stability parameter, and (d) hourly averaged outdoor PM concentration from GRIMM monitors at all six homes and the ADEQ BAM monitor (logarithmic PM axis).

In Figure 6, data from March 14–16, 2009, are plotted showing two nighttime periods with elevated PM levels coinciding with stable ABL time periods. The first begins just after sunset (approximately 18:30) on March 14 at 19:00 when z/L becomes positive (indicated by first gray box), and the maximum PM at each location occurs between 20:00 and 21:00. At 22:00 the wind speed increased and the ABL becomes temporarily unstable again, leading to a decrease in PM, similar to a washout event (Sriramasamudram, 2009). This shows the importance of quantitatively determining ABL stability and not just relying on the assumption that nighttime periods have a stable ABL. The ABL becomes stable again at 00:00 on March 15 and the PM levels increase and remain elevated until 12:00, when the wind speeds begin to increase above 1 m sec⁻¹. The second nighttime period shown in the figure starts at 19:00 on March 15 (second gray box), and depicts a typical CAP situation. The nighttime ABL is stable and the wind speeds remain below 1 m sec⁻¹, leading to an accumulation of PM until the morning when the CAP is destroyed by surface heating.

PM data measured at the highest elevation, Homes 2 and 4, were not available during the time period shown in Figure 6; therefore, another time period with data from Home 4 was plotted for March 28–29, in Figure 7. With the exception of the ADEQ location, the PM concentrations at all home locations are a similar order of magnitude during the day from 07:00 on March 28 until 18:00 when the ABL stability changes (indicated by gray box). The PM values at Home 4 and at the ADEQ locations are particularly interesting to investigate for this time period because they change considerably during the transition from an unstable to stable ABL. The ADEQ PM₁₀ increases from 20 to 640 μg m⁻³, while at Home 4 the opposite occurs with PM₃ decreasing from a daytime maximum of 802 μg m⁻³ down to 51 μg m⁻³. The two locations represent the lowest (ADEQ) and highest (Home 4) elevations of the monitoring locations in the AQINO study with an elevation difference of approximately 161 m.

Discussion

As mentioned in the Methods section, all PM sampling inlets in the AQINO study were placed at locations to monitor concentrations indicative of resident’s exposure, both for outdoor ambient PM and for indoor PM. Performing a one-way analysis of variance (ANOVA) test in MATLAB (anova1) on the natural log of PM₃ concentrations and grouping by location suggests that the samples come from different distributions for indoor and outdoor PM₃ (p < 0.001). This indicates a large spatial heterogeneity for the outdoor PM concentrations, which, based on previous work in the United States–Mexico border region (Holmes et al., 2009; Chow and Watson, 2001; Jeon et al., 2001; Li et al., 2001; Jaramillo, 2005), is to be expected. Additionally, there are large differences in the magnitude of PM concentrations measured at each home location, a difference that requires the data to be plotted on a logarithmic scale for visual comparison (see Figures 4, 6, and 7). This is most likely due to the presence of extremely local sources for the PM concentrations, or what is referred to as microscale sources in the Results section.

While the sampling inlets were placed as far away as possible from sources of PM the influence of microscale sources on the PM concentrations is evident in the measurement data. It is important to distinguish the difference between these microscale sources and the sources that are considered to be local, that is, train, highway, border crossing, and industry. Examples of microscale sources (Figure 8) include tail-pipe emissions from parked vehicles, outdoor cigarette smoking, pets and other animals, areas with increased pedestrian traffic on ground with loose dirt, neighborhood trash burning, recycling of indoor biomass burning, and so on. The presence of these source types is common in this region because of the accelerated growth since the 1970s, putting a strain on urban planning with many residents having limited access to public utilities (NAQTF, 2005; Peña, 2005). The concentrations measured during the AQINO experiment are more representative of PM from these sources due to the lower height of the sampling lines, while the ADEQ regulatory monitoring network sensors are placed at much higher heights above the ground (>8 m, vs. 1.2 to 3.4 m for the AQINO study). The primary importance of the microscale sources in this paper is due to the large vertical and horizontal PM gradients near the source. This is shown in Veranth, Pardyjak, and Seshadri (2003), where the vertical gradient of PM due to fugitive dust is greatest in the first 5 m above the ground (source). They also indicate that the vertical distribution of the gradient evolves downwind of the source, becoming more well-mixed, and that the dust flux is largest closest to the source (3 m versus 95 m downwind in their study) as a result of near-source deposition. The ADEQ monitor is located at a height greater than 5 m and farther than 2 km from the unpaved roads, neighborhood trash burning, and other microscale sources in Nogales, Sonora. Hence, the concentrations measured at the regulatory monitoring site are not representative of the microscale sources. Rather, they are representative of the regional and local sources of PM in Ambos Nogales.

Figure 8. Examples of microscale sources in the AQINO field study: (left) Neighborhood trash burning, (middle) Loose dirt at a home entrance, with a pet dog, (right) Unpaved roads. This photo also depicts the elevation difference for homes in the region.

The results presented here show the importance of accounting for spatial variability in PM when assessing adverse health effects linked to air pollution, due to influences from both micrometeorological factors and local sources of air pollution. In addition to the horizontal spatial heterogeneity of PM, this study shows the importance of elevation above a valley floor in complex terrain due to the influence of ABL stability on PM mixing. Figure 9 shows a boxplot of the ADEQ PM₁₀ concentrations binned according to wind speed for the duration of the AQINO study. While the anemometer used at the ADEQ is not accurate for wind speeds less than 1 m sec⁻¹, the two lowest wind speed bins (0.5 and 1 m sec⁻¹) can both be considered calm winds. The plot shows that the highest PM is associated with periods of calm winds, which is in accordance with the results shown in the previous section for nighttime periods with increased atmospheric stability.

Figure 9. (top) ADEQ BAM PM₁₀ box plot versus wind speed (ADEQ anemometer); and (bottom) wind speed histogram.

Additionally, indicators for indoor exposure such as the I/O ratio that is commonly used to relate the influence of IAP on health need to be interpreted with caution, as the values may not reflect the actual indoor air exposures. One unique air pollution indicator cannot be used as an input to the statistical analysis performed in epidemiological research for exposure estimates. Regarding air pollution metrics and exposure estimates, this work also indicates that even in a relatively small region (9 km by 6 km) there is a large variability in human exposure to air pollution. This is particularly the case for homes utilizing biomass fuels for cooking, where the indoor PM concentrations are higher than for homes with gas stoves. Using the ANOVA test with gas and biomass grouping for all homes shows higher PM₃ concentrations for biomass fuel and that the samples come from different distributions (p < 0.001).

Another interesting aspect of the AQINO study was the use of human participants. In a discussion at an international symposium on recommendations for the study of SES and air pollution health effects (Bell et al., 2005), community involvement was presented as a key feature for a successful study. The benefits of community involvement include better cooperation of participants in the study, local knowledge, enhanced understanding of results and related health issues, improved public perception of the research community, and instigating an interest in the use of the results by local policymakers. In the current study, home schedule documentation for Homes 3 and 4 appears to be more complete than for the other homes; see the indoor home activity plots presented in Holmes et al. (2011). It is hypothesized that this may be due to the increased presence of the researchers in those two homes because they were selected for the carbon collection instrumentation and required several additional home visits.

Summary

To visualize and quantify the extent of spatial variability in PM for the homes in the AQINO field study, a boxplot of PM concentrations versus the approximate atmospheric stability (estimated by time of day based on results from Figures 6 and 7) is shown in Figure 10. During the AQINO study, the maximum PM concentrations occur outdoors for both size ranges, and it is significant to note that the maximum occurs during periods with a stable atmospheric boundary layer (ABL). For the indoor PM, the maximum corresponds to an unstable ABL, which for Figure 10 is estimated as the time period between 05:00 and 18:00, and relates to daytime when the majority of indoor activity occurs. Therefore, the highest median and largest range of indoor PM occur during the unstable ABL periods due to the influence of the increased human activity indoors (median PM₃ [25th and 75th percentile]: 178 μg m⁻³ [99 and 420], 108 μg m⁻³ [48 and 186], and 238 μg m⁻³ [116 and 501] for stable, neutral, and unstable ABL).

Figure 10. Box plots summarizing the spatial variability of hourly PM data as a function of approximate atmospheric stability for all six homes in the AQINO study: (top) Outdoor PM; (bottom) Indoor PM. Total data (T): All, Stable (S): 21:00–02:00, Neutral (N): 03:00–04:00 and 19:00–20:00, Unstable (U): 05:00–18:00. Outliers are not included; therefore, the value of the maximum concentration is indicated on the plot.

It is hypothesized that the local meteorology and terrain (e.g., cold air pools and elevation changes) lead to periods with elevated PM and an increased spatial variability in the Nogales basin. This is evident during the stable ABL periods, where the maximum outdoor concentrations occur and also coincide with the maximum median for the AQINO field study (median PM₃ [25th and 75th percentile]: 251 μg m⁻³ [121 and 534], 201 μg m⁻³ [97 and 414], and 250 μg m⁻³ [130 and 543] for stable, neutral, and unstable ABL). The stable ABL corresponds to nighttime periods, from 21:00 to 02:00, and therefore the impact of microscale sources is expected to be minimal because people are sleeping. Hence, the PM behavior is governed by regional air pollution sources, meteorology, and terrain influences. During the daytime, the median outdoor PM is elevated due to the presence of PM sources that correlate to increased anthropogenic activity. It is recommended that future experiments in similar terrain with meteorology favorable to the formation of cold air pools utilize models to design the experiment, so sensors can be placed in the optimal location to capture the spatial variability of the air pollution.

Human activity patterns that are correlated to day of the week also appear to impact the PM variability. For example, low PM concentrations on Sunday due to decreased activity and reduced impact from microscale and regional sources such as the border crossing. Figure 10 shows that for indoor PM, the neutral ABL periods have the lowest PM. This is because those time periods relate to periods with minimal indoor activity: 03:00 to 04:00 because people are sleeping and 19:00 to 20:00 because it is customary in this region to eat the final meal of the day later into the evening, around 21:00 or 22:00. Overall, the results from the AQINO study indicate that there is spatial and temporal variability of indoor and outdoor PM in the region, leading to large differences in personal exposures for residents in Nogales, Sonora, Mexico.

Acknowledgments

The AQINO field study was completed with assistance from University of Utah students Scott Speckart and Daniel Alexander. In addition, the authors are grateful to Dr. Diane Austin from University of Arizona and her colleagues in Nogales, Sonora, for assistance in finding study participants and Dr. Rodney Larson at the University of Utah for assistance with the survey; to Dr. John Veranth and David Wagner from University of Utah and Dr. David L. Johnson from University of Oklahoma for loaning equipment for use during the experiment; and to Bill Roe from Grimm Technologies, Inc., for calibrating instruments on short notice. Also, they are grateful to the Arizona Department of Environmental Quality for providing data from the Nogales, AZ, monitoring site.

Funding

The authors acknowledge the support of the Southwestern Consortium for Environmental Research and Policy (SCERP).

Additional information

Notes on contributors

Heather A. Holmes

Heather A. Holmes is an assistant professor in the Department of Physics at the University of Nevada, Reno.

Eric R. Pardyjak

Eric R. Pardyjak is a professor in the Department of Mechanical Engineering at the University of Utah.