Personal particulate matter exposures and locations of students in four neighborhoods in Accra, Ghana


Air pollution exposure and places where the exposures occur may differ in cities in the developing world compared with high-income countries. Our aim was to measure personal fine particulate matter (PM2.5) exposure of students in neighborhoods of varying socioeconomic status in Accra, Ghana, and to quantify the main predictors of exposure. We measured 24-hour PM2.5 exposure of 56 students from eight schools in four neighborhoods. PM2.5 was measured both gravimetrically and continuously, with time-matched global positioning system coordinates. We collected data on determinants of exposure, such as distances of homes and schools from main roads and fuel used for cooking at their home or in the area of residence/school. The association of PM2.5 exposure with sources was estimated using linear mixed-effects models. Personal PM2.5 exposures ranged from less than 10 μg/m3 to more than 150 μg/m3 (mean 56 μg/m3). Girls had higher exposure than boys (67 vs 44 μg/m3; P-value=0.001). Exposure was inversely associated with distance of home or school to main roads, but the associations were not statistically significant in the multivariate model. Use of biomass fuels in the area where the school was located was also associated with higher exposure, as was household’s own biomass use. Paved schoolyard surface was associated with lower exposure. School locations in relation to major roads, materials of school ground surfaces, and biomass use in the area around schools may be important determinants of air pollution exposure.


Ambient air pollution is responsible for an estimated 3.2 million annual deaths worldwide.1 The highest levels of air pollution in the world now occur in cities in Asia, the Middle East, and Africa.2 Although there are some similarities in the sources and spatial patterns of air pollution in cities in developing countries compared with those in high-income countries (e.g., traffic-related pollution), there are also significant differences (e.g., biomass use for cooking).3, 4, 5, 6, 7, 8, 9, 10 For example, biomass use, which is common in both rural and urban areas,11,12 accounts for between one third and one half of fine particulate matter (PM2.5; particles below 2.5 micrometers in aerodynamic diameter) pollution in different neighborhoods of Accra, Ghana, with another 10–30% due to traffic and road dust.10 These factors are also important determinants of the spatial patterns of air pollution between and within neighborhoods, and of its concentrations in the household environment.5,8,13 Time–location–activity patterns in the developing world are also different from high-income countries due to differences in the built environment, transportation, and sociocultural factors. As a result, both the levels of air pollution exposure and places where the highest exposures occur may differ in cities in the developing world compared with high-income countries.

A number of studies have examined personal exposure to air pollutants for school children in high-income countries, with an emphasis on the role of traffic-related pollution.14, 15, 16, 17, 18 Although a few personal PM exposure studies exist in cities of developing countries,19,20 the great majority of exposure studies are from rural areas, and assessed the role of household cooking and heating as a source of exposure. To the best of our knowledge, only rural exposure studies exist in sub-Saharan Africa21,22 or studies that have measured school pollution but not personal exposure,23,24 despite the fact that the urban population in this region is growing faster than any other in the world.25

In this paper, we report personal PM2.5 exposures of students in four neighborhoods in Accra, Ghana. In addition to being among the few studies of personal exposure to air pollution in cities of developing countries, the paper makes a number of contributions. First, by simultaneously collecting time-stamped continuous data on location and exposure, we are able to identify time and places where exposure tends to be high. Second, the study covers poor, middle-income, and wealthy neighborhoods, allowing for an assessment of social inequalities in exposure. Third, we collected data on household and community determinants of exposure and assessed their associations with measured exposures. Fourth, we examined the correlation between personal exposure and neighborhood ambient concentrations, which provides information on the measurement error when ambient monitoring data are used instead of personal exposure in epidemiological studies in these settings.


This study was approved by the Institutional Review Boards of the Harvard School of Public Health and the Noguchi Memorial Institute for Medical Research at the University of Ghana.

Study Location

Our study was conducted in four neighborhoods in Accra, Ghana: James Town/Usher Town (JT), Asylum Down (AD), Nima (NM) and East Legon (EL) (Figure 1). The study neighborhoods were selected such that they lay on a nearly straight line from the coast to the northern boundaries of the Accra Metropolitan Area, and had varying socioeconomic status (SES) based on data from the 2000 Population and Housing Census. JT and NM are densely populated low-income communities where most residents use biomass for cooking at home and for cooking food to sell on the street.13 AD is a middle-class neighborhood, and EL is an upper-class, sparsely populated residential neighborhood where most families live on large plots of land in modern low-rise homes. Fewer people use biomass fuels in AD and EL than in JT and NM. Accra’s Central Ring Road borders AD. Smoking is very uncommon in Ghana, with prevalence below 10% nationally and in our sample.10,26 Those who smoke commonly smoke outside the house.

Figure 1

Study neighborhoods, schools, and homes. The colored biomass use map is delimited by enumeration areas (EA); each EA has approximately the same population size; hence, the area of an EA is inversely related to population density.

Study Design

Between January and August 2008, we measured personal PM2.5 exposure and locations of 56 students who were between the ages of 10 and 17 years. The students were from two public junior high schools in each study neighborhood. The schools were visited prior to the start of measurement to explain the aims of the study and request participation; at least 90% of students in all schools agreed to participate. The study subjects were selected by the school teachers, with approximately the same number of male and female students. Subjects were then enrolled following written consent from their parents. A minimum of six students from each school participated in the study. The locations of the schools and the students’ residences are shown in Figure 1. Exposures were measured using portable PM monitors, and locations were recorded using global positioning system (GPS) devices. Both instruments were placed in backpacks worn by the subjects. The subjects were asked to keep the backpack as close to them as possible at all times.

We conducted measurements for each student over a 24-h period. The measurements were performed only on weekdays when the students attended school. A random one third (20 out of 56) of our subjects had repeated measurements, with an average of 2.5 measurements per subject. The repeated samples were collected the day after the first measurement, except when the first measurement was done on a Friday or before a holiday. We used structured questionnaires to collect information about the students’ activities throughout the day, time spent near household cooking fire, the fuel used by the family for cooking and kitchen characteristics.

PM Measurement Methods

We used a combination of integrated gravimetric and continuous real-time monitors to measure PM2.5 exposure.

Integrated gravimetric PM2.5

We used external elutriators connected to Personal Exposure Monitors (PEMs) (Harvard School of Public Health, HSPH, Boston, MA)27 with a D50 of 2.5 μm (aerodynamic diameter) at 1.8 liters per minute (l pm) (±10%) and an internal level greased impaction surface. Inside the PEMs, PTFE filters with ring (Pall Life Sciences, Teflo, 0.2 μm pore size, 37 mm diameter) were back-supported by Whatman drain discs. PEMs were connected by Tygon PVC tubing to a Casella Apex Lite personal sampling pump (Casella USA, Amherst, NH) drawing air at 1.8 lpm. To conserve battery life, pumps were programmed to draw air for 1 out of every 6 min for a total of 290 min over the 24-h sampling period. Air flow rates were checked at the beginning and end of each sampling period using a calibrated rotameter. The PEM and pump were placed inside the backpack with the elutriator nozzle protruding through an opening in the backpack.

All filters were weighed pre- and post sampling on a Mettler Toledo MT5 microbalance maintained at HSPH Laboratory, after being conditioned in a temperature- and relative humidity (RH)-controlled environment (20.5±0.2 °C, 39±2% RH) for at least 24-h, and statically discharged via a polonium source. In both pre- and post weighing, samples were weighed twice; if these two masses were more than 5 μg apart, a third weighing was carried out. After the third weighing, the average of the two measured masses within 5 μg of each other was used for calculating concentrations. After every batch of 10 samples, the zero, span, and linearity of the balance were checked via a set of class ‘S’ weights.

Continuous PM2.5

We used DustTrak (DT) model 8520 monitors (TSI, Shoreview, MN) for continuous measurement of PM2.5 exposure. PM2.5 exposure was measured every second, averaged, and recorded at 1-min intervals. The DTs were operated at a flow rate of 0.8 lpm, with an upstream external mini-PEM27 used as the size selective inlet for PM2.5. In the mini-PEM, a level greased well served as the impaction surface. The DTs were calibrated to a zero filter prior to each 24-h sampling period to avoid drifts.

Following earlier studies,4,5,8,13 we standardized the minute-by-minute PM records for RH using the relationship from a previous study.28 The RH-standardized PM2.5 data were then adjusted using a correction factor (CF), calculated as the ratio of the co-located integrated (gravimetric) PM2.5 measurement to the average of the minute-by-minute continuous measurements over the same time period.4,5,8,13 We calculated the CF separately for each subject, using her/his own gravimetric and continuous measurements. Twenty five percent of the minute-by-minute continuous data were missing because the instrument malfunctioned, for example, due to laser or battery failure; another 1% was excluded due to concerns about data validity, for example, when the connecting tubing was bent in a way that limited airflow.

Ambient PM2.5

We operated four simultaneous rooftop ambient gravimetric and continuous monitors, one in each study neighborhood (Figure 1). Measurements were done for 48 h every 6 days. The fixed sites were located in the residential areas in each neighborhood with relatively modest influence from direct traffic. Details on ambient measurements are provided elsewhere.4 Personal PM2.5 exposures were measured on the same days as the ambient 48-h monitoring as much as possible.

Location Data

We used a Garmin eTrex Vista GPS device to record coordinates for each subject’s residence and school, which were used to calculate average distances from the nearest main road in Arc-Map 10.1 (ESRI); road classification was based on geo-coded road map of Accra to reflect similar traffic density across our study neighborhoods. We also measured each subject’s location at 1-min intervals using a GPS device placed in the outer pocket of the backpack that he/she wore.

Forty five percent of time–location data were missing because satellite signal was weak, most of which occurred indoors; another 1% was outliers due to measurement error. We addressed these issues in two steps. First, we repositioned outlier coordinates on the subject’s walking path, using the median location of five ordered coordinates containing the outlier, retaining the temporal ordering of the data. Second, we imputed missing coordinates between 2200 and 0500 hours, when the subjects were likely to be at home based on the information from the time–location–activity questionnaires, using the home location. This reduced the proportion of missing coordinates to 14% of total minute-by-minute data.

Meteorological Variables

Data on meteorological variables (RH, number of hours with rain in each 24-h measurement period, and wind speed) were obtained from measurements at the Kotoka International Airport in Accra, as detailed in previous publications.4,5,8,13

Household Fuel Use and Community Socioeconomic Status

We used a 10% sample of the Ghana 2000 Population and Housing Census to calculate the following variables for census enumeration areas (EA) that contained subjects’ residences and schools:

  • proportion of households that use charcoal or wood for cooking; and

  • average SES, calculated as described previously based on housing characteristics, water and waste systems, and ownership of durable assets.8,13

Data Management and Statistical Analysis

We synchronized date and time on the DT and GPS units. Continuous PM2.5 concentrations and location data were compiled into a single data set by matching on date and time, with each record representing a unique date, time, location, and PM concentration. Weather variables were incorporated into the data set using date and time.

We used the location data for all subjects to analyze and visualize where subjects spent the most time in each study neighborhood. Specifically, we calculated the density of time spent at any point in the neighborhood using the number of minutes that was spent at that point by all students, smoothed using a two-dimensional normal density kernel. We also graphically present the students’ PM2.5 exposures by location. In these graphs, we averaged all continuous PM2.5 values within 20 m × 20 m grids to indicate places that on average are associated with higher/lower exposure.

We used regression analysis to examine the association of average daily personal PM2.5 exposure with its potential individual, household, and neighborhood determinants. We repeated the regressions both with and without adjustment for the neighborhood 48-h average ambient PM2.5 concentration. The former implies that neighborhood ambient PM2.5 may be (partially) independent of the local sources included in the model, for example, due to regional influence; the latter implies that neighborhood ambient PM2.5 itself is due to the local sources. To accommodate repeated measurements from subjects in the same school, we estimated the following linear mixed-effects model.

  • PMpersonal exposure=24-h integrated personal PM2.5 exposure (μg/m3)

  • X=a vector of covariates, including:

    • gender (male; female)

    • subject’s household fuel (biomass; non-biomass)

    • percent of households using biomass in the EA corresponding to the residence and school of the subject

    • day of the week on which measurement was conducted

    • weighted average of distances from the residence or school to main roads

    • average household SES in the EA corresponding to the subjects’ residence or school

    • schoolyard surface (paved and paved broken; packed and loose dirt)

    • time spent by subjects near household cooking fire during the 24-h personal measurement period (always or sometimes; none)

  • Weather =a vector of the following weather variables:

    • number of hours since last precipitation before the start of the 24-h personal exposure measurement period

    • average wind speed (m/s) during the 24-h exposure measurement period

  • PMambient=integrated ambient PM2.5 concentration (μg/m3) at a non-traffic rooftop site in the subject’s neighborhood. We used linear interpolation to estimate ambient PM2.5 on exposure measurement days when we had no ambient data.

  • b=school-level random intercept29

  • λ=subject-level random intercept29

  • ɛ=vector of within-school and within-subject errors

  • β, γ, and δ=regression coefficients

The school-level random intercept helps remove the influence of unobserved factors that affect all measurements in each school. Both ambient and personal PM2.5 concentrations were log-transformed to ensure that model residuals were normally distributed. Residual diagnostics suggested a better model fit when distance to main roads and weather variables were also log-transformed (as done in a previous study8).

All analyses were done using the open-source statistical package R version 3.0.0 (R Project for Statistical Computing, Vienna, Austria).


We collected 85 24-h integrated and 82 continuous personal PM2.5 samples from 56 students in eight schools in the four study neighborhoods. Demographic characteristics of the students and the factors that may affect their air pollution exposure are summarized in Table 1. There were 38 male and 18 female students in our study, with a mean age of 13.5 years. Sixty-two percent of subjects resided in households that cooked with biomass fuel outdoors in an open-air shared courtyard; the majority of these used charcoal with only 7% cooking with firewood. Six students lived in households that purchased their food and did not cook on the measurement day. All subjects were from households in which no one smoked.

Table 1 Demographic and exposure characteristics of the study subjects and school locations

Average personal PM2.5 exposure across all four neighborhoods was 56.0±33.5 μg/m3, with individual exposures ranging from less than 10 μg/m3 (during the rainy season) to more than 150 μg/m3 (just after dry and dusty season). Exposure was above 25 μg/m3, the WHO guideline for 24-h ambient PM2.5, for nearly 90% of the students.30 Students in AD had the lowest exposure (geometric mean 36.9 μg/m3) and those in NM the highest (57.5 μg/m3) (Table 2). Exposure was significantly higher among girls than boys (67 vs 44 μg/m3; P-value=0.001); higher exposure among girls persisted in the multivariate model even after adjusting for time spent near the cooking fire (Table 3). For those students with repeated measurement, mean difference between the first measurement and subsequent measurements was 9.2 μg/m3; the mean absolute difference was 32.2 μg/m3.

Table 2 Geometric mean personal PM2.5 exposure and ambient PM2.5 levels, by neighborhood and other characteristics
Table 3 Coefficients for multivariate analysis of the association of personal PM2.5 with individual, household, neighborhood, and meteorological variables

Over one third of the students lived within 100 m of busy roads (Figure 1 and Table 1). These subjects had higher PM2.5 exposure (geometric mean 51.4 μg/m3) than those who lived farther away from busy roads (45.1 μg/m3). Similarly, personal exposures at three schools that were<50 m from busy roads were significantly higher than at the five schools located farther away from busy roads (72 vs 44 μg/m3; P-value<0.001). Association of exposure with distances of schools and home from main roads was not significant in multivariate analysis (Table 3).

On weekdays, the students typically only traveled between their home and school, with few detours. The majority (~90%) commuted to school on foot, with distances ranging from a few meters to >3 km. During their commute, the students’ walking paths traversed different areas and road types, including busy highways/roads, local roads, residential alleys and foot paths, and markets. Across all four neighborhoods, students were most likely to be in motion (defined as change in position >5 m, which was the average movement of GPS data during night hours) at about 0800 hours (i.e., immediately before school) and around 1600 hours (immediately after school).

Figure 2 shows that students primarily spent their weekdays at home, at school, or walking between the two, with schools having the highest density of student time. We observed differences in movement patterns between the low- and high-SES neighborhoods. In the high SES, sparsely populated, residential neighborhood of EL, children spent almost all of their out-of-school time at their home. In JT, a low-income and densely populated neighborhood, the children spent time in locations other than their home, but within their own neighborhood. When divided by the time of the day, 73% of the time spent at home by these students was between 2200 and 0600 hours; 1% between 0800 and 1500 hours, and the remaining 26% at other times of the day. In contrast, 85% of all the time spent at school was between 0800 and 1500 hours. Only about a third of total measurement minutes were spent at locations other than home and school.

Figure 2

Density of time spent by the students at different locations. The graph shows the extent of time spent by all students, with the darker density demonstrating more time spent or a larger number of students spending time at a point.

The average of minute-to-minute PM2.5 concentrations along the children’s walking paths are shown in Figure 3. When taken over the course of the day, exposures were highest between around 0800 and 1200 hours, declining slowly to their lowest levels between 2000 hours and around midnight, when they rose again. This pattern is consistent with the temporal pattern of ambient pollution, reported elsewhere.4 Geometric mean of continuous exposure data decreased with increasing distance from main roads: 51.3 μg/m3 for distances <50 m, 47.7 μg/m3 for distances between 50 m and 100 m, 37.7 μg/m3 for distances between 100 m and 200 m, and 32.7 μg/m3 for distances >200 m. By location, average exposures were about the same at home (geometric mean 36.5 μg/m3) and school (39.1 μg/m3), but were higher at other locations (43.2 μg/m3), which mostly included walking to and from school. Taking into account the average levels of exposure and the time spent in each location, 41% of the total exposure occurred at home, 25% at school, and the remaining 34% at other locations. We tested the relative contributions of time and space to total exposure; neighborhood of residence and distance from closest main road together explained 10.4% of the variance of the continuous exposure data; only~3% was explained by time.

Figure 3

Spatial pattern of PM2.5 exposures of the study subjects, by neighborhood. The graph shows average of all minute-by-minute exposures, with averaging done in 20 m × 20 m grids. Annual geometric mean PM2.5 concentrations at residential monitoring sites were 50.5 μg/m3 in JT, 31.1 μg/m3 in AD, 28.4 μg/m3 in NM, and 23.2 μg/m3 in EL. Annual geometric mean PM2.5 concentration at a roadside monitoring site was 35.2 μg/m3 in NM.4

When considered by neighborhood, the students’ exposures at their schools were lowest in AD, but walking along major roads in this neighborhood was associated with high exposure levels. In JT, exposure was >100 μg/m3 at homes and along alleys, where a large number of biomass stoves are used for cooking street food.5,8 The places with highest exposures in NM were around the local market and central bus station, both located on a busy road (Figure 3).

To provide more detail on movement and exposure patterns, Figure 4 shows location by time and exposure for three example students. The 99th percentile of PM2.5 exposure for these students was >350 μg/m3. A student in NM walking along a secondary road was exposed to slightly higher PM2.5 during his morning commute to school than during his return commute through residential alleys (Figure 4a). A student in AD experienced higher exposure at or near school than at home or while commuting (Figure 4b). A third student from JT had high PM2.5 exposure throughout the day irrespective of her location (Figure 4c).

Figure 4

Minute-by-minute location and PM2.5 exposure for three example subjects in (a1, a2) NM, (b1, b2) AD, and (c1, c2) JT.

Personal exposure to fine particles was on average 23% higher than ambient concentrations with geometric means of 47.5 μg/m3 and 38.5 μg/m3, respectively (Figure 5). We found moderate correlation between personal exposure and neighborhood ambient concentrations (r=0.42; 95% CI 0.23–0.58). In NM, personal exposures were lower than ambient levels, with a geometric mean personal-to-ambient ratio of 0.60. Most personal PM2.5 exposures in JT and EL were higher than the ambient levels, resulting in geometric mean personal-to-ambient ratios of 1.24 and 1.55, respectively. Personal exposures were similar to ambient PM2.5 levels in AD.

Figure 5

The relationship between personal PM2.5 exposure and ambient PM2.5.

In unadjusted analysis, personal PM2.5 exposure was inversely related to the SES of the EA of residence and school, with geometric mean of exposure being 58.4 μg/m3 for students from schools in below-median-SES EAs and 36.3 μg/m3 for those in above-median-SES EAs (Table 2). Similarly, students living in low-SES EAs had geometric mean exposure of 55.3 μg/m3 vs 41.5 μg/m3 for those living in high-SES EAs (Table 2). There was no consistent or significant association with SES in the multivariate model (Table 3).

In multivariate analysis, personal PM2.5 exposure was higher in schools with dirt schoolyard surface than those which were paved, but the finding was only significant in the model that adjusted for ambient PM2.5 (Table 3). Household biomass use in the EA where the school is located was significantly associated with higher personal PM2.5 exposure. A student attending a school located in an EA where all households use biomass fuel would have 241% (95% CI 41–728%) and 153% (95% CI 1–531%) higher personal PM2.5 exposure than his/her counterpart attending school in an EA where biomass was not used, respectively, in the model with and without adjustment for neighborhood ambient PM2.5. There was no statistically significant association between personal PM2.5 exposure and biomass fuel use in the EA of residence. Using biomass at home was associated with higher PM2.5 exposure, by 31–38% in the different models. There was also little association between PM2.5 exposure and day of week.


While cities in the developing world have the highest PM concentrations, there has been limited data on human exposure, especially for children and adolescents. Our study provides a detailed analysis of the movements and exposures of school children in a growing metropolitan area. We found higher exposure in lower-SES neighborhoods, an influence from biomass use at home and around the school, and from the construction of the schoolyard surface. We also found that boys had lower exposure than girls, even after adjustment for time spent near the cooking fire.

Due to the absence of similar data, especially from the developing world, our results could only be broadly compared with other exposure studies. Personal PM2.5 exposure for school children in Accra neighborhoods were more than double those in USA17,18,31 and in Europe.14,32 In USA and the Netherlands, higher personal PM2.5 exposure was associated with proximity of homes and schools to major roads.14,17,32

There are a number of innovations and strengths to our study. We combined geo-referenced data about the neighborhood (from the census and road map), the students’ homes and schools, and their movements and exposures to have rich data on exposure to air pollution and its individual, household, and community determinants in a city of a developing country. This in turn permitted mapping locations throughout the day, and identifying times and places where high exposures occurred. The data were from students in eight schools across four neighborhoods with varying SES (Figure 1), allowing for analysis in relation to community SES. Simultaneous ambient and personal exposure enabled us to examine the relationship between the two.

The data used in this study also have a number of limitations that are common to many field research studies. Equipment malfunction led to the loss of some of the continuous location and exposure data. We could impute missing GPS data for the night period using the subject’s home location, to reduce the missing data to only 14%. DT monitors use light scattering technique, which is subject to error. Although we systematically corrected the continuous PM data, the steps involved introduced additional uncertainty. Further, it would have been ideal to have data on time spent near specific pollution sources, which may have been additional predictors of personal exposure. Logistical difficulties, including distance between study neighborhoods, restricted our ability to conduct personal measurements simultaneously in all four study neighborhoods. Further, our data covered only 8 months of the year and therefore, could not be used to assess seasonality of exposure. For the same reason, we did not have 24-h PM data from the students’ homes and schools, as we did for neighborhood ambient pollution. Finally, carrying the backpacks fitted with the monitors may have modified the students’ behavior despite their statements that this was not the case. GPS data confirm self-reported activities that school attendance occurred as usual.

Our findings indicate that household fuel use and school location may be determinants of children’s air pollution exposure in Accra. The role of biomass fuel is further supported by findings that it may contribute between 38% and 48% of total PM mass in Accra10 and be a determinant of the spatial pattern of air pollution within neighborhoods.5,8 If the role of biomass burning as an important determinant of school children’s exposure is established in further studies, it should motivate focus on policies that specifically address urban biomass use and create incentives and conditions for transition to cleaner fuels such as LGP.13 Similarly, the role of schoolyard surface can be further investigated through intervention studies that involve exposure measurement before and after paving schoolyards. If these studies show a significant reduction in students’ exposure, existing schools can be modified to reduce exposure by paving and regular cleaning of schoolyards and roads around them. More broadly, as Ghana makes strides toward universal primary education (Millennium Development Goal 2), new schools will also inevitably be built. There is a need for evidence base that inform locations of new schools and for their structure and materials, that for example, ensure sufficient distance between main roads and schools33 to curb students’ air pollution exposure.


  1. 1

    Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H et al. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010. Lancet 2012: 380: 2224–2260.

    Article  Google Scholar 

  2. 2

    Brauer M, Amann M, Burnett RT, Cohen A, Dentener F, Ezzati M et al. Exposure assessment for estimation of the global burden of disease attributable to outdoor air pollution. Environ Sci Technol 2012: 46: 652–660.

    CAS  Article  Google Scholar 

  3. 3

    Arku RE, Vallarino J, Dionisio KL, Willis R, Choi H, Wilson JG et al. Characterizing air pollution in two low-income neighborhoods in Accra, Ghana. Sci Total Environ 2008: 402: 217–231.

    CAS  Article  Google Scholar 

  4. 4

    Dionisio KL, Arku RE, Hughes AF, Vallarino J, Carmichael H, Spengler JD et al. Air pollution in Accra neighborhoods: spatial, socioeconomic, and temporal patterns. Environ Sci Technol 2010: 44: 2270–2276.

    CAS  Article  Google Scholar 

  5. 5

    Dionisio KL, Rooney MS, Arku RE, Friedman AB, Hughes AF, Vallarino J et al. Within-neighborhood patterns and sources of particle pollution: mobile monitoring and geographic information system analysis in four communities in Accra, Ghana. Environ Health Perspect 2010: 118: 607–613.

    CAS  Article  Google Scholar 

  6. 6

    Padhi BK, Padhy PK Assessment of intra-urban variability in outdoor air quality and its health risks. Inhal Toxicol. 2008: 20: 973–979.

    CAS  Article  Google Scholar 

  7. 7

    van Vliet E, Kinney P Impacts of roadway emissions on urban particulate matter concentrations in sub-Saharan Africa: new evidence from Nairobi, Kenya. Environ Res Lett 2007; 2: 045028.

    Article  Google Scholar 

  8. 8

    Rooney MS, Arku RE, Dionisio KL, Paciorek C, Friedman AB, Carmichael H et al. Spatial and temporal patterns of particulate matter sources and pollution in four communities in Accra, Ghana. Sci Total Environ 2012: 435-436: 107–114.

    CAS  Article  Google Scholar 

  9. 9

    Etyemezian V, Tesfaye M, Yimer A, Chow JC, Mesfin D, Nega T et al. Results from a pilot-scale air quality study in Addis Ababa, Ethiopia. Atmos Environ 2005: 39: 7849–7860.

    CAS  Article  Google Scholar 

  10. 10

    Zhou Z, Dionisio KL, Verissimo TG, Kerr AS, Coull B, Arku RE et al. Chemical composition and sources of particle pollution in affluent and poor neighborhoods of Accra, Ghana. Environ Res Lett 2013: 8: 044025.

    Article  Google Scholar 

  11. 11

    Bailis R, Ezzati M, Kammen DM Mortality and greenhouse gas impacts of biomass and petroleum energy futures in Africa. Science 2005: 308: 98–103.

    CAS  Article  Google Scholar 

  12. 12

    Bonjour S, Adair-Rohani H, Wolf J, Bruce NG, Mehta S, Pruss-Ustun A et al. Solid fuel use for household cooking: country and regional estimates for 1980-2010. Environ Health Perspect 2013: 121: 784–790.

    Article  Google Scholar 

  13. 13

    Zhou Z, Dionisio KL, Arku RE, Quaye A, Hughes AF, Vallarino J et al. Household and community poverty, biomass use, and air pollution in Accra, Ghana. Proc Natl Acad Sci USA 2011: 108: 11028–11033.

    CAS  Article  Google Scholar 

  14. 14

    van Roosbroeck S, Wichmann J, Janssen NA, Hoek G, van Wijnen JH, Lebret E et al. Long-term personal exposure to traffic-related air pollution among school children, a validation study. Sci Total Environ 2006: 368: 565–573.

    CAS  Article  Google Scholar 

  15. 15

    Borgini A, Tittarelli A, Ricci C, Bertoldi M, De Saeger E, Crosignani P Personal exposure to PM2.5 among high-school students in Milan and background measurements: the EuroLifeNet study. Atmos Environ 2011 8// 45: 4147–4151.

    CAS  Article  Google Scholar 

  16. 16

    Spira-Cohen A, Chen LC, Kendall M, Lall R, Thurston GD Personal exposures to traffic-related air pollution and acute respiratory health among Bronx schoolchildren with asthma. Environ Health Perspect 2011: 119: 559–565.

    Article  Google Scholar 

  17. 17

    Spira-Cohen A, Chen LC, Kendall M, Sheesley R, Thurston GD Personal exposures to traffic-related particle pollution among children with asthma in the South Bronx, NY. J Expo Sci Environ Epidemiol 2010: 20: 446–456.

    CAS  Article  Google Scholar 

  18. 18

    Crist KC, Liu B, Kim M, Deshpande SR, John K Characterization of fine particulate matter in Ohio: indoor, outdoor, and personal exposures. Environ Res 2008: 106: 62–71.

    CAS  Article  Google Scholar 

  19. 19

    Du X, Kong Q, Ge W, Zhang S, Fu L . Characterization of personal exposure concentration of fine particles for adults and children exposed to high ambient concentrations in Beijing, China. J Environ Sci 2010; 22: 1757–1764.

    CAS  Article  Google Scholar 

  20. 20

    Jiang R, Bell ML A comparison of particulate matter from biomass-burning rural and non-biomass-burning urban households in northeastern China. Environ Health Perspect 2008: 116: 907–914.

    Article  Google Scholar 

  21. 21

    Ezzati M, Saleh H, Kammen DM The contributions of emissions and spatial microenvironments to exposure to indoor air pollution from biomass combustion in Kenya. Environ Health Perspect 2000: 108: 833–839.

    CAS  Article  Google Scholar 

  22. 22

    Dionisio KL, Howie SR, Dominici F, Fornace KM, Spengler JD, Adegbola RA et al. Household concentrations and exposure of children to particulate matter from biomass fuels in The Gambia. Environ Sci Technol 2012: 46: 3519–3527.

    CAS  Article  Google Scholar 

  23. 23

    Kistnasamy EJ, Robins TG, Naidoo RN, Batterman S, Mentz GB, Jack C et al. The relationship between asthma and ambient air pollutants among primary school students in Durban, South Africa. Int J Environ Health 2008; 2: 365–385.

    CAS  Article  Google Scholar 

  24. 24

    Naidoo RN, Robins TG, Batterman S, Mentz G, Jack C Ambient pollution and respiratory outcomes among schoolchildren in Durban, South Africa. S Afr J Child Health 2013: 7: 127–134.

    CAS  Article  Google Scholar 

  25. 25

    United Nations Department of Economic and Social Affairs (Population Division) 2011.

  26. 26

    World Health Organization Global Status Report on Noncommunicable Diseases 2010. World Health Organization (WHO): Geneva, 2011).

  27. 27

    Demokritou P, Kavouras IG, Ferguson ST, Koutrakis P Development and laboratory performance evaluation of a personal multipollutant sampler for simultaneous measurements of particulate and gaseous pollutants. Aerosol Sci Technol 2001 2001/01/01 35: 741–752.

    CAS  Article  Google Scholar 

  28. 28

    Chakrabarti B, Fine P, Delfino R, Sioutas C Performance evaluation of an active personal DataRAM PM2.5 Mass Monitor (Thermo Anderson pDR-1200) designed for continuous personal exposure measurements. Atmos Environ 2004: 38: 3329–3340.

    CAS  Article  Google Scholar 

  29. 29

    Laird NM, Ware JH Random-effects models for longitudinal data. Biometrics 1982: 38: 963–974.

    CAS  Article  Google Scholar 

  30. 30

    WHO Air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulfur dioxide. 2005.

  31. 31

    Wu C-F, Delfino RJ, Floro JN, Quintana PJE, Samimi BS, Kleinman MT et al. Exposure assessment and modeling of particulate matter for asthmatic children using personal nephelometers. Atmos Environ 2005; 39: 3457–3469.

    CAS  Article  Google Scholar 

  32. 32

    Van Roosbroeck S, Jacobs J, Janssen NAH, Oldenwening M, Hoek G, Brunekreef B . Long-term personal exposure to PM2.5, soot and NOx in children attending schools located near busy roads, a validation study. Atmos Environ 2007; 41: 3381–3394.

    CAS  Article  Google Scholar 

  33. 33

    Amram O, Abernethy R, Brauer M, Davies H, Allen RW Proximity of public elementary schools to major roads in Canadian urban areas. Int J Health Geogr 2011: 10: 68.

    Article  Google Scholar 

  34. 34

    Nakagawa S, Schielzeth H A general and simple method for obtaining R2 from generalized linear mixed-effects models. Methods Ecol Evol 2013: 4: 133–142.

    Article  Google Scholar 

Download references


ME is supported by a UK MRC Strategic Award.

Author information



Corresponding author

Correspondence to Majid Ezzati.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Arku, R., Dionisio, K., Hughes, A. et al. Personal particulate matter exposures and locations of students in four neighborhoods in Accra, Ghana. J Expo Sci Environ Epidemiol 25, 557–566 (2015).

Download citation


  • Africa
  • air pollution
  • adolescent health
  • exposure
  • traffic pollution

Further reading


Quick links