Measurements of particle number size distributions and optical properties in urban Shanghai during 2010 World Expo: relation to air mass history

2.1. Location of sampling site

The intensive field campaign was performed between 15 April and 22 June, 2010. The sampling site was located in the eastern part of Shanghai, and approximately 8 km far to the northeast of World Expo Park. All the instruments were housed in the observatory, which was situated on the roof of a sixth floor building of Shanghai Pudong Environmental Monitoring Station (31.2267°N, 121.5382°E). The sampling inlets extended vertically from the laboratory window, which were ~25 m in height above the ground level. The surroundings of this site were mainly a residential and business area. Two major roads with heavy traffic at the west and south of the site were respectively 500 m away. There are no other significant local source emissions around. Hence, this site is assumed as representative for the Shanghai urban background aerosol.

2.2. Instrumentation

Number size distributions of atmospheric particles from 15 to 600 nm (mobility diameter) were measured by a SMPS (Scanning mobility particle sizer) system (TSI model 3936, TSI Inc., St. Paul, MN, USA). It consists of a long Differential Mobility Analyzer (DMA, TSI model 3081) transmitting only particles with a certain size, and a Condensational Particle Counter (CPC, TSI model 3022) for counting the particle number concentration. One inlet impactor with 0.071 cm diameter nozzle was installed ahead the whole system to remove big particles. The flow rate was 0.3 L/min for the sample air and 3.0 L/min for the sheath air. Particle number size distribution was taken every 5 minutes. The size-dependent diffusion losses in connecting tubes were corrected as compiled by Wiedensohler et al. (2012).

A Multi-angle Absorption Photometer (MAAP, model Carusso 5012, Thermo, Inc., Waltham, MA USA) was used to obtain the aerosol light absorption coefficient at 670 nm (b _ap,_670nm). The b _ap,_670nm is determined from the simultaneous measurement of radiation passing through and scattered back from a particle-loaded filter. The measurement is operated at three detection angles to resolve the influence of light-scattering aerosol components on the angular distribution of the back scattered radiation (Petzold and Schönlinner, 2004). The flow rate of the MAAP was calibrated once per week. The uncertainty of MAAP at 99% confidence level has been estimated as 12% by Petzold and Schönlinner (2004), and the minimum detection limit is lower than 100 ng/m³ at 2 minutes average.

Aerosol total light coefficient at a wavelength of 532 nm (b _sp,532nm) was measured by a commercial integrating nephelometer (M9003, Ecotech Pty Ltd. Knoxfield, Australia) (Heintzenberg and Charlson, 1996; Müller et al., 2011). Before the field campaign, calibration was carried out with filtered particle-free air and CO₂ as the low and high span gases, respectively. According to the manual, the lowest detectable limit is 0.4 Mm⁻¹.

Measurement of the SO₂ concentrations was carried out by using a commercial analyser (EC9850, Ecotech Pty Ltd. Knoxfield, Australia), and an automatic meteorology station was operated to measure meteorology parameters including temperature, relative humidity and pressure.

During the whole measurements, ambient aerosol particles were sample via a 3/4-inch stainless steel tubing with a PM₁₀ sampling head, and then split to SMPS and MAAP. Total suspended particles were collected by the nephelometer via a 1/4-inch stainless steel tubing. The relative humidity for each device was kept within 40% by using the silica gel diffusion dryer. The frequency for changing the dryer was about 1 week.

2.3. Data analysis

2.3.2. Mode fitting

The measured particle number size distributions were fitted and parameterised by a multiple log-normal distribution function (Seinfeld and Pandis, 2006):

4

$\frac{dN}{d\text{log}Dp}=\frac{N_i}{\sqrt{2\pi log\sigma }}exp\left[-\frac{{(logDp-log\overline{D{p}_i})}^2}{2{(log\sigma )}^2}\right]$

Here, N _i and $\overline{D{p}_i}$ are the total number concentration and mean diameter of mode i, respectively. σ is the geometric mean standard deviation of the distribution. In this study, we applied an automatic algorithm for fitting every measured particle number size distribution (N: 19872) using MATLAB. The algorithm does not need a user decision for the initial input parameters, while it requires only the maximum number of possible modes. Detailed information could be found in Hussein et al. (2005). Figure 1 displayed one example of fitting result for mean particle number size distribution during the entire measurement period.

Fig. 1

As a result, three modes were identified corresponding to the nucleation mode, Aitken mode and accumulation mode, respectively. The statistical analysis showed that the corresponding mean diameters were 21 nm, 56 nm and 186 nm, respectively, with the standard deviation of 6 nm, 21 nm and 69 nm. Hence, the diameter size ranges for the nucleation, Aitken and accumulation modes were chosen as 16 to 30 nm, 30 to 100 nm and 100 to 600 nm in this study.

3.1. Particle number concentrations and size distributions

The temporal variability of hourly average number and volume concentrations of total particles (16–600 nm) during the entire measurement period are represented in Fig. 2a. The total aerosol number concentrations were observed from 3000 to 43 300 cm⁻³, with a mean value of 12 700±6200 cm⁻³. The particle volume concentrations were calculated from the observed number size distributions with an assumption of spherical particles. The values varied largely in the range of 3–60 µm³/cm³ during the campaign, with an average of 16±8 µm³/cm³. With the assumption of particle density as 1.7 g/cm³, this value was consistent with the PM₁ (29.2 µg/m³), which was achieved both by High–Resolution Time-of-Flight Aerosol Mass Spectrometer (HR-ToF-AMS) and Single Particle Soot Photometer (SP2) during the same campaign (Huang et al., 2012). The average number concentrations were 3100 cm⁻³ for nucleation mode, 6700 cm⁻³ for Aitken mode and 2900 cm⁻³ for accumulation mode. The corresponding median values were 2600 cm⁻³, 5800 cm⁻³ and 2400 cm⁻³, respectively (Fig. 2b).

Fig. 2

The total particle number concentration was lower than those measured in the other urban environments in North America and Europe, such as Pittsburgh (22 000 cm⁻³; Stanier et al., 2004), New York (22 100 cm⁻³; Bae et al., 2010),Leipzig (11 600–21 400 cm⁻³; Wehner and Wiedensohler, 2003), Atlanta (23 100 cm⁻³; Woo et al., 2001), New Delhi (20 000–250 000 cm⁻³; Mönkkönen et al., 2005). However, higher accumulation mode particles were observed in urban background sites of Shanghai compared with the other sites except for New Delhi, representing the more polluted situation. The summary of size-segregated particle number concentrations measured in various atmospheric environments of China is listed in Table 1. The mean total number concentration at Shanghai was close to the recent study conducted in the winter of 2008 (Du et al., 2012) and was much lower than the observations in the urban sites of Beijing (32 800 cm⁻³; Wu et al., 2008), Guangzhou (29 000 cm⁻³; Yue et al., 2010a) as well as the suburban of Shanghai (30 200 cm⁻³; Gao et al., 2009) and Nanjing (23 300 cm⁻³; Herrmann et al., 2014), even the rural sites like Yufa (17 000 cm⁻³; Yue et al., 2009) and Xinken (16 300 cm⁻³; Liu et al., 2008). However, the results were 3–7 times higher than the marine environment such as Yellow Sea (ship measurement, 4000 cm⁻³; Lin et al., 2007), as well as the regional background stations Waliguan (high–altitude mountain, 3816 m a.s.l., 2000 cm⁻³; Kivekas et al., 2009). Note that the differences may be due to the instrumentation techniques (the detections of lower and upper diameters), which may bring the diverse integrated size range. In addition, the different sampling locations and seasons would also cause the variability. In principle, the particle number concentrations in urban areas were twice than those in rural sites and approximately one order of magnitude higher than those in the marine/background environments. On average, the total particle count was dominated by ultrafine particles (Dp<100 nm, 77%), and in which Aitken mode particles account for 53%. This was comparable to the results found in Beijing regions (48–53%) (Wu et al., 2008; Yue et al., 2009).

The diurnal variations of aerosol number concentrations in each mode are shown in Fig. 2c. The nucleation mode particles peaked around noon local time, which was similar to the gaseous sulphuric acid (Wang et al., 2013a). This is reasonable because it has been identified as a key precursor in the particle nucleation process (Yue et al., 2010b; Wang et al., 2011; Zhang et al., 2012). The Aitken mode particles showed a three-peak diurnal pattern, which was obvious in the midday (12:00) and vague both in the morning (7:00) and late afternoon time (18:00). The midday peak might be due to the growth of newly formed particles by condensation and coagulation processes (Wu et al., 2008), while the morning and late afternoon peaks are more related to the local traffic emissions. The accumulation mode particles did not show the significant change with time.

3.2. Optical properties

The time evolutions of particle light scattering and absorption coefficients during the campaign period are illustrated in Fig. 3a. The values (mean±standard deviation) were 210±140 Mm⁻¹ and 26±20 Mm⁻¹ at Shanghai, respectively. b _sp,532nm values were approximately a factor of 2 higher than the other urban site, including Atlanta (summer, 121 Mm⁻¹, 530 nm; Carrico et al., 2003), Mexico City (spring, 105 Mm⁻¹, 550 nm; Marley et al., 2009) as well as New Delhi (summer, 110 Mm⁻¹, 550 nm; Soni et al., 2010). b _ap,532nm values, however, were typically much lower than Mexico City (37 Mm⁻¹, 550 nm; Marley et al., 2009) and New Delhi (62 Mm⁻¹, 550 nm; Soni et al., 2010), indicating less local emissions at Shanghai urban. Compared to the megacities in China, it is interesting to find that both higher b _sp and b _ap were shown in Beijing (summer, 300±270 Mm⁻¹ and 55±32 Mm⁻¹, 550 nm; Liu et al., 2009) and Guangzhou (autumn, 418±159 Mm⁻¹ and 91±60 Mm⁻¹, 540 nm; Andreae et al., 2008), even the surrounding regions, such as Wuqing (summer, 379±251 Mm⁻¹ and 43±27 Mm⁻¹, 550 nm; Ma et al., 2011) and Xinken (autumn, 333±138 Mm⁻¹ and 70±42 Mm⁻¹, 550 nm; Cheng et al., 2008). This phenomenon suggested that the pollutions in Beijing and Pearl River Delta (PRD) regions might be determined both by local and regional transport emissions.

Fig. 3

The aerosol single scattering albedo ω _0,532nm values at Shanghai ranged from 0.69 to 0.97, with a mean value of 0.88. Previous studies (Reid et al., 1998; Carrico et al., 2003; Cappa et al., 2009) demonstrated that ω ₀ in mid-visible wavelength range varies from 0.79 (freshly biomass burning) to 0.99 (clean marine aerosol) depending on the air mass type, which could be generally characterised as the low value denotes the primary source, while the high value indicates the regional background aerosol. In this study, the values were higher than that at New Delhi (Soni et al., 2010) and Xinken (Cheng et al., 2008), which was significantly influenced by local sources, and comparable with Atlanta (Carrico et al., 2003) and Gosan (Kim et al., 2005).

Figure 3b exhibited b _sp,532nm as a function of particle number concentrations in different modes, including the particles in the size range of 300–500 nm (insert graph). In our results, the relations between b _sp,532nm and the number concentration of nucleation and Aitken mode particles both displayed an envelope type pattern, with high values of b _sp,532nm being always associated with low nucleation mode particle number concentration, and no clear correlation can be seen for the nucleation and Aitken modes. On the contrary, the high correlation (R=0.69) was observed between b _sp,532nm and the number concentration of accumulation mode particles. This result was consistent with the Mie theory. However, if we look at specific size range (300–500 nm), the correlation (R=0.87) was found strongly, indicating that the aerosol particles in this certain diameter range had the highest scattering efficiency at 532 nm.

Figure 3c presented the relationship between the ω ₀ and the aerosol load (represented by b _sp,532nm). It had been found that ω ₀ generally decreased with increasing particle absorption coefficient, especially in the same lever of aerosol load. This indicated the variation of ω ₀ was more sensitive to b _sp than b _ap at Shanghai. Meanwhile, the decreasing rate of ω ₀ was smaller when the aerosol load was higher. One possible explanation was that the increasing aerosol pollution was always associated with more aged air mass and more internally mixed BC, thus the mass fraction of light-scattering materials (e.g., sulphate and most of the OC) will get higher. This conclusion had been confirmed by previous study (Cheng et al., 2008).

3.3. Back trajectory cluster analysis

The air quality of the megacity, such as Beijing and Shanghai, is not only determined by the local source emission, but may also be influenced by the regional transport depending on the meteorological air mass history (Takegawa et al., 2009; Huang et al., 2012). Hence, to explore the impacts of the air mass history, the HYSPLIT4 model was used to follow the history of air masses arriving at Shanghai. As a result, four-cluster analysis was the optimum solution according to the change in total spatial variance, with the mean back trajectory of each cluster presented in Fig. 4. Each of the clusters represented between 37 (cluster N) and 108 (cluster W) back trajectories. The mean back trajectories showed significant differences in direction and length. Cluster N and W arrived from the northern and western sections of mainland China with different air mass speed, while cluster NE and E came from northeasterly and eastern marine areas, respectively. The air mass came from west was found to be the most frequent cluster, accounting for 38% on average of all back trajectories.

Fig. 4

Figure 5 illustrated the mean number, surface and volume size distributions for each cluster. The number size distributions of different air mass types mostly peaked at around 25–60 nm, whereas the maximum of surface and volume size distributions were shown around 200–300 nm and 300–400 nm, respectively. The mean total particle number concentrations for the air masses came from northern and western directions through the continent and were 16 200 cm⁻³ and 14 300 cm⁻³, respectively, which were higher than when the air masses came from the East China Sea (Table 2). In addition, higher particle volume concentrations were observed associated with the higher number concentration, with mean values of 16 µm³/cm³ and 21 µm³/cm³, respectively, for cluster N and cluster W. On the contrary, the air masses coming from northeastern and eastern marine area contained less pollutant, with mean particle volume concentrations of 14 µm³/cm³ and 11 µm³/cm³, respectively. The highest particle volume concentration was observed in cluster W, which might be explained by the higher accumulation mode particles (3700 cm⁻³), accounting for 26% of the total number concentration compared with that (~20%) in other clusters. In summary, this particle volume concentration loading distribution pattern indicated that the westerly continental air mass represented the most severe pollutant regional transport condition in Shanghai. This finding was consistent with the previous study (Huang et al., 2012). However, it should also be pointed out that the structure of local emission might influence the results.

Fig. 5

Figure 6 showed the mean diurnal evolutions of particle number size distributions, averaged for each of the trajectory clusters mentioned above. Significant differences were recognisable: the cluster N with fast moving air mass transport from north showed a clear pattern that the ultrafine mode particles formed in the morning (10:00) and growth into Aitken mode, indicating the new particle formation event. In contrast, the particle number size distribution of other three clusters showed different patterns, which was only observed the appearance of burst at noon. The condensation sink (CS) describes how rapidly vapour molecules could condense onto the pre-existing particles, which could be used representing the pre-existing particles concentration (Kulmala et al., 2001; Dal Maso et al., 2005). Its value could be calculated using eq. (5):

5

Fig. 6

where D is the diffusion coefficient of the condensing vapour, β is the transitional regime correction factor, Dp is the aerosol particle diameter and N is their number concentration. However, it should be noticed that in this study the CS values were calculated based on the dry particle number size distributions. The CS values did not show the significant differences between cluster N (0.022 s⁻¹) and cluster W (0.024 s⁻¹), which were much higher than the cases where air masses came from marine areas, 0.017 s⁻¹ and 0.012 s⁻¹ for cluster NE and E (Table 2), respectively. However, the mean SO₂ concentration was 11.4 ppb for cluster N, which was higher than the observations in cluster NE (6.2 ppb), cluster E (4.7 ppb) and cluster W (7.7 ppb). The higher SO₂ concentration for cluster N might cause the different evolution of particle number size distribution. As a result, the mean particle number concentration in nucleation mode was 4700 cm⁻³ in cluster N, which was 1.2, 1.7 and 1.3 times higher than those in cluster NE, E and W, respectively. In addition, the lower temperature and relative humidity were observed in cluster N, with the mean values of 19% and 46%, respectively. Such meteorological conditions caused by the fast moving air mass from northerly direction favours the occurrence of nucleation. All of the above showed that air mass had an important effect on NPF events with significant subsequent growth. This result was consistent with the observations in previous studies (Wehner et al., 2008; Shen et al., 2011).

In terms of optical parameters, the variations of particle light absorption and scattering coefficients for four clusters were associated with the number concentration of accumulation mode particles (Table 2), as analysed in Section 3.2. The polluted west back trajectory had the highest b _ap,532nm and b _sp,532nm among four clusters, with the mean and standard deviation values of b _ap,532nm and b _sp,532nm of 35±19 and 328±197 Mm⁻¹, respectively, indicating the slow air mass came from the continent area had a stronger extinction effect. However, the aerosol single scattering albedo ω ₀ did not show the same variation trend as b _ap,532nm and b _sp,532nm. The North China Plain, which is the main route of cluster N, is a major black carbon emission area (Wang et al., 2012). Hence, the lowest values (ω ₀=0.84) were observed when the fast air mass was moving from northern China (cluster N), compared with the other three clusters (ω ₀ ~ 0.88–0.89).