Characteristics of Organic Carbon and Elemental Carbon in Atmospheric Aerosols in the Urban Area in Beibei , a Suburb of Chongqing

To investigate the pollutive characteristics of atmospheric carbonaceous aerosols in Beibei District, a suburb of Chongqing Municipal City, graded aerosol samples were continuously collected by cascade impactors over one year, from March 2014 to February 2015. Carbonaceous aerosols, including organic carbon (OC) and elemental carbon (EC), were detected by a DRI 2001A carbon analyzer. The results showed that the average annual concentrations of OC and EC in the PM2.1 were 16.3 ± 7.6 and 1.8 ± 0.7 μg m, and in the PM9.0 were 25.0 ± 9.6 and 3.2 ± 1.2 μg m, respectively. On the one hand, a more seasonal distribution of OC concentrations appeared in the PM2.1 (winter > spring > summer > autumn) than in the PM9.0 (winter ≈ spring ≈ summer > autumn); furthermore, whereas the OC significantly positively correlated with wind speed for both the PM2.1 (P < 0.05) and the PM9.0 (P < 0.01), it was negatively correlated with relative humidity (P < 0.05) for the latter. On the other hand, both the PM2.1 and the PM9.0 exhibited a more seasonal distribution of EC concentrations (winter > spring > summer > autumn), and the EC in the PM2.1 significantly negatively correlated with temperature (P < 0.05). Additionally, both the OC and the EC were concentrated mainly in the fine particles (< 2.10 μm), and the size distributions of the OC in all four seasons displayed a bi-modal pattern that peaked in the size ranges of 0.43– 0.65 μm (for fine particles) and 4.7–5.8 μm (for coarse particles), whereas the EC displayed a uni-modal pattern that peaked in the size range of 4.7–5.8 μm (for coarse particles). Furthermore, the correlations between the OC and the EC were analyzed, and the SOC (secondary organic carbon) in the PM2.1 was estimated using the primary OC/EC ratio. It was found that the OC highly significantly correlated with the EC (P < 0.01), with the average annual concentration of the SOC being 6.3 ± 5.9 μg m, which accounted for 33.5 ± 22.6% of the OC. Analyzing the sources of the pollutive atmospheric aerosol in Beibei further demonstrated that it mainly originated in biomass burning, gasoline-vehicle exhaust, and coal combustion.


INTRODUCTION
Carbonaceous aerosols, as an important component of atmospheric aerosols, are typically classified as EC (elemental carbon) and OC (organic carbon).EC, also known as black carbon or soot, including pure carbon and graphite carbon, originates mainly from coal combustion, vehicle emission and wood burning (Liu et al., 2015).In contrast, OC can be directly emitted from sources such as POC (primary organic carbon) and SOC (secondary organic carbon), which are formed from the products of atmospheric chemical reactions through low vapor pressure, proper temperature and sunlight in the atmosphere.Atmospheric aerosols not only have important effects on the extinction of solar radiation, but also affect atmospheric long-wave radiation.In general, EC is a strongly absorbing aerosol, which has significant influence on visibility, accounting for about 20% of the visibility reduction compared with other aerosol particles (Deng et al., 2008), while OC can scatter solar radiation (Kirkevag et al., 2013).As a result, both of them can reduce atmospheric visibility and affect the global climate change by changing the radiation characteristics of cloud and precipitation process (Jacobson et al., 2001).Moreover, EC and OC can also enter human's lungs through respiration and cause chronic respiratory diseases, even trigger lung cancer (Yang et al., 2003;Pope et al., 2006;Mauderly et al., 2008).
In recent years, the analysis of the pollution characteristics of carbonaceous aerosols in atmosphere has become a hot topic around the world (Kumar and Yadav, 2016;Chen et al., 2017a).The distributions of carbon fractions in indoor and outdoor total carbon (TC, OC + EC) pointed to the contributions of motor vehicle exhaust and coal-combustion, and about 90% of carbonaceous particles in indoor air resulted from penetration of outdoor pollutants, while indoor sources contributed only 10% of the indoor carbonaceous particles (Cao et al., 2012).At eight sites in four Chinese cities (Hong Kong, Guangzhou, Shenzhen and Zhuhai) of the Pearl River Delta Region, the OC and EC in PM 2.5 and PM 9.0 were strongly correlated (Cao et al., 2003).The distributions of eight carbon fractions (OC 1 , OC 2 , OC 3 , OC 4 , EC 1 , EC 2 , EC 3 and OPC [optically detected pyrolyzed carbon]) indicated that the biomass burning, coal combustion and motor-vehicle exhaust were all contributed to these carbonaceous aerosols in Tianjin, China (Gu et al., 2010).
Chongqing, known as a "fog city" for the serious air pollution, is an important industrial mega-city in southwest China.At present, a number of studies on the carbonaceous aerosol pollution in the atmosphere of Chongqing had been reported.For example, the seasonal characteristics of OC and EC in PM 2.5 were in the order of winter > fall > spring > summer in the hinterland of Wanzhou, Chongqing, located in the Three Gorges Reservoir region along the Yangtze River (Zhang et al., 2015).The OC and EC concentrations in PM 10 were significantly correlated to each other both in spring and in autumn, Chongqing, implying the existence of similar emission sources such as coal combustion, and the SOC was presented as the OC/EC ratios generally exceeded 2.0 (Ye et al., 2007).However, the pollution sources of carbonaceous aerosols were little reported in those studies.Moreover, the particle size distribution was also rarely reported.Therefore, in this study, carbonaceous aerosol samples were collected by the Anderson sampler inertial impactor over a one-year period from March 2014 to February 2015, and the data measured from these samples were analyzed to reveal the characteristics of seasonal changes in OC and EC and their particle size distribution.In addition, the main pollution sources of OC and EC in the Beibei Suburb were determined by the correlation and principal component analysis, in order to provide a rational basis for the control and management of the aerosol pollution caused by OC and EC.

Site Location and Sampling Method
The sampling site (29°48ʹ43″N, 106°24ʹ58″E) locates at the rooftop (282 m a.s.l.) of the College of Resources and Environment building (35 m above ground) at the Southwest University campus in the Beibei suburb, Chongqing Municipal City, China (Fig. 1).The Beibei suburb is surrounded by the Zhongliang Mountain (1,000 m a.s.l. in the peak) in the south and the Jinyun Mountain (951 m a.s.l. in the peak) in the north.The local air circulation in the sampling site is little affected by nearby educational and residential buildings with some small ups and downs.The site, representing a typical urban environment in the Beibei suburb of the Chongqing City, is about 30 m to a main city traffic road and about 800 m to the south of the highway intersection and has no direct industrial pollution sources within a 5-km distance.

Carbonaceous Aerosol Analysis
The samples were analyzed for OC and EC using DRI Model 2001 Thermal/Optical Carbon Analyzer.And the IMPROVE thermal/optical reflectance (TOR) protocol (Chow et al., 1993(Chow et al., , 2001(Chow et al., , 2004b) ) was used for the carbon analysis.In a non-oxidizing helium (He) atmosphere, a 0.5025 cm 2 punch aliquot of a sample film was stepwise heated from 140°C (OC 1 ), 280°C (OC 2 ), 480°C (OC 3 ), through 580°C (OC 4 ).The ramp-up to the next temperature or atmosphere began when the flame ionization detector (FID) response returned to a baseline or a constant value under the condition that the time spent in any segment (OC 1 , OC 2 , etc.).When this condition had been reached in the OC 4 segment, the 2% O 2 /98% He was introduced and peaks were integrated at 580°C (EC 1 ), 740°C (EC 2 ), and 840°C (EC 3 ), respectively.The carbon that evaporated at each temperature was catalyzed into carbon dioxide (CO 2 ) by manganese dioxide, then reduced to methane (CH 4 ) for quantification with a flame ionization detector.As the temperature increased in the inert helium, some of the organic carbon was pyrolyzed into black carbon, which made the carbon and EC peaks difficult to be distinguished.Therefore, the 633-nm He-Ne laser was used to monitor the light intensity of the filter film during the measurements.The changes in the light intensity clearly indicated the starting point of oxidation to EC.To ensure a distinction between OC and EC, the seven fractions (OC 1 , OC 2 , OC 3 , OC 4 , EC 1 , EC 2 and EC 3 ) were separately reported in the data sheet.In addition, the IMPROVE protocol defined OC as OC 1 + OC 2 + OC 3 + OC 4 and EC as EC 1 + EC 2 + EC 3 , respectively.

Quality Assurance and Quality Control (QA/QC)
All samples were collected on polyester fiber filters (81 mm in diameter) during the study period.The polyester fiber filters were pre-fired (2 h at 800°C) to remove all organic materials and weighed before and after sampling using a micro-balance with a sensitivity of ± 0.01 mg.The filters were conditioned in a dryer at 25 ± 3°C under a relative humidity of 22 ± 3% for 72 h before each weighing.After reweighing, the sampling filters were immediately stored in a freezer at -20°C to avoid the loss of semi-volatiles.The samplers were cleaned using an ultrasonic bath before sampling.In addition, the sampling flow rates were calibrated before each sampling and monitored using a flow meter during each sampling.
Moreover, the analyzer was calibrated with known quantities of CH 4 every day.Replicate analyses were performed at the rate of one per group of ten samples.Sixteen blank filters were also analyzed, and the sample results were corrected by the average of the blank concentrations.The difference determined from the replicate analyses was less than 5% for the total carbon (TC, OC + EC) and 10% each for OC and EC (Cao et al., 2003).

Statistic Analyses
The figures were plotted by the Origin 9.1 software.Statistical analyses (correlation and principal component analyses) were performed with a SPSS 17.0.Correlation between meteorological factors and OC or EC as well as the OC and EC in PM 2.1 were made by the Pearson correlation coefficients to infer the influence of meteorological factors on OC, EC and the origin of carbonaceous aerosols in atmosphere.Moreover, the principal component analysis (PCA) was used to classify and analyze the seven carbonaceous aerosols of OC 1 , OC 2 , OC 3 , OC 4 , EC 1 , EC 2 and EC 3 in the PM 2.1 to explore the main sources of carbonaceous aerosol in the Beibei suburb.

OC and EC as a Function of Particulate Size
The size of particle was the most basic physical characteristic of atmospheric particulates, which can determine the residence time, transmission distance and mechanism of particle removal in the atmosphere (Spracklen et al., 2009).So, the characteristics of OC and EC size distribution were important to the study on the formation and transformation of atmospheric aerosols, the characteristics of dry and wet deposition, and the global effects of aerosols (Mcmurry et al., 1989).
During the observation period, about 61.5% of OC and 50% of EC were concentrated in fine particles (Fig. 2).And OC has the highest concentration in the size of 0.43-0.65 µm (4.8 ± 2.2 µg m -3 ) and 0.65-0.1 µm (4.9 ± 2.7 µg m -3 ), which accounted for 18.0% and 18.4% of the TOC (the total of OC in all particle sizes) concentration, respectively.There was the highest concentration of EC in the size of < 0.43 µm (0.7 ± 0.4 µg m -3 ), which accounted for 20.1% of TEC (the total of EC in all particle sizes) concentration.
In order to illustrate the particle size distribution of carbonaceous aerosol in Beibei suburb, logarithmic model was taken to analyze the particle size distribution of OC and EC, and using the concept of differentiation, defined as: In the formula, dM represents the mass concentration of particles in the range of lgD p to lgD p + dlgD p ; M(lgD p ) represents the distribution function of mass concentration; D p is the average value between the particle size segments, then taking dM/dlgD P as the horizontal axis and D P as the vertical axis to obtain the OC and EC distribution in different size particles.
The size distributions of OC in four seasons all showed a "bi-modal" according to the Fig. 3, with peaks in the range size of 0.43-0.65 µm in fine particle, and 4.7-5.8µm in coarse particle.While EC was more complicated, EC showed a "uni-modal" with peak in the range size of 4.7-5.8µm in coarse particle, and in fine particle, EC was messy, with no obvious peak characteristics in four seasons.

Estimation of Secondary Organic Carbon
It has been recognized for decades that SOC related to haze, visibility, climate and heath (Griffin et al., 2002).The ratio of OC/EC was used to identify the emissions and transformation characteristics of carbon particles, and SOC existed when the ratio exceeded 2.0 (Turpin et al., 1995;Chow et al., 1996).Moreover, the larger the ratio of OC/EC was, the more serious the SOC pollution was, and this approach had been adopted by numerous studies (Gu et al., 2010;Li et al., 2015).
During the whole observation period, the average ratio of OC/EC (9.3 ± 3.7) was greater than 2.0, and the highest OC/EC ratio (11.4 ± 5.4) appeared in summer while the lowest (8.0 ± 2.2) in autumn (Table 2), indicating the existence of SOC in four seasons.
Although there was no simple direct analytical technique available, only several indirect methodologies have been applied in the evaluation of SOC formation in ambient aerosols (Turpin and Huntzicker, 1991;Pandis et al., 1992;Turpin and Huntzicker, 1995).According to Turpin et al. (1990), the production of SOC in PM 2.1 can be calculated from the following equation: where OC tot was the OC in PM 2.1 , (OC/EC) pri was the ratio Fig. 2. Averaged concentrations of organic carbon (OC) and elemental carbon (EC) in each particle size, and percentages of OC and EC in each particle size to TOC (the total OC in all particle sizes) and TEC (the total EC in all particle sizes) in Beibei District, a suburb of Chongqing Municipal City, Southwest China.7.9 ± 7.2 32.9 ± 18.2 8.6 ± 2.0 Annual 6.3 ± 5.9 33.5 ± 22.6 9.3 ± 3.7 Data are means ± SD (spring: n = 5; summer: n = 5; autumn: n = 5; winter: n = 4).
of OC and EC in PM 2.1 , produced during the primary emission process.However, the value of (OC/EC) pri related to the emission of each pollution source, which was not easily to be calculated.Turpin and Huntzicker (1991) found that the lowest ratio of OC/EC could approximately equal to (OC/EC) pri in a certain period.Thus, Eq. ( 2) can be rewritten as: The ratio of OC/EC was affected by meteorology of the region, diurnal and seasonal fluctuations in emissions and aerosol transportation (Ali et al., 2016), the average of the three minimum OC/EC ratios (5.48) in PM 2.1 during the whole sampling time was selected to replace (OC/EC) pri , and was substituted into Eq.(3) to estimate the concentration of SOC in each sample, then calculate the average concentration of SOC for the whole year and each season.

Temporal OC and EC
The concentrations of OC in both PM 2.1 and PM 9.0 in Beibei were accounted for ~90% of the TC, making it the predominant carbon contributor (Table 1).Also in Beibei, the annual average concentrations of OC and EC in PM 2.1 or PM 9.0 in this study were 16.3 ± 7.6 and 1.8 ± 0.7 µg m -3 or 25.0 ± 9.7 and 3.2 ± 1.3 µg m -3 , respectively (Table 1), while OC and EC in a similar size of PM 10 ten years ago were 54.1 ± 15.6 and 6.2 ± 2.0 µg m -3 (Ye et al., 2006).Similarly, the concentrations of OC and EC in PM 2.5 were 23.6 and 8.7 µg m -3 in 2013 in Wanzhou suburb of Chongqing (Zhang et al., 2015), which might be due to more industrial pollution emissions from the construction of Wanzhou economic and technological development area, where is about 220 km away of east from Beibei.Meanwhile, the average concentrations of OC and EC in PM 2.5 were 19.0 ± 13.3 and 4.6 ± 2.6 µg m -3 between May 2012 and April 2013 in Chengdu, where is about 250 km away of west from Beibei (Chen et al., 2014).These results indicated that the air pollution in Beibei had been obviously alleviated in recent years, though the sizes of those referred aerosols were different with present study (PM 2.1 and PM 9.0 ) and previous studies from Beibei (PM 10 ), Wanzhou (PM 2.5 ) and Chendu (PM 2.5 ).
The concentrations of OC and EC presented different seasonal characteristics because of the variations in climate and pollution sources (Lang et al., 2017).During one year observation period, greater OC concentration in PM 2.1 over the four seasons ranked as winter (20.7 ± 12.0 µg m -3 ) > spring (18.8 ± 8.05 µg m -3 ) > summer (16.0 ± 3.2 µg m -3 ) > autumn (10.5 ± 2.4 µg m -3 ).While Chen et al. (2017b) found that the traditional activity of preserving meat with smoking in southwestern China was a major source of particulate pollution in winter.Therefore, the higher concentration of OC in winter might be due to more biomass being burned to produce smoking for meat preservation, the lower temperature and higher atmospheric stability, which could lead to the difficult proliferation of pollutants, while the more motor-vehicle exhaust under low temperature in winter was probably another reason.For PM 9.0 , the concentration of OC showed the seasonal distribution as summer (29.4 ± 6.0 µg m -3 ) ≈ spring (28.2 ± 9.9 µg m -3 ) ≈ winter (27.0 ± 13.3 µg m -3 ) > autumn (15.8 ± 2.3 µg m -3 ) (Table 1).More rainfall was probably one of the reasons why the concentration of OC in autumn (62.1 mm) was lowest among the four seasons (18.2, 27.6 and 2.9 mm in spring, summer and winter).Compared with the OC concentration in PM 2.1 , the OC concentration in PM 9.0 in spring and summer was rapidly increased, probably due to the increase of coarse-grained biomass aerosol in spring and summer, like pollen (Pavuluri et al., 2013).Whether in PM 2.1 or PM 9.0 , the concentration of EC was the highest in winter, which may be caused by the lower temperature and higher atmospheric stability in winter, and lowest in autumn, which may be caused by more rainfall in autumn.
During one-year observation period, the different seasonal characteristic of OC and EC may be related to the meteorological factors, such as precipitation (PRE), temperature (TEM), relative humidity (RHU), sunshine radiation (SSD) and wind speed (Win).In PM 2.1 and PM 9.0 , the relationships of the concentration of OC, EC or TC with PRE, TEM, RHU, SSD and Win had been estimated through correlation analysis (Table 3).The table revealed that OC and TC have a significant positive relationship with Win (P < 0.05) in PM 2.1 , and their relationship in PM 9.0 with Win were fairly good (P < 0.01), though the relationships were statistically significant, the range of wind speed was quite small (0.9-1.7 m s -1 ) during the entire sampling period.Thus the wind speed does influence the concentration of OC and TC at Beibei but not strongly.EC is negatively correlated with TEM (P < 0.05) in PM 2.1 , meaning that as the air temperature decreases, the concentration of EC will increase instead, which reflects the highest concentration of EC appeared in winter.There were no significant correlations between EC and five meteorological factors in PM 9.0 .Whitby et al. (1972) first described the basic size distribution of atmospheric particles by Minnesota Aerosol Analyzing System.The size distributions of atmospheric particles were described as three parts: (1) Aitken (< 0.1 µm), (2) accumulation mode (0.1-2 µm), and (3) coarse mode (> 2 µm).Then John et al. (1990) found that there were two sub-modes (the condensation and droplet mode) in the accumulation mode, the mode with the peak at 0.2 µm was "condensation mode," which might mainly be produced by the gas-phase reaction.A mode with the peak at 0.7 µm was the "droplet mode," which might be mainly produced by the fine particle nucleation reaction in droplets (John et al., 1990).

OC and EC as a Function of Particulate Size
The particles in the accumulation mode can stay a long time in the air, and not easy to be degraded.At the same time, the size of particles in the accumulation mode was similar to the wavelength of solar shortwave radiation, indicating that the particles have a high light extinction, and an important impact on global climate and human health (Yu et al., 2010).During the whole observation period, about 61.5% of OC and 50% of EC were concentrated in the fine particles (< 2.10 µm) (Fig. 2), while, the particles with the size of 0.1-2 µm were in accumulation mode, indicating that most of OC and EC were mainly concentrated in accumulation mode, which can greatly influence the global climate and human health.
The particle size of OC showed a "bi-modal" distribution in all four seasons (Fig. 3).The fine particle in the size of 0.43-0.65 µm was in Aitken mode, which may be caused by the vehicle exhausts (Yu et al., 2009) and biomass combustion (Glaser et al., 2005), and the coarse particle in 4.7-5.8µm was in coarse mode, which may be caused by the dust, soil suspended matter (Bi et al., 2005) and biological aerosols, such as microorganisms, plant fragments, pollen, etc. (Lan et al., 2011).While EC in four seasons only shows a coarse peak in the size of 4.7-5.8µm, which may be caused by the direct combustion emissions of the EC mixed with the surface dust generated by tire brake wear, road dust, industrial emissions and construction activities (Salma et al., 2002).

Estimation of Secondary Organic Carbon
The annual average concentration of SOC was 6.3 ± 5.9 µg m -3 , which accounted for 33.5 ± 22.6% of OC, indicating that SOC was an important component of carbonaceous aerosols (Table 2).Such results were lower than those (SOC: 9.0 ± 10.5 µg m -3 ; SOC/OC: 32.3 ± 7.9%) in Wanzhou (Zhang et al., 2015), reflecting that the pollution of SOC in Beibei was less serious than that in Wanzhou.
Previous research results show that the formation of SOC is from VOCs via two processes: condensable organic compounds through oxidation reaction and the nucleation and condensation of vapors (Pandis et al., 1992).The photochemical activity and atmospheric temperature therefore play important roles in the SOC formation.An investigation into the effects of atmospheric temperature on SOC formation showed that SOC in PM 2.5 had a significant negative correlation with temperature (R 2 = 0.42) (Niu et al., 2012).Lower temperatures were favorable for absorption and condensation of semi-volatile organic compounds on existing particles (Pandis et al., 1992;Odum et al., 1996).However, the mechanism of SOC formation during wintertime is still arguable.For example, the photochemical reaction resulting in SOC formation was significant in winter (Huang et al., 2014).The water-soluble organic products through a gasphase photochemistry dissolving into the aqueous phase could react further to form low volatility products that were largely partitioned in the particle phase (Lim et al., 2010), but gaseous oxidant concentrations decreased significantly, suggesting a reduced production of secondary aerosols through gas-phase reactions in winter (Zheng et al., 2015).Thus it might be reasonable to infer that high average SOC concentrations (7.9 ± 7.2 µg m -3 ) in winter was estimated as a result of lower temperature and high VOCs emissions in this study.However, it is difficult to conclude which is the dominant pathway for the chemical processes (OH chemistry, aqueous-phase chemistry or NO 3 chemistry).Compared with results in winter, there was an overall trend towards lower SOC concentration (7.4 ± 4.5 µg m -3 ) but a higher percentage of SOC in the OC (44.5 ± 20.6%) in summer, which might be caused by the more intense solar radiation during the summer and it will provide favorable conditions for photochemical activity and SOC production.

Sources of Carbonaceous Aerosols
The ratio of OC/EC can be used not only to evaluate the existence of SOC, but also to analyze the main pollution sources and the emission characteristics of carbonaceous aerosols (Gu et al., 2010).When the ratios of OC/EC ranged 1.0-4.2,2.5-10.5 and 32.9-81.6,which indicated that the pollution sources were vehicle exhaust emission, coal emission and cooking emission, respectively (Schauer et al., 1999a(Schauer et al., , b, 2001(Schauer et al., , 2002a, b;, b;He et al., 2004;Chen et al., 2006).The ratios of OC/EC in the four seasons in Beibei were 7.0 ± 8.5, 7.4 ± 4.5, 3.1 ± 2.3, 7.9 ± 7.2 respectively, which were all larger than 2.0.The high ratios of OC/EC suggested that the measured OC was not only come from the direct emissions of particles as primary pollutants but also in the form of SOC formed by chemical reactions or produced by residential biomass burning and coal combustion (Shen et al., 2017).In addition, it could be a heavy-duty diesel truck exhaust if the ratio of OC/EC was 0.8 (Hildemann et al., 1991).Thus the value over 0.8 of the OC/EC ratios in all the four seasons (Table 2) would indicate a less heavy-duty diesel truck exhaust in the Beibei suburb, which was consistent with no direct industrial pollution sources within a 5-km distance of the sampling site in this study.Turpin and Huntzicker (1991) found that the relationship between OC and EC could be further used to distinguish the source of carbon particles if OC had a good correlation with EC, which indicated that they came from the same source.Therefore, the correlation analysis of OC and EC could probably distinguish the source of carbon aerosol particles (Turpin et al., 1990).From Fig. 4, a significant correlation between OC and EC in PM 2.1 in the Beibei suburb (R 2 = 0.43, P < 0.01) could indicate the same source for both OC and EC.Chen et al. (2006) reported that the air pollution could mainly derive from coal emissions when the ratio of OC/EC was between 2.5 and 10.5.As a result, the main air pollution source could be the coal  emission as the slope of the correlation curve of OC and EC in PM 2.1 was 7.27 in the Beibei suburb.
One of the unique features of the IMPROVE protocol was that it can provide the concentrations of seven fractions for carbonaceous aerosol particles (Chow et al., 1993(Chow et al., , 2001)).The abundances in each seven fractions of carbonaceous aerosols had been utilized to distinguish different emission sources of carbonaceous aerosols (Chow et al., 2003(Chow et al., , 2004a)).
The concentrations of OC 1 , OC 2 , OC 3 , OC 4 , EC 1 , EC 2 and EC 3 in PM 2.1 were 0.8 ± 0.5, 6.0 ± 2.9, 3.6 ± 1.6, 2.4 ± 1.3, 4.7 ± 3.7, 0.7 ± 0.3 and 0.1 ± 0.2 µg m -3 , accounting for 4.2%, 33.3%, 19.8%, 13.3%, 26.1%, 3.9% and 0.4% of the TC, respectively (Fig. 5).In general, OC 2 , OC 3 , OC 4 and EC 1 were the most abundant species in Beibei.Previous studies had showed that the abundance of OC 2 , OC 3 , OC 4 , EC 1 and EC 2 might be associated with motor vehicle exhaust, EC 1 was the characteristic component of gasoline vehicle exhaust emissions, while EC 2 and EC 3 were enriched in diesel vehicle emission profile (Watson et al., 1994;Chow et al., 2004;Cao et al., 2005).The concentrations of OC 2 , OC 3 , OC 4 and EC 1 in PM 2.1 were higher than the others, and the concentrations of EC 2 and EC 3 were the lowest (Fig. 5), indicating that the main pollution of carbonaceous aerosols in Beibei was the motor vehicle exhaust, especially gasoline vehicle emission.Cao et al. (2005) found that OC 2 accounted for 46.9% of TC in coal combustion samples, 29.2% in biomass burning samples and 30.5% in motor vehicle samples, and OC 1 was the characteristic emission component of biomass combustion (contributed 36.8% to TC in biomass burning samples).While the concentration of OC 1 in spring was higher than that in other three seasons (Fig. 5), and the annual average concentration of OC 2 was highest among the seven carbon components, indicating that more biomass was burned in spring, and the coal combustion was one of the main pollution sources of carbonaceous aerosols in Beibei.Chow et al. (2004a) also found that OC 3 (accounted 43% for OC) and EC 1 (accounted 93% for EC) were enriched in the cooking profile, while the concentrations of OC 3 and EC 1 were higher in spring and winter than those in summer and autumn (Fig. 5).This reflected that cooking was another pollution source of carbonaceous aerosols in Beibei, and more cooking emission in winter and spring might be related to more hot pot (a traditional typical gourmet food in Chongqing) was eaten in winter and spring.
To distinguish the different emission sources of carbonaceous aerosols in Beibei, PCA (principal component analysis) was used to analyze the seven fractions of carbonaceous aerosol in PM 2.1 samples by the SPSS software.The factors whose eigenvalue greater than 1 were extracted, and the factor value of different components was differentiated to make it easy to factor analysis by the method of orthogonal rotation.Then, the factor values used to identify the factor characteristic components were underlined, three factors with eigenvalues greater than 1 were finally obtained (Table 4).
Firstly, the contribution rate of Factor 1 was the highest in three factors and accounted for 55.3% of all seven fractions, with OC 3 , OC 4 and EC 1 as the main contributors, while OC 3 , OC 4 , EC 1 and EC 2 were associated with motor vehicle emission, and EC 1 was the characteristic component of gasoline vehicle exhaust, so the Factor 1 might be connected with the gasoline vehicle exhaust.Secondly, the contribution of Factor 2 accounted for 16.2% of all seven fractions, with OC 1 and OC 2 as the main contributors, while Cao et al. (2005) found that OC 2 accounted for 46.9% of TC in coal combustion samples and OC 1 was the characteristic emission component of biomass combustion (contributed 36.8% to TC in biomass-burning samples), so the Factor 2 might be consisted with coal emission.Thirdly, the contribution of Factor 3 accounted for 15.0% of all seven fractions, which was close to Factor 2, was mainly affected by OC 1 , which might be connected with biomass combustion.Thus the main pollution sources of carbonaceous aerosols in Beibei were gasoline vehicle exhaust, coal emission and biomass combustion.

CONCLUSION
The carbonaceous aerosol was investigated on the campus of Southwest University in Beibei District, a suburb of Chongqing, China.During the period of sampling, the average annual concentrations of the OC and EC in the PM 2.1 and PM 9.0 were 16.3 ± 7.6 and 1.8 ± 0.7 µg m -3 , and 25.0 ± 9.6 and 3.2 ± 1.2 µg m -3 , respectively.The distribution of OC concentrations in the PM 2.1 was highly seasonal (winter > spring > summer > autumn), and increased biomass burning, for the purpose of preserving meat via smoking, may have accounted for the highest concentrations occurring in winter, whereas increased rainfall may have accounted for the lowest concentrations occurring in autumn.Meanwhile, the concentrations of OC in the PM 9.0 in winter were close to those in spring and summer but higher than those in autumn.By contrast, the PM 2.1 exhibited a more seasonal distribution of EC concentrations (winter > spring > summer > autumn), with the highest and lowest concentrations, occurring in winter and autumn, respectively, being attributable to the same factors that affected the OC.For PM 2.1 , significantly positive and negative correlations existed between the OC and the wind speed and between the EC and the air temperature; however, for PM 9.0 , significantly positive or negative correlations existed between the OC and the wind speed (P < 0.01) and between the OC and the relative humidity (P < 0.05), but no significant correlation was detected between the EC and any of the five meteorological factors.In addition, the OC and the EC were mainly concentrated in the fine particles (< 2.1 µm), and the former showed a bi-modal size distribution in all four seasons.Fine particles in the size range of 0.43-0.65 µm may have originated in vehicle exhaust and biomass combustion, whereas coarse particles in the size range of 4.7-5.8µm may have originated in dust, suspended soil matter, and biological aerosols, such as microorganisms, plant fragments, and pollen.The EC in all four seasons only exhibited a peak in the coarse mode, in the size range of 4.7-5.8µm, which may have been caused by direct combustion emissions of EC mixing with surface dust generated by tire/brake wear, road dust, industrial emissions, and particles from construction.Furthermore, the OC and the EC were significantly correlated (P < 0.01), and the average annual concentration of the SOC was 6.3 ± 5.9 µg m -3 , accounting for 33.5 ± 22.6% of the OC.Finally, analyses of the sources of atmospheric aerosol indicated that the pollution in Beibei was mainly derived from the exhaust of gasoline vehicles, biomass combustion, and coal combustion.

Fig. 3 .
Fig. 3. Size distribution of OC and EC in four seasons.

Fig. 4 .
Fig. 4. Correlation analyses between organic carbon (OC) and elemental carbon (EC) in PM 2.1 in Beibei District, a suburb of Chongqing Municipal City, Southwest China.

Table 1 .
Concentrations of elemental carbon (EC), organic carbon (OC), and TC (total carbon (TC, OC + EC)) in PM 2.1 and PM 9.0 in the Beibei suburb of Chongqing Municipal City, Southwest China.

Table 2 .
Concentrations of secondary organic carbon (SOC), ratios of SOC/organic carbon (OC) and OC/EC (elemental carbon) in PM 2.1 during four seasons in 2014 in the Beibei suburb of Chongqing Municipal City, Southwest China.

Table 3 .
Correlation analyses between meteorological factors and organic carbon (OC), elemental carbon (EC) or total carbon (TC) in the PM 2.1 and PM 9.0 in Beibei suburb of Chongqing Municipal City, Southwest China.

Table 4 .
Principal component analyses for seven carbonaceous aerosols in Beibei suburb of Chongqing Municipal City, Southwest China.