Chemical and stable carbon isotopic composition of PM 2 . 5 from 1 on-road vehicle emissions in the PRD region and implication for 2 vehicle emission control policy 3 4

19 Vehicle emission is a major source of urban air pollution. In recent decade, the 20 Chinese government has introduced a range of policies to reduce the vehicle emission. 21 In order to understand the chemical characteristics of PM2.5 from on-road vehicle 22 emission in the Pearl River Delta (PRD) region and to evaluate the effectiveness of 23 control policies on vehicles emission, the emission factors of PM2.5 mass, elemental 24 carbon (EC), organic carbon (OC), water-soluble organic carbon (WSOC), 25 water-soluble inorganic ions (WSII), metal elements, organic compounds and stable 26 carbon isotopic composition were measured in the Zhujiang Tunnel of Guangzhou, 27 the PRD region of China in 2013. Emission factors of PM2.5 mass, OC, EC, and 28 WSOC were 92.4, 16.7, 16.4, and 1.31 mg vehicle km respectively. Emission 29 factors of WSII were 0.016 (F) ~ 4.17 (Cl) mg vehicle km, totally contributing 30 about 9.8% to the PM2.5 emission. The sum of 27 measured metal elements accounted 31 for 15.2% of the PM2.5 emission. Fe was the most abundant metal element, with an 32 emission factor of 3.91 mg vehicle km. Emission factors of organic compounds 33 including n-alkanes, PAHs, hopanes, and steranes were 91.9, 5.02, 32.0 and 7.59 μg 34 vehicle km, respectively. Stable carbon isotopic composition δC value was 35 measured and it was -25.0‰ on average. An isotopic fractionation of 3.2‰ was found 36 during fuel combustion. Compared with a previous study in Zhujiang Tunnel in year 37 2004, emission factors of PM2.5 mass, EC, OC, WSII except Cl, and organic 38 compounds decreased by 16.0-93.4%, which could be attributed to emission control 39 policy from 2004 to 2013. However, emission factors of most of the metal elements 40 increased significantly, which could be partially attributed to the changes in motor oil 41 additives and vehicle conditions. There are no mandatory national standards to limit 42 metal content from vehicle emission, which should be a concern of the government. A 43 snapshot of the 2013 characteristic emission of PM2.5 and its constituents from 44 on-road vehicular fleet in the PRD region retrieved from our study was found to be 45 useful for the assessment of past and future implementation of vehicle emission 46 control policy. 47 48


Introduction
Vehicle emission is a major source of urban air pollution and it accounts for approximately 14~ 50% of total fine particle mass in urban areas (Sheesley et al., 2007;Wang et al., 2008;Yu et al., 2013).The environmental and health effects of vehicle emission have been our concern during the last decades.Numerous studies have been conducted to characterize vehicular particulate matter (PM) emission in many countries, with respect to emission factors, chemical composition, and size distribution (Chiang and Huang, 2009;Laschober et al., 2004;Pio et al., 2013).The characteristics of vehicle emission in China were studied by tunnel experiments, dynamometer tests or road monitoring (He et al., 2008;Jin et al., 2014;Song et al., 2012).Because of the differences in fuel qualities, engine conditions, and operation practices, the PM emission from vehicles varied from region to region and time to time.
The Pearl River Delta (PRD) region is located in the southern coast of China, noted for its rapid urbanization and industrialization in the last few decades, has experienced serious atmospheric pollution.Vehicle emission accounts for approximately 25-30% of total fine PM in the PRD region (http://epaper.southcn.com/nfdaily/html/2014-01/03/content_7261687.htm).Peer reviewed papers had reported emission factors and chemical characteristic of PM 2.5 from vehicle emission in the PRD region, by means of tunnel studies in Zhujiang Tunnel (Guangzhou) and Wutong Tunnel (Shenzhen) (He et al., 2006;He et al., 2008;Huang et al., 2006b).However, the sampling in these studies was conducted in 2004.
During the past decade, the Environment Protect Agency of Guangdong Province revised the "Motor vehicle exhaust pollution prevention and control regulations of Guangdong Province" in 2008 and released the "PRD Regional Air Quality Management Plan" and "A Clean Air Plan" in 2010, to improve the relevant air quality through policies and measures.The emission standards for newly registered vehicles were tightened to China IV and the better quality of gasoline and diesel were supplied in 2013.Therefore, the characteristics of PM emission from vehicles in the PRD region might have changed throughout these years.
Tunnel experiments and chassis dynamometer tests were widely used to measure various pollutants emitted from vehicles (He et al., 2006;Heeb et al., 2003).However, dynamometer test has the defect that it can not account for vehicle fleet composition and emissions characteristics related to break and tire wear and re-suspension of road dust (Thorpe and Harrison, 2008).Tunnel studies have been demonstrated to be a suitable setup to measure PM emissions from on-road mixed fleets (Chiang and Huang, 2009;Laschober et al., 2004;Pio et al., 2013).
This study was carried out in a roadway tunnel located in the PRD region.We report here the emission factors of PM 2.5 mass, organic carbon (OC), elemental carbon (EC), water-soluble inorganic ions (WSII), metal elements, water-soluble organic carbon (WSOC), organic compounds and stable carbon isotope.WSOC has the potential to modify the hygroscopicity of particles, PM size and cloud condensation nuclei activities (Shulman et al., 1996), but it is often ignored in previous studies owing to the hydrophobic nature of the organic aerosol from primary vehicle emission.
Stable carbon isotope (δ 13 C) is very useful for tracing sources (Lopez-Veneroni, 2009;Widory, 2006), and it was also less reported for vehicular exhaust emissions (Ancelet et al., 2011;Widory, 2006).The objectives of this study are: (1) to obtain comprehensive information on the chemical and stable carbon isotopic composition of PM 2.5 emission from on-road vehicles in the PRD region; (2) to compare our results with the previous study conducted in the same tunnel in 2004; (3) to evaluate the effectiveness of the implementation of vehicle emission control policies from 2004 to 2013 in the PRD region.Although the fleet composition in this tunnel was probably different from the vehicle composition in the PRD region, it does not affect the conclusions in this paper.

Tunnel sampling
PM 2.5 samples were collected during August 10 to 14, 2013 from the roadway tunnel (Zhujiang Tunnel) located in Guangzhou City, China.It has two bores, each of which has three lanes with traffic in the same direction, as shown in Fig. 1.Two high-volume PM 2.5 samplers (GUV-15HBL1, Thermo, USA) were placed at a distance of 75 m from the entrance and 75 m from the exit, respectively.The vehicle speed in the Zhujiang Tunnel was 18 to 45 km h -1 , with an average vehicle speed of 33.4 km h -1 during the sampling.The air samples were drawn at about 1.13 m 3 min −1 through the quartz fiber filters (QFFs,20.3 cm×25.4 cm,Whatman).Other special devices such as diffusion denuders and foam plugs were not used due to the difficulties in applying these devices.Consequently, volatilization losses or adsorption artifacts may occur on the filter for semi-volatile organic compounds especially for the low molecular weight compounds because these compounds exhibit high volatility (Kavouras et al., 1999).However, the calculation of emission factors were based on the concentration differences between the exit and entrance of the tunnel, the potential losses or adsorption artifacts of semi-volatile organic compounds would be partly deducted.Field blank samples were also collected by loading filters into the samplers but without pulling air through.The ventilation system of the tunnel was turned off during the sampling period, thus the dispersion of air pollutants in the tunnel was mainly brought from the piston effect arising from the traffic flow.The sampled filters were wrapped with annealed aluminum foil and stored in a refrigerator at −40 °C till analysis.The meteorological parameters were synchronously recorded.
A video camera was placed at the exit to record the passing vehicles during the sampling periods.The videotapes were then used to determine the vehicle counts and to classify the vehicles into three categories, namely, diesel vehicles (DV) (heavy-duty trucks, light-duty trucks and large passenger cars), gasoline vehicles (GV) (small cars and motorcycles), and liquefied petroleum gas vehicles (LPGV) (bus and taxies).The average traffic density during sampling was 1797 per hour with DV, GV and LPGV proportion of 13.7 ± 2.7%, 59.8 ± 8.8% and 26.5 ± 7.9% respectively.
More details of the vehicle counts and meteorological conditions are summarized in Table S1 of the Supplement.

Chemical analysis
The PM 2.5 mass concentrations were determined gravimetrically by weighing the quartz filters before and after sampling in an electronic hygrothermostat for 24 h at 25 °C and 50% relative humidity (RH).Then, samples were analyzed for OC/EC, WSOC, WSII, metal elements, organic compounds and stable carbon isotope.The experimental methods of the chemical analysis are available in the Supplement.

Calculation of emission factor
Average emission factor (EF) was calculated for each sampling period on the basis of the concentration difference between the exit and entrance of the tunnel by the following equation (Handler et al., 2008): where EF is the emission factor of a species in unit of mg vehicle -1 km -1 , N is the number of vehicles passing through the tunnel, L is the distance between inlet and outlet sampling locations, C out and C in are the measured species concentration at the tunnel outlet and inlet, respectively, and V is the corresponding air volume calculated from the cross-sectional area of the tunnel, the average wind speed, and the sampling duration of each filter.The average concentrations of all measured species at the inlet and outlet sampling locations and the corresponding emission factors in this study are presented in Tables S2-4 of the Supplement.

PM 2.5 mass, OC, EC, WSOC, WSII, metal elements
The PM 2.5 mass emission factors ranged from 79.8 to 107 mg vehicle -1 km -1 , with an average of 92.4 ± 8.9 mg vehicle -1 km -1 .Average OC and EC emission factors were 16.7±1.9and 16.4±2.1 mg vehicle -1 km -1 , respectively, and they accounted for 19 ± 2% and 18 ± 2% of PM 2.5 mass emission.The ratio of OC to EC in the Zhujiang Tunnel ranged from 0.77 to 1.35, with an average of 1.03.Previous studies have shown that the OC/EC ratio is useful to separate gasoline engine emissions from diesel emissions.Higher values (> 2) are associated with GV and LPGV exhaust, and lower values (0.3 to ∼0.9) associated with DV exhaust (Cadle et al., 1999;Cheng et al., 2010;Gillies and Gertler, 2000).Therefore, the low OC/EC ratios in this study, which are closer to that from DV exhaust, indicate that diesel vehicles played an important role in the PM 2.5 emission although the proportion of DV was only 13.7% during the sampling.Additionally, it should be noted that emissions of EC from heavy duty trucks are expected to be relatively low under the low speed operating conditions in the tunnel (Kweon et al., 2002).Therefore, the ratio could be lower at the actual driving condition of vehicle fleet with a higher speed on the road.The concentration of WSOC in the inlet location was 6.21 μg m -3 (Table S2 of the Supplement) with a percentage of 31.1% of OC, which is close to that of ambient air (Ding et al., 2008;Ho et al., 2006).While in the outlet location, the concentration of WSOC was 8.00 μg m -3 , with a percentage of 17.9% of OC.The WSOC had been reported to contribute on average 20% to OC in the exit of Marseille roadway tunnel (El Haddad et al., 2009), in which background influence was included.The calculated emission factor of WSOC in this study ranged from 0.5 to 2.8 mg vehicle -1 km -1 , with an average of 1.31 mg vehicle -1 km -1 , which consisted of 7.84% of OC.Such a WSOC fraction is considerably lower than that previously measured for biomass burning particles (71%) (Mayol-Bracero et al., 2002).However, it could influence the hygroscopicity of particles and the formation of secondary aerosols (Ho et al., 2006;Rogge et al., 1993b;Weber et al., 2007) and is worthy of more attention and in-depth research.
PM2.5 mass was also obtained by summing OM, EC, geological component, sea salt, and major water soluble inorganic ions (NH 4 + , SO 4 2-, NO 3 -).OC was multiplied by 1.4 to estimate mass of OM (He et al., 2008).The geological component of 35 mg vehicle -1 km -1 was estimated based on the Al emission data as present in Table S3 of the Supplement.A typical road dust Al composition is 9% on average (Tiittanen et al., 1999).Sea salt of 9 mg vehicle -1 km -1 was estimated by Na assuming sea salt contains 32% of Na.Thus, the average PM2.5 reconstructed mass was 91.8% of the gravimetric value.This discrepancy can be attributed to the uncertainties in the weighing process, the estimation methods and uncalculated components.

Organic compounds
The average emission factors and abbreviated names of 67 individual organic compounds identified in the Zhujiang Tunnel, including n-alkanes, polycyclic aromatic hydrocarbons (PAHs), hopanes, and steranes are listed in Table S4 of the Supplement.These organic compounds accounted for 0.59% of the OM and 0.11% of the PM 2.5 mass emissions, The distributions of organic molecular markers associated with PM 2.5 are known to be source indicative despite of their small mass fractions (Schauer et al., 1996;Simoneit, 1986).n-Alkanes are an important class of organic compounds in atmospheric aerosols, and their homologue distribution may indicate different pollution sources (Rogge et al., 1993a).In this study, the n-alkane traces were dominated by C11-C36 with no odd-even carbon number predominance and the maximum was at C24, consistent with the characteristics of vehicle emissions reported by Simoneit (Simoneit, 1984(Simoneit, , 1985)).The emission factors of of individual n-alkanes were in the range of 0.22 (C13)~ 13.3 (C24) μg vehicle -1 km -1 (Table S4 of the Supplement).
There has been a worldwide concern to PAHs due to their known carcinogenic and mutagenic properties.PAHs are thought to be the result of incomplete combustion.
Totally 15 priority PAHs (the results of naphthalene have not been discussed in this study due to its low recovery) were identified and quantified.The emission factor of total PAHs varied from 4.56 to 5.54 μg vehicle -1 km -1 in this study.The emission factor of benzo[a]pyrene (BaP), which is often used as an indicator of PAHs and regarded by World Health Organization as a good index for whole PAHs carcinogenicity, was in the range of 0.37 to 0.46 μg vehicle -1 km -1 .The emission factors for other compounds ranged from 0.006 (acenaphthene) to 0.89 (pyrene) μg vehicle -1 km -1 (Table S4 of the Supplement).Pyrene was the most abundant compound, followed by chrysene (CHR), benzo[ghi]perylene (BghiP) and benz[a]anthracene (BaA), which is different from biomass burning and coal combustion (Huang et al., 2014;Shen et al., 2012).PAHs diagnostic ratios have been used as a tool for identifying pollution emission sources including ANT/(ANT+PHE), FLA/(FLA+PYR), BaA/(BaA+CHR), BbF/(BbF+BkF), IcdP/(IcdP+BghiP) and BaP/(BaP+BghiP) (Tobiszewski and Namiesnik, 2012;Yunker et al., 2002;Zhang et al., 2005).We summarized PAHs ratios mentioned above in Fig. 2 for three combustion sources including vehicle emission, biomass burning and coal combustion.
On the whole, the six ratios in this study are similar to the other tunnel experiments, though environmental conditions of tunnels are different to some extent.It is also suggested that the ratio of FLA/(FLA+PYR) and IcdP/(IcdP+BghiP) might be more suitable to distinguish vehicle emission from biomass burning and coal combustion.
Hopanes and steranes are known molecular markers of aerosol emissions from fossil fuel utilization (Simoneit, 1985).Rogge et al. (1993a) and Schauer et al. (1996) had shown that these petroleum biomarkers can be used to trace motor vehicle exhaust contributions to airborne PM in southern California atmosphere.Fourteen major hopanes homologues with emission factors ranging from 0.46 ~ 9.14 μg vehicle -1 km -1 and twelve steranes homologues ranging from 0.31 ~ 0.97 μg vehicle -1 km -1 were identified in this study.17α(H),21β(H)-hopane (HP30) was the most abundant component with the emission factor of 9.14 μg vehicle -1 km -1 .The emission factor of total hopanes was 32.0 μg vehicle -1 km -1 .Emissions of the S hopanes for the extended 17α(H),21β(H)-hopane homologues > C31 were always higher than those of the corresponding R pairs.All these characteristics of hopanes in the Zhujiang Tunnel are consistent with those in gasoline and diesel exhausts (Rogge et al., 1993a;Simoneit, 1985) and in other tunnel studies (see Fig. S1 of the Supplement).Emission factors of individual sterane ranged from 0.31 to 0.97 ng vehicle -1 km -1 , and the sum of their emission factors was 7.58 μg vehicle -1 km -1 .The most abundant homologue was C29αββ-stigmastane (20R) (29αββR), followed by 29αααS and 29αββS.

Stable carbon isotope
Stable carbon isotope analysis of vehicle emissions in Zhujiang Tunnel yielded δ 13 C values ranging from -25.5 to -24.7‰ with an average value of -25.0 ± 0.2‰, and is comparable to previously reported ranges of -29 to -24.6‰ (Table 1) for vehicular fuel emission.Generally, the variation in δ 13 C Fuel could affect the δ 13 C of hydrocarbons (Keppler et al., 2004;Yamada et al., 2009).In the PRD region, the δ 13 C value of gasoline was on average -28.6 ± 0.6‰, and the δ 13 C value of diesel was -27.8 ± 0.2‰, and small variation of fuel δ 13 C was observed (Hu et al., 2014).We calculated the isotopic differences between δ 13 C PM2.5 and δ 13 C Fuel , which represents the apparent isotopic fractionation occurring during fuel burning.It expressed as Δ 13 C (‰), and is defined by the following equation (Yamada et al., 2009).
Comparing the stable isotopic carbon value of vehicular fuel emission with other particulate emission sources (see Table 1), it is found that different emission sources showed different stable carbon isotopic composition.For total carbon in PM2.5 sample, δ 13 C (‰) of coal and fuel oil combustion are -23.9‰and -26.0‰ respectively, while that of vehicle emission is -25.9~-25.0‰.Obviously, the δ 13 C (‰) of vehicle emission is not significantly different from that of coal and fuel oil combustion.However, they are obviously different from other sources, like dust particle (-21~-18.4‰),C3 plant (-19.3~-13‰), and C4 plant (-34.7~-27‰).Therefore, δ 13 C might be used to distinguish the fossil fuel combustion from the other sources.

Comparison with previous study conducted in the same tunnel
To investigate the variation of chemical emission characteristics from vehicles in the PRD region over the past decade, we compared the chemical emission characteristics of this study with that of previous study (He et al., 2008) for the same tunnel in 2004 (see Figs 3 and 4).2).Additionally, comparing the fleet composition of 2013 with 2004 in Zhujiang Tunnel, we found that the proportion of DV and GV decreased while that of LPGV increased.LPG is a type of clean energy, and LPGV is known to emit much less PM mass than GV and DV, while DV emits the most PM mass (Allen et al., 2001;Myung et al., 2014;Yang et al., 2007).LPG could be combusted more completely than gasoline and diesel.Changes mentioned above contributed greatly to the decrease of emission factors of OC and EC (31.3% and 66.9%) and PM 2.5 mass (16.0%) from 2004 to 2013.However, the emissions of PM 2.5 mass, OC and EC are still quite higher than those measured in other countries (see Table S5 of the Supplement).The implication of these high emission levels is that both the fuel quality and engine technologies in the PRD region need to be further improved.
It is also found from Fig. 3 3).The emission levels of nitrate and ammonium were about one-tenth of those observed in 2004, possibly because NOx emission standard is tightened from 2004 to 2013 (Table 2) leading to the less production of ammonium nitrate.Emission factor of chloride is quite higher than that obtained from Zhujiang Tunnel in 2004 and other tunnels.Chloride was found up to 74 mg vehicle -1 km -1 in PM 10 in the Howell Tunnel, due to the application of salt to melt ice on roadways in the winter (Lough et al., 2005).However, it is not applicable in Guangzhou City.The good correlation between Cl -and Na + (r 2 =0.992) indicates the re-suspension of sea salt particles combined with vehicle emission PM might be a major source (He et al., 2008).
Emission factors of most of the metal elements increased in Zhujiang Tunnel from 2004 to 2013 except Cd and Pb.Na emission increased 3.18 mg vehicle -1 km -1 in 2013 than in 2004.Na correlated weakly with Cl -(r 2 =0.374) and Na + (r 2 =0.429).This indicates that Na emission had other sources and it was not only from the re-suspension of sea salt particles.The other four most abundant elements including Fe, Ca, Mg, K increased by one to three times.As they are major components of crustal material, resuspended road dust might result in their increase.However, the wind speed in 2013 was found to be higher but not significantly than that in 2004 (3.8 m s -1 in 2013 versus 3.0 m s -1 in 2004).This minor difference in wind speed could not account for the large increase.Furthermore, examination of the number of vehicles per hour in 2013 and 2004 suggests that there were fewer vehicles per hour in 2013.
Therefore, a more plausible explanation is that there was a lot more dust on the road in 2013.Other sources would cause the increased emissions of these elements, such as oil additive (Mg, Ca, Cu, Zn. (Cadle et al., 1997)) and the wear of engines (Fe) (Cadle et al., 1997;Garg et al., 2000), brakes and tires (Al, Fe,Cu, Mn, Cd, Ni, Pb and Zn.(Garg et al., 2000;Pio et al., 2013)).Additionally, emissions of Zn, Cu, Mn, Cr, Ni, V, As, Co, U, and Tl increased by 0.5 to 4.5 times.Although the sum of these elements did not exceed 0.5% of PM 2.5 mass, they are important for health effects.Lower emission factor of Pb (0.01±0.0007 mg vehicle -1 km -1 ) in 2013 than in 2004, could be a result of the phase out of leaded gasoline across China in the late 1990s.
Figure 4 shows comparison of organic compounds emission in Zhujiang Tunnel between 2004 and 2013.The n-alkane homologues exhibited a smooth hump-like distribution with the most abundance at C24, as shown in Fig. 4a.Such a distribution pattern was similar to patterns observed in Zhujiang Tunnel 2004.However, there are some differences.Firstly, the highest abundant n-alkane shifted from C23 in 2004 to C24 in 2013.This difference might be explained by the shift of gas-particle partitioning as alkanes of < C26 are semi-volatile.However, a significant T-test (p = 0.14) showed that the temperatures were not significant different between this study (33.0 ± 2.3℃) and that in 2004 (31.8 ± 1.0 ℃).Thus the differences due to different C max between this study and the study in 2004 can not be regarded as a result of temperature differences.Furthermore, C max was found to be C24 in every test of this study, although the temperature ranged from 28.6 to 36.1 ℃.It was reported that the n-alkane in the highest abundance was C20 for DV and C25 or C26 for GV in dynamometer tests (Rogge et al., 1993a;Schauer et al., 1999Schauer et al., , 2002)) S6 of the Supplement).This could be attributed to the variation of fleet composition between 2004 and 2013.PAHs emitted from LPGV are about one-third of that from GV (Yang et al., 2007), while DV emit more PAHs than GV (Phuleria et al., 2006).Therefore, higher proportion of LPGV and lower proportion of DV resulted in the lower emission factor of PAHs in 2013 than that in 2004.Emission factors of hopanes also decreased from 2004 to 2013, the percentage of decrease ranged from 56.2% to 68.7%.However, the distributions of hopane series derived from different tunnel studies were very similar (see Fig. S2 of the Supplement).This seems to suggest that the hopane emission characteristics might be independent of the fleet composition.This is a reasonable result given that hopanes originate from the lubricating oil used in DV, GV and LPGV rather than from the fuel (He et al., 2008;Phuleria et al., 2006).Owing to more units in heavy-duty vehicle need lubrication, emission factors of hopanes attributable to heavy-duty vehicle were higher than that to light-duty vehicle (Phuleria et al., 2006).Reduction of the proportion of heavy-duty vehicle (bus, heavy-duty truck, large passenger cars) proportion in fleet composition in 2013 (11.3%) compared that in 2004 (20%) might be the reason that emission factors of hopanes decreased.

Implication for vehicle emission control policy
Vehicle emission control strategies and policies adopted by Guangdong Province can be classified as emission control on vehicles and fuel quality improvements or alternative fuel utilization.The PM emission standards for newly registered vehicles were tightened from China II in 2004 to China IV in 2013 (Table 2).The reduction of on road high PM emitting vehicles, the phasing in of lower PM emitting vehicles, and more environmental friendly vehicles on road with better advanced engines following the implementation of these emission standards were effective for decreasing PM emission.Emission factors of PM decreased by 16% from 2004 to 2013.Also for NOx, the emission limit was reduced to about half from 2004 to 2013.This change in emission standard which limit NOx emission was a major factor for emission factors of nitrate and ammonium decreased by about 90%.On the other hand, National Standard has been revised several times to improve fuel quality to adapt to stringent vehicle emission standards (Table 3).(Wu et al., 2012)) increased by 25.2% from 2004 to 2013 (Fig. 5b).Consequently, we have demonstrated that more stringent emission standards and higher quality of fuel or more utilization of clean fuels will be necessary to offset the impacts on the growth in vehicle population and to improve air quality in the PRD region.Additionally, owing to no mandatory national standards to limit metal content from vehicle emission, the emission of most metal increased from 2004 to 2013 (Fig. 3 and 5b).In China, heavy metals, including As, Cr, Cu, Ni and Tl, had been listed as key substances that should be preferentially monitored in the atmospheric environment (SEPA, 2003).These increases of metal elements should raise the awareness of the government due to their health concern.BaP/(BaP+BghiP) and IcdP/(IcdP+BghiP) ratios for three source emissions.The vehicle emission (VE) composition is from data collected in roadway tunnel((this study), (He et al., 2008), (Ancelet et al., 2011), (He et al., 2006), (Ho et al., 2009), (Oda et al., 2001).The biomass burning profiles are obtained from 9 straws (Shen et al., 2011), 26 firewood (Shen et al., 2012), 3 plant leaves and branches (Sheesley et al., 2003) and 2 biomass briquettes burning (Sheesley et al., 2003).The coal combustion profiles are selected from 5 coals imitate combustion (Shen et al., 2011)and main coal-mining regions in China (Zhang et al., 2008Zhang et al., 2008).

Conclusions
Fig. 3 shows that PM 2.5 mass, OC, EC decreased significantly from 2004 to 2013.The reason can be partly attributed to the implementation of pollution control measures for Chinese vehicle emission.During this 9-year period, vehicle emission standard have raised two levels (from China II in 2004 to China IV in 2013) (Table to 2013.Improvement of fuel quality resulted in sulfate emission factor decrease from 3.18 to 0.61 mg vehicle -1 km -1 , since the amount of sulfur in fuel is slashed by 81.5~ 95 % in China IV (2013) when compared that in China II (2004) (Table PM 2.5 samples were collected between August 10 and 14, 2013 in Guangzhou Zhujiang Tunnel to acquire a comprehensive snapshot of the chemical characteristics of vehicle emission.The average emission factors of PM 2.5 mass, EC, OC, WSOC, WSII, metal elements, organic compounds and stable carbon isotope were measured.Stable carbon isotope δ 13 C value indicates an isotopic fractionation of 3.2‰ occurred during fuel combustion.Compared with a previous study in Zhujiang Tunnel in year 2004, emission factors of PM 2.5 mass, EC, OC, and major WSII decreased due to control policy induced changes throughout the nine years from 2004 to 2013, that is, change of fleet composition, implementation of more stringent gasoline and diesel emission standards, improvement in fuel quality, and clean fuel used for taxi and buses.The shift in n-alkanes distribution and decreased PAHs emission were due to the lower proportion of DV in 2013 than in 2004, and the decrease in emission factor of hopanes were due to the reduction of proportion of heavy duty vehicles.Our study shows that control polices for vehicles emission by the government were effective to decrease the emission factors of PM 2.5 , EC, OC, and WSII from on-road vehicular fleets.However, the increase in emission of most metal elements should raise the awareness of the government, since metal elements, especially heavy metals could affect human health.Also, to offset the impacts in the growth of vehicle population and to improve air quality in the PRD region, more stringent emission and aggressive control policies are necessary.

Fig. 4 .
Fig. 4. Comparison of organic compounds emissions in the Zhujiang Tunnel sampling

Fig. 5 .
Fig. 5. (a) Growth in total vehicle population in Guangzhou during 2004-2013.(b) . As the emissions collected in tunnel studies present a composite result of emissions from a mixed vehicle fleet, the lower fraction of DV in 2013 was more likely the cause for the shift of the C max .Secondly, emission factors of C16-C26 in 2013 were quite lower than those in 2004, while this trend reversed gradually after C27.Emission factors of the PAHs decreased by 67.6% ~ 93.4%.BaP equivalents (BaPeq) emission factors decreased by 88.1% from 2004 to 2013 (Table In general, our results suggest that these strategies are effective to reduce emission factors of PM 2.5 mass, as well as OC, EC, WSII, and organic compounds in PM 2.5 .However, the total vehicle population increased year by year.As shown in Fig.5a, the total vehicle population increased by 49.1% from 2004 to 2013.Total emission of vehicle exhaust of PM 2.5 mass (calculated as emission factors multiply by annual average driving distance per car and vehicle population