LETTER • OPEN ACCESS You may also like
Spatial-temporal patterns of ambient fine - How protective is China’s NationalAmbient Air Quality Standards on short-
term PM2.5? Findings from blood pressureparticulate matter (PM ) and black carbon (BC) measurements of 1 million adults2.5 Tianjia Guan, Tao Xue, Jian Guo et al.
pollution in Accra - Chemical composition and sources of
particle pollution in affluent and poor
neighborhoods of Accra, Ghana
To cite this article: Abosede S Alli et al 2021 Environ. Res. Lett. 16 074013 Zheng Zhou, Kathie L Dionisio, Thiago G
Verissimo et al.
 - Impacts of transboundary air pollution andlocal emissions on PM2.5 pollution in the
Pearl River Delta region of China and the
View the article online for updates and enhancements. public health, and the policy implications
X Hou, C K Chan, G H Dong et al.
This content was downloaded from IP address 197.255.69.32 on 08/12/2021 at 09:33
Environ. Res. Lett. 16 (2021) 074013 https://doi.org/10.1088/1748-9326/ac074a
LETTER
Spatial-temporal patterns of ambient fine particulate matter
OPEN ACCESS (PM2.5) and black carbon (BC) pollution in Accra
RECEIVED
15 February 2021 Abosede S Alli1, Sierra N Clark2,3, Allison Hughes4, James Nimo4, Josephine Bedford-Moses4,
REVISED Solomon Baah4, JiayuanWang1, Jose Vallarino5, Ernest Agyemang6, Benjamin Barratt3,7, Andrew Beddows3,7,
28 May 2021
Frank Kelly3,7, George Owusu6, Jill Baumgartner8,9, Michael Brauer10,11, Majid Ezzati2,3,12,
ACCEPTED FOR PUBLICATION
2 June 2021 Samuel Agyei-Mensah
6 and Raphael E Arku1,∗
1
PUBLISHED Department of Environmental Health Sciences, School of Public Health and Health Sciences, University of Massachusetts, Amherst,
25 June 2021 MA, United States of America
2 Department of Epidemiology and Biostatistics, School of Public Health, Imperial College, London, United Kingdom
3
Original content from MRC Center for Environment and Health, Imperial College London, London, United Kingdom
4
this work may be used Department of Physics, University of Ghana, Legon, Ghana
under the terms of the 5 Harvard T.H. Chan School of Public Health, Boston, MA, United States of America
Creative Commons 6 Department of Geography and Resource Development, University of Ghana, Legon, Ghana
Attribution 4.0 licence. 7 NIHR HPRU in Environmental Exposures and Health, Imperial College London, London, United Kingdom
Any further distribution 8 Institute for Health and Social Policy, McGill University, Montreal, Canada
of this work must 9
maintain attribution to Department of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, Canada
10
the author(s) and the title School of Population and Public Health, The University of British Columbia, Vancouver, Canada
of the work, journal 11 Institute for Health Metrics and Evaluation, University of Washington, Seattle, United States of America
citation and DOI. 12 Regional Institute for Population Studies, University of Ghana, Legon, Ghana
∗ Author to whom any correspondence should be addressed.
E-mail: rarku@umass.edu
Keywords: air pollution, fine particulate matter, black carbon, air quality, Ghana, sub-Saharan Africa
Supplementary material for this article is available online
Abstract
Sub-Saharan Africa (SSA) is rapidly urbanizing, and ambient air pollution has emerged as a major
environmental health concern in growing cities. Yet, effective air quality management is hindered
by limited data. We deployed robust, low-cost and low-power devices in a large-scale measurement
campaign and characterized within-city variations in fine particulate matter (PM2.5) and black
carbon (BC) pollution in Accra, Ghana. Between April 2019 and June 2020, we measured weekly
gravimetric (filter-based) and minute-by-minute PM2.5 concentrations at 146 unique locations,
comprising of 10 fixed (∼1 year) and 136 rotating (7 day) sites covering a range of land-use and
source influences. Filters were weighed for mass, and light absorbance (10−5m−1) of the filters was
used as proxy for BC concentration. Year-long data at four fixed sites that were monitored in a
previous study (2006–2007) were compared to assess changes in PM2.5 concentrations. The mean
annual PM2.5 across the fixed sites ranged from 26 µg m−3 at a peri-urban site to 43 µg m−3 at a
commercial, business, and industrial (CBI) site. CBI areas had the highest PM2.5 levels (mean:
37 µg m−3), followed by high-density residential neighborhoods (mean: 36 µg m−3), while
peri-urban areas recorded the lowest (mean: 26 µg m−3). Both PM2.5 and BC levels were highest
during the dry dusty Harmattan period (mean PM2.5: 89 µg m−3) compared to non-Harmattan
season (mean PM −32.5: 23 µg m ). PM2.5 at all sites peaked at dawn and dusk, coinciding with
morning and evening heavy traffic. We found about a 50% reduction (71 vs 37 µg m−3) in mean
annual PM2.5 concentrations when compared to measurements in 2006–2007 in Accra. Ambient
PM2.5 concentrations in Accra may have plateaued at levels lower than those seen in large Asian
megacities. However, levels are still 2- to 4-fold higher than the WHO guideline. Effective and
equitable policies are needed to reduce pollution levels and protect public health.
© 2021 The Author(s). Published by IOP Publishing Ltd
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
1. Introduction 2. Methods
Global PM2.5 exposures are gradually declining, but 2.1. Study location
there is little data from sub-Saharan Africa (SSA), The GAMA is the industrial and administrative cen-
where there are increasing concerns about air pol- ter of Ghana and one of the fastest growing metro-
lution in cities [1]. The urban population in SSA politan areas in SSA with ∼5 million residents and
has increased by over 400% since 1980 to about 450 an annual growth rate of 4.2% [24]. The GAMA con-
million people in 2017, making it the world’s fastest sists of Accra Metropolitan Area (AMA) at its core,
urbanizing region [2]. Urban residents in SSA have the port city of Tema to the east and 11 other adjoin-
access to increasing infrastructure, technology, and ing districts [25, 26]. The GAMA is in a tropical cli-
services for improved quality of life [3, 4]. However, mate zone with high average monthly temperatures
the sprawl has been largely unplanned in terms of and relative humidity (RH) ranging between 25 ◦C
land use factors. Environmental protection policies and 33 ◦C (77–90 ◦F) and 77%–85%, respectively
have also not kept pace with urban growth [3, 5], [25]. The GAMA has two major seasons: the rainy
making air quality a growing public health concern (May–October) period, and the dry period compris-
in cities [6–8]. Yet, cities in SSA lack ground-level ing the Harmattan (November–February) character-
air quality monitoring as exists in North America, ized by north-easterly trade winds from the Sahara
Europe, and parts of Asia [9, 10]. This lack of system- Desert [27].
atic monitoring is an obstacle to understanding the
within-city patterns, sources and health impacts of air
pollution, which are essential for designing effective 2.2. Study design
air quality policies [7, 11, 12]. This work was conducted within the multi-country
Exposure to elevated levels of fine particulate and multi-city ‘Pathways to Equitable Healthy Cities’
matter (PM2.5) and black carbon (BC), a compon- study (http://equitablehealthycities.org/), which aims
ent of PM, presents economic and health risks to to provide scientific evidence on how urban develop-
urban residents in SSA and elsewhere [13–15]. Evid- ment and policies can be managed to enhance health
ence suggests that BC is associated with higher health equity.
effects per unit when compared to PM mass, and As previously described [27], we designed a year-
is an indicator of the health risks related to emis- long campaign to examine the spatial (land-use fea-
sions from combustion sources [16]. As SSA urban- tures) and temporal (daily, weekly, monthly and
izes, there is an urgent need for detailed air monit- seasonal) variations in ambient PM2.5 and BC by
oring data in cities to inform interventions to protect sampling at a combination of fixed (∼1 year, n = 10
the health and wellbeing of the population. In partic- sites) and rotating (7 d, n = 136) sites. This design
ular, city-wide data on BC in SSA cities are limited allowed for detailed assessment of both the tem-
[14, 17, 18]. poral (using fixed site data) and spatial (using rotat-
In Accra, Ghana’s largest city and capital, air pol- ing site data) variability of PM2.5 and BC over the
lution emissions are characterized by diverse mixture study area. Further, this design allowed us to optim-
of combustion and non-combustion sources, includ- ally use a finite number of monitoring equipment to
ing biomass fuels, road dust and vehicle emissions capture data across the entire geographical extent of
[17, 19, 20]. Like other cities in SSA, rapid urbaniza- the study area. We used a structured form to collect
tion in Accra is intensifying industrial and economic information on land-use features at each monitoring
activities as well as increasing the demand for trans- site [27]. The sites were subsequently grouped into
portation, new fleet of vehicles, and energy, all with four land-use classes: commercial, business, indus-
major implications for air quality, exposure patterns trial (CBI); high-density residential; medium/low-
and health inequalities [21, 22]. density residential; or peri-urban (see supplementary
We aimed to collect detailed spatial and tem- text S1 for additional details). We originally planned
poral data and characterize within-city variations in a 12 month field campaign to collect data at 150 sites
PM2.5 and BC in the Greater AccraMetropolitan Area starting April 2019; however, the fieldwork was sus-
(GAMA) of Ghana. In a large-scale measurement pended for six weeks (31st March–18th May 2020)
campaign, we collected year-long data on PM2.5 and due to the COVID-19 pandemic lockdown in Accra
markers of BC from a network of diverse motoring and self-isolation of field team members. After the
sites. The data and analysis provide comprehensive lockdown was lifted and daily activities returned to
and granular information on air pollution variations pre-lockdown status, we conducted additional three
in a sprawling SSA city. We also analyzed changes weeks of measurement (19th May–11th June 2020)
in PM2.5 concentrations over a decade by comparing at all fixed sites along with 12 rotating sites, res-
annual data with those in a previous smaller study ulting in close to 12 months of data from 10 fixed
(2006–2007) [23]. and 136 rotating sites (see figure S2 (available online
2
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
at stacks.iop.org/ERL/16/074013/mmedia) for meas- in a temperature and RH controlled laboratory
urement timeline). (23 ± 2 ◦C, 35 ± 2% RH) at The University of
The 10 fixed sites were operated continuously, British Columbia. Further information on the UPAS
collecting weekly and 1 min averages throughout and filter handling can be found elsewhere [27, 33].
the measurement campaign at key locations selected An additional 27 duplicate (20% of sites) integrated
based on population density, road networks, neigh- samples and 28 field blanks were collected at rotat-
borhood socioeconomic status (SES) and household ing sites, including three post-COVID-19 lockdown
biomass fuel use data from the national census [28]. duplicates and blanks. The average of the duplicate
To compare changes in annual mean PM2.5 levels measurements was taken and final PM2.5 concentra-
within the last decade, four of the 10 fixed sites were tions were blank corrected. Quantitative information
placed at the exact locations monitored by Dionisio on blanks and duplicates are in the supplementary
and colleagues [23].We also collected 1-week samples text (figure S3).
at each of the 136 rotating sites, which were selec-
ted with a stratified random sampling scheme based
2.5. Continuous PM
on land-use, with more emphasis placed on AMA 2.5
We deployed a low-cost Zefan real-time continuous
where the majority of the population live [27]. The
monitor (www.zfznkj.com/) to measure PM con-
sites were initially computer-generated and the actual 2.5
centrations at 1 min intervals. The Zefan relies on
sampling locations that were as close as possible to
a light scattering technique to assess PM using
the computer-generated ones were identified by the 2.5
Plantower sensors (model PMS7003), which have
field team. The median distance (interquartile range,
been evaluated with reference monitors (i.e. FDMS
IQR) between the original computer-generated loca-
8500 and TEOM 1400ab) over 6–12 month peri-
tions versus the actual sites monitored was 181 (67–
ods [34, 35]. While this technique provides accurate
407) m. During the field campaign, the rotating sites
temporal pattern in measured PM concentrations, its
were sampled in groups of five each measurement
magnitude is inexact as PM mass are only inferred
week alongside the fixed sites.
from particle characteristics (e.g. number, size and
refractive index), which can be affected by weather
2.3. PM2.5 measurement and analytical methods conditions (e.g. RH and temperature) [34, 36].
We measured both real-time (1 min interval) and
Following previous studies [23, 37], we correc-
integrated gravimetric (weekly averages) PM2.5 con- ted the minute-by-minute continuous PM measure-
centrations using portable battery operated low-cost
ments by a correction factor (CF) calculated such that
and low-power monitors that were placed in protect-
the average of continuous PM measurements was
ive cases fastened onmetal poles at about 4 m (±1m) 2.5
equal to the integrated gravimetric PM concentra-
above ground [27]. We included in our analysis only 2.5
tion at the same location over the same 7 d measure-
samples frommonitors that operated for⩾75%of the
ment period. This was done to ensure that the aver-
measurement period (i.e. at least 5 out of 7 d to cap-
age weekly continuous measurements were the same
ture both weekdays and weekends) and had an aver-
as the gravimetric which has less error than optical
age flow rate within 10% of the intended rate.
sensors. We calculated unique CFs per site for each
7 d period. The median (IQR) of the CFs were 0.84
2.4. Integrated PM2.5 (0.69–1.13), similar to CFs previously reported for a
Weekly integrated PM2.5 was measured using the different optical sensor in Accra [23, 37].
Ultrasonic Personal Aerosol Sampler (UPAS) (Access
We tested minute-by-minute monitor-to-
Sensor Technologies, Fort Collins, USA) [29] oper-
monitor precision by running all monitors alongside
ated at 1 litre per minute (lpm). The UPAS has been
each other over a 24 h period prior to the commence-
demonstrated to have a close agreement with ref-
ment of field campaign [27]. Further, we conducted
erence monitors [29–31] over a wide range of con-
mid-campaign (in January 2020) monitor-monitor
centrations (10–1600 µg m−3) in diverse settings.
precision by co-locating the instruments at one of
However, a recent field evaluation suggested that
the fixed sites for a week to assess potential drift over
overloading could occur at filter masses above 650 µg
the course of the campaign. Finally, post-campaign,
[32], an issue that could be avoided by using the
we co-located two Zefan sensors with a U.S. fed-
duty-cycle feature on the UPAS in highly polluted
eral reference monitor located at the U.S. embassy
environments. To avoid overloading filters and to
in Accra. We did not see any within- or between-
also conserve battery power, the UPAS was operated
monitor bias in the sensor performance pre-, mid-,
at 50% duty cycle, drawing air 30 s every minute for a
and post-campaign.
total of 5040min over the 7 d sampling period. PM2.5
mass was collected on 2 µm pore size 37 mm bar-
coded Teflon membrane filters (https://mtlcorp.com/ 2.6. Black carbon
filters/) and weighed pre- and post-sampling Black carbon (BC) aerosols are known indicators
using a MTL AH500 automated robotic scale of combustion-related constituents of PM emissions
(www.mtlcorp.com/#/filter-weighing) maintained and contribute to global warming [38, 39]. Recent
3
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
( )
epidemiological studies also indicate associations week; (CFixed Site) and CFixed Sitej are the average
between BC and adverse health outcomes [16, 40]. PM2.5 or BC in the corresponding jth measurement
Thus, we used the absorption coefficient (light week and annual average PM2.5 or BC at all fixed
absorbance) (10−5m−1) of the post-weighed PM [( ) ( )]2.5 sites respectively, and CFixed Site / CFixed Site is
filters, estimated by applying an image-based reflect- j
ance method [41], as a marker for BC concentra- the TAF.
tions [42, 43]. The image-based reflectance method
closely correlates (r2 = 0.98) to elemental carbon 3.2. Temporal analysis
(EC) concentrations by thermo-optical reflectance, We examined the temporal patterns in the data by
with 1 absorbance unit (1× 10−5m−1) equivalent to season (Harmattan vs non-Harmattan), days of the
1.67 µg m−3 EC [41]. week (plus weekday vs weekend), and time of day
(diurnal) using data from the fixed sites.We also eval-
3. Data analysis uated changes in annual PM2.5 levels over a decade
(2006–2007 vs 2019–2020) by comparing fixed site
We collected 99 313 h (10 fixed sites = 78 890 and data obtained from the same four residential loca-
136 rotating sites= 20 423) of valid real-time and 654 tions sampled in a previous study [23].
(fixed sites = 518 and rotating sites = 136) weekly All analyseswere done using the statistical analysis
integrated gravimetric PM2.5 samples. Of these, 21
package R, version 3.6.1 [45], and an alpha of 0.05was
(fixed) and 10 (rotating) integrated samples were used as cut-off of significance.
excluded from analysis either due to failure to meet
inclusion criteria or for quality control reasons (e.g. 4. Results
blocked airflow and SD card malfunction), leaving a
total of 623 weekly (497 fixed and 126 rotating sites) 4.1. Spatial patterns in PM2.5 and BC
gravimetric samples for analysis. concentrations
The measurement locations and the measured con-
3.1. Spatial analysis centrations relative to the World Health Organiz-
We used data from the rotating sites to assess the spa- ation (WHO) air quality guideline are shown in
tial patterns of PM2.5 and BC across the city by the figure 1. The season adjusted mean (standard devi-
four site-types: CBI, high-, and medium/low-density ation, SD) integrated PM2.5 and BC concentrations
−3
residential and peri-urban. To provide more detail on across the rotating sites were 31 (10) µg m and
−5 −1
influence of traffic related sources on PM2.5 pollution 5 (2) × 10 m respectively. PM2.5 concentration
in the GAMA, we grouped the samples collected at at every rotating site was higher than the WHO
−3
rotating sites according to the type (major, secondary annual guideline of 10 ug m , while 99%, 71%
and minor) and surface material (paved, mixed and and 31% of the sites exceeded the interim target
−3 −3
unpaved) of the road near the monitoring site. Since 3 (IT-3, 15 µg m ), IT-2 (25 µg m ) and IT-1
−3
monitoring at rotating sites occurred in groups of five (35 µg m ), respectively (figure 1). The mean
sites per week (i.e. samples were not collected simul- PM2.5 and BC levels at rotating sites varied by land-
taneously at all rotating sites, nor evenly by site-types use. The highest PM2.5 concentrations were in CBI
−3
during each measurement week), we accounted for areas (mean: 37; range: 23–67 µg m ) and high-
potential influence of time trend/season on the spa- density residential neighborhoods (mean: 36, range:
tial patterns of the measured concentrations to allow 21–67 µg m
−3) (p < 0.01). Peri-urban sites had
for comparison across sites. We adjusted for potential the lowest concentrations (mean: 26, range: 16–
−3
time trends at the rotating sites by applying weekly 56 µg m ) after medium/low-density neighbor-
−3
specific temporal adjustment factor (TAF) using data hoods (mean: 28, range: 15–54 µg m ) (figure 2).
from the ten fixed (year-long) sites. For each meas- Similarly, BC concentrations were two times higher
−5 −1
urement week, a TAF calculated as the ratio of the in CBI areas (mean: 7, range: 1–14 × 10 m )
mean PM or BC across all fixed sites for that week compared with peri-urban sites (mean: 3, range: 1–2.5
6 × 10−5m−1to the mean annual PM or BC across all fixed sites ) (table 1). In general, average PM2.5 2.5
was used to adjust the samples collected at the rotating concentrations were slightly higher at sites along
sites in that particular week [44]. The season adjusted major and secondary roads compared with sites near
concentration C adjusted( i)j of the ith rotating site for the
minor roads, but not by road surface. We observed
jth measurement wee[k was calculated as:
similar patterns for BC. Overall, the relative differ-
( ) ( )] ences in BC across land use factors were much lar-adjusted ger than the relative differences in PM concen-
(C ) = (C ) / CFixed Site
2.5
/ CFixed Sitei j i j (1)j trations, suggesting that PM2.5 in the GAMA may
not be affected by community/local sources (such as
where (Ci)j is the PM2.5 or BC concentration meas- vehicle tailpipe emissions and trash burning) asmuch
ured at the ith rotating site in the jth measurement as BC.
4
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
Figure 1. Year-long (fixed) and week-long (rotating) monitoring locations. The colors indicate the integrated PM2.5 concentration
relative to the World Health Organization (WHO) Air Quality Guidelines (AQG) for PM2.5 (IT= interim target). The average
concentrations at the fixed sites represent the overall mean of 52 weeks, while the rotating sites represent seasonally-adjusted
values (also representing estimated annual means). Major and secondary/tertiary roads are from OpenStreetMap (downloaded
2019) and the GAMA boundary from Ghana Statistical Service.
Figure 2. Season-adjusted mean PM2.5 concentration at rotating sites by land-use categories. The solid and dashed horizontal
lines show WHO annual (10 µg m−3) and 24-h (25 µg m−3) AQG for PM2.5, respectively.
5. Temporal patterns at CBI areas (figure 3). Similarly, annual mean BC
concentrations were lowest at the peri-urban site and
5.1. Annual and seasonal patterns in PM2.5 highest at CBI sites (table 2).
Mean (SD) annual PM2.5 concentrations across the By season, the mean PM2.5 and BC con-
ten year-long (fixed) sites was 37 (40) µg m−3 centrations during the Harmattan (89 µg m−3
and ranged from site-type specific annual means of and 12 × 10−5m−1) were 4- and 2-fold higher
26 µg m−3 at the peri-urban site, 32–40 µg m−3 at than the non-Harmattan period (23 µg m−3 and
medium/low-density residential sites, 35–40 µg m−3 6 × 10−5m−1), respectively (figure 4). The absolute
at high-density residential sites, and 37–43 µg m−3 mean difference in PM2.5 concentrations between the
5
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
Table 1. Season-adjusted PM2.5 and BC concentrations at rotating sites by land-use categories.
PM (µg m−32.5 ) BC (1× 10−5m−1)
Site type (no. of sites) Mean (SD) Range Mean (SD) Range
All rotating sites (n= 126) 31 (10) 15–67 5 (2) 1–14
CBI (n= 23) 37 (10) 23–67 7 (3) 1–14
High-density (n= 28) 36 (10) 21–67 6 (2) 2–10
Medium/low-density (n= 47) 28 (7) 15–54 4 (1) 1–8
Peri-urban (n= 28) 26 (11) 16–56 3 (1) 1–6
Figure 3.Mean annual PM2.5 concentrations (bars; colored by site-type) and mean concentrations by season (Harmattan vs
non-Harmattan). The solid horizontal line shows the WHO annual AQG of 10 µg m−3. The dotted line represents the magnitude
of the difference between seasonal non-Harmattan and Harmattan mean concentrations. CBI: Commercial, business and
industrial areas. Sites: N1 West at Lapaz (N1W) and Tema Motorway (TMW) are at the west and east ends of the multi-lane N1
motorway; Asylum Down (AD) is on the Ring Road Central; Jamestown (JT) and Nima (NM) are low-income, densely populated
and high biomass use neighborhoods in south and middle of AMA; Taifa (TF) is an emerging neighborhood north of the city;
Labadi (LA) is an indigenous Ga community along on the Coast; East Legon (EL) is a high-income neighborhood next to the
University of Ghana Campus. Previously residential streets in EL now host large corporate, commercial and small business
ventures; Ashaiman (ASH) is an emerging neighborhood next to the port city of Tema; and University of Ghana Hill (UGH) is
located on top of the quiet Legon Hill.
Harmattan and non-Harmattan periods at each site Harmattan period is noteworthy as it indicates that
were between 56 and 71 µg m−3 (figure 3). While meteorological conditions likely magnify local emis-
the absolute levels were higher during the Harmat- sions and lead to higher concentrations.
tan, both periods showed substantial relative spatial
variability. The peri-urban site recorded the highest 5.2. Day of the week pattern
seasonal mean difference in PM2.5 concentrations Using the minute-by-minute continuous data, we
while sites in high-density residential neighborhoods found no differences in mean PM2.5 concentra-
recorded the lowest. For each measurement month, tions between day of the week (Monday–Sunday)
the peri-urban site consistently registered the low- nor between weekdays and weekends in the GAMA,
est PM2.5 and BC levels (figure 4). Like PM2.5, BC regardless of whether the data were from the fixed or
levels also increased during the Harmattan months rotating sites or both (p > 0.05). Although Sundays
(figure 4(b)), and both showed higher variability showed slightly lower mean PM2.5 overall, the mean
in the Harmattan as indicated by the sample SD. difference (3 µg m−3) was not significant (p= 0.57).
The overall observed doubling of BC levels in the The absence of between-day of the week variation in
6
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
Table 2. Annual and seasonal PM2.5 and BC concentrations at fixed (yearlong) sites by land-use categories.
PM2.5 (µg m
−3) BC (1× 10−5m−1)
Site type (no. of sites) Season Mean (SD) Range Mean (SD) Range
Fixed sites (n= 10) Annual 37 (40) 6–266 7 (4) 1–25
Harmattan 89 (64) 24–266 12 (5) 3–25
Non-Harmattan 23 (7) 6–52 6 (3) 1–18
CBI (n= 3) Annual 40 (41) 17–266 11 (4) 3–25
Harmattan 94 (67) 28–266 16 (5) 5–25
Non-Harmattan 26 (5) 17–52 10 (3) 3–17
High-density (n= 2) Annual 38 (37) 16–231 7 (3) 3–21
Harmattan 87 (63) 26–231 12 (4) 5–21
Non-Harmattan 26 (6) 16–41 6 (2) 3–12
Medium/low-density (n= 4) Annual 36 (41) 11–245 6 (4) 1–22
Harmattan 88 (64) 24–245 10 (4) 3–22
Non-Harmattan 21 (7) 11–51 5 (2) 1–18
Peri-urban (n= 1) Annual 26 (41) 6–217 3 (3) 1–14
Harmattan 81 (71) 25–217 7 (4) 3–14
Non-Harmattan 12 (4) 6–26 2 (1) 1–4
Figure 4.Weekly integrated PM2.5 (A) and BC (B) concentrations at the fixed sites averaged by site-types across measurement
months. Bars are standard deviations of the weekly measurements in that month. The horizontal line in (A) shows the WHO
annual AQG of 10 µg m−3.
PM2.5 in the GAMAwas consistent across all land-use PM2.5 concentrations at all sites rose around 03:00
categories (figure S4). daily, peaking at about 06:00, followed by a gradual
decline to their lowest values around 10:00. Levels
5.3. Diurnal patterns remained fairly stable between 10:00 and 15:00, after
PM2.5 concentrations from all sites showed strong which the concentration slowly increased with a rel-
bimodal variability across time of day, and was con- atively smaller peak around 18:00–19:00. There was
sistent over land-use areas and by season (figure 5). about an hour delay in the timing of the peaks
7
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
Figure 5. Diurnal patterns of PM2.5 concentration across land-use categories. The minute-by minute measurements from all 4592
site-days over the measurement period were averaged. The solid and dashed lines represent Non-Harmattan and Harmattan
seasons, respectively. The horizontal line represents the WHO 24 h AQG of 25 µg m−3.
during the Harmattan and the smaller early even- residential neighborhoods of JT and NM, where
ing peak was less pronounced compared to the non- PM2.5 levels decreased on average by ∼60%. We
Harmattan period. In general, average PM2.5 con- observed a smaller reduction (35%) in the middle-
centrations at nighttime (18:00–05:59) were slightly income AD, but slight increase (21%) in high-income
higher than daytime levels (37 vs 34 µg m−3). Dur- EL where there are a mix of residences and corpor-
ing these periods, biomass is burned in some neigh- ate, commercial and small businesses. The observed
borhoods for residential and small-scale commercial increase in high-income EL could also come from an
purposes, such as cooking street food and bakery overall increase in local commercial activities.
operation.
6. Discussion
5.4. Change in PM2.5 concentration since 2006/2007
In 2006/2007, Dionisio and colleagues [23] recor- We conducted a large-scale measurement campaign,
ded large variability (with wide SDs) in mean annual and a detailed analysis of the spatial and temporal
PM2.5 in four residential neighborhoods of varying patterns of ambient PM2.5 and BC pollution in the
SES and biomass use within the AMA, with values SSA city of Accra (1500 km2). We found a reduc-
ranging from28µgm−3 in the affluent neighborhood tion in PM2.5 pollution when compared with a dec-
of East Legon (EL), and 57 µg m−3 in middle-income ade ago, but the present levels exceed local and inter-
Asylum Down (AD), to >70 µg m−3 in low-income, national public health guidelines by ∼2–4 folds. Our
densely populated Nima (NM) and Jamestown (JT) data show that PM2.5 pollution in Accra is becom-
(figure 6). In the current study (2019/2020), themean ing more uniform across communities, similar to cit-
annual PM2.5 concentrations were lower at the same ies in Europe and North America where PM2.5 is a
locations, and ranged from 34 µg m−3 at EL to regional pollutant and not as affected by community
40 µg m−3 in JT. This suggests a reduction (and sources as in the past. Nonetheless, there remain some
more uniformity/plateau) in PM pollution in the city. disparities in PM2.5 and BC concentrations within
The largest reductions were observed in high-density the city with significant seasonal variations. The CBI
8
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
Figure 6. Comparison of mean annual PM2.5 concentrations between 2006/2007 (sample range: 12–1292 µg m−3) and 2019/2020
(sample range: 13–245 µg m−3) measurement campaigns. Bars are standard deviation of all measurements in that study period,
including Harmattan. The horizontal line shows the WHO annual AQG (10 µg m−3).
(mostly influenced by traffic) and high-density res- as well as higher PM2.5 and BC concentrations at
idential (mostly influenced by traffic and biomass industrial and high-density residential sites in Ibadan,
use) areas were 35%–50% more polluted relative to Nigeria [48]. Within the sub-region, mean annual
peri-urban sites, which typically experience relatively PM2.5 concentrations in our study are higher than
lower traffic, commercial and industrial activities. annual averages observed for equally sprawling cities
Within-year changes in local meteorology produced like Ibadan, Nigeria (24–33 µg m−3) [48]. In global
distinct seasonality in PM2.5 and BC pollution, with comparisons, mean annual PM2.5 in the GAMA were
concentrations during the Harmattan about twice substantially higher than those found in large cities
that of the non-Harmattan period. Diurnal concen- of high-income countries such as New York, USA (5–
trations of PM2.5 peaked at dawn and dusk at times 11 µg m−3) [49] and London, UK (5–15 µg m−3)
that coincided with the morning/evening traffic rush [50], but lower than annual averages in Asian cities
and biomass use hours. such as Beijing, China (53–112 µg m−3) [51] and
In this city-wide analysis, our findings are con- Delhi, India (122–148µgm−3) [52]. Althoughwe did
sistent with previous smaller studies conducted in the not study the composition and relative contribution
AMA that also reported higher PM2.5 and BC concen- of different sources to PM2.5 pollution, the high
trations at locations with persistent road-traffic and BC levels observed at CBI and high-density areas
in densely populated neighborhoods [17, 23]. Sim- suggest that vehicle emissions and biomass burning
ilar to our results, studies in other large SSA cities are important determinants of PM2.5 pollution in
have reported higher PM2.5 concentrations in loca- the GAMA. Our observed city-wide spatial patterns
tions with high road-traffic volumes in the CBI areas aligns with the work of Zhou et al, 2013, which doc-
of Nairobi, Kenya [46] and Kampala, Uganda [47]; umented major contributions from traffic, road dust,
9
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
and biomass burning to PM2.5 and BC pollution in 7. Conclusion
the Accra city core.
Elevated PM2.5 during the Harmattan season As urbanization in SSA continues and cities are faced
is expected across West Africa given the influ- with the challenge of managing air quality from
ence of transported mineral dust from Sahara diverse sources [8], data on local air pollution and
desert [5, 17, 23, 48, 53–55]. However, the observed sources are urgently needed to enable evidence-based
increase in BC concentrations, a product of incom- policy efforts to protect public health. To avoid sim-
plete combustion, in the Harmattan season also sug- ilar poor air quality challenges seen in Asian cities,
gests that changes to local meteorological conditions systematic air quality management plans are needed
during this period (e.g. high temperature, low wind- to further reduce current air pollution levels. Success-
speed and absence of precipitation) may produce ful air pollution mitigation efforts will require atten-
stagnant conditions that substantially amplify local tion to land-use planning and accounting for season-
anthropogenic emissions [56]. The daily PM2.5 cycle ality. Besides the direct impact of Harmattan on PM
of bimodal pattern with peaks in the mornings and pollution, changes in the local meteorology during
evenings, provides further support for the influence this period suggests almost no room for worsening
of rush hour traffic, biomass combustion as well as emissions from local sources during this period. Our
pollution build-up due to temperature inversion and study provides compelling evidence for systematic air
variations in meteorological conditions between day pollution monitoring as well as implementation of
and nighttime hours and seasons [23, 57]. It is likely Ghana’s air quality policy initiatives aimed at protect-
that the observed improvements in PM2.5 pollution, ing health and improving air quality in the GAMA.
especially in high-density neighborhoods that also
tended to have high-biomass use, was due to gradual Data availability statement
reductions in biomass use. Both behavioral and
policy changes accompanying economic improve- The data that support the findings of this study are
ments might have brought about reduction in local available upon reasonable request from the authors.
community emissions. For instance, there is evidence
of downward trend in the proportion of households
utilizing biomass fuel for cooking, with a significant Acknowledgments
switch from predominantly wood (more polluting)
to charcoal and gas, which are less polluting [58]. In This work is supported by the Pathways to Equit-
terms of policy, Ghana currently has in place penalties able Healthy Cities grant from the Wellcome Trust
on the importation of used and old vehicles to curb [209376/Z/17/Z]. For the purpose ofOpenAccess, the
traffic emissions in general [20]. Therefore, incentiv- author has applied a CC BY public copyright license
izing transition to cleaner fuels could further improve to any Author Accepted Manuscript version arising
air quality in the GAMA [6, 59, 60]. With sustained from this submission. This work is also supported by
economic and urban expansion, vehicle ownership aGCRFDigital Innovation forDevelopment in Africa
in Ghana is increasing by 10% annually [61] and network grant fromUKRI [EP/T029145/1].We thank
the GAMA accounts for 60% of the total num- the Accra residents who permitted us to install mon-
ber of registered vehicles [21, 22]. Without invest- itors on their property and the staff at Physics Depart-
ments in infrastructure (e.g. improved road net- ment, University of Ghana for their support in organ-
works) and environmental management programs, izing the laboratory used during this project.
this growth could lead to higher vehicular emissions
than observed previously [17], which will worsen air ORCID iDs
quality over time. Attaining cleaner air in Accra (i.e.
meeting WHO guideline levels) is likely to require Sierra N Clark https://orcid.org/0000-0002-8592-
implementation of Ghana’s proposed traffic-related 3466
air pollution reduction strategies such as the bus- Allison Hughes https://orcid.org/0000-0002-
rapid transit system, development of vehicle emission 9912-6935
standards, and maintaining the current penalties on Raphael E Arku https://orcid.org/0000-0001-
importation of old vehicles while providing incent- 8914-8463
ives and rebates on new cars [21, 22]. Given that
PM2.5 and BC pollution are worse during Harmat- References
tan, there is a specific need for additional air quality
[1] Shaddick G, Thomas M L, Mudu P, Ruggeri G and Gumy S
management plans in this period. Land stabilization 2020 Half the world’s population are exposed to increasing
interventions (such as covering road surfaces with air pollution npj Clim. Atmos. Sci. 3 1–5
dust suppressants, and sweeping/ washing roads) and [2] United Nations, Department of Economic and Social Affairs
building extensive green walls of forest to act as pro- 2018 Population division (UN-DESA). world urbanization
prospects The 2018 Revision (available at: https://
tective barrier around the city can reduce dust particle population.un.org/wup/Publications/Files/WUP2018-
load [62]. Report.pdf) (Accessed 29 May 2020)
10
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
[3] Güneralp B, Lwasa S, Masundire H, Parnell S and Seto K C [24] Ghana Statistical Service (GSS) 2010 Population and housing
2018 Urbanization in Africa: challenges and opportunities census final results 2012 (available at: www.ghanahero.com/
for conservation Environ. Res. Lett. 13 015002 ARCHIVE_1/Mission-Ghana-Planning-Inventory/2010_
[4] Turok I and McGranahan G 2013 Urbanization and POPULATION_AND_HOUSING_CENSUS_FINAL_
economic growth: the arguments and evidence for Africa RESULTS.pdf) (Accessed 14 March 2019)
and Asia Environ. Urban 25 465–82 [25] Addae B and Oppelt N 2019 Land-use/land-cover change
[5] Naidja L, Ali-Khodja H and Khardi S 2018 Sources and levels analysis and urban growth modelling in the Greater Accra
of particulate matter in North African and sub-Saharan Metropolitan Area (GAMA), Ghana Urban Sci. 3 1–20
cities: a literature review Environ. Sci. Pollut. Res. [26] Akubia J E K and Bruns A 2019 Unravelling the frontiers of
25 12303–28 urban growth: spatio-temporal dynamics of land-use change
[6] Amegah A K and Agyei-Mensah S 2017 Urban air pollution and urban expansion in greater Accra metropolitan area
in sub-Saharan Africa: time for action Environ. Pollut. (Ghana) Land 8 1–23
220 738–43 [27] Clark S N et al 2020 High-resolution spatiotemporal
[7] Petkova E P, Jack D W, Volavka-Close N H and Kinney P L measurement of air and environmental noise pollution in
2013 Particulate matter pollution in African cities Air Qual. sub-Saharan African cities: pathways to equitable health
Atmos. Health 6 603–14 cities study protocol for Accra, Ghana BMJ Open
[8] Hanif I 2018 Impact of economic growth, nonrenewable and 10 e035798
renewable energy consumption, and urbanization on carbon [28] Ghana Statistical Service (GSS) 2010 Population and
emissions in sub-Saharan Africa Environ. Sci. Pollut. Res. Housing Ghana Stat Serv pp 1–117 (available at:
25 15057–67 www.statsghana.gov.gh/docfiles/2010phc/Census2010_
[9] Katoto P D M C et al 2019 Ambient air pollution and health Summary_report_of_final_results.pdf%0Ahttp://
in sub-Saharan Africa: current evidence, perspectives and a unstats.un.org/unsd/demographic/sources/census/
call to action Environ. Res. 173 174–88 2010_phc/Ghana/Provisional_results.pdf)
[10] Egondi T, Muindi K, Kyobutungi C, Gatari M and Rocklöv J [29] Volckens J, Quinn C, Leith D, Mehaffy J, Henry C S and
2016 Measuring exposure levels of inhalable airborne Miller-Lionberg D 2017 Development and evaluation of an
particles (PM2.5) in two socially deprived areas of Nairobi, ultrasonic personal aerosol sampler Indoor Air 27 409–16
Kenya Environ. Res. 148 500–6 [30] Burrowes V J et al 2020 Comparison of next-generation
[11] Amegah A K 2018 Proliferation of low-cost sensors. What portable pollution monitors to measure exposure to PM 2.5
prospects for air pollution epidemiologic research in from household air pollution in Puno, Peru Correspondence
sub-Saharan Africa? Environ. Pollut. 241 1132–7 Indoor Air 30 445–58
[12] Simwela A, Xu B, Mekondjo S S and Morie S 2018 Air quality [31] Pillarisetti A et al 2019 Measuring personal exposure to fine
concerns in Africa: a literature review Int. J. Sci. Res. Publ. particulate matter (PM 2.5) among rural Honduran women:
8 588–94 a field evaluation of the ultrasonic personal aerosol sampler
[13] Altieri K E and Keen S L 2019 Public health benefits of (UPAS) Environ. Int. 123 50–3
reducing exposure to ambient fine particulate matter in [32] Afshar-Mohajer N, Foos R, Ramachandran G and Volckens J
South Africa Sci. Total Environ. 684 610–20 2021 Field evaluation of the ultrasonic personal aerosol
[14] Curto A et al 2019 Predictors of personal exposure to black sampler (UPAS) for respirable dust exposure in a taconite
carbon among women in southern semi-rural Mozambique mine Ann. Work Exposures Health 65 127–35
Environ. Int. 131 1–14 [33] Arku R E, Birch A, Shupler M, Yusuf S, Hystad P and
[15] Nducol N, Siaka Y F T, Yakum-Ntaw S Y, Manga J D and Brauer M 2018 Characterizing exposure to household air
Vardamides J C 2021 Preliminary study of black carbon pollution within the Prospective Urban Rural Epidemiology
content in airborne particulate matters from an open site in (PURE) study Environ. Int. 114 307–17
the city of Yaoundé, Cameroon Environ. Monit. Assess. 193 [34] Bulot F M J et al 2019 Long-term field comparison of
1–11 multiple low-cost particulate matter sensors in an outdoor
[16] Janssen N A H et al 2011 Black carbon as an additional urban environment Sci. Rep. 9 1–13
indicator of the adverse health effects of airborne particles [35] Cho E M et al 2019 Reliability of low-cost, sensor-based fine
compared with PM10 and PM2.5 Environ. Health Perspect. dust measurement devices for monitoring atmospheric
119 1691–9 particulate matter concentrations Int. J. Environ. Res. Public
[17] Zhou Z et al 2013 Chemical composition and sources of Health 16 1–10
particle pollution in affluent and poor neighborhoods of [36] Badura M, Batog P, Drzeniecka-Osiadacz A and Modzel P
Accra, Ghana Environ. Res. Lett. 8 044025 2018 Evaluation of low-cost sensors for ambient PM2.5
[18] Aboh I J K et al 2009 Identification of aerosol particle monitoring J. Sens. 2018 1–16
sources in semi-rural area of Kwabenya, near Accra, [37] Arku R E et al 2015 Personal particulate matter exposures
Ghana, by EDXRF techniques X-Ray Spectrom. and locations of students in four neighborhoods in Accra,
38 348–53 Ghana J. Exposure Sci. Environ. Epidemiol. 25 557–66
[19] Dionisio K et al 2010 Within-neighborhood patterns and [38] Van Vliet E D S and Kinney P L 2007 Impacts of roadway
sources of particle pollution: mobile monitoring and emissions on urban particulate matter concentrations in
geographic information system analysis in four sub-Saharan Africa: new evidence from Nairobi, Kenya
communities in Accra, Ghana Environ. Health Perspect. Environ. Res. Lett. 2 045028
118 607–13 [39] Saraswat A, Apte J S, Kandlikar M, Brauer M, Henderson S B
[20] Rooney M S et al 2012 Spatial and temporal patterns of and Marshall J D 2013 Spatiotemporal land use regression
particulate matter sources and pollution in four models of fine, ultrafine, and black carbon particulate matter
communities in Accra, Ghana Sci. Total Environ. in New Delhi, India Environ. Sci. Technol. 47 12903–11
435–436 107–14 [40] Segersson D et al 2017 Health impact of PM10, PM2.5 and
[21] Musah B I, Peng L and Xu Y 2020 Urban congestion and black carbon exposure due to different source sectors in
pollution: a quest for cogent solutions for Accra City IOP Stockholm, Gothenburg and Umea, Sweden Int. J. Environ.
Conf. Ser.: Earth Environ. Sci. 435 1–8 Res. Public Health 14 11–4
[22] Ghana Environmental Protection Agency (GEPA) 2018 The [41] Jeronimo M et al 2020 Analysis of black carbon on filters by
greater Accra metropolitan areas air quality management image-based reflectance Atmos. Environ. 223 1–8
plan pp 1–64 [42] Cyrys J et al 2003 Comparison between different
[23] Dionisio K L et al 2010 Air pollution in Accra traffic-related particle indicators: elemental carbon (EC),
neighborhoods: spatial, socioeconomic, and temporal PM2.5 mass, and absorbance J. Exposure Anal. Environ.
patterns Environ. Sci. Technol. 44 2270–6 Epidemiol. 13 134–43
11
Environ. Res. Lett. 16 (2021) 074013 A S Alli et al
[43] Quincey P 2007 A relationship between black smoke index event in an urban environment Environ. Monit. Assess.
and black carbon concentration Atmos. Environ. 41 7964–8 190 1–19
[44] Rivas I et al 2014 Child exposure to indoor and outdoor air [54] He C, Breuning-Madsen H and Awadzi T W 2007
pollutants in schools in Barcelona, Spain Environ. Int. Mineralogy of dust deposited during the Harmattan season
69 200–12 in Ghana Geogr. Tidsskr 107 9–15
[45] R Core Team 2020 R: A language and environment for [55] Abiye O E et al 2020 Chemical footprints of harmattan dust
statistical computing R Foundation for Statistical and traffic corridor particulates monitored at two
Computing, Vienna, Austria (available at: www.r- environmentally distinct geopolitical zones in Nigeria
project.org/) Environ. Sci. Pollut. Res. 27 39317–329
[46] Kinney P L et al 2011 Traffic impacts on PM2.5 air quality in [56] Querol X et al 2019 Monitoring the impact of desert dust
Nairobi, Kenya Environ. Sci. Policy 14 369–78 outbreaks for air quality for health studies Environ. Int.
[47] Kirenga B J et al 2015 The state of ambient air quality in two 130 1–16
ugandan cities: a pilot cross-sectional spatial assessment Int. [57] Onyango S, Parks B, Anguma S and Meng Q 2019
J. Environ. Res. Public Health 12 8075–91 Spatio-temporal variation in the concentration of inhalable
[48] Akinlade G O et al 2015 Spatial and temporal variations of particulate matter (PM10) in Uganda Int. J. Environ. Res.
the particulate size distribution and chemical composition Public Health 16 1–12
over Ibadan, Nigeria Environ. Monit. Assess. 187 1–14 [58] Arku R E et al 2016 Geographical Inequalities and social and
[49] Squizzato S, Masiol M, Rich D Q and Hopke P K 2018 environmental risk factors for under-five mortality in Ghana
PM2.5 and gaseous pollutants in New York State during in 2000 and 2010: bayesian spatial analysis of census data
2005–2016: spatial variability, temporal trends, and PLoS Med. 13 1–14
economic influences Atmos. Environ. 183 209–24 [59] Das I, Pedit J, Handa S and Jagger P 2018 Household air
[50] Greater London Authority 2020 Air pollution monitoring pollution (HAP), microenvironment and child health:
data in London (available at: www.london.gov.uk) (Accessed strategies for mitigating HAP exposure in urban Rwanda
23 Jul 2020) Environ. Res. Lett. 13 045011
[51] Kong L and Tian G 2020 Assessment of the spatio-temporal [60] Sharma A and Chung C E 2014 Climatic benefits of black
pattern of PM2.5 and its driving factors using a land use carbon emission reduction when India adopts the
regression model in Beijing, China Environ. Monit. Assess. US onroad emission level Future Cities Environ.
192 1–19 69 1–13
[52] Gorai A K, Tchounwou P B, Biswal S S and Tuluri F 2018 [61] World Bank Group 2014 Rising through cities in ghana:
Spatio-temporal variation of particulate matter(PM2.5) ghana urbanization review overview report (available at:
concentrations and its health impacts in a Mega City, Delhi www.worldbank.org) (Accessed 13 August 2020)
in India Environ. Health Insights 12 1–9 [62] Enete I 2012 Harmattan dust: composition, characteristics
[53] Adeniran J A, Aremu A S, Saadu Y O and Yusuf R O 2018 and effects on soil fertility in Enugu, Nigeria Br. J. Appl. Sci.
Particulate matter concentration levels during intense haze Technol. 2 72–81
12