Research article

Air pollution in Kigali, Rwanda: spatial and temporal

variability, source contributions, and the impact of
car-free Sundays

R Subramanian 1,2,7,*, Abdou Safari Kagabo3, Valérien Baharane3, Sandrine Guhirwa3,

Claver Sindayigaya3, Carl Malings 1,2,8, Nathan J Williams 4,7, Egide Kalisa 5, Haofan Li1,
Peter Adams 1, Allen L Robinson 1, H. Langley DeWitt6, Jimmy Gasore3,7, and
Paulina Jaramillo 4,7
1
Center for Atmospheric Particle Studies, Carnegie Mellon University, Pittsburgh, PA 15218, USA
2
Observatoire Sciences de l’Univers-Enveloppes Fluides de la Ville à l’Exobiologie, CNRS UMS3563, France
3
Department of Physics, University of Rwanda, Kigali, Rwanda
4
Department of Engineering and Public Policy, Carnegie Mellon University, Pittsburgh, PA, USA
5
Institute for Applied Ecology New Zealand, School of Science, Auckland University of Technology, Auckland 1142, New Zealand
6
Center for Global Change Science, Massachusetts Institute of Technology, Cambridge, MA, USA
7
Kigali Collaborative Research Center, Kigali, Rwanda
8
NASA Postdoctoral Program Fellow, Goddard Space Flight Center, Greenbelt, Maryland, MD 20771, USA
*Corresponding author, E-mail: [email protected] [email protected]

Received: 3 March 2020 - Reviewed: 8 April 2020 - Accepted: 17 August 2020

https://doi.org/10.17159/caj/2020/30/1.8023

Abstract
Ambient air pollution, particularly fine particulate mass (PM2.5) and ozone (O3), is associated with premature human mortality and
other health effects, but monitoring is scarce to non-existent in large parts of Africa. Lower-cost real-time affordable multi-pollutant
(RAMP) monitors and a black carbon (BC) monitor were deployed in Kigali, Rwanda to fill the air quality data gap here. PM2.5 data were
corrected using data from a coincident, short-term campaign that used standard filter-based gravimetry, while gas data were verified
by collocation with reference carbon monoxide (CO) and O3 monitors at the Rwanda Climate Observatory at Mt Mugogo, Rwanda. Over
March 2017-July 2018, the ambient average PM2.5 in Kigali was 52 µg/m3, significantly higher than World Health Organization (WHO)
Interim Target 1. Study average BC was 4 µg/m3, comparable to mid-sized urban areas in India and China and significantly higher
than BC in cities in developed countries. Spatial variability across various urban background sites in Kigali appears to be limited,
while PM2.5 at Mt Mugogo is moderately correlated with PM2.5 in Kigali. A sharp diurnal profile is observed in both PM2.5 and BC, with
the Absorption Angstrom Exponent (AAE) indicating that the morning peak is associated with rush-hour traffic-related air pollution
(TRAP) while the late evening peak can be attributed to both traffic and domestic biofuel use. PM2.5 in the dry seasons is about two
times PM2.5 during the following wet seasons while BC is 40-60% higher. Local sources contribute at least half the ambient PM2.5 during
wet seasons and one-fourth during dry seasons. Traffic restrictions on some Sundays appear to reduce PM2.5 and BC by 10-12 µg/m3
and 1 µg/m3 respectively, but this needs further investigation. Dry season ozone in Kigali can exceed WHO guidelines. These lower-
cost monitors can play an important role in the continued monitoring essential to track the effectiveness of pollution-control policies
recently implemented in Rwanda.

Keywords
fine particulate matter, ozone, black carbon, sub-Saharan Africa, urban air pollution, vehicular emissions, biofuel emissions, low-cost
sensors

Introduction in 2016, ambient air pollution caused about three thousand

Ambient air pollution, especially fine particulate mass (PM2.5) deaths in Rwanda (Brauer et al., 2012; WHO, 2018). However,
and ozone (O3), has been associated with premature human such estimates can be uncertain because exposure is inferred
mortality (Dockery et al., 1993; Jerrett et al., 2009; Laden et al., from satellite estimates. There has been no long-term ground-
2006). The World Health Organization (WHO) estimates that based monitoring in major cities like Kigali to validate estimated

CLEAN AIR JOURNAL © 2020. The Author(s). Published under a

Volume 30, No 2, 2020 1 Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 2 of 15

exposures. The situation is similar in many other sub-Saharan the OBM analysis of the pseudo-observations comparable to the
African countries (Kalisa et al., 2019; Petkova et al., 2013). This model-computed regional/local divide.
lack of monitoring due to resource limitations also hampers
scientific understanding of the sources contributing to air Traffic restrictions such as car-free days, low-emission zones
pollution in these countries, which is essential to formulating (LEZ), and “odd/even” policies have been used in cities worldwide
effective environmental management policies. to reduce air pollution. The evidence that such policies reduce
human exposure is mixed. No effect was observed on air
In Rwanda, biomass use accounts for 85% of energy quality in Mestre-Venice (Masiol et al., 2014). London’s LEZ saw
consumption, in the form of wood and charcoal (MININFRA, decreases in nitrogen dioxide (NO2), but no significant changes
2018). Wood is used in rural households and charcoal in urban in PM2.5 or PM10 (Mudway et al., 2019). Extensive traffic control
households. Petroleum in the form of transportation fuel, measures in Beijing during the 2008 Olympics reduced median
liquified petroleum gas (LPG) for household cooking, and black carbon (BC) concentrations by as much as 50% (Wang et
electricity generation accounts for 13% of national energy al., 2009). The Rwandan government, to promote social welfare,
consumption (MININFRA, 2018). As of 2018, 77% of on-road has recently implemented “car-free” Sundays, wherein major
vehicles (excluding motorcycles) were manufactured before roads are blocked off and people take part in group exercises on
2005 (Duhuze, 2018). As a result, air pollution in Kigali can car-free streets starting at 7 AM and ending either at 10 AM or at
be significantly higher at roadside locations than at urban noon. In 2017, the first Sunday of each month was designated
background locations (Kalisa et al., 2018). Rwandan air quality “car-free”, while in 2018 that was expanded to the first and third
is also influenced by regional forest fires and seasonal weather Sundays of each month. Additionally, to curb the use of older
patterns (DeWitt et al., 2019). Rwanda’s electricity generation imported cars and harmonize duty structures with other East
(218 MW) is composed of 45% hydropower, diesel and heavy fuel African countries, in 2017 the Rwandan government increased
oil at 27%, methane (14%), peat (7%), and solar (6%). Diesel is duties on cars, with the increase depending on the vehicle age
used to fuel peaking power stations and for backup generation (RRA, 2017). The vehicle import rate dropped by 20% in the
during power outages, which can be significant contributors to first half of 2017, likely connected to the higher import duties
ambient air pollution (Farquharson et al., 2018; Subramanian et (Ngabonziza, 2017). Monitoring is required to quantify the
al., 2018). Some of these sources were identified by Henninger impact of these policies on air quality in Kigali.
(2013) using scanning electron microscopy of filter samples.
Here, we show how low-cost sensors can improve scientific
To more accurately quantify source contributions (a requirement understanding of air quality in resource-challenged countries.
for effective air quality management), source apportionment is The high time resolution of the RAMPs and BC monitor enables
often conducted with chemical mass balance (CMB) or receptor an examination of diurnal patterns in each season, which is not
modeling (e.g. positive matrix factorization) using organic possible with integrated daily filter samples. We apply OBM to
molecular markers (Shrivastava et al., 2007; Subramanian et al., the RAMP and BC datasets to get a preliminary estimate of the
2007) or aerosol mass spectrometry (AMS) (Zhang et al., 2011). regional and local contribution to Kigali air pollution in each
However, such studies require extensive sample collection season. Measurements at multiple “urban background” sites
and offline analysis or long-term deployment of expensive within Kigali allow us to examine intra-urban variability in air
equipment, especially if sources can vary between seasons. pollution at non-roadside locations across Kigali, unlike the
Pikridas et al. (2013) find that an observation-based method urban background/roadside comparison by Kalisa et al. (2018).
(OBM), which uses the temporal pattern of pollution measured Long-term monitoring with RAMP monitors allows examination
with even a low-cost PM monitor, closely replicates the regional/ of the seasonal variability in ambient air pollution across
urban divide based on PM2.5 composition measurements inside multiple dry and wet seasons. Furthermore, we evaluate the
and upwind of Patras, Greece. Diamantopoulou et al. (2016) use impact of the “car-free Sunday” policy on air pollution. The multi-
regional air quality modeling to simulate observations and find wavelength aerosol light absorption from the BC monitor helps

Table 1: Site descriptions and deployment periods for the measurements reported in this study.
Site UR-CST CMU-Africa Gacuriro Belle Vue RCO
Type Urban Urban Urban Urban Rural
Latitude -1.96279 -1.94455 -1.9219 -1.92563 -1.58625
Longitude 30.06473 30.08961 30.09389 30.0924 29.56568
Instrument and deployment period
RAMP #140 March-June 2017 July-Dec 2017
RAMP #145 March-June 2017 July-Dec 2017
RAMP #152 July 17-July 18
BC-1054 July 17-April 18

CLEAN AIR JOURNAL

Volume 30, No 2, 2020 2 © 2020. The Author(s). Published under a
Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 3 of 15

us qualitatively identify contributions from biomass burning the instrument inoperable. The Gacuriro and Belle Vue sites are
and fossil fuel combustion. Comparison of urban and rural air about 5 km from the UR-CST site.
pollution provides insight into the impact of urbanization on air
quality. We end with recommendations on ways to implement The RAMPs and gas sensor calibration
and improve such studies in Rwanda and other countries in the The RAMPs (Figure S1) were manufactured by Sensevere (now
Global South. owned by Sensit Technologies, Valparaiso, IN, USA) and cost
about US$ 3,000 each at the time of purchase (base unit without
an external PM sensor). The RAMP monitors and calibration
methodologies are described in previous work (Malings et al.,
2019, 2020; Subramanian et al., 2018; Zimmerman et al., 2018).
Briefly, the RAMP uses passive Alphasense (UK) electrochemical
sensors to measure CO, nitrogen dioxide (NO2), O3, and other
gases. The raw signals of the RAMP electrochemical gas sensors
(collected at 4 times per minute) are processed and averaged to
provide hourly ambient concentrations using generalized RAMP
(gRAMP) calibration models (Malings et al., 2019) developed in
Pittsburgh, Pennsylvania, USA. The gRAMP calibration models
are based on data from several RAMP monitors collocated
with reference gas monitors at the CMU campus in Pittsburgh
in 2017 and were shown to transfer better to other locations
in Pittsburgh than calibration models developed for individual
RAMPs. For CO, a quadratic regression (QR) gRAMP model is used
and for O3, a hybrid random forest/linear regression (“hybrid-
RF”) gRAMP model is used.

For local verification of CO and O3, RAMP #140 was collocated

with reference monitors for CO and CO2 (Picarro G2401) and O3
(Teledyne T400) at the Rwanda Climate Observatory (RCO) on
the summit of Mt Mugogo (about 70 km from Kigali and 2590 m
above sea level, DeWitt et al. (2019)) over July-December 2017.
Due to instrument malfunctions, collocated measurements are
available for only 60 days of this six-month period. “Mugogo”
linear regression, QR, and hybrid-RF models are developed
using a subset (four weeks) of the collocation data; as shown
Figure 1: Measurement sites in Rwanda (a) and specifically in Kigali (b)
where RAMPs were deployed for varying periods over 2017-2018. Maps in Zimmerman et al. (2018), a four week period is sufficient to
generated using Google Earth Pro. develop calibration models for these electrochemical sensors.
The remaining collocation data (32 days) are set aside to
provide an independent or unseen data set for testing model
Methodology performance, a practice previously established by our group
in Zimmerman et al. (2018) and Malings et al. (2019). The
Sampling locations performance of the Mugogo and the Pittsburgh gRAMP models
Starting in March 2017, RAMP monitors were deployed in Kigali is shown in Table 2. For CO, the gRAMP models showed slightly
at multiple locations (Figure 1) as summarized in Table 1. Kigali worse correlation (Pearson r) and normalized mean absolute
terrain is composed of hills and valleys; the RAMPs in this study error (CvMAE) than the Mugogo-based calibration models. For
were all located at hilltop sites. The Gacuriro and Belle Vue Estate O3, the gRAMP model is comparable to the better-performing
locations are both residential neighborhoods about 0.5 km Mugogo QR and hybrid-RF models for r and CvMAE. RAMP-
apart. The Carnegie Mellon University (CMU-Africa) site is about specific Mugogo models show lower bias than the gRAMP
2.5 km from Gacuriro, located in a commercial building complex models. During the dry season, most of the data from the gRAMP
near a major road and six stories above ground (this was the ozone model are within ±30% of the reference monitor data at
campus till late 2019). RAMP #152 and a Met-One 10-wavelength RCO (Figure S2).
BC monitor (BC-1054) were deployed in July 2017 on the roof
of the University of Rwanda’s five-story College of Science & RAMP-specific models trained on one RAMP may not transfer as
Technology building (UR-CST), where Kalisa et al. (2018) had well to other RAMPs as the gRAMP models (Malings et al., 2019).
collected filter-based samples for their urban background As most of the Kigali O3 data is from RAMPs not collocated at
location over April-June 2017. This manuscript focuses on RAMP Mugogo, the results presented here for ambient CO and O3 are
data collected at UR-CST between July 2017-July 2018. The BC based on the gRAMP models. As we do not have local verification
data ends in April 2018 as the pump malfunctioned, which made for NO, NO2, or SO2, these data are not presented here.

CLEAN AIR JOURNAL © 2020. The Author(s). Published under a

Volume 30, No 2, 2020 3 Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 4 of 15

Table 2: Summary statistics showing the performance of Pittsburgh-based gRAMP and RAMP-specific Mugogo calibration models, both tested on 32 days
of collocation at the Mugogo site. (These testing data were not used to build the Mugogo calibration models.)

CO Ozone
Model
Pearson r CvMAE Bias (ppb) Pearson r CvMAE Bias (ppb)
Pittsburgh gRAMP 0.82 0.19 15.7 0.57 0.24 6.02
Mugogo linear 0.87 0.16 -4.22 0.16 0.24 1.10
Mugogo QR 0.92 0.13 -1.83 0.63 0.20 0.30
Mugogo hybrid-RF 0.90 0.16 -6.28 0.56 0.22 0.32

PM2.5 measurements absorption enhancement due to BC mixing state (Bond et al.,

PM2.5 is measured using a Met-One neighborhood PM monitor 2006), which means the BC mass concentrations reported here
(NPM) paired with each RAMP. The NPM is a nephelometer with may be overestimates. The Absorption Angstrom Exponent (AAE)
a PM2.5 cyclone, an inlet heater to reduce humidity effects, and was calculated based on all ten wavelengths as the negative
a pump (flow rate 2 lpm). The PM2.5 data are processed using slope of the relationship between the wavelength-dependent
methods developed based on collocations of over two dozen attenuation and the wavelength in a log-log space (Moosmüller
NPMs with US Environmental Protection Agency (EPA) federal et al., 2011). For fresh combustion aerosol mainly composed of
equivalent method (FEM) beta attenuation monitors (BAM) at black carbon and for particles small relative to the wavelength,
an urban background location and a source-impacted location AAE is expected to be near unity. A compilation of studies
in Pittsburgh (Malings et al., 2020). Either a physical (assumed (Lack and Langridge, 2013) on fossil fuel emissions and urban
composition and hygroscopic growth with scaling to BAM pollution with BC as the dominant absorber yielded an average
values) or an empirical approach (a quadratic regression with AAE of 1.1±0.3 (± one standard deviation). The AAE for light-
raw data, temperature, and RH as the variables) is used to absorbing organic compounds (brown carbon or BrC) associated
convert as-reported NPM readings to “BAM-equivalent” PM2.5 with biomass burning or humic-like substances is higher, with
mass concentration (reported at 35% RH and 22 °C); both estimates ranging from 2 to 6 (Kirchstetter et al., 2004; Sun et al.,
methods performed similarly in Pittsburgh. For Rwanda, we use 2007). Hence, higher values of AAE can be used to qualitatively
the Pittsburgh-based physical approach, then apply a further identify periods when non-BC aerosol components such as
localized correction to account for differences in aerosol size light-absorbing dust or biomass burning containing BrC are
distribution and composition between Kigali and Pittsburgh. significant contributors to aerosol light absorption. We do not
This localized correction of the “BAM-equivalent” PM2.5 data attempt to quantify the contributions of fossil fuel and biomass
is based on a comparison of the Gacuriro RAMP with 24-hour burning emissions here as there can be significant uncertainty
integrated filter-based measurements in April 2017 collected associated with such methods, as described in DeWitt et al.
by Kalisa et al. (2018) at UR-CST about 5 km away (described (2019) and references therein.
in the SI). Briefly, the “BAM-equivalent” PM2.5 data (corrected
for hygroscopic growth but not for aerosol differences between
Kigali and Pittsburgh) are strongly correlated (correlation Results and discussion
coefficient, r2 = 0.77) with the filter-based PM2.5 for non-working
days (weekends, holidays, and car-free Sundays). However, Intra-urban variability across Kigali
the comparison is more scattered for working days (weekdays The UR-CST and Belle Vue Estate sites are about 5 km apart and
that are not holidays), which might indicate some intraurban have the largest paired data set across both dry (July-September)
variability related to local activities such as traffic and industrial and wet (October-November) seasons. An orthogonal distance
emissions that may be more prominent on working days. regression (ODR) fit of the paired hourly average PM2.5
Overall, the working day “BAM-equivalent” RAMP PM2.5 values concentrations yields a slope 0.996±0.012 and effectively zero-
are scaled up by 1.69 and the non-workday “BAM-equivalent” intercept. An ordinary least-squares fit (not forced through zero)
RAMP PM2.5 data are scaled up by 1.39 (Figure S3). has a correlation (r2) of 0.61. The UR-CST and Belle Vue Estate
sites in Kigali are typical of urban background hilltop locations
BC mass and AAE measurements (Figure 1), unlike the urban background/roadside comparison
The BC-1054 monitor (Met One Instruments, Inc.) deployed at where Kalisa et al. (2018) found significant differences. The
the UR-CST site measures light attenuation by a filter sample much smaller dataset of 337 paired hourly average PM2.5
at ten wavelengths between 370-950 nm. The as-reported BC measurements at the Gacuriro and CMU-Africa sites yields a
mass concentrations from this monitor were processed using slope of 0.967±0.03 (and an effectively zero intercept) with the
manufacturer-provided software (BC Load Correction 1.3.1), ODR fit. For the sake of simplicity (as the succeeding analysis
which corrects for known filter-loading artifacts (Kirchstetter focuses on longer-term comparisons), the measurements at all
and Novakov, 2007) using the algorithm developed by Virkkula four sites are averaged into a single time series providing hourly
et al. (2007). However, no correction is made for potential light average PM2.5 values from March 2017 to July 2018.

CLEAN AIR JOURNAL

Volume 30, No 2, 2020 4 © 2020. The Author(s). Published under a
Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 5 of 15

A comparison of the O3 measured at the matched residential and A comparison of the overlapping months shows a reduction of
university locations (Gacuriro and CMU-Africa; Belle Vue Estate almost 40% from 2017 to 2018 for March, April, and May, and
and UR-CST) shows that most paired values are in reasonable 51% for July, while the June monthly average is practically
agreement around the 1:1 line (Figure S4). An ODR fit of the unchanged. However, June and July 2018 had the least data
1-hour O3 values measured at Gacuriro and CMU-Africa yields a coverage (as noted in the preceding paragraph), and fewer than
slope of 0.97±0.01 with negligible intercept. A comparison of the 300 hours of data are available for June 2017. The March-May
UR-CST and Belle Vue estate O3 measurements is complicated by comparisons are more robust with 1,683 and 1,942 hours of data
the significantly higher O3 apparently measured at the UR-CST available in 2017 and 2018 respectively. April 2017 had a larger
site during the mid-July to mid-September period. The reason fraction of non-working day data (43%) compared to the other
for these differences is unclear, as temperature and relative five months (30-34%), but similar reductions are seen when
humidity (two variables likely to affect sensor performance) the comparison is restricted to working days. An examination
measured at the two urban background locations are identical. of spatially-resolved reanalysis data (Siebert et al., 2019) for
This issue of potentially high O3 at UR-CST needs to be rainfall (Rwanda Meteorology Agency, 2019) in Kigali (Figure S6)
investigated further; here, we take a conservative approach shows that the UR-CST site during MAM 2018 experienced 38
and present the data set as a unified O3 time-series (as we do rainy days (≥5 mm/day), compared to 16 such rainy days in MAM
for PM2.5), except the UR-CST measurements are excluded from 2017 at Gacuriro. Total rainfall was also substantially higher in
the current analysis. This restricts our analysis to the March- 2018; 612 mm at UR-CST in MAM 2018 compared to just 180 mm
November 2017 period (as there is only a week of data for at Gacuriro in MAM 2017. The more frequent and heavier rainfall
December 2017 from the Belle Vue estate site), which are the in MAM 2018 could at least partly explain the significantly lower
two 2017 wet seasons (March-April-May or MAM and ON) and the PM2.5 in MAM 2018 compared to MAM 2017.
long dry season (June-July-August-September, JJAS).

Unlike O3 and PM2.5, differences in CO concentrations were

sometimes observed at the residential locations and the campus
sites in Kigali (Figure S5 of the SI). The highest concentrations
were observed usually at the campus sites, which might reflect
the greater traffic seen by these sites compared to residential
neighborhoods. However, the Gacuriro neighborhood did
experience two hours when CO exceeded 3 ppm. Overall, the
study average CO concentrations were similar at UR-CST and Belle
Vue Estate at 0.446±0.322 ppm and 0.447±0.283 ppm (average
and standard deviation) respectively. CO was 0.504±0.499 ppm
at CMU-Africa and 0.404±0.320 ppm at Gacuriro. Concentrations Figure 2: Monthly average PM2.5 measured in Kigali over the course of this
at rural Mugogo were even lower, at 0.225±0.097 ppm. However, study. Red bars indicate 2017 and blue bars indicate 2018 data. Error bars
are one standard deviation of the hourly measurements in that month.
the measured CO concentrations at all locations were low and Solid round markers show the number of hourly data points collected in
far below US EPA standards (35 ppm for 1 hour). The WHO has that month. Horizontal dashed lines indicate WHO annual guideline and
not set guideline values for CO. The higher standard deviations the WHO’s first interim target (IT 1).
at the urban sites compared to the rural site suggests that the
urban areas are more likely to see large spikes of CO. Working day/non-Working day and
seasonal differences in PM2.5
Average monthly patterns of PM2.5 and Figure 3 shows the distribution of hourly PM2.5 values measured
changes from 2017 to 2018 in Kigali across each of the four seasons: long wet season (MAM
The unified Kigali time series for ambient PM2.5 is summarized 2017 and MAM 2018), long dry season (JJAS 2017, henceforth
as monthly average PM2.5 values, covering the period between JJAS), short wet season (ON 2017, henceforth ON), and short dry
March 2017 and July 2018 (Figure 2). The least data were collected season (December 2017 and January-February 2018, henceforth
DJF.) Note that August and September are both considered “dry”
in June-July 2018 (100 and 64 hours) due to maintenance
here. As Figure 2 showed, the monthly average PM2.5 value for
issues. For all other months, between 298-744 hours of data are
September is closer to that of August and June, and noticeably
available, averaging 634 hourly values per month (88% of a 30-
higher than the average PM2.5 for October and November.
day period). The monthly average PM2.5 values range from 25 µg/
Additionally, September 2017 saw just 17 mm of rainfall,
m3 in May 2018 to 102 µg/m3 in July 2017, all higher than the
compared to 104 mm of rainfall in October-November 2017
WHO annual guideline of 10 µg/m3; most months also exceed
(Figure S6). This suggests that our classification is appropriate
the WHO’s first interim target (WHO IT 1) of 35 µg/m3. The study
for 2017.
average hourly PM2.5 in Kigali is 52.4±33.7 µg/m3; the large
standard deviation suggests significant temporal variability that
Figure 3 shows that ambient PM2.5 levels are higher in the dry
will be explored further in later sections. seasons than in the wet seasons. The hourly data collected with
the RAMPs on non-working days (nWD) in MAM 2018 and ON

CLEAN AIR JOURNAL © 2020. The Author(s). Published under a

Volume 30, No 2, 2020 5 Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 6 of 15

shows a pear-shaped pattern, in contrast to the hourglass shape

of MAM 2017 nWD data; this is also seen in the working day (WD)
data. This suggests that there are fewer high-concentration
periods in the later wet season months. WD PM2.5 values are
higher than the nWD concentrations for all seasons by 19-
27%. However, as discussed earlier, WD PM2.5 values are scaled
upwards by 1.69 based on a comparison with the Kalisa et al.
(2018) wet season filter measurements while nWD PM2.5 values
are similarly upscaled by 1.39 - a difference of 22%. Thus, this
weekday-weekend effect could reflect Kalisa et al.’s results for
April 2017 and needs to be investigated further for other months
of the year.

Table 3 summarizes the seasonal averages for working days.

The dry season average PM2.5 concentrations are about two
times the following wet season average; to be exact, JJAS/ON is
1.89 (95% confidence intervals 1.82-1.97) and DJF/MAM 2018 is
2.11 (95% CI 2.03-2.19). This was also observed for non-working
days. Higher pollution in the dry season was also observed by
Kalisa et al. (2018) for PM2.5 in Kigali (though in campaigns of 2-4
weeks in each season) and by DeWitt et al. (2019) for BC at Mt
Mugogo, who attribute the higher dry season concentrations to
regional biomass burning. Kalisa et al. measured a median daily
PM2.5 of 126 µg/m3 (IQR 113-141 µg/m3) over June 15-30, 2017.
The median for our JJAS hourly measurements (which does
not include data for the June 15-28 period) was 66.2 µg/m3 (IQR
45.4-95.4 µg/m3). It seems that the June 15-28 period saw higher
pollution based on Kalisa et al.’s results, but other parts of the Figure 3: Distribution of 1-hour average PM2.5 in each of the seasons over
the course of this study, grouped into (A) working day and (B) non-working
same dry season were relatively cleaner based on our results, day measurements. March-May (MAM) and October-November (ON) are
showing the importance of long-term monitoring. the wet seasons; June-September (JJAS) and December-February (DJF)
are the dry seasons. The boxes show the interquartile range (IQR), i.e. the
BC and AAE in Kigali 25th and 75th percentiles of data; the notch in each box shows the median
value; and the whiskers mark the 2nd and 98th percentile of all data. The
A total of 6,850 hours of BC data were collected at UR-CST violin plot shows the relative distribution of all data. Y-axis truncated at
between July 6, 2017-April 24, 2018. The equivalent BC mass 200 µg/m3 for visual clarity.
concentration (as measured at 880 nm) hourly averages ranged
from 0.14 µg/m3 to 49.6 µg/m3, with a study average of 4.04±2.86 BC working day seasonal averages are summarized in Table 3.
µg/m3. This value is significantly higher than the values Non-working day seasonal averages were 0.2-0.9 µg/m3 lower.
observed in urban areas of developed countries; e.g. in the early The difference between the wet and dry seasons (dry season
2000s, urban ambient BC was 2 and 1 µg/m3 respectively in the BC 40-60% higher) in Kigali is lower than the factor-of-four
states of New Jersey and California in the USA after decades of difference observed in similarly-tropical Pune (Kolhe et al.,
reductions (Kirchstetter et al., 2017). The Kigali BC values are 2018) or even at the rural Mt Mugogo site in Rwanda, where
comparable to mid-sized urban areas in China and India. In Hefei DeWitt et al. (2019) found dry and wet season BC different by
(central China), annual average BC in 2012-2013 was 3.5 µg/m3 almost a factor-of-three.
(Zhang et al., 2015). In Pune (western India), the average BC over
2015-2016 was 3.9 µg/m3 (Kolhe et al., 2018). The study average AAE was 1.53, with most values between 0.9-
2.1 (Figure S7). As shown in Table 3, AAE was slightly lower in

Table 3: Seasonal PM2.5 and BC statistics for this study in Kigali, Rwanda
Average diurnal
WD PM2.5 average Regional WD WD BC average
Season minimum WD AAE
± SD (µg/m3) (A) PM2.5 (%) C = B/A ± SD (µg/m3)
PM2.5 (µg/m3) (B)
MAM 2017 49.9 ± 31.7 21.3 42.7 N/A N/A
JJAS 81.2 ± 42.3 57.3 70.6 5.20±3.56 1.47±0.15
ON 42.9 ± 21.0 22.8 53.2 3.20±2.24 1.57±0.16
DJF 66.0 ± 31.6 50.7 76.8 4.52±2.96 1.52±0.16
MAM 2018 31.3 ± 17.7 22.8 73.0 3.30±2.40 1.59±0.19

CLEAN AIR JOURNAL

Volume 30, No 2, 2020 6 © 2020. The Author(s). Published under a
Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 7 of 15

the dry seasons than in the wet seasons. Monthly average AAE influenced by domestic biofuel use. However, compared to
values ranged from 1.39-1.59, lower than the 1.5-1.9 range seen previous studies showing AAE values of ~1.1±0.3 for fossil fuel
at rural Mt Mugogo (DeWitt et al., 2019); this could indicate a BC-dominated pollution, the AAE values observed in Kigali are
greater contribution from vehicular sources to ambient BC in often higher, suggesting that there is always some biomass
urban Kigali. Higher AAE (>1.8) values were almost always only burning influence (associated with BrC).
seen at BC concentrations below 5 µg/m3 (Figure S7), while
higher BC concentrations showed AAE values around 1.4 or Local and regional contributions to PM2.5
less, which suggests that fossil fuel combustion is associated
with higher BC levels, but the background BC is dominated by
and BC in Kigali
We use OBM (Diamantopoulou et al., 2016; Pikridas et al., 2013) to
biomass burning.
estimate the regional and local contributions to ambient PM2.5in
Kigali. During the dry season, a higher regional background
Diurnal profiles of PM2.5 and BC is likely, as transported pollution may not be rained out. The
Figure 4 shows the average diurnal pattern of PM2.5 and BC for
background can include regional biomass burning, dust, and
the five seasons (four for BC) during the study period for which
other upwind emissions as well as secondary PM2.5. The OBM
we have a large number of measurements. Only working day
assumes that the minimum value of a seasonal diurnal profile
data are shown for visual clarity; a similar pattern is observed
based on hourly measurements is the seasonal average regional
for the non-working day measurements. As noted earlier, PM2.5
contribution. Then, the seasonal average local contribution is
concentrations are substantially higher in the dry seasons than
the difference between the seasonal average ambient PM2.5 and
in the wet seasons. A morning peak (likely related to traffic) is
the seasonal average regional contribution. These estimates of
observed between 8 AM-10 AM local time for all seasons except
regional contribution assume the ambient concentrations at the
MAM 2017. Concentrations then fall during the day (as the
minima are entirely regional, but there may be local vehicular
boundary layer height increases) before rising back up in the
and domestic biofuel emissions all day. Thus, the results are the
evening, likely a combination of evening emissions and lower
upper bound of the regional contribution and the lower bound
boundary layer heights. The MAM 2017 night-time highs (and
for the local contribution.
to some extent the JJAS night-time highs) remain at that level
until the morning traffic peak. For the other three seasons, the
morning traffic peak is higher than the night-time high values.
While the BC concentrations follow a similar diurnal profile, in
all seasons the maxima occur during the morning rush hours,
with concentrations lowest in the afternoon. In further contrast
to the PM2.5 dry/wet seasonal differences, the wet and dry
season BC concentrations are much closer to each other, though
dry season BC is still higher.

The highest PM2.5 concentrations are observed at night-time

during MAM 2017 and JJAS, with average PM2.5 around 80 µg/
m3 between 8 PM-12 AM in the wet season (MAM 2017) and
around 100 µg/m3 between 7 PM-11 PM in the dry season (JJAS).
In contrast, the highest average PM2.5 levels in other seasons
were observed during the morning rush hour - around 60 µg/m3
between 7 AM-10 AM in ON, around 90 µg/m3 between 8 AM-10
AM in DJF, and around 45 µg/m3 between 8 AM-10 AM in MAM
2018. The BC maxima all occur during the morning rush hour
between 7 AM-8 AM except for DJF, when levels are marginally
higher between 8 AM-9 AM than during the preceding hour.

The morning and evening peaks at similar times for PM2.5 and
BC indicate that these PM peaks are related to combustion
emissions. Figure 5 shows the diurnal variation in AAE during
the dry and wet seasons. In both cases, AAE is lower between 6
AM to 9 AM, when BC mass concentrations are higher; the lowest
median AAE is 1.4 between 7 AM-8 AM during the dry season.
However, the night-time peaks in PM2.5 and BC are associated Figure 4: Average diurnal patterns of (A) PM2.5 and (B) BC for each season
with higher AAE values, when hourly medians approach 1.6- over the course of this study. Data restricted to working days, with similar
1.7. These differences suggest that while the morning BC and patterns observed for non-working days.
PM2.5peaks are mostly associated with fossil-fuel vehicular
emissions, the night-time BC and PM2.5 peaks are additionally

CLEAN AIR JOURNAL © 2020. The Author(s). Published under a

Volume 30, No 2, 2020 7 Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 8 of 15

Figure 5: Diurnal variations in Angstrom Absorption Exponent (AAE) during

the dry and wet seasons suggest vehicular emissions and domestic biofuel
use contribute to ambient air pollution at different times of the day. The
boxes show the IQR (25th and 75th percentiles of data); the notch shows the
median; whiskers indicate the 9th and 91st percentiles. The corresponding
hourly boxes for the two types of seasons are offset for visual clarity.

Table 3 summarizes the seasonal statistics required to calculate

the average regional contribution to Kigali ambient PM2.5 for
each season. The minimum average PM2.5 values were observed
between 3-4 PM local time in all seasons. About half the ambient
PM2.5 in Kigali appears to be from regional sources in the wet
season, and the regional share rises to over 70% in the dry
seasons. Put another way, local sources contribute at least half
the observed PM2.5 during the 2017 wet seasons, and one-fourth
of ambient PM2.5 during the two dry seasons. The MAM 2018 wet
season appears to be an anomaly compared to the two other
wet seasons, but in terms of absolute PM2.5 concentrations, the
regional component is about the same in MAM 2018 as during
the 2017 wet seasons.

Following a similar procedure for BC, the regional contribution

on working days is 54-55% in the JJAS and DJF dry seasons, but
only 33% and 40% in the wet MAM 2018 and ON seasons. The
regional contribution to Kigali BC on non-working days is 7-8%
higher during JJAS, ON, and MAM 2018, which likely reflects
lower local emissions (including car-free Sundays). Overall, local
sources are larger contributors to ambient BC than to PM2.5 in
Kigali, ranging from about half in the dry season to two-thirds in
the wet seasons.

The impact of Sunday car-free hours on

urban background air pollution
An estimate of the impact of traffic-related air pollution (TRAP)
at non-roadside, urban background locations can be obtained
by comparing the RAMP and BC measurements on “car-free”
Sundays with the measurements on all other Sundays. In this
study period, PM2.5 (BC) measurements were made on 16 (11)
“car-free” Sundays that can be compared to data from 42 (30) Figure 6: Impact of limiting vehicular traffic on Sunday mornings in Kigali,
Rwanda. Panel (A) shows the average hourly PM2.5 measured over 16 car-
regular Sundays. However, in Rwanda, the car-free policy is free Sundays and 42 regular Sundays between April 2017-May 2018. Panel
only in effect on Sunday mornings. The time-resolved data from (B) shows the same PM2.5 data as a scatter plot. Panel (C) shows a scatter
the RAMPs and BC monitor allows investigation of the benefits plot of the corresponding BC concentrations. Error bars are the standard
error.
of this policy for the specific periods when such policies are in
effect.

CLEAN AIR JOURNAL

Volume 30, No 2, 2020 8 © 2020. The Author(s). Published under a
Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 9 of 15

Figure 6 shows a comparison of the hourly average PM2.5 8-hour average occurs on August 30, 2017 between 10 AM-6 PM.
between regular Sundays and car-free Sundays. The average All 8-hour averages over 50 ppb occur during the daytime, for
hourly PM2.5 concentration is lower in the morning period by an 8-hour periods beginning usually at 10 AM or 11 AM.
average of 7 µg/m3. In the afternoon and evening periods, the
diurnal patterns converge, indicating no significant difference in
pollution. The reduction in ambient PM2.5 is highest between 7
AM-9 AM, when PM2.5 on regular Sundays is highest; during these
Conclusions and recommendations
hours the average PM2.5 is reduced by 10-12 ug/m3 on car-free for future work
Sundays or 20% of PM2.5 at the same time on regular Sundays. We have presented the results of a long-term ground-based
BC concentrations are lower by about 1 µg/m3 between 7 AM-10 monitoring campaign, the first of its kind in Kigali, Rwanda.
AM on car-free Sundays, which is 18%-28% of the corresponding Lower-cost and relatively low maintenance RAMP monitors
concentrations on regular Sundays. However, there are also were used for this study, with local verification and correction
similar differences for pre-7 AM “business-as-usual” times on of sensor calibrations by collocation with reference monitoring
these Sundays, and so the observed reductions during the during an overlapping campaign and at RCO.
morning car-free periods may not be directly attributable to the
car-free policy. O3 pollution in Kigali was usually below WHO guidelines, but the
50-ppb threshold could be exceeded in the dry season. Periods
Ozone in Kigali and Mugogo of high ozone can be identified using low-cost sensors and a
Only 942 hours of RAMP data are available from rural Mt Mugogo general calibration, though local calibrations improve sensor
for JJAS (dry season) and 881 hours for ON (wet season), performance. The RAMP PM2.5 (before filter-based correction)
compared to the 2,400 and 1,464 hours of O3 measurements in correlates strongly with filter-based PM2.5 on non-working days
Kigali for the same seasons. The maximum 8-hour average O3 when TRAP and other working-day contributions are lower, but
value at rural Mugogo as measured by the RAMP using the gRAMP the RAMP PM2.5 values were still a significant underestimate.
calibration models were 34.4 ppb in dry JJAS and 33.0 ppb in These differences suggest that the size distribution of PM2.5 in
wet ON – similar, unlike the significantly higher O3 observed in Kigali is quite different from that in Pittsburgh, with substantial
urban Kigali during the dry season (JJAS) compared to the wet contributions from sub-300 nm particles (where low-cost optical
season (MAM and ON). However, as discussed previously, the sensors are less sensitive). Future studies with lower-cost
gRAMP model (hybrid-RF) is positively biased at Mt Mugogo monitors should include collocated filter-based measurements
during ON. The RCO reference monitor at this site shows a or short-term intensive studies with aerosol sizing instruments
seasonal difference, with maximum 8-hour O3 concentrations of (e.g. a scanning mobility particle sizer, SMPS) to account for
40.6 ppb in JJAS and 30.6 ppb in ON (restricting the comparison such differences.
to common periods with the RAMP data.) This is consistent with
previous findings at this site (DeWitt et al., 2019). The RAMP- The Health Effects Institute (2019) (HEI) using data from
specific Mugogo QR model, which is not seasonally biased, the Global Burden of Disease Study 2017 estimates that the
shows maximum 8-hour O3 concentrations of 33.9 ppb in JJAS population-weighted annual average PM2.5 in Rwanda was 43
and 26.5 ppb in ON – below the reference values but capturing µg/m3 in 2017 (and the same for 2015-2016). Our study found
the seasonality. A closer examination of the data (for common that the study average PM2.5 in Kigali is 52±34 µg/m3 at residential
periods) indicates that both the reference monitor and the or university locations, which can be considered as “urban
gRAMP model show the maximum 8-hour O3 on the same day background” sites. Kalisa et al. (2018) show that pollution levels
of the dry season - September 2, 2017. The RAMP O3 maximum can be significantly higher at the roadside in Kigali than at the
(34.4 ppb) occurs between 8 AM-4 PM, during which time the urban background locations where our measurements were
reference monitor average was 40.3 ppb. made. Air pollution in the low-lying valleys of Kigali can also
be higher (Henninger, 2013). RAMP measurements at Musanze
The average 1-hour O3 concentrations in Kigali during March- in late 2017 showed PM2.5 concentrations at this rural site were
November 2017 were 16 ppb in the wet season (MAM 2017 and moderately correlated (r2 = 0.54) with Kigali PM2.5 and about
ON) compared to 22 ppb in the dry season (JJAS). The maximum 20% lower (not shown). These results suggest that the HEI
O3 values, 66 ppb (1-hour average) and 57 ppb (8-hour average), population-weighted average for Rwanda and other similar
were observed in the dry season. Previous measurements satellite-based estimates could be underestimates that need to
over Kigali during aircraft takeoff and landings in the MOZAIC be updated with ground-based monitoring, like the campaign
campaign (Sauvage et al., 2005) over 1997-2003 also showed presented here.
significantly higher O3 concentrations in the dry season in the
lower troposphere. Three recent studies – this paper, DeWitt et al. (2019), and Kalisa
et al. (2018) – have found that air pollution is significant in urban
The current US EPA standard for 8-hour average O3 is 65 ppb and rural Rwanda, with considerable spatial variability due to
and the WHO guideline value for 8-hour average O3 is 50 ppb. O3 local conditions and sources. Hence, future studies with low-
in Kigali is higher than the WHO guideline on 10 days over our cost monitors should include a variety of locations, such as
sampling period, of which six are in July. However, the highest low-lying areas and roadside locations. Our results should be

CLEAN AIR JOURNAL © 2020. The Author(s). Published under a

Volume 30, No 2, 2020 9 Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 10 of 15

considered a lower estimate of the PM2.5 pollution that Kigali

residents are exposed to – which is concerning since the average
Acknowledgments
This research was supported by the College of Engineering,
monthly concentrations in our 16+ month study were above the
the Department of Engineering and Public Policy, and the
WHO annual guideline and often over WHO’s first interim target.
Department of Mechanical Engineering at Carnegie Mellon
University via discretionary funding support for Paulina
Long-term monitoring over five seasons shows that differences
Jaramillo and Allen Robinson. This work also benefited from
in PM2.5 between working days and non-working days are
State assistance managed by the National Research Agency
smaller than the differences between the dry and wet seasons.
under the “Programme d’Investissements d’Avenir” under the
The average dry season PM2.5 levels are about two times the
reference “ANR-18-MPGA-0011” (“Make our planet great again”
succeeding wet season PM2.5 while the comparable seasonal
initiative). The results and conclusions of this paper are the sole
differences in BC are about 40-60%, indicating that transported
responsibility of the authors and do not represent the views of
non-BC (e.g. dust) or low-BC (e.g. forest fires) pollution is
the funding sources.
important in the dry season. Particulate pollution in Kigali has
a distinct diurnal profile, as the morning rush-hour usually
results in the maximum BC and PM2.5 due to traffic pollution,
and minima occur in the afternoons. The higher AAE values
Author contributions
Conceptualisation: RS, ALR, PJ; methodology: RS, JG, HLD;
associated with the overnight BC and PM2.5 peaks suggest that
data collection: ASK, VB, SG, CS, NJW, EK; data analysis and
domestic biofuel use could also be a significant contributor to
validation: RS, CM, ASK, SG, HL, EK; writing/initial draft: RS;
overnight air pollution.
writing/revisions: all authors; student supervision: JG, RS, PJ, PA;
project leadership and management: RS; funding acquisition:
Overall, local sources could contribute half the PM2.5 and two-
ALR, PJ, RS
thirds of the BC in the wet seasons in Kigali, which means
that local pollution control policies can significantly improve
Kigali’s air quality. In MAM 2018, heavier local rainfall may have
helped reduce ambient PM2.5 levels, though the January-July
Data availability
The RAMP data (gRAMP-based gas concentrations and “BAM-
2018 average PM2.5 was 43.7 µg/m3, still above the WHO’s first
equivalent” PM2.5 mass concentrations, without filter-based
interim target for annual average PM2.5. New pollution control
correction) and BC data are available at https://doi.org/10.6084/
policies are being implemented in Rwanda, including doubling
m9.figshare.8074436.v1
the number of car-free Sundays and increased duties on older
imported used cars. Continued monitoring is essential to
evaluate the impact of these policies.
References
Bond, T.C., Habib, G., Bergstrom, R.W., 2006. Limitations in the
Air pollution in Kigali is also influenced by substantial regional
enhancement of visible light absorption due to mixing state.
contributions especially in the dry season; for example, the dry
Journal of Geophysical Research: Atmospheres 111. https://doi.
season daily minimum PM2.5 was over 50 µg/m3, or over 70%
org/10.1029/2006JD007315
of average PM2.5. About half the BC during the dry season also
appears to be regional, which suggests a significant biomass
Brauer, M., Amann, M., Burnett, R.T., Cohen, A., Dentener, F.,
burning influence in the regional/background contribution in
Ezzati, M., Henderson, S.B., Krzyzanowski, M., Martin, R.V., Van
line with DeWitt et al. (2019). Hence, controls on regional biomass
Dingenen, R., van Donkelaar, A., Thurston, G.D., 2012. Exposure
burning are essential to reducing PM2.5 concentrations in Kigali.
Assessment for Estimation of the Global Burden of Disease
Studies with more advanced instrumentation (such as an AMS)
Attributable to Outdoor Air Pollution. Environ. Sci. Technol. 46,
would help better quantify the time-resolved contribution of
652–660. https://doi.org/10.1021/es2025752
different local and regional sources to air pollution in Rwanda,
which in turn would help identify additional air quality
DeWitt, H.L., Gasore, J., Rupakheti, M., Potter, K.E., Prinn, R.G.,
interventions.
Ndikubwimana, J. de D., Nkusi, J., Safari, B., 2019. Seasonal
and diurnal variability in O3, black carbon, and CO measured at
This study was carried out in collaboration with the University
the Rwanda Climate Observatory. Atmospheric Chemistry and
of Rwanda and RCO (co-authors on this paper) and with the
Physics 19, 2063–2078. https://doi.org/10.5194/acp-19-2063-
support of local residents who hosted the RAMPs. Towards the
2019
end of this study, the Rwanda Environmental Management
Authority (REMA) acquired a reference monitoring station as
Diamantopoulou, M., Skyllakou, K., Pandis, S.N., 2016.
well as eight RAMPs to establish their own air quality monitoring
Estimation of the local and long-range contributions to
network. Low-cost monitors can significantly reduce the costs
particulate matter levels using continuous measurements in a
of air quality monitoring for developing countries, but local
single urban background site. Atmospheric Environment 134,
support and buy-in are keys to success.
1–9. https://doi.org/10.1016/j.atmosenv.2016.03.015

CLEAN AIR JOURNAL

Volume 30, No 2, 2020 10 © 2020. The Author(s). Published under a
Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 11 of 15

Dockery, D.W., Pope, C.A., Xu, X., Spengler, J.D., Ware, J.H., Fay, Kolhe, A.R., Aher, G.R., Ralegankar, S.D., Safai, P.D., 2018.
M.E., Ferris, B.G., Speizer, F.E., 1993. An association between air Investigation of aerosol black carbon over semi-urban
pollution and mortality in six U.S. cities. N. Engl. J. Med. 329, and urban locations in south-western India. Atmospheric
1753–1759. https://doi.org/10.1056/NEJM199312093292401 Pollution Research 9, 1111–1130. https://doi.org/10.1016/j.
apr.2018.04.010
Duhuze, R.N., 2018. Fuel and Vehicle Standards: Vehicle
Inspection, compliance and enforcement - case of Rwanda. Lack, D.A., Langridge, J.M., 2013. On the attribution of black and
Presented at the UNEP “Africa Clean Mobility Week” conference, brown carbon light absorption using the Ångström exponent.
Nairobi, Kenya. Atmospheric Chemistry and Physics 13, 10535–10543. https://
doi.org/10.5194/acp-13-10535-2013
Farquharson, D., Jaramillo, P., Samaras, C., 2018. Sustainability
implications of electricity outages in sub-Saharan Africa. Nature Laden, F., Schwartz, J., Speizer, F.E., Dockery, D.W., 2006.
Sustainability 1, 589. https://doi.org/10.1038/s41893-018-0151- Reduction in fine particulate air pollution and mortality:
8 Extended follow-up of the Harvard Six Cities study. Am. J.
Respir. Crit. Care Med. 173, 667–672. https://doi.org/10.1164/
Health Effects Institute, 2019. State of Global Air 2019 [WWW rccm.200503-443OC
Document]. URL https://www.stateofglobalair.org/data/#/air/
plot (accessed 5.1.19). Malings, C., Tanzer, R., Hauryliuk, A., Kumar, S.P.N., Zimmerman,
N., Kara, L.B., Presto, A.A., R. Subramanian, 2019. Development
Henninger, S.M., 2013. When Air Quality Becomes Deleterious—A of a general calibration model and long-term performance
Case Study for Kigali, Rwanda. Journal of Environmental evaluation of low-cost sensors for air pollutant gas monitoring.
Protection 04, 1. https://doi.org/10.4236/jep.2013.48A1001 Atmospheric Measurement Techniques 12, 903–920. https://doi.
org/10.5194/amt-12-903-2019
Jerrett, M., Burnett, R.T., Pope, C.A., Ito, K., Thurston, G., Krewski,
D., Shi, Y., Calle, E., Thun, M., 2009. Long-Term Ozone Exposure Malings, C., Tanzer, R., Hauryliuk, A., Saha, P.K., Robinson,
and Mortality. New England Journal of Medicine 360, 1085– A.L., Presto, A.A., Subramanian, R., 2020. Fine particle mass
1095. https://doi.org/10.1056/NEJMoa0803894 monitoring with low-cost sensors: Corrections and long-term
performance evaluation, Aerosol Science and Technology, 54:2,
Kalisa, E., Archer, S., Nagato, E., Bizuru, E., Lee, K., Tang, N., 160-174, https://doi.org/10.1080/02786826.2019.1623863
Pointing, S., Hayakawa, K., Lacap-Bugler, D., 2019. Chemical
and Biological Components of Urban Aerosols in Africa: Masiol, M., Agostinelli, C., Formenton, G., Tarabotti, E., Pavoni,
Current Status and Knowledge Gaps. International Journal of B., 2014. Thirteen years of air pollution hourly monitoring in a
Environmental Research and Public Health 16, 941. https://doi. large city: Potential sources, trends, cycles and effects of car-free
org/10.3390/ijerph16060941 days. Science of The Total Environment 494–495, 84–96. https://
doi.org/10.1016/j.scitotenv.2014.06.122
Kalisa, E., Nagato, E.G., Bizuru, E., Lee, K.C., Tang, N., Pointing,
S.B., Hayakawa, K., Archer, S.D.J., Lacap-Bugler, D.C., 2018. MININFRA, 2018. Energy Sector Strategic Plan, 2018/19-2023/24.
Characterization and Risk Assessment of Atmospheric PM2.5 and Moosmüller, H., Chakrabarty, R.K., Ehlers, K.M., Arnott, W.P.,
PM10 Particulate-Bound PAHs and NPAHs in Rwanda, Central- 2011. Absorption Ångström coefficient, brown carbon, and
East Africa. Environ. Sci. Technol. 52, 12179–12187. https://doi. aerosols: basic concepts, bulk matter, and spherical particles.
org/10.1021/acs.est.8b03219 Atmospheric Chemistry and Physics 11, 1217–1225. https://doi.
org/10.5194/acp-11-1217-2011
Kirchstetter, T.W., Novakov, T., 2007. Controlled generation of
black carbon particles from a diffusion flame and applications in Mudway, I.S., Dundas, I., Wood, H.E., Marlin, N., Jamaludin,
evaluating black carbon measurement methods. Atmospheric J.B., Bremner, S.A., Cross, L., Grieve, A., Nanzer, A., Barratt,
Environment 41, 1874–1888. https://doi.org/10.1016/j. B.M., Beevers, S., Dajnak, D., Fuller, G.W., Font, A., Colligan, G.,
atmosenv.2006.10.067 Sheikh, A., Walton, R., Grigg, J., Kelly, F.J., Lee, T.H., Griffiths,
C.J., 2019. Impact of London’s low emission zone on air quality
Kirchstetter, T.W., Novakov, T., Hobbs, P.V., 2004. Evidence that and children’s respiratory health: a sequential annual cross-
the spectral dependence of light absorption by aerosols is sectional study. The Lancet Public Health 4, e28–e40. https://
affected by organic carbon. Journal of Geophysical Research: doi.org/10.1016/S2468-2667(18)30202-0
Atmospheres 109. https://doi.org/10.1029/2004JD004999
Ngabonziza, D., 2017. Car Imports to Rwanda Drop by 20% As
Kirchstetter, T.W., Preble, C.V., Hadley, O.L., Bond, T.C., Apte, J.S., High Taxes Bite – KT PRESS [WWW Document]. URL https://
2017. Large reductions in urban black carbon concentrations in the ktpress.rw/2017/05/car-imports-to-rwanda-drop-by-20-as-
United States between 1965 and 2000. Atmospheric Environment high-taxes-bite/ (accessed 5.1.19).
151, 17–23. https://doi.org/10.1016/j.atmosenv.2016.11.001

CLEAN AIR JOURNAL © 2020. The Author(s). Published under a

Volume 30, No 2, 2020 11 Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 12 of 15

Petkova, E.P., Jack, D.W., Volavka-Close, N.H., Kinney, P.L., 2013. Wang, X., Westerdahl, D., Chen, L.C., Wu, Y., Hao, J., Pan, X., Guo,
Particulate matter pollution in African cities. Air Qual Atmos X., Zhang, K.M., 2009. Evaluating the air quality impacts of the
Health 6, 603–614. https://doi.org/10.1007/s11869-013-0199-6 2008 Beijing Olympic Games: On-road emission factors and
Pikridas, M., Tasoglou, A., Florou, K., Pandis, S.N., 2013. black carbon profiles. Atmospheric Environment 43, 4535–4543.
Characterization of the origin of fine particulate matter in a https://doi.org/10.1016/j.atmosenv.2009.06.054
medium size urban area in the Mediterranean. Atmospheric
Environment 80, 264–274. https://doi.org/10.1016/j. WHO, 2018. Global Health Observatory: Deaths - by country
atmosenv.2013.07.070 [WWW Document]. WHO. URL http://apps.who.int/gho/data/
node.main.BODAMBIENTAIRDTHS?lang=en (accessed 5.1.19).
RRA, 2017. EAC Proposed Depreciation Schedule [WWW
Document]. URL https://www.rra.gov.rw/index.php?id=219 Zhang, Q., Jimenez, J.L., Canagaratna, M.R., Ulbrich, I.M., Ng,
(accessed 5.1.19). N.L., Worsnop, D.R., Sun, Y., 2011. Understanding atmospheric
organic aerosols via factor analysis of aerosol mass
Rwanda Meteorology Agency, 2019. Reanalysis data [WWW spectrometry: a review. Anal Bioanal Chem 401, 3045–3067.
Document]. URL https://meteorwanda.gov.rw/index.php?id=2 https://doi.org/10.1007/s00216-011-5355-y
(accessed 5.1.19).
Zhang, X., Rao, R., Huang, Y., Mao, M., Berg, M.J., Sun, W.,
Sauvage, B., Thouret, V., Cammas, J.-P., Gheusi, F., Athier, G., 2015. Black carbon aerosols in urban central China. Journal
Nédélec, P., 2005. Tropospheric ozone over Equatorial Africa: of Quantitative Spectroscopy and Radiative Transfer, Topical
regional aspects from the MOZAIC data. Atmospheric Chemistry issue on optical particle characterization and remote sensing
and Physics 5, 311–335. https://doi.org/10.5194/acp-5-311-2005 of the atmosphere: Part I 150, 3–11. https://doi.org/10.1016/j.
jqsrt.2014.03.006
Shrivastava, M.K., Subramanian, R., Rogge, W.F., Robinson, A.L.,
2007. Sources of organic aerosol: Positive matrix factorization of Zimmerman, N., Presto, A.A., Kumar, S.P.N., Gu, J., Hauryliuk, A.,
molecular marker data and comparison of results from different Robinson, E.S., Robinson, A.L., R. Subramanian, 2018. A machine
source apportionment models. Atmospheric Environment 41, learning calibration model using random forests to improve
9353–9369. https://doi.org/10.1016/j.atmosenv.2007.09.016 sensor performance for lower-cost air quality monitoring.
Atmospheric Measurement Techniques 11, 291–313. https://doi.
Siebert, A., Dinku, T., Vuguziga, F., Twahirwa, A., Kagabo, D.M., org/10.5194/amt-11-291-2018
delCorral, J., Robertson, A.W., 2019. Evaluation of ENACTS-
Rwanda: A new multi-decade, high-resolution rainfall and
temperature data set—Climatology. International Journal of Appendix
Climatology 0. https://doi.org/10.1002/joc.6010
Further investigation of the performance
Subramanian, R., Donahue, N.M., Bernardo-Bricker, A., Rogge, of Pittsburgh gRAMP calibration models
W.F., Robinson, A.L., 2007. Insights into the primary–secondary
and regional–local contributions to organic aerosol and PM2.5
and comparison with local “Mugogo”
mass in Pittsburgh, Pennsylvania. Atmospheric Environment 41, models
7414–7433. https://doi.org/10.1016/j.atmosenv.2007.05.058 We further explore the performance of the Pittsburgh hybrid
RF gRAMP model by comparing it with the Mugogo QR model,
Subramanian, R., Ellis, A., Torres-Delgado, E., Tanzer, R., Malings, which had the best overall performance for O3 among the
C., Rivera, F., Morales, M., Baumgardner, D., Presto, A., Mayol- Mugogo models. Figure S2 shows scatter plots of the RAMP-
Bracero, O.L., 2018. Air Quality in Puerto Rico in the Aftermath reported concentrations (as 8-hour running averages) for the
of Hurricane Maria: A Case Study on the Use of Lower Cost Air two calibration approaches against the reference monitor,
Quality Monitors. ACS Earth Space Chem. 2, 1179–1186. https:// separated into the dry (June-July-August-September, JJAS) and
doi.org/10.1021/acsearthspacechem.8b00079 wet (October-November, ON) seasons. However, the gRAMP
model is positively biased in ON, with most of the data above
Sun, H., Biedermann, L., Bond, T.C., 2007. Color of brown carbon: the 1:1 line and clustered around the +30% line. In other words,
A model for ultraviolet and visible light absorption by organic the gRAMP model overpredicts O3 during the wet season. The
carbon aerosol. Geophysical Research Letters 34. https://doi. Mugogo model (panels B and D) performs better (expected since
org/10.1029/2007GL029797 it was trained on just under half of the underlying 15-minute
dataset), yielding O3 concentrations clustered around the 1:1
Virkkula, A., Mäkelä, T., Hillamo, R., Yli-Tuomi, T., Hirsikko, line.
A., Hämeri, K., Koponen, I.K., 2007. A Simple Procedure for
Correcting Loading Effects of Aethalometer Data. Journal of the
Air & Waste Management Association 57, 1214–1222. https://doi.
org/10.3155/1047-3289.57.10.1214

CLEAN AIR JOURNAL

Volume 30, No 2, 2020 12 © 2020. The Author(s). Published under a
Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 13 of 15

RAMP PM2.5 data compared to filter-based comparison of working day data between filter-based PM2.5
and the RAMP PM2.5 is worse (r2 of 0.08); this could reflect local
24-hour “reference” concentrations variations on specific days as the sites are 5 km apart. We use
For local calibration verification, we compare data from RAMP the slope (1.69±0.18) of a fit forced through zero to scale up all
#140 (deployed in Gacuriro) with the filter-based 24-hour PM2.5 weekday RAMP PM2.5 data. Due to RAMP malfunctions, no RAMP
concentrations reported by Kalisa et al. (2018) for the UR-CST data were collected for the dry season days (June 15-29) when
site (about 5 km from Gacuriro) in April 2017 (wet season). Kalisa Kalisa and co-workers collected filter-based measurements, so
et al. (2018) reported that in Kigali, workday PM2.5 was almost the wet season scaling factors are used for all RAMP data.
50% higher than non-workday PM2.5. The RAMP and filter-based
PM2.5 are strongly correlated (Figure S3) on eight out of ten non-
workdays, with a correlation coefficient (r2) of 0.77 and a slope Reference
of 1.39±0.06 (forced through the origin). On April 2 (Sunday), Kalisa, E., Nagato, E.G., Bizuru, E., Lee, K.C., Tang, N., Pointing,
there was construction activity on the CST site close to the filter S.B., Hayakawa, K., Archer, S.D.J., Lacap-Bugler, D.C., 2018.
sampler, which could explain high PM levels that would not be Characterization and Risk Assessment of Atmospheric PM2.5 and
seen in Gacuriro. It is not clear what was different about April PM10 Particulate-Bound PAHs and NPAHs in Rwanda, Central-
15 (Saturday), when the filter-based PM2.5 was comparable to East Africa. Environ. Sci. Technol. 52, 12179–12187. https://doi.
or higher than the highest workday PM levels. In this paper, all org/10.1021/acs.est.8b03219
weekend RAMP PM2.5 values are scaled upwards by 1.39. The

Figure S1: The RAMPs and the external Met-One NPMs in the lab (left) and a unit deployed in Kigali, Rwanda (right).

CLEAN AIR JOURNAL © 2020. The Author(s). Published under a

Volume 30, No 2, 2020 13 Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 14 of 15

Figure S2: Comparison of RAMP calibration models for ozone with reference monitor data at the Mt Mugogo Climate Observatory. The “Pittsburgh gRAMP”
model (A,C) is developed on collocations in Pittsburgh, PA, USA. The “Mugogo QR” model (B,D) is based on a collocation with the reference monitor at
Mugogo; the data shown here includes both training (28 days) and testing (32 days) data. JJAS (A,B) is the dry season and ON (C,D) is the wet season.

Figure S3: Developing scaling factors for optical RAMP PM2.5 measurements
by comparison with filter-based 24-hour PM2.5 measurements (reported by
Kalisa et al. 2018). Solid blue circles indicate workdays. Solid red circles
show eight non-workdays that are used to develop the non-workday
scaling factor, which excludes two non-workdays that experienced
significantly higher filter-based PM2.5 than even workdays (shown by the
open circles.) The filter measurements were conducted at the University of
Rwanda campus, while these RAMP measurements were conducted about
5 km away in the Gacuriro neighborhood.

CLEAN AIR JOURNAL

Volume 30, No 2, 2020 14 © 2020. The Author(s). Published under a
Creative Commons Attribution Licence.
Research article: Air pollution in Kigali, Rwanda Page 15 of 15

Figure S4: Ozone measurements in Kigali over the course of this study. (Left) Scatter plot shows most of the paired measurements across the city agree with
each other. (Right) However, there are periods when the UR-CST O3 is much higher than at Belle Vue estate.

Figure S7: Higher BC concentrations are associated with lower AAE values.

Figure S5: Intraurban variability of 1-hour CO concentrations across

Kigali.

Figure S6: Spatially-resolved rainfall data using the ENACTS-Rwanda

methodology, from the Rwanda Meteorological Agency. MAM 2018 saw
significantly more rainfall and more rainy days than other periods.

CLEAN AIR JOURNAL © 2020. The Author(s). Published under a

Volume 30, No 2, 2020 15 Creative Commons Attribution Licence.