Characteristics of Volatile Organic Compounds in the Pearl River Delta Region, China: Chemical Reactivity, Source, and Emission Regions

: Volatile organic compounds (VOCs) are important precursors of photochemical ozone and secondary organic aerosol (SOA). Here, hourly variations of ambient VOCs were monitored with an online system at an urban site (Panyu, PY) in the Pearl River Delta region during August–September of 2020 in order to identify reactive VOC species and major sources of VOCs, OH loss rate (L OH ), SOA formation potential (SOAFP), and corresponding emission source regions. The average concentration of VOCs at PY was 31.80 ± 20.82 ppbv during the campaign. The C 2 –C 5 alkanes, aromatics, and ≥ C 6 alkanes contributed for the majority of VOC, alkenes and aromatics showed the highest contribution to L OH and SOAFP. Further, m/p-xylene, propene, and toluene were found to be the top three most reactive anthropogenic VOC species, with respective contributions of 11.6%, 6.1%, and 5.8% to total L OH . Toluene, m/p-xylene, and o-xylene constituted a large fraction of calculated SOAFP. Seven major sources were identiﬁed by using positive matrix factorization model. Vehicle exhaust made the most signiﬁcant contribution to VOCs, followed by liqueﬁed petroleum gas and combustion sources. However, industrial-related sources (including industrial solvent use and industrial process emission) had the largest contribution to L OH and SOAFP. By combining source contribution with wind direction and wind speed, the regions of different sources were further identiﬁed. Based on high-resolution observation data during ozone pollution, this study clearly exhibits key reactive VOC species and the major emission regions of different VOC sources, and thus beneﬁts the accurate emission control of VOCs in the near future.


Field Sampling
The ambient VOC samples were taken at an urban site in Guangzhou located on the roof of the six-story building of Panyu (PY) middle school (Figure 1; 113.366°E, 22.913°N). The sampling site location was in the mixed residential and commercial area of PY district, Guangzhou, situated in the central PRD, approximately 26 km and 28 km southeast of central Guangzhou and Foshan, respectively, 47 km northeast of Jiangmen city, 42 km west of Dongguan city and 80 km northwest of downtown Shenzhen. The site was selected as not only a typical urban site, but also a representative regional site in the PRD.

Online Measurement of VOCs and Other Pollutants
Ozone pollution is more serious during autumn in the Pearl River Delta. From 17 August to 7 September 2020, thirteen episodes occurred when the maximum daily 8 h average concentration of O3 exceeded the National Ambient Air Quality class II standard (160 μg m −3 ) and the maximum hourly O3 level reached 281 μg m −3 . In order to explore the characteristics of VOCs during long-term ozone pollution, a total of 55 VOCs in ambient air, including 28 alkanes, 10 alkenes, 16 aromatics, and ethyne, were analyzed using an online system (GC-866; Chromatotec Inc., France) from 17 August to 7 September 2020 within a 1 h interval. The VOC samples were first collected at a flow rate of 40 mL min −1 for 5 min, then the VOCs were analyzed with two different gas chromatographs equipped with a flame ionization detector (GC-FID), one for C2-C6 compounds (AirmoVOC C2-C6; Chromatotec Inc., France) and another for C6-C12 compounds (AirmoVOC C6-C12; Chromatotec Inc., France). For the C2-C6 channel, the higher molecular weight compounds in samples were trapped with Carbotrap C and Carbotrap B and then the non-retained compounds were trapped at −8℃ by a cell with the Peltier effect. For the C6-C12 channel, the sample was passed through a fine tube and C6-C12 compounds were extracted with porous substances according to their affinity. The second trap was then rapidly heated to 220 °C and kept at that temperature for 240 s until the VOCs were transferred to the GC-FID system. The C2-C6 VOCs were separated using an ultimetal column (25 m × 0.55 mm I.D), while C6-C12 VOCs were separated using a capillary metallic column (MXT30 CE; 30 m × 0.28 mm × 1.0μm). Detailed information about the instruments is described elsewhere [47,48].
Other air pollutants including SO2, NO2, CO, and O3 were measured by 43i pulsed fluorescence, 42i chemiluminescence, 48i infrared radiation and the 49i UV photometric method using Thermo Environmental Instruments (Thermo Fisher Scientific, USA) during a campaign with a 1 h interval, respectively. The hourly concentrations of PM2.5 were measured by the β-absorption method using a BAM 1020 instrument (Met One Instrument, USA).

Quality Control and Quality Assurance
Stringent quality control and quality assurance measures were taken for the online system during the campaign. The calibration standards were generated by dynamically diluting the Photochemical Assessment Monitoring Station (PAMS, Spectra Gases Inc., New Jersey,USA) gas mixture. Then, the calibration curves were generated by analyzing above standards the same way as the filed samples and verified every two weeks. The responses of the GC-866 system were auto-calibrated every day with a 5 ppbv mixed standard in order to ensure that responses were lower than ± 10% of the calibration curve. The method detection limits (MDLs) for VOC species ranged from 0.01 to 0.56 ppbv.

The OH Loss Rate of VOCs
The LOH of VOCs represents the chemical reactivity of each VOC species, and is calculated by multiplying the mixing ratio level of VOCs by its OH radical rate constant, as shown in Equation (1): where LOHi represents the loss rates of i VOC species that react with OH radical, VOCi represents the concentration levels of i VOC species, and KOH i represents the rate constant for the reaction of the i VOC species with OH radical; the KOH i for 56 VOCs species were obtained from the results of Carter [49].

Secondary Organic Aerosol Formation Potential (SOAFP)
Some VOC species are important precursors of SOA, and the SOAFPs from VOC species were estimated in the present study as shown in Equation (2): where SOAFPi represents the SOA formed from i VOC species, VOCi represents the concentration levels of i VOC species in ambient air, Fvocri represents the fraction of reacted i VOC species, and FACi denotes the fractional aerosol coefficient for i VOC species [50][51][52].

Positive Matrix Factorization
PMF is a multivariate factor analysis tool that decomposes a matrix of sample data into two matrices, factor contributions (g) and factor profiles (f). The method is reviewed briefly here and described in greater detail elsewhere [53,54]. VOC sources were resolved according to the data matrix, X, which decomposes observed VOCs to g (factor contributions) and f (factor profiles) as represented in Equation (3). The PMF solution minimizes the objective function based on uncertainties, as shown in Equation (4).
The input files of PMF included the concentration of VOCs and user-provided uncertainty associated with the observed VOCs. Data values below the MDL were substituted with MDL/2; missing data values were substituted with median concentrations. If the concentration was less than or equal to the MDL provided, the uncertainty was calculated using the equation Unc = 5/6 MDL; if the concentration was greater than the MDL provided, the uncertainty was calculated as Unc = [(Error faction × mixing ratio) 2 +(MDL) 2 ] 1/2 . The number of factors in PMF was selected based on Q/Qexp, as shown in Section 3.5 [55]. Figure 2 shows the time series of observed O3, VOCs, NO2, PM2.5, as well as the meteorological parameters (temperature, humidity, wind speed, wind direction) at the PY site during the campaign. Overall, O3 showed a higher level from 21 August to 6 September, with maximum hourly mean concentrations above 160 μg m −3 , and the maximum hourly O3 level was 281 μg m −3 at 18:00 on 22 August. Furthermore, thirteen episode days (59%) occurred during the 22-measurement days with the maximum daily 8 h average concentration of O3 exceeding the National Ambient Air Quality class II standard of 160 μg m −3 . This demonstrates that O3 pollution was serious during August-September in the PRD region. Meanwhile, the synchronous increase of O3 and PM2.5 during 20-22 August may be due to the reaction of VOCs. During the campaign, the mixing ratio of VOCs ranged from 6.59 to 144.38 ppbv, averaged as 31.80 ± 20.82 ppbv. C2-C5 alkanes were the most abundant VOC species, with a mean concentration of 15.33 ± 9.66 ppbv, and amounted to 48.2% of the total measured VOCs during sampling ( Figure 3). The respective levels of aromatics, ≥C6 alkanes, alkenes, and alkyne were 7.33 ± 5.76, 4.60 ± 4.60, 3.03 ± 1.31, and 1.51 ± 0.81 ppbv, which contributed 23.0%, 14.5%, 9.5%, and 4.8% of the observed VOCs ( Figure 3). The mixing ratio of VOCs averaged 31.80 ± 20.82 ppbv in this study. Compared with the data reported in other studies from around the world (Table 1), the mixing ratio of VOCs in urban Guangzhou in this study was comparable with previously reported levels in Wuhan, China during January 2016 (32.07 ppbv) and Colorado from March-May 2015 (32.10 ppbv) [56,57] [44,64], possibly due to the implementation of pollution control strategies in the past five years. The percentage of C2-C5 alkanes in the present study is similar to that in Beijing, Tianjin, Liuzhou, Paris, and Mexico City [58,59,62,63,65]. The percentage of alkenes is lower than reported in most other Chinese cities [59,66], while the proportion of aromatic hydrocarbons is higher than in Tianjin, Chengdu, and Colorado [57,59,67]. The difference in the level and composition of VOCs among cities is related to the geographical conditions and the urban energy and industrial structure. In general, the concentration of VOCs in Guangzhou is at a higher level compared to worldwide, as the Panyu site in this study is located in the southern part of urban Guangzhou with higher vehicle emission intensity, and it is also surrounded by industrial cities such as Foshan, Zhongshan, and Dongguan, with more contribution from industry-related emissions. , and the sum of the top ten VOCs contributed 68.9% of the total VOCs. Among the above-mentioned VOCs, propane, i-butane, and nbutanes are the three most prominent species from liquefied petroleum gas (LPG) emission [68], toluene and m/p-xylene are mainly from industrial solvents and vehicle exhaust [37], i-pentane and n-pentane are typical tracers of oil evaporation [69], and ethyne, ethane, and ethylene are mainly obtained from the emissions of various combustion sources [70]. Consequently, VOCs in Guangzhou are closely related to the use of LPG and industrial solvents, vehicle exhaust, oil evaporation, and combustion sources. Figure 4 shows the diurnal variations in VOCs and their main components (C2-C5 alkanes, ≥C6 alkanes, alkenes, and alkyne). The mixing ratio of VOCs during the night was higher than during the day, with the highest at 8:00 and lowest at 15:00. From 21:00 to 6:00, the concentration of VOCs remained at 30-40 ppbv, then it increased and reached the daily maximum at 8:00 with 44.20 ± 26.92 ppbv. This result clearly indicates the influence of vehicle exhaust emissions during the morning rush hour. After 8:00, with the elevated temperature and planetary boundary layer height, the atmospheric diffusion conditions gradually improve, and more VOCs are consumed due to the occurrence of significant photochemical activity around noon, leading to a continuous decrease in VOCs from 8:00 to 15:00. After 15:00, the mixing ratios of VOCs gradually increased due to poor diffusion conditions and decreased photochemical activity. C2-C5 alkanes, ≥C6 alkanes, aromatics, and alkyne exhibited similar diurnal variation to that of VOCs, with daily maximum and minimum values appearing at 08:00 and 15:00, respectively, and the concentration during the day much lower than at night. Alkenes exhibited different diurnal patterns compared with VOCs. Owing to the influence of vehicle exhaust emissions during the morning rush hour, the mixing ratio of alkenes increased rapidly from 2.31 ± 1.32 ppbv at 06:00 to 3.84 ± 2.39 ppbv at 09:00, then gradually decreased to 2.53 ± 0.80 ppbv. Except for the period from 6:00 to 12:00, the mixing ratio of alkenes was maintained at 2.00-2.40 ppbv. The photochemical reactivity of alkenes is faster than those of alkanes and aromatics [38]; however, the concentration of alkenes did not decrease around noon, which indicates that there may be more alkene emissions during the day in Guangzhou or nearby regions.

LOH and SOAFP
VOCs are composed of a variety of species, and their atmospheric reactivity varies significantly. The reactivity of VOCs refers to the ability of VOCs to participate in atmospheric chemical reactions, and the reaction of VOCs with OH radical is the key factor affecting ozone formation in polluted atmosphere [38,71]. Therefore, in this study, the LOH was calculated, and the reactivity of VOC species was identified. Some VOC species are also important precursors of SOA; thus, the SOAFP was calculated and the important VOC species capable of forming SOA were identified. Figure 5 exhibits the correlation between VOCs and LOH. During the campaign, VOCs were significantly correlated with LOH (r 2 = 0.87), and LOH increased with increased VOC mixing ratio. The LOH/VOC ratio was 0.1673, as shown in Figure 5; furthermore, the average OH radical reaction rate constant of VOCs was calculated to be 6.25 × 10 −12 cm 3 molecule −1 s −1 , which was higher than that of toluene (5.63 × 10 −12 cm 3 molecule −1 s −1 ) [38]. This result indicates that the reactivity of the atmosphere was relatively strong during autumn in Guangzhou; however, it was lower than that of Shanghai during summer (8.11 × 10 −12 cm 3 molecule −1 s −1 ) [72]. LOH ranged from 1.96 to 28.34 s −1 during the campaign, with an average LOH of 7.71 ± 3.73 s −1 . Alkenes contributed 40.3% of LOH, followed by aromatics (36.5%), C2−C5 alkanes (12.0%), and ≥C6 alkanes (10.7%) (Figure 3b). Compared with the composition of VOCs, the percentages of alkenes and aromatics increased from 9.5% and 23.0% for VOCs to 40.3% and 36.5% for LOH, respectively. However, the percentages of C2-C5 alkanes and ≥C6 alkanes decreased, from 48.2% and 14.5% for VOCs to 12.0% and 10.7% for LOH, indicating that alkenes and aromatics contributed more to ozone formation; therefore, priority control of reactive alkenes and aromatics is key to ozone pollution control. However, it is noteworthy that 22.7% of total LOH was attributed to alkanes; even though the reactivity was relatively low, emission control of alkanes cannot be ignored.

Ratio Analysis
The ratios between different VOCs species have been widely used to assess the sources of VOCs. i-Pentane and n-pentane are the important alkanes. The i-Pentane/npentane (i-P/n-P) ratio was 2.4 in a tunnel experiment, and the ratio ranged from 1.5 to 2.6 for gasoline evaporation and 0.5 for diesel evaporation [69,73]. The ratio of i-P/n-P ranged from 0.5 to 3.4, and averaged 1.75 ± 0.6 during the campaign, as shown in Figure 6a, indicating that vehicle exhaust and oil evaporation (including gasoline and diesel evaporation) were responsible for the ambient pentanes. Ethyne is a typical tracer for combustion emissions; therefore, the ratios of VOCs species to ethyne reflected the impact of other sources relative to combustion [41,73]. The ratios of propane/ethyne and i-butane/ethyne were 2.6 ± 0.8 and 1.2 ± 0.5 in the present study, and these values were between the ratios reported in a previous study carried out in a suburban and urban site in Guangzhou in 2009; the ratio of n-butane/ethyne (2.4 ± 1.1) was higher here than in the earlier study [41]. Benzene is a Group I human carcinogen and its use as an industrial solvent has been banned. Benzene is mainly sourced from combustion sources such as vehicle exhaust and boilers [73,74]. Toluene in ambient air is significantly influenced by both industrial solvent use and vehicle exhaust [73]. Notably, the toluene/benzene (T/B) ratio is widely used to identify the major sources of aromatics. Previous studies have shown that a T/B < 1.0 indicates that the main sources of aromatics are biomass and coal burning [75] while 2.0 indicates vehicle emissions and tunnel study [73,76,77], different types of industrial and solvent use emissions have shown a higher T/B ratio [73]. Figure 6b shows that the T/B ratio at the Panyu site ranged from 2.1 to 52.2 and averaged 10.0 ± 7.3, which indicates the significant impact of industrial solvent emissions. This result is consistent with the results of research carried out in the upwind city of Dongguan, wherein it was reported that a T/B>5 was influenced by industrial solvent emissions [78,79]. Furthermore, benzene is significantly correlated with ethyne, which is the typical tracer of combustion sources, indicating the influence of combustion-related sources on ambient aromatics.

Source Identification
In this study, PMF was further used to identify the contributing sources of VOCs during the campaign. Thirty-one VOC species with high abundance and typical source tracers were used for PMF. The PMF was operated with a factor number increasing from 4 to 8. When the factor number was 7, the Q(Robust) and Q(True) were converged to 2156.08, and the PMF results well interpreted the measured VOCs, with a slope and r 2 value of 1.01 and 0.99, respectively. Therefore, the seven-factor solution was selected as the best solution from PMF.
The source profiles resolved by PMF are shown in Figure 7. Factor 1 was rich in C5-C7 alkanes; 2,3-dimethylbutane, 3-methylpentane, and 2,3-dimethylpentane were typical tracers of gasoline vehicle exhaust [58]; Higher carbon alkanes and aromatics including 3methylhexane, 2,2,4-trimethylpentane, and toluene were mainly from diesel vehicle exhaust [76]; thus, factor 1 was considered vehicle exhaust emission. Factor 2 was identified as LPG emission, which is distinguished by a high loading of the LPG tracers propane, ibutane, and n-butane [68]. Factor 3 showed dominance of i-pentane and n-pentane, and was believed to be oil evaporation emission [69]. Factor 4 was industrial solvent use emission, with higher proportions of ethylbenzene, m/p-xylene, and o-xylene; higher toluene to benzene (T/B) ratio was also resolved in factor 4, which is consistent with the influence of solvent use [40]. Factor 5 was regarded as industrial process emission, as it was distinguished by the presence of ethylene, propene, trans-2-butene, trans-2-pentene, and styrene; while ethylene and propene are important raw materials in chemical production [80], butenes are mainly obtained from process emissions [81], and styrene is one of the raw materials of resin and rubber. Source 6 was recognized as a combustion source, as it is associated with ethyne, ethane, propene, benzene, and some C6-C7 alkanes [73], which are markers of coal/biomass combustion. Factor 7 was identified as biogenic emission, as it consisted of a high proportion of isoprene, which is tracer of biogenic emission [82]. Based on the PMF results (Figure 8a), vehicle exhaust was the largest source of VOCs, contributing 24.6%, followed by LPG (15.8%), combustion sources (15.4%), industrial solvent use (14.8%), and oil evaporation (11.3%). Industrial process emissions and biogenic emissions contributed 9.7% and 8.4% of VOCs, respectively. Industrial-related sources, including industrial solvent use and industrial process emissions, together contributed 24.5% of VOCs.

Sources of LOH and SOAFP
The contribution of different sources to LOH and SOAFP was further calculated (Figure 8b, c). Surprisingly, biogenic emissions contributed the most to LOH at 20.7%, which was comparable with the percentage of isoprene to LOH (21.0%) shown in Section 3.3. Notably, in this study the ambient samples were collected during August and September, when biogenic emission intensity was the greatest and significantly higher than in other months. Vehicle exhaust was found to be the most important anthropogenic source, contributing 16.3% of LOH (Figure 8b), followed by industrial process emissions (16.2%), industrial solvent use (15.7%), combustion sources (14.8%), LPG (10.0%), and oil evaporation (6.3%). In general, vehicle exhaust emissions dominate LOH in urban cites. For instance, motor vehicle exhaust contributed for 50% and 19.4% of LOH in Pasadena and Wuhan [83,84], respectively, higher than the percentage of vehicle exhaust to LOH reported in the present study. In Houston, vehicle and petrochemical refinery emissions accounted for 42% of LOH [85], similar to the sum of vehicle exhaust and industrial-related emissions in urban Guangzhou (48.2%).
Industrial solvent use, combustion sources, and vehicle exhaust ranked as the first three most important sources of SOAFP, contributing 30.5%, 18.3%, and 18.1%, respectively (Figure 8c), followed by LPG (12.0%) and industrial process emissions (9.1%). Oil evaporation and biogenic emissions contributed a mere 6.2% and 5.8% of SOAPF, respectively. The SOAFP of industrial solvent use in this study is consistent with industrialized Ningbo [86], and also similar with the result that industrial sources contributed most to SOAFP based on emission inventory study [87]. Notably, the contribution of vehicle exhaust emissions to SOAFP was lower than industrial-related sources, probably due to SOAFP being limited in this study to C2-C12 nonmethane hydrocarbons (NMHCs). Moreover, chamber simulations revealed that intermediate-volatility organic compounds (IVOCs), in particular those emitted from diesel vehicles, act as important precursors for SOA [88,89].
In summary, vehicle exhaust was found to be the largest source of VOCs. However, industry-related sources (including industrial solvent use and industrial process emissions) had the largest contribution to LOH and SOAFP, followed by vehicle exhaust and combustion sources. Therefore, strengthening emission control based on reactivity instead of concentration should be considered in future policies.

Emission Regions of Different Sources
Furthermore, the emission region of different sources was analyzed by combining the wind direction and wind speed. This type of analysis can help to clarify source areas and transmission channels for future accurate control of different VOC sources (Figure 9).
The hotspot of vehicle exhaust emissions was mainly concentrated in the northerly direction with wind speed <1 m s −1 , possibly related to vehicle exhaust emission in central urban areas of Guangzhou and Panyu district under northerly winds. There were two peaks for LPG emission in the west-southwest and west-northwest with wind speed of 1.5 m s −1 , as well as a strip southeast of the Panyu site. For oil evaporation, the hotspot was concentrated in a southern direction with low wind speed (<0.5 m s −1 ). Similar to vehicle exhaust, industrial solvent use emissions were mainly centered on the north-northeast region with wind speed <1 m s −1 , and several hotspots were also observed with relatively high wind speed (2-3 m s −1 ), indicating transport from other cities such as Dongguan and Foshan, the most important manufacturing centers of furniture and commodity products in the PRD region. The industrial process emissions region was distributed in the northwest with wind speed of 0.5-2 m s −1 , indicating the joint influence of local emissions and regional transport in the northwest. A hotspot was observed in the northwest for combustion sources. Different from other sources, biogenic emissions were distributed in multiple directions with high wind speed, and mainly emitted from the surrounding forest-covered areas.

Conclusions
To explore the reactive VOC species and identify the major sources of VOCs, LOH, SOAFP, and corresponding emission regions in the urban PRD during severe ozone pollution, the ambient VOCs were monitored with a high-resolution online system in central PRD (PY site) from August to September 2020. The results showed that the average mixing ratio of VOCs during the campaign was (31.80 ± 20.82) ppbv. Furthermore, C2-C5 alkanes, aromatics, ≥C6 alkanes, alkenes, and alkyne contributed 48.2%, 23.0%, 14.5%, 9.5%, and 4.8% of the measured VOCs, respectively; propane, n-butane and toluene were the three most abundant VOC species.
The chemical reactivity of VOCs in Guangzhou was equivalent to that of toluene, with calculated average OH radical reaction rate constants of VOCs of 6.25 × 10 −12 cm 3 molecule −1 s −1 . Alkenes and aromatics showed the highest contribution to LOH, and the key active anthropogenic compounds were m/p-xylene, propene, and toluene. Aromatics attributed 93.1% of SOAFP, and toluene, m/p-xylene, o-xylene, ethylbenzene, and 1,2,3-trimethylbenzene were the most abundant species. Consequently, priority control of reactive alkenes and aromatics is key to ozone and PM2.5 pollution prevention and control.
The sources of VOCs were investigated using PMF. Vehicle exhaust, LPG, and combustion emission were three most important sources at the PY site, accounting for 24.6%, 15.8%, and 15.4% of observed VOCs, followed by industrial solvent use (14.8%), oil evaporation (11.3%), industrial process emissions (9.7%), and biogenic emissions (8.4%), respectively. Moreover, the LOH and SOAFP were mostly contributed by industry-related sources (including industrial solvent use and industrial process emissions), vehicle exhaust, and combustion sources, indicating that control of industry-related sources and vehicle exhaust emissions are key to reducing the chemical reactivity of VOCs in Guangzhou. The transport channels and emission regions of different sources were further identified by combining source contribution with wind direction and wind speed, and different hotspot areas for different sources were explored. Based on high resolution observed VOC data, this study demonstrated the key reactive VOC species and identified the major sources and corresponding emission regions. It is thus of benefit for the accurate emission control of VOCs in the PRD region, and aids in the formulation of an emission control strategy for VOCs in similar cities around the world.