PM_2.5 Chemical Compositions and Aerosol Optical Properties in Beijing during the Late Fall

Abstract

Daily PM_2.5 mass concentrations and chemical compositions together with the aerosol optical properties were measured from 8–28 November 2011 in Beijing. PM_2.5 mass concentration varied from 15.6–237.5 μg∙m⁻³ and showed a mean value of 111.2 ± 73.4 μg∙m⁻³. Organic matter, NH₄NO₃ and (NH₄)₂SO₄ were the major constituents of PM_2.5, accounting for 39.4%, 15.4%, and 14.9% of the total mass, respectively, while fine soil, chloride salt, and elemental carbon together accounted for 27.7%. Daily scattering and absorption coefficients (σ_sc and σ_ap) were in the range of 31.1–667 Mm⁻¹ and 8.24–158.0 Mm⁻¹, with mean values of 270 ± 200 Mm⁻¹ and 74.3 ± 43.4 Mm⁻¹. Significant increases in σ_sc and σ_ap were observed during the pollution accumulation episodes. The revised IMPROVE algorithm was applied to estimate the extinction coefficient (b_ext). On average, organic matter was the largest contributor, accounting for 44.6% of b_ext, while (NH₄)₂SO₄, NH₄NO₃, elemental carbon, and fine soil accounted for 16.3% 18.0%, 18.6%, and 2.34% of b_ext, respectively. Nevertheless, the contributions of (NH₄)₂SO₄ and NH₄NO₃ were significantly higher during the heavy pollution periods than those on clean days. Typical pollution episodes were also explored, and it has been characterized that secondary formation of inorganic compounds is more important than carbonaceous pollution for visibility impairment in Beijing.

1. Introduction

The atmospheric visibility in China has been deteriorating with economic growth during the past 40 years [1,2]. Visibility impairment is resulted from light scattering and absorption by atmospheric particles and gases, especially from the scattering by the particles of similar size range as the wavelength range of visible light. Numerous studies have indicated that the fine particles caused most of the visibility impairment, while the influence of gas and coarse particles on visibility degradation was commonly weak [3,4]. Moreover, meteorological parameters, such as wind, rain, and temperature, especially the relative humidity, have their contributions as well [5]. Additionally, atmospheric particles have significant impacts on climate change, which is one of the greatest sources of uncertainty in estimating the direct radiative forcing [6,7]. Generally, inorganic and organic aerosols have a cooling effect on climate by scattering light, while black carbon (BC) has a warming effect by absorbing light.

Previous studies usually determined the chemical compositions and optical properties of atmospheric aerosols separately [8,9,10]. Yang et al. [11] compared the characteristics of PM_2.5 in representative megacities of China. Results showed that five major species including organic carbon (OC), elemental carbon (EC), SO₄²⁻, NO₃⁻, and NH₄⁺ amounted to 54%–59% of PM_2.5 mass in Beijing, Chongqing and Guangzhou, and the percentages of total carbon and secondary inorganic ions were very close, implying that both primary and secondary particles had a significant contribution to the PM_2.5 mass. Recently, Zhang et al. [12] indicated that secondary inorganic aerosols, mineral dust and organic matter (OM) each accounted for about 20% of PM_2.5 in Beijing, respectively, suggesting both primary and secondary components of PM_2.5 in Beijing were equally important.

The parameters of light extinction (b_ext) can be measured directly using optical instruments such as an integrating nephelometer for the light scattering coefficient (σ_sc), or an aethalometer for the absorption coefficients (σ_ap). Optical properties of PM_2.5 have been conducted in Beijing, Shanghai, and Guangzhou [13,14,15,16,17]. In recent years, a few studies have focused on the relationship between the chemical compositions and optical properties of aerosols [18,19]. Results showed that NH₄⁺, SO₄²⁻, NO₃⁻, and OC are the main contributors to aerosol scattering, and the light absorption coefficients had strong linear correlations with EC in Shanghai. Furthermore, as an alternative method, the Interagency Monitoring of Protected Visual Environments (IMPROVE) formula could be used to estimate b_ext based on the chemical compositions of particulate matter (PM). The original and revised IMPROVE algorithms for estimating b_ext were developed from the particle data at 21 rural/remote sites with low scattering coefficients. Furthermore, the IMPROVE formula usually assumes externally mixed status of PM and fixed mass extinction efficiency for each species [20]. However, the actual mass absorption efficiency (MAE) and mass scattering efficiency (MSE) were not constant due to large temporal and spatial variations of chemical compositions of PM. Thus, it is necessary to evaluate the applicability of the IMPROVE formula to the calculation of b_ext in more polluted urban Beijing in China.

In the present study, aerosol optical properties including σ_sc and σ_ap, as well as the chemical compositions of PM_2.5 were measured in Beijing during the late fall of 2011. The applicability of the IMPROVE formula to the calculation of b_ext was then evaluated, and the contribution of PM_2.5 chemical compositions to visibility impairment was discussed. In addition, the formation mechanisms of typical pollution episodes during the late fall were also explored.

2. Experimental Section

2.1. Sampling

Samples were collected from 8–28 November 2011 at the campus of Tsinghua University (39°98ʹN, 116°32ʹE) in urban Beijing, about 600 m north of the Fourth Ring Road. The campus is mainly surrounded by residential areas without significant factory emissions. Beijing is connected to the industrialized cities of the Great North China Plain in the South, and surrounded by the Yanshan Mountains in the west, north, and northeast.

Daily 23 h integrated PM_2.5 samples were collected using a five-channel Spiral Ambient Speciation Sampler (SASS, MetOne Inc., Grants Pass, OR, USA) with a flow rate of 6.7 L∙min⁻¹. The first channel was used for PM_2.5 mass and elemental analysis with a 47 mm Teflon filter. The second channel collected the particles for the analysis of water-soluble inorganic ions with a 47 mm Teflon filter. The third channel was used to collect PM_2.5 on quartz filters for organic and elemental carbon analysis.

2.2. Gravimetric and Chemical Analysis

The PM_2.5 mass concentrations were determined using an electronic balance with a detection limit of 1 µg (Sartorius, Göttingen, Germany) after stabilizing at a constant temperature (22 ± 5 °C) and relative humidity (40% ± 5%) for 24 h.

Four anions (SO₄²⁻, NO₃⁻, Cl⁻, and F⁻) and five cations (Na⁺, NH₄⁺, K⁺, Mg²⁺, and Ca²⁺) were determined in aqueous extracts of the filters by Ion chromatography (ICS-1000 and ICS-2000 for anion and cation, respectively, Dionex, Sunnyvale, CA, USA). To extract the water-soluble ions from the Teflon filters, each sample was extracted twice with 7.5 ml Milli-Q water via an ultrasonic bath for 20 min, and then filtered through a 0.45 μm PTFE syringe filter and stored in a refrigerator at 4 °C until analysis.

A 0.5 cm² punch from each quartz filter was analyzed for OC and EC using a DRI Model 2001 Thermal/Optical Carbon Analyzer (Atmoslytic Inc., Calabasas, CA, USA), following the IMPROVE thermal optical reflectance (TOR) protocol [21].

Crustal elements including Al, Si, Ca, Fe, and Ti were analyzed by Energy Dispersive X-ray fluorescence spectrometry (Epsilon 5 ED-XRF, PANalytical Company, Almelo, The Netherlands) on Teflon filters. Quality assurance/Quality Control (QA/QC) procedures of the XRF analysis procedure were described by Xu et al. [22].

2.3. Quality Control

The fresh quartz filters were pre-heated at 450 °C in a muffle furnace for 6 h to remove any volatile components before sampling. Furthermore, after collection, the samples were sealed in clean plastic bags, and were stored in a freezer at −18 °C before chemical analysis to minimize the evaporation of volatile components. Before and after sampling, the Teflon filters in the first channel were weighed after being equilibrated for 24 h. The artifacts during the sampling and analysis were estimated by a field blank filter.

2.4. Measurements of Aerosol Optical and Meteorological Parameters

BC mass concentration, σ_sc and meteorological data were measured on the roof of the Physics Building about 30 m above the ground in Peking University, which is about 1 km away from the sampling sites of Tsinghua University. An automatic weather station (Vaisala Ltd., Helsinki, Finland) was used to record wind speed (WS), wind direction, relative humidity (RH), temperature (Temp), and visibility (VR).

σ_sc was monitored using a single wavelength (525 nm) integrating Nephelometer (M9003, Ecotech, Melbourne, VIC, Australia). This instrument drew ambient air through a heated inlet tube to maintain RH in the Nephelometer chamber below 60%. The scattering intensity over angles from 7° to 170° was measured and integrated to yield the scattering coefficient. Zero calibration was performed every two days with particle-free air to subtract the Rayleigh scattering, while span calibration was carried out every month using R-134 gas.

BC mass concentration was measured with an Aethalometer (AE-16, Magee Scientific, Berkeley, CA, USA). The principle of this instrument to calculate the BC concentration is based on the attenuation of an incident beam at a wavelength of 800 nm caused by the particles loaded in the quartz filter. No size-selective inlet was used for both the nephelometer and aethalometer. Considering the negligible contribution of coarse particles to light extinction, the measured σ_sc can be approximately attributed to the PM_2.5.

2.5. Data Analysis

2.5.1. Reconstruction of PM_2.5 Mass

PM_2.5 components can be grouped as follows: secondary inorganic aerosols (SNA), OM, EC, fine soil (FS), and chloride salt (CS). SNA is the sum of SO₄²⁻, NO₃⁻, and NH₄⁺, and OM is derived from multiplying OC concentrations by a factor of 1.6 to account for unmeasured atoms according to Xing et al. [23], which demonstrated that the calculated OM/OC mass ratio in summer was relatively high (1.75 ± 0.13) and in winter was lower (1.59 ± 0.18) in PM_2.5 collected from 14 Chinese cities. Beijing is far from the coastal oceans, and sea salt is not transported to Beijing, thus it has a minor contribution to PM_2.5 in Beijing. The CS was considered instead of sea salt, and estimated by summing concentrations of Cl⁻, K⁺, and Na⁺ according to Zhang et al. [12].

The concentrations of FS are often estimated by assuming the oxides of the elements mainly associated with soil (Al₂O₃, SiO₂, K₂O, CaO, FeO, Fe₂O₃, and TiO₂), which is calculated as follows [24]:

[FS] = 2.20[Al] + 2.49[Si] + 1.63[Ca] + 2.42[Fe] + 1.94[Ti]

(1)

2.5.2. Reconstruction of the Light Extinction Coefficient

According to the revised IMPROVE algorithm, the reconstructed b_ext is shown from the following equation assuming an externally mixed aerosol [20]:

$$\begin{array}{l}{b}_{ext}=b_{sp}+b_{ap}+b_{ag}+b_{sg}\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\approx 2.2\times f_s(RH)\times [Small{(N{H}_4)}_{2}S{O}_4]+4.8\times f_L(RH)\times [L\text{arg}e{(N{H}_4)}_{2}S{O}_4]\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}+2.4\times f_s(RH)\times [SmallN{H}_{4}N{O}_3]+5.1\times f_L(RH)\times [L\text{arg}eN{H}_{4}N{O}_3]\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}+2.8\times [SmallOM]+6.1\times [L\text{arg}eOM]\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}+10\times [EC]+1\times [Fine\hspace{0.17em}Soil]+1.7\times f_{ss}(RH)\times [Sea\hspace{0.17em}Salt]+0.6\times [P{M}_{2.5~10}]\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}+0.33\times [N{O}_2(ppb)]+Rayleigh\hspace{0.17em}Scattering\end{array}$$

(2)

The algorithm divides the concentrations of SO₄²⁻, NO₃⁻, and OM into small and large-sized fractions. The size modes are described by log-normal mass size distributions with geometric mean diameter and geometric standard deviations. The fraction of a component in the large- or small-sized mode was estimated by an empirical approach [20]. The apportionment of the total concentrations of (NH₄)₂SO₄ into the concentrations of the small and large size fraction in PM_2.5 is accomplished using the following equations:

$$[L\text{arg}e{(N{H}_4)}_{2}S{O}_4]=\frac{[Total{(N{H}_4)}_{2}S{O}_4]}{20\mu g.m^{-3}}\times [Total{(N{H}_4)}_{2}S{O}_4],\hspace{0.17em}for[Total{(N{H}_4)}_{2}S{O}_4]<20\mu g.m^{-3}$$

(3)

$$[L\text{arg}e{(N{H}_4)}_{2}S{O}_4]=[Total{(N{H}_4)}_{2}S{O}_4],\hspace{0.17em}for[Total{(N{H}_4)}_{2}S{O}_4]>20\mu g.m^{-3}$$

(4)

$$[Small{(N{H}_4)}_{2}S{O}_4]=[Total{(N{H}_4)}_{2}S{O}_4]-[L\text{arg}e{(N{H}_4)}_{2}S{O}_4]$$

(5)

Similar equations are used to apportion total NH₄NO₃ and total OM concentrations into small and large size fractions. The water growth adjustment term f_s(RH), f_L(RH) for small and large size distribution of (NH₄)₂SO₄ and NH₄NO₃, and f_ss(RH) for sea salt are used according to the water growth curves provided by Pitchford et al. [20].

According to the revised IMPROVE method, SO₄²⁻ and NO₃⁻ are assumed to be fully neutralized by NH₄⁺ in the forms of (NH₄)₂SO₄ and NH₄NO₃, respectively. Therefore, (NH₄)₂SO₄ mass is estimated by the SO₄²⁻ mass multiplied by a factor of 1.38, and the NH₄NO₃ mass is estimated by the NO₃⁻ mass multiplied by a factor of 1.29.

In order to compare with the reconstructed b_ext calculated using Mie theory at 550 nm, σ_sc measured at 525 nm with the integrating nephelometer should be converted to that at 550 nm according to the method by Jung et al. [16].

$${\sigma }_{sc}(550nm)={\sigma }_{sc}(525nm)*{\left(\frac{\lambda (550nm)}{\lambda (525nm)}\right)}^{-\alpha }$$

(6)

where α is the scattering Angström exponent, an average α value of 1.18 determined in Beijing during the summer of 2012 by Tian et al. [18] was used in the present study.

σ_ab at 550 nm was calculated based on BC concentration following the equation:

σ_ab = K × [BC]

(7)

where K is the conversion factor, which was set to 8.1 m²∙g⁻¹ in this study according to a previous study [25].

The single scattering albedo (SSA) is defined as the ratio of the aerosol scattering coefficient to the extinction coefficient at a known wavelength, as derived from the formula:

$$SSA=\frac{{\sigma }_{sc}}{{\sigma }_{sc}+{\sigma }_{ab}}$$

(8)

3. Results and Discussion

3.1. PM_2.5 Chemical Compositions

The time series of daily PM_2.5 mass concentrations and the meteorological parameters, including RH, temperature, and WS are shown in Figure 1. The mass concentrations of PM_2.5 ranged from15.6–237.5 μg∙m⁻³ and averaged 111.2 ± 73.4 μg∙m⁻³. Compared with other studies conducted in urban Beijing, the average PM_2.5 concentration in this study was lower than that measured during autumn of 2006 (194.2 μg∙m⁻³) [26] and 2009 (135 μg∙m⁻³) [12], while comparable with that observed during the same season in 2005 (115.0 μg∙m⁻³) and 2012 (106.9 μg∙m⁻³) [27]. There were sixty percent of days with daily PM_2.5 mass concentration exceeding the China Ambient Air Quality Standards(75 μg∙m⁻³). The highest PM_2.5 mass concentration occurred on 26 November, which was associated with the high relative humidity and low wind speed.

Figure 1. Daily variations of PM_2.5 mass concentration and meteorological parameters.

Figure 1. Daily variations of PM_2.5 mass concentration and meteorological parameters.

The temporal variations of nine water-soluble inorganic ions (WSIIs) are presented in Figure 2. The average concentration of total nine ions was 46.9 ± 33.8 μg∙m⁻³, accounting for 41.5% of PM_2.5 mass concentration. NO₃⁻ was the most abundant species in WSIIs with an average concentration of 14.7 ± 11.2 μg∙m⁻³, followed by SO₄²⁻ (12.2 ± 9.63 μg∙m⁻³), NH₄⁺ (9.13 ± 7.26 μg∙m⁻³), and Cl⁻ (6.62 ± 4.62 μg∙m⁻³), accounting for 28.9%, 25.6%, 17.7%, and 14.6% of WSIIs, respectively. The rest of K⁺ (1.66 ± 1.41 μg∙m⁻³), Na⁺ (0.95 ± 0.52 μg∙m⁻³), Ca²⁺ (0.81 ± 0.32 μg∙m⁻³),F⁻ (0.61 ± 0.34 μg∙m⁻³), and Mg²⁺ (0.25 ± 0.18 μg∙m⁻³) each had a minor contribution to the WSIIs, totally accounting for 13.2% of WSIIs. SNA typically constituted 33.5%–87.1% of the total WSIIs and 15.3%–46.0% of PM_2.5, respectively.

Figure 2. Daily variations of water-soluble ions and [NO₃⁻]/[SO₄²⁻] ratios.

Figure 2. Daily variations of water-soluble ions and [NO₃⁻]/[SO₄²⁻] ratios.

NO₃⁻ and SO₄²⁻ are mainly formed by atmospheric reactions of precursor gases such as NO_x and SO₂. Generally, SO₂ emits from coal combustion, while NO_x is the result of any type of combustion such as coal-fired power plants and automobiles. The mass ratio [NO₃⁻]/[SO₄²⁻] has been used to identify the influence of the stationary and mobile sources of sulfur and nitrogen [28]. The [NO₃⁻]/[SO₄²⁻] ratio ranged from 0.41–2.42, with an average value of 1.19. It was higher than that of values (around 0.68) measured in Beijing from 2001–2006 [26,29,30,31], but rather more comparable to those observed in recent years [12]. As illustrated in Figure 2, the ratio was usually lower during weekend days (12, 18–19 November) than on workdays, indicating that the higher [NO₃⁻]/[SO₄²⁻] ratio in the present study was probably associated with the rapid increase of motor vehicles in recent years. According to the statistics from the China Vehicle Emission Control Annual Report in 2013, the amount of vehicles reached 5 million in Beijing by 2012, which was an increase of about three times compared with the amount of vehicles in 2001 [32].

Previous study showed that Cl⁻ might be derived from coal combustion when the Cl⁻/Na⁺ equivalent concentration ratios were larger than the mean ratio (1.17) for sea water. The ratios of Cl⁻/Na⁺ were in the range of 1.6–11.6 with a mean value of 6.63 during the study period, implying that Cl⁻ may be originated from coal combustion rather than sea spray [12].

The equivalent molar ratio of total cations to total anions (CE/AE) ranged from 0.71–1.40, with an average value of 0.95 ± 0.14 during the study period. Figure 3 illustrates the scatter plots of the sum of cations versus anions. Results showed that the slope was slightly lower than 1, implying that the fine particles collected in the study period were weakly acidic. Moreover, the ratios of [NH₄⁺]/[SO₄²⁻ + NO₃⁻] were close to 1, demonstrating that SO₄²⁻ and NO₃⁻ were fully neutralized by NH₃. Therefore, the dominant chemical form of SO₄²⁻ was (NH₄)₂SO₄ rather than NH₄HSO₄, which can be estimated by the SO₄²⁻ mass concentration multiplied by a factor of 1.38, while NO₃⁻ existed as NH₄NO₃, and can be estimated by the NO₃⁻ mass concentration multiplied by a factor of 1.29.

Figure 3. Relationships of equivalent concentrations of cations versus anions (a) and [NH₄⁺] versus [SO₄²⁻ + NO₃⁻] (b).

Figure 3. Relationships of equivalent concentrations of cations versus anions (a) and [NH₄⁺] versus [SO₄²⁻ + NO₃⁻] (b).

As illustrated in Figure 4, OC varied from 2.1 to 64.3 μg∙m⁻³, averaging 27.5 ± 19.9 μg∙m⁻³, while EC ranged from 0.86 to 19.6 μg∙m⁻³, averaging 9.62 ± 6.24 μg∙m⁻³. The contribution of OC and EC to PM_2.5 were 24.5% and 8.96%, respectively. Such levels of OC and EC were close to those observed in the same season in recent years [10,26,30,33], whereas lower than those measured ten years ago [29,34].

Figure 4. Variations of OC, EC, and OC/EC (a) as well as primary organic carbon (POC), secondary organic carbon (SOC), and SOC/OC (b).

Figure 4. Variations of OC, EC, and OC/EC (a) as well as primary organic carbon (POC), secondary organic carbon (SOC), and SOC/OC (b).

The relationships between OC and EC can be used to identify the origins of carbonaceous particles [35,36]. As shown in Figure 5, strong correlations between OC and EC were observed with a correlation coefficient of 0.97, indicating that OC and EC were likely derived from the same major primary sources during the campaigns. On the other hand, the OC/EC ratios did not vary distinctly during the study period, especially during the space heating days. The ratios ranged from 1.96–3.52, averaging 2.79, and were very close to the value of 2.7 from coal combustion suggested by Watson et al. [37]. This pointed to the fact that OC and EC likely originated mainly from coal combustions. Furthermore, the mean OC/EC ratio was higher than 2, indicating that SOC might be present during the study period [38].

Figure 5. Relationships of OC and EC concentrations.

Figure 5. Relationships of OC and EC concentrations.

The method of EC-tracer has been widely used to estimate the SOC concentration since it was first introduced by Castro et al. [9,39]. This approach suggested that samples having the lowest OC/EC ratio contained almost exclusively POC. Then, the concentration of SOC can be estimated by the following formula:

POC = EC × (OC/EC)_min

(9)

SOC = OC – POC

(10)

where (OC/EC)_min was the value of the lowest OC/EC ratio. Based on the (OC/EC)_min of 1.63, the SOC concentrations varied from 0.02–25.9 μg∙m⁻³ with an average value of 9.03 μg∙m⁻³. As illustrated in Figure 4, it is interesting to note that the concentrations of SOC were still high with low temperature during the study period except on 10, 12 and 22 November. This may be caused by the combination of the high precursor emission due to the largely increased coal combustion for residential heating and low wind speed (averaging 0.79 m∙s⁻¹), which was favorable for the pollutants accumulation and formation of secondary organic aerosol. The low temperature was not favorable for the gas to particle conversion, whereas the frequent inversion conditions were likely favorable for the formation of SOC [40].

Daily variations of crustal elements are shown in Figure 6. Five crustal elements have a similar variation as the PM_2.5 mass concentrations. Their concentrations varied significantly from day to day. The average concentration for Al, Si, Ca, Fe, and Ti was 0.51 ± 0.26, 1.01 ± 0.58, 0.74 ± 0.36, 1.10 ± 0.64, and 0.06 ± 0.03 μg∙m⁻³, respectively. Increasing wind speed could be expected to increase the concentrations of crustal elements, but the concentration of the five elements had a weak correlation with the wind speed (R < 0.4) in the present study. However, when the wind speed exceeded 1.5 m∙s⁻¹ (11 and 18 November), the concentrations of crustal elements were higher than those on any other day. The FS mass concentration was estimated by summing the above five crustal elements plus oxygen for the normal oxides as Equation (1). The average mass concentrations of FS were 7.66 ± 3.62 μg∙m⁻³, ranging from 1.99–15.1 μg∙m⁻³, and accounting for 9.42% of PM_2.5.

Figure 6. Time series of crustal elements concentration.

Figure 6. Time series of crustal elements concentration.

3.2. PM_2.5 Mass Balance

The reconstructed PM_2.5 mass concentrations were close to the measured ones with strong correlation (Figure 7), indicating that the reconstruction of PM_2.5 could be reasonable. Nevertheless, a few biases were observed in the reconstructed PM_2.5 mass. Water absorption of the water-soluble components may lead to positive biases and overestimate the PM_2.5 mass, while the volatilization of NH₄NO₃ and volatile organic matter may result in negative biases. Moreover, the conversion used to estimate OM from OC also caused an uncertainty in calculating the PM_2.5 mass.

Figure 7. Scatter plots of measured and reconstructed PM_2.5 mass concentrations.

Figure 7. Scatter plots of measured and reconstructed PM_2.5 mass concentrations.

Figure 8 presents the reconstructed chemical compositions in PM_2.5. On average, the fractions of major chemical compositions followed the order of OM > NH₄NO₃ > (NH₄)₂SO₄ > FS, CS, and EC. OM was the most abundant component in PM_2.5 (averaging 45.8 ± 31.7 μg∙m⁻³), accounting for 39.4% of PM_2.5. The contribution of FS (averaging 7.66 ± 3.62 μg∙m⁻³), CS (averaging 9.47 ± 6.24 μg∙m⁻³), and EC (averaging 9.97 ± 6.28 μg∙m⁻³) to PM_2.5 was similar, each approximated to 9%. The percentage of SNA (30.2%) was much higher than the three species of FS, CS, and EC, but slightly lower than that of OM. The percentages of (NH₄)₂SO₄ and NH₄NO₃ (averaging 16.7 ± 13.2 and 19.0 ± 14.4 μg∙m⁻³, respectively) were 14.9% and 15.4%, respectively. Compared with the results determined over the same period in earlier years [11], it is noted that the percentage of SNA in our study decreased by 3%–10% compared with that measured during 2003–2007, while the OM fraction rose about 5%.

Figure 8. Daily variations of the reconstructed chemical composition in PM_2.5.

Figure 8. Daily variations of the reconstructed chemical composition in PM_2.5.

3.3. Analysis of Aerosol Optical Properties

The time series of daily averaged optical properties including σ_sc, σ_ap, and SSA are shown in Figure 9. Daily σ_sc ranged from 31.1–667 Mm⁻¹, with a mean value of 270 ± 200 Mm⁻¹, while σ_ap was in the range of 8.24–158.0 Mm⁻¹, with a mean value of 74.3 ± 43.4 Mm⁻¹. The mean σ_sc value was considerable lower than that measured in urban Beijing in 2009 and during 2005–2006 [41,42], but higher than that obtained at a suburban site (Changping) and rural site (Shangdianzi) [15]. Compared with the results mentioned above, the mean σ_ap value was lower than that measured in 2009 as well, but higher than that during 2005–2006. The increased σ_ap in recent years is likely attributable to the rapid increase of vehicle pollution, since vehicular exhaust was one of the primary factors affecting aerosol absorption. The mean value of SSA was 0.76, which was comparable with the results determined in Beijing during 2005–2006 [42] and in 2009 [41].

MSE is an important parameter for estimating radiative forcing of aerosols and chemical extinction budgets for visibility impairment. Generally, there are two methods to estimate MSE, i.e., measurement method and multilinear regression method [43]. MSE was defined as the ratio of measured σ_sc to aerosol mass concentration according to the measurement method. One alternative method can also be used by regression of the measured σ_sc against aerosol mass concentration. Since the RH in the nephelometer was maintained below 60% and the PM_2.5 mass concentrations were measured at RH of 40%, those days with RH below 50% were selected for regressing to minimize the impact of particle hygroscopic growth on σ_sc. According to the measurement method, daily MSE varied from 1.70–3.02 m²∙g⁻¹, with a mean value of 2.32 ± 0.44 m²∙g⁻¹. It is noted that a strong correlation between the measured σ_sc and PM_2.5 mass concentration was observed with a high correlation coefficient of 0.988 (Figure 9). The slope was 2.67 from a linear regression, which was slightly higher than the value obtained by the measurement method, also found by Titos et al. [43]. Compared to that measured in urban Beijing by Zhao et al., and Jing et al. [41], the relative lower MSE in our study was likely related to the heavily polluted events.

Figure 9. The time series of daily averaged aerosol optical properties (a) and relationship between the measured σ_sc and PM_2.5 concentration (b).

Figure 9. The time series of daily averaged aerosol optical properties (a) and relationship between the measured σ_sc and PM_2.5 concentration (b).

3.4. Chemical Apportionment of the Aerosol Optical Parameters

In order to appoint the contribution to the visibility impairment, b_ext was reconstructed based on the chemical compositions of aerosol. In the present study, the extinction effect by fine particles was studied, while the contributions of gases were excluded because they only accounted for a small fraction (about 2%–4%) of b_ext [44]. The impact of sea salt on b_ext was ignored since Beijing is about 150 km away from the East China’s coastal oceans. Moreover, the contribution of coarse mass to b_ext was not included because of lack of the concentrations of coarse matter. Then, the revised IMPROVE formula of Equation (2) was modified as follows:

$$\begin{array}{l}{b}_{ext}=2.2\times f_s(RH)\times [Small{(N{H}_4)}_{2}S{O}_4]+4.8\times f_L(RH)\times [L\text{arg}e{(N{H}_4)}_{2}S{O}_4]\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}+2.4\times f_s(RH)\times [SmallN{H}_{4}N{O}_3]+5.1\times f_L(RH)\times [L\text{arg}eN{H}_{4}N{O}_3]\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}+2.8\times [SmallOM]+6.1\times [L\text{arg}eOM]\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}+10\times [EC]+1\times [Fine\hspace{0.17em}Soil]\end{array}$$

(11)

The measured and reconstructed b_ext are illustrated in Figure 10. It is found that the measured b_ext were considerably lower than the reconstructed value, especially during the heavily pollution levels. The deviation varied from 18.1%–140%, with an average value of about 70%. Jung et al. [45] also found that the b_ext was overestimated by 36.7% based on the revised IMPROVE algorithm. However, a few other studies using the IMPROVE formula found that there existed a good correlation between the measured and reconstructed b_ext, and the slopes were close to 1.0 in Shanghai [19] and Guangzhou [4]. Compared to the results conducted in Shanghai, a lower MAE (7.7) was used to calculate b_ap. If the same value of 7.7 was used in the present study, the biases would decrease by 8%. Additionally, the MSE used to calculate b_ext in Guangzhou was much higher than the value in the present study. Thus, it can be deduced that a lower MSE than the value adopted according to the revised IMPROVE algorithm in the present study should be used to reconstruct b_ext, which may result in a reconstructed b_ext approximately equal to the measured b_ext. Although more locally-derived MSE and MAE were necessary for effectively reconstructing the b_ext, we did not obtain these values in the present study due to lack of the amount of in situ and sampling data.

Figure 10. The temporal variations of measured and reconstructed b_ext (a) and the correlation between measured and reconstructed optical parameters (b).

Figure 10. The temporal variations of measured and reconstructed b_ext (a) and the correlation between measured and reconstructed optical parameters (b).

Although the reconstructed b_ext was higher than that measured one, both were correlated well (R = 0.989), so did b_sp and b_ap (Figure 11). Therefore, the relative contribution of each chemical composition to b_ext can also be analyzed by the modified IMPROVE algorithm. As shown in Figure 11, OM, (NH₄)₂SO₄, NH₄NO₃ and EC were the dominant contributors, accounting for 97.7% of b_ext together, while the contribution of FS was small, accounting for only 2.3% of b_ext. On average, OM was the largest contributor to the b_ext, accounting for 44.7% of b_ext, while SNA accounted for 34.4% of b_ext. (NH₄)₂SO₄ and NH₄NO₃ contributed 16.4% and 18.0% of b_ext, respectively. Our results were different from those determined in Beijing in previous studies conducted in summer [16,18,46], which showed that (NH₄)₂SO₄ and NH₄NO₃ were the largest contributor to the b_ext.

Figure 11. Relative contributions of each chemical composition in PM_2.5 to b_ext.

Figure 11. Relative contributions of each chemical composition in PM_2.5 to b_ext.

3.5. Typical Pollution Episodes

As shown in Figure 1, four obvious pollution episodes were observed during the campaign, with the visibility deteriorating to less than 10 km. They were observed on 9–11 November, 14–17 November, 20 and 21 November, and 23–27 November. Obviously, in the first pollution period, pollutants accumulated gradually from 8–10 November, with daily PM_2.5 mass concentration increasing from 36.3–98.8 μg∙m⁻³, and then decreasing dramatically to 17.1 μg∙m⁻³. All the chemical components increased with PM_2.5 mass, especially for NO₃⁻, which increased by six times compared with the value on 8 November. Furthermore, it can be found that during the pollution accumulation period, the wind speed was less than 0.5 m∙s⁻¹. On 11 November, strong wind was favorable for dispersion of the pollutants and accumulation of the crustal material. Moreover, 12 November was weekend, and the reduction of vehicles may also contribute to the lower concentration of PM_2.5. From an extinction perspective, in the first pollution stage, σ_sc increased from 71.6–177.2 Mm⁻¹, while σ_ap varied little, implying that the visibility degradation was mainly caused by the rapid increase of SNA.

In the second pollution period from 12–19 November, daily PM_2.5 mass concentration increased from 17.1–215.5 μg∙m⁻³, with the maximum value occurring on 15 November, and then plunging to 32.0 μg∙m⁻³ within three days. As illustrated in Figure 1, it can be found that the wind speeds were very low from 12–15 November, and in favor of the accumulation of the pollutants. On the other hand, the RH increased from 30% to 82%, which was favorable for the formation of SNA. Residential heating starting in the middle of November might be the primary reason for the heavier pollution in the second stage. As shown in Figure 2, the concentration of Cl⁻ on 13 November had a dramatic increase compared with the previous days, approximately to 10 times higher than that on 12 November. Chloride may be essentially contributed by coal combustion in Beijing during the heating season. Thus, the high PM_2.5 mass concentration was mainly associated with coal combustion. In fact, the concentrations of OC and EC during 13–27 November were much higher than those before 12 November. The sharp increase of OC and EC also verified the influence of coal combustion on the increase of PM_2.5 mass concentration. Unlike the first pollution stage, σ_ap on the most heavy pollution day (15 November) was about 15 times higher than that on the clean days. Meanwhile, σ_sc rose from 31.0–439.2 Mm⁻¹, with a similar growth rate as σ_ap. As shown in Figure 11, during the second pollution episode, the contribution of OM and EC to b_ext decreased from 54.2% and 25.6% to 32.7% and 11.5% as the pollutants accumulated, respectively, while the contribution of SNA to b_ext increased from 18.3%–54.8% on the accumulation period. On 16 November, although the mass concentration of PM_2.5 reduced by 40%, the visibility was still less than 3 km, which was ascribed to the largest contribution of SNA to b_ext. As presented in Figure 11, the contribution of SNA reached up to 73.7% whereas that of OM and EC decreased to 26.7% to b_ext. The PM_2.5 mass concentration had a significant decrease on 18 November due to rain and strong wind (Figure 1).

Based on the analysis of a typical pollution episode, it can be concluded that the secondary formation of aerosol was more important than the carbonaceous pollution for the haze formation in Beijing. In the other two pollution periods, a similar trend of the chemical composition to that during the second pollution stage was observed. In general, the pollution accumulation was in accordance with the increase of the SNA, OC and EC concentrations under stable weather conditions until arrival of strong wind.

4. Conclusions

During the heating period from 8–28 November 2011, aerosol optical properties as well as chemical compositions were investigated simultaneously in Beijing. Daily PM_2.5 mass concentration varied from 15.6–237.5 μg∙m⁻³ and presented a mean value of 111.2 ± 73.4 μg∙m⁻³. Among the chemical components in PM_2.5, NO₃⁻ was the most abundant species in WSIIs with an average concentration of 14.7 ± 11.2 μg∙m⁻³, followed by SO₄²⁻, NH₄⁺, and Cl⁻, accounting for 28.9%, 25.6%, 17.7%, and 14.6% of WSIIs, respectively. The rest of K⁺, Na⁺, Ca²⁺, F⁻, and Mg²⁺ have a minor contribution to the WSIIs, accounting for 13.2% of WSIIs together.

The mean σ_sc, σ_ap and SSA values at 550 nm were 270 ± 200 Mm⁻¹, 74.3 ± 43.4 Mm⁻¹ and 0.76 during the entire observation period, respectively. Both of the σ_sc and σ_ap increased significantly during the pollution accumulation episode. The b_ext were estimated by the revised IMPROVE formula based on the chemical compositions of PM_2.5. Compared with the measured σ_sc and σ_ap, the reconstructed b_ext was overestimated, but had a strong correlation with a high correlation coefficient of 0.989. OM was the largest contributor, accounting for 44.7% of b_ext, followed by NH₄NO₃, (NH₄)₂SO₄, with minor contribution from soil dust (2.3%).

Pollution episodes in Beijing were strongly influenced by both emissions and meteorological conditions. Pollutant was accumulated in calm or weak winds while diffused under strong wind conditions. Additionally, the coal combustion for residential heating was another major reason for the heavy pollution during the sampling period. Four typical pollution episodes during the study period were observed, it was found that NH₄NO₃ and (NH₄)₂SO₄ were the largest contributor to the b_ext rather than carbonaceous components during the pollution accumulation episodes, implying that the secondary inorganic pollutants were more important than the carbonaceous pollution for heavy pollution formation. Therefore, the reduction of their precursors such as SO₂, NO_x and NH₃ could effectively improve the visibility in Beijing.