Polycyclic Aromatic Hydrocarbons in Urban Air: Concentration Levels, Patterns, and Source Analysis in Nairobi, Kenya

This study describes polycyclic aromatic hydrocarbon (PAH) concentrations, patterns, and possible sources from atmospheric filter samples collected from three different areas in the city of Nairobi, Kenya. Total median concentrations for the 25 PAHs detected were higher in the traffic area (201 ng/m⁻³), followed by the residential area (141 ng/m⁻³), and lowest in the industrial area (128 ng/m⁻³). Results from the three sampled areas show that the percentage contributions of carcinogenic PAHs were approximately 30% of the total PAH concentrations reported. Some PAH isomer ratios differentiated traffic sources from non-traffic sources. Principal component analysis showed four significant principal components accounting for 82% of the variance. The first principal component (35%) was associated with fuel burning. The second principal component (27%) was associated with traffic emissions (diesel and gasoline). The other two principal components, which accounted for 12% and 8%, could not be interpreted with certainty. In order to interpret all the principal components in relation to sources, further collection of data is needed. More data points would have helped in further resolving the sources because data analysis models recommend more than 30 data sets.

Keywords:

• Polycyclic aromatic hydrocarbons
• concentration levels
• source analysis
• Nairobi
• principal component analysis

Introduction

Polycyclic aromatic hydrocarbons (PAHs) are ubiquitous environmental pollutants formed during the incomplete combustion of organic matter (Harrison et al., 1996; Mackay and Hickie, 2000). A number of high-molecular-weight PAHs are known carcinogens, mutagens, and some are suspected endocrine disrupters (Boström et al., 2002). To control their emission, source identification methods have been studied. Certain PAH concentration ratios to benzo[e]pyrene (B[e]P) have been used to differentiate PAH sources successfully (Ohura et al., 2004; Sanderson et al., 2004). Different sources of PAHs (e.g., gasoline and diesel engine cars, oil combustion from power plants, and domestic fuel burning) have been shown to generate different PAH profiles (Bzdusek et al., 2004; Li et al., 2003; Simcik et al., 1999). Profiles obtained from environmental monitoring might differ from actual source data due to such factors as mixing of PAHs from different sources, degradation during transport, weather conditions, and type of fuel used in a particular source (Dachs et al., 2002; Dickhut et al., 2000; Simcik et al., 1997). Therefore, site-specific PAH signatures help relate the compositional profiles from suspected sources to local and regional atmospheric conditions.

In Nairobi, Kenya, factors such as industrial activities, motor vehicles, and population have been on the rise recently, making air pollution a high priority. However, the lack of data on major air environmental pollutants that affect human health is an obstacle for source identification. PAHs in Nairobi could originate from different sources including industrial activities, motor vehicles, and household burning of fuel. The main types of fuel used in industrial area include industrial diesel, fuel oil, sawdust, pulverized coke, and used oil (UNEP, 1999), whereas power generation uses diesel. Motor vehicles in the city do not have catalytic converters because of the use of leaded fuel but this fuel will be phased out in 2006. The main types of petroleum products used in Nairobi are liquefied petroleum gas (LPG), premium and regular gasoline, automotive gas oil (diesel), industrial diesel, fuel oil, and bitumen. Crude oil imported in Kenya is low in sulphur and the refinery has the capability of reducing the sulphur content (UNEP, 1999). In the sampled areas, trash burning is limited because most of the solid waste is collected and dumped in the only solid waste dumpsite in Nairobi. Some waste in this site is openly burned; however, this could be a major PAH source because the open burning was located far from the sampled areas. Residential area samples were collected in a heavily populated area where most households use kerosene, charcoal, wood, and, to a lesser extent, LPG for cooking.

Due to the stability and health concerns of particle-bound PAHs, particulate matter in air was collected using quartz filter and analyzed. The aim for the present research was to report the concentration levels and patterns of these PAHs, including the high-molecular-weight fractions, which are potentially toxic and lacking in many PAH studies. Therefore, this study analyzed 30 different PAHs from atmospheric particle samples, collected for 2 to 3 hours from three different areas covering residential, industrial, and heavily trafficked areas during the period of August 25 to September 12, 2003, in Nairobi, Kenya. To the best knowledge of these authors, little or no data is available on PAH levels in Nairobi.

Materials and Methods

Sampling sites were located in traffic, industrial, and residential areas in Nairobi, Kenya (Figure 1). There were two traffic area sampling points—one was located near the junction of a major highway (Uhuru Highway) that links the port city of Mombasa to the city of Nairobi and other neighboring countries. The other traffic area was located near a junction where traffic from residential areas enters the central business district. Major activities in the vicinity where the industrial samples were collected include food factories, wood processing, tire manufacturing, and power generation. The residential area was located in a densely populated area where the major types of domestic fuel used are charcoal, wood, and kerosene. These fuels are used for both cooking and heating. During the month of September, the nights are cold and therefore some households heat their homes with charcoal.

Figure 1 Sampling sites in the city of Nairobi, Kenya.

Table 1 shows the sampling information during air data collection in Nairobi. Atmospheric particle samples were collected on quartz filters (QR100, Advantec, Japan; cut to 3.2-cm diameter and preheated to 450°C for 4 hours), using a mini-pump (MP-∑500, Sibata, Tokyo, Japan) at 5 L· min⁻¹ from the three different areas. Among the total of 24 samples, the volumes of air collected ranged from 0.47 to 0.93 m³. The filters sampled were folded and stored in aluminum foil under 4°C until analysis. The traffic area filters darkened a few minutes after start of sampling. The intensity of blackness was highest in traffic area samples. The industrial area filters were not as dark as traffic area samples whereas the residential area filters were light with a tinge of brown to black.

Table 1 Sampling information in Nairobi, Kenya

	Sampling period		Temperature °C		Humidity %		Volume
Sampling area	Time	Date	max	min	at 3 pm	at 9 am	L
Traffic	11.30–14.30	25-Aug-03	25	12	49	77	865
Traffic	15.30–19.30	03-Sep-03	23	13	54	85	902
Traffic	13.30–16.30	04-Sep-03	26	11	36	72	879
Traffic	10.00–13.00	06-Sep-03	27	11	39	98	914
Traffic	09.00–12.00	09-Sep-03	23	13	52	71	931
Traffic	09.00–12.00	02-Sep-03	21	13	52	80	463
Traffic	09.00–12.00	05-Sep-03	27	14	36	84	902
Traffic	09.30–12.30	10-Sep-03	29	14	45	80	911
Traffic	10.00–13.00	11-Sep-03	25	12	45	72	892
Traffic	09.00–12.00	12-Sep-03	25	12	57	72	901
Industrial	14.00–17.00	02-Sep-03	21	13	52	80	454
Industrial	10.00–13.00	03-Sep-03	23	13	54	85	902
Industrial	09.00–12.00	04-Sep-03	26	11	36	72	914
Industrial	14.00–17.00	05-Sep-03	27	14	36	84	865
Industrial	12.00–15.00	09-Sep-03	23	13	50	71	889
Industrial	14.00–17.00	10-Sep-03	29	14	45	80	901
Industrial	14.00–17.00	11-Sep-03	25	12	45	72	877
Industrial	13.00–16.00	12-Sep-03	25	12	57	72	470
Residential	19.00–21.30	02-Sep-03	21	13	52	80	767
Residential	19.30–22.00	03-Sep-03	23	13	54	85	762
Residential	19.00–21.00	04-Sep-03	26	11	36	72	492
Residential	17.00–20.00	08-Sep-03	24	13	52	87	853
Residential	17.00–20.00	09-Sep-03	23	13	50	71	895
Residential	19.30–21.00	10-Sep-03	29	14	45	80	707

The sampling time was enough to quantify PAHs collected in the filters, the concentration of PAH seems to be high due to speed at which the filters turned black. In addition, sampling was carried out during high traffic time, industrial production, and peak cooking time. Therefore, the samples were representative of the sampling area.

The PAH analysis was adapted from previously published methods (Mandoulet et al., 2000; Pleil et al., 2004; Shu et al., 2003; US EPA, 1996). Briefly, the filter samples were spiked with 100 μL of 1.25 μg/mL of deuterated PAH mixture containing naphthalene-d₈, acenaphthene-d₁₀, phenanthrene-d₁₀, chrysene-d₁₂, perylene-d₁₂, and benzo[ghi]perylene-d₁₂ prior to extraction (EPA 8270 surrogate mix, Sigma-Aldrich, USA). Automated Soxhlet equipment (Soxtherm, Gerhardt, Germany) was used in extracting the PAHs from the filter samples, using a 1:1 mixture of acetone:hexane solution (140 mL) at 140°C programmed for 2 hours and 30 minutes. The extract was cleaned up by a Sep-Pak silica cartridge (Waters, MA, USA) with 10–12 mL of 20% acetone:hexane mixture. The purified extract was then concentrated to 0.9 mL by N₂ purge at room temperature and 100 μL of 1 μg/mL internal standard (pyrene-d₁₀) was added to the vial. The sample was then stored in the freezer below 4°C until analysis with gas chromatography/mass spectometry (GC/MS). Two μL of the extracted sample in the vial was injected into a GC/MS (Agilent HP6890 series GC and HP5973 series MS) equipped with a DB-5 capillary column (60 m × 0.25 mm i.d.; 2.5 μm film thickness, J&W Scientific, Folsom, CA) and detected by selected ion monitoring (SIM) mode. The temperature program was operated as follows: initial temperature of 90°C, held for 2 minutes, ramped at 30°C·min⁻¹ to 200°C, then 5°C·min⁻¹ to 320°C and held isothermally at 320°C for 16 minutes. Injector and auxiliary transfer line temperatures were 280°C and 290°C, respectively. Carrier gas (helium 99.99% purity) was maintained at a constant flow rate of 1.4 mL/min⁻¹ throughout the analysis. The overall program lasted for 46 minutes and was able to analyze 30 different PAHs.

All target PAHs were calibrated using 30 PAH native standards (Quebec Ministry of Environmental PAH mix, Sigma-Aldrich, USA, and Accustandards, USA), and six surrogate standards. Internal standard (pyrene-d₁₀) was spiked in all the extracts before GC/MS analysis. Quality assurance program was carried out by: 1) setting up a six-point calibration curve from six different concentrations of the native and surrogate matrix (0.04, 0.1, 0.2, 0.5, 1.0, 1.5 μg/mL), 2) spiking the samples with a known concentration of the six surrogate standards before extraction, 3) running a 0.2 μg/mL matrix mixture of all the PAH standards in each GC/MS run to check calibration curve, and 4) running two filter blanks using the method previously described, which and showed very low or no peaks for the target PAHs. Detection limits for the PAHs calculated from S/N ratio multiplied by 3 ranged from 0.04 to 0.1 ng/m⁻³. Good correlation coefficients of over 0.99 were obtained between standard concentrations and peak areas for all the native and surrogate standards. The percentage recoveries for the surrogate standards were used to correct the concentration of the PAHs. Average recoveries for the surrogate standards were 52%, 54%, 82%, 74% and 77% for acenaphthene-d₁₀, phenanthrene-d₁₀, chrysene-d₁₂, perylene-d₁₂, and benzo[ghi]perylene-d₁₂, respectively. Recoveries of naphthalene-d₈ were below 30% due to its volatility.

Of the 30 target analytes, 25 were reportable in aerosol samples (Table 2). Namely, five PAHs, naphthalene, 2,6-dimethyl naphthalene, acenaphthylene, and acenaphthene concentrations, were not reported because of the low recoveries of naphthalened₈ surrogate standard. The concentration of 2-methyl anthracene was not reported because of matrix interference. Internal standard method and Agilent special ChemStation programs, such as retention time locking (RTL), were used to determine PAH concentration of the samples. PAHs were deemed detected when retention time and qualifier ion ratio of the standard matched that of a sample exactly. A deviation of less than or around 5 seconds for retention time and less than 10% of the qualifier ion ratio was allowed for some PAHs. A chromatogram of some of the PAHs that were measured in one of the samples is shown in Figure 2.

Table 2 Ranges and median concentrations of 25 PAHs (ng·m ⁻³) in Nairobi, Kenya

	Traffic area (n = 10)		Industrial area (n = 8)		Residential area (n = 6)
PAHs (abbreviation)	Range	Median	Range	Median	Range	Median
Fluorene (Flu)	0.6–16.4	2.1	0.8–9.7	1.8	0.6–7.3	2.5
Dibenzothiophene (Dbt)	0.6–6.6	3.8	2.1–7.9	3.3	2.8–7.8	4.1
Phenanthrene (Phe)	3.2–24.8	9.0	5.3–16.2	7.8	6.1–13.3	7.8
Anthracene (Ant)	0.5–8.1	2.4	1.2–3.3	1.7	1.4–8.7	5.0
3,6-Dimethyl phenanthrene (3,6-dm Phe)	0.5–7.7	3.0	1.0–2.9	1.6	1.0–4.7	1.8
Fluoranthene (Fth)	2.2–18.6	12.3	3.6–14.1	7.6	3.8–16.5	11.9
Pyrene (Py)	2.7–34.0	22.7	5.5–23.2	11.5	5.0–26.6	17.8
Retene (Ret)	0.8–13.3	4.5	1.6–5.7	2.7	1.9–11.0	4.4
Benzo[c]phenanthrene (B[c]P)	0.4–11.0	2.1	0.8–2.8	1.7	1.0–4.0	3.2
Benzo[a]anthracene (B[a]A)	1.0–17.3	10.4	3.1–12.6	6.2	4.9–22.5	10.8
Chrysene (Chy)	1.0–19.0	12.5	3.8–13.7	7.8	4.8–19.0	10.7
Benzo[b]fluoranthene (B[b]Fth)	0.8–35.8	8.1	3.8–31.0	6.4	5.3–9.1	8.1
7,12-dimethyl benz[a]anthracene (7,12-dmB[a]A)	0.1–2.9	0.3	0.2–4.0	0.3	0.3–11.5	0.5
Benzo[j+k]fluoranthene (B[j+k]Fth)	0.6–13.5	4.3	1.8–4.5	3.1	3.4–10.8	4.8
Benzo[e]pyrene (B[e]P)	1.1–21.0	12.4	4.7–11.0	7.1	5.2–11.9	7.3
Benzo[a]pyrene (B[a]P)	0.9–16.1	10.1	2.7–10.8	5.0	4.6–16.1	8.1
3-Methyl cholanthrene (3-Me Chol)	0.1–8.1	0.4	0.4–4.6	0.6	0.1–3.5	0.4
Indeno[1,2,3-cd]pyrene (IP)	1.4–17.0	12.7	4.1–12.9	8.0	6.0–11.0	7.4
Dibenzo[a,h]anthracene (Db[a,h]A)	0.3–9.9	0.8	0.2–1.8	0.6	0.5–2.9	1.4
Benzo[ghi]perylene (B[ghi]P)	1.9–43.1	31.3	9.0–24.6	15.8	7.4–13.8	7.6
Dibenzo[a,l]pyrene (Db[a,l]P)	0.4–10.0	1.6	0.7–3.8	1.2	1.0–3.3	1.8
Dibenzo[a,e]pyrene (Db[a,e]P)	0.3–18.8	2.4	1.1–3.3	1.7	0.8–2.4	1.4
Coronene (Cor)	1.2–29.2	19.1	5.2–14.7	8.9	2.8–8.1	5.1
Dibenzo[a,i]pyrene (Db[a,i]P)	0.2–4.3	0.4	0.2–0.6	0.4	0.0–2.1	0.6
Dibenzo[a,h]pyrene (Db[a,h]P)	0.2–4.5	0.4	0.1–1.5	0.5	0.3–2.1	0.6
Total	23.5–328.9	200.9	66.9–185.9	128.2	86.0–201.7	140.9

Figure 2 GC/MS chromatogram showing some of the PAHs detected in a traffic sample.

Statistical analysis was performed using statistical software (StatSoft Inc, Tulsa, USA). PAH concentration levels were reported for their range, median, and Pearson's correlation coefficients. The data were analyzed for sources by isomer ratio analysis and principal component analysis (PCA). Nonparametric tests, such as the Kruskal-Wallis test, were used to detect inherent differences in PAH concentration data obtained from the sampled areas.

Results and Discussion

Concentration Levels

Range and median concentrations of each of the PAH species and their totals from 24 filter samples in Nairobi are shown in Table 2. In the traffic area the most abundant PAHs were: Py, B[b]Fth, B[ghi]P, and Cor. Their concentrations ranged from 2.7–34.0 ng/m⁻³, 0.8–35.8 ng/m⁻³, 1.9–43.1 ng/m⁻³, and 1.2–29.2 ng/m⁻³, respectively. In the industrial area the same PAHs were abundant, but their concentrations were lower. They ranged from 5.5–23.2 ng/m⁻³ (Py), 3.8–31 ng/m⁻³ (B[b]Fth), 9–24.6 ng/m⁻³ (B[ghi]P), and 5.2–14.7 ng/m⁻³ (Cor). In the residential area Py, B[a]A, and Chy were more dominant and their concentrations ranged from 5–26.6 ng/m⁻³, 4.9–22.5 ng/m⁻³, and 4.8–19 ng/m⁻³, respectively. Concentrations of B[b]Fth, 5.3–9.1 ng/m⁻³, B[ghi]P, 7.4–13.8 ng/m⁻³, and Cor, 2.8–8.1 ng/m⁻³ in residential areas were lower compared with traffic and industrial areas. Total median concentrations for the 25 PAHs were higher in the traffic area (200.9 ng/m⁻³), followed by the residential area (140.9 ng/m⁻³), and lowest in the industrial area (128.2 ng/m⁻³). These results are expected because traffic emissions have been shown to be a major source of PAHs in urban areas (Park et al., 2002; Guo et al., 2003). For specific PAH markers for traffic emissions such as B[ghi]P and Cor, median concentrations were approximately two and four times higher in trafficked areas than in industrial and residential areas, respectively.

Median concentrations of carcinogenic PAHs such as B[a]P were high in the trafficked areas (10.1 ng/m⁻³) and lower in the industrial (5.0 ng/m⁻³) and residential areas (8.1 ng/m⁻³). These concentrations were higher than that recommended by World Health Organization (WHO) guidelines for Europe of 0.1 ng/m⁻³ (Boström et al., 2002) and a newly established annual average value of 1.0 ng/m⁻³ for PM10-associated B[a]P (European Parliament, 2004). However, it should be kept in mind that direct comparison of PAH concentrations among the three sites is not allowed because sampling in the traffic area was conducted during the rush morning hours, whereas in the industrial and in the residential areas, sampling was carried out during different time periods with possibly lower PAH emissions and/or with atmospheric conditions more favorable for pollutant dispersion. PAHs that had average concentrations lower than 0.7 ng/m⁻³ included 7,12-dmB[a]A, 3-Mchol, Db[a,i]P, and Db[a,h]P. Concentration levels of Db[a,l]P were reported, but it should be noted that a number of other dibenzopyrene isomers elute at the same retention time, complicating the certainty of detecting this PAH. Concentration data in this study for PAHs such as Fth, Py, B[b]Fth, B[e]P), and Cor were comparable with those presented by Fang et al. (2004) in Taiwan and Smith et al. (1996) in Pakistan.

Average percentage contributions of individual PAHs to the total of 25 PAHs measured are presented in Figure 3. Statistical analysis (Kruskal-Wallis test) showed no significant differences between the sampling areas for three-ring (H = 0.64, df = 2, P < 0.01), four-ring (H = 2.15, df = 2, P < 0.01), five-ring (H = 1.87, df = 2, P < 0.01), and six-ring PAHs (H = 7.63, df = 2, P < 0.01). This could be partly due to mixing of the PAH sources or a strong influence of one source in the sampled areas.

Figure 3 Average percentage contribution of the 25 PAHs from sampling areas.

Carcinogenic PAHs

PAHs that have been classified as possible carcinogens include: B[a]A, B[b]Fth, 7,12-dmB[a]A, benzo[j+k]fluoranthene (B[j+k]Fth), B[a]P, indeno[1,2,3-cd]pyrene (IP), dibenz[a,h]anthracene (Db[a,h]A), Db[a,l]P, dibenzo[a,e]pyrene (Db[a,e]P), Db[a,i]P, and Db[a,h]P (Boström et al., 2002; Tsai et al., 2004; WHO/IPCS, 1998). Results from the three sampled areas show that the percentage contribution of carcinogenic PAHs is approximately 30% of the total PAHs concentrations reported. This percentage is slightly higher than that reported by Tsai et al. (2004) for medium- and high-molecular-weight PAHs from samples collected in a trafficked area.

Source Analysis

Isomer Ratios

Sources of PAHs have been explained using isomer ratios (Guo et al., 2003; Manoli et al., 2004; Ohura et al., 2004; Park et al., 2002). Table 3a and Table 3b show the range and average isomer ratios observed in this study, and source isomer ratios reported in literature, respectively. The average ratios for Fth/(Fth+Py), B[e]P/(B[e]P+B[a]P), and IP/(IP+B[ghi]P) ranged between 0.37–0.41, 0.48–0.58, and 0.33–0.45, respectively. These values are comparable with those reported by Kavouras et al. (1999) of 0.41 ± 0.1, 0.45 ± 0.27, and 0.32 ± 0.22, for vehicle emissions (gasoline and diesel engines). The ratios calculated were almost constant in the three different areas sampled indicating that a single source or multiple constant sources which had the constant isomer ratio might have influenced the three sampling areas. The average ratio of B[ghi]P/B[a]P was 2.89 in industrial area, 2.81 traffic area, and 1.39 in residential area. The high values obtained in industrial and heavily trafficked areas indicate that gasoline emissions were major sources in these areas (Rogge et al., 1993).

Table 3a Summary of isomer ratios of selected PAHs

	Traffic area		Industrial area		Residential area
Isomer ratios	Range	Average	Range	Average	Range	Average
Fth/(Fth + Py)	0.26–0.46	0.37	0.32–0.45	0.39	0.08–0.42	0.41
B[a]A/Chy	0.76–1.13	0.91	0.61–1.50	0.87	0.77–1.25	1.06
B[a]A/(B[a]A +Chy)	0.43–0.53	0.48	0.38–0.60	0.46	0.43–0.56	0.51
B[e]P/(B[e]P+B[a]P)	0.49–0.69	0.56	0.45–0.69	0.58	0.36–0.63	0.48
B[ghi]P/B[a]P	1.65–4.12	2.81	2.28–3.42	2.89	0.47–3.04	1.39
B[ghi]P/B[e]P	1.16–3.34	2.20	1.27–2.79	2.09	0.63–1.81	1.32
IP/(IP+B[e]P)	0.42–0.56	0.51	0.44–0.60	0.51	0.41–0.59	0.51
Cor/B[e]P	0.46–2.09	1.35	0.82–1.67	1.20	0.32–1.07	0.75
IP/(IP+B[ghi]P)	0.27–0.46	0.33	0.29–0.46	0.34	0.31–0.53	0.45

Table 3b Reference isomer ratios of selected PAHs from previous research

	Traffic sources		Combustion
Isomer ratios	Diesel	Gasoline	Oil	Wood	Others industrial
Fth/(Fth + Py)	0.38 ^a , 0.60–0.70 ^d	0.17 ^a , 0.40 ^b	0.62 ^a
	0.41 ± 0.10 ^k
B[a]A/Chy	0.53 ± 0.06 ^f			0.79 ± 0.13 ^f	0.60 ± 0.06 ^f
B[a]A/(B[a]A +Chy)	0.38–0.64 ^d	0.43 ^b	0.32 ^a	0.50 ^b
B[e]P/(B[e]P+B[a]P)	0.64 ^a , 0.70 ^b	0.87 ^a	0.68 ^a	0.34 ^g
	0.45 ± 0.27 ^k
B[ghi]P/B[a]P	1.16 ^a , 1.2–2.2^b	1.72 ^a , 2.5–3.3 ^b	0.40 ^a		1.69–8.63 ^a
Cor/B[e]P	1.8 ^i , 1.54 ± 0.16 ^j				0.35–0.48 ^h
IP/(IP+B[ghi]P)	0.32 ± 0.22 ^k
	0.35–0.70 ^b	0.21–0.22 ^c	0.82 ^a	> 0.5 ^e
, , , , , , , , , ,

The average ratio for Cor/B[e]P was low in residential (0.75) relative to industrial (1.20) and heavily trafficked (1.35) area, indicating that sources other than traffic influenced PAH levels in residential area. These results can be compared with the one presented by Nielsen (1996) who reported a value of 1.54 ± 0.16 for traffic emissions. Larsen and Baker (2003) reported a higher ratio of 1.8 than the values recorded in this study. Samples that recorded a value > 1 for the Cor/B[e]P ratio were interpreted as influenced by traffic sources. Within this region were 80% of the traffic area samples, 62% of industrial area samples, and 16% of residential area samples. Samples that recorded a value < 1 for the same ratio were considered to be from other sources than traffic since five out of six residential samples and three out of eight industrial samples were within this region. However, one must take into account the extensive use of diesel in industrial area as a contributor to the high percentage observed and PAH weathering as a contributing factor to the percentage shown in residential area.

A plot of B[a]A/(B[a]A+Chy) versus IP/(IP+B[ghi]P) ratios shows differences among the sampling areas (Figure 4). The plot shows an approximate positive linear relationship for residential area samples (R = 0.93, P < 0.005) an indication that both pyrogenic and petrogenic combustion influence the area (Tsapakis and Stephanou, 2005; Yunker et al., 2002). For the latter ratio, seven traffic samples were within the range of 0.25–0.35, whereas three were within the range of 0.35–0.70; they were interpreted as gasoline and diesel engine emissions, respectively, based on literature values. For example, Khalili et al. (1995) reported a range of 0.21–0.22 for gasoline engine vehicles, whereas Manoli et al. (2004) reported a range of 0.16–0.39 for the same ratio in ambient air in Thessaloniki, Greece, suggesting a stronger contribution from gasoline than diesel engine emissions. Three residential area samples recorded values of between 0.49–0.77, which Yunker et al. (2002) associated with wood combustion.

Figure 4 Plot of isomer ratios B[a]A/(B[a]A + Chy) versus IP/(IP + B[ghi]P).

PAH Weathering

A brief report on the meteorological conditions of Nairobi indicates that during the months of August and September it is mostly cloudy and sunshine occurs between 4.2 and 5.9 hours in a day. Also in May through September, east and southeast winds of 10–15 mph are dominant. The winds pass through industrial and traffic areas located south of the residential area (approximately 10–15 km from industrial and traffic areas). It is expected that weathering will affect PAHs as they move through air from traffic and industrial areas to residential area. Correlation analysis showed that two PAHs (Cor and B[ghi]P, source markers for diesel and gasoline engines) had a high ratio in all the sampled areas, suggesting a common source. These authors therefore expect weathering of these two PAHs as they move through air from industrial and traffic areas to residential area. Four ratios, namely IP/(IP+B[e]P), IP/(IP+B[ghi]P), Cor/B[e]P, and B[ghi]P/B[e]P, were used in interpretation of weathering. The average ratio of IP/(IP+B[e]P) was the same in the all the areas sampled areas, probably because these PAHs have sources in the three areas and/or they are more stable, therefore undergoing less weathering, whereas the average ratio for IP/(IP+B[ghi]P) was almost the same in traffic (0.33) and industrial areas (0.34) relative to residential area (0.45), which was slightly higher. This result is probably due to B[ghi]P degradation from industrial and traffic area sources to the residential area. The Cor/B[e]P ratio was low in the residential area (0.75) compared with industrial (1.20) and traffic (1.35) areas. Coronene is a marker of traffic pollution; it is probable that this PAH was degraded as it moved from the traffic area to residential area, given the low ratio in residential area. The average B[ghi]P/B[e]P ratio between traffic and industrial area relative to residential areas was higher than that of Cor/B[e]P due to the stability of Cor relative to B[ghi]P in terms of photochemical reaction (Esteve et al., 2004). Baymer and Hites (1988) indicated that dark (i.e., high carbon content) substrates stabilize PAHs to photolytic breakdown because they absorb more light, making less light available for photolysis. The intensity of blackness of the filter samples was highest for the traffic area followed by industrial area samples but the residential area samples were less dark. The main routes of PAH removal from the atmosphere could be through dilution, dry deposition, and photochemical degradation (Dachs et al., 2002). Therefore, it is probable that dilution of PAHs occurs as they move in the air mass followed by photochemical loss by OH, NO₂ scavenging (Dachs et al., 2002; Esteve et al., 2004; Hayakawa et al., 2002; Simcik et al., 1997).

Correlation Analysis

Correlation between different PAHs has been used to explain the sources of PAHs (Ohura et al., 2004, Schauer et al., 2003). Miguel et al. (1998) and Larson and Baker (2003) showed high correlation existed between Cor and B[ghi]P for diesel and gasoline emissions sources. Table 4 shows that the correlation coefficient between Cor and B[ghi]P was high (R = 0.97, n = 10) for the trafficked area, suggesting diesel and gasoline engine emissions as major PAH sources in the area. High correlations between B[a]A and Py (R = 0.98, n = 8) and between B[a]P and Fth (R = 0.92, n = 8) were recorded for industrial area samples. These PAHs were associated with oil combustion (Larsen and Baker, 2003; Harrison et al., 1996; Yunker et al., 1996). In residential area samples, high correlations were recorded between chrysene (Chy) and B[a]A (R = 0.93, n = 6); Py and Fth (R = 0.93, n = 6); and Chy and Fth (R = 0.90, n = 6). These PAHs have been associated with domestic fuel burning (Hays et al., 2003; Lee et al., 2004; Simcik et al., 1999; Rogge et al., 1993).

Table 4 Correlations of selected PAHs in the sampling areas

PAH	TA	IA	RA
B[ghi]P - Cor	R = 0.97 n = 10	R = 0.97, n = 8	R = 0.91, n = 6
Py - Fth	R = 0.62, n = 10	R = 0.93, n = 8	R = 0.93, n = 6
Py - B[a]A	R = 0.32, n = 10	R = 0.98, n = 8	R = 0.58, n = 6
Py - B[a]P	R = 0.23, n = 10	R = 0.93, n = 8	R = 0.33, n = 6
Fth - B[a]P	R = − 0.16, n = 10	R = 0.92, n = 8	R = 0.45, n = 6
Chy - B[a]A	R = 0.76, n = 10	R = 0.52, n = 8	R = 0.93, n = 6
Chy - Fth	R = − 0.04, n = 10	R = 0.65, n = 8	R = 0.90, n = 6
IA = industrial area, n = number of samples, R = correlation coefficient, RA = residential area, TA = trafficked area

Generally, all samples from different areas showed a high correlation for six- and seven-ring PAHs, suggesting diesel or gasoline contribution. The differences in Table 4 were observed because traffic area samples did not show higher correlations for four- and five-ring PAHs, whereas industrial area samples showed better correlations between four- and five-ring PAHs, relative to residential area samples that showed good correlations for the four-ring PAHs.

Principal Component Analysis (PCA)

PCA was employed to distinguish PAH sources and describe inherent differences in PAH profiles. To avoid bias in PCA, concentrations of fluoranthene to coronene were used because recovery standards used to correct their concentrations were high. Concentrations of 7,12-dmB[a]A, 3-Mchol, Db[a,l]P, B[c]P, Db[a,i]P, and Db[a,h]P were low and therefore not included in PCA because spurious effects could be introduced. Preceding PCA, individual PAH concentrations were transformed to individual PAH percentages, divided by the total PAH concentrations measured in each sample and multiplied by 100 (Johnson et al., 2002; Kim et al., 2004). The obtained matrix of 14 individual PAHs by 24 samples was then used for PCA.

PCA extracted four factors with eigenvalues > 1, which accounted for 82% of the variance in total. Varimax rotation did not further improve the resolution of components The first principal component accounting for 35% of the variance had a high loading for B[a]A (0.87), Chy (0.76), Fth (0.73), Py (0.69), and B[a]P (0.68) (Table 5); these PAHs have been associated with domestic fuel burning (Rogge et al., 1993; Lee et al., 2004) and stationary sources in industrial areas (Fang et al., 2004). This principal component was interpreted as domestic and industrial fuel burning. The second principal component had a positive moderate loading for B[b]Fth (0.65) and Db[a,h]A (0.73) high loading for B[j+k]Fth (0.91), high negative loading for B[ghi]P (0.78), and moderate negative loading for Cor (0.68). Collectively, this pattern was attributed to traffic emissions. These PAHs are markers of diesel and gasoline engine vehicles (Miguel et al., 1998; Sanderson et al., 2004; Schauer et al., 2003). The third and fourth principals accounting for 12% and 8% could not be interpreted fully. However, the factors may be linked to wood combustion because of the positive loadings for Ret (0.62), and moderate loadings for IP (0.45) (Didyk et al., 2000; Li and Kamens, 1993; Lehndorff and Schwark, 2004). The result would have been more conclusive if other wood source markers such as 1,7-dimethylphenanthrene and 2,6-dimethylphenanthrene were measured (Larsen and Baker, 2003). Figure 5a and Figure 5b show the score and loading plots of the first versus the second PC. These plots show the influence of fuel burning source markers (B[a]A, Chy, Py, B[a]P, and Fth) in the residential area. This is probably due to influence from industrial sources in addition to residential area sources, a fact reinforced by the extensive use of fuel in the industrial area in addition to the residential areas where charcoal is used both for cooking and space heating.

Figure 5a PCA scores of the individual PAHs normalized by total percentage.

Figure 5b PCA loadings of the individual PAHs normalized by total percentage.

Table 5 PCA loadings for particle bound PAHs in Nairobi

PAHs	Factor 1	Factor 2	Factor 3	Factor 4
Fth	0.73	−0.09	0.25	0.06
Py	0.69	−0.64	0.21	0.04
Ret	0.10	0.53	0.62	0.36
B[a]A	0.87	0.10	−0.39	0.00
Chy	0.76	0.12	−0.36	−0.24
B[b]Fth	−0.58	0.65	0.21	−0.24
B[j+k]Fth	−0.01	0.91	−0.26	−0.05
B[e]P	−0.64	0.23	−0.28	−0.56
B[a]P	0.68	0.02	−0.60	0.20
IP	−0.48	−0.27	−0.12	0.45
Db[a,h]A	−0.14	0.73	−0.23	0.48
B[ghi]P	−0.57	−0.78	−0.17	0.01
Db[a,e]P	−0.57	0.20	−0.40	0.32
Cor	−0.64	−0.68	−0.25	0.13
Variance attributed to each factor
	4.9	3.8	1.6	1.2
Percentage	35.0	27.0	12.0	8.0

Varimax-rotated PCA of log-transformed PAH concentrations that have been Z-score normalized were similar to percentage normalized results. Four principal components accounting for 82% of the variance were obtained from both methods. PC2 and PC3 explaining 40% and PC1 27% of the variance were interpreted as fuel burning and traffic emissions, respectively. These values correspond with PC1 and PC2 from the previously calculated results. Therefore, from PCA of the major sources of PAHs in Nairobi are fuel burning and traffic emissions with moderate inputs from other sources.

Conclusions

In summary, PAH concentrations in the traffic areas were higher than in the other two sampling areas. Specific PAH markers such as B[ghi]P and Cor concentrations were higher in trafficked areas. B[a]P concentration was higher than the recommended value of WHO guidelines in Europe. Contribution of carcinogenic PAHs to the 25 total PAHs reported in Nairobi was approximately 30%. PAH profile data showed no statistically significant differences between the sampling areas. Isomer ratio results such as Cor/B[e]P were able to differentiate traffic sources from non-traffic sources, whereas the B[ghi]P/B[a]P ratio showed that traffic and industrial areas were influenced more by gasoline and diesel engines relative to residential areas. Environmental weathering of the PAHs was probably through wind dilution followed by photochemical loss, a fact shown from the difference in the previously mentioned isomer ratios from traffic, industrial, and residential areas. High correlations of some PAHs such as Cor and B[ghi]P in trafficked areas, B[a]A and Py in industrial areas, and B[a]A and Chy in residential areas were associated with traffic emissions, oil, and fuel combustion, respectively. Overall source analysis from isomer ratios and PCA indicate that fuel burning and traffic emissions were the major sources of PAH levels in the three sampled areas in Nairobi and other sources were not outstanding or could not be explained because of PAH source mixing.

Acknowledgement

This work was partly supported by the 21st Century COE Program “Environmental Risk Management for Bio/Eco-system” of the Ministry of Education, Culture, Sports, Science and Technology of Japan.

Notes

^a Manoli et al. (2004)

^b Rogge et al. (1993)

^c Khalili et al. (1995)

^d Sicre et al. (1987)

^e Yunker et al. (2002)

^f Dickhut et al. (2000)

^g Li and Kamens (1993)

^h Ohura et al. (2004)

ⁱ Larsen and Baker. (2003)

^j Neilsen (1996)

^k Kavouras et al. (1999)