The field campaign was conducted in the January 2013, and the sampling site was on the rooftop of a building in the Chinese Research Academy of Environmental Sciences (CRAES, 40∘2′29.46′′ N, 116∘24′51.00′′ E), which is located outside the 4 km north of the 5th Ring Road and 15 km from the city center as shown in Fig. 1.

The wind rose of Beijing in January is depicted in Fig. 2, and the corresponding calm wind frequencies are listed in Table 1. Firstly, the prevailing wind directions (frequency higher than 10 %) were SW (12.50 %), ENE (11.33 %), N (10.94 %) and WSW (10.74 %). Compared with the statistical results of Zhao et al. (2013) for 2009 and 2010, the frequency of WSW, SW and SSW were higher in this study. Su et al. (2004) concluded that southwest transport pathway was one of the three major pathways for outside pollutants being transported to Beijing. The CMAQ model simulation of secondary aerosols around Beijing in summer of 2003 by Liu et al. (2005) suggested that when wind from southern and southwestern directions prevails in Beijing, high concentrations of vapor and NH3 was brought in, which would enhanc the efficiency of the secondary aerosol production. Secondly, the average temperature in January 2013 (-5.1 ∘C) was lower than that in 2009 and 2010 (-2.0 ∘C) (Zhao et al., 2013). Lower temperature would lead to more demand for fossil fuel, and more air pollutants could be emitted. Thirdly, the calm wind frequency (6.64 %) was higher than that in 2009 and 2010 (3.21 %), which was perhaps an important reason for the air pollution episode in January 2013.

The hourly concentrations of Cl-, NO3-, SO42-, Na+, NH4+, K+, Mg2+ and Ca2+ associated with PM2.5 were simultaneously measured by an ambient ion monitor (Model URG 9000B, URG Corporation, USA). This instrument draws air in through a PM2.5 sharp-cut cyclone at a volumetric-flow controlled rate of 3 Lmin-1 to remove the larger particles from the air stream. Then the air stream is drawn through a liquid diffusion denuder where water-soluble acid gases (e.g., SO2, NH3, and HNO3) are removed. In order to achieve high collection efficiencies, the particles-laden air stream next enters an aerosol super-saturation chamber to enhance particle growth. An inertial particle separator collects these enlarged particles, which are dissolved in water solution and then injected into the ion chromatograph. The detection limits of ions were shown in Table 2.

Hourly concentrations of PM2.5 were obtained from Chaoyang Meteorological Bureau, which is 10 km away from the CRAES sampling site.

Data were analyzed using SPSS 17.0 for Windows (SPSS Inc., 2008) for the correlation analysis and linear regression. EPA PMF 4.2 (USEPA, 2011) was applied to identify the potential contributors to the ion species. A detailed introduction to Positive Matrix Factorization (PMF) is shown in the Supplement.

Water-soluble ions comprise a large part of aerosol particles and play an important role in the atmosphere. Na+, Mg2+, and Ca2+ are mainly from crustal sources, such as re-suspended road dust, soil dust, and construction dust, and SO42-, NO3-, and NH4+ represent the secondary pollution sources from the transformation of their precursors of NH3, SO2 and NOx (Wang et al., 2005). Cl- is usually considered to be from coal combustion (He et al., 2001), and K+ is from biomass burning (Duan et al., 2004).

The concentrations of ionic species in the whole sampling period were shown in Table 3. Hourly concentrations of individual ions from AIM measurements varied significantly in a broad range. Emission intensities of gaseous precursors, oxidation or conversion rate, local atmospheric mixing, and regional transport all would affect their concentration levels (Hu et al., 2014).

Wang et al. (2014a) reported that five air pollution episodes were identified in the Beijing-Tianjin-Hebei (Jing-Jin-Ji) area in this time period, and the two most severe episodes occurred during 9–15 and 25–31 January. This study also analyzed the continuous variations of ions in Fig. 3, finding peak values of SO42-, NO3- and Cl- were observed on four periods: 10–15, 18–20, 21–24, and 26–30 January. NH4+ and K+ had only peak concentrations during 10–15 January. Concentrations of Na+, Ca2+ and Mg2+ showed no significant variations before 28 January. In the last several days, the concentrations of Ca2+ and Mg2+ decreased dramatically. Most constructive activities related with the emission of Ca2+ and Mg2+ were suspended by government during this episode. This would perhaps be the reason of lower concentrations of Ca2+ and Mg2+ after 28 January.

To better understand the ion species characterizations in January 2013, the main ion concentrations measured by manual sampling in the winter of Beijing from other study were summarized in Table 4. We can find that the concentrations of SO42-, NO3-, Cl- and NH4+ were comparable in different studies.

Huang et al. (2014) collected PM2.5 samples during 4–9 January (light air pollution) and 10–15 January (heavy air pollution), and analyzed daily concentrations of ions. We calculated the mean ions concentrations during corresponding time periods as shown in Table 5. The concentrations of Cl-, NO3-, SO42- and NH4+ in these two studies are comparable, except the NH4+ during light air pollution period (4–9 January). Large differences were observed for some ions with relatively low levels, such as Na+, K+, Mg2+, Ca2+.

Chinese Ministry of Environmental Protection promulgated “Technical Regulation on Ambient Air Quality Index (on trial)” in 2012, and regulated the AQI values and corresponding concentrations of air pollutants. The regulation divided the daily concentrations of PM2.5 into six classes: 0–35, 35–75, 75–115, 115–150, 150–250 and larger than 250 µgm-3, corresponding AQI as follows: 0–50 (excellent), 51–100 (good), 101–150 (light pollution), 151–200 (medium pollution), 201–300 (heavy pollution) and more than 300 (severe pollution). According to the PM2.5 concentration classification, this study divided the ions concentrations into six categories and compared their distribution trend in Fig. 4. Except Mg2+ and Ca2+, all ions concentrations showed an increasing trend with the deterioration of PM2.5 pollution, indicating the ion species (except Mg2+ and Ca2+) contributed to the increase of PM2.5 concentration.

To further investigate the relative abundance variation trend, percentage of ion species in different PM2.5 concentration ranges were calculated as shown in Fig. 5. Cl-, NO3-, Na+, K+, Mg2+ and Ca2+ showed a declining trend along with the increase of PM2.5 levels, while SO42- and NH4+ took up increasing proportions of total ions (the increasing of NH4+ abundance was not significant, from 12.7 % under the PM2.5 concentration less than 35 µgm-3 to 18.5 % under the PM2.5 concentration larger than 250 µgm-3). The different variations of species percentage in total ions indicate that SO42- and NH4+ may contribute more to the PM2.5 pollution. This result is similar with the conclusion of Wang et al. (2014a), who considered the increasing proportion of sulfate enhanced particle hygroscopicity and thereby accelerating formation of the haze pollution.

The mass ratio of [NO3-]/[SO42-] has been used as an indicator of the relative importance of stationary vs. mobile sources of sulfur and nitrogen in the atmosphere. Higher [NO3-]/[SO42-] mass ratios indicate the predominance of mobile sources over stationary sources of pollutants (Arimoto et al., 1996).

The average mass ratio of [NO3-]/[SO42-] during the observation period was 0.68±0.44, which was comparable with the results of Wang et al. (2014a), but higher than the value measured during the winters of 2001–2003 in Beijing (0.49). Wang et al. (2005) suggested that the contribution of mobile sources (e.g., motor vehicles) increased in Beijing, in accord with the adjustment of the energy structure in recent years.

The [NO3-]/[SO42-] ratio was also classified according to the PM2.5 concentration ranges. As shown in Fig. 6, we can find that the [NO3-]/[SO42-] ratio decrease with the increase of PM2.5 level. When the PM2.5 concentrations were lower than 75 µgm-3, the ratio were larger than 1, indicating predominance of mobile source over stationary source of pollutants. When PM2.5 concentration increased, the stationary sources would contribute more pollutants than mobile sources. The results also confirmed that SO42- was one of the most important compositions which contributed to the PM2.5 heavy pollution.

The diurnal variations of ion species in PM2.5 in the sampling period were shown in Fig. 7. On the whole, SO42-, NO3-, Cl-, K+ and NH4+ exhibited similar diurnal patterns with a broad nighttime maximum and a relatively low concentration at daytime, while the variations of Na+, Mg2+ and Ca2+ were small. Compared with another highly resolved measurements for ion species at the same site during summer of 2008 (Gao et al., 2013), the diurnal cycle of SO42- in this study was completely opposite, i.e., the highest concentration of SO42- occurred at daytime and the lowest at night during summer campaign of 2008. Also, for some ions, such as SO42-, Cl- and K+, high levels at night and low concentrations at daytime could be the characterizations of coal combustion. The sampling site is located outside the 5th ring of Beijing, and there still exists considerable coal combustion boilers for domestic heating. The emission from these coal combustion facilities would be one of the important reasons that led to the diurnal variations.

Gao et al. (2013) found the SO42- concentrations in summer rapidly increased from the early morning to the late afternoon synchronously with the enhancement of solar radiation and the O3 and SO2 concentrations. In this study, the O3 concentration was low, especially in the pollution episodes (<15 µgm-3) (Wang et al., 2014a). Such low O3 levels cannot supply enough oxidizing capacity in the atmosphere; therefore, other oxidation reactions, regional transport at night or the impact of meteorological may also be responsible for the high levels of SO42-, NH4+ and NO3-.

The chemical forms of those major ions, i.e. SO42-, NO3-, Cl, NH4+, Ca2+, and K+, in aerosols in Beijing were identified by bivariate correlations. Verma et al. (2010) and Wang et al. (2005) used bivariate correlations to identify the possible chemical forms of ions in PM2.5. Since concentrations of all the ions are not normally distributed (k-s test, p<0.05), Spearman Correlation analysis was applied. Table 5 showed the correlation coefficients among these major ions.

Ammonia is an important alkaline gas in the atmosphere. Ammonia in air is neutralized first by H2SO4 to form (NH4)2SO4 or NH4HSO4, and then the remaining is neutralized by reaction with HNO3 to form NH4-NO3.

In this study, among all cations, NH4+ was highly significantly correlated with NO3- and SO42- as shown in Table 6. The slope of the regression between NH4+ and SO42- (µeq vs. µeq) for the whole data set was 1.33, indicating the complete neutralization of SO42- by NH4+, and suggested that (NH4)2SO4, instead of NH4HSO4, was the major species formed by SO42- and NH4+. Moreover, the slope between NH4+ and the sum of SO42- and NO3- (µeq vs. µeq) was 0.99, further revealing NH4+ completely neutralized by both SO42- and NO3-. Thus only Cl- can supply negative charge to balance the Na+, K+. Considering the correlation coefficients of Cl- between Na+, K+, were comparable, the mixture of NaCl and KCl were likely to be the major form of ions (NH4)2SO4 and NH4NO3.

The mass concentrations of all ions species were input into EPA PMF 4.2 model to identify the potential sources. Four factors were isolated, representing potential sources, including secondary nitrate, secondary sulfate, coal combustion and biomass burning, and fugitive dust. The factor profiles of PM2.5 bound ion species at the monitoring site are shown in Fig. 8.

Factor 1 and factor 2 were considered as secondary nitrate and sulfate, with high loading of NO3- and NH4+ in factor 1, as well as SO42- and NH4+ in factor 2. NO3- is mainly converted from ambient NOx, which is emitted by both vehicle exhaust and fossil fuel combustion. The precursor of aerosol SO42- is SO2, which may originate from biomass burning and fossil fuel combustion. Cl- and K+ were found to be the main contributors to factor 3, originating mainly from coal and biomass combustion. Factor 4 is identified as fugitive dust, such as soil dust, constructive dust, and paved or unpaved road dust, including high contributions of Ca2+ and Mg2+.

The factor contributions to the ion concentrations are illustrated in Fig. 9. Factors 1 and 2 were the biggest contributors to NO3- and SO42-, respectively. Both factors also comparably contribute to the concentration of NH4+. Factor 3 was the main source for Cl- and K+, while factors 1 and 2 also took some portions for Cl- and K+, respectively. Na+, Mg2+ and Ca2+ were mostly contributed by factor 4.