The State of Ambient Air Quality in Two Ugandan Cities: A Pilot Cross-Sectional Spatial Assessment

Air pollution is one of the leading global public health risks but its magnitude in many developing countries’ cities is not known. We aimed to measure the concentration of particulate matter with aerodynamic diameter <2.5 µm (PM2.5), nitrogen dioxide (NO2), sulfur dioxide (SO2), and ozone (O3) pollutants in two Ugandan cities (Kampala and Jinja). PM2.5, O3, temperature and humidity were measured with real-time monitors, while NO2 and SO2 were measured with diffusion tubes. We found that the mean concentrations of the air pollutants PM2.5, NO2, SO2 and O3 were 132.1 μg/m3, 24.9 µg/m3, 3.7 µg/m3 and 11.4 μg/m3, respectively. The mean PM2.5 concentration is 5.3 times the World Health Organization (WHO) cut-off limits while the NO2, SO2 and O3 concentrations are below WHO cut-off limits. PM2.5 levels were higher in Kampala than in Jinja (138.6 μg/m3 vs. 99.3 μg/m3) and at industrial than residential sites (152.6 μg/m3 vs. 120.5 μg/m3) but residential sites with unpaved roads also had high PM2.5 concentrations (152.6 μg/m3). In conclusion, air pollutant concentrations in Kampala and Jinja in Uganda are dangerously high. Long-term studies are needed to characterize air pollution levels during all seasons, to assess related public health impacts, and explore mitigation approaches.


Introduction
On the 25 March 2014, the World Health Organization (WHO) released new estimates of the contribution of air pollution to global mortality showing that seven million deaths were attributable to air pollution worldwide in the year 2012 (3.7 million due to ambient air pollution (AAP) and 4.3 million due to indoor air pollution (IAP)) [1]. This number represents a doubling from the air pollution mortality rates estimated by WHO in the year 2004 [1,2].
Air pollution is thus one of the leading global public health risks. Health problems commonly associated with air pollution exposure include: respiratory diseases (e.g., chronic obstructive pulmonary disease, asthma, lung cancer and acute respiratory infections in children) and cardiovascular diseases (such as ischemic heart disease and stroke) [2]. Adverse health effects associated with air pollution exposure are particularly severe among vulnerable populations (e.g., people with respiratory diseases like asthma), older people, and children. Available data also show that airpollutionhasthepotential to impair lung growth as a result of perinatal exposures thus threatening the health of entire generations [3][4][5][6]. Although over 3000 substances are known to potentially contaminate air [7], the WHO has identified particulate matter (PM), nitrogen dioxide (NO2), carbon monoxide (CO), sulfur dioxide (SO2) and ozone (O3) as the pollutants with greatest public health importance [2]. The United States (US) National Ambient Air Quality Standard (NAAQS) [8] designates all of the above plus airborne lead (Pb) as criteria pollutants.
Data on the magnitude of air pollution in African cities is limited, particularly as it relates to Sub-Saharan Africa [11]. The WHO database provides an average PM2.5 value for Africa of 78 μg/m 3 annual mean (which is almost three times the set limit) [12]. A detailed review of the database shows that 18 African studies, seven of which were from South Africa, were used in generating this average, indicating a dearth of data on air pollution for the African continent. In most of the African studies, PM concentrations exceed WHO limits.
Data from African cities on gas phase pollutants are even sparser [11]. Available reports, however, indicate that concentrations of gas phase pollutants are low [13][14][15][16][17][18][19][20][21][22]. Carmichael et al., in an extensive study in Africa, Asia and South America, found that concentrations of gas phase pollutants of SO2, NO2 and ammonia were generally lower in the tropical regions than non-tropical regions of the studied countries [23].
For Uganda, data on air pollution is nearly nonexistent. To date, there is only one publication available, from our group, showing a PM concentration of 100 μg/m 3 from a single pilot study measured in one district of Kampala [24]. The 2010 Uganda State of the Environment Report acknowledges this lack of air pollution data for Uganda [25]. The current study expands on our previous air pollution assessment efforts [22] and provides novel data on ambient concentrations of four key air pollutants (PM2.5, NO2, SO2, and O3) at various sites in Kampala and Jinja.

Study Design
This is a cross-sectional, spatial, pilot assessment of ambient air concentrations of PM2.5, NO2, SO2, and O3 at different sites in three different land use areas in Kampala and Jinja during the period from 30 June to 27 July, 2014.

Study Sites and Monitoring Approaches
Air pollutant monitoring was conducted in Kampala and Jinja. Kampala, the capital city of Uganda, covers an area of 197 km 2 and is spread over 22 hills at an altitude of 1120 m above sea level. The city's day and night population is 3 million and 1.72 million people, respectively [26]. The day population represents Kampala residents and commuters entering the city from outside regions for work, education and business. Annual rainfall in Kampala ranges from 1750-2000 mm with peak wet seasons from March to May and from September to November. The dry seasons are between June and July, and December and February. The average annual temperature is 21.9° C and relative humidity ranges from 53-89% [27].
Jinja is the second largest city in Uganda, located 80 km east of Kampala and covers a land area of 28 km 2 at an altitude of 1230 m above sea level. Jinja has a day population of 300,000 people and a night population of 89,700 people [26]. The annual rainfall averages 1125 mm [28].
Air pollutant sampling sites in both cities were selected to represent different topographies and land use areas: commercial, industrial, and residential. According to the land classification system of local climate zones (LCZ), all of the sampling sites belong to the following categories: open low-rise or sparsely built [29]. Representative photographs of these sites are shown in Figure 1. Sites for PM monitoring were fewer than sites for gas phase pollutant monitoring, as equipment for PM sampling was limited. Areas of the cities characterized by high commercial activities such as trading, small-scale manufacturing and high traffic were selected as commercial land use areas. Industrial land use areas were in designated industrial areas of the cities. In Kampala, the industries surrounding the monitoring sites were involved in textile, steel and food products, while in Jinja we observed food products, plastics

c) d)
and steel industries. Land use areas defined as residential were divided into two categories, those with paved/tarmac roads and those with unpaved/murram roads. Meteorological parameters (temperature and humidity) and O3 were monitored at one central commercial site in Kampala city. Meteorological data (2012-2014) were also retrieved from National Weather Services to compare the year-round meteorological conditions and the meteorological conditions during sampling.

Air Pollutant Sampling Methods
PM2.5 concentrations were measured over periods of 24 hours from 30 June to 27 July 2014 with a real-time aerosol monitor, DUSTTRACK II-8530 (TSI Inc, Shoreview, MN) at 18 sites (15 in Kampala and 3 in Jinja) that can assess PM2.5 concentrations in a range from 1 µg/m 3 to 400 mg/m 3 . Prior to all measurements, the DUSTTRACK monitor was calibrated using the federal reference method, and zerocalibrated prior to each sampling session. All data were saved on the monitor until the end of study when it was downloaded into an excel database for analysis in Stata 11.2.
Concentrations of NO2 and SO2 were measured with Combo diffusion tubes (NO2 and SO2, Ormantine, FL, USA) at 28 study sites (22 in Kampala and 6 in Jinja). At each study site, two duplicate diffusion tubes were secured on the outside walls of selected buildings, 3-5 meters above ground. Sampling sites were selected to reflect different land use patterns and geographic topography, and each building was at least 3 meters away from immediate emission sources. The sampling height was selected for the safety of the passive samplers and was within the USEPA ambient monitoring siting criteria (i.e., < 15 m). Each passive diffusion tube was exposed to ambient air for two weeks. Sampling starting and sampling end times were recorded. Two traveling blanks were included for each city. Prior to, and following, sampling periods, the samplers were stored at 4 °C. Combo diffusion tubes were shipped to Gradkos laboratory in England where NO2 and SO2 analyses were conducted on a Dionex ICS1100 ICU10 ion chromatography system (Thermo Fisher Scientific Inc., Waltham, MA, USA). O3 was measured with a federal equivalent real-time monitor (POM, 2B Technologies, CO, USA) that was calibrated before the study period, and cleaned daily during the sampling period.

Meteorological Measurements
Temperature and humidity were monitored daily for the first seven days of the study period with a real-time monitor (HOBO U23, OnSet, MA, USA). The monitor was calibrated prior to sampling, and cleaned daily during the sampling period.

Data Analysis
Data from all monitors were downloaded directly into a Microsoft Excel database and analyzed using Stata 11.2 (StataCorp LP, College Station, TX). Descriptive statistics were used to summarize all pollutant concentrations and meteorological data. Mean pollutant concentrations were compared between land use areas and cities by t-tests. A p-value of <0.05 was considered statistically significant. Concentrations of travelling blanks (i.e., tubes not exposed to sampling environments) were subtracted from all measured SO2 and NO2 concentrations prior to statistical analysis.

Ethics Approval
The study protocol was approved by the Mulago Hospital Research and Ethics Committee and the Uganda National Council for Science and Technology.

Temperature and Humidity
Temperature and relative humidity were measured at the central monitoring location during the first week (30 June to 5 July, 2014) of sampling. The mean (± standard deviation (SD)) ambient temperature was 24.7 ± 1.9 °C (maximum 26.4 °C, minimum 21.1 °C). Mean humidity was 63.5% ± 5.7 (maximum 74.4%, minimum 58.0%). We did not monitor temperature and humidity data further due to limited equipment availability. Archived temperature and humidity data from the National Weather Services (at Entebbe Airport) was used instead and is presented in Table 1 showing largely constant weather conditions during the entire air pollution monitoring period.   Table S1. The mean 24-hour PM2.5 concentrations calculated for all study sites was 132.1 μg/m 3 . The concentration measured by the real-time monitor was comparable to the filter-based approach. At the central monitoring site (city center), the PM2.5 mass concentrations were 90.4 µg/m 3 obtained from the filter-basedapproachvs. 94.0 µg/m 3 obtained from the real-time monitor. By city, 24-hour mean PM2.5 concentrations in Kampala were higher than in Jinja, but the difference did not reach statistical significance (138.6 μg/m 3 and 99.3 μg/m 3 , p = 0.20). By land use, PM2.5 and nitrogen dioxide pollution levels are shown in Table 2. The highest 24-hour mean PM2.5 concentrations were observed at the industrial (156 μg/m 3 ) followed by residential areas with unpaved roads (152.6 μg/m 3 ) and commercial land use areas (129.4 μg/m 3 ). Residential and office areas with paved roads had the lowest mean PM concentrations (88.3 μg/m 3 ). Compared to residential areas with paved roads, residential areas with unpaved roads had significantly higher mean 24-hour PM2.5 concentrations (152.6 μg/m 3 vs. 88.3 μg/m 3 , p = 0.045). Considering both industrial and commercial areas as nonresidential, no significant differences were noted between nonresidential and residential areas (131.0 μg/m 3 vs. 132.8 μg/m 3 , p = 0.93). The 24-hour mean PM2.5 concentration at a site in Jinja with unpaved roads was comparable to sites with unpaved roads in Kampala (161 μg/m 3 vs. 151.4 μg/m 3 ). The mean PM2.5 concentration at sites with paved roads in Jinja, however, was lower than that at similar sites in Kampala (68.5 μg/m 3 vs. 108.0 μg/m 3 ).

Gas Phase Pollutants
Duplicate diffusion tubes were used for sampling SO2 and NO2 at each monitoring site. Each tube can simultaneously collect NO2 and SO2. The concentrations of NO2 and SO2 at each monitoring site were calculated as the average of the readings of the two tubes.
NO2 and SO2 concentrations were measured at a total of 28 monitoring sites (22 in Kampala and six in Jinja). In Kampala, one of the SO2 duplicate tubes could not be retrieved at two monitoring sites. In Jinja, both NO2 tubes could not be retrieved at one site and one SO2 tube only was retrieved at one site. Therefore, 27 NO2 and SO2 sampling tubes were available for analysis (22 from Kampala and five from Jinja). The mean monitor exposure time was 330.34 (±25.54) hours or 13 days and 19 hours. The overall precision, expressed as coefficient of variation based on 22 pairs of co located sampling, was 14.0%.  supplementary  Table S2. The mean two-week NO2 concentration was 24.9 µg/m 3 .
By city, Kampala air was characterized by a higher mean total NO2 concentration than that of Jinja (26.69 µg/m 3 vs. 17.49 µg/m 3 , p = 0.07). The spatial variations of NO2 concentration levels in Kampala are shown in Figure 2b.
NO2 concentrations by land use are shown in Table 2. The highest NO2 concentrations were observed in commercial land use areas (32.19 µg/m 3 ) and the lowest in residential land use areas with paved roads (18.39 µg/m 3 ). The mean NO2 concentrations in commercial (including industrial) land use areas were significantly higher than in residential land use areas (32.19 µg/m 3 vs. 19.69 µg/m 3 , p = 0.002). NO2 concentrations measured in residential land use areas did not significantly differ from those in industrial land use areas (19.

Sulfur Dioxide
SO2 was measured at all study sites and in all land use areas where NO2 was measured. Levels of detectable SO2 concentrations are shown in supplementary Table S3. SO2 concentrations were below detection limit (<0.03 µg/m 3 ) at 20 of the 27 monitoring sites. The two-week mean SO2 concentration at all monitoring sites was 3.79 ± 3.0 µg/m 3 .
By city, there was only one monitoring site in Jinja where SO2 concentrations reached detectable levels. This site, in an industrial land use area of Jinja, had a higher SO2 concentration level (7.3 µg/m 3 ) than that in the industrial land use area of Kampala (<0.69 µg/m 3 ). Six other monitoring sites in Kampala also showed SO2 concentrations above detection limit (mean 3.11 µg/m 3 ). Comparing SO2 concentrations by land use, concentrations were highest in industrial land use areas (7.39 µg/m 3 ), followed by commercial land use areas (3.69 µg/m 3 ), then residential land use areas with paved roads (2.79 µg/m 3 ), or unpaved roads (2.39 µg/m 3 ).

Ozone
The mean one-hour O3 concentration at the monitoring site established at Lourdel Road, Wandegeya in Kampala measured over a period of seven days was 11.4 μg/m 3 (±4.8 μg/m 3 ). O3 concentrations varied considerably across the sampling period ( Figure 3).

Discussion
This pilot study demonstrates presence of high PM concentrations and low gas phase air pollutant levels in Kampala and Jinja between 30 June and 17 July, 2014. The observed mean PM2.5 concentration of 132.1 μg/m 3 (5.3-fold above the limit defined by WHO) across all monitoring sites in the current study is comparable with the mean PM2.5 concentration of 104.9 μg/m 3 described in an earlier singlesite pilot study from a district in Kampala [24]. As expected, particulate air pollution levels were found to be greatest in areas with high commercial/industrial land use and unpaved roads.
The observed mean PM2.5 concentration of 132.1 μg/m 3 in the current study clearly exceeds the mean PM concentration of 78 μg/m 3 calculated for the African region and reported in the WHO 2014 publication of the global state of air quality from 1600 cities in 91 countries [12]. Within the East African region, however, the mean PM concentration (132.1 μg/m 3 ) observed in the current study is comparable with that reported from Nairobi/Kenya (128.7 μg/m 3 ) and significantly higher than that reported from Dar es Salaam/Tanzania (26 μg/m 3 ) [17,30].
Sources of particulate air pollution described in the studies of African cities are typically emissions from vehicles, re-suspended dust from unpaved roads, smoke from indoor biomass fuel use and garbage burning, and industrial sites [14,31,32]. During the current study, we observed source emissions of dust and soil blown by wind from unpaved roads, black smoke exhausts from cars, trucks and busses and smoke from burning household garbage in both Kampala and Jinja. High PM levels in residential land use areas with unpaved roads without industrial activity or high traffic volume, suggest that re-suspended dust significantly contributes to high PM levels. High PM levels in commercial land use areas with high traffic volume and paved roads, in contrast, suggest that vehicle emissions represent another significant source of PM in Kampala and Jinja. Dust from unpaved roads in the suburbs of both cities appears to be carried by human activities into areas with paved roads.
Ambient air PM composition has been reported in some African cities [16,24,33]. Our earlier pilot study in Kampala found that more than 90% of PM2.5 studied at a sampling site in the Mpererwe district of Kampala was comprised of crustal species (probably re-suspended soil dust) and carbonaceous aerosol [24]. In Dar es Salaam, a study of PM collected close to a vehicle traffic site found carbon to be the main component suggesting vehicular emissions as its main source [33]. In Nairobi, Gaita et al. found that vehicle traffic, mineral dust, industrial activity, combustion and a mixed factor (composed of biomass burning, secondary aerosol and aged sea salt) were the main sources of PM air pollution [32]. Mineral dust and traffic were responsible for approximately 74% of PM2.5 mass [32]. Based on our findings and observations in the current pilot study we speculate that re-suspended dust and vehicular emissions are the primary sources of PM2.5 in Kampala and Jinja and may also be significant contributors to air pollution in other African cities.
We also assessed the concentration of three key gas phase pollutants (NO2, SO2 and O3) in Kampala and Jinja. Even though concentrations of NO2, SO2 and O3 were below WHO guideline levels (200 μg/m 3 one-hour mean, 20 μg/m 3 24-hour mean and 100 μg/m 3 eight-hour mean, respectively [2]) we recognize that our findings cannot be directly compared with WHO air quality standards due to differences in averaging times. However, our findings are comparable with findings from other gas phase pollutant studies in Africa [13][14][15]17]. We do not know why gas phase pollutants in our study and other studies from Africa are low. Climatic conditions in the studied areas may facilitate adsorption of gas phase pollutants onto PM. As PM concentrations were found to be high in the current study and other studies in Africa this may explain the observed low concentrations of gas phase pollutants.
Conclusions from the current cross sectional study have to be made considering that temporal variations of air pollutant concentrations could not be assessed, which is a major limitation of this pilot study, especially for PM2.5. Due to the limitation of the number of real-time instruments (i.e., only one DustTrack), we could not measure PM2.5 concentrations at multiple sites at the same time. Therefore, we were unable to differentiate spatial variation from temporal variation in PM2.5 concentrations. However, the meteorological data in Table 1 suggests that the weather conditions during our measurements were quite consistent, minimizing the possibility of the impact of weather on PM2.5 concentrations. Given that this study was conducted within a short period of time, source emission profiles in different locations will not be expected to change dramatically. Therefore, the measured PM2.5 spatial variation at minumum suggests the heterogeneous pattern of PM2.5 in Kampala.
In addition, PM2.5 concentrations (132.1 μg/m 3 ) observed in this study (dry season) are consistent with findings from our earlier study, in which PM2.5 (104.3 μg/m 3 ) was collected in December 2013 (also a dry season). Meteorological conditions during this pilot study were typical in Kampala for dry seasons, as shown in Table 1. We are aware of the impact of weather and seasonal variations on air pollutant concentrations and expect PM2.5 concentrations to be different in rainy seasons. Future studies will have to expand air pollution monitoring to other cities and parts of Uganda, cover all weather seasons and begin exploring air pollution effects on public health, in particular lung health in urban populations of Uganda.

Conclusions
This study suggests that high level PM air pollution is prevalent in urban and suburban areas in Uganda, with PM2.5 concentrations above 100 µg/m 3 in multiple locations in Kampala. Land use characteristics define ambient PM2.5, NO2 and SO2 concentrations. Long-term exposures to the observed high levels of air pollution likely represent a major risk to public health in Kampala and Jinja. Long-term studies are needed to assess air pollution levels during the course of multiple weather seasons and the health impact in exposed populations.