Variations of Particle Size Distribution , Black Carbon , and Brown Carbon during a Severe Winter Pollution Event over Xi ’ an , China

Real-time particulate matter (PM) size distributions, 4-hour time resolution, PM2.5, carbonaceous materials, and their optical properties were measured during a severe pollution event in Xi’an, China. High PM2.5/PM10 ratios were observed on both pollution (0.83) and non-pollution (0.73) days, emphasizing the abundance of fine particles during sampling days. The particle number (PN) first peaked with a wide size range (30–100 nm) before morning rush hours (approximately 01:00–05:00) on pollution and non-pollution days, demonstrating that PN was governed by the accumulation of freshly emitted diesel particles and characterized by distinct aerosol condensation growth. By contrast, the second peak time and size range differed between pollution and non-pollution days because of different formation mechanisms. The lightabsorbing coefficients of both black carbon (BC, babs-880nm, BC) and brown carbon (BrC, babs-370nm, BrC) were high on pollution days and decreased to approximately half of those values on non-pollution days, indicating that the degree of light absorption is reduced by rain. The diurnal variation in babs-880nm, BC pollution peaked with traffic on January 1 and 2. By contrast, it remained in relatively stable and high ranges (120–160 Mm) in the second period (January 3–5) without traffic peaks, illustrating that the dominant sources changed even during the same pollution period. High values of both babs-370nm, BrC and babs-880nm, BC coincided in the afternoon and evening due to emissions from primary sources, and abundant aqueous secondary organic carbon, respectively. A highly variable mass absorption coefficient of BrC also indicated the variety of fuel combustion sources of primary BrC in Xi’an.


INTRODUCTION
Urban pollution is an atmospheric phenomenon of high particulate matter (PM) loadings that leads to substantial visibility impairment at a relative humidity (RH) of less than 90% (Hyslop, 2009).Typically, the levels of PMs, including organic aerosol (OA) and black carbon (BC), and precursor gases increase substantially on pollution days (Odman et al., 2009;Zheng et al., 2015;Fang et al., 2017).Severe pollution has attracted considerable scientific interest because of its effects on public health and climate change (Pope et al., 2002;Nel, 2005;Law and Stohl, 2007;Peplow et al., 2014;Von Schneidemesser et al., 2015).Given the effect of urban air pollution, most studies have focused on examining the concentrations and chemical compositions of gaseous pollutants and PMs.However, studies have confirmed that PM size is a key indicator of particle formation and growth (Ketzel et al., 2004;Liu et al., 2008;Yue et al., 2016).Hitchins et al. (2000) indicated that 50% of total particle number (PN) concentrations in the 20-500-nm range decreased with an increase in distance from pollution sources.Different size distributions were found in PM chemical species and can be used to classify the sources and variations of particles with the PN (Liu et al., 2008;Ma et al., 2012;Tian et al., 2014).For instance, both nucleation and Aitken mode particles (3-130 nm) indicate fresh emissions, such as emissions from gasoline and diesel engines (Zhang and Wexler, 2004;Shi et al., 2010), whereas accumulation mode and larger particles mainly form from a series of chemical transformations (Zhang et al., 2004).Thus, investigating how particle size and number change during pollution periods can be beneficial.
Current climate models and field studies indicate that aerosol light absorption is a critical component in climate forcing (Bond and Bergstrom, 2006), because light-absorbing carbonaceous (LAC) particles are not only the essential components of PM (Ramanathan and Carmichael, 2008) but also the dominant absorbers of solar radiation in the atmosphere.In airborne carbonaceous species, two types of LACs are present: BC (at visible and infrared wavelengths) and brown carbon (BrC; at near-ultraviolet wavelengths).
Both change Earth's radiative balance and drastically reduce visibility.Primary or pure BC exhibits a graphitelike structure with small-sized (10-50 nm) spherules and is a primary pollutant with high chemical stability emitted mainly from internal combustion (Wentzel et al., 2003;Guha et al., 2015;Sarkar et al., 2015).However, pure BC is rarely found in the atmosphere and agglomerates rapidly after emission by adsorbing organic and inorganic vapors through typical aging processes.This leads to complex physical and light absorption properties (Kotzick and Niessner, 1999;Levitt et al., 2007;Liu and Smallwood, 2011;Lin et al., 2016;Zhan et al., 2016;Zhou et al., 2018).BC and BrC have entirely different morphologies, optical and chemical properties, and emission sources.In contrast to BC, BrC is large with a diameter of 100-400 nm, and its fraction may contain hundreds of organic compounds with varying atmospheric behaviors (Andreae and Gelencsér, 2006;Alexander et al., 2008).
Xi'an (34°16′N, 108°54′E, 400-m above sea level) is a megacity located in the center of Northwest China with a population density of 870 per km 2 .Because of its geographical location in the Guanzhong Plain, Xi'an is surrounded by the Qinling Mountains in south and the Loess Plateau in north and experiences frequently stagnant conditions; the city's air pollution can be easily trapped and exacerbated (Shen et al., 2009(Shen et al., , 2014)).According to atmospheric visibility data, airborne emissions in Xi'an have decreased over the last 10 years (Shen et al., 2016); however, winter pollution remains severe.In this study, a high-resolution analysis of particle size distribution, gas precursors, PM levels, and their carbon species was conducted.In addition, aerosol properties, such as concentration, size distribution, and light absorption, were characterized and compared between pollution and nonpollution days.

Observational Site
We performed field observations in a representative site in the southeast area of downtown Xi'an (34°16′N, 108°54′E; Fig. 1).Both online and offline instruments were located on a rooftop (approximately 15-m high) at Xi'an Jiaotong University that is exposed to complex emission sources (Shen et al., 2012;Zhang et al., 2015).Residential areas and the campus of Xi'an Jiaotong University are to the north and east, respectively.Two heavily used roads, the South Second Ring Road and Xingqing Road, are to the south and west, respectively.Many diesel-powered trucks and buses ply on these roadways at night.Xi'an, as well as other cities in northern China, experienced heavy pollution in winter in recent years and suffered through the longest stretch of stifling air pollution ever recorded in the country.The poor-quality air during this stretch consisted of exceedingly high concentrations of particles less than 2.5 µm in diameter (PM 2.5 , surpassed 500 µg m -3 ), and this stretch ended through precipitation and high winds 1 week later.According to the ambient air quality standards (AAQS) of China (GB3095-2012: http://kjs.mep.gov.cn/hjbhbz/bzwb/dqhjbh/qhjzlbz/201203/t20120302 _224165.htm)and the measured data in this study, pollution day was defined as that the 24-h average concentrations of both PM 2.5 and PM 10 were twice times higher than the national AAQS-Grade II value (PM 2.5 , 75 µg m -3 ; PM 10 , 150 µg•m -3 ).In this study, two episodes are identified and discussed during our sampling days: January 1-5, 2017 (pollution days), and January 6, 2017 (non-pollution day).

Time-integrated PM 2.5 Collection and Carbonaceous Species Analysis
A total of 42 PM 2.5 samples with time resolution of 4 hours were collected on Whatman 47 quartz filters (Whatman Inc., Maidstone, UK) by using a PQ200 ambient air particulate sampler (BGI Inc., USA) at a flow rate of 16.7 L min -1 .The filter samples were equilibrated for 24 hours at 20-23°C in a chamber at relative humidity (RH) of 35% and 45% before and after sample collection, respectively.They were weighed at least three times on a high-precision (± 1 µg) microbalance (ME-5, Sartorius Inc., Germany) to determine PM mass.After weighing, all the samples were stored in a freezer at approximately 20°C to prevent the evaporation of volatile compounds until analysis (Shen et al., 2010).
A 0.5-cm 2 punch of each sample was analyzed for elemental carbon (EC) and organic carbon (OC) of PM 2.5 following the Interagency Monitoring of Protected Visual Environments thermal and optical reflectance protocol by using a DRI Model 2001 thermal and optical carbon analyzer (Atmoslytic Inc., California, USA).OC fractions (OC1, OC2, OC3, and OC4 at 120°C, 250°C, 450°C, and 550°C, respectively, in a helium atmosphere) and EC fractions (EC1, EC2, and EC3 at 550°C, 700°C, and 800°C, respectively, in 2% oxygen and 98% helium) were also collected.During volatilization of OC, a part of the OC was converted pyrolytically to EC (this fraction of OC was called OP) (Chow et al., 2005).Therefore, OC is the sum of OC1, OC2, OC3, OC4, and OP, and EC is the sum of EC1, EC2, and EC3 minus OP.Additional quality assurance procedures are described in detail by Cao et al. (2005).

Collection of Particle Size Distribution Data
The size distribution (18.1-532.5 nm) and concentration of aerosol particles were continuously monitored using a scanning mobility particle sizer (SMPS) during sampling periods.The SMPS consists of a single-stage impactor (with a cutoff diameter of 0.5-0.7 µm), a differential mobility analyzer (Model 3082), and a condensation particle counter (Model 3775).These particle sizers are routinely used to accurately measure sizes of nanoparticles suspended in liquids.The aerosol flow rate of 3.3 L min -1 was carefully selected to be sufficiently high to remove large particles from aerosols in the impactor but sufficiently low to provide the residence time necessary to measure and separate particles.This time consisted of a 30-s scan, 6-s retrace, and 10-s purge (Chen et al., 2016).A more detailed description of the SMPS is provided by Gulijk et al. (2004).

BC and BrC Measurements
The concentration and optical parameters of BC were determined using a seven-wavelength Aethalometer-31 (Model-AE31, Magee Scientific Inc., USA).The flow rate of AE31 was calibrated prior to deployment.The sampling air flow rate was 5.0 L min -1 .AE31 was programmed to automatically determine aerosol light absorption at seven wavelengths (370, 470, 520, 590, 660, 880, and 950 nm) in µg m -3 at 5-min intervals over the sampling period from a location near PM 2.5 samplers.The accuracy of data from AE31 is affected by two factors: a nonlinear response as loading levels on the filter media increase (loading factor) and the effects of light scattering by fiber filter substrates (Collaud Coen et al., 2010).These factors were considered and corrected in the study of Shen et al. (2017).
The light absorption coefficient (b abs ) is the most important parameter for BC and BrC determination.The absorption at 880 nm is assumed to be contributed only by BC, whereas absorption at a lower wavelength (370 nm) is assumed to be contributed by both BC and BrC.Therefore, b abs at 880 nm refers to b abs-880nm, BC .The BrC absorption was derived from the filter-based aerosol absorption spectrum analysis by subtracting BC absorption (Olson et al., 2015).First, we projected the absorption coefficient (b abs ) of BC at 880 nm to full spectrum (370, 470, 520, 590, 660, 880, 950 nm) using AAE = 1.Then, the b abs-370nm, BC is subtracted from the b abs values at 370 nm (b abs-370nm ) to estimate the b abs-370nm, BrC .The detailed descriptions can be found in Olson et al. (2015) and Zhang et al. (2017).
The value of Absorption Ångström exponent (AAE), referred to as a power exponent of wavelength, is nearly constant (~1) for BC but is 1.2 or much higher for BrC.The relationship between wavelength-dependent AAE and b abs of BC and BrC were described as: b abs = K × λ -AAE (here K refers to a constant value and λ denotes wavelength of BrC), and the AAE _Full spectrum was determined by statistical regression fitting the b abs data covering the UV to the near-IR ranges (includes 370, 470, 520, 590, 660, 880, and 950 nm) with a power law equation.
The mass absorption coefficient (MAC) is a primary parameter used to characterize the optical properties of BC.The MAC can build the relationship between optical properties and mass concentrations of BC and is expressed in m 2 g -1 .All BC data were normalized with EC data from the same station to produce EC-equivalent BC values.BC calculation used site-specific MAC values.The MAC-BrC (for aqueous BrC) was calculated by dividing b abs-370nm, BrC by the OC filter-based concentration (Olson et al., 2015).We can calculate MAC-BrC and MAC-BC using the following equations: Both EC and OC were obtained from the DRI Model 2001 thermal and optical carbon analyzer mentioned in section 2.2.

Online Aerosol, Trace Gases, and Meteorology Measurement
Hourly concentrations of nitrogen dioxide (NO 2 ), sulfur dioxide (SO 2 ), and ozone (O 3 ) gases and time-integrated PM 2.5 and PM 10 were collected from the data center of the Ministry of Environmental Protection of China (http://datacente r.mep.gov.cn).Meteorological parameters, including temperature, RH, visibility, and wind speed/direction (WS/WD), were obtained from the Shaanxi Meteorological Bureau at the "Xingqing residential area" meteorological station located approximately 3 km north of the sampling site.

Brief Introduction of Severe Pollution Days
The daily average PM level, along with the meteorological parameters (temperature, RH, WS, and visibility), between January 1, 2017, and January 6, 2017, are listed in Table 1.The daily mean visibility on pollution days ranged from 1.1 to 1.6 km and improved to 3.1 km on non-pollution days.The visibility in Xi'an was still much lower than that in other Chinese megacities during winter pollution periods (e.g., 3.3 km in Guangzhou (Tan et al., 2009) and 2.6 km in Beijing (Yang et al., 2015)), confirming that this pollution period was particularly severe.Fig. 2 depicts two peaks of diurnal variations of visibility observed during the sampling periods.The first peak was 3 km in the early morning of January 2, and this peak was heavily influenced by the wind which originated from northeastern directions (WD: 30-80°) was high (> 3.5 m s -1 , Fig. 2).The second peak was 3.5 km in the afternoon of January 5 and was mainly caused by the appearance of wet depositions, the rising PBL, and high WS (> 4 m s -1 ) from northeastern direction (WD: 40-90°).
On pollution days, the average concentrations of PM 2.5 and PM 10 were 396.2 and 476.7 µg m -3 , respectively; these were 5.2 and 3.2 times higher than the Chinese National Ambient Air Quality Standard (GB3095-2012) of 75 for PM 2.5 and 150 µg m -3 for PM 10 (An et al., 2013).Low WS values (< 2 m s -1 ) indicated that high PM loadings were accompanied by poor diffusion conditions.High PM levels always corresponded to low visibility under a high RH (approximately 86%), implying that the hygroscopic growth of PM-soluble components leads to substantial visibility degradation (Cheng et al., 2013;Liu et al., 2013).By contrast, the mean levels of PM 2.5 and PM 10 on nonpollution days were only half of those on pollution days.The mean PM 2.5 /PM 10 ratio remained high on both pollution and non-pollution days (0.83 vs. 0.73), indicating that airborne particles were mainly present as PM 2.5 during sampling days (Fig. 3(a)).
The average mass concentration of SO 2 on pollution days was 43.4 µg m -3 , which was 1.6 times higher than that on non-pollution days.The mean concentrations of typical vehicle emissions, namely NO 2 , carbon monoxide (CO), and BC, were 111.7, 4089, and 13.7 µg m -3 on pollution days compared with 53.1, 3277, and 7.8 µg m -3 on nonpollution days.Meanwhile, these traffic markers showed similar diurnal patterns on pollution days, peaking during the heavily polluted time periods of morning (09:00-11:00) and evening (19:00-21:00) rush hours.After rain (05:00), the concentrations of NO 2 , CO, and BC rapidly decreased on January 6; however, the concentration of BC declined faster than that of CO, indicating that BC in Xi'an was coated by more hydrophilic groups and was more susceptible to wet depositions near the end of the heavy pollution (Dalirian et al., 2017).The O 3 concentration showed large daily variability, varying from 5.3 to 22.4 µg m -3 with an average value of 14.2 µg m -3 ; however, it remained lower than that during other seasons (Wang et al., 2012), indicating weak atmospheric oxidation capacity on pollution days.

Number and Size Distribution
Fig. 4 shows the comparison of size distributions (ranging from 18.5 to 532.5 nm) and mean concentrations between pollution and non-pollution days in Xi'an.The number of pollution particles peaked at 50-70 nm and can be identified as Aitken mode particles, which are slightly larger than the size of freshly emitted particles (approximately 30 nm), emphasizing that PM pollution was heavily influenced by fresh emissions and grew under stable atmospheric conditions (Zhang et al., 2017).The mode of non-pollution particles was between 100 and 170 nm, and half of them were in the Table 1.Meteorological parameters and online PM concentrations during sampling days.
During pollution days, diurnal PN concentrations peaked before morning rush hours (approximately 01:00-05:00) at a wider size range (30-100 nm) mainly because of the accumulation of freshly emitted particles and the condensation growth of new particles from diesel vehicles.This demonstrated that freshly emitted diesel particles during this time contributed substantially to the abundant PN at fixed size range (around 100 nm), providing some justification for the "Government Strengthening Management on Vehicle Exhaust" in Xi'an.During the second peak of pollution PN, both Aitken and accumulation modes were observed in the afternoon (15:00-18:00).This peak was not coincided with the variations of fresh vehicle emissions such as CO, NO 2 , and BC.In fact, low O 3 level implied the weak atmospheric oxidation.Therefore, this peak should mainly form from the condensation process under the conditions of relatively highly pre-existing PM levels (Maricq, 2007;Gao et al., 2009;Perez et al., 2010).
The diurnal patterns of non-pollution PN distribution were also observed in Fig. 4(b2).The two peaks occurred during early morning (02:00-04:00) and evening rush hours (18:00-20:00).In the first peak, the size range was similar to pollution particles, but the time interval of the PN was shorter.By contrast, the evening peak of nonpollution PM was different from that of pollution PM and appeared at a larger size range (80-200 nm).This was caused by the rapid growth of intensive traffic particles and condensation mechanisms under a high RH (Zhang et al., 2017).

Time-integrated Offline PM 2.5 Carbonaceous Concentrations
The total carbon (TC) concentration was calculated as 40.8 µg m -3 on pollution days and 17.2 µg m -3 on nonpollution days.TC, which includes EC and OC, constitutes a substantial fraction of PM 2.5 mass (24.4% on pollution days and 23.1% on non-pollution days) and heavily influences ambient optical properties (Andreae and Gelencsér, 2006;Bond and Bergstrom, 2006).The variations of TC/PM 2.5 changed slightly during sampling days (Fig. 5(a)).These phenomena emphasized that carbon materials were key constituents of PM 2.5 , and their sources appeared to be stable during sampling days.A strong correlation (R = 0.8, P < 0.001) was observed between OC and EC, indicating that TC maintains a fairly constant level as a primary pollutant in Xi'an (Cao et al., 2007).The time-integrated percentages of the eight carbon fractions are listed in Fig. 5(b).During sampling days, EC1, OC3, and OC4 were the most abundant carbon fractions and contributed approximately 60% of TC.This emphasizes the contribution from coal burning (Cao et al., 2005;Shen et al., 2010).The abundance of gasoline emissions led to high levels of OC1 and OC2, but low EC2 and EC3 indicated fewer diesel emissions.Additionally, the OC/EC ratios in this study varied from 3.0 to 5.8, indicating the presence of secondary organic carbon (SOC) (Zhang et al., 2015(Zhang et al., , 2017)).In addition, the substantial OC2 and OC3 concentrations were always associated with SOC (Gu et al., 2010).SOC concentration was estimated based on the EC-tracer method in Turpin and Huntzicker (1995).Thus, the estimation of SOC occupied approximately 30% of OC during severe pollution periods.

Characteristics of LAC Optical Properties
Earlier studies have defined LACs as pure LAC aerosols that include the strongly light-absorbing refractory BC particles at 880 nm and the partial OAs of BrC, which can preferentially absorb light in near-ultraviolet and blue wavelengths (Andreae and Gelencsér, 2006;Habib et al., 2008;Kirchstetter and Thatcher, 2012;Lei et al., 2018).Since AAE is a highly sensitive approach to classify the presence of BrC, the average AAE _Full spectrum values during pollution and non-pollution periods were 1.22 ± 0.02 and 1.30 ± 0.05, emphasizing the strong presence of BrC during our sampling periods in Xi'an.The average values of b abs-880nm, BC and b abs-370nm, BrC were 119.9 and 76.3 Mm -1 for pollution day PM 2.5 and 66.2 and 49.4 Mm -1 for non-pollution day PM 2 , respectively (Fig. 5(c)).This illustrates the abundance of BC and BrC particles.The changes in b abs-880nm, BC experienced three periods.In the initial pollution period (January 1-2), the b abs-880nm, BC showed a large variability and peaked at the morning (07:00-10:00) and evening (18:00-21:00) rush hours, emphasizing that b abs-880nm, BC was more sensitive to traffic density at the beginning of the severe pollution.During the midterm pollution period of late January 3 to January 5, b abs-880nm, BC rose sharply on the morning of January 3 and then displayed relatively stable and high ranges (120-160 Mm -1 ).These phenomena indicated that the high b abs-880nm, BC values during the second pollution period were mainly influenced by the accumulation of multiple primary sources (i.e., industrial, heating, and traffic emissions) under stable meteorological conditions and were less associated with the variations of traffic flow.The NO 2 /SO 2 mixing ratio can be a useful tool to examine the relative contribution of vehicular sources.The emission ratio of NO 2 to SO 2 peaked at the high level of 3.5 on January 3 (primary pollution period) and decreased to 1.9 on January 5 (midterm pollution period), indicating that vehicle emissions and stationary sources both substantially contributed to severe pollution days.The increased accumulation from certain stationary sources in the middle of the pollution period contributed to the high b abs-880nm, BC (Han et al., 2011;Tian et al., 2016).
In comparison with b abs-880nm, BC values, the peaks of b abs-370nm, BrC mainly concentrated in the late-night period of 0:00-3:00 from January 1 to 3 and afternoon on pollution days.The afternoon peaks of b abs-370nm, BrC associated with increased b abs-880nm, BC values, primarily because of the accumulation of primary anthropogenic emissions during daytime.Therefore, residential coal and biomass burning were key contributors to heavy UV light-absorption.The midnight peaks of b abs-370nm, BrC stress that nighttime PM also contained strong UV absorption (Saleh et al., 2013).Aqueous SOC formation in midnight during high RH condition should be one of important contributors for high b abs-370nm, BrC .Previous literatures reported that the conditions of stagnant air, high RH, and low temperature in winter haze days also favored the partitioning of SVOCs to the particle phase through aqueous reactions (Strader et al., 1999;Shrivastava et al., 2008;Chen et al., 2010).In addition, acidic aerosols also promote the formation of SOC.For instance, formation of high-molecular-weight (MW) SOC products has been observed during the OH-initiated oxidation of aqueous 3,5-dihydroxybenzoic acid (Kroll and Seinfeld, 2008); phenolic/methoxyphenols compounds forms SOC via efficient OH addition to the aromatic ring from aqueous reactions (Sun et al., 2010); and those of gaseous biomass precursors (isoprene and α-pinene) on acidic aerosol particles' (which involves sulfate radicals (SO 4 -)) surfaces (Limbeck, 2003;Noziere et al., 2010;Perri et al., 2010).Recent study by our group also revealed that one third of OC on average during haze days was formed through aqueous reactions under the high RH and stronger PM 2.5 acidity in Xi'an (Zhang et al., 2015).
The MAC can be used to understand light-absorbing abilities.The MAC-BC values calculated at 880 nm using Eq.(2) were 15.3 and 11.2 m 2 g -1 for pollution and nonpollution days, respectively.The winter MAC-BC values exhibited wider variations, differing by a factor of up to four during sampling periods because of variable BC sources in winter.The MAC was not governed solely by absorbing efficiency; their sources also influenced it.Cheng et al. (2011) indicated that intensive biomass emissions reduce MAC-BC values; however, the abundant diesel exhaust during severe pollution days displayed relatively high MAC-BC values.Compared with the MAC-BC, the MAC-BrC (calculated at 370 nm in Eq. ( 1)) showed lower levels and less variability (2.2 m 2 g -1 for pollution PM and 2.4 m 2 g -1 for non-pollution PM).Thus, BrC is a weaker absorber than BC.A wide MAC-BrC range (Fig. 5(d); 1.2-2.8m 2 g -1 ) covered ranges produced by regions with predominant biomass burning, such as Philadelphia (Jeong et al., 2004) (2.4 m 2 g -1 ), Beijing (Cheng et al., 2011) (0.71-1.79 m 2 g -1 ), and the Amazon basin (Hoffer et al., 2006) (0.5-1.5 m 2 g -1 ) but was much higher compared with the traffic-dominated region of the southeastern United States (Bond, 2004) (0.31-0.70 m 2 g -1 ).As a result, these variations of MAC-BrC were possibly caused by the variety of fuel combustion sources during severe pollution periods.

CONCLUSION
The concentrations of gas pollutants (NO 2 , SO 2 , and CO), BC, and PM 2.5 were highest on pollution days and nearly double of those on non-pollution days.Based on SMPS analysis, both pollutive and non-polluting particles showed relatively large size ranges in this study and included the Aitken and accumulation modes.Diurnal variations of the pollutants' PN distribution emphasized the influences of fresh vehicle emissions and particle condensational growth.The condensation of much larger PM was observed in non-pollution periods.Among the carbon species, a considerable portion (approximately 60%) consisted of EC1, OC3, and OC4 on both non-pollution and pollution days, which emphasized the effect of coal burning during sampling days.Moreover, the high OC/EC (3.0-5.8) and SOC/TC (approximately 30%) ratios in this study indicated the noteworthy influence of SOC.The substantial OC2 and OC3 concentrations and high O 3 levels in the afternoon also support this conclusion.With regard to the light-absorbing properties, the diurnal pattern of b abs-880nm, BC showed different peak times during pollution days, strongly suggesting that BC was heavily influenced by traffic emissions in the early pollution period and accumulated from other primary sources (i.e., heating and industrial emissions) in the midterm period.In contrast, the peaks of b abs-370nm, BrC corresponded to the periods of photochemical and aqueous SOC.The ranges of both MAC-BC and MAC-BrC vary noticeably, confirming that the dominant sources of BC and BrC change during a severe pollution period.

Fig. 2 .Fig. 3 .
Fig. 2. Temporal variations of meteorological parameters with time resolution of 1 h during a severe pollution period in Xi'an.(a) Visibility; (b) RH; (c) Wind speed and (d) Temperature.

Fig. 4 .
Fig. 4. Comparison of measurements between pollution and non-pollution days: (a) Particle size distribution, the solid lines represent lognormal fits; (b) Time series of mean diameter; (c) Diurnal variations of mean number concentrations (dN/dlogDp).

Fig. 5 .
Fig. 5. Time series of (a) TC/PM 2.5 ratios (µg m -3 ), (b) concentrations of major PM 2.5 carbon species, (c) b abs-880nm, BC and b abs-370nm, BrC , and (d) MAC-BC and MAC-BrC during sampling days.Three colored regions represent the different pollution stages: primary pollution period, midterm pollution period, and late pollution period.