Introduction

Air pollution is the main environmental health hazard worldwide1,2. For Africa, around one million premature deaths/year are associated with ambient air pollution, making it more deadly than major diseases like diarrhoea, HIV or malaria2,3,4,5. Rapid and unplanned urbanisation, characterised by inadequate support infrastructure and policies, contribute to increasing air pollution levels in sub-Saharan Africa (SSA) cities6,7,8,9. The SSA cities are growing at an annual average rate of 4.2% compared to a global average of 1.7%, and the urban SSA population is thereby projected to double by the year 204010. This calls for effective pollution control measures and policies to address the air quality challenges in the region, in order to reduce the health burden and ensure sustainable urbanisation.

Despite the strong impact on human health and regional climate, air pollution in SSA remains understudied9,11,12. The number of observation sites are few, and measurements are typically conducted only in short campaigns, while long-term, year-round observations are central for estimating health and environmental impact13,14,15,16. Meanwhile, the available data show high pollution levels and an increasing temporal trend7,17,18,19,20,21,22. Projections suggest that anthropogenic emissions in SSA will increase drastically over the coming decades7,23. This signifies a major risk to human respiratory health and regional climate, hence the need to combat air pollution. For instance, high exposure to PM2.5 and black carbon in Nairobi is an eminent public health risk16,22,24. Efforts to address air pollution challenges are however hindered by limited understanding of the relative contribution from different emission sources16,25,26.

Black Carbon (BC), a primary aerosol from fossil fuel combustion and biomass burning, is one of the more toxic species in PM2.5 and a potent climate warmer27,28,29. The sources of BC are poorly constrained in general, but more so in the SSA region due to limited observational data7,9,23,30. Chemical transport and climate models, as well as mitigation efforts, largely rely on bottom-up emission estimates in evaluating environmental effects. These estimates typically report large uncertainties, reflecting both activity estimates (fuel consumption data) but perhaps mainly emission factor estimates—emission per unit fuel combusted31,32,33. The activity estimates for SSA are comparably ill-constrained, and it is not generally clear that emission factors typically developed for other regions apply equally well to SSA7,23,30,34. Taken together, the uncertainties in estimates of bottom-up emissions are particularly large for SSA.

An alternative, or complement to bottom-up source-segregated emission estimates, is source quantification from source-diagnostic measurements in the ambient atmosphere. For BC, radiocarbon (14C) analysis has proven a particularly effective tracer for differentiating between fossil and biomass burning sources25,26,32,35. BC from fossil emissions is completely depleted in 14C (Δ14Cfossil = −1000‰), whereas biomass sources have a distinct Δ14C signature relative to atmospheric CO2 at the time of carbon fixation (Δ14Cbiomass ≈ +50 ‰, see discussion below)36. The 14C signature is determined specifically on BC isolates, and thus do not share the potential ambiguity of other chemical source tracers, which typically have different atmospheric fate, and are actually not parts of the BC pool. In addition, since BC is recalcitrant to physical and chemical transformations, it is not influenced by atmospheric processes and maintains its original 14C isotopic composition37. Therefore, the two source categories - fraction fossil and fraction biomass burning—can be differentiated at high precision. This technique has been applied for BC source apportionment in remote regions38,39 and heavily polluted cities40,41, yet so far not for the SSA region.

Here, we apply the radiocarbon technique to quantify the sources of BC in SSA. Our study site is located in Nairobi, which is one of the largest and most rapidly growing cities in SSA. Year-round (March 2014 to February 2015) PM2.5 filter samples were collected at an urban background site. Together with BC data, we present other PM2.5 components; water-soluble inorganic ions and organic carbon (OC) concentrations. The Δ14C signatures for BC allow us to quantify the relative contribution of biomass burning versus fossil sources, and assess how the sources vary seasonally.

Results and discussion

Meteorology and fires

The meteorology of Nairobi is governed by the East African monsoon system, with major (March to May) and minor (October and November) wet seasons and intersecting dry periods42. Overall, there is a limited seasonality in different meteorology parameters during the study period. For instance, comparably scanty rainfall was recorded even during the typical wet seasons, while winds were predominantly north-easterly (SI Fig. S3). On the other hand, the BT analyses show apparent seasonal shifts in air mass origin as a function of the annual oscillation of the inter-tropical convergence zone (ITCZ, Fig. 1). During Boreal winter, the air masses are largely of north-eastern origin, while south-easterly (SE) air masses are intercepted between March and November.

Fig. 1: Fire regimes and airmass back trajectories (BTs, blue lines) during the sampling campaign in Nairobi (white circle with a black outline).
figure 1

a During the December-February period, BTs arriving at Nairobi are mainly from the northeast sector. Meanwhile, large-scale fires mainly occur downwind, with little expected influence on PM2.5 composition in Nairobi. b. During March-November, the fire regime has largely shifted to the south. During this period, BTs arriving at Nairobi partially overlap with the fires, suggesting a potential impact on the air quality. The fire data were derived from the NASA Moderate Resolution Imaging Spectroradiometer (MODIS) satellite product for March 2014 to February 2015 sampling period. The 5-days air mass BTs were computed every 6 h, with an arrival height of 100 m above ground level (1890 m a.s.l).

Full size image

The dry African monsoon periods are associated with large-scale savannah and forest fires, clearly observable from space (Fig. 1). These fires mainly occur north of the equator (e.g. South Sudan, the Central African Republic and Cameroon) between December to February, while the fire regime shifts to the south during the Boreal summer dry period. Even though the majority and more intense fires during Boreal summer occur on the African south-west coast, in D.R. Congo and Angola, considerable fires also occur in Nairobi’s upwind, Madagascar and Tanzania (Fig. 1). Given the air mass transport pathways, the fires in these East African countries may influence the aerosol regime in Nairobi through long-range transport, while the analogous phenomenon during the Boreal winter dry period is less likely, as the fires occur downwind.

PM2.5 composition

During the present year-round campaign, fine particulate air pollution (PM2.5) in Nairobi was found to be continuously elevated, with an annual average of 27 ± 6 μg m−3. This is five times higher than 2021 WHO recommended annual mean PM2.5 guideline value at 5 μg m−3 while surpassing the WHO 24-h limit of 15 μg m−3 in all sampling days43. Overall, limited seasonal PM2.5 variability was observed in Nairobi, with slightly elevated values during the dry period (Fig. 2 and Table 1), and in agreement with a previous study18. No clear links between the observed concentrations and specific meteorological parameters, e.g. estimated PBL height and precipitation, were found. This suggests that variability in these meteorological parameters were not strong enough to impact the urban background aerosol loadings significantly, while atmospheric processing and emissions also modulate the loadings (SI Fig. S3). The largest component of PM2.5 was carbonaceous aerosols (CA; 64 ± 11%) with organic aerosols (OA = 2.1 × OC44) contributing 49 ± 7% of the PM2.5 loadings, while BC accounts for 15 ± 4%.

Fig. 2: Year-round variation of PM2.5, BC, and source-diagnostic ratios.
figure 2

a Shows PM2.5 and BC concentrations, b shows OC/BC and BC/PM2.5 ratios, c shows SO42−/BC and NH4+/BC ratios and d presents K+/BC and NO3−/BC ratios. The predominant air masses arriving at Nairobi are marked out on top, and the transition periods are highlighted in light brown colour.

Full size image
Table 1 Seasonally averaged PM2.5 and BC concentrations (μg m−3), Δ14CBC signatures (‰) and fraction fossil contribution (ƒfossil), shown with mean and standard deviation, for the year-round campaign in Nairobi.
Full size table

The contribution of water-soluble inorganic ions (WSII) amounted to 13 ± 5% of the PM2.5 loadings and was dominated by SO42− (1.8 ± 0.8 μg m−3; 7 ± 3%). Overall, a larger seasonality was observed in WSII components compared to carbonaceous aerosols, and elevated WSII concentrations were observed during the dry periods, likely attributable to prevailing meteorology and emissions variability, e.g. long-range transport from downwind fires (SI Fig. S1). Besides CA and WSII, 25 ± 5% of the PM2.5 remained unaccounted for, likely reflecting aerosol-bound water and elemental components.

The sea-salt contribution to WSII was estimated to be less than 2%, using Na+-based estimates, while Na+ correlated well with Mg2+ (R2 = 0.79, p < 0.01), another tracer for sea salt (SI Note S2). SO42− and NH4+ generally have different emissions source profiles but tend to form a stable salt, and are often found to be well correlated45. Here, a strong linear relation between SO42− and NH4+ (R2 = 0.75, p < 0.01, and R2 = 0.84, p < 0.01 on removing a single outlier), and a molar slope of NH4+ vs SO42− at 2.2, suggest the formation of (NH4)2SO4. Overall, the PM2.5 concentration and composition compare well with previously reported urban background values in other SSA cities (SI Table S1).

Black carbon concentrations

The BC levels at the urban background site in Nairobi were steadily high at 3.9 ± 1.2 μg m−3, through the year-round study period (Fig. 2). Similarly to PM2.5, no distinct seasonal trend was found for BC. The observed annual mean, and lack of seasonality, is comparable to previously reported values in a two-year study at the same sampling location (May 2008–April 2010; 3.9 ± 0.8 μg m−3)18, and measurements at an industrial background site in the city (a range of 2.3–7.8 μg m−3)46. However, these background values are much lower than the BC exposure levels reported at the curbsides and bus termini inside the city of Nairobi, which were found above 20 μg m−3 and accounted for over 30% of PM2.5 mass, as pointed out in the Nairobi City County air quality action plan and reference therein22,47,48,49. The persistently high BC concentrations thus pose a health risk to the over four million city residents.

The BC concentrations in Nairobi are comparable to reported levels in other SSA cities (Fig. 3; SI Table S1). Urban background levels above 3 μg m−3 are recorded in different cities across the region, based on the available shorter-term studies17,18,19,20,21,22,46,50,51. Although direct comparisons between cities are complicated by, e.g. intra-city variability and meteorology, the observed BC levels are comparable to those reported in urban background environments in megacities in South and East Asia, but higher than in European and North American cities (Fig. 3 and SI Tables S2, S3). While absolute concentrations are key for exposure and effects, ratios are less dependent on meteorological parameters such as PBL height and ventilation, and thus offer a more conserved means for comparing pollution characteristics at urban background locations in different cities. For the herein investigated cities, we observe that the BC/PM2.5 ratio in Nairobi and other SSA cities are in general elevated (~15%) compared to other global cities, suggesting a unique urban pollution regime and aerosol composition in the region (Fig. 3). Or in other words, the ambient BC pollution is particularly severe in SSA urban environments, which is aggravated by the fact that BC is a particularly toxic PM2.5 component.

Fig. 3: Comparison of BC/PM2.5 ratio (columns; %) and the corresponding BC concentrations (diamonds, with error bar as the standard deviation) reported in major cities globally.
figure 3

Different colours/shades are used to represent cities in different regions for ease of comparison. Nairobi* refers to this study. A detailed compilation and references are provided in SI Table S2.

Full size image

Source marker ratios

The ratios and correlations between different PM2.5 species can give insights into sources and trends. BC and OC are co-emitted from incomplete combustion but at varying emission factors for each, between different sources. Therefore, the OC/BC ratio has sometimes been used as a source-diagnostic marker, where elevated ratios often are interpreted as biomass-influenced regimes15. However, OC is also formed from secondary sources and is less recalcitrant than BC in the atmosphere, making OC/BC a non-conservative source tracer52,53. In this study, the OC/BC ratio range from 1.0 to 2.7 with little seasonal variability, and partially disparate origins, qualitatively indicating comparably low biomass burning contributions (Fig. 2 and SI Fig. S2).

Additionally, ratios of certain inorganic species may also be indicative of different emission sources. For instance, K+/EC ratio can be used as a wood/biomass burning marker. Here, we observed elevated K+/EC ratios during the dry boreal summer period, coinciding with the arrival of air masses through savanna-burning dominated regions in the south and post-harvest season (Fig. 2 and SI Fig. S2). Meanwhile, the NO3−/EC is often used as a tracer for traffic emissions and lightning strikes. However, recent studies suggest highly elevated levels of NO3− also in African wildfires15,54, and here we see increases in NO3−/EC during the dry period, although less pronounced when compared to the K+/EC trend, noting that the latter is more specific for wood burning (Fig. 2). Taken together, we find an imprint on Nairobi air of long-range air mass transport from boreal summer season large-scale African fires, but not to the extent that is seriously altering the air quality.

The source-diagnostic Δ14C signature in Nairobi BC

The Δ14C signature is a unique tracer for quantitatively constraining the relative contributions from biomass burning vs fossil combustion to BC, with high precision. Here, we find that BC aerosols in Nairobi were highly depleted in 14C throughout the study period (Δ14C = −840 ± 34‰; Fig. 4). Unlike biomass (Δ14Cbiomass = +57 ± 52‰), fossil fuels are radiocarbon dead (Δ14Cfossil = −1000‰). Therefore, the highly depleted Δ14C values indicate a dominant contribution from fossil fuel combustion. The Δ14C signature remained comparably constant throughout, with no distinct seasonal or temporal trend, and implying a minimal shift in BC source strength (Fig. 4).

Fig. 4: Year-round variations in BC source contribution in Nairobi.
figure 4

a Presents the Δ14C signatures of BC (Δ14CBC) and the corresponding fractional contribution of biomass burning (ƒbiomass), and b shows the Δ14C resolved fossil vs biomass sourced BC concentrations in Nairobi. Δ14C of −1000‰ implies 100% fossil. Roughly every second sample was analyzed for 14C signatures. The prevailing are masses received in Nairobi, and the transition periods are indicated.

Full size image

A high influence of fossil emissions on BC levels, and Nairobi’s air quality in general, is hereby realised. The year-round averaged fossil fraction accounted for 85 ± 3% of the BC emissions (Eq. 1). This translates to an annual average of 3.4 ± 1.1 μg m−3 of BC from fossil fuel combustion emissions and 0.6 ± 0.1 μg m−3 from biomass burning. Similar to BC concentrations, but different from biomass burning markers such as K+/BC, there was no clear seasonality in the fossil and biomass burning fractions and their respective concentrations (Fig. 4). Therefore, the BC sources were predominantly local, and constant through the year, with minimal influence from regional biomass burning episodes. The differences in the seasonal trends between 14C-derived biomass BC concentrations and K+/BC is explained by the large variability in biomass burning emission factors for different components (e.g. K+ is mainly a marker of wood-burning, while much of BC in Africa is also from burned grasses), but is also by the differences in atmospheric processing and atmospheric transport15. The elevated fossil fuel combustion contributions to BC found here is in qualitative agreement with previous conclusions for Nairobi18,22,55. However, while the present isotope-based study specifically source apportions BC, previous studies used different chemical tracers to apportion PM2.5, making the approaches and results largely complementary.

The high fossil contribution to BC reported here is consistent with the lack of an effective transport policy in Nairobi, leading to heavy traffic congestion16,55,56. Nairobi city is estimated to accommodate over a third of the 3.1 million registered vehicles in Kenya, while 68% of the fuel imports are consumed in the transport sector16,57. The fleet’s fuel economy, consisting mainly of second-hand imported vehicles and two-wheelers, is two to three times lower than in developed countries56,58,59,60,61. Besides, higher altitude in Nairobi (1690 m asl.) could increase vehicular emissions due to lower absolute O2 levels, which give a less efficient combustion62,63. The industry and commerce sector consumes 16% of the petroleum fuels, while coal contributes under a percentage of Kenya’s energy mix57. Meanwhile, in informal settlements and low-income neighbourhoods, the use of kerosene cookstoves and biomass fuels e.g. charcoal—we find a background biomass signature—is still prevalent and a potential BC source30,64.

Similar to Nairobi, urban sites around the world generally exhibit a higher fraction of BC from fossil origins (SI Table S3). 14C-derived fossil contributions accounting for over 75% of BC are reported in North American65,66 and European Cities67,68,69,70, while the actual BC concentrations in these cities are lower than in SSA cities. East Asian cities also have elevated fossil contributions and BC concentrations in the same range as Nairobi, but are much larger in size and activity levels25,41,71,72. For some South Asian cities, e.g. Delhi and Dhaka, there is a clear seasonal impact from upwind biomass burning activities, as detected by 14C in BC40,73. Although we do find elevated levels of biomass burning tracers, e.g. K+/EC, during periods of regional fires upwind of Nairobi, the signal is not detectable in the 14C-signal in BC, likely reflecting differences in emission factors and atmospheric fate.

Outlook

Air pollution is a major impediment towards resilient and sustainable cities in the SSA region, e.g. challenging the UN Sustainable Developmental Goals. A near-universal measure of air quality is PM2.5, with clear health guidelines defined by the WHO43. Despite increasing efforts, observational data on the magnitude, composition and sources of PM2.5 are still scarce in SSA6,8,9. Here, we report that the annual average PM2.5 levels in Nairobi are over five times the WHO recommended limits43. The overall air quality situation is further compounded by high indoor pollution levels, especially in informal settlements24,30,64. Especially, the relative contributions of BC in PM2.5 are highly elevated (~15%). Unfortunately, this appears to be a common feature among several SSA cities, that sets the region aside from other continents. While BC is particularly toxic, it is also a strong climate warming forcer of regional climate. Taken together, this suggests that BC should be specifically emphasised when discussing SSA air quality and when designing additional air quality measurement programmes, with potential benefits on climate.

The high fossil contributions to BC in Nairobi reported here most likely reflect traffic emissions, including high-emitting vehicles. The fossil contributions to BC in other SSA cities remains to be investigated, but traffic in SSA overall share the characteristics of Nairobi, making it a prime suspect when it comes to contributions to regional air quality6,9,56,58,59. An increasing trend in traffic use over a larger SSA region is consistent with the satellite-based observation of NOX54,74.

Overall, rapid urbanisation and population growth in SSA will, if left unregulated, lead to a rapidly deteriorating air pollution problem23,75. The region’s carbonaceous aerosol emissions are expected to contribute to 50% of the global emissions by 20307,76. Even for urban Nairobi, the contribution from carbonaceous aerosols to PM2.5 is major (64 ± 11%), while for background sites it is even larger15. Beyond human respiratory health, such emissions may interfere with the regional climate in unfavourable ways29. For instance, interference with the African monsoon system will have major consequences for floods, droughts, and freshwater supply, negatively affecting the region’s largely agrarian economies77,78.

Overall, this study stresses the importance of regional initiatives to combat air pollution and BC emissions in particular. Investing in an efficient public transport system, promoting non-motorised transport, and enforcing fuel and emission standards appear to be, although socio-economically challenging, feasible strategies to counter the current trajectory.

Methods

PM2.5 Sampling

The sampling site was located on the rooftop of one of the University of Nairobi buildings (1.279°S, 36.817°E; 1690 m above sea level; 17 m above ground level), in a park-like environment near the Nairobi city centre. Nairobi—Kenya’s capital city—hosts around 4.4 million residents, with a daytime population of over six million people79. Based on a compilation of literature studies, the ‘Nairobi City County Air Quality Action Plan (2019–2023)’ identified traffic, refuse to burn, and industrial emissions as key pollutant sources in the city16. However, the sampling site experienced neither direct influences from emission sources such as industries, traffic hotspots and dumpsite/refuse to burn, nor obstruction to the free flow of air masses. The surrounding roads usually experience low vehicle density, with restricted use for public service vehicles - locally known as ‘Matatus’ - and heavy trucks. Therefore, the site was considered representative of the background air quality in the city, as previously described18,80.

A high-volume sampler (model DH-77, Digitel A.G., Switzerland; flowrate of ~30 m3 h−1) with PM2.5 inlet was installed and used to collect 24-h filter samples every 6th day—to ensure all days of the week are represented—on prebaked (450 °C for 6 h) quartz fibre filters (Millipore, 150 mm diameter). Monthly field blanks were also collected by loading the filter into the sampler without starting the pump. In total, 66 filter samples were collected between March 2014 and February 2015.

Concentration measurements

PM2.5 mass concentrations were determined gravimetrically by dividing mass loading (difference in filter mass before and after sampling) by sampled air volume. The mass determination was performed in a specially-constructed temperature- and humidity-controlled room (T = 20 ± 1 °C, Rh = 40 ± 5%). Before weighing, filters were equilibrated for 24 h.

Water-soluble inorganic ions (WSII) were analyzed using the Dionex Aquion ion chromatography instrument (Thermo Finnigan LLC), applying a previously described analysis protocol53. The aerosol BC (quantified as mass-based equivalent-elemental carbon, EC) and organic carbon (OC) concentrations were measured with a thermal-optical transmission (TOT) analyzer (Sunset Laboratory, Tigard, Oregon) using the NIOSH 5040 protocol81. The instrument response was calibrated using a sucrose standard, while the analytical precision was ascertained using analyses traceable to the NIST Urban Dust Standard Reference Material, SRM-8785. The OC values were blank corrected by subtracting an average of the field blanks (0.9 ± 0.3 μg cm−2; equivalent to 0.02 μg m−3). No BC was detected in the field blanks (n = 13). Analysis of triplicates was conducted to check measurement precision and sample homogeneity, and showed a mean relative standard deviation of 3% for OC and 2% for EC, well within the instrumental error at 5 and 6%, respectively. For WSII, the average relative s.d. is <5% for all measured ions.

Δ14C source apportionment

Roughly every second sample (excluding blanks) for the year-round study period was selected for isolation and cryo-trapping of BC (n = 28) for 14C analysis, using a modified Sunset TOT instrument and a previously described protocol25,26. In brief, the BC deposited on the filter is thermally separated from OC and combusted into CO2. The produced CO2 is then diverted and purified online through silver wool and magnesium perchlorate traps to remove halogens and moisture, respectively. Subsequently, the CO2 is cryo-trapped in liquid N2 and flame sealed in glass ampoules. Ag and CuO were preadded into the glass ampoules, and combusted for 6 h at 400 °C, to remove gas impurities that may interfere with isotopic analyses26. The trapped CO2 samples were then analysed for Δ14C signatures using an accelerator mass spectrometer (AMS), at the Radiocarbon Laboratory at Uppsala University, Sweden.

To quantify the fractional contributions from biomass burning (ƒbiomass) versus fossil fuel combustion (ƒfossil = 1 − ƒbiomass), we applied the isotopic mass balance equation:

$${\Delta }^{14}{C}_{{{{{{\mathrm{sample}}}}}}}={f}_{{{{{{\mathrm{biomass}}}}}}}\cdot {\Delta }^{14}{C}_{\mathrm {biomass}}+(1-{f}_{{{{{{\mathrm{biomass}}}}}}})\cdot {\Delta }^{14}{C}_{{{{{{\mathrm{fossil}}}}}}}$$
(1)

where Δ14Csample represents the radiocarbon signature of the sample, and Δ14Cfossil is the fossil signature at −1000‰. The Δ14Cbiomass endmember may vary between +20‰ and +225‰, reflecting the Δ14C signature for atmospheric CO2, as influenced by 1960s nuclear bomb tests, fossil CO2 emissions, and global carbon recycling. As such, annual plants carry the ambient Δ14C signatures (+20‰ for 2014/2015), while more long-lived organic matter (e.g. trees) may be more enriched in 14C36,82. For this study, a regionally parametrised SSA biomass endmember for 2015, Δ14Cbiomass = +57 ± 52‰, was used15.

Pyrolized carbon, formed during combustion of OC in the helium phase of the NIOSH 5040 protocol, may be inadvertently trapped with BC fraction. Here we estimate that such phenomena may potentially shift the Δ14C of BC by a maximum of 30‰, which is within the uncertainty margin of the isotope measurements (SI Note S1).

Meteorology, air mass back trajectories and satellite products

Local meteorology parameters (e.g. wind speed and direction, temperature and rainfall) were obtained from the meteorological station at Jomo Kenyatta International airport (JKIA) in Nairobi and complemented with meteorological parameters retrieved from the Global Data Assimilation System (GDAS). To investigate potential influence from other geographical locations on air pollution in Nairobi, 5-days air mass back trajectories (BT) were computed every 6 hours, with an arrival height of 100 m above ground level (1890 m a.s.l). For BT analysis, the NOAA Hybrid Single-Particle Lagrangian Integrated Trajectory Model (HYSPLIT, version 4) and GDAS (1° × 1°) archived meteorological datasets were used83. Remote sensing fire-spot detections were retrieved from the NASA Fire Information for Resource Management Services (FIRMS) database, based on retrievals from the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite product84.