Biomass burning contribution to ambient air particulate levels at Navrongo in the Savannah zone of Ghana

Abstract

The concentrations of airborne particulate matter (PM) in Navrongo, a town in the Sahel Savannah Zone of Ghana, have been measured and the major sources have been identified. This area is prone to frequent particulate pollution episodes due to Harmattan dust and biomass burning, mostly from annual bushfires. The contribution of combustion emissions, particularly from biomass and fossil fuel, to ambient air particulate loadings was assessed. Sampling was conducted from February 2009 to February 2010 in Navrongo. Two Gent samplers were equipped to collect PM₁₀ in two size fractions, coarse (PM_10-2.5) and fine (PM_2.5). Coarse particles are collected on a coated, 8-μm-pore Nuclepore filter. Fine particle samples were sampled with 47-mm-diameter Nuclepore and quartz filters. Elemental carbon (EC) and organic carbon (OC) concentrations were determined from the quartz filters using thermal optical reflectance (IMPROVE/TOR) methods. Elements were measured on the fine-particle Nuclepore filters using energy-dispersive x-ray fluorescence. The average PM_2.5 mass concentration obtained at Navrongo was 32.3 μg/m³. High carbonaceous concentrations were obtained from November to March, the period of Harmattan dust and severe bush fires. Total carbon was found to contribute approximately 40% of the PM_2.5 particulate mass. Positive matrix factorization (PMF) suggested six major sources contributing to the PM_2.5 mass. They are two stroke engines, gasoline emissions, soil dust, diesel emissions, biomass burning, and resuspended soil dust. Biomass combustion (16.0%) was identified as second most important source next to soil dust at Navrongo.

Implications:

Black carbon has now been recognized as playing a major role in radiation forcing, and one major source of biomass burning is slash-and-burn agriculture. Such practices are used in Sub-Saharan Africa and the time for field burning corresponds to the time when there is also substantial transport of Saharan dust to West Africa. This paper shows the major role of biomass burning on fine particle concentrations in northern Ghana and provides useful data on the EC concentrations in this area.

Introduction

Ambient atmospheric fine particle concentrations and compositions play a major role in pollution transport, visibility degradation, and climate change. High concentrations of PM_2.5 and PM₁₀ (i.e., particles of aerodynamic diameters of 2.5 μm and 10 μm or less, respectively) in the atmosphere have been shown to have significant influence on human health worldwide (Dockery, 1997). To better understand the role of fine particles in these processes, it is important to quantify anthropogenic and natural components and their regional and seasonal variations, as well as to identify their possible sources.

Pollutants of major public health concern include particulate matter (PM₁₀ and PM_2.5) and gases such as NOx, SO₂, O₃, and CO (Akhter et al., 2004; Begum et al., 2007) that have known severe health implications (Dockery and Pope 1994; Pope, 2000; Schwartz, 2001). Fine particles in the atmosphere have both anthropogenic and natural origins. Anthropogenic sources are largely due to combustion processes: motor vehicle emissions, fossil fuel burning, large industrial processes, and biomass burning. Natural sources include windblown soils, volcanic emissions, sea spray, and lightning-induced biomass burning (Cohen et al., 2010).

African savanna fires almost all result from human activities and produce as much as one-third of the total global emissions from biomass burning (Bond et al., 2013). Emissions from African savanna burning are known to be transported over the mid-Atlantic, south Pacific, and Indian oceans and thus have global climate implications (Levine, 1991; Gatari et al., 2006; Cohen et al., 2010). The high acidity of the precipitation in the tropics is known to be attributed to biomass burning. As a result of the enormous quantities of biomass burned each year in Africa and the Amazon, ozone levels in the burning season are close to those in the industrialized world (Weler, 2003). Ozone concentrations in Africa are almost high enough to be toxic to plants and are likely to be responsible for crop losses.

On a global average basis, Ramanathan and Carmichael (2008) estimate that black carbon alone produces an atmospheric forcing of +2.6 W/m² and a surface forcing of –1.7 W/m² (combined forcing of +0.9 W/m²). A recent review (Bond et al., 2013) has suggested that black carbon is the second most important source of warming after CO₂. Thus, the importance of particulate carbon in air quality, human health, and the ecosystem underscores the need to assess and control its levels in the atmosphere.

The objectives of this work are to determine seasonal variations (Harmattan and non- Harmattan) of PM_2.5 mass and carbon concentrations in ambient air and the possible sources of the particulates with emphases on biomass contribution.

Experimental Methods

Sampling

The sampler was located at the Navrongo Campus of the University for Development Studies (UDS). The sampling location was in Navrongo, a town located at about 11 km south of the main border between Ghana and Burkina Faso (Figure 1). Located in the Guinea savannah belt, the place is typically Sahelian (hot and dry), with the vegetation consisting mostly of semi-arid grassland interspersed with short trees and flat terrain. The sampling period covered 1 year from February 2009 to February 2010 with 110 sampling days. However, the sampling schedule was not uniform across the year, and more samples were collected during the periods in which high particulate matter mass concentrations were expected to be observed.

Figure 1. Map showing the monitoring site in Navrongo, Ghana.

The region has two main seasons, wet and dry. The wet season extends from April to October, with heavy rainfall mainly occurring between June and October. The highest level of rainfall is recorded in the months of August and September (Figure 2b). The dry season is subdivided into the Harmattan (November–mid-February) and the dry hot season (mid-February–April). Monthly temperatures range from 20°C to 40°C, with the mean minimum and maximum temperatures estimated at 22.8°C and 34.4°C, respectively. Navrongo is located at 10° 53′ 5″ N, 1° 5′ 25″ W and is at an altitude of 200–400 m above sea level. The farming seasons occur from the month of May to October. Tractor and animal ploughs are the major means of preparing lands for farming.

Figure 2. (a) Monthly mean PM_2.5 mass concentrations. (b) Monthly mean rainfall measured in Navrongo.

The major anthropogenic sources of particulate pollution in Navrongo are fuelwood consumption in homes, biomass burning from bushfire and open refuse dumps (open burning), fossil fuel combustion (from automobile exhaust), dust from construction, and resuspension. There are no heavy industrial activities that emit pollutants into the ambient air. The period of bush burning occurs from the month of November to February within the dry (Hamattan) season. The natural source of particulates is the windblown dust, which includes Harmattan dust. Few scattered (sporadic) burnings of farmlands occur in the farming period.

The sampling equipment was placed about 2.0 m above the ground level. The sampling was conducted using a Gent stacked filter unit (SFU) system that collects two size fractions. The sampler operates at a flow rate of 16–17 L/min and has been described in detail by Hopke et al. (1997). Two samplers were co-located about 1.5 m from each other. Coated 8-μm-pore nuclepore polycarbonate filters were used for the collection of coarse particles in both samplers. The fine particulate matter samples were collected on 0.4-μm-pore Nuclepore polycarbonate and quartz fiber (Pall Tissuquartz) filters in the second sampler. The quartz fiber filters were prefired for at least 3 hr to get rid of background carbonaceous compounds before the sampling (Chow et al., 1993; Yang et al., 2005). The sampling interval covered 24 hr.

Analysis

The sample filters were conditioned in desiccators for at least 24 hr and each weighed before and after sampling to obtain the net weight (mass) of the collected sample. The Nuclepore polycarbonate filters were used for the elemental analysis by employing an energy-dispersive x-ray fluorescence (EDXRF) technique using the Spectro X-Lab 20000 EDXRF system. The following elements were identified; Na, Mg, Al, Si, P, S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Ga, Ge, Se, Zn, As, Br, Rb, Sr, Mo, Ag, Cd, Sn, W, and Pb.

The quartz filters were used for the analysis of the carbonaceous species (i.e., organic carbon [OC] and elemental carbon [EC]) using the Interagency Monitoring of Protected Visual Environments/Thermal Optical Reflectance (IMPROVE/TOR) method for eight carbon fractions (Chow et al., 1993, 2004, 2008). Organic carbon (OC) fractions were volatilized by four temperature steps (OC1 at 120°C, OC2 at 250°C, OC3 at 450°C, and OC4 at 550°C) in a helium environment. Pyrolyzed OC (OP) was oxidized at 550°C in a mixture of 2% oxygen and 98% helium environment until the original intensity of the reflectance is attained. This is followed by elemental carbon (EC) fractions measured in the oxidizing environment (EC1 at 550°C, EC2 at 700°C, and EC3 at 850°C). OP was subtracted from EC1 and utilized as an independent variable in this study since the reported EC1 concentrations in IMPROVE/TOR method include OP concentrations. Thus, EC1 in this study did not include OP. These fractions have been found to be useful in separating various motor vehicle sources in the United States (Kim et al., 2004; Begum et al., 2005).

Source apportionment

The fundamental principle of receptor modeling is to identify the source to receptor relationship by mass balance and conservation. Multivariate receptor models are based on the analysis of the correlation between measured concentrations and source emission of chemical species. The assumption is that highly correlated compounds come from the same source and the chemical characteristics of the sources are constant throughout the measurement period. The positive matrix factorization (PMF) proposed by Paatero and Tapper (Paatero and Tapper, 1994; Paatero, 1997 ; Lee et al., 2003 ) is an advanced receptor modeling technique that apportions the measured aerosol concentrations to their sources. The model forces nonnegative factor elements using a weighted least-squares fit (Lee et al., 2003). PMF depends on error estimates for each measured data value, and the approach of Polissar et al. (1998) was used for estimating the concentration values and their associated error estimates including below detection limit values and missing data. Below the method detection limit (MDL) values were replaced by half of the MDL values, and their uncertainties were set at five-sixths of the MDL values. There were no missing values in this data set. In this study, one-third of the MDL was added to the analytical uncertainty for each value above the MDL (Polissar et al., 1998). Uncertainties in the OC and EC data were calculated by error propagation in the volume of air sampled and the variations in the replicate analyses of collected samples.

For the selection of the input variables, the signal-to-noise (S/N) ratios were calculated (Paatero and Hopke, 2003). Variables were defined to be good, weak, or bad depending on S/N ratio values. Species with S/N ratios less than 0.2 were considered bad and therefore rejected. Species that have signal/noise (S/N) ratios (Tables 1 and 2) between 0.2 and 2 were considered weak variables and their estimated uncertainties were increased by a factor of 2 or 3 to reduce their weight in the solution as recommended by Paatero and Hopke (2003). Although species with S/N above 2 could be considered as good, the percentage of data above detection limit was used as a complimentary criterion for the choice of strong variables since S/N is very sensitive to sporadic values much higher than the level of noise (Amato et al., 2009). The species identified as bad and rejected based on their S/N values were EC3, P, Co, Ga, Ge, Se, As, Mo, Ag, Cd, Sn, and W. The identified weak variables OC1 and EC2 were downweighted by a factor of 3.

Table 1. Summary statistics for the species measured at Navrongo (μg/m³)

Species^a	Arithmetic mean^b	Standard deviation^b	Maximum^b	Minimum^b	Detection limit	Number of BDL Values (%)	S/N ratio
PM2.5 (μg/m^3)	32.4	14.96	88.01	7.14	1.30	0.00	32.70
OC1 (μg/m^3)	1.32	1.05	6.70	0.18	0.42	0.00	0.80
OC2 (μg/m^3)	4.79	2.48	19.49	0.57	0.42	0.00	8.00
OC3 (μg/m^3)	3.04	1.47	10.76	0.42	0.42	0.00	11.00
OC4 (μg/m^3)	1.11	0.63	3.27	0.17	0.42	0.90	2.10
EC1 (μg/m^3)	0.56	1.39	14.14	0.02	0.06	1.80	26.00
EC2 (μg/m^3)	0.46	0.32	2.20	0.00	0.06	2.70	0.90
OP (μg/m^3)	1.11	0.80	5.05	0.06	0.06	0.90	2.20
Na (μg/m^3)	0.42	0.22	1.01	0.10	0.15	0.00	6.70
Mg (μg/m^3)	0.17	0.14	0.61	0.005	0.008	0.00	13.10
Al (μg/m^3)	1.46	1.10	5.04	0.17	0.09	0.00	121.50
Si (μg/m^3)	3.03	2.87	1.13	0.1	0.02	0.00	153.20
S	201.0	75.0	411.0	25.0	53.8	0.00	130.40
Cl	706.0	1200.0	5410.0	18.0	29.7	0.00	94.30
K	508.0	317.0	1420.0	10.0	22.4	0.00	77.10
Ca	602.0	578.0	2500.0	55.0	14.7	0.00	89.50
Ti	133.0	105.0	489.0	9.0	3.7	0.00	45.60
V	2.0	2.0	9.0	0.0	0.8	0.00	2.70
Cr	19.0	6.0	61.0	6.0	6.0	0.00	23.90
Mn	44.0	26.0	145.0	13.0	1.4	0.00	39.90
Fe	1570	876	4510	501	52.4	0.00	72.50
Ni	57.0	18.0	185.0	21.0	14.1	0.00	17.80
Cu	189.0	61.0	615.0	66.0	6.3	0.00	34.40
Zn	110.0	37.0	344.0	37.0	24.6	0.00	43.10
Br	26.0	9.0	84.0	7.0	3.7	0.00	21.70
Rb	3.0	2.0	9.0	1.0	0.6	5.40	1.20
Sr	14.0	7.0	41.0	1.0	1.5	0.90	3.40
Pb	31.0	6.0	98.0	11.0	8.9	0.00	13.70
Notes: ^aAll values in ng/m^3 except as noted for specific species.
^bBlank values were deducted from the initial concentration values.

Table 2. PM_2.5 mass concentration values and the composition of the identified species at Navrongo compared with that of other studies (ng/m³)

Species	This study	Ashaiman (Accra, Ghana)^a	Kwabenya (Accra, Ghana)^b	Al Zamalek (Egypt)^c
PM2.5 (μg/m^3)	32.4	21.56	40.8	61.9
OC (μg/m^3)	11.36	6.31
EC (μg/m^3)	0.95	2.04
Na (μg/m^3)	0.42	0.74	—	—
Mg (μg/m^3)	0.17	0.09	—	—
Al (μg/m^3)	1.46	0.81	0.83	0.12
Si (μg/m^3)	3.03	1.16	0.17	0.30
S	201	391.3	462	—
Cl	706	144.5	123	—
K	508	486.8	377	500
Ca	602	287.2	144	200
Ti	133	59.0	37.5	—
V	2.0	2.9	1.9	—
Cr	19.0	17.3	2.7	—
Mn	44.0	27.4	7.6	—
Fe	1570	987.3	289	200
Ni	57.0	54.4	3.2	—
Cu	189	180.1	4.1	—
Zn	110	164.4	6.5	—
Br	26.0	32.4	5.9	—
Rb	3.0	2.1	1.4	—
Sr	14.0	7.2	2.9	—
Pb	31.0	43.9	2.5	700
Notes: ^aOfosu et al. (2012),
^bAboh et al. (2009),
^c Abu-Allaban et al. (2002).

Results and Discussion

Observed concentrations

The average PM_2.5 mass concentration and that of the identified species can be seen in Table 1. The average PM_2.5 obtained for the sampling period was 32.4 μg/m³, which is higher than the World Health Organization (WHO) annual mean value of 10 μg/m³ and 24-hr mean of 25 μg/m³. The results of the species concentrations in this work compare well with other works in Ghana and Egypt (Table 2) (Ofosu et al., 2012; Aboh et al., 2009; Abu-Allaban et al., 2002).

The mean monthly PM_2.5 mass concentration (Figure 2a) and the corresponding rainfall levels (cm) (Figure 2b) show that the months with high particulate matter concentrations had relatively lower rainfall levels. November to January, which falls within the Harmattan period, had no rainfall.

The mean, maximum, and minimum of the ratios of carbon fractions (EC and OC) to total carbon (TC) and particulate mass are shown in Table 3. On average, OC and EC components in TC were found to be 92.3% and 7.4%, respectively. The average amount of TC in PM_2.5 particulate mass was also found to be 41.2%. The mass concentration values for PM_2.5 and carbon compounds for the Harmattan and non-Harmattan periods are shown in Table 4. The total mean carbon concentration increased by a factor of 1.70 during the Harmattan period.

Table 3. Ratio of carbon fractions to total carbon and PM_2.5 particulate mass

Species fractions (%)	Mean	Min	Max
EC/OC	8.6	1.5	26.9
OC/TC	92.3	22.6	99.8
EC/TC	7.4	1.5	36.8
OC/Mass	37.9	4.9	87.8
EC/Mass	3.9	1.1	13.4
TC/Mass	41.2	4.9	89.3

Table 4. Concentration values of PM_2.5 and carbon species (μg/m³) for Harmattan and non-Harmattan periods

	Harmattan (H)	Non-Harmattan (n-H)	Ratio (H/n-H)
Mass	58.36	34.18	1.71
OC1	2.03	0.90	2.25
OC2	6.15	3.98	1.55
OC3	3.91	2.53	1.55
OC4	1.40	0.93	1.50
EC1	1.10	0.25	4.49
EC2	0.65	0.34	1.93
OP	1.53	0.85	1.80
OC	14.76	9.16	1.61
EC	1.73	0.57	3.06
TC	16.48	9.71	1.70

The correlations between the five major elements of soil (Al, Si, Ca, Ti, and Fe) seen in Table 5 were very high; Si-Al (r ² = 0.98), Si-Ca (r ² = 0.98), Si-Ti (r ² = 0.96), and Si-Fe (r ² = 0.94). Potassium is seen to be correlated with the other crustal elements Ca, Ti, Mn, and Fe such that there is a significant contribution of potassium from soil. However, the K correlations are lower than for the other elements, indicating the presence of another source. Sulfur and potassium (S-K, r ² = 0.58) are moderately correlated, which can be typical of diesel vehicles and other combustion sources. Sulfur occurs as secondary sulfate generally originating from SO₂ emissions from combusting sources. However, diesel vehicles burning high-sulfur (0.25% to 0.30% S) fuels such as those typically found in developing countries do emit primary sulfate. A clear-cut correlation of sulfur with the key elements could not be expected, as they may involve several different sources. A strong correlation could not be seen between K-EC and K-OC because potassium could be coming from different sources, including soil in a typical savannah area.

	Al	Si	S	K	Ca	Ti	Mn	Fe	Cu	Zn	Pb	OC	EC
Al	1.00
Si	0.98	1.00
S	0.41	0.36	1.00
K	0.86	0.86	0.58	1.00
Ca	0.95	0.98	0.36	0.87	1.00
Ti	0.99	0.96	0.39	0.84	0.94	1.00
Mn	0.95	0.93	0.42	0.89	0.93	0.96	1.00
Fe	0.97	0.94	0.44	0.88	0.94	0.97	0.99	1.00
Cu	0.29	0.21	0.37	0.37	0.25	0.26	0.46	0.45	1.00
Zn	0.36	0.27	0.42	0.42	0.31	0.32	0.50	0.50	0.97	1.00
Pb	0.29	0.20	0.39	0.37	0.23	0.27	0.45	0.44	0.91	0.88	1.00
OC	0.20	0.19	0.34	0.43	0.23	0.18	0.37	0.33	0.67	0.65	0.61	1.00
EC	0.02	0.05	0.20	0.32	0.09	0.01	0.18	0.13	0.42	0.39	0.37	0.86	1.00

Source apportionment

Soil.

Since mobilization of Saharan dust is expected to be a major constituent of the ambient PM at Navrongo, a separate analysis of soil has been made. Windblown soil is composed mainly of the oxides of Mg, Al, Si, Ca, Ti, and Fe with many other trace elements. The summation of these five major oxides account form more than 85% of the total soil composition (Weast and Astle, 1982; Begum et al., 2011; Lee et al., 2001). The soil concentration (Malm et al., 1994) can be estimated as:

1

The square brackets signify concentration values.

This equation assumes that the two common oxides of iron Fe₂O₃ and FeO occur in equal proportions. The factor of 2.42 for iron also includes the estimate for K₂O in soil through the (K/Fe) = 0.6 ratio for sedimentary soils (Begum et al., 2011). The mean mass of soil as determined from the preceding equation was 15.9 μg/m³ and this represented 46.2% of the mean PM_2.5 mass of 32.4 μg/m³.

PMF analysis.

PMF was implemented using the U.S. Environmental Protection Agency (EPA) PMF version 3. The parameter FPEAK was used to control rotations. By setting a nonzero of FPEAK, the EPA PMF yielded more physically realistic solutions. A six-source model provided the most physically reasonable source profiles in a variety of source number solutions and FPEAK values and also on examination of the scaled residuals and the Q value. The values of the scaled residuals were symmetrically distributed and with the exception of a few cases were within ±3 for all the species (Amato et al., 2009).

The Q values were plotted against the FPEAK values to explore the rotational space where only the small changes in the Q values are observed. In this study, FPEAK values between –1 and 1 were examined and an FPEAK value of 0.1 was selected. The reconstructed PM_2.5 mass concentrations were estimated by the sum of the contributions from PMF resolved sources. The comparisons between the predicted and measured PM_2.5 mass concentrations show that the resolved sources effectively reproduced the measured values (R ² = 0.92).

Figures 3 and 4 show the source profiles obtained for the six-factor PMF solution and the time series of contributions from each source respectively. The six factors (Table 6) were identified as two-stroke engines, gasoline emissions, soil dust, diesel emissions, biomass, and resuspended dust.

Table 6. Average source contributions to PM_2.5 mass concentrations at Navrongo

Source		Mean source contribution (μg/m^3)	Standard deviation (μg/m^3)	Fractional source contribution (%)
1	Two-stroke engines	3.4	2.4	9.9%
2	Gasoline emissions	3.7	2.8	10.9%
3	Soil dust	12.3	6.2	35.9%
4	Diesel emissions	3.9	2.5	11.5%
5	Biomass	5.4	4.4	15.8%
6	Resuspended road dust	5.5	5.4	16.0%

Figure 3. Source profiles deduced from the PMF analysis of data for Navrongo, Ghana.

Figure 4. Source contributions deduced from the PMF analysis of data for Navrongo, Ghana.

The first source has relatively high levels of OC2 OC3, Ca, Fe, and Zn with traces of Cu, Ni, Mn, Br, and Pb. Exhaust emissions from two-stroke engines are well known for high levels of emitted Zn and some amount of Ca. Zn and Ca are additives in motor oil that is mixed with the gasoline in two–stroke engines (Ofosu et al, 2012). This source can, therefore, be attributed to two-stroke engines. The source did not show distinct periodic variations in contribution as shown in the time-series plot of Figure 4.

The second profile represents gasoline vehicular emission sources. It consists mainly of OC2, OC3, OC4, OP K, Ca, and Fe, with Mn, Ni, Cu, Zn, Br, and Pb present in traces. The sulfur content in this source is seen to be relatively low as compared to that of diesel. The manganese level obtained in this source is also considerably low. This confirms the fact that the manganese levels in air are within the natural background levels since the introduction of methylcyclopentadienyl manganese tricarbonyl (MMT) as an additive in gasoline in Ghana in 2004 (Ofosu et al, 2012). This source did not show significant variations in mass contribution throughout the sampling period (Figure 4).

The major species contributing to the third source include Na, Mg, Al, Si, Ti, Fe, Ca, and K. This source component can be assigned to soil dust. The source contribution plot shows high contributions from November 2009 to January 2010. This period falls within the Harmattan season, which comes with high levels of dust (Ofosu et al, 2012).

The fourth profile was identified to be diesel emissions. Diesel emissions can be characterized by OC, EC, and traces of Na, Ca, S, Zn, Cu, and Ni, which are all present in this profile (Kim et al., 2004; Begum et al., 2005). EC is the primary pollutant emitted by diesel combustion (Gray and Cass, 1998). The EC1 and EC2 values in this source are greater than that of gasoline, and this serves as the basis for its identification. This source did not show significant variations in mass contribution throughout the survey period (Figure 4). There were also no significant differences between weekday and weekend contributions in contrast with developed countries where there are typically lower diesel emission contributions on weekend days (e.g., Kim et al., 2004).

Biomass is mostly identified by OC, EC, K, and traces of Na, Cl, Cu, and Br that are seen in the fifth profile. OC, EC, and K serves as the basis for the identification of biomass burning. The source contributions were found to be relatively high from November 2009 to February 2010 as seen in Figure 4. The relatively high contribution of this source found within the observed period is due to the annual occurrence of bush fires during the dry season (November to March). Thus, annual savannah bush fires during the dry season contribute significantly to particulate loading in the ambient air. Again there were no differences in weekend/weekday contributions. In the United States, higher biomass burning contributions are observed on weekends when people burn wood in fireplaces and stoves (Wang et al., 2012).

The sixth source has relatively high levels of Na, Mg, Al, Si, Ti, Fe, Ca, and K similar to that of the third profile but also has significant levels of OC2, OC3, and OC4. This source can be attributed to resuspended road dust.

The average contributions of each source to the PM_2.5 mass concentrations are summarized in Table 6. Soil dust together with resuspended road dust contributed the greatest part of the PM_2.5 mass (51.9%). This compares well with percentage soil dust contribution of 46.2% determined from the equation for soil ( Eq. 1).

Navrongo town lies on the major road linking Ghana and Burkina Faso. The high contribution of vehicular exhaust (diesel and gasoline) emissions can mostly be attributed to the vehicular movements (for goods and passenger) to and from neighboring countries that lies to the north of Ghana. Biomass burning, seen to be the next largest single source contributor after soil dust, is mostly from perennial Savannah bush fires. It can be seen from Figure 5 that the biomass contribution to ambient PM_2.5 particulates level increases considerably in the dry (Harmattan) season when there are bushfires.

Traces of Si, Al, Mg, and Ti present in the gasoline and diesel vehicle emissions profiles may be from associated road dust (Hwang and Hopke, 2007). Significant levels of Ca and Zn can be observed in all the vehicular emissions (diesel, gasoline and two stroke engines). This result can be due to the use of old vehicles with engines that have not been well maintained. The Zn and Ca are additives in motor oil (Alander et al., 2005). Poorly functioning engines, particularly those with leaky piston rings, burn more lubricating oil, releasing Zn and Ca as visible white clouds in many cases. The relatively high levels of Fe in vehicular emissions can partly be due to brake wear and muffler ablation (Ofosu et al., 2012).

Conclusion

This study identified the major sources of the ambient PM_2.5 at the Navrongo in the Sahel Savannah zone in Ghana. The sources were qualitatively identified and the source contributions were quantitatively estimated. The six sources identified were two-stroke engines (9.9%), gasoline emissions (10.9%), soil dust (35.9%), diesel emissions (11.5%), biomass burning (15.8%), and resuspended soil dust (16.0%). Crustal dust contributed the greater part of the ambient air PM values among the six identified sources. It comprised soil dust and resuspended dust (soil dust intermixed with vehicular emissions). The Harmattan dust could have accounted for the prevalence of crustal dust in the ambient air PM. Biomass is seen to be the next largest single source contributor after soil dust and is mostly from home cooking and perennial Savannah bush fires. The Harmattan bush fires therefore contribute significantly to the biomass particulate mass concentrations in ambient air.

Acknowledgment

The authors gratefully acknowledge the technical staff of the Center for Air Resources Engineering and Science at Clarkson University, Potsdam, NY, who contributed to the elemental and carbonaceous analyses. They also acknowledge the authorities of UDS at Navrongo campus for their cooperation during the sampling.