Aerosol Size Spectra and Particle Formation Events at Urban Shanghai in Eastern China

Aerosol number size distributions between 10 nm and 10 μm were measured from October 2008 to February 2009 in Shanghai, China. The average particle number, surface and volume concentrations were 1.3 × 10 1/cm, 6.4 × 10 μm/cm and 64 μm/cm, respectively. Aitken particles dominated the total number of particles, and accumulated particles were the greatest contributor to particle surface area. Particle number size distributions could be characterized by multi-lognormal functions. The average number size distributions of aerosols revealed a clear diurnal pattern of two peaks within 30–60 nm corresponding to the morning and afternoon traffic rush hours. All size particles had two peaks in mean number concentrations during the rush hours, and 10–20 nm and 20–50 nm particles had one additional peak in late morning. The new particle formation events were found on four days out of 73. The apparent formation rates varied from 0.2 to 0.5 cm/s, and the growth rates of newly formed particles were 3.3–5.5 nm/h. Overall, the new particle formation events had a significant impact on particle size spectra in the nucleation and Atiken modes, but insignificant effects on particle surface and volume concentrations.


INTRODUCTION
Aerosols have important effects on the atmosphere, climate system, and human health.Aerosol particles interact directly with the incoming solar radiation by scattering and absorption to light (Charlson et al., 1992), and affect indirectly the Earth's radiation budget by acting as cloud condensation nuclei (Rosenfeld, 1999).Indeed aerosol is one of the greatest sources of uncertainty in assessing global climate forcing (IPCC, 2007).
The radiative effects of tropospheric aerosols depend on the physical and chemical properties of particles, and their non-uniform spatial and temporal distributions (Seinfeld and Pandis, 1998).Particle size and size distribution are important aspects of aerosol physical properties, and particle number concentration should get more attention than mass concentration.Over the past decade, much research has been done on urban aerosol number size distributions, including long-and short-term measurements (Junker et al., 2000;Wählin et al., 2001;Wehner et al., 2002;Wehner and Wiedensohler, 2003;Stanier et al., 2004;Held et al., 2008;Weber, 2009;Lonati et al., 2011).Most of researches on aerosol size spectra were done in the developed countries, less studies were carried out in the newly industrialized countries (Dunn et al., 2004;Mönkkönen et al., 2005;Wehner et al., 2008;Wu et al., 2008;Zhang et al., 2001Zhang et al., , 2008)).
The nucleation and subsequent growth of new aerosol particles consists of complicated processes (Kulmala and Kerminen, 2008), and contributes a certain extent to the changes of fine aerosol loading, aerosol physical, chemical and optical properties as well as activity as cloud condensation nuclei (Kulmala, 2003).Kulmala et al. (2004a) and Holmes (2007) have reviewed a number of studies on new particle formation (NPF) and growth within various areas and discussed the implications of different mechanisms and related impact factors.Increasing evidence reveals that new particles are formed by nucleation of non-or low-volatile gas-phase compounds emitted from either biogenic or anthropogenic sources, followed by growth into small particles (Kulmala et al., 2000;Wehner et al., 2005).Sulfuric acid is known as a major nucleating component in the atmosphere to form the critical nucleus which then grows to a detectable size (Kulmala et al., 2004c;Kerminen et al., 2004;Zhang et al., 2004b;Zhang, 2010b).
The NPF is highly sensitive to several factors including preexisting aerosol surface areas, aerosol aging, ambient gasphase concentrations as well as weather conditions (Nilsson et al., 2001;Boy and Kulmala, 2002;Kerminen et al., 2004).Identification of NPF and growth processes within urban atmosphere is more difficult than within rural or remote atmosphere (Kulmala et al., 2004a;Holmes, 2007).Numerous studies have focused on and devoted considerable efforts to NPF and growth over worldwide urban areas (Harrison et al., 2000;McMurry et al., 2000;Shi et al., 2001;Woo et al., 2001;Alam et al., 2003;Dunn et al., 2004;Stanier et al., 2004;Zhang et al., 2004a;Shi et al., 2007;Hussein et al., 2008;Olofson et al., 2009).To date, only a few measurements on NPF events have been performed in the urban or suburban areas of China (Wehner et al., 2004;Wu et al., 2007;Liu et al., 2008;Gao et al., 2009;Yue et al., 2009;Yao et al., 2010;Shen et al., 2011;Zhang et al., 2011), which leads to an unclear outline of urban NPF and growth in this region.
In order to understand particulate matter in the atmosphere over urban areas in the Yangtze River Delta (YRD) of China, measurements concerning on aerosols were operated in Shanghai.The main purpose of this paper is to characterize particle size spectra, and analyze NPF events and their potential contributions to aerosol size spectra evolution.

Sampling Time and Location
In situ measurements were conducted continuously on the roof of a 6-storey teaching building about 20 m above the ground (31°18′N, 121°29′E) within the campus of Fudan University in Shanghai from October 2008 to February 2009.The sampling site is located in an urban district prevailing education, inhabitation and commerce activities with potential pollution sources of residential and traffic emissions, roughly 20 km northeast of the city center.All dates and times reported are local time (LT) of 8 h ahead of UTC.

Instrumentation and Measurements
Particle size-segregated monitoring provides a quantitative measure of ambient aerosol number concentration in individual size stage and total particle size distribution.In this study, a set of Wide-range Particle Spectrometer (WPS TM , model 1000XP, MSP Corporation, USA) was used to measure the number size distribution of aerosols in diameter from 10 nm to 10 μm.The WPS consists of a high-resolution Differential Mobility Analyzer (DMA), a Condensation Particle Counter (CPC) and a wide-angle Laser Particle Spectrometer (LPS).The sampled air flow enters the instrument through a common inlet at a flow rate of 1.0 L/min, of which 0.70 L/min is sampled into the LPS and the remaining 0.30 L/min is sampled for the DMA followed by the CPC.The DMA has a cylindrical geometry with an annular space for the laminar aerosol and sheath air flows, which critical dimensions were optimized to obtain size classification of particles between 10 and 500 nm with a maximum voltage of 9000 volts and a minimum voltage of 10 volts when operating with a sheath flow rate of 3 L/min.The CPC is of thermal diffusion type, with a saturator maintained at 35°C and a light-scattering droplet counter to count mobility-classified particles (counting accuracy ±10%) from the DMA.The CPC has a dual reservoir to prevent the working fluid from being contaminated by water due to moisture condensation in the condenser.The LPS is a singleparticle wide-angle optical sensor to detect particles from 0.4 to 10 μm in diameter (counting efficiency nearly 1.0), in which particles are drawn into the aerosol inlet at a flow rate of 0.70 L/min and focused with a 3 L/min flow of sheath air towards the center of a ribbon-shaped laser beam from a laser diode.Details of the WPS have been described by Liu et al. (2010) including main components, calibration and standardization, and measurement examples etc.
The ambient air passed through a black conductive tube of 1.5-m length and 0.65 cm i.d. at a flow rate of 1.0 L/min (transmission efficiency 0.90, Kumar et al., (2008)), and then was dried by one active carbon dryer (RH less than 40%) before entering the WPS.Before and after the campaign, the DMA was calibrated with Standard Reference Materials (SRM) from the U.S. National Institute of Standards and Technology (NIST) such as uniform size polystyrene latex (PSL, Duke Corporation) spheres of SRM 1691 (mean diameter 0.269 µm) and SRM 1963 (0.1007 µm) to verify proper DMA transfer function and accurate particle sizing traceable to NIST (sizing accuracy ±3%).The LPS was calibrated for size and optical resolution with four NIST traceable PSL sphere sizes (mean diameter 0.701, 1.36, 1.6 and 4.0 μm).Similar WPS instruments have been employed to collect available data on aerosols under polluted conditions in the Chinese YRD (Gao et al., 2009;Zhang et al., 2010a).In present study, we chose the sample mode with 60 channels for DMA and 24 channels for LPS, and the WPS took about 3 min for one complete scanning over the entire size range with 2 sec for each channel.The datasets were manually edited to remove invalid data resulting from instrumental problems and analyzed primitively using a software (WPS Commander) provided by the manufacturer.
Ozone was measured using an ozone analyzer with UV photometer (model 49i, Thermo Fisher Scientific, Co., Ltd), and nitrogen oxides were monitored by a NO-NO 2 -NO x analyzer (model 42i,Thermo Fisher Scientific,Co.,Ltd).Quality control of zero check was performed automatically and span checks were performed every day.Filters were replaced every two weeks, and calibration was made every three months.Detailed information on these instruments and application is available elsewhere (Zhang et al., 2010a).Data on concentrations of these major gaseous pollutants were collected at a time resolution of 1 min.Meteorological parameters such as temperature, relative humidity (RH), wind and visibility etc. were from the Baoshan weather station (professional weather station nearest to the site) of the Shanghai Meteorological Bureau.Daily averages of PM 10 and SO 2 concentrations were inverted from Air Pollution Index (API) provided by the Shanghai Environmental Monitoring Center.

Particle Size Spectra
Overall, a total of seventy-three day ground-based measurements on atmospheric particle size distributions were achieved during the period of 14 October 2008 to 21 February 2009.Based on particle diameter, herein we classify particle size classes as nucleation (10-20 nm), Aitken (20-100 nm), accumulation (100-1000 nm) and coarse (1-10 µm) modes.Table 1 summaries statistics of particles, trace gases and meteorology factors during the campaign, in which the period of October to November is defined as autumn, and the period of December to next February is defined as winter.During the whole measurement period, on average, temperature was less than 15°C, RH was above 60%, wind speed was close to 2.5 m/s and atmospheric visibility was lower than 16 km.Such stagnant meteorological conditions were facile and usually contributed to the occurrence of air pollution (e.g.haze) with relatively higher daily average PM 10 , resulting in about 20% of the observation days exceeding the second grade of China's National air quality standard (150 µg/cm 3 ).Daily average SO 2 concentrations were 0.06 mg m -3 , and the averages of other major gaseous pollutants were 15-21 ppb for O 3 and 16-27 ppb for NO x , which were comparable to measurements conducted in shanghai from January 2006 to May 2007 (Geng et al., 2008).
As shown in Table 1, the average number concentrations of particles integrating for full sizes were up to approximate 1.3 × 10 4 1/cm 3 in both autumn and winter.In fact, the temporary values of these particle number concentrations showed high variability with a range of 5.4 × 10 2 to 3.4 × 10 4 1/cm 3 in autumn and 1.2 × 10 3 to 9.3 × 10 4 1/cm 3 in winter.In terms of number per volume, particles of Aitken and accumulation modes were predominant in the entire particle group with fractions of 62-66% and 25-30%, respectively, however, nucleation mode particles accounted for only about 7-9% and coarse particles were negligible 0.1%.Total particle surface and volume concentrations were calculated by integrating the number size distributions of size-resolved particles assuming particle sphericity and density 1.0 g/cm 3 , which is reasonable and comparable to 1.0-1.5 g cm -3 in Riverside, California and 1.2-1.8g/cm 3 in Queensland, Australia (Morawska et al., 1999;Spencer et al., 2007), and their averages were 6.6-7.7 × 10 2 µm 2 /cm 3 and 54-66 µm 3 /cm 3 for autumn and winter.Accumulation mode particles dominated exceeding 80% of the total particle surface area concentration, and contributed about 30-45% of the total particle volume concentration.
In total, the plots of particle number concentration as a function of particle diameter were similar over the entire 5month period, as did the scattering of particle surface and volume concentrations versus particle diameter.The peak diameters of particle number size distributions were almost within the size range of Aitken mode, and the peak diameters of particle surface and volume size distributions mostly fell in the range of accumulation mode.Fig. 1 shows mean particle number, surface and volume size distributions averaged over autumn and winter.All of these particle size spectra could be characterized successfully by integrating multi-lognormal functions, three-mode for number size distribution and two-mode for surface and volume size distributions.For example, the geometric mean diameters (GMD) of particle number size distributions fitted by multi-lognormal functions centered at around 24, 60, 223 nm in autumn and 27, 48, 148 nm in winter.These GMDs were 546, 2479 nm and 372, 2172 nm for surface size distributions and 488, 4800 nm, and 398, 6129 nm for volume size distributions in autumn and winter, respectively.The fitting parameters of geometric standard deviation (GSD) were roughly in magnitude of 1.3-3.2 for particle number, surface and volume size distributions.
Fig. 2 illustrates mean diurnal variations of particle number size distributions and particle concentrations during a diel time in autumn and winter.Two peaks of size-resolved particle number concentrations were located within 30-60 nm and obviously found at the time of 7:00-9:00 and 18:00-21:00 in autumn, and 6:00-10:00 and 17:00-20:00  in winter, which were consistent with the known morning and afternoon traffic rush hours in urban areas.This diurnal pattern was possibly attributed to strong vehicle emissions at the traffic times and meteorological conditions to hamper pollutant dispersion such as inversion and lower mixing height, comparable to previous studies on urban pollutions (Alam et al., 2003;Pitz et al., 2003;Wehner and Wiedensohler, 2003;Hussein et al., 2004).The diurnal variation of mean total particle number, surface and volume concentrations also confirmed this phenomenon (Fig. 2).
Fig. 3 provides an overview of the diurnal variations of average number concentrations of nucleation, Aitken and accumulation particles during the whole campaign.As total particle number concentration shown in Fig. 2, the mean number concentrations of particles in sizes ranges of 50-100 nm, 100-500 nm and 500-10000 nm showed a similar diurnal pattern with two peaks corresponding to the traffic rush periods.Besides of these peaks related to traffic exhaust contributions, the mean number concentrations of 10-20 nm particles appeared one additional peak at the time of 9:30-13:00, markedly in autumn.The most possible explanation is NPF and growth processes, which may contribute to certain extent the increase of particle loadings especially in nucleation mode, and the evolution of particle size spectra (Boy and Kulmala, 2002;Gao et al., 2009).Additionally, the 20-50 nm particles also appeared a relatively moderate peak in number concentration at the time of 10:00-13:30, along with 10-20 nm particles in particular of autumn.This result implicates that the formation and subsequent growth of new particles tightly links to Atiken particles, and impacts the loading of particles in nucleation and Atiken modes other than larger sizes.Therefore, NPF events were analyzed in the following section.

New Particle Formation Event
According to previous studies, the key criterion for discerning NPF event is the acute burst of nucleation mode particles, namely newly formed particles up to detectable size 3 nm exceeding the given value in number concentration although somewhat different in various environments, and lasting one long time of several hours (Birmili and Wiedensohler, 2000;Kulmala et al., 2004a).Moreover, for excluding disturbance of the direct emission of primary particles particularly anthropogenic particles in urban areas (Shi et al., 2001), the apparent 'banana' shape of particle number concentration as a function of time and particle diameter, and preexisting particle surface concentration, precursor gas and meteorological conditions etc. have been used for supplementary criterions to identify and characterize atmospheric NPF events (Boy and Kulmala, 2002;Heintzenberg et al., 2007;Olofson et al., 2009).In this study, we adapted these criterions for identifying NPF events.
Although nucleation mode particle group in absence of 3-10 nm particles because of instrument lower detection limits, in this study the particles of 10-20 nm were still used to represent a part of growing newly formed particles, which are able to trace and describe NPF events under urban environments as demonstrated by previous studies (Gao et al., 2009).Under lower preexisting particle level and favorable meteorological conditions for facilitating pollutant dispersion, one burst of newly formed nucleation particles and their subsequent growth at the rate of a few nanometers per hour, was viewed as one NPF event if a distinct nucleation mode of aerosol particles lasting from initial outbreak to maximum in number concentration for at least 1.5 h, and the maximum number concentration of 10-20 nm particles was larger than 3.0 × 10 3 1/cm 3 .On the other hand, one slight nucleation particle burst lasting a short time but without the growth of 10-20 nm particles to larger sizes (e.g.20-50 nm), were not included in the NPF events as non-NPF events due to the calculated formation rate was questionable in accuracy.And, we identified days as 'non-event' whenever no or sporadic new particles is formed.Our pattern was almost comparable to the classification of Hussein et al. (2008) with somewhat difference, who defined sporadic occurrence of newly formed particles as 'undefined' case.On a whole, the NPF events were found on 4 days during the entire period.Table 2 provides statistical parameters for characterizing NPF events, and related trace gas concentrations and meteorological factors.The increase rate of particle number concentrations in nucleation mode (dN 10-20 /dt), namely apparent particle formation rate, was a measure of the time evolution of aerosol number concentrations (N 10-20 ) in the size range of 10-20 nm from initial burst to maximum concentration.This apparent formation rate is theoretically less than the actual nucleation rate estimated using measurements of 3 nm particles because of lack of 3-10 nm particles and neglect of nm-size particle coagulation sink (Pirjola et al., 1998).The particle growth rate (GR) was the growth of nucleation particle geometric mean diameter as a function of time, which was calculated by using the method reported by Kulmala et al. (2001Kulmala et al. ( , 2004b) )   ) Tim e 10-20 nm 20-50 nm 50-100 nm 100-500 nm 500-10000 nm W inter Fig. 3. Mean diurnal variation of the number concentration of 10-20 nm, 20-50 nm, 50-100 nm, 100-500 nm and 500-10000 nm particles averaged over autumn and winter.
Table 2. Summary of environment conditions, 10-20 nm particle number concentration and temporal variation rate, and particle growth rates (GR) during the new particle formation (NPF) events and non-event days.DT is the duration time of particle formation event starting from nucleation mode particle burst to its maximum number concentration, surface is total particle surface concentration, and N 10-20 is an average number concentration of 10-20 nm particles, dN 10-20 /dt is the mean variation rate of 10-20 nm particles dependent on time.As shown in Table 2, the apparent formation rate spanned from 0.2 to 0.5 cm 3 /s, which was within the range of 0.01-10 cm 3 /s measured in the boundary layer, and far smaller than the new particle formation rates reported at other urban locations such as Atlanta (20-70 cm 3 /s), St. Louis (1-80 cm 3 /s), New Delhi (3.3-13.9cm 3 /s) and Beijing (3.3-81.4cm 3 /s) (Kulmala et al., 2004a;Mönkkönen et al., 2005;Wu et al., 2007).Recently, Gao et al. (2009) reported the apparent formation rate 1.2-2.5 cm 3 /s observed in summer in Taicang, one small city near Shanghai, was slightly larger than this study.The growth rate was about 3.3-5.5 nm/h, within the range of 0.1-20 nm/h reported by vast studies in worldwide polluted and clean environments, which was higher than 2-3 nm/h over the Boreal forest regions (Kulmala et al., 2001), and comparable to 1.2-5.6 nm/h for 2006 summer in Beijing (Yue et al., 2009) and 3.6-7.4nm/h in Taicang (Gao et al., 2009), and slightly lower than 0.1-11.2nm/h for 2004-2005 in Beijing (Wu et al., 2007), 0.5-9 nm/h in Mexico City (Dunn et al., 2004) and 11.6-18.1 nm/h in New Delhi (Mönkkönen et al., 2005).Generally, in comparison with non-events, during the event periods environment conditions favored the NPF due to photochemistry reactions such as relatively lower preexisting particle surface concentration (non-/event ratio: 0.57, following the same), air temperature (0.7), RH (0.9) and NO x concentration (0.33), and higher wind speed (1.3), atmospheric visibility (1.4) and O 3 concentration (1.26).This point was consistent with the reviews to previous measurements (Boy and Kulmala, 2002;Holmes, 2007;Kulmala and Kerminen, 2008).

Date
One example of the NPF event on 27 November 2008 is shown in Fig. 4. Mixed layer depths were calculated using the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT4) model (Draxler and Rolph, 1998) event.The burst of nucleation mode particles happened at around 10:00 after sunrise, denoted as the sharply increasing of 10-20 nm particle number concentrations.And then, the particle number size distributions showed a clear 'banana' shape lasting from 11:00 up to 22:00, denoted as high levels of 20-50 nm particle number concentrations due to the growth of newly formed particles to larger sizes.Before the start of the NPF event, wind speed and the mixed layer depth of boundary atmosphere increased due to thermal convection during early morning hours and continued the whole event (Fig. 4(c)).These environmental conditions were useful for particles suppressed below the inversion layer in the night and primary particles (size nearly 30 nm) from vehicle emissions to disperse in the atmosphere more easily.Thus, this situation of relatively clear atmosphere is favorable for particle nucleation and new nucleated particles to reduce loss by sinking on preexisting particle surfaces.As shown in Fig. 4(b), particle surface concentrations decreased straightly for the whole day, especially just before the event particle surface concentration dropped from 670 µm 2 /cm 3 at 8:20 to 520 µm 2 /cm 3 at 10:00.Such low particle surface area concentrations could weaken the removal of preexisting particles to the newly formed particles as reported by others (Kulmala et al., 2001(Kulmala et al., , 2008)).During the event, the number concentrations of 10-20 nm particles increased very fast from 760 1/cm 3 at 10:00 and reached the maximum value of over 3200 1/cm 3 at 12:20.Following the increase of nucleation mode particles, the number concentrations of 20-50 nm particles also increased significantly about three times higher than that before the event, and its initial time (11:00) delayed about 1 hour compared with the kickoff of nucleation particle burst (10:00).As a whole, the apparent formation rate and the growth rate of new particles during this event were about 0.24 cm 3 /s and 4.5 nm/h, respectively.Fig. 4(c) and Table 2 also show that high atmospheric visibilities, low NO x and high O 3 concentrations provide facile conditions for NPF in urban areas such as strong solar radiation, neglectable effects of particles directly emitted from anthropogenic sources, and sufficient precursor gases for photochemical reactions.For the event period, the NPF largely influenced the particle size spectra of aerosols rather in nucleation and Atiken mode particles (Fig. 5), but its impacts on particle surface and volume concentrations were insignificant.
As proven by many studies, the atmospheric aerosol formation and growth is tied strongly with chemistry, particularly for sulfuric acid and other condensable vapors of low volatility (Boy and Kulmala, 2002;Kerminen et al., 2004;Zhang, 2010b).Therefore, we will take efforts in future works on chemistry including gaseous compounds participating in aerosol formation and the chemical composition and other properties of nucleated particles.

CONCLUSIONS
Continuous measurements were carried out from October 2008 to February 2009 in Shanghai to characterize the evolution of urban aerosol size spectra under autumn and wintertime conditions.The average number concentrations of particles were approximate 1.3 × 10 4 1/cm 3 , and Aitken and accumulation mode particles were dominant and accounted for over 87% of the particle number per air volume.Accumulation particles were a major contributor to the surface area of total size particles with a fraction of exceeding 80%.The maximum diameter of particle number size distributions was within Aitken size range.The average particle number size distributions were characterized and fitted successfully by integrating three-mode multi-lognormal functions.
The NPF events were distinctively observed on four days.The apparent formation rates of these events varied from 0.2 to 0.5 cm 3 /s, and the growth rates were about 3.3-5.5 nm/h which were comparable to the values reported in previous studies conducted at various urban locations.Our study has enriched data on the particle size spectra evolution of urban aerosols and shed light on NPF under urban conditions.

Fig. 1 .
Fig. 1.Mean particle number, surface and volume size distributions averaged over autumn and winter.

Fig. 2 .
Fig. 2. Mean diurnal variation of particle number size distribution (dN/dlogDp) and mean particle number, surface and volume concentrations integrating in all sizes, which were averaged over autumn and winter.

Fig. 4 .
Fig. 4. Change of (a) particle size distribution, (b) number concentrations of 10-20 nm and 20-50 nm particles and surface concentration integrating for all sizes, and (c) NO x , O 3 and mixed layer depth as a function of time on 27 November 2008, a day with a well-defined and typical formation event.

Table 1 .
Averages of ground-based aerosol and trace gas concentrations, aerosol size distribution parameters, and meteorology factors from October 2008 to February 2009.GMD is the geometric mean diameter of particle number size distribution fitted by integrating multi-lognormal functions, RH is relative humidity, Vis is atmospheric visibility, and WS is wind speed.