Interactive comment on “ Submicron aerosols at thirteen diversified sites in China : size distribution , new particle formation and corresponding contribution to cloud condensation nuclei production ”

Understanding the particle number size distribu- tions in diversified atmospheric environments is important in order to design mitigation strategies related to submicron particles and their effects on regional air quality, haze and human health. In this study, we conducted 15 different field measurement campaigns between 2007 and 2011 at 13 indi- vidual sites in China, including five urban sites, four regional sites, three coastal/background sites and one ship cruise mea- surement along eastern coastline of China. Size resolved particles were measured in the 15-600 nm size range. The median particle number concentrations (PNCs) were found to vary in the range of 1.1-2.2 ◊ 10 4 cm 3 at urban sites, 0.8-1.5◊ 10 4 cm 3 at regional sites, 0.4-0.6 ◊ 10 4 cm 3 at coastal/background sites, and 0.5 ◊ 10 4 cm 3 during cruise


Introduction
Atmospheric particles play an important role in the degradation of visibility and changing the balance of global radiative forcing (Dusek, 2006;IPCC, 2007), besides showing adverse impacts on human health (Heal et al., 2012).Size of atmospheric particles, ranging from 1.5 nm up to hundreds of micrometers, is a key factor for evaluating environmental effects of particles (Kulmala et al., 2013;Buseck and Adachi, 2008;Kumar et al., 2014).For example, particle diameter is considered to be more important than chemical composition for cloud-nucleating ability (Dusek, 2006).Smaller particles may have greater potential of health impacts compared with their larger counterparts (WHO, 2013).Particles smaller than 100 nm in diameter (ultrafine particles) have deeper deposition in human body and are able to induce more intense oxidative stress in cells (Nel et al., 2006).Accumulation mode particles, on the other hand, have high light extinction efficiency and can explain degradation of visibility in severe air pollution event to a great extent (See et al., 2006).Meanwhile, particle size distribution offers information on type, origin, and atmospheric transformation of the particles (Buseck and Adachi, 2008;Harrison et al., 2011).Therefore, the knowledge of size distributions of submicron particles, including their temporal and spatial variability, is crucial in characterizing human exposure, estimating climate effects, and designing monitoring strategies for both developed and developing countries (Kumar et al., 2014).
Measurements of particle number distribution (PND) have been extensively conducted in many European and US sites in the past 2 decades (Asmi et al., 2011a;Bigi and Ghermandi, 2011;Borsos et al., 2012;Kumar et al., 2010;Wehner and Wiedensohler, 2003), but at a much lesser extent in developing countries (Monkkonen et al., 2005;Wu et al., 2008;Kumar et al., 2011;Wang et al., 2013a, b).Temporal and spatial variation of PND in these developed countries has been widely recognized (Asmi et al., 2011a).Whilst these particles arise from a number of non-vehicle exhaust sources (Kumar et al., 2013b), primary emission from road vehicles is thought to be a main source of particle number concentration (PNC) in urban areas (Kumar et al., 2010).
The new particle formation (NPF), associated with a rapid burst of nucleation mode particles, results in an increase of cloud condensation nuclei (CCN) number concentration after growth (Wiedensohler et al., 2009).The NPF events were observed in many atmospheric environments in the world, especially in relatively clean atmosphere (Kulmala et al., 2013).Though the first study on NPF events during polluted episodes was conducted in the megacity of Beijing (Wehner et al., 2004;Wu et al., 2007), the occurrence of NPF events is only reported at a few sites in China up to now (Du et al., 2012;Liu et al., 2008;Wang et al., 2013c;Wehner et al., 2004;Wiedensohler et al., 2009;Herrmann et al., 2014).Although it is a common interest that regional NPF events are a main source of atmospheric CCN production (Kuang et al., 2009), most of the estimation on the contribution of NPF events to CCN concentration till date are only based on model simulations using regional/global scale models with extra NPF mode (Merikanto et al., 2009;Yu et al., 2012).The only few measurement studies that attempted to quantify the strength of this nucleation process merely calculated the enhancement of CCN concentration along with NPF events (Yue et al., 2011;Kerminen et al., 2012;Kuang et al., 2009;Asmi et al., 2011b).These enhancement results are likely to be greatly influenced by other factors such as change of boundary layer as well as primary emissions.More measurements and new measurement-based approaches are therefore needed to estimate the contribution of NPF to corresponding CCN concentration in diversified environments.
China has been experiencing unprecedented modernization and urbanization process since the 1980s.The rapid industrial revolution has intensively occurred in China during the past 30 years, providing both the chance to become "the world factory" and the challenge of severe air pollution problem.The aerosol pollution in recent years at both local and regional scale has attracted great attention, as it results in the heavy smog or haze episodes and might lead to potential health effects (Xu et al., 2013).The aerosol pollution characterizes regional property in major economic developed regions with megacity or city clusters, for instance, Bohai Sea rim region, Yangtze River delta and Pearl River delta.To understand the feature of aerosol pollution in these regions requires multi-site measurements within each pollution region.
In this study, we have therefore conducted PND measurements at 13 different sites in China in order to provide comprehensive understanding of the effects of primary emissions, regional pollutants transportation and new particle formation on PNDs.These sites are classified into four main categories -urban, regional, coastal and ship cruising -representing the typical atmospheric environments in China.Unique features of each site category, including the particle size distributions, seasonal and diurnal variation, are discussed in this paper.Special focus is given to the NPF events and their contribution to CCN concentration at each of the thirteen sites.

Sites description
A total of 15 field measurement campaigns of PNDs were conducted at 13 different sites between 2007 and 2011.These sites are broadly classified in the following four types according to the atmospheric environments (see Table 1), including five urban sites, four regional sites, three coastal/background sites and one ship cruise measurement along Eastern coastal China.As shown in Fig. 1, most of these sites were situated in the most developed and largest city cluster regions in China.
Table 1.Summary of information providing description of sampling periods at each measurement site.Please note that the words in parenthesis against each site represent a short name of each site, and subscripts u, r, c and s indicate urban, regional, background/coastal and cruise site types, respectively.The location of all the thirteen measurement sites.Red, yellow and green color dots and blue line represent the sites types of urban, regional, costal and cruise.Black circles show the three largest urban areas in China.

Urban sites
The first urban site, Guangzhou (GZ u ; 23.13 • N, 113.26 • E) was located on the roof of the Guangdong Provincial Environmental Monitoring Center in the downtown of Guangzhou city, at a height of about 50 m above street level (Yue et al., 2013).This site is representative of a typical ambient condition in Guangzhou urban areas (Zhang et al., 2008).
The second urban site, Shanghai (SH u ; 21.53 • E, 31.23 • N), was located on the roof of a six-floor building of Shanghai Pudong Environmental Monitoring Station in the eastern part of Shanghai urban areas.The surroundings of this site were mainly residential and business buildings (Huang et al., 2012).
The third urban site, Urumchi (UR u ; 87.58 • N, 43.83 • E), was located on roof of the Urumchi Environmental Monitoring Center in the downtown of Urumchi city.Urumchi city is the capital of Xinjiang Uyghur Autonomous Region, with the biggest desert in China to the north and mountains to the south.Sampling inlet at UR u site was located at ∼ 20 m above the street level.The surroundings were mainly residential and business buildings.There was one main road 200 m away to the west and another 400 m away to the east.
The fourth urban site, Wuxi (WX u ; 31.56 • N, 120.29 • E), was located on the roof of a five-floor building in the center of Wuxi city.This site was surrounded by the residential buildings.No obvious stationary sources existed nearby and the nearest main road was about 300 m away to the east.The sampling inlet was at ∼ 15 m above the ground level.
The fifth urban site, Jinhua (JH u ; 29.1 • N, 119.69 • E), was located on the roof of the Jindong Environmental Building in the east part of the city of Jinhua.The sampling inlet was at ∼ 25 m above the street level.There was no industry source nearby and the nearest main road was located ∼ 300 m away to the west.As the wind direction is always from the northeast during the measurements (October), the influence of this main road on our site can be ignored.

Regional sites
The first regional site, Heshan (HS r ; 22.71 • N, 112.93 • E), was an urban outflow site of Guangzhou megacity in central Pearl River delta (PRD), with a distance from Guangzhou downtown 50 km.It was located on the top of a small hill (40 m), about 7 km away from Heshan downtown areas, and was far from any industrial sources.The surrounding areas of the site were dominated by farmlands and forests.Biomass burning events were observed occasionally in the farmlands.This site can be well representative of the air pollution outflow from the polluted central PRD urban areas (Gong et al., 2012).The second regional site, Kaiping (KP r ; 22.33 • N, 112.54 • E), was located ∼ 120 km away from the city of Guangzhou to the southwest.Under the influence of the Asian monsoon, the dominant air mass comes to PRD from the northeast in fall.Hence, the Kaiping site could be assumed to be a downwind receptor site.Instruments were placed on the third floor of the building at the Kaiping supersite (∼ 10 m above the ground level), which is surrounded by shrubs and eucalyptus forest (Wang et al., 2013c).The site was free of any significant local pollution emissions.A detailed geographic description of this measurement site can be seen elsewhere (Huang et al., 2011).The third regional site, Jiaxing (JX r ; 30.8 • N, 120.8 • E), was a suburban site between Shanghai and Hangzhou.It was located on the roof of school building in a small town, 15 m above the ground level and 8 km away from the downtown of Jiaxing city, which was situated in the middle of Shanghai megacity and Hangzhou, capital of Zhejiang province.The fourth regional site, Yufa (YF r ; 39.51 • N, 116.3 • E), was about 40 km to the south of Beijing downtown area.This site was located on top of a building (∼ 20 m above the ground level) at the campus of Huangpu College.There were no industrial sources around this site, except the farm land and a residential area (Guo et al., 2010).

Three coastal/background sites
The first coastal site, Baguang (BG c ; 22.65 • N, 114.54 • E), was located on the roof of a three-floor building (∼ 10 m above sea level) at the seaside in a small peninsula in the southern China, 50 km away from the city of Shenzhen to the east.No stationary source or traffic source was found nearby the site.To the east and south of the site was the South China Sea.This site can well represent the background atmosphere of southern China during autumn.
The second coastal site, Wenling (WL c ; 28.40 • N-121.61• E), was situated in a flat ground area in a peninsula in the southeast of China, surrounding by only farmland.The sampling inlet was about 4 m above the ground level.The East China Sea was up to 2 km away from the site to the northeast and southeast.The city of Taizhou was about 30 km away to the northwest of the site.Wind direction ranged from north to east during the whole campaign at WL c site, resulting in a variety of air masses that can be encountered from modified clean maritime air mass to polluted continental air mass.
The third coastal site, Changdao (CD c ; 37.99 • N, 120.70 • E), was located at ∼ 50 m above the sea level on a hill in the north coast of Changdao island.This island lies offshore to the east edge of central eastern China and laid between the Jiaodong and the Liaodong Peninsula in the Bohai Sea.A detailed geographic description of the measurement site is presented in (Hu et al., 2013).

Ship cruise measurement
Ship cruise measurement (ES s ) was carried out on the "Dong Fang Hong 2", which is a multi-functional marine research vessel (http://eweb.ouc.edu.cn/4b/61/c4169a19297/page.htm).The observatory was located on the 6th floor of the "Dong Fang Hong 2", which was about 15 m above the sea level.The vessel sailed from Qingdao of Shandong province (24.5 • N, 118.1 • E) on 17 March, reached the southernmost area of the cruise near Wenzhou of Zhejiang province on 27 March, and returned at Qingdao (24.5 • N, 118.1 • E) on 9 April.The study area of the whole cruse covered both the East China Sea and Yellow Sea of China (see Fig. 1).

Instrumentation
The Scanning Mobility Particle Sizer (SMPS, TSI Inc.) system was used to obtain PNDs in the 15-600 nm (mobility diameter) size range at all the 13 sites.The SMPS system includes one differential mobility analyzer (DMA) and one condensation particle counter (CPC).The time resolution of this system was 5 min and the flow rates of sample and sheath air were 0.3 and 3.0 L min −1 , respectively.The relative humidity of sample air was kept below 40 % by a silica diffusion dryer within the inlet lines and sheath air cycles.Sizedependent particle losses inside the instruments as well as in the sampling tubes are calculated (Willeke, 1993), and the data are corrected by the obtained correction parameters for each site.

Parameterization
PNDs at each site are parameterized by a multiple log-normal distribution function.Each mode is described by the following function (Seinfeld and Pandis, 1998): where N i , µ i , and σ i are the total number concentration, mean diameter and geometric mean standard deviation of the distribution of mode i, respectively.The task of the fitting program is to minimize the residual part Q, which is described as follows: The growth rate (GR) of newly formed particles and condensational sink (CS) are calculated for NPF events.The CS determines the how quickly of the gaseous molecules can condense onto the pre-existing aerosols and can be calculated by using Eq. ( 3) (Kulmala et al., 2001(Kulmala et al., , 2012)).
Here D is the diffusion coefficient, n(D p ) represents the dry particle size distribution function, β M is the transitional correction factor for the mass flux, and N i is the particle number concentration in the size section i.The CS value in NPF days is calculated as the mean value, between 09:00 and 12:00 h (local time), of the day.It should be noted that the CS values calculated here are based on the dry particle number size distributions, which may underestimate the real CS values in ambient humidity.GR is calculated using the Eq. ( 4) (Kulmala et al., 2012): where D p,m is a mean geometric diameter resulted from the log-normal fitting of the particle number size distribution.
The methods for calculating both CS and GR are described by Wu et al. (2007).
The critical diameter (D p,crit ) at which 50 % of the particles are activated at the super saturation (S c ) can be calculated based on the knowledge of ambient Kappa (κ) value expressed by Eq. ( 5) (Petters and Kreidenweis, 2007). (5) where κ is the hygroscopicity parameter used to model the composition-dependence of the solution water activity; σ s/a is the droplet surface tension (assumed to be that of pure water with a value of 0.0728 N m −2 ); M w is the molecular weight of water; ρ w is the density of liquid water (g cm −3 ); R is the universal gas constant (J mol −1 K −1 ), and T is the absolute temperature (K).

Spatial and seasonal variability of PND
The statistics of PNC in the 15-600 nm size range are given in Table 2. Average and median PNCs at all sites were in the range of 0.5-2.8× 10 4 cm −3 and 0.4-2.2× 10 4 cm −3 , respectively.The median PNC at urban (1.1-2.2 × 10 4 cm −3 ) and regional sites (0.8-1.5 × 10 4 cm −3 ) were 2 times larger than those at the coastal/background sites (0.4-0.6 × 10 4 cm −3 ) and during cruise measurement (0.5 × 10 4 cm −3 ).The highest PNCs were observed at UR u site due to frequent NPF events, as well as intensive primary emissions, such as coal combustion for local industries, heating supply and residents' use (Li et al., 2008).The average observed PNCs at Chinese urban sites (1.8 × 10 4 cm −3 ) were higher than those in European cities which were reported as 1.6-7.0× 10 3 cm −3 and 1.6 ± 0.8 × 10 4 cm −3 by Borsos et al. (2012) and Kumar et al. (2013a), respectively, suggesting a much larger exposure risks to Chinese population compared with those in European cities.Meanwhile, the average PNCs in low aerosol-loading area (coastal/background sites) in China were still two-fold higher than those in European countries (Asmi et al., 2011a).This suggests that the high aerosol loading in the background areas of China increases the background aerosol concentration.
Peak diameters for measured size distributions are calculated in the range of 24-123 nm at all sites (Table 2).The peak diameters were found higher at regional sites (89 nm on average) than at urban sites (69 nm on average).Aging of particles during their transport from urban to regional area likely leads to the growth of particle diameters (Moffet and Prather, 2009), which leads to the largest particle peak diameters at regional sites despite the frequent NPF events at these sites.Besides, particle emissions from biomass burning in regional areas have diameter larger than 100 nm (Reid et al., 2005), and may influence the diameter of ambient particles at regional sites.At coastal sites, the peaks of particle size distributions were not as sharp as those seen at other sites (Fig. 2).The PNCs show a wide variation near the peak diameters, as both clean background episodes and polluted episodes caused by transportation occurred in these areas (Hu et al., 2013).
The median particle size distribution at most of the sites can be fitted into three modes.The peak diameters of these modes were in the range of 132-327 nm for the first mode (which can be roughly recognized as accumulation mode), 59-116 nm for the second mode (which can be roughly recognized as Aitken mode), and 17-38 nm for the third mode (which can be roughly recognized as nucleation mode or the second Aitken mode) (Table 2).At HS r , CD r and ES s sites, the fitting exercises give only the first and second modes.
Strong seasonal variations were found at JX u and WX u sites, where measurements were conducted both in summer and winter.Higher PNCs in nucleation mode and lower PNCs in accumulation mode were observed in summer measurements.
Table 2. Particle diameter, percentile of total PNCs and mode fit parameters for mean size distribution during each measurement at each site.σ is the geometric standard deviation of the mode, N m is the mode number concentration, D p,m and is the geometric mean diameter of the fitted log-normal mode.Please note that the subscripts "win" and "sum" represent winter and summer, respectively.

Particle number concentration in different size ranges
The PNCs in the 15-600 nm range are separated into three different size groups, which are smaller than 25 nm (N 15−25 ), 25-100 nm (N 25−100 ) and 100-600 nm (N 100−600 ).These three size groups are used to represent three modes of nucleation, Aitken, and accumulation, respectively.The number concentration in each mode can be approximated by integrating certain size distribution using the Equation ( 7): where N a−b (t) represents PNCs in the a-b nm range at a certain time interval, D p is the geometric mean diameter of the size interval, and n i is the measured PNC in a particular size interval (cm −3 ).
As illustrated in Fig. 3, the PNCs in each size range show a good log-normal distribution in most cases, but there are still several major differences among the distribution of PNCs in different size ranges at individual sites.The N 100−600 at urban and regional sites were much larger than those at coastal sites and during cruise measurement, though these values did not show obvious differences between urban and regional sites, indicating the regional feature of secondary aerosols in the whole region of city clusters.Mean N 25−100 values appeared to be up to 10-times higher than those in 100-600 nm size range at urban sites, but no significant differ-ence was found at other sites (Fig. 3), suggesting the emission from road vehicles at urban sites.The distributions of both N 15−25 and N 25−100 show a very "narrow" distribution at most studying sites compared with European countries (Asmi et al., 2011a), mainly because there was seldom clean episodes with very few particles in the atmosphere in China.Even if the air mass was coming from the clean remote continental or oceanic region, the new particle formation and high flux of primary particle emission would increase the N 15−25 and N 25−100 within a short time (Wu et al., 2008).The PNCs of nucleation mode (N 15−25 ) particles at coastal sites were lower than those of N 100−600 and N 25−100, especially in autumn and winter, suggesting that either the nucleation rate was low, or high proportion of nucleation mode particles coagulated onto the surfaces of larger particles.Besides, a "wide" distribution (Fig. 3c) of N 15−25 indicates that though the nucleation rate was low most of time at coastal sites, a few NPF events with high concentration in nucleation mode still occurred.

Diurnal variation of PND
Figure 4 shows the diurnal variations of the average PND at all the measurement sites.These figures are sorted by seasons and site types.Diurnal PND at nearly all the measurement sites presents a build-up of high concentration belts with peak diameter between 70 and 150 nm.These belts were much more obvious during night time, when there was neither NPF event nor large amount of primary emissions, and should be recognized as the accumulation mode.As smaller particles in nucleation mode and Aitken mode have large diffusion coefficient which provides them a large possibility to coagulate onto the surfaces of larger-sized particles, the residence time for particles in the boundary layer are short (Davidson and Wu, 1990).On the other hand, accumulation mode particles had much longer atmospheric life and expe-rienced longer aging process in the atmosphere, resulting in the appearance of this accumulation belt.The higher concentration and larger diameter of particles in this accumulation belts may reflect higher degree of aging of ambient particles.For example, at some sites such as YF r and KP r , the peaks of these layers were about 100-120 nm, larger than those at the urban sites of SH u and WX u (about 80-90 nm).Besides, particles at all other sites in the three large city groups (see the black circle in Fig. 1) had these high concentration belts, suggesting the regional pollution feature within the whole city group area.In contrast, no such accumulation belt was found at UR u site, as the PND at UR u site was largely influenced by the large primary emissions and NPF events.
There were large concentrations of nucleation mode particles in the middle of the day at most of studying sites (Fig. 4), showing the contribution of the NPF events to the PNC during the whole measurements.The contributions to the PNCs were much more obvious in summer time due to much more NPF events.On the other hand, the PNCs of nucleation mode were not very high in the middle of the day during all the winter measurements and some of the autumn measurements since there were no NPF events found during these campaigns (Table 3).Compared with urban and regional sites, influence of PNC at coastal/background sites by NPF events was much lower due to the lower concentration of precursor gases (e.g.SO 2 ) at these sites.
Besides the peak of NPF events around the noon, there were another two concentration peaks during morning (between 07:00 and 09:00 h) and evening (between 17:00 and 20:00 h) rush hours at most of the urban sites, with the size of the peaks about 30-70 nm.The appearance of these peaks was due to the primary emissions from vehicles as well as the diurnal change of the height of boundary layers (Lin et al., 2009).In the meanwhile, at regional sites there were also two concentration peaks at the same time of day.The difference is that at regional sites the sizes of these peaks were much larger (about 70-150 nm) than at urban sites.This reveals the fact that biomass burning in the regional farmland contribute large amount of primary particles (Huang et al., 2011;Reid et al., 2005).Besides, such morning and evening peaks were not found at coastal sites, confirming that there was no primary emission near these coastal sites.

Frequencies and parameters of NPF events
Obvious NPF events were recognized during most of our measurements.The criterion for discerning NPF events in this study includes three steps.First, there should be a burst of PNC in nucleation mode (below 25 nm in diameter) (Birmili and Wiedensohler, 2000).Second, primary emitted species, such as black carbon and CO concentrations should not enhance significantly.Third, the event should last for more than 2 h, with the increase in particle diameter.As the lower cut-off diameter of our measurements (15 nm) was much larger than the size of nucleation clusters (1.5-2 nm) (Kulmala et al., 2013), the NPF events might occur 1 to 2 hours before we observed them.Two typical types of NPF events (the "banana" and the "apple", which were classified by Wu et al., 2007) were observed at a number of our sites.The frequencies of NPF events at urban and regional sites (38 %) were similar to previous studies that focused on the NPF events in some other sites in China (Wu et al., 2007;  The values outside the bracket represent the average CS or GR; the values inside the bracket represent the maximum and minimum CS or GR; "none" means that there is no NPF events found in the whole measurement; "-" means that the GR value can not be calculated (only for the site of UR u ), or there is no such information in the reference paper.Yue et al., 2013) and were much higher than at coastal sites and events in some other sites in China (Fig. 5).
Sulfuric acid is a well-known precursor for nucleation (Boy et al., 2005;Zhang, 2010;Yue et al., 2010;Andreae, 2013).High concentrations of SO 2 and OH provide strong source of sulfuric acid at urban and regional sites, and promote the NPF events.On the other hand, the aerosol loading in the ambient atmosphere acts as the condensation and coagulation sink and thus restrains the NPF event.Though the high concentration of pre-existing particles can hinder the NPF in urban and regional areas, there will also be high precursor concentration (such as SO 2 , and VOCs) and strong oxidation (performed as high O 3 and OH radical concentration) in these areas, which will favor the NPF.At urban and regional sites in our study, the high precursor concentration and strong oxidation may play a more crucial role than the high concentration of pre-existing particles, leading to the high frequency of NPF events.On the other hand, in the clean coastal area in China, concentrations of precursors as well as the particles are not as high as in urban area as both gases and particles will be dilute during the transport from polluted region to clean coastal area.Moreover, unlike the particles, the gaseous precursors may also undergo continually atmospheric oxidation and transform into particle phase during the transport, which makes the reduction of gaseous precursors larger than the reduction of particles, resulting in fewer NPF events at coastal sites.Besides, other reasons such as meteorological conditions might also affect the frequency of NPF events at diversified sites.In polluted areas, the NPF events are often expected to occur since the air mass is coming from the clean continental background and is not greatly influenced by the polluted urban air (Wu et al., 2007).At coastal sites, however, no NPF events were noted when the air mass was coming from the ocean side with clean air, as this cleaner air mass was not carrying enough NPF precursors such as H 2 SO 4 or low-volatile VOCs to favor the nucleation events.This is substantiated by the fact that SO 2 concentrations in the air coming from ocean side were much lower than those coming from the continent (e.g., at BG c site, average SO 2 concentration in the air mass from ocean and continent were 1.4 ppbv and 2.8 ppbv, respectively) and supported by other studies at coastal site (Yu et al., 2014).
The CS values of NPF events are calculated as the mean CS value, between 09:00 to 12:00 h (local time), of the day (Eq.3).In general, CS in all NPF days were in the range of 0.9-5.6 × 10 −2 s −1 at urban sites, 0.3-8.6 × 10 −2 s −1 at regional sites, 1.0-2.6 × 10 −2 s −1 at coastal/background sites and 0.8-1.1 × 10 −2 s −1 during cruise measurement (see Table 3).The upper limits of CS at urban and regional sites were far larger compared with those at coastal/background sites and during cruise measurement and at other sites in western countries (Table 3), which further confirms the fact that NPF events can occur under high aerosol loading in polluted areas in China (Wu et al., 2007).On the other hand, the lower limit of CS at different sites showed no obvious difference, indicating that there are "clean case" NPF events even in these polluted urban areas (Wu et al., 2007).
The GRs of newly formed particles (calculated from 15 nm to 30 nm) ranged from 4.2 to 18.1 nm h −1 at urban sites, 3.2 to 21 nm h −1 at regional sites, and 1.6 to 7.5 nm h −1 at both coastal sites and during cruise measurement (Table 3).The highest GR was found at YF r site (regional) with value of 21 nm h −1 .Average GRs at urban and regional sites were about twice those at coastal sites and during cruise measurement (Table 3), indicating that the higher concentrations of gaseous precursors in the polluted areas not only favor the formation of particles, but also accelerate the growth rate as long as the nucleation particles are formed.
Both the CS and GR results in our study are comparable to other studies performed in Beijing (Wu et al., 2007), Backgarden (Yue et al., 2013), and Xinken (Liu et al., 2008), in China.CS values in our study are generally higher than those at European and American sites and smaller than those at some sites in developing countries, such as New Delhi (see Table 3).The GR values show no obvious differences between China and other countries (both in developed and developing countries) at the same site types (see Table 3).

Evaluation of the contribution of NPF to potential CCN
The basic idea to estimate the production of potential CCN from NPF events is based on the assumption that particles larger than a certain diameter could be served as CCN.As the newly formed particles consist mainly ammonium sulfate and secondary organics (Yue et al., 2010) and primary emission particles contain a lot of black carbon and hydrophobic organics (Medalia and Rivin, 1982;Reid et al., 2005), the κ value for newly formed particles and primary emission particles may be of large difference.Some previous studies in China have concluded that the average κ value of all ambient particles is about 0.3 for many environments in China (Gunthe et al., 2011;Yue et al., 2010;Rose et al., 2010), but few has provided the κ value of newly formed particles in the atmosphere.In this study, the hygroscopicity parameter κ for the newly formed particles is taken as 0.43 at about 100 nm based on the assumption of one third of chemical components of these particles are organics and others were inorganics (Gunthe et al., 2011;Yue et al., 2010).Therefore, the only task left is to achieve the PND of newly formed particles.
In this study, mode fitting method is used to distinguish the newly formed particles from others.Diurnal half-hourly average PND data are used for the log-normal fitting.A total of three (or four) lognormal modes are achieved in the fitness exercise (see Eq. 1) on the PND data.These include up to two NPF modes (with initial diameter about 15-20 nm, which may grow up to 50 nm later), one Aitken mode (40-100 nm) as well as one accumulation mode (100-500 nm). Figure 6 illustrates an example of the mode fitting result at the site of KP r , which is a typical regional site with high frequency of NPF events.Half-hourly fitting results in the afternoon from 11:00 to 16:00 h at KP r site with two NPF modes and two pre-existing modes are shown in Fig. 6.At the site of KP r , NPF events were observed to occur between mainly between 09:00 and 11:00 h every day, resulting in a sharp and narrow nucleation mode peak at about 19 nm at 11:00 h (Fig. 6).In the following 5 hours, the peak diameter of NPF mode gradually grew from 19 to 50 nm, with the concentration of NPF mode decreasing from 7500 to 5000 cm −3 at the same time.Furthermore, when the peak diameters of these lognormal distribution NPF modes reached 50 nm, there would be a large amount of particles with diameter exceed 80 nm or even 100 nm.These particles were large enough to possibly act as CCN under a certain S c .The contribution of NPF to potential CCN production can then be calculated by integrating the distribution of the NPF mode particle with diameter larger than D p,crit .As it is challenging to extract NPF mode when the NPF peaks are submerged in the size distribution of ambient particles, in order to achieve accurate fitness results, prominent NPF peak in the size distribution is required.Therefore, our subsequent discussions focused on the time period from 13:00 to 17:00 h in the afternoon, as the time is neither too early for the formation of new particles, nor too late for primary emission to perform a dominate role in the ambient PND.
The contributions of NPF events and growth to potential CCN concentration in all measurements are calculated through this approach.Results of this exercise are illustrated in Table 4. Two values of super saturations are assumed here as 0.5 and 0.2 %, with critical diameter of 50 nm and 91 nm for newly formed particles, and 60 nm and 100 nm for other particles, respectively.Contributions of NPF to potential CCN concentration varied greatly during different measurements, ranging from 0 to 66 % when S c is 0.5, and 0 to 24 % when S c is 0.2 (Table 4).The highest values were found in the summer at WX u site.During summertime, the average contributions values at urban and regional sites were 50 % (S c = 0.5) and 18 % (S c = 0.2), which were much higher than those in other seasons.The strong oxidation condition in summer favors the formation of new NPF as well as the growth of newly formed particles, accelerating the newly formed particles to perform as CCN.On the contrary, the contribution is 0 % during two winter measurements, as no NPF events were observed during these two measurements.
Weighted average contributions for different size types are roughly calculated.Annual average contribution of NPF to CCN at urban, regional and coastal sites were found to be 33 %, 19 %, 7 % (at S c = 0.5), and 11 %, 6 %, 0 % (at S c =0.2), respectively.Good correlation (R 2 = 0.7) between growth rate of NPF and contribution to CCN among all sties was found, indicating that growth rates of NPF events are the decisive factor in the conversion of newly formed particle to possible CCN (Yue et al., 2011).
Although there were two NPF events during the cruise measurement, no contribution to CCN concentration was found by these NPF events, because the diameter of these secondary particles did not meet the critical diameter for CCN.Zhang et al. (2012) demonstrated that NPF in marine environment is not likely to produce particles of the size of CCN.Our study further confirms that the NPF events show little impact on the concentration of CCN in the marine area near polluted continent.It may need more than one day for the newly formed particles to perform as CCN.
In previous studies, people always use the CCN concentration enhancement factor to describe the contribution of NPF events to corresponding CCN (Yue et al., 2011;Kuang et al., 2009;Lihavainen et al., 2003).It is obtained by dividing the concentration of CCN-sized particles after a NPF event by that prior to the event.The average CCN enhancement factors after the NPF events at S c = 0.2 (D p,crit = 100 nm) were calculated to be 1.5 at a remote site in Northern Finland (Lihavainen et al., 2003) and in the range of 1.5-2.5 in Beijing (Yue et al., 2011).Similar conclusion was obtained in Botsalano, but with great seasonal variation (Laakso et al., 2013).The factors in these studies are higher than in our "-" represents that there is no measurement at that type of sites in the certain season; "0 %" represent that there is no NPF event found in the measurement; "< 1 %" represent that there are NPF events found in the measurement, but the relative contributions of these NPF events to the total CCN concentration are smaller than 1 %.
study because they only considered NPF days.The values would be much lower and similar to our result if the enhancements were multiplied by the frequencies of NPF events.Besides, the method used in these studies did not consider the influence of primary emissions and change of boundary layer height on CCN concentration (Laakso et al., 2013) and thus may overestimate the enhancements.The mode fit method used in this study can exclude the influence of primary emission by adding one Aitken mode in the fitting result and therefore can provide more accurate contribution.However, there are still some uncertainties and limitations in this study.First, we can only calculate the contribution to CCN production by the NPF event in the same day.Some of the preexisting particles before NPF events may also come from the NPF in previous days, but these can only be recognized as pre-existing particles rather than NPF particles through our approach.This is likely to result in underestimation of the contribution.Second, even though our method has its advantage distinguishing the nucleation particles from non-NPF particles, some non-NPF particles may have a chance to be recognized as nucleation particle when we perform the processing of diurnal average PNCs, which may lead to overestimation of results.Third, as discussed above, the approach requires very clear NPF peaks in the size distributions in order to achieve accurate fitness results.So the focusing time period is constrained between 14:00 and 17:00 h, when the NPF mode is clear and evening rush hour has not come.The contribution in the following hours or the following days cannot be precisely calculated though this approach.As it is crucial to evaluate the contribution of NPF events to CCN production in the following days, more studies on this field are needed in the future.Nevertheless, based on field measurements, our work provides annual and seasonal average contribution of NPF events to CCN production for the first time in diversified Chinese atmosphere.The estimation of contribution of NPF events to corresponding CCN might be particularly useful for the validation of global climate models.

Summary and conclusions
This paper presents size-resolved measurements of submicron aerosol in the 15-600 nm size range during the 15 individual field campaigns, taken between 2007 and 2011, at thirteen different sites in China.These sites include five urban sites, four regional sites, three coastal/background sites and one ship cruise measurement, and are mainly located within the three largest city clusters of China.The particle number size distributions are fitted into three modes (nucleation, Aitken, accumulation) at most of the sites.The median PNC at urban (1.1-2.2 × 10 4 cm −3 ) and regional sites (0.8-1.5 × 10 4 cm −3 ) were two-times larger than those at coastal and background (0.4-0.6 × 10 4 cm −3 ) sites and during cruise measurement (0.5 × 10 4 cm −3 ).
Primary emission from road vehicles as well as regional biomass burning appears to have a large contribution to the ambient PNCs.High emissions from road vehicles at urban sites resulted in up to 10-times higher N 25−100 than those of N 100−600 .PNCs in both size ranges were nearly identical at all other sites.Vehicular emissions at urban sites and biomass burning emission at regional sites led to two concentration peaks in the early morning and evening hours.
Regional secondary aerosol pollution is a main feature for the submicron aerosol pollution in China.No obvious differences in the N 100−600 particles were observed between urban and regional sites.Diurnal variations show a build-up of high concentration accumulation belt at all the regional sites.
The occurrence frequencies of the NPF events in high aerosol-loading environment of China were found to be higher than those in less aerosol-loading environments.High gaseous precursors and strong oxidation at urban and regional sites not only favor the formation of particles, but also accelerate the growth rate after the nucleation mode particles are formed.The average GR of nucleation mode particles were 8.0-10.9nm h −1 at urban sites, 7.4-13.6nm h −1 at regional sites and 2.8-7.5 nm h −1 at coastal sites and during cruise measurement.The NPF events in the less aerosol-loading environment were found to be greatly influenced by pollutant transport.
Average contributions of NPF events to potential CCN are calculated in this study.Contributions of NPF events to potential CCN at 0.2 super-saturation in the afternoon of all measurement days were 11 % and 6 % at urban sites and regional sites, respectively.On the other hand, NPF events at coastal sites and during cruise measurement had little impact on potential CCN.
Our study presents a unique data set of aerosol size distributions and general concepts of the features of submicron particle pollution along with the fundamental drivers of particulate pollution in China.Besides, our estimation of contribution of NPF events to corresponding CCN might be particularly useful for the validation of global climate models.
The Supplement related to this article is available online at doi:10.5194/acp-14-10249-2014-supplement.

Figure 2 .
Figure 2. Median distribution (solid black line), 16th and 84th percentile distribution (shaded areas) and fitted modes (green, blue and red scatter dots) at all the measurement sites.

Figure 3 .
Figure 3. Distribution of PNC in different size ranges at (a) urban sites, (b) regional sites, (c) coastal sites and ship measurement.The red, blue and green line represents the distribution of N 15−25 , N 25−100 and N 100−600 , respectively.

Figure 4 .
Figure 4. Diurnal variation of PNDs during measurements at different types of sites during different seasons.The color represents dN/d log D p (cm −3 ).All diurnal variation figures use the same axis and color bar.

Figure 5 .
Figure 5. Frequencies of NPF events at different sites.The color of red, yellow, green and blue represent the site type of urban, rural, coastal and cruise, respectively.

Figure 6 .
Figure 6.Lognormal fitting result of diurnal average PND at KP r site.The six line and symbol pictures show the mode fit results for the diurnal variation from 11:00 to 16:00 at KP r site.Totally two nucleation modes and two pre-existing modes are found by mode fitting.

Table 3 .
Summary of parameters of NPF events at different sites.

Table 4 .
Summary of contribution of NPF events to potential CCN during the time between 14:00 to 17:00 LT in all measurements.