Particle Climatology in Central East China Retrieved from Measurements in Planetary Boundary Layer and in Free Troposphere at a 1500m-High Mountaintop Site

Particle number size distribution (PNSD) is an important variable for evaluating the effect of aerosols on climate. In this study, the PNSD in the size range of 3 nm–2.5 μm was measured over a 20-month period at Mt. Tai station, located at ~1500 m asl. in central east China (CEC). The mean particle number concentrations in the nucleation (Nnuc, 3–25 nm), Aitken (NAit, 25–100 nm), accumulation modes (Nacc, 100–1000 nm) and in total measured particle size range were 3200 cm, 5200 cm, 3400 cm, and 11800 cm, respectively. New particle formation (NPF) events determined from the PNSD data occurred on 32% of measured days, with particle formation rate and growth rate of 4.0 ± 3.7 cm s and 6.1 ± 2.5 nm h, respectively. Time periods of 12:00–17:00 and 23:00–7:00 local time were selected to represent periods when the air mass at the station was dominated by planetary boundary layer (PBL) and free troposphere (FT) air, respectively. The diurnal variation of particle number concentration was influenced mostly by NPF events and evolution of the PBL. When NPF event days were excluded, the particle number concentration also experienced a seasonal variation with maximum in summer and minimum in winter. This seasonality was influenced by seasonal variations in PBL evolution and by air mass advection. The results of this study characterize the regional particle climatology in central east China in terms of particle number and size.


INTRODUCTION
Particles with diameters ranging from 0.1 µm to 1 µm are highly effective at scattering solar radiation via the direct aerosol effect (Haywood and Boucher, 2000).Particles larger than about 0.1 µm may also act as cloud condensation nuclei (CCN) and contribute to indirect radiative effects (Köhler, 1937;Twomey, 1977).These effects are thought to result in a negative global mean radiative forcing by aerosol particles (IPCC, 2013).However, large uncertainties still remain due to the absence of adequate information about spatio-temporal variability of global aerosol concentration, aerosol physico-chemical properties, and aerosol-cloud interaction (IPCC, 2013).
Particle number size distribution (PNSD) displays a large variability with the particle diameter ranging from a few nanometers to tens of micrometers, depending on different emission sources and atmospheric processes.The number concentration of particles also varies from less than 100 particles per cubic centimeter in remote areas to several tens of thousands in polluted urban areas (Kulmala et al., 2004).
All particle populations measured in ambient conditions are affected by particle sources of varying geographical extent, and the particles are typically well mixed throughout the turbulent planetary boundary layer (PBL), making it difficult to separate between different particle sources.A high-altitude mountain station is located above PBL for at least some part of the time, allowing a view of the particle population on larger geographical scale.The location of the site in free troposphere (FT) air allows the evaluation of transport from further away and exchanges with the planetary boundary layer (Gao et al., 2005).Compared to other methods of studying FT air, the advantage of mountaintop observatories is that they provide the opportunity to conduct long-term measurements in the FT.
In this work, particle number size distribution measurements (in particle diameter range 3 nm-2.5 µm) were conducted at the summit of Mt.Tai in central east China (CEC) for a 20-month period.Mt.Tai is surrounded by the most economically developed regions in China, where the air quality is typically poor.Reported CO concentrations at Mt. Tai are generally higher than at other rural midlatitude mountaintop sites in the Northern Hemisphere (Gao et al., 2005).There have been some previous atmospheric experiments at the site, such as the Mt.Tai Experiment 2006 (Li et al., 2008).Fu et al. (2008) revealed that the FT air at Mt. Tai was largely influenced by field burning of agricultural waste in the North China plain (NCP) during the summer harvest season.The chemical composition and mass size distribution of PM 1 (particles with diameter smaller than 1 µm) at Mt. Tai indicated relatively aged and wellmixed aerosols (Zhang et al., 2014).However, no long-term continuous particle number size distributions measurements at Mt. Tai have been conducted before this study.The objective of this work is to describe the characteristics of PNSD in CEC region, as measured at Mt. Tai under PBL and FT conditions.

Site Description
The measurements were made at Taishan National Basic Meteorological Observation Station located at the summit of Mt. Tai (36.25°N, 117.10°E, 1534 m asl.), and lasted from July 2010 to February 2012.Mt.Tai is located in the middle of Shandong Province and surrounded by Jiangsu, Anhui, Hebei and Henan provinces (Fig. 1).Hebei, Henan and Shandong belong to the largest provinces with many cities exceeding 5 million in population.The terrain around Mt. Tai consists of lower hills and mountains with peak heights < 1000 m asl. to the west of Mt.Tai and densely populated lowland areas to other directions.Distances to both Bohai Sea and Yellow Sea are roughly 200 km.The summit of Mt.Tai overlooks the city of Tai'an (population: 5.6 million), which is located only 10 km to the south.The city of Ji'nan (population: 6.9 million) is situated 60 km to the north.As a World Heritage site, Mt.Tai receives many tourist visits from March to October, as well as during traditional festivals in the Chinese lunar calendar.Consequently, local emissions from small restaurants and temples on the mountain are occasionally substantial local pollution sources (Inomata et al., 2008).

Aerosol Instrumentation
Aerosols were sampled through a PM 10 impactor and a PM 2.5 cyclone in sequence, and the total flow rate through the sample inlet was 16.7 L min -1 .The relative humidity of the sample air was maintained below 30% with an automatic regenerating adsorption aerosol dryer in the inlet line (Tuch et al., 2009).This procedure ensured comparability between different aerosol measurements in the same inlet.PNSD for mobility diameters between 10 and 680 nm was measured every 5 min from July to December in 2010 using a scanning mobility particle sizer (SMPS) system, which consisted of an electrostatic classifier controller platform (TSI Model 3080), a long differential mobility analyzer (DMA, TSI 3081), and a condensation particle counter (CPC3010, TSI).
From beginning of December 2010 the SMPS system was replaced by a twin differential mobility particle sizer (TDMPS; Birmili et al., 1999) and an aerodynamic particle sizer (APS, TSI Model 3321), which together measured particles with sizes from 3 nm to 2.5 µm.The instrumentation used in this campaign is the same as in an earlier publication (Shen et al., 2011), where one can look for more detailed information.The raw mobility distribution data were converted to PNSD data following the inversion routine described by Pfeifer et al. (2014).The PNSD data were corrected for particle losses inside the DMA and in the sampling line, following the method described by Wiedensohler et al. (2012).
Compared to the TDMPS, the SMPS measurements had a narrower measurable particle diameter range (10-680 nm), which excluded the smallest and largest particles and therefore underestimated the total particle number concentration.The contribution of particles with diameter larger than 680 nm to the total number concentration was less than 1% (based on the TDMPS data in this study) and could therefore be ignored.The number concentration of particles between 3 and 10 nm was highly dependent on new particle formation (NPF) events.Particles in this size range can contribute significantly to the total particle number concentration during NPF events.Therefore, the particle number concentrations from the SMPS were underestimated during NPF days relative to the results from TDMPS.

Meteorological Conditions
Meteorological data including temperature (T), relative humidity (RH), wind direction (WD) and wind speed (WS) were obtained from an automatic weather station (type DZZ4, Jiangsu Radio Scientific Institute CO., LTD, China).The temperature at the observation site varied from 18 ± 3°C (mean ± standard deviation) in summer (Jun., Jul. and Aug.) to -7 ± 5°C in winter (Dec., Jan. and Feb.).RH was higher (90 ± 20%) in summer, and lower (40 ± 20%) in spring (Mar., Apr. and May) and winter.The dominant wind directions were from southwest and northeast.Heavy rain occurred mostly in summer: the total summer precipitation was 647 mm accounting for 74% of the annual total precipitation (875 mm).According to the meteorological data with 1h resolution, foggy or cloudy periods were often observed, mainly during summer and fall (Sep., Oct. and Nov.).
The air masses at Mt. Tai during the measurement period have been identified by Zhang et al. (2014).Air masses arriving at Mt. Tai slowly in PBL accounted for 60% of all air mass back-trajectories, being most dominant in the summer and fall.These air masses arrived typically from north, south or south-east over severely polluted areas and brought both industrial and agricultural pollutants to the site (Fu et al., 2008).The remaining 40% of the backtrajectories represented air masses that travelled faster and at higher altitude over northern and north-western China.This type of air mass was dominant in winter and spring, bringing drier and cleaner air to Mt. Tai.

Data Preparation
A 20-month dataset was evaluated in this investigation.The locally contaminated time periods, including the Spring Festival (Jan. 1 st -15 th , 2011), national holidays, weekends during peak season and some traditional holidays in Chinese lunar calendar were removed from the data.These periods comprised nearly 120 days during the entire measurement period.Also periods of instrument malfunction or calibration were removed.After this, periods of short time peaks in the particle number concentration were identified and removed as probable local contamination.Even after the removal of such periods it is possible that there is still some local contamination left in the data, which could not be separated with the available information.The data left to be analyzed accounted for 75% of the measurement period.Moreover, there were periods with high frequency of clouds.Based on the meteorological data statistics, any time period when the relative humidity was higher than 95% was considered as the site potentially being inside a cloud (Schmeissner et al., 2011).When the site was inside a cloud only cloud interstitial particles were sampled because of the cut-off diameter of the PM 2.5 cyclone.Due to the removal processes of in-cloud and below-cloud scavenging (Schmeissner et al., 2011), the PNSD data derived for potentially in-cloud conditions was separated from the dataset.According to meteorological data with a 1-h time resolution the average frequency of potentially in cloud conditions during the measurement period was ~16%, leaving 59% of the total time period for further analysis as valid data.The monthly valid data coverage and types of excluded data are given in Fig. 2.
All particle number concentration data in this study are reported for standard pressure and temperature (1013 hPa and 273 K) to provide comparability with other measurements.The times of day are given in local time (UTC time + 8h) unless specifically mentioned otherwise.

Particle Number and Mass Concentration Analysis
The measured particles were divided into four size ranges (later called modes) based on their mobility diameter: nucleation mode (3-25 nm), Aitken mode (25-100 nm), accumulation mode (100-1000 nm) and coarse mode (> 1 µm) (Dal Maso et al., 2005).In some parts of the data analysis the particle number concentrations in the different size ranges were analyzed separately when analyzing phenomena relevant to only some of the size ranges.
The mass concentration of PM 2.5 can also be retrieved from PNSD data, if information on particle density is available, and particle shape is known or assumed.In the case of this analysis the particles were assumed to be spherical, and particle density was estimated from the reported mass concentrations of chemical component (organics, (NH 4 ) 2 SO 4 and NH 4 NO 3 ) of submicron particles at Mt. Tai during the same time period of the PNSD measurements (Zhang et al., 2014).The densities were combined following the volume average mixing rule and assuming the volume fractions of different components to be independent of particle size.As a result, the calculated density of aerosol particles at Mt. Tai was 1.56 g cm -3 .

Identification of New Particle Formation (NPF) Events
NPF event in the observed data were identified and classified based on the methodology presented by Dal  Maso et al. (2005).For a day to be classified as NPF event day with this method a distinct new mode of particles has to appear in the nucleation mode, subsequently grow to larger sizes, and prevail for a few hours.As the lowest detectable particle size by the TDMPS was 3 nm, the formation rate of new particles (J 3 ) was defined as the flux of particles into the observable size range.The parameters characterizing the NPF event, including J 3 , growth rate of the formed particle mode (GR), and condensation sink (CS) were calculated based on the methods suggested by Kulmala et al. (2004Kulmala et al. ( , 2012)).When calculating GR, the geometric mean diameter of the new particle mode, D p , was derived by converting the PNSD in lognormal modes with a least squares fitting algorithm DO-FIT_4.20 (Hussein et al., 2005).

Data Separation to Planetary Boundary Layer (PBL) and Free Troposphere (FT) Conditions
At the summit of a mountain, the atmospheric conditions are often influenced by the diurnal evolution of PBL height.The frequency with which a mountain site is influenced by PBL air is dependent on the mountain size, shape, geographical location, and local climate (Kleissl et al., 2007).Diurnal variations in aerosol mass concentrations at Mt. Tai have been attributed to upslope/downslope air transport and PBL development (Zhou et al., 2009).In this study the free troposphere (free atmosphere) is defined as the portion of the earth's atmosphere, above the planetary boundary layer, in which the effect of the earth's surface friction on the air motion is negligible (AMS, cited 2014).This definition includes also the nighttime residual layers (RL) from previous days as parts of the free troposphere.There are several ways to examine whether the observed air is dominantly from PBL or from FT, but most of those require auxiliary measurements that were not available for this study (e.g., Baltensperger et al., 1997;Weingartner et al., 1999;Collaud Coen et al., 2011).
Mean diurnal cycles of meteorological parameters during each season were evaluated.During each season temperature increased and wind speed decreased for a period lasting roughly from 6 to 18 at local time.The length of this period was slightly longer in summer and shorter in winter.The mean diurnal pattern of absolute humidity was calculated for each season and it showed systematically higher values during daytime (Fig. 3), indicating that there is a systematical source of water vapor during the day.Apart from the effect of evaporating cloud droplets such a source requires the site to be within PBL.From these data it can be assumed that the site is influenced by PBL air at daytime during all seasons, the time periods from 10 to 17 and from 23 to 7 at local time showing the most and least influence of PBL air, respectively.
If the NPF event days are not included and air mass is assumed to be horizontally homogeneous, the main factor affecting the diurnal variation of accumulation mode particle number concentrations (N acc ) is the diurnal evolution of the PBL height.The particle number concentration data from non-NPF days was separated and normalized for each individual day separately by the mean of the total particle number concentrations during that day to get rid of the daily fluctuation.It showed that the normalized N acc typically started to increase around 8 in the morning and reached to the maximum value at 12-18 during all seasons (Fig. 3).The same homogeneity between the seasons was also present in the median values.
Based on the diurnal cycle of the calculated absolute humidity and the diurnal cycle of normalized N acc , the time period 12:00-17:00 was selected to represent the PBL condition during all seasons, while the time period of 23:00-07:00 was selected to represent the FT condition.Using the same time periods for all days reveals the general pattern of PBL evolution, but some time periods during individual days can be misclassified.Therefore the results presented for PBL air can contain some contamination from FT air and vice versa.
Meteorological sounding data in 2011 was derived from a nearby station, Zhangqiu, (36.68°N, 117.55°E, 123 m asl.), which is located about 60 km northeast of Mt.Tai.There were soundings twice a day, at 8 and 20 local time (LT).The height of PBL and RL can be identified based on the potential temperature profile during the cases with no deep inversion or cloud, which accounted for 40% of the sounding data.During the daytime, the PBL height is defined as the location of the sharpest changes with height in potential temperature.The nighttime PBL is defined as the surface layer where the potential temperature increases with the height and the RL is the stable layer that is almost isolated form the ground (Yi et al., 2001).The mean values ± standard deviation of PBL and RL heights at 8 and 20 LT for each season are given in Table 1.It could be concluded that according to the sounding data the Mt.Tai site was above the PBL at both 8 and 20 LT.The station was above the identifiable RL in spring and winter, but quite close to the identifiable RL in summer and fall.The sounding data supports that the times we chose for defining PBL and FT are reasonable.

Overview of the PNSD and Particle Number Concentration
The mean value ± standard deviation of particle number concentration in nucleation mode (N nuc ), Aitken mode (N Ait ), and accumulation mode (N acc ) for the whole measurement period were 3200 ± 4400 cm -3 , 5200 ± 4400 cm -3 and 3400 ± 2600 cm -3 , respectively, with the total particle number concentration being 11800 ± 6200 cm -3 .The mean values of N nuc , N Ait and N acc in PBL were 5400 cm -3 , 6500 cm -3 and 3700 cm -3 , while in FT 400 cm -3 , 2600 cm -3 and 2800 cm -3 , respectively.The particle number concentration for the coarse mode (≥ 1 µm) was usually less than 5 cm -3 , and it was less than 20 cm -3 even during episodes when air carried significant amounts of desert dust from Gobi desert (Zhang et al., 2003).In this work, the coarse mode particle number concentration is not discussed in further detail.
Mass concentrations of PM 2.5 at PBL and FT conditions were calculated as 52 ± 45 µg m -3 and 34 ± 30 µg m -3 , respectively.The mass concentration within PBL was about 50% higher than that in FT condition.
Several long-term measurements of particle number concentration have previously been conducted at high altitude stations (Table 2).These studies addressed processes including exchange of mountain/valley wind, PBL/FT patterns, long-range transport of particles and NPF events, all of which can alter PNSD characteristics.For comparison the values from the different sites are presented at both ambient conditions and reduced to sea level pressure 1013 hPa in Table 2 (Seinfeld and Pandis, 1998).Even when the particle number concentrations are reduced to sea level pressure, it is clear that sites at higher elevation typically experience lower particle loading.Generally, the reported particle number concentrations in FT conditions at other sites are in the range from some hundreds to one thousand particles cm -3 , and several thousands of particles cm -3 in PBL conditions at lower elevation sites.Compared to these values the aerosol number concentrations at Mt. Tai are very high.The other high altitude sites where reported mean particle number concentrations are more than 10% of those measured at Mt. Tai are Puy de Dôme in France (1465 m asl., Venzac et al., 2009) during summer, Storm peak in USA (3210 asl., Lowenthal et al., 2002) during spring when NPF events are frequent, several sites in Indian Himalayas (Komppula et al., 2009) and Mt.Waliguan in western China (3816 m asl., Kivekäs et al., 2009).Even the mean aerosol particle number concentration at Shangdianzi (SDZ), a low altitude site in North China plain (293 m asl., Shen et al., 2011) was lower than that measured at Mt. Tai in PBL conditions.The high particle number concentrations at Mt. Tai are partly produced by frequent new particle formation events and partly by the high particle emissions and persistent high pollution levels in the region surrounding Mt.Tai.

New Particle Formation Parameters
During the measurement period, NPF events were observed frequently (130 days), which accounted for nearly 32% to the sum of valid and potentially in-cloud data.This occurrence frequency was close to the number reported at SDZ (36%, Shen et al., 2011).In order to illustrate the NPF event characteristics, the analysis of a typical NPF event is given below as an example (Fig. 4).This event occurred on May 3 rd 2011.The newly formed particles with diameter around 5 nm were first observed at around 7:40 in the morning.Before that time the nucleation mode particle concentration was less than 500 cm -3 and CS was ~0.01 s -1 .N nuc sharply increased to over 8000 cm -3 in less than an hour, corresponding to the net increase rate (dN nuc /dt) of 3.7 cm -3 s -1 .The particle loss rate due to coagulation (F coag ) was 0.2 cm -3 s -1 .Thus, the value of J 3 was 3.9 cm -3 s -1 for this NPF event.The contour plot of PNSD (Fig. 4(a)) shows the growth of the newly formed particle mode to follow the shape of "banana", corresponding to relative smooth and linear growth of the mode mean diameter D p (Fig. 4(d)).The newly formed particles were observed to grow to 80 nm in diameter within the day, in which size they can serve as the potential CCN in the atmosphere (Komppula et al., 2005), even though in polluted conditions the required particle diameter for cloud droplet activation is expected to be higher (Wiedensohler et al., 2009).During the growth period (8-21), the mean observed GR of the new particle mode was 5.6 nm h -1 derived by linear fitting to D p (t).
Generally for NPF events at Mt. Tai the onset of new particle formation was typically observed between 8 and 10, with a few exceptions at earlier than 8 or around 11.The NPF event frequency exhibited seasonal variation; it was highest (75%) in spring and relatively low during other seasons (13%, 23% and 16% for summer, fall and winter, respectively).Sulfuric acid has been proved to play an important role in the nucleation process (Kulmala et al., 2004) and the concentration of its precursor gas, SO 2 , is reported to be typically high at Mt. Tai in spring time (Zhou et al., 2009).The frequency peak is also related to meteorological factors favoring new particle formation (e.g., clear, dry, and sunny days) in spring (Birmili et al., 2003;Wu et al., 2007;Shen et al., 2011).The spring peak in event frequency is qualitatively similar to results obtained at other stations, e.g., Hyytiälä, Finland (Kulmala et al., 2001) and SDZ in NCP (Shen et al., 2011).The mean CS (Dal Maso et al., 2005) just prior the onset of the NPF event was 0.02 ± 0.01 s -1 .This value is similar to values obtained at SDZ station (Shen et al., 2011), but almost a magnitude higher than the CS values observed at several remote stations, e.g., in the South African savannah (0.004 s -1 ; Vakkari et al., 2011) and Hyytiälä (0.002 s -1 ; Dal Maso et al., 2005).It was almost an order of magnitude higher than that at other mountain sites as listed in Table 3.
The mean observed GR of the new particle mode during the measurement period was 6.1 ± 2.5 nm h -1 .This is within the range of GR values observed at rural locations (1-10 nm h -1 ; Kulmala et al., 2004).Moreover, the GR observed at Mt. Tai was higher than that at SDZ (4.3 nm h -1 ; Shen et al., 2011).Other reported growth rates at high altitude sites (Table 3) are typically 0.4-12 nm h -1 .The mean GR at Mt. Tai was similar to most of the other high altitude sites, apart from those sites with strong marine characteristic (Izaña, Mauna Loa).However, different size ranges used for calculating J and GR cause uncertainty in the comparison results.At Mt. Tai the GR was highest in summer (9.0 ± 1.0 nm h -1 ) and lowest in winter (5.5 ± 3.2 nm h -1 ).Similar seasonal GR cycles have been reported for other sites (e.g., Birmili et al., 2003;Shi et al, 2007;Mazon et al., 2009).At rural and background sites the GR during NPF events has been shown to correlate with the seasonality of precursor vapors and photo-chemical reactions over large areas (Kulmala et al., 2004).Kivekäs et al. (2014) have also demonstrated that the observed growth rate of the newly formed particle mode at a stationary measurement site can actually result from phenomena further upwind of the site.

Diurnal Variation in Particle Number Concentration
The mean diurnal variation of particle number concentration in each mode was calculated for days with NPF events and non-NPF event days separately, and both are shown in Fig. 5.The separation of the dataset was done in order to segregate the influence of NPF events and the PBL-driven diurnal pattern of particle number concentration.The days with a distinct nucleation mode of particles but without observable consequent particle growth were classified as "undefined" days, which were not included in NPF days or non-NPF days (Dal Maso et al., 2005).
During NPF days there was a clear diurnal pattern in all particle modes.It started with N nuc beginning to increase at around 8 in the morning and peaking at noon (Fig. 5(a)).The onset time of the increase matches with both the intrusion time of PBL air at the site (Fig. 3) and with the typical onset time of NPF events (Fig. 4).The concentrations of N Ait and N acc started to increase around 10 in the morning and peaked at 15 (N Ait , Fig. 5(b)) and 18 (N acc , Fig. 5(c)) in Boundary layer evolution also affects these cycles, which explains the simultaneous start of increase in N Ait and N acc .
During non-NPF days the diurnal evolution of particle number concentration was mainly affected by PBL evolution.As expected, at nighttime the N nuc values during non-NPF days were similar to those during NPF days.The diurnal pattern of N nuc was similar during non-NPF and NPF days, but the amplitude of N nuc during non-NPF days was an order of magnitude lower than that during NPF days (Fig. 5(a)).N nuc started to increase in the morning and reached maximum around noon.The N nuc increase during non-NPF days can be caused by non-perfect separation with the undefined days, although the undefined days have been excluded from non-NPF days.The diurnal cycle in N Ait during non-NPF days was also weaker than during NPF days (Fig. 5(b)).Local emission from Tai'an city, which is located at the foothill of Mt.Tai, could be an important source of Aitken mode particles at the site.In N acc the cycle was similar to that during NPF days, but peaking earlier at around 14 (Fig. 5(c)), which is in line with the diurnal boundary layer evolution.The peak value of N acc during non-NPF days was slightly higher than that during NPF days, indicating that typically during non-NPF days there was a larger surface area (higher CS) in the atmosphere inhibiting the NPF events to take places.

Seasonal Variation of PNSD in PBL and FT Conditions
The mean value and standard deviation of particle number and mass concentrations in PBL and FT conditions in different seasons for NPF days and non-NPF days are given in Table 4.In FT condition, the PNSD showed a lower total particle number and mass concentrations, representing somewhat less polluted regional background air.The seasonal variation of number and mass concentration for days without NPF days in both PBL and FT conditions showed a pattern of highest particle number concentration in summer and lowest in winter.This was consistent with the seasonal air mass transport and emission patterns.The air masses moving slowly through the high emission areas in CEC and NCP favored accumulation of particles in at Mt. Tai (Zhang et al., 2014).During summer the higher PBL also pumps more particles from PBL to Mt. Tai summit altitude, which explains the higher particle number and mass concentrations during summer also in FT.During non-NPF days, the particle number and mass concentrations were about 30% and 25% higher in PBL than those in FT during spring and summer, and the difference was more significant in fall and winter, with number and mass concentrations about 50% and 45% higher in PBL than in FT.It was also found that average PNSDs in PBL and FT were  Furthermore, the number concentration of particles larger than 200 nm in diameter did not vary much between the different seasons in PBL and FT, being (the seasonal mean) from 1400 to 1800 cm -3 in PBL condition and from 1000 to 1300 cm -3 in FT condition, respectively.When NPF event days were considered, the season with the highest particle number concentrations was spring.This was due to the much higher frequency of NPF events during spring compared to the other seasons.The particle number concentration in PBL was about two times higher than that in FT condition, and even 30-120% higher than that in PBL without NPF days.This indicates the NPF events could produce very high number concentrations of small particles at Mt. Tai.Even during NPF days PM 2.5 was highest in summer and it was normally lower during NPF days than during days without NPF.As the NPF events typically occurred in air with low pre-existing CS, the results for days without NPF events are somewhat biased towards more polluted air and slowly moving air masses with higher particle mass concentration.

CONCLUSIONS
Aerosol particle number size distribution (PNSD) in the size range of 3 nm-2.5 µm was continuously measured at Mt. Tai, a mountaintop station in central east China from July 2010 to March 2012.The mean number concentration of submicron particles, reduced to 1013 hPa, was 11800 cm -3 .The typical particle number concentrations at Mt. Tai were significantly higher than that at other high elevation sites around the world.
New particle formation (NPF) events were observed on 32% of the days, most frequently in spring.The mean particle formation rate (J 3 ) of 4.0 cm -3 s -1 and growth rate (GR) of 6.1 nm h -1 .It was found that the formation rate and condensation sink could be even an order of magnitude higher than most studies at the other high elevation stations.The newly formed particle modes were observed grow into the size where they can act as potential cloud condensation nuclei within a day.This could be very important for aerosol-cloud interactions, especially at the sites with high RH.During NPF days, N nuc , and N Ait were much higher than those during non-NPF days, indicating significant contribution of NPF event to particle number concentrations in those size ranges.However, N acc was found to be higher during non-NPF days.
Based on the diurnal cycles of absolute humidity as well as of normalized accumulation mode particle number concentrations, time periods 12:00-17:00 and 23:00-07:00 local time were selected to represent PBL and FT conditions, respectively.This selection was supported by the meteorological sounding data.The mean particle number concentrations of N Ait and N acc in PBL conditions were about twice of those in FT conditions, but N nuc was more than 10 times higher in PBL than in FT.This does not mean that NPF is limited to PBL, but rather that NPF is limited to daytime and the newly formed particles grow fast enough to grow out from nucleation mode before 23:00 local time.
The particle number concentrations also exhibited pronounced seasonal variations in both PBL and FT air, being highest in summer and lowest in winter during non-NPF event days.This could be linked to seasonality of air mass transport pathways, emission sources and PBL evolution.The particle number and mass concentrations were about 30-200% and 25-50% higher, respectively, in PBL than those in FT, depending on the season and NPF events.
The high particle number and mass concentrations observed at Mt. Tai were found to be systematical.The high N Ait and N acc concentration also in FT conditions revealed that this is a large scale regional phenomenon, and that the diurnal cycle in PBL height can pump large amounts of particles into free troposphere, especially in low moving air masses in the summer.This indicates that any actions to improve air quality at CEC must be taken at regional level, and that the effects and their climate implications observed at Mt. Tai can be extrapolated to larger areas in CEC.The observations at Mt. Tai can provide the opportunity to conduct long-term measurements in the FT, as well as PBL in central east China, which are important for evaluating the effect of aerosols on climate and for validating the performance of atmospheric chemical transport models.

Fig. 1 .
Fig. 1.(a) The map of North China Plain and central east China showing the location of Mt.Tai and the major cities with population larger than 5 million.The circles indicate the population in unit of million inhabitants.(b) The topography of the area surrounding Mt.Tai within 60 km distance.

Fig. 2 .
Fig. 2. Monthly data coverage and types of data during the measurement period.

Fig. 3 .
Fig. 3.The seasonal mean diurnal pattern of absolute humidity and mean and median normalized particle number concentrations of accumulation mode on days without NPF events (a-d).

Fig. 4 .
Fig. 4. The time evolution of PNSD (a), CS (b), N nuc (c) and D p obtained by fitting procedure (d) on May 3 rd 2011.The solid red lines show the linear fits used to derive dN nuc /dt and GR.The dashed black lines indicate the time periods that were selected for the linear fits.

Fig. 5 .
Fig. 5.The mean diurnal variation of particle number concentrations in nucleation mode (a), Aitken mode (b) and accumulation mode (c) for NPF days and for days without NPF events.The error bars present the standard deviation.Please note the logarithmic N nuc scale in plot (a).

Table 1 .
The mean value and standard deviation of height of PBL and RL (unit: m asl) in different seasons derived from the meteorological sounding data.

Table 2 .
Comparison of mean particle number concentrations at Mt. Tai and other sites

Table 3 .
Values of formation rates, growth rates and condensation sink observed in this and in other studies at high altitude sites.

Table 4 .
Seasonal mean and standard deviation of particle number concentration and mass concentration of PM 2.5 in PBL and FT conditions during non-NPF days and NPF days, respectively.