and growth at ﬁve Particle formation and growth at ﬁve rural and urban sites

Ultraﬁne particle (UFP) number and size distributions were simultaneously measured at ﬁve urban and rural sites in Southern Ontario, Canada as part of the Border Air Quality and Meteorology Study (BAQS-Met 2007). Particle formation and growth events at these ﬁve sites were classiﬁed based on their strength and persistence as well as the 5 variation in geometric mean diameter. Regional nucleation and growth events and local short-lived strong nucleation events were frequently observed at the near-border rural sites, upwind of industrial sources. Surprisingly, the particle number concentrations at one of these sites were higher than the concentrations at a downtown site in a major city, despite its high tra ﬃ c density. Regional nucleation and growth events 10 were favored at intense solar irradiance and less polluted cooler drier air. The most distinctive regional particle nucleation and growth event during the campaign was observed simultaneously at all ﬁve sites, which were up to 350 km apart. Although the ultraﬁne particle concentrations and size distributions generally were spatially hetero-geneous across the region, a more uniform spatial distribution of UFP across the ﬁve 15 areas was observed during this regional nucleation event. Thus, nucleation events can cover large regions, contributing to the burden of UFP in cities and potentially to the associated health impacts on urban populations. In addition, particle formation in southwestern Ontario appears to more often be related to anthropogenic gaseous emissions, although biogenic emissions may at times contribute. Local short-lived nu- 20 cleation events at


Introduction
Over the last decade, formation of particles and their subsequent growth in the atmosphere have been found to occur in many different environments (e.g., Alam et al., 2003;Jeong et al., 2004;Dunn et al., 2004;Wehner et al., 2004;Mönkkönen et al., 2005;Mejia et al., 2007;Charron et al., 2007). Measurement of the number and size 5 distributions of ultrafine particles (UFP, d p < 100 nm) has made possible the identification of particle nucleation events, which are found to exhibit large spatial and temporal variability. The concentrations of sulphuric acid and organic compounds, precursors to particle nucleation and growth, are believed to be important factors, and these may vary across locations and meteorological conditions Boy et al., 2008). However, particle formation events can occur on a large scale. Stanier et al. (2004) found that particle nucleation events occurred at both an urban background site in Pittsburgh, Pennsylvania, USA and an upwind, rural site 38 km away. Wehner et al. (2007) investigated particle formation events at an urban site in Leipzig and at three rural areas located ∼10 km to 50 km from this urban background site in Germany; this 15 study revealed that new particle formation events at the four sites took place nearly simultaneously with similar intensities. For evaluation on a larger spatial scale, several studies have investigated simultaneous measurements at rural and remote background sites at distances of ∼200 km to ∼1500 km in Finland, Sweden, and Estonia (Tunved et al., 2003;Vana et al., 2004;Dal Maso et al., 2007;Hussein et al., 2009). Hussein et as part of the Border Air Quality and Meteorology Study in the summer of 2007 (BAQS-Met 2007). The border area in this region has Canada's highest PM 2.5 mass concentrations (Environment Canada, 2007). The high pollutant concentrations coupled with the unusual meteorology induced by the nearby Great Lakes was an important-motivation for this study. In addition, the study offered a unique opportunity to explore ultrafine par- 15 ticle nucleation and growth at rural and urban sites surrounded by numerous local and regional sources of gaseous pollutants and particulate matter (PM). It was also suspected that the meteorology and local topography influence the spatial and temporal distributions of particle number concentrations in this region.
We used simultaneous measurements of UFP number and size distributions at multi- 20 ple sites to explore the spatial extent of new particle formation and to attempt to identify sources and processes contributing to UFP across the study region. The specific objectives of this study were to distinguish between regional and local nucleation events based on simultaneously measured particle number concentrations at five rural and urban sites influenced by regional and local industrial as well as traffic sources; to 25 evaluate their relationship with meteorological parameters and atmospheric pollutant concentrations; and to determine the degree of inter-site homogeneity in terms of the occurrence of events and consistency in the temporal variability of UFP number concentrations. The overall goals were to improve understanding of the influence of local Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | and regional nucleation events on UFP concentrations in urban and rural areas and provide information needed to resolve the impact of UFP on the health of urban and rural residents.
2 Experimental methods 2.1 Monitoring sites 5 The BAQS-Met 2007 intensive field campaign was conducted to investigate the effects of transport and lake breezes on pollutant levels in southwestern Ontario, as well as to determine the sources of these pollutants. Measurement of ambient pollutants and meteorological parameters was simultaneously conducted from 20 June to 8 July 2007 at five sites: Harrow, Bear Creek, Ridgetown, Egbert, and Toronto. As shown 10 in Fig. 1, the monitoring sites at Harrow, Bear Creek, and Ridgetown are geographically located in a border airshed region shared by Canada and the United States, so local air quality was typically impacted by regional and long-range transported pollutants from both Canada and the US. Wind rose plots (Fig. 2) indicated that the wind at the three near-border sites was predominantly from the south-west, the direction of 15 numerous US and Canadian industrial point sources. The Harrow site, for example, located within 5 km of the shore of Lake Erie in an agricultural area, had dominant southwesterly winds with strong wind speeds; it was influenced by long-range transported pollutants from the US and, more locally, by emissions from the Detroit/Windsor industrial region. The monitoring site at Ridgetown is located ∼10 km from the Lake electrometers to measure the number and size distributions of particles from 6 nm to 560 nm in 32 channels every second. The high time resolution data make it possible to detect rapid changes in particle number and size distributions. However, due to distortions in the particle counting and sizing efficiency of the FMPS, data for particles in the size range of 8 nm −100 nm were corrected as suggested by Jeong and Evans (2009). 20 In addition, size distributions of particles larger than 100 nm measured by the FMPS were adjusted based on polystyrene latex (PSL) calibration particles and the results of laboratory comparison experiments with the SMPS used for Toronto. After correction, the FMPS and SMPS showed good agreement (Jeong and Evans, 2009). The SMPSs used at the other sites consisted of Differential Mobility Analyzers (DMA) and Conden-from 14 nm to 25 nm, neglecting any coagulation loss. The number concentrations of particles (14 nm-25 nm) at the start and end of a period during which a linear increase was observed were used to quantify the apparent formation rate. Geometric mean diameters (GMD) for each size distribution were used to examine particle growth processes. The observed growth rate (GR) of the newly formed particles was quantified 15 by fitting the GMD of particles smaller than 100 nm during the particle formation event:

Measurements of particle number and size distributions
To quantify the condensational growth and sink of newly formed particles, the condensation sink (CS) was estimated based on the analysis by Kulmala et al. (2001 and Dal Maso et al. (2005). In brief, the condensation sink, the rate of loss of molecules 20 onto existing particles, was obtained by integrating over the size distribution as follows where D is the diffusion coefficient, β M is the transitional regime correction factor, d pi is the particle diameter of size class i , and N i is the particle number concentration in 11621 Introduction size class i . The transitional correction factor was expressed as (Fuchs and Sutugin, 1971) β M = K n + 1 1.33α −1 K n 2 + 1.33α −1 K n + 0.38K n + 1 = K n + 1 1.33K n 2 + 1.71K n + 1 ( 3) with Knudsen number K n=2λ/d p , where λ is the mean free path of the gas molecules under standard conditions (∼67 nm), and the accommodation coefficient α is assumed 5 to be unity.

Wind sector and spatial variation analyses
To ascertain the likely geographic direction of local sources of UFP, a conditional probability function analysis (CPF, Ashbaugh et al., 1985) was conducted based on particle number concentrations in the 14 nm to 100 nm size range. The CPF provided a metric 10 of the probability that high UFP concentrations were related to specific wind directions, with high values typically associated with local sources. The CPF was calculated as where m ∆θ is the number of times the particle number concentration exceeds a certain threshold criterion (the 75th percentile in this study) when the wind arrives from a given 15 wind sector (∆θ, 15 degrees), and n ∆θ is the total number of times the wind comes from the same sector. The spatial distribution of particle number concentrations at the five sites was determined using correlation analysis and coefficients of divergence (COD) for the ten pairs: Harrow-Ridgetown ( where N i j and N i k represent the hourly i th concentration measured at sampling sites j and k, respectively, and p is the number of samples (Wongphatarakul et al., 1998).
The COD provided information about the relative inter-site uniformity, with a COD of 5 zero indicating no inter-site variability, and a high COD value (approaching 1) indicating high variability. In addition to the analysis of spatial uniformity, inter-site associations for different particle size ranges were determined by estimating Spearman rank correlation coefficients (r) for site pairs.
3 Results and discussion 10

Comparison of particle number and size distributions
Number concentrations of 14 nm-100 nm particles across the five sites are presented in Table 2. Thirty-minute average particle concentrations comprising a total of 567 samples for each site were used for this comparison. The highest average was found at Harrow, an agricultural area, while the number concentrations at Ridgetown and Egbert were the lowest. Interestingly, the particle number concentration at the agricultural site was higher than that at the downtown metropolitan site in Toronto, despite its high traffic density. A similar trend was observed when the particle size range was expanded to 8 nm-100 nm: the average particle number concentration was 10 500±6000 cm by factors of 6-14 than those for the other sites, indicating higher background concentrations at Toronto due to traffic emissions. As shown in Table 2, there was a factor of ∼85 difference between the 90% and 10% percentile concentrations for Harrow as compared with ∼7 for Toronto. Harrow was impacted by intermittent, strong and shortlived nucleation events, causing the UFP concentration to rise sporadically from low to 5 high values. In contrast, Toronto was continuously affected by a baseline of <25 nm particles from vehicular exhaust. Figure 3 presents the average size distributions for UFP at the five sites. While the size distribution in Toronto exhibits a near-single mode at ∼30 nm with a minor peak at the lowest size, a distinct bimodal distribution is observed for Harrow, with modes 10 of ∼14 nm and ∼39 nm. The 14 nm mode was caused by strong particle nucleation events, with persistent subsequent growth of newly formed particles. During the BAQS-Met campaign, several strong peaks in the ∼39 nm size mode were found for Harrow without growth events (e.g., 27 June). This was likely due to the transport of slightly aged particles formed upwind of Harrow. High particle number concentrations corre- 15 sponded with high wind speeds (∼10 m s −1 ), most often originating from the southeast (Figs. 2 and 4). Hence, major power plants located in Toledo, Ohio and ∼60 km southwest of the Harrow site were possible sources of these slightly aged (<∼2 h) particles in the Harrow area.
Bimodal distributions were commonly observed at the monitoring sites, except for 20 Toronto. A number of studies conducted in urban areas have observed unimodal particle size distributions of particles larger than 10 nm (Harrison et al., 1999;Despiau and Croci, 2007;Krudysz et al., 2009 , 2002a, b;Zhang et al., 2004). The number and size distributions measured at the Toronto site, located approximately 20 m from the curbside of a busy road, may represent the evolution of freshly formed motor vehicle exhaust particles. The unimodal distribution in downtown Toronto was likely due to the influence of vehicle exhaust particles with high volatility, which may undergo rapid changes in particle size distributions 5 through evaporation and/or condensation (Biswas et al., 2007). The UFP at Bear Creek exhibited a mode at ∼52 nm, and the cutoff at 14 nm may have prevented the detection of a second mode. As shown in Fig. 4c, high concentrations of UFP were associated with southwest and north winds, indicating the possible impact of industrial emissions from the Toledo region and of oil refinery plants in Sarnia, Ontario, located ∼50 km 10 north of the site. As shown in Fig. 3, the particle size distribution at the remote site (Egbert) was close to the average for Ridgetown. Although average concentrations were lower than at other sites, the size distributions observed at Ridgetown and Egbert were bimodal, with a larger size mode of ∼60 nm. This might have been due to less frequent nu-15 cleation and longer growth processes, without abrupt changes in meteorological conditions. The mesoscale weather phenomena were different for Ridgetown, where the weather was often cloudy and rainy during the measurement period. These meteorological conditions might explain the lower concentrations at Ridgetown compared to those at Harrow and Bear Creek, despite the close proximity of the three measure-20 ment sites. The CPF plot ( Fig. 4b) for the particle number concentrations at Ridgetown shows that high particle number concentrations were observed when winds were from the northwest, while the CPF analysis ( Fig. 4d) for high particle number concentrations at Egbert showed a stronger contribution from the south, indicating anthropogenic influences from urban areas. 25 In summary, the size distribution and number concentrations can vary substantially between rural sites and appear to be influenced by the frequency and intensity of nucleation and growth events. Such events can produce number concentrations that exceed those near high traffic. Introduction

Classification of particle nucleation and growth events
Size distribution time series at the five sites are depicted in Fig. 5. The color scale represents the number concentration (d N/d logDp) of particles in each size class. The types of nucleation and growth events at the five sites can be classified by visual inspection of the particle evolution. Atmospheric particle nucleation studies have suggested different types of nucleation events and criteria for characterizing them (e.g., Boy and Kulmala, 2002;Dal Maso et al., 2005;Jeong et al., 2006;Qian et al., 2007). However, the classification of nucleation events is still somewhat subjective as standardized criteria do not exist. In this study, criteria based on temporal changes in the number concentration and geometric mean diameter (GMD) were used to classify events, an approach more objective than one relying alone on visual inspection of the growth curves. The nucleation and growth events were classified based on their intensity and persistence into Class I, Class II, and Class N. Any continuous increase in the number concentration of particles smaller than 25 nm for at least 1 h from 07:00 a.m. to 15 03:00 p.m. (EDT) was considered to be a particle nucleation event. Class I represents the strongest nucleation and growth events, involving increases in the number concentration of particles smaller than 25 nm and continuous growth to larger mode particles. An event was designated as Class I if it lasted for at least 3 h, with a gradual increase in the GMD. Events with a high particle formation rate and little or no subsequent growth 20 of nucleated particles were classified as Class II. Non-nucleation days comprise Class N. Figure 6 shows examples of the diurnal variation in the number concentration of particles smaller than 25 nm and the GMD of particles <100 nm during the two types of nucleation events. The difference between the maximum and the minimum concentra- was typically observed at the beginning of a Class I event as a result of the formation of new particles, while a gradual increase in the GMD was seen during the subsequent growth of these newly formed particles (Fig. 6a). The elapsed time between the start point (the smallest GMD) and the end point of the period of linear increase in the GMD was used to estimate the growth rate (GR, nm h −1 ) of nucleated particles. The growth 5 event in Fig. 6a was observed for around 8 h, and the corresponding growth rate was 3 nm h −1 . There was another particle burst during the nucleation event on 23 June at Harrow. This event occurred at around 11:00 a.m., with the change in wind direction from the east (∼100 • ) to the south (∼200 • ). The air parcel from Lake Erie (to the south) showed lower number concentrations of nucleation mode particles and more mature 10 size distributions than those corresponding to the first peak. The abrupt change of wind direction at around 11:00 a.m. was consistent with Environment Canada's determination of a lake breeze event. Typically, the wind direction observed at the Harrow site fluctuated, and the particle nucleation and growth events were strongly affected by this meteorological variability. The effect of lake breezes on particle number concentration 15 is discussed in Sect. 3.3. During the Class II event on 25 June, the number concentration of particles smaller than 25 nm increased from 1600 cm −3 at 10:10 a.m. to 52 000 cm −3 at 11:00 a.m., with an abrupt change in GMD variations. The wind direction changed from the south to the east during this strong particle formation event. This Class II event was likely 20 associated with local-scale particle formation taking place in a plume. The atmospheric parameters responsible for particle nucleation and growth events are discussed in more detail later in this section. The types of particle nucleation events at the five sites were compiled (Table 3). Nucleation events were harder to resolve in Toronto than at the other sites due to the 25 high variability of the number concentrations arising from the high traffic density at this downtown site. Thus, the classification at the Toronto site was based on moving averages of particle number concentrations. In addition to the classification of nucleation events, meteorological parameters and SO 2 concentrations between 07:00 a.m. and Introduction  (Table 3). While there were several days when events were observed at more than one site, there was only one day, 2 July, when Class I events were observed at all five sites; 26 June was the only non-event day for all five sites. During the period from 20 June to 7 July 2007 the hottest temperature was recorded on 26 June with constant wind from southwest directions (209-260 degree) 5 across the region. These will be referred to hereafter as the "event day" and the "nonevent day". Class I nucleation and growth events were most frequently observed at the Harrow site, while no Class II events were observed at Toronto or Egbert. The high frequency of both regional (Class I) and local (Class II) nucleation events in Harrow resulted in the highest average number concentration at this site.

Effect of meteorological parameters
Variations in temperature, relative humidity (RH), wind speed (WS), wind direction (WD), and solar irradiance were compared based on the occurrence of nucleation events at the five sites. The ambient temperature during the Class I event days was, on average, ∼4 • C lower than on non-event days at Harrow. The temperature was also 15 lower on event days by ∼4 • C at the other four sites. On average, the RH was about 12% lower on event days (Class I) compared with non-event days (Class N). As shown in Table 3, much higher average RH was observed at Ridgetown (86±6%) than the average of 56±10% observed at Harrow and Bear Creek, the two closest sites. A negative correlation between the occurrence of nucleation events and RH has previously 20 been noted (Boy and Kulmala, 2002;Bonn and Moortgat, 2003;Jeong et al., 2006;Boy et al., 2008). Bonn et al. (2002) proposed that new particles are formed from stabilized Criegee intermediates produced during the ozonolysis of organic molecules. Boy et al. (2008) suggested that the stabilized Criegee intermediates decompose into more volatile products by reacting with water molecules. It is also possible that the growth 25 of pre-existing particles due to the uptake of water vapors may decrease particle formation rates (Charron et al., 2007). The inhibiting effect of water vapor on particle nucleation is a possible explanation for the frequent observation, in this field study, of 11628 Introduction nucleation and growth events on days with lower RH and lower ambient temperature. In addition to the higher RH, more frequent rain might partially explain the lower average number concentration of UFP at Ridgetown. The RH at Egbert (62±12%) was also slightly higher than average, possibly contributing to the lower frequency of Class I type nucleation events at this site.

5
Wind direction also appeared to influence the occurrence of particle nucleation and growth events, whereas no dependence on wind speed was found. The dominant wind direction (>50%) for non-event days was the southwest and south, directions often associated with above-average concentrations of PM 2.5 . Comparisons were made of the air mass origin at all five sites during the non-event day (26 June) and the event 10 day (2 July), based on back trajectories obtained using the National Oceanic and Atmospheric Administration (NOAA) HYbrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model with EDAS meteorological data (Draxler and Rolph, 2003). Back trajectories were calculated for 48 h, starting at 11:00 a.m. (EDT), at a height of 100 m above ground level. The air masses for the event day, 2 July 2007 (Fig. 7a), originated 15 from northern Canada and contained less polluted, cooler, drier air. This is consistent with the finding of Nillson et al. (2001), who observed that nucleation events occurred in arctic and polar air masses that corresponded with cold air advection. In contrast, during the non-event day, the air masses came from the south and passed over industrial regions containing power plants in Ohio (Fig. 7c). Generally, hotter, more humid air 20 masses were associated with higher concentrations of PM 2.5 . Nucleation events could be inhibited by high PM 2.5 mass concentrations due to the scavenging of condensable vapors by pre-existing particles (Kerminen et al., 2001).
The variability of particle number concentrations was also influenced by the lake breeze effect discussed earlier. Frequent lake breezes at Harrow, Ridgetown, and Bear 25 Creek were identified using observations from satellite and surface mesonets operated by Environment Canada (Table 4). Most of the events were associated with elevated particle number concentrations in the size range of 14 nm-100 nm. For example, on 6 July, particle number concentrations during lake breezes increased by a factor of 7 around noon at Harrow, by a factor of 16 around 02:00 p.m. at Ridgetown, and by a factor of 31 around 05:00 p.m. at Bear Creek, as compared to 30-min average particle concentrations prior to the lake breezes. During the events, GMDs for the UFP size distributions at the three sites ranged from 38 nm to 45 nm, indicating that somewhat aged air masses were impacting the sites.

5
On average, solar irradiance which exhibited a mean of 730±56 Wm −2 during Class I event days was 11% and 33% stronger than during Class II days and non-event days, respectively. It appears that Class I nucleation events corresponded closely to solar radiation intensity, which is suggestive of the formation of hydroxyl radicals through photochemical processes. During Class I event days, new particle formation began 10 when solar irradiation was as low as 140 Wm −2 (e.g., 23 June at Harrow). The highest initial value for solar irradiation before a Class I event was 860 Wm −2 on 29 June at Ridgetown. Clearly, intense solar radiation was necessary for Class I nucleation events. Alam et al. (2003) reported an analysis of nucleation events at a suburban site in Birmingham, UK, in which particle formation events were also strongly associated 15 with higher levels of solar irradiance ranged from 170 Wm −2 to 750 Wm −2 .

Effect of plume emissions
Overall average SO 2 concentrations at Harrow and Bear Creek were 4.6±3.2 ppb and 4.2±4.9 ppb, respectively, whereas substantially lower SO 2 averages were observed at Toronto (1.0±1.6 ppb) and Egbert (0.6±0.9 ppb). High SO 2 was associated with 20 the more frequent occurrence of Class II events at the Harrow and Bear Creek sites. The average SO 2 concentrations for Class II events were 7.2±3.4 ppb at Harrow and 7.1±6.6 ppb at Bear Creek, whereas the SO 2 averages for Class I were 3.0±2.8 ppb and 2.2±2.9 ppb, respectively. As shown in Fig. 6b, a sharp increase in particle number concentrations (5700 to Ridgetown (1300 to 2800 cm −3 ) at noon and 01:00 p.m., respectively, with a high concentration of SO 2 (∼14 ppb) at Bear Creek. Abrupt changes in wind direction (300 • to 150 • ) and wind speed (0.5 to 5.3 m s −1 ) were observed at Ridgetown due to the lake breeze effect, whereas a slight change in wind direction (250 • to 220 • ) with a stronger 5 wind speed (1.6 to 2.9 m s −1 ) was observed at Bear Creek during the event. The back trajectory analysis for the 25 June Class II event showed that all trajectories arriving at the Harrow, Bear Creek, and Ridgetown sites originated from the Ohio River Valley, while air masses at Harrow also passed over the northern Ohio industrial region near Lake Erie (Fig. 7b), indicating the impact of fossil fuel combustion. Hence these and the other Class II particle formation events were likely caused by sulphuric acid nucleation resulting from oxidation of SO 2 in a plume. Anthropogenic emissions of SO 2 contributed to the occurrence of nucleation events in southwestern Ontario. A high concentration of SO 2 in a plume could increase the available condensable vapor concentrations without increasing the condensation sink. Sulphuric acid is a key 15 species associated with nucleation events (Berndt et al., 2005) and events have previously been associated with high SO 2 concentration plumes, in studies conducted near stationary sources (e.g., Dunn et al., 2004;Stanier et al., 2004).
No new particle formation was observed on 25 June at the Egbert and Toronto sites. In these regions, the concentration of SO 2 also increased when the wind came from 20 the southwest on 25 June, although concentrations were much lower than those at Harrow and Bear Creek. A substantial increase in the concentration of >60 nm particles was observed at noon on 25 June at the Toronto site, along with the highest SO 2 concentration (8 ppb) during the measurement period. At Egbert, an increase in SO 2 (7 ppb) was also observed around 11:00 a.m. without a significant change in nucleation 25 mode particle number concentrations. The HYSPLIT results indicated that air masses at both sites originated from the Ohio industrial region (Fig. 7b). However, the plume emissions originating from the Ohio River Valley were advected into Toronto and Egbert areas after approximately three days of atmospheric transport. The growth and dilution Introduction of particles during transport may explain the SO 2 peaks with the presence of larger particles (>60 nm) observed on 25 June at Egbert and Toronto.

Analysis of a regional nucleation event
Particle formation events (Class I) were observed on 2 July at all five sites, up to ∼350 km apart, indicating a regional-scale nucleation event (Table 3). In the urban 5 area, interference from local sources such as motor vehicle exhaust was lower since 2 July was a holiday, which likely made it easier to detect the event at that location. On 23 and 24 June, a weekend, nucleation and growth events were also observed at all sites except Ridgetown (missing data). The particle burst that occurred on 2 July was the most distinctive nucleation and growth event observed during the measure-10 ment period (Fig. 8). Particle nucleation started between 07:00 a.m. and 10:00 a.m., occurring first at Ridgetown and Egbert. During the nucleation event, the wind was consistently from the north (Toronto and Egbert) and the east (Harrow, Ridgetown, and Bear Creek). The back trajectories in Fig. 7a show that these latter air masses also originated from northern Ontario, indicating that these sites were impacted by clean 15 cool air. Wind speeds at the Egbert site were between 3.6 m s −1 and 4.2 m s −1 during the nucleation burst, whereas very calm winds (1.1 m s −1 ) were observed before the nucleation event (∼07:00 a.m.). Initiation of the nucleation event was likely related to the arrival at Egbert of this clean air, containing few pre-existing particles. Due to the northerly winds, the ambient temperature on the event day was typically colder than on 20 other days, with moderate relative humidity. The colder temperature on the nucleation event days following the influence of cool air masses is in agreement with observations in many different locations in Finland, USA, and UK (Nilsson et al., 2001;Stainer et al., 2004;Charron et al., 2007). Hussein et al. (2009) reported that regional-scale new particle formation events at five Nordic background sites were favored in similar air 25 masses from the northern Atlantic Ocean. Most of these air masses are considered clean and not strongly influenced by anthropogenic activity except ship emissions. In regards to SO 2 concentrations, nucleation events at all sites exhibited among the 11632 Introduction lowest concentrations (≤1 ppb), suggesting that on 2 July, the region was less influenced by anthropogenic emissions. Charron et al. (2007) observed that on large-scale nucleation event days SO 2 concentrations were generally below 1 ppb with air masses from the northern Atlantic Ocean. While even this low SO 2 concentration may be sufficient to produce the ∼10 6 molecules cm −3 of H 2 SO 4 (g) needed to initiate nucleation, 5 it was likely too low to have supported the observed growth rates presented in Table 5 ( Sipilä et al., 2010). Thus, biogenic sources are believed to have played a role at the rural sites, with organic compounds contributing to growth and possibly biogenic SO 2 (g) contributing to nucleation. Biogenic emissions are typically influenced by many factors, including ambient temperature. Even though the large forested area in northern Ontario represent a major biogenic source, the colder temperature would have limited their emissions. Longer-term studies are needed to help resolve the relative contributions of anthropogenic and biogenic emissions to nucleation and growth in this region. This information is needed to help understand the possible influences of newly formed particles on regional climate. The conditions associated with the regional nucleation event 15 on 2 July were characterized in terms of the particle formation rate, the growth rate, and the condensation sink (CS) ( Table 5). As shown in Fig. 8c, the condensation sink varied over the duration of the event. In this study, two values for the condensation sink were estimated for each of the five sites: the 1-h average prior to the nucleation event (CS i ), and the 24-h average (CS aveg ). The highest formation rates for particles smaller 20 than 25 nm were observed at Egbert and Harrow, while the lowest rate was observed at the urban site (Toronto). A high CS i at the Toronto site might be responsible for its low formation rate: a high condensation sink will scavenge potential condensable vapors responsible for particle nucleation and growth. As expected, the CS i values at the five sites on the morning of the event day were lower by a factor of 1.1-2.9 than those 25 on the non-event day, 26 June. The occurrence of the highest formation rate with the lowest CS aveg at Egbert indicates the presence of rapid particle formation events without effective particle growth. Even though the lowest CS i was observed at Ridgetown preceding the nucleation event, the new particle formation rate was the second lowest of the five sites. An episode of rain was recorded (01:00 a.m.-07:00 a.m.) before the nucleation event at Ridgetown, during which ambient particles and gaseous species associated with precursors of condensable vapors may have been scavenged, resulting in the depletion of condensable vapors along with the low initial condensation sink. Growth rates (GR) at the five sites ranged from 2.9 nm h −1 to 6.7 nm h −1 . The highest 5 growth rate was observed at the downtown Toronto site. These values are comparable to those found at urban and rural sites in Germany (Wehner et al., 2007), as well as an urban (Alam et al., 2003), and a rural site in the UK (Charron et al., 2007). Wehner et al. (2007) found slightly higher GR ranging from 4.2 nm h −1 to 5.5 nm h −1 at an urban site than those identified at rural sites. At a rural site in the UK, the measured 10 GR varied between 2 nm h −1 and 6 nm h −1 , while the GR ranged from 3.6 nm h −1 to 9.0 nm h −1 in the urban area.  reported that the GR is higher in polluted areas than in clean environments and suggested a more significant contribution of sulphuric acid to particle growth in urban areas. However, the GR is often many times higher than the rate of the growth due to sulphuric acid alone; for example, most 15 of the nucleated particles in a polluted area of Mexico City were composed of organics and nitrate (Kuang et al., 2008;Smith et al., 2008). Given the similar levels of SO 2 and solar intensities on 2 July at all sites, and that Toronto had the highest growth rate yet lowest nucleation rate on this day, it is likely that other components, such as less volatile condensable organics from local sources, contributed to growth at this urban 20 site.

Temporal and spatial variability of UFP
Coefficients of divergence (COD) as a function of particle size were calculated for each site pair based on hourly average measurements over the 18 measurement days (Fig. 9a). On average, COD values were greater than 0.4 for all sizes, with a mean of 0.42, suggesting a moderate degree of spatial heterogeneity over this large region.
There was an inverse linear relationship between COD and particle size, indicating that spatial heterogeneity increased as particle size decreased. was consistent with the spatial variability reported at California sites within 0.4-11 km of each other (Krudysz et al., 2009), although the California study showed less heterogeneity than what was observed in the present study; this is likely due to the smaller spatial scales involved. With respect to the correlation analysis, increased correlation coefficients for particles larger than 40 nm were observed, while the highest correlation 5 coefficient of 0.71 was observed between Bear Creek and Ridgetown for 93 nm particles. The correlation between Egbert and Toronto, ∼80 km south of the Egbert site, was generally very poor (r < 0.1), except for larger particles (>80 nm), for which the r value was 0.52. This high temporal variability was presumably driven by the differences in source types between the rural and urban sites and the prevailing meteorological conditions during the campaign. A larger multi-year comparison of nucleation at these two sites is now underway. Notably, particles smaller than 25 nm showed low r values with high COD values. The differences in the strength and frequency of nucleation events and primary emissions from local sources, as well as the size-dependent fate of UFP in the atmosphere, would result in high spatial divergence for small particles. 15 To constrain the influence of particle nucleation events on spatial variations, the COD and r values were calculated on the nucleation event day (2 July) and non-event day (26 June) (Fig. 10). The COD values for particles in the 14 nm-100 nm size range during the nucleation event and non-event days were 0.41 and 0.55, respectively, while the corresponding correlation coefficients were 0.43 (event, p < 0.05) and 0.14 (non-20 event), indicating more uniform spatial distribution of UFP across the five areas during the regional nucleation event. Interestingly, enhanced spatial homogeneity was observed for smaller particles (<80 nm), with a greater improvement for particle sizes between 20 nm and 30 nm, relative to COD values during the non-event day. While the mean COD for the event day at all sites was 0.46 for 25 nm particles, much lower in- 25 dividual COD values (0.27-0.38) were found for Harrow, Ridgetown, and Bear Creek. In addition, the highest correlation coefficient for 25 nm particles (a mean r of 0.62) was observed during the event day, with a higher correlation (r = 0.9) among Harrow, Ridgetown, and Bear Creek. Thus, although the concentrations and size distri-Introduction butions of ultrafine particles usually were spatially heterogeneous, regional nucleation and growth events could greatly increase the spatial homogeneity and inter-site correlation of these particles.

Conclusions
Number concentrations and size distributions of particles in the size range of 14 nm-5 100 nm were simultaneously measured at five rural and urban sites in Southern Ontario during the BAQS-Met campaign. On average, the number concentrations of particles were higher at a near-border agricultural site, Harrow, than the concentration at the downtown Toronto site. This was likely due to the influence of intermittent nucleation events, which caused the UFP concentration to rise from low to high values. However, the highest median number concentrations of particles in the smaller size range 14 nm-25 nm were observed at Toronto, most likely due to persistent production of particles <25 nm in motor vehicle exhaust. While the mean size distribution for Toronto exhibited a unimode at ∼30 nm, mean bimodal distributions in rural areas were caused by nearby nucleation and growth of smaller particles and transport of larger particles from upwind 15 sources. Wind sector and back trajectory analyses suggested that strong peaks in particle number concentrations at Harrow were often associated with anthropogenic emissions. Observed nucleation and growth events at these five sites were classified in terms of the strength and persistence of number concentration of particles smaller than 25 nm 20 and geometric mean diameter variations during the events. The most distinctive type of nucleation and growth event, Class I, was frequently observed at the Harrow site. Strong and short-lived Class II events occurred more often at Harrow, Ridgetown, and Bear Creek than at the downtown Toronto site. The frequent occurrence of regional and local-scale nucleation events at Harrow resulted in the highest of the average number 25 concentrations for these five locations. During Class I nucleation events, the mean temperature and RH were 19±3 • C and 57±15%, respectively. On Class I event days, ACPD 10,2010 Particle formation and growth at five rural and urban sites the solar radiation flux was approximately 33% higher than on non-event days. Back trajectory analyses suggested that Class I nucleation and growth events were associated with less polluted cooler drier air from northern Ontario, whereas the air masses during non-event days passed over distant industrial regions to the south (e.g., Ohio River Valley). Class II events at the near-border sites were associated with high SO 2 , 5 which likely originated from nearby industrialized regions. A nucleation and growth event on 2 July was detected at all five sites, indicating that the event had a spatial extent of at least 350 km. This event was associated with unpolluted air from northern Ontario having relatively low SO 2 concentrations and a low condensation sink. The unpolluted air associated with this event suggested that biogenic sources contributed 10 to the growth and possibly the nucleation. The spatial homogeneity of particle number and size distributions measured simultaneously at the five sites within 40-350 km of each other was assessed by calculating Spearman correlation coefficients and coefficients of divergence. Coefficients of divergence among the five sites were greater than 0.4 for all particle sizes, with higher COD 15 for smaller particles (<25 nm), indicating that spatial heterogeneity increased as particle size decreased. The inter-site temporal correlations were very weak for smaller particles (<25 nm), but they increased slightly for larger particles. However, during a regional nucleation event on 2 July, homogeneity was much higher, especially for the closer Harrow, Ridgetown, and Bear Creek sites: for 20 nm-30 nm particles, the lowest 20 COD values were found along with the highest correlation coefficients.
In summary, regional nucleation events were observed even in Toronto, a major urban centre. Thus, particles from regional nucleation processes can contribute to the burden of ultrafine particles in cities, and potentially to the associated health impacts on urban populations. Nonetheless, the high number concentrations at rural sites sug-25 gest that care be taken, for now, in the use of particle number as a metric of population exposure. Further research is required to establish the differences in toxicity of particles from nucleation events and those from vehicle emissions. Many short-lived strong particle formation events observed during this three-week campaign were as-Introduction sociated with trans-border air flows from the south and south-west, and with elevated SO 2 . Hence, particle formation and growth in southwestern Ontario are in general related to anthropogenic emissions. A regional nucleation event associated with unpolluted air suggested that biogenic emissions can also contribute to particle formation and growth. These anthropogenically and biogenically-formed particles may contribute 5 to the production of cloud condensation nuclei, thus influencing regional climate. Studies of longer duration will help resolve the relative contributions of anthropogenic and biogenic emissions to nucleation and growth in this region. Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version
ACPD 10,2010 Particle formation and growth at five rural and urban sites