An investigation of nucleation events in a coastal urban environment in the southern hemisphere.

The occurrence of and conditions favourable to nucleation were investigated at an industrial and commercial coastal location in Brisbane, Australia during five different campaigns covering a total period of 13 months. To identify potential nucleation events, the difference in number concentration in the size range 14-30 nm (N 14-30 ) between consecutive observations was calculated using first-order differencing. The data showed that nucleation events were a rare occurrence, and that in the absence of nucleation the particle number was dominated by particles in the range 30-300 nm. In many instances, total particle concentration declined during nucleation. There was no clear pattern in change in NO and NO 2 concentrations during the events. SO 2 concentration, in the majority of cases, declined during nucleation but there were exceptions. Most events took place in summer, followed by winter and then spring, and no events were observed for the autumn campaigns. The events were associated with sea breeze and long-range transport. Roadside emissions, in contrast, did not contribute to nucleation, probably due to the predominance of particles in the range 50-100 nm associated with these emissions.


Introduction
Ambient aerosols play an important role in atmospheric processes affecting the human and natural environments. They affect air quality, reduce visibility, and induce climate change by directly scattering and/or absorbing the incoming solar radiation (Charlson et al., 1992;Kim et al., 2006), or indirectly by acting as cloud condensation nuclei (Hobbs, 1993). Aerosol particles are emitted from a variety of anthropogenic and natural sources either directly into the atmosphere or as secondary particles by gas-toparticle formation process.
There is growing interest in studying and analysing the mechanisms of formation of secondary particles. The development of new instruments during the 1990s to measure the particle size distribution of nanoparticles (< 50 nm) has enabled scientists to observe the formation and growth of new particles (see Kulmala et al. (2004) for review). Nucleation events, the formation of large numbers of particles at the lower end of the measurable size range, have been observed in different environments around the world. For example, they have been reported in remote (e.g. Tunved et al., 2003), urban (e.g. Jeong et al., 2004;Zhang et al., 2004) and coastal areas (e.g. Vaattovaara et al., 2006) and at various latitudes in the upper troposphere and the lower stratosphere (Lee et al., 2003).
It has been shown that the probability of nucleation was increased by elevated sulphur dioxide (SO 2 ) concentrations . This gas is mainly emitted from anthropogenic sources such as the combustion of sulphur-containing fossil fuel (Stern, 2005). Therefore, aerosol nucleation in the atmosphere would be expected to be enhanced by human activities (see also Curtius (2006) for discussion). In urban air, morning nucleation events have been found to be consistent with peaks in traffic (Jeong et al., 2004). In contrast, in coastal environments, higher concentrations of nucleation mode particles have been observed during entries of clean air rather than of polluted air (O'Dowd et al., 2002). This is also confirmed by a Finish study (Spracklen et al., 2006), which found that particle concentrations in remote continental regions are dominated by nucleated particles whereas in polluted continental regions are dominated by primary particles. This paper aims to analyse the frequency of and the atmospheric conditions favourable for nucleation events at coastal urban location in Brisbane, Australia.
Monitoring was conducted during four campaigns of two weeks duration each, and a campaign of four weeks duration, covering a total period of 13 months. The objective was to investigate which meteorological conditions enhanced the probability of nucleation and to investigate any patterns in gaseous concentrations leading to the events. The monitoring site was located between the road and railway line. The road and railway line follow a straight almost north-south path in the vicinity of the site. The sampling inlet was located in a flat area, 8 m E of the centre line of the road, 12 m W of the centre of the railway line and three metres above the ground surface. The site is surrounded by mangrove swamps to the N, and the Pacific Ocean to the E. An oil refinery is located in the south-western quadrant, with its closest point approximately 500 m from the sampling area. A seaport facility occupies most of the island, and is located approximately 1 km N of the site. Due to the freight activity in the area, traffic is dominated by diesel trucks.

Description of the instruments and monitoring procedure
Particle size distribution in the range 14-800 nm was measured with a TSI 3934 Scanning Mobility Particle Sizer (SMPS). The SMPS consists of a TSI 3071A electrostatic classifier (EC) and a TSI 3010 condensation particle counter (CPC). The EC classifies particles according to size while the CPC counts the number of particles in each size channels. An interfacing computer controls the process of measurement and stores the data supplied by the counter. Monitoring was conducted continuously at five-minute intervals.
Other air quality parameters measured were NO x , and SO 2 . NO x concentration was measured with an Ecotech 9841 NO x analyser, and SO 2 concentration was measured with an Ecotech 9850 SO 2 .
Meteorological conditions were measured with a Monitor Sensors MS1 portable weather station. The weather station monitored wind direction and speed as well as a range of other meteorological variables including temperature, relative humidity, atmospheric pressure and solar radiation intensity.

Preparation of the database for analysis
The most important marker of a nucleation event is a significant increase in the concentration of nuclei mode particles in the time series. Once a significant increase was identified, additional characteristics of the data such as traffic are examined to rule out primary particle sources. This procedure is similar to the one described by Stanier et al. (2004).
The majority of secondary particles are in the size range < 30 nm. This means that nucleation events are characterised by an increase of particle number concentration in the 14-30 nm size range (N 14-30 ). The first step was to divide the size distribution into 14-30, 30-50, 50-100, 100-300 and 300-800 nm and to calculate the total concentration of particles in each size class through the general formula: and α is a correction factor obtained from the equation: The SMPS data covered 64 channels per decade. Calculation of the log differences between consecutive size channels gave an average α-value of 0.015625.

Data analysis techniques
To identify the increase in the concentration of the N 14-30 , the difference between two successive observations was calculated. In other words, the first-ordered difference of the time series was obtained through the general equation: is positive, then it indicates that a nucleation event has likely occurred; if it is negative, then it indicates that particle loss has occurred due to growth by coagulation or condensation, or removal due to diffusion, dilution, evaporation or any other removal mechanism. To assist in the interpretation of the results, contour plots were added.
To identify the preconditions for nucleation process, atmospheric conditions during the events were recorded. In addition, hourly average NO, NO 2 and SO 2 concentrations prior to and during the event were recorded and the differences were calculated. This was done to determine whether any of these gases were particle precursors in the study area. Also, wind direction data were divided into following wind sectors (in degrees clockwise from N): • Railway ( The analysis was done to identify the sources contributing to nucleation.

Mean variation of number size distribution
The diurnal patterns of number size distribution in each of the different campaigns are shown in Figure 2. The patterns are in general different for each campaign although the two autumn patterns are very similar. Due to the length of the series, there is no enough evidence to conclude that there is seasonality in the patterns. In most cases, particle number peaks above 30 nm. Peaks below 30 nm are observed in the winter campaign between 6:00 and 8:00 local time.
In both autumn campaigns, the majority of the particles are in the approximate range 30-120 nm. Throughout the day, particle number peaks at around 50-70 nm reaching maximum concentrations between 12:00 and 15:00. High concentrations are also observed in the morning although they are lower than those in the early afternoon.
During the winter campaign, there is greater variation in number size distribution. The highest concentrations are found between 7:00 and 9:00 peaking at around 20-30 nm, and between 18:00 and 22:00, peaking at around 50-60 nm. During the morning peak, there is a high concentration in the ultrafine range thus indicating that the observed peak in the nuclei mode can be explained by increased emissions.
The spring campaign follows a similar pattern to the winter campaign although the peak concentrations are lower. A small peak at about 20 nm is observed between 7:00 and 9:00, reaching maximum concentrations at 8:00. Although this peak is accompanied by high concentrations of larger ultrafine particles, this peak appears to have another source than the larger particles.
During the summer campaign, particle number peaks at around 35-45 nm between 2:00 and 4:00, 60-80 nm between 6:00 and 10:00, and 40-50 nm between 15:00 and 19:00. The maximum concentrations are comparable to those in the winter campaign although the pattern is visibly different.
To assist in the interpretation of the daily pattern, average particle number concentration of the N 14-30 , N 30-50 , N 50-100 , N 100-300 and N 300-800 were plotted against time of the day ( Figure 3). With the exception of the summer campaign, which shows fluctuations in number concentration throughout the day, all size classes follow similar daily patterns peaking at around 6:00-8:00. The highest values occur in winter and the lowest occur in spring and summer.
With the exception of winter, throughout the day the majority of particles are in the range 30-300 nm. During the autumn campaigns, the largest contributor to the particle number is the N 50-100 , followed by the N 30-50 and N 100-300 having similar contributions.
The contribution of the N 14-30 is visibly much lower, particularly in the autumn 2006 campaign.
During the winter campaign, the N 14-30 is the maximum contributor to the particle number between 5:00 and 11:00, reaching a maximum of approximately 7.0 x 10 3 cm -3 at around 7:00. These peaks are characterised by relatively high number concentrations in the range 30-100 nm, reaching concentrations of about 10.0 x 10 3 cm 3 (N 30-50 and N 50-100 combined). For most of the rest of the day, the maximum contributor is the N 50-100 , followed by the N 30-50 . Unlike the autumn campaigns, the N 14-30 is a significant contributor to the particle number, reaching concentrations of about 3.0 x 10 3 cm -3 .
In the spring, the maximum contributor is the N 50-100 reaching maximum concentrations of 2.1 x 10 3 cm -3 between 5:00 and 8:00. The N 30-50 is the second largest contributor reaching a maximum of about 1.9 x 10 3 cm -3 during these times.
The corresponding N 14-30 concentration is about 1.5 x 10 3 cm -3 , becoming the third largest contributor, but declining sharply after these hours, down to less than 500 cm -3 at midday.
In summer, there are several fluctuations in number concentrations throughout the day, but the particle number is generally dominated by the N 30-50 reaching their maxima at about 3.0 x 10 3 cm -3 . For most of the day, the N 50-100 remains the second largest contributor, with maximum concentrations of 2.5 x 10 3 cm -3 at around 6:00.
The N 14-30 is the third largest contributor, except between 8:00 and 9:00, where it is the largest with concentrations of about 2.0 x 10 3 cm -3 , only slightly larger than the N 50-100 concentration.
To shed more light on the factors affecting the daily variation in particle number, average traffic numbers were plotted against time of the day (Figure 4). The traffic pattern is more or less similar for the different seasons. There were problems of traffic data quality with the autumn 2006 campaign and therefore the results cannot be considered representative of the actual levels. Traffic volumes peak between 8:00 and 14:00, reaching levels between 120 vehicles h -1 in the winter and 180 vehicles h -1 in the autumn 2007.
To further assist in the interpretation of the results, average wind directions for each campaign were plotted ( Figure 5). The graphs show that for both autumn campaign the wind exhibits similar daily patterns of variation. During the autumn and winter the wind originates predominantly from the south, in other words, from the refinery and mangrove swamps sectors. In spring, between 1:00 and 8:00, the wind originates from the road and port sectors. During the summer, for most of the day, the wind direction is from the railway sector. Table 1 provides a summary of the conditions observed during nucleation events and the increase in the N 14-30 . Nucleation events are defined as those where significant increases in the N 14-30 are greater than more than 50% increase in total particle number. These events were further classified as "weak", "moderate" and "strong" following a criteria similar to the one used by Stanier et al. (2004): dN 14-30 /dt < 4,000 cm -3 h -1 was classified as weak, dN 14-30 /dt 4,000-15,000 cm -3 h -1 was classified as moderate, and dN 14-30 /dt > 15,000 cm -3 was classified as strong. In a similar manner to Stanier et al. (2004) dN 14-30 /dt is the nucleation rate but the number of nuclei clusters growing to a detectable size (i.e. above 14 nm).

Conditions that favoured nucleation
The majority of the events took place in summer, followed by winter and two events in spring. No events were registered during the autumn campaigns. Increases in the concentration of the N 14-30 did not necessarily result in a significant increase in total particle number. Air temperature ranged from 18.5 o C to 28.8 o C with an average of 23.7 o C. Solar radiation levels were high, ranging from 46 Wm -2 to 1063 Wm -2 and averaging 394 Wm -2 . Based on the observed data, a t-test was done to test the hypothesis that solar radiation levels were significantly higher in summer than in winter. The test, however, proved that the difference was insignificant (p = 0.340).
This is not surprising since sampling took place in a subtropical location, where the differences in sunlight and cloud cover between summer and winter are not as large as in more temperate climate. Relative humidity ranged from 51.7% to 77.9% averaging 70.7% in winter and 56.9% in summer. In this case, the t-test showed significant differences in relative humidity values (p < 0.01).
Sometimes, the change in total concentration was negative which indicates that nucleation events occur when the concentration of larger particles decreased. In most instances, the change in the N 14-30 was greater than the change in total concentration, indicating that the events were correctly identified, in other words, the increases in N 14-30 were no associated with significant increases in the concentration of the larger particles. Table 2 shows the concentration of NO, NO 2 and SO 2 before and during the events.
No clear or dominant pattern is observed in the change of NO and NO 2 concentrations. In most cases, SO 2 concentrations, declined during the events. There were only a few exceptions, the most notable one in 19 January 2007 at 13:15, when the SO 2 concentrations rose 12.9 x 10 -2 ppb during the event.

Discussion
The size distribution of submicrometre particles in the size range 14-800 nm was examined in order to investigate the occurrence of nucleation events. The analysis was done by analysing the evolution of the size distribution during each day of the investigation, as well as the analysis of the time series of the concentration of sizefractionated particles using first-order differencing techniques.
Most nucleation events took place for the wind direction from the railway sector, in other words, when the wind moved towards the road. As mentioned earlier, traffic is dominated by diesel vehicles and therefore when the wind originates from the road the particle number is dominated by the 50-100 nm size fraction. Hitchins et al. (2000) found that when the wind moved towards the road, even at the closest point, the total particle concentration was similar to the background. Since passing of trains was rare, wind from the railway sector is dominated by the background and therefore this sector is associated with clean air masses. Sampling took place in a marine environment and therefore the influence of sea breeze is significant around the site, particularly at this sector, which was the closest to the ocean.
There were a few events associated with wind blowing from the port and the refinery sectors, and none from the road. This is indicative of long range transport. Several studies have shown that nucleation can occur more or less uniformly in air masses that extend several kilometres (e.g. Kulmala et al., 2001;O'Dowd et al., 2007). This means that newly formed particles can grow to a detectable size (i.e. 14 nm) over several kilometres, despite dilution.
Even though the wind does not blow from the road during nucleation events, for all campaigns, the N 14-30 peaked in the morning ( Global radiation has been identified to influence new particle production (Boy & Kulmala, 2002;O'Dowd et al., 1999). Solar radiation levels are generally higher in summer and lower in winter and as a result American and European studies have found nucleation events to be more frequent in summer (e.g. Qian et al., 2007).
Conversely, other studies have found that nucleation events are less frequent winter than in spring or autumn (e.g. Stanier et al., 2004) and sometimes they are even absent during the winter (Wehner & Wiedensohler, 2003). In contrast, a South Korean study found that nucleation events occurred more frequently in winter (Lee et al., 2008).
The difference between the Korean study and the European and American studies is that the Korean study was done in a coastal environment whereas those mentioned earlier were conducted in urban areas.
In the present study, although nucleation events were more frequent in summer than in winter, they were less frequent in spring, almost nonexistent, and none were observed in the autumn. The statistical test showed that the difference in solar radiation levels during these events was almost the same between summer and winter.
Furthermore, many of these events were observed at much lower radiation levels than those observed in East St. Louis (Qian et al., 2007). In that study, nucleation was observed when the median solar radiation intensity was 680 W m -2 and no nucleation events were observed at below 450 W m -2 (Qian et al., 2007). In the present investigation, nucleation events were observed even when solar radiation intensity was as low as 46 W m -2 . This compares well with the conditions observed in coastal environments (e.g. Lee et al., 2008). The occurrence of new particle production depends not only on the presence of intense solar radiation but also on the properties of the present air mass (Wehner & Wiedensohler, 2003). Therefore, although the importance of solar radiation upon new particle formation cannot be dismissed, there are other forces influencing the process, for example, the type of environment.
This study found that new particle production took place at relative humidity values of around 50% to 80% and air temperatures of 18.0 o C to around 29 o C. The relative humidity conditions for nucleation are similar to those reported by O'Dowd et al.
(2002) whilst the temperature in this study are higher than those reported in that study.
Micrometeorological processes promote an increase in surface vapour flux by providing additional water vapour, which increases the nucleation probability along with possible nucleation precursor species (O'Dowd et al., 2002). Similarly, increases in turbulent fluctuations in temperature and humidity can also significantly increase the probability of nucleation (Easter & Peters, 1994;Pirjola et al., 2000;Nilsson & Kulmala, 1998).
The analysis of the daily pattern of variation in size distribution and the concentration of size-fractionated particles indicates that the particle number is dominated by the range 50-100 nm, which is consistent with the size distribution observed for diesel exhaust emissions (Morawska et al., 1998). This is not surprising since the site is dominated by diesel traffic and sampling took place very close to the road. The dominance of this size range has the effect of increasing the available surface area.
The presence of a high particle surface area prevents nucleation due to diffusion of small particles and condensed material to the surface of larger particles (e.g. Friedlander et al., 1991) As a result, nucleation events were rare and of very short duration.
Particle number concentration was higher in winter caused by different meteorological conditions reducing the available volume of air for their dispersion.
The peak in their concentrations, with the exception of summer, occurred at around 6:00-9:00 consistent with increases in traffic volumes. Although traffic levels peak at around midday, particle number is lower than during the morning rush.
It was observed that often the formation of new particles was accompanied by decreases in the concentration of larger particles. This is shown by the negative change in total concentration during the events even though the N 14-30 increases significantly. Spracklen et al. (2006) showed that significant reductions in primary particle emissions may lead to an increase in total particle concentration because of the coupling between particle surface area and the rate of new particle formation. In the present study, new particle formation did not necessarily increase the total particle concentration.
Gaseous sulphuric acid has been identified as a key compound for the formation and growth of new particles and is formed by the reaction of SO 2 with the hydroxyl radical . Urban (e.g. Qian et al., 2007;Woo et al., 2001) and coastal studies (Lee et al., 2008) found that SO 2 concentrations increased during nucleation events. In contrast, this study found that with only some notable exceptions, SO 2 concentrations declined during the events. No reasonable explanation can be found for this discrepancy except that in the present study nucleation events were associated with cleaner air masses, or that the SO 2 is being consumed during nucleation.

Summary and conclusions
The evolution of particle size distribution in the range 14-800 nm was analysed in order to determine the frequency of nucleation events during four campaigns of two weeks duration and one campaign of four weeks, covering a total period of 13 months.
This study found that nucleation events were rare and of very short duration and did not contribute to large concentrations. The majority of the events took place in summer. There were no events during the autumn campaigns and events in winter were more frequent than in spring. These events occurred randomly independent of time of the day although they were more frequent during the daylight hours. During these events, there was no significant difference in solar radiation levels. Therefore, the differences in occurrences of nucleation between summer and winter are not explained by differences in solar radiation levels.
In many cases, the formation of new particles did result in significant reductions in the total particle concentration. This is because there are differences in size distributions and number concentrations between the road sector and the railway sector. The road sector is dominated by diesel traffic and therefore there is a much higher concentration of larger particles whereas the railway sector is mostly associated with the background emissions largely influenced by sea breeze.
Nucleation occurred in most cases when the wind originated from the railway sector followed by the port sector. Trains passing by were a rare occurrence and therefore train emissions had little influence in railway sector. No nucleation was observed when the wind originated from the road. Therefore, nucleation events were associated with long range transport and cleaner air masses. This was because more polluted air from the road was associated with direct emissions of larger particles providing a greater available surface area for condensation, thus reducing the probability of nucleation from this sector.
NO and NO 2 did not play any role in nucleation. There was no clear pattern of change in their concentrations during nucleation events. In contrast, in most cases, SO 2 concentrations declined during the events. The only plausible explanation is that nucleation was favoured by cleaner air conditions or that SO 2 was being consumed during nucleation. However, there were a few occasions when SO 2 increased therefore its role in nucleation is not very clear.