Characteristics of concentrations and chemical compositions for PM 2 . 5 in the region of Beijing , Tianjin , and Hebei , China

In order to study the temporal and spatial variations of PM2.5 and its chemical compositions in the region of Beijing, Tianjin, and Hebei (BTH), PM2.5 samples were collected at four urban sites in Beijing (BJ), Tianjin (TJ), Shijiazhuang (SJZ), and Chengde (CD), and also one site at Shangdianzi (SDZ) regional background station over four seasons from 2009 to 2010. The samples were weighted for mass concentrations and analyzed in the laboratory for chemical profiles of 19 elements (Al, As, Ba, Ca, Cd, Co, Cr, Cu, Fe, K, Mg, Mn, Ni, P, Pb, Sr, Ti, V, and Zn), eight watersoluble inorganic ions (Na, NH+4 , K , Mg2+, Ca2+, Cl, NO−3 , and SO 2− 4 ), and carbon fractions (OC and EC). The concentrations of PM2.5 and its major chemical species were season dependent and showed spatially similar characteristics in the plain area of BTH. The average annual concentrations of PM2.5 were 71.8–191.2 μg m−3 at the five sites, with more than 90 % of sampling days exceeding 50 μg m−3 at BJ, TJ, and SJZ. PM2.5 pollution was most serious at SJZ, and the annual concentrations of PM2.5, secondary inorganic ions, OC, EC, and most crustal elements were all highest. Due to stronger photochemical oxidation, the sum of concentrations of secondary inorganic ions (NH+4 , NO − 3 , and SO 2− 4 ) was highest in the summer at SDZ, BJ, TJ, and CD. Analysis of electric charges of water-soluble inorganic ions indicated the existence of nitric acid or hydrochloric acid in PM2.5. For all five sites, the concentrations of OC, EC and also secondary organic carbon (SOC) in the spring and summer were lower than those in the autumn and winter. SOC had more percentages of increase than primary organic carbon (POC) during the winter. The sums of crustal elements (Al, Ca, Fe, Mg, Ti, Ba, and Sr) were higher in the spring and autumn owing to more days with blowing or floating dust. The concentrations of heavy metals were at higher levels in the BTH area by comparison with other studies. In Shijiazhuang and Chengde, the PM2.5 pollution was dominated by coal combustion. Motor vehicle exhausts and coal combustion emissions both played important roles in Tianjin PM2.5 pollution. However, motor vehicle exhausts had played a more important role in Beijing owing to the reduction of coal consumption and sharp increase of cars in recent years. At SDZ, regional transportation of air pollutants from southern urban areas was significant.


Introduction
The region of Beijing, Tianjin, and Hebei (BTH) is one of the important city agglomerations in China, accounting for 10.9 % of national GDP and 7.8 % of national population in 2010.As result of economic development, the structures of air pollution sources and fuel consumption have constantly been changing.At present, coal is still the primary fuel in most areas of BTH and widely used for industrial processes and daily life, and more coal is combusted for heating in the winter.However, coal has been gradually being replaced by natural gas and electricity in urban areas of Beijing since 1999(Beijing Environmental Protection Bureau, 2011).The total amount of motor vehicles for BTH went from 2. photochemical reaction, more polluted and hazy days resulting from reactive gas and fine particle have appeared and become increasingly conspicuous.In many urban areas of BTH, more than 100 days per year have been in haze, and the annual average visibility has been lower than 15 km in recent years (Zhao et al., 2011(Zhao et al., , 2012)).The average number concentration of particles less than 1000 nm could reach up to 30 000 cm −3 in the urban area of Beijing, which is much higher than that in some foreign studies (Wu et al., 2008).The scattering effect of fine particles is the most important influence factor for visibility, and the scattering coefficients of aerosols accounted for more than 85 % of extinction coefficients in Beijing and surrounding areas (Yan et al., 2008;Garland et al., 2009;Ma et al., 2011).Several studies of aerosol health effects carried out in Beijing have also revealed the relationships between fine particle pollution and morbidity and mortality (Zhang et al., 2000;Guo et al., 2009;Kipen et al., 2010;Wu et al., 2010).
The new national ambient air quality standard was issued in February 2012.The standard for PM 2.5 was first implemented in some key regions and cities in 2012, and will not be implemented nationwide until 2016.So data for PM 2.5 concentration and composition could not be systematically acquired in most parts of China by now.In China, studies for PM 2.5 have been gradually carried out since 2000, and mainly focused on the urban areas of the Pearl River Delta (Cao et al., 2004;Hagler et al., 2006;Duan et al., 2007;Liu et al., 2008), the Yangtze River Delta (Ye et al., 2003;Yang et al., 2005;Wang et al., 2008;Li et al., 2011), and the BTH region.
In BTH, most of the related studies about PM 2.5 chemical compositions have been carried out in Beijing.Some studies have given the general characteristics of PM 2.5 chemical compositions and discussed their seasonal variations, correlations, or sources (He et al., 2001;Sun et al., 2004;Song et al., 2006).Some studies have focused on the concentrations, correlations, sources or formation of some specific species (such as inorganic ions, carbonaceous components, or organic matters) of PM 2.5 in Beijing (Yao et al., 2002;Dan et al., 2004;Huang et al., 2006;Wang et al., 2009;Pathak et al., 2009;Ianniello et al., 2011).In addition, the size distributions of aerosol chemical species (Yao et al., 2003;Cheng et al., 2009;Guo et al., 2010;Li et al., 2012), aerosol number concentrations or new particle formation processes (Wu et al. 2007(Wu et al. , 2008;;Yue et al. 2009Yue et al. , 2011;;Shen et al., 2011;Zhang et al., 2011;Gao et al., 2012), and aerosol optical characteristics or mixing state (Cheng et al., 2009;Deng et al., 2011;Ma et al., 2011Ma et al., , 2012;;Chen et al., 2012) have also been discussed for Beijing and the surrounding region.
PM 2.5 , or fine particles, have been systematically analyzed in many studies in Beijing.But PM 2.5 has seldom been spontaneously sampled and chemically analyzed at several sites in different regions of BTH.Related published data for PM 2.5 in urban areas of Tianjin and Hebei are lacking.There are no overall analyses for PM 2.5 chemical compositions for the region of BTH yet.To better control the regional aerosol pollution and carry out some further related studies in BTH, it is quite necessary to obtain the information of regional PM 2.5 concentrations and their chemical compositions, and also to know their spatial and temporal variations.
In this study, PM 2.5 was sampled at five sites in BTH over four seasons.The main objectives of this paper are to (1) characterize spatial and seasonal concentration variations for PM 2.5 and its chemical compositions, (2) study the interrelations between different compositions, (3) evaluate primary and secondary sources of PM 2.5 , and (4) analyze variations of PM 2.5 chemical compositions in Beijing in recent years.The reactive gases (NO, NO 2 , SO 2 , O 3 , CO), light scattering and absorption, and meteorological parameters were also measured at some sites.The further analysis about sources, chemical composition variations, formation processes, optical characteristics and transportations for PM 2.5 in different pollution episodes will be discussed in other papers.

Sampling sites and meteorological conditions
As shown in Fig. 1, five sites were selected in the BTH area.Beijing and Tianjin are two municipalities that are directly administrated by the central government and are surrounded by Hebei province.Shijiazhuang is the capital of Hebei.Beijing, Tianjin, and Hebei are located in the North China Plain.Shangdianzi is a regional background station of atmospheric compositions used as the representative site of the North China Plain, which is 100 km northeast of the urban area of Beijing and located in the boundary of the plain area.It is also one of the Global Atmosphere Watch (GAW) stations, and the detailed instructions for Shangdianzi had been previously reported (Zhao et al., 2009).Chengde is one city located in the northern mountainous area of Hebei.Four sites were separately located in the urban areas of Beijing, Tianjin, Shijiazhuang, and Chengde, and all placed on the roofs of local meteorological bureaus.Another sampling site was set at Shangdianzi.In this paper, the sampling sites for Shangdianzi, Beijing, Tianjin, Shijiazhuang, and Chengde are abbreviated as SDZ, BJ, TJ, SJZ, and CD.
The seasonal wind roses of five sites during all four sampling periods are depicted in Fig. 2, and the corresponding calm wind frequencies are listed in Table 1.The prevailing wind directions at SDZ, BJ, and SJZ were all dominated by the nearby topography, and wind directions at CD were closely related to the direction of the urban valley area surrounded by mountains.The wind directions in TJ were influenced by land and sea breeze (Fig. 1).In the BTH area, wind is mainly from the south in the summer and from the north in the winter.The topography not only governs the wind directions but also decides the regional transportation of air pollutants.The average values of temperature, relative humidity, wind speed, and observed visibility for four sampling periods at the five sites were also listed in Table 1 (the dates of four sampling periods, see Sect.2.2).The meteorological conditions were similar in the BTH area; for all four seasons, temperature ranked in the order of CD < SDZ < BJ < TJ < SJZ, wind speed was lowest at SJZ, and visibility was lowest at TJ; for all five sites, temperature ranked in the order of winter < autumn < spring < summer, relative humidity was highest and visibility was lowest in the summer, wind speed was highest in the spring, and calm wind was most frequent in the autumn.

Sample collection and chemical analysis
Aerosol sampling was performed in spring (6 April to 1 May 2009), summer (9 July to 4 August 2009), autumn (11 October to 4 November 2009), and winter (14 January to 8 February 2010).PM 2.5 was sampled each day from 09:00 to 08:00 the next day.Owing to precipitation or occasional sampler instability, more than 25 days were taken to assure at least 20 valid samples for each period at each site.Two parallel medium-volume samplers (TH-150A, 100 L min −1 ) were used at each site for collecting PM 2.5 on 90 mm polypropylene filters and quartz fiber filters.Polypropylene filters were used for analysis of inorganic elements.Quartz fiber filters were used for carbon and water-soluble inorganic ion components.Each polypropylene filter was cut into fragments and put into a conical flask.Then 16 mL of HNO 3 and 4 mL of HClO 4 were added into the flask and heated by an electric stove until there was ∼ 3 mL residual left.After cooling, the solution was filtered and decanted into a test tube, and diluted to 15 mL with ultrapure water.Then an ICP-AES (ULTIMA, JY) was used for analysis of 19 elements (Al, As, Ba, Ca, Cd, Co, Cr, Cu, Fe, K, Mg, Mn, Ni, P, Pb, Sr, Ti, V, and Zn).
A 1/4 piece of each quartz fiber filter was put into a glass tube, and 10 mL of de-ionized water was added.After a 15 min ultrasonic bath at room temperature, the solution was drawn into a 5 mL syringe, filtered by a syringe filter, and injected into a polymeric vial with filter cap.The polymeric vials were put into a Dionex AS-DV Autosampler and analyzed by ion chromatography (ICS-1000, DIONEX) for water-soluble inorganic ions (Na + , NH + 4 , K + , Mg 2+ , Ca 2+ , Cl − , NO − 3 , and SO 2− 4 ).A 0.5 cm 2 punch from each quartz fiber filter was analyzed by a thermal optical carbon analyzer (DRI-2001A) for eight carbon fractions, following the IMPROVE A protocol (Chow et al., 1993(Chow et al., , 2001(Chow et al., , 2007)).

Quality control
Quartz fiber filters were all baked at 800 • C for 3 h prior to sampling to remove adsorbed organic vapors; and polypropylene filters were also heated at 80 • C for 2 h to remove volatile materials.Before and after sampling, the filters were weighed after being equilibrated for 24 h in a desiccator.All the samples were stored in a refrigerator under 3 • C and analyzed for chemical species in less than a month.Compared with quartz fiber filters, polypropylene filters have higher moisture sensitivity and lower collection efficiency (over 90 % of PM 2.5 mass by quartz fiber filters for most polypropylene filters).So the PM 2.5 concentrations were from gravimetric analysis of quartz fiber filters.The concentrations of all species were calculated by mass divided by sampling air flows.
Adsorption of gases and vapors from the airstream will lead to positive artifacts for measurements of sulfates, nitrates, and organic compounds; evaporation of sulfates, nitrates, and organic compounds into the passing airstream will lead to negative artifacts.The artifacts were often estimated by a field blank filter or a collocated filter placed behind the sampling filter.For quartz fiber filters, positive and negative biases of organic gases had been found to   approximately cancel each other under some circumstances (Watson and Chow, 2002;Eatough et al., 2003).Different batches of quartz fiber filters may also have different equilibrium properties for organic gases (Kirchstetter et al., 2001).So the artifacts of organic matter were not taken into account when calculating the concentrations of carbon fractions.The quartz fiber filters used in this study had lower alkalinity (less than 25 µeq g −1 ), so the adsorption artifacts of oxides of nitrogen and nitric acid could be essentially eliminated.Volatilization loss of NH 4 NO 3 from the filter has been found during sampling, especially during the warm mouths and during the warmest periods of the day (Zhang and McMurry, 1992;Ashbaugh and Eldred, 2004;Chow et al., 2005).Because of the limitation of the sampler, the volatilized nitrate and volatilized ammonium were difficult to predict and estimate.The NO − 3 concentrations in this paper all represented non-volatilized parts of nitrate.Some materials can be dissolved into the solution during pre-treatment of laboratory analysis, such as extraction, digestion, filtration, or injection.For example, we found that the glass tube for extraction dissolved out small amounts of Na + and that the syringe filters for solution filtration also had Na + and SO 2− 4 leaching.Thus, over two blank filters were also analyzed when analyzing each set of samples (samples for one season at one site), following the same procedures for both inorganic elements and water-soluble inorganic ions analysis.And the blank concentrations of inorganic elements and water-soluble inorganic ions were subtracted to get actual concentrations.The detection limits for inorganic elements and water-soluble inorganic ions were determined by three times the standard deviation of over 20 analyses of pure solvent used for extraction.The actual concentrations of each inorganic element or water-soluble ion in valid samples were all much higher than the detection limit and the standard variation of blank concentrations.After being baked at 800 • C, the residual carbon on the quartz fiber filters was negligible, and the detection limits for OC and EC were three times the standard deviation of over 20 pre-baked blank filter analyses.

PM 2.5 concentrations
The PM 2.5 concentrations for all samples at the five sites (Fig. 3) show similar fluctuations.The annual average concentrations were 71.8-191.2µg m −3 , ranked in order as SDZ < CD < BJ < TJ < SJZ.The annual concentrations exceeded 100 µg m −3 at BJ, TJ, and SJZ.SDZ is a background station without significant pollution sources within 30 km of the site.However, the PM 2.5 concentration at SDZ was still high, mainly owing to the regional transportation from the southern urban area.According to our PM 2.5 continuous monitoring data, the PM 2.5 concentrations were quite higher when southwest wind was blowing than the moment when wind was blowing from other directions at SDZ and also at BJ (Zhao et al., 2009).The distributions of PM 2.5 concentrations at the five sites (Fig. 4) also show that the PM 2.5 pollution was quite severe in the BTH area.The PM 2.5 concentrations exceeded 50 µg m −3 on more than 90 % of sampling days at BJ, TJ, and SJZ, and even exceeded 100 µg m −3 on more than 88 % of sampling days at SJZ.There was a four-day heavy pollution episode from 16 to 19 January 2010, with a higher degree of atmospheric stability, a high level of PM 2.5 , and lower visibility.However, PM 2.5 was only sampled at SJZ on 19 January because of problems with the sampler.At four urban sites, the maximum concentrations during four seasons all appeared in this pollution episode.The PM 2.5 concentration on 19 January reached 759.6 µg m −3 at SJZ.According to the winter concentration trends in Fig. 3, the maximum concentration would be higher, and the winter average concentration could reach about 300 µg m −3 at SJZ if all samples were obtained during that episode.
At SDZ, new particle formation events occurred most frequently in the spring, and the number concentrations were also highest for nucleation mode, Aitken mode, and accumulation mode (Shen et al., 2011).In addition, more windy days and large tracts of bare land caused more contribution of fugitive dust in the spring.So the maximum concentration appeared and the seasonal concentration was also highest in the spring.
In the winter, more particles were emitted from coal combustion for heating; thus the seasonal PM 2.5 concentrations at SJZ and CD were highest in the winter.PM 2.5 was relatively stable for BJ in all four seasons, because the energy used for heating and industrial processes was mainly electricity and natural gas except for limited residential coal consumption in the urban areas of Beijing.The finding that PM 2.5 at TJ was highest in the summer is unexpected and was attributed to the summer having more days with a stable atmosphere and higher concentrations of secondary inorganic ions (see Sect. 3.2.2).

Spatial concentration variations
The annual average concentrations of PM 2.5 species and seasonal average concentrations of major components therein at all five sites are listed in Table 2 and Table 3, respectively.We calculated the correlation coefficients among all chemi-cal species of samples from all four seasons at each site.Inorganic elements of Al, Ca, Fe, Mg, Ti, Ba, and Sr had strong correlations (mostly more than 0.80) with each other at each site, reflecting the common crustal sources.The NH + 4 , NO − 3 , and SO 2− 4 in the atmosphere were basically formed by a gasto-particle process as a result of chemical reactions of precursor gases (Seinfeld and Pandis, 2006;Guo et al., 2010;Ianniello et al., 2011), so we call them "secondary inorganic ions" in this paper.
In Shijiazhuang, a great deal of coal was combusted every year for industrial processes and daily life, the amount of cars had increased rapidly year by year, and a lot of buildings were under construction or demolition during sampling periods.The average wind speed was also lowest at SJZ.All of these conditions above made the aerosol pollution in Shijiazhuang become worrying, and the annual concentrations of secondary inorganic ions, OC, EC, and most crustal elements at SJZ were highest for all five sites.
At SDZ, the annual concentrations of PM 2.5 , crustal elements, OC, and EC were lowest, but the annual concentrations of NH + 4 , NO − 3 , and SO 2− 4 were higher than at CD.According to the prevailing wind directions of BJ and SDZ, the regional transportation of aerosol from southern urban areas to SDZ should not be discounted.The level of O 3 , which is the most important photochemical oxidant, was higher at SDZ than at BJ (Liu et al., 2008).Except for the regional transportation, the higher O 3 level could also facilitate the formation of secondary inorganic ions at SDZ by photochemical oxidation of reactive gases.
The annual average OC and EC concentrations were both in the order of SDZ < BJ < TJ < CD < SJZ at all five sites.The levels of OC and EC at CD were even slightly higher than at BJ and TJ, mainly owing to the simpler energy and pollution structure dominated by coal combustion.The annual OC/EC in PM 2.5 was relatively consistent in BTH, 2.56-2.87 at each site.
The annual average concentrations of crustal elements (Al, Ca, Fe, Mg, Ti, Ba, and Sr) were 2.87-8.27µg m −3 at the five sites, only accounting for 4.0-5.1 % of PM 2.5 mass.Crustal elements were mainly from fugitive dust influenced by the overall level of urban sanitation, greening cover, human earthmoving activities, and dust control.

Secondary inorganic ions
For environments with an abundance of NH 3 , the NH 4 HSO 4 and (NH 4 ) 2 SO 4 are sequentially generated; the fine particulate SO 2− 4 is not volatile.Ammonium nitrate is often formed in areas with high ammonia and nitric acid concentrations and low sulfate concentrations (Seinfeld and Pandis, 2006).Secondary inorganic ions (NH + 4 , NO − 3 , and SO 2− 4 ) were major components of PM 2.5 in BTH, which comprise 24-43 % of annual average PM 2.5 mass at the five sites.The mass ratios of secondary inorganic ions to PM 2.5 were all highest in the summer (42-61 % of summer PM 2.5 ) (Fig. 5).At each site, the NH + 4 and SO 2− 4 had similar seasonal trends and highest concentrations in the summer (Table 3).In the BTH area, the correlation coefficients (CCs) between NH + 4 and SO 2− 4 were obviously higher than CCs between NH + 4 and NO − 3 and CCs between NO − 3 and SO 2− 4 (Table 4).This means sulfate had been better neutralized by NH + 4 on the filters.The SO 2− 4 in the summer was evidently higher than NO − 3 and also much higher than SO 2− 4 in other seasons at each site.However, SO 2 was usually lowest in the summer, not only because of less fuel combustion but also owing to the highest conversion rate of SO 2 to SO 2− 4 resulting from the strongest photochemical oxidation in the summer (Table 5).Furthermore, the mean concentration of NH 3 in the summer period of Beijing was several times higher than in the other period (Ianniello et al., 2010), and the better thermal stability of sulfate also facilitated the accumulation of SO 2− 4 .The NO − 3 concentrations had not evidently fluctuated as SO 2− 4 .Nitrate is more sensitive to temperature, and higher temperature does not favor the formation of nitrate.According to previous studies in Beijing, a large portion of nitrate, chloride, and ammonium could evaporate from the filters, especially in the summer (Ianniello et al., 2011).
The annual averages of NO − 3 /SO 2− 4 were 0.89, 1.07, 0.75, 0.85, and 0.45 for SDZ, BJ, TJ, SJZ, and CD, respectively.The lowest NO − 3 /SO 2− 4 at CD could also reflect the dominating coal combustion sources for particles.Higher NO − 3 /SO 2− 4 at BJ and SDZ showed that a greater fraction of particles resulted from motor vehicle exhaust.
In some studies, all NO − 3 and SO 2− 4 were assumed to be in the form of NH 4 NO 3 and (NH 4 ) 2 SO 4 (or NH 4 HSO 4 ), and the calculated and measured NH + 4 agreed well (De-Bell et al., 2006;Louie et al., 2005).We also calculated the molar concentrations of positive electric charges of NH + 4 (PEC=NH + 4 /18) and negative electric charges of NO − 3 and SO 2− 4 (NEC=NO − 3 /62+SO 2− 4 /96×2).If all SO 2− 4 was assumed to be in the form of HSO − 4 , then NEC=(NO − 3 /62+ SO 2− 4 /96).At the five sites, the correlations between PEC and    NEC were significant.The seasonal average PEC and NEC at the five sites were depicted in Fig. 6.Except for autumn and winter seasons at CD, NH + 4 was far from enough to match NO − 3 and SO 2− 4 in all four seasons at each site.However, the sum of positive electric charges of all analyzed cations (Na + , NH + 4 , K + , Mg 2+ , Ca 2+ ) well balanced the negative charges of SO 2− 4 in most cases.It indicated the existence of acid.As mentioned above, ammonium nitrate is often formed in areas with high ammonia and nitric acid concentrations, and low sulfate concentrations.According to thermodynamic results from laboratory experiment, NH 4 NO 3 would not be expected to be formed in BTH under summer conditions (Ianniello et al., 2011).But in this study, the highest concentrations of NO − 3 had formed in acidic and ammoniumpoor aerosol in the summer.In the study of Ianniello et al. (2011), very similar seasonal trends for secondary inorganic ions were reported.In the study of Pathak et al. (2009), high concentrations of NO − 3 were also found in the ammoniumpoor PM 2.5 samples at the Beijing site in the summertime.According to the discussions of these above literatures, the nitrate was most probably formed via the heterogeneous hydrolysis of N 2 O 5 on the surface of the moist and acidic aerosols.In addition, the internal mixture of nitrate with sulfate made the deliquescent particles able to easily absorb HNO 3 and reduced the thermodynamic dissociation constant Kn for NH 4 NO 3 .Moreover, the dissociation of NH 4 NO 3 and NH 4 Cl probably resulted in some residue of HNO 3 and HCl on the filters.

Carbonaceous components
The carbonaceous aerosol pollution exhibited obvious seasonal characteristics.For all five sites, the OC and EC concentrations in the autumn and winter were higher than those in the spring and summer (Fig. 7), and OC had higher rates of increase than EC.According to the study of Chow et al. (2004), PM 2.5 emissions from coal-burning sources had high OC abundances and high OC/EC values.The seasonal  fluctuations in the level of OC and EC reflected that more aerosol was emitted from more fuel combustion for heating in the winter.In the winter, SJZ had the highest seasonal OC concentration (47.2 µg m −3 ), and CD had the highest seasonal EC concentration (13.1 µg m −3 ) for the five sites.As mentioned above, the OC and EC concentrations would be much higher in the winter at SJZ if all samples were obtained during that heavy pollution episode.Similar to seasonal trends of OC and EC, the OC/EC values were also lowest in the summer and higher in the spring and autumn at each site, and highest in the winter at every site except CD.
In this study, an EC tracer method was used to calculate the seasonal concentrations of secondary organic carbon (SOC) (Turpin and Huntzicker, 1995;Castro et al., 1999;Cao et al., 2004Cao et al., , 2007)), and four to five samples with the lowest OC/EC ratios were used to calculate coefficients in the POC (primary organic carbon) expression during a given season at each site.The data from days with blowing dust, floating dust, or precipitation were removed before the analysis.The seasonal SOC concentrations and SOC/OC are listed in Table 6.Similarly to OC and EC, SOC and POC (OC-SOC) concentrations were also higher in the autumn and winter and lowest in the summer.However, SOC had more percentages of increase than POC during the winter.In the study of Cao et al. (2007), the SOC concentrations were also higher in the winter for several northern Chinese cities.Compared with secondary inorganic ions, the levels of SOC showed opposite seasonal trends.The stable atmosphere and low temperature, which appeared more frequently during the autumn and winter, could facilitate the accumulation of air pollutants and accelerate the condensation or adsorption of volatile organic compounds onto pre-existing aerosol.Because the profiles of particulate organic compounds were not obtained, the actual processes of secondary formation for OC in the winter could not be revealed in this paper and needed to be deeply discussed.A more detailed description of SOC calculation has been published in another paper (Zhao et al., 2013).

Inorganic elements
The sums of crustal elements (Al, Ca, Fe, Mg, Ti, Ba, and Sr) were higher in the spring and autumn at SDZ, BJ, TJ, and SJZ and only higher in the spring at CD (Fig. 5).According to the recorded weather phenomena, there were two or three days with blowing or floating dust during spring and autumn sampling periods.The sampled aerosol was mainly soil particles from local or other places during those days.Heavy metals can cause serious bodily damage, especially to children (Fergusson, 1990).At present, there is not a concentration standard for heavy metals in ambient particulate matter.The annual or seasonal concentrations of heavy metal elements in PM 2.5 in some other studies are summarized in Table 7.By comparison, all analyzed heavy metals were  evidently detected in urban areas of mainland China; the concentrations of heavy metals were at higher levels, even at SDZ background station.In the BTH area, BJ, TJ, and SJZ had higher heavy metal concentrations, and Zn, Pb, Mn, Cu, Cr, and As at all five sites were all much higher than those in other studies.Except for the lowest concentrations in the summer, there was no common seasonal trend for each element at five sites.Pb and Zn were the most abundant heavy metals in PM 2.5 in the BTH area.The annual average concentrations of Pb plus Zn were 0.266-0.975µg m −3 at all five sites, and were close to 1.0 µg m −3 at TJ and SJZ.Pb had stronger correlations with Zn, K, and EC at each site (Table 8).Pb and Zn have always been found together from minerals to industrial productions and processes.EC is essentially a primary pollutant, and K is mainly from biomass burning and decomposition.The stronger correlations among Pb, Zn, and K indicate their common sources of metal processing and fuel or biomass combustion.

Temporal variations of Beijing PM 2.5 chemical compositions
PM 2.5 has been systematically studied in Beijing in some previous studies.The seasonal average of PM 2.5 chemical species of Beijing from two previous studies (He et al., 2001;Sun et al., 2004) were listed in Table 9.He et al. (2001) sampled PM 2.5 at two urban sites of Beijing (Chegongzhuang and Tsinghua) from 1999 to 2000.The Chegongzhuang site in the study by He et al. (2001) and the BJ site in this study were only 2.4 km apart.The BNU site in the study by Sun et al. (2004) was 6.4 km away from the BJ site in this study.So the PM 2.5 data of the Chegongzhuang site in He et al. (2001) was more comparable with data of this study.Compared with studies of He et al. (2001) and Sun et al. (2004), the concentration levels of secondary inorganic ions and carbon components had significantly varied.The annual concentration of SO 2− 4 and NO − 3 increased from 14.47 and 10.3 µg m −3 in study of He et al. (2001) to 19.07 and 20.47 µg m −3 in this study.The NO − 3 in all four seasons had evidently increased, reflecting the influences by NO x emissions from the sharp increase in motor vehicles.The SO 2− 4 in the spring and summer increased from 10.15 and 17.14 µg m −3 to 16.42 and 33.76 µg m −3 , respectively.However, the SO 2− 4 in the autumn and winter decreased from 12.55 and 24.87 µg m −3 to 11.53 and 14.23 µg m −3 , respectively.The opposite seasonal trends of SO 2− 4 show the evidence of the decreased coal consumption in the winter and the raised level of photochemical reaction in the summer over the past ten years.In addition, NH + 4 and NO − 3 were also varied from lower in the summer and highest in the winter to lower in the winter and highest in the summer.The sum of concentrations of secondary inorganic ions (NH + 4 , NO − 3 , and SO 2− 4 ) was highest in the summer at SDZ, BJ, TJ, and CD, due to stronger photochemical oxidation.Except for autumn and winter seasons at CD, NH + 4 was far from enough to match NO − 3 and SO 2− 4 in all four seasons at each site.So nitric acid or hydrochloric acid probably existed in PM 2.5 on the filters.
For all five sites, the OC and EC concentrations in the spring and summer were lower than those in the autumn and winter, reflecting more aerosols emitted from fuel combustions for heating.Compared with secondary inorganic ions, the levels of SOC showed opposite seasonal variations.The stable atmosphere and low temperatures appearing more frequently during the autumn and winter could facilitate the accumulation of air pollutants and accelerate the condensation or adsorption of volatile organic compounds.The sums of crustal elements (Al, Ca, Fe, Mg, Ti, Ba, and Sr) were higher in the spring and autumn at SDZ, BJ, TJ, and SJZ and only higher in the spring at CD owing to more days with blowing or floating dust.The concentrations of heavy metals were at higher levels in the BTH area, even at SDZ background station.Pb and Zn were the most abundant heavy metals in PM 2.5 .Pb had stronger correlations with Zn, K, and EC at each site, indicating their common sources of metal processing and fuel or biomass combustion.
Compared with results in studies of Beijing carried out ten years before, the concentrations of secondary inorganic ions had changed from highest in the winter to highest in the summer, and the levels of OC and EC in all four seasons had decreased.So the characteristics of aerosol pollution in Beijing have changed from a simple type of coal combustion to a complex type dominated by motor vehicle exhausts.
56 million in 2000 to 11.01 million in 2010.Owing to direct emissions from motor vehicles and secondary formation by Published by Copernicus Publications on behalf of the European Geosciences Union.P. S. Zhao et al.: Concentrations and chemical compositions for PM 2.5

Figure 2 .Fig. 2 .
Figure 2. Wind roses of all four sampling periods at each site.

Figure 4 .
Figure 4. Seasonal PM 2.5 concentrations at five sites.The boxes and whiskers denote the minimums, 25 percentiles, medians, 75 percentiles, and maximums, the dots denote the mean values.

Figure 4 .
Figure 4. Seasonal PM 2.5 concentrations at five sites.The boxes and whiskers denote the minimums, 25 percentiles, medians, 75 percentiles, and maximums, the dots denote the mean values.

35Figure 5 .
Figure 5. Seasonal concentrations of secondary inorganic ions, TC, and crustal elements at five sites.

Figure 6 .
Figure 6.Molar concentrations of positive electric charges of NH 4 + (PEC) and negative electric charges of NO 3 -and SO 4 2-(NEC).

Fig. 5 .
Fig. 5. Seasonal concentrations of secondary inorganic ions, TC, and crustal elements at the five sites.

Table 1 .
Seasonal meteorological conditions at five sampling sites.
Max. for winter at SJZ was 759.6 μg m -3 .Seasonal PM 2.5 concentrations at the five sites.The boxes and whiskers denote the minimums, 25 percentiles, medians, 75 percentiles, and maximums; the dots denote the mean values.* Max. for winter at SJZ was 759.6 µg m −3 . *

Table 2 .
Annual averages concentrations of PM 2.5 species at each site (µg m −3 ).

Table 3 .
Seasonal average concentrations of major components at the five sites (µg m −3 ).

Table 6 .
Seasonal SOC concentrations and SOC / OC ratios by site.

Table 7 .
Comparisons of heavy metal element concentrations in different studies a (µg m −3 ).
a The significant figures of concentration values follow respective paper; "-" represent element concentrations that were not analyzed or given.b Sampling period.c Sampling site.

Table 8 .
Correlation coefficients between Pb and Zn, Pb and K, and Pb and EC.