李致宇 ，張 新 ，王啟元 ，André S. H. Prév?t，曹軍驥 ，韓月梅

1.中國科學院地球環(huán)境研究所 黃土與第四紀地質國家重點實驗室，中國科學院氣溶膠化學與物理重點實驗室，西安 710061

2.中國科學院第四紀科學與全球變化卓越創(chuàng)新中心，西安 710061

3.中國科學院大學，北京 100049

4. Laboratory of Atmospheric Chemistry, Paul Scherrer Institute (PSI), 5232 Villigen, Switzerland

Atmospheric submicron particulate matters are of great importance to affect air quality, human health,and Earth’s climate (Kanakidou et al, 2005; Hallquist et al, 2009; Kuang et al, 2016). Severe haze episodes occurred frequently in the urban areas of China in the past decade, which were to a large extent caused by the presence of submicron particles (Huang et al, 2014;An et al, 2019). Atmospheric submicron particles can substantially reduce the visibility during haze episodes through scattering and absorbing visible lights (Kang et al, 2013). Understanding the physical and chemical properties of submicron particles is essential to accurately assess their adverse environmental and health effects. Atmospheric submicron particles are composed of a variety of chemical species such as organic matter, sulfate, nitrate, ammonium, and black carbon (Williams et al, 2010; Ng et al, 2011; Sun et al,2012; Minguillón et al, 2015). Organic matter,accounting for a substantial portion of submicron particles, can be produced directly from primary emissions and also formed secondarily via gas to particle conversion (Sun et al, 2013; Elser et al,2016; Sun et al, 2016a). Secondary inorganic and organic aerosols increased rapidly during heavily polluted haze episodes, in particular at wintertime, and contributed substantial additional airborne particulate mass loadings (Chakraborty and Gupta, 2010; Zhang et al, 2015a; Huang et al, 2016). Characterizing the formation and transformation processes of submicron secondary aerosols is helpful to develop effective strategies for air pollution controlling.

Fenwei Basin, situated in the middle of China,has raised considerable attention for severe air pollution in recent years (Xu et al, 2012; Shen et al,2014; Wang et al, 2015). Fenwei Basin is comprised mainly by the Fenhe Basin in Shanxi Province and the Weihe Basin in Shaanxi Province. The narrowand crescent-shaped terrain of Fenwei Basin is in particular unfavorable for the transport and dispersion of air pollutants. Fenwei Basin has been designated officially as one of the three key regions of air pollution controlling in China since 2018. Xi’an is the largest while one of the most polluted cities among 11 primary cities in Fenwei Basin. The composition and source of atmospheric particles have been studied extensively in Xi’an City in prior studies(Cao et al, 2005; Cao et al, 2012; Wang et al, 2015).Most of those studies, however, were conducted on the basis of aerosol filter sampling and subsequent offline chemical analysis with low time resolution.For instance, coal combustion was identified as an important source of atmospheric fine particles in Xi’an and contributed to more than half of the particle light scattering property (Cao et al, 2012; Xu et al,2012). The major sources of submicron particles were combustion related primary emissions and secondary aerosols during wintertime in Xi’an (Shen et al, 2010;Huang et al, 2014). Characterization of submicron particles from real time measurements can gain further insights into the fast atmospheric processes regarding their formation and transformation.

This study aims at providing detailed knowledge on the composition, source, and atmospheric process of atmospheric submicron particles from online measurement during summertime in order to develop more accurate air pollution control strategies in Xi’an City. In early summer of 2017, an aerosol chemical speciation monitor (ACSM) was deployed for online measurement of chemical components of submicron particles at a suburban site of Xi’an. The air quality status in the studied period was assessed based on the air pollutants data from a national air quality monitoring station nearby the sampling site. The compositional properties of submicron particles were characterized and discussed based on the ACSM measurement. The source of submicron organic aerosols was identified through positive matrix factorization analysis of organic mass spectra.Furthermore, the influences of meteorological parameters on the mass loading and chemical composition of submicron particles were analyzed statistically and discussed.

1 Field measurements

1.1 Sampling location

A field measurement of atmospheric submicron aerosol particles was conducted at Chanba ecological district (34°19′48″N, 109°03′00″E) in Xi’an, Shaanxi Province, China from 24 May to 6 June 2017. The sampling site was situated in an open suburban area of Xi’an and there were no high-rises nearby. The east third ring road of Xi’an was approximately 200 m in the west of the sampling site. This suburban area was surrounded by tourist attractions and was part of the International Horticultural Exposition (2011) Park of Xi’an. This sampling site represented a typical suburban environment downstream of a megacity.

1.2 ACSM operation and data processing

Atmospheric aerosols were sampled through a stainless steel tubing (9.525 mm) at a flow rate of 3 L ? min?1. A cyclone with a particle cutoff size of 2.5 μm (model URG-2000-30ED, URG Corp.) was installed prior to the inlet line to remove coarse particles. Downstream the cyclone, the aerosol flow was dried by passing through a nafion dryer (model MD-110-48S, Perma Pure Inc.) to prevent possible influence of variable relative humidity on the sequent measurements. The composition of nonrefractory submicron particles (NR-PM1), including organics, sulfate, nitrate, ammonium, and chloride,was measured in real time using a quadrupole aerosol chemical speciation monitor (Q-ACSM,Aerodyne Research Inc.) (Ng et al, 2011). The massto-charge ratio (m/z) of 10 to 150 was determined by the Q-ACSM. The scanning frequency of mass spectrum was 500 ms and a full-spectrum scan took approximately 70 s. An automatic valve was switched back and forth between the ambient aerosol and background measurements (30 s for each), from which the mass spectra of submicron particles were obtained. The ionization efficiency (IE) of the ACSM instrument was calibrated at the end of the study. For the IE calibration, ammonium nitrate particles were generated using an aerosol atomizer (model 9302,TSI Inc.) and size-selected at 300 nm by a differential mobility analyzer (model 3080, TSI Inc.) and then measured using the ACSM and a condensation particle counter (model 3772, TSI Inc.).

The measurement data were acquired using the standard software of ACSM DAQ (v1.4.4.5) at an interval of 30 min. The data were processed using ACSM Local tool (v1.5.3.5) with Igor Pro (v6.37,WaveMetrics Inc.). A relative ionization efficiency(RIE) of 5.5 was applied for ammonium based on the IE calibration. The default values of RIE were used for other species, that is, 1.4 for organics, 1.2 for sulfate, 1.1 for nitrate, and 1.3 for chloride. A collection efficiency of 0.5 was used for the ACSM data processing herein, given that the mass fraction of nitrate was low (mean 9% ± 6%) and the particles were dried and not acidic (the ratio of the measured to the predicted amount of ammonium needed to neutralize the inorganic anion was on average 1.0 ± 0.4) (Middlebrook et al, 2012). The resulted mass concentrations of NR-PM1showed reasonable agreement with those of PM2.5from the national air quality monitoring data (section 2.2). The mass spectra of organic aerosols derived from the ACSM measurement were analyzed further using positive matrix factorization (PMF) evaluation tool (PET v3.05) (Ulbrich et al, 2009; Zhang et al, 2011).The (m/z) from 12 to 120 was used for the PMF analysis, given that (m/z) 127 to 129 were influenced by naphthalene signals and also (m/z) 120 to 150 contributed minorly to the total organic signals.

Hourly meteorological data, including temperature, relative humidity (RH), wind speed and wind direction, and precipitation, measured at Jinghe station in Xi’an (34°25′48″N, 108°58′12″E) during the studied period were obtained from the National Meteorological Information Center of China (http://data.cma.cn). Moreover, the hourly mass concentrations of primary air pollutants and air quality index (CO,O3, SO2, NO2, PM2.5, PM10, and AQI) were obtained from the China National Environmental Monitoring Centre (http://www.cnemc.cn/) at a national air quality monitoring station that was approximately 500 m in the northwest of the sampling site.

2 Results and discussion

2.1 Overview of air quality status

Fig. 1 presents the diurnal variations in the mass concentrations of primary gaseous (CO, O3, SO2, and NO2) and particulate (PM2.5and PM10) pollutants from 24 May to 6 June, 2017. The mass concentrations of gaseous pollutants CO, O3, SO2, and NO2were on average 958 μg ? m?3, 85 μg ? m?3, 13 μg ? m?3, and 56 μg ? m?3, respectively. These values were below the 1st national ambient air quality standard of China(2012) in 24-hour (or 8-hour for O3) averages (Fig. 1c).Ozone was the dominant pollutant for 10 of the 14 days, while NO2was dominant for one day. High concentrations of ozone were observed during the daytime and peaked in the afternoon. In contrast,NO2was abundant mostly in the nighttime. The mass concentration of PM2.5was on average 33.3 μg ? m?3during the studied period. This value was slightly lower than the 1st national ambient air quality standard of China (35 μg ? m?3), while it was higher than the PM2.5standard of the World Health Organization(25 μg ? m?3). The mass concentration of PM10was 71.6 μg ? m?3, which was between the 1st and 2nd national standard of China for PM10in 24-hour average(i.e., 50 μg ? m?3and 150 μg ? m?3, respectively; Fig. 1c).

Fig. 1 Diurnal profiles of the mass concentrations of gaseouspollutants (a) and particulate matters and the dimensionless air quality index (b) at the nearest national air quality monitoring station during the studied period, statistical analysis of the mass concentrations of gaseous and particulate air pollutants and the air quality index for the entire period (c)

The air quality index (AQI) representing the extent of air pollution is also presented in Fig. 1.The AQI values of below 50, 50 to 100, and above 100 represent the excellent, good, and unhealthy air quality, respectively. A mean value of AQI was 70 across the studied period. The AQI values of below 50, 50 to 100, and above 100 accounted for 24%, 57%,and 19% of the total days, respectively. Therefore,the air was at least not heavily polluted during the studied period in Xi’an, in particular less polluted by atmospheric fine particulate matters of PM2.5. The low mass concentrations of air pollutants during the summertime in Xi’an were distinct compared to those in the heavily polluted wintertime with large amounts of coal combustion and wood burning for heating and secondary pollutants (Shen et al, 2010; Cao et al,2012).

2.2 Chemical composition of NR-PM1

The mass concentration of NR-PM1was on average (30.1 ± 15.4) μg ? m?3during the studied period.The distributive and cumulative frequencies of the mass concentration of NR-PM1are presented in Fig.2a. The mass concentration of NR-PM1was abundant mostly in the range of 20.0 μg ? m?3to 30.0 μg ? m?3(accounting for 30%), followed by those in the range of 10.0 μg ? m?3to 20.0 μg ? m?3(21%) and 30.0 μg ? m?3to 40.0 μg ? m?3(20%). The frequency was small for the mass concentrations of below 10.0 μg ? m?3and above 40.0 μg ? m?3(overall 18%). Fig. 2b presents a comparison of the mass concentrations between the NR-PM1measured by the ACSM and the PM2.5obtained from the nearby national air quality monitoring station. The mass concentrations of NR-PM1correlated moderately to those of PM2.5and with a correlation coefficientrof 0.65. The mass concentration ratio of NR-PM1/ PM2.5was on average 0.8, suggesting that PM1constituted a substantial portion of PM2.5. The NR-PM1/ PM2.5ratio was similar to that of 0.71 for Baoji (a city beside Xi’an) in spring 2014 (Wang et al, 2017). The ratio was slightly higher than that of 0.64 for Beijing in summer 2011 (Sun et al, 2012). The differences among individual cities could result from different factors such as emission sources, formation processes,meteorological conditions, and regional transport influence.

Fig. 2 Distributive and cumulative frequencies of the mass concentration of NR-PM1 derived from the ACSM measurement(a), a comparison between the mass concentration of NR-PM1 measured by the ACSM and those of PM2.5 from the nearby national monitoring station for the entire period (b)

The temporal variations in the chemical composition of NR-PM1and meteorological parameters for the entire study are presented in Fig. 3.The temperature varied from 15℃ to 38℃ (mean(24 ± 6)℃) and the relative humidity varied from 10%to 99% (mean 55% ± 29%) (Fig. 3a). The wind was below 3 m ? s?1(mean (1 ± 0.6) m ? s?1) in the speed across the entire period and was dominated by the south and southeast directions (Fig. 3b). A heavy rainfall occurred continually in the latter period from 3 to 6 June. The mass concentration of organic matter was in the range of 2.9 μg ? m?3to 53.7 μg ? m?3(mean(18.9 ± 10.9) μg ? m?3). The mean mass concentrations of sulfate, nitrate, ammonium, and chloride were 5.3 μg ? m?3, 2.8 μg ? m?3, 2.9 μg ? m?3, and 0.1 μg ? m?3,respectively. Organics was the most dominant component and accounted for on average 63% of the total mass concentration of NR-PM1, followed by sulfate (18%), ammonium (10%), nitrate (9%), and chloride (0.4%).

Fig. 3 Time series of temperature and relative humidity (a), wind speed, wind direction, and precipitation (b), and the mass concentrations (c), and the mass fractions of NR-PM1 species (d) during the study period

A comparison of the mass concentration and the composition of NR-PM1among multiple locations in China is summarized in Tab. 1. The mass concentration of NR-PM1observed herein during the summertime was 4.6 times lower than those reported for the wintertime in Xi’an (Zhong et al, 2020), suggesting the air was less polluted in summertime. The mass concentration of NR-PM1was also much lower than those of other locations in wintertime (Zhang et al, 2015b; Qin et al,2017; Wang et al, 2017; Han et al, 2019; Huang et al,2019). On the other hand, the mass concentration of NR-PM1was comparable with those of Shanghai and Hong Kong in summer and fall seasons in early years(Huang et al, 2012; Sun et al, 2016a; Sun et al, 2016b).The low concentrations of NR-PM1observed in Xi’an during the studied period might be partly resulted from the less production of air pollutants and the favorable meteorological condition for their dispersion.Moreover, the composition of NR-PM1in this study was characterized by the high mass fractions of organic matter compared to those in Xi’an during wintertime and other locations. Possible explanations include the contribution sources and the formation and evolution processes of submicron particles were different at individual locations.It is also possible that the lower mass concentrations of NR-PM1herein were to a large extent resulting from the decreased mass concentrations of secondary inorganic species of sulfate, nitrate, and ammonium in summertime when without substantial combustions for heating compared to those in wintertime.

Tab. 1 Comparison of the mass concentration and chemical composition of NR-PM1 measured by ACSM and aerosol mass spectrometer among multiple locations in China

The diurnal profiles of the mass concentration and mass fraction of NR-PM1species are presented in Fig. 4. The mass concentration of organics fluctuated slightly across the day. The highest peak of organics was observed at around 22∶00 local time (LT),which might be associated to the large abundance of air pollutants because of the reduced planetary boundary layer height at nighttime. There was also a small peak from 08∶00 to 14∶00 LT, which was possibly contributed by the traffic emissions and also photochemical processes (see further details in section 2.3 regarding the diurnal variations of organic factors). A large increase in the mass concentration and mass fraction of sulfate was observed from 08∶00 to 14∶00 LT, suggesting that sulfate was produced greatly from atmospheric photochemical processes. Ammonium exhibited a similar variation trend as those of sulfate and peaked at around 11∶00 LT, indicating that ammonium was greatly combined and thus neutralized by sulfate. Nitrate was abundant mostly at nighttime, which was likely associated to the reduced planetary boundary layer height for vertical dilution and the lower temperature favorable for nitrate partitioning to the particle phase. A small peak of nitrate was also observed in the morning rush hour,indicating the contribution of the oxidation reactions of NOxfrom traffic vehicle emissions.

2.3 Sources of submicron organic aerosols

Given that organic matter accounted for a great portion of submicron particles during the studied period (i.e., on average 63%), the source apportionment herein was focused primarily on organic aerosols. Two organic factors were retrieved from the PMF analysis of submicron organic mass spectra, including a hydrocarbon-like organic aerosol component (HOA) and an oxygenated organic aerosol component (OOA). Other factors such as biomass burning, coal combustion, and cooking related organic aerosols, which have been reported previously at nearby sites in winter and spring seasons (Elser et al,2016; Wang et al, 2017), were not identified from the current study. This could be explained by the insignificant contributions of other sources to total organic aerosols during the summertime. This is also consistent with the low mass concentration of organic aerosols during the studied period in Xi’an compared to those at other sites (Tab. 1). The PMF solutions with more than two factors added very little information but split the primary two factors of HOA and OOA.

Fig. 5 presents the mass spectra, mass loadings,and diurnal profiles of the two PMF factors. The mass spectra of HOA was characterized by the dominant peaks of hydrocarbon fragments(e.g.,m/z29, 43, and 57) and(e.g.,m/z27, 41, and 55).HOA accounted for on average 43% of total submicron organic aerosol mass. The temporal variation of the mass loading of HOA exhibited clear diurnal cycles and was to a certain extent similar to those of NO2that was associated mostly to traffic vehicle emissions.The mass loading of HOA was decreased significantly during the rainfall period from 4 to 6 June. The diurnal profile of HOA peaked in the morning (around 06∶00 to 09∶00 LT) and nighttime. The lowest mass loading appeared at around 15∶00 to 17∶00 LT, suggesting that HOA was hardly influenced by photooxidation processes and thus likely associated to primary emissions. Taken into account of these typical characteristics together with the fact that the sampling location was away from residential areas and was close to a main road with large traffic volumes, HOA might be composed primarily of vehicle emissions.

The mass spectrum of OOA had prominent peak atm / z44 (mainly), which accounted for 21% of the total signal intensity.is the major fragment ion from the electrical ionization of the carboxyl group compounds and can reflect the oxidation degree of organic aerosols (Takegawa et al, 2007). OOA accounted for on average 55% of total organic aerosol mass. The temporal variation of OOA mass loading correlated highly with the summed sulfate and nitrate concentrations (rof 0.75), suggesting that OOA was associated to secondary formations. The mass loading of OOA increased prominently in the daytime and peaked at around 12∶00 LT, suggesting that OOA was influenced mainly by secondary photochemical processes. A linear regression analysis of PMF factors as functions of oxidized gaseous species (NO2+ O3)is presented in Fig. 6. The mass loading of OOA correlated more strongly with those of the oxidized gases (rof 0.49), suggesting that OOA was formed substantially from the oxidation of the precursors by the atmospheric oxidants. In contrast, HOA correlated poorly with oxidized gaseous species (rof 0.21),which is consistent with the less oxidized feature of primary vehicle emissions.

Fig. 4 Diurnal profiles of the mass concentrations (a) and the mass fractions (b) of NR-PM1 species during the studied period

Fig. 5 Two factors of HOA and OOA resolved from the PMF analysis: mass spectral profiles (a, d),factor mass loadings (b, e), and diurnal variations in the mass loadings (c) and mass fractions of the two factors (f)

Fig. 6 Correlations of the mass loadings of HOA (a) and OOA (b) with the mass concentrations of atmospheric oxidized gaseous species (the sum of NO2 and O3)

2.4 Meteorology influenced atmospheric processes

Fig. 7 presents the statistical analyses on the mass concentrations of organics, sulfate, and nitrate as functions of meteorological parameters of temperature,relative humidity, and wind speed. Note that other factors, such as local emissions and regional transport,might also to a certain extent influence the mass concentrations of individual components, whereas they were not taken into account herein. In contrast to those of organic and nitrate without clear trends with the change of temperature, the mass concentration of sulfate increased constantly from 3.5 μg ? m?3to 8.1 μg ? m?3with the increase of temperature from 15℃to 39℃ (Fig. 7a). An enhanced ozone concentration was also corresponded to the increasing temperature.A possible explanation is that photochemical reactions were more intense at high temperature and promoted the transformation of gaseous precursors to sulfate(Wang et al, 2016). The mass concentration of organics increased gradually with temperature from 15℃ to 25℃ and was nearly constant from 25℃ to 39℃. A slight decrease in the mass concentration of nitrate was observed at variable temperatures from 20℃ to 39℃, which is possibly due to the evaporation of nitrate compounds at a higher temperature. The variable dependence of the mass concentrations of organics and nitrate on temperature suggests that multiple competitive pathways, such as compensation by volatilization, influenced their overall productions,in addition to the enhanced photochemical reactions at higher temperatures.

An increase in the mass concentration of nitrate was obtained with the relative humidity increased from 18% to 78% (Fig. 7b). A higher amount of liquid water content would present in particles at higher relative humidities and thus might serve as an effective medium for the heterogeneous reactions of precursors such as N2O5to produce nitrate salts(Cheng et al, 2016; Wu et al, 2018). The nitrate concentration was, however, decreased for RH above 90%, which was possibly resulted from the removal by precipitation. The influences of relative humidity on organics and sulfate were not significant,suggesting that the local aqueous reactions were at least not their dominant formation pathways.Interestingly, the mass concentration of organics was decreased up to 38% constantly with increasing wind speed from 0 m ? s?1to 2.4 m ? s?1(Fig. 7c). A decrease trend was also observed for sulfate from 5.8 μg ? m?3to 2.7 μg ? m?3. This result suggests that organics and sulfate were originated substantially from local sources and were transported and dispersed regionally when at higher wind speeds. A negligible change, however,was observed for nitrate with increasing wind speed,indicating that nitrate associated to regional transport was almost comparable to those formed locally during the studied period.

In order to gain further insights into the influence of meteorology on the mass concentration and composition of submicron particles, four typical scenarios were identified during the entire period, including high-organic, high-nitrate, highsulfate, and rainfall periods (see Fig. 3). The four scenarios were defined based on the characteristics in the composition of NR-PM1and meteorological conditions, as summarized in Tab. 2. The highorganic period was observed from 25 to 29 May,during which organics constituted on average 70% of total mass concentration of NR-PM1. The temporal variation trends of organic mass concentrations and meteorological parameters exhibited clear diurnal cycles. The mass concentrations of inorganic species were generally low during this period.These results represented a characteristic profile of submicron particles during clear days in Xi’an at summertime.

Tab. 2 The mass concentrations and mass fractions of NR-PM1 species and the meteorological parameters of temperature, relative humidity, and wind speed averaged over the entire period and the periods with high-organics, highnitrate, high-sulfate, and rainfall

The high-nitrate period was characterized by the increased mass concentration and mass fraction of nitrate from 29 to 30 May. The mass concentration of nitrate was on average 1.9 times high compared to those of sulfate. A light precipitation occurred and the relative humidity was on average 85% during the high-nitrate period. This result suggests that the formation of nitrate was favorable at high relative humidities, which was possibly associated to the enhanced aqueous reactions and the less volatilization at this condition. The high-sulfate period from 1 to 3 June was characterized by a great increase in the mass concentration of sulfate, which was on average 3.3 times those of nitrate. The high-sulfate period corresponded to higher temperatures (mean 29℃)compared to other periods (mean 17℃ to 27℃).The high temperature could lead to the enhanced photochemical reactions for the production of sulfate and organics, which contributed greatly to the total NR-PM1mass concentration. A heavy continuous rainfall period occurred from 3 to 6 June with an average relative humidity of 98% and temperature of 17℃. The mass concentrations of each NR-PM1species decreased substantially compared to other periods, which was possibly resulted from the removal of air pollutants through wet deposition. The mass fractions of inorganic species (the sum of sulfate,nitrate, and ammonium) increased to on average 56%of NR-PM1mass. Nevertheless, organics was still the most dominant component. The difference in the composition of NR-PM1might result from the different air mass origins during the rainfall events compared to those of other periods. Another possible explanation is that the precipitation not only had strong effects on the removal of particulate matters but also unfavorable for the formation of secondary organic aerosols due to the removal of precursor gases (as seen in Fig. 5b and 5e).

3 Summary

Atmospheric submicron particles were measured in real time using an ACSM at the Chanba ecological district of Xi’an, China from 24 May to 6 June, 2017.The chemical composition, source, and atmospheric process of submicron particles were characterized based on the ACSM measurements. The mass concentration of NR-PM1was on average (30.1±15.4) μg ? m?3for the entire period, which was much lower compared to those during the heavily polluted wintertime. Organic matter was the most dominant component and accounted for on average 63% of NRPM1mass, followed by sulfate (18%), ammonium(10%), nitrate (9%), and chloride (0.4%). Two factors including HOA and OOA were resolved from the PMF analysis of organic mass spectra. HOA was likely primarily governed by vehicle emissions. OOA correlated better with atmospheric oxidants (NO2+ O3)and secondary inorganic species, suggesting OOA was largely formed secondarily by atmospheric oxidation processes. The mass concentration and composition of NR-PM1were affected greatly by meteorological conditions. A higher temperature was favorable for the formation of sulfate possibly because of the enhanced photochemical reactions.A higher relative humidity promoted the formation of nitrate presumably through aqueous phase reaction of nitrogen oxides. The decreased mass concentration of organic matter was obtained both in rainfall period and with the increase of wind speed,suggesting the wet deposition and regional transport played important roles for the removal of submicron organic aerosols. The characteristics of atmospheric submicron particles obtained from the current study in the early summertime, contrasting to the heavily polluted wintertime, should be taken into account for developing further air pollution control strategies in Xi’an and the surrounding area, in particular for periods without substantial coal combustion for heating.