The sampling campaign was conducted at the industrial site of Yopougon with an area of 153 km² [14]. This measurement site (red dot) corresponds to GPS the coordinates 5˚23'18" North and 4˚4'35" West (Figure 1). The collection of particles was carried out three times a week with LVS/LV-S6-RV Sven Leckel sampler, from April 2018 to July 2019 [13]. The meteorological parameters including temperature, relative humidity and wind speed were provided by the Société d’Exploitation et de Développement Aérportuaire, Aéronautique et de Météorologique-Cote d’ivoire (SODEXAM).

PM concentrations were determined by gravimetric mean. Black carbon measurements were performed using an EEL Smoke Stain reflectometer (Model 43 M, Diffusion Systems Ltd 43) Figure 2. A light is source shines its light on the filter, and the reflected light is measured by photocells located in a black housing. The reflector reading is obtained directly from the universal digital readout and converted to output voltage. Both methods used are described elsewhere [13]. The particulate matter collected on the filters was quantitatively analyzed for trace elements by an Energy Dispersive X-ray Fluorescence (EDXRF) spectrometer of type X-123 (Figure 3). This spectrometer is composed of a fast SDD detector (25 mm diameter and 130 eV resolution), a mini X-ray tube with silver anode (30 kV and 25 µA), an excitation and emergence angle of 67.5˚, an X-ray tube-sample distance of 33.9 mm and a 15.9 mm detector-sample. The samples were irradiated during 300 seconds and the obtained X-ray spectra were processed using the XRS-FP software of CrossRoads Scientific Company. Then, the elemental contents given in µg/g were converted into airbone concentrations in µg/m³. This EDXRF method gives elemental concentrations with a typical error margin of 10%, which includes statistical counting errors of the detected elements in the sample. Twenty-two elements, namely Na, Mg, Al, Si, P, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Rb, Sr, Zr and Pb were detected and quantified.

Principal Component Analysis (PCA) with varimax rotation was used to estimate and identify the possible sources of coarse and fine particles. Thus, the chemical elements with higher concentrations in each factor were interpreted as fingerprints of emission source that it represents. In the present study, SPSS software was used to perform multivariate factor analysis.

The mean values concentration of fine (PM_2.5) and coarse (PM₁₀) particulates fractions were 48.83 µg/m³ and 77.34 µg/m³ respectively. The corresponding highest concentration was equal to 96.5 µg/m³ and 94.1 µg/m³. The time serie plots of the particulate matter (PM) in both size particles and their respective content in BC are presented in Figure 4 and Figure 5.

From the beginning of the great rainy season (May 2018) until October 2018, the fine particles showed an inverted behaviour to that observed for coarse particles (Figure 5 and Figure 6). After this period, the fine particles increased and reached one significant peak at the beginning of the great dry season (December 2018) with a concentration of 93.5 µg/m³. Thereafter it decreased considerably until almost at the end of the great rainy season (June 2019 37.64 µg/m³) before increasing slightly.

A variation of BC in both sizes (PM_2.5 and PM₁₀) was observed during the study period. A considerable peak of BC was recorded in PM₁₀ in December 2018 (Figure 4). This could be justified by the industrial stacks releases into the air during this period of the year.

The times series plot indicated that the monthly concentrations of PM₁₀ increased during the great rainy season (May 2018 to July 2018) samplings, decreased significantly from August 2018 to September 2018, followed by a slight increase at the beginning of the small rainy season. This seasonal trend could be attributed, in part, to the meteorological conditions. It was found that the higher values of PM₁₀ corresponded to lower temperatures and higher wind speed values and vice versa except for April 2019 to May 2019 where PM₁₀, temperature and wind speed had the same evolution (Figure 6).

Table 1 presents the results of PM studies of some African cities. The PM values found in these studies exceeded largely the WHO standards (PM_2.5: 15 µg/m³/24h; PM₁₀: 45 µg/m³/24h) [15].

It can be noted that the values of our study are lower than those found in these different African countries, but they remain higher than the international standards [15].

The elemental compositions, their average concentrations in PM_2.5 and PM₁₀ and standard deviations are presented in Table 2. Thirteen elements were determined for all the samples in coarse and fine particles. The average concentrations of these elements ranged from 0.0010 µg/m³ for Zr in PM_2.5 to 0.487 µg/m³ for Ca in PM₁₀.

Figure 7 shows the concentrations level of the elements detected in fine and coarse particles at the industrial area of Yopougon, Abidjan. For both fine and coarse particles, Zr was the element with the lowest concentration (0.0010 ± 0.0005 µg/m³). However, the highest concentration was recorded for K (0.202 ± 0.080 µg/m³) in fine particles and Ca (0.487 ± 0.188 µg/m³) in coarse particles. A comparison of the metal concentrations in both particles indicated that the elements of crustal origin (Al, K, Ca, Mn and Zr) were more prevalent in PM₁₀ than PM_2.5. Whole, the elements from anthropogenic sources (Cr, Ni, Cu, Zn and Pb) were less prevalent in fine particulates.

To further assess dominant source categories and quantify their contributions for coarse and fine aerosols, the principal component analysis (PCA) with varimax rotation was used. For coarse fraction (Table 3), four principal components (PCs) were extracted that, accounting for over 83% of the explained variance. The PCA results of PM₁₀ showed that the first factor (PC1), with the maximum percentage of variance (27.18%), had high loadings of Mn, Cl, BC, Cu, Zr and Pb. PC1 could show a combined contribution of industrial activities and non-exhaust emissions. This factor includes the contribution from vehicle non-exhaust sources traced by Mn and Cu, but it also receives significant mass contributions by combustion species like BC and Cl. BC can be found in combustion emissions [18] [19] and Cl can be considered as an elemental tracer for coal combustion [20] and industrial activities mainly composed of Zr and Pb which reflects the influence of ceramic industry processes [21]. PC2 mainly consisted of Zn, Ni, and Pb with 21.23% of the total variance, which was recommended as fingerprints for the metal processing industry. Oliveira et al pointed out industrial sources that had a strong contribution from Zn, Pb and Mn could be related to waste incineration or metallurgy [22]. PC3 showed a high loadings of Ca, Al, K, and Fe elements with a crustal origin, and explained 18.37% of the total variance. This factor is interpreted as a mixture of several sources including soil resuspension, urban works and regional mineral dust. PC4 explained 17.06% of the variance with high loadings of S, Cr and K. It would point out the role of waste combustion sources typically burning of woods [23].

*S.D. is Standard Deviation.

For PM_2.5, two principal components (PCs) were extracted by PCA analysis, accounting for over 81% of the explained variance. The PCA results of PM_2.5 (Table 4) showed that the first factor (PC1), with the maximum percentage of variance (66.85%), had high loadings of Al, S, Cl, K, Ca, Cr, Mn, Fe, Ni, Cu, Zn, Zr, and Pb. This factor looked like a combination of different sources, according to their characteristic tracers, as Non-exhaust (Pb, Zn), Industry (Cu, Ni, and Cr), Combustion (S, Ni) and Crustal (Al, Fe) [24]. Therefore, this factor 1 could be assigned to sources such as scrap packaging, stainless steel, ceramics, electronic products, mechanical wear of electrical components, vehicles parts and metal parts of used equipment. The second factor, which explains 14.42% of the total variance, was strongly correlated with the BC. This factor originated from exhaust emissions [25] [26].