The time series of meteorological and aerosol data is as follows: wind direction (WD), wind speed (WS), relative humidity (RH), gaseous pollutants (NO2, SO2, O3, and CO), and chemical components (organics, sulfate, nitrate, ammonium, and chloride). The average mass concentration of PM2.5 and PM1 was 225 and 101 μg m–3, respectively. The maximum mass loading of PM2.5 (524 μg m–3) and PM1 (250 μg m–3) happened on 20 December, which was about 7 times that of the China National Ambient Air Quality Standard (24-h mean of 75 μg m–3) and 20 times that of the World Health Organization Standard (24-h mean of 25 μg m–3). Compared with the severe air pollution cases in January 2013 (680 μg m–3) and winter 2015 (626 μg m–3) (Sun et al., 2016), the PM2.5 concentration in the current case was a little lower than that observed during those severe haze episodes. Although the maximum concentration for this case was not higher than that in 2013 and 2015, the severe air pollution lasting for 7 days forced the Beijing municipal to issue the red alert for smog–haze. The meteorological conditions were stagnant during the entire HPEs, as shown in Fig. 1. WS was constantly low, less than 2 m s–1, while the RH remained above 60% and even higher than 80% during the most severe pollution period. The stagnant conditions together with high RH were likely the primary external factor facilitating formation of severe pollution in winter time of Beijing. The same conclusion was also obtained by Sun et al. (2016).
Figure 1. Temporal evolutions of (a) meteorological data, (b) gas pollutants, and (c) particle chemical components in Beijing during 15–23 December 2016.
Comprehensive factors such as the meteorological situations, variations of gaseous pollutants, photochemical reactions, local emissions, and regional transport played different roles in the evolution of HPEs. At the beginning and end of the observation experiment, wind speed reached 4 m s–1, which swept the particles and pollutions away and led to very low mass concentrations for the chemical species and gaseous pollutants (NO2, SO2, and CO) except for higher O3 during these two periods. On the contrary, the chemical components PM1, NO2, and CO demonstrated an increasing trend, and PM1 reached a maximum level on 20 December 2016 under the low wind speed. The upward trends of CO and NO2 resulted from the constant primary emission as well as the stagnant environment. Amounts of particles scattered much more solar radiation back to space, weakening the photochemical reaction. Furthermore, the weak photochemical reaction slowed down O3 production, lessened the NO2 participation into O3 production, and resulted in low levels of O3 and continuous increase in NO2 concentrations. The growing concentrations of chemical components were linked with the accumulation of existing particles by coagulation processes, gas–particle transformation processes, and endless condensation processes from the low-volatility components or gas phase. In a word, interactions between photochemical reactions, gas–particle transformation, and meteorological conditions made the entire pollution process complex.
We noted that the SO2 concentration was relatively stable during the observation, and even decreased on 19 December. This trend was accompanied by increased sulfate concentration. The reversed variations of SO2 and sulfate indicate that additional SO2 was involved in the gas–particle transformation and into the particle phase.
The RH in the whole observation period exhibited fluctuations from 30% to 90%, and the increase of RH was always accompanied by particulate mass concentration increments. Furthermore, the maximum NR-PM1 concentration occurred under the highest RH. RH showed a stronger effect on sulfate formation in wintertime (Sun et al., 2013), which is often associated with fog events and the ability of the precursor SO2 to form sulfate. Additional details on sulfate formation are discussed in later sections.
According to the PM2.5 mass concentration and the temporal evolution of size distributions of organics, sulfate, nitrate, and ammonium, seven different episodes including two non-polluted episodes (NPE1 and NPE2) and five heavy polluted episodes (HPE1–5) were identified. The average mass concentrations of the five main components, sum of organics, sulfate, nitrate, ammonium, and chloride (OSNAC), and PM2.5 in the seven cases are listed in Table 1. The pie charts for the five main components in the seven episodes are also displayed in Fig. 2.
Case Organics Sulfate Nitrate Ammonium Chloride ΣOSNAC PM2.5 HPE1 39.6 8.3 13.4 7.3 3.0 71.6 154.5 HPE2 60.3 14.2 21.2 11.4 3.9 111.0 221.4 HPE3 63.0 20.7 30.0 16.8 5.9 136.4 309.2 HPE4 86.2 31.8 41.6 22.0 6.6 188.2 432.7 HPE5 75.3 30.7 43.9 19.0 4.7 173.6 409.9 NPE1 2.6 1.0 0.5 0.5 0.1 4.7 8.6 NPE2 5.8 2.3 0.6 0.9 0.1 9.7 20.7
Table 1. Average mass concentration (μg m–3) during the different cases
Table 1 shows that the concentrations of OSNAC in the five HPE cases are 71.6, 111.0, 136.0, 188.2, and 173.6 μg m–3, respectively, while those in the two NPE cases were 4.7 and 9.7 μg m–3 at the beginning and end of the pollution events. HPE4 is the most serious polluted episode, being about 2.6 times that of HPE1 and even 40 times that of NPE1.
The mass concentration of PM2.5 varied widely among the evolution processes. Five formation rates of PM2.5 were calculated based on the mass increments of PM2.5 from the lowest value to the highest value during the five stages, marked as F1, F2, F3, F4, and F5 in Fig. 2. The PM2.5 formation rates in the five stages were 26, 22, 22, 32, and 67 μg m–3 h–1, respectively. Similar formation rates at the first three stages were obtained, indicating a similar secondary aerosol formation mechanism. But the formation rates in F4 and F5 were higher than those in the former stages, and reached the highest in F5. It is noted that the concentrations of CO and NO2 in F4 and F5 also increased much more than those in other three stages. As primary emission makers are coal combustion burning and traffic emission, these two gaseous pollutants are worthy being considered. Meanwhile, the sulfate exhibited dramatic increases during F4 and F5 stages, too. The aqueous-phase reaction of sulfate particularly under high NO2 conditions would be the main reason for this higher formation rate (Wang et al., 2016). On the other hand, the formation rates in F4 and F5 in this study were much higher than those observed during severe haze episodes in October 2014 (2.0–4.4 μg m–3 h–1) (Yang et al., 2015), but lower than that during the January 2013 haze episodes (88 μg m–3 h–1) (Wang et al., 2014; Zhang et al., 2015), and the highest formation rates in winter 2015 were 115 and 165 μg m–3 h–1 (Sun et al., 2016). The lower formation rates in this case compared with those in 2013 and 2015 can be mainly attributed to the weak primary source emission, weak photochemical reactions with lower O3 concentration, and less regional transport under the stagnant meteorological condition.
The proportions of the five main species in OSNAC are plotted in pie charts (Fig. 2). Totally, no dramatic difference in chemical composition was noted among the five HPEs, whereas about half was obtained from organics, 20% nitrate, 15% sulfate, 10% ammonium, and 4% chloride. The chemical constitution pattern was similar to the amount mentioned in former literatures (Huang et al., 2010; Sun et al., 2010, 2013, 2016; Zhang et al., 2011, 2013). The pattern infers the same source types, major atmospheric processes, and formation mechanisms. Organics corresponded to 43%–55% in the polluted cases and even 60% in the non-polluted episodes. Another significant species (nitrate) was around 20% of OSNAC in the polluted events, but only 10% in the clean episodes. Sulfate was found to be more important in the NPEs than in the HPEs, with a proportion of 20% in NPEs and 10% in HPEs. The clean period was more inclined to new particle formation process, with sulfate and organics as the necessary species involved in the formation (Kulmala et al., 2004; Wu et al., 2007; Shen et al., 2011; Zhang Y. M. et al., 2011). This situation resulted in the higher contribution to the NR-PM1. Compared with organics, sulfate, and nitrate, the ammonium played a relatively stable role in chemical compositions; its percentages in both HPEs and NPEs were 11%.
The relationship between mass concentration and diameter follows the cubic formula. A slight growth of small particles in size would significantly increase the mass concentration. To investigate the variations of particle size throughout the polluted periods, examination of HPEs is helpful for understanding the heavy haze and its formation mechanism. Figure 3 displays the average mass-resolved size distributions of organics, sulfate, nitrate, and ammonium of the selected seven episodes. The black dashed lines indicate the peak positions of nitrate. The size distributions of organics were broader than those of the other inorganic species, and organics dominated a large part of the small particles. The peak sizes of the chemical species displayed an accumulation mode of around 400–600 nm. This result indicates that most particles were aged in the atmosphere.
Figure 3. Average size distributions of organics, sulfate, nitrate, and ammonium particles during HPEs and NPEs. The black dashed lines denote the peak positions of nitrate.
Different size distribution patterns were also found between NPEs and HPEs (Fig. 3) . In the two NPEs, all the particles less than 100 nm consisted largely of organics, whereas organics, sulfate, nitrate, and ammonium contributed to the particle composition measured between 200 and 800 nm. No sulfate but merely organics was monitored for small particles. These results suggest that the newly acquired small particles may have been derived from primary organic aerosol (POA) emission, such as vehicles or/and the secondary transformation process of volatile organic components (VOCs). However, sulfate is seldom involved in the original stage of the new particle formation. Accumulation mode sulfate, nitrate, and ammonium particles might be formed by condensation processes of their precursor gases SO2, NO2, and NH3 that were condensed on the surface of existing organic particles.
In terms of the five HPEs, the peak sizes of all the species grew gradually (Fig. 3). For example, the peak size of nitrate in HPE1 was 400 nm, with an average mass concentration of 13.4 μg m–3. In HPE2, the peak size grew to around 500 nm, and the mass loading was 21.2 μg m–3. Meanwhile, the size peaks and mass concentrations in HPE3, HPE4, and HPE5 were 580 nm and 30 μg m–3, 600 nm and 41.6 μg m–3, and 600 nm and 43.9 μg m–3, respectively. Oxidization, condensation, and coagulation processes, as well as heterogeneous reactions might be significant factors for particle growth. This topic remains a considerable challenge for intercepting the formation mechanism of heavily polluted episodes (Holmes, 2007; Chan and Yao, 2008; Yue et al., 2010; Liang et al., 2016; Wang et al., 2016).
The above section mentioned that the mass concentrations of the NR-PM1 and other species increased continuously during the whole HPEs. In fact, daily variations in the mass concentrations were also evident. The diurnal patterns of different NR-PM1 species are influenced by multiple factors, such as boundary layer development, photochemical activities, local emissions, regional transport, and meteorological conditions (Aiken et al., 2008; Jimenez et al., 2009; Zhang et al., 2009, 2013; Tie et al., 2015). Thus, examination of diurnal cycle is useful in the study of the particle formation mechanism.
The average diurnal cycles of main chemical species, precursor gases, and RH are plotted in Fig. 4. Organics exhibited a pronounced tri-modal diurnal pattern, which peaked from midnight to early morning for 7 h (2300–0500 BT), at noon, and in the evening. The development of boundary layer resulted in the descent of organics from 0600 BT in the morning. The noon peak might be linked to enhanced cooking emissions during lunch, and the evening peak was likely related to multiple factors such as dinner cooking and extensive traffic emission during the rush hours. This diurnal cycle of organics is fairly similar to those reported in 2006 in Beijing (Sun et al., 2010) and the winter diurnal cycles (Zhang et al., 2012).
Figure 4. Diurnal cycles of chemical species mass concentrations in (a) NR-PM1; (b) sulfate, relative gases, and RH; and (c) nitrate and nitrogen dioxide.
For sulfate and chloride, obvious peaks in the early morning and evening were observed, and the concentrations decreased to the lowest during noon time. The decreased boundary layer and the oxidization process cannot intercept the early morning peak. To clarify the possible reasons for this pattern, the diurnal cycles of sulfate, SO2, O3, and RH were compared in Fig. 4b. Sulfate followed the diurnal cycles of RH but appeared an inverse variation to SO2 and O3. This phenomenon infers three things. First, the SO2 is not emitted locally but transported from distant locations by some mechanisms. Second, the aqueous reaction in the atmosphere might be a critical means to form particulate sulfate because of the relative similar patterns of sulfate and RH. Most of SO2 is transformed to a particulate sulfate. Third, the oxidization process is not important to the sulfate. For instance, nitrate began to increase at 1000 BT and climbed to the first peak at noon, lasting for 5 h. Then, the process reached a second peak at 2000 BT. The noon peak was related to the oxidization reactions, and the evening peak was linked with the rush hour. Figure 4c displays the diurnal cycles of nitrate and NO2. It is revealed that the nitrate, NO2, and O3 involved similar diurnal patterns. The oxidization process of NO2 was the main approach to nitrate formation. Although the “even–odd tag forbidden rule” was implemented, the total traffic number remained large. Traffic was considered as the main local emission source of NO2, and NO2 was emitted continuously. The severely polluted environment weakened the solar radiation and lessened the amount of NO2 involved in O3 formation. This condition resulted in increased NO2 participants in the oxidization process and thus produced a large nitrate amount.
Organic species were apparently an important aerosol component in the studied haze event. AMS mass spectral matrix of OA with Positive Matrix Factorization (PMF; Paatero and Tapper, 1994) uses the analysis and evaluation tool described by Ulbrich et al. (2009). The PMF analysis based on the Q-AMS MS dataset in our experiment was first investigated for 1–7 factors and the rotational parameter (FPEAK) varying from –1 to 1 (step = 0.5) (see supplemental material Fig. S1, available online). The scaled residual values were also employed to make decision for the factor numbers (Fig. S2). The rotation was not applied in our data because there was no improvement for the Q/Qexp values (Fig. S2). More detailed information for the final decision of optimum factor numbers were discussed in supplement files (Figs. S3–S7).
Based on the complex consideration and analysis, five distinct OA components were identified in this experiment in accordance with their unique MS profiles, correlation with tracers, and diurnal variations (Ulbrich et al., 2009). Cooking organic aerosol (COA), hydrocarbon organic aerosol (HOA), and coal combustion organic aerosol (CCOA) can be regarded as surrogates of primary organic aerosol (POA), whereas the two oxygenated organic aerosol (OOA I and OOA II) are SOA (Jimenez et al., 2009; Ng et al., 2010).
Cooking emission is a major organic aerosol source in Beijing because of unique Chinese cooking habits and culture (He et al., 2004). A high m/z 55/57 ratio is widely used as an indicator for the presence of COA at various urban sites on the basis of previous studies (Mohr et al., 2009; Sun et al., 2010, 2013; Xu et al., 2016; Zhang et al., 2017). In this study, the high ratio of m/z 55/57 and high fraction of m/z 55 in identified COA mass spectra were observed. As a kind of local emission, two pronounced peaks occurred at approximately 1200 and 1800 BT, consistent with the traditional meal times, in the diurnal cycle of this component, especially during the clean periods when the peaks were very obvious in the time series of COA (Fig. 5). On average, the COA concentration during the whole experiment was 8.9 μg m–3, accounting for 18% of the total organic mass. This percentage is similar to the result in Beijing in 2006 (Sun et al., 2010) and 2008 (Huang et al., 2010; Zhang et al., 2013). The result also coincides with those in other cities, such as Lanzhou (Zhang et al., 2017).
Figure 5. (a) Mass profiles, (b) time series and related species (right axis in grey lines), and (c) diurnal cycles of subtype organics.
HOA has been widely identified by several previous studies in Beijing (Sun et al., 2010, 2013; Zhang J. K. et al., 2014). Most of these studies showed similar MS patterns to those of primary combustion sources, such as gasoline and diesel vehicle exhausts (Mohr et al., 2009; Canagaratna et al., 2010). In this study, the MS pattern of HOA was primarily characterized by high m/z 43. Other m/z 29, 41, 43 55, and 57 were also distinguishable (Fig. 5a). The mean mass loading of HOA in this experiment was 7.8 μg m–3 and about 16% of OA. This percentage is very close to the values in the previous studies in Beijing (Huang et al., 2010; Sun et al., 2013). In the haze red-alert period, all the traffic flows worse than European III were banned. Furthermore, the odd–even tag limitation regulation reduced emissions and resulted in the decreased proportions of HOA in OA. On the other hand, in the time series, the diurnal cycle of another oxygenated organics aerosol (OOA II) displayed a similar pattern to that of nitrate and NO2, indicating that OOA II would be a kind of oxygenated organics from aged HOA.
Coal combustion is the most important source of winter heating in northern China. CCOA, another important POA factor, accounted for 23% of the total organic mass in this study. The mean mass concentration was 11.4 μg m–3. The mass spectrum of CCOA contained high-intensity signals at m/z 41, 43, 44, 55, 57, 69, 91, and 115, similar to those in other previous studies (Hu et al., 2013; Zhang J. K. et al., 2014), especially for the distinct signals at the latter two ion fragments, which had been used as good tracer ions for CCOA (Fig. 5). Another m/z 60 was also prominent in the CCOA MS profile, as the typical tracer of biomass burning emission. In this case, our reported CCOA might be a kind of mixtures of both coal combustion and biomass burning organic aerosols. This mixture requires more profound research. The temporal variation of CCOA showed a similar trend with the time series of chloride. Same results were also discussed in other works (Xu et al., 2016; Zhang et al., 2017). As CO is another common marker of coal combustion, diurnal cycles of CO in gaseous phase and chloride displayed higher mass loading at night and lower at daytime, which was similar to this resolved factor. Thus, taking this factor as coal combustion organic aerosol (CCOA) was reasonable (Fig. 5c).
The two OOA subtypes were distinguished by prominent contributions at m/z 43 and 44 in the MS profiles. These contributions indicate prevalence of the oxidized organic compounds. A relatively higher signal at m/z 43 corresponded to the less oxidized OA components and was referred to as OOA I (Fig. 5) (Sun et al., 2010; Ng et al., 2011). A relatively higher signal at m/z 44 than that at m/z 43 generally corresponded to the more oxidized OA component and was referred to as OOA II. The time series of the two OOA components correlated well with those of the inorganic secondary species (sulfate and nitrate). OOA I correlated well with sulfate, whereas OOA II correlated well with nitrate in this study. On average, OOA I accounted for 19% of the total OA mass, with 9.7 μg m–3 in the mass concentration. OOA II occupied 24% of the total OA, which was also the highest contributor of the total OA. Overall, the two OOA components were 43% of the total OA. This result is comparable to some other findings during wintertime (Sun et al., 2013; Zhang et al., 2013, 2017; Xu et al., 2016).
As discussed previously, aqueous reaction became a dominant process for the sulfate formation, whereas the oxidized reactions considerably influenced the behavior of nitrate. The diurnal cycle of OOA II matched well with that of nitrate and NO2, exhibiting a higher concentration between 1500 and 2000 BT. Nitrate and NO2 as markers of traffic source were widely accepted, and the good correlations among them imply that the OOA II was likely a kind of aged/oxygenated organic of HOA from traffic emission. It is well known that Beijing is surrounded on three sides by mountains , and on one side by a plain. Traffic emission, as a dominant local source, was continuous. During the heavy polluted episodes, stagnant wind brought the local traffic emitted pollution to and from the urban area and mountain sides; the particles were also aged or oxygenated during the entire polluted process, resulting in the transformation of HOA to OOA II. In all, during the HPEs, the majority of OA was derived from primary organics, including COA, HOA, and CCOA, and the different kinds of secondary OA transformed from CCOA and HOA, which also originated from the POA.
Given the PMF results, the OAs were separated into five subtypes of components, to infer different sources or evolution processes. During the entire polluted period, different OAs would perform in different ways. To survey the roles of these OAs, average mass concentrations of the five organics during the seven periods and the whole period are shown in Table 2. The percentages of these five organic subtypes in OA are plotted in Fig. 6.
COA HOA CCOA OOA I OOA II HPE1 5.3 7.4 14.4 3.6 11.1 HPE2 10.4 10.5 16.2 6.6 17.1 HPE3 13.9 10.3 13.2 9.9 16.8 HPE4 15.7 13.3 23.2 18.6 17.5 HPE5 11.7 11.8 15.5 22.1 18.0 NPE1 1.0 0.3 0.3 0.4 0.8 NPE2 1.2 0.5 0.5 0.6 1.7 All 8.9 7.8 11.4 9.7 11.7
Table 2. Average mass concentrations (μg m–3) of OA components in different cases
Table 2 shows that all the OA components display an increasing trend from HPE1 to HPE5. The POAs (including COA, HOA, and CCOA) reached the highest values in HPE4 and decreased in HPE5. The stagnant meteorological conditions from HPE1 to HPE4 and continuous emissions from local sources drived the accumulations of COA, HOA, and CCOA. In HPE5, a gentle wind blew some of the accumulations away. This occurrence reduced their mass loadings. For the two types of SOAs, the maximum concentrations were noted at HPE5, indicating that the oxidization process from VOCs or POAs to SOAs was advanced constantly. For the OA increasing rates, the COA, HOA, and OOA II amounts increased faster from HPE1 to HPE2. Moreover, the CCOA and OOA I grew much further from HPE3 to HPE4. The OOA II amount was relatively constant from HPE2 to HPE5. During the HPEs, O3 remained at a low level and weakened the atmosphere’s oxidizing ability. The OOA II was generally considered as a representative of additional oxidized organic aerosols. Thus, we concluded that the oxidization or aging process during the HPEs was insignificant.
During the different episodes, the five OA components played distinct roles (Fig. 6). In HPE1, CCOA was the dominant aerosol, occupying about 35% of OA. By contrast, OOA I occupied the largest proportion in HPE5. The high concentration of CCOA in HPE1 possibly came from the regional transport at the beginning of the haze red-alert event. As the meteorological conditions stagnated further, the pollution worsened. The regional transport of CCOA also slowed down. As a result, the source of CCOA reduced in the later HPEs. With the pollution being serious, continuous emissions of COA and HOA from local cooking and traffic activities occurred because of their increasing contribution to the total OA. On the two clean periods, COA and OOA II were about 35% and 30%, respectively, in NPE1, and 25% and 38%, respectively, in NPE2. Low particulate matter concentration, intensive solar radiation, and high O3 concentration may provide an optimum environment for the photochemical aging process during NPE2. This aspect led to a high contribution of OOA II.