Source Apportionment of Fine and Coarse Particulate Matter in Industrial Areas of Kaduna , Northern Nigeria

This study was conducted to investigate the sources of fine and coarse airborne particulate matter in an urban environment in Nigeria. A total of 278 samples were collected over a twelve-month period from two industrial areas (Kudenda agricultural processing AP and Refinery industries NNPC) in Kaduna, Northern Nigeria for PM2.5 and PM2.5–10 on nuclepore polycarbonate filters using a Gent sampler. Elemental concentrations and black carbon analyses were performed using X-Ray fluorescence (XRF) and optical transmissometry respectively. The annual average concentrations for PM2.5 at each site (Kudenda and NNPC) were 135.7 μg m and 37.2 μg m and for PM2.5–10, concentrations were 269.2 μg m and 97.4 μg m, respectively. These values exceeded the Nigerian Annual National Ambient Air Quality Standard (NAAQS) of 15 μg m for PM2.5 and 60 μg m for PM10. Positive matrix factorization (PMF) was used to identify sources and quantify their contribution to pollutants at the sampling sites in one of the most industrialized cities in Nigeria. Four sources were resolved for both PM2.5 and PM2.5–10 and were identified as: Residual oil 49% (17.16 ± 0.04 μg m), Soil 29% (10.36 ± 0.26 μg m), Continental dust 18% (6.20 ± 0.18 μg m), and Motor vehicles emissions 4% (1.56 ± 0.02 μg m) for PM2.5 while that for PM2.5–10 were Soil 50% (27.37 ± 1.03 μg m), Continental dust 21% (11.55 ± 0.26 μg m), Vehicular emissions 18% (9.87 ± 0.03 μg m), and Petrochemical 11% (6.23 ± 0.02 μg m). About 82% and 79% were attributed to anthropogenic sources for both fine and coarse samples, respectively. Continental dust was associated with northwesterly and northerly regional transport. Residual oil combustion was the predominant fine PM source and was attributed to fuel oil combustion for power generation and process energy within the local industrial areas. Although transported continental dust is an important source, the majority of the airborne PM in this industrial area was the result of local emissions.


INTRODUCTION
Polluted air is a major health hazard with the World Health Organization estimating globally that more than 7 million premature deaths are attributable to the direct impact of outdoor and indoor air pollution (WHO, 2014).The quality of air in the environment especially in developing countries is likely to be responsible for more than half of the health risk posed by pollution.Improvements in pollution monitoring through development of equipment and statistical techniques during the last several decades have steadily enhanced the ability to measure and potentially mitigate the health effects of air pollution in advanced countries (Harrop, 2003;Kiran et al., 2006).
In Nigeria, there has been rapid increase in population in the last 30 years from about 73 million in 1980 to 168.8 million in 2012 with its associated rural to urban migration such that 51% of the total population are estimated to be living in Nigerian cities and towns (UNWPP, 2012).It is projected that by the year 2030, 80% of the total Nigerian population would be living in urban areas (UNFPA, 2008).As is the case in Nigeria as well as many other developing countries, large population concentrations and the rapid growth of urban centers pose serious problems of congestion, unemployment, poverty, violence, and environmental degradation, the latter typified by significant air pollution.Consistent with results from cities around the world, air pollution in major cities in Nigeria is primarily attributed to the emissions from vehicular transport, cottage and large scale industry, energy production and domestic sources (Efe, 2008).In particular, residents in Kaduna, an industrial city in northern Nigeria, are exposed to significant ambient air pollution concentrations due to increased vehicular emissions, substantial petrochemical refining activities within city limits, use of diesel-fueled generators for commercial energy supply for the chemical-based industries some even located in residential areas as well as reliance on smallscale petrol powered generators for domestic electricity supply (Ndoke et al., 2006).
In Nigeria, air pollution has been identified to be especially related to 3 main sources namely; industrial and oil industries (Iyoha 2009;Nwaichi andUzazabona, 2011), traffic (Ndoke et al., 2006;Olowoporoku 2011), and the effects of the north-easterly winds that bring the "Harmattan" to Northern Nigeria (Dimari et al., 2008).There has been remarkable industrial progress in Nigeria over the last three decades with the establishment of many cottage industries.This rapid industrialization has not been matched with proper planning for the control of environmental pollution problems that are usually associated with such industrial development (Adejumo et al., 1994;Owoade et al., 2013).Severe air pollution has been identified in overcrowded cities such as Lagos, Port Harcourt, Kaduna, Kano, and Abuja (Ifeanyichukwu, 2002).Earlier work in Nigeria focused on the levels of pollutants in ambient air environment near industries and later on vehicular emissions in southern Nigeria without much focus on Northern Nigeria (Obiajunwa et al., 2002;Owoade et al., 2009).A detailed investigation of elemental constituents in fine airborne particulate matter (PM 2.5 ) as well as its source apportionment is crucial to achieving effective air pollution control in northern Nigeria.This study focuses on qualitative and quantitative chemical characterization of the major particulate air pollutants in the Kaduna metropolis to provide a quantitative source apportionment.This work contributes to the baseline data on environmental pollution for the northern part of the country especially identification and apportionment of sources of pollution for industrial areas.

Study Area
The study was conducted in Kaduna in northwestern Nigeria.Kaduna with an approximate size of 8000 m 2 and a 2013 population of 1.8 million people is located at an altitude of 635 m (2000 feet) above the sea level and has typical Guinea Savannah vegetation.Maximal annual rainfall of 1000-1500 mm (40-60 inches) is recorded in the months of April to October with a distinct dry season from November to March.During the dry season, Kaduna experiences a cool 'Harmattan' period because of northeasterly winds that are characterized by hazy, dusty conditions and low temperatures.Two sampling sites were strategically located that captured the major industrial areas within Kaduna Metropolis (Fig. 1).One site was located at the Kaduna Refinery and Petrochemical Company (Latitude: 10°27'55.75''Nand Longitude: 7°29'43.64''E), the third largest Nigerian National Petroleum Corporation (NNPC) refinery.Products from the refinery include: liquefied petroleum gas (LPG), gasoline, residual oil, diesel oil, kerosene, lubricating oil, and sulfur.Additional products from the lubricating oil complex are base oils, asphalt (bitumen) and waxes.The second sampling point was in Kudenda (Latitude 10°28'4.20''N,and Longitude 7°24'7.45''E),which is one of the other major industrial areas of Kaduna.Industries in this location include: Flour mills, grains processing industry, IBBI Breweries and Sun Glass producing industry, Unifoam Industry, and WANO gas industry.Unifoam and WANO gas industry are located within a 500 m radius of each other.
Corresponding meteorological variables were measured at both sampling sites using an Oregon Scientific Advanced Weather Station.It comprised of a wind vane for measuring wind direction, cup anemometer for wind speed, barometer for atmospheric pressure and rain gauge for precipitation volume.Fig. 2 provides the wind roses for both sites for the sampling periods.

Sampling
Two (2) low-volume Gent samplers were deployed to the study sites which are 2 km apart for a 12-month period.Kudenda was sampled from March to August 2013 and NNPC was sampled November 2013 to February 2014.Samples were collected over 24 hours, four times a week (on Mondays, Wednesdays, Fridays and Saturdays).The Gent sampler equipped with a stacked filter unit (SFU) (Maenhaut et al., 1994;Hopke et al., 1997) collected two PM fractions (PM 2.5 and PM 2.5-10 ).Air was sampled at an average rate of 16 L min -1 (approx.) at a height of 1.6 meters, which is taken to be the average height of an adult.
The filters were equilibrated at a relative humidity between 32% and 45% and at a temperature between 20 and 25C for at least 24 hours and then weighed using a microbalance (Sartorius model CP2P-F) with a minimum resolution of 0.001 mg.After the sampling period, each filter was re-equilibrated and reweighed to provide measured mass concentrations.A total of 278 samples were analyzed for the two PM fractions: 102 each of fine and coarse filters at Kudenda and 37 each of fine and coarse at NNPC.

Chemical Analysis
Black carbon measurements were performed using a Magee Scientific Optical Transmissometer (OT21).The transmission intensity of light of 880 and 370 nm wavelengths was used to determine the light absorption.The manufacturer's recommended absorption coefficients, 16.6 m 2 g -1 and 39.5 m 2 g -1 , were used for 880-nm and 370nm channels, respectively.The black carbon concentration was estimated for each filter in µg m -3 .The loaded filters were analyzed for 20 elements using Energy Dispersive X-Ray Fluorescence analytical technique (Spectro X-Lab Pro) (Sunder Raman et al., 2008).The measured elements were: Na, Mg, Al, Si, S Cl, K, Ca, Ti, V, Mn, Fe, Ni, Cu, Zn, As, Br, Rb, Sr and Pb.

PMF
Source apportionment analysis was performed using  EPA PMF v5.The model provides a flexible modeling approach that effectively uses the information in the data.
In PMF, the data matrix X of dimension n rows and m columns, where n and m are the number of samples and species, respectively, can be factored into matrices; G (n × p) and F (p × m), where p represents the number of factors extracted (Paatero et al., 2014).The solution to the PMF problem depends on estimating uncertainties for each of the data values used in the PMF analysis.The approach of Polissar et al. (1998) has been used to estimate the concentration values and their associated error estimates.
Because of small number of samples collected at the NNPC site and prior work by Kara et al. (2015) showing the value of multiple site data.Given an area with significant local sources, sampling from multiple locations such that a given source cannot easily impact both sites simultaneously provides edge points (Henry, 2003) and thus, additional resolution in the analysis.

Trajectory Analysis
To investigate the history of air masses arriving at the sampling site, 7-day backward trajectories analysis were calculated using the UK Universities Global Atmospheric Modeling Programme (UGAMP) trajectory model (Methven, 1997).The model is driven by the 6-hourly European Centre for Medium-Range Weather Forecast (ECMWF) wind analysis data.Twenty-five (25) trajectories are released at 12:00 UTC from the sampling site at a pressure of 900 hPa on each sampling day with a time-step of 0.6 hours.

Conditional Probability Function (CPF)
The conditional probability function model (Ashbaugh et al., 1985;Kim et al., 2003) is a method to estimate the direction of local sources.Local meteorological data were used for the conditional probability function calculations in order to determine the local sources.The wind direction range was divided into 32 sectors of 11.25° per sector.The CPF for each sector is defined as: where mc i is the number of occurrences of fractional contribution exceeding a threshold (e.g., those ranking top u p in the total time series) in direction sector, i, and nc i is total number of wind direction occurrences in this sector.The fractional source contribution was used instead of the original source contribution to avoid the influence of atmospheric dilution.In order to ensure statistical reliability, a threshold for the minimal number of wind direction occurrences in each sector, n t, is needed.If nc i < n t, the corresponding CPF i was set to zero.Winds with the speeds below a threshold, w t of 1 m s -1 were excluded from analysis.
The sources are likely to be located in the directional sectors with high CPF values.It should be noted that CPF is not suitable for distant sources since air particles may travel through circuitous pathways.In this study, the key parameters k, u p , n t , and w t , were set to 32, 25%, 15 and 1.0 m s -1 .

RESULTS AND DISCUSSION
The annual average concentrations for PM 2.5 for the Kudenda and NNPC sites were 135.7 ± 4.5 and 37.2 ± 1.7 µg m -3 , respectively.These values exceeded the Nigerian Annual National Ambient Air Quality Standard (NAAQS) of 15 µg m -3 .PM 2.5-10 mean values were 269.2 ± 6.8 and 97.4 ± 2.4 also exceeded Nigerian standard (NAAQS) of 60 µg m -3 .The average values of PM 2.5 /PM 2.5-10 ratio for the two sites were 0.50 ± 0.16, and 0.38 ± 0.20, respectively, indicating an almost even distribution of PM between these fractions as shown in Table 1.Among the quantified species, Na, Al, Si, K, Ca, and Fe were found to be most abundant and are mostly crustal species.Emissions from crustal sources are mostly in the PM 2.5-10 , where 10-30% of the PM 10 mass were crustal species.Vehicular emissions were likely to be the principal source for Pb, As, Br, and Zn in the ambient air near the major road (Pacyna, 1998;Singh and Jaques, 2002;Sternbeck et al., 2002;Adachi and Tainosho, 2004).

Fine and Coarse PM Apportionments
PMF was run multiple times for the data for a specific size fraction combined from both sites and from the three sampling periods.To obtain optimal number of sources, 4-10 factors were tested.The Q values, the resulting source profiles, and the scaled residuals were studied.After a thorough evaluation and interpretation of each model run, the optimum numbers of factors were chosen based on the most physically reasonable result and adequate fit of the model to original data for both PM 2.5 and PM 2.5-10 .The optimal factor number that was chosen was four (4) for both PM 2.5 and PM 2.5-10 samples.The FPEAK parameter was varied from -0.8 to 0.4 to refine the source profiles (Santoso et al., 2011).The optimum solution was chosen to be that with FPEAK = -0.4based on G-space edges showing no correlations among the resolved sources.Figs. 3 and 4 present the identified source profiles for fine and coarse fraction for the two industrial sampling sites in Kaduna while Figs. 5 and 6 present the time series plots of the estimated daily contributions from each source.Conditional probability function (CPF) was used to identify the local sources as shown in Figs.7 and 8.

Identified Sources and Source Profiles
Residual oil was identified only in PM 2.5 and is characterized by high loadings of Na, K, V, Ni, Cu, Zn, As, Br, Pb and BC as shown in Fig. 4. Key species suggested for this factor includes Ni, and V (Schauer et al., 1996).This factor contributed 49% (17.16 ± 0.04 µg m -3 ) of the PM 2.5 mass concentration as shown in Table 2.This factor contained larger amount of V and Ni representing the influence of the residual oil combustion source (Sun et al., 2007).Querol et al. (2009) reported that concentrations of V are elevated in the Mediterranean region owing to increased consumption of fuel oil for power generation and industrial emissions.This source was expected in the study area because of the constant use of heavy duty power generators since most of the time when the sampling was taking place there was no grid supply of electricity.In the study area, there is refinery which burn the residual oil to provide process heat needed to refine the oil.
The second factor is local soil with source profile dominated by high loading of Na, Mg, Al, Si, K, Ca, Ti, Mn, Fe, Rb and Sr as shown in Figs. 3 and 4 for both the PM 2.5 and PM 2.5-10 samples (Begum et al., 2010;Stone et al., 2010).The presence of K in both factors indicated that the local soil is strongly contaminated by the biological materials and may also be incorporating the contribution from biomass burning.The factors contributed 29% (10.36 ± 0.26 µg m -3 ) and 50 % (27.37± 1.03 µg m -3 ) of the total mass of fine and coarse samples, respectively (Table 2).
The third identified source is motor vehicles/vehicular emissions for both PM 2.5 and PM 2.5-10 .These profiles were enriched in Na, S, Cl, V, Zn, Cu, As, Br, Rb, and Pb with minimal concentration of BC as shown in Figs. 3 and 4, respectively.The high S in this profile might be because of high S Nigeria diesel fuel content (1330 ppm) similar to what has been observed in Indonesia (Santoso et al., 2008).
Elemental markers of motor vehicle emissions include Cu, Zn, Pb, Ni, Mn, and BC (Almeida et al., 2006;Begum et al., 2010;Stone et al., 2010;Chow et al., 2004).The motor vehicles sources were classified further into gasoline and diesel sources that were resolved in fine PM while vehicular emissions including two-stroke engines were resolved in the coarse PM.Since heavy trucks are involved in the transportation of agricultural products within the sampling site, vehicular emissions were expected to be the major contributor in PM mass.The source contribution of vehicular emissions in this study was found to be 4% (1.56 ± 0.02 µg m -3 ) in PM 2.5 and 18 % (9.87 ± 0.03 µg m -3 ) in PM 2.5-10 mass (Table 2).
The fourth factor which is continental dust mainly consists of very high percentage of Na, Mg, Al, Si, K, Ca, Ti, Mn, Fe, Cu, Zn, Br, Sr as shown in Figs. 3 and 4 for the PM 2.5 and PM 2.5-10 samples, respectively.The major elements used as tracers for continental dust include Si, Al, Fe, Ca, and K (Han et al., 2006;He et al., 2001).
Table 2 shows that the factors contributed 18% (6.20 ± 0.02 µg m -3 ) and 21 % (11.55± 0.03 µg m -3 ) to the mass concentrations of both fine and coarse PM samples respectively.This impact of transported dust is because Kaduna is a northern state in Nigeria and there are strong seasonal northeasterly winds from the Sahara Desert.A high contribution to the coarse fraction is the result of the strong winds that aided long-range transport from the Bodele depression of the Chad Basin, far from the sampling sites in Kaduna, Nigeria.
Figs. 9(a)-9(i) show 7-day backward trajectory plots for periods between March 4 to April 3, 2013 with color coding      of the atmospheric pressure as the air mass travel from the source to the sampling location.Plots g and h from the 21 st and 30 th of March, 2013, respectively, shows that the dust was contributed predominantly by air masses from the South and South-East over the Atlantic.PM samples during this period showed elevated concentrations of maritime aerosol as indicated by the Na and Cl concentrations.However, there was not a sufficient impact to allow a separate marine aerosol factor to be resolved.
Figs. 9(a), 9(b), 9(c), 9(d), 9(e), 9(f), and i trajectories of air masses during the Harmattan period, show air masses from North Africa and the Sahel region as major contributors to the PM sampled during period.These results explain the very high mass concentrations in both the fine and coarse fractions measured during the peak of Harmattan periods.The trajectory analysis shows that the air masses from the source of the aerosol to the sampling location during the Harmattan periods are between 650 hPa and 900 hPa.The last factor, the petrochemical industry, was only found in fine samples.Its profile was characterized by high loadings of Na, Cl, V, Cu, Zn, As, Br, Pb, and BC as shown in Fig. 4.This source is characterized by the typical residual oil tracers V, Ni, and Sn (Pandolfi et al., 2011).The source contribution was very low 11 % (6.23 ± 0.02 µg m -3 ) to the PM 2.5-10 .At the time of the PM sampling, the production activity was very low and only operating at about 20% of their production capacity.In addition to this facility, many heavy industries are located about 5 km away in the north-north-east direction from the monitoring site and also use residual oil as a source of process heat and locally generated electricity.

Seasonal Variations of Source Contributions in PM 2.5 and PM 2.5-10
Seasonal variations of source contributions to the PM 2.5 and PM 2.5-10 samples are shown in Figs. 5 and 6.The sampling was not conducted simultaneously because at NNPC, we were not allowed to start sampling until October, 2013.At Kudenda where here is an agricultural processing industry, it was observed that there were reasonably high peak contributions of residual oil combustion during the sampling periods when compared with that at NNPC sampling site.This result was expected for PM 2.5 because of the extensive use of heavy duty power generators.Most of the time during the sampling period, there was no grid supply of electricity.The contribution of soil was also high in both the PM 2.5 and PM 2.5-10 samples from March 2013 at Kudenda.The high contributions might be a result of the unpaved road that led to the industrial area and many heavy duty vehicles travelling through the area.In contrast, the NNPC site was an area where all of the roads were paved.Vehicular emissions and motor vehicles showed higher peaks at NNPC for PM 2.5-10 than PM 2.5 samples, which might be as a result of many petrol tankers loading fuel from the refinery.The CPF plots in Figs.7 and 8 indicate that this source was affected more by the northerly wind for both fine and coarse filters this might be so because most of the industries are located towards northern part of Kaduna city.Also, for vehicular emissions, there are more contributions from Kudenda than from NNPC as shown in Figs. 5 and 6.This difference is because of seasonal variations given that sampling was performed at Kudenda mostly during the dry season while that at NNPC was during the rainy season.For the soil source, Figs. 5 and 6 show that the high contribution was mainly in the dry season which peak in early March, 2013 for both PM 2.5 and PM 2.5-10 samples at Kudenda.These results are in agreement with the CPF results showing the northerly wind (Figs. 7 and 8).The seasonal variation of the source contribution for continental dust for both PM 2.5 and PM 2.5-10 samples also shows similar trend as the soil source that was mainly during the dry season at Kudenda.This effect is due to strong winds that aid long-range transport from the Bodele depression of the Chad Basin (Figs. 5 and 6).
Finally, for petrochemical industry, there was no definite pattern in the seasonal variation of the contribution because it depends on the production schedule.Its contribution was high when production increases and during the sampling period in 2013, the production level was very low.

CONCLUSIONS
The annual average concentrations for PM 2.5 at each site (Kudenda and NNPC) were 135.7 µg m -3 and 37.2 µg m -3 and for PM 2.5-10 were 269.2 µg m -3 and 97.4 µg m -3 , respectively.These values exceeded the annual National Ambient Air Quality Standard (NAAQS) of 15 µg m -3 for PM 2.5 and 60 µg m -3 for PM 2.5-10 .Source apportionment of these data was performed using USEPA PMF v 5. Four sources were identified for both PM 2.5 and PM 2.5-10 samples, respectively.Residual oil combustion was the predominant fine PM source and was attributed to fuel oil from power generation within the local industrial areas.Transported Harmattan dust featured prominently in the coarse PM samples.This study helps stakeholders and policy makers to understand the influences of regional and local sources on PM 2.5 and PM 2.5-10 in this urban area.

Fig. 1 .
Fig. 1.Map of Kaduna state with sampling location and wind rose in Nigeria.

Fig. 6 .
Fig.6.PMF-derived source contributions for the PM 2.5-10 data for the two sampling sites in Kaduna.

Fig. 7 .
Fig. 7. CPF plots for source contribution resolved by PMF for PM 2.5 samples from industrial sites in Kaduna (A = Residual Oil, B = Soil, C = Vehicular Emissions).

Fig. 8 .
Fig. 8. CPF plots for source contribution resolved by PMF for PM 2.5-10 samples from industrial sites in Kaduna (A = Soil, B = Vehicular emissions C = Petrochemical).

Table 1 .
Mean ± standard deviation of particulate matter (PM 2.5 and PM 10 ) (concentration in µg m -3 ) with the PM ratio from two industrial sites.

Table 2 .
Mean source Contribution derived from the PMF modeling for PM 2.5 and PM 2.5-10 .