Characteristics of Carbonaceous Aerosol in PM2.5 at Wanzhou in the Southwest of China

Hourly organic carbon (OC) and elemental carbon (EC) concentrations in PM2.5 were measured from June 2013 to May 2014 in Wanzhou, the second largest city in the Chongqing Municipality, in the southwest of China. Results show that the annual average concentrations of OC and EC were 13.16 ± 7.98 and 3.12 ± 1.51 μgC·m−3, respectively. Clear seasonal variations of OC and EC concentrations were observed, with their concentrations at minima in summer and maxima in winter. The diel concentration profile of OC and EC presented a bimodal pattern, which was attributed to the cooperative effects of local meteorological conditions and source emissions. The daily average OC/EC ratio ranged from 2.05 to 8.17 with an average of 4.15 for the whole study period. Strong correlations between OC and EC were found in winter and spring, indicating their common sources, while their correlations were poorer in summer and autumn, indicating that the influence of biogenic emissions and secondary organic carbon (SOC) were significant during those seasons. The estimated SOC concentrations were 2.19 ± 1.55, 7.66 ± 5.89, 5.79 ± 3.51, and 3.43 ± 2.26 μgC·m−3, accounting for 29.2%, 52.7%, 27.4%, and 30.5% of total organic carbon in summer, autumn, winter, and spring, respectively. The analysis of back trajectories suggested that high PM2.5, OC, and EC concentrations were associated with air masses originating from or passing over several industrial centers and urban areas in western and northwestern China. Air trajectories from the southeast with short pathways were the dominant trajectories arriving at Wanzhou, indicating that local sources had a big influence on PM2.5, OC, and EC concentrations.


Introduction
Carbonaceous aerosol-which contributes 20-60% to PM 2.5 mass in urban settings [1][2][3][4]-has been found to have a large impact on environmental conditions, such as visibility and climate change [5][6][7][8]. Carbonaceous aerosol is usually separated into elemental carbon (EC) and organic carbon (OC). EC has a graphitic-like structure and comes mainly from incomplete combustion processes. EC has a strong absorptivity of solar radiation and plays an important role in visibility reduction and aerosol radiative forcing [5,9,10]. EC-containing particles are also associated with adverse health effects [11,12]. OC consists of thousands of organic compounds, containing polycyclic aromatic hydrocarbons, polychlorinated biphenyls, and other hazardous components, which have possible mutagenic and carcinogenic effects. OC can be directly emitted from sources (primary OC, POC) or produced from the gas-to-particle conversion of volatile organic compounds (secondary OC, SOC) [13].

Experimental Methods
Particulate carbonaceous aerosol was measured using a Sunset Laboratory semi-continuous OC/EC analyzer (Model 4, Sunset Laboratory Inc., Forest Grove, OR, USA) with a thermal-optical transmittance (TOT) protocol for pyrolysis correction. The ambient air was first sampled into a PM2.5 cyclone inlet with a flow rate of 8 L·min −1 , and then passed through a carbon parallel-plate diffusion denuder (from Sunset Lab.) to remove volatile organic compounds that may cause a positive bias in the OC concentrations measured [39]. The OC and EC were collected on a quartz fiber filter with an effective collection area of 1.13 cm 2 . The analyzer was programmed to collect aerosol for 45 min at the start of each hour, followed by the analysis of carbonaceous species during the remainder of the hour.
After sampling, the instrument oven was first purged with helium, and then the filter was heated according to a NIOSH-type temperature protocol [40], with a slight modification to the heating temperature according to previous studies [1,26]. The oven temperature was first increased to a maximum of 870 °C, thermally desorbing OC and pyrolysis carbon (PC) into a manganese dioxide (MnO2) oxidizing oven. Then the temperature was reduced to 550 °C; the carrier gas was switched to 10% oxygen in helium. A second temperature ramp was then initiated in the oxidizing gas stream and EC was oxidized off the filter and into the oxidizing oven. As the carbon components flowed through the MnO2 oven they were converted to CO2 and detected by a Non-Dispersive Infrared Absorption (NDIR) CO2 sensor. OC and EC were automatically quantified by dividing their peak areas by the internal calibration peak made by methane gas (5% CH4 in helium). The correction for the PC converted from OC to EC was performed by monitoring the transmittance of a pulsed He-Ne diode laser beam at 670 nm through the quartz filter during the sample analysis. Control calibrations by sucrose were conducted regularly and instrumental blanks (measured by an automatic reanalysis of the just analyzed filter) were determined once per day at 0:00 o'clock with the following average values: OC = 0.20 ± 0.10 μg; EC = 0.00 μg (the midnight sample collection were 15 min shorter). Dynamic blanks were also measured by adding a quartz fiber filter that was not prefired upstream of the denuder and using the same sampling time as during the measurement (in total 60 dynamic blank samples were collected) with the following averages: OC = 0.43 ± 0.25 μgC·m −3 ; EC = 0.01 μgC·m −3 . The uncertainty of this system is reported to be 5% [41].
The sampling period was chosen to represent the four seasons, as follows: summer, 18 June 2013 to 31 August 2013; autumn, 1 September 2013 to 30 November 2013; winter, 1 December 2013 to 28

Experimental Methods
Particulate carbonaceous aerosol was measured using a Sunset Laboratory semi-continuous OC/EC analyzer (Model 4, Sunset Laboratory Inc., Forest Grove, OR, USA) with a thermal-optical transmittance (TOT) protocol for pyrolysis correction. The ambient air was first sampled into a PM 2.5 cyclone inlet with a flow rate of 8 L·min −1 , and then passed through a carbon parallel-plate diffusion denuder (from Sunset Lab.) to remove volatile organic compounds that may cause a positive bias in the OC concentrations measured [39]. The OC and EC were collected on a quartz fiber filter with an effective collection area of 1.13 cm 2 . The analyzer was programmed to collect aerosol for 45 min at the start of each hour, followed by the analysis of carbonaceous species during the remainder of the hour.
After sampling, the instrument oven was first purged with helium, and then the filter was heated according to a NIOSH-type temperature protocol [40], with a slight modification to the heating temperature according to previous studies [1,26]. The oven temperature was first increased to a maximum of 870 • C, thermally desorbing OC and pyrolysis carbon (PC) into a manganese dioxide (MnO 2 ) oxidizing oven. Then the temperature was reduced to 550 • C; the carrier gas was switched to 10% oxygen in helium. A second temperature ramp was then initiated in the oxidizing gas stream and EC was oxidized off the filter and into the oxidizing oven. As the carbon components flowed through the MnO 2 oven they were converted to CO 2 and detected by a Non-Dispersive Infrared Absorption (NDIR) CO 2 sensor. OC and EC were automatically quantified by dividing their peak areas by the internal calibration peak made by methane gas (5% CH 4 in helium). The correction for the PC converted from OC to EC was performed by monitoring the transmittance of a pulsed He-Ne diode laser beam at 670 nm through the quartz filter during the sample analysis. Control calibrations by sucrose were conducted regularly and instrumental blanks (measured by an automatic reanalysis of the just analyzed filter) were determined once per day at 0:00 o'clock with the following average values: OC = 0.20 ± 0.10 µg; EC = 0.00 µg (the midnight sample collection were 15 min shorter). Dynamic blanks were also measured by adding a quartz fiber filter that was not prefired upstream of the denuder and using the same sampling time as during the measurement (in total 60 dynamic blank samples were collected) with the following averages: OC = 0.43 ± 0.25 µgC·m −3 ; EC = 0.01 µgC·m −3 . The uncertainty of this system is reported to be 5% [41].
The sampling period was chosen to represent the four seasons, as follows: summer, 18 28 February 2014; and spring, 1 March 2014 to 31 May 2014. Due to instrument malfunction, data from 12 January to 7 February 2014 was excluded from our results. A total of~300 days of data was collected on the instrument during this study. The meteorological parameters were also collected by an automatic weather station (WS500-UMB, GER) during the sampling period. As the sampling interval was set at 5 minutes, data for temperature, relative humidity (RH), and wind speed were averaged in the same periods as OC and EC.

Results and Discussion
3.1. Seasonal Variations of OC and EC in PM 2.5 As shown in Figure 2, OC and EC exhibited similar daily variations in Wanzhou. The daily average concentration of OC ranged from 3.84 to 44.09 µgC·m −3 with an average concentration of 13.16 ± 7.98 µgC·m −3 , while EC concentration varied from 0.76 to 8.34 µgC·m −3 with an average concentration of 3.12 ± 1.51 µgC·m −3 . OC and EC concentrations in this study are compared with those measured in other Chinese cities in Table 1. In spite of the incomparability of OC and EC data between the methods of NIOSH TOT, Interagency Monitoring of Protected Visual Environments (IMPROVE), thermal optical reflectance (TOR), and the two-step thermal procedure, total carbon (TC, TC = OC+ EC) measurements by those methods agree well [42,43]. The level of TC in Wanzhou was found to be similar to the urban sites in Shanghai, Quanzhou, and to the suburban site in Xiamen, but lower than the urban sites in Chongqing, Chengdu, Beijing, Tianjing, Guangzhou, and Wuhan in China. TC concentration contributed 25.3% to PM 2.5 mass in Wanzhou. This fraction was comparable to that in most Chinese sites (17.6-41.4%).     Conversion of OC to organic matter (OM) is recognized as one of the most critical factors of uncertainty in mass closure calculations [44]. Turpin and Lim [45] suggested the use of different conversion factors according to site type: 1.6 ± 0.2 for urban areas, 1.9-2.3 for aged aerosols, and 2.2-2.6 for biomass burning. Taking into account that our site is an urban station, we used a value of 1.6 (OM = 1.6 × OC). Total carbonaceous aerosol (TCA) was calculated by the sum of EC and organic matter (OM). As Table 1 shows, TCA accounted for 38% of PM 2.5 , implying that TCA represented a major fraction of the PM 2.5 mass in Wanzhou. This fraction was in accordance with findings in other Chinese sites (26.7-51.3%).  Figure 3 shows that monthly average OC values varied by nearly fourfold, from a low of 6.45 ± 1.19 µgC·m −3 (July 2013) to 26.84 ± 9.63 µgC·m −3 (January 2014), while the monthly average EC values varied by nearly threefold, from a low of 1.93 ± 0.40 µgC·m −3 (July 2013) to 5.08 ± 1.96 µgC·m −3 (January 2014). Both monthly OC and EC concentrations increased during autumn and peaked during winter. The seasonal concentrations of OC and EC are summarized in Table 2. The seasonal average OC concentrations in PM 2.5 were 7.47 ± 2.37 µgC·m −3 for summer, 14.81 ± 8.39 µgC·m −3 for autumn, 21.11 ± 9.22 µgC·m −3 for winter, and 11.25 ± 4.65 µgC·m −3 for spring, while the average EC concentrations during the four seasons were 2.12 ± 0.51, 3.53 ± 1.51, 4.32 ± 1.86, and 2.76 ± 1.12 µgC·m −3 , respectively. The OC and EC concentrations exhibited similar seasonal variations, with the descending order: winter > autumn > spring > summer. The seasonal characteristics of carbonaceous aerosols in Wanzhou can be explained as the combined impact of climatic conditions and local emissions. Though there was no central-heating period in winter as with northern cities in China, local emissions (coal and wood burning, small stoves used, etc.) also increased rapidly due to decentralized heating on cold days. At the same time, there was an increased combustion source attributed to the residents using wood-burning to cook their traditional bacon [37]. In addition, the urban area of Wanzhou is located in a valley surrounded by mountains that trap the pollutants within. As can also be seen from Table 2, the wind speed was lowest (0.78 m·s −1 in winter, 0.84, 0.95, 0.79 m·s −1 in spring, summer, and autumn, respectively) and RH was relatively higher in winter (78.0%, 80.2%, 69.2%, and 80.6% in winter, spring, summer, and autumn); these unfavourable geographical and meteorological conditions limited the dilution and dispersion of pollutants. In autumn, the enhanced emissions from biomass burning and frequent stable weather conditions mainly resulted in higher carbonaceous concentrations. In spring, the higher wind speeds facilitated the dispersion of pollutants. In the summer, rainfall was much more abundant (60-70% of annual precipitation in the summer, about 1181.5 mm on average) [51], and carbonaceous aerosol could be efficiently removed by wet scavenging [36]. The seasonal variations of OC and EC in this study were similar to other studies conducted in China [3,27,31,50].
Atmosphere 2018, 9, x FOR PEER REVIEW 6 of 14 by wet scavenging [36]. The seasonal variations of OC and EC in this study were similar to other studies conducted in China [3,27,31,50].

Diel Variations of OC and EC in PM2.5
The diel variations of OC and EC in different seasons are illustrated in Figure 4a,b. Pronounced EC diel variations were observed in different seasons, a bimodal pattern with a morning rush peak (8:00-10:00) and an evening peak (19:00-20:00). This bimodal trend was consistent with the local meteorological conditions and source emission. The EC distinct morning peak clearly corresponded to the morning traffic rush hour. The EC evening peak could be attributed to afternoon traffic emissions and cooking, combined with low mixing heights. The boundary layer usually becomes deeper in the afternoon because of strong solar radiation and turbulent eddies, providing a larger volume for diluting pollutants; hence the EC trough can be observed during this period (14:00-16:00).  The diel variations of OC and EC in different seasons are illustrated in Figure 4a,b. Pronounced EC diel variations were observed in different seasons, a bimodal pattern with a morning rush peak (8:00-10:00) and an evening peak (19:00-20:00). This bimodal trend was consistent with the local meteorological conditions and source emission. The EC distinct morning peak clearly corresponded to the morning traffic rush hour. The EC evening peak could be attributed to afternoon traffic emissions and cooking, combined with low mixing heights. The boundary layer usually becomes deeper in the afternoon because of strong solar radiation and turbulent eddies, providing a larger volume for diluting pollutants; hence the EC trough can be observed during this period (14:00-16:00). In comparison with EC, OC also showed a bimodal pattern, but there had broad maxima from 7:00 to 12:00 in summer, autumn, and spring. This phenomenon may be caused by the relative importance of sources of OC other than traffic existing at the site, such as secondary organic aerosol formation. In winter, the OC displayed a clearly bimodal pattern, with morning peak and evening peak occuring at 12:00 and 19:00-20:00 local time, respectively. The midday peak could be due to the longer existence of the inversion layer in the mountain area in winter. This phenomenon was also observed in EC diel variation; the EC morning peak time was delayed by about one to two hours during the winter. The later onset of the morning buildup and the earlier onset of the evening accumulation in winter compared to summer corresponded with the later sunrises and earlier sunsets in winter. The highest concentrations of both EC and OC were observed during the winter evenings and nights when there was the highest need for residential heating. However, the main factor for the higher night-time OC and EC concentrations (in all seasons) was probably a lower on average nocturnal mixing boundary layer in comparison with the daylight hours. Therefore, regardless of the season, the diel minima of EC and OC were found during the afternoon with both the highest temperature and a thicker mixing boundary layer.
night-time OC and EC concentrations (in all seasons) was probably a lower on average nocturnal mixing boundary layer in comparison with the daylight hours. Therefore, regardless of the season, the diel minima of EC and OC were found during the afternoon with both the highest temperature and a thicker mixing boundary layer.
The OC/EC ratio was rather variable within the day (Figure 4c). The diel variations of OC/EC ratios exhibited two peaks around the time 12:00 and 19:00 in summer, autumn, and spring. The midday OC/EC maximum can be due to the formation of SOC when the photochemical activity was intensive under favourable meteorological conditions. The evening OC/EC maximum may be owing to the surface inversion formed after sunset, trapping more secondary organic carbon in the shallow boundary layer. Unlike other seasons, the OC/EC ratio exhibited a third peak at 3:00 in winter. One possible reason for this phenomenon was residential heating (coal and wood burning, small stoves used, etc.) used during the cold winter night. The diel variations of OC/EC ratios in different seasons showed minimum values during the morning rush hour (8:00-9:00) when OC and EC concentrations build up, probably due to the increased contributions from primary emissions. In general, the OC/EC ratio in winter was higher than the other three seasons. There may be two reasons for this phenomenon: one is attributed to the increased combustion emissions as depicted in Section 3.1, the other one is that the adverse meteorological conditions in winter limited the dilution and dispersion of pollutants, and created conditions for the condensation or adsorption of volatile organic compounds.   The OC/EC ratio was rather variable within the day (Figure 4c). The diel variations of OC/EC ratios exhibited two peaks around the time 12:00 and 19:00 in summer, autumn, and spring. The midday OC/EC maximum can be due to the formation of SOC when the photochemical activity was intensive under favourable meteorological conditions. The evening OC/EC maximum may be owing to the surface inversion formed after sunset, trapping more secondary organic carbon in the shallow boundary layer. Unlike other seasons, the OC/EC ratio exhibited a third peak at 3:00 in winter. One possible reason for this phenomenon was residential heating (coal and wood burning, small stoves used, etc.) used during the cold winter night. The diel variations of OC/EC ratios in different seasons showed minimum values during the morning rush hour (8:00-9:00) when OC and EC concentrations build up, probably due to the increased contributions from primary emissions. In general, the OC/EC ratio in winter was higher than the other three seasons. There may be two reasons for this phenomenon: one is attributed to the increased combustion emissions as depicted in Section 3.1, the other one is that the adverse meteorological conditions in winter limited the dilution and dispersion of pollutants, and created conditions for the condensation or adsorption of volatile organic compounds.

Relationship between OC and EC
As mentioned above, EC comes mainly from incomplete combustion emissions such as coal consumption, vehicle exhaust, and biomass burning. OC originates from both primary and secondary sources. Since primary OC and EC are mostly generated from the same sources, EC can be used as a tracer for primary combustion generated OC [52]. The relationship between OC and EC is very useful in assessing the original sources of carbonaceous particles [32,53]. As shown in Figure 5, the higher correlations between EC and OC were in winter (0.95) and spring (0.91). The poorer correlation (but still high) was observed in the summer (0.77). The strong winter and spring correlations indicated the presence of common dominant sources for OC and EC (e.g., biomass burning, coal combustion, and motor vehicle exhaust) in the vicinity of the site. On the contrary, the lower correlation in summer and autumn indicated some independent OC sources without the presence of EC such as biogenic and secondary organic aerosols [26]. The regression slopes varied from 3.11 to 4.71, implying that there is a seasonal variability in emission sources and SOA contributions.
As mentioned above, EC comes mainly from incomplete combustion emissions such as coal consumption, vehicle exhaust, and biomass burning. OC originates from both primary and secondary sources. Since primary OC and EC are mostly generated from the same sources, EC can be used as a tracer for primary combustion generated OC [52]. The relationship between OC and EC is very useful in assessing the original sources of carbonaceous particles [32,53]. As shown in Figure 5, the higher correlations between EC and OC were in winter (0.95) and spring (0.91). The poorer correlation (but still high) was observed in the summer (0.77). The strong winter and spring correlations indicated the presence of common dominant sources for OC and EC (e.g., biomass burning, coal combustion, and motor vehicle exhaust) in the vicinity of the site. On the contrary, the lower correlation in summer and autumn indicated some independent OC sources without the presence of EC such as biogenic and secondary organic aerosols [26]. The regression slopes varied from 3.11 to 4.71, implying that there is a seasonal variability in emission sources and SOA contributions.
The OC/EC ratio has been used to study the emission and transformation characteristics of carbonaceous aerosol. Typical emission sources of carbonaceous fractions include vehicle exhaust (2.5-5.0) [54], coal combustion (2.5-10.5) [55], and biomass burning (4.3-79.7) [56,57], respectively. It should be noted that the OC/EC ratios presented above were all measured by the TOT method, and were comparatively higher than the values by TOR [58]. As shown in Figure 2 and Table 2, the daily average OC/EC ratio varied between 2.05 and 8.17 with an average of 4.15 for the whole study period. The OC/EC ratio showed seasonal variations with a high value in winter (5.03) and a low value in summer (3.55). This seasonal pattern of higher wintertime OC/EC ratio was also observed in PRDR [33] and Shanghai [34]. The elevated OC/EC ratio in winter could be attributed to several reasons. Firstly, coal combustion and biomass emission for winter heating contribute more to OC than EC, and also increase the emission of volatile organic precursors. Peng et al. [38] found that TC in PM2.5 had a stronger relationship with water-soluble potassium (K + ) in winter than in summer, suggesting biomass burning contributed significantly to the carbon pollution in Wanzhou. Secondly, a stable atmosphere and low temperatures can facilitate the accumulation of air pollutants and create conditions for the condensation or adsorption of volatile organic compounds. Thirdly, the low mixing layer height in winter would enhance the SOC formation [59].

Estimation of Secondary Organic Carbon Concentration
Appel et al. [60] pointed out that high OC/EC ratios indicated the formation of SOC from gaseous precursors through the photochemical process. Chow et al. [53] observed that the ratios of OC/EC The OC/EC ratio has been used to study the emission and transformation characteristics of carbonaceous aerosol. Typical emission sources of carbonaceous fractions include vehicle exhaust (2.5-5.0) [54], coal combustion (2.5-10.5) [55], and biomass burning (4.3-79.7) [56,57], respectively. It should be noted that the OC/EC ratios presented above were all measured by the TOT method, and were comparatively higher than the values by TOR [58]. As shown in Figure 2 and Table 2, the daily average OC/EC ratio varied between 2.05 and 8.17 with an average of 4.15 for the whole study period. The OC/EC ratio showed seasonal variations with a high value in winter (5.03) and a low value in summer (3.55). This seasonal pattern of higher wintertime OC/EC ratio was also observed in PRDR [33] and Shanghai [34]. The elevated OC/EC ratio in winter could be attributed to several reasons. Firstly, coal combustion and biomass emission for winter heating contribute more to OC than EC, and also increase the emission of volatile organic precursors. Peng et al. [38] found that TC in PM 2.5 had a stronger relationship with water-soluble potassium (K + ) in winter than in summer, suggesting biomass burning contributed significantly to the carbon pollution in Wanzhou. Secondly, a stable atmosphere and low temperatures can facilitate the accumulation of air pollutants and create conditions for the condensation or adsorption of volatile organic compounds. Thirdly, the low mixing layer height in winter would enhance the SOC formation [59].

Estimation of Secondary Organic Carbon Concentration
Appel et al. [60] pointed out that high OC/EC ratios indicated the formation of SOC from gaseous precursors through the photochemical process. Chow et al. [53] observed that the ratios of OC/EC exceeding 2.0 can be used to identify SOC formation. In this study, all the values of OC/EC were larger than 2.0, indicating the formation of SOC during the sampling period. Castro et al. [61] suggested using the minimum OC/EC ratios in the EC tracer method to calculate SOC; the equation is as follows: where (OC/EC) min is the minimum of the OC/EC ratios observed during the sampling time.
An estimation of SOC in Wanzhou was made according to Equation (1). It must be noted that this method can give only semi-quantitative information because of the high uncertainties associated with them. As discussed in Sections 3.2 and 3.3, the ratios of OC/EC can be affected by the meteorology and local sources. Taking these into consideration, the (OC/EC) min observed for each season was used in calculation: the values were 2.52 in summer, 2.05 in autumn, 3.57 in winter, and 2.84 in spring. These values were comparable with those of Guangzhou (2.3-4.5) [33], Shanghai (2.7-4.1) [34], and Xiamen (2.4) [50]. The seasonal variations of SOC concentrations and SOC/OC ratios are depicted in Figure 6. The seasonal variations of the estimated SOC concentrations were 2.19 ± 1.55 µgC·m −3 , 7.66 ± 5.89, 5.79 ± 3.51, and 3.43 ± 2.26 in summer, autumn, winter, and spring, respectively. The relative proportion of the estimated SOC in OC was highest in autumn (52.7%) and roughly comparable in spring (30.5%), summer (29.2%), and winter (27.4%). This implied that the relative contribution of SOC to OC in PM 2.5 was only comparable to that of POC to OC in autumn, while in the other three seasons, OC was dominated by POC. High SOC/OC ratio in autumn in Wanzhou could be due to the abundant sunlight and less precipitation as well as a stagnated atmospheric condition which provided favorable conditions for photochemical production of SOC and accumulation of pollutants. In wintertime, the averaged SOC concentrations were also higher compared to those in summer and spring, but its abundance in OC was smallest. The rather high absolute SOC contribution in winter could be due to the low mixing heights and stagnated weather conditions favoring the formation and accumulation of secondary organic aerosols. However, the stagnating weather conditions usually lead to high concentrations of all pollutants (both primary and secondary OC). The similar seasonal variation trend was also found in Beijing [42], Shanghai [34], and Tianjin [31].
wintertime, the averaged SOC concentrations were also higher compared to those in summer and spring, but its abundance in OC was smallest. The rather high absolute SOC contribution in winter could be due to the low mixing heights and stagnated weather conditions favoring the formation and accumulation of secondary organic aerosols. However, the stagnating weather conditions usually lead to high concentrations of all pollutants (both primary and secondary OC). The similar seasonal variation trend was also found in Beijing [42], Shanghai [34], and Tianjin [31].

Back Trajectory Analysis
In order to investigate the effect of air mass pathway on the characteristics of carbonaceous aerosol, 3-day back trajectories arriving at Wanzhou at a height of 100 m above the ground level were calculated using a Geographical Information System based software [62] and gridded meteorological

Back Trajectory Analysis
In order to investigate the effect of air mass pathway on the characteristics of carbonaceous aerosol, 3-day back trajectories arriving at Wanzhou at a height of 100 m above the ground level were calculated using a Geographical Information System based software [62] and gridded meteorological data from the US National Oceanic and Atmospheric Administration (NOAA). The trajectory computations were carried out once a day, with a start time of 00 UTC. The run time of every trajectory was 72 h. The three-dimensional vertical velocity field was used to calculate the vertical motion of the air parcel. Five clusters were obtained by the clustering algorithm for the sampling times, and the cluster-mean trajectories are shown in Figure 7. The air masses associated with cluster 1 originated from the middle part of Guangxi Province and then crossed the eastern part of Guizhou Province. Cluster 2 originated from the northern Tibetan Plateau, and moved southeasterly over the Qaidam basin, north of Sichuan Province, with a relatively higher travelling height. Cluster 3 originated from northeastern China, and passed over the middle of Henan and the northwest of Hubei. Cluster 4 came from the middle of Sichuan Province, and passed over several urban districts of Sichuan. Cluster 5 represented the short trajectories that came from the southeast areas around Wanzhou with a relatively lower travelling height.
Statistics of PM 2.5 , OC, EC, and TC daily average concentrations associated with each cluster's backward trajectories are summarized in Table 3. The cluster-averaged PM 2.5 , OC, and EC levels were highest in cluster 4, i.e., 86.69, 17.09, and 4.23 µg·m −3 , respectively. This is mainly because air masses of cluster 4 passed over several industrial centers and urban areas, such as Chengdu, a highly populated as well as industrially dense city [48], and Suining, Guang'an, and Dazhou. These cities are rich in mineral resources, such as oil, natural gas, and coal mines, etc. Cluster 2 and cluster 5 showed similar carbonaceous aerosols concentrations, but the PM 2.5 concentration was higher in cluster 2. Cluster 2 passed though sources of Asian dust and several industrial cities, therefore its pathways were likely to bring aerosols to Wanzhou. Cluster 5 mainly came from local sources and was the primary type of air masses arriving in Wanzhou. There are large-scale industrial complexes including chemical industrial plants and the power plants in the southeast areas of Wanzhou. Thus, the southeast wind brought in a large amount of pollutants from the industrial activities from this region's sources. The cluster-averaged low concentrations appeared in cluster 1 and cluster 3. The clean air masses from the south and northeast areas would disperse air pollution in Wanzhou.
Province. Cluster 2 originated from the northern Tibetan Plateau, and moved southeasterly over the Qaidam basin, north of Sichuan Province, with a relatively higher travelling height. Cluster 3 originated from northeastern China, and passed over the middle of Henan and the northwest of Hubei. Cluster 4 came from the middle of Sichuan Province, and passed over several urban districts of Sichuan. Cluster 5 represented the short trajectories that came from the southeast areas around Wanzhou with a relatively lower travelling height.  Table 3. The cluster-averaged PM2.5, OC, and EC levels were highest in cluster 4, i.e., 86.69, 17.09, and 4.23 μg·m −3 , respectively. This is mainly because air masses of cluster 4 passed over several industrial centers and urban areas, such as Chengdu, a highly populated as well as industrially dense city [48], and Suining, Guang'an, and Dazhou. These cities are rich in mineral resources, such as oil, natural gas, and coal mines, etc. Cluster 2 and cluster 5 showed similar carbonaceous aerosols concentrations, but the PM2.5 concentration was higher in cluster 2. Cluster 2 passed though sources of Asian dust and several industrial cities, therefore its pathways were likely to bring aerosols to Wanzhou. Cluster 5 mainly came from local sources and was the primary type of air masses arriving in Wanzhou. There are large-scale industrial complexes including chemical industrial plants and the power plants in the southeast areas of Wanzhou. Thus,

Conclusions
In this study, OC and EC concentrations in PM 2.5 were measured hourly with a semi-continuous thermal optical analyzer at an urban site in Wanzhou (in southwest China) during four seasons from June 2013 to May 2014. The annual average concentrations of OC and EC were 13.16 ± 7.98 µgC·m −3 and 3.12 ± 1.51 µgC·m −3 . OC and EC showed seasonal variations as expected, with the highest values in winter and the lowest concentrations in summer, which was mainly associated with seasonal changes in source emissions and mesoscale meteorology. Pronounced and similar diel variations of EC concentrations in different seasons were found, which both displayed a bimodal pattern with peaks appearing in the morning and evening rush hour. The diel variations of OC concentrations also showed a similar bimodal pattern to EC in summer, autumn, and spring, but the morning peaks were broad. In wintertime, the diel OC concentration exhibited a clear bimodal pattern with the morning peak occurred at midday. Additionally, OC had its diel maxima during the night in all of the seasons. The seasonal and diel variations of OC and EC concentration were mostly dominated by the seasonal and diel variability of meteorological conditions and source emissions. Strong relationships between OC and EC concentrations were found in winter and spring, indicating their common sources. The correction was lower in summer and autumn, implying the presence of other sources. The secondary organic carbon concentrations in PM 2.5 were estimated by using an EC-tracer method and the results showed that SOC concentrations were on average about 2.19 ± 1.55, 7.66 ± 5.89, 5.79 ± 3.51, and 3.43 ± 2.26 µgC·m −3 , accounting for 29.2%, 52.7%, 27.4%, and 30.5% of total organic carbon in summer, autumn, winter, and spring, respectively. This result demonstrated that the relative contribution of SOC to OC in PM 2.5 was only comparable to that of POC to OC in autumn, while in the other three seasons, OC was dominated by POC. Five transportation pathways that contributed to OC, EC, and PM 2.5 in Wanzhou were identified by trajectory clustering. The air masses with high PM 2.5 and carbonaceous concentrations originated from or passed over several industrial centers and urban areas in western and northwestern China. By contrast, the lowest PM 2.5 and carbonaceous concentrations observed were those coming from south and northeast areas, which brought in clean air masses to Wanzhou with a longer pathway. However, air trajectories from the southeast were the dominant trajectories, which indicated the influence of local sources by their short pathway.